Geo-Sustainnovation for Resilient Society: Select Proceedings of CREST 2023 (Lecture Notes in Civil Engineering, 446) 9819992184, 9789819992188

Citation preview

Lecture Notes in Civil Engineering

Hemanta Hazarika · Stuart Kenneth Haigh · Babloo Chaudhary · Masanori Murai · Suman Manandhar   Editors

Geo-Sustainnovation for Resilient Society Select Proceedings of CREST 2023

Lecture Notes in Civil Engineering Volume 446

Series Editors Marco di Prisco, Politecnico di Milano, Milano, Italy Sheng-Hong Chen, School of Water Resources and Hydropower Engineering, Wuhan University, Wuhan, China Ioannis Vayas, Institute of Steel Structures, National Technical University of Athens, Athens, Greece Sanjay Kumar Shukla, School of Engineering, Edith Cowan University, Joondalup, WA, Australia Anuj Sharma, Iowa State University, Ames, IA, USA Nagesh Kumar, Department of Civil Engineering, Indian Institute of Science Bangalore, Bengaluru, Karnataka, India Chien Ming Wang, School of Civil Engineering, The University of Queensland, Brisbane, QLD, Australia Zhen-Dong Cui, China University of Mining and Technology, Xuzhou, China

Lecture Notes in Civil Engineering (LNCE) publishes the latest developments in Civil Engineering—quickly, informally and in top quality. Though original research reported in proceedings and post-proceedings represents the core of LNCE, edited volumes of exceptionally high quality and interest may also be considered for publication. Volumes published in LNCE embrace all aspects and subfields of, as well as new challenges in, Civil Engineering. Topics in the series include: • • • • • • • • • • • • • • •

Construction and Structural Mechanics Building Materials Concrete, Steel and Timber Structures Geotechnical Engineering Earthquake Engineering Coastal Engineering Ocean and Offshore Engineering; Ships and Floating Structures Hydraulics, Hydrology and Water Resources Engineering Environmental Engineering and Sustainability Structural Health and Monitoring Surveying and Geographical Information Systems Indoor Environments Transportation and Traffic Risk Analysis Safety and Security

To submit a proposal or request further information, please contact the appropriate Springer Editor: – Pierpaolo Riva at [email protected] (Europe and Americas); – Swati Meherishi at [email protected] (Asia—except China, Australia, and New Zealand); – Wayne Hu at [email protected] (China). All books in the series now indexed by Scopus and EI Compendex database!

Hemanta Hazarika · Stuart Kenneth Haigh · Babloo Chaudhary · Masanori Murai · Suman Manandhar Editors

Geo-Sustainnovation for Resilient Society Select Proceedings of CREST 2023

Editors Hemanta Hazarika Kyushu University Fukuoka, Japan Babloo Chaudhary National Institute of Technology Karnataka Surathkal Mangalore, India

Stuart Kenneth Haigh University of Cambridge Cambridge, UK Masanori Murai Shimizu Corporation Tokyo, Japan

Suman Manandhar Kyushu University Fukuoka, Japan

ISSN 2366-2557 ISSN 2366-2565 (electronic) Lecture Notes in Civil Engineering ISBN 978-981-99-9218-8 ISBN 978-981-99-9219-5 (eBook) https://doi.org/10.1007/978-981-99-9219-5 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore Paper in this product is recyclable.

Organization

Organizing Committee Chairperson Prof. Hemanta Hazarika, Kyushu University, Japan

Honorary Chairpersons Prof. Masayuki Hyodo, Professor Emeritus, Yamaguchi University, Japan Prof. Osamu Kusakabe, International Press-in Association, Japan Prof. Takaji Kokusho, Professor Emeritus, Chuo University, Japan Prof. Kazuya Yasuhara, Professor Emeritus, Ibaraki University, Japan

Co-chairpersons Prof. Gopal Santana Phani Madabhushi, University of Cambridge, UK Prof. Mitsu Okamura, Ehime University, Japan Prof. Anand Jagadeesh Puppala, Texas A&M University, USA Prof. Takayuki Shimaoka, Kyushu University, Japan Mr. Kunihiko Tanaka, Japan Foundation Engineering Co., Ltd., Japan Mr. Hiroshi Yamada, NITTOC Construction Co., Ltd., Japan Prof. Noriyuki Yasufuku, Kyushu University, Japan

v

vi

Organization

Secretary General Dr. Yoshifumi Kochi, K’s Lab Inc., Japan

Co-secretary General Dr. Naoto Watanabe, KFC Ltd., Japan

Members Mr. Shunichi Ikeda, Toa Grout Kogyo Co., Ltd., Japan Mr. Shinichiro Ishibashi, Nihon Chiken Co., Ltd., Japan Dr. Tomohiro Ishizawa, National Research Institute for Earth Science and Disaster Resilience, Japan Dr. Kentaro Kuribayashi, Eight-Japan Engineering Consultants Inc., Japan Ms. Nanae Maeda, Corp Seiko, Japan Dr. Suman Manandhar, Kyushu University, Japan Mr. Daisuke Matsumoto, Japan Foundation Engineering Co., Ltd., Japan Dr. Kenta Mizuno, Wakachiku Construction Co., Ltd., Japan Dr. Masanori Murai, Shimizu Corporation, Japan Dr. Atsunori Numata, Soil and Wood Research Institute, Japan Mr. Yoshikazu Ochi, Kawasaki Geological Engineering Co., Ltd., Japan Dr. Mai Sawada, Tokyo Institute of Technology, Japan Dr. Tadaomi Setoguchi, Association of Disaster Experts in Kyushu Region, Japan Prof. Daisuke Suetsugu, University of Miyazaki, Japan Mr. Tsuyoshi Tanaka, Tokyo City University, Japan Dr. Kentaro Yamamoto, Oita University, Japan

Steering Committee Chairperson Prof. Yasuhide Fukumoto, Kyushu University, Japan

Organization

vii

Co-chairpersons Prof. Haruichi Kanaya, Kyushu University, Japan Prof. Naoaki Suemasa, Tokyo City University, Japan

Secretary General Dr. Masanori Murai, Shimizu Corporation, Japan

Co-secretary General Mr. Tsuyoshi Tanaka, Tokyo City University, Japan

Members Prof. Takenori Hino, Saga University, Japan Mr. Shinichiro Ishibashi, Nihon Chiken Co., Ltd., Japan Dr. Yoshifumi Kochi, K’s Lab Inc., Japan Dr. Kenta Mizuno, Wakachiku Construction Co., Ltd., Japan Ms. Tomomi Nishi, Fukuoka University, Japan Dr. Atsunori Numata, Soil and Wood Research Institute, Japan Mr. Yoshikazu Ochi, Kawasaki Geological Engineering Co., Ltd., Japan Dr. Osamu Otsuka, KFC Ltd., Japan Prof. Kazunari Sako, Kagoshima University, Japan Mr. Taisuke Sasaki, Nihon Chiken Co., Ltd., Japan Dr. Tadaomi Setoguchi, Association of Disaster Experts in Kyushu Region, Japan Prof. Daisuke Suetsugu, University of Miyazaki, Japan Prof. Yuichi Sugai, Kyushu University, Japan Dr. Kentaro Yamamoto, Oita University, Japan

Technical Committee Chairperson Prof. Stuart Kenneth Haigh, University of Cambridge, UK

viii

Organization

Co-chairpersons Prof. Toshiro Hata, Hiroshima University, Japan Prof. Shinya Inazumi, Shibaura Institute of Technology, Japan Dr. Eiji Kohama, Port and Airport Research Institute, Japan Prof. Neelima Satyam, Indian Institute of Technology Indore, India Prof. Yuichi Sugai, Kyushu University, Japan

Secretary General Dr. Babloo Chaudhary, National Institute of Technology Karnataka Surathkal, India

Co-secretary Generals Dr. Kohei Araki, National Institute of Technology, Tokuyama College, Japan Dr. Suman Manandhar, Kyushu University, Japan Dr. Sugeng Wahyudi, NITTOC Construction Co., Ltd., Japan

Members Dr. Lutfian Rusdi Daryono, NITTOC Construction Co., Ltd., Japan Dr. Zentaro Furukawa, Okayama University, Japan Dr. Yi He, Southwest Jiaotong University, China Dr. Toshiyuki Himeno, National Institute of Technology, Oita College, Japan Dr. Keisuke Ishikawa, Tokyo Denki University, Japan Dr. Atsushi Koyama, University of Miyazaki, Japan Dr. Masaaki Katagiri, Nikken Sekkei Ltd., Japan Dr. Tomokazu Ozawa, moAI, Japan Mr. Kenichi Sakai, Fukuoka Prefecture, Japan Mr. Yasutaka Tabuki, Fukuoka Prefecture, Japan Mr. Tsutomu Tsuchiya, Chemical Grouting Co., Ltd., Japan Dr. Kyohei Ueda, Kyoto University, Japan Prof. Kenji Watanabe, University of Tokyo, Japan Dr. Kentaro Yamamoto, Oita University, Japan

Organization

Logistics Committee Chairperson Dr. Yoshifumi Kochi, K’s Lab Inc., Japan

Co-chairperson Mr. Daisuke Matsumoto, Japan Foundation Engineering Co., Ltd., Japan

Secretary General Mr. Shinichiro Ishibashi, Nihon Chiken Co., Ltd., Japan

Members Mr. Takashi Fujishiro, Institute of Geo-Disaster Prevention, Japan Mr. Shunichi Ikeda, Toa Grout Kogyo Co., Ltd., Japan Ms. Nanae Maeda, Corp Seiko, Japan Ms. Mika Murayama, Kyushu University, Japan Dr. Osamu Otsuka, KFC Ltd., Japan Mr. Taisuke Sasaki, Nihon Chiken Co., Ltd., Japan Mr. Akira Takeda, Okumura Corporation, Japan

Special Committee on SDGs Chairperson Dr. Atsunori Numata, Soil and Wood Research Institute, Japan

ix

x

Organization

Co-chairperson Dr. Naoki Sakai, National Research Institute for Earth Science and Disaster Resilience, Japan

Secretary General Dr. Tomohiro Ishizawa, National Research Institute for Earth Science and Disaster Resilience, Japan

Co-secretary General Dr. Mai Sawada, Tokyo Institute of Technology, Japan

Members Prof. Chandan Ghosh, National Institute of Disaster Management, India Dr. Yusaku Isobe, IMAGEi Consultant Corporation, Japan Dr. Ik Joon Kang, Kyushu University, Japan Prof. Taizo Kobayashi, Ritsumeikan University, Japan Dr. Kentaro Kuribayashi, Eight-Japan Engineering Consultants Inc., Japan Dr. Wei Feng Lee, Ground Master Construction Co., Ltd., Taiwan Dr. Nguyen Thi Hoai Linh, Kyushu University, Japan Dr. Mamiko Maeno, Fukuoka City Council, Japan Ms. Alena Raymond, University of California Davis, USA Mr. Tsuyoshi Tanaka, Tokyo City University, Japan Dr. Tran Thi Thanh Thuy, Ho Chi Minh City, Vietnam

Special Committee on Diversity Chairperson Prof. Fauziah Ahmad, Universiti Sains Malaysia, Malaysia

Organization

xi

Co-chairpersons Dr. Naoko Kitada, Geo-Research Institute, Japan Ms. Kyoko Ohta, Fukuoka Prefectural Assembly, Japan Dr. Atsuko Sato, Civil Engineering Research Institute for Cold Region, Japan Dr. Satoquo Seino, Kyushu University, Japan

Secretary General Dr. Ryoko Senda, Kyushu University, Japan

Co-secretary General Dr. Tomohiro Ishizawa, National Research Institute for Earth Science and Disaster Resilience, Japan

Members Dr. Amizatulhani Abdullah, Universiti Malaysia Pahang, Malaysia Ms. Tomomi Nishi, Fukuoka University, Japan Dr. Tadaomi Setoguchi, Association of Disaster Experts in Kyushu Region, Japan

Sponsorship Committee Chairperson Mr. Hideaki Moriya, NITTOC Construction Co., Ltd., Japan

Co-chairpersons Dr. Yusaku Isobe, IMAGEi Consultant Corporation, Japan Dr. Nozomu Kotake, JAFEC USA, USA

xii

Organization

Dr. Chakravarthy R. Parthasarathy, Sarathy Geotech and Engineering Services Pvt. Ltd., India Dr. Yasuo Shirai, Kiso-Jiban Consultants Co., Ltd., Japan Dr. Lin Wang, Chuo Kaihatsu Corporation, Japan

Secretary General Dr. Masanori Murai, Shimizu Corporation, Japan

Co-secretary Generals Dr. Akira Ishikawa, Shimizu Corporation, Japan Dr. Tadaomi Setoguchi, Association of Disaster Experts in Kyushu Region, Japan

Members Mr. Takayoshi Inukai, Zeta Sekkei Inc., Japan Mr. Daisuke Matsumoto, Japan Foundation Engineering Co., Ltd., Japan Dr. Kenta Mizuno, Wakachiku Construction Co., Ltd., Japan Mr. Katsuji Takematsu, Japan Foundation Engineering Co., Ltd., Japan Mr. Tsuyoshi Tanaka, Tokyo City University, Japan

Public Relations Committee Chairperson Dr. Kenta Mizuno, Wakachiku Construction Co., Ltd., Japan

Co-chairpersons Prof. Netra Prakash Bhandary, Ehime University, Japan Dr. Kiyoharu Hirota, Kokusai Kogyo Co., Ltd., Japan

Organization

xiii

Secretary General Mr. Tsuyoshi Tanaka, Tokyo City University, Japan

Co-secretary Generals Dr. Kentaro Kuribayashi, Eight-Japan Engineering Consultants Inc., Japan Dr. Guojun Liu, Takenaka Corporation, Japan Dr. Tadaomi Setoguchi, Association of Disaster Experts in Kyushu Region, Japan

Members Mr. Yoshikazu Ochi, Kawasaki Geological Engineering Co., Ltd., Japan Ms. Nanase Ogawa, GIKEN Ltd., Japan

Youth Endeavor Committee Chairperson Dr. Kentaro Yamamoto, Oita University, Japan

Co-chairpersons Dr. Keita Lee, Nippon Koei Co., Ltd., Japan Prof. Kazunari Sako, Kagoshima University, Japan Prof. Daisuke Suetsugu, University of Miyazaki, Japan Dr. Norimasa Yoshimoto, Yamaguchi University, Japan

Secretary General Dr. Taichi Hyodo, Toyama Prefectural University, Japan

xiv

Co-secretary Generals Dr. Zentaro Furukawa, Okayama University, Japan Dr. Shintaro Miyamoto, National Defense Academy, Japan

Members Mr. Yuta Ichikawa, Shimizu Corporation, Japan Mr. Katsuya Ogo, Nippon Koei Co., Ltd., Japan

Organization

Preface

Geo-Sustainnovation for Resilient Society is a compilation of carefully selected and peer-reviewed papers from the 2nd International Conference on Construction Resources for Environmentally Sustainable Technologies (CREST 2023). This conference, hosted by Kyushu University, Fukuoka, Japan, was held in Fukuoka International Congress Center, Fukuoka, from November 20 to 22, 2023. It was co-hosted by the University of Cambridge, International Society for Soil Mechanics and Foundation Engineering (ISSMGE), Japan Society of Civil Engineers (JSCE), Japanese Geotechnical Society (JGS), The Japan Landslide Society, ISSMGE Technical Committee 307 (TC307), ISSMGE Technical Committee 215 (TC215), ISSMGE Asian Regional Technical Committee 1 (AsRTC1), ISSMGE Asian Regional Technical Committee 3 (AsRTC3), International Press-in Association (IPA), and Global Society for Smart Geo-Sustainnovation (GLOSS). Further support to this conference was extended by the Kyushu Regional Development Bureau, Ministry of Land, Infrastructure, Transport and Tourism, Japan, Fukuoka Prefecture, Fukuoka City, Consulate General of India Osaka-Kobe, National Research Institute for Earth Science and Disaster Resilience (NIED), UN-HABITAT Regional Office for Asia and the Pacific, Fukuoka, Kyushu Branch of Japanese Geotechnical Society (JGS), The Society of Materials Science Japan (JSMS), Japan Federation of Construction Contractors (JFCC), Japan Civil Engineering Consultants Association (JCCA), Japan Geotechnical Consultants Association (JGCA), Organization of Geowaste Technology for a Recycled Based Society, Applied Slope Engineering Association (ASERG), and the Local Resilience Research Institute (LRRI). The principal aim of CREST 2023 was to disseminate knowledge and foster discussions concerning issues related to natural disasters and disasters associated with anthropogenic activities. Additionally, it sought to proffer solutions through utilization of alternative resources, groundbreaking technologies, and adaptable disaster mitigation strategies. All these endeavors converged toward the common objective of establishing a resilient and sustainable society from a geoengineering perspective. The conference’s themes spanned across a wide range of interdisciplinary areas that align with the Sendai Framework for Disaster Risk Reduction (DRR). The holistic approach, based on Sendai framework for DRR, includes the integration of disaster xv

xvi

Preface

risk reduction into development strategies, thereby cultivating resilience in nations and communities against future disasters. The conference was centered on the themes of sustainability, the promotion of innovative concepts, and advancements in the design, construction, and maintenance of geotechnical structures. All these efforts were geared toward contributing to climate change adaptation and disaster resilience, in alignment with the sustainable development goals (SDGs) set forth by the United Nations (UN). The conference aspired to unite scientists, researchers, engineers, and policymakers from around the world, creating a platform for robust debate and discussion on these pressing issues. The proceedings of CREST 2023 are thoughtfully partitioned into five volumes. This volume encompasses eight fundamental issues, namely, (i) Information-based (AI, IoT, VR, etc.) measures for natural disaster mitigation, (ii) Application of DX and i-Construction, (iii) Physical and numerical modelling of disasters and disaster mitigation techniques, (iv) Smart energy harvesting techniques, (v) Disaster and environment, (vi) Community outreach through soft type disaster mitigation measures, (vii) Education for sustainable development goals, and (viii) Measures for achieving Society 5.0 goals. This book comprises a total of 36 contributions, this volume is set to be presented during the conference by authors and speaker shailing from all corners of the globe. Each manuscript underwent rigorous reviews, subjected to evaluation by at least two reviewers chosen from an international panel of experts. The publication of Geo-Sustainnovation for Resilient Society was made possible through the unwavering dedication of the core members of the technical committee of CREST 2023, along with the contributions of the staff from the Research Group of Adaptation to Global Geo-Disaster and Environment, Kyushu University, Japan. The editors extend their heartfelt gratitude to all those who have played a part in this endeavor. The editors also wish to convey their heartfelt appreciation to all the reviewers for generously dedicating their time and efforts to meticulously reviewing the manuscripts, thereby enhancing the overall content. The editors hold the hope that this book will prove valuable to students, researchers, professionals, and policymakers. The editors also firmly believe that, in the years to come, the knowledge encapsulated within this volume will contribute significantly to the realization of the UN’s Sustainable Development Goals. Fukuoka, Japan Cambridge, UK Surathkal, India Tokyo, Japan Fukuoka, Japan

Hemanta Hazarika Stuart Kenneth Haigh Babloo Chaudhary Masanori Murai Suman Manandhar

Acknowledgments

Financial Supports The Organizing Committee of the 2nd International Conference on Construction Resources for Environmentally Sustainable Technologies (CREST 2023) and the editors of this book gratefully acknowledge the financial support provided by the following organizations: The Maeda Engineering Foundation, Tokyo; Fukuoka Convention and Visitors Bureau Fukuoka; Japan Tourism Agency, Ministry of Land, Infrastructure, Transport and Tourism, Government of Japan; and Consulate General of India, Osaka-Kobe. The editors also would like to acknowledge all our sponsors (Diamond, Gold, and Silver), without which holding off the conference and this publication would not have been possible.

Panel of Reviewers The manuscript for each chapter included in this book was carefully reviewed for the quality and clarity of technical contents by at least two members from the review panel, consisting of the following international experts. The editors wish to express their sincere gratitude to all the reviewers for their valuable time and efforts. Dr. Christelle Nadine Abadie, University of Cambridge, UK Dr. Kohei Araki, National Institute of Technology, Tokuyama College, Japan Dr. Arup Bhattacharjee, Jorhat Engineering College, India Dr. Babloo Chaudhary, National Institute of Technology Karnataka Surathkal, India Dr. Jitesh T. Chavda, Sardar Vallabhbhai National Institute of Technology Surat, India Dr. Lutfian Rusdi Daryono, NITTOC Construction Co., Ltd., Japan Dr. Rohit Divyesh, TechFab India Industries Ltd., India Prof. Yasuhide Fukumoto, Kyushu University, Japan Dr. Zentaro Furukawa, Okayama University, Japan xvii

xviii

Acknowledgments

Prof. Zheng Han, Central South University, China Dr. Yi He, Southwest Jiaotong University, China Prof. Shinya Inazumi, Shibaura Institute of Technology, Japan Dr. Yusaku Isobe, IMAGEi Consultant Corporation, Japan Dr. Yoshifumi Kochi, K’s Lab Inc., Japan Prof. Adapa Murali Krishna, Indian Institute of Technology Tirupati, India Dr. Guojun Liu, Takenaka Corporation, Japan Dr. Suman Manandhar, Kyushu University, Japan Dr. Shintaro Miyamoto, National Defense Academy of Japan, Japan Prof. Kasinathan Muthukkumaran, National Institute of Technology Tiruchirappalli, India Prof. Toshihiro Noda, Nagoya University, Japan Dr. Hemanth Noothalapati, Shimane University, Japan Dr. Rolando Orense, University of Auckland, New Zealand Dr. Osamu Otsuka, KFC Ltd., Japan Prof. Rathish Kumar Pancharathi, National Institute of Technology Warangal, India Prof. Muhammad Akmal Putra, NITTOC Construction Co., Ltd., Japan Prof. Mizuki Sakai, National Institute of Technology, Nagano College, Japan Dr. Ryoko Senda, Kyushu University, Japan Dr. Meghna Sharma, National Institute of Technology Hamirpur, India Dr. Kazuaki Tanaka, Kyushu Institute of Technology, Japan Prof. Tetsuo Tobita, Kansai University, Japan Dr. Sugeng Wahyudi, NITTOC Construction Co., Ltd., Japan Prof. Akihiko Wakai, Gunma University, Japan Dr. Khonesavanh Vilayvong, Kyushu University, Japan

Contents

Information Based (AI, IoT, VR etc.) Measures for Natural Disaster Mitigation A Stratigraphic Classification Estimation Method by the D-Layer Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Satoshi Murakami An Approach for Evacuation Vulnerability Assessment with Consideration of Predicted Evacuation Time . . . . . . . . . . . . . . . . . . . . Zishuang Han, Kohei Kawano, Ibrahim Djamaluddin, Takumi Sugahara, Hiroyuki Honda, Hisatoshi Taniguchi, and Yasuhiro Mitani Development and Applicability Assessment of a Tunnel Face Monitoring System Against Tunnel Face Collapse . . . . . . . . . . . . . . . . . . . . Ryohei Hase, Yasuji Mihara, Dohta Awaji, Rieko Hojo, and Shoken Shimizu Development of Real-Time Measuring System of Tip Position with Deep Mixing Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hong-Son Nguyen, Yuji Adachi, Minoru Yamada, Toshiaki Takaue, Masato Kobayashi, Moriyasu Takada, and Shinya Inazumi Evaluation of Landslide Triggering Mechanism During Rainfall in Slopes Containing Vertical Cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muhammad N. Hidayat, Hemanta Hazarika, Takashi Fujishiro, Masanori Murai, Suman Manandhar, Yan Liu, Yasuhide Fukumoto, Haruichi Kanaya, and Osamu Takiguchi Landslide Risk Prediction and Regional Dependence Evaluation Based on Disaster History Using Machine Learning and Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fudong Ren and Koichi Isobe

3

11

23

35

47

57

xix

xx

Contents

Machine Learning for Estimation of Surface Ground Structure by H/V Spectral Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tetsuo Tobita and Wataru Yamamoto

69

Regular Deformation-Based Landslide Potential Detection with DInSAR—A Case Study of Taipei City . . . . . . . . . . . . . . . . . . . . . . . . . . Kuo-Lung Wang, Jun-Tin Lin, and Shih-Yuan Lin

81

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient Tunnel Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Motoki Sato, Nobusuke Hasegawa, Hiroki Ohtsuka, and Erika Kukisawa

89

Application of DX and i-Construction A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Shih-Hao Cheng, Min-Ju Li, Ricky K. N. Wong, Takeshi Iwakubo, and Shinichi Ueno Aso Ohashi Bridge Area Slope Disaster Reconstruction Project Executed by Unmanned Construction Method Using i-Construction . . . . 113 Shinichi Nomura, Naoto Yamagami, and Tsuyoshi Nakade Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 K. Rangaswamy and Regi P. Mohan Physical and Numerical Modelling of Disasters and Disaster Mitigation Techniques A Plasticity Model of Binary Mixtures for Liquefaction Simulation Considering the Equivalent Granular Void Ratio . . . . . . . . . . . . . . . . . . . . . 145 Fu-Hsuan Yeh, Yi-Qian Lu, and Louis Ge Development of an Advanced Landslide Simulation Using Clustering Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 Kazuo Matsuura and Yasuhide Fukumoto Effect of Microbial Strains Through Triaxial Test on Bio-Treated Granular Soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Yu-Syuan Jhuo, Pin-Hsiu Liu, Chang-Ping Yu, and Louis Ge Experimental Study of Warm Permafrost Mechanical Property Under Cyclic Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Liwei Song, Junfang Liu, Xiaomin Liu, and Xin Zhao Influence of Initial Static Shear Stress on the Dynamic Response of Embedded Cantilever Retaining Wall with Saturated Backfill . . . . . . . 185 Anurag Sahare, GyuChan Choi, and Kyohei Ueda

Contents

xxi

Mechanical Behaviors of MICP Treated Binary Granular Materials Under One-Dimensional Consolidation . . . . . . . . . . . . . . . . . . . . . 197 Yu-Syuan Jhuo, Yi-Qian Lu, Chang-Ping Yu, and Louis Ge Numerical Simulations for Seismic Response of Laterally Loaded Pile Foundation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Tanmoy Barik and Babloo Chaudhary Response of Offshore Wind Turbine Foundation Subjected to Earthquakes, Sea Waves and Wind Waves: Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Subodh Kumar, Babloo Chaudhary, Manu K. Sajan, and P. K. Akarsh Seismic Response Characteristics of Loess Slope in Seasonally Frozen Regions Using Shaking Table Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 Jinchang Chen, Ailan Che, Zhijian Wu, and Lanmin Wang Seismic Responses of Rubble Mound Breakwater: Numerical Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 P. K. Akarsh, Babloo Chaudhary, Manu K. Sajan, and Subodh Kumar Stability Analysis of Rubble Mound Breakwaters Under Tsunami Overflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 Manu K. Sajan, Babloo Chaudhary, P. K. Akarsh, and Subodh Kumar Disaster and Environment Anisotropy of Pressure Generated by Water Absorption of Deformation-Constrained Granular Bentonite Specimens . . . . . . . . . . . 257 Yuta Ichikawa, Hideo Komine, and Shigeru Goto Development and Geographical Evaluation of Slope Failure Risk Index Due to Changes of Soil Moisture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 Honoka Asada and Kohei Araki Drought and Management Approach for Sustainability of Tropical Reservoir Ecosystem: A Case Study of Ubolratana Reservoir, the Most Productive Reservoir in the Northeast Thailand . . . . . . . . . . . . . 281 Jeeraya Muangsringam and Charumas Meksumpun Evaluating Liquefaction Strength Prediction Method for Low Improvement Ground Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 Koji Yamamoto, Kenichi Sato, Takuro Fujikawa, and Chikashi Koga Landslide Analysis Using Geographic Information System (GIS) in South Tapanuli Regency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Ika Puji Hastuty and Fauziah Ahmad River-Basin Classification for Flood Risk Assessment in Indonesia . . . . . 309 Adityawan Sigit and Morihiro Harada

xxii

Contents

Synthesis of Management Measures for Water Resource Sustainability and Resilient Society: A Case Study of Mae Klong Watershed Through Four Decade Evidences . . . . . . . . . . . . . . . . . . . . . . . . . 321 Charumas Meksumpun Community Outreach Through Soft Type Disaster Mitigation Measures Developing Effective Flood Inundation Maps for Risk Communication and Evacuation Planning in Obuse, Japan . . . . . . . . . . . . 333 Mizuki Sakai, Yugo Hashimoto, Naoki Todoroki, Yoshinori Furumoto, and Toru Oomiya Disaster Preparedness and Perception of Earthquake Risk: A Case of Samoeng Nuea Sub-District and Samoeng Tai Sub-District, Chiang Mai, Thailand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 Thapthai Chaithong, Kanyaporn Parung, Jirapa Khoysungnoen, Panyawi Tanseenawanon, and Shupisara Jingsungnoen Estimating the Effects of Community Disaster Management Plan on Disaster Risk Reduction Literacy Using Propensity Score Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 Takumi Sugahara, Shinya Fujimoto, Hiroyuki Honda, Hisatoshi Taniguchi, Tabihito Fujihara, and Yasuhiro Mitani Research on Methods for Determining and Understanding the Soundness of Retaining Walls Using Image Analysis . . . . . . . . . . . . . . . 367 Kenki Owada, Anurag Sahare, Kazuya Itoh, and Naoaki Suemasa Education for Sustainable Development Goals A Decision Support Tool for the Sustainability Rating Index for the Maintenance of Low-Volume Rural Roads in India . . . . . . . . . . . . . 381 Raji Reddy Myakala and S. Shankar Educational Journeys: Student Perception of School Life in Disaster Recovery Contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 Emily Gall Orillon

About the Editors

Prof. Hemanta Hazarika is currently a professor in the Graduate School of Engineering and Department of Interdisciplinary Science and Innovation, Kyushu University, Fukuoka, Japan. Professor Hazarika’s research activities include disaster prevention and mitigation, soil-structure interaction, stability of soilstructures during earthquakes and tsunami, ground improvement, application of recycled waste and lightweight materials in constructions, stability of cut slopes, and landslides and their protection. He has more than 350 technical papers in various international journals, international conferences, workshops and symposia to date. He has also authored two textbooks on soil mechanics and their Japanese versions. In addition, Prof. Hazarika served as the editor of six books on diverse topics. Prof. Hazarika has several years of experience in teaching, research as well as geotechnical practice and consulting both within and outside Japan. He is currently a foreign expert of the world’s first research center in Palu, Indonesia on liquefaction research called “The National Research Center for Liquefaction.” Currently, Prof. Hazarika is the chairman of Asian Technical Committee on “Geotechnical Mitigation and Adaptation to Climate Changeinduced Geo-disasters in Asia-Pacific Regions” of International Society of Soil Mechanics and Geotechnical Engineering (ISSMGE). He is also Founding President of a general incorporated corporation called GLOSS (Global Society for Smart Geo-Sustainnovation).

xxiii

xxiv

About the Editors

Prof. Stuart Kenneth Haigh is a professor in Geotechnical Engineering at Cambridge University and a fellow of Trinity College. In addition, Prof. Stuart Kenneth Haigh is Assistant Director of the Schofield Centre. He has a wide range of research interests centered around physical modelling at 1g and utilizing centrifuges and numerical modelling using FE and MPM. His 25 years of experience in developing complex centrifuge experiments includes developing earthquake actuators, high-g robots, and an in-flight tunnel boring machine. He also delivered the Geotechnic Lecture on Mobilizable strength design for flexible retaining walls in November 2021. He has published more than 100 papers in several journals of repute. Dr. Babloo Chaudhary is an assistant professor in the Department of Civil Engineering at National Institute of Technology Karnataka (NITK) Surathkal, India. He completed his Ph.D. from Kyushu University, Japan. He was a postdoctoral fellow at Kyoto University, Japan. His research activities include geo-disaster prevention and mitigation, dynamic soil-structures interaction, coastal geotechnics, renewal energy, ground improvement, energy geotechnics. His expertise extends from physical model tests, including 1g model tests and centrifuge model tests, to numerical simulation in the domain of geotechnical Engineering. He has published around 100 technical papers in various international and national journals, conferences, workshops and symposia. Dr. Chaudhary is recipient of Shri M. S. Jain Biennial Award given by Indian Geotechnical Society in 2017. Dr. Masanori Murai has had a distinguished career as a senior engineer with Shimizu Corporation, a leading construction company in Japan, after receiving his Ph.D. from Kochi University in Japan. Throughout his career, he has specialized in the field of geotechnical engineering and has played a key role in the design, construction, and maintenance of a wide range of infrastructure projects. In addition to his work on infrastructure projects, Dr. Murai has also been involved in research and development activities related to geotechnical engineering. He has published numerous papers in peerreviewed journals and presented at conferences around

About the Editors

xxv

the world, sharing his expertise and insights with other professionals in the field. Overall, Dr. Murai’s career at Shimizu Corporation has been characterized by a strong focus on safety, sustainability, and innovation. His expertise in geotechnical engineering and commitment to excellence have made him an asset to the company and the field of engineering as a whole. Dr. Suman Manandhar is a Research Fellow in the Department of Civil Engineering at Kyushu University, Japan. He completed his post-doctoral research from Kyushu University and enhanced his career from a lecturer to guest/visiting associate professor at the Institute of Lowland and Marine Research, Saga University, Japan. He also served in different academic and professional sectors in Nepal. He is one of the founder members of Global Institute for Interdisciplinary Studies (GIIS) in Nepal. He has published more than 70 papers in refereed journal and proceedings in different SCI and Scopus-indexed journals. His major research areas cover risk sensitive land use plan and multi-hazard assessments, geo-disasters, DRRM, foundation, ground improvement, stability analysis, liquefaction analysis, numerical modeling, and compaction on embankment slopes.

Information Based (AI, IoT, VR etc.) Measures for Natural Disaster Mitigation

A Stratigraphic Classification Estimation Method by the D-Layer Neural Networks Satoshi Murakami

Abstract Information of borehole logs in a geotechnical information database has non-uniformity of spatial density of investigation locations, their depth and different appearance frequency of soils. So, it has characteristic of information bias. The purpose of this study is to propose a new idea of a global error function of Neural Networks (NNs) for stratigraphic classification estimation and its machine learning procedure in consideration with the information bias. Ordinary NN method (ONN) treats all the input data without discrimination in machine learning because the error function of NN is defined as a residual square sum in each data without taking account of information bias. In this study, Layer NN method (LNN) of which a new error function depends on an averaged residual squared sum in each stratum and distance of estimation location. LNN with and without consideration of dependence on the distance has been indicated as Distance Correlation Layer NN method (DLNN) and Simple Layer NN method (SLNN), respectively. In DLNN, the dependence on the distance expresses by Inverse Distance Weighting (IDW) method. Applicability and performance of ONN, SLNN, and DLNN in 3 cross sections in Fukuoka plain have been investigated by using a geotechnical information database. The proposed method that is DLNN has obtained higher correct estimation rate in each 3 cross sections than ONN and SLNN. And, it could eliminate or reduce incorrect estimations in the others. Therefore, the results have shown the proposed method (DLNN) is extremely effective for stratigraphic classification estimation by Neural Networks. Keywords Machine learning · Artificial neural network · Stratigraphic classification

S. Murakami (B) Fukuoka University, Fukuoka, Fukuoka, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_1

3

4

S. Murakami

1 Introduction Geotechnical Information Database (Geo-DB) can be used for local vulnerability assessment against geo-disaster. However, almost of geo-information of borehole logs in the Geo-DB has no information on stratigraphic classifications. Therefore, stratigraphic classification must first be performed for all geotechnical information if vulnerability assessment requires stratigraphic classification. So, the availability of a pre-stratified Geo-DB would be extremely useful for the assessment. The purpose of this study is to confirm the applicability of Artificial Neural Networks (NN) for stratigraphic classification estimation. The study proposes a machine learning procedure in consideration with non-uniformity of spatial density of investigation locations, their depth and different appearance frequency of soils. The proposed method has been applied to Geo-DB in Fukuoka plain, Japan. The results have shown the proposed method is extremely effective for stratigraphic classification estimation by NN.

2 Neural Networks for Stratigraphic Classification Estimation Using the geotechnical information (depth, soil type, N-value, etc.) in the BDs, we consider strata classification by Artificial Neural Network (NN). We focus on BD No. p shown in Fig. 1. The number of input data corresponding to layer number k is denoted as N p,k . This geotechnical information is used as the input value, and the output value is calculated from the input value through an intermediate layer. In the ordinary NN (ONN), the sum of the error functions for all training data (referred to as the overall error function E in this study) is considered and the NN

Fig. 1 Machine learning of NN for p-th BD and its error function

A Stratigraphic Classification Estimation Method by the D-Layer …

5

model that yields good estimation results is determined by minimizing it. E is defined as E=

P K N p,k 1  E ( p,k,n) N p=1 n=1

(1)

k=1

The ONN method uses the data equally to make decisions during learning. However, in the case of BD, the survey depth is not uniform, and the number of data tends to become sparser as the depth increases. The engineer recognizes this bias in the data and understands the characteristics of each layer. The stratigraphic classification is then determined by looking at the surrounding stratigraphic conditions. Therefore, this study considers incorporating the engineer’s manner by redefining the overall error function, which is different from the ONN method. The following overall error function E, which is calculated by averaging to eliminate bias in the data for each stratum in the BD, has been defined by introducing a weight coefficient r p,k with respect to the distance between the locations of BD. N p,k K P 1   r p,k  E ( p,k,n) E= K k=1 p=1 N p,k n=1

(2)

 where Pp=1 r p,k = 1. The method using weight coefficients that depend on the distance is called the Distance-Layer NN method (DLNN method), while the case with equal weight coefficients is called the Simple Layer NN method (SLNN method) [1]. The SLNN method is a device to eliminate the bias by averaging the teacher data that differ in number per stratum, and the DLNN method further reduces the difference with the teacher data closer to the estimation location. By minimizing the overall error function in this way, we expect the NN model to average the data and reduce the bias, so that the information on BDs closer to the estimated location has smaller errors. In this study, L2 regularization is applied in machine learning. The training data are first divided into teacher data and test data, and L2 regularization parameter λ is determined by contrasting the estimated value of the test data obtained from the machine learning results using the teacher data with the assumed λ with the measured value of the test data. Next, using the determined λ, the best IDW parameter used in the DLNN method is also selected by comparing the estimated value of the test data obtained from the machine learning results using the teacher data with several assumed IDW parameters and the actual measured value of the test data.

6

S. Murakami

3 Overview of the Objective Region and Preparation of Training and Validation Data The objective region is the Fukuoka Plain in Japan. The BDs were taken from the Kyushu Geotechnical Information Shared DB (2005, 2012, 2018), and the BDs along the Fukuoka Urban Expressway from Chidoribashi JCT (a) to Tsukiguma JCT (b), where continuous geotechnical information exists as shown in Fig. 2. For the 125 BDs extracted from the DB, stratigraphic classification in the area has been performed using the AIST-Borehole Log Analysis ver. 2.1 provided by the National Institute of Advanced Industrial Science and Technology (AIST) in Japan. The stratigraphic classification results are shown in Fig. 3. The 125 BDs were divided into five parts, four as training data (Dataset A, B, C, and D) and the remaining one as validation data (Dataset X). The trends of the data for study and validation are shown that there is no difference in the percentage of data occupied by each layer, the Nakasu and Arae layers have a large number of data and account for the majority of the data, while the Kanatake gravel layer tends to be extremely small, about 10–15% of the total data [2].

Fig. 2 Cross section line and BD locations in the objective region

Fig. 3 Geological cross section

A Stratigraphic Classification Estimation Method by the D-Layer …

7

4 Results and Discussions Using four sets of training data (Dataset A, B, C, and D), three sets are used as learning training data and the remaining one set is used as test data for cross-validation, and λ has been determined by the ONN and SLNN methods. Given λ = 10–3 , 10–4 , 10–5 , and 10–6 , λ has been decided on 10–6 and 10–5 , which had the smallest difference between the percent-age correct by the teacher data and the percentage correct by the data for testing, for the ONN and SLNN methods, respectively. Next, the IDW parameter m in the DLNN method has been determined using the results from the SLNN method. In the present study, the search radius R is fixed as 500 m. Based on the results of three comparisons with m = 1.0, 2.0, and 4.0, m has been set as 2.0. Using the results of the above studies, the data for validation has been estimated. The number of intermediate layers in the NN model is set to 9, and 50 million machine learning cycles using ADADELTA [3] are used for the update calculation. The results of the distribution of correct and incorrect answers and the percentage of correct answers for the data for validation by the ONN, SLNN and DLNN methods are shown in Figs. 4, 5 and 6, respectively. ONN and SLNN have approximately the same value of the correct answer rate for the training data. The value of the correct answer rate of the verification data slightly decreases in the both case of ONN and SLNN method. However, the percentage of correct data for the validation of the ONN and SLNN methods is about 10% higher for the DLNN method, indicating that the strata classification estimation by the DLNN method, which attempted to incorporate engineers’ methods, is superior to the ONN and SLNN methods.

Fig. 4 Estimation results by ONN

8

S. Murakami

Fig. 5 Estimation results by SLNN

Fig. 6 Estimation results by DLNN

5 Conclusion The purpose of this study is to clarify the applicability of the NN method to stratigraphic classification estimation using BDs, and a cross section of the Fukuoka Plain has been selected as the objective region. Although the ONN method, which uses all the data equally as a basis for decision making during training, gave a correct estimation rate of more than 80%, the DLNN method, which averages the data to reduce bias and expects the information from BDs closer to the estimated location to have a smaller error rate, gave a higher rate. This suggests that it is necessary to consider the information bias of the geotechnical information in order to obtain better estimation results. Acknowledgements Part of this research was supported by Grant-in-Aid for Scientific Research (C) (JP20H00266) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

A Stratigraphic Classification Estimation Method by the D-Layer …

9

References 1. Murakami S (2020) The D-layer neural networks method for stratigraphic classification estimations. In: Proceedings of the 55th geotechnical engineering conference: DS-2-04 (in Japanese) 2. Murakami S, Ryu S, Miwa S (2020) Effectiveness of the D-layer neural networks method on stratigraphic classification estimations—case study of a cross section of ground in Fukuoka plain. In: Proceedings of the 55th geotechnical engineering conference: 22-12-2-06 (in Japanese) 3. Zeiler MD (2012) ADADELTA: an adaptive learning rate method. https://arxiv.org/pdf/1212. 5701.pdf

An Approach for Evacuation Vulnerability Assessment with Consideration of Predicted Evacuation Time Zishuang Han , Kohei Kawano , Ibrahim Djamaluddin , Takumi Sugahara , Hiroyuki Honda , Hisatoshi Taniguchi , and Yasuhiro Mitani

Abstract Heavy rainfall is a frequent and widespread severe weather hazard that may cause flood damage and human casualties. Since heavy rainfall is a progressive disaster, its scale and hazardous areas can be foreseen beforehand. Therefore, evacuating people from hazardous buildings to shelters in advance is an efficient effort to reduce casualties, but a scientific basis is still required. This paper proposes an approach for assessing each building’s evacuation vulnerability based on predicted evacuation time, aiming to support evacuation decision-making under heavy rainfall. As such, this paper applies Dijkstra’s algorithm to find the evacuation route from each building to accessible shelters. Moreover, a prediction model based on the random forest algorithm is developed to estimate their time-varying evacuation time. Road spatial and temporal characteristics that may affect evacuation time are used when developing the model. Finally, the proposed approach is implemented in Joso City, Japan, to verify its feasibility. As a result, the proposed approach accurately predicts and visualizes the evacuation time between each building and its optimal evacuation shelter. It also visually identifies the hard-to-evacuate buildings. The results indicate that the proposed approach can effectively reflect evacuation vulnerability and support heavy rainfall evacuation decision-making, which proves its validity and practicality. Keywords Heavy rainfall disaster · Vulnerability assessment · Evacuation time prediction Z. Han (B) · T. Sugahara Department of Civil Engineering, Graduate School of Engineering, Kyushu University, Fukuoka, Japan e-mail: [email protected] K. Kawano · H. Honda · H. Taniguchi · Y. Mitani Disaster Risk Reduction Research Center, Graduate School of Engineering, Kyushu University, Fukuoka, Japan I. Djamaluddin Faculty of Engineering, Hasanuddin University, Makassar, Indonesia © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_2

11

12

Z. Han et al.

1 Introduction Due to global warming, the frequency and intensity of heavy rainfall disasters are increasing globally. Heavy rainfall is a widespread severe weather hazard that may cause flood damage and enormous losses to human lives [1–3]. Since heavy rainfall is a progressive disaster, although it cannot be prevented entirely, its scale and hazardous areas can be foreseen beforehand. Therefore, an appropriate evacuation can help minimize the disaster’s impact and mitigate human casualties [4, 5]. Local municipalities are thus expected to issue the appropriate evacuation notifications to the appropriate region at the appropriate time. However, local municipalities need to make quick judgments based on various information. Moreover, there is a possibility of issuing inappropriate information through subjective judgments, which may threaten residents’ safety [6]. Therefore, providing a quantitative basis for local municipalities to make evacuation decisions is vital. Vulnerability is a significant component of disaster risk management. It refers to a person’s or group’s exposure, susceptibility to natural hazards, and inability to cope with the hazard’s impact [7–9]. Regarding previous studies, vulnerability research mainly focused on what kind of damage, loss, or recovery ability there would be in a hazardous situation [7, 9, 10]. As these methods mainly evaluate vulnerability after disasters, they are unsuitable for supporting evacuation decision-making. This study aims to develop a method to quantitatively assess vulnerability and provide an evacuation decision-making basis for local municipalities during heavy rainfall. In this study, the predicted evacuation time is used to quantify evacuation vulnerability. Moreover, an evacuation time prediction method is proposed to estimate the time-varying vulnerability of each building during heavy rainfall. The remainder of this paper begins with a detailed description of the proposed evacuation time prediction method, which includes selecting evacuation routes and predicting evacuation times. The following section discusses the application of the proposed method in Joso City. Finally, the conclusion is outlined at the end.

2 Methodology To develop a method to evaluate each building’s vulnerability under heavy rainfall disasters, this study first applies Dijkstra’s algorithm to select all the shortest evacuation routes from each building to all possible evacuation shelters under the impact of heavy rainfall. Then, this study uses the random forest algorithm to predict evacuation time on each selected route, and the minimum of those is the final solution. The following subsections further explain the process in detail.

An Approach for Evacuation Vulnerability Assessment …

13

2.1 Evacuation Route Selecting In this part, a Dijkstra’s algorithm is first applied to find the shortest route from each building to all possible evacuation shelters without considering heavy rainfall. The Dijkstra’s algorithm is a graph search algorithm proposed in 1959 by Edsger Dijkstra that has been widely applied in disaster evacuation route selection [11, 12]. Typically, there is more than one evacuation shelter accessible for each building. Therefore, this study targets all accessible evacuation shelters as goals and applies Dijkstra’s algorithm to solve all the shortest routes from each building to all possible evacuation shelters. Next, the result of Dijkstra’s search is merged with the time-varying inundation area. Then, the routes passing through inundation areas are updated as impassable routes. Correspondingly, the passable routes are selected as the optional evacuation routes, and buildings without optional routes are considered hard-to-evacuate. The target here is to find optional evacuation routes for each building. Therefore, the method is repeated multiple times until the optional evacuation routes for all buildings are found.

2.2 Evacuation Time Predicting This study assumes that residents use private vehicles during the evacuation process. Then, the random forest algorithm is used to predict the evacuation time of private vehicles on the selected evacuation route. Random forest is a tree-structured ensemble learning method proposed by Leo Breiman in 2001, which has high robustness, high performance, and high practicability [13, 14]. In our previous study, we used only road spatial characteristics as input variables and applied the random forest classification algorithm to predict evacuation time, which is straightforward but limited in reflecting the time-varying characteristics [15], because both the spatial characteristics and time characteristics significantly impact travel time prediction [16–18]. This study incorporates spatial variables with temporal variables and applies random forest regression algorithm to address the remaining problems. The detailed information and definition of the selected temporal and spatial variables can be seen in Table 1. Taking the example of predicting evacuation time for route i at time t, the first row T it is the output of the proposed prediction model: evacuation time of route i at time t. The remaining rows are the corresponding input variables, which include road geometric characteristics variables of route i, and time characteristics variables at time t as indicated in Table 1. After establishing the evacuation time prediction model, this section predicts the evacuation time along each route obtained in Sect. 2.1. For each building, the minimum of these predicted times is determined as the final evacuation time. Accordingly, these routes and shelters are the optimization evacuation routes and shelters.

14

Z. Han et al.

Table 1 Variable description Variable

Definition

T it

Evacuation time of the evacuation route i at time t

L

Length of the route i

W 1 –W 5

Different road widths in route i (W 1 : over 19.5 m, W 2 : 13–19.5 m, W 3 : 5.5–13 m, W 4 : 3–5.5 m, W 5 : under 3 m)

S 1 –S 7

Different road slopes in route i (S 1 : under − 9%, S 2 : − 9 to − 7%, S 3 : − 7 to − 5%, S 4 : − 5 to 5%, S 5 : 5–7%, S 6 : 7–9%, S 7 : over 9%)

R1 –R4

Different road curvature radii in route i (R1 : under 15 m, R2 : 15–60 m, R3 : 60–150 m, R4 : over 150 m)

Z 1 –Z 2

Percentage of road i within different zones (Z 1 : agriculture zone, Z 2 : forest zone)

IS

Number of straight intersections in road i

IL

Number of left turns intersections in road i

IR

Number of right turns intersections in road i

Sig

Number of signalized intersections in road i

TOD

Time of day (0–23, represents the time from 0:00 to 24:00 with every 1-h time step) at time t

DOW

Day of week (1–7, represents Monday through Sunday) at time t

To rigorously evaluate the prediction accuracy of the proposed model, the mean absolute error (MAE) and the mean absolute percentage error (MAPE) are selected as the evaluation criteria in this study. As calculated below, MAE measures the average difference between the actual and predicted values, while MAPE measures the percentage difference. These two indexes can indicate models’ predictive accuracy, where lower MAE and MAPE represent high accuracy of the prediction results. Additionally, to better measure the overall prediction performance, the distribution of absolute prediction time error (APTE) is also analyzed in this study. n 1  real pre  yi − yi  n i=1 pre  n  1   yireal − yi  × 100% MAPE = n i=1 yireal

MAE =

pre

(1)

(2)

Here, yireal and yi represent the actual and predicted values of evacuation travel time; n represents the number of measurements.

An Approach for Evacuation Vulnerability Assessment …

15

3 Case Study 3.1 Study Area and Data Collection The proposed method was applied to a real-world case study in Joso City to demonstrate its performance and applicability. Joso City, located in southwestern Ibaraki Prefecture of Japan (as shown in Fig. 1), has about 59,969 residents distributed over a total area of 123.64 km2 . In 2015, Joso City was hit by unprecedented rainfall during the Kanto–Tohoku heavy rainfall, which broke the levees along the Kinugawa River that runs through Joso City and inundated one-third of the city area [19, 20]. The major data types used in this study included travel time data, road network geometric information, and location data of building and evacuation shelters in Joso City. In our previous study [15], we collected travel time data through vehicle driving, which is accurate but limited in collected time and difficult to apply on a large scale. In this study, to address the remaining problems, the travel time data were collected from the Google Maps distance matrix application programming interface (API), a web-based service that provides travel distance and time [21–23]. The road network geometric information was downloaded from the Geospatial Information Authority of Japan (GSI), a national organization that conducts national surveying and mapping activities in Japan [24]. The location data of building were collected from the Zenrin company, a famous Japanese map publisher [25]. The location data of the evacuation shelter were collected from the web page of Joso City [26].

Fig. 1 Map of study area

16

Z. Han et al.

Fig. 2 Prediction results of the test routes from November 18, 2021, to November 24, 2021

Table 2 Performance of the prediction model Prediction accuracy MAE (min) MAPE (%)

Mon 0.79

Tue 0.68

Wed 0.62

Thu 0.8

Fri 0.69

Sat 0.78

Sun 0.76

8.29

7.68

7.22

8.69

7.64

8.31

8.17

APTE within 1 min (%)

73.95

73.95

73.95

73.95

73.95

73.95

73.95

APTE within 5 min (%)

99.82

99.82

99.82

99.82

99.82

99.82

99.82

3.2 Model Building and Prediction Accuracy Assessing This paper utilized 231 routes between evacuation shelters in Joso City to build the evacuation time prediction model. Then, travel time data of the selected routes were collected from November 18, 2021 (Thursday) to November 24, 2021 (Wednesday), with 1-h being the collection interval. Temporal and spatial variables were also created and accounted for model prediction. To further verify the effectiveness of the proposed model, the training and testing scheme was designed as follows: 80% (185 routes) of the historical data were used as train data, and the remaining 20% (46 routes) of the data were used as test data in the model. Figure 2 shows the prediction results of average evacuation time on the selected 46 routes from November 18, 2021, 00:00 to November 24, 2021, 23:00, which covers a full week. As listed in Table 2, the MAE and MAPE were relatively low in this prediction. Over the seven days, the average value of MAE is only 0.73 min with a MAPE of 7.99%, and about 99.82% of APTE were controlled within 5 min. Overall, the prediction accuracy of the proposed model was acceptable.

3.3 Evacuation Time Predicting in Joso City To further illustrate the applicability of the proposed method, this section applied it to a specific heavy rainfall scenario in Joso City. The scenario was established according

An Approach for Evacuation Vulnerability Assessment …

17

to the 2015 Japan Heavy Rainfall Event that occurred in Joso City, and the disaster was assumed to begin at 06:00 AM. The corresponding inundation area data divided the study area into 48,637 meshes with a size of 50 m * 50 m. As the inundation area data were updated at 10-min intervals, the time step of evacuation time prediction here was also set as 10 min. Then, the evacuation time of each building was predicted from 10 min before the disaster (05:50 AM) to 90 min at the disaster (07:30 AM). Figure 3 shows the predicted evacuation time at 05:50 AM (t = − 10 min). It can be seen there were no hard-to-evacuate buildings, and the longest evacuation time was 9.4 min. Therefore, if residents begin to evacuate at 05:50 AM (10 min before the disaster), all of them can evacuate to evacuation shelters before the disaster

Fig. 3 Evacuation time prediction results before the disaster time of 10 min

18

Z. Han et al.

occurs. Table 3 summarizes the basic statistics of predicted evacuation time at various disaster times t, including the percentage of evacuation time within 2 min, 4 min, 6 min, 10 min, and hard-to-evacuate observations. Figure 4 visualizes the inundation evolution process in the study area, and the time-varying evacuation time is also presented. It can be seen in Table 3 and Fig. 4 that since the disaster occurred, the rate of hard-to-evacuate buildings has remained stable at first, then sharply increased from t = 30 min, and finally reached 43.65% at t = 90. The main reason was that the inundation area rapidly expanded and spread from the Kinugawa River. As indicated by the black circle in Fig. 4, the inundation area trapped more buildings as the disaster progressed, making safe evacuation increasingly impossible. In addition, it can be found that the hard-to-evacuate buildings’ locations do not exactly coincide with the inundated areas. Some of these buildings that are not in the inundated areas seem temporarily safe, but it will be challenging to be evacuated safely due to the inundation of evacuation routes. It should be noted that the condition would become more serious with the disaster evolution. Thus, these areas should take special care and issue evacuation as early as possible. Overall, the predicted results explicitly reflected the time-varying evacuation time under this heavy rainfall scenario. The prediction maps clearly showed the evacuation time distribution and hard-to-evacuate buildings’ location. Correspondingly, local municipalities can accurately grasp the required evacuation time for each building, then provide a more scientific judgment of when and where to begin evacuation. As a result, the proposed evacuation time can be used to quantitatively assess vulnerability during the disaster evolution. Table 3 Basic statistics of predicted travel time data Evacuation time (min)

Within 2 min (%)

Within 4 min (%)

Within 6 min (%)

Within 8 min (%)

Within 10 min (%)

Hard to evacuate

t = − 10

14.22

55.41

97.68

99.99

100.00

0.00

t=0

14.20

54.72

96.78

99.09

99.09

0.91

t = 10

14.16

54.62

96.69

99.00

99.00

1.00

t = 20

13.88

53.54

95.51

97.82

97.82

2.18

t = 30

13.88

53.51

95.48

97.79

97.79

2.21

t = 40

13.61

52.80

93.11

95.18

95.18

4.82

t = 50

10.31

43.29

84.08

89.05

89.17

10.83

t = 60

9.59

40.50

79.57

84.41

84.54

15.46

t = 70

7.51

32.39

65.50

70.18

70.30

29.70

t = 80

5.41

26.90

55.99

60.45

60.57

39.43

t = 90

4.85

24.84

51.83

56.22

56.35

43.65

An Approach for Evacuation Vulnerability Assessment …

19

Fig. 4 Evacuation time prediction results at the disaster time of 0, 30, 60, and 90 min

3.4 Discussion In this paper, an evacuation prediction model is developed and further validated in Joso City. Table 2 shows that 73.95% of APTE are within 1 min, but still 0.18% are beyond 5 min. However, it should be noted that the available data set for training in this study is small. The prediction accuracy can be improved if more data are used for model training. In addition, when local municipalities need to make evacuation decisions, most of the weather information they referred is released every five or ten minutes. Thus, the performance is still acceptable even though some of the results appear to be inaccurate. In general, the proposed model accurately predicted the

20

Z. Han et al.

overall trend of evacuation time in each day of the week and each hour of the day. It performed well in predicting evacuation time for various temporal and spatial characteristics. In this paper, an empirical study based on real-world data from 2015 Japan’s Heavy Rainfall Event in Joso City is also carried out to demonstrate the application of the proposed method. The result has shown that the proposed method performed well in evacuation time prediction and hard-to-evacuate building identification, which can effectively reflect the time-varying vulnerability during the disaster evolution. There are two aspects this research will focus on in future studies along this research line. First, this study only considers the temporal and spatial correlations when predicting travel time on evacuation routes. Objectively, time is determined not only by temporal–spatial characteristics but also by external conditions such as weather and individual behavior. Therefore, further studies will incorporate these conditions to predict evacuation time more accurately. Second, as previously stated, the transport mode during evacuation is assumed as a private vehicle, but in reality, residents will also evacuate on foot. Therefore, another critical study in future is to estimate evacuation conditions under different transport modes and further assess the vulnerability.

4 Conclusion This study proposed an evacuation time prediction method to quantify vulnerability, in which a Dijkstra’s algorithm and a random forest regression algorithm are applied. The case study conducted in Joso City reveals the validity and practicality of this method. The proposed method can not only find the evacuation route under timevarying inundation areas, but also predict corresponding evacuation time considering route spatial and temporal characteristics. The predicted results and maps can provide a quantitative basis and further assist local municipalities in making evacuation decisions. Acknowledgements This work was supported by Council for Science, Technology and Innovation (CSTI), Government of Japan, Cross-ministerial Strategic Innovation Promotion Program (SIP), “Enhancement of National Resilience against Natural Disasters” (Funding agency: National Research Institute for Earth Science and Disaster Resilience), and JST SPRING, Grant Number JPMJSP2136.

References 1. Garner AJ, Mann ME, Emanuel KA, Kopp RE, Lin N, Alley RB, Horton BP, DeConto RM, Donnelly JP, Pollard D (2017) Impact of climate change on New York City’s coastal flood hazard: increasing flood heights from the preindustrial to 2300 CE. Proc Natl Acad Sci 114:11861–11866

An Approach for Evacuation Vulnerability Assessment …

21

2. Jha AK, Bloch R, Lamond J (2012) Cities and flooding: a guide to integrated urban flood risk management for the 21st century. World Bank Publications (2012) 3. Wang X, Kinsland G, Poudel D, Fenech A (2019) Urban flood prediction under heavy precipitation. J Hydrol 577:123984. https://doi.org/10.1016/j.jhydrol.2019.123984 4. Hosseini O, Maghrebi M, Maghrebi MF (2021) Determining optimum staged-evacuation schedule considering total evacuation time, congestion severity and fire threats. Saf Sci 139:105211. https://doi.org/10.1016/j.ssci.2021.105211 5. Lin P, Lo SM, Huang HC, Yuen KK (2008) On the use of multi-stage time-varying quickest time approach for optimization of evacuation planning. Fire Saf J 43:282–290 6. Lindell MK, Prater CS, House DH (2022) Cascadia subduction zone residents’ Tsunami evacuation expectations. Geosciences 12:189. https://doi.org/10.3390/geosciences12050189 7. Nasiri H, Mohd Yusof MJ, Mohammad Ali TA (2016) An overview to flood vulnerability assessment methods. Sustain Water Resour Manag 2:331–336 8. Rehman S, Sahana M, Hong H, Sajjad H, Ahmed BB (2019) A systematic review on approaches and methods used for flood vulnerability assessment: framework for future research. Nat Hazards 96:975–998 9. Blaikie P, Cannon T, Davis I, Wisner B (2014) At risk: natural hazards, people’s vulnerability and disasters. Routledge, London 10. Proceedings of the seminars on flood vulnerability analysis and on the principles of floodplain management for flood loss prevention (1984) Presented at the seminar on flood vulnerability analysis (1982: Bangkok) 11. Dijkstra EW (1959) Others: a note on two problems in connexion with graphs. Numer Math 1:269–271 12. Zhu Y, Li H, Wang Z, Li Q, Dou Z, Xie W, Zhang Z, Wang R, Nie W (2022) Optimal evacuation route planning of urban personnel at different risk levels of flood disasters based on the improved 3D Dijkstra’s algorithm. Sustainability 14:10250. https://doi.org/10.3390/su141610250 13. Breiman L (2001) Random forests. Mach Learn 45:5–32 14. Liu Y, Wu H (2017) Prediction of road traffic congestion based on random forest. In: 2017 10th international symposium on computational intelligence and design (ISCID), pp 361–364 15. Kohei K, Daichi I, Takumi S, Yuko Y, Hisatoshi T, Yasuhiro M (2021) Study on decision making support of evacuation guidance with a novel estimation of evacuation time considering lead time of disaster strike. Proc Inst Soc Saf Sci 39:401–409 16. Cheng J, Li G, Chen X (2019) Research on travel time prediction model of freeway based on gradient boosting decision tree. IEEE Access 7:7466–7480 17. Qiu B, Fan W (2021) (David): machine learning based short-term travel time prediction: numerical results and comparative analyses. Sustainability 13:7454. https://doi.org/10.3390/su1313 7454 18. Sharmila RB, Velaga NR, Kumar A (2019) SVM-based hybrid approach for corridor-level travel-time estimation. IET Intell Transp Syst 13:1429–1439 19. Ohki M, Tadono T, Itoh T, Ishii K, Yamanokuchi T, Watanabe M, Shimada M (2019) Flood area detection using PALSAR-2 amplitude and coherence data: the case of the 2015 heavy rainfall in Japan. IEEE J Sel Top Appl Earth Obs Rem Sens 12:2288–2298 20. Levee collapse of the Kinu river: How did Joso city residents evacuate? https://www.nhk.or.jp/ bunken/english/research/domestic/20160801_6.html. Last accessed 13 Oct 2022 21. Get started with the distance matrix API. https://developers.google.com/maps/documentation/ distance-matrix/start. Last accessed 13 Oct 2022 22. The bright side of sitting in traffic: crowdsourcing road congestion data, https://googleblog.blo gspot.com/2009/08/bright-side-of-sitting-in-traffic.html. Last accessed 13 Oct 2022 23. Goudarzi F (2018) Travel time prediction: comparison of machine learning algorithms in a case study. In: 2018 IEEE 20th international conference on high performance computing and communications; IEEE 16th international conference on smart city; IEEE 4th international conference on data science and systems (HPCC/SmartCity/DSS), pp 1404–1407 24. About GSI | GSI HOME PAGE. https://www.gsi.go.jp/ENGLISH/page_e30003.html. Last accessed 13 Oct 2022

22

Z. Han et al.

25. ZENRIN. https://www.zenrin.co.jp/english/index.html. Last accessed 13 Oct 2022 26. Location of evacuation shelters/ Joso City Homepage. http://www.city.joso.lg.jp/jumin/anzen/ bosai/1419259176253.html. Last accessed 13 Oct 2022

Development and Applicability Assessment of a Tunnel Face Monitoring System Against Tunnel Face Collapse Ryohei Hase, Yasuji Mihara, Dohta Awaji, Rieko Hojo, and Shoken Shimizu

Abstract To avoid damage caused by rock collapse in tunnel faces, it is necessary to sound an alarm before a rock collapse occurs so that workers have time to evacuate from the tunnel face. However, the area around the tunnel face is a confined workspace where multiple workers and machines are mixed together, and visual monitoring by observers alone is not sufficient to ensure safety and time. To solve these problems, we have developed a system to identify cracks on the surface of shotcrete using deep learning based on semantic segmentation, a method of image analysis, to realize constant monitoring as a technology to support monitoring and judgment (Tunnel Face Monitoring System) (Uke et al. in ITA-AITES World Tunnel Congress, 2022 [1]). At mountain tunnel construction sites where this system is applied, work behaviors such as crack observing are likely to change depending on the work environment and surrounding conditions. Therefore, the observation of tunnel face was used as the target task to verify the effect of environmental changes on crack detection time. This paper reports on the AI construction method and the usefulness of this system, which was verified quantitatively and objectively in a demonstration case conducted with the budget of a project commissioned by the Ministry of Economy, Trade and Industry (METI) in 2021. Keywords Tunnel face monitoring supportive system · Virtual reality (VR) · AI model · Tunnel construction

R. Hase (B) · Y. Mihara · D. Awaji Shimizu Corporation, 2-16-1, Kyobashi, Chuo-Ku, Tokyo, Japan e-mail: [email protected] R. Hojo Nagaoka University of Technology, 1603-1, Tomioka-Cho, Nagaoka, Niigata, Japan S. Shimizu National Institute of Occupational Safety and Health, Japan, 1-4-6, Umezono, Kiyose, Tokyo, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_3

23

24

R. Hase et al.

1 Introduction To avoid danger from rockfall at tunnel construction sites, it is necessary to sound an alarm before a rockfall occurs and allow time for workers to evacuate the tunnel face. Traditionally, cracks have been determined as a source of danger by human visual inspection. However, there are many areas where visual inspection is difficult due to safety, time constraints, and economic burdens, such as the use of temporary scaffolding and elevating vehicles. As a solution, we have developed the Tunnel Face Monitoring System, which uses an AI-equipped camera to assist in identifying cracks on the surface of shotcrete [1]. Figure 1 shows an overview of the Tunnel Face Monitoring System. Below is a description of the functions of the Tunnel Face Monitoring System and the construction of AI-based crack detection.

1.1 Crack Detection Function Cameras are installed on the sidewalls near the tunnel face so that images of the tunnel face can always be acquired when workers need to enter the tunnel face area. By connecting the camera images to an on-site analysis PC (in the control panel) for real-time analysis, cracks on the surface of shotcrete, one of the precursor phenomena to tunnel face collapse, can now be constantly monitored and detected as soon as they occur. Video information is analyzed by an AI-equipped PC and projected on a screen such as a tablet, allowing the user to observe the shotcrete surface of the tunnel face with their own eyes and quickly confirm crack detection by comparing the analysis results with video taken from a blind spot [1]. Cracks in the shotcrete surface can be detected by real-time analysis of the video being filmed using AI that has learned crack patterns through deep learning. There are two advantages to applying deep learning:

Alarm equipment

Erector or Drilling jumbo operator

Worker

Al-Enabled Camera

Observer Tunnel face monitoring supportive system control panel

Fig. 1 Conceptual diagram of the tunnel face monitoring supportive system

Development and Applicability Assessment of a Tunnel Face …

25

. It is resistant to noise such as people and machines entering the video. . It can be systematized with simple equipment, and the calculation process can be completed on-site. On the other hand, it should be noted that the reliability and validity of the detected cracks must be fully verified.

1.2 Alarm Transmission Function When a crack is detected, which is a predictive sign of tunnel face collapse, an alarm can be quickly issued to alert workers to evacuate. Since cracks need to be notified to workers quickly and reliably, LED lights are used to illuminate the tunnel surface clearly, and alarm sounds are used to provide a broad and clear warning. Even under noisy conditions (blower, dust collector, heavy machinery, etc.) at the tunnel site, the light and alarm sound reliably convey crack detection information to workers.

1.3 Construction of AI to Detect Cracks in Shotcrete and Crack Detection Performance The system is capable of identifying cracks on shotcrete surfaces using a learning method based on deep learning, a type of AI. To detect cracks from captured images, the system employs an image analysis method based on semantic segmentation, which is one of the learning methods using deep learning. In [1], to select the optimal network, models trained with each of the three semantic segmentation methods, ContextNet [2], PixelNet [3], and BiSeNet [4], were created with the same training and validation data, and the accuracy on the validation data was compared. PixelNet was selected as the AI model due to its high accuracy for cracks on the surface of shotcrete [1]. The AI model was used to detect cracks in images of shotcrete collapse during hole through of tunnel. As a result, it was confirmed that the AI model trained with the drawn cracks had sufficient detection accuracy for actual cracks. Based on these results, we believe that the AI model used in this system can immediately detect cracks even in situations where workers and heavy machinery move in complex ways, such as near tunnel faces.

26

R. Hase et al.

2 Methodology 2.1 Evaluation Test of Crack Detection Capacity Using Crack Occurrence Video Videos of cracks drawn on tunnel faces under various environmental conditions were made, and the time it took to recognize the cracks was measured. The environmental conditions were time of crack occurrence, location, and brightness (see Table 1). A total of 27 videos (5 min each) were created using a combination of these conditions and were randomly and continuously displayed on a monitor screen (80 in.) setup in a room with a brightness of 20 Lx. The number of pixels and transparency of cracks in the movies increased with time (number of pixels: 1–4, transparency: 10– 100%) to simulate crack growth. Five subjects (indicated by “*” in Table 2) aged 25– 48 years (2–26 years of service) with experience working at tunnel construction sites participated in the experiment. The experiment was conducted under the assumption that each subject was observing the tunnel face on the screen as the observer. The video was stopped when the subject recognized a crack on the tunnel face in the video, and the time displayed on the screen was recorded as the crack detection time. Table 1 Environmental conditions varied in the experiment

Crack initiation time (min)

1.0

2.5

Crack initiation location

Upper left Crown

Video brightness (%)

0

− 25

3.5 Upper right − 50

Table 2 Tunnel site experience and job positions of the subjects Subject ID Field experience and job position A

Experienced two mountain tunnel sites as supervisor (in charge of geology)

B*

Experienced two mountain tunnel sites as supervisor (in charge of excavation for about 1 year)

C

Chief construction site manager (six sites, total excavation length approx. 6 km)

D*

Tunnel site supervision experience but excluding excavation supervision

E

Chief construction site manager (six sites, total excavation length approx. 7 km)

F*

Experienced one mountain tunnel site as supervisor (in charge of excavation for about half year)

G*

Chief construction site manager (four sites, total excavation length approx. 3 km)

H*

Head of construction site (ten sites, total excavation length approx. 7 km)

Development and Applicability Assessment of a Tunnel Face …

27

2.2 Method for Evaluating the Applicability of Introducing a Tunnel Face Monitoring System Using VR Experiments The experiment was conducted in a construction simulation facility (width: 15.2 m, length: 33.2 m, height: 10 m) at the National Institute for Occupational Safety and Health, Japan. Subjects aged 25–48 years, with 2–26 years of work experience working for a construction company (see Table 2 for field experience and job category), participated in the experiment. This experiment was conducted to detect cracks on the surface of shotcrete using VR that reproduced the site at the time of erecting steel supports [Fig. 2(4)]. A tunnel face 15.3 m wide and 9.8 m high was virtually set up at a distance of 26 m from the starting line [Figs. 2(1) and 2(2)], and erector [3.3 m wide × 12 m long (excluding the arm) × 7.5 m high] was placed in the center of the front face of the tunnel face [Fig. 2(3)]. The experiment consisted of detecting five types of cracks (center (diagonal), center (horizontal), crown, upper left, and upper right) that appeared on the shotcrete surface of the tunnel face in a VR that simulated the environment in the tunnel site (steel support erection). The following four conditions were set up by the Tunnel Face Monitoring System support method, and the above cracks at arbitrary locations were detected under each condition: (1) visual monitoring (without the Tunnel Face Monitoring System); (2) assuming an AI-equipped camera was installed on the right sidewall; (3) assuming an AI-equipped camera was installed on the left sidewall; (4) assuming an AI-equipped cameras was installed on both the right and left sides. The same Tunnel Face Monitoring System was used as in the video-based experiment. The crack detection time for this system was calculated from the video images of cracks on the tunnel face used in the VR experiment, and the calculated detection time was reflected in the VR experiment. In addition to the five types of cracks described above, two types of pseudo-cracks (cracks that are detected by this system and alerted by the

Fig. 2 Situation in the experimental facility (1, 2) and in the VR virtual environment (3–5)

28

R. Hase et al.

tablet but do not appear on the tunnel face in the VR) were added to avoid excessive reliance on this system under conditions where this system assists in observing. By incorporating random pseudo-cracks to bring the environment in the VR closer to the actual site environment, this experiment confirmed the effectiveness of implementing this system as an adjunct to tunnel face observation, rather than relying entirely on this system. In the experiment, the time from the start until the crack was detected and the location where the subject detected the crack were measured.

3 Results and Discussion 3.1 Crack Detection Time in Experiments Using Video

-25%

-50%

Brightness

0%

Crack detection time Crack detection time Crack detection time (min.) (min.) (min.)

Figure 3 shows the crack detection time results for the five subjects (B–H) under each environmental condition. It can be seen that the crack detection time for both the system and the subjects changed significantly with changes in brightness. It was observed that as the brightness decreased, the crack detection time was extended regardless of the location of the crack. This trend was similar for both the subjects and the system, indicating that the system retains the same visual and discriminative abilities as the subjects. On the other hand, there was no significant relationship between crack initiation time and crack detection time. Since this experiment used a short video of about 5 min, it is thought that the subjects’ concentration on observing the tunnel surface did not have a significant effect on the crack detection time. 1.2 1 0.8 0.6 0.4 0.2 0 1

2.5

3.5

1

2.5

3.5

1

2.5

3.5

1

2.5

3.5

1

2.5

3.5

1

2.5

3.5

1.2 1 0.8 0.6 0.4 0.2 0 1.2 1 0.8 0.6 0.4 0.2 0

AI Subjects : B D F G H Location of crack occurrence : Upper left Crown Upper right

1

2.5

3.5

1

2.5

3.5

1

2.5

3.5

Crack initiation time (min.)

Fig. 3 Crack detection time versus crack initiation time for each environmental condition

Development and Applicability Assessment of a Tunnel Face …

29

3.2 Comparison of Rack Recognition Performance Between Subjects and This System To compare the crack detection times of the subject and this system, the rate of change was calculated based on the crack detection time of this system expressed in Eq. (1). (Rate of change based on the system) =

(Subject’s crack detection time) (System’s crack detection time)

(1)

Rate of change relative to the crack detection time of this system

To compare the crack detection times of the subject and this system, the “rate of change based on the Tunnel Face Monitoring System” was calculated by dividing the crack detection time of the subject by the crack detection time of this system. The relationship between this rate of change and brightness is shown in Fig. 4. There was no difference in the rate of change based on this system due to differences in brightness. Focusing on the location of the cracks, the average rate of change was about 0.8 for the cracks on the upper left, indicating that the subject detected the cracks earlier than this system. On the other hand, for the upper right and crown cracks, the average rate of change was greater than 1.0, indicating that the subjects were slower to detect cracks than this system. In particular, for the crown crack, the system detected it about 1.3 times faster than the subject. All cracks drawn in this experiment grew to approximately the same width and length over time. In addition, the system observed the tunnel face uniformly and detected cracks in almost constant time. And the subjects monitored the entire tunnel face with almost no eye movement, suggesting that they were observing under the same conditions as the system. From the above, it can be inferred that the subjects were affected by the pattern (luminance difference) of the shotcrete around the location of the cracks. Although the detection time varied depending on the location of cracks, this system is considered to have the same monitoring capability as an observer subject. Subjects: 3

B

D

F

G

H 3.5

3.5

3.5

Upper Left

2.5

3

Crown

3

2

2

2

1.5

1.5

1.5

1

1

1

0.5

0.5

0.5 0

0

0 0

0

0 -25 -25 -25 -50 -50 -50

Brightness

Upper Right

2.5

2.5

0

0

0 -25 -25 -25 -50 -50 -50

Brightness

0

0

0 -25 -25 -25 -50 -50 -50

Brightness

Fig. 4 The rate of change based on this system due to differences in brightness for each location of crack occurrence

30

R. Hase et al.

Luminance

200 160 120 80

200 160 120 80 40

40 150 110

Pixel

70

30

0

160

200 pixel

120 80 40 200 160 120 80 40 0

Upper Right

150 pixel

200

Original image

Luminance

200 pixel

240

Crown

150 pixel

Original image

Luminance

150 pixel

Upper Left

240 200

Original image

200 pixel

160 120 80 40 200 160

30

70

Pixel

110 150

120 80 40 0

30

70

110

150

Fig. 5 Luminance distribution around each crack

3.3 Effect of Drying Condition of Shotcrete at Crack Initiation Site on Crack Detection Time In the experiments using video images, subjects tended to detect cracks in the crown area the slowest, followed by the upper right area and then the upper left area. This tendency may suggest that the luminance of the shotcrete surface affected the subjects’ crack detection time. Therefore, we used ImageJ [5] to obtain the luminance distribution for a fixed area (W:200pixel × H:150Pixel) near the cracks used in this experiment. (Fig. 5). The results showed that there were luminance differences around the cracks that occurred in the crown and the upper right part, and that it took time for the subjects to detect the cracks. In particular, the crack at the crown, which took the longest time to detect, had a larger overlap length between the crack and the area with the luminance difference. This suggested that more time was required for the observer to detect cracks in the areas with luminance differences due to differences in the drying conditions of the shotcrete surface. On the other hand, under the condition of ± 0 video brightness, the crack detection time of this system was the slowest for cracks occurring on the upper left, followed by cracks on the upper right and crown, suggesting that the influence of the difference in luminance of the shotcrete surface is small. These results suggest that this system may be able to detect cracks earlier than the observer under conditions with luminance differences, such as differences in the dryness of the shotcrete surface.

3.4 Observation Density Since the rate of change based on this system in the VR experiment ranged from 1.6 to 6.4 with the observation assisted by this system, the crack detection time of this system in the VR experiment was much faster than the crack detection time of the subjects. Regarding the judgment of pseudo-cracks, most subjects correctly judged the cracks as pseudo-cracks (94% correct rate of pseudo-crack judgment), rather than

Development and Applicability Assessment of a Tunnel Face …

31

completely trusting the results of this system, indicating that the subjects actually judged the cracks visually. Therefore, it can be inferred that under the conditions of the VR experiment, the subjects were intensively visually checking the crack detection points warned by the system. Based on the above discussion, it can be considered that the crack detection time by this system includes the fastest time that the subject was able to recognize a crack, and from this time, the time required to observe the entire tunnel face can be calculated according to the observation distance of the tunnel surface. The time to detect cracks was found to vary from subject to subject. This may be due to the fact that the time to detect a crack varies depending on the movement of the observer’s gaze. Figure 6 is a conceptual diagram of the time required to observe the tunnel face. As the distance from the crack to the observer increases, the detection time is expected to increase. The earliest the crack is detected after it is visible tends to be the lower limit and the latest the upper limit. The difference between the upper and lower limits is considered to be the time required to observe the entire tunnel face at the distance from the crack to the tunnel face. The area of the tunnel face that can be observed in 1 s is the observation density, and the time required to observe the entire tunnel face (◿T = T1−L − T2−L ) was used to calculate the observation density. Figure 7 shows the relationship between the distance from the subject to the crack and the crack detection time obtained in this experiment. The dashed lines in the figure are the approximate lower (blue dashed line) and upper (red dashed line) limits at each distance. These trends indicate that as the distance between the subject and the crack increases, the crack detection time increases. The approximate straight line for the lower limit of the crack detection time is highly dependent on the detection time when the system is installed, while the approximate straight line for the upper limit is highly dependent on the detection time when the system is not installed. Using these approximate lines, the observation density was calculated as in Eq. (2).

Fig. 6 Image of the observer’s line of sight during monitoring and conceptual diagram of time required for tunnel face observation

32

R. Hase et al.

Fig. 7 Distance to crack versus crack detection time

(Observation area of tunnel face) (Time required to observe the entire tunnel face) (Observation area of tunnel face) = (Upper limit) − (Lower limit) = S/1.0L + 50 (2)

(Observation Density) =

Here, S is the observed area of the tunnel face (m2 ) and L is the distance from the crack to the subject (m). The area of the tunnel face used in this experiment was 72.8 m2 , and this value can be applied to Eq. (2) to obtain the observation density depending on the distance from the crack to the subject. If the distance from the crack to the subject is 10.5 m, the observed density is calculated to be 1.2 m2 /s. On the other hand, in order to calculate the observation density by this system, it is necessary to exclude areas that are blind spots. In this experiment, the area observed by this system, taking blind spots into account, was 45.9 m2 , or approximately 50% of the tunnel face area. The time from the transmission of the video taken by this system to the detection of a crack is 3.5 s. Substituting these values into Eq. (2) to obtain the observation density of this system, a value of 13.1 m2 /s was obtained. This value is approximately 10.9 times greater than that of the subject, indicating the usefulness of this system.

3.5 Risk Reduction Effects of Implementing a Tunnel Face Monitoring System The risk reduction effect of this system with and without assist was confirmed using the crack detection time. Figure 7 shows the comparison results of this system with and without assist. The results show that this system reduces the crack detection time by about 16 s at 10.5 m, the average observation distance from the subject to the crack, because the area to be monitored is fixed. Since the typical jogging speed of an adult male is said to be 8 km/h, a 16-s reduction in crack detection time translates to a distance of about 36 m from the tunnel face. Rock collapse at the tunnel face is basically a rock collapse phenomenon, and under the conditions of

Development and Applicability Assessment of a Tunnel Face …

33

the VR experiment, approximately 15.3 m (1D) is considered to be the impact zone. Therefore, installing a Tunnel Face Monitoring System with observation capabilities equal to or greater than that of an observer is expected to improve worker safety by allowing workers to be well away from the area of impact.

4 Conclusion The results of the video experiment confirmed that the Tunnel Face Monitoring System was capable of continuous observation and had the same observation capability as a subject under intensive observation, although the detection time varied slightly depending on the location of cracks. Authors confirmed the possibility that the Tunnel Face Monitoring System could detect cracks earlier than the observer, especially when there were differences in luminance, such as when the shotcrete was not yet dry. In a VR environment simulating the actual tunnel site, the system was able to observe the tunnel face at an observation density 10.9 times greater than that of the subject playing the role of observer, confirming the high usefulness of the system based on the results of the VR experiment. In addition, it was confirmed that the introduction of the Tunnel Face Monitoring System, which assists in crack detection, can reduce the time it takes for the observers to detect a crack by approximately 16 s. It was also found that if evacuation instructions can be issued immediately after crack detection, the possibility of avoiding dangers such as tunnel collapses increases.

References 1. Uke S, Awaji D, Hemmi R, Ihara H, Mihara Y (2022) Development of AI crack detection system for alert against tunnel face collapse. In: ITA-AITES world tunnel congress 2. Rudra PP, Ujwal B, Stephan L, Christopher Z (2018) Contextnet. Exploring context and detail for semantic segmentation in real-time. cs.CV. 5. https://doi.org/10.48550/arXiv.1805.04554 3. Changqian Y, Jingbo W, Chao P, Changxin G, Gang Y, Nong S (2018) Bisenet: bilateral segmentation network for real-time semantic segmentation. cs. CV. https://doi.org/10.48550/arXiv.1808. 00897 4. Aayush B, Xinlei C, Bryan R, Abhinav G, Deva R (2017) Pixelnet: representation of the pixels, by the pixels, and for the pixels. cs. CV. https://doi.org/10.48550/arXiv.1702.06506 5. Caroline AS, Wayne SR, Kevin WE (2012) NIH ImageJ: 25 years of image analysis. Nat Methods 9:671–675

Development of Real-Time Measuring System of Tip Position with Deep Mixing Methods Hong-Son Nguyen, Yuji Adachi, Minoru Yamada, Toshiaki Takaue, Masato Kobayashi, Moriyasu Takada, and Shinya Inazumi

Abstract The accuracy control during the construction in the soil improvement work was mainly the empirical control of the construction manager such as attitude of the construction machine and experience of the operator. There is a growing need for technology that able to measure and control the track at the tip of a mixing tool in real time and with high accuracy in deep construction or construction in close proximity to existing structures. The authors developed a deep depth measurement system capable of measuring the tip position of a mixing tool in real time during construction for the deep cement mixing method and applied it to actual construction work for the first time. In this system, a group of biaxial inclinometer, communication infrastructure, and battery installed inside the casing rod and the underwater and underground set of communication equipment made it possible to share the tip position information acquired by the operator and related staffs through wireless communication without affecting the rod connection as well as disassemble work. In the actual construction depth of 43.7 m deep, the amount of eccentricity at the time of construction relative to the design position was about 40 cm or less, which was within the range of the construction error initially assumed, and the construction was completed safely without affecting the existing underground drainage structure. This article is going to describe the detailed of system configuration and present to measurement results of the first applied construction site. Keywords Tip position · Real-time measurement · Wireless communication · Deep cement mixing

H.-S. Nguyen (B) · Y. Adachi · M. Yamada Hazama Ando Corporation, Tokyo, Japan e-mail: [email protected] T. Takaue · M. Kobayashi Aoyama Kiko Co. Ltd., Tokyo, Japan M. Takada System Construction Co. Ltd., Tokyo, Japan S. Inazumi Shibaura Institute of Technology, Tokyo, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_4

35

36

H.-S. Nguyen et al.

1 Introduction In recent years, there has been an increase in foundation works, including largescale ground improvement, for the widening and extension of railways and highways, the construction of railway for new linear-central bullet trains, and for disaster prevention purposes. Particularly in deep construction, where the construction depth exceeds 40 m, there is concern that the quality of the ground improvement, such as bearing capacity, deformation control and impermeability, may deteriorate if the drilling trajectory deviates from its designed position. In addition, when construction is carried out in close proximity to buried structures in service, it is also necessary to ensure a certain separation. For these reasons, there is a growing need for technology that enables real-time visualization, sharing and analysis of various construction information, including the tip position of the drilling rod in the ground. Conventionally, it was common to measure the verticality of the visible rod on the ground and calculate the tip position assuming the same angle for the underground part. Some attempts to use tilt sensor was installed at the lower end of the rod, and all paths were wired for communication using a fixed tube inside the rod, but there were problems with waterproofing and reduced workability due to the work required to connect the communication cable at the rod joint. Furthermore, as the construction depth increases, the number of joints increases, and there are problems that affect the construction workability, such as increasing the connection work at the joints. The authors have developed a deep tip measurement system that can measure the tip position of the mixing tool in real time during construction for the relative stirring deep mixing method, which enables construction at great depths [1]. The system aims to enable operators and construction workers to control the tip position of the mixing tool in real time during operation by means of a biaxial inclinometer, battery installed inside the casing rod of the construction equipment and a series of communication devices. This paper describes the features of the system and examples of its application to actual construction work.

2 Overview of Cement Deep Mixing Method The deep cement stabilization method (DCS Method) [2] (hereinafter referred to as this method) is a deep cement mixing method using a relative mixing function to forcibly mix the original ground and solidified cementitious material while discharging slurry cement or cementitious solidified material into the ground. This is a deep cement mixing method that creates a homogeneous large-diameter improvement column. Figure 1 shows an overview of the construction machinery and the shape of the mixing tool. The mixing tool at the tip of the machine are of the relative mixing type, whereby the outer and inner blades rotate in opposite directions, preventing the ground from twisting and turning together, and enabling the creation of high-quality improved bodies.

Development of Real-Time Measuring System of Tip Position …

37

Fig. 1 Construction machine and mixing blade of the relative deep mixing method

This is a deep cement mixing method for large diameters and depths, suitable for hard ground with a gravel diameter of up to 300 mm and an SPT (N) value of less than 40 that is located above the soft layers or sandwiched between soft ground layers. Maximum construction depth is 50 m, and improvements are made by dividing the construction into several casing rods. While attaching sensors to the casing rods has been considered, problems with data communication and construction workability have hindered practical application.

3 Configuration and Features of the Measurement System Figure 2 shows the measurement principle of the system and the data wireless communication method. This method can be used to improve the ground to a maximum depth of about 50 m by connecting 10-m casing rods. A biaxial inclinometer was installed for each of these rods. The inclination was measured for each casing rod, and the drilling position in the direction of depth was calculated from the casing rod length and the amount of inclination. The inclinometer was placed in a storage box inside the rod as an integrated unit with a communication infrastructure and battery. The communication infrastructure was used for wireless communication between the nodes in the 920 MHz radio frequency band. The communication method for communicating the measured data of the inclinometer was by means of a communication cable wired inside the casing rod and wireless communication via an antenna at the joints of the casing rod. Data are delivered sequentially from the deepest part to the shallowest communication base (multi-hop communication).

38

H.-S. Nguyen et al.

The uppermost communication unit is always located above ground, from where it is wired to a dedicated data processing terminal in the operator’s room. The success of underground and underwater wireless communication at the joints of the casing rods, which was previously considered difficult, has enabled stable measurement and data communication without being affected by the casing rod preaching operation. Figure 3 shows the situation of the antennas installed at the joints (convex and concave sections). The joints are inserted and fixed using special bolts. The distance between the antennas is 7 mm when the joint is inserted. A full-scale underwater wireless communication experiment was performed, and the results indicated that stable wireless communication is possible if the separation between the antennas is less than 60 cm in freshwater [1]. Figure 4 shows an overview of the construction machine equipped with this system. It consists of a casing storage box, a ground communication device (transmitter/receiver) and a global navigation satellite system (GNSS) antenna. The battery is intended to provide approximately one month of continuous use during construction. This system can be linked to the Pile and Ground Improvement Construction Information Visualization System (3D Pile Viewer) [3, 4], which has already developed since 2016, to enable real-time visualization of the amount of displacement at the mixing tool’s tip position relative to the design coordinates of the soil–cement

Fig. 2 Principles of measurement and data communication method

Development of Real-Time Measuring System of Tip Position …

(a) Convex part of joint

Fig. 3 Antenna mounted on the joint

Fig. 4 Construction machine equipped with this system

(b) Concave part of joint

39

40

H.-S. Nguyen et al.

column. This displacement amount is determined by adding the amount of surface displacement caused by the slope of the construction machine and the construction position to the amount of underground displacement calculated from the inclinometers installed in the casing rods. If the construction control values are exceeded, the operator and those involved in the construction can be notified and anomalies can be identified at an early stage, thereby reducing risks in construction, and ensuring quality.

4 Applicable Example The system was first applied to a large-depth ground improvement project using deep cement stabilization method. The standard cross-section of this work is displayed in Fig. 5 and the general plan view in Fig. 6. The improvement work is to increase the shear strength of the existing high landfill of approximately 83 m height that involves of deposition of soft silt (Fc1, Fc2), sandy soil (Fs), gravelly soil (Fg), and existing embankment (Bg), with the aim of seismic reinforcement and stabilization of the landfill. The diameter of the ground improvement body was 1600 mm and was carried out in two areas: the upper section and the lower section. The number of improved columns was approximately 1400 in total, making this a large-scale ground improvement project. The maximum improvement depth of the lower section was 43.7 m, which was a large-depth construction. In addition, as the lower end of the lower section of the improvements passed through the existing in-service drainage tunnel, it was necessary to ensure a certain distance between the top and side of the tunnel to avoid affecting this tunnel. A plan view of the lower section of the improvement and the existing tunnel section is shown in Fig. 7. A distance of 1.4 m at the side and 1.0 m at the top was established as the range of influence of the ground improvement construction on the drainage tunnel, and this newly developed system was used to control the mixing tool so that they did not enter the tunnel within this range. The construction machines applied in this construction work are shown in Fig. 4. The management screens in the construction control room located inside the project area and the machine operator’s room are shown in Fig. 8. The system uses dedicated software to display the posture and displacement of the casing rod during construction in real time. The underground measurement results are combined with the ground surface displacement using a 3D Pile Viewer to display the trajectory of the tip position displacement relative to the design coordinates, and the information can be shared in real time with the construction related people in charge on a browser. Figure 9 shows the 3D Pile Viewer screen visualizing the construction status of the improvement columns near the buried drainage tunnel in service. The coloring of the improvement columns represents the electrical current value of the rotary drive motor, which indicates the rotational load generated during penetration procedure. The redder the color, the higher the resistance of the ground. The red frame at the

Development of Real-Time Measuring System of Tip Position …

41

Fig. 5 Standard cross-section of the construction works

Fig. 6 General plan of the construction works

bottom of the 3D Pile Viewer screen shows that the tunnel is being constructed at a specified distance from the tunnel in service. Figure 10 shows a graph of the horizontal displacement at the tip of the improvement columns adjacent to the tunnel in service. In this and other deep cement mixing

42

H.-S. Nguyen et al.

Fig. 7 Plan view of improvement columns and drainage tunnel

(a) Dedicated PC management screen

(b) 3D Pile View screen to check tip position

Fig. 8 Status of tip position monitoring via management screen

methods, the vertical inclination during construction is empirically said to be less than 1/100 of the depth. Based on this, a control standard value of 40 cm or less was set for the allowable horizontal displacement at a construction depth of 40 m. Based on the measurement results, the displacement at the tip position of the composed

Development of Real-Time Measuring System of Tip Position …

43

Fig. 9 Real-time visualization with 3D Pile Viewer

component of all the improvements could be managed so that the displacement was within the control standard value. In addition, the construction could be completed safely without any impact on the existing buried waterway tunnel due to the ground improvement construction.

44

(a) Distribution with dept in X, Y direction depth

H.-S. Nguyen et al.

(b) Plan view of tip positions at final

Fig. 10 Measured displacement of improvement columns tip adjacent to the drainage tunnel

5 Conclusion A large-depth tip measurement system equipped with a high-precision inclinometer and underground/underwater wireless communication function has been developed and applied to large-depth deep cement mixing ground improvement construction. The system enables the position of the mixing tool’s tip during construction to be confirmed with high accuracy and enables real-time management of deviations between the actual construction position and the design position, and the displacement from the existing buried structures. The system will continue to increase the number of application cases and contribute to construction risk reduction and more efficient construction management. The system will also be expanded to the others ground improvement works, such as piling and excavation retaining wall, etc. Efforts are being made to modify the technology so that the system can contribute to ensuring the workmanship and quality of works involving deep excavation, improving the reliability of construction management, and increasing the efficiency of on-site operations, and the use of the system in automated construction is also being considered.

Development of Real-Time Measuring System of Tip Position …

45

References 1. Nguyen HS, Adachi Y, Yamada M, Takaue T, Kobayashi M, Takada M, Inazumi S (2020) Development of a real-time position measurement management system for this method. In: 14th Japanese ground improvement symposium, Japan 2. Japanese Society of Materials Science and Technology (2015) Technical evaluation certificate No. 1006 of the deep cement stabilization method. Update 2nd version 3. Kizuki T, Sawaguchi H, Imai T, Takaue T, Tsuchiya J, Inazumi S (2018) Introduction of a realtime construction management system in the application of large diameter and deep depth deep cement mixing method. In: 13th Japanese ground improvement symposium, Japan 4. Nguyen HS, Adachi Y, Kizuki T, Maeba H, Inazumi S (2020) Integration of information and communication technology (ICT) into cement deep mixing method. Int J GEOMATE 19(74):194–200

Evaluation of Landslide Triggering Mechanism During Rainfall in Slopes Containing Vertical Cracks Muhammad N. Hidayat , Hemanta Hazarika, Takashi Fujishiro, Masanori Murai, Suman Manandhar, Yan Liu, Yasuhide Fukumoto, Haruichi Kanaya, and Osamu Takiguchi

Abstract Landslides are one of the most destructive natural hazards that lead to the loss of lives and infrastructures. Understanding the landslide mechanism is one of the important parts of mitigating landslide hazards. The occurrence of landslides can be significantly influenced by the presence of vertical cracks. These cracks act as potential weak points in the slope, making it more susceptible to failure. In this paper, slope model experiments were carried out in the laboratory using K–7 sand incorporating rainfall intensity of 70 mm/h via a sprinkler system. The slope model was simulated with a slope angle of 40°. The main objective of the study was to evaluate the mechanism of slope deformation patterns in two different types of slopes, with vertical cracks and without vertical cracks. The soil moisture content and pore water pressure affect the instability of slope models. The results showed that the distribution of landslide activity was retrogressive and more vulnerable on a slope having vertical cracks compared to a slope without vertical cracks. Keywords Landslide · Crack · Flume test · Rainfall · Slope failure

M. N. Hidayat (B) · H. Hazarika · S. Manandhar · Y. Fukumoto · H. Kanaya Kyushu University, Fukuoka, Japan e-mail: [email protected] T. Fujishiro Geodisaster Prevention Institute, Kitakyushu, Japan M. Murai Shimizu Corporation, Tokyo, Japan Y. Liu Wuyi University, Jiangmen, China O. Takiguchi ALSENS Inc., Sagamihara, Kanagawa, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_5

47

48

M. N. Hidayat et al.

1 Introduction The global landslide database shows that from 2004 to 2016, more than 4800 landslide events occurred, resulting in more than 55,000 fatalities [1]. Landslides are destructive natural disasters that occur frequently, especially during the rainy season, and cause great loss of life and property. For example, landslides occurred in Fukuoka and Oita prefecture, Japan, in July 2017 due to the high intensity of rainfall [2, 3]. Landslides also occurred in Bengkulu, Indonesia, in April 2019, resulting in more than 20 casualties and the destruction of infrastructures [4]. The examples of landslides in Japan and Indonesia represent only a fraction of the numerous natural hazards that require immediate attention and mitigation. Rainfall-induced landslides are a form of natural hazard that can cause severe damage and are caused by several factors. The primary cause is often a combination of factors such as heavy rainfall, soil saturation, increased pore water pressure, steep slope angle, etc. Laboratory-scale studies on the mechanism of rainfall-induced landslides have been demonstrated by researchers. Chueasamat et al. [5] studied the effect of surface sand layer density and rainfall intensity on slope failure. Cogan and Gratchev [6] conducted experiments to identify the effects of different triggering factors for rainfall-induced shallow landslides, while Ahmadi-Adli et al. [7] investigated the triggering mechanism and rainfall threshold of landslides in unsaturated soil slopes. The presence of cracks in the soil also plays an important role in slope failure. Rainwater can easily infiltrate the soil through cracks. The accumulation of water weakens the shear strength of the soil and increases the likelihood of landslides [8, 9]. The cracks in the soil could be caused by weathering and erosion, earthquake and seismic activity, human activities, and so on. Understanding the behavior of rainfallinduced landslides on cracked slopes provides for a more accurate assessment of landslide risk. In this study, slope model experiments were performed comparing a slope containing multiple vertical cracks with a slope model without vertical cracks. The main objective of this study was to evaluate the mechanism of slope deformation patterns in a slope with vertical cracks using the parameters of soil moisture content, pore water pressure, and velocity of deformation. During the experiments, the slopes were monitored using the landslide early warning system (LEWS) developed in the laboratory [10].

Evaluation of Landslide Triggering Mechanism During Rainfall …

49

2 Methodology 2.1 Model Preparation Two series of model experiments were performed in the laboratory. The first model (Model 1) was a slope model without cracks, while the second one (Model 2) was a model with preset vertical cracks on the top part of the slope. The dimension of the slope model is 926 mm in length, 400 mm in width, and 500 mm in height. The slope angle was set to 40°, and the initial water content was 10%. Artificial rainfall produced a controlled rainfall for the study by using a rainfall simulator. The rainfall simulator was attached to a flume tank with a height of 480 mm above the soil surface. The rainfall simulator set on the flume tank produces a 70 mm/ h rainfall intensity through nine nozzles. The vertical cracks in Model 2 were made by inserting a long bolt into the soil. Each crack was cylindrical in shape, 5 mm in diameter, and 100 mm deep. The setting of the vertical cracks was two by four with a total of eight vertical cracks located at the top and on the slope. The detailed layout of the slope model is shown in Fig. 1. Silica sands of grade seven (K–7 sand) were used in this study. The soil properties are shown in Table 1.

Fig. 1 Schematic view of the slope model: a sensors placement and b 3-D view of slope model and the location of the camera

Table 1 Soil parameter of the experiment

Parameter Mean grain size (mm) Dry unit weight of soil (kN/m3 )

Value 0.17 13.73

Void ratio

0.866

Coefficient of uniformity

2.96

Specific gravity

2.62

50

M. N. Hidayat et al.

Two sets of cameras were used to capture the images of the slope model from the front and side. The images captured from the camera had 10 s intervals and were being used for particle image velocimetry (PIV) analysis.

2.2 System Configuration The monitoring devices for the experiments were based on the previous research that had been done in the laboratory [10, 11], which are moisture sensors, pore water pressure sensors, acceleration sensors, and cameras. Moisture sensors were used to measure the soil moisture level by detecting the capacitive change that touches the surface of the sensors. The moisture sensors were connected to the IoT boards and sent the data to the server. Pore water pressure sensors were used to measure the pore water pressure in the soil. Low-power consumption acceleration sensors connected to a display monitor were used to measure the soil deformation with 0.05° accuracy.

2.3 Particle Image Velocimetry Particle image velocimetry (PIV) was used to analyze the image recorded during the experiment. PIV can detect the velocity of soil particles from the captured image and obtain the displacement vector. It consists of two patches, which are the test patch and the detection patch (Fig. 2). The first image is captured by the camera and split into a grid of test patches. The detection patch from the second image is used to search for test patches. The size of the detection patch is larger than the size of the test patch. Finally, the correlation between each test patch and its matching detection patch is estimated, and each test patch’s displacement vector is observed. The velocity vector was obtained by capturing the correlation between the detection patch and test patch on every image taken from the beginning until the end of the experiment.

3 Results This section illustrates the monitoring results of changes in soil moisture content, pore water pressure, and soil deformation at the rainfall intensity of 70 mm/h. The moisture sensor (MS), pore water pressure sensor (PWP), and acceleration sensor (AS) were installed in the slope model to evaluate the mechanism of slope failure with and without vertical cracks. Figures 3, 4 and 5 show the experimental results. In both models, the soil moisture sensors MS1–MS4 detected a more rapid increase in moisture content compared to the other sensors (Fig. 3). These moisture sensors were located near the soil surface. The last moisture sensors responding to the increase

Evaluation of Landslide Triggering Mechanism During Rainfall …

51

Fig. 2 Displacement vector in PIV analysis: a the image captured by the side camera, b the first image that shows the detection and test patches before the soil movement was recorded, and c the second image that records the movement of the soil particle captured by the detection and test patches

of soil moisture content were MS5 and MS6 because both sensors were located the farthest from the surface. As the rainwater infiltrates through the soil pore, water accumulates at the bottom and generates pore water pressure. The pore water pressure was measured by the pore water pressure sensor P1–P3 (Fig. 4). P1 was located at the bottom and closer to the slope toe; thus, P1 responds faster to the increase of pore water pressure. The acceleration sensor monitors the deflection of the slope and records 10 data per second. The first landslide in Model 1 was recorded when the soil slope deflected by 0.27°, while in Model 2 the slope deflected by 13.2° (Fig. 5). The maximum deflection angles recorded in Model 1 and 2 were 10.7° and 99.7°, respectively. From Fig. 5, it can also be observed that the slope displacement process started from the small movement of the soil when the first landslide occurred. Afterward, both models show a rapid increase in the deflection angle at the third landslide.

Fig. 3 Moisture content monitoring result under rainfall intensity of 70 mm/h

52

M. N. Hidayat et al.

Fig. 4 Pore water pressure reading

Fig. 5 Soil deflection angle recorded for both model experiments

Two cameras were used to monitor the slope from the front and side (Fig. 1b). During the slope model experiments, multiple failures were recorded on both slope models from the toe to the crest of the slope. The slope model experiment found that there were more slope failures recorded on Model 2 in comparison with Model 1, particularly near the crest of Model 2, which has vertical cracks in it. The cracks allowed rainwater to easily seep into the soil. As the soil becomes saturated, its shear strength decreases, making the slope more prone to instability and sliding along the preexisting cracks. Consequently, Model 2 experiences a higher number of failures due to these factors. Five typical shallow landslides that exhibit the characteristics of the landslides in each experiment were chosen among the numerous shallow landslides (Figs. 6 and 7). The first failure in Model 1 started from the eroded slope toe at 100 min and progressed to 120 min after the rainfall began, while in Model 2, it started from

Evaluation of Landslide Triggering Mechanism During Rainfall …

53

Fig. 6 Velocity vector in Model 1

105 to 134 min. Both models exhibit similar results in the PIV analysis, i.e., as time progresses, the velocity vectors increase. As the failure extends from the toe of the slope to the crest, the velocity and volume of displaced soil also increase. However, the magnitude of velocity in both models was different. Model 2 has a smaller magnitude of velocity compared to Model 1 due to the presence of numerous vertical cracks. As the number of slope failures increases, the volume of displaced soil in each failure decreases, resulting in a lower velocity.

4 Discussion Several failures occurred during the experiments. In Model 1, the first, third, and fifth failures occur 100, 108, and 120 min, respectively, after the initiation of the rainfall. The first failure occurred when MS1 and MS6, located at the bottom layer, recorded the saturated state of the soil. As the rain continues, the rest of the moisture sensors begin to record the saturated state. A similar result was shown in Model 2. The failure time for the first, third, and fifth failures in Model 2 were 105, 115, and 134 min, respectively, after the rainfall started. During rainfall, water infiltrates into the soil through the soil pores, causing the pore water pressure to increase. As the rain continues, the sensors detect the

54

M. N. Hidayat et al.

Fig. 7 Velocity vector in Model 2

increasing pore water pressure in the soil. The pore water pressures in Model 2 have a higher value than in Model 1. The reason is that during rainfall, the existence of cracks provides favorable pathways for water to infiltrate into the soil, increasing the pore water pressure. The P1 value in both models shows a temporary decrease before increasing again. The P1 of Model 1 decreased after the third landslide. This was caused by the slope failure moving the soil near P1 so that the infiltrating water no longer accumulated but flowed with the moving soil. On the other hand, P1 in Model 2 decreased after the first landslide. When the first landslide occurred in Model 2, the accumulated water in the soil flowed out through the failure zone, reducing the pore water pressure. There is a large difference in the value of the maximum deflection angle for the two acceleration sensors. This was caused by the difference in the methods of placing the acceleration sensors. The AS in Model 1 was placed horizontally in the soil, while the AS in Model 2 was placed vertically with the AS cable at the bottom. The idea of changing the AS position in Model 2 was to let the AS move freely without any constraints. The AS movement in Model 1 was constrained because when the slope failure occurred, the AS cable held it back. Conversely, in Model 2, the AS cable was not holding it back, allowing the sensors to move more than 90°. The detail of the AS positioning is shown in Fig. 8. In both slope models, the initiation of slope failure occurred at the base of the slope. Then, the failure gradually propagated upwards to the higher points of the slope. This

Evaluation of Landslide Triggering Mechanism During Rainfall …

55

Fig. 8 The positioning of the acceleration sensor: a the acceleration sensor in Model 1 was placed horizontally, and b the acceleration sensor in Model 2 was placed vertically

Fig. 9 The slope failure progresses from the toe to the top of the slope

pattern of slope failure indicates a retrogressive mechanism, where instability starts at the bottom and progressively moves toward the top as shown in Fig. 9.

5 Conclusion In this research, the slope model experiments were conducted using silica sand grade seven (K–7) to understand the failure mechanism of slopes. A slope model containing vertical cracks under rainfall conditions was conducted. The result is then compared with the slope without vertical cracks. The monitoring device used was a landslide early warning system (LEWS) device previously developed in the laboratory, which are moisture sensors, pore water pressure sensors, and acceleration sensors. The slope model experiment concludes that:

56

M. N. Hidayat et al.

(1) Vertical cracks can provide a direct pathway for rainwater to enter the soil and seep into the ground. This is evidenced by the high value of pore water pressure and the presence of numerous slope failures in the area around the vertical cracks. (2) The presence of vertical cracks contributed to a gradual slope failure process in the slope model, resulting in a gradual and slower movement of soil material. (3) The rapid increase in pore water pressure in the vertical cracks in the silica sand slope model progressively ruptured from the toe and extended upslope, forming the retrogressive failure. Acknowledgements This study was financed by The Maeda Engineering Foundation, Japan, and supported by the Institute of Mathematics for Industry, Kyushu University, Japan. The authors also would like to express gratitude to former student Ms. Yurika Taguchi and the former staff Dr. Vilayvong Khonesavanh of Kyushu University for their help during the experiment.

References 1. Froude MJ, Petley DN (2018) Global fatal landslide occurrence from 2004 to 2016. Nat Hazard 18:2161–2181 2. Ministry of Land, Infrastructure, Transport, and Tourism Japan (2017) White paper on land, infrastructure, transport and tourism in Japan 2017, p 378 3. Hazarika H, Yamamoto S, Ishizawa T, Danjo T, Kochi Y, Fujishiro T, Okamoto K, Matsumoto D, Ishibashi S (2020) The 2017 July Northern Kyushu torrential rainfall disaster—geotechnical and geological perspectives. In: Krishna A, Katsumi T (eds) Geotechnics for natural disaster mitigation and management. Developments in geotechnical engineering. Springer, Singapore 4. Mase LZ, Perdana A, Hardiansyah, Amri K, Bahri S (2022) A case study of slope stability improvement in Central Bengkulu landslide in Indonesia. Transp Infrastruct Geotechnol 9:442– 466 5. Chueasamat A, Hori T, Saito H, Sato T, Kohgo Y (2018) Experimental tests of slope failure due to rainfalls using 1g physical slope models. Soils Found 58:290–305 6. Cogan J, Gratchev I (2019) A study on the effect of rainfall and slope characteristics on landslide initiation by means of flume tests. Landslides 16:2369–2379 7. Ahmadi-adli M, Huvaj N, Toker NK (2017) Rainfall-triggered landslides in an unsaturated soil: a laboratory flume study. Environ Earth Sci 76:735–748 8. Zhang G, Wang R, Qian J, Zhang J-M, Qian J (2012) Effect study of cracks on behavior of soil slope under rainfall conditions. Soils Found 52:634–643 9. Mukhlisin M, Khiyon KN (2018) The effects of cracking on slope stability. J Geol Soc India 91:704–710 10. Liu Y, Hazarika H, Kanaya H, Takiguchi O, Rohit D (2023) Landslide prediction based on low-cost and sustainable early warning systems with IoT. Bull Eng Geol Env 82:177–190 11. Liu Y, Hazarika H, Takiguchi O, Kanaya H (2021) Developing a sustainable system for early warning against landslides during rainfall. In: 1st international symposium on construction resources for environmentally sustainable technologies, CREST 2020, pp 917–926. Springer, Singapore

Landslide Risk Prediction and Regional Dependence Evaluation Based on Disaster History Using Machine Learning and Deep Learning Fudong Ren and Koichi Isobe

Abstract Around the world, many people were killed and injured by landslides, and thousands of houses and buildings were destroyed. Therefore, the evaluation and analysis of landslide hazards are very important for human, environmental, cultural, economic and social sustainability. This paper discusses the susceptibility of landslides and the regional dependence, in other words differences in characteristics between regions, based on ArcGIS and AI technologies such as machine learning and deep learning. Considering previous studies and various kind of data, 11 factors related to historical landslide: elevation, slope, aspect, curvature, lithology, soil, land-use, precipitation, distance from rivers, faults, and roads are selected in this research. The above characteristic data are aggregated and processed through ArcGIS to train the models and verify their performance, and finally check their versatility by applying them to the other regions. In machine learning, KNN, DT and MLP (ANN) models outperformed LR and RF models. In deep learning, the performance of CNN depends on the structure of the model, i.e., layer depth, input window size, and training strategy. It is necessary to optimize the deep learning model structure to show the better performance than machine learning. A model that is optimized in a specific region may not necessarily be applicable in another region as it is observed that the LR model and the DT model in the HKD area performed far less than the results for the other area (regional dependence). In this study for the three regions, the KNN model and the MLP model comprehensively outperform the other models from the aspect of the regional dependence. Keywords Landslide · AI · Risk assessment

F. Ren (B) Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Hokkaido, Japan e-mail: [email protected] K. Isobe Faculty of Engineering, Hokkaido University, Sapporo 060-8628, Hokkaido, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_6

57

58

F. Ren and K. Isobe

1 Introduction Landslide is a natural phenomenon, which has caused huge losses and some even devastating disasters to industrial and agricultural production and people’s lives and property. Therefore, the evaluation and analysis of landslide hazards and landslide sensitivity mapping are particularly important for human environmental, cultural, economic, and social sustainability. Looking back at the development history of landslide disaster research and evaluation in the world, we can roughly find that landslide disaster research and evaluation mainly have the following characteristics: (1) large amount of data (a large amount of data is needed for landslide disaster research in a certain area), (2) data complex structure (both spatial data and attribute data), (3) multi-source, ambiguity, indeterminism, and randomness of disaster information. In view of the above characteristics, the information processing and spatial comprehensive analysis model becomes more complicated. Traditional methods are inadequate for such model processing and analysis, which may lead to a decrease in the reliability of the results and lack of intuitiveness and operability in expressing the results. Numerical models and Geographic Information System (GIS) [1], which integrate spatial data and attribute data processing, management, analysis, query, and input–output functions, are powerful tools for landslide risk assessment and analysis. In particular, recent case studies apply machine learning models to evaluate and analyze landslide areas, such as artificial neural network (ANN) [2], logistic regression (LR) [3, 4], decision tree (DT) [4], support vector machine (SVM) [5, 6], and KNN [7, 8], and other models have been used to analyze landslide sensitivity. The purpose of this study is to discusses the susceptibility of landslides and the regional dependence, in other words differences in characteristics between regions, and the model versability, based on various kind of machine learning models and deep learning models by applying to three different target areas and comparing these results.

2 Study Area and Used Data This paper studies the regional differences in the characteristics of landslides in different target areas and for the purpose of studying earthquake landslides in the future. In this research, three different study areas are selected: Niigata Prefecture (NIG area), Iwate Prefecture, Miyagi Prefecture (IWT-MYG area), and Hokkaido (HKD area) in Japan as shown in Fig. 1. Landslides can be caused by a variety of reasons, and there are complex links between each of them. By studying the causes of landslides in the previous research, 11 features that are most intuitively linked to landslides are selected for this research: (1) internal (topography, lithology, and geological structure) [9–12], (2) external factors: (a) induced reasons: earthquake and precipitation [13–16], (b) human causes [16–18], etc. The dataset and their source are given in Table 1.

Landslide Risk Prediction and Regional Dependence Evaluation Based …

59

Fig. 1 Location of study area and distributions of historical landslides. a NIG study area, b IWTMYG study area and c HKD study area

3 Methodology The detailed workflow of this study is shown in Fig. 2. Firstly, database construction as given in Table 1 is processed and combined in ArcGIS. Secondly, the Excel data and tiff data processed in ArcGIS are input into the machine learning models and the deep learning models. Subsequently, the model performance is evaluated based on a series of evaluation indicators such as area under the receiver operating characteristic (ROC) curve (AUC), accuracy, precision (P-score), recall (R-score), and F1-score as shown below. Finally, the results are visualized to get the landslides susceptibility map (LSM) results.

3.1 Machine Learning (LR, DT, RF, KNN, ANN, and MLP) Logistic regression (LR) is an extremely easy to understand model that is equivalent to y = f (x), showing the relationship between the independent variable x and the dependent variable y. Decision tree (DT) is a tree structure consisting of a series of nodes, and each node represents a feature and corresponding decision rules. It is the process of classifying or regressing instances based on features, i.e., dividing the data into sub-regions

60

F. Ren and K. Isobe

Table 1 Features and data range in the study area Code

Dataset

Resource

NIG area

IWT-MYG area

HKD area

A0

Historical landslides (HLS)

NIED

0, 1

0, 1

0, 1

A1

Elevation (m)

DEM

8 ~ 1027

1 ~ 1625

0 ~ 1002

A2

Slope (°)

DEM

0 ~ 52.05

0 ~ 63.72

0 ~ 72.20

A3

Aspect (north = 0, clockwise)

DEM

− 1 ~ 360

− 1 ~ 360

1066 ~ 1308

A4

Curvature

DEM

− 7.15 ~ 6.98

− 17.92 ~ 22.81

− 64.40 ~ 54.20

A5

Soil

MLIT

1, 2, 3, … 17

1, 2, 3, … 17

1, 2, 3, … 17

A6

Lithology

GSJ

1, 2, 3, … 28

1, 2, 3, … 43

1, 2, 3, … 20

A7

Land-use

MLIT

1, 2, 3, … 12

1, 2, 3, … 12

1, 2, 3, … 12

A8

Average rainfall (mm)

JMA

2169.54 ~ 2795.69

1047.28 ~ 2308.10

1066.00 ~ 1308.00

A9

Distance to MLIT river (m)

0 ~ 2085.1

0 ~ 3931.3

0 ~ 1885.4

A10

Distance to GSJ fault (m)

0 ~ 22,314.6

0 ~ 15,008.5

0 ~ 10,731.7

A11

Distance to MLIT road (m)

0 ~ 6787.8

0 ~ 24,985.5

0 ~ 8999.9

NIED National Research Institute for Earth Science and Disaster Resilience GSJ Geological Survey of Japan, National Institute of Advanced Industrial Science and Technology MLIT Ministry of Land, Infrastructure, Transport, and Tourism JMA Japan Meteorological Agency

(sub-trees) based on a feature, and then recursively dividing the sub-regions until a condition is met, then stopping the division and acting as a leaf node, or continuing recursively if the condition is not met. Random forest (RF) is an algorithm that integrates multiple decision trees through the idea of ensemble learning. Its basic unit is a decision tree. K-nearest neighbor (KNN) is one of the simplest machine learning algorithms, which can be used for classification and regression, and is a supervised learning algorithm. It is based on that if the majority of the K most similar (i.e., most neighboring) samples in the feature space belong to a particular class, then that sample also belongs to that class. Artificial neural network (ANN) is subdivided into single layer type and multiple layers type. Each layer contains several neurons. Each neuron is connected by a directed arrows with variable weights. The network is trained through repeated learning of known information and through gradual adjustments. An ANN with multiple hidden layers is called a multilayer perceptron (MLP) [2].

Landslide Risk Prediction and Regional Dependence Evaluation Based …

61

Fig. 2 Flowchart of this study

3.2 Deep Learning (DNN and CNN) Deep neural network (DNN) is a multilayer unsupervised neural network and uses the output features of the previous layer as the input of the next layer for feature learning. After layer-by-layer feature mapping, the features of existing spatial samples are mapped to another feature space to learn better feature representations for existing inputs. Deep neural network (DNN) is built on the basis of the perceptron model. The perceptron model is a multi-input and single-output model. Convolutional neural network (CNN) is an artificial neural network. The structure of CNN can be divided into three layers: (1) convolutional layer whose main function is to extract features, (2) max pooling layer which is the layer of down sampling without damaging the recognition results, and (3) fully connected layer which is classification layer. The features obtained through the convolution and pooling layers are classified at the fully connected layer.

4 Results The distribution of the features in this research area is similar, but the different characteristics are observed in the elevation, rainfall, and lithology shown in Fig. 3. The elevation of landslide in the IWT-MYG area is higher than that in the NIG area and the HKD area. This shows that the IWT-MYG area locates in more mountainous area. The average precipitation in NIG area is much larger than that in the other two regions. This is because the dry and cold northwest monsoon of the Eurasian

62

F. Ren and K. Isobe

continent moves southward through the Sea of Japan, and the cold air becomes humid, forming precipitation on the windward slope of western Japan. In terms of lithology, the landslides in the three regions are distributed in a wide variety of lithology. Table 2 presents the results by some kinds of machine learning models. According to the results of AUC alone among the three regions, the KNN model shows the best performance for the NIG region (AUC = 0.931), and DT model indicates the best performance for the IWT-MYG region (AUC = 0.961), and the ANN results are the best for the HKD area (AUC = 0.969). For all models, the results obtained for the IWT-MYG area and the HKD area are better than for the NIG area. However, there is possibility that AUC alone cannot evaluate the performance of the model comprehensively, so accuracy (ACC), precision (P-score), and recall (R-score) are also introduced to further evaluate the model performance. The P-score is not high for all three regions because the model does not perform well in predicting negative samples, resulting in an increase in FP. Accuracy is defined as the proportion of correctly predicted sample number to the total sample numbers. Therefore, ACC results include the following two tendencies; (1) in case that the P-score is low, and the ACC is low, it means that the model on actual negative examples shows poor performance, and the decrease in the number of correctly predicted samples (TP + TN), resulting in low ACC results, (2) in case that the P-score is low, but the ACC is relatively good, it means that the imbalance of positive and negative examples in the dataset. Although the prediction for the actual negative examples is not good, the actual negative examples number is large enough in the data used in this series of training, so the number of correct predicted samples (TP + TN) will not be greatly affected, resulting in relatively good ACC results. The R-score is holistically better in the results in all regions, except the R-score of the LR model in three regions, which shows that the LR model has poor detection performance for actual positive examples, resulting in a decrease in TP. Accuracy (ACC) =

TP + TN TP + FP + TN + FN

Precision (P-score) =

TP TP + FP

Fig. 3 Relationship between landslide distribution and features

(1) (2)

Landslide Risk Prediction and Regional Dependence Evaluation Based …

TP TP + FN

(3)

2 × P score × R score P score + T R score

(4)

R score (R-score) = F1 score (F score) =

63

From the DNN model results given in Table 2, considering all the evaluation indicators, the performance of the DNN model in the IWT-MYG area is the best, the result of recall is relatively poor in the NIG area, and the DNN model’s ability to detect the actual positive cases in the NIG area is relatively weak. It indicates the current DNN model structure is unfordable for the NIG region. In the CNN model, three different input sizes are tried, but the results of all evaluation indicators are similar. Therefore, the performance of the model in different regions cannot be evaluated and compared, so the loss and accuracy are used to further Table 2 Results obtained by the machine learning and deep learning models Area

Model

ACC

P-score

R-score

F-score

AUC

NIG

LR

0.704

0.278

0.617

0.383

0.744

IWT-MYG

HKD

DT

0.873

0.547

0.866

0.670

0.873

RF

0.654

0.281

0.864

0.422

0.654

KNN

0.832

0.466

0.884

0.610

0.931

ANN

0.827

0.458

0.854

0.596

0.919

MLP

0.811

0.433

0.866

0.577

0.914

CNN

0.876

0.815

0.973

0.887

0.877

DNN

0.945

0.709

0.427

0.533

0.706

LR

0.802

0.331

0.495

0.396

0.769

DT

0.908

0.600

0.893

0.718

0.961

RF

0.756

0.323

0.783

0.457

0.844

KNN

0.848

0.458

0.863

0.598

0.927

ANN

0.880

0.526

0.828

0.644

0.938

MLP

0.876

0.516

0.857

0.644

0.942

CNN

0.841

0.771

0.968

0.858

0.842

DNN

0.973

0.809

0.726

0.765

0.857

LR

0.873

0.202

0.635

0.307

0.873

DT

0.952

0.480

0.918

0.630

0.952

RF

0.726

0.413

0.805

0.551

0.839

KNN

0.912

0.327

0.924

0.483

0.967

ANN

0.913

0.328

0.914

0.482

0.969

MLP

0.898

0.314

0.928

0.465

0.958

CNN

0.861

0.799

0.964

0.874

0.861

DNN

0.982

0.784

0.629

0.698

0.811

64

F. Ren and K. Isobe

judge the model performance. As shown in Fig. 4, the loss and accuracy results of the three study areas. Firstly, in the NIG area, the results of the three different input sizes are similar, but in the convergence process, the convergence of loss is different. There is noise in the convergence process of three different sizes, but the degree is different. The main reason for the noise is that the current model structure is not the optimal model for the dataset, and as the input size gradually becomes smaller the noise gradually decreases, while indicating that the current model structure is suitable for datasets with small input sizes in the NIG area. But there is still room for optimization even with large input sizes. In the IWT-MYG study area, there is no phenomenon similar to the NIG area. When the input size is 3 and 7, both accuracy and loss converge quickly and there is almost no noise, indicating that the current model with the input size 3 and 7 is very suitable for IWT-MYG study area, but in case that the model with the input size is 11, it tends to overfit them. Its main possible reason for this phenomenon is that the number of learning iterations is too much (Overtraining). In other words, the current model structure with the input size 11 can learn in fewer iterations. The same overfitting phenomenon also occurs for the model with input size 7 and 11 in the HKD datasets because of the same reasons shown above. From this discussion, it can be said that the optimum input size exists according as the landslides scale included in the dataset.

Fig. 4 Loss and accuracy in CNN (top to bottom: NIG area, IWT-MYG area, and HKD area, left to right: window size 11, 7, and 3)

Landslide Risk Prediction and Regional Dependence Evaluation Based …

65

Fig. 5 LSM based on the machine learning models (KNN and MLP)

In the landslide susceptibility map (LSM) as shown in Figs. 5 and 6, the sensitivity is divided into five equal parts, while showing each sensitivity represents a different risk level and expressing the results on the coordinates. Judging from the evaluation indicators and the results of LSM, for all three regions, the performance of the KNN model and the MLP model is better than that of the other models, meaning that it is suitable to apply to the practical use. Comparing the results of machine learning and deep learning, the LSM result of deep learning only extracts the area where landslide occurred and surrounding small area as a high-risk area, and is more inclined to determine the location of landslide, while the LSM results of machine learning are more biased toward sensitivity analysis of the location around the landslide. Comparing the results from machine learning and deep learning by using various evaluation indicators, although there is little difference in AUC, it can be found that the deep learning model is better than machine learning in both P-score and R-score. It indicates that the deep learning is stronger than machine learning in both subjectively positive and objectively positive predictions.

5 Conclusions This study mainly discusses the performance of various types of machine learning and deep learning in the three regions by using some evaluation indicators, while showing the predictive performance, the applicability of the models, and the regionality of the study area. For the versatility of model, in machine learning, it is currently difficult to explain the generality of the model. This part will be discussed in depth in further studies. But in the deep learning, whether it is DNN or CNN, the same structure is used in the three regions, CNN performs slightly better than DNN model, but the prediction

66

F. Ren and K. Isobe

Fig. 6 LSM based on the deep learning models (CNN and DNN)

performance is similar, it can be concluded that the optimized deep learning model can be quickly used to different areas with only simple adjustments. For the same area, the performance of deep learning is better than machine learning. Although the AUC results are similar, accuracy, P-score, and R-score are all better than machine learning, which fully shows that deep learning is better than machine learning. The predictive ability of positive examples is stronger than machine learning models, but the high predictive ability also has certain shortcomings. Compared with machine learning, deep learning requires longer learning time and better computer hardware. The performance of different models is also different. For different regions, although the same model is used, the results obtained are different. In machine learning model, combining evaluation metrics and LSM, the results of HKD are better than IWT-MYG regions, and the results of IWT-MYG regions are better than NIG regions. In the learning model, there is noise in the datasets of the three input sizes in the NIG region, while in the IWT-MYG and HKD regions, the input size 3 and 7 datasets are suitable for the current model, while in the input size 11 datasets appear overfitting, all of these illustrate the differences between regions, that is the landslides scale dependency.

References 1. Jiang Q (2005) Evaluation based on ArcGIS for slopes geohazards. Doctoral dissertation, Chengdu University of Technology 2. Kawabata D, Bandibas J (2009) Landslide susceptibility mapping using geological data, a DEM from ASTER images and an artificial neural network (ANN). Geomorphology 113(1– 2):97–109

Landslide Risk Prediction and Regional Dependence Evaluation Based …

67

3. Mondal S, Mandal S (2018) RS & GIS-based landslide susceptibility mapping of the Balason River basin, Darjeeling Himalaya, using logistic regression (LR) model. Georisk Assess Manage Risk Eng Syst Geohazards 12(1):29–44 4. Kadavi PR, Lee CW, Lee S (2019) Landslide-susceptibility mapping in Gangwon-do, South Korea, using logistic regression and decision tree models. Environ Earth Sci 78(4):116 5. Lee S, Hong SM, Jung HS (2017) A support vector machine for landslide susceptibility mapping in Gangwon Province, Korea. Sustainability 9(1):48 6. Lin Q, Liu Y, Liu L, Wang Y (2017) Earthquake-triggered landslide susceptibility assessment based on support vector machine combined with Newmark displacement model. J Geo-Inf Sci 76(19):1–9 7. Sameen MI, Pradhan B, Bui DT, Alamri AM (2020) Systematic sample subdividing strategy for training landslide susceptibility models. CATENA 187:104358 8. Adnan MSG, Rahman MS, Ahmed N, Ahmed B, Rabbi M, Rahman RM (2020) Improving spatial agreement in machine learning-based landslide susceptibility mapping. Remote Sens 12(20):3347 9. Liu R, Sun N, Tang G (2010) Analysis of geological environment and causes of landslides in Guangdong. Trop Geogr 1:13–17 10. Ray RL, Jacobs JM (2007) Relationships among remotely sensed soil moisture, precipitation and landslide events. Nat Hazards 43(2):211–222 11. Larsen IJ, Montgomery DR (2012) Landslide erosion coupled to tectonics and river incision. Nat Geosci 5(7):468–473 12. Tibaldi A, Ferrari L, Pasquarè G (1995) Landslides triggered by earthquakes and their relations with faults and mountain slope geometry: an example from Ecuador. Geomorphology 11(3):215–226 13. Cao C, Wang Q, Chen J, Ruan Y, Zheng L, Song S, Niu C (2016) Landslide susceptibility mapping in vertical distribution law of precipitation area: case of the Xulong hydropower station reservoir, Southwestern China. Water 8(7):270 14. Catani F, Casagli N, Ermini L, Righini G, Menduni G (2005) Landslide hazard and risk mapping at catchment scale in the Arno River basin. Landslides 2(4):329–342 15. Yano A, Shinohara Y, Tsunetaka H, Mizuno H, Kubota T (2019) Distribution of landslides caused by heavy rainfall events and an earthquake in northern Aso Volcano, Japan from 1955 to 2016. Geomorphology 327:533–541 16. Sun L, Ma B, Pei L, Zhang X, Zhou JL (2021) The relationship of human activities and rainfallinduced landslide and debris flow hazards in Central China. Nat Hazards 107(1):147–169 17. Glade T (2003) Landslide occurrence as a response to land use change: a review of evidence from New Zealand. CATENA 51(3–4):297–314 18. Bozzano F, Cipriani I, Mazzanti P, Prestininzi A (2011) Displacement patterns of a landslide affected by human activities: insights from ground-based InSAR monitoring. Nat Hazards 59(3):1377–1396

Machine Learning for Estimation of Surface Ground Structure by H/V Spectral Ratio Tetsuo Tobita and Wataru Yamamoto

Abstract A new method for efficiently identifying ground motion observation site is proposed by the deep learning with a convolutional neural network (CNN). This method transforms the acceleration spectra at an observation site into unique color spectra corresponding to their amplitudes, which are then learned by the deep learning. First, the seismic H/V spectral ratios of earthquake motions of 50 gal or less obtained at eight K-NET stations are tested. As a result, the K-NET sites were identified with an accuracy of more than 95%. Misclassified earthquake motions were unique ones such as those having large source distances or hypocenter depth being deep. Next, the microtremor H/V spectral ratios obtained at the same K-NET sites were input to the CNN which was solely trained by the seismic H/V spectral ratio as above. The accuracy varied from 0 to 95% and on average 50%. Although results are encouraging, the method requites further developments. Keywords Earthquake · Microtremor · H/V spectrum · Deep learning · Convolution neural network

1 Introduction Large earthquakes, such as the 2011 off the Pacific coast of Tohoku, Japan, earthquake and the 2016 Kumamoto, Japan, earthquake, have caused substantial damages to human society in recent years. In the near future, the Nankai Trough mega-earthquake is predicted to occur, which is expected to cause devastating damages over a wide area in the western part of Japan. In order to minimize such damages, it is desirable to know how the local site behaves under such a large earthquake. Distributions of earthquake intensity at local sites are estimated with various methods. For example, to estimate T. Tobita (B) Kansai University, Suita, Osaka, Japan e-mail: [email protected] W. Yamamoto Oyo Cooperation, Saitama, Saitama, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_7

69

70

T. Tobita and W. Yamamoto

seismic intensity in the entire Japan Island, J-SHIS [1] uses the distance attenuation relationship to predict the intensity at the engineering bedrock (Vs > 400 m/s). Then, combining microtopography and empirical site amplification method, seismic intensity of the ground surface is estimated. The method is convenient. However, it is an indirect method and incapable of predicting dominant frequencies at a particular site which is one of the important seismic local site characteristics. For disaster mitigation, it is desirable to obtain seismic local site characteristics of a wide area in a uniform and appropriate manner. In assessing detailed damages, level of ground deformation is often evaluated by numerical analysis in which ground models are often created from results of in situ survey above the seismic bedrock. When geotechnical information is insufficient, additional investigations must be carried out, which include borehole investigations and/or the PS logging. However, they cannot be easily conducted due to the need for space, time, and cost. One of the geophysical exploration methods that are relatively inexpensive and have a short survey period is the method of microtremor observation. In this method, a natural period of the ground and site characteristics are estimated from microtremors. For example, Kanai and Tanaka [2] proposed a simple method to determine the ground type based on the mean period and maximum period or the dominant period and maximum amplitude using the microtremors. Nakamura [3] showed that the peak frequency of the H/V spectral ratio, which is defined as the ratio of the horizontal and vertical components of the microtremor spectrum, corresponds to the natural frequency of the surface soil at the observation site. The H/V spectral ratio has since been widely used to estimate the depth of the engineering bedrock and the thickness of sedimentary layers. There have been many studies on the application of machine learning to seismology and earthquake engineering, e.g., [4–7]. In addition, the JSCE Earthquake Engineering Committee has been actively conducting research in the “Research Subcommittee on Utilization of AI and IoT Technologies for Disaster Prevention and Mitigation (Chairperson: Y. Maruyama).” In this study, an attempt is made to identify observation sites by combining wave spectra with image recognition technology based on deep learning (convolutional neural network, hereafter CNN) [8, 9], which is a type of machine learning technology that has been dramatically improved in recent years. The method is to identify the observation site by using seismic motions observed by the K-NET. In this method, the acceleration spectrum of the surface layer is converted into a color spectrum image by color coding [10] the amplitude values in the manner shown in Fig. 1. This color spectrum is used as an input to the CNN to identify the ground model and observation site through image analysis. In this study, the amplitude of the H/V spectral ratio is normalized to the maximum value, so the color spectrum does not reflect the absolute value of the amplitude. Therefore, at present, the only information included in the color spectrum is the relative amplitude values at each frequency.

Machine Learning for Estimation of Surface Ground Structure by H/V …

71

Fig. 1 Transformation of a curved spectrum into a color spectrum image

2 Classification of the K-NET Site by Observed Ground Motions In this section, an attempt is made to classify the eight K-NET stations [11] by ground motions, including five sites near Osaka City and three sites near Atsuma Town, Hokkaido (Fig. 2). A total of 1677 seismic waves with a maximum acceleration of 50 gal or less, which are estimated to be less affected by ground nonlinearity, were selected for the period from January 1996 to June 2021. Figure 3 shows the frequency distribution of the source depth, source distance, magnitude, and epicentral location of these waveforms. In this study with the observed seismic motions, the color spectra are created with the seismic H/V spectral ratios shown in Fig. 4. Figure 5 shows representative 18 color spectra per observation site. The figure shows that the arrangement of the colors differs slightly even at the same observation site due to the effects of propagation characteristics, source characteristics, and nonlinearities in the ground. Table 2 shows the number of color spectral images for training (train) and validation (test) for each observation site. The MATLAB [12] is used as a platform of the machine learning with hyperparameters shown in Table 1. Classification results using the CNN show that the average accuracy is 95.6%, indicating that each observation site can be identified almost accurately using the color spectrum (Table 2). The learning curve and loss function shown in Fig. 6 both converge to a constant value, indicating that the learning process is successful. Table 3 shows the number of correct responses to the validation images for the most accurate learning model. All of the validation images were correctly identified at the three sites in Hokkaido, but there were 18 waves of misclassified

72

T. Tobita and W. Yamamoto

Fig. 2 Investigated K-NET sites in Osaka (a) and Hokkaido (b) [1]

(a) Osaka

(b) Hokkaido

seismic H/V spectral ratios at the four sites in Osaka. The misclassified seismic H/ V spectral ratios for OSK002 were nine waves as shown in Table 3. Table 4 shows the probability of classification of each site for these nine waves. In the same table, “No.” in the left column indicates the given sequential number of the nine seismic H/V spectral ratios, and the number in the “Category” column indicates the probability that each seismic H/V spectral ratio is classified to each site. In the same table, focusing on the seismic H/V spectrum of No. 1, the probability that this seismic H/V spectrum ratio is classified as OSK002 is 44%, and the probability that it is classified as OSK005 is 55%. In the case of this seismic H/V spectrum ratio, it is assumed that the characteristics of OSK002 were captured to some extent in the classification process, but it was classified as OSK005. On the other hand, focusing on the seismic motion of No. 2, the probability of being misclassified as OSK005 is 100%. Figure 7 compares the seismic H/V spectral ratios of No. 2, No. 4, and No. 7, which were misclassified with high probability, with that of OSK002 (average of randomly selected ten waves) used in the training. The figure shows that the seismic H/V spectra of No. 2 and No. 4 are significantly different from that of OSK002. However, the seismic H/V spectral ratio of No. 7 seems to be similar to the OSK002. Further investigation of the results shows that the earthquake No. 4 misclassified as OSK001 is a deep earthquake with a source depth of 333 km, which might have had some peculiar characteristics among the seismic

Machine Learning for Estimation of Surface Ground Structure by H/V …

73 (a)

(d)

(c)

(b)

Fig. 3 Strong motion records utilized in this study (1677 waves in total) [12]: epicentral locations (a), histograms of depth (b), source distance (c), and magnitude (d)

HKD127

HKD128

HKD184

OSK001

OSK002

OSK003

OSK004

OSK005

Fig. 4 Examples of seismic H/V spectrum at each location (randomly selected ten waves)

74

T. Tobita and W. Yamamoto

Fig. 5 Examples of the color spectrum of the seismic H/V spectrum

Table 1 Hyperparameters for the CNN taken in this study

Solver

SGD

MiniBatchSize

32

MaxEpochs

20

InitialLearnRate

0.001

Shuffle

Every-epoch

Table 2 Number of color spectra used in the training and testing phase Category HKD127 HKD128 HKD184 OSK001 OSK002 OSK003 OSK004 OSK005 Train

175

175

175

176

138

127

151

57

Test

75

75

75

76

59

55

64

24

Total

250

250

250

252

197

182

215

81

4.0

100

Training

Training

Test

Loss

Accuracy

Test 50

2.0

(b)

(a)

10 0

400 Iteration

800

0 0

400 Iteration

800

Fig. 6 Example of the a accuracy and b loss curves for classification of K-NET sites with the seismic H/V spectrum

motions used in the study. The earthquake No. 7, misclassified as OSK004, also has a hypocentral distance of 505 km and was also found to be a peculiar earthquake among the earthquake motions used in the study. However, the source distance and depth of

True class

75

0

0

0

0

0

0

0

HKD127

HKD128

HKD184

OSK001

OSK002

OSK003

OSK004

OSK005

HKD127

1

0

0

0

0

0

75

0

HKD128

Predicted class

0

0

0

0

0

75

0

0

HKD184

0

1

0

1

76

0

0

0

OSK001

Table 3 True and predicted number of the K-NET sites with the seismic H/V spectrum

3

1

1

50

0

0

0

0

OSK002

0

0

52

4

0

0

0

0

OSK003

0

62

2

1

0

0

0

0

OSK004

20

0

0

3

0

0

0

0

OSK005

0.83

0.97

0.95

0.85

1.00

1.00

1.00

1.00

Recall

Machine Learning for Estimation of Surface Ground Structure by H/V … 75

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.01

0.00

0.00

3

4

5

6

7

8

9

0.00

0.00

0.00

0.00

0.00

2

0.00

HKD128

0.00

HKD127

Category

1

Eq. No.

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

HKD184

0.03

0.00

0.00

0.04

0.00

0.94

0.00

0.00

0.00

OSK001

0.47

0.47

0.00

0.25

0.36

0.04

0.32

0.00

0.44

OSK002

0.48

0.52

0.01

0.69

0.64

0.01

0.00

0.00

0.01

OSK003

Table 4 Misclassified nine earthquakes for OSK002 and their degree of coincidence

0.00

0.00

0.96

0.00

0.00

0.00

0.00

0.00

0.00

OSK004

0.02

0.01

0.01

0.00

0.00

0.00

0.67

1.00

0.55

OSK005 OSK002

True class

OSK003

OSK003

OSK004

OSK003

OSK003

OSK001

OSK005

OSK005

OSK005

Predicted class

76 T. Tobita and W. Yamamoto

Machine Learning for Estimation of Surface Ground Structure by H/V …

77

Fig. 7 Average of seismic H/V spectrum of OSK002 and misclassified ones (No. 2, No. 4, and No. 7)

the No. 2 earthquake motion were not significantly different from those used in the study, suggesting that the earthquake motion has other unidentified characteristics. This indicates that the method can also be used to efficiently extract unusual seismic motions by focusing on the misclassified ones.

3 Classification of Microtremor Observation Sites Using a Trained CNN with Seismic H/V Spectral Ratios Here, an attempt is made to identify microtremor observation sites by the trained CNN with the seismic H/V spectral ratio. The microtremor observation at the eight K-NET stations described in the previous chapter were carried out with the device JU-410 (Hakusan Kogyo). Field measurements of about 15 min long were made at each site. The seismic H/V spectral ratios and the microtremor H/V spectral ratios at the eight stations are shown in Fig. 8. The H/V spectral ratios shown in the figures are the average of all seismic motions used in the study for the seismic H/V spectral ratios, and the average of 20 waves of H/V spectral ratios during 20 consecutive seconds without noise for the microtremor H/V spectral ratios. The figure shows that the frequency range of the peaks of both spectral waveforms is located in the frequency range of 2–10 Hz, and therefore, we attempted to identify the observation sites of the microtremors using the model that learned the color spectrum of the seismic H/ V spectra in this frequency range. The validation results shown in Table 5 indicate that the average accuracy is about 50%, but the accuracy of HKD127 is as high as 90% and OSK004 is 95%. This is due to the similarity of the normalized color spectra of the H/V spectra in the frequency band of 2–10 Hz as shown in Fig. 8. In OSK005, the shape of the H/ V spectra in the frequency range of 2–10 Hz seems to be similar (Fig. 8), but the H/

78

T. Tobita and W. Yamamoto

HKD127

HKD128

HKD184

OSK001

OSK002

OSK003

OSK004

OSK005

Fig. 8 Comparison of the (averaged over entire records) seismic and microtremor H/V spectrum ratio for the eight K-NET stations

V spectra of OSK001 and OSK005 are not similar. The reproducibility of OSK005 was low at 20%, although the shapes of the two spectra seemed to be similar in the frequency range of 2 ~ 10 Hz. Table 5 shows that OSK005 tends to be misclassified as OSK001 and OSK002. Although results are encouraging, the method requires further developments.

4 Conclusions A new method for identifying ground motion observation sites using images is proposed, in which the H/V spectrum of the ground is transformed into a unique color spectrum to be analyzed by the convolutional neural network (CNN), a type of machine learning. The ultimate goal of this study is to quickly predict a ground structure from microtremor H/V spectra with high accuracy by taking advantage of existing database of the H/V spectra and the ground structures. As an initial step, the seismic H/V spectra and the borehole information at the K-NET stations were utilized as a database. Conclusions obtained are as follows: (1) By using the color spectra of the seismic H/V spectra of maximum acceleration of 50 gal or less observed at eight K-NET stations (five in Osaka and three in Hokkaido), the sites are identified with the average accuracy of 95.6%. The misclassified spectra are of a deep earthquake with a source depth of 333 km and an earthquake with an epicentral distance of 505 km. From this, the method can also be used to efficiently extract unusual seismic motions.

True class

18

0

2

8

0

0

0

0

HKD127

HKD128

HKD184

OSK001

OSK002

OSK003

OSK004

OSK005

HKD127

0

0

0

0

2

0

10

0

HKD128

Predicted class

0

0

0

0

0

9

0

0

HKD184

6

0

0

3

0

0

0

0

OSK001

7

0

5

8

8

0

1

1

OSK002

2

1

12

1

1

2

5

1

OSK003

1

19

0

1

1

7

2

0

OSK004

Table 5 Identified and unidentified K-NET sites by the trained CNN with the seismic H/V spectrum between 2 and 10 Hz

4

0

3

7

0

0

2

0

OSK005

0.20

0.95

0.60

0.40

0.00

0.45

0.50

0.90

Recall

Machine Learning for Estimation of Surface Ground Structure by H/V … 79

80

T. Tobita and W. Yamamoto

(2) The correlation between the seismic H/V spectral ratio and the microtremors H/V spectral ratio at the K-NET stations is confirmed. By focusing on the frequency range between 2 and 10 Hz, where the seismic and microtremor H/ V spectral ratios seem to coincide, an attempt was made to classify the sites using a model trained solely on the color spectrum of seismic H/V spectra. The average accuracy was poor at 50%, and the accuracy of each station was biased. The above results indicate that the method can be used to classify the K-NET stations based on the shape of the seismic H/V spectrum. However, at present stage, it is not yet possible to identify the observation sites with high accuracy using a CNN trained solely on the seismic H/V spectral ratios. In order to achieve this, it may be necessary to investigate in detail the difference between the microtremor and seismic H/V spectral ratios. Furthermore, if the database which links a microtremor H/V spectrum and a ground structure at various sites is made available, it may be possible to quickly estimate the ground structure by the microtremor H/V spectral ratios with the proposed method.

References 1. Japan Seismic Hazard Information Station (J-SHIS). National research institute for earth science and disaster resilience (NIED). https://doi.org/10.17598/nied.0010. Accessed 2022.11.1 2. Kanai K, Tanaka T (1961) On microtremors VIII. Bull Earthq Res Inst Univ Tokyo 39(1):97– 114 3. Nakamura Y (1989) A method for dynamic characteristics estimation of subsurface using microtremor on the ground surface. Q Rep RTRI 30(1):25–33 4. Tsunekawa H (1997) A neural network prediction model of the principal motions of earthquakes based on the preliminary tremors. Jpn Soc Fuzzy Theory Syst 9(4):551–559 5. Emami SMR, Harada T, Iwao Y (1996) Prediction of peak horizontal acceleration using an artificial neural network model. Struct Eng/Earthq Eng JSCE 13(2):111s–118s 6. Yanase M, Maruyama Y (2019) Estimation of the locations of liquefaction occurrences based on covariance structure analysis and support vector machine. J JSCE A1 75(4):I_133–I_143 7. Kubo H, Kunugi T, Suzuki W, Suzuki S, Aoi S (2020) Hybrid predictor for ground-motion intensity with machine learning and conventional ground motion prediction equation. Sci Rep 10:11871. https://doi.org/10.1038/s41598-020-68630-x 8. Krizhevesky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. Proc NIPS 1097–1105 9. Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. In: Fleet D, Pajdla T, Schiele B, Tuytelaars T (eds) Computer vision—ECCV 2014. ECCV 2014. Lecture notes in computer science, vol 8689. Springer, Cham 10. Endo R (2016) Definition of a function to calculate RGB of light. Incorporated nonprofit organization natural science. https://www.natural-science.or.jp/article/20160513143413.php. Accessed 2022.11.1 11. National Research Institute for Earth Science and Disaster Resilience. NIED K-NET, KiKnet. National Research Institute for Earth Science and Disaster Resilience. https://doi.org/10. 17598/NIED.0004. Accessed 2022.11.1 12. The Math Works, Inc. (2022) MATLAB

Regular Deformation-Based Landslide Potential Detection with DInSAR—A Case Study of Taipei City Kuo-Lung Wang , Jun-Tin Lin, and Shih-Yuan Lin

Abstract Seventy percent of Taiwan consists of hillsides, and due to early urbanization in Taipei City, the safety of buildings and roads on hillsides is of paramount importance, especially given the increasing population density. This research aims to use remote sensing methods for the early detection of potential slope deformation. The current process typically involves visual image interpretation and on-site exploration, but SAR remote sensing technology can also be utilized to observe surface deformation trends. This technology can be used for potential landslide investigation, slope deformation, surface subsidence, seismic deformation observation, volcanic deformation observation, surface subsidence observation in mining areas, etc., to obtain long-term deformation trends. This study employs DInSAR technology to conduct a multi-sequence investigation and analyze possible variable slopes. Using the number of deformation points and the amount of deformation in the slope unit, the potential classification of the slope is carried out by comparing and analyzing the in-situ survey method and observation data. The DInSAR data is combined with both ascending and descending orbits to understand the possibility of surface deformation in Taipei City. The analysis results are consistent with the findings after on-site investigation, and the program has been adopted by the Taipei city government for several years. Keywords SAR · Slope deformation · Landslide hazard prevention

K.-L. Wang (B) · J.-T. Lin National Chi Nan University, Nantou 545, Taiwan e-mail: [email protected] S.-Y. Lin Taipei City Government, Taipei 110, Taiwan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_8

81

82

K.-L. Wang et al.

1 Introduction To understand the potential risks associated with slopes on hillsides in Taipei City, the city conducts continuous surveys, observations, and patrols every year to explore areas with potential hazards and comprehensively understand whether sensitive slopes have the potential for slippage. In addition to the current methods, it is feasible to use remote sensing detection methods to conduct multi-sequence investigations and analyze possible variable slopes in advance to obtain a comprehensive surface deformation trend. This can help in a warning and reducing losses. Remote sensing investigations should be used to analyze the slope’s continuous variation signs, geological features, major preservation objects, etc. This study adopts a widearea detection method for analyzing disaster prevention management strategies in Taipei City.

2 Literature Review Differential interferometric synthetic aperture radar (DInSAR) technology utilizes the phase change information of radar to calculate surface deformation, building upon the InSAR technology. It has been widely employed in various fields in recent years, including seismic displacement measurement [5], mine monitoring, stratum subsidence [2], volcanic deformation, glacier displacement, and slope sliding [6], among others. Various approaches have been developed to enhance the accuracy of DInSAR, such as the removal of atmospheric effects, orbit correction, and the installation of corner reflectors [9]. On the other hand, the use of multi-sequence differential interferometric images has been further developed based on DInSAR technology, such as the short baseline method SBASInSAR technology, PSInSAR technology that utilizes the characteristics of permanent scatterers, and SqueeSAR technology. However, despite the advantages of DInSAR technology in interpreting surface microscopic image deformation, it also has multiple sources of error and computational limitations. Slope sliding monitoring is challenging due to factors such as severe terrain fluctuations, atmospheric effects, dense vegetation coverage, and excessively fast slope sliding speed. Traditional DInSAR technical analysis has limitations in this regard. However, the MTInSAR technology of time series analysis has been applied to the monitoring of slope slippage, with common methods including PSInSAR, SBASInSAR, etc. For example, Ferretti et al. [3] used PSInSAR to monitor slopes in the Ancona region of Italy, while Hilley et al. [4] monitored the Berkley area in the United States using the same technique. Berardino et al. [1] proposed using SBASInSAR to monitor the slope deformation of Maratea, Italy. Wang et al. [7] verified the comparison of DInSAR results with GNSS data, which wrong orbit data will lead to an improper result and the deformation rate could be controlled by rainfalls.

Regular Deformation-Based Landslide Potential Detection …

83

Various studies have applied different multi-sequence analysis techniques to analyze the situation of a slope sliding, in addition to the cases mentioned above. However, regardless of the analysis mode, it remains challenging to solve the geometric deformation caused by terrain and the low correlation of vegetation coverage areas. Moreover, fast-sliding slopes are prone to de-correlate topographic changes, adding to the difficulty of slope sliding monitoring in Taiwan due to terrain constraints. Therefore, establishing an effective monitoring model is an important research topic for researchers. Due to the better vegetation conditions in Taiwan, the one with the longer wavelength has better results in terms of accuracy and analysis feasibility when performing differential interferometric SAR analysis. However, the Stripmap mode used in Japan for high-resolution images of Taiwan has been decreasing in recent years. Even some areas only have one Stripmap image per year. When the time is too long, the surface changes too much and the coherence becomes poor. It is difficult to use it for longterm or event-based detection. This research is based on the Sentinel-1 satellite of the European Space Agency. Sentinel-1 satellite is two satellites, and the shortest imaging interval can be shortened to 6 days. For regular monitoring, a 12-day revisit cycle is sufficient. It can effectively detect potential sliding blocks.

3 Study Area 3.1 Data for Analysis In this case, the remote sensing analysis targets the potential sliding slopes in Taipei City, and the changes in the past five years are observed. L-band is the most suitable for mountain observation, but unfortunately, the Japanese ALOS satellite only has about 2–3 images per year in Taiwan. The European Sentinel-1 is an Earth observation satellite in the European Space Agency’s Copernicus Program (GMES). It consists of two satellites (Sentinel-1a and 1b) equipped with C-band synthetic aperture radar (SAR). The radar waves are 5.6 cm long, with spatial resolutions of 5 × 5, 5 × 20, 25 × 100, and 5 × 20 m, and image widths from 80 to 400 km. Sentinel-1 provides continuous imagery (day, night, and all weather) for monitoring surface movement and surveying land features such as forests, water sources, and soil. Its most significant feature is that data can be downloaded from the European Space Agency website. In Taiwan, only VV and VH polarization data are available. The imaging period in Taiwan is from the end of October 2014 to the present, with a revisit period of 6–12 days. The images of ascending orbit 69 and descending orbit 105 in Taiwan area are shown in Fig. 1.

84

K.-L. Wang et al.

Fig. 1 The coverage of ascending orbit 69 and descending orbit 105 in Taiwan

3.2 Large-Scale Potential Sliding Block Detection Based on the SBAS method using the differential interferometric synthetic aperture radar analysis and interferometric stacking technique, multi-sequence surface deformation data can be obtained to effectively detect the sliding block, including its range and scale. However, the large-scale analysis of surface deformation information may include surface subsidence and uplift, which need to be screened to confirm the location of the potential sliding block. Initially, the subsidence area is selected as the potential sliding block. After obtaining large-scale surface deformation and subsidence data, the cause of slippage is judged. As slippage in Taiwanese hillsides is often a continuous creep, it is difficult to predict destructive disasters. Therefore, this study selects the deformation rate acceleration position generated in the multi-series data to identify potential sliding events caused by external factors such as torrential rain or earthquakes, as shown in Fig. 2. Sentinel-1 images from the past 5 years are used, with an image every 12 days. To avoid slope changes caused by noise, checking every 96 days effectively reduces the chance of misjudgment. The analysis flow is shown in Fig. 3. The analysis produces two types of potential slip information: long-term potential deformation slope and event-type slip controlled by heavy rain or earthquakes. However, the latter may also be man-made, requiring site examination and confirmation. Taking Wang et al.’s [8] Neihu District screening as an example, setting the ratio of the quarterly deformation rate to the long-term deformation rate in the analysis process may be one of the key screening conditions. Take track number 69 as an example, analyze the terrain change rate checks in 12 seasons during the 3-year period. The cumulative results of the ratio points of one season to the long-term deformation rate are consistent in each season. The double tangent method is used for the bending situation of the cumulative results. The selectable rate-of-change

Regular Deformation-Based Landslide Potential Detection …

85

Fig. 2 Schematic diagram of rate-of-change screening

Fig. 3 Large-scale potential sliding slope screening process

screening value ranges from 1.05 to 1.35. In order to obtain consideration for each change event prudently, the screening value of the rate of change is selected to be 1.1.

4 Results and Discussions The analysis period for this study spans 5 years from March 2017 to March 2022. The analysis uses the short baseline method, and the time period between two images should not exceed 72 days during the screening process. This setting is most effective for shorter wavelengths and enhances the image correlation with Sentinel-1. The short baseline method of multipath analysis provides an opportunity for repeated verification. The spatial baseline length for screening is set at 150 m to ensure the final product’s accuracy. Since different orbits may provide varying topographical

86

K.-L. Wang et al.

advantages for slope detection, the detection points and results will be used to filter the orbit results based on the slope unit’s direction after the analysis. This study focuses on identifying potential sliding blocks in Taipei City using the short baseline method, which yields a point density of more than 1000 points per square kilometer. This high density of points is beneficial for sliding block detection. In the subsequent processing of the analysis results, effective deformation detection is conducted for each orbital direction. As mentioned previously, the moving window method is utilized to compare the deformation rate with the long-term deformation rate, and it is recorded once when the detection is successful. The number of effective points located in the slope unit is summed up. The number of points per unit area, the average displacement of the effective points, and the average total displacement of the effective points in the slope unit are calculated. The results are then sorted accordingly. To achieve the original analysis objective, the average total displacement of effective points in each slope unit was visualized in Fig. 4. The colored slope units represent the threshold value of 4 mm/month for monthly vertical deformation that was reached during the analyzed period. This value was set based on the typical attention value for slope sliding displacement of 2 mm/month and the requirement for continuous deformation. The analysis results revealed that the slope units were relatively scattered and did not form a complete sliding block. Another significance of using this deformation standard is that when calculated every 3 months, the deformation exceeding 12 mm will be screened out. And 12 mm is the drift value that may occur in the process of using the SBAS method, because short-term analysis may cause misjudgment. For the sake of prudence, it may be necessary to increase the displacement threshold.

5 Conclusion Remarks This research develops a method to detect ground changes and obtain the amount and rate of vertical surface deformation. Based on the analysis of the past five years, reasonable results have been obtained under long-term monitoring. The main conclusions and suggestions are as follows: 1. When using the point-count method per unit area, the displacement may be too small to determine whether it is a priority site for survey or processing. When using the point-displacement method, the advantage of the slope unit to form a sliding block cannot be presented. The long-term average displacement of the detection point can show the actual problematic points during the presentation. 2. When detection is an event, the moving average method is used for detection. After multiple detection statistics screening, more confirmed surface changes can be screened out compared to fixed detection every season.

Regular Deformation-Based Landslide Potential Detection …

87

Fig. 4 Deformation-based landslide potential map of Taipei City

3. There are many limitations when using SAR analysis. This study attempts to overcome these limitations and present a large-area slope unit change situation to identify possible sliding blocks. However, large-area detection may lead to omissions. If the analysis target is focused on the edge of the slope and the gradient is not high, the changing trend and sliding mechanism analysis in the key slope should be obtained from the analysis of the key slope on a small-scale basis. 4. The results of the current analysis are presented using a five-year period for final results. However, the conditions for landslides change gradually over time. It is recommended to present landslide deformation potentials in terms of annual or even quarterly landslide deformation potential if possible. In this way, the dynamic changes of each landslide site can be detected.

88

K.-L. Wang et al.

References 1. Berardino P, Fornaro G, Lanari R, Sansosti E (2002) A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Trans Geosci Remote Sens 40:2375–2383 2. Chang C-P, Wang C-T, Wang H-C, Chen K-S (2004) Application of DInSAR in monitoring the metropolitan land-surface deformation: Jungli industry park as an example. J Photogramm Remote Sens 9(3):9–14 3. Ferretti A, Prati C, Rocca F (2000) Nonlinear subsidence rate estimation using permanent scatterers in differential SAR interferometry. IEEE Trans Geosci Remote Sens 38(5):2202–2212 4. Hilley GE, Burgmann R, Ferretti A, Novali F, Rocca F (2004) Dynamics of slow-moving landslides from permanent scatterer analysis. Science 304(5679):1952–1955 5. Tong X, Smith-Konter B, Sandwell DT (2014) Is there a discrepancy between geological and geodetic slip rates along the San Andreas Fault System? J Geophys Res Solid Earth 119:2518– 2538 6. Wang K-L, Lin J-T (2014) Investigation of potential landslide with differential interferometry using ALOS PALSAR—an example of May river watershed. J Photogramm Remote Sens 19(3):253–265 7. Wang K-L, Lin J-T, Chu H-K, Chen C-W, Lu C-H, Wang J-Y, Lin S-H, Chi C-C (2021a) Highresolution LiDAR digital elevation model referenced landslide slide observation with differential interferometric radar, GNSS, and underground measurements. Appl Sci 11(23):11289 8. Wang K-L, Lin J-T, Wu C-C, Chen C-F, Chiu T-W, Hsieh M-H (2021b) Study on sliding displacement classification of slope land—an example of Neihu District, Taipei City. J Chin Inst Civ Hydraul Eng 33(2):139–149 9. Zhou Y, Li C, Ma L, Yang MY, Liu Q (2014) Improved trihedral corner reflector for highprecision SAR calibration and validation. In: 2014 IEEE geoscience and remote sensing symposium, Quebec City, QC, pp 454–457

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient Tunnel Maintenance Motoki Sato, Nobusuke Hasegawa, Hiroki Ohtsuka, and Erika Kukisawa

Abstract In tunnel inspection, close visual inspection and hammering sound test are usually carried out to check the deformation and deterioration of lining concrete. These inspections are carried out using elevated work vehicles to do the close visual check and hammering test. The deformation cracks and deterioration area found during the inspection are marked on the concrete lining by chalk. These marks are then recorded as sketches. Since these inspection works need to be done by a large number of technicians and are time-consuming works, it is necessary to improve the efficiency of the works and keep the quality of the sketching to avoid differences due to the sketching skills of technicians. In order to achieve these purposes, we developed two technologies. The first one is to obtain developed diagrams of lining surface automatically from point cloud data acquired by laser scanners. The second one is to extract cracks automatically from developed diagrams using AI (artificial intelligence). We are able to improve the efficiency and quality of inspection works with these technologies. Keywords Tunnel inspection · Laser scanner · AI · Compression sensing technology · Close-up visual inspection · Sketch

M. Sato (B) · N. Hasegawa · H. Ohtsuka · E. Kukisawa OYO Corporation, Saitama, Saitama, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_9

89

90

M. Sato et al.

1 Introduction In Japan, many road tunnels were constructed during the period of high economic growth from the 1950s to the 1970s. These tunnels are now aging, and concrete deterioration of lining and structural stability become serious problems. In order to ensure safety of tunnel users, road administrators in Japan are obliged to conduct periodic inspections once every five years for more than 10,000 road tunnels [1]. However, the shortage of inspectors with appropriate skills and knowledge has become serious, and it will become more serious in future, because the number of tunnels require maintenance increases year by year and the number of inspectors decreases year by year due to the declining birthrate and aging population. Therefore, in order to reduce the burden of inspection works, the Ministry of Land, Infrastructure, Transport and Tourism has issued a guideline [2] for utilizing inspection support technology and encouraged local governments to introduce new technologies [3]. Figure 1 shows the flow of inspection works. The inspection works are divided into on-site work and office work. In on-site work, inspectors use an elevated work vehicle to check deformation (cracks, lifting and peeling of lining concrete) by closeup visual inspection and hammering sound test. The checked deformation is marked on lining, and the marks are sketched on paper by inspectors. These on-site works are carried out under traffic regulation. In office work, sketches are drawn manually on CAD drawings, soundness diagnosis is made from the state of deformations, and the results are reported as inspection record and stored in database. We have developed a system (we call the system the 3-D laser measurement system [4]) to improve efficiency by automating the work from sketches by inspectors to CAD drawings. In this paper, we describe the measurement system using a 3-D laser scanner and an automatic detection of chalk-marked cracks using AI (artificial intelligence).

Fig. 1 Flow of tunnel inspection work

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient …

91

2 Development of 3-D Tunnel Laser Measurement System [4] 2.1 Acquisition of Images by 3-D Laser Scanner 3-D laser scanner acquires 3-D point cloud data. It is easy to obtain a developed diagram from 3-D point cloud data since each point cloud data has coordinates. If chalk-marked cracks can be extracted from 3-D laser scanner images instead of human sketches, the following can be expected: (1) The sketch work time is shortened. (2) The risk of overlooking the lining marking in sketches is reduced, because the sketch is performed within limited time, including the shadow of a projector. (3) The positioning accuracy of marking is enhanced. Figure 2 shows a comparison of a camera image and a 3-D laser scanner image. The 3-D laser scanner is FARO Focus 3D Laser Scanner (FARO Technologies, Inc.). This is a 3-D laser scanner that is widely used on the market. Left image is obtained by a camera, and right image is obtained by 3-D laser scanner. The 3-D laser scanner images also detect cracks marked with chalk. However, it is a little blurry compared to the camera image. Acquiring data of the entire circumference of the tunnel with a 3-D laser scanner is much faster than photographing with a camera, and if the chalkmarked cracks can be extracted from the images obtained by the 3-D laser scanner, the advantage will be great.

Fig. 2 Camera image (left) and 3-D laser scanner image (right)

92

M. Sato et al.

Fig. 3 Comparison of image

2.2 Utilization of Compression Sensing Technology Since the images acquired by the 3-D laser scanner are a little blurry compared to the camera images, we decided to apply compression sensing technology to detect chalkmarked cracks clearly. The compression sensing technology is used in various field as a technology for obtaining super-resolution images. Figure 3 shows a comparison of the original image and the image obtained by the compression sensing technology, and the result of automatic crack extraction by AI. The red lines are the crack automatically extracted by AI. Cracks are more pronounced by applying compression sensing technology. However, although chalked cracks can be seen in images using compressed sensing technology, there are some cracks that cannot be extracted by AI. It is a future task to be able to extract these.

3 Overview of 3-D Tunnel Laser Measurement System 3.1 Method Figure 4 shows the flow from data acquisition to CAD drawings. After acquiring 3-D point cloud data with 3-D laser scanner, a developed diagram is created, crack enhancement processing using compression sensing technology is performed, cracks are detected by AI, and CAD drawings are created. The process is performed almost automatically. Each process is described below.

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient …

93

Fig. 4 The flow of analysis procedure

Field Measurement. Table 1 shows the specification of measurement, and Fig. 5 shows the situation of the measurement. Measurements are made every lining span (approximately 10 m). Analysis (Automated by AI) (1) Automatically, it generates a 2-D lining development diagram from the 3-D laser scanner-acquired 3-D point cloud data. In this process, a clear image can be obtained by correcting the shadows appearing on the edges of the image using compressed sensing technology. (2) Since deformations (such as cracks) marked with chalk on the lining concrete are recorded in the lining development images which are obtain by the above processing, AI technology extracts the deformation information (only cracks). The AI system used here is trained in advance using training data. (3) Extracted deformations (cracks) can be highlighted on the lining development diagram and automatically output as CAD data. Drawing Deformation Development The procedure for creating a deformation development diagram is shown below. Table 1 Measurement specifications Measurement frequency

Once every 10 m

Measurement time

About 2 min per 10 m of a two-lane road tunnel

Usage time

Up to 5 h (battery replaceable)

Measurement environment

5 ~ 40 °C No thick fog inside the tunnel (No condensation on the measuring instrument lens)

Instrument

Weight: 7.2 kg Dimensions: width 1.2 × depth 1.1 × height 1.75 m (When using a tripod)

Measurement error

± 2 mm (for a distance of 25 m)

94

M. Sato et al.

Fig. 5 On-site measurement situation

(1) Draw lining span frames, span numbers, etc., on CAD. (2) Allocate the CAD data of the cracks generated by the above methods to a predetermined layer because CAD layer for drawing tunnel development diagrams is defined and add other deformation information (such as numerical value of crack width). (3) Allocate CAD data to the inspection report, etc., as a deformation development diagram.

3.2 Features The features of the system are as follows: (1) By performing laser scanner measurements during tunnel inspection work, sketching work can be omitted, making inspection work more efficient. (2) High-precision 2-D lining development diagram can be created by using 3-D point cloud data. (3) By acquiring 3-D point cloud data, deformation of tunnel arches, side walls, and road surfaces can be captured. (4) The use of commercially available 3-D laser scanners makes it easier to procure equipment and eliminates the need for specialized measurement skills. (5) By using an infrared laser, illuminating the tunnel lining surface during measurement would not be needed, and the backlight from the tunnel lighting does not interfere with the measurement.

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient …

95

* Trial calculation based on tunnel length of 400m

Fig. 6 Time saving by 3-D laser measurement system

4 Result Obtained from System Development 4.1 Efficient Sketching This system finishes sketching work in about two minutes while the sketching work of a traditional tunnel inspection takes about 10 min per a tunnel span. In total, the work time can be reduced to about 1/5. In addition, since the efficiency of creating deformation development diagrams can be improved, the overall inspection work time can be reduced by about 10% in case of an average road tunnel (about 400 m length) (Fig. 6). Shortening the sketching time contributes to shortening the time required for traffic regulation, which is expected to reduce the inconveniences of drivers and passengers and CO2 emissions by shortening engine idling time.

4.2 Improved Accuracy of Deformation Development Diagram In Fig. 7, the cracks extracted by AI on the sketch diagram are shown with green lines, and the background is the result of conventional sketch work. The many positions of the cracks drawn in the sketch and measured by this system do not match. By using this system, it has become possible to accurately visualize the shape and position of deformation.

96

M. Sato et al.

Fig. 7 Higher accuracy of deformation map by utilizing new technology

5 Conclusion Using the point cloud data of the tunnel measured by the laser scanner, we built a system that creates a developed diagram of the tunnel lining. We have succeeded in automatically extracting the deformation by recording the result of marking the deformation found in the inspection in the developed diagram and recognizing the chalk-marked cracks with AI. In the creation of developed diagram, we were able to create clearer images by utilizing compressed sensing technology. The results are accurate inspection and increased the efficiency of inspection work. In future, we will continue to develop technology to extract deformed parts that have not been marked.

Utilization of AI-Based Diagnostic Imaging for Advanced and Efficient …

97

References 1. Road Bureau, Ministry of Land, Infrastructure, Transport and Tourism (2019) Road tunnel periodic inspection guidelines, p 1 2. Ministry of Land, Infrastructure, Transport and Tourism (2019) Guidelines for the use of new technologies (draft), p 2 3. Policy Bureau, Ministry of Land, Infrastructure, Transport and Tourism (2021) Guide for introducing new technologies in infrastructure maintenance (draft)—introducing new technologies is not difficult, p 2 4. Ministry of Land, Infrastructure, Transport and Tourism (2021) Inspection support technology performance catalog

Application of DX and i-Construction

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok Shih-Hao Cheng, Min-Ju Li, Ricky K. N. Wong, Takeshi Iwakubo, and Shinichi Ueno

Abstract The Tube-a-Manchette (TAM) grouting is commonly used to construct a waterproof base grouting. It is conducted by injecting grout at a fixed injection rate simultaneously through several grout ports with known spatial coordinates in the ground. The pressure response from the ground can be monitored digitally at each injection port through pressure and flow rate recorders. Also, the recorded data are illustrated in 3D coordinates through computer platforms. Typically, the larger the pressure response from the ground, the better the grouting effect. Finally, compared to the previous experience-based evaluation method, the smart grouting system can interpret the TAM grouting results more systematically and rationally in real-time. Keywords Smart grouting · Base grout · Shaft bottom plug · Grouting pressure · Datalogger system

1 Introduction Over the past few decades, cut-and-cover methods are usually adopted for underground infrastructure construction because of the cost-effectiveness, construction space, and construction sequence aspects. However, when excavating extremely deep and when the groundwater table is high, the thickness of the impermeable layer under the excavation base is the critical factor that influences construction safety. To improve construction safety, base grouting, as a compensation measure, is often used S.-H. Cheng (B) National Taiwan University of Science and Technology, Taipei, Taiwan (R.O.C.) e-mail: [email protected] M.-J. Li · R. K. N. Wong SANSHIN CORPORATION, Tokyo, Japan T. Iwakubo SANSHIN CONSTRUCTION (THAILAND), Bangkok, Thailand S. Ueno TOMEC CORPORATION, Tokyo, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_10

101

102

S.-H. Cheng et al.

to increase the thickness of the impermeable layer and enhance the excavation base’s strength and water tightness. This study presents a smart grouting record developed with the Tube-a-Manchette (TAM) grouting system. Also, a deep excavation project in Thailand was used to demonstrate the application of automatic grouting pressure management. Given grouting quality control, such a system can real-time diagnoses the grouting conditions and take countermeasures. The automatic recorded system is based on real-time grouting pressure and flow management. Therefore, the digital record can reduce the impact of paper on the environment, also compared with the conventional grouting system in this paper [1, 2].

2 Site Condition The dimensions of this deep excavation were 12 m in length and 8 m in width. The excavation depth was 34.2 m and was retained by a 57.5-m-deep diaphragm wall, as shown in Fig. 1. The bottom-up construction sequence was adopted for this excavation. The groundwater level observed was around 1.2 m below the ground surface. A base grouting work was to be conducted above and below the base of the diaphragm wall to cut off the groundwater seepage from the sand layer and to reduce the risk of groundwater uplift failure. TAM grouting method was adopted in this project. As Fig. 1 shows the subsoil condition of the project in Thailand, the groundwater level at the site is 11 m below the surface, and the soil profile was formatted with clay and sand interlayer. Among them, the first layer (at a depth of 0–17.5 m) is a soft clay layer. The second layer (at a depth of 17.5–28 m) is the 1st stiff clay layer. The third layer (at a depth of 28–39 m) is the 1st sand layer, and the SPT-N value was 28. The fourth layer (at a depth of 39–47 m) is the 2nd stiff clay layer, and the SPT-N value was found to be 15–38. The last layer is the 2nd sand layer, and the SPT-N value was approximately 47.

3 Outline of TAM Grouting Method TAM grouting method was adopted for the base grouting in this study. The grouting site uses the grouting volume for quantitative control. Grouting pressure at each step was recorded while the design grouting volume grouted into the ground. In this case, when the ground was excavated to more than 11 m below the ground surface, the groundwater pressure became more significant than the pressure from the ground surface, causing the excavation base to uplift and piping. In this case, the bottom plug grouted below the excavation base is about 576 m3 . After the grouting is completed, the permeability coefficient must satisfy the design requirements (k = 1 × 10−6 ~ 5 × 10−7 m/s).

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok

103

Fig. 1 The soil profile of the grouting site

The work used two stages of grouting: the initial grouting used cement and bentonite grout (CB grout) with an injection volume of 3 ~ 10% soil porosity for the 2nd sand layer. Similarly, the soil layer was injected during the secondary grouting, a solution-type sodium silicate grout (with MK reagent) with a 30 ~ 37% soil porosity volume. Table 1 shows the summary of the base grouting work. The initial grouting using CB grout was aimed at filling and compacting the voids in the ground. Such grout compressed the sandy soil and made the improved zone homogenization. Using sodium silicate grouting in secondary grouting reduces the permeability and increases the thickness of the impermeable layer. Table 1 Summary of the grouting zone and grouting parameters in this study Area

TAM grouting stage

Soil volume (m3 )

Grout ratio (%)

Quantity (nos.)

Drilling length (m/ hole)

Grout flow (L/min)

Base

Initial/CB grout

576

3 ~ 10

60

57.2

7 ~ 12

Secondary/ SL grout

30 ~ 37

7 ~ 12

104

S.-H. Cheng et al.

4 TAM Grouting System 4.1 Conventional TAM Grouting System In the past, the management of TAM grouting was based on quantitative management methods. The grouting volume was set according to the soil properties and conditions and whether the grouted zone reached the designed permeability requirement. The flow measurement controls the grouting volume, as shown in Fig. 2. After the grouting, a grouting report and accumulative volumes must be prepared. It takes time to sort out and analyze the grouting effect in cases with large number of grouting holes. Figure 3 shows the manually recorded condition on the grouting site. However, manual recording is not synchronized with grouting time, it is difficult to know the real-time flow and pressure changes when confirming changes in the grouting process. It may increase the risk of recording errors.

Fig. 2 Traditional flow measurement control system

Fig. 3 Conventional manual recording on the field

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok

105

Two grouting stages were adopted in this base grouting work: the suspension type cement and bentonite grout (CB grout) was used in the first-stage grouting; the solution-type sodium silicate grouting (SL grout, with MK reagent) was used in the second-stage grouting. The porosity of sand layer was about equal to 33–47%. The goal of the base grouting was to fill 90% of the soil porosity. So the grout volume equal to 30–42% by soil volume was to be injected in this project. The grout volume injected was further divided to 10% by soil volume for CB grouting and 30% by soil volume for SL grouting.

4.2 A Smart TAM Grouting System/Grout Datalogger System/TDS-CG Applied a digital management system into the grouting procedure can effectively reduce the manual recorded error and increase the efficiency of recorded data processing. Figure 4 shows the smart TAM grouting system embedded in a grouting procedure that can automatically record the flow data on the computer platform. Typically, the conventional flow measurement is used to record a single grouting hole. In contrast, the smart grouting system can be integrated into 8 grouting holes in the datalogger system, saving manpower and recorded time. Figure 5 shows the equipment of the smart TAM grouting system in the grouting site. Compared with the conventional method, the new system can reduce human errors through computerized records and enhance quality control. Also, labor can be reduced more than in the traditional approach.

Fig. 4 Layout of a smart TAM grouting system embedded in the grouting procedure

106

S.-H. Cheng et al.

Fig. 5 The equipment of the smart TAM grouting system

5 Quality Control and Grout Result 5.1 Grout Pressure Distribution After the grouting is completed, the pressure distribution is according to the average pressure injected in each stage. Using the color to distinguish the grouting pressure can determine the grouting pressure at the injection hole in each step (0.33 m) by comparing them with each other to evaluate the grouting effects and detect the insufficiently improved area [3]. In the data recording of the new system, the realtime data can also be used to create a pressure contour distribution using the 3D drawing software, which is more beneficial for figuring out the pressure distribution. Figure 6 shows the grouting pressure distribution at different depths after the secondary grouting. External splitting and other conditions may result in low pressure during the grouting process. Therefore, use the pressure distribution to re-check grouting (re-grouting) for specific areas so that the improvement of the soil layer at the same depth can achieve a more uniform effect, as shown in Fig. 7.

Fig. 6 Pressure distributions of secondary grouting for the base grouting

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok

107

Fig. 7 Pressure distributions of re-check grouting for the base grouting

5.2 Drilling Alignment of the Grouting Holes In order to improve the construction quality, the drilling accuracy is also strictly controlled. Due to different soil conditions, it is easy to cause the drilling skew and misalignment in large-depth drilling. In this case, the drilling accuracy and the displacement of the grouting pipe affected by the grouting pressure are also collected as a reference. To facilitate the measurement of drilling alignment, the shape acceleration array (SAA device) was used to measure the verticality of grouting holes. The SAA device comprises a series of electromechanical sensors with a length of 0.5 m. The sensors are connected in series with flexible joints, which can be customized to any size [3]. The outer diameter (ϕ23 mm) of the SAA device was sufficiently small to be inserted directly into the drill rod. Thus, removing the drill rods is unnecessary when measuring the drill alignment. Photographs of the SAA device components are shown in Fig. 8.

Fig. 8 a Photographs of the SAA components. b Measurement of inner TAM pipe

108

S.-H. Cheng et al.

As shown in Fig. 9, the drilling alignment for the grouting was measured by inserting the inner drill rod after the drilling was completed. The drilling depth is approximately 57.2 m, and the drilling accuracy is about 1/150 (according to the XY combined direction). Table 2 shows the result of the accuracy measurement on three terns about drilling, before and after grouting. Using the inner TAM pipe to measure the displacement, collect monitoring data before and after grouting to know the influence of the lateral pressure of grouting on adjacent holes. Figure 10 shows one alignment measured result before and after grouting. There is an apparent displacement in the bottom grouting area, about 0.05–0.08 m. The displacement was caused by lateral pressure during injection at the proximal grout ports.

Fig. 9 Measured deviation profiles of grouting hole

Table 2 Summary of the accuracy measurement Item

Drilling (inner drill rod)

Installing (inner T.A.M. pipe)

After grouting (inner T.A.M. pipe)

Min

1/220

1/190

1/158

Max

1/105

1/80

1/55

Avg.

1/140

1/120

1/105

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok

109

Fig. 10 Compared with hole alignment between before and after grouting

5.3 Field Pumping Test To verify the effectiveness of this base grouting project, on-site pumping test and rising head test were carried out to measure the permeability of grout zone in the excavation site. The pumping test was carried out by pumping out the adequate amount of water to keep the phreatic surface of groundwater table of the whole test area falling at the same rate by Cheng et al. [1]. The coefficient of permeability (k) of the test site was calculated using Eq. (1) and the data gathered from the pumping test. k=

Q×D A × h

(1)

where Q: pumping rate, A: area of the test site, D: thickness of base grouting zone, h: drop of ground water surface in the observation wells. In general, the measured k-values were in the order of magnitude of 3.80 × 10−5 cm/s. It met the requirement set by the owner that k-value of the base grouted zone must be < 5 × 10−5 cm/s, which is about two orders of magnitude lower than that of the in-situ sandy soil before grouting.

110

S.-H. Cheng et al.

Fig. 11 Configuration of the pumping test

6 Conclusions TAM grouting with automatic grouting pressure recorded system was carried out as the field test for the excavation case. The following conclusions can be drawn from the findings of this study: 1. In this study, real-time records can be obtained from TAM grouting system with digitally recorded functions to facilitate quality control and examination of grouting behavior. In addition, the automated recorded system can save data processing time. 2. Each grouting hole has its spatial coordination, so the grouting information inside the grouting area can be presented in a 3D pressure contour. Compared with the traditional single-hole pressure distribution, it can supply better information about foundation injection. In addition, supplementary grout in a more rational way to enhance the water tightness of the base grout can also serve as a better reference. 3. The pumping test was used to check the effectiveness of grouting. The test results show that the permeability coefficient was 3.80 × 10−5 cm/s, which met the design standard (Fig. 11).

A Smart TAM Grouting System for the Shaft and Base Grouting in Bangkok

111

References 1. Cheng SH, Liao HJ, Hatakeyama K, Wong RKN, Iwakubo T (2015) T.A.M. grouting to reduce artesian water pressure acting on the base of an excavation. In: Proceedings of international conference on soft ground engineering. Singapore, pp 873–881 2. Hayashi K, Matsubayashi Y (1996) Chemical grouting contributed to the urban development in Hiroshima city. In: Proceedings of grouting and deep mixing. USA, pp 299–302 3. Cheng SH, Liao HJ, Yamazaki J, Wong RKN, Iwakubo T (2020) Alignment of vertical and inclined jet grout columns for waterproofing. Geotech Test J 43(2):325–338

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project Executed by Unmanned Construction Method Using i-Construction Shinichi Nomura, Naoto Yamagami, and Tsuyoshi Nakade

Abstract In 2016, a large section of a slope collapsed in the Aso Ohashi area in the wake of the 2016 Kumamoto Earthquake (main shock). We accepted a project to restore the damaged infrastructure from this huge disaster as quickly as possible. Rapid infrastructure recovery from natural disasters is an important and urgent task as well as an important social issue. Since infrastructure needing to be restored from an unexpected large disaster is very difficult and challenging, it is necessary to adopt a comprehensive approach combining various technical resources. We had already accumulated years of experience in the field of unmanned construction projects to cope with this challenging project, and developed a network-compatible unmanned construction system, never attempted before. With this approach, we reconstructed the collapsed steep slope depending only on remote operation, by making the affected site off-limits to human entry. The project combined 3D models and ICT in all processes including surveys, design, construction, and measurement, and introduced i-Construction, utilizing concurrent engineering, making it possible for us to achieve successfully, safely, and quickly recover from the disaster. Keywords The 2016 Kumamoto earthquake · Unmanned construction system · i-Construction · Wireless LAN · Concurrent engineering

1 Introduction A huge slope collapse took place in the Aso Ohashi area, in the wake of the Kumamoto Earthquake on April 16, 2008 (referred to as this earthquake)—a huge disaster sweeping away National Route 57, the JR Hohi Main Line, and National Route S. Nomura · N. Yamagami Kyushu Regional Development Bureau, Ministry of Land, Infrastructure, Transport and Tourism, Fukuoka, Japan T. Nakade (B) Kumagai Gumi Co., Ltd, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_11

113

114

S. Nomura et al.

325 and the Aso Ohashi bridge which were located below the slope (Fig. 1). At the upper part of the collapsed slope, open cracks and sharp drops were everywhere. The landslide slopes presented themselves as steep cliffs, with a risk of further collapse due to rainfall and aftershocks. Under these circumstances, the Ministry of Land, Infrastructure, Transport and Tourism, initiated a “directly managed emergency project related to erosion control disaster,” and launched an emergency countermeasure project in May 2016 to prevent secondary disasters which are likely to occur with the collapse of large unstable sediments remaining on the slopes. To study the measures to stabilize the collapsed sloped and restore the transportation infrastructure, a “Technical Study Group for Restoration of the Aso Ohashi Area” was formed consisting of experts and officials relating to erosion control, roads, and railroads. The reconstruction project, receiving advice for examination from the group, had been advanced and completed by July, 2017. Since 2017, a special plan for permanent slope measures for erosion control— managed directly by the government—had been started. Thus, the slope countermeasure work was finished in August 2020, and National Route 57 was opened to the public in October of the same year. After the elapse of four and a half years since the Kumamoto Earthquake, the key reconstruction projects of transportation infrastructure in the Aso area were completed.

Fig. 1 Collapsed site of the slope (view of the whole site)

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

115

2 Emergency Countermeasures 2.1 Overview of Emergency Countermeasures The countermeasures below were taken as key emergency solutions, in the affected collapsed area to prevent secondary disasters due to collapse of instable sediments remaining in the upper section of the collapsed slope (Fig. 2). Earth-retaining embankment: This is an earthwork intended to prevent secondary disasters which would be caused by collapse of the unstable sediments which remain on the upper part of the slope. The goal of this project was to construct a soil cement embankment in two stages, 3 m high and 5 m wide, top and bottom, considering the height stones might potentially bound and the workability of unmanned construction machinery. Removing of unstable sediments at the head of the slope (hereinafter referred to as “rounding”): For this purpose, heavy equipment is used in remote mode to remove the surface layer (andosols)—a topographical convex formed around the sliding cliff at the top of the slope—as well as rock debris sediment changed into soil, floating rocks, and boulder rocks. Because of the presence of open cracks and differences in grade surrounding the collapsed site in the upper section of the collapsed slope, there was risk of further collapse due to rainfall and aftershocks, all reconstruction work within the collapsed site was unmanned. Since it is impossible for humans to gain access to such a steep and

Fig. 2 Overview of an emergency countermeasure project

116

S. Nomura et al.

huge collapsed slope, we took the following approach for safe and rapid emergency disaster response. • Unmanned construction technology should be upgraded further to cope with the need for a network. • Introduction of a comprehensive i-Construction • Integration of survey, design, and construction processes

2.2 Network-Compatible Unmanned Construction The requirement for reconstruction was that the earth-retaining embankment be promptly constructed and that a large number of unmanned construction machines be put into operation in limited areas of the collapsed site for operation from a safe remote location. Under these circumstances, however, we were faced with various issues in conventional type unmanned construction as detailed below. • Setting up a wireless environment was time-consuming. • Difficulty in switching wireless stations. • Operation of equipment becomes unstable due to jamming and interference between wireless stations. To solve these issues, we introduced a network-guided unmanned construction system which was developed at the time of the Great East Japan Earthquake as a technological solution enabling remote operation within a range of several tens of kilometers. Using this framework concept, all data for operation videos and GNSS (satellite positioning system) are transmitted over IP (Internet Protocol), making it possible not only to connect various devices, but also to utilize wireless resources effectively and to facilitate relays for large volume, high-speed transmission. In addition, the use of fiber optic cables enabled stable, long-distance remote operation, and facilitated free and easy arrangement of control rooms. Such sophistication of the system makes it possible to control unmanned machines and to create a stable operating environment by centrally managing image data, GNSS, and other information. (Fig. 3, Photo 1). For this project, we maximized the use of wireless stations in order to combine optical cables, high-speed wireless access systems, and various types of wireless LANs, and installed an ultra-remote operation room at a kilometer from the construction site, which made it possible to operate up to 14 construction machines in the collapsed site (Photo 2).

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

Fig. 3 Concept of network-compatible unmanned construction

Photo 1 Operation at the remote operation room

117

118

S. Nomura et al.

Photo 2 Unmanned machines and equipment at the collapsed site

3 Introduction of a Comprehensive i-Construction System Considering the situation of a large land collapse site, an emergency construction measure required that humans be forbidden from entering the site, and that it was necessary to perform the measures promptly, we introduced an i-Construction system covering a series of surveys, design, construction, and management as shown in Fig. 4—a construction project based upon a three-dimensional topographic model.

3.1 Investigation It is important to understand the local topography and geological structure after the disaster for designing slope disaster prevention measures. The field survey was extremely difficult since workers were not allowed in the collapsed area, open cracks were distributed in the head, as well as the thick growth of bamboos taller than the height of a human. In such a difficult situation, we were forced to proceed with the survey by making full use of aerial laser surveying and UAV photogrammetric surveying.

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

119

Fig. 4 Overview of an i-Construction project

3.2 Design In designing the earth-retaining embankment, the planar arrangement and structure of the earth-retaining embankment were determined based on the simulation results of rock falls from the slope head, using the aforementioned three-dimensional topographic model on aerial laser surveys and considering the workability of unmanned construction equipment (Fig. 5). In designing the rounding, as a first step, we used ortho images and a 3D topographic model of the top of the collapsed area obtained with oblique directional photography in the UAV survey to identify the condition of floating and boulder stones—possible sources of falling rocks—and to set up the bottom edge of the rounding within the range where these stones would fall. The construction range of the rounding was determined considering the stability gradient of the cut-off slope (1:1.2) and the condition of the back cracks.

3.3 Start-Up of an Unmanned Construction System The unmanned construction system, with a requirement for being rapidly put in operation, was set up in the following two phases.

120

S. Nomura et al.

Fig. 5 3D model of the earth-retaining embankment

Phase 1 (High-Performance Remote-Control Room) The construction system was set up on May 5, 2016, and put in operation at the site. In the first phase, the system was set up with two backhoes, two rough terrain vehicles, and a bulldozer that might be ready for operation in remote mode. Normally, for operation of an unmanned system, it is necessary to provide approximately ten days for the installation and setting-up for the operation of equipment. However, the introduction of a high-performance remote-control room (Photo 3) enabled us to start unmanned construction in as little as three days. In this case, the room was completed in advance with equipment installation and their settings before being transported to the site. By doing so, we were able to begin with the access road leading to the point for starting the earth-retaining embankment as quickly as possible. Phase 2 (Enhanced Remote-Control Room) The second phase preparations were made during the time the unmanned construction was carried out from the high-performance remote-control room, while augmenting the remote-control construction equipment and facilities. The network-compatible operation room described above was installed near 1 km downstream (Photo 4), where the risk of secondary disasters is low. Switching between operation rooms was performed while maintaining compatibility with each wireless environment, and work continuity was ensured by short switching times of ~ 30 min. The second phase began operation in June 2009, with equipment capable of operating 14 construction machines and 21 cameras.

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

121

Photo 3 High-performance remote-control room

Photo 4 Enhanced remote-control room

3.4 Introduction of Construction ICT (Computerized Construction) For this project, the unmanned construction system was equipped with two highprecision GNSS receivers and multiple sensors, using a machine guidance system and machine control (bulldozer earth removal board control system) which are able to centrally manage location information, ICT construction data, and 3D design data in the control box (Fig. 6). The construction machine is normally equipped with a

122

S. Nomura et al.

Fig. 6 Introduction of construction ICT in the process of unmanned construction

controller, but for this project, in order for the controller to be able to be installed in an enhanced remote operation room, a special convertor was developed and used, making guiding, and setting of data, machines and equipment in a remote mode. The plane shape of the embankment is a shape with several folding points and was constructed using operation cameras and on-vehicle cameras in combination with a machine guidance system. The machine guidance was used not only for guiding the shaping of slopes, but also for indicating a scope for mixing improvement materials as well as served as a survey device for completed shapes of construction. In the ground surrounding the construction site, there is a lot of accumulated andosols, and it was feared that the unmanned construction machine might settle and slide if it rained. To cope with this issue, the leveling function of the machine guidance was used to confirm the inclination of the unmanned construction machine (Photo 5). If the cameras are not placed appropriately, they could not recognize delicate angles from video images only, and determining the tilt of the machine must be made based on the operator’s experience. Therefore, the angle of the level was grasped quantitatively which enabled supplying the operator with important information for decisions, and to avoid major problems.

3.5 Construction of the Drainage Structure for Earth-Retaining Embankment Since in the collapsed area, a tremendous amount of surface water flows down during and after rain, it is important to build adequate drainage for the earth-retaining embankment. On the other hand, in the process of unmanned construction, the

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

123

Photo 5 Level-based guidance for unmanned construction machine

drainage on general pipeline structures raised some issues in terms of installation, construction precision, and workability especially in the process of burying drainage pipes and box culverts and installation of slits. To cope with these issues, we employed a process which was able to be implemented in parallel with the embankment work and was likely to pose less problems in workability in terms of transportation and placement of materials, installation, and adopted a specially designed drainage structure as shown in Fig. 7, laying large gravel (about 20 cm) evenly in parts of the embankment, while the downstream side was protected by placement of bagged cobbled stones (Photo 6). Fig. 7 Concept of the construction of drainage

124

S. Nomura et al.

Photo 6 Placement of bagged cobble stones

The use of this process enabled us to build a drainage structure with a planned drainage capacity by unmanned construction. With regard to rainfall during construction, this system was equipped with adequate drainage functions, which we hope are extended to a future applicability for similar projects (Photo 7). Photo 7 Completed status of earth-retaining embankment

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

125

Photo 8 Construction of the rounding

3.6 Construction of the Rounding The area along the rounding was constructed with a high-altitude slope excavator hung via cable from the anchor which was installed at the top. The construction area was divided in three work blocks, in a manner so that adjacent construction machines would not interfere with each other, and the work procedure was planned in order that multiple machines in service are not aligned in an up-and-down direction. Since standing trees at the site had thin trunks, the project adopted anchors of the type to be embedded into the earth, and they were subjected to pull-out tests for endurance. Considering that the project was a remote-control unmanned construction at the top of the collapsed ground, the operator was able to have an enhanced visual monitoring capability by confirming the status of the site, watching the images from the movable cameras mounted on the high-altitude slope excavator and from the fixed cameras which are installed in the vicinity of the construction site (Photo 8).

3.7 Management Since surveys by human workers were not possible in the collapsed area, we used the machine guidance system mentioned above to confirm the completed shapes for earth-retaining embankment construction. We introduced a machine guidance system which is able to manage various data centrally in the control box, leading to improved operation for high-precision construction management.

126

S. Nomura et al.

Fig. 8 Topography of a collapsed land site in ortho images

In addition, UAV surveying was conducted in line with weather events such as heavy rainfall and the progress of countermeasure work. The status of sediment movement and gully erosion in the collapsed area was determined by visual inspection using image and differential analysis before and after each event. Figure 8 compares the three ortho images: one taken immediately after the earthquake but before the countermeasures; one after the rainfall in June, two months after the earthquake; one at the time when the main countermeasures were completed. From these images, newly occurred collapses due to rainfall and topographic alteration were confirmed, the data of which were reflected in countermeasures. Figure 9 shows the differential analysis by UAV surveying after rounding, and it is possible to confirm the condition of earth removal by rounding and the condition of accumulated sediments in the collapsed area.

4 Integral Process of Survey, Design, and Construction A project management system for emergency countermeasure work is shown in Fig. 10. Faced with the necessity for rapid design and construction at the time of an unprecedented slope collapse, it was determined that we needed to assemble the knowledge and expertise relating to erosion control technology from a wide range of

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

127

Fig. 9 Deferential analysis between before and after the rounding

locations in Japan. To cope with this issue, the Kyushu Regional Development Bureau promptly set up a technical study group of experts, with Kumagai-Gumi Co., Ltd. in charge of design and construction to carry out comprehensive site management for the project, collaborating with a consulting firm that is familiar with aerial surveying, erosion control, and other technologies. In addition, considering that the scale of the disaster was extremely large, and massive unmanned construction was necessary to prevent secondary disasters, and it was decided that the project required a high level of knowledge about construction technologies. Therefore, the contractor party also established a committee of experts. In parallel with design and construction, the Kyushu Regional Development Bureau also conducted a detailed geological survey, while a specialist firm of geology and survey separately proceeded with the installation of devices for dynamic movement observation at the site to monitor the ground. In order to coordinate the activities by various parties, the three-way meeting was held at the Kyushu Regional Development Bureau on a weekly basis, where the client and contractor worked together to share information in real time, enabling highly flexible and rapid decision making in project management.

128

S. Nomura et al.

Fig. 10 Management system for the emergency countermeasure project

5 Conclusion For the construction project of emergency countermeasures on the site affected by an unprecedented disaster, we successfully completed the slope countermeasure work in August 2020, despite aftershocks and torrential rains, seeking the best way through repetition of investigation, design, and construction (Photo 9). On October 3 of the same year, the “Slope Protection Completion and National Route 57 Opening Ceremony” was held, and a monument was erected to commemorate this massive slope failure, naming it “Sugaru Collapse” as a reminder of the earthquake that is part of the history of the man’s battle against severe nature (Photo 10). This project introduces 3D models and ICT technology and concurrent engineering, which consists in integrating the processes from survey and measurement to design, construction, and management, thereby speeding up decision making and shortening the recovery period. We are now in a world facing an increasing number of natural disasters due to severe climate changes. However, the recovery method described above is expected to be used widely as an effective disaster restoration system. We are deeply grateful for the warm support, cooperation, and words of encouragement from local residents who were eagerly awaiting the reconstruction of the Aso region as soon as possible, as well as for the chance to take part in the path to the development of the Aso region, overcoming many a difficulty and challenge. We wish for the recovery and further development of the Aso region.

Aso Ohashi Bridge Area Slope Disaster Reconstruction Project …

129

Photo 9 Panoramic view of the area surrounding the slope collapse reconstruction project in the Aso Ohashi area

Photo 10 Monument

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material K. Rangaswamy

and Regi P. Mohan

Abstract Soft soils are fragile in supporting the paved infrastructures, causing severe differential settlements and lateral displacements. Due to higher water contents, the soft subgrade soil becomes more compressible with overlying road pavement layers. Therefore, it is necessary to stabilize the soft subgrade soils to improve the geotechnical properties before constructing the road infrastructure. Nowadays, the ground profiles stabilize after using various commercially available nano-chemicals. Compared with traditional stabilizing methods, non-traditional nano-chemical treatment has the advantages of lower cost and better environmental protection. The present study initially characterized the clay soil for its basic geotechnical properties. Further, a series of experiments were conducted to study the effect of organosilane nano-chemicals on consistency limits and CBR strength with different curing periods. Herein, nanotechnology-based organosilane compound (Terrasil) is introduced into the cemented clay soil at 1% cement binder in different amounts of 0.03–0.05% for stabilization. XRD and SEM tests were conducted to discern the latent mechanisms and study the soil matrix’s physicochemical changes. The focus of this study was to determine the optimal dosage of nano-chemical corresponding to maximum CBR strength at different curing periods. The results indicated that the strength of stabilized clay soil increased with the increase of nano-chemicals up to a specific dosage and further decreased. Optimal results were obtained for soil treated with 0.045% nano-chemical and 1% cement. XRD and SEM results postulated the evidence of the furtherance in cementitious activity within the soil matrix. Keywords Soft clay · Organosilane nano-chemical · CBR

K. Rangaswamy (B) · R. P. Mohan Department of Civil Engineering, National Institute of Technology Calicut, Kozhikode, Kerala 673601, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_12

131

132

K. Rangaswamy and R. P. Mohan

1 Introduction 1.1 Background Low-strength, highly compressible soils with clay or silt on-site make the soil unsuitable for construction activities and pose a problem to civil engineers. It is uneconomical to get the entire soil deposits replaced. The engineering properties of these problematic soils need to be altered to achieve desired values. Ground improvement techniques, including compaction, dewatering, treatment with admixtures, reinforcement techniques, etc., are available to treat the soft clay grounds. An economical solution to stabilize soils with low strength and high deformation is chemical soil stabilization, in which admixtures are used to improve the properties of soils. Most traditional additives, like lime, fly ash, rice husk ash, etc., require high energy demand for their production, which contributes to global warming [1, 2]. Besides, these have a high cost of excessive maintenance, poor soil structure, leaching away during Indian monsoons, secondary chemical pollution, and detrimental environmental impacts associated with construction [3]. All the above factors have tended to reduce the application of stabilization with traditional chemicals over the past decade. At present days, science has advanced far better from the olden days. From an economic point of view, many alternative admixtures are being used nowadays to quicken the process of soil improvement and to avoid delays in construction activities. Thus, many more non-traditional additives are developed from the advanced search for new materials. Research is being carried out in this area using alternative stabilizing agents such as bio-enzymes, nano-chemicals, nano-materials, ionic polymers, etc., to understand their behavior in soil [4–6].

1.2 Stabilization Using Nano-Materials Many studies are being conducted on different nano-materials, nano-chemical compounds, etc., to find their effectiveness as soil stabilizers. Unlike traditional materials such as lime and cement, the nano-materials are permanent, non-toxic, biologically and chemically inert, and have excellent durability. The studies conducted by [7, 8] indicated that adding nano-materials to soil leads to an improvement in the strength, Atterberg’s limits and a decrease in the permeability properties of soil. Due to its high specific surface area, even a tiny amount of nano-material addition could cause a significant change in the physical and chemical properties of soil [5]. A study on the effect of different nano-materials (nano-clay, nano MgO, and nano alumina) on engineering characteristics showed that an increase in nano-clay content increases Atterberg’s limits due to the higher specific surface area of nanoparticles. However, the increase in nano MgO and nano alumina decreased Atterberg’s limits [9]. California Bearing Ratio (CBR) strength values increased considerably with

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material

133

nano-chemical addition, and the design of pavement resulted in a reduced thickness of about 25% lower than that with untreated soil due to reactions of clay soil with nano-chemical [10–13]. Thus, nano-material stabilization will be of great significance in sustainable ground improvement. From all these studies, it is observed that nanotechnology-based stabilization makes the soil slightly less plastic and improves the strength characteristics of the clay soil to a higher degree. Thus, the main focus of this study is to minimize the use of environmentally detrimental traditional stabilizers like cement, lime, etc.

2 The Objective of the Study This paper discusses the influence of an organosilane-based nano-chemical addition on the geotechnical engineering properties of soft clay soil for different curing periods and varying dosages of nano-chemicals. The scope of this paper is limited to studying the variation in the consistency limits and CBR strength of a soft soil cemented with 1% cement binder stabilized with nano-chemical after 7 to 28 days of curing period.

3 Materials and Methods 3.1 Materials The natural untreated clay soil used in the study has been procured locally from Pantheerakavu, Kozhikode, Kerala, India, at a depth of 1.5–2 m from the ground surface. The native soil used in the present study is termed virgin soil (VS). Different experimental tests were conducted in the laboratory according to Indian Standard specifications. The soil contains 60% silt and 22% of clay fractions. Liquid and plastic limits are 79 and 44%, respectively. It starts to shrink with the addition of water content above 27%. Soil is classified as highly compressible silt. The maximum proctor compaction dry density and OMC were found to be 12.6 kN/m3 and 39%, respectively. The soil exhibits an unconfined compressive strength of 50 kPa. The nanotechnology-based organosilane nano-chemical used in the present study is Terrasil, manufactured by Zydex industries in Vadodara, Gujarat, India. It is a concentrated liquid in golden brown color and is water-soluble and designated as a nano-chemical (NC) in the present study. It has chemical compounds of hydroxyalkyl–alkoxy–alkylsilane (65–70%), benzyl alcohol (25–27%), and ethylene glycol (3–5%). Based on economic considerations and to minimize environmental pollution, cement usage in the present study is restricted to 1% by weight of soil to act as a binder. Several published papers proved that the strength of nano-chemicalstabilized clay soils is significantly more enhanced in the presence of even a tiny

134

K. Rangaswamy and R. P. Mohan

amount of cement content instead the shear strength of clay soils stabilized by nanochemicals/materials alone. Nano-chemical can act as an accelerator in the hydration of cement, which imparts strength development due to accelerated pozzolanic reactions in nano-chemical treated soil in the presence of cement binder. The strength of clay soils stabilized by nano-chemicals alone is significantly lesser than the strength improved by the combination of cement and nano-chemical additives.

3.2 Sample Preparation and Testing The treated soil samples were prepared as follows: Initially, the binder cement at 1% by weight of soil is introduced into the air-dried pure clay and mixed in a dry state to distribute uniformly. Later the specific dosage of nano-chemical is injected into the OMC of the soil to obtain the nano-chemical solution. Nano-chemical solution has sprayed directly on the dry soil–cement mixture and mixed thoroughly to spread uniformly. A similar procedure is adopted in the field application of road pavement stabilization by the Zydex industries [14]. The present study chooses the trial dosages of nano-chemicals in the range of 0.03–0.05% by the weight of clay soil. The designations used for the composition of different soil mixes are VS: Virgin soil; VS + C1: Virgin soil + 1% binder; VS + C1 + NC1: Virgin soil + 1% Binder + 0.03% NC; VS + C1 + NC2: Virgin soil + 1% Binder + 0.04% NC; VS + C1 + NC3: Virgin soil + 1% Binder + 0.045% NC; VS + C1 + NC4: Virgin soil + 1% Binder + 0.05% NC. Consistency Limit Tests: The prepared soil mixtures were cured for up to 28 days after keeping them in the sealed polythene cover and desiccator. The cured soil samples were dried, and the consistency limit tests were conducted to examine the effect of nano-chemical on the liquid, plastic, and plasticity of clay soils with 1% cement binder. The tests were conducted as per the respective Indian Standard code procedures. CBR Tests: The present study has conducted the modified proctor compaction (MPCT) tests to examine the influence of nano-chemicals on the variation of compacting density and optimum moisture content of clay soil. The samples have cured for three days before conducting the MPCT tests [15], and it has found that there were slight changes in the maximum dry density (MDD) and OMC since the replacement of soil is very minimal with 1% cement and up to 0.05% nano-chemical by the weight of soil. Hence, the treated soil specimens for the CBR testing have been prepared at a maximum dry density of virgin soil for comparison with the parent clay soil and to maintain uniformity. The unsoaked CBR tests have carried out on the nano-chemical treated soil specimens with 1% cement binder after being cured at 7, 14, 21, and 28 days to examine the influence of nano-chemical on the CBR strength variation of clay soil. The curing period allows the chemical reactions to take place and examine the effect of curing on strength improvement. CBR test specimens were stored under humid conditions with a covered curing technique to prevent moisture

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material

135

loss and cure the samples. The test soil specimens at particular nano-chemical dosage have been cast in CBR molds of constant volume (150 mm diameter and 175 mm height) at compaction density of virgin soil and cured for the specified curing period. CBR testing has been conducted at a strain rate of 1.25 mm/min according to Indian Standards [16]. The soaked CBR tests have also been performed with 4 days of soaking after curing the soil specimens.

4 Results and Discussion A series of Atterberg’s limit and CBR tests had performed on various nano-chemical soil mixtures to determine the effect of nano-chemical on the index and CBR strength of clay soil mixtures cured from 7 to 28 days.

4.1 Consistency Limits The effect of curing time on a variation of consistency limits of nano-chemical treated soil mixtures is shown in Figs. 1, 2 and 3. Figure 1 predicts the variation of liquid limits, demonstrating that the liquid limit of nano-chemical treated soil mixtures decreases with curing times till 28 days; however, the rate of decrease is almost negligible after 21 days. Figure 2 indicates that the plastic limit of Terrasil treated soil mixtures increases with the curing times, making the material more deformable, similar to lime-stabilized soil. However, the plasticity index of nano-chemical treated soil mixtures is significantly reduced with curing times, as shown in Fig. 3. Hence, it is realized that the chemical interactions would take up to 21–28 days of curing, and the nano-chemical soil mixtures became less plastic enough and sustained further. The present study also found the shrinkage limits of nano-chemical treated soil mixtures decreased (24–18%) with curing times, indicating the stabilized soil becomes more deformable up to the plastic state. Figures 1, 2 and 3 also explore the effect of nanochemical dosage on the consistency limits of clay soil used in the present study. It shows that liquid limit and plasticity indices decrease with the addition of 0.045% nano-chemical and further increase irrespective of curing times. Plastic limits are raised to 0.045% of nano-chemical added into the soil and decrease further. Table 1 presents the percentage of variation in consistency limits of nano-chemical treated soil mixtures cured at 28 days. It shows the soil mixture contains 0.045% nano-chemical and 1% cement binder, providing maximum reduction of liquid limit and plasticity indices about 31.2 and 98.5% decrease over the liquid limit and plasticity index of virgin clay soil, respectively. The maximum increase of the plastic limit of nano-chemical treated soil is about 22.2% at 0.045% nano-chemical with 1% cement binder.

136

K. Rangaswamy and R. P. Mohan

Fig. 1 Variation of liquid limit of nano-chemical soil mixtures with curing times

90

Liquid limit, %

85 80

VS VS+C1+NC1 VS+C1+NC2 VS+C1+NC3 VS+C1+NC4

75 70 65 60 55 0

5

10

15

20

25

30

Curing time, days

Fig. 2 Variation of plastic limit of nano-chemical soil mixtures with curing times

54

Plastic limit, %

52 50 48

VS VS+C1+NC1 VS+C1+NC2 VS+C1+NC3 VS+C1+NC4

46 44 42 40 38 0

5

10

15

20

25

30

Curing time, days

40 35

Plasticity index, %

Fig. 3 Variation of plastic index of nano-chemical soil mixtures with curing times

VS VS+C1+NC1 VS+C1+NC2 VS+C1+NC3 VS+C1+NC4

30 25 20 15 10 5 0 0

5

10

15

20

25

30

Curing time, days

4.2 CBR Strength The unsoaked and soaked CBR strength variations of nano-chemical soil mixtures with curing times are presented in Figs. 4 and 5, respectively. The figures illustrate that

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material

137

Table 1 Percentage variation in consistency limits of nano-chemical treated soil mixtures Soil mixtures

Consistency limits, % (28 days of curing) LL

% Dec

PL

% Inc

PI

% Dec

Virgin clay

79

VS + C1 + NC1

57.4

27.3

52.7

19.7

4.68

86.6

VS + C1 + NC2

55.4

29.8

53.4

21.3

2

94.3

VS + C1 + NC3

54.3

31.2

53.8

22.2

0.5

98.5

VS + C1 + NC4

55.1

30.2

53

20.4

2.1

94

–

44

–

35

–

the CBR strength is increased with the increase of curing up to 28 days irrespective of nano-chemical treated soil mixtures. The rise in CBR is due to the liquid limits and plasticity index of nano-chemical treated soils decreasing with curing. Figures 4 and 5 also explore the effect of nano-chemical dosage on the CBR strength of clay soil used in the present study. It shows that CBR strength increases with the addition of 0.045% nano-chemical and further decreases irrespective of curing times. Hence, the nano-chemical with 0.045% is the optimum dosage that gives the maximum CBR strength of clay soil. The high CBR strength of 0.045% nano-chemical treated soil mixture is due to the soil becoming as low plastic with less PI than the other treated soil combinations. Table 2 presents the percentage of variation in CBR strength of nano-chemical treated soil mixtures cured at 28 days. The unsoaked and soaked CBR strength of treated soil mixtures has been found to increase by about 13–22 times the unsoaked and soaked CBR strength of clay, respectively, with the addition of an optimum dosage of 0.045% nano-chemical into the clay soil. The variation in reduction of CBR strength of unsoaked and soaked nano-chemical-stabilized soil mixtures is about 5–7% at 28 days of curing. The increase rate in the CBR of clay soil with the addition of 0.03% NC is very rapid, and further, the increase is gradual till the optimum dosage of 0.045% NC. The CBR strength improvement of nanochemical-treated clay in the presence of 1% cement binder is more significant than the strength of clay soil treated with nano-chemicals alone. The results comply with published papers that investigated the effect of nano-chemicals on the strength of clay soils [12, 17].

4.3 Microstructural Characterization Scanning Electron Microscopy: Fig. 6 shows the FESEM images of virgin clay soil, and Fig. 7 shows the FESEM image of the nano-chemical treated soil mixture (VS + C1 + NC3) containing the optimum dosage of nano-chemical (0.045%) after seven days of curing.

138

K. Rangaswamy and R. P. Mohan

Fig. 4 Variation of unsoaked CBR strength of nano-chemical soil mixtures with curing times

50

Unsoaked CBR, %

45 40 35 30

VS VS+C1+NC1 VS+C1+NC2 VS+C1+NC3 VS+C1+NC4

25 20 15 10 5 0 0

5

10

15

20

25

30

Curing time, days

Fig. 5 Variation of soaked CBR strength of nano-chemical soil mixtures with curing times

50 45

Soaked CBR, %

40 35 30

VS VS+C1+NC1 VS+C1+NC2 VS+C1+NC3 VS+C1+NC4

25 20 15 10 5 0 0

5

10

15

20

25

30

Curing time, days

Table 2 Percentage variation in CBR strength of nano-chemical treated soil mixtures Soil mixtures

CBR, % (28 days of curing) Unsoaked

Virgin clay

3.6

% Inc

Soaked

% Inc

–

2.1

–

VS + C1 + NC1

42

1066

38.8

1747

VS + C1 + NC2

48

1233

45

2042

VS + C1 + NC3

49.9

1286

47.2

2148

VS + C1 + NC4

45.5

1164

42.8

1938

It can be seen from the figures that there is a significant change in the soil fabric of the treated soil mixture prepared at the optimum dosage of nano-chemical. In nanochemical treated soil, the particles are bound together and are more closely packed or agglomerated, with a noticeable reduction in void spaces. It occurs because van der Waals forces develop strongly between the nanoparticles. Thus, adding nanochemical considerably reduced the porosity as it fills the micro- and nano-pores between soil particles.

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material

139

Fig. 6 FESEM image of untreated soil after 7 days of curing

Fig. 7 FESEM image of the treated soil mixture (VS + C1 + NC3) after 7 days of curing

5 Mechanism of Stabilization The main ingredient of nano-chemical used in the present study is hydroxyalkyl– alkoxy–alkylsilanes. The considerable improvement in unconfined compression strength is due to the strong reaction of nano-chemical with silanol groups of soil and the resulting siloxane bondage [18], which forms a breathable in-situ membrane. The nano-chemical reacts with the soil particles and makes the surfaces waterproof, stiffening the soil and thus increasing the bond between soil particles [14]. Beyond optimum dosage, the strength decreases due to the loosening of the interaction between soil and nano-chemical in the presence of excess non-active nano-chemical.

140

K. Rangaswamy and R. P. Mohan

6 Conclusions A comprehensive experimental study was conducted on the soil before and after stabilizing with 1% cement and a nanotechnology-based organosilane compound to determine the effect of nano-chemical addition on the CBR strength of the clay soil, including consistency limits. The salient conclusions drawn from the present study are as follows: • Adding nano-chemical up to 0.045% into the clay soil with 1% cement binder shows the minor liquid limit and plasticity index and increases at higher dosages of nano-chemical. Nano-chemical treated soils are becoming less plastic with curing times, stabilizing after 21–28 days. • The CBR strength of clay soil, along with 1% cement as a binder, increased with an increase in the nano-chemical dosage up to 0.045% of the weight of clay soil and is about 22 times higher than the soaked CBR strength of untreated virgin soil at a 28-day curing period. • The CBR strength of nano-chemical-stabilized soil after the 21–28-day curing period reveals a higher strength improvement than at less curing periods. It may be due to improving the chemical interactions of nano-chemical to flocculate and agglomerate the soil particles, thereby stiffening the soil with curing times. • The FESEM images of nano-chemical treated soil showed that interparticle voids had reduced considerably, and the structure had changed to flocculated with the addition of nano-chemical. In summary, treating weak paved subgrade soil with an organosilane-based nano-chemical at a dosage of 0.045% of the weight of clay soil along with 1% cement content is the best combination to achieve higher CBR strength. The shear strength achieved by this stabilization method is sufficient to improve so that locally available field soil deposits could enhance the shear strength and can be used in pavement construction works. Thus, typical weak soils unsuitable for road pavement construction could stabilize by adding organosilane-based nano-chemical compounds.

References 1. Ewa DE, Egbe EA, Akeke GA (2016) Effects of nano-chemical on geotechnical properties of ogoja subgrade. J Res Inform Civil Eng 13(1):820–831 2. Taha MR, Khan TA, Jawad IT, Firoozi AA (2013) Recent experimental studies in soil stabilization with bio enzymes-a review. Electr J Geotech Eng 18:3881–3894 3. Buazar F (2019) Impact of biocompatible nano silica on green stabilization of subgrade soil. Sci Rep 9(1):1–9. https://doi.org/10.1038/s41598-019-51663-2 4. Carroll D, Starkey HC (1971) Reactivity of clay minerals with acids and alkalies. Clays Clay Mineral 19(5):321–333. https://doi.org/10.1346/CCMN.1971.0190508 5. Zhang G (2007) Soil nanoparticles, and their influence on engineering properties of soils. Geotech Spec Publ 26:1–13. https://doi.org/10.1061/40917(236)37

Nano-Chemical Stabilization of Soft Soil as a Paved Subgrade Material

141

6. Sol-Sanchez M, Castro J, Urena CG, Azanon JM (2016) Stabilization of clayey and marly soils using industrial wastes: ph and laser granulometry indicators. Eng Geol 200:10–17. https://doi. org/10.1016/j.enggeo.2015.11.008 7. Noll MR, Bartlett C, Dochat TM (1992) In situ permeability reduction and chemical fixation using colloidal silica, NATIONAL OUTDOOR ACTION CONFERENCE 1992. Las Vegas, pp 443–457 8. Yonekura R, Miwa M (1993) Fundamental properties of sodium silicate-based grout. In: Geotechnical conference 1993. Singapore, pp 439–444 9. Priyadharshini R, Arumairaj PD (2015) Improvement of bearing capacity of soft clay using nanomaterials. Int J Sci Res 4(6):218–221 10. Patel AN, Mishra CB, Pancholi VV (2015) Scientifically surveying the usage of terrasil chemical for soil stabilization. Int J Res Adv Technol 3(6):77–84 11. Prakash MM, Shravan K (2015) Effect of nanochemicals on properties of the black cotton soil. In: National conference on sustainable infrastructure development 2015, Chandigarh 12. Mrudul UV, Damodariya SM (2016) Laboratory investigation of soil stabilized using terrasil. Int J Sci Res Develop 3(12):537–539 13. Roshni S, Jeyapriya SP (2017) Experimental study on the use of nanochemical and cement in the modification of subgrade. Int J Sci Eng Res 8(3):1868–1872 14. ZYDEX industries Home page. https://zydexindustries.com/soil-stabilization/. Accessed 10 July 2022 15. IS 2720-Part 7 (1980) Methods of test for soils-laboratory determination of water content-dry density relation using heavy compaction. Bureau of Indian Standards, New Delhi 16. IS 2720-Part 16 (1987) Methods of test for soils-laboratory determination of CBR strength. Bureau of Indian Standards, New Delhi 17. Lekha BM, Goutham S, Ravi Shankar AU (2013) Laboratory investigation of soil stabilized with nano-chemical. In: Proceedings of Indian geotechnical conference. Roorkee, pp 1–7 18. Arkles B, Steinmetz JR, Zazyczny J, Mehta P (1992) Factors contributing to the stability of alkoxysilanes in an aqueous solution. J Adhesion Sci Technol 6(1):91–104. https://doi.org/10. 1163/156856192X00133

Physical and Numerical Modelling of Disasters and Disaster Mitigation Techniques

A Plasticity Model of Binary Mixtures for Liquefaction Simulation Considering the Equivalent Granular Void Ratio Fu-Hsuan Yeh , Yi-Qian Lu, and Louis Ge

Abstract Most natural soil deposits are non-homogeneous granular materials consisting of coarse-grained particles with finer non-plastic particles filling the voids. Experimental results on binary mixtures are needed to assess the performance of constitutive models considering non-homogeneous granular materials. This study aims to calibrate and inspect an advanced constitutive model based on the generalized plasticity framework from monotonic and cyclic triaxial tests on binary mixtures. A state parameter-based model uses the global void ratio (e) and state parameter (ψ) as a unifying framework. The parameters e and ψ were substituted by e* and ψ * to check the performance of the framework in constitutive modeling, and both simulations were compared. Samples of binary mixtures (the coarse-grained particles: Ottawa sand; the fine-grained particles: Vietnam silica fine sand) were used to simplify the complex non-homogeneous granular materials where their particle shapes and size ratio have been considered to check the applicability of the proposed constitutive model. The triaxial monotonic tests were first used to check the model and calibrate the model parameters. The calibrated parameters from monotonic tests were applied to predict the amplitude of shear strain and the number of cycles until liquefaction in the undrained cyclic triaxial tests. This model cannot reproduce very well in the simulations of undrained cyclic triaxial test results, but the results show that the inclusion of e* and ψ * could be considered in the simulations under monotonic loading when accounting for the effects of fines content. Therefore, to accurately predict liquefaction behaviors, more laboratory data must be used to improve this unified framework. Keywords Binary mixtures · Generalized plasticity · Equivalent granular void ratio F.-H. Yeh · Y.-Q. Lu · L. Ge Department of Civil Engineering, National Taiwan University, 10617 Taipei, Taiwan Present Address: F.-H. Yeh (B) Department of Civil and Construction Engineering, National Taiwan University of Science and Technology, 106335 Taipei, Taiwan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_13

145

146

F.-H. Yeh et al.

1 Introduction Recent historical cases indicated that sandy soil containing a high amount of finegrained materials (or fines content) may also liquefy (e.g., [1]). Their strength and volume change behavior relative to their fines content has been a research topic in recent years. The effect of fines content in granular soils on the liquefaction potential and its static strength behavior has been studied. Lade and Yamamuro [2] indicated that sand’s fines content affected its liquefaction potential. They found that complete static liquefaction may occur in sand with high relative density and high fines content. One factor that influences soil behavior may be void space. The physical and mechanical properties of granular mixtures may be affected by fines in voids interacting with the host coarse sand or vice versa (e.g., [3]). Since mixtures containing fines have varied mechanical properties, Thevanayagam [4] presented a preliminary study on the shear strengths of sandy soil containing fines and proposed an explanation of the variation between sand–silt mixtures by postulating the intergranular void ratio as its source. The hypothesis shows that coarse grains dominate the behavior of mixtures with small fines content [5, 6]. A binary mixture, defined as a mixture with two different grain sizes, can simplify the complex composition of a granular mixture. Our major concern is finding an advanced constitutive model to predict the stress– strain–volume change behaviors of soil. Finding a proper stress–dilatancy relationship is essential to help interpret the dilatancy behavior of mixtures or granular materials. Manzanal et al. [7] proposed a model based on generalized plasticity that considers the state parameter proposed by Been and Jefferies [8] and the dilatancy behavior proposed by Li and Dafalias [9]. This model can reproduce the soil behavior under static and cyclic conditions [10]. The dilatancy relationship and the state parameter can characterize loose or dense states of soil. Been and Jefferies [8] first introduced the concept of the state parameter to describe soil behavior; the parameter reflects the influence of void ratio and stress level on a critical state line through a series of triaxial tests. The state parameter depends on the current state void ratio and critical state void ratio, so determining the critical state void ratio is necessary. Therefore, a critical void ratio line in e − p space accounting for the effect of fines that incorporates the concept of the equivalent granular void ratio proposed by Thevanayagam et al. [5] is adopted. This study investigates the state parameter-based generalized plasticity model incorporating the equivalent granular void ratio in predicting laboratory data results with 0 and 15% fines contents of binary mixtures. First, the model parameters were calibrated by fitting the experimental data results under monotonic loading, including shear stress versus axial strain, shear stress versus mean effective stress, and excess pore pressure versus axial strain. Then, to check whether one set of parameters is sufficient for predicting the behaviors of the material, the calibrated parameters were used to predict the cyclic undrained triaxial test results. Finally, check whether the concepts of e* and ψ * could be used when calculating liquefaction onset and accounting for the effects of fines content.

A Plasticity Model of Binary Mixtures for Liquefaction Simulation …

147

2 Equivalent Granular Void Ratio and State Parameter The equivalent granular void ratio e* proposed by Thevanayagam et al. [5] accounts for the effect of fines content as follows: e∗ =

e + (1 − b) f c 1 − (1 − b) f c

(1)

Here, e is the global void ratio, f c is the fines content of mixtures, and the parameter b is the fraction of fines, ranging from 0 to 1. In the state parameter-based generalized plasticity model, the current state void ratio e should substitute by the equivalent granular void ratio e* . If f c equals zero, the equivalent granular void ratio reduces to the global void ratio as coarse sand. The global void ratio represents the current state that should consider the fines content effect, if any. The original state parameter (ψ = e − ec ) is proposed by Been and Jefferies [8]. The variables e, ec , and ψ were replaced by e* , ec∗ , and ψ * , where ψ * is the equivalent state parameter generated by the equivalent granular critical state void ratio ec∗ and equivalent granular void ratio e* at the current state.

3 State Parameter-Based Generalized Plasticity Model After Manzanal et al. [7] The framework of the state parameter-based generalized plasticity model after Manzanal et al. [7], unifying the inclusion of the state parameter and the definition of the hardening modulus concepts, is introduced. The incremental stress is related to the incremental strain through the tangential stiffness tensor following the general relation: dσ = Dep : dε De : ngL/U ⊗ n f : De Dep = De − HL/U + n : De : ngL/U

(2)

Here, Dep and De are the elastoplastic and elastic stiffness tensors, respectively. The plastic modulus H L/U is the scalar function for either loading or unloading, where subscripts L and U represent loading and unloading, respectively. The loading direction tensor nf and the direction of plastic flow ngL/U are required. The loading direction vector nf is directly defined in the model as follows: ⎡

⎤T df

, nf = ⎣ 1 + d 2f



1 1+

⎦ d 2f

(3)

148

F.-H. Yeh et al.

The plastic flow for loading and unloading are defined as follows: ⎡ ngL = ⎣ 

⎤T dg 1 + dg2

1

⎦  1 + dg2

,

and ngU

 ⎡     dg   , ⎣ = −    1 + dg2 

⎤T 1

⎦  1 + dg2 (4)

Here, the value of d g is related to dilatancy behavior, and the definition of d f is similar to d g . The study follows the model [7] on fully saturated states.

3.1 Elastic Behavior The elastic moduli’s empirical equations to consider the void ratio and confining pressure formed by two elastic constants are defined as follows: G es = G es0 ·

(2.97 − e∗ )2   · p · patm (1 + e∗ )

(5)

K ev = K ev0 ·

(2.97 − e∗ )2   · p · patm (1 + e∗ )

(6)

Here, Ges0 and K ev0 are the initial material constants, p is the mean effective stress, patm is the atmospheric pressure, and the void ratio e* changes with the varying loading–unloading stage.

3.2 Dilatancy and Plastic Flow The stress–dilatancy equation relating to the state parameter considering the phase transformation line proposed by Li and Dafalias [9] is defined as follows: dg =

d0 (ηPTS − η), ηPTS = Mg · exp m · ψ ∗ Mg

(7)

Here, M g is the slope of the critical state line in the p − q space, ηPTS is the stress ratio at the point of phase transformation, and d 0 and m are the model parameters. The loading–unloading direction tensor nf is defined in a similar form as ng . The definition of d f and d g is also formed alike, but d f uses M f instead of M g : df =

d0  · M f · exp m · ψ ∗ − η Mf

(8)

A Plasticity Model of Binary Mixtures for Liquefaction Simulation …

149

All critical state data, including coarse sand and binary mixture, can be depicted by a unique relationship referred to as the critical state line proposed by Li and Wang [11]. This equation can be used as the equivalent granular void ratio to determine the critical state void ratio in e-p space: ec∗ = e − λ

p patm

ζc (9)

Here, e is the initial critical state void ratio at low confining pressure, λ and ζ c are model constants. A non-associated flow rule is followed to present the behavior of cohesionless soils, which means M g = M f . Manzanal et al. [7] proposed a relationship between M f and M g : Mf = h1 − h2 · Mg

e∗ ec∗

β (10)

Here, β, h1, and h2 are model parameters. M f /M g cannot exceed 1.

3.3 Plastic Modulus for Loading and Unloading Under loading conditions, the plastic modulus H L denotes the portion of shear and volumetric strain as follows: HL = H0 ·



p  · patm · H f · (Hv + Hs ) · HDM

(11)

H 0 incorporates the equivalent granular void ratio at the current state and the equivalent critical state void ratio as follows:  H0 = H0 · exp −β0 ·

e∗ ec∗

β  (12)

Here, H0 , β0 , and β are model parameters. H f is one of the plastic coefficients describing the relation between admissible internal states η and inadmissible external states ηf as follows:  

1 η 4 · Mf , ηf = 1 + Hf = 1 − ηf α

(13)

Here, α is a model parameter. H v depends on the equivalent state parameter incorporating the peak stress ratio ηp as follows:

150

F.-H. Yeh et al.

 Hv = Hv0 · η p − η

(14)



η p = Mg · exp −βv · ψ ∗

(15)

Here, H v0 and β v are model parameters. H s reflects the accumulation of plastic shear strains and is defined as follows: Hs = β1 · exp(−β0 · ξs )

(16)

Here, β 0 and β 1 are model constants. H s explains the reproduction of softening behavior without causing immediate failure and the residual state at the critical state line. In Manzanal’s study, the definition of H DM and ζ is the same as Pastor’s model [12]. H DM reflects the material memory during reloading as follows:

HDM =

ζmax ζ

γ (17)

Here, γ is a model parameter, and the past stress history updates ζ max to keep the maximum of ζ. The plastic modulus H U under unloading can be defined as follows:

HU = HU0

ηU Mg

−γU

     ηU   ηU   ≥1   < 1; HU = HU0 for  for  Mg  Mg 

(18)

Here, H U0 and γ U are material parameters.

4 Inspection Based on Laboratory Data 4.1 Experimental Setup: Specimen Preparation Ottawa sand was selected as coarse grain for the specimens without considering fines content. For the coarse-grained dominated binary mixture at a fines content of 15%, the specimens were prepared using 15% Vietnam silica fine sand mixed with Ottawa sand. The mean grain size ratio between Ottawa sand and Vietnam fine silica sand was about 7.5 (D50 /d 50 = 0.7197/0.0965). The physical properties of Ottawa sand are as follows: its specific gravity Gs is 2.66, maximum void ratio emax is 0.751, minimum void ratio emin is 0.557, uniformity coefficient C u is 1.20, and coefficient of curvature C c is 0.97. The physical properties of Vietnam silica fine sand are as follows: Gs is 2.66, C u is 1.09, and C c is 0.98. Both selected grains of sand is zero plasticity index PI. Table 1 shows the experiments used for predictions, including six

A Plasticity Model of Binary Mixtures for Liquefaction Simulation …

151

Table 1 Program of laboratory tests used for predictions Test

e0,ini (−)

Dr0,ini (%)

e0 (−)

Dr0 (%)

p0 (kPa)

CD_FC0_150

0.665

45

0.652

51

150

e0∗ (−)

FC (%) 0

–

CD_FC0_300

0.634

60

0.617

69

300

0

–

CD_FC0_450

0.659

48

0.637

59

450

0

–

CU_FC0_150

0.661

47

0.648

53

150

0

–

UCT_FC0_150

0.644

55

0.622

67

150

0

CD_FC15_150

0.428

51

0.412

58

150

15

0.683

–

CD_FC15_300

0.434

49

0.434

49

300

15

0.681

CD_FC15_450

0.434

49

0.401

63

450

15

0.665

CU_FC15_150

0.431

50

0.427

52

150

15

0.668

UCT_FC15_150

0.474

32

0.466

36

150

15

0.714

drained monotonic triaxial tests (CD), two undrained monotonic triaxial tests (CU), and two undrained cyclic triaxial tests (UCT).

4.2 Parameter Calibration The parameter set of the adopted model is presented in Table 2. This model requires 18 parameters for monotonic loading, one for the determination of equivalent void ratio (b), and three for cyclic loading. The calibration of parameters can be divided into five groups, consisting of elastic stiffness constants (K ev0 and Ges0 ), critical state (ϕ c , e , λc , and ζ c ), plastic flow (m, h1 , and h2 ), plastic modulus (H0 , β0 , β, H v0 , β v , β 0 , β 1 , and α), and unloading/reloading (γ DM , γ U , and H U0 ). The parameters defined by the equivalent granular critical void ratio line (Fig. 1) were calibrated by the triaxial test results under monotonic loading. Table 2 Calibrated material parameters of the constitutive model Elasticity

Critical state

K ev0 [–] Ges0 [–] b [–]

ϕ c [°]

292

135

0.035*

Plastic flow

e [–] λc [–]

28.4/30*

ζ >c [–] d 0 [–] m [–]

0.751 0.0242 0.7

1.0

250

β0 [–] 1.8

1.25

0.50

Unloading/reloading

Plastic modulus H0 [–]

h1 [–] h2 [–]

5.5/0.55*,**

β [–]

H v0 [–]

β v [–]

β 0 [–]

β 1 [–]

α [–]

γ DM [–]

γ U [–]

5.0

100/20*

2.3

0

0

0.45

4

4

Note β 0 = β 1 = 0 on fully saturated states; additional adjustment for different conditions: 15%;** for FC = 0% under cyclic loading

H U0 [–] 20,000 * for

FC =

152

F.-H. Yeh et al.

Fig. 1 Initial and critical states of all monotonic tests in e or e* –log(p ) space

4.3 Element Test Inspections Inspection Based on Drained and Undrained Triaxial Tests: Monotonic Loading. Figure 2a–c provides the simulations of three tests of Ottawa sand with similar initial relative densities at various confining pressures. With a single set of material parameters, this model reproduced material behavior for the Ottawa sand very well in the critical state and critical void ratio, shear strength, and dilatancy. Figure 2d– f presents the simulations for the binary mixture with 15% fines content with the approximately same set of parameters, except three parameters (ϕ c , m, and H v0 ) were adjusted. However, parts of the simulations in the deviatoric stress and the volumetric strain have some discrepancies. This model slightly underestimates the peak shear strength for higher initial effective pressures, overestimates the volumetric strain in lower confining pressure, and underestimates the volumetric strain in medium and higher confining pressure. Figure 3 shows the respective numerical simulations compared to the experimental data in the spaces of shear stress versus axial strain (q − ε1 ), excess pore water pressure versus axial strain (u − ε1 ), and effective stress space (p − q). It presents that the model can reasonably predict the experimental test results. The model reproduced all eight simulated tests under monotonic loading with different drainage conditions, initial void ratios, and initial mean pressures in general. The calibrated model parameter sets can be used to predict the triaxial tests under cyclic loading. Inspection Based on Undrained Triaxial Tests: Cyclic Loading. In the simulations, the accumulation of axial strain occurred only toward the triaxial extension side (Figs. 4 and 5). Figure 4 shows that after a certain number of cycles, the stress paths from the simulations repeatedly pass through a butterfly-shaped loop. When simulating cyclic triaxial tests for a sample with 15% fines content, this model cannot reproduce elastic and plastic strains, dilatancy behavior, or effective stress relaxation (Fig. 5). This is due to the fact that the equivalent granular void ratio for the binary mixture with 15% fines content is 0.714 (the maximum equivalent granular void ratio is 0.751), which means that the sample is very loose at the initial state. Hence, the

A Plasticity Model of Binary Mixtures for Liquefaction Simulation …

153

Fig. 2 Simulations of drained monotonic triaxial tests

Fig. 3 Simulations of undrained monotonic triaxial tests

simulation demonstrates that the sample liquefies quickly. Based on the parameter set calibrated by monotonic triaxial test results, this model cannot sufficiently reproduce elastic and plastic strains, dilatancy behavior, or effective stress relaxation in the simulations of undrained cyclic triaxial tests for the samples with zero and 15% fines contents.

154

F.-H. Yeh et al.

Fig. 4 Simulation of undrained cyclic triaxial test for the Ottawa sand with FC = 0% fine sand

Fig. 5 Simulation of undrained cyclic triaxial test for the sample with FC = 15% fines content

5 Conclusion Material parameters should be determined and calibrated before geotechnical analysis and evaluation in numerical modeling. This study adopted the unified state parameter-based generalized plasticity model, incorporating the equivalent granular void ratio and critical state. The model parameters were calibrated from all monotonic triaxial test results and were used to predict the amplitude of shear strain and the number of cycles until liquefaction in the undrained cyclic triaxial tests. Only one equivalent granular critical void ratio line was used to describe all test results for the binary mixtures with the fines contents at 0 and 15%. The results show that the model can reproduce the experimental results under monotonic loading by considering the inclusion of an equivalent granular void ratio and equivalent state parameter. However, the model cannot capture the liquefaction behaviors in the simulations of undrained cyclic triaxial tests, such as liquefaction onset, the amplitude of shear strain, and the number of cycles. Hence, more triaxial tests on binary mixtures with different loading and drainage conditions, portions of fines contents, initial void ratios, and initial mean pressures need to be carried out to prove that the scenario of the unified state parameter-based generalized plasticity model incorporating the equivalent void ratio concept is workable. Acknowledgements This research work was made possible by financial support from the National Science and Technology Council, Taiwan, R.O.C., by providing funding with respect to a

A Plasticity Model of Binary Mixtures for Liquefaction Simulation …

155

project entitled “Mitigation of underground life pipelines in liquefiable soil—incorporating new investigation techniques” under project 110-2221-E-002-046-MY3.

References 1. Cubrinovski M (2013) Liquefaction-induced damage in the 2010–2011 Christchurch (New Zealand) earthquakes. In: Proceedings of the 7th conference of the international conference on case histories in geotechnical engineering. Missouri University of S&T, Chicago, pp 1–11 2. Lade PV, Yamamuro JA (1997) Effects of non-plastic fines on static liquefaction of sands. Can Geotech J 34(6):918–928 3. Salgado R, Bandini P, Karim A (2000) Shear strength and stiffness of silty sand. J Geotech Geoenviron 126(5):451–462 4. Thevanayagam S (1998) Effect of fines and confining stress on undrained shear strength of silty sands. J Geotech Geoenviron 124(6):479–491 5. Thevanayagam S, Shenthan T, Mohan S, Liang J (2002) Undrained fragility of clean sands, silty sands, and sandy silts. J Geotech Geoenviron 128(10):849–859 6. Rahman MM, Lo SR, Baki MAL (2011) Equivalent granular state parameter and undrained behaviour of sand–fines mixtures. Acta Geotech 6(4):183–194 7. Manzanal D, Merodo JAF, Pastor M (2011) Generalized plasticity state parameter-based model for saturated and unsaturated soils. Part 1: saturated state. Int J Numer Anal Methods Geomech 35(12):1347–1362 8. Been K, Jefferies MG (1985) A state parameter for sands. Geotechnique 35(2):99–112 9. Li XS, Dafalias YF (2000) Dilatancy for cohesionless soils. Geotechnique 50(4):449–460 10. Manzanal D, Bertelli S, Lopez-Querol S, Rossetto T, Mira P (2021) Influence of fines content on liquefaction from a critical state framework: the Christchurch earthquake case study. Bull Eng Geol Environ 80:4871–4889 11. Li XS, Wang Y (1998) Linear representation of steady-state line for sand. J Geotech Geoenviron 124(12):1215–1217 12. Pastor M, Zienkiewicz OC, Chan AHC (1990) Generalized plasticity and the modeling of soil behavior. Int J Numer Anal Methods Geomech 14(3):151–190

Development of an Advanced Landslide Simulation Using Clustering Technology Kazuo Matsuura and Yasuhide Fukumoto

Abstract Vegetation has been used as an ecological measure to mitigate landslides. Some computational studies have focused on soil enforcement by plant roots. Generally, root system architecture is important on the relative influence of different soil failure modes. However, root system architectures considered in the past literature were simple shaped, such as straight bars or chained cylinders, and did not have branched shape as in the real root systems. In this study, the motion of soil particles interacting with the root growth model by Clausnitzer and Hopmans is computed in the framework of the discrete element method, and the effects on soil reinforcement are investigated. The time evolution of the soil under shear deformation applied impulsively at the initial time is reproduced. It is found that the distribution of soil particles interacting with the roots increases and also roots expand as time evolves. Keywords Landslide · Discrete element method (DEM) · Clustering

1 Introduction About 70% of Japan’s land area is covered by mountains and hills, and mountain slopes are steep and prone to collapse due to the small land size. The rivers flowing out of the mountain water sources are also steep, and thus, river overflows also likely to occur. In addition, climate change increases the risk of landslide disasters. Vegetation has been used as an ecological measure to mitigate landslides [1– 12]. Root effects on soil failure criteria (Mohr–Coulomb failure criteria) and slope failure have been reported. Among them, in [8], numerical simulations of threedimensional direct shear tests were performed using standard implicit finite element and individual element methods for slope stability due to root reinforcement. The K. Matsuura (B) Ehime University, 3 Bunkyo-Cho, Matsuyama 790-8577, Ehime, Japan e-mail: [email protected] Y. Fukumoto Institute of Mathematics for Industry, Kyushu University, Fukuoka 819-0395, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_14

157

158

K. Matsuura and Y. Fukumoto

numerical simulations were performed on low cohesion homogeneous soils, and the root model consisted of 36 linear, unbranched, thin roots planted in three parallel lines. Bourrier et al. [9] proposed a numerical model of direct shear tests of non-rooted and rooted granular soils based on the discrete element method (DEM). On the other hand, root system architecture is important on the relative influence of different failure modes. However, root system architectures considered in the past literature were simple shaped, such as straight bars or chained cylinders, and did not have branched shape as in the real root systems. The objective of this study is to incorporate into computation a root structure with a more realistic branching structure based on a root growth model and to investigate the effects of root strengthening when the soil is subjected to shear using DEM.

2 Numerical Method 2.1 Overview of the Computational Model In this study, a root model with a branched structure, which is generated by a root growth algorithm and represented by a point cloud, is placed among spherical soil particles as an initial condition. Simple shear in one direction is applied to the soil in order to investigate the effect of the presence of the root structure on resistance to shear deformation. In the framework of DEM, soil particles and root nodes are all represented by particles, and interaction between soil and roots leads to evaluating forces acting on the particles due to contact between them [12, 13]. When the radius of a particle is r p , the density is ρ, the velocity vector of a particle is v, the force vector acting on a particle is F, and time is t, the equation of motion for a single particle is as follows: dv 4 = F/m p , m p = ρ πr 2p dt 3

(1)

Here, r p = 0.5 cm, ρ = 2.6 × 10−3 g/cm3 [14]. In this study, it is assumed that the properties of particles constituting soil and roots are same for simplification.

2.2 Root Growth Model The root model used in this study is the three-dimensional root model proposed by Clausnitzer and Hopmans for investigating transient root growth and soil water flow [15, 16]. The model solves the Richards equation for the three-dimensional soil water head by finite element method (FEM) for soil water flow and solves the evolution equation for the evolution of the root tips with taking experimental data into account.

Development of an Advanced Landslide Simulation Using Clustering …

159

The model was constructed by dividing the model hierarchy into three levels: with and without transpiration and water uptake by the roots taken into account.

2.3 Root–Soil Interaction Interaction between roots and soil, (a) root segment–soil particle interaction, (b) root segment–root segment interaction, (c) restoring forces inside a single root segment, and (d) soil particle–soil particle interaction, is particularly considered. (a) Root Segment–Soil Particle Interaction Roots were represented by collection of points, and two neighboring points on the same segment are connected by a straight edge. To evaluate interaction between a segment AB and a soil particle P, a situation in which the soil particle collides with the edge was considered. First, the distance between particle P and segment AB was evaluated. Collision is considered to occur when the shortest segment from the center of particle P to segment AB is orthogonal to segment AB, and the distance is less than 2r p . Here, r p is the radius of a particle. When the distance between the segment and center P is denoted as d p , the normal stress σ r −s at the contact region is approximated as follows:    σr −s = d p r p Er −s .

(2)

Here, E r −s is the Young’s modulus for the soil and root materials. When the contact area is considered as a circle, the area Ar −s is approximated as follows: Ar −s = πrc2 .

(3)

Therefore, the repulsive force at the contact region is approximated as follows:  F r −s = −eσr −s Ar −s m p .

(4)

Here, e is the unit vector from the center of particle P to the central point of collision on segment AB. After the normal contact force Fr −s is evaluated, − 1/2 Fr −s is added to points A and B on segment AB, and Fr −s is added to point P, respectively. (b) Root Segment–Root Segment Interaction Each straight segment constituting the root is assumed to be short. The distance between the centers of gravity of segments A and B is denoted as d AB . The forces Fr −r acting on the centers of gravity were evaluated similarly to Eq. (4). The force Fr −r at the center of gravity was equally distributed to the end points of a segment. (c) Restoring Forces Inside the Root Segment

160

K. Matsuura and Y. Fukumoto

By defining the equilibrium distance d AB,eq of each segment before any root deformation at the initial state, the restoring force inside a segment is evaluated as follows: F r es = −kr es (d AB − d AB,eq )

(5)

(d) Soil–Soil Particle Interaction Contact forces between particles are evaluated similarly to Eq. (4).

3 Results and Discussion Figure 1 shows root growth for t = 200, 400, and 600 days. Bifurcated root structures are generated in the computed domain. The root system of 400 days is embedded in the initial soil field of the main computation which computes interaction between roots and sheared soil. In the main computation, roots were initially distributed in the soil, and the following shear was applied impulsively. v s = v s + v0 (z/40), v0 = 20 cm/s

(6)

Here, vs is the velocity of a solid particle. The velocity of landslide is small around the onset of the slide. The small value of v0 = 20 cm/s is assumed in this study.

(a) t=200 days

(b) t=400 days

(c) t=600 days

Fig. 1 Root growth (from z = 40 mm to z = 0 mm) for t = 200, 400, and 600 days. Different colors show different connected paths, i.e., cluster elements, in the root system. Some different paths are shown in a same color due to the limitation of the colors

Development of an Advanced Landslide Simulation Using Clustering …

161

Figure 2 shows the time evolution of the distribution of soil particles and roots at times t = t, 50t, and 100t. Here, t = 1e − 2 [s]. In regions where there are no roots, the soil is deformed by the simple shear. On the other hand, when there are roots in the soil, the soil undergoes shear and local internal displacement around the roots due to interaction with the soil. As time evolves, the distribution of soil particles interacting with the roots increases and also roots expand. Thus, a realistic root distribution based on the root growth model by Clausnitzer and Hopmans and its interaction with the soil under shear were successfully simulated. In future, more precise physical properties for roots and soil, deformation models due to collision, and quantitative evaluation of soil strengthening will be introduced.

(i)

t=1Δt

(ii) t=50Δt (a) Soil displacement

(iii) t=100Δt

(i)

t=1Δt

(ii) t=50Δt (b) Root node displacement

(iii) t=100Δt

Fig. 2 Interaction between sheared soil and roots

162

K. Matsuura and Y. Fukumoto

4 Conclusions In this study, the motion of soil particles interacting with the root growth model by Clausnitzer and Hopmans was computed in the framework of the discrete element method, and the effects on soil reinforcement were investigated. The time evolution of the soil under shear deformation applied impulsively at the initial time was reproduced, and it was found that the distribution of soil particles interacting with the roots increases and also roots expand as time evolves. Acknowledgements This study is a result of a Short-term Cooperative Research (Project Research) “Advanced Simulation of Landslide Disasters Using Clustering Techniques,” of Institute of Mathematics for Industry, Kyushu University.

References 1. Endo T (1980) Effect of tree roots upon the shear strength of soil. JARQ 14(2):112–115 2. O’Loughlin C, Ziemer RR (1982) The importance of root strength and deterioration rates upon edaphic stability in steepland forests. In: Proceedings of I.U.F.R.O. Workshop P.1.07–00 ecolgy of subalpine ecosystems as a key to management. 2–3 August 1982, Corvallis, Oregon. Oregon State University, Corvallis, Oregon, pp 70–78 3. Abe K, Ziemer R (1991) Effect of tree roots on a shear zone: modeling reinforced shear stress. Can J For Res 21:1012–1019 4. Ekanayake JC et al (1997) Tree roots and slope stability: a comparison between PINUS ¯ RADIATA and KANUKN. N Zeal J For Sci 27(2):216–233 5. Li ZA, Cai X (2007) Effects and mechanisms of plant roots on slope reinforcement and soil erosion resistance: a research review. Chin J Appl Ecol 18(4):895–904 6. Ireson AM, Butler AP (2009) Modelling plant root system development in response to soil water status: a review. Imperial 17:1–26 7. Kim D et al (2010) Estimating soil reinforcement by tree roots using the perpendicular root reinforcement model. Int J Eros Contrl Eng 3(1):80–84 8. Mao Z et al (2014) Evaluation of root reinforcement models using numerical modelling approaches. Plant Soil 381:249–270 9. Bourrier F et al (2013) Discrete modeling of granular soils reinforcement by plant roots. Ecol Eng 61P:646–657 10. Wang H et al (2018) Model test of the reinforcement of surface soil by plant roots under the influence of precipitation. Adv Mater Sci Eng 53:1–12 11. Masi EB et al (2021) Root reinforcement in slope stability models: a review. Geosciences 11(212):1–24. https://doi.org/10.3390/geosciences11050212 12. Zhao T (2017) Coupled DEM-CFD analyses of landslide-induced debris flows. Springer, New York 13. Mikio S (2012) Numerical simulation of granular flows. Maruzen publishing, Chiyoda 14. Kenji I (2018) Soil mechanics, 3rd edn. Maruzen publishing, Chiyoda 15. Clausnitzer V, Hopmans JW (1994) Simultaneous modeling of transient three-dimensional root growth and soil water flow. Pant Soil 164:299–314 16. Somma F, Clausnitzer V, Hopmans JW (1997) An algorithm for three-dimensional, simultaneous modeling of root growth, transient soil water flow, and solute transport and uptake

Effect of Microbial Strains Through Triaxial Test on Bio-Treated Granular Soil Yu-Syuan Jhuo, Pin-Hsiu Liu, Chang-Ping Yu, and Louis Ge

Abstract Microbiologically induced calcium carbonate precipitation (MICP) is a sustainable and environmental protection technique for ground improvement. In this study, the bacteria had determined Bacillus pasteurii, and the sources have different paths. One of the microbial strains was purchased from the bioresource collection and research center (BCRC) of the food industry research and development institute (FIRDI). The other microbial strains were collected from the sewage of Dihua sewage plant and then isolated and purified from the sewage sample. Calcium chloride had determined the calcium sources, grouted with a single-phase solution in the undrained specimen, and ensured that the specimen was completely infiltrated by bacterial solution. Curing time had determined at 7 and 14 days. The treated specimens were examined by the triaxial static consolidation drained test. Under effective confining pressure at 50,100, and 200 kPa, the results indicate that after 7 days of curing, the volume change of YU-SP was greater that of BCRC-SP. Nonetheless, a stable increase in peak strength was observed with increasing curing time in BCRCSP treated specimens, revealing distinct mechanical behaviors attributed to variations in bacterial strain sources. Keywords Microbiologically induced calcium carbonate precipitation · Consolidation drained (CD) test · Different microbial strains

1 Introduction Ground improvement is commonly utilized to increase the resistance of soil liquefaction at a given site. Taiwan is located on the circum-Pacific belt, where earthquakes occur frequently. The soil formation of western Taiwan is mostly due to sedimentary process, and the soil mainly consists of sand and slit. When an earthquake takes place, the soil particles tend to be re-arranged due to seismic load, resulting in the pore water pressure build-up and the effective stress decrease. Therefore, the Y.-S. Jhuo · P.-H. Liu (B) · C.-P. Yu · L. Ge National Taiwan University, Taipei, Taiwan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_15

163

164

Y.-S. Jhuo et al.

decrease of the bearing capacity of soil often leads to the differential settlements and sand boiling. For soil liquefaction prevention, conventional techniques of ground improvement, including vibration, compaction, mixing, and grouting, are associated with limitations and disadvantages. These techniques are time-consuming and expensive, and the case of chemical grouting and admixture is non-environmentally friendly. Therefore, the technique of microbiologically induced calcium carbonate precipitation (MICP) has been developed. It is considered as more environment-friendly and economical. Until now, it has various studies on different calcium sources, relative densities, cyclic injection, etc. The performance of MICP-treated soil would be influenced by several types, and the main factors are described as follows: (1) Bacterial solution concentration would affect the quantity of calcium carbonate; (2) the suitable temperature for gently MICP depositing the calcium carbonate is between 20 and 40 °C; (3) the most suitable pH value for the growth of bacteria is above 9 that the pH would certainly influence crystal appearance; (4) different calcium sources have been exhibited in various properties; (5) soil properties that have abundant mineral and suitable particle size would generally contribute to the improved cementation effect; (6) grouting technology would directly affect soil structure and solidification [2]. However, by utilizing the MICP in the field, soil samples can be collected at the location for bacterial identification. Based on the results of the collection and analysis, the optimal pH and temperature for bacterial activity have been determined [10]. Different calcium sources have different influences on the treated specimen [3, 4]. The treated specimen with increased cyclic injection could achieve the peak strength which is close to the dense soil by compaction [9]. The factor that would influence unconfined compression strength (UCS) is saturation and the content of calcium carbonate: The saturation of the treated specimen positively correlates with the calcium carbonate content, while with the increase of qucs and modulus of elasticity, the negative correlates with saturation [8]. The MICP-treated specimen showed brittleness, increased strength and peak compression value occurred earlier with higher calcium carbonate content in the UCS test [7]. The curing time increases with the shear velocity which indicates the longer curing time would obtain the greater effect of the strengthening of treated specimen [5]. The cyclic injection promotes four times where deviator stress–axial strain curve appears non-linear curve, considering the specimen was yielded [9]. The main reaction equations of MICP are as follows [6]: Ca2+ + Cell → Cell − Ca2+ urease

(1)

2− CO(NH2 )2 −→ 2NH+ 4 + CO3

(2)

Cell − Ca2+ + CO2− 3 → Cell − CaCO3 ↓

(3)

Effect of Microbial Strains Through Triaxial Test on Bio-Treated …

165

Fig. 1 Particle size analysis of Vietnam silica sand

2 Materials and Methodology 2.1 Sand In this study, Vietnam silica sand (C 306) was used. From the particle size distribution curve, the D60 , D50 , D30 , and D10 are 0.20, 0.18, 0.17, and 0.14 mm, respectively. It is a poorly graded sand (SP) based on the unified soil classification system. The coefficient of uniformity is 1.49, and the coefficient of curvature is 1.02. The special gravity is 2.66. The maximum and minimum dry densities are 1.69 and 1.39 and 0.911 and 0.574, respectively (see Fig. 1).

2.2 Bacterial In this study, the bacteria had determined Bacillus Pasteurii, and the sources have different paths. One of the microbial strains was purchased from the Bioresource Collection and Research Center (BCRC) of the Food Industry Research and Development Institute (FIRDI), which is named BCRC-SP in the study. The other was obtained from Professor Chang-Ping Yu’s laboratory at the graduate institute of environmental engineering at NTU. For the sake of simplicity, the source is abbreviated to YU-SP in the following discussion. For the cultivation of YU-SP, the microbial strains were collected from the sewage of Dihua sewage treatment plant. The culture medium utilized a rich medium that could rapidly brood the microbial strains. Streak a plate three times to obtain isolated colonies. The target of isolated colonies had been confirmed and transferred to the pure culture medium. For detailed instructions, please refer to this paper [1].

166

Y.-S. Jhuo et al.

2.3 Urea and Calcium Sources The type of urea is a reagent grade. The chemical formula of urea is CO(NH2 )2 , the molecular weight of urea is 60.06 g/mol, and the minimum purity is 99.0%. The concentration of urea in this study was controlled at 0.5 M. The source of calcium is calcium chloride known as the technical grade, the chemical formula is CaCl2, and the molecular weight is 111.1 g/mol. The properties of calcium chloride are high solubility in water and benefits to prepare the MICP specimen. Until now, calcium chloride has been found effective in accelerating concrete setting, but it could not be used in reinforced concrete due to the risk of steel bar corrosion from chloride. Additionally, the special gravity of calcium chloride is 2.151. The concentration of calcium chloride in this study was controlled at 0.5 M.

2.4 Treatment Solution Preparation The study attempted to use a single-phase treatment solution with surface percolation injection. Calcium source, urea, and bacterial solution are all components of the treatment solution. The single-phase treatment solution is a mixture of three components at the same time before injection. The pH value of bacterial solution determines the time required to induce calcium carbonate. Bacillus pasteurii is used to induce calcium carbonate by converting urea to ammonia and carbon dioxide. The cultivation was conducted under aerobic conditions. The ingredients of culture medium for BCRC-SP and YU-SP are displayed in Table 1. They both utilized the same method to prepare the bacterial solution and the procedure for preparing MICP-treated specimens. 1 ml of thawed frozen bacteria was added to 55 ml of culture solution after the frozen bacteria tube was removed from the −80 °C refrigerator. A shaker with 100 rpm was used to incubate the bacteria in the culture solution for 24 h at room temperature. The optical density (OD600 ) was measured between 1.200 and 1.500, and the pH value was about 9.45–9.55, so it was named “bacterial solution 1.” 5 ml of “bacterial solution 1” was then added to the culture medium (see Table 1) with urea free. Following 24 h of incubation, the OD600 value is measured between 2.300 Table 1 Composition of culture medium

BCRC-SP

YU-SP

5% Urea

5% Urea

15 g Agar

15 g Agar

3 g Yeast extract

3 g Beef extract

5 g Peptone

5 g Peptone

1 L Distilled water

1 L Distilled water

Effect of Microbial Strains Through Triaxial Test on Bio-Treated …

167

and 2.600, and the pH value is between 9.20 and 9.25, and it is called “bacterial solution 2.” The “Bacterial Solution 2” is added to the culture solution with urea at 15% volume and mixed well. The OD600 value was measured between 0.320 ~ 0.450, and the pH value is around 8.65 ~ 8.80. This step could make sure that the pH value of the bacterial solution is low enough to inhibit calcium ions combine with carbonate ions instantly when mix bacterial solution with urea and calcium, and avoid the accumulation of a large amount of calcium carbonate on the top of the specimen during the percolation process, which may contribute to the uneven improvement of the specimen.

2.5 Specimen Preparation For the specimen remolding of MICP-treated, the dry unit weight was achieved to be 1.527 g/cm3 and that the relative density was determined to be 50%. According to ASTM D7181-11, triaxial sand specimens had a diameter of approximately 7.3 cm and a height of 15 cm, resulting in a height-to-diameter ratio of more than 2.0. Paper molds with an open end were used for remolding in the study in order to prepare many MICP-treated specimens at one time (Fig. 2a). The other side of the mold is sealed with metal cap (Fig. 2b). The cap was drilled a few holes and sealed with tape (Fig. 2c). This is to ensure that it is undrained during injection and it can be drained after a few days of curing. Before remolding, the paper mold was sprayed with silicon oil to remove the paper mold easily. Five layers of sand were filled into the mold layer by layer. Filter papers were placed below and on the specimen (Fig. 3a). Fig. 2 Paper molds used for treated specimen: a drilled a few holes on cap (sealed the holes with tape); b create a bacterial solution; c pour 300 ml of bacterial solution into the remold specimen

168

Y.-S. Jhuo et al.

Fig. 3 Procedures of injection: a put the filter paper on surface; b create a bacterial treatment solution; c pour 300 ml of bacterial solution into the remold specimen

The amount of treatment solution (Fig. 3b) injected into the specimen is equal to the calculated saturated water content of the sand specimen. This is injected into the sand specimen in the undrained condition to ensure the sand was soaked in the solution (Fig. 3c). The amount of treatment solution was 300 ml for each triaxial sand specimen. The tape covering the drilled holes was removed after a few days of curing to let the treated specimen air-dry. The treated specimens were cured in a container that had a humidity of 85–90% at room temperature.

2.6 Triaxial Tests In this study, the saturated consolidated drained (CD) triaxial tests were conducted to estimate the mechanical behaviors of pure sand and MICP-treated soil specimens. In order to obtain reliable data, the specimens would be performed in saturated condition, in accordance with ASTM D7181-11. Except for load cell manufactured by SensoLink Co, the other triaxial equipment were manufactured by GDS. The MICP-treated specimens were handled carefully and directly transferred to the triaxial plate. The whole test would be divided into three stages: saturation, consolidation, and shearing. First, the saturation phase has two stages: manual saturation followed by back pressure saturation. Apply manual saturation until the air in the specimen had been exhausted by manual saturation. The back pressure was then increased along with the cell pressure, maintaining a constant effective cell pressure of 20 kPa until the specimen should be considered saturation if the value of Skempton’s coefficient B is equal to or > 0.95. The purpose of the consolidation of the test is to allow the specimen to achieve equilibrium in a drained state at the effective consolidation stress for which strength is determined with 50,100 and 200 kPa. In this study, the shearing rate is 0.185 mm/min. The shear

Effect of Microbial Strains Through Triaxial Test on Bio-Treated …

169

loads continued to be recorded up to 20% of axial strain to observe the peak of the destructive stress and residual stress.

3 Results and Discussion 3.1 Calcium Carbonate Content of BCRC-SP and YU-SP 30 g of the totally dried MICP-treated sand specimen was sampled for acid digestion, which is a method for examining the calcium carbonate content. Calcium carbonate was decomposed by 4 M hydrogen chloride (HCl) into calcium ions and carbonate ions that were dissolved in the solution. The remaining sand was washed with distilled water many times until the pH returned to 6.80 and then oven-dried. The weight difference between before and after acid washing is the content of calcium carbonate. The calcium carbonate of YU-SP is higher than that of BCRC-SP, and both the contents depend on the curing time.

3.2 Triaxial CD Test Results Stress–Strain Curves and Volume Changes. BCRC-SP- and YU-SP-treated specimen results of triaxial test are shown in Figs. 4, 5 and 6. Both of the treated specimens were at concentration 0.5 M and confining pressure 100 kPa. For the stress–strain curve of BCRC-SP, the peak strength is increasing with curing time, and the strain of peak strength at different curing times is larger than untreated specimens. Although the specimen treated with YU-SP also displays a higher value on the strain of peak strength, the peak strength is not significantly different from untreated specimen. The volume changes of YU-SP-treated specimens are smaller than untreated specimen. Both MICP-treated specimens have lower volume change at the contraction stage during shearing. Summary of peak strength and strain for CD test at 50, 100, 200 confining pressure is shown in Table 2. The peak strengths of BCRC-SP-treated specimen at 14-day curing are higher than 7-day curing, and the peak strength happened corresponds to the axial strain which is lower than 7-day curing. The 60-day curing displays brittle behavior during shearing. The results of YU-SP-treated specimens are not related to curing time at different confining pressure. In the BCRC-SP, the peak strength strain occurred early in the treated specimen. Additionally, as curing time increased, there was a faster rate of strength enhancement, resulting in greater performance. Regarding volume change and axial strain (Fig. 6b) in YU-SP, the curves at 7 and 14 days exhibited stable decreases in volume change with longer curing times, indicating a tendency towards brittle behavior. These outcomes demonstrate different mechanical behaviors.

170

Y.-S. Jhuo et al.

Fig. 4 Calcium carbonate content of two bacterial strains at 7- and 14-day curing

(a) 400 350

Deviator stress (kN/m2)

Fig. 5 BCRC-SP at different curing times: a Deviator stress–axial strain curve, b volume change–axial strain curve

300 250 200 150 100

Pure Sand_CD100kPa BCRC_Chloride_0.5M_7D1C_CD100kPa BCRC_Chloride_0.5M_14D1C_CD100kPa

50 0

(b)

7.0

Volume change (%)

6.0 5.0 4.0 3.0 2.0 1.0

Pure Sand_CD100kPa BCRC_Chloride_0.5M_7D1C_CD100kPa BCRC__Chloride_0.5M_14D1C_CD100kPa

0.0 -1.0 0

5

10

15

20

25

Axial strain (%)

Cohesion and Friction Angle. Table 3 shows the summary of strength parameters for triaxial CD test. The c and phi of untreated specimen are 0 kPa and 39.52°. Both of the MICP-treated specimens have significant increase on cohesion, which are around 11 kPa. However, the friction angles of treated specimens decrease slightly compared to the untreated specimen. The BCRC-SP-treated specimen seems to have a relation with curing time.

Effect of Microbial Strains Through Triaxial Test on Bio-Treated … (a) 400 350

Deviator stress (kN/m2)

Fig. 6 YU-SP at different curing times: a Deviator stress–axial strain curve, b volume change–axial strain curve

171

300 250 200 150 100

Pure Sand_CD100kPa YU_Chlorde_0.5M_7D1C_CD100kPa YU_Chloride_0.5M_14D1C_CD100kPa

50 0

(b) 7.0

Volume change (%)

6.0 5.0 4.0 3.0 2.0 1.0

Pure Sand_CD100kPa YU_Chlorde_0.5M_7D1C_CD100kPa YU_Chloride_0.5M_14D1C_CD100kPa

0.0 -1.0 0

5

10

15

20

25

Axial strain (%)

Table 2 Summary of peak strength and strain for CD test Confining pressure (kPa) Microbial strains 50

Peak strength (kPa)

200.24 209.30

Peak strength (kPa) Strain of peak strength (%)

200

YU-SP

7

Strain of peak strength (%) 100

BCRC-SP

Curing times

Peak strength (kPa) Strain of peak strength (%)

14 4.39

60 3.28

7 – –

14

203.94 201.16 5.03

4.08

354.99 371.75 346.35 359.52 364.13 5.22

5.33

666.85 697.90 6.87

5.24

3.83 – –

4.92

6.06

670.17 671.24 6.87

6.12

172 Table 3 Summary of strength parameters for CD test

Y.-S. Jhuo et al.

Condition BCRC YU PS

Curing times

Strength parameter c

ϕ

7

11.8

37.683

14

11.2

38.224

7

11.9

37.604

14

10.9

37.748

0

0

39.52

4 Conclusion Different Microbial Strains. For the condition of 14-day curing and 100 kPa effective confining pressure, YU-SP is greater than BCRC-SP in terms of reducing the volume change of the treated specimen. The calcium carbonate content of YU-SP is higher than BCRC-SP. However, it seems to have little effect on improving peak strength, but it may reduce volume change. Therefore, it appeared that different strain sources would also become an influential factor on mechanical properties. Curing Times. The BCRC-SP-treated specimen that has been placed for over 60 days was observed in comparison with the one of 7-day curing. The peak strength decreased from 5.22 to 3.83% when the curing time increased from 7 to over 60 days. It was found that the longer curing time would ensure to gently cement the soil particle together with calcium carbonate after depositing. Moreover, the specimen would be solid. The Shape of the Calcium Carbonate Crystallization. The calcium carbonate crystallization of YU-SP is as granular as fine-grained soil and filled among soil particles. The curing time could be reduced to achieve the same amount of volume change as BCRC-SP’s. For the purpose of demonstrating the effectiveness of the cementitious between YU-SP and BCRC-SP, the XRD analysis will be compared with the results of the triaxial tests.

References 1. Shih CC (2022) Evaluating the application of simple 3D bioprinting in microbial immobilization: a case study of ureolytic bacteria and heavy metal removal. Master Thesis (National Taiwan University) 2. Tang CS, Yin LY, Jiang NJ, Zhu C, Zeng H, Li H, Shi B (2020) Factors affecting the performance of microbial induced carbonate precipitation (MICP) treated soil: a review. Environ Earth Sci 79(5):1–23 3. Kadhim FJ, Zheng J (2017) Influences of calcium sources, concentration of the cementation solution and type of sand on microbial induced carbonate precipitation. Int J Civil Struct Environ Infrastruct Eng Res Develop 7(2):23–32

Effect of Microbial Strains Through Triaxial Test on Bio-Treated …

173

4. Ako˘guz H, Çelik S, Bari¸s Ö (2019) The effects of different sources of calcium in improvement of soils by microbially induced calcite precipitation (MICP). Sigma J Eng Nat Sci 37(3):953–965 5. Chen HJ, Tsai TM, Wu CH (2021) Research on the feasibility of strengthening the soil structure by biomineralization. J Chin Instit Eng 44(3):214–222 6. DeJong JT, Fritzges MB, Nüsslein K (2006) Microbially induced cementation to control sand response to undrained shear. J Geotech Geoenviron Eng 132(11):1381–1392 7. Van Paassen LA, Ghose R, Van der Linden TJ, van der Star WR, Van Loosdrecht MC (2010) Quantifying biomediated ground improvement by ureolysis: large-scale biogrout experiment. J Geotech Geoenviron Eng 136(12):1721–1728 8. Cheng L, Cord-Ruwisch R, Shahin MA (2013) Cementation of sand soil by microbially induced calcite precipitation at various degrees of saturation. Can Geotech J 50(1):81–90 9. Gao Y, Hang L, He J, Chu J (2019) Mechanical behaviour of biocemented sands at various treatment levels and relative densities. Acta Geotech 14(3):697–707 10. Sivakumar G (2020) Experimental study on near-surface stabilization of slopes in cold region using bio-mediated, doctoral dissertation (Hokkaido University). Geotechnical Engineering

Experimental Study of Warm Permafrost Mechanical Property Under Cyclic Load Liwei Song , Junfang Liu , Xiaomin Liu, and Xin Zhao

Abstract In order to study the dynamic response of the warm permafrost subgrade under traffic loads, the low temperature dynamic triaxial test was conducted on remolded soil which extracted from Yakeshi highway in Inner Mongolia Autonomous Region and gained the hysteretic loop curves, while the deviatoric stress changes from 0 to 50 kPa. The cyclic loading frequency is 6 Hz and the test temperature is − 1.5 °C. The result shows that each deviatoric stress–strain hysteresis loop can be divided into two parts, and data points can be described by sine function separately for each part. The phase angle of sine function is changing with the number of loading cycles and the sample always keeps stable. The warm permafrost hysteresis loop area linearly reduces and the dynamic modulus exhibits a pattern of initial increase, followed by a decrease, and then a subsequent increase. Keywords Warm permafrost · Hysteretic loop · Cyclic load · Mechanical property

1 Introduction The warm permafrost located in northwestern and northeastern China is a puzzle for construction project. Its characteristics directly affect the structure design and construction in cold region. For the highway subgrade, with increasing of the traffic cyclic loading times, the accumulative deformation is obvious and may affect normal transportation, so it is important to study the mechanical characteristics of warm permafrost under dynamic load. The warm permafrost is the soil that the temperature ranges from − 1.5 to 0.0 °C, which is also called phase change zone permafrost. L. Song (B) · X. Liu China Construction Sixth Engineering Bureau Co. Ltd., Tianjin, China e-mail: [email protected] J. Liu Inner Mongolia University of Technology, Hohhot, Inner Mongolia Autonomous Region, China X. Zhao Hebei University of Engineering, Handan, Hebei, China © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_16

175

176

L. Song et al.

The warm permafrost is sensitive to the temperature change and its mechanical characteristics are more complicated than normal permafrost. If the warm permafrost is just under road, it may have some bad impact on transportation. Xu [1] pointed out the influence parameters to permafrost damping ratio, such as test temperature, confining pressure, axial vibration frequency, and cyclic times. Zhao [2] found the changing law between dynamic mechanical parameters and load frequency, temperature and water content for the frozen silty clay and frozen fine sand embedded in Qinghai-Tibet railway. Gao [3] found the influence parameters of dynamic modulus and damping ratio by generalizing hyperbolic model for the remolded soil in Qinghai-Tibet railway. Shi [4] found that the warm permafrost elastic modulus decreases with temperature increasing and increases with the cyclic frequency increasing for the soil embedded in Qinghai-Tibet railway. The dynamic damping ratio increases with load frequency increasing, and decreases with temperature dropping. Luo [5] proposed a new method to calculate damping value of the large plastic deformation permafrost. Gao [6] gained the relationship between dynamic strength and residual strain for high content permafrost by the triaxial test. Zhang [7] studied the dynamic strength of frozen clay, and pointed out that main reasons of controlling the dynamic strength are off-energy accumulation and rising temperature. Jiao [8] studied the law between hysteresis loop change and deformation accumulation under cyclic load for warm permafrost, and pointed out that the maximum stress amplitude is key to the accumulation strain and hysteretic loop area. Otherwise, some other scholars [9–15] conducted series of research for the permafrost and gained lots of conclusions. Yakeshi city is located in the northeast of Inner Mongolia Autonomous Region, and a larger area of warm permafrost is distributed in this city. Because a normal car’s weight is about 14 kN, the axial load amplitude is 50 kPa in the test. The triaxial tests were conducted to study the dynamic characteristics of remolded warm permafrost and may provide some reference to the design and construction work for roads and railways.

2 Test Instrument, Samples, and Test Program 2.1 Test Instrument and Loading Mode. The tests were conducted on the STX-100 dynamic triaxial instrument made by American corporation GCTS, as shown in the Fig. 1. The maximal axial force of the test instrument is 25 kN, and the maximal axial loading frequency is 10 Hz. The maximal loading distance is 50 mm, and the maximal confining pressure is 2 MPa. The testing machine chamber is connected to a constant low temperature circulatory system, the sample can be tested at a preset temperature, which can be changed from − 50 to 200 °C. The precision of temperature control is ± 0.1°C.

Experimental Study of Warm Permafrost Mechanical Property Under …

177

Fig. 1 The low temperature dynamic triaxial test equipment

Fig. 2 Loading process of cyclic triaxial test

Deviator stress / kPa

Using cyclic sine load to simulate the traffic vehicle, the confining pressure is 25 kPa, and the axial pressure is changing from 25 to 75 kPa during the test. At the first stage, both the axial pressure and confining pressure increase by a same rate until the confining pressure reaches preset value. At the second stage, the sample is maintained at steady state for consolidation, and then the axial cyclic load is applied, as shown in the Fig. 2.

C

A B Dynamic loading stage Consolidation stage time / s

178

L. Song et al.

Table 1 Test scheme Temperature /°C

Frequency /Hz

Confining pressure/kPa

Dynamic stress amplitude/ kPa

− 1.5

6

25

50

2.2 The Specimen Production The sample soil was extracted from the highway subgrade in Yakeshi city, then is paved to lamella on desk to air seasoning in the laboratory. After being sieved by 2 mm geotechnical screen, the soil density is kept at 1.74g/cm3 with 16.5% water content, then the soil is made into samples. According to the ‘Standard for geotechnical testing method, GB/T 50123-2019’, the sample diameter is 38 mm and height is 76 mm. The sample is fixed in the freezing chamber of triaxial instrument and kept at − 1.5 °C for 12 h, then start applying load on the sample.

2.3 Test Diagram The low temperature triaxial test parameters are listed in the Table 1. According to the ‘Standard for geotechnical testing method, GB/T 50123-2019’, the sample failure standard is that the summation of elastic strain and plastic strain reaches 5%. The test will automatically stop at 5000 loading cycles, if the total strain does not exceed 5%.

3 The Test Results and Results Analysis 3.1 The Deviator Stress-Stain Curves Under Cyclic Loading The deviator stress–strain curves for the 100th–110th, 1000th–1010th, 10000th– 10010th, 30000th–30010th, 49990th–50000th loading cycles are shown in Fig. 3. With the number of loading cycles increasing, the hysteretic curves become denser and denser, but the cyclic loop shape always keeps stable and moves to the right direction gradually. Under the dynamic cyclic loading, the plastic deformation increment decreases gradually, and the hysteretic curves appear to be parallelograms. The hysteretic curves appear viscoelastic-plastic characteristics at the early loading stage, and the maximal axial strain of warm permafrost is less than 0.46% at last. The sample plastic deformation increases with an increase in the number of load cycles. For the holes and cracks are compressed, the plastic deformation increment of sample decreases gradually, until the plastic deformation stops increasing and the sample remains stable at last.

Experimental Study of Warm Permafrost Mechanical Property Under …

179

Fig. 3 The hysteretic loop while σ 3 = 25 kPa and |σ 1 − σ 3 |max = 50 kPa

3.2 The Hysteretic Curves Fitting. The 1000th, 5000th, 10000th, 25000th, 50000th hysteretic curves are divided into upper part and lower part by a line respectively, as shown in the Fig. 4. For the 1000th hysteretic curve, the data points above the dividing line can be fitted by sine function, as well as the other points, as shown in the Fig. 5 and in Fig. 6. The deviatoric stress changes with strain by sine function, as shown by the following equation. σ1 − σ3 = 58 + 13.7 · sin(62εd + 9.8)

(1)

where σ1 and σ3 (kPa) are the axial stress and confining pressure respectively, and the correlation coefficient R 2 = 0.92. The deviatoric stress changes with strain by sine function for the data points below the dividing line, as shown by the following equation: σ1 − σ3 = 40.9 + 14.4 · sin(62εd + 15.5)

(2)

180

L. Song et al.

Fig. 4 The deviatoric stress–strain curve for the 1000th loading

80

The hysteretic curve data

σ1-σ3 / MPa

70 60 50 40 30 20 0.21

0.22

0.23

0.24

0.25

0.26

0.27

0.28

0.29

εd(%)

Fig. 5 The fitting curve of data points above the dividing line

Date points

75

Fitting curve

σ1-σ3 / MPa

70 65 60 55 50 45 40 0.22

0.23

0.24

0.25

0.26

0.27

0.28

0.29

0.27

0.28

0.29

εd(%)

σ1-σ3 / MPa

Fig. 6 The fitting curve of lower part of deviatoric stress–strain

60

Date points

55

Fitting curve

50 45 40 35 30 25 20 0.22

0.23

0.24

0.25

0.26

εd(%)

Experimental Study of Warm Permafrost Mechanical Property Under …

181

where σ1 and σ3 (kPa) are the axial stress and confining pressure respectively, and the correlation coefficient R 2 = 0.92. By the same way, the 5000th, 10000th, 25000th, 50000th hysteretic curves can be fitted by sine functions, as shown by the following equation: σ1 − σ3 = A + B · sin(62εd + C)

(3)

where σ1 , σ3 (kPa), and εd are the axial stress, confining pressure, and axial strain, respectively. All the parameters (A, B, and C) of fitting function are listed in the Table 2. The hysteretic curve areas are calculated respectively for the 1000th, 5000th, 10000th, 25000th, and 50000th loading cycles, and the curve area linearly decreases with number of loading cycles increasing, as shown in the Fig. 7. Table 2 The parameters of fitting function Cyclic time

Data points

A

B

The correlation coefficient R2

C

1000

Above dividing line

58

13.7

9.8

0.92

1000

Below dividing line

40.9

14.4

15.5

0.92

5000

Above dividing line

56.3

14.4

− 13.2

0.87

5000

Below dividing line

40.8

11.8

− 7.6

0.92

10,000

Above dividing line

56.8

14.8

− 0.3

0.92

10,000

Below dividing line

42.8

15.6

− 0.3

0.93

25,000

Above dividing line

56.5

13.9

− 0.12

0.8

25,000

Below dividing line

37.6

12.6

− 0.13

0.89

50,000

Above dividing line

55.8

13

− 0.1

0.92

50,000

Below dividing line

40.6

15.2

−0.4

0.96

Fig. 7 The relationship between hysteretic loop area and number of loading cycles

The hysteretic loop area

0.0140

Hysteretic loop area

Fitting curve 0.0135

0.0130

0.0125

0.0120 0

10000

20000

30000

40000

Number of loading cycles

50000

182

L. Song et al.

In the loading cycles process, the soil is compressed, and becomes denser and denser, and the pores between solid particles were reduced. The relationship between hysteretic loop area and number of loading cycles is shown by the following equation: S = −3.73 × 10−8 · N + 0.01384

(4)

where S is hysteretic loop area, and N is number of loading cycles. The correlation coefficient R 2 = 0.85.

3.3 The Dynamic Modulus and Dynamic Damping Ratio By Ref. [13] and Fig. 8, the dynamic modulus E d and dynamic damping ratio λ can be calculated by Eq. (5) and Eq. (6). σd εd

(5)

1 Sshadow π SABC

(6)

Ed = λ=

where σd (kPa) is axial stress, εd (%) is axial train, Sshadow is the shadow area in the Fig. 8, and SABC is the area of ABC in the Fig. 8. For the 1000th, 5000th, 10000th, 25000th, and 50000th hysteretic loop curves, the value of dynamic modulus and dynamic damping ratios were calculated, respectively, by the before-mentioned method, and all the values are listed in the Table 3. The sample basically maintains stable in this test. At the first stage of loading, the particles were compressed gradually, and it is denser and denser with number Fig. 8 The single hysteretic loop curve

Experimental Study of Warm Permafrost Mechanical Property Under …

183

Table 3 The dynamic modulus and damping ratios Hysteretic times

1000

5000

10,000

25,000

50,000

E d /MPa

90

90.9

86.4

93.8

96.1

λ

0.33

0.29

0.26

0.26

0.25

of loading cycles increasing, and the dynamic modulus also increases during this process. When the number of loading cycles reaches 10,000, a little of particles cracked, and the dynamic modulus decreases. For the continuous cyclic loading, the new cracks and pores were reduced, and the dynamic modulus increases again until the test ended. For the soil is compressed to be denser and denser, its energy dissipation capacity decreases, so the damping ratio decreases in the test.

4 Conclusion The warm permafrost sample was tested with 25 kPa confining pressure when the test temperature is − 1.5 °C, and the axial cyclic stress changes from 25 to 75 kPa, then gained all the 50,000 deviatoric stress–strain curves. (1) The hysteretic loop curves of warm permafrost appear to be parallelograms and strain hysteresis is obvious. (2) The hysteretic loop curves of warm permafrost always maintain stable and the strain of sample increases gradually, but the total axial strain is less than 0.46% at last. (3) Each hysteretic loop curve can be divided into two parts by a line and data points in each part can be described by sine function, respectively. (4) The hysteretic loop area linearly decreases with the number of loading cycles increasing. (5) The soil particles undergo three stages: compaction, particle cracking, and densification. The dynamic modulus exhibits a pattern of initial increase, followed by a decrease, and then a subsequent increase. The dynamic damping ratio decreases gradually in the test. Acknowledgements This research was funded by the China State Construction Engineering Corporation, grant no. CSCEC-2020-Z-46.

References 1. Xu C, Xu X, Qiu M (2002) Experimental study on dynamic damping ratio of frozen soil under cyclic loading. J Harbin Univ C.E. Architect 35(006):22–25

184

L. Song et al.

2. Zhao S, Zhu Y, He P (2003) Testing study on dynamics parameters of frozen soil. Chin J Rock Mech Eng 22(S2):2677–2681 3. Gao Z, Lai Y, Xiong E (2010) Experimental study of characteristics of warm and ice-rich frozen clay under cyclic loading. Rock Soil Mech 31(6):1744–1751 4. Shi Y, He P, Bian X (2006) Testing study on dynamic parameters of warm frozen soil of Qinghai-Tibet railway. Subgrade Eng 5:93–95 5. Luo F, Zhao S, Ma W (2015) Analysis of dynamic properties of frozen clay by morphological characteristics of hysteresis curves. Rock Soil Mech S1:299–304 6. Gao Z, Shi J, Zhang S (2009) Experimental study of dynamic strength characteristics and residual strain of ice rich frozen soil. J Glaciol Geocryol 31(6):1143–1149 7. Zhang S, Lai Y, Li S (2008) Dynamic strength of frozen soils. Chin J Geotech Eng 30(004):595– 599 8. Jiao G, Zhao S, Ma W (2013) Evolution laws of hysteresis loops of frozen soil under cyclic loading. Chin J Geotech Eng 35(7):1343–1349 9. Liu J, Cui Y, liu X (2020) Dynamic characteristics of warm frozen soil under direct shear test-comparison with dynamic triaxial test. Soil Dynam Earthquake Eng 133:106–114 10. Zhao X, Yang G, Tian J (2017) Research on the Compression and deformation characteristics of freezing-thawing cycles of high moisture content of loess. Chin J Underground Space Eng 13(4):899–904 11. Leng W, Zhai B, Xu F (2020) Probabilistic model of cumulative plastic strain of coarse-grained soil fill based on large-scale dynamic triaxial tests. J Vibr Shock 39(15):214–249 12. Huang B, Fu X, Zhang B (2015) Test technology and normalized characteristics of dynamic elastic modulus and damping ratio. Chin J Geotech Eng 37(4):659–666 13. He F, Wang X, Liu D (2017) Experimental study on dynamic characteristic parameters of undisturbed frozen sandy soil of Qinghai-Tibet railway. J China Rail Soc 39(6):112–117 14. Cui H, Wang W, Shao B (2022) Experimental Research on freezing and thawing characteristics of improved coarse grain filling material for high-railway foundation in seasonal freezing area. J Safety Environ 220(2):770–777 15. Shan R, Song L, Bai Y (2014) Model study of damage evaluation of frozen rock wall under blasting load. Chin J Rock Eng 33(10):1945–1952

Influence of Initial Static Shear Stress on the Dynamic Response of Embedded Cantilever Retaining Wall with Saturated Backfill Anurag Sahare , GyuChan Choi, and Kyohei Ueda

Abstract Earthquake-induced lateral spreading usually takes place under a stress state governed by the presence of initial static shear stress coupled with time varying cyclic stresses. However, due to the unavailability of high-quality element test data conducted with an initial static shear bias, usually the numerical modeler ignores the presence of static shear stresses on a soil element. This paper presents the initial calibration framework and the subsequent numerical insights on the deformation mechanism for an embedded cantilever retaining wall subjected to dynamic loading, when the initial static shear stress is comprehensibly considered in the constitutive model framework. For the sake of effective comparison, calibrations were also conducted without any static shear stress bias. A cocktail glass model was used to represent the soil elements, for which the initial calibration was conducted based on the results of cyclic direct simple shear tests. The constitutive model was able to capture important features arising due to the initial static shear stress including considerable lesser degradation in the shear modulus of soil due to limited generation of excess pore pressure under the subsequent undrained cyclic loading. Post initial calibration, a system level performance was conducted to evaluate the effects of static shear in terms of excess pore pressure generation and sheet-pile deformation mechanism. The simulations revealed the occurrence of predominant dilative responses representing the soil stiffening during the cyclic shearing without the application of initial static shear bias as compared to a case with initial static shear stress leading to a different deformation mechanism. Keywords Constitutive model · Initial static shear stress · Earthquakes

A. Sahare (B) Advanced Research Laboratories, Tokyo City University, Tokyo, Japan e-mail: [email protected] G. Choi Department of Civil and Earth Resources Engineering, Kyoto University, Kyoto, Japan K. Ueda Disaster Prevention Research Institute, Kyoto University, Kyoto, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_17

185

186

A. Sahare et al.

1 Introduction Liquefaction-related retaining wall damages have been frequently reported during the previous earthquakes particularly in Japan e.g., [1–3]. The substantial increase in excess pore water pressure during the 1995 Kobe earthquake was one of the main reasons leading to the collapse of multiple retaining structures which were originally designed according to the conventional Mononobe-Okabe method as highlighted by Iai [4]. The Mononobe-Okabe method, which is originally based on the idea of a pseudo-static analysis, cannot consider some of the important mechanism involving soil-structure interactions including the time-varying kinematic aspects of the cyclically loaded soil. The saturated soil, when continuously compressed and cyclically loaded under undrained conditions, may also manifest dilatant mechanism which cannot be captured under such simplified analysis. These mechanisms will have consequential effects on the deformation of a sheet-pile which will lead to displacements far different than a rigid block movement normally idealized in a rigid block theory. Understanding of these complex mechanisms demands numerical modeling by adopting a properly calibrated constitutive model from element tests and centrifuge models. It is also important for the constitutive model to effectively capture important soil features like dilatancy and soil anisotropy when cyclically loaded. To this end, Liquefaction Experiments and Analysis Projects was established, which is a joint research collaboration venture bringing researchers from different parts of the world together to study the soil liquefaction mechanism and its effects on soilstructure interactions using high-quality centrifuge modeling, element testing and numerical modeling based on various complex constitutive models [see special issue in SDEE e.g., [5–7]. As a part of LEAP-UCD 2017 Project, 24 centrifuge model tests involving a liquefiable sloping ground (made of Ottawa F-65 sand) were performed at 9 different centrifuge facilities around the world. This was further followed by LEAPASIA-2019 Project, which led to the development of 24 additional tests which also led in validating the generalized scaling laws. Eventually, a well-defined 3-D regression surface was established based on 48 high-quality centrifuge tests, which defined relationship between the relative density of soil, lateral displacement of soil and the applied earthquake characteristic in terms of PGAeff (see [7]). Apart from centrifuge modeling, LEAP also aims at understanding the phenomenon of soil liquefaction and its effects on soil-structure interactions using numerical modeling by employing various high end constitutive models. These include the strain space multiple mechanism model, PDMY02, CycLiqCP, PM4Sand and SANISAND-SFR. For this purpose, initially, the numerical modelers are provided with the element test data with soil sample being sheared under different levels of cyclic shear stress or vertical confining stress. However, it is quite adamant that the cyclic response of such saturated sand may also be governed by the initially applied static shear stress, which is particularly relevant for cases involving lateral soil spreading. Numerical modelers often ignore the presence of such initially applied static stress into their constitutive model calibration, when dealing with liquefactionrelated issues. The key objective of this paper is to understand the cyclic response of a

Influence of Initial Static Shear Stress on the Dynamic Response …

187

saturated soil when sheared under the presence of static shear bias with combinations of different confining stress and static stresses. Finally, a bigger picture is to answer the question on how the system level performance is governed by the presence of initial static shear stress involving a cantilever wall with a saturated backfill. Within this realm, recently under the sidelines of LEAP-GWU-2022, cyclic direct simple shear tests (CDSS) were performed at the George Washington University (GWU) with initial static shear stress. These tests were provided to the simulation teams which provide the unique opportunity to assess the effects of such static shear on the system level performance.

2 Constitutive Model The strain space multiple mechanism model or a cocktail glass model was utilized for the research work related to this paper. The granular material within this model is idealized based on virtual simple shear mechanism oriented in arbitrary directions correlating the micromechanical structures to macroscopic deformations. Due to this feature, the model can consider the simple shear mechanisms in different virtual directions under complicated loading paths while considering the rotation of the principal stress axes. Recently, Iai et al. [8] extended the theoretical formulation of the micromechanical and macroscopic strain energy to incorporate the effects of dilatancy in a cocktail glass model, which is defined as a summation of contractive and dilative components of dilatancy. The model can simulate the cyclic behavior of granular materials under isotropic and anisotropic stress conditions. Overall, the model parameters are classified based on volumetric, shear and dilatancy mechanism. Further details about the various constitutive model parameters and their respective derivations can be seen in [8, 9]. The strain space multiple mechanism model is available in a finite element program called as FLIP ROSE (Finite element analysis program of Liquefaction Process/Response of Soil-structure systems during Earthquakes) and is widely used to solve the geotechnical-related problems involving soil-structure interactions and has also been consistently used during the previous LEAP projects.

3 Calibration of Element Tests with Static Shear (Class-A Predictions) The numerical model for the constitutive model calibration comprised of two elements and four nodes. The soil elements were modeled using the strain space multiple mechanism model, while the generation of pore water during undrained shearing was considered using the pore pressure elements. Initially, the constitutive model parameters were set as equivalent to those obtained during the previous LEAP

188

A. Sahare et al.

project, which were then subsequently altered to consider the effects of initial static shear bias with the newly conduced cyclic direct simple shear tests (CDSS) [10]. The soil as initially specified is Ottawa F-65 sand. For numerical models with the initial static shear bias, initially, the consolidation stress was applied in the form of vertical stress, followed by the application of initial static shear stress with perfectly drained conditions. Lastly, a cyclic load was applied under undrained conditions and the soil sample was sheared leading to generation of excess pore water pressure. The constitutive model parameters obtained after simulation considering a static shear bias at a vertical confining stress of 40 and 100 kPa are shown in Table 1. The constitutive model parameters were also obtained from the element tests, which were sheared without any initial static shear bias, to understand the different mechanisms arising. As an example, Fig. 1 shows the stress–strain relationship, estimated stress path, evolution of shear strain and excess pore pressure under undrained Table 1 Calibrated constitutive model parameters for the strain space multiple mechanism model Parameter

Unit

Parameter designation

Simulation and prediction

ρt

t/m3

Mass density

2.032

pa

kPa

Reference confining pressure

75

K L/Ua

kPa

Bulk modulus

78,235

rK

–

Reduction factor of bulk modulus for liquefaction analysis

0.5

lK

–

Power index of bulk modulus for liquefaction analysis

2

Gma

kPa

Shear modulus

30,000

Internal friction angle for plane strain

41.5

Upper bound for hysteretic damping factor

0.24

φ PS f hmax

–

φp

Phase transformation angle

28

εdcm

–

Limit of contractive component

0.85

r

εdc

–

Parameter controlling contractive component

1.7

rεd

–

Parameter controlling dilative and contractive components

0.12

q1

–

Parameter controlling initial phase of contractive component

3

q2

–

Parameter controlling final phase of contractive component

0.35

S1

–

Small positive number to avoid zero confining pressure

0.005

c1

–

Parameter controlling elastic range for contractive component

2.00

qus

–

Undrained shear strength (for steady state)

0

Influence of Initial Static Shear Stress on the Dynamic Response …

189

Fig. 1 Calibrated stress path, stress–strain relationship, shear strain and excess pore pressure generation with the cyclic loading under an applied vertical stress of 40 kPa and a cyclic stress of 0.15 (without static shear bias)

cyclic loading under an applied vertical stress of 40 kPa and a cyclic stress ratio of 0.15, without the application of any static shear bias. The constitutive model is found to replicate the desired soil features when cyclically loaded in undrained conditions, with nearly similar number of cycles to generate the desired shear strain and the excess pore pressure as measured experimentally. Figure 2 shows the calibrated stress path, stress–strain relationship, generation of shear strain and excess pore pressure with the cyclic loading under an applied vertical stress of 100 kPa and a cyclic stress ratio of 0.15 with a static shear bias of 40 kPa. Both the measured and experimental responses suggest the substantial increase in the liquefaction resistance due to the application of static shear, with the measured and computed stress path being far away from the critical state line. Due to this, soil does not liquefy with excess pore pressure ratio remaining far below 1.0 even after the application of a large cyclic load. Although the computed responses are slightly different from the measured one’s, the important features of sand under the application of static shear are very well captured. The detailed analysis of the sand features under the application of static shear bias on the element level will be discussed in our future publications.

190

A. Sahare et al.

Fig. 2 Calibrated stress path, stress–strain relationship, shear strain and excess pore pressure generation with the cyclic loading under an applied vertical stress of 100 kPa and a cyclic stress of 0.15 (with initially applied static stress of 40 kPa)

4 Influence of Initial Static Shear Stress on the System Level Performance Involving a Retaining Wall (Class-B Predictions) The finite element models were developed using a computer program FLIP-ROSE in the 2-D domain comprising of 1733 nodes and 3022 elements. The strain space multiple mechanism model was used to model the soil elements, whereas the sheetpile was modeled using the linear beam elements, having a Young’s modulus of 69 GPa and a poisson’s ratio of 0.33. The interaction of the retaining wall with the surrounding soil was comprehensively considered using the joint elements to represent the separation and slippage effects arising during the cyclic loading. As specified earlier, in order to assess the effects of initial static shear bias, the constitutive model calibration was carried out separately for element tests performed under the application of static shear and for the element tests executed without any static shear bias. Figures 1 and 2 were obtained with similar parameters (shown in Table 1), but for further discussions, it was important to calibrate the parameters giving specific importance to the element tests with static shear. The bulk modulus and the shear modulus of the soil was changed to represent the relative density adopted in the centrifuge tests performed at various universities as a part of LEAP project. However, it is to be noted that the estimated responses are not compared with any of the measured centrifuge data, which is beyond the scope of this paper. Figure 3 shows the numerical model with a retaining wall, where the bottom sand layer is a dense compacted sand having a relative density of around 90%, while the backfill soil is relatively

Influence of Initial Static Shear Stress on the Dynamic Response …

191

loose (relative density of 66%) which was air pluviated in the centrifuge tests. The liquefaction and dilatancy parameters were applied for this layer, while the bottom layer was analyzed as non-liquefiable in numerical analysis. The numerical analysis was performed in two stages; initially, a self-weight analysis was conducted to establish the static gravitational loading within the model followed by dynamic analysis. The input motion was a specified motion for the LEAP project, which is a sinusoidal waveform having multiple cycles of loading at a constant excitation frequency of 1 Hz throughout the motion as shown in Fig. 4. Acceleration, soil displacement, stress–strain responses, excess pore water pressure responses, sheet-pile’s lateral displacement and rotation were computed from dynamic analysis. However, due to space limitations and considering the scope of this paper, the effects of the initial static shear stress (considered 15 kPa) were evaluated in terms of excess pore pressure response and the sheet-pile response during the cyclic loading. It is also important to ascertain the effects arising due to this mechanism near to the structure and at the free field. Keeping this in mind, Fig. 5 depicts the excess pore pressure responses at the free field for the two models with constitutive model being calibrated differently depending on the application of initial static shear stress or not (see nodes P4, P5 and P6 in Fig. 3). The numerical model,

Fig. 3 FEM meshing of the numerical model involving a retaining wall Fig. 4 Input motion considered for the dynamic analysis

192

A. Sahare et al.

which was calibrated without the application of initial static shear bias, shows large suction response within the excess pore pressure during the cyclic loading closer to the ground surface, with P4 and P5 depicting generation of negative response at some intervals within the cyclic loading. On the other hand, the sand toward the bottom of a backfill layer gets compressed leading to generation of identical positive excess pore pressure in both the numerical models. Figure 6 shows the comparison of the excess pore pressure between the two models near to the structure. One can clearly see the significant differences observed in Fig. 6 as compared to Fig. 5, with the numerical model without any prior static shear stress leading to predominant suction or dilative responses. Interestingly, this negative phase generation of excess pore pressure commences with the initiation of cyclic loading. This may represent the soil applying considerable amount of lateral resistance in the backfill to the sheetpile rotation toward seaside, which may also correspond to limited mobilization of passive resistance at the right side of the wall.

Fig. 5 Excess pore pressure responses at the free field for numerical models calibrated with and without the application of initial static shear stress

Influence of Initial Static Shear Stress on the Dynamic Response …

193

Fig. 6 Excess pore pressure responses near to the structure for numerical models calibrated with and without the application of initial static shear stress

On the other hand, when initial static shear stress is considered, the soil gets compressed along the depth of model with the generation of positive excess pore pressure. As compared to the free field, lesser excess pore pressure develops at P3 near to the sheet-pile at the same depth as compared to P6, which signifies soil being comparatively stiffer near to the structure. It is also important to note that soil achieves its peak excess pore pressure value under the application of initial 5 cycles of loading, post which the soil is found to oscillate while maintaining the similar peak excess pore pressure per cycle of loading till the end of shaking for the model with initial static shear stress. The consequences of this soil deformation mechanism during the cyclic loading can be assessed in terms of sheet-pile’s lateral displacement for the two models. Perhaps, due to prevalent dilative response captured near to the structure for a model without the consideration of initial static shear bias, a significantly smaller sheet-pile lateral displacement is found to be induced during the cyclic loading. On the other hand, since the soil gets compressed relatively for

194

A. Sahare et al.

Fig. 7 Sheet-pile lateral displacement during the cyclic loading

model with the consideration of initial static shear bias, a much larger deformation could be seen. It is also important to note that the different deformation mechanism for the two models, with numerical model having the consideration of initial static shear bias, manifests a progressive deformation per cycle of loading, leading to a maximum value at the end of shaking. On the other hand, sheet-pile is found to attain the peak displacement value at around 20 secs, post which the backfill soil resistance becomes very large and thereby not allowing any further accumulation of displacement. This can be idealized with large negative excess pore pressure values at the same time instance (Fig. 7).

5 Conclusion The following conclusions can be drawn, • The initial element level calibration of the constitutive model revealed the capabilities of the strain space multiple mechanism model to simulate the cyclic soil response under different conditions. Without the application of static shear bias, the model performed very well. However, when the initial static shear stresses are considered, the model manifested a slightly different response, although important soil features were reasonably captured. • The system level performance indicated the importance of the consideration of initial static shear bias, with models calibrated without the consideration of static shear stress, led to a predominant dilative response from the outset of shaking. On the other hand, numerical models calibrated under the initially applied static stresses indicated the soil’s compressibility under the applied cyclic loading. These two different mechanisms are found to have significances on the sheet-pile lateral deformation during the cyclic loading, which is the most important consideration in the seismic-related design of any retaining structure.

Influence of Initial Static Shear Stress on the Dynamic Response …

195

References 1. Iai S, Urakami T, Mutoh Y, Kikuchi M (1998) Liquefaction at coastal area and performance of a quaywall during 1989 Off East Chiba prefecture earthquake. Technical Note of Port and Harbour Research Institute. No. 616 (in Japanese, 1988) 2. Tsuchida H (1990) Japanese experience with seismic design and response of port and harbor structures. In: Proceedings of POLA Seismic Workshop on Seismic Engineering, Port of Los Angeles, pp 138–164 (1990) 3. Whitman RV, Christian JT (1990) Seismic response of retaining structures. Port of Los Angeles, pp 427–452 4. Iai S (1998) Rigid and flexible retaining walls during Kobe Earthquake. In: International Conference on case histories in geotechnical engineering, Missouri 5. Sahare A, Tanaka Y, Ueda K (2020) Numerical study on the effects of rotation radius of geotechnical centrifuge on the dynamic behavior of liquefiable sloping ground. Soil Dyn Earthq Eng 138:106339 6. Vargas RR, Ueda K, Tobita T (2020) Centrifuge modeling of the dynamic response of a sloping ground-LEAP-UCD-2017 and LEAP-ASIA-2019 tests at Kyoto University. Soil Dyn Earthq Eng 140:106472 7. Tobita T, Ueda K, Vargas RR, Ichii K, Okamura M, Sjafruddin AN et al (2022) LEAP-ASIA2019: Validation of centrifuge experiments and the generalized scaling law on liquefactioninduced lateral spreading. Soil Dyn Earthq Eng 157:107237 8. Iai S, Tobita T, Ozutsumi O, Ueda K (2011) Dilatancy of granular materials in a strain space multiple mechanism model. Int J Numer Anal Methods Geomech 35:360–392 9. Ueda K, Iai S (2021) Noncoaxiality considering inherent anisotropy under various loading paths in a strain space multiple mechanism model for granular materials. Int J Numer Anal Methods Geomech 45(6):815–842 10. ElGhoraiby MA, Manzari M (2021) Cyclic behavior of sand under non-uniform shear stress waves. Soil Dyn Earthq Eng 143:106590

Mechanical Behaviors of MICP Treated Binary Granular Materials Under One-Dimensional Consolidation Yu-Syuan Jhuo, Yi-Qian Lu, Chang-Ping Yu, and Louis Ge

Abstract The mechanical behaviors of binary granular materials are significantly different due to the amount of their fines content, which is normally categorized as coarse-grains dominant or fine-grains dominant. This study attempts to apply microbial technology to improve the strength and stiffness of binary granular materials. The technique uses microbial urea hydrolysis to induce calcite, which is also called microbiological-induced calcite precipitation (MICP). A series of one-dimensional consolidation tests were carried out to examine the compressive behavior of the materials before and after MICP treatment. The fines content evaluated in this study included 0, 15, and 30, while the relative density is 40%, and the grain size ratio of coarse grains to fine grains is 12. The single-phase treatment was introduced to minimize the sand being disturbed during the MICP process, where natural percolation from the bottom of the specimen was applied. The curing time of 3, 7 days was selected and each specimen in the consolidation ring was inundated with water for 24 h before the test. This study attempted to improve the binary mixtures with the MICP technique, to investigate the effect of MICP on the one-dimensional consolidation results. The results were the opposite of what was predicted. Keywords Binary packing material · Microbiological induced calcite precipitation (MICP) · Compression · Fines content

1 Introduction In the vast soil of the earth, the soil consists of a mixture of many complex coarse and fine grains, and the use of binary granular materials helps simplify the in-situ condition and problems. It is composed of two kinds of grains and its mechanical behaviors are significantly different due to the amount of their fines content which is normally categorized as coarse-grains dominant or fine-grains dominant [1]. The maximum and minimum void ratio decreases to a valley with the increase of fine Y.-S. Jhuo (B) · Y.-Q. Lu · C.-P. Yu · L. Ge National Taiwan University, Taipei, Taiwan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_18

197

198

Y.-S. Jhuo et al.

grain content, where this transition zone falls between 20 and 40%, depending upon material characteristics and particle size distribution. After the transition zone, the void ratio increases with the increase of fine grains [2]. The population of the Taiwan metropolitan area has gradually increased and reached a saturated state, where many buildings and infrastructure are constructed at sites with poor geological conditions. Taiwan is frequently struck by earthquakes, which makes the technology of ground improvement extremely important. Microbiologically induced calcium carbonate cementation (MICP) as the functions of pore filling and particle capturing have begun with the process of sediment rock formation. In recent years, with the rising awareness of environmental protection, the application of this function to issues related to geotechnical engineering has been regarded as a sustainable method [3, 4]. The principle of MICP is to use urease to decompose urea into ammonia and carbon dioxide, the reaction is as Eq. 1. An alkaline substance, such as ammonia, raises the pH value of proximal environments and promotes the formation of calcium carbonate by combining calcium ions with carbonate ions (Eq. 2). 2− − CO(NH2 )2 + 3H2 O → 2NH+ 4 + 2OH + CO3

(1)

− Ca2+ + HCO2− 3 + OH → CaCO3 ↓ +H2 O

(2)

The use of MICP can make the soil particles cemented to provide better mechanical properties while maintaining the original non-disturbed characteristics of the soil [5]. In this study, binary granular materials were treated by the technique of MICP to simulate the complex behavior of soil improvement. The MICP technique may effectively improve the compressibility of soil against natural disasters, where the resulting void ratio changes and the relevant sand compression parameters can be obtained using a one-dimensional compaction test.

2 Materials and Methods 2.1 Coarse Grains and Fines The coarse-grained sand used in this study was Ottawa sand (ASTM C-109). The specific gravity of the Ottawa sand is 2.657. The coefficient of uniformity and coefficient of curvature are 1.462 and 0.958, respectively, which is classified as well-graded sand (SW). The fine-grained sand was purchased from Chin-Ching Silica Sand Co., Ltd., which is Vietnam silica sand with a sub-angular shape. The specific gravity is the same as the coarse grain. A binary mixture was formulated using designated weights of coarse grains and fines. The sand passed through a 0.840 mm (No. 30) sieve and retained on a 0.590 mm (No. 30) sieve were selected as coarse grains, which were mixed with the fines that passed through a 0.088 mm (No. 170) sieve

Mechanical Behaviors of MICP Treated Binary Granular Materials …

199

and retained on a 0.075 mm (No. 200) sieve. The particle distribution of coarse grain and fines is presented in Fig. 1. This binary mixture has a particle ratio of 8.77. The fines content for one-dimensional consolidation was 0, 15, and 30%. The plot of emax and emin versus fines content is shown in Fig. 2. To observe the volume change of the specimen under various normal stress, the target relative density of each specimen for the consolidation test was set at 40%. 1

Ottawa sand

Percent Finer by Weight (%)

0.9

Vietnamese quartz sand

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 100

10

1

0.1

0.01

0.001

Diameter, d (mm) - log scale

Fig. 1 Particle size distribution curve of two kinds of particles a Ottawa sand b Vietnamese quartz sand 0.8 0.7

Void ratio

0.6 0.5 0.4 0.3 0.2

e emax emin

0.1 0 0

15

Fines content(%)

Fig. 2 Void ratio at a relative density of 0, 15, and 30%

30

200

Y.-S. Jhuo et al.

2.2 Bacteria Bacillus pasteurii, commonly found in soil and water in nature, can produce a large amount of intracellular urease. It can catalyze the hydrolysis of urea, produce ammonia ions and carbonate ions, and precipitate calcium carbonate in an alkaline environment. The Bacillus pasteurii used in the study was purified from the wastewater samples collected from the Dihwa Sewage Treatment Plant in the lab of the graduate institute of environmental engineering, NTU [6]. The ingredients of a culture medium for cultivating bacteria are beef extract 3 g, peptone 5 g, and urea 5% in 1 L of liquid medium. This is followed by adding 15 g/L of agar to make a solid medium.

2.3 Preparation of Treatment Solution The treatment solution contains bacterial solution, urea, and calcium source. This study used a single-phase treatment solution for treating the specimens, which means that mix three ingredients at once before injecting them into the specimens. The pH value of the bacteria solution is usually around 9.45–9.50. For single-phase injection, calcium carbonate would be formed instantly, when bacterial solution mixes with a calcium source, and affect the uniformity of treated specimens. The brand name of the used urea is BioShop Canada Inc., and its molar mass is 60.06 g·mol−1 . This study chose calcium nitrite as a calcium source. The chemical was purchased from Emperor Chemical Co., Ltd., and the molar mass is 164.00 g·mol−1 . The proportion of urea and calcium nitrate was 1 to 1, and a concentration of 1 M was applied.

2.4 Testing Method Specimen Preparation. To allow the sand specimen to be tested evenly to absorb the MICP solution in the specimen ring, a plastic petri dish made of Polystyrene (PS) (Fig. 3a) was prepared in advance, and a plastic sheet of Polyvinyl Chloride (PVC) (Fig. 3b) was used according to the size of the petri dish to create a plastic chassis (Fig. 3c) in the petri dish of facilitates the transfer of the specimen from the petri dish and avoids the disturbance of the specimen in the specimen ring. According to each specimen ring, the upper filter paper and lower filter paper are numbered and weighed (Fig. 4a). The size of the specimen ring is about 5 cm in diameter and about 1.9 cm in height, and the initial relative density is controlled at 40%. The volume and weight of the sand specimen that needs to be filled in the specimen ring use the Polystyrene (PS) plastic petri dish prepared in advance, put the plastic sheet made of Polyvinyl Chloride (PVC), and put the corresponding

Mechanical Behaviors of MICP Treated Binary Granular Materials …

201

Fig. 3 a plastic petri dish made of polystyrene (PS); b plastic sheet of polyvinyl chloride (PVC); c plastic chassis

number of lower filter paper (Fig. 4b) under it. The specimen ring is filled with the corresponding required weight and the sand pattern is filled slightly above the height of the specimen ring by 0.1–0.3 cm using the extension ring, taking into account that when the sand specimen is inhaled into the MICP, the sand specimen will drop in height due to liquid inhalation. Pour the MICP liquid into the plastic petri dish, and let it stand for some time, wait for the sand specimen to be equal to the specimen ring (Fig. 4c), remove the extension ring, mark the curing time, and test time on the plastic petri dish, and finally put it into the sealed box (Fig. 4d) and wait for the curing time. The Specimen on the Consolidometer. The cultured specimen is transferred from the plastic petri dish (Fig. 5a), pulled vertically using a plastic chassis of Polyvinyl Chloride (PVC), and transfer out, and to prevent the specimen from being disturbed,

Fig. 4 a specimen ring, upper filter paper, and lower filter paper; b put the corresponding number of lower filter paper; c pour the MICP liquid into the plastic petri dish; d put the specimen into the crisper box

202

Y.-S. Jhuo et al.

Fig. 5 Move the specimen from the petri dish: a pull up vertically to transfer out; b and c flip 180° and slide out parallel to the plastic chassis

turn it 180 degrees and slide it out parallel to the plastic chassis to complete the transfer to the step (Fig. 5b, c). Put the specimen into the consolidometer, add water, and place it for 24 h before starting the test. One-Dimensional Consolidation Test. The pressure is applied to the specimen by the static load through the lever arm principle, the lever ratio is 9.82, and loading, unloading, and reloading, the axial stress accumulated in the loading stage are 2.467, 7.402, 19.738, 44.410, 69.083, 118.428, 217.118 kPa, a total of 7 steps. After unloading all the iron blocks one by one, and then adding the iron blocks back one by one, the two orders of iron blocks are applied to 321.719 and 522.793 kPa, and finally all are unloaded one by one. In this test, the Humboldt HM-2432.3F recorder is used to record the time and vertical displacement of each stage of the iron block. The material used in this experiment is no cohesion, the saturation is 1, and the volume of the specimen changes rapidly after being compressed. The recording time can be set to 0.1 s, and the displacement can be accurately recorded to 0.001 mm, which can effectively record the changes in the specimen during the compression process.

3 Results and Discussions This study attempted to investigate the behavior of consolidation of binary mixtures which treated with MICP. The fines content was 0, 15, and 30%, and the curing time was 3, 7, and 14 days.

3.1 Normalized Void Ratio Versus Stress The loading, unloading, and reloading curves on the plot of normalized void ratio versus pressure in log scale were presented in Fig. 6, with the legend simplified as follows: calcium nitrite was abbreviated as N, the cultivation duration for MICP was simplified as 3D, 7D, and 14D representing three, seven, and fourteen days, respectively. The content of fine particles was simplified as FC0, FC15, and FC30 representing fine particle contents of 0%, 15%, and 30%, respectively. The result

Mechanical Behaviors of MICP Treated Binary Granular Materials …

203

of 0% fines content is shown in Fig. 6a. In the case of fines content of 0%, the untreated specimen has less change in the normalized void ratio during the test. For MICP-treated specimens, the changes in normalized e are significant at the loading stage. 14 days of curing resulted in the greatest change, while 3 days resulted in the least change. For fines content of 15%, a similar trend to that observed for 0% fines content was observed in Fig. 6b. However, the trend was different for the MICPtreated specimen with 30% fines content in Fig. 6c. Specifically, the greatest change in normalized void ratio occurred after 7 days of curing, while the least change occurred after 14 days of curing.

3.2 Calcium Carbonate Content Calcium carbonate content was obtained through acid digestion, which used 4 M Hydrochloric Acid (HCl) to decompose the calcium carbonate formed in the treated specimen into calcium ions and carbonate ions, since the treatment solution was injected through the capillary in this study. A difference in weight between before and after acid washing indicates calcium carbonate content. The results of the calcium carbonate content analysis are presented in Fig. 7, which shows that the calcium carbonate content is decreasing with fines content. The higher the calcium carbonate content, the greater the change in normalized void ratio, and the higher the void ratio after reloading for the 15 and 30% fines content.

3.3 Compression and Swelling Indexes The results shown in Fig. 8 indicate that the distribution of compression index (Cc) was slightly different than what was initially expected. It was anticipated that an improvement in microbial-induced calcium carbonate precipitation (MICP) would lead to a decrease in Cc. However, the obtained result was contrary to the expectation, showing a larger change in Cc compared to the untreated specimen. This result was obtained by averaging multiple sets of experimental data, which enhances its reliability and robustness. In addition, we observe that the trend of Cc increases with the increase of fine particle content, which is consistent with the observation that “the values of Cc increase and attain the peak value at approximately FC 25%” [7]. The distribution map of swell index (Cs) also indicates a certain regularity, with the 7-day swelling index being less than the 14-day swelling index, and distributed points being very close.

204

Y.-S. Jhuo et al.

(a)

1

Nornalized Void ratio

0.95

0.9

0.85

0.8

FC0 N3DFC0

0.75

N7DFC0 N14DFC0

0.7 1

(b)

10

100

1000

1

Nornalized Void ratio

0.95

0.9

0.85

0.8 FC15 N3DFC15

0.75

N7DFC15 N14DFC15

0.7 1

(c)

10

100

1000

1

Nornalized Void ratio

0.95

0.9

0.85

0.8 FC30 N3DFC30

0.75

N7DFC30 N14DFC30

0.7 1

10

100

1000

Log P (kPa)

Fig. 6 Plot of normalized e versus log p showing loading, unloading, and reloading curves: a FC0; b FC 15; c FC30

Calcium carbonate (g) in 10 g treated sand

Mechanical Behaviors of MICP Treated Binary Granular Materials …

205 N3D

0.08

N7D

0.07

N14D

0.06 0.05 0.04 0.03 0.02 0.01 0 0

15 Fines content (%)

30

Fig. 7 Calcium carbonate (g) in 10 ml solution 0.015

0.055 0.05 0.045 0.04

0.01

0.03

Cs

Cc

0.035

0.025 0.02

NO MICP N3D

0.005

0.015

N7D

0.01 0.005

N14D

0 0

15

Fines content (%)

30

0 0

15

30

Fines content (%)

Fig. 8 Cc and Cs distribution graph

4 Conclusion There is a characteristic of binary mixtures that the void ratio changes with fines content under a given packing condition (relative density). The particles’ assembly is unstable in the transition zone. This study attempted to modify the binary mixtures with the MICP technique, through investigating the effect of MICP from the onedimensional consolidation tests. The results were the opposite of what was predicted. The specimen’s volume changed significantly during consolidation after MICP treatment at a shorter curing time. After a certain curing time, the volume change is smaller, especially at 30% fines content. Following the same method of modifying the binary mixture specimen, the calcium carbonate content is determined by the fines content. However, since the treatment solution was injected through the capillary in this study, the amount of treatment solution left in the specimen was limited. This results in less calcium carbonate formation.

206

Y.-S. Jhuo et al.

In conclusion, the unexpected volume changes observed during consolidation after MICP treatment could be attributed to several factors. Firstly, the limited distribution of the treatment solution through the capillary may have resulted in less calcium carbonate formation, which contributed to the formation of voids and increased porosity, leading to volume changes during consolidation. Secondly, the formation of calcium carbonate may have altered the surface properties of sand particles, affecting the particle–particle interactions and the consolidation behavior of the mixture. Despite these factors, further investigation and analysis would be necessary to confirm the exact cause of the unexpected volume changes observed during consolidation after MICP treatment.

References 1. Ng T-T, Zhou W, Ma G (2022) Numerical study of a binary mixture of similar ellipsoids of various particle shapes and fines contents. Int J Geomech 22(4):14 2. Pillitteri S, Opsomer E, Lumay G, Vandewalle N (2020) How to size ratio and segregation affect the packing of binary granular mixtures. Soft Matter 16:9094–9100 3. DeJong JT, Mortensen BM, Martinez BC, Nelson DC (2010) Bio-mediated soil improvement. Ecol Eng 36:197–210 4. Khodadadi TH, Kavazanjian E, Paassen L, DeJong J (2017) Bio-grout materials: a review. Geotech Spec Publ (GSP) 288:1–12 5. Xu Y-R, Xu Y, Wang A-X (2022) One-dimensional compression characteristics of uniformly graded sand under high stresses. Granular Matter 24:60 6. Shih CC (2022) Evaluating the application of simple 3D bioprinting in microbial immobilization: a case study of ureolytic bacteria and heavy metal removal. Master thesis 7. Jhuo Y-S, Yeh Y-H, Ge L (2020) Shear strength and volume change behavior of binary granular mixtures. J GeoEng 15(2):103–108

Numerical Simulations for Seismic Response of Laterally Loaded Pile Foundation Tanmoy Barik and Babloo Chaudhary

Abstract Pile foundations are frequently used to safely transfer enormous loads from superstructures to deeper soil layers in order to support structural weight. Due to earthquakes, piles are subjected to dynamic loadings, which causes significant damage to the foundation as well as the superstructure. Laterally loaded piles vulnerable to earthquakes have a complex dynamic reaction. Earthquake motions shake the foundation ground which can make it unstable. Due to this unpredictable behaviour, the research has been carried out for the last few decades on it but still the behaviour is not understood completely. Therefore, an attempt has been made in this study to investigate the dynamic response of the pile foundation due to earthquakes by performing numerical study using Finite Element (FE) program, PLAXIS 3D. In addition to it, several parameters have been studied to understand the effects of several parameters like soil properties, earthquake features (acceleration, frequency etc.) on pile behaviour, considering soil-pile interaction. It can be stated that this numerical simulation will be useful for the researchers and practicing engineers working on this domain. Keywords Pile foundation · Dynamic response · Earthquake load · Finite element analysis · PLAXIS 3D

1 Introduction Pile foundations are often used to support large superstructure loads in places such as power plants, petrochemical complexes, refineries and compressor stations. In addition to static loads, piles are also subjected to dynamic lateral loads from operating machinery, wind movements and earthquakes. Therefore, the dynamic response of lateral loads has attracted the attention of researchers and design consultants [1]. Till now, various methods have been developed for the analysis of laterally loaded piles (both single and groups) subjected to dynamic loadings, using the results obtained T. Barik (B) · B. Chaudhary National Institute of Technology Karnataka, Surathkal, Karnataka, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_19

207

208

T. Barik and B. Chaudhary

from shaking table and centrifuge tests, as well as by means of numerical analysis [2–4]. Amongst them, model test results of Boominathan and Ayothiraman [5] represented a comparison of bending moment response of single and group piles under static and dynamic loading conditions. Muthukkumaran and Subha [6] performed numerical analysis to understand the response of single pile under seismic condition. It was concluded from the study that for a constant slope and depth of liquefiable soil layer, the behaviour of the increment rate of lateral displacement and bending moment of the pile is inversely proportional with the L/D ratio of pile [6]. For pile group, the maximum bending moment as well as the maximum displacement were encountered larger in magnitude at middle and back rows as compared to the front rows in level ground and sloping ground surfaces as represented by the literature of Deendayal et al. [7]. The study of Dong et al. [8] based on seismic response of laterally loaded pile deduced that the effect of seismic intensity on the displacement of soil is higher as compared to the effect of lateral soil pressure. The present study aims to investigate the effects of earthquakes on the behaviour of pile foundation under lateral loads. Finite Element Method (FEM)-based software PLAXIS 3D has been used for the numerical analyses.

2 Description of the Experiment Used for Validation For validation of the numerical model, study performed by Hussein and Albusoda [9] was used. Hussein and Albusoda [9] performed shake table test to analyse the dynamic responses of pile foundation. The equipment used in this experiment consists of a laminar box of 800 × 800 mm dimension with maximum acceleration 3 g and payload 1000 kg with a total shaking table self-weight is equal to 300 kg. Natural frequency varies from 0 to 100 Hz. A laminar shear box that was constructed from 12 square steel laminates has been used in this study to hold the tested soil [9]. Karbala sand is predominantly prevailed in the soil profile, considered for the testing purpose. The air-dried sand sample (Relative densities of 30 and 70%) was screened through sieve no. 10 [9]. Embedded depth of the model aluminum hollow circular pile was 400 mm with an outer and inner diameter of 16 mm and 13 mm, respectively. The length to diameter ratio considered for the test was equal to 25. A combined static vertical and horizontal load (Self-weight of pile cap: 65 N and lateral loads acting: 7 N) was applied on the pile. Kobe earthquake (1995) motion was used in this study for understanding the responses of pile under the action of strong earthquake motion [9]. Figure 1 represents the soil-pile system that has been employed for the test.

Numerical Simulations for Seismic Response of Laterally Loaded Pile …

209

Fig. 1 Schematic diagram of the soil-pile system

3 Numerical Modelling of Soil Pile System FEM-based software PLAXIS 3D [10] was used to simulate the numerical analyses in this study. Major factors that play important role in FEM-based numerical modelling are- finalizing the soil constitutive model and modelling of the pile considering soilpile interaction. Based on the field conditions and applicability, above-mentioned factors were decided during the numerical modelling.

3.1 Soil Properties Foundation ground consisted of two layers of soils; i.e., dense sand and loose sand. Water table was present at great depth. The soil properties, used for this study, are taken from the literature by Hussein and Albusoda [9] and are shown in Table 1. Mohr–Coulomb constitutive model was used in the present analyses.

3.2 Pile Properties In this study, the pile is modelled using the embedded beam element. The diameter and thickness of the model aluminum pile were 16 mm and 1.5 mm, respectively. The soil-pile interaction was defined through the embedded beam elements itself, since it

210

T. Barik and B. Chaudhary

Table 1 Soil properties used in the numerical analysis (after Hussein and Albusoda [9]) Soil

Material properties

Loose sand

Dense sand

Drainage type

M-C model

M-C model

Depth of soil layer (m)

0–0.32

0.32–0.75

17

18

Bulk unit weight (γ ) (kN/m3 ) (kN/m3 )

18

19.5

Relative density (%)

30

70

Modulus of elasticity (E ref ) (kN/m2 )

10,000

10,000

Saturated unit weight (γ sat )

Poisson’s ratio (vnu )

0.33

0.33

Cohesion (C  ref) (kN/m2 )

1

1

Angle of internal friction (ϕ  ) (°)

32

35

Dilatancy angle (°)

2

2

Table 2 Pile Properties used in the numerical analysis (after Hussein and Albusoda [9])

Pile properties

Values

Material type

Elastic

Modulus of elasticity (kN/m2 )

67 × 106

Unit weight

(kN/m3 )

27

Beam type

Predefined (circular tube)

Dia (m)

0.016

Thickness (m)

0.0015

Axial skin resistance (T max ) (kN/m) 1.0 × 1012 Base resistance (F max ) (kN)

0.06

has the inbuilt feature of interaction properties. Axial skin resistance of the pile was considered depth dependent for the analysis. Strength reduction factor (Rinter ) was defined as 0.5 for the current analysis. The pile properties that have been considered in the present study are mentioned in the Table 2.

3.3 Earthquake Motion Time history of Kobe Earthquake (1995) [11] was used as earthquake input motion in this study. The main earthquake data is mentioned in Table 3.

Numerical Simulations for Seismic Response of Laterally Loaded Pile … Table 3 Earthquake Data (Kobe, 1995)

211

Earthquake parameters

Values

Magnitude (Mw)

6.9

Peak ground acceleration (PGA) (g)

0.82

Duration of shaking (s)

48

Epicenter depth (km)

17.9

3.4 Boundary Conditions Absorbent boundary was used in dynamic stage to resist the inflow of reflected waves of earthquake. In this type of boundary, viscous dampers are automatically applied in the boundary by PLAXIS itself.

3.5 Mesh Generation A FE Mesh as shown in the Fig. 2 has been generated in PLAXIS 3D. The characteristics of a finite element mesh are defined by its coarseness. In this study, medium coarseness has been considered as the percentage convergence for medium meshing was found to be greater as compared to coarser, finer or very finer meshing. Coarser and finer meshing has underestimated and overestimated the result, respectively, in the present analyses. After generating the mesh, FE analysis was executed. Earthquake loads were applied using the ‘prescribed displacement’ option. Dynamic analysis mode was used as the calculation type to simulate the earthquake-induced load. In the earthquake analysis phase, displacements of previous stages were reset to zero to analyse the effects of earthquake individually. Fig. 2 Generated mesh of the soil-pile system

212

T. Barik and B. Chaudhary

4 Results and Discussion In this present study, a 3D numerical analysis was carried out to understand the behaviour of pile foundation under the action of strong earthquake motion. FEMbased software PLAXIS 3D was used to develop the numerical model. Bending moment response that was found from the simulation result was validated with the results given by Hussein and Albusoda [9]. It can be observed from the Fig. 3 that the results from the present study and the experimental results of Hussein and Albusoda [9] are matching with a good accuracy. Hence, the numerical model developed for the present study is successfully validated and can be used for further analyses and parametric studies.

4.1 Effects of L/D Ratio Bending moments (BM) were determined for different values of L/D ratios. It can be seen in Fig. 4 that maximum positive BM increased from 5.1 N-m to 11.8 N-m and 13.3 N-m, whilst L/D ratio was increased from 25 to 30 and 40, respectively. The maximum negative BM was increased from 4.3 N-m to 7.9 N-m and 10.1 N-m, whilst L/D ratio was increased from 25 to 30 and 40, respectively. It can be inferred from the results that as L/D ratio was increased, consequently, length also got increased and as BM always shows proportional behaviour with the length, as a result, peak bending moment was also increased. It can also be stated that with the increase of pile length, stiffness of the pile was reduced and thereby peak bending moment was increased.

0 6

4

2

0

-2

-4

-6

Length of the Pile (mm)

-100 -200 -300 -400 -500

PLAXIS 3D

-600

Hussein & Albusoda

-700 Bending Moment (N-m)

Fig. 3 Validation plot (pile length vs. bending moment)

Numerical Simulations for Seismic Response of Laterally Loaded Pile …

213

0 15

10

0

-5

-5 5

-10

-15

-10

L/D Ratio

-15 -20 L/D = 25

-25

L/D = 30

-30 -35

L/D = 40

-40 Bending Moment (N-m)

Fig. 4 Variation of bending moment with L/D ratio

4.2 Effects of Peak Earthquake Acceleration In order to see the effects of seismic acceleration on the behaviour of the pile, BMs were determined for different values of input peak accelerations. It can be observed from Fig. 5 that peak positive BM decreased from 5.1 N-m to 3.4 N-m and 2.3 N-m, whilst peak acceleration was decreased from 0.82 g to o.065 g and 0.55 g, respectively. Similarly, maximum negative BM was decreased from 10.1 N-m to 7.9 N-m and 4.3 N-m, whilst peak acceleration was increased from 0.82 g to 0.65 g and 0.55 g, respectively. It can be concluded from the figure that as the intensity of the earthquake decreased, the load induced by earthquake motion was also reduced, which led the pile to bend less as compared to the former case.

214

T. Barik and B. Chaudhary

Fig. 5 Variation of bending moment with pile length for varying earthquake acceleration

5 Conclusion In this present study, a FEM-based numerical analysis was performed for better understanding of the behaviour of the pile foundation under the action of earthquake. The developed numerical model was well validated with the literature by Hussein and Albusoda [9]. Furthermore, the effects of L/D ratio and earthquake input acceleration on the BM response of pile were studied and the following conclusions were made. • Whilst increasing the L/D ratio, it was observed that when L/D ratio is increased from 25 to 30, positive BM was increased by 131% and when L/D was increased from 30 to 40, positive BM was increased by 14%. Hence, it can be concluded that after a certain limit of increase of L/D ratio, the BM shows negligible reduction in magnitude. • Whilst decreasing earthquake peak acceleration value, it was observed that when acceleration was decreased from 0.82 to 0.65 g, positive BM was decreased by 33.33% and whilst acceleration was decreased from 0.65 to 0.55 g, positive BM was decreased by 32.3% in magnitude. It can be concluded that this reduction of BM was occurring due to the reduction of seismic intensity. Adjacent soil layer intends to deform less with the decrease of the peak acceleration, thereby results in a lesser magnitude of BM.

References 1. Boominathan A, Ayothiraman R (2006) Dynamic response of laterally loaded piles in clay. Proc Inst Civ Eng Geotech Eng 159:233–241

Numerical Simulations for Seismic Response of Laterally Loaded Pile …

215

2. Albusoda BS, Salem AK (2016) The effect of interaction on pile-raft system settlement subjected to earthquake excitation. Appl Res J 2(4):205–214 3. AL-Recaby MK (2016) Dynamic response to lateral excitation of pile group model in sandy soil. Doctoral dissertation, Ph.D. thesis, University of Technology 4. Elewi AS (2017) Response of single pile and pile groups to lateral soil movement. Doctoral dissertation, Ph.D. thesis, University of Technology 5. Boominathan A, Ayothiraman R (2007) An experimental study on static and dynamic bending behaviour of piles in soft clay. Geotech Geol Eng 25:177–189 6. Muthukkumaran K, Subha IP (2010) Effect of earthquake induced lateral soil movement on piles in a sloping ground. In: Indian Geotechnical conference—2010, GEOtrendz, 16–18 Dec 2010, IGS Mumbai Chapter & IIT Bombay 7. Deendayal R, Muthukkumaran K, Sitharam TG (2015) Response of laterally loaded pile groups in sloping ground. In: Proceeding of Indian geotechnical conference 8. Dong Y, Feng Z, He J, Chen H, Jiang G, Yin H (2019) Seismic response of a bridge pile foundation during a shaking table test, shock and vibration, vol 2019, 16p 9. Hussein RS, Albusoda BS (2021) Experimental and numerical analysis of laterally loaded pile subjected to earthquake loading. Modern applications of geotechnical engineering and construction. Springer, Lecture Notes in Civil Engineering 112 10. PLAXIS 3D Reference manual, Tutorial, Material models manual and Scientific manual, Bentley Communities 11. Time history of Kobe Earthquake (1995) Strong-motion virtual data centre

Response of Offshore Wind Turbine Foundation Subjected to Earthquakes, Sea Waves and Wind Waves: Numerical Simulations Subodh Kumar, Babloo Chaudhary, Manu K. Sajan, and P. K. Akarsh

Abstract Offshore wind turbines are an economical and sustainable method for generating renewable energy over extended periods. They efficiently harness wind power and are strategically located far from residential areas in the sea, resulting in minimal noise pollution. These towering structures rely on wind as their primary energy source and are installed at varying water heights from shallow to medium depths. The critical aspect of ensuring the stability of the foundation for such massive and tall structures becomes particularly important, especially in regions prone to earthquakes. This research paper focuses into the influence of wind loading on offshore wind turbine platforms, with specific emphasis on the suction caisson foundation. To assess the effects of wind loads, numerical analyses were performed using the finite element software PLAXIS. The findings reveal that horizontal deflection and shear stress increase as the angle of internal friction and unit weight decrease. Additionally, the study conducts parametric analyses to explore the impact of other variables on the behaviour of the turbine. These conclusions emphasize the significance of designing resilient foundations for offshore wind turbines, considering factors such as wind loads, soil characteristics, and structural parameters. This ensures their long-term stability and effectiveness as a sustainable source of energy. Keywords Suction caisson · Wind loadings · Offshore wind turbine · Stability · Renewable energy

S. Kumar (B) · B. Chaudhary · M. K. Sajan · P. K. Akarsh National Institute of Technology Karnataka, Surathkal, Karnataka, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_20

217

218

S. Kumar et al.

1 Introduction In modern era, energy requirement is increasing day by day and to meet the demand, the non-renewable resources are not sufficient. Non-renewable sources are depleting very fast and are insufficient to meet the requirement at these days. To meet such huge energy goal, utilization of wind energy is playing a key role. Offshore wind turbine is a good option for wind energy generation, due to uninterrupted and good speed of wind. Amongst the renewal energy, wind energy is one of the best options. The harvesting of energy from wind in some countries has even surpassed the hydroelectric power generation. For using wind energy efficiently, we need a very high and large tower-like structure called wind turbine. India ranks fifth globally in terms of installed capacity with 16.1 GW, and it stands as the second-largest in Asia after China [1, 2]. In April 2012, an Indian wind energy expert highlighted that due to the growing size of wind power equipment and the emergence of offshore wind turbines, the nation possesses an untapped potential of 3,000 GW [3]. The wind power turbines can be installed in onshore and offshore regions. In offshore environment, wind speed is strong and less variable when comparing to onshore conditions. Therefore, in offshore area, we can install large turbines to get much greater power output and the biggest advantage of this is no use of land, so that we can go up to a larger area for turbines installation, as we know in developing countries, we have very less land area available for such requirement. The main challenges for installation of wind turbines are foundation parts, which have to be given at different depth inside sea water and different soil conditions (types). In offshore environment, foundation has to face many phenomena like cyclic loading from wind and sea waves [4–13], which are not yet studied sufficiently and a lot of research is required yet. Currently, gravity-based shallow foundations or some combination of one or more driven piles are being used in practice. The foundation must be engineered to withstand substantial overturning forces and endure million number of cycles of lateral loads whilst retaining its structural integrity over its designated 25-year lifespan. Various foundation types are employed to take the loads coming from offshore wind turbines, including monopod foundation, such as gravity bases and caissons, as well as robust single piles referred to as monopiles. Additionally, multi-pile carrying structures like tripods and jackets are utilized in this context [14–16]. The choice of a specific foundation type typically relies on the unique soil conditions at the wind farm site and other factors, such as the availability of suitable installation and transportation equipment. Additionally, the ease of mass production and storage also holds a vital role in determining the most suitable foundation type for a project.

Response of Offshore Wind Turbine Foundation Subjected …

219

Fig. 1 Details of models developed in PLAXIS 3D

2 Methodology The Finite Element Method (FEM) serves as a numerical technique for solving differential or integral equations and finds applications in numerous physical problems governed by such equations. Essentially, this method involves the assumption of piecewise continuous functions in a manner that minimizes errors within the solution. Here, we have used PLAXIS 3D software [17] for our numerical analysis. A soil model was modelled by taking a borehole and defining the soil depth up to 3 times the length (3L), whilst the model’s length was set at 6 times of the diameter (6D). The width taken of model was established as 3 times of the diameter (3D), as illustrated in Fig. 1. Numerical modelling was conducted using a typical bucket foundation configuration placed in dense sand. A vertical dead load (V) of 10,000 kN, a standard load for a 5 MW offshore wind turbine, was applied at the centre of the bucket lid. The dimensions included a diameter (D) of 14 m and a skirt length (L) of 10.5 m, representing an embedment ratio of L/D = 0.75 [18] (Fig. 2). In the case of caisson foundations, high rate of application of loading saturated soil is known to exhibit partially undrained behaviour. However, for the purposes of this analysis, it was understood that the soil behaves in a drained manner, with hydrostatic pressure dissipating rapidly after the application of load. The caisson foundation itself was represented as a plate element in form of steel in the modelling process, with a skirt thickness of 30 mm, a common practice in bucket design. The steel properties used in the model included an elastic modulus (E) of 210 GPa, which have Poisson’s ratio of 0.3, and a unit weight of 77 kN/m3 . To account for the bucket’s rigid behaviour, due to stiffeners on the bucket lid, a caisson lid with a thickness of 100 mm and a very high number of elastic modulus (E = 210 × 106 GPa) was incorporated. The analyses done numerically were conducted in multiple calculation phases, starting with the initial phase that considered the initial modelled structure and state of stress. Geostatic stresses have been computed in initial phase by applying static loads to the soil elements, where earth pressure coefficient at rest is given by k o = 1 – sin ϕ [11]. Parallelly, caisson and tower were also modelled as

220

S. Kumar et al.

Fig. 2 Validated numerical model

considering plate material made of steel. Soil-structure interaction was also modelled after taking interface concept along the perimeters of the caisson skirt and the inner side of the caisson lid, which is in contact with soil. The parameters used for the dense sand in the Hardening Soil (HS) model included a friction angle of 41°, a tangent stiffness of 95,800 kN/m2 , and a weight density of 19.5 kN/m3 . The Hardening Soil model is an innovative method for simulating soil behaviours, providing a more accurate representation of soil stiffness by considering three different input stiffnesses: E50, Eur, and Eoed. It is much different than Mohr– Coulomb model, because the HS model takes the consideration for stress-dependent stiffness moduli, with all stiffnesses increasing with variation of pressure and a set value of a standard stress, generally taken as 100 kPa (1 Bar). To compute the interaction between the caisson and the nearby soils, interface elements were employed, and the soil hardening model parameters were used with default values (Table 1).

Response of Offshore Wind Turbine Foundation Subjected …

221

3 Result and Discussions 3.1 Effect of Soil’s Internal Friction Angle Increasing the angle of internal friction by 5° has led to a 42.5% decrease in lateral displacement at the caisson’s top, as illustrated in Fig. 3. The displacement contours also emphasize that settlement is particularly pronounced in the vicinity of the suction caisson foundation’s lid. Furthermore, this increase in the angle of internal friction has resulted in a 17% reduction in shear stress concentration along the caisson skirt. Table 1 Soil properties

Parameters Unit weight

Medium dense sand (kN/m3 )

19.5

Secant stiffness (kN/m2 )

125,100

Tangent stiffness (kN/m2 )

95,800

Effective cohesion

C

(kN/m2 )

0.1

Effective angle of friction ϕ  (°)

41

Relative density (%)

60

Fig. 3 Load displacement curve for various φ values

222

S. Kumar et al.

Fig. 4 Load displacement curve for various unit weight of sand

3.2 Effect of Soil’s Unit Weight Raising the unit weight of the soil by 5.5 kN/m3 has led to a 29% decrease in lateral displacement at the caisson’s upper section, as depicted in Fig. 4. The displacement patterns also underscore that settlement is particularly notable in the vicinity of the suction caisson foundation’s lid. Additionally, the increase in unit weight has resulted in a reduction of shear stress concentration along the caisson skirt by 28.80%.

3.3 Effect of Caisson Material Type Utilizing materials such as steel for the caisson has led to a 46.7% reduction in lateral displacement at the caisson’s top, as demonstrated in Fig. 5. The displacement patterns also reveal that settlement is most noticeable near the lid of the suction caisson foundation. Substituting aluminum for steel as the caisson foundation material has resulted in a 31.44% decrease in shear stress concentration along the caisson skirt.

Response of Offshore Wind Turbine Foundation Subjected …

223

Fig. 5 Load displacement curve for various E of caisson material

4 Conclusion Offshore wind turbines are immense and towering structures that come with a significant price tag. Any damage to these structures can result in the need for a substantial amount of funds for repairs or reinstallation. Earthquakes alone have the potential to cause extensive damage to turbine structures, and when combined with wind and sea waves, the nature of failure and the extent of destruction become highly unpredictable. Therefore, it is imperative to implement robust countermeasures to prevent significant damage to wind turbines. Local scour at caisson foundations of offshore wind turbines is a major critical structural stability issue. Scouring is a big problem alone to the foundation and earthquake can add more to this in a complex way. So, we need some strong countermeasures to resist the scouring near the foundation and seabed. The Peak Ground Acceleration (PGA) exerts a significant influence on the settlements, horizontal displacements, and rotation of offshore wind turbines (OWT). As the PGA increases, the settlements, horizontal displacements, and rotations of OWTs also experience a corresponding rise, potentially exceeding permissible limits. It is essential to thoroughly evaluate the impact of earthquakes in the design of OWTs situated in high seismic regions. In instances of substantial deformation, the study recommends implementing specific countermeasures to enhance the earthquake resilience of OWTs. As in our study, numerical simulation has been carried out for caisson foundation of OWT subjected to cyclic loadings. From the numerical analysis conducted, it can be concluded that increase in unit weight of soil and increase of internal friction of soil leads to less lateral displacement of caisson foundation inside the soil. Utilizing materials such as steel for the caisson has reduced lateral displacement compared

224

S. Kumar et al.

to concrete. However, replacing steel with aluminum further decreases foundation lateral displacement.

References 1. Thapar S, Sharma S (2020) Factors impacting wind and solar power sectors in India: a surveybased analysis. Sustain Prod Consum 21:204–215 2. Feyzollahzadeh M, Mahmoodi MJ, Nikarvesh SM, Jamali J (2016) Wind load response of offshore wind turbine towers with fixed monopile platform. J Wind Eng Ind Dyanm 158:122– 138 3. Khan MF, Khan MR (2013) Wind power generation in India: evolution, trends and prospects. Int J Renew Energy Dev 2:175–186 4. Gao B, Zhu W, Zhang Q, Ye G (2022) Response of suction bucket foundation subjected to wind and earthquake loads on liquefiable sandy seabed. Soil Dynam Earthq Eng 160:1–28 5. Wang X, Zeng X, Li X, Li J (2020) Liquefaction characteristics of offshore wind turbine with hybrid monopile foundation via centrifuge modelling. Renew Energy 145:2358–2372 6. Sun XY, Luan MT, Tang XW (2010) Study of horizontal bearing capacity of bucket foundation on saturated soft clay ground. Rock Soil Mech 31(2):667–672 7. Lian JJ, Sun LQ, Zhang JF, Wang HJ (2011) Bearing capacity and technical advantages of composite bucket foundation of offshore wind turbines. Trans Tianjin Univ 17(2):132–137 8. Byrne BW, Houlsby GT (2003) Foundations for offshore wind turbines. Phil Trans R Soc A 361(1813):2909–2930 9. Bhattacharya S, Biswal S, Aleem M, Amani S, Prabhakaran A (2021) Seismic design of offshore wind turbines: good, bad and unknowns. Energies 1–29 10. Lombardi D, Cox JA, Bhattacharya S (2011) Long-term performance of offshore wind turbines supported on monopiles and suction caissons. In: Dynamic soil properties and behavior, conference proceedings, pp 1–6 11. Achmus M, Akdag CT, Thieken K (2013) Load-bearing behavior of suction bucket foundations in sand. Appl Ocean Res 43:157–165 12. Wang P, Zhao M, Du X, Liu J, Xu C (2018) Wind, wave and earthquake responses of off shore wind turbine on monopile foundation in clay. Soil Dynam Earthq Eng 113:47–57 13. Sadorsky P (2021) Wind energy for sustainable development: driving factors and future outlook. J Clean Prod 289:125779 14. Zhang JF, Zhang XN, Yu C (2016) Wave-induced seabed liquefaction around composite bucket foundations of offshore wind turbines during the sinking process. J Renew Sustain Energy 8(2):023307 15. Lehane BM, Pedram B, Doherty JA, Powrie W (2014) Improved performance of monopiles when combined with footings for tower foundations in sand. J Geotech Geoenviron Eng 140:3– 10 16. Negro V, López-Gutiérrez JS, Esteban MD, Matutano C (2014) Uncertainties in the design of support structures and foundations for offshore wind turbines. Renew Energy 63:125–132 17. Brinkgreve RBJ, Engin E, Swolfs WM (2013) Plaxis bv, Delft 18. Bagheri P, Son won S, Kim Man J (2017) Investigation of the load-bearing capacity of suction caissons used for offshore wind turbines. Appl Ocean Res 67:148–161

Seismic Response Characteristics of Loess Slope in Seasonally Frozen Regions Using Shaking Table Test Jinchang Chen, Ailan Che, Zhijian Wu, and Lanmin Wang

Abstract The seasonal freeze–thaw effect and frequent earthquakes caused extensive and serious landslides in the Loess Plateau of China. To investigate the dynamic response characteristics of loess slope under the effect of freeze–thaw cycles and earthquakes, large-scale shaking table tests of the freeze–thaw and non-freeze–thaw loess model slopes were performed. The amplification law of acceleration in the slope was clarified by root mean square (RMS) amplification factor and the evolution law of energy and the predominant frequency of the slopes were revealed by variational mode decomposition and Hilbert transform methods (VMD-HT). Result shows that the RMS amplification factor of the surface of model slope subjected to freeze–thaw cycles was significantly greater than that of the slope without the effect of freeze–thaw cycles. Hilbert spectra shows that the Hilbert energy in the slope body is uniformly distributed within 50 Hz under 0.05 g, and low frequency energy (below 20 Hz) gradually dominates in the slope under 1.2 g. The maximum Hilbert energy in the surface of slope suffered freeze–thaw effect larger than it in the slope without the effect under the same input intensity, which suggested that the slope suffered freeze–thaw effect would experience the higher seismic energy. The results reflect that the slope after freeze–thaw cycles more easily damaged when subjected to the earthquake. The study can provide meaningful reference for seismic landslide hazard risk prevention in the region subjected to seasonal freeze–thaw cycles. Keywords Loess slope · Freeze–thaw cycles effect · Shaking table test · Dynamic response characteristics

J. Chen · A. Che (B) School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, 800 Dongchuan-Road, Shanghai 200240, China e-mail: [email protected] Z. Wu College of Transportation Science and Engineering, Nanjing Tech University, Nanjing 210009, China L. Wang Lanzhou Institute of Seismology, China Earthquake Administration, Lanzhou 730000, China © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_21

225

226

J. Chen et al.

1 Introduction The loess area in China exceeds 4.4 × 105 km2 , of which more than 90% is in seasonally frozen soil area. The seasonal repeated freeze–thaw cycle causes water content and structure of the loess surface changes significantly. The loess transits from the unfrozen state to the frozen state or the reverse cycle, and its physical and mechanical properties change dramatically, which would trigger landslides during the freeze–thaw period [1]. Meanwhile, earthquakes frequently occurred in the area. The existing field survey data show that seven major earthquakes with magnitude of 8 and above have occurred in the loess plateau area in history, and extremely serious regional seismic geological disasters have been caused in areas [2]. Studies have proved that the internal causes of the instability of frozen soil slopes are water and heat transfer of soil [3, 4] and the changes in physical properties [5, 6], the main external causes are meteorological conditions, hydrogeological conditions, and cover soil layer conditions [7–9]. The freeze–thaw process has dual effects, that is, it has compaction effect on loose soil and loosening effect on dense soil [10]. The increase of freeze–thaw cycles will continuously reduce the compressive strength and durability of the soil [11]. Some researchers have found that freeze–thaw cycles have periodic effects on loess. The resonance frequency of loess will drop sharply during the melting stage, but the resonance frequency of soil will gradually increase during the freezing process [12, 13]. Dynamic response characteristics of slope under earthquake are often investigated by shaking table test and simulation. Compared to the simulation, the experiment is more intuitive, and its conclusion is obtained through real phenomena, so the conclusion is highly reliable. The dynamic response of loess slope under the coupled effect of rainfall and earthquake was investigated through shaking table model test [14]. The energy and damage evolution of slope under frequent earthquake was clarified based on shaking table test using variational mode decomposition and Hilbert transform methods [15]. It shows that the predominate frequency of slope would gradually decrease with the accumulation damage degree. Most of loess area in China suffered freeze–thaw and the earthquakes frequently occurred there, which triggered serious landslides disaster. However, the dynamic response of loess slope suffered freeze–thaw effect under the action of earthquakes is rarely studied. Large-scale shaking table comparative tests of the freeze–thaw and non-freeze–thaw loess model slopes were performed. The dynamic response of them was investigated, which can provide meaningful reference for the work related to the failure mechanism of the landslides.

Seismic Response Characteristics of Loess Slope in Seasonally Frozen …

227

2 Methodology 2.1 Root Mean Square (RMS) Acceleration Amplification Factor The amplification factor of acceleration was often used to reflect the magnification of the inertial force within the slope. However, it is founded that the acceleration response would increase or decrease at a certain moment under the effect of surrounding environment. So, the factor cannot effectively show the magnification law of acceleration. In this study, RMS acceleration amplification factor was used to evaluate the amplification effect of acceleration of slope structure. The factor is defined as Eq. (1). It is the ratio of RMS acceleration of monitoring points to base point (shaking table in the shaking table test). It can overcome the randomness limits of maximum acceleration, which can reflect the whole and average level of acceleration. The calculation process is based on the MATLAB platform.  β= 

1 td

1 td

td

 21 a(t)2 dt

0

td

 21

(1)

b(t)2 dt

0

where β represents the RMS amplification factor of acceleration, a(t) is the acceleration of the monitoring points, t d is the duration, and b(t) is the acceleration of shaking table.

2.2 Variational Mode Decomposition (VMD) Each intrinsic mode function (IMF) of signal after VMD can be defined as Eq. (2), which is an amplitude modulated and frequency modulated signal. u k (t) = Ak (t) cos(φk (t))

(2)

where uk (t) is the IMF, ϕ k (t) is the phase of Input Signal, and Ak (t) is the envelope of the Input Signal. VMD method is a constrained variational problem to obtain the IMFs of Input Signal, as it shown in Eq. (3). The augmented Lagrangian function was introduced to solve the problem [16].

228

J. Chen et al.



 2

     j − jωk t   ∂t δ(t) + ∗ u k (t) e min   {u k },{ωk } πt 2 k  subject to u k (t) = f (t)

(3)

k

where f (t) is original signal, {uk } = {u1 ,…,uK }, {ωk } = {ω1 ,…,ωK }, δ(t) is a pulse function.

2.3 Hilbert Transform (HT) Each IMF acquired by VMD is well behaved in HT. HT is defined as Eq. (4). ⎛ ∞ ⎞

u(τ ) ⎠ 1 ⎝ v(t) = H [u(t)] = P dτ π t −τ

(4)

−∞

where v(t) represents the result of Hilbert transformation, H is the Hilbert transformation, and P represents the value of the Cauchy Principal. The analytic signal can be defined as Eq. (5). The result of Hilbert transformation is the imaginary part of it and the IMF is the real part of it. z(t) = u(t) + j v(t) = a(t)e jθ(t) a(t) =



u 2 (t) + v 2 (t)

θ (t) = arctan

v(t) u(t)

(5) (6) (7)

where z(t) is the analytic signal, a(t) is the amplitude, and θ (t) is the phase. The Hilbert spectrum H(t, ω) is defined as Eq. (8). It displayed the time, instantaneous frequency, and amplitude of signal on the same plane. H (t, ω) =

K 

ai (t, ωi )

(8)

1

ω(t) = where H(t, ω) is the Hilbert spectrum.

dθ (t) dt

(9)

Seismic Response Characteristics of Loess Slope in Seasonally Frozen …

229

3 Shaking Table Test The layout of the test is shown in Fig. 1a. The loess used in the shaking table test is from Lanzhou New Area, Gansu Province. The elastic modulus, cohesive strength, and internal friction angle are 58.78 MPa, 70.8 kPa, and 27.9°, respectively. Two slope models have similar physical and mechanical properties. The whole slope is divided into 9 layers to compact and each layer with thickness of 10 cm. The degree of compaction of each layer is 0.85. The size of each model is 2.84 × 1.45 × 0.9 m. In the test, two constant temperature cold baths were used and S-shaped copper pipes were laid on the slope to freeze the model 1 (Fig. 1b, c). The interval of copper tubes was 10 cm and the freezing temperature is − 20 °C. The freeze–thaw process is monitored by the pre-embedded temperature sensor. After the fourth freeze–thaw cycle, shaking table tests were performed on two slope models. The distribution of acceleration sensors is shown in Fig. 2 17. Acceleration sensors were used in the test, which were mainly distributed at the surface or along the elevation of the slope in order to reflect the amplification and the energy characteristics of acceleration of freeze–thaw and non-freeze–thaw loess model slopes. 20 temperature sensors were laid in the slope to monitor the change of temperature in the slope. Four freeze–thaw cycles were conducted, as shown in Fig. 3. The temperature changed from − 2.28 to 16 °C within 100 mm, and from 4 to 16 °C at 250 mm. The freeze–thaw effect mainly occurred on the surface of slope.

(a)

(b)

(c)

Fig. 1 Test models and Freeze–thaw equipment a Test model, b constant temperature cold baths, c S-shaped copper pipes

Fig. 2 Distribution of sensors (Acceleration sensors (Left), Temperature sensors (Right))

230

J. Chen et al.

Fig. 3 Temperature change during four freeze–thaw cycles

Tangyu wave in Wenchuan earthquake with intensity from 0.05 to 1.2 g was loaded. Figure 4 shows the wave form and the corresponding Fourier spectra of it. Two loading conditions (0.05 g and 1.2 g) were mainly investigated. The duration is 76 s.

(a) Wenchuan Tangyu wave Fig. 4 Loading wave

(b) Fourier spectra

Seismic Response Characteristics of Loess Slope in Seasonally Frozen …

231

4 Results and Discussions 4.1 Distribution Characteristics of RMS Acceleration Amplification Factor RMS acceleration amplification factor was calculated and the distribution of it in slope body was shown in Fig. 5. It exhibited that the amplification effect of acceleration increases with the increase of elevation and up to maximum at the slope shoulder for model 1 and model 2. Compared to the slope without freeze–thaw (model 2), the slope suffered freeze–thaw (model 1) has larger value. For model 1, the maximum value is 2.9, which is 52% larger the model 2 under 0.05 g (Fig. 5a, c). Under 1.2 g, the value of model 1 is 64% larger than model 2 (Fig. 5b, d). It indicates that the slope suffered freeze–thaw subjected to greater inertia force and more easily damaged at the slope shoulder in the earthquake. With the increases of intensity, the amplification value shows a downward trend. The maximum value decreases from 2.9 to 2.3 for model 1 and decreases from 1.9 to 1.4 for model 2, when the intensity increases from 0.05 to 1.2 g. It reflects that the damage degree of the slope increased and the structure of it was not complete. RMS acceleration amplification factors along slope surface and height were furtherly investigated, as shown in Fig. 6. The amplification shows a similar trend at the surface of slope for model 1 and model 2 with the increase of intensity (Fig. 6a). The trend shows a slight difference when compared to the slope suffered and without freeze–thaw. For the slope without freeze–thaw, it is clear that the maximum value occurred at A1. For the slope suffered freeze–thaw, A1 also has a large value. With the increase of height, the amplification effect shows a similar increasing trend. The maximum value occurred at A4. A4 and A7 in the slope under the effect of freeze– thaw increased greatly compared to the slope without freeze–thaw. It suggested that the slope surface suffered freeze–thaw subjected to large amplification effect.

(a) Model1(0.05g)

(b) Model1(1.2g)

(c) Model2(0.05g)

(d) Model2(1.2g)

Fig. 5 Distribution of RMS acceleration amplification factor (mm)

232

J. Chen et al.

(a) Amplification factor along surface

(b) Amplification factor along height

Fig. 6 Amplification factor along surface and height

4.2 Characteristics of IMFs VMD result of A3 under 0.05 g is shown in Fig. 7a. The original signal was decomposed into three IMFs with frequency from low to high. The corresponding Fourier spectra of IMFs is shown in Fig. 7b. It shows that the predominate frequency of each IMF is corresponding to the predominate frequency of original signal, which are 4.03 Hz, 15.49 Hz, and 29.89 Hz, respectively. The predominate frequency of each IMF has well incoherence with each other. Correlation coefficient and Variance contribution rate of each IMF to the original signal are shown in Fig. 8. It reflects that the IMF2 has the most relevant to the original signal and the correlation coefficient up to 0.8. IMF2 also has the largest variance contribution rate, which is approximately 60%. It indicates that the IMF2 contains mainly information of the original signal. Based on the VMD method, meaningful components of the seismic signal can be obtained.

(a) VMD result

(b) Fourier spectra of IMFs

Fig. 7 VMD result and Fourier spectra of IMFs of A3 under 0.05 g of model 1

Seismic Response Characteristics of Loess Slope in Seasonally Frozen …

233

Fig. 8 Correlation coefficient and variance contribution rate of each IMF

4.3 Distribution Characteristics of Hilbert Energy Hilbert spectra of the A3 under 0.05 g and 1.2 g of model 1 is shown in Fig. 9. Under 0.05 g, the predominate frequency is distributed within 50 Hz and the energy is approximately uniformly distributed within 50 Hz. With the increase of intensity, the predominate frequency gradually concentrated below 20 Hz and the energy is also concentrated within 20 Hz. Low frequency energy gradually dominates in the slope. It is mainly the components of input signals. It is suggested that the slope body changed from free vibration to forced vibration due to the large intensity of input signal and the great damage degree. Similar results also occurred in the model 2, as shown in Fig. 10. It reflects that the energy of the signal has a trend to concentrate on the low frequency with the damage and intensity increase for the slope suffered or without freeze–thaw effect. The maximum Hilbert energy in model 1 is 16.7% and 71.4% larger than it in model 2, when the input intensity is 0.05 g and 1.2 g, respectively. It suggested that slope suffered freeze–thaw effect experiences the higher seismic energy.

234

J. Chen et al.

(a) 0.05g

(b) 1.2g

Fig. 9 Hilbert spectra of A3 under different intensity (model 1)

(a) 0.05g

(b) 1.2g

Fig. 10 Hilbert spectra of A3 under different intensity (model 2)

5 Conclusion In order to clarify the freeze–thaw effect on the dynamic response of slope, a comparative shaking table test of the freeze–thaw and non-freeze–thaw loess model slopes was performed. The dynamic response was investigated by RMS amplification factor and VMD-HT. The conclusions are as follows: (1) The surface of slope suffered freeze–thaw has a larger RMS amplification of acceleration compared to the slope without freeze–thaw effect. It suggested that the slope suffered freeze–thaw subjected to greater inertia force and more easily damaged in the earthquake. With the increases of intensity, the amplification value shows a downward trend. It reflects that the damage degree of the slope increased with the increase of intensity.

Seismic Response Characteristics of Loess Slope in Seasonally Frozen …

235

(2) The Hilbert energy in the slope body is uniformly distributed within 50 Hz under low intensity. With the increase of intensity, low frequency energy (below 20 Hz) gradually dominates in the slope suffered or without freeze–thaw effect. The slope would change from free vibration to forced vibration due to the large intensity and the accumulation of damage degree. The maximum Hilbert energy in model 1 is larger than it in model 2 under same input intensity, which suggested that the slope suffered freeze–thaw effect experiences the higher seismic energy. Acknowledgements This research work was supported by the Major Science and Technology Special Program of Yunnan Province (No. 202303AA080010), Key Program of National Natural Science Foundation of China (No. 42330704)

References 1. Kong J, Zhuang J, Peng J, Ma P, Zhan J, Mu J, Wang J, Zhang D, Zheng J, Fu Y, Wang S, Du C (2023) Failure mechanism and movement process of three loess landslides due to freeze-thaw cycle in the Fangtai village, Yongjing County, Chinese Loess Plateau. Eng Geol 315:107030 2. Chen J, Wang L, Pu X, Li F, Li T (2020) Experimental study on the dynamic characteristics of low-angle loess slope under the influence of long- and short-term effects of rainfall before earthquake. Eng Geol 273:105684 3. Shoop SA, Bigl S (1997) Moisture migration during freeze and thaw of unsaturated soils: modeling and large scale experiments. Cold Reg Sci Technol 25(1):33–45 4. Ban Y, Lei T, Liu Z (2016) Comparison of rill flow velocity over frozen and thawed slopes with electrolyte tracer method. J Hydrol 630–637 5. Harris C, Smith JS, Davies MC (2008) An investigation of periglacial slope stability in relation to soil properties based on physical modelling in the geotechnical centrifuge. Geomorphology 93(3):437–459 6. Luo X, Jiang N, Fan X (2015) Effects of freeze–thaw on the determination and application of parameters of slope rock mass in cold regions. Cold Regions Sci Technol 32–37 7. Kawamura S, Miura S (2011) Stability evaluation of volcanic slope subjected to rainfall and freezing and thawing action based on field monitoring. Adv Civil Eng 1–14 8. Kawamura S, Miura S (2013) Rainfall-induced failures of volcanic slopes subjected to freezing and thawing. Soils Foundations 53(3):443–461 9. Wang S, Qi J, Yu F (2016) A novel modeling of settlement of foundations in permafrost regions. Geomechanics Eng 10(2):225–245 10. Viklander P (1998) Permeability and volume changes in till due to cyclic freeze thaw. Can Geotech J 35:471–477 11. Takeshi K, Aly A, Toshihide S (2012) Effect of freeze-thaw cycles on durability and strength of very soft clay soil stabilised with recycled Bassanite. Cold Regions Sci Technol 82:124–129 12. Weber S, Fah D, Beutel J (2018) Ambient seismic vibrations in steep bedrock permafrost used to infer variations of ice-fill in fractures. Earth Planetary Sci Lett 501:119–127 13. James SR, Knox HA, Abbate RE (2019) Insights into permafrost and seasonal active-layer dynamics from ambient seismic noise monitoring. J Geophys Res-Earth Surf 124:1798–1816 14. Pu X, Wang L, Wang P, Chai S (2020) Study of shaking table test of seismic subsidence loess landslides induced by the coupling effect of earthquakes and rainfall. Nat Hazards 103(1):923– 945 15. Chen J, Che A, Wang L (2023) Cumulative damage evolution rule of rock slope based on shaking table test using VMD-HT. Eng Geol 314:107003

236

J. Chen et al.

16. Dragomiretskiy K, Zosso D (2014) Variational mode decomposition. IEEE Trans Signal Process 62(3):531–544

Seismic Responses of Rubble Mound Breakwater: Numerical Analyses P. K. Akarsh, Babloo Chaudhary , Manu K. Sajan, and Subodh Kumar

Abstract Rubble mound breakwater is a coastal structure, which is constructed to provide tranquil conditions in and around the port areas. Generally, the rubble mound structures are subjected to vigilant waves throughout the year. After the earthquakes of Kobe (1995), Kocaeli (1999), Tohoku (2011) etc. it is observed that the breakwaters can collapse due to failure of foundation and by seismic activity. Hence, in order to assess this problem, the current investigation deals with the study of rubble mound breakwaters and it is behavior against the seismic forces using numerical analysis. A finite element software PLAXIS is used for the numerical simulations. For study, a prototype has been selected and numerical model developed is a conventional rubble mound breakwater. In countermeasure model, the sheet piles in the foundation soil on extreme side of mound were considered. The numerical analyses have been done for constant seismic loading and soil properties. The parameters like vertical settlement and horizontal displacement were determined at different nodes. The vertical settlement was observed to be predominant in the crest region and it was reduced by 38% in countermeasure model. The displacement contours were significantly seen in core and armor units. The horizontal displacement of mound was seen by lateral movement of outer layers and it was 23% lesser for sheet pile reinforced model. Keywords Rubble mound breakwaters · Settlement · Horizontal displacement · Seismic loading · Excess pore water pressure · Sheet piles

1 Introduction The waves generated in the mid of the sea travel toward the coastal shores. Part of the energy is dissipated by wave action and rest by hitting against the shore area. To protect the shore area and ports and harbors, an offshore structure called breakwater is built. These breakwaters provide tranquil condition and resist the vigilant P. K. Akarsh (B) · B. Chaudhary · M. K. Sajan · S. Kumar Department of Civil Engineering, National Institute of Technology Karnataka, Surathkal 575025, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_22

237

238

P. K. Akarsh et al.

waves traveling toward coastal structures. Rubble mound breakwaters are one such breakwater, trapezoidal in shape and heterogeneous assemblage of natural rubble consisting of quarried rocks in the core and natural or artificial armor as a protection cover. These breakwaters dissipate wave energy by 60–70% absorbing and rest by reflection. Most of the energy will be absorbed by outer armor units, the inner layers prevent transmission of wave energy and wave energy dissipated through voids of breakwater [1–3]. The breakwaters are subjected to vigilant wave forces throughout the year. Along with this, they might also be subjected to other environmental forces like wind forces, earthquake forces, etc. After the Indian Ocean earthquake of 2004, the important coastal structures including breakwaters were damaged in Srilanka, Indonesia, India, Male etc., [4, 5] and during Great East Japan earthquake of 2011, the foundations of many breakwaters like breakwater at Iwate prefecture, Hattaro breakwater etc., were collapsed [6, 7]. When breakwater foundations are subjected to earthquake, there will be excessive pore water pressure in the soil and shear strain in the structure leading to deformation or settlement of the breakwater. Sever tremors create liquefaction condition. Due to this earth vibration, there will be subsequent tsunami approaching the breakwaters [8, 9]. If these structures are resilient during an earthquake, the effect of Tsunami would be lesser. Hence, it is necessary to study the behavior of seismically induced behavior of rubble mound breakwaters. Unfortunately, only a few researchers around the world have studied about the seismic behavior of rubble mound breakwater. Kiara et al. [10] developed both physical and numerical model to investigate the seismic behavior of rubble mound breakwaters on solid bed and loose sand. In the model I, the armors were placed interlocking to each other and in model II, the armors were placed in riprap manner. In the numerical approach, they considered (1) geotechnical parameters under seismic action along with analytical expression to calculate the hydrodynamic pressure and (2) fundamental oscillation and excitation frequency of surrounding water. Memos et al. [11] performed the shaking table experiment and analyzed the seismic behavior considering hydrodynamic pressure on the slope of breakwaters with soft foundations. They used boundary element code or hydrodynamic code (Elastic) along with the existing geotechnical code (QUAD4M) to check the hydrodynamic pressures along the inclined faces and height wise accelerations of flexible mound. Cihan et al. [12] analyzed the dynamic reaction of homogeneous rubble mound breakwater for dry and wet conditions with the help of physical models by shaking table and numerical models by PLAXIS 2D. They said that with the increase in the base accelerations, the settlement, damage level and volumetric strain of core and armor increased. The three-dimensional FEM model is developed by Ye and Jeng [13] using Biot’s dynamic poro-elastic theory to investigate the consolidation and dynamic behavior of rubble mound. Yang and Jin [14] examined seawater-breakwater-foundation interaction against the seismic waves using arbitrary Lagrangian–Eulerian approach (mathematical) and LS-DYNA finite element program to know the dynamic response. Chaudhary et al. [15] have used sheet piles and geogrids in the foundation of composite type breakwater and the mechanism was elucidated using FLAC application. Jafarian et al. [16] investigated seismic response of the breakwater placed on thick sediments

Seismic Responses of Rubble Mound Breakwater: Numerical Analyses

239

of loose liquefiable soil. Two models were considered one with conventional one and the other to counteract the liquefaction, the stone columns were used. The data of centrifuge test was validated with the two-dimensional numerical model created by FLAC. The use of stone column dissipates the excess pore water pressure rapidly and soil can restore its resistance. In the current study, a numerical investigation is done by developing a model in FEM software called PLAXIS 2D. The seismic behavior of conventional model is studied considering parameters like settlement, displacement etc. Also from literatures, it is noted that the countermeasures for mitigating earthquake-induced damage of earthquake are limited in numbers. Hence, along with the conventional model, a countermeasure model consisting of sheet piles in the foundation is developed to improve the resilient behavior of breakwater. The effectiveness of the developed countermeasure is compared. The model description and results have been discussed in subsequent sections.

2 Model Description In this study, the numerical modeling of rubble mound breakwater was carried out by finite element two-dimensional approach utilizing a software called PLAXIS 2D. One of the cross-sections of prototype breakwater based in Vizhinjam International Seaport, India was considered as conventional breakwater. The superstructure of breakwater consists of outer armor on sea side, underlayers on harbor side, inner core and a crown wall. The superstructure was built on the two layers of foundation soil i.e., upper loose sand and lower dense sand. The superstructure was placed center and suitable dimensions of loose and dense sand were considered, so that there shall not be any boundary effect during seismic analysis. The schematic representation of conventional model drawn in PLAXIS 2D is shown in Fig. 1a. Initially, the boundaries of two layers of foundation soil were drawn, then the trapezoidal cross-section of superstructure was drawn to prototype dimensions. The properties of materials used in superstructure and foundation soil were defined by Hardening soil model and it is depicted in Table 1. The properties of armor and core materials were considered from the values given by Cihan et al. [17] and the details about sand layers were taken from Deghoul et al. [18]. The crown wall with shear key was constructed based on linear elastic model with modulus of elasticity (E) = 27 × 106 kN/m2 and unit weight (γ ) = 25 kN/m3 . The defined geometry was then meshed using 15 nodded triangular elements. The grid was selected fine size in order to minimize numerical distortion of transmission of seismic wave and element size selected was as small as 1/10 to 1/ 8 of the wavelength of highest frequency [19]. In the similar way, the countermeasure model was drawn to same dimensions and soil properties. The sheet piles were provided on extreme sides of superstructure and within the foundation soil, protruding until it was into dense sand (24.7 m). The sheet piles were assigned with EA (axial stiffness) value of 6 × 106 kN/m. Suitable interfaces were provided between sheet pile boundaries and soil layer. The drawn

240

P. K. Akarsh et al.

Fig. 1 Model breakwater developed in PLAXIS 2D: a Conventional model and b Countermeasure model

geometry was then meshed using 15 nodded triangular elements with coarse factor of 0.04. The schematic representation of countermeasure model drawn in PLAXIS 2D is shown in Fig. 1b. The mean sea level was taken as water level and kept constant to 14.8 m for both the models by providing seepage boundaries along the outlines of slope of mound and seabed. The effect of hydrodynamic forces created by dynamic loadings is taken into account by Westergaard’s added mass implementation [20]. The seismic loading was provided in the form of sinusoidal input wave form. The sinusoidal wave was assigned a frequency of 8 Hz, amplitude intensity of 0.4 g for 10 s duration and it was applied at the base of the model. The acceleration measured at the base of model is shown in Fig. 2. The dynamic analyses were done by providing absorbing boundary conditions on both the sides and at bottom of seabed. The boundary conditions were assigned with free-field elements in x-direction and complaint base in the lower y-direction. The foundation soil was assigned with a Rayleigh damping ratio of 5%, which is commonly used for most of the geologic materials [21, 22]. The parameters like settlement and horizontal displacement were measured on the top nodes of crown wall.

Seismic Responses of Rubble Mound Breakwater: Numerical Analyses

241

Table 1 Properties of materials used to develop breakwater model Parameters

Loose sand

Dense sand

Core

Under layer I

Under layer II

Armor

Unsaturated unit weight, γ (kN/m3 )

17

17.5

16

17

17

18

Saturated Unit weight, γsat (kN/m3 )

17.5

18

17

18

18

20

Secant young’s modulus, E50 ref (kN/ m2 )

16.4 × 103

60.4 × 103

110 × 103

100 × 103

105 × 103

95 × 103

Ref. unloading–reloading modulus, Eur- ref (kN/ m2 )

49.14 × 103

18.1 × 103

230 × 103

220 × 103

225 × 103

210 × 103

Cohesion, C (kN/m2 )

0.5

0.5

–

–

–

–

Angle of internal friction, φ (°)

33.2

37

41

43

44

46

Dilatancy angle, ψ (°)

7

21

10

10

10

10

Poisson rate, V

0.3

0.3

0.2

0.2

0.2

0.2

Power for stress dependency, m

0.5

0.5

0.5

0.5

0.5

0.5

Failure ratio

0.9

0.9

0.9

0.9

0.9

0.9

References

Deghoul et al. [18]

Cihan et al. [17]

Fig. 2 Input acceleration recorded at the base of model

242

P. K. Akarsh et al.

3 Results and Discussion 3.1 Settlement The settlement of breakwater was measured at V1 and V2 locations. The behavior of conventional and countermeasure model under the 0.4 g amplitude seismic loading for conventional and countermeasure at respective location is shown in Fig. 3a, b. The conventional model settled 347 mm at V1 and 338 mm at V2 locations, whereas the countermeasure model displaced downwards 219 mm and 208 mm at respective locations (negative value indicates that displacement was downwards). These values are observed at the end of given time, 10 s. During seismic loading, V1 settled greater than V2 because of lesser stability. The vertical settlement was observed to be predominant in the crest region due to the subsidence of lower layers of mound into the soil. The average settlement of countermeasure model was reduced by 37.6% as compared to conventional one. Due to inclusion of sheet pile, there is lateral confinement of sand and as a result, sand between the sheet pile will get more densify. This would improve the resistance of seabed against vertical deformation, which can be clearly understood from the displacement contours as shown in Fig. 4. In conventional model, the displacement is predominantly observed in the central portion of model and less in boundaries (Ref. Figure 4a). The movement of y-directional contours was seen maximum in core and armor layers and minimum toward lower sand regions. Where as in countermeasure model, the displacement contour valves were seen lesser (Ref. Fig. 4b).

3.2 Horizontal Displacement The horizontal displacement of breakwater was measured at H1 and H2 location. To explain the behavior of conventional and sheet pile included breakwater, the variation of horizontal displacement with time is considered as shown in Fig. 5. The average total horizontal displacement shown by conventional and countermeasure model was 354 mm and 273 mm, respectively (positive value indicates that displacement was right side). Therefore, on an average, the horizontal displacement was reduced by 23% for the sheet pile reinforced model, when sinusoidal input motion carrying 0.4 g amplitude is subjected. The displacement was due to the volumetric strains created by the combined action of seismic forces and hydrodynamic forces [23]. In the countermeasure model, the movement of sand in horizontal direction is restricted by sheet pile inclusion in the foundation soil and so is the lateral displacement superstructure in spite of above-mentioned forces. Hence, the sheet pile inclusion can act as barrier or cut off wall in restraining lateral movement of sand.

Seismic Responses of Rubble Mound Breakwater: Numerical Analyses

243

Fig. 3 Settlement vs Time at: a V1 and b V2 locations for conventional and countermeasure model

4 Conclusion In the current study, numerical analyses of rubble mound breakwater have been conducted to know the behavior under seismic loadings. In countermeasure model, the sheet piles were used to know the effectiveness and following conclusions were drawn, a. The settlement of core material in conventional model was due to the densification of loose sand and the average settlement was reduced by 38% for the countermeasure model.

244

P. K. Akarsh et al.

(a)

(b)

Fig. 4 Displacement contours: a Conventional model and b countermeasure model

b. The displacement contours were found more predominant in the upper core region and it was decreasing toward the dense sand indicating that settlement was greater in crest region. The use of sheet piles has reduced the displacement contours values. c. The horizontal displacement in conventional model was due to the combined effect of seismic inertial forces and hydrodynamic forces. The inclusion of sheet pile has decreased the lateral movement of breakwater by 23% due to lateral confinement of foundation soil by sheet piles.

Seismic Responses of Rubble Mound Breakwater: Numerical Analyses

245

Fig. 5 Average Horizontal displacement versus Time for conventional and countermeasure model

References 1. CIRIA C683 (2007) The rock manual: the use of rock in hydraulic engineering, vol 2, p 550 2. Tulsi KR (2016) Three-dimensional method for monitoring damage to dolos breakwaters, Master of Science thesis, Stellenbosch University 3. Akarsh PK, Chaudhary B (2023) Review of literature on design of rubble mound breakwaters. In: Nandagiri L, Narasimhan MC, Marathe S (eds) Recent advances in civil engineering. Lecture Notes in Civil Engineering, vol 256. Springer, Singapore 4. Ergin A, Balas CE (2006) Damage risk assessment of breakwaters under Tsunami attack. Handbook of environmental chemistry, vol 5: Water Pollution 39(2):231–243 5. Sheth A, Sanyal S, Jaiswal A, Gandhi P (2006) Effects of the December 2004 Indian Ocean tsunami on the Indian mainland. Earthq Spectra 22 6. Hazarika H, Kasama K, Suetsugu D, Kataoka S, Yasufuku N (2013) Damage to geotechnical structures in waterfront areas of northern Tohoku due to the March 11, 2011 Tsunami disaster. Indian Geotechn J 43(2):137–152 7. Arikawa T, Sato M, Shimosako K, Hasegawa I, Yoem GS, Tomita T (2012) Failure mechanism of Kamaishi breakwater due to the great east Japan earthquake Tsunami. In: Proceedings of the 33rd conference on coastal engineering, Santander, Spain, pp 1–13 8. Chaudhary B, Hazarika H, Murakami A, Fujisawa K (2019) Development of resilient breakwater against Earthquake and Tsunami. Int J Geomech 19(1) (ASCE) 9. Akarsh PK, Chaudhary B, Sajan MK, Sah B, Kumar S. Novel technique to mitigate the earthquake-induced damage of rubble mound breakwater. Geotext Geomembr https://doi.org/ 10.1016/j.geotexmem.2023.11.001 10. Kiara A, Memos C, Tsiachris A (2001) Some practical aspects on the seismic behavior of rubble-mound breakwaters, Ports Conference 2001, American Society of Civil Engineers, Virginia, United States 11. Memos CD, Kiara A, Pavlidis E (2003) Coupled seismic response analysis of rubble-mound breakwater. In: Water and maritime engineering, ICE, 23–31

246

P. K. Akarsh et al.

12. Cihan K, Yuksel Y, Berilgen M, Cevik EO (2012) Behavior of homogenous rubble mound breakwaters materials under cyclic loads. Soil Dyn Earthq Eng 34:1–10 13. Ye JH, Jeng DS (2013) Three-dimensional dynamic transient response of a poro-elastic unsaturated seabed and a rubble mound breakwater due to seismic loading. Soil Dyn Earthq Eng 44:14–26 14. Yang X, Jin X (2015) Dynamic analysis of a rubble-mound breakwater subject to earthquake including seawater–structure foundation interaction. In: Engineering for the maritime environment, 1–17 15. Chaudhary B, Hazarika H, Murakami A, Fujisawa K (2018) Geosynthetic-sheet pile reinforced foundation for mitigation of earthquake and tsunami induced damage of breakwater. Geotext Geomembr 46:597–610 16. Jafarian Y, Bagheri M, Khalili M (2019) Earthquake-induced deformation of breakwater on liquefiable soil with and without remediation: case study of Iran LNG Port. In: Shu S, He L, Kai Y (eds) New developments in materials for infrastructure sustainability and the contemporary issues in geo-environmental engineering, sustainable civil infrastructures. Springer Nature 17. Cihan K, Yuksel Y (2011) Deformation of rubble-mound breakwaters under cyclic loads. Coast Eng 58:528–539 18. Deghoul L, Gabi S, Hamrouni A (2020) The influence of the soil constitutive models on the seismic analysis of pile-supported wharf structures with batter piles in cut-slope rock dike. Studia Geotechnica et Mechanica 42(3):191–209 19. Kuhlemeyer RL, Lysmer J (1973) Finite element method accuracy for wave propagation problems. J Soil Mech Foundations Division ASCE 99(4):21–27 20. PLAXIS 2D (2019) V20: hydrodynamic pressure in dynamic analysis: Westergaard’s added mass model validation cases. Bentley 21. Chaudharya B, Hazarika H, Nishimura K (2017) Effects of duration and acceleration level of earthquake ground motion on the behavior of unreinforced and reinforced breakwater foundation. Soil Dyn Earthq Eng 98:24–37 22. Ebrahimian B, Hosseinpanahi A (2019) Numerical investigation of seismic behavior of rubble mound breakwaters rested on liquefiable seabed foundations. In: The 4th Iranian conference on geotechnical engineering, Tehran, Iran 23. Ghalandarzadeh A, Ghalandarzadeh S, Abdi F (2021) Experimental study on the seismic deformations of rockfill breakwaters. Soil Dyn Earthq Eng 147

Stability Analysis of Rubble Mound Breakwaters Under Tsunami Overflow Manu K. Sajan, Babloo Chaudhary, P. K. Akarsh, and Subodh Kumar

Abstract Rubble mound (RM) breakwaters are the most commonly constructed breakwaters across the globe. Even though the breakwaters are designed to withstand to dynamic wave loadings, a natural disaster such as tsunami could impart additional loadings beyond the designed limits and thereby reduce the stability of the structure. Unfortunately, several RM breakwaters were severely damaged or even collapsed under the impact of past tsunamis such as the 2004 Indian Ocean tsunami and 2011 Great East Japan tsunami. The failure of these breakwaters would lead to the inundation of tsunami waves to the coastal areas causing devastating damages to life and property. Therefore, it is relevant to make the RM breakwaters resilient against tsunami impacts, so that the breakwater can either completely prevent or at least reduce the impact height of tsunami waves. In order to design a RM breakwater resilient against tsunami, the failure mechanisms under tsunami overflow conditions have to be properly understood. The present study thus aims to numerically evaluate the stability of RM breakwaters under tsunami overflow conditions. The cross-section details of the North breakwater at the Ennore Port, Chennai, India have been modelled at full scale in the finite element software Plaxis. The model was then subjected to a tsunami overflow condition. The corresponding deformations and stability of the RM breakwater were estimated. It was observed that the stability of the breakwater was considerably reduced under tsunami overflow conditions. Keywords Rubble Mound breakwaters · Tsunami · Numerical modelling

1 Introduction The response of coastal structures to the tsunami overflow conditions has gained attention among the researchers since the impact of 2011 Great East Japan tsunami [1–5]. Several coastal structures such as breakwaters, dykes, quay walls along many ports were severely damaged during the tsunami [6–8]. The experimental studies M. K. Sajan (B) · B. Chaudhary · P. K. Akarsh · S. Kumar National Institute of Technology Karnataka, Surathkal, Karnataka, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_23

247

248

M. K. Sajan et al.

backed up with numerical investigations on the mechanism involved in the failure of caisson type breakwaters had highlighted the influence of seepage and scouring during a tsunami overflow [9–11]. The coupled effect of seepage of water through the rubble foundation and scouring of exposed foundation under the impact of overflowing water jet has compromised the stability of the caisson units [12, 13]. However, similar studies on the estimation of stability of rubble mound type breakwaters under tsunami overflow condition are limited. The impact of 2011 Great East Japan tsunami has destroyed breakwaters across the coasts of Japan which were mainly caisson type. Whereas, the 2004 Indian Ocean tsunami had damaged many breakwaters along the coasts of Indonesia, Sri Lanka and India which were mainly rubble mound type [14, 15]. In caisson type breakwaters, the impervious caissons block tsunami wave which results in considerable level difference between seaside and harbour side water level that leads to severe seepage through the foundation. Even though the rubble mound breakwaters are highly pervious, the rate of water rising in seaside during a tsunami would always be larger than the permeability of the mound, which would also lead to a considerable level difference between sea levels on either side. The turbulent seepage of water through mound results in dislocation of rubbles which eventually reduces the stability of breakwater. In rubble mound breakwaters, the crown wall units were observed to be displaced under and overturned under the increased lateral thrust of tsunami bore [16, 17]. This eventually leads to scouring of amour blocks in along harbour side slope. In rubble mound breakwaters, the stability of breakwater has a dependency on the seepage force exerted by water percolating through the mound structure. In normal storm waves, the energy of impacted wave is dissipated through the mound by the natural interlocking mechanism between heavier rubbles. On the contrary, large volume of water would be seeping through the mound for a prolonged duration during a tsunami overflow. Therefore, the enormous energy transferred from water that seeps through the mound could exceed the inertial potential energy of each rubble and thereby destabilize the breakwater. In the present study, the stability analysis is performed in the finite element software Plaxis for different scenarios of tsunami overflow. The height of tsunami bore was varied to understand the change in the stability of the rubble mound. The deformation of nodes along crest, seaside slope and harbour side slope were analyzed to evaluate the degree of damage that occurred on the structure.

2 Methodology The geometrical details of the north breakwater at the Ennore port were used to model the RM as shown in Fig. 1. The breakwater was modelled on top of a seabed layer which was divided into two layers. The top layer was modelled as loose sandy soil and bottom as a hard layer. The geometry was then assigned with materials as per the properties mentioned in the material data set tabulated in Table 1 [18]. Previous research works by Cihan and Yuksel (2013) have accurately validated the numerical modelling of RM breakwaters using soil constitutive models in Plaxis [18].

Stability Analysis of Rubble Mound Breakwaters Under Tsunami Overflow

249

Fig. 1 Cross-sectional details of the RM breakwater

Therefore, the present study also adopts similar properties for numerical modelling of the RM breakwater. A fifteen nodded triangular meshing was assigned to the whole model with finer meshing towards within the breakwater. The numerical model was discretized into 6485 nodes and 786 elements. The flow conditions were then defined to simulate the rise of tsunami in the seaside. Tsunami was modelled as a reached level of water above withdrawn sea level. The water level during tsunami on either seaside or harbour side was plotted in the flow conditions module. Tsunami height was increased up to 15 m above the withdrawn sea level of 5 m by defining head functions. The model was then analyzed under staged construction tab with calculation type as fully coupled flow deformation analysis.

3 Results and Discussions The response of breakwater subjected to different height of tsunami impact has been presented in terms of the displacement in crown wall unit, changes in pore water pressure (PWP) and changes in factor of safety. The impact of tsunami wave is represented by the dimensionless parameter ht /hw where ht represents the height of tsunami bore above withdrawn sea level and hw represent the withdrawn sea level.

3.1 Displacement of Crown Wall The increase in height of tsunami bore exerted additional thrust on the RM towards the harbour side. This was evidently seen in the lateral displacement measured from a node located at the centre top of crown wall unit as shown in Fig. 2. As the impact height of tsunami wave increased, the corresponding displacement of crown wall also increased. It was observed that a 15 m high tsunami wave can cause a 90% more displacement on the crown wall unit. A detailed look into the curves representing each impact height shows that once the height of tsunami exceeds the height of breakwater the crown wall gets displaced rapidly. An increase in the displacement of crown wall was observed as the tsunami impact continued over a longer period.

250

M. K. Sajan et al.

Table 1 Input parameters for numerical modelling Material

Armour

Core

Crown wall

Top Sand layer

Bottom sand layer

Model

Hardening Soil

Hardening Soil

Linear Elastic

Mohr–Coulomb

Mohr–Coulomb

Drainage Type

Drained

Drained

Non-porous

Drained

Drained

γu (kN/m3 )

18

16

50

17

17

γd

(kN/m3 )

20

17

50

18

18

eint

0.5

0.5

0.5

0.5

0.5

E50 ref (kN/ m2 )

130 × 103

110 × 103

30 × 106

30 × 103

55 × 103

Eoed ref (kN/ m2 )

120.7 × 103

110 × 103

Eur ref (kN/ m2 )

270 × 103

230 × 103 0.15

0.3

0.35

υ’

0.2

0.2

power(m)

0.5

0.5

c’

0

0

1

1

φ’

45

41

35

42

5

10

0

0

ψ nc

0.316

0.35

Data Set

Standard

Standard

Standard

Standard

Soil Type

Coarse

Coarse

Coarse

Coarse

kx (m/s)

1

0.8

8.25 × 10−5

0.8

10−5

K0

ky (m/s)

1

8.25 ×

8.25 × 10−5 10−5

3.2 Effect of Pore Water Pressure The changes in PWP measured from the node located near the toe of the breakwater clearly indicated that with higher ht /hw ratios, the PWP hiked tremendously. This increase in PWP exerts higher magnitude of stresses on the primary armour units, which in turn displaces these units. Such dislocation of armour units from the mound would expose the lighter inner layers and thereby compromising the stability of breakwater. It is evident from Fig. 3 that as the impact height of tsunami increases the PWP in the RM rises considerably. At the highest tsunami impact, a notable 80% hike was observed in the PWP. Also, increased PWP was observed to be prominent throughout the overflow duration of tsunami wave. The destabilizing forces induced by higher PWP in the mound have an added effect in the displacement of crown wall detailed in Fig. 2.

Stability Analysis of Rubble Mound Breakwaters Under Tsunami Overflow

251

0.30 ht/hw =0.6 ht/hw =1.2 ht/hw =1.8 ht/hw =2.4 ht/hw =3.0

Lateral displacement of crown wall (m)

0.25

0.20

0.15

0.10

0.05

0.00 0

5

10

15

20

25

30

Time (min)

Fig. 2 Displacement of crown wall unit under various tsunami impact 200 ht/hw =0.6

180

ht/hw =1.2

ht/hw =1.8

ht/hw =2.4

ht/hw =3.0

160

PWP at toe (kPa)

140 120 100 80 60 40 20 0 0

5

10

15 Time (min)

20

25

Fig. 3 Changes in PWP at the RM toe under different impact heights of tsunami

30

252

M. K. Sajan et al.

2.2

2.0

ΣMsf

1.8

1.6

ht/hw =0.6 ht/hw =1.2 ht/hw =1.8 ht/hw =2.4 ht/hw =3.0

1.4

1.2

1.0 0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

|u| (m)

Fig. 4 Stability analysis of the RM breakwater under different tsunami impacts

3.3 Stability of Breakwater Detailed stability analysis was performed on each case of tsunami impact on the breakwater structure. It was observed that the increased height of tsunami wave reduced the stability of the structure as shown in Fig. 4. The factor of safety which is denoted as Msf was observed to get reduced by 50% with the increase in height of tsunami bore. It was observed that as the tsunami height reached near the breakwater height, in cases where ht /hw = 2.5 and 3, the stability of breakwater was reduced more. Therefore, it can be inferred an overflow of tsunami wave over the breakwater causes more destabilizing forces and might even lead to the collapse of the structure. It was evidently seen in the reduction of factor of safety of the breakwater close towards unity during the overflow of the highest tsunami wave.

4 Conclusions An impact of tsunami wave on RM breakwaters could induce large destabilizing forces for prolonged duration. The present study numerically analyzes the stability of the RM breakwater under different scenarios replicating the impact of tsunami waves of different wave heights. The large magnitude of lateral thrust exerted by tsunami wave on the RM breakwater was evidently seen in the displacement of

Stability Analysis of Rubble Mound Breakwaters Under Tsunami Overflow

253

crown wall unit. In cases where impact height of tsunami wave exceeds the height of breakwater, a drastic reduction in the stability of breakwater was observed. This reduction in stability occurs due to the excess PWP developed within the breakwater under a high head tsunami. In addition to that, under a tsunami overflow scenario, the PWP at the toe region of breakwater increases rapidly which reduced the stability of armour layer. The destabilization of armour layer could expose the protected core layer which eventually leads to failure of breakwater. The following inferences could be drawn out of the present study. • When the impact height of a tsunami wave exceeded the height of breakwater, a 90% additional lateral displacement was observed in the crown wall unit. • A tsunami impact with ht /hw = 3 has generated an 80% higher PWP near the toe of breakwater. The larger seepage forces through the mound under higher PWP have reduced breakwater stability. • The combined effect of intense lateral thrust and seepage pressures has resulted in a 50% reduction in breakwater safety factor. The impact of a tsunami wave could be thus considered as one of the most critical loading conditions of RM breakwaters. Therefore, RM breakwaters must be designed to withstand the destabilizing forces induced by tsunami waves especially for coastal regions vulnerable to tsunami impacts.

References 1. Arikawa T, Sato M, Shimosako K, Hasegawa I, Yeom GS, Tomita T (2012) Failure mechanism of kamaishi breakwaters due to the great east Japan earthquake Tsunami. In: Proceedings of the coastal engineering conference, pp 2–6. https://doi.org/10.9753/icce.v33.structures.16 2. Kato F, Suwa Y, Watanabe K, Hatogai S (2012) Mechanisms of coastal dike failure induced by the Great East Japan Earthquake Tsunami. Coastal Eng Proc 1(33):40. https://doi.org/10. 9753/icce.v33.structures.40 3. Mitsui J, Matsumoto A, Hanzawa M, Nadaoka K (2016) Estimation method of armor stability against tsunami overtopping caisson breakwater based on overflow depth. Coastal Eng J 58(4). https://doi.org/10.1142/S0578563416400192 4. Suzuki K, Zikuhara S, Tatewaki K, Hosokawa Y (2018) Clarifying the stability of armour blocks behind the caisson against tsunami after abrupt change of breakwater width. In: Coasts, marine structures and breakwaters 2017. ICE Publishing, pp 1131–1140. https://doi.org/10. 1680/cmsb.63174.1131 5. Esteban M, Thao ND, Takagi H, Jayaratne R, Mikami T, Shibayama T (2015) Stability of breakwaters against tsunami attack. In: Handbook of coastal disaster mitigation for engineers and planners. Elsevier Inc, pp 293–323. https://doi.org/10.1016/B978-0-12-801060-0.00015-0 6. Sugano T, Nozu A, Kohama E, Shimosako K, Kikuchi Y (2014) Damage to coastal structures. Soils Found 54(4):883–901. https://doi.org/10.1016/j.sandf.2014.06.018 7. Takagi H, Bricker JD (2015) Damage and the effect of breakwaters on inundation extent case study of the 2011 Great East Japan Earthquake and Tsunami. In: Handbook of coastal disaster mitigation for engineers and planners. Elsevier Inc. https://doi.org/10.1016/B978-0-12-801 060-0.00018-6

254

M. K. Sajan et al.

8. Koshimura S, Hayashi S, Gokon H (2014) The impact of the 2011 Tohoku earthquake tsunami disaster and implications to the reconstruction. Soils Found 54(4):560–572. https://doi.org/10. 1016/j.sandf.2014.06.002 9. Zen K, Kasama K, Kasugai Y, Dong S (2013) Failure of rubble mound beneath caisson due to Earthquake-Induced Tsunami. https://doi.org/10.1115/OMAE2013-10059 10. Bricker JD, Nakayama A, Takagi H, Mitsui J, Miki T (2015) Mechanisms of damage to coastal structures due to the 2011 Great East Japan Tsunami. In: Handbook of coastal disaster mitigation for engineers and planners. Elsevier Inc, pp 385–415. https://doi.org/10.1016/B978-0-12-801 060-0.00019-8 11. Chaudhary B, Hazarika H, Murakami A, Fujisawa K (2019) Development of resilient breakwater against Earthquake and Tsunami. Int J Geomech 19(1):1–17. https://doi.org/10.1061/ (asce)gm.1943-5622.0001314 12. Sassa S, Takahashi H, Morikawa Y, Takano D (2016) Effect of overflow and seepage coupling on tsunami-induced instability of caisson breakwaters. Coast Eng 117:157–165. https://doi. org/10.1016/j.coastaleng.2016.08.004 13. Takahashi H, Sassa S, Morikawa Y, Takano D, Maruyama K (2014) Stability of caisson-type breakwater foundation under tsunami-induced seepage. Soils Found 54(4):789–805. https:// doi.org/10.1016/j.sandf.2014.07.002 14. Kozai K, Kobayashi E, Kubo M, Morita K (2005) Tsunami disaster assessment of ports in India by using quickbird images. In: Asian association on remote sensing—26th Asian conference on remote sensing and 2nd Asian space conference, ACRS 2005, vol 1, pp 59–63 15. Narayan JP, Sharma ML, Maheshwari BK (2006) Tsunami intensity mapping along the coast of Tamilnadu (India) during the deadliest Indian Ocean tsunami of December 26, 2004. Pure Appl Geophys 163(7):1279–1304. https://doi.org/10.1007/s00024-006-0074-6 16. Guler HG, Arikawa T, Oei T, Yalciner AC (2015) Performance of rubble mound breakwaters under tsunami attack, a case study: Haydarpasa Port, Istanbul, Turkey. Coast Eng 104:43–53. https://doi.org/10.1016/j.coastaleng.2015.07.007 17. Aniel-Quiroga Í, Vidal C, Lara JL, González M, Sainz Á (2018) Stability of rubble-mound breakwaters under tsunami first impact and overflow based on laboratory experiments. Coast Eng 135(February):39–54. https://doi.org/10.1016/j.coastaleng.2018.01.004 18. Cihan K, Yuksel Y (2013) Deformation of breakwater armoured artificial units under cyclic loading. Appl Ocean Res 42:79–86. https://doi.org/10.1016/j.apor.2013.05.002

Disaster and Environment

Anisotropy of Pressure Generated by Water Absorption of Deformation-Constrained Granular Bentonite Specimens Yuta Ichikawa, Hideo Komine, and Shigeru Goto

Abstract Bentonite is being considered for use as a buffer material in the geological disposal of high-level radioactive waste, and studies of its swelling pressure have been conducted. However, most of the previous studies have measured only the vertical pressure and not the lateral pressure. In this study, the vertical and lateral pressures generated by water absorption in a granular bentonite specimen under deformation constraints were measured. The results showed that the lateral pressure converged to a value slightly greater than the vertical pressure, with a difference of about 200 kPa for a dry density of 1.6 Mg/m3 and about 100 kPa for a dry density of 1.4 Mg/m3 . The larger specimen volume required more time to converge, but if the dry density were equal, there was no significant difference in the convergence value of the pressure. This suggests that the use of specimens with small dimensions is effective in obtaining pressure properties of granular bentonite in a short time. Keywords Bentonite · Swelling pressure · Anisotropy

1 Introduction A project for the geological disposal of high-level radioactive waste is underway. Figure 1 shows a schematic diagram of geological disposal. The waste containers (overpack) will be buried at a depth of more than 300 m [1]. A buffer material will be placed between the overpack and the surrounding bedrock. Many functions are required of buffer materials, such as low permeability and self-sealing properties, to control radionuclide migration, and bentonite is considered to be a material that meets these requirements. Bentonite buffer materials are believed to swell when in contact with groundwater, exerting pressure on the overpack and surrounding Y. Ichikawa (B) Shimizu Corporation, 16-1, Kyobashi 2-chome, Chuo-ku, Tokyo, Japan e-mail: [email protected] H. Komine · S. Goto Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_24

257

258

Y. Ichikawa et al.

Fig. 1 Concept of a geological disposal facility

bedrock. Laboratory experiments on the swelling pressure of bentonite have been conducted to evaluate the design of the overpack and the effect of the pressure on the surrounding bedrock [1–5]. However, most of them measured only the pressure in the vertical direction of the specimen (water supply direction) and did not measure the lateral pressure. On the other hand, a study using Japanese powder bentonite Kunigel-V1 reported that vertical and lateral pressures converged to the same degree [6, 7]. In this study, we measured the vertical and lateral pressures generated by water absorption in granular bentonite specimens under deformation constraints and compared and discussed the two.

2 Material and Specimen Preparation Method Table 1 shows the fundamental properties of the granular bentonite used (Kunigel GX manufactured by KUNIMINE INDUSTRIES CO., LTD.) [8]. Since the optimum water content of granular bentonite is approximately 18%, the samples in this study were watered and cured at the target water content of 20% [9]. Figure 2 shows the appearance of the samples after water addition and curing. Figure 3 shows the method used to prepare granular bentonite specimens. Bentonite was placed and compacted in either 60 mm or 150 mm diameter rings with integrated pressure gauges. The 60 mm diameter ring incorporates three pressure gauges located 10 mm from the bottom of the ring, with a pressure-sensing surface diameter of 7.6 mm, a maximum capacity of 3000 kPa, and a minimum scale of 1.5–1.6 kPa. The 150 mm diameter ring incorporates three pressure gauges

Anisotropy of Pressure Generated by Water Absorption … Table 1 Fundamental properties of sample

259

Maximum particle diameter (mm)

10

Particle shape type

Granular Sodium

Type Soil particle density

(g/cm3 )

2.65

Liquid limit (%)

344.4

Plastic limit (%)

23.6

Plasticity index

320.8

Content of montmorillonite (%)

48.6

Fig. 2 Appearance of sample

at 15 mm or 30 mm from the bottom of the ring, with a pressure-sensing surface diameter of 14.8 mm, a maximum capacity of 2000 kPa, and a minimum scale of 0.5–0.6 kPa. The procedure for preparing the specimens is as follows. First, the ring, bottom plate, and top plate were bolted together and the pressure gauge was started. This is to measure the lateral pressure generated by vertical compaction. Next, the bentonite was placed in the ring and compacted into two, three, or five layers, with each layer approximately 10 mm thick, depending on the height of the specimen. Table 2 shows a list of cases of specimens prepared by the above procedure. In each experiment, the specimen height was determined as follows. For the 60 mm diameter ring, the length from the top of the ring to the top of the specimen was measured at five locations as shown in Fig. 3 (left) and calculated from the average value. For the 150 mm diameter ring, a square bar was placed on the top plate as shown in Fig. 3 (right), and the length from the bar to the top of each layer was measured at five locations and calculated from the average value. On the other hand, the length up to the final layer (fifth layer) was measured at 17 locations.

260

Y. Ichikawa et al.

Fig. 3 Specimen preparation method

Table 2 List of experimental cases Case

Specimen dry density (Mg/m3 )

Specimen diameter (mm)

Specimen height (mm)

Length from bottom of specimen to pressure gauge (mm)

A

1.6

60

20

10

B

1.6

150

50

30

C

1.6

150

30

15

D

1.4

60

20

10

E

1.4

150

50

30

3 Water Supply Method and Method of Measuring Pressure In this experiment, the pressure generated by supplying distilled water to a specimen with constrained deformation is measured. Water was supplied to the specimen by connecting the water supply port on the bottom plate to a double tube burette, as shown in Fig. 4.

Anisotropy of Pressure Generated by Water Absorption …

261

Fig. 4 Experimental device

Since swelling deformation could not be completely prevented, the vertical deformation was measured with a displacement transducer with a maximum capacity of 25 mm and a minimum scale of 0.002 mm. Of the pressure generated by the above procedure, the lateral pressure was measured with the same pressure gauge used when the specimens were fabricated, and the vertical pressure was measured with a load cell with a maximum capacity of 50 kN and a minimum scale of 0.012 kN. Before the start of water supply, low pressure was applied in the vertical direction to check the contact between the load cell and the specimen.

4 Experimental Results and Discussion 4.1 Pressure Generated During Specimen Preparation and Initial Phase of Water Supply Figures 5 and 6 show the pressure generated during specimen preparation and initial phase of water supply. The lateral pressure increased with vertical compaction and decreased slowly after the completion of compaction. It can be seen that the pressure does not show an immediate upward trend after the start of water supply. This may

262

Y. Ichikawa et al.

Case A

Case B

Case C

Fig. 5 Pressure generated during specimen preparation and initial phase of water supply (Case with a dry density of 1.6 Mg/m3 )

be due to the fact that the water supply causes not only swelling but also skeletal changes in the specimen.

4.2 Pressure Generated After Water Supply Figures 7, 8, and 9 show the experimental results for a dry density of 1.6 Mg/m3 (case A, B, C). Residual pressures due to compaction were existed in the lateral direction at the start of water supply as described in previous chapter. In the vertical direction, on the other hand, low pressure was generated by tightening the clump. From Figs. 7, 8, and 9, it was found that a larger specimen volume requires more time for convergence, but if the dry density is equal, there is no significant difference

Anisotropy of Pressure Generated by Water Absorption …

263

Fig. 6 Pressure generated during specimen preparation and initial phase of water supply (Case with a dry density of 1.4 Mg/m3 )

in the convergence value of the pressure. Lateral pressure converges to about 1.2–1.5 times the vertical. Figures 10 and 11 show the experimental results for a dry density of 1.4 Mg/m3 (case C, D). In both cases, the lateral convergence was about 1.1 times the vertical convergence, and as in the 1.6 Mg/m3 case, it took longer for the larger specimen Fig. 7 Pressure after water supply (Case A)

264

Y. Ichikawa et al.

Fig. 8 Pressure after water supply (Case B)

Fig. 9 Pressure after water supply (Case C)

volume to converge, but if the dry density was equal, there was no significant difference in the convergence value of the pressure. Vertical displacement due to swelling was within 0.6 mm in all cases at dry density of 1.6 and 1.4 Mg/m3 . Figure 12 shows a comparison of the vertical and lateral pressures at the end of the measurement. It can be seen that there is no significant difference in the convergence value of the pressure if the dry density is equal. This suggests that the

Anisotropy of Pressure Generated by Water Absorption …

265

Fig. 10 Pressure after water supply (Case D)

Fig. 11 Pressure after water supply (Case E)

use of specimens with small dimensions is effective in obtaining pressure properties of granular bentonite in a short time. In all cases, the lateral pressure converged to a value slightly greater than the vertical pressure, with a difference of about 200 kPa for a dry density of 1.6 Mg/m3 and about 100 kPa for a dry density of 1.4 Mg/m3 . The reason for the slightly higher lateral pressure is thought to be the effect of the initial residual pressure.

266

Y. Ichikawa et al.

Fig. 12 Comparison of vertical and lateral pressure convergence values

5 Conclusion In this study, granular bentonite was compacted and specimens with dry densities of 1.6 Mg/m3 and 1.4 Mg/m3 were prepared, and the vertical and lateral pressures generated by the water absorption of the specimens with constrained deformation were measured. As a result, the lateral pressure converged to a value slightly greater than the vertical pressure, with a difference of about 200 kPa for a dry density of 1.6 Mg/m3 and about 100 kPa for a dry density of 1.4 Mg/m3 . The larger specimen volume took longer to converge, but if the dry density were equal, there was no significant difference in the convergence value of the pressure. This suggests that the use of specimens with small dimensions is effective in obtaining pressure properties of granular bentonite in a short time.

Anisotropy of Pressure Generated by Water Absorption …

267

References 1. Japan Nuclear Cycle Development Institute (1999) Project to establish the scientific and technical basis for HLW disposal in Japan -project overview report—JNC TN1400 99-022 2. Komine H, Ogata N (2004) Predicting swelling characteristics of bentonites. J Geotechn Geoenvironmental Eng 130(8):818–829 (American Society of Civil Engineers (ASCE)) 3. Mitachi T (2008) Mechanical behavior of bentonite-sand mixtures as buffer materials. Soils Found 48(3):363–374 4. Komine H, Yasuhara K, Murakami S (2009) Swelling characteristics of bentonites in artificial seawater. Can Geotechn J 46(2):177–189 5. Hiramoto M, Kobayashi Y, Aoyagi S, Miyanomae S (2008) A study for improvement of the credibility of evaluation method for the long-term dynamical behavior of near-field rock mass, 2; A study for the influence of inner pressures caused by buffer material and overpack, JAEAResearch 2008-013 6. Kudo K, Tanaka Y, Yokokura T, Kitamura I (2005) Discussions on the method for swelling pressure measurement of the compacted bentonite sample. In: The 40th annual meeting of the Japan National conference on geotechnical engineering 7. Tanaka Y, Watanabe Y (2019) Modelling the effects of test conditions on the measured swelling pressure of compacted bentonite. Soils Found 59:136–150 8. Komine H (2011) Fundamental study on swelling pressure characteristics of granular bentonite—comparison of swelling pressure of granular bentonite of which maximum diameters are 10mm and 2mm-. In: The 46th annual meeting of the Japan National conference on geotechnical engineering 9. Sugiura K, Komine H, Yasuhara K, Murakami S (2009) Influence of initial water content on swelling characteristics of bentonite ore. In: The 44th annual meeting of the Japan National conference on geotechnical engineering

Development and Geographical Evaluation of Slope Failure Risk Index Due to Changes of Soil Moisture Honoka Asada and Kohei Araki

Abstract The development of for sediment disaster prediction system for the community requires observational method for slope failure and mapping of slope failure risk index by using GIS. In this study, first, we prepared a field for experimental studies on slope failure. Second, a slope stability analysis is performed for surface slides, considering changes in cohesion associated with soil moisture, to derive the slope safety factor and probability of failure with changes in moisture content by volume. In addition, this study method is discussed to represent slope failure risk index due to changes in soil moisture on topographic maps. The occurrence of a small slope failure was captured on June 4, 2021, from 2:00 a.m. to 3:00 a.m. The stability analysis of the experimental slope field showed slope failure probability of more than 60% at the slope failure time. Slope failure probabilities were greater than 70% when the moisture content by volume was greater than approximately 35%. Slope failure probabilities were greater than 70% when the moisture content by volume was greater than approximately 35%. In the analysis of the area around National Institute of Technology, Tokuyama College, the analysis results were mapped by using GIS. At 25% moisture content by volume, the safety factor became less than 1.5, and the slope failure probability began to increase significantly. Slope failure probability was rapidly increased when the safety factor approached 1. Keywords Slope failure · Soil moisture · GIS

H. Asada (B) Advanced Course of Environmental and Civil Engineering, NIT, Tokuyama College, Yamaguchi 745-8585, Japan e-mail: [email protected] K. Araki Department of Civil Engineering and Architecture, NIT, Tokuyama College, Yamaguchi 745-8585, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_25

269

270

H. Asada and K. Araki

1 Introduction The frequency of heavy rains is increasing due to climate change such as global warming. Slope failures triggered by heavy rainfall are occurring in Japan. These have developed into sediment disasters. The average annual number of sediment disasters is 1450 in Japan (2012–2020) [1]. The torrential rain in July 2020 was one of the largest wide area disasters. It caused 961 sediment disasters in 37 prefectures [2]. The number of designated areas such as sediment-related disaster warning areas is 678,309 in Japan [3]. A community system is required to announce the safety of steep slopes. Also, it is necessary to identify the slope with high failure risk and to improve the accuracy of slope stability analysis. The development of for sediment disaster prediction system for the community requires observational method for slope failure and mapping of slope failure risk index by using GIS. In this study, first, we prepared a field for experimental studies on slope failure. Second, a slope stability analysis is performed for surface slides, considering changes in cohesion associated with soil moisture, to derive the slope safety factor and probability of failure with changes in moisture content by volume. In addition, this study method is discussed to represent slope failure risk index due to changes in soil moisture on topographic maps.

2 Development of Slope Failure Risk Index 2.1 Evaluation of the Relationship Between Soil Shear Strength Parameters and Precipitation (Matsuo [4] Method) Matsuo [4] method is used to evaluate the relationship between soil shear strength parameters (cohesion c and internal friction angle ϕ) and precipitation. Figure 1. Example of cohesion to degree of saturation relationship (silt sand) [4]. According to the experiments conducted by Matsuo and Ueno [5], regardless of the soil type, the cohesion peaks at a certain degree of saturation. Hereafter referred to as maximum cohesion cmax . Matsuo found that the difference between the degree of saturation at the maximum cohesion (hereafter referred to as inflection degree of saturation Sri ) and the optimum saturation Sropt has a strong negative correlation with the mean particle size D50 in Eq. (1). Equation (2) of the relationship between initial void ratio e0 and inflection degree of saturation was derived for various soil types. The degree of saturation is Eq. (4). The relationship between the ratio of cohesion to maximum cohesion, inflection degree of saturation, and degree of saturation was examined, and a linear relationship was found on average, which is expressed in Eq. (5). The wet unit weight γt is expressed by Eq. (6).

Development and Geographical Evaluation of Slope Failure Risk Index …

271

Fig. 1 Example of cohesion to degree of saturation relationship (silt sand) [4]

Sropt = −79.09D50 + 37.86

(1)

Sri = −100.0e0 +(Sropt − Sropt + 100emin )

(2)

Sr = {(1 + e0 )/ e0 }θ

(3)

cmax = 0.103(0.0086)e0 D50 Uc

(4)

μc = {1.0 − 0.018 (Sr − Sri )}cmax

(5)

γt = γd +(θ /100)γw

(6)

where Uc is the uniformity coefficient and γd is the dry unit weigh γw is the unit weight of water.

2.2 Slope Stability Analysis Methods A slope stability analysis is performed to determine the safety factor Fs . The safety factor in partial saturation is obtained using Eq. (7), and the safety factor in saturation is obtained using Eq. (8). Then, the dependence of the cohesion μc and the wet unit weight γt on soil moisture is evaluated. Sr < 100 Fs = Sr = 100 Fs =

μc + γt Z cos2 i tan ϕ τf = τ γt Z cos i sin i

(7)

μc + γsub Z cos2 i tan ϕ τf = τ γsat Z cos i sin i

(8)

272

H. Asada and K. Araki

Fig. 2 Normal distribution

where τf is shear strength to resist slip, τ is shear strength to cause slip, Z is surface thickness, i is inclination angle, ϕ is internal friction density, γsub is submerged unit weight, and γsat is saturated unit weight.

2.3 Calculation of Slope Failure Probability Considering the variation of cohesion force c, a normal distribution f(c) (mean μc , standard deviation σc (Eq. (9)) is introduced in Fig. 2. Find the cohesion cF with a safety factor set to 1, and calculate the probability that it will be less than or equal to cF , with c as a variable [6]. Then, let us define the probability of slope failure as in Eq. (10). σc = 0.182cmax

(9)

cF Pf =

f (c)dc

(10)

−∞

3 Field Experiment 3.1 Maintenance of Experimental Sites A field experimental slope (gradient of slope 30°, height 5 m, covered with 0.9 m wide and 0.1 m thick of Table 1 decomposed granite soil) was constructed at Tokuyama National College of Technology. Observation cameras, soil moisture meters, and rain gauges were used to determine slope dynamics.

Development and Geographical Evaluation of Slope Failure Risk Index …

273

Table 1 Soil parameters Decomposed granite soil

Soil parameter e0

Initial void ratio Soil particle density

(Mg/m3 )

ρs

Optimum moisture content

wopt

Maximum dry density (Mg/m3 )

ρd max

Average particle size (mm)

D50

Uniformity coefficient

Uc

0.448 2.65 10.8 1.92 1.10 21.4

Matsuo [4] 0.940 2.62 15.9 1.77 0.420 65.0

3.2 Observation and Measurement Results Picture-1 shows the imaging results. In Picture-1, the area is divided into seven plots, but in this paper, only the Unprotected areas (the first Unprotected soil-1 from the left, the fourth Unprotected soil-2) are discussed. Picture-1 (a) is at 1:35 a.m. on June 4, 2021, and (b) is at 2:59 am. Gully erosion (circled area) occurred at Unprotected soil-1 for about 2:00 a.m. to 3:00 a.m. on June 4, 2021. Figure 3 shows the relationship between precipitation and water ratio for Unprotected soil-1 and Unprotected soil-2 from 9:00 a.m. on June 3 to 9:00 p.m. on June 4, 2021. The maximum precipitation of 32 mm was observed at 2:00 a.m. on the 4th, and the maximum water ratio of 45.1% was recorded at 4:00 a.m. on Unprotected soil-1. The maximum water ratio reached 45% due to cracking caused by erosion of slope and an increase in the void ratio. Also, it found the water ratio fluctuates with precipitation, and gully erosion occurs when the water ratio begins to increase rapidly.

Picture 1 Imaging result

274

H. Asada and K. Araki

Fig. 3 Relationship between precipitation and water ratio

4 Comparison and Discussion of Safety Factor and Slope Failure Probability To examine the validity of the analysis method, slope failure risk was analyzed and evaluated based on data from a field experimental slope on Unprotected soil-1 where a small-scale collapse occurred on June 4, 2021, at approximately 2:00 to 3:00 a.m.

4.1 Focus on Water Ratio Figures 4 and 5 show the variation of the safety factor and slope failure probability over time for gradient of slope 30°, 40°, and 50°. In Fig. 4, the safety factor decreased as the water ratio increased, reaching 0.759 at a gradient of slope 30° around 3:00 a.m. when the erosion of slopes occurred. In Fig. 5, an increase in the probability of collapse was observed from around 2:00 a.m. to 5:00 a.m. on the 4th. The results of the analysis show the dangerous value at the time of the erosion, suggesting the validity of the analysis method. Also, the amount of change in both the safety factor and the failure probability is greater for slopes with a slope gradient of 30°–40° than for slopes with a slope gradient of 40°–50°. Therefore, precipitation and soil moisture have a significant effect on slopes with slopes between 30° and 40°.

4.2 Focus on Gradient of Slope The analysis was performed using a water ratio of 43.19% on Unprotected soil-1 at 3:00 a.m. on June 3, from a gradient of slope 5°–60°. Figure 6 shows the relationship between gradient of slope and safety factor. The gradient of slope decreased significantly from 0° to 20°, and the safety factor was less than 1 at 20°. Also, safety factor tended to converge above 30°. Figure 7 shows the relationship between gradient of

Development and Geographical Evaluation of Slope Failure Risk Index …

275

Fig. 4 Variation in safety factor

Fig. 5 Variation in slope failure probability

slope and slope failure probability. In Fig. 7, the gradient of slope increased sharply above 10°. From these results, we thought that the slope failure probability is a more sensitive response than the safety factor for steep slopes of 30° or more. Fig. 6 Relationship between gradient of slope and safety factor

276

H. Asada and K. Araki

Fig. 7 Relationship between gradient of slope and slope failure probability

5 Geographical Evaluation Studies The gradient of slope from the Fundamental Geospatial Data was used; the surface layer thickness was set to 1.1 m [7] and the internal friction angle to 25° [8], and the slope was analyzed and mapped in GIS according to Matsuo’s representative values (shown in Table 1). Figure 8 shows the distribution of the safety factor and Fig. 9 shows the distribution of the slope failure probability. Figures (a) through (d) in Figs. 8 and 9 show the results of analysis of water ratio from 25 to 40% in 5% increments. From Figs. 8 and 9a, the slope failure probability begins to increase when the safety factor is less than 1.5. Figures 8 and 9(d) show that the slope failure probability increases significantly when the safety factor approaches 1. The response of the slope failure probability is more sensitive in the danger zone around a safety factor of 1. Therefore, the slope failure probability is suitable for evaluation in the danger zone around the safety factor of 1.

6 Summary In this study, we attempted to understand real phenomena using online measurement devices and geographical evaluation of steep slopes using GIS, with an awareness of the introduction of IoT technology, for the development of a community-based landslide prediction system. First, a field observation slope was constructed at Tokuyama National College of Technology, and a small-scale slope failure was imaged by an observation camera on June 4, 2021, from around 2:00 am to 3:00 am. We measured precipitation and soil moisture (water ratio) and found that a rapid increase in water ratio above 30% was a possible cause of the slope failure. Second, we developed slope failure risk index (safety factor and slope failure probability) based on the method of Matsuo [4]) assuming changes in cohesion and evaluated the health of the surrounding slopes by GIS using the representative values of Matsuo [4]. The validity of the analytical method was examined by using field experiment data in the analysis, and it was considered that the slope failure probability was an indicator that showed a more significant response on steep slopes than the safety factor.

Development and Geographical Evaluation of Slope Failure Risk Index …

277

Fig. 8 Safety factor distribution

It is necessary to continue to collect data through field observations and field experiments. It is also necessary to establish a method for estimating the probability of slope failure based on precipitation and soil moisture, predicting the time of failure, and evaluating the scale of the failure.

278

H. Asada and K. Araki

(a) water ratio 25%

(b) water ratio 30%

(c) water ratio 35%

(d) water ratio 40%

Fig. 9 Slope failure probability distribution

Acknowledgements The research was supported by JSPS Grant-in-Aid for Scientific Research 21K04263. The authors would like to thank Professor Shunji Ue, Associate Professor Keiji Kuwajima, and Head of Technology Yasushi Fukuda of National Institute of Technology, Tokuyama College for their invaluable cooperation in conducting this research. The authors would like to express their gratitude to them.

References 1. MLIT Homepage. https://www.mlit.go.jp/river/sabo/jirei/r3dosha/r3doshasaigai.pdf. Last accessed 21 Oct 2022.(In Japanese) 2. MLIT Homepage: https://www.mlit.go.jp/river/sabo/jirei/r2dosha/r2doshasaigai.pdf. Last accessed 21 Oct 2022.(In Japanese) 3. MLIT Homepage. https://www.mlit.go.jp/mizukokudo/sabo/content/001510225.pdf. Last accessed 21 Oct 2022.(In Japanese) 4. Matsuo M (1984) Reliability in geotechnical engineering design, pp 232–235.(In Japanese) 5. Matsuo M, Ueno M (1978) Study on reliability-based design for prevention of slope failure. Proc Jpn Soc Civil Eng 276:77–87.(In Japanese) 6. GSI Homepage. https://fgd.gsi.go.jp/download/menu.php. Last accessed 21 Oct 2022.(In Japanese) 7. Osanai N, Tomita Y, Akiyama K, Matsusita T (2009) Realty of cliff failure disaster. Techn Note Natl Inst Land Infrastructure Manage 530:75.(In Japanese)

Development and Geographical Evaluation of Slope Failure Risk Index …

279

8. Yamaguchi Prefecture Shunan Civil Engineering Office: Report on the 2010 Landslide Disaster Precaution Area, etc., Foundation Survey Commissioned Work Zone No. 5 (Kume, Shunan City), 1–4 (2011).(In Japanese)

Drought and Management Approach for Sustainability of Tropical Reservoir Ecosystem: A Case Study of Ubolratana Reservoir, the Most Productive Reservoir in the Northeast Thailand Jeeraya Muangsringam

and Charumas Meksumpun

Abstract Occurrences of drought disasters in tropical reservoir ecosystem are commonly due to limitations of water inflow that are related to impacts from climate changes. Research on integration of fifty-five years of information on the Ubolratana reservoir, the most productive reservoir in the northeast Thailand, indicated high fluctuated levels of inflows and storage volumes. In particular, the storage volumes during 2019–2020 were lower than the “Minimum Storage Level” determined for reservoir’s sustainability. Such phenomenon reflects severe drought crisis in the reservoir area. Results implied that changes in the inflow had remarkable effect on economically important fishery resources. Results of the drought periods also indicated that quantities of organic suspended solids of reservoir water had decreased and may consequently influence biological production potentials of the reservoir ecosystem. In addition, decreased inflow and water storage contributed to the increased water retention time and, accordingly, enhanced frequent increases of potentially toxic blue green algae, in particular in the lacustrine zones. Overall, the research implied that drought disaster can deteriorate the quality of foods and habitats for aquatic living and, thus, decrease stability of ecosystem production. In management coping with such conditions, it is necessary to accelerate upstream rehabilitation in order to promote sufficient water volume. The inflow should be enhanced appropriately, while the outflow should be carefully managed, for conservative management of the reservoir resources. Proactive measures for habitat and water quality remediation should be developed and evaluated further. Keywords Drought · Reservoir Ecosystem Sustainability · Hydrological Impact · Proactive Management

J. Muangsringam Major of Fishery Science, Kasetsart University, Bangkok 10900, Thailand C. Meksumpun (B) Faculty of Fisheries, Kasetsart University, Bangkok 10900, Thailand e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_26

281

282

J. Muangsringam and C. Meksumpun

1 Introduction Occurrences of drought disasters in tropical reservoir ecosystem are commonly due to limitations of river inflow that are related to impacts from climate changes. Climate change can recently increase the variability, frequency, and severity of impacts from both flood and drought phenomena [1]. Climate change can also alter hydrological and ecological regimes, including their fluctuations of related resource productions. Hydrological changes can result in low water volume within a reservoir as well as reduced inflow rates and can influence biological processes in the reservoir ecosystem. In general, simultaneous increases in temperature and nutrients during droughts can contribute to enhance phytoplankton and zooplankton abundance through bottom-up mechanisms [2–4]. Reservoir plankton community can change predictably with water resource quality and respond remarkably to variations in the environment [5]. In some lacustrine areas, historically low levels of water were associated with remarkable changes in zooplankton community [6, 7]. In addition, fish may be smaller and more omnivorous with increasing temperatures [8]. Depending on future eco-hydrological changes, fish population may considerably deteriorate in impacted reservoir ecosystem. Thus, understanding of mechanism and key drivers of spatial and temporal variabilities will be essential to managing reservoir ecosystems during the period of drought stress. In many tropical countries, reservoir fishery productions are an important source of protein-rich food for local communities [9–11]. Nevertheless, the fishery productions are negatively affected by a decrease of water storage and an increase of water temperature, both of which potentially affect the survival and reproductive strategies of aquatic animals [12]. The Ubolratana reservoir is the second-largest reservoir in Thailand. It has a maximum surface area of 410 km2 and a storage volume of 4640 million cubic meters [13]. It is a relatively shallow (average depth of 5.6 m), but highly productive mesotrophic to eutrophic reservoir [14, 15]. Fishery production from the Ubolratana reservoir is among the most important in northeastern Thailand, based on yield per area [16]. The maximum fish harvest from this reservoir was 2122 tons during the first 10 years of water storage. Nevertheless, the reservoir ecosystem has faced severe drought conditions since 2018. Very low precipitation and inflows have caused the water storage volume to be drastically decreased, compared to levels in the previous decade [17]. Such drought conditions have been shown to have an impact on water quality and living resources of the reservoir ecosystem. The water storage volume during November 2019 decreased to a level lower than its minimum storage criteria and had the lowest recorded volume of ca 18% [13]. Several fishery resources in Ubolratana reservoir, therefore, have been impacted and need to be conservatively managed. In this study, therefore, temporal changes in resource abundance throughout the drought years were analyzed along with related eco-hydrological factors. The present paper provides information concerning population response to the drought conditions and related environmental characteristics of the reservoir

Drought and Management Approach for Sustainability of Tropical …

283

habitats. Future conservation approaches for economically important fish stock and their sustainable productions are also discussed.

2 Methodology The Ubolratana reservoir is shallow mesotrophic reservoir [15] located in the northeastern part of Thailand (Fig. 1). The reservoir receives water from the Phong and Phaniang rivers in the northern region, and from Choen River in the southern region. The climate of the whole area is influenced by two tropical monsoons; the southwest monsoon from May to October (rainy season), and the northeast monsoon from November to April (dry season) [17]. In this study, related hydrological information (water storage volume and inflow) was reviewed [13]. Water storage volumes of the reservoir have apparently fluctuated. During 2018–2020, the water storage volume was lower than the prescribed “minimum storage volume” (580 million m3 ) of the reservoir. Contrastingly, during 2021, the water storage volume remarkable increased to the level of 110% of reservoir capacity (Fig. 2). In addition, related hydrological data were collected: rainfall was from the Loei Air Monitoring Station [17]. Related secondary information was gathered from various scientific documents reported during the fifty-five years of the reservoir operation. Information from the integrated research program on “Synthesis of Reservoir Management System for Aquatic Resource Conservation and Sustainable Utilization: A Case Study of Ubolratana Reservoir” carried out during 2017–2019 was also simultaneously analyzed and synthesized for appropriate management approach, together with critical analysis of fish capture information throughout the operation years [16]. Descriptive statistics (means and standard deviations) were applied to depict the changes over reservoir operation periods. Differences in parameters among periods

Fig. 1 The Ubolratana reservoir is located in Khon Kaen province in the north-eastern part of Thailand. The reservoir can be divided into riverine, transition, and lacustrine zones along the distances from the major two river inlets; the Phong and Choen River channels

284

J. Muangsringam and C. Meksumpun

Fig. 2 Water storage volumes of the Ubolratana reservoir during 1972–2021 (modified from [13]). (The upper and lower grey lines indicate the prescribed the “maximum storage” and the “minimum storage” volumes of the reservoir)

and locations were evaluated by paired sample T-test (p < 0.05). Correlations between densities of aquatic resources including fish captures and environmental parameters were analyzed using Spearman’s rank correlation coefficient (p < 0.05) (SPSS version 27).

3 Results and Discussions During 1972–2021, the results of inflows from the major river inlet, the Phong River channel, indicated annual fluctuations of the inflow levels (Fig. 2). Among the periods from the start of reservoir operation until recent years, the drought phenomena have revealed to occur several times with the interval of about 8–10 year. Nevertheless, severe drought conditions had occurred during the recent years (2018–2020). During the severe droughts, the water storage volume was lower than the prescribed “minimum storage volume” (580 million m3 ) of the reservoir.

Drought and Management Approach for Sustainability of Tropical …

285

In this study, monthly variations of the inflow were comparatively analyzed with the water storage volumes of the reservoir ecosystem. During the rainy season (May to October), the increases in inflow were noticed, in particular during mid rainy season (May to July) (Fig. 3). Nevertheless, the water storage volumes were later increased, during late rainy season (September to October). During continuous drought periods of 2018–2020, the lowest levels of the water storage volume were examined. Along those periods, marginal and submerge vegetation in coastal zones of the reservoir ecosystem were remarkably deteriorated and died out due to long-term exposure to sunlight. Decreased inflow can limit the increase of local fish population, in particular during the spawning period of migratory fish species [11]. Analysis of fish catches during the past five decades of the Ubolratana reservoir implied a gradual decrease of captures with the rates of about 1.5% per year [18]. In the reservoir ecosystem, the dominant fish species were examined to be major groups of silver barb, River sprat, Smith’s barb, and Mud carp. Catches of these fish groups during the early rainy season (May to July) of 2017, the year with comparatively high storage volume, were reported for the whole reservoir to be 80,487 kg, 72,010 kg, 65,456 kg, and 57,303 kg, respectively.

Fig. 3 Monthly levels of inflows and water storage volumes of the Ubolratana reservoir during 2014–2021

286

J. Muangsringam and C. Meksumpun

During the early rainy season (May to July) of 2017, the air temperature and rainfall were reported to be 27.5–28.7 C and 563–663 mm, respectively. Thereafter, the drought conditions had been occurring. In the drought years of 2018–2019, the air temperature and rainfall during the early rainy season were reported to be 28.5–30.3 C and 213–452 mm, respectively [13, 17]. After the severe drought had passed, the fish catches during 2020 decreased 3–8 times to 10,940 kg, 12,798 kg, 16,608 kg, and 17,223 kg, respectively. During 2020, some occurrences of Tilapia species were also examined. The results indicated that not only the quantitative decreases of economically important fish species, the qualitative changes in fish composition had occurred. Such changes should be carefully considered for development of suitable conservative management. In this study, relationships between the inflow and fishery productions during 2016–2020 were analyzed. The quantities of silver barb (SVB) and Mud carp (MUC) catches during the early rainy season (May to August) had significant relationships with the inflows (INF), as the equations; SVB = 88 INF+5795 (r = 0.76) and MUC = 53 INF+8705 (r = 0.81), respectively. Such findings imply that the level of monthly inflow of > 25 million m3 should be maintained in order to enhance stability of fishery resources. Results also indicated that changes in water inflow had remarkable effect on suspended solids (Fig. 4). During the years 2020–2022 after severe drought, the quantities of organic matters in suspended solids of reservoir water were decreased, in particular in the riverine zone of the reservoir ecosystem. Apparent decreases during the rainy season (Fig. 4; September) can consequently influence physicochemical characteristics of the water habitats and, thus, impact on production potentials of migratory fishes in the reservoir ecosystem. In addition, decreased inflow and water storage had increased water retention time and, thus, induced frequent increases of blue green algae, in particular in the lacustrine zones of reservoir ecosystem. Drought condition in reservoir can also

Fig. 4 Temporal changes of total organic suspended solids (Org-TSS) of the riverine, transition, and lacustrine zones of the Ubolratana reservoir during 2020–2022

Drought and Management Approach for Sustainability of Tropical …

287

lead to akinetes formations in the heterocystous cyanobacteria and can stimulate potentially toxic cyanobacteria blooms [19, 20]. During the drought through 2018– 2020 of the Ubolratana reservoir ecosystem, the densities of phytoplankton also greatly fluctuated. Densities of potentially toxic cyanobacteria were examined to increase when the water levels were low [15], possibly due to suitable conditions of increased ammonium concentrations and temperature. The highest total density occurred during dry season in the transition zone. Nevertheless, in the riverine and transition zones of the reservoir, density of phytoplankton had significant negative relationships with inflow and total suspended solids. The overall views from the previous studies also implied the need to monitor risk condition of the water resources from the increasing abundance of potentially toxic cyanobacteria during continuous drought conditions [15]. Moreover, the increases in inflow and suspended solids in particular areas during the rainy season may contribute to decrease potentially toxic cyanobacteria. Oppositely, the decreases in the inflow and suspended solids can enhance the blooming of the potentially toxic cyanobacteria in the reservoir ecosystem. The functions of such hydrological and water quality factors have been emphasized.

4 Conclusion Overall, the research indicated that problem of drought disaster can decrease stability of ecosystem production and, thus, deteriorate the quality of foods for aquatic living. The research also explained functions of hydrological factors and water quality that impact on fishery resources. The whole results implied crucial need to control hydrological drivers of the reservoir for preventing deterioration in production during such environmental conditions. For management coping with such conditions, it is necessary to accelerate upstream rehabilitation to promote sufficient water volume. In addition, the outflow should also be carefully managed. Proactive measures for habitat and water quality remediation should be further developed and evaluated with fish production information. Acknowledgements This study was supported by the National Research Council of Thailand, the Graduate School of Kasetsart University, the Kasetsart University Research and Development Institute, and the Department of Fishery Biology, Faculty of Fisheries, Kasetsart University. Deep thanks are also due to all members of the Sediment and Aquatic Environment Research Laboratory, Department of Fishery Biology, for their kind help in field surveys. The authors acknowledge Dr. Santi Poungcharean for his kind recommendation in fish population analysis. Comments from anonymous reviewers were much appreciated.

288

J. Muangsringam and C. Meksumpun

References 1. Naz BS, Kao SC, Ashfaq M, Gao H, Rastogi R, Gangrade S (2018) Effects of climate change on streamflow extremes and implications for reservoir inflow in the United States. J Hydrol 556:359–370 2. Wolfinbarger WC (1999) Influences of biotic and abiotic factors on seasonal succession of zooplankton in Hugo reservoir, Oklahoma, U.S.A. Hydrobiologia 400:13–31 3. Feuchtmayr H, Moss B, Harvey I, Moran R, Hatton K, Connor L, Atkinson D (2010) Differential effects of warming and nutrient loading on the timing and size of the spring zooplankton peak: an experimental approach with hypertrophic freshwater mesocosms. J Plankton Res 32:1715–1725 4. Striebel M, Singer G, Stibor H, Andersen T (2012) Trophic overyielding: phytoplankton diversity promotes zooplankton productivity. Ecology 93(12):2719–2727 5. Stamou G, Kassipai M, Moustaka-Gouni M, Michaloudi E (2021) The neglected zooplankton communities as indicators of ecological water quality of Mediterranean lakes. Limnetica 40(2):359–373 6. Havens KE, East TL, Beaver JR (2007) Zooplankton response to extreme drought in a large subtropical lake. Hydrobiologia 589:187–198 7. Winder M, Reuter JE, Schladaw SG (2009) Lake warming favors small-sized planktonic diatom species. In: Proceedings of the royal society B. Royal society, United Kingdom, pp 427–435 8. Jeppesen E, Meerhoff M, Holmgren K, González-Bergonzoni I, Teixeira-de Mello F, Declerck SAJ (2010) Impacts of climate warming on lake fish community structure and potential effects on ecosystem function. Hydrobiologia 646:73–90 9. Guo Z, Li Z, Liu J, Zhu F, Perera HACC (2011) Status of reservoir fisheries in China and their effects on environment. In: Tropical and sub-tropical reservoir limnology in China. 2nd edn. Springer, Netherlands 10. Amarasinghe US, De Silva SS (1999) Sri Lankan reservoir fishery: a case for introduction of a co-management strategy. Fish Manage Ecol 6:387–399 11. Asawamanasak P, Meksumpun C (2022) Impacts of rainfall and related hydrological factors on spawning characteristics of Silver Barb (Barbonymus gonionotus Bleeker, 1849) in a tropical reservoir ecosystem. J Fisheries Environ 46(1):1–12 12. Tundisi JG, Matsumura-Tundisi T (2012) Limnology. Braz J Biol 69(1):395–439 13. Electricity Generating Authority of Thailand (EGAT) http://watetele.egat.co.th/ubolratana/. Last accessed 1 Aug 2022 14. Muangsringam J, Meksumpun C, Whanpetch N, Poungcharean S, Sangmek P (2021) Changes in the Asian clam population in a tropical mesotrophic reservoir during severe drought. J Fisheries Environ 45(3):1–13 15. Mengchouy R, Meksumpun C (2022) Spatial and temporal dynamics of water quality and potentially toxic cyanobacteria during drought conditions in a mesotrophic reservoir ecosystem. J Fisheries Environ 46(2):1–14 16. Department of Fisheries (DOF) (2022) Annual reports of 2018–2022. Total Fish Production of Khon Kaen province. Department of Fisheries, Bangkok, Thailand 17. Thai Meteorological Department (TMD). https://www.tmd.go.th/index.php. 2022. Last accessed 10 June 2022 18. Meksumpun C, Mengchouy R, Buakaew K, Nimsantijaroen W (2019) The Ubolratana reservoir and meteorological changes during 50 years: trend and risk of air temperature on fishery resources. In: 8th Phayao research conference 2019. Phayao University, Phayao, pp 1429–1441 19. Mowe MAD, Mitrovic SM, Lim RP, Furey A, Yeo DCJ (2015) Tropical cyanobacteria blooms: a review of prevalence, problem taxa, toxins and influencing environmental factors. Limnology J 74(2):205–224 20. Moura ADN, Aragão-Tavares NKC, Amorim CA (2018) Cyanobacterial blooms in freshwater bodies from a semiarid region, Northeast Brazil: a review. J Limnol 77(2):179–188

Evaluating Liquefaction Strength Prediction Method for Low Improvement Ground Materials Koji Yamamoto, Kenichi Sato, Takuro Fujikawa, and Chikashi Koga

Abstract A limit exists to what can be captured by extending the current liquefaction strength evaluation methods in the case of a massive earthquake such as the Great East Japan Earthquake. Therefore, developing new technologies for predicting and countermeasures against liquefaction damage based on instead of new concepts is essential. One of the current liquefaction prediction methods is the FL method. This method requires liquefaction strength determined by an undrained cyclic triaxial test. However, environmentally friendly low-modification liquefaction countermeasures with low cement addition, such as the premixing treatment method, are primarily used in reclaimed port areas, and cement is commonly used in ground improvement methods. However, it is challenging to determine the liquefaction strength of ground materials that have been improved and solidified over time, due to the loading capacity of a typical undrained cyclic triaxial test apparatus. Therefore, Koga et al. (Abstracts of Master’s Thesis, Chuo University, pp 1–4 (2013)) investigated a method of determining the liquefaction strength of low-improved ground materials using the cone index test, a simple method of measuring liquefaction strength. This paper reports the results of a study on a prediction method to determine the strength of sandy materials with different grain sizes after cementation and stability based on the liquefaction strength of low-mixing and early-age materials. Keywords Cone Index · Undrained cyclic triaxial test · Liquefaction Strength Prediction

1 Introduction In recent years, the damage caused by natural disasters has steadily increased, and the disasters have become increasingly vast and complex due to climate change caused by global warming and increased seismic activity. Kazama et al. [2–4] propose that liquefaction damage caused by huge earthquakes, such as the Great East Japan K. Yamamoto (B) · K. Sato · T. Fujikawa · C. Koga Department of Civil Engineering, Fukuoka University, Fukuoka 814-0180, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_27

289

290

K. Yamamoto et al.

Earthquake, has limitations in being captured by an extension of current liquefaction strength assessment methods. Therefore, developing technologies for predicting and counteracting liquefaction damage based on new concepts is essential. In line with national land resilience, low-modification liquefaction countermeasures with low cementitious solidifier additions, typified by premixing treatment methods, are used specifically for reclaimed land in harbor areas. The liquefaction strength of such ground-improved materials is typically assessed by consolidated undrained shear tests [5] in the laboratory. However, the liquefaction strength of geomaterials that solidify over time is challenging to determine easily due to the testing equipment’s loading capacity, and other methods must be considered. Therefore, Koga et al. [1] investigated a method of determining the liquefaction strength of low-improved ground materials using the cone index test, a simple method of measuring liquefaction strength. This paper reports the results of a study on a prediction method to determine the strength of sandy materials with different grain sizes after cementation and stabilization based on the liquefaction strength of low-mix and young early-age materials.

2 Experimental Study 2.1 Experimental Sample and Specimen Fabrication Methods This study’s soil samples were two types of decomposed granite soils with a grain size range of less than 2 mm and silty sand. Figure 1 shows the particle size distribution of the three types. (Hereafter referred to as decomposed granite soil A, cut to less than 75 µm, and decomposed granite soil B, which contained less than 75 µm.) Ordinary Portland cement was used as the solidifying material. Table 1 shows the experimental conditions. Various amounts of cement were added to each sample’s dry mass. Each experiment’s specimens were prepared at w = 10% and compacted by the tamping method (5 layers) in a prescribed mold (diameter 7.5 cm, height 15 cm) to achieve a relative density of 60%. The specimens were cured air-cured at a constant temperature of 20 °C. If the specimens were not sufficiently solidified, they were frozen for one day.

2.2 Testing Procedure Undrained Shear Test (JGS 0523) Undrained shear tests [5] were conducted using sinusoidal stress control at a loading rate of 0.1 Hz. Liquefaction was determined when the bialtitude axial strain (DA = 5%) was reached, and the specimens’ liquefaction strength ratio, RL, was the cyclic stress amplitude ratio, σ d /2σ  0 , corresponding to several cycles (N = 20). The

Evaluating Liquefaction Strength Prediction Method for Low … 0.075 silt

0.250

fine sand

Percentage passing (%)

100

0.850

2

291

4.75

medium coarse fine sand sand gravel

19 medium gravel

75 coarse gravel

Decomposed granite soilA (2mm 75 m) Decomposed granite soilB (2mmbelow) Silica sand

80 60 40 20 0 0.001

0.01

0.1 1 Particle size (mm)

10

100

Fig. 1 Grain size accumulation curve

Table 1 Specimen preparation conditions Soil samples

Relative density (%)

Moisture content (%)

Cement mixing ratio (%)

Curing days

Decomposed granite soil A

60

10

0,0.5,0.7,1.0,1.2,1.5,2.0

3,7,14,28

Decomposed granite soil B

60

10

0,0.5,1.0,1.5,2.0,2.5

7,14,28

Silica sand

60

6.2

0,2.0,2.5,3.0

7,14,28

pore water pressure coefficient B values of more than 0.96 were confirmed for all conditions, and the study was conducted under sufficiently saturated conditions. Cone Index Test (JIS a 1228) Cone penetration tests [6] were conducted in a mound with a diameter of 10 cm and a height of 12 cm under the conditions shown in Table 1, adjusted to a moisture content ratio of w = 10%, and prepared by the tamping method (three layers) by thrusting to achieve a relative density of 60%. Cone penetration tests were conducted using a cone penetrometer with a tip base area of 3.24 cm2 to determine the cone index, qc , from the average penetration value resistance at penetration volumes of 5, 7.5, and 10 cm when the specimen was penetrated at a speed of 1 cm/s. The improvement effect was evaluated from the average values of three tests in each condition. The improvement effect was evaluated from the average of the three tests in each condition.

292

K. Yamamoto et al.

3 Results and Discussions 3.1 Cyclic Shear Properties of Cement Improved Soil Figure 2a–c show the effective stress path diagrams for decomposed granite soil A after 14 days of curing at three cement mixing rates (C = 0.5, 1.0, and 2.0%). As the amount of cement mixing to the improved soil increased, the cement solidification effect occurred, and the cyclic shear resistance increased. For the cement mixing rate C = 2.0% after 14 days of curing, at cyclic stress (CSR) (σ d /2σ  0 ) = 0.48, the limit of the cyclic loading capacity used in this experiment, the effective stress was insignificant during loading due to the effect of the specimen’s cement solidification. The effective stress does not decrease during the loading process due to the specimen’s solidification by mixing the cement, indicating that liquefaction has not occurred. Figure 3a–c show the effective stress paths of decomposed granite soil B after seven days of curing at cement mixing rates of C = 0, 1.0, and 2.5%. Here the number of curing days for decomposed granite soil B is different from that of decomposed granite soil A because of the effect of grain size and the specimens’ different solidification development. As the cement mixing rate increases, the cement solidification effect occurs, and the cyclic shear resistance increases. Liquefaction is reached after 28 days of curing at a cement mixing rate of C = 1.0%. Figure 4a–c show the effective stress path diagrams after seven days of curing for three conditions with silica sand at cement mixing rates of C = 1.5, 2.0, and 3.0%. Here, grain size affects the silica sand, and the effective stress path diagrams for comparison differ from those of decomposed granite soils A and B because the specimens’ solidification occurred differently from those of decomposed granite soils A and B. Similarly, silica sand shows an increase in cyclic shear resistance due to the solidification effect of cement as the amount of cement added increases. Figure 5 shows the relationship between axial strain and deviator stress at 14 days of curing for three cement addition rates (C = 1.0, 1.5, and 2.0%) using decomposed granite soil A. Shear deformation is suppressed as curing, and the solidification progresses with the increasing cement mixing rate, making it challenging to deform 100

-50 -100

C=0.5% t=14day CSR=0.24

0

20 40 60 80 100 2 120 Mean effective stress p'(kN/m )

(a) C=0.5%

2

2

Deviator stress q (kN/m )

0

Decomposed granite soil A

Decomposed granite soil A

50

50

c

c

c

50

100

Deviator stress q (kN/m )

Decomposed granite soil A

2

Deviator stress q (kN/m )

100

0 -50 -100

C=1.0% t=14day CSR=0.34

0

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

(b) C=1.0%

Fig. 2 Effective stress path diagram for decomposed granite soil A

0 -50 -100

C=2% t=14day CSR=0.48

0

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

(c) C=2.0%

Evaluating Liquefaction Strength Prediction Method for Low … 100

50

0 -50 -100

C=0% t=0day CSR=0.12

0

50

c

0 -50 -100

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

decomposed granite soil B

2

Deviator stress q (kN/m )

2

decomposed granite soil B

c

c

50

100

Deviator stress q (kN/m )

Decomposed granite soil B

2

Deviator stress q (kN/m )

100

293

C=1.0% t=7day CSR=0.26

0

(a) C=0%

0 -50 -100

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

C=2.5% t=7day CSR=0.50

0

50 100 150 200 2 Mean effective stress p'(kN/m )

(b) C=1.0%

(c) C=2.5%

Fig. 3 Effective stress path diagram for decomposed granite soil B

100

100

0 -50 -100

C=1.5% t=7day CSR=0.32

0

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

(a) C=1.5%

2

Silica sand

50

c

50

c

c

50

Deviator stress q (kN/m )

Silica sand

2

Deviator stress q (kN/m )

Silica sand

2

Deviator stress q (kN/m )

100

0 -50 -100

C=2.0% t=7day CSR=0.38

0

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

(b) C=2.0%

0 -50 -100

C=3.0% t=7day CSR=0.45

0

20 40 60 80 100 120 2 Mean effective stress p'(kN/m )

(c) C=3.0%

Fig. 4 Effective stress path diagram for silica sand

up to DA = 5% double amplitude axial strain, determining the liquefaction. Therefore, attention was focused on the cone penetration test to determine the strength by other methods, which can determine the strength of the ground in a simplified manner.

Fig. 5 Axial strain and axial differential stress of decomposed granite soil A

294

K. Yamamoto et al.

3.2 Liquefaction Strength of Sand with Cement Modification Figure 6 shows the liquefaction strength curves for decomposed granite soils A and B (DA = 5%) at a cement mixing rate of C = 1.0% for different curing days. The liquefaction strength of both decomposed granite soils increased as curing days progressed, with a significant increase in strength at 28 days. Figure 7 shows the relationship between the number of curing days and liquefaction strength of decomposed granite soil A at cement mixing rates of C = 0, 0.5, 0.7, 1.0, 1.5, and 2.0%. The liquefaction strength increased with the increasing cement mixing rate and curing days, confirming the solidification effect of cement modification. The liquefaction strength at a cement mixing rate of C = 2.0% can be determined up to seven days of curing, but it is challenging to determine after 14 days. Thus, using the standard cyclic undrained shear test, liquefaction determination is no longer possible in the strength assessment of low-improvement liquefaction control methods for sandy soils using cementitious solidifiers.

Fig. 6 Liquefaction strength curves for each total number of cure days (DA = 5%) Fig. 7 Liquefaction strength of decomposed granite soil A at different cement mixing rates (DA = 5%)

Evaluating Liquefaction Strength Prediction Method for Low …

295

Fig. 8 Relationship of cone index to the number of curing days

3.3 Liquefaction Prediction Methods Using Cone Index Tests Cement mixing and curing solidification made it challenging to determine the liquefaction strength due to the limits of the loading stress of the undrained shear test machine. Therefore, we focused on the cone index test, which can determine the strength of materials simply, and evaluated the liquefaction strength by the penetration resistance value. Figure 8 shows the relationship between the number of curing days and the cone index for decomposed granite soils A and B and silica sand. The cone index of both decomposed granite soils increased with the curing days and the cement addition rate due to the onset of the solidification effect. Therefore, Fig. 9a–c summarize the results of the relationship between the cone index and liquefaction strength for all curing days for both amplitude axis strains to estimate the liquefaction strength after the solidification stabilization process (3, 7, 14, and 28 days) using the cone index. The results are summarized for DA = 1, 2, and 5%, respectively. An excellent correlation exists between the liquefaction strength and cone index for two decomposed granite soils of different grain sizes and silica sand of different materials, irrespective of the deformation of both amplitude axis strains. This relationship indicates that liquefaction strength can be estimated from penetration resistance values performed under identical conditions. For example, the liquefaction strength at a cement mixing rate of C = 2.0% and DA = 5% at 28 days of curing, for which the liquefaction strength could not be determined this time, is the value of the cone index (qc = 11,200 kN/m2 ), the value of the cone test results shown in Fig. 8. Therefore, from Fig. 9c, the liquefaction strength is approximately 0.48. A liquefaction strength of more than 0.5 is considered good. Therefore, further studies will show a simplified estimation method based on the relationship between liquefaction strength and cone values for young timber.

296

K. Yamamoto et al.

(a)

(b)

(c)

Fig. 9 Relationship between liquefaction strength and cone index for all curing days

4 Conclusion 1. Low-modified decomposed granite soils and silica sand with a cement mixing rate of C = 2.0% could not be sufficiently evaluated for liquefaction strength due to the loading limits of the experimental apparatus as they were cured. 2. The relationship between the liquefaction strength and cone penetration resistance of decomposed granite soil and silica sand correlated well, regardless of the deformation magnitude of both amplitude axial strains of cyclic shear for each type of decomposed granite soil and different materials. 3. The possibility of assessing the liquefaction strength of cement-improved ground using the cone index test. The study revealed the possibility of using the v-cone index test to assess the liquefaction strength of cement-improved soils.

References 1. Koga Y (2013) Effective confining pressure in the relationship between liquefaction strength and cone penetration resistance of fine-grained sand. Abstracts of Master’s Thesis, Chuo University, pp 1–4 2. Kazama M et al (2015) Issues in liquefaction research as seen in liquefaction damage from the Great East Japan Earthquake. Trans Japan Soc Earthq Eng 15(7) (Special Issue). Author F, Author S (2016) Title of a proceedings paper. In: Editor F, Editor S (eds) Conference 2016, LNCS, vol 9999. Springer, Heidelberg, pp 1–13 3. Kazama M (2010) Geotechnical disaster by the 2011 off the Pacific coast of Tohoku Earthquake and geotechnical issues for reconstruction. In: Proceedings of the technical conference on the great East Japan Earthquake, Damage and lessons learned from the huge earthquake and huge tsunami, pp 41–65. Author, F (2010) Contribution title. In: 9th international proceedings on proceedings. Publisher, Location, pp 1–2 4. Kazama M (2011) Overview of the 2011 Tohoku earthquake damage and geotechnical engineering issues. Geotech Eng J 7(1):1–11 5. Undrained Shear Test (JGS 0523) (2015) Method for unconsolidated-undrained triaxial compression test on soils with porewater pressure measurements (JGS 2015) 6. JIS A 1228 (2020) Cone index test method for compacted soil

Landslide Analysis Using Geographic Information System (GIS) in South Tapanuli Regency Ika Puji Hastuty

and Fauziah Ahmad

Abstract Landslide is a natural disaster that often occurs in Indonesia. Landslide is defined as a movement of soil or rock substances, or a mixture of the two substances that move down out of the slopes which is caused by stability disturbances of soil and rock making up the slopes. Losses that can be incurred include loss of life, damage to residences, loss of property, and psychological impacts. In this study, the development of landslide-prone area map was carried out in South Tapanuli Regency based on the estimation model of Pusat Penelitian Tanah dan Agroklimat, Department of Agriculture (2004). The parameters used in this study were rainfall, rock type, slope, land cover, and soil type. Scoring and overlapping method was used to obtain landslide susceptibility levels by using QGIS 3.24 application. According to the study results, there were 4 landslide susceptibility levels in South Tapanuli Regency, namely low landslide susceptibility (142.0093 km2 ), moderate landslide susceptibility (1528.4674 km2 ), high landslide susceptibility (2314.9484 km2 ), and very high landslide susceptibility (379.5909 km2 ). There were 3 districts dominated by moderate landslide susceptibility levels, namely Aek Bilah, Batang Angkola, and Muara Batang Toru districts. Then, there were 10 districts classified in high landslide susceptibility level, namely West Angkola, South Angkola, East Angkola, Angkola Sangkunur, Arse, Marancar, Saipar Dolok Hole, Sayur Matinggi, Sipirok, and Tano Tombang Angkola. Only Batang Toru district classified in very high landslide susceptibility level. Keywords Landslide · Geographic information system · Hazard

I. P. Hastuty Universitas Sumatera Utara, Road Dr. Mansyur No. 9, Medan, Indonesia F. Ahmad (B) Universiti Sains Malaysia, Nibong Tebal Seberang Perai Selatan Penang, Malaysia e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_28

297

298

I. P. Hastuty and F. Ahmad

1 Introduction 1.1 Longsor Di Sumatera Utara Landslide is a natural disaster that often occurs during rainy season in Indonesia [1-3] According to Lembaga Badan Nasional Penanggulangan Bencana [4], there were 73 landslides occurred in North Sumatra from 2009 to 2019. North Sumatra is a province that has a high number of landslide-prone area. Based on Indeks Risiko Bencana Indonesia (IRBI) in 2020, there were 22 regencies/cities classified as high landslide-prone area and 11 districts/cities classified as moderate landslide-prone area. Therefore, North Sumatra is included as the province whose needs attention regarding its landslide potential. Landslides cause a lot of losses such as damage to housing, infrastructure, and land which hamper the economic sector. The closed access road also has an impact on hampering the economy of society around the landslide area. South Tapanuli Regency is one of the landslide-prone area located in North Sumatra. The area of South Tapanuli Regency is 4,355.35 km2 , while the height ranges from 0 to 1,985 m above sea level. Rainfall in South Tapanuli Regency tends to be irregular throughout the year which is one of the factors in the occurrence of landslides. The highest and lowest rainfall occurs in November (120 mm) and February (54 mm) [5]. According to 2020 population census, the population of South Tapanuli was estimated at 300,911 people with a population density of 69 people per km2 . The number of households was 65,253 households where each household was inhabited by an average of 4 people. This number indicated that the large number of residents affected the availability of land which had an impact on slope loading. In South Tapanuli Regency, there were 12 districts with moderate to high ground movement potential, while the other districts were on moderate level [6]. Reviewing from the aspects that have been discussed previously, it can be concluded that most of South Tapanuli Regency has a high potential for landslides. If a landslide occurs, it will have an impact on the residents around the area, especially in the economic sector, and land use. According to [7], the factors that cause landslides are geology, water, strength, geotechnic parameters, construction method, dynamic of strength, and geometric of hills.

1.2 Landslide Hazard Evaluation There are two ways that can be done to evaluate landslide hazard, namely direct observation and utilizing technological tools. Direct observation is carried out by observing and assessing potential areas such as identifying causal factors. Another way is to utilize technological tools by analyzing maps such as bedrock maps, geological maps, topographic maps, and so on using Geographic Information System (GIS).

Landslide Analysis Using Geographic Information System (GIS) …

299

Geographic Information System is a computer-based information system that is used to process and store geographic data or information. GIS, in the form of spatial data, is geographically oriented data that uses a location with a certain coordinate system as its reference basis. So that, GIS applications can display locations, conditions, trends, patterns, and modeling. This ability is what differentiates GIS from other information systems.

2 Research Methods The method used in this study was scoring and overlay method where the secondary data obtained would be processed with scoring and weighting system. This study was carried out in South Tapanuli Regency located between 0°58 35 –2°07 335 North Latitude and 98°42 505 –99°34 165 East Longitude. The mapping step used scoring and overlay method with QGIS 3.24 software. The parameters required for this research were rainfall, rock type, slope, land cover, and soil type. The data that has been scored and weighted was overlaid to get the total score which became a value that represented landslide susceptibility levels. The greater the total score, the higher the susceptibility level. The output generated from this study was a map of landslide susceptibility levels in South Tapanuli Regency, where the classification was divided into 4 classes namely: low, medium, high, and very high based on the total score. Utilization of GIS applications in landslide hazard mapping can simplify analysis work while saving time and costs. The weighting in this study referred to the estimation of Pusat Penelitian Tanah dan Agroklimat (Puslittanak) [8].

2.1 Indicator Weight Determination Each indicator has its own weight percentage, this weight will determine the landslide susceptibility levels. The level of susceptibility to landslides is divided into 3 levels, namely high susceptibility level with a score of 3 (three), moderate susceptibility level with a score of 2 (two), and low susceptibility level with a score of 1 (one). The weight value of each indicator is as follows in Table 1.

3 Result of Study The weighting in this study referred to the estimation of Pusat Penelitian Tanah dan Agroklimat (Puslittanak), Department of Agriculture in 2004. According to Puslittanak [8], there are 5 types of parameters that cause landslides, namely rainfall, rock type, slope, land cover, and soil type. Puslittanak, Department of Agriculture

300

I. P. Hastuty and F. Ahmad

Table 1 Weighting value Parameter

Classification

Weight

Skor (%)

Rainfall (mm/year)

< 1500

1

30

Rock type

Soil type

Slope angle (%)

Vegetation

1501–2000

2

2001–2500

3

2501–3000

4

> 3000

5

Alluvial

1

Sediment

2

Volcanic

3

Alluvial

1

Yellowish latosol

2

Brownish latosol

3

Andosol, podzolic

4

Regosol

5

45

5

Ponds, reservoirs, waters

1

City/settlement

2

Forests and plantations

3

Shrubs

4

Moors, paddy fields

5

20

20

20

10

(2004) stated that rainfall has the highest role in the occurrence of landslides with a weight of 30% of the total weighting. The overlay method was carried out by overlapping all the parameter layers that have been weighted, with the results of scores and the weights multiplication of each parameter as a reference. Then, the multiplication values were summed up to obtain new data, namely the total score which became the value of landslide susceptibility levels. According to Puslittanak, Department of Agriculture (2004), the estimation formula used to identify landslide susceptibility level is as follows: Skor Total = 0, 3FCH + 0, 2FBD + 0, 2FKL + 0, 2FPL + 0, 1FJT

(1)

where are FCH = Rainfall factor, FBD = Rock type factor, FKL = Slope angle factor, FPL = Landcover factor, FJT = Soil type factor. Figure 1 is the percentage of district area in South Tapanuli and Fig. 2 shows that South Tapanuli Regency has 14 districts namely Aek Bilah, West Angkola, South

Landslide Analysis Using Geographic Information System (GIS) …

301

Angkola, East Angkola, Angkola Sangkunur, Arse, Batang Angkola, Batang Toru, Marancar, Muara Batang Toru, Saipar Dolok Hole, Sayur Matinggi, Sipirok, and Tano Tombang Angkola. Rainfall was a parameter that causes landslides which has the highest weight percentage of 30%. The rainfall distribution map was sourced from the North Sumatra Bappeda Geoportal. From Fig. 3, it can be seen that South Tapanuli Regency was dominated by moderate rainfall (2001–2500 mm/year) with 59.51% of South

Fig. 1 Pie chart presentase per district

Fig. 2 Administration map

302

I. P. Hastuty and F. Ahmad

Fig. 3 Rainfall map

Tapanuli Regency Area. Very wet rainfall (3001–3500 mm/year and 3501–4000 mm/ year) occurred in an area of 21.75% of South Tapanuli Regency Area. Wet rainfall (2501–3000 mm/year) occurred in an area of 14.88% of South Tapanuli Regency Area. Dry rainfall (1501–2000 mm/year) occurred in the smallest area, only 3.90% of South Tapanuli Regency Area. Based on Fig. 4, it can be seen that the rocks in South Tapanuli Regency consist of alluvial, sedimentary, and volcanic rocks. Alluvial rocks have an area of 19.09% of South Tapanuli Regency Area. Sedimentary rocks cover a fairly wide area with 25.08% of South Tapanuli Regency Area. Volcanic rocks have the widest area with 55.83% of South Tapanuli Regency Area. Based on DEM (Digital Elevation Model) analysis with Q-GIS application sourced from the DEMNAS website, 10.07% of South Tapanuli Regency has a slope of 0–8% (gentle slope), 24.40% has a slope of 8%–15% (moderate slope), 21.72% has a slope of 15–30% (steep slope), 29.11% has a slope of 30–45% (extremely steep slope), and 14.53% has a slope > 45%. These results can be seen in Fig. 5. Figure 6 shows the land cover in South Tapanuli Regency which consists of mining, settlements, primary dryland forest, secondary dryland forest, secondary swamp forest, plantation forest, plantations, shrubs, swamp shrubs, dryland farming, mixed dryland farming, open land, and paddy fields. The type of land cover dominating South Tapanuli Regency was mixed dryland farming with an area of 31.83% of the total area.

Landslide Analysis Using Geographic Information System (GIS) …

Fig. 4 Rock type map

Fig. 5 Slope map

303

304

I. P. Hastuty and F. Ahmad

Fig. 6 Land cover map

Figure 7 shows the soil type in South Tapanuli Regency which consists of Humic Acrisols, Orthic Acrisols, and Dystric Fluvisols. Of the entire area of South Tapanuli Regency, Humic Acrisols have an area of 83.61%, Orthic Acrisols have an area of 0.02%, and Dystric Fluvisols have an area of 16.37%. Based on the total weight calculation of the parameter results at the research location, the lowest and highest scores were 1.80 and 4.50 (Table 2.). The highest score and lowest score were used to obtain classification intervals for 4 landslide susceptibility levels, namely low, moderate, high, and very high susceptibility levels. The interval value obtained using interval equation in Eq. (2) was 0.675. Based on the results of each level of landslide susceptibility in South Tapanuli Regency, it can be seen that high susceptibility level has the widest area with 53.03%, while low susceptibility level has the smallest area of 3.25%. Each parameter was given a score based on the level division, then the score was multiplied by the weight according to their respective contributions. The result of the multiplication became a reference attribute that would be totaled when the overlay was done. The sum of the reference attributes result for each layer produced new data, namely the total weight which was the value of landslide susceptibility levels. The results can be seen in Fig. 8 where out of the 14 districts in South Tapanuli Regency, there were 3 districts whose area was dominated by moderate landslide susceptibility level, namely Aek Bilah, Batang Angkola, and Muara Batang Toru Districts. While the 10 districts were dominated by high landslide susceptibility level, namely West Angkola, South Angkola, East Angkola, Angkola Sangkunur, Arse, Marancar, Saipar Dolok Hole,

Landslide Analysis Using Geographic Information System (GIS) …

305

Fig. 7 Soil type map

Table 2 Landslide susceptibility intervals

No

Interval

Tingkat kerawanan

1

1.8–2.4

Rendah

2

2.5–3.1

Sedang

3 4

3.2–3.8 3.9–4.5

Tinggi Sangat Tinggi

Sayur Matinggi, Sipirok, and Tano Tombang Angkola. Only Batang Toru Regency was dominated by a very high landslide susceptibility level.

306

I. P. Hastuty and F. Ahmad

Fig. 8 Landslide susceptibility map

4 Conclusion Based on the results and analysis of the research, it can be concluded as follows: 1. South Tapanuli Regency has the characteristic parameters of landslides in the form of: a. Rainfall was dominated by moderate rainfall (2001–2500 mm/year), about 59.51% of the total area of South Tapanuli Regency. b. Rock type was dominated by volcanic rock formations, about 55.83% of the total area of South Tapanuli Regency. c. The slope was dominated by slope > 45%, about 29.11% of the total area of South Tapanuli Regency. d. Land cover was dominated by dry mixed farming, about 31.83% of the total area of South Tapanuli Regency. e. The soil type was dominated by Humic Acrisols, about 83.61% of the total area of South Tapanuli Regency. 2. The landslide susceptibility map in South Tapanuli Regency has been obtained using the Geographic Information System (GIS), where: a. Low landslide susceptibility level has an area of 3.24% of the total area of South Tapanuli Regency.

Landslide Analysis Using Geographic Information System (GIS) …

307

b. Moderate landslide susceptibility level has an area of 34.83% of the total area of South Tapanuli Regency. c. High landslide susceptibility level has an area of 52.75% of the total area of South Tapanuli Regency. d. Very high landslide susceptibility level has an area of 8.65% of the total area of South Tapanuli Regency. 3. Out of the 14 districts in South Tapanuli Regency, there were 3 districts whose area was dominated by moderate landslide susceptibility level, namely Aek Bilah, Batang Angkola, and Muara Batang Toru Districts. While the 10 districts were dominated by high landslide susceptibility level, namely West Angkola, South Angkola, East Angkola, Angkola Sangkunur, Arse, Marancar, Saipar Dolok Hole, Sayur Matinggi, Sipirok, and Tano Tombang Angkola. Only Batang Toru Regency was dominated by a very high landslide susceptibility level.

References 1. Sarya G, Andriawan AH, Ridho A, Seputro H (2014) Intensitas Curah Hujan Memicu Tanah Longsor Dangkal di Desa Wonodadi Kulon. Jurnal Pengabdian LPPM UNTAG Surabaya 1(1):65–71 2. Sitepu F, Selintung M, Harianto T (2017) Pengaruh Intensitas Curah Hujan dan Kemiringan Lereng Terhadap Erosi Yang Berpotensi Longsor. Jurnal Penelitian Enjiniring 21(1):23–27 3. Hidayat R, Zahro AA (2020) Penentuan Ambang Curah Hujan untuk Memprediksikan Kejadian Longsor, pp 1–10 4. Badan Nasional Penanggulangan Bencana (2019) Bencana Menurut Jenisnya di Indonesia Tahun 2009 s/d 2019. Psychological Bulletin 5. Badan Pusat Statistk (2021) Kabupaten Tapanuli Utara Dalam Angka. BPS-Tapanuli Utara 6. Badan Geologi Homepage. https://vsi.esdm.go.id/index.php/gerakan-tanah/peringatan-dini-ger akan-tanah/4086-mei-2023 7. Goodman RE, Scott Kieffer D (2000) Behavior of rock in slopes. vol 126, pp 675–684 8. Pusat Penelitian dan Pengembangan Tanah dan Agroklimat (2004) Laporan Akhir Pengkajian Potensi Bencana Kekeringan, Banjir dan Longsor di Kawasan Satuan Wilayah Sungai CitarumCiliwung, Jawa Barat Berbasis Sistem Informasi Geografi. Puslittanak, Bogor

River-Basin Classification for Flood Risk Assessment in Indonesia Adityawan Sigit

and Morihiro Harada

Abstract The risk of water-related disasters is increasing globally due to climate change, but varies depending on land use, population, and watersheds. This study presents a classification method for flood risk assessment that allows appropriate mitigation measures to be taken based on regional characteristics. A case study was conducted in Indonesia, which has unique geographical characteristics, such as the Asian Monsoon Zone and high rainfall. First, a database on the topography, climate, and population demographics of each river basin will be maintained in the GIS by using an open-source dataset. We extract areas that may be inundated by some types of floods by applying simple assumptions to topographic dataset such as Digital Elevation Maps (DEM), Grid Point Values (GPV), and Height Above Nearest Drainage (HAND). The results of aggregating the population exposed to hazardous areas by subdistrict showed a very large range in the exposure ratio, from 0.0 to 0.88. The analysis, which also considered the poor ratio of each subdistrict, revealed a ranking of the poor population exposed to flooding. These results suggest useful information for proposing effective mitigation tailored to local characteristics. These results demonstrate the advantages of focusing on the characteristics of the exposed society in water-related risk assessment. It is hoped that the classification can guide readers in determining targeted disaster prevention and mitigation policies. Keywords River-basin · Classification · Flood risk assessment

A. Sigit (B) Department of Civil Engineering, Islamic University of Indonesia, Yogyakarta, Indonesia e-mail: [email protected] A. Sigit · M. Harada Department of Engineering, Gifu University, Gifu, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_29

309

310

A. Sigit and M. Harada

1 Introduction Climate change-induced flood disasters are occurring all over the world and will grow in magnitude in the years ahead. Flood is one of the most frequent natural disasters with reported average annual losses and fatalities between 1980 and 2012 exceeding $23 billion and 5900 fatalities, respectively, making it the most destructive disaster affecting communities worldwide [1]. These risks have a negative effect on national economic growth. The dynamics of hazard, exposure, and vulnerability in flood risk result in regional differences and global trends [2]. For example, some areas in Japan will receive between 1.2 and 1.3 times more rainfall by 2050, and the potential economic losses related to flooding will increase significantly in the future due to climate change [3]. While in Indonesia, especially Java Island, climate change and urban expansion will allegedly increase the risk of riverbank flooding by 76% by 2030, and the implementation of adaptation measures is increasingly urgent [4]. As the threat of flooding increases in various regions of the world, mitigation measures are needed in the face of this. However, there are gaps in disaster mitigation in each country. Disaster management system done by developed countries may not be adopted by developing and underdeveloped countries [5]. Indonesia as a developing country, the lack of hydrological data for flood hazard analysis needs in most parts of the country leads to research difficulties and high disaster assessment costs [6]. In addition, Indonesia has a diversity of characters in society. Within the diversity of society, there is some socioeconomic diversity that will affect government policies in each region, especially in the disaster management process. This is mainly because, for some communities, the cost of disaster management systems is not cheap, and sometimes the implementation is not effective. The application of the disaster management system to the exposure and vulnerable land-use sites is not effective considering certain areas have limited costs that are directly proportional to the limitations of their socio-economic conditions [7]. Therefore, there needs to be some strategy for disaster management, especially in how to effectively assume the flood hazard. The initial stage in the disaster management system includes risk assessment activities. To contribute to effective adaptation strategies, flood risk assessments need to focus on the relationship between hazard locations and the society to which it is exposed. However, many hazard maps published in many countries and regions only show hazard assumptions. Indonesia, with its highly diverse society, needs to be analyzed not only the hazards but also the social conditions to which it is exposed. To develop adaptation strategies, it is necessary to classify areas to analyze the relationship between hazard and exposure. Not only for each administrative region, but also to identify its local characteristics so that effective measures can be developed for each region. From this background, several research points can be drawn: (1) assuming floodprone areas. The lack of detailed hydrological data for flood analysis, which makes physical-complex simulations based on assumptions of external forces, needs to be

River-Basin Classification for Flood Risk Assessment in Indonesia

311

addressed to make disaster assessment more economical. And (2) classify exposed areas by comparing flood hazards and community characteristics.

2 Study Area and Data 2.1 Study Area The Serayu River, Central Java Province, Indonesia, often floods and causes inundation during the rainy season every year. The river crosses five districts, namely Wonosobo, Banjarnegara, Purbalingga, and Banyumas, until it empties into the Indian Ocean in Cilacap. Based on disaster event data from the Local Disaster Management Agency, the location of the distribution of flood events that occurred from 2016 to 2020 can be seen in Fig. 1. INDONESIA

CENTRAL JAVA

Fig. 1 Flood event 2016–2020 of Serayu River, Serayu Watershed, Central Java, Indonesia [8]

312

A. Sigit and M. Harada

Based on the flood event map, it can be seen that Serayu Watershed has several locations that are flooded. This can also be used as an estimate that in the next rainy season some of these areas have the potential for flooding caused by overflowing flow from the river. It is necessary to overcome flooding in several locations that are potentially flooded every year as a flood disaster management in flood-prone areas in Serayu River.

2.2 Data In this study, the data is grouped into 2 main groups, namely Hazard Data and Exposure Data (see Table 1). Hazard. The usual method of flood risk assessment for a watershed; calculate the frequency of flooding, determine the magnitude of flooding, simulate it with 1D or 2D, and then produce the inundated area map. Indonesia, which has a lack of data makes this method difficult and requires a lot of simulation costs. In this study, we use a more effective method by assuming a map of inundated areas through comparison of topographic data. The data are records of inundated areas in 2016–2020, Digital Elevation Map (DEM) and Height Above the Nearest Drainage (HAND) in the Serayu River. HAND and DEM data are taken from MERIT global hydrography datasets. Exposure. Information of socio-economic data was adjusted to the Geo-Information System (GIS) topography of the subdistrict region and then produced the GIS Database of Socio-Subdistrict. Socio-economic and Subdistrict Region information of Central Java was taken from the Indonesian Central Bureau of Statistics of Central Java.

3 Method In this study, the research steps are described as the following Fig. 2. Table 1 Data and material research Data and materials

Result

Hazard 1. Inundation area records

Map of inundated area by DEM-HAND

2. Topographic data

Relationship

Exposure 1. Socio-economic information

GIS database of Socio-Subdistrict

2. Subdistrict region

Relationship

River-Basin Classification for Flood Risk Assessment in Indonesia

313

Fig. 2 Research flowchart

3.1 Hazard In determining the hazardous area, inundation depth is required. An inundation depth was assumed from Grid Point Value (GPV) analysis by the elevation difference of raster data between HAND and DEM (see Fig. 3). Vector of recorded data flood events from 2016 to 2020 at Serayu Watershed was used as the boundary of the inundation area.

HAND

DEM GPV of DEM and HAND

Inunda.on Area

Fig. 3 Example of GPV analysis of Serayu Watershed inundation area

314

A. Sigit and M. Harada

3.2 Exposure Analysis of exposure used data comparison between population characteristics and exposed area. In this study, the population characteristics used the subdistrict population and socio-economic data. The exposed area is limited to subdistricts focusing on 29 districts of Central Java Province. These districts of Central Java Province are shown in Table 2.

3.3 Districts of Central Java Province The following table is the districts in Central Java province with their population, poverty population, and civilian per capita district income in 2020.

4 Results 4.1 Determining the Hazardous Area with the HAND Raster To assume the hazardous area, we are focusing on some limited area of inundation. From this relationship, we could get the relationship between the HAND value, Elevation, and Inundation areas. From the GPV analysis, we obtained the results shown in Fig. 4. According to Fig. 4, inundation area containing the high HAND has a strange value area. Over 80% of inundation areas are located under 300 m altitude. The low-land floodplains are low inundated, while in high ground, inundations are higher. There are several possibilities that explain this situation. One of them is the possibility that the shape of the inundation area is rough, so it contains areas that are not inundated, so there is the possibility of some larger inundation height values. With the irregularity of the data we received, we used a reasonable value by dividing the elevation (highland and lowland). The elevation range of inundation depth is limited to elevations below 300 m. Thus, we can assume a relationship between elevation (DEM data) and inundation (HAND data) (Fig. 5). For the determination of the plan’s inundation height, we see that the maximum value is too big (see Fig. 6). We think it is not a correct value. So, we focused on the quartile values in each elevation range separates every 30 m and the HAND values so that we could get a relationship (see Fig. 7). Judging from the results of the quartile analysis, we focus on the medium value of 50%. The values obtained include 1.9, 4.5, and 5 m, making it suitable to serve as the threshold value. In this study, we set the value between 2 and 5 m. With these values, the risk area of inundation is analyzed.

River-Basin Classification for Flood Risk Assessment in Indonesia

315

Table 2 Socio-economic information of Central Java Province 2020 District Name

Capital

Area (km2 )

Population

Number of poor people

Income per capita ($)

Cilacap

Cilacap

2124.47

1,944,857

198,600

4279.86

Banyumas

Purwokerto

1335.30

1,776,918

225,840

2222.97

Purbalingga

Purbalingga

677.55

998,561

149,480

1880.49

Banjarnegara

Banjarnegara

1023.73

1,017,767

144,950

Kebumen

Kebumen

1211.74

1,350,438

211,090

1648.78

Purworejo

Purworejo

1091.49

769,880

84,790

1829.54

Wonosobo

Wonosobo

981.41

879,124

137,630

1679.72

Magelang

Mungkid

1102.93

1,299,859

146,340

1762.59

16.06

121,526

9270

5036.92

Magelang City Magelang Boyolali

Boyolali

1008.45

1,062,713

100,590

2332.86

Klaten

Klaten

658.22

1,260,506

151,830

2395.72

Sukoharjo

Sukoharjo

489.12

907,587

68,890

2874.20 2142.61

Wonogiri

Wonogiri

1793.67

1,043,177

104,370

Karanganyar

Karanganyar

775.44

931,963

91,720

Sragen

Sragen

941.54

976,951

119,380

1392.43

Surakarta (or Solo) City

Surakarta

46.01

522,364

47,030

6461.77 1392.43

Grobogan

Grobogan

2013.86

1,453,526

172,260

Blora

Blora

1804.59

884,333

103,730

Rembang

Rembang

887.13

645,333

100,080

Pati

Pati

1489.19

1,324,188

127,370

Kudus

Kudus

425.15

849,184

64,240

Jepara

Jepara

1059.25

1,184,947

91,140

2418.59 1671.94

Demak

Demak

900.12

1,203,956

146,870

1597.88

Semarang

Ungaran

950.21

1,053,094

79,880

3248.27

Salatiga City

Salatiga

57.36

192,322

9690

4848.69

Semarang City Semarang

373.78

1,653,524

79,580

7258.02

Temanggung

Temanggung

837.71

790,174

77,330

1957.79

Kendal

Kendal

1118.13

1,018,505

97,490

3077.03 1972.44

Batang

Batang

788.65

801,718

70,570

Pekalongan City

Pekalongan

45.25

307,150

22,160

Pekalongan

Kajen

837

968,821

91,860

Pemalang

Pemalang

1118.03

1,471,489

209,030

Tegal City

Tegal

39.68

273,825

19,550

2490.46 1243.82 (continued)

316

A. Sigit and M. Harada

Table 2 (continued) District Name

Capital

Area (km2 )

Population

Number of poor people

Income per capita ($)

Slawi

876.1

1,596,996

117,500

1739.29

Brebes

1902.37

1,978,759

308,780

1683.75

32,800.69

36,516,035

3,297,100

2500

100%

2000

80%

1500

60%

1000

40%

500

20% 0%

0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500

Number of grids included in the inundation area

Totals

Cumulative Percentage

Tegal Brebes

Elevation range [m]

1400 1200 1000 800 600 400 200 0

100% 80% 60% 40% 20% 0%

Cumulative Percentage

Number of grids included in the inundation area

Fig. 4 DEM and HAND relationship

30 60 90 120 150 180 210 240 270 300 Elevation range [m] (Less than 300m only)

Fig. 5 DEM and HAND relationship below 300 m elevations

4.2 Exposure Analysis Using that hazardous area information, we overlap some socio-economic information, so it makes some information to classify characteristics of its subdistrict. From the results of the comparison of the ratio of poor district residents and the ratio of the risk of being affected by inundation between 2 and 5 m, it was found that the highest exposure is Demak District with a ratio of 0.88, and the highest exposed poor people are Brebes, Kebumen, Rembang and Wonosobo with a ratio around 0.16.

River-Basin Classification for Flood Risk Assessment in Indonesia

317

350 HAND value [m]

300 250 200 150 100 50 0 -50

0

50

100 150 200 DEM value [m]

250

300

350

Fig. 6 DEM and HAND events below 300 m elevations

Quartile values of HAND value [m]

1000.0

First quartile value Median value

100.0 10.0

Thied quartile value Maximun value

1.0 0.1

Elevation range [m] Fig. 7 Quartile values of DEM and HAND

In Fig. 8, the highest exposure rate observed is Demak district, and Demak also has high poor ratio, so it is ranked as number 2. And Brebes district has the second highest exposure ratio that has been observed, but the exposed poor ratio is the highest, so it is ranked as number 1 (Fig. 9).

5 Summary It is known from the results that the districts where the poor are most at risk of inundation in Central Java Province are Brebes and Demak. The results of this study provide useful information to propose effective mitigation tailored to local characteristics. The results demonstrate the advantages of focusing on the characteristics of exposed communities in water-related risk assessment. It is hoped that the results

318

A. Sigit and M. Harada

Fig. 8 Result of poor people at risk of being inundated in Central Java Province

Fig. 9 Districts with poor population with exposure ratio

River-Basin Classification for Flood Risk Assessment in Indonesia

319

of this study can serve as input for local policymakers in managing inundation risks and determining targeted disaster prevention and mitigation policies.

References 1. Münchener Rückversicherungs Gesellschaft. https://www.munichre.com/content/dam/mun ichre/contentlounge/website-pieces/documents/Press-release-NatCat-first-halfyear-2015.pdf 2. IPPC (2022) Impacts of 1.5°C global warming on natural and human systems. Global Warming 1.5 °C 3. Tezuka S et al (2014) Estimation of the effects of climate change on flood-triggered economic losses in Japan. Int J Disaster Risk Reduction 4. Muis S et al (2015) Flood risk and adaptation strategies under climate change and urban expansion: a probabilistic analysis using global data. Sci Total Environ 538:445–457 5. Ali HM et al (2021) Planning and assessment approaches towards disaster resilient hospitals: a systematic literature review. Int J Disaster Risk Reduction 61 6. Maki S et al (2022) A deep reinforced learning spatiotemporal energy demand estimation system using deep learning and electricity demand monitoring data. Appl Energy 324:119652 7. Nugroho A et al (2022) Impacts of village fund on post disaster economic recovery in rural Aceh Indonesia. Int J Disaster Risk Reduction 70 8. Balai Besar Wilayah Sungai Serayu-Opak (BBWS-SO) (2021) Laporan Akhir: Kegiatan Monitoring Banjir Wilayah Sungai Serayu Bogowonto dan Progo Opak Serang

Synthesis of Management Measures for Water Resource Sustainability and Resilient Society: A Case Study of Mae Klong Watershed Through Four Decade Evidences Charumas Meksumpun

Abstract Watersheds in tropical countries recently face problems either quantitative or qualitative aspects due to climate impacts and anthropogenic sources. Wellprepared adaptive measures and related guidelines are needed in order to maintain resource productivity and enhance utilization sustainability for our future well-being society. In Thailand, Mae Klong watershed has a large catchment area of 30,171 square kilometers covering nine provinces. The watershed possesses comparatively higher slope, with upstream areas in high mountains in western Thailand, and several lower-reach river channels flowing into estuarine zones along the Upper Gulf of Thailand. Plenty of river channels and small creeks provide remarkable nourished ecosystems. In particular, the lower reach in Samut Songkram province has channel structures that have long been well-utilized for community transportation and horticultural productions. Consideration on increasing problems of flood disasters and water quality deteriorations, development of suitable measures and related action plans for remediation and sustainable management of water resources is needed. Thus, integrated science–social research project has been initiated since 2021. The project methodology was performed through analysis of academic information during the past forty years, interviews of related stakeholders, opened-panel meetings among community leaders, and holistic synthesis for effective management approach. During the project platform, adaptive and alternation approaches for management of water problems (i.e., floods, water quality deterioration, and marine water intrusions) were discussed and analyzed. Accordingly, measures for water resource sustainability and resilient society were synthesized and proposed through integrated implementation techniques of Downstream, Bottom-up, and Parallel Management Approaches. In the overall views of the research, area-based sensitive indicators should be carefully considered and further developed. Moreover, enhancement of “Awareness on Worth of Ecosystem Sustainability” was considered to be critical necessary to achieve community cooperation, stakeholder responsibility, and also society well-being. C. Meksumpun (B) Kasetsart University, Bangkok 10900, Thailand e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_30

321

322

C. Meksumpun

Keywords Management measures · Water quality · Aquatic resource sustainability · Tropical watershed

1 Introduction Watersheds in tropical countries recently face problems either quantitative or qualitative aspects due to climate impacts and anthropogenic sources. Well-prepared adaptive measures and related guidelines are needed in order to maintain resource productivity and enhance utilization sustainability for our future well-being society. In the western part of Thailand, the Mae Klong watershed has a large catchment area of 30,171 km3 covering nine provinces (Fig. 1). The watershed possesses comparatively higher slope with upstream areas in high mountains. In the downstream areas, there are plenty of river channels flowing through flood plain into the Upper Gulf of Thailand. Several river channels contribute to the nourishment of aquatic ecosystems. In particular, the lower-reach watershed in the Samut Songkram province of 416 km2 has more than 300 river channel structures and a lot of internal creeks that have long been utilized for community transportation and horticultural productions [1]. Nevertheless, flood events in the lowland zones also cause remarkable problems to nearby community. Recent problems of eutrophication in the lower-reach river [2], together with evidences of deteriorated water quality [3, 4], have been frequently noticed. Consideration on increasing problems of water quality deteriorations and flood disasters, development of suitable measures and related action plans for remediation and sustainable management of water resources is needed [1, 4]. To cope with such issues, an integrated science–social research project has been initiated. Outputs from this research can be further applied for development of suitable strategic approach. Guideline for necessary researches for the watershed sustainable development was proposed.

2 Methodology This research was carried out through analysis of science–social research papers reported during the previous forty years. Additionally, interviews of representative stakeholders in relations to the Mae Klong watershed utilizations (i.e., horticulture, aquaculture, salt production, and mangrove restoration) were performed in the local communities during 2021–2022. In addition, opened-panel meetings among community leaders were carried out at Amphawa Town, located in the lower-reach river in the Samut Songkram province, in order to identify recent problems of the watershed and clarify their priority for further effective management. During the research, adaptive and alternation approaches for management of water problems (i.e., floods, water quality deterioration, and marine water intrusions) were mainly discussed.

Synthesis of Management Measures for Water Resource Sustainability …

323

Mae Klong watershed

Samut Songkram

Fig. 1 Boundary of the Mae Klong watershed located in the western part of Thailand

According to the overall information together with know-hows and recommendations from all stakeholders, holistic management approach was synthesized. Thereafter, integrated management guidelines were proposed to the community leaders and related organizations for further application approach.

3 Results and Discussions In this study, necessary information for sustainable management of water resource was clarified and synthesized. The important features concerning on status of water environments, their spatiotemporal changes, and eutrophication carrying capacities of the water areas were examined and summarized as follows.

324

C. Meksumpun

3.1 Water Environments and Key Drivers of Changes Recent status of water environments of the Mae Klong watershed based on the guideline of surface water quality standard [5] is in good to moderate water quality status, in particular in the upper- and middle-reach river. Nevertheless, deteriorated water quality was reported for the lower-reach river habitats and the estuarine zones of the Samut Songkram province, the lowest region of the Mae Klong watershed [4]. Such occurrences are majorly caused by agro-industrial waste discharges, domestic runoff, and severe waste discharges of pig farming industry from nearby Ratchaburi province [6]. Nutrients were found to contribute to the hypertrophic condition of the lower river and the estuary [4]. The baseline level of phytoplankton chlorophyll a in the upper river was determined to be ∼5 µg/L, while levels exceeded 20 µg/L in the estuarine zone [3, 4]. Such the chlorophyll a levels were highly significant related to dissolved inorganic nitrogen (DIN) and orthophosphate phosphorus (P). To maintain the water resource eutrophy in a good condition, the DIN and P levels in the lower river of the Mae Klong watershed of about 20 µmol/L and 1 µmol/L, respectively, were suggested as the upper limits [3]. Nevertheless, recent DIN levels (2018–2019) of the lower river have increased more than 1.5 times and exceeded the criteria for river conservation. According to eco-hydrological variations of the watershed, moreover, changes of total suspended solids (TSS) in the water columns were noticed (Table 1). TSS levels were fluctuated and somewhat increased, in particular in the estuarine zone of the lower Mae Klong watershed. The highest TSS level of 339 mg/L was examined during early-loading period [4]. From 2005 to 2019, the increasing trend of TSS of the river system has been revealed. In addition, contaminations of copper and zinc in the estuarine sediments reached the levels of 40 and 119 ppm, respectively, during the surveyed period of 2019 [4]. Such levels of sedimentary copper and zinc were about two times higher than the levels of 2007 (Table 2). Overall, the increases of chlorophyll a, TSS, and the heavy metals, thus, imply gradual deterioration of the Mae Klong watershed environmental status. Table 1 Total suspended solids (mg/L) in the Mae Klong Estuary during 2005–2018

Year of study

Early-loading period

2005

31–110

2006

4–43

4–65

2007–2008

8–96

36–65

2017–2018

55–339

28–196

Mid-loading period 9–19

Synthesis of Management Measures for Water Resource Sustainability …

325

Table 2 Heavy metal concentrations (ppm) in the Mae Klong Estuary during 2007–2019 Year of study

Cadmium

Lead

Copper

Zinc

2007

nd–0.8

7.3–34.6

2.0–14.5

10.9–40.9

2009

2.4

29.4

16.6

59.9

2016

nd

54.8

12.6

47.5

2019

nd

22.2–50.6

4.0–40.4

17.8–119.3

3.2 Nutrient Transfers and Carrying Capacity Analysis Among the whole scientific databases, clarification of watershed longitudinal changes and transfer efficiency of DIN and P seems to be of most importance for zonation management approach. Application of BOX model for nutrient analysis along several sectional reach throughout river [3, 7] provided useful information on the “pollution source and transfer regime” of the river system. Accordingly, the sensitive zones can be identified, and the remediation management for both area and time aspects can be appropriately designed. For the Mae Klong River, research evidences have revealed that the “inflow rate” of the river in each zone is the key driver controlling eutrophication status and river pollution remediation potential. Appropriate amounts of inflow can enhance nutrient and pollutant dilution and hence impact on the control of phytoplankton production of the river system [3, 6]. The previous reports also emphasized the impact of “self-contamination” which implied the importance of the internal loads that were higher than the external loads (i.e., loads from the upper reach river). In addition, the internal loads around the lowest watershed region of the Samut Songkram province was examined to be highest (24 tons DIN/day and 4 tons P/day, respectively) [3]. Nutrients in the form of orthophosphate phosphorus (P) have been noticed to increase with the increase of community waste discharges. Comparatively higher P levels were usually examined in the lower river and in the estuary [4, 8]. Since recent P levels exceeded water quality standard of Thailand (1.45 µmol/L) [5, 9], urgent management, i.e., the control of agro-industrial and food industrial activities, is recommended for conservation and remediation of the water resources [10]. In addition, analysis on the impact of inflow on estuarine fishery resources [8] also reflected the importance of inflow in the lower-reach watershed ecosystem. When compared to the other three rivers flowing into the Upper Gulf of Thailand, inflow rates of the Mae Klong River, either during the dry- or rainy-seasons, were in comparatively higher levels. The highest inflow was examined to be 753 m3 /S (Table 3) [4]. Such high inflow rates contributed to a good estuarine circumstance and, accordingly, can help to maintain a better condition of fishery resource and aquaculture productions, in particular clam and oyster productions in the Mae Klong Estuary [1, 4, 8].

326

C. Meksumpun

Table 3 Inflows (m3 /S) from major rivers into the Upper Gulf of Thailand during 2017–2018 Rivers

Apr-2017

Jul-2017

Dec-2017

Apr-2018

Mae Klong

452.1

369.8

290.0

753.2

Jul-2018 369.8

Tha Chin

95.0

174.0

9.0

180.5

174.0

Chao Phraya

94.6

417.4

185.4

160.1

417.4

Bang Pakong

42.7

222.3

42.9

80.3

222.3

3.3 Topographical Characters of Creeks and Channels in the Lower-Reach Watershed The Mae Klong watershed in particular the lower region in the Samut Songkram province is so called “the last Eastern Venice” of Thailand because there are plenty of natural river channels with a lot of beautiful creeks (Fig. 2). Such channel characteristics contribute to the exchange and remediation of the river water. Circulation of the water in the channel system also acts as remediation pathway for improvement of water quality because of the effectiveness of nutrient accumulation by phytoplankton and the sedimentation of particulate matters due to submerged vegetations. Along the channels, the increase of dissolved oxygen can be occurred as well. Besides the improvement of the water quality, the roles of channels in the lower reach of the watershed have revealed to prevent problems from flood disasters. Horizontal distributions of flood water mass can make great help in flood control and

Fig. 2 Topographical characters of creeks and channels in the lower-reach watershed

Synthesis of Management Measures for Water Resource Sustainability …

327

decrease inflow speed to acceptable levels for surrounded community. Although several constructions (i.e., diversion dams, large building, and reports) in the lower river areas caused dramatical decreases of the channels, the importance of the channels is still realized by local community. Conservation management of the channels is, therefore, always carried out mainly from local people and non-governmental efforts. More contributions from governmental sectors are crucial needed.

3.4 Synthesis of Conservative Management Approaches In the research, conceptual scheme for management achievement was proposed. Moreover, measures of Downstream, Bottom-up, and Parallel Management Approaches for water resource sustainability and resilient society were synthesized and proposed for further integrated implementation. These measures were developed based upon the sustainable development goals for well-being and resilient society. Top-down Management Approach. Top-down Management Approach was synthesized mainly for governmental sectors with responsibility in water problem solving. The approach was focused mainly for the problems of (1) salinity intrusion, (2) flood, (3) water quality, and (4) long-term water management. From knowhow integrations, the Salinity Intrusion Management Approach should concern the important issues such as the administration of freshwater with deep consideration of impacts from tidal phenomena together with the control of irrigation gates. In addition, remediation and development of natural channels was strongly recommended. The Flood Management Approach should concern the important issues such as the re-construction or enlargement of irrigation gates, the development of suitable operation system, and the development of natural channels with potential in receiving and distributing the water mass. The Water Quality Management Approach should concern the control for sufficient inflow water mass into the lower-reach watershed, critical motoring of point-source pollution zones, remediation of water and sediment quality along the front part of the irrigation gates, and enhancement of social perception in environmental deterioration for cooperative management. In addition, the Long-term Water Management Approach should deeply concern the following aspects: • • • • • • •

Development of effective watershed measures and administration system; Analysis of management approach together with alternative ways; Development of small community-based project rather than large-area project; Development of social awareness on environmental deteriorations; Development of waste water remediation system in local areas; Management of garbage and water pollution problems; Promotion of eco-worth rather than economic-value evaluation.

Bottom-up Management Approach. The Bottom-up Management Approach was synthesized mainly for community sectors with responsibility in problem solving.

328

C. Meksumpun

The approach can contribute to the development of community-based competency. The important aspects are as follows. • • • • • • • • •

Enhancement of communication among community and governmental sectors; Encouragement of community-based water management; Development of research on river structures and aquatic ecosystems; Conservation management of river channels for a good water condition; Remediation of aquatic vegetations including mangrove habitats; Enhancement of environmental conservation and habitat restoration; Development of network from upstream to lower watershed communities; Enhancement of cultural conservation together with ecosystem conservation; Development for long-term eco-sustainability rather than short-term incomes.

Parallel Management Approach. The Parallel Management Approach was synthesized mainly for research and academic sectors. The approach can contribute to know-how and implication development. The necessary research aspects are included the issues such as (1) river carrying capacity (both quantitative and qualitative aspects), (2) future needs of water uses, (3) impacts from social and cultural changes on watershed management, (4) effective indicating factors for water resource conditions, (5) effective techniques for community competency development, and (6) suitable city plan for watershed conservation and sustainable development.

4 Conclusion In this research, the environmental status of the Mae Klong watershed has been depicted to gradually degraded along the times during the past 40 years. The deterioration of water quality was induced by either the increases of wastewater discharges or the decreases of natural river channels during the recent periods of town construction and development. Anthropogenic changes of the watershed topographical structures also enhanced flood disaster severity. In order to maintain a good water circumstance for well-being community, the Top-down, Bottom-up, and Parallel Management Approaches were synthesized. From the overall views of the research, areabased monitoring indicators should be carefully considered and further developed. Moreover, enhancement of “Awareness on Worth of Ecosystem Sustainability” was of importance. Such perception is critical necessary to achieve community cooperation, stakeholder responsibility, and our well-being society. Acknowledgements This study was partly supported by the Department of Fishery Biology, Faculty of Fisheries, Kasetsart University, the Kasetsart University Research and Development Institute, and the Graduate School of Kasetsart University. The author acknowledges all members of the Mae Klong Conservation Organization for their invaluable recommendations. Deep thanks are also due to Jeeraya Muangsringam for her kind helps in field surveys and encouragements.

Synthesis of Management Measures for Water Resource Sustainability …

329

References 1. Chiravate S (2014) Kon Mae Klong, 7th edn. S. Asia Press Ltd., Bangkok 2. Channual P (2007) Heavy metal dynamics: a case study on relationship between heavy metals and sediment qualities in Mae Klong River. Master Thesis. Marine Science. Kasetsart University, Bangkok, Major Field 3. Thongdonphum B, Meksumpun S, Meksumpun C (2011) Nutrient loads and their impacts on chlorophyll a in the Mae Klong River and estuarine ecosystem: An approach for nutrient criteria development. Water Sci Technol 64(1):178–188 4. Faculty of Fisheries (2019) The final report on development of socio-ecological based effective fishery management policy for good governance in sustainable fishery of the Inner Gulf of Thailand. Faculty of Fisheries, Kasetsart University, Bangkok 5. Department of Pollution Control (2006) Water quality standards for surface waters of Thailand. Water Quality Management Office. Department of Pollution Control, Bangkok 6. Meksumpun C (2022) Mae Klong and the conservative development. Faculty of Fisheries, Kasetsart University, Bangkok 7. Thaipichitburapa P, Meksumpun C, Meksumpun S (2010) Province-based self-remediation efficiency of the Tha Chin river basin, Thailand. Water Sci Technol 62(3):594–602 8. Meksumpun C (2020) Pathways of mackerel in the Gulf of Thailand: the impacts from aquatic environmental problems. Kasetsart University, Bangkok, Faculty of Fisheries 9. Chongprasith P, Wiliratanadilok W, Utoomprurkporm W (1999) ASEAN marine water quality criteria for phosphate. ASEAN-Canada CPMS-II Cooperative Programme on Marine Science. Department of Pollution Control, Bangkok 10. OECD (1982) Eutrophication of waters: monitoring, assessment and control. Organization of Economic Co-operation and Development, Paris

Community Outreach Through Soft Type Disaster Mitigation Measures

Developing Effective Flood Inundation Maps for Risk Communication and Evacuation Planning in Obuse, Japan Mizuki Sakai, Yugo Hashimoto, Naoki Todoroki, Yoshinori Furumoto, and Toru Oomiya

Abstract Obuse is a small town in the northern part of Nagano Prefecture in Japan surrounded by the Shinano and Matsukawa rivers. The typhoon Hagibis resulted in the flooding of the Shinano River and caused enormous damages to Obuse. In this study, we aim to devise a method based on which local government officials can use flood inundation maps to convey detailed flooding information to residents for smalland medium-sized rivers. To connect the information provided in the maps to actual evacuation actions, videos were created based on current flood inundation maps, and flood inundation simulation was performed based on the use of the International River Interface Cooperative (iRIC) software and Nays2DFlood solver. A terrain model was created for the simulations by using a digital elevation model (DEM). Two types of inundation analyses were performed for different terrain models in different drainage conditions. Videos were created from the data by using the geographical information system software (ArcGIS) when the levee broke at 58 sites along the Matsukawa River. In the iRIC simulation, the reproducibility was sufficient when a 10 × 10 m mesh was used, a simple correction was added to the levee parts of DEM, and a drainage pumping station was installed. In the simulations, it is possible to adjust the rainfall intensity and river discharge in an area; therefore, situationally rainfall and flood damages can be foreseen. New inundation information can be provided and can be used for evacuation planning because the information includes the spatiotemporal behavior of floodwater. Keywords Flood inundation map · Evacuation plan · Risk communication · Nonstructural measures M. Sakai (B) · N. Todoroki · Y. Furumoto National Institute of Technology, Nagano College, Nagano 381-8550, Japan e-mail: [email protected] Y. Hashimoto Fujitec Construction Consultants, Nagano 381-2244, Japan T. Oomiya General Affairs Division, Obuse Town, Nagano, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_31

333

334

M. Sakai et al.

1 Introduction Disaster measures include “nonstructural measures,” such as refuge behavior and awareness generation. One of the methods used to implement nonstructural measures involves the provision of information to residents through risk information maps called “hazard maps” in Japan. The Ministry of Land, Infrastructure, Transport, and Tourism (MLIT), river administrators (MLIT and prefectural governments), and prefectural river authorities publish flood inundation map and designate areas that may be inundated in the event of flooding as flood inundation areas, based on the flood fighting act. Municipalities prepare and disseminate flood hazard maps for residents based on flood inundation area maps. A flood hazard map is an important tool used to disseminate flood risk information to residents. In September 2015, a heavy rainfall disaster occurred in the Kanto and Tohoku districts; houses collapsed were washed away by floodwaters caused by levee breaks, and many people delayed evacuation and were isolated in the flooded areas, and the importance of early evacuation was recognized. According to survey results in 2020, the recognition rate of flood hazard maps was approximately 30%, and public awareness was an issue [1]. However, the 2022 survey results showed that the recognition rate of the flood hazard maps was increasing to values which exceeded 70%; it included both printed-paper hazard maps and digital-hazard maps accessed via the Internet [2]. To achieve risk communication and evacuation planning for residents, it is important to develop effective flood hazard maps, and promote the understanding on various occasions [3]. Obuse is a small town in the northern Nagano Prefecture in Japan. The town had a population equal to 10,999 (in 2019) and a population density equal to 580 persons per km2 . The total area of the town is 19.12 km2 . The typhoon Hagibis led to the 2019 disaster that caused flooding of the Shinano River and enormous damage in the town of Obuse. Although no direct human suffering occurred, many houses were flooded, and the scales of damage were equal to 36 in above floor inundations in houses (total number, 118) and stores, 24 in underfloor inundations, and 58 in warehouses. In addition, approximately1.3 km2 were submerged in the riverbed and the fields in the area. Evacuation orders were issued to 673 households in six districts and included 2070 people. In the town of Obuse, a survey was conducted to assess the evacuation behavior [4]. In total, 50% of the residents understood the contents of the hazard map, 30% had seen it but did not understand it, and approximately 15% had never seen it. Nearly half of the people in some areas felt threatened when the alert level was highest and an evacuation order was issued; nevertheless, approximately 82.1% of the inhabitants did not evacuate, while 14.6% and 3.3% were evacuated to shelter and vertical evacuation inside the house, respectively. In the heavy rain caused by the typhoon Hagibis, there were many flood damages in small- and medium-sized rivers without hazard maps. Municipalities also need to understand the flood risks of these rivers and provide them to residents. The flood damage caused by the typhoon Hagibis in the town of Obuse was attributed to the Shinano River. However, the town is surrounded by three rivers, and residents are very anxious about flooding owing to the Matsukawa River. The Matsukawa River is a branch river of the Shinano River

Developing Effective Flood Inundation Maps for Risk Communication …

335

Japan

Nagano pref.

obuse town and Matsukawa river basin

Fig. 1 Outline map of the town of Obuse. Shown are the Matsukawa River, its geographical location, and catchment area used for the analyses

that flows from the Shiga Plateau to the town of Obuse via the village of Takayama and has a length of 38 km (Fig. 1). The flood inundation map was issued by the Nagano Prefecture six months after the typhoon Hagibis. The latest hazard map was provided in October 2020 based on the inundation map, and information on Matsukawa was also included. Most of flood hazard maps are provided in printed-paper form, and data contained therein can efficiently provide the areal extent of a flood event but do not provide spatiotemporal floodwater behaviors. In particular, the Matsukawa River contains currents flowing at high speeds, and the spatiotemporal behavior should be presented from the perspective of the Matsukawa River in the town of Obuse, which is in an alluvial fan. In this study, we conducted a flood simulation analysis for the Matsukawa River and examined the effectiveness of the analysis in comparison with the flood depth and spatiotemporal behavior results of the predicted flood inundation map published by the prefecture. We aimed to provide information of the digitalhazard maps include spatiotemporal behavior of flood by simple simulation to convey detailed information on flooding to residents about for medium-sized rivers.

2 Materials and Methods 2.1 Inundation Analysis Based on iRIC Inundation analysis was conducted by using the software iRIC (version 2.3) in conjunction with the Nays2DFlood solver. The iRIC analysis software for river flow and riverbed variation used the multidimensional surface-water modeling system (MD_SWMS), developed by the United States Geological Survey (USGS), and the

336

M. Sakai et al.

river interface corporative (RIC Nays) developed by the Foundation of the Hokkai-do River Disaster Prevention Research Center. Nays 2D Flood is a flood flow analysis solver that relies on unsteady two-dimensional planar flow simulations based on the use of boundary-fitted coordinates as the general curvilinear coordinates. Various analysis parameters can be set in iRIC. In an urban area, buildings may obstruct the flow of floodwater, thus causing the flow to concentrate on roads. In this case, the analysis was performed on a small computational grid, wherein the computational load increases. To avoid this, in the simulation of iRIC, it was used continuity equations and equations of motion that took into account the influences of buildings. These formulas reflect the building occupancy rate and the drag force (Cd) owing to the building. The data needed for an overflow calculation by Nays 2D Flood were topographic data, and data of inflow discharge and roughness of each river or each inflow point. In this study, the data used for analysis were collected according to the data collection method of Nays2DFlood_Solver manual [5]. Each dataset used here-in is described below. Analytical Data Collection and Conditions of Model and Hydraulic Parameters. Collection of Digital Elevation Model for Topographic Data. The digital elevation model (DEM) data were obtained from fundamental geospatial data provided by the Geospatial Information Authority of Japan [5] with a 5 m mesh covering the entire catchment area of the Matsukawa River. These were used as topographical data of the floodplain. In addition, the catchment area of the Matsukawa River was calculated by using the geographical information system (GIS) software (Arc GIS pro (2.6). ESRI). Data Collection for Calculation of the Building Occupancy and Setting of the Coefficient of Roughness. Land use data and building areas were obtained from the digital national land information download service [6] provided by the Ministry of Land, Infrastructure, Transport, and Tourism. Land use data were obtained as a land use refinement mesh, and seventeen Manning’s roughness coefficients of the floodplain were set based on this. From the building area, the building occupancy ratio, which indicates the area ratio of buildings within the analytical grid, was calculated. Creation of the Hydrograph for Analysis. Hydrographs for inundation analysis are created from hyetographs obtained by extending the actual, maximum precipitation. However, to confirm the accuracy of the results of models, it was necessary to use a discharge similar to the hydrograph used in the preparation of the flood inundation area map (implemented by the Nagano Prefecture) for the iRIC analysis conducted in this study. Therefore, the hydrograph was matched with the shape of the water level data that were used to create the inundation area map. Figure 2a shows the hydrograph used for the analysis by the iRIC. Analytical Condition of Two Models. Two types of inundation analysis were performed based on the following conditions. The assumed scale was the maximum rainfall, which is similar to the scale of the assumed probable maximum flood inundation area map provided by the Nagano Prefecture. It was calculated for the catchment area of Matsukawa River and was equal to 763 mm/day. The study area was set to

Developing Effective Flood Inundation Maps for Risk Communication …

Shinoi river drainage pump sta.on

2000 Dischage (m3/sec)

337

1500

Fig.1

5000 m

1000 Inflow point

500 0 0

5

10

15

Time (hour)

20

25

Ainoshima drainage pump sta.on Inunda.on analysis area with grids

Fig. 2 a Left side: hydrograph used for inundation analysis in the Matsukawa River. b Right side: scope of computational grid in the international river interface cooperative analysis, river inflow points, and installation locations of drainage pump stations

include the maximum range of the inundation area when the target river overflowed. In this study, we set up a computational grid which included the range predicted by the flood inundation area map of the Matsukawa River provided by the Nagano Prefecture (Fig. 2b). Model No. 1 used the DEM data that were downloaded as the terrain data without change. Drainage from a drainage pump station was not taken into consideration. In Model No. 2, we also considered the two drainage pumping stations (with drainage capacities equal to of 20 m3 /s) in Fig. 2b. The analyzed results were output in a format that can be displayed by using the geographic information system ArcGIS provided by ESRI and compared with the prefecture’s analysis data.

2.2 Visualization of Temporal Changes of Inundation Area The inundation map for each catchment provided by the prefecture was created according to the manual of the Ministry of Land, Infrastructure, Transport, and Tourism. Data on numerous variables, such as flood depth, flow velocity, and flood duration, were provided. Based on this map, municipalities create hazard maps indicating the maximum inundation depth for each point. This information about flooding was also provided as calculation data not only for paper maps. Therefore, it is possible to visualize the time course of flooding by using the data. We confirmed the Nagano Prefecture’s inundation data for Matsukawa River, and the results for 58 levee breakpoints and overflows without levee breakages are presented in the forms of movies created by ArcGIS. For comparison with the iRIC analysis, the results for overflow without levee breach were used. We created a video that expressed the spatiotemporal extent of the flooded area up to 480 min after the levee breach at 10 min intervals and used the flood depth.

338

M. Sakai et al.

3 Results and Discussion 3.1 Comparison of Hazard Map Data and iRIC Results by Maximum Inundation Depth Figure 3a shows the estimated maximal inundation depth (map created by the authors based on the data provided by Nagano Prefecture) referred to as the “prefectural inundation areas map,” and the results of maximum inundation depth in each computational grid for Model No. 1 (b) and Model No. 2 (c), respectively. The prefectural inundation area map (a) was created by using the maximum inundation depth of the grid during the time at which flooding occurred; this was shown to residents in municipalities as a hazard map. The iRIC analysis results for Model 1 (b) and Model 2 (c) are similar to the prefectural inundation area map (a). The direction of inflow from Matsukawa is indicated by an arrow in (a). This point is located at similar locations in (b) and (c), and the floodwater from the inflow point spreads radially toward the coast of the Shinano River. In coastal areas of Shinano River, areas with deep inundation extend in a belt-like shape. In addition, there is a deep and broad inundation area in the northeastern part of the town of Obuse. This reflects the topography of this town (which is located in an alluvial fan), indicating that the altitude decreases from the foot of Mt. Karita to the Shinano River. In the case of flooding of the Matsukawa River, it was reproduced such that the flow occurred from the lower part of the coast of the Shinano River upstream along the Matsukawa River toward the center of the town. The analyzed results of Model 1(b) shows deeper of inundation at the confluence of the Matsukawa and Shinano Rivers, and the paddy fields and upland fields in the northeastern part of the town than map (a). One reason for this is thought to be the model that does not consider drainage. The results of Model 2(c) improved these points by installing a drainage pump station, and the results were closer to the prefectural inundation area data. Especially, comparing the areas with flood depth ≥ 0.5 m, which is an indicator that evacuation is necessary in the event of a disaster, the results were similar. Accordingly, it was confirmed that the results were improved by considering the drainage pump station. However, even in the case of Model 2(c), some differences from the prefectural data were confirmed in the paddy area in the northeastern part of the town. The reason for this will be described in the next paragraph, which confirms the temporal trend of the flood of the analyzed results.

3.2 Comparison of iRIC Analysis Results and Prefecture Data Based on the Temporal Trend of the Flood The analyzed results by iRIC can show the temporal trend of flooding. Although the prefectural data are also limited, it is possible to confirm the spread at a specific

Developing Effective Flood Inundation Maps for Risk Communication …

339

unit: m

(a)

(b)

(c)

Fig. 3 Comparison between a estimated maximal inundation depth (map created by the authors based on the data provided by Nagano Prefecture), b analyzed inundation depth in the flood by Model No. 1, and c analyzed inundation depth in the flood by Model No. 2

time after the onset of flooding by processing with GIS. We compared these results. Inundation analyses by iRIC and prefectures were conducted based on discharge and river water levels, respectively. Therefore, to compare the two results, the time course of flooding was confirmed by matching the results of iRIC with the start time of inundation in the prefectural flood inundation area map data. In the hydrograph in Fig. 2a, which was used for iRIC analyses, the time at which the discharge reached the planned discharge of 620 m3 /s coincided with the inundation onset time of the prefectural flood inundation area map data. It was confirmed that the hydrograph used for the simulation was valid. Figure 4 shows the extent of flooded areas at 90, 180, 270, and 360 min after the onset of flooding in the cases of (a), (b), and (c), respectively. In Model 1(Fig. 4b), the temporal trend of flood expansion was similar with that in Fig. 4a, it was found that even a simple analysis can express the spread as a function of time from the onset of flooding. However, the inundation depth of the floodplain was deeper. Other differences in Model 1 with Fig. 4a include that inundation depth did not change even after 270 min when the hydrograph gradually decreased, and the size of the inundation area did not decrease. To simplify the iRIC analysis as much as possible, the obtained DEM data were used as topographical data without modification, and the drainage pumping stations were not considered. Therefore, even if the hydrograph gradually diminished, the water spread over the inundation area was not drained, and the inundation area did not decrease. Elevations that used as topology data were obtained from the original DEM data even at points at which sluice gates and box culverts were located; topographical features, such as embankments and bridges above them, are the elevations. In other words, the floodwater that was originally drained through the embankment by the sluice gate and box culvert was not drained by these points. The huge box culverts that connect to the main roads have been modified by correct data, and the topography of the upper layer has been removed; however, the small gates and box culverts have not been modified. The original DEM data had four such drainage failure points within the analysis area. Therefore, in Model 2, we installed two drainage pump stations and modified the DEM data of the upper layer (as shown in Fig. 5) to drain the water in the areas where

340

M. Sakai et al.

Fig. 4 Comparison between a estimated time temporal of inundation depth and area (map created by the authors based on the data provided by the Nagano prefecture), b temporal trend of inundation depth and area based on Model No. 1, and c temporal trend of inundation depth and area based on Model No. 2

drainage was inadequate. In the results of Model 2 (Fig. 4c), drainage was achieved; therefore, the results after 180 min showed that the flooded areas were similar to Fig. 4a in all areas. Even after 270 min, the inundation depth was excessive in the southwest flood area in Fig. 4b but was reproduced accurately in Fig. 4c. The two points that differed from Fig. 4a are the flooding occurred at intersections with roads and railways along the Matsukawa River, and that the tendency of flooding in the paddy fields in the northeastern part of the town of Obuse did not decrease which continued for 360 min. On the road which passed under the railway bridge along the Matsukawa River, the DEM data were processed according to the current situation; however, this was not reflected in the analysis grid and the movement of water could

Developing Effective Flood Inundation Maps for Risk Communication … Drainage is inadequate by embankment of DEM data.

341

Modified DEM data.

Fig. 5 Correction example of digital elevation model (DEM) data in areas where the embankment sluice was not modified and drainage was obstructed

not be expressed. Therefore, it did not flow down well, the inundation depth was higher than that in Fig. 4a, and it behaved as if it spread sideways. In the fields in the northeastern part, it was considered that the 10 m computational grid could not reflect drainage phenomena, such as small agricultural drainage channels and side ditches. However, in an actual disaster, there is a possibility that the drainage pump station and drainage channels will not function normally. In this analysis, if the spread of flooding can be expressed accurately with minimal processing, it is possible to present it as information that leads to evacuation behavior.

3.3 Accuracy Check of the Model We compared the estimated water depths of Models 1 and 2 with the inundation maps provided by the prefectures and evaluated them quantitatively. The results of iRIC were overlapped on the inundation depth of the prefectural data, the estimated value at the same point was extracted, and the values were compared. The numbers of points in (a) and (b) at which the analysis mesh intersected at inundation depths ≥ 0.5 m—indicating cases which may require evacuation—were equal to 45,845 and 45,792, respectively. Correlation was confirmed for this point. Linear approximation curves with the intercepts set to zero are drawn in Fig. 6. From the approximation curves, the correlation coefficient R is equal to 0.934 when the intercept was set to zero in Model No. 1, the determination coefficient R2 is equal to 0.873, and the slope is equal to 1.48. Even if the prefectural inundation data show that the height is less than 2 m, there are many points in Model 1 at which the estimated values were > 4 m, and this was found to be excessive. The approximation curve with the intercept set to zero in Model 2 had a slope equal to 1.041, a high-correlation coefficient R equal to 0.972, and a high determination coefficient R2 equal to 0.945. In both models, the prefectural data showed that there were many points in which the estimated values were underestimated, particularly in areas with deep water. In iRIC simulations, it was necessary to make residents aware of evacuation through the simplicity of the work and the obtained results. The operation performed in Model 2 was simple;

342

M. Sakai et al. 7

6 5

y = 1.4817x R² = 0.2662

4 3 2 1 0

iRIC result model No.2

iRIC result model No.1

7

6 5

y = 1.0431x R² = 0.7615

4 3 2 1 0

0 1 2 3 4 5 6 7 Maximun depth in prefectual data

0 1 2 3 4 5 6 7 Maximun depth in prefectual data

Fig. 6 Correlation between maximum inundation depth (m) in prefectural data and simulation results

however, the accuracy improved compared with that for Model 1. It is thus of value as it provides information on disaster situations.

4 Conclusions To connect the information of inundation map to actual evacuation actions, we first created videos from the information of the existing flood inundation map. Flood inundation simulations were then conducted with the use of the software iRIC in conjunction with the use of the solver Nays2DFlood. A terrain model was created from the DEM for simulations. Two types of inundation analyses were performed at different terrain models and drainage conditions. We confirmed that the current inundation map was made from the results of inundation simulation for 58 sites in the Matsukawa River levee. Videos were created by using the software GIS from the simulation when the levee breaks occurred at each site. In these videos, residents can observe the spread of the flooded area at the shortest interval of 10 min from the river levee break. In the iRIC simulations, the reproducibility was sufficient when a 10 × 10 m mesh was used, simple processing was added to the DEM for levees, and an additional drainage pump station was installed. The rainfall intensity and river discharge at the area was adjustable in the simulation. Therefore, it was easier to express the image of rainfall and flood damage to the residents. We can provide new inundation information for use in evacuation planning, including spatiotemporal floodwater behavior information. The results of this study can be utilized for risk communication in the disaster prevention seminar that has been held every year since 2020 in the town of Obuse. Furthermore, it is expected to be used for evacuation planning pertaining to flood risk management by local governments. Acknowledgements This work was supported by JSPS KAKENHI Grant Number 22K04661.

Developing Effective Flood Inundation Maps for Risk Communication …

343

References 1. Mobile Society Research Institute. NTT DOCOMO. https://www.moba-ken.jp/project/disaster/ disaster20220519.html. Last accessed 4 Oct 2022. In Japanese 2. Mobile Society Research Institute. NTT DOCOMO. https://www.moba-ken.jp/project/disaster/ disaster20200123.html. Last accessed 4 Oct 2022. In Japanese 3. About the ideal flood hazard map from the perspective of residents: Report of the Flood Hazard Map Exploratory Committee (2016). In Japanese 4. Questionnaire on Typhoon No. 19 in Obuse Town. https://www.town.obuse.nagano.jp/docs/bou sai.html 5. Nays2DFlood_SolverManual 6. Basic map information download service. Geospatial Information Authority of Japan. https:// fgd.gsi.go.jp/download/menu.php. Last accessed 4 Oct 2022 7. Digital national land information download service. Ministry of Land, Infrastructure, Transport and Tourism. https://nlftp.mlit.go.jp/ksj/

Disaster Preparedness and Perception of Earthquake Risk: A Case of Samoeng Nuea Sub-District and Samoeng Tai Sub-District, Chiang Mai, Thailand Thapthai Chaithong , Kanyaporn Parung, Jirapa Khoysungnoen, Panyawi Tanseenawanon, and Shupisara Jingsungnoen

Abstract The Wiang Haeng active fault, part of the Muang Haeng active fault group, is the newest active earthquake fault found in Thailand. The Department of Mineral Resources announced its discovery in 2019. This active fault has been identified as an NS normal fault with a length of approximately 100 km. The recently reported discovery of the fault may influence people’s awareness and perception of earthquake risk and disaster preparedness in the local region. The purpose of this study was to assess the perception of local people’s earthquake hazard risk and disaster preparedness in light of this active fault. A study was conducted on a sample of 649 households in the Samoeng Nuea and Samoeng Tai sub-districts of the Samoeng district of Chiang Mai, Thailand, which are located on the fault line. The results show that local people perceive of the risk of the active fault line that runs through their village. Keywords Active fault · Earthquake · Risk perception · Disaster preparedness

1 Introduction Local people in the Northern region of Thailand have experienced recent earthquakes, with the epicentre of several earthquakes located in Thailand and the neighbouring countries of Myanmar and Laos. For example, on 24 March 2011, a strong earthquake known as the Tarlay Earthquake, with a magnitude of 6.8 Mw, occurred in Tarlay, Myanmar (Latitude 20.705ºN, Longitude 99.949ºE) [1]. Moreover, on 5 May 2014, the Thai Meteorological Department reported a moderate earthquake, known as the Mae Lao earthquake, with a local magnitude of 6.3 ML in Chiang Rai province in Northern Thailand (Latitude 19.656ºN, Longitude 99.670ºE) [2]. These earthquakes

T. Chaithong (B) · K. Parung · J. Khoysungnoen · P. Tanseenawanon · S. Jingsungnoen Department of Geography, Kasetsart University, Bangkok 10900, Thailand e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_32

345

346

T. Chaithong et al.

caused soil liquefaction and damaged or destroyed several buildings [2]. As a consequence of these earthquake events, related government agencies and the local people in the affected areas are conscious of the risk of earthquakes. For instance, local schools and a university in the earthquake-risk zones have collaborated to develop a local earthquake disaster curriculum for primary and secondary school students [3]. This curriculum provides essential earthquake knowledge for students, which can build understanding and improve their chances of surviving an earthquake. However, the local people who live in areas far from the affected areas, but still reside in high seismic hazard zones, may have a different perception of earthquake risk. Hence, it is significantly important to study the perceptions of local people regarding the risk of earthquakes. Understanding people’s risk perception is a key component of improving risk management. It leads to the creation of awareness campaigns and guidelines for earthquake preparedness. Previous research has studied the concept of risk perception. These studies have examined the opinions that people express when they are asked to evaluate the risk of hazardous events or disasters [4]. Risk perception studies have been conducted for a variety of hazardous events such as floods, earthquakes, nuclear power plant accidents, or outbreaks of diseases. When considering a perceived risk, there are many factors that can influence a person’s judgement of the risk. People’s perception of risk depends on the nature of the risk, personal factors (such as age, sex and cultural or educational background) and external factors (such as the availability of scientific information, the economic situation of the individuals or the political decision-making practices in the community) [5]. In researching risk perception related to natural hazards, the characteristics of the natural hazard and the stakeholders are the main variables in these studies. The purpose of the current study is to assess, via questionnaire, local people’s perception of earthquake hazard risk and disaster preparedness in the Samoeng Nuea and Samoeng Tai sub-districts of the Samoeng district of Chiang Mai, North Thailand.

2 Study Area The study area included the Samoeng Nuea and Samoeng Tai sub-districts located in the Samoeng district of Chiang Mai, North Thailand. The Department of Mineral Resources, which recently analysed the study area, found that it occupies two seismic hazard zones, which are classified as strong (level VI) and very strong (level VII), according to the severity of an earthquake on the Mercalli scale. The Wiang Haeng active fault is a part of the Muang Haeng active fault group that runs through the Samoeng Nuea and Samoeng Tai sub-districts in a north–south direction and cuts through the middle of the community. The fault slip rate is approximately 0.01– 0.11 mm per year, and the length of the Wiang Haeng active fault is approximately 100 km. The maximum possible earthquake magnitude based on a paleo-earthquake magnitude estimation is 6.8 M. Figure 1 shows the fault line and seismic hazard zone in the study area [6].

Disaster Preparedness and Perception of Earthquake Risk: A Case …

Fig. 1 Seismic hazard zone and active fault line in the study area

347

348

T. Chaithong et al.

Topographically, the study area is composed of high mountains, and the local people live in the plains between the valleys. The maximum population density, found in the Samoeng Tai sub-district, is approximately 159 people per km2 . The minimum population density is approximately eight people per km2 . Compared to the Samoeng Nuea sub-district, Samoeng Tai has an overall higher population density. Moreover, considering the commercial and urban areas, the Samoeng Tai sub-district has a higher economic activity than the Samoeng Nuea sub-district. In regard to land use, forests, field crops and shifting cultivation are the majority of the land use types in the study area, with shifting cultivation most frequently found in the highlands. Figure 2 shows (a) elevation, topography and the active fault line, (b) land use and (c) population density in the study area.

3 Method In order to achieve the objective of the study, a questionnaire was constructed to survey the local people in the study area. There are a total number of 3567 households in the study area. Taro Yamane’s simplified formula to calculate sample sizes for finite populations was used in this study. It assumed a 95% confidence level (error 5%). The sample size was determined to be 360 households; however, the study collected survey data from a total of 649 households by conducting door-to-door interviews. There were three parts in the questionnaire as given in Table 1. The first part collected general socioeconomic information of the participants and related personal data. The second part assessed actual disaster preparedness for earthquakes and the third part assessed the participants’ personal perceptions of earthquake risk and preparedness.

4 Results and Discussion The questionnaire was given in interview form to 649 householders in the study area. Table 2 presents the demographic and socioeconomic data of the participants. In regard to the demographic and socioeconomic variables, the valid sample included 340 (52.39%) females, of which 331 (51%) were aged > 60 years, and 309 (47.15%) had a primary school education. For length of residence in the area, most of the sample had lived in the area since birth. The data showed that the majority of the households consisted of elderly people who lived with grandchildren, known as skipped generation families. When conducting in-depth interviews with the participants, the householders mentioned that the working-age adults—their children—worked in the Chiang Mai municipality or in Bangkok. This is a general phenomenon in Thailand and developing countries [7]. Only the grandparents live with their grandchildren because the parents have left them behind in the rural area and moved to the urban centres. Children and older adults constitute vulnerable populations, which are a priority consideration in disaster preparedness and management [8]. Communication

Disaster Preparedness and Perception of Earthquake Risk: A Case …

a) elevation and topography

349

b) land use

c) population density Fig. 2 a Elevation, topography and active fault line, b land use and c population density in the study area

350

T. Chaithong et al.

Table 1 Structure of the questionnaire Part

Questions

Point range

Socioeconomic

1. Sex

Male/female/NA

2. Age

Year/NA

3. Educational backgrounds

1. Uneducated backgrounds 2. Primary school 3. Junior high school 4. Senior high school 5. Vocational certificate 6. High vocational certificate 7. Bachelor’s degree or higher 8. Non-formal and informal education

4. Occupation

Basic disaster preparedness

Basic disaster preparedness

Perception of earthquake risk

5. Number of household members

…person

6. Length of residence in the area

…years

7. Do you have first aid equipment for emergencies?

0 = no 1 = yes

8. Does your family have food reserves for emergencies?

0 = no 1 = yes

9. Does your residence have heavy objects placed high above your head?

0 = no 1 = yes

10. Does your residence have furniture fixed to the wall?

0 = no 1 = yes

11. Have you ever received knowledge or training about earthquake disasters from any government agencies?

0 = no 1 = yes

12. Do you have disaster-related insurance that covers disasters caused by earthquakes?

0 = no 1 = yes

13. Do you know the location of the nearest evacuation centre in the event of a natural disaster?

0 = no 1 = yes

14. Have you ever participated in an earthquake evacuation drill?

0 = no 1 = yes

15. Did you know that the Wiang Haeng fault line runs through the Samoeng district?

0 = no 1 = yes

16. How concerned are you about living near the 1…5 Wiang Haeng fault? Not concern very concern 17. When an earthquake strikes, to what extent do you think you are prepared to handle it?

1…5 Not survive survived completely (continued)

Disaster Preparedness and Perception of Earthquake Risk: A Case …

351

Table 1 (continued) Part

Questions

Point range

18. How much do you think your property will be threatened in the event of an earthquake?

1…5 Undamaged complete loss

19. Do you think your family is prepared for future disasters?

1…5 No preparation well prepared

20. Do you think a local administrative authority 1…5 such as a sub-district administrative No potential high organization or a municipality in your area potential has the potential to manage an earthquake event in your area?

barriers and literacy were challenging issues during the interviews because the most of participants spoke a Northern Thai language as known as Kam Mueang and some of the participants were illiterate. While most of them can comprehend the central Thai language, the risk of communication issues for local people is also a concern. Because language and media access are key for improving risk perception and risk communication, it is important that authorities are aware of these circumstances [9]. Regarding occupation, the participants included 286 (44.07%) farmers and 192 (29.59%) self-employed individuals. In comparing between land use types and occupation, the survey found that there is a corresponding relationship, and field crops and shifting cultivation are the key types of agricultural land utilization. Self-employed individuals included mainly owners of community-based shops, such as grocery stores, local restaurant, barbershops and salons. To find out about actual disaster preparedness for earthquakes, the questionnaire asked about basic preparedness for survival in the early stages of a disaster and the presence of an evacuated centre in the local community. According to analysis, approximately 346 participants (53.31%) had a prepared first aid kit at home, 182 (28.04%) had prepared food reserves for emergencies and just one (1) (0.15%) reported having received knowledge or training about earthquake disasters. None of the participants had ever participated in an earthquake evacuation drill, but 343 (52.85%) knew the location of the nearest evacuation centre in the community. Moreover, none of the participants had a plan for purchasing insurance that covers the effects of natural disasters. Taking this into account, we can summarize that the risk capacity of the communities should be urgently developed to increase earthquake disaster risk knowledge. When conducting in-depth interviews with the participants, most of them knew or had heard the news about the Mae Lao earthquake in 2014; however, they thought that the epicentre of the Mae Lao earthquake was far from their community. Figure 3 shows the results of basic preparedness questions. As for participants’ perceptions of earthquake risk, 502 (77.35%) reported not knowing about the existence of the Wiang Haeng fault in the study area. The participants’ concern after learning about the Wiang Haeng fault that runs along their

352 Table 2 Demographic and socioeconomic data of the participants (n = 649)

T. Chaithong et al.

Number

Percentage

Male

309

47.61

Female

340

52.39

23

3.55

Variable Gender

Age 18–31 32–45

128

19.72

46–59

167

25.73

> 60

331

51.00

144

22.19

Education Uneducated backgrounds Primary school

306

47.15

Junior high school

92

14.18

Senior high school

55

8.47

Vocational certificate

13

2.00

High vocational certificate

9

1.39

Bachelor’s degree or higher

23

3.54

Non-formal and informal education

7

1.08

Occupation Farmer

286

44.07

Unemployed

77

11.86

Employee

93

14.33

Self-employed

192

29.59

Government official

1

Fig. 3 Results of basic preparedness questions (n = 649)

0.15

Disaster Preparedness and Perception of Earthquake Risk: A Case …

353

Table 3 Average rating and standard deviation of the perception of earthquake risk Questions

Mean

Standard deviation

1. How are you concerned about the Wiang Haeng fault near the living area?

3.94

1.25

2. When an earthquake strikes, to what extent do you think you can handle it?

3.88

0.95

3. How much do you think your property will be threatened in the event of 3.60 earthquake occurs?

1.28

4. Do you think your family prepared for future disasters?

3.89

1.03

5. Do you think a local administrative organization such as a sub-district administrative organization or a municipality in your area has the potential to manage with an earthquake event in your area?

3.78

1.17

villages was high, with an average rating of 3.94 of 5. When asked about their family’s preparedness and ability to respond in the event of an earthquake, the participants evaluated themselves as able to handle an earthquake, with an average rating of 3.88 out of 5. When asked about property damage due to an earthquake event, the participants evaluated the potential level of severity of property damage as moderate, with an average rating 3.60 out of 5. Moreover, the participants assessed their family’s preparedness for future disaster as quite well (average rating 3.89 of 5). Regarding the role of the sub-district’s administration for disaster management, the participants’ expressed a high level of trust (average rating 3.78 of 5) in the administration’s ability to manage and respond to an earthquake disaster. Considering to the perception of earthquake risk by age, it showed that participants of all ages were quite worried when they knew about the active fault lines near their villages. According to the analysis, the results show that local people with a bachelor’s degree level of education or higher perceived a significant risk and are aware of the information about the active fault lines. Table 3 presents the average rating and standard deviation of the perception of earthquake risk.

5 Conclusion This study focused on assessing the risk perception and basic disaster preparedness for earthquake hazards in two sub-districts in Northern Thailand. According to analysis, synthesis and interpretation, it has been concluded that the study area, which is in a rural location, has a high population of people vulnerable to natural disasters as a result of the mitigation of working aged adults to work in urban areas. More than a half of participants were unaware of the existence of an earthquake fault in their community. In addition, the participants thought that earthquakes could only have an indirect effect on their area. For basic disaster preparedness, none of the participants had ever participated in an earthquake evacuation drill, but a majority

354

T. Chaithong et al.

knew the location of the nearest evacuation centre. However, the food and water for emergency event, more than a half of participant do not prepare and concern about them. For perception of the earthquake risk, the participants perceive of the risk of active fault line that run though the village. In the overview, the participants evaluate themselves that they can response the earthquake hazard quite well. The analysis showed that the most vulnerable groups were the elderly with below compulsory education levels.

References 1. Mase LZ, Likitlersuang S, Tobita T, Chaiprakaikeow S, Soralump S (2020) Local site investigation of liquefied soils caused by earthquake in Northern Thailand. J Earthq Eng 27(7):1181–1204 2. Foytong P, Ornthammarath T Empirical seismic fragility functions based on field survey data after the 5 May Mae Lao (Northern Thailand) earthquake. Int J Disaster Risk Reduction 42:101344 3. Jantakoon J (2015) Disaster education: learning approach to disaster preparedness activities (Part 1). J Educ Naresuan Univ 16(4):188–201 4. Slovic P, Fischhoff B, Lichtenstein S (1982) Why study risk perception? Risk Anal 2(2):83–93 5. World Health Organization (2002) Establishing a dialogue on risks from electromagnetic fields. Geneva, Switzerland 6. Department of Mineral Resources Homepage. http://www.dmr.go.th/n_more.php?c_id=11. Last accessed 12 Oct 2022 7. Ingersoll-Dayton B, Tongchonlatip K, Punpuing S, Yakas L (2018) Relationships between grandchildren and grandparents in skipped generation families in Thailand. J Intergenerational Relat 16(3):256–274 8. Marshall J, Wiltshire J, Delva J, Bello T, Masys AJ (2020) Natural and manmade disasters: vulnerable populations. In: Masys A, Izurieta R, Reina Ortiz M (eds) Global health security. Advanced sciences and technologies for security applications. Springer, Cham 9. Bouder F (2014) Risk perception and communication. In: Teodorescu HN, Kirschenbaum A, Cojocaru S, Bruderlein C (eds) Improving disaster resilience and mitigation—IT means and tools. NATO science for peace and security series C: environmental security. Springer, Dordrecht

Estimating the Effects of Community Disaster Management Plan on Disaster Risk Reduction Literacy Using Propensity Score Analysis Takumi Sugahara , Shinya Fujimoto , Hiroyuki Honda , Hisatoshi Taniguchi , Tabihito Fujihara , and Yasuhiro Mitani

Abstract Improving the disaster risk reduction literacy of residents is crucial for reducing the damage caused by serious disasters. This study aimed to investigate the impact of developing a community disaster management plan (CDMP) on residents’ disaster risk reduction literacy. To formulate the CDMP, a disaster risk communication workshop was conducted to collect information on existing disasters and discuss disaster risks in the community. The results were compiled into a risk map and timeline to support proactive disaster management activities by residents. The study assessed the effect of participation in the workshop on disaster risk reduction literacy, using a post-workshop questionnaire to survey participants. Propensity score analysis was employed to infer causal effects. The results showed that full participation in the workshop significantly increased participants’ understanding of disaster threats and improved disaster risk reduction literacy. Keywords Risk communication workshop · Disaster risk reduction literacy · Propensity score analysis

1 Introduction Abnormal weather conditions caused by global warming have led to frequent occurrences of torrential rains that exceed expectations, resulting in loss of life due to floods and landslides. In Japan, local governments provide support to encourage T. Sugahara (B) Graduate School of Civil Engineering, Kyushu University, Fukuoka, Japan e-mail: [email protected] S. Fujimoto Graduate School of Sociology, Doshisha University, Kyoto, Japan H. Honda · H. Taniguchi · T. Fujihara · Y. Mitani Disaster Risk Reduction Research Center, Graduate School of Engineering, Kyushu University, Fukuoka, Japan © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_33

355

356

T. Sugahara et al.

evacuation before and after a disaster strikes. However, in a large-scale disaster, local governments may be unable to provide support due to damage or cease to function, reaching what the Japanese government calls “the limit of public help” [1]. To protect lives from disasters, “self-help” and “mutual help” are crucial. In Itoshima City, Fukuoka Prefecture, hazard maps have been created since 2012 as a soft type of disaster mitigation measure. However, these maps are only effective if residents who see them understand the risks described and the intentions behind their creation and act accordingly in the event of a disaster [2]. Disaster prevention education that only provides information is not enough to lead to coping behavior during a disaster [3]. Therefore, improving residents’ disaster risk reduction literacy, including disaster prevention and response skills, is necessary to prevent damage through residents’ countermeasures. Disaster risk reduction literacy refers to the ability to understand threats, make necessary preparations, and take appropriate actions against disasters when they occur [4]. To improve disaster risk reduction literacy, it is essential to provide residents with knowledge that can be used during a disaster, such as information on the threats that exist in the community and what preparations are necessary. Community-driven disaster prevention requires action plans beyond disaster prevention education that provides information. This study aims to examine and implement a disaster risk communication workshop (RC) method to improve the disaster risk reduction literacy of local residents. The method involves creating a risk map reflecting the disaster risks in the community and discussing a timeline, which is an action plan in case of a disaster, to recognize the risks in residents’ living areas. By creating these maps and timelines, a community disaster management plan (CDMP) is completed. This paper aims to demonstrate the effect of participation in the creation of CDMP through an RC on residents’ disaster risk reduction literacy.

2 Disaster Risk Communication Workshop 2.1 Design of a Disaster Risk Communication Workshop Methodology The aim of this study’s disaster risk communication (RC) workshop is to improve the disaster risk reduction literacy of participants, as defined by Hayashi, which includes understanding threats, proactive measures, and confidence in acting during disasters [4]. To address each element, we designed a workshop methodology that first focuses on improving “threat understanding” by increasing awareness of the disaster risks in the community. While hazard maps are one way to learn about disasters, they do not include vulnerability information, so the authors conducted field surveys and workshops with local residents to collect information on risky areas and the residences of people who need help during evacuation. The results were visualized on a risk map. To improve disaster risk reduction literacy elements of “advance

Estimating the Effects of Community Disaster Management Plan …

357

Fig. 1 Disaster risk communication workshop flow and CAUSE model

preparedness” and “confidence in actions during a disaster,” we planned a workshop to create a timeline. The timeline is an organization of “when, who, and what” in each phase leading up to the occurrence of a disaster, considering the situation in the residential area. This approach was used successfully after Hurricane Katrina in 2005 and Hurricane Sandy in 2012 [5], and our study uses my-timeline to enhance self-help and a district timeline with defined roles and rules to enhance mutual help in the living area. To conduct effective RC, we used the CAUSE model proposed by Rowan, which involves a five-step process of (1) credibility, (2) awareness, (3) understanding, (4) solutions, and (5) enactment [6]. By conducting workshops along this process, we aimed to achieve risk acceptance and handling it. Figure 1 shows the RC method we designed based on this model.

2.2 Conduct Disaster Risk Communication Workshops Areas of Study. The development of CDMP using the resident-participatory workshop method proposed in this study is constrained by workshop size, leading to limited participation. Therefore, to ensure resident input in the planning process and reduce disparities between total population and workshop participants, the study areas selected are Itoshima and Onojo cities in Japan, which have smaller populations. Itoshima City, located in the northwest part of Fukuoka, covers 215.7 km2 including mountains and coastline, and faces an aging population issue with 29.6% aged over 65 years. It experienced heavy rainfall and subsequent river overflow and landslides in 1953. In recent years, urbanization has increased the number of houses near rivers and mountainous areas, leading to an increase in flooding and landslides.

358

T. Sugahara et al.

Onojo City, located in the central part of Fukuoka, covers 26.89 km2 and has concerns about earthquake damage due to the Kego fault. In 2005, the city was hit by an earthquake with a magnitude of 7.0, damaging around 200 houses. Population in these areas is subdivided into wards with several hundred to several thousand residents (see Table 1). Creating a Risk Map. In this study, a “risk map” is developed by incorporating vulnerable areas such as hazardous locations in residential areas and housing into the conventional hazard map. The process to create the map involves a series of activities ranging from preliminary field surveys to a second workshop, as illustrated in Fig. 1. During the preliminary field survey, the local residents, city hall personnel, and experts examine the areas that have experienced disasters in the past and the vulnerable areas. In the subsequent workshop, more residents are invited to participate, and the risky areas are marked on a B0-sized paper map and discussed. The experts explain the geographical features and disaster characteristics to the residents to raise their awareness of the disaster risks in their living areas. Finally, temporary evacuation sites for different types of disasters, such as flooding, are identified, and evacuation routes to those sites are discussed. An example of a risk map created during the study is presented in Fig. 2. Formulate a Timeline. After creating the risk maps, the third and fourth workshops are dedicated to developing a timeline. In the third workshop, participants create a my-timeline to gain an understanding of what a timeline entails. They begin by listing the necessary actions to be taken in the event of a disaster. Next, they develop a my-timeline by linking each phase of the disaster to the required actions. In the fourth workshop, based on the actions considered in the my-timeline, participants develop a local timeline by considering the actions they cannot undertake alone, as well as the guidelines for using temporary evacuation sites. The final timeline created is shown in Fig. 2. Table 1 Ward population and disaster characteristics Study area

Ward

Population

Household count

Disaster characteristics

Itoshima City

Ihara

850

260

Located in a mountainous area, the site is at risk of river flooding and landslide disasters

Onojo City

Omaru

321

60

Takasu

816

268

Suenaga

150

125

Inokuchi

2002

730

Otoganadai

1599

580

Otogana

3297

1010

Otoganahigashi

1873

800

Kamabuta

3792

1370

The location of the site on a fault zone makes it vulnerable to damage from earthquakes

Estimating the Effects of Community Disaster Management Plan …

359

Fig. 2 Example of a risk map and timeline

3 Methodology for Measuring Effectiveness of Workshop Participation on Disaster Risk Reduction Literacy 3.1 Objectives and Survey Data for Disaster Risk Reduction Literacy Assessment The effectiveness of the RC is evaluated by assessing whether the disaster risk reduction literacy of residents who participated in the RC has improved. Propensity score analysis [7] is used for this evaluation. The data used in the analysis were collected through a questionnaire distributed after the risk map was completed. Therefore, the effect of developing the timeline has not been confirmed. Questionnaires are distributed to all households in the target area, and the head of the household is asked to fill out the answers on behalf of the household, as the head of the household usually participates as a representative in the RC workshops. The questionnaires are collected by community centers in the ward, completing the collection two months after distribution. Table 2 presents the timing and collection rate of the questionnaires in the ward. While it is unclear whether all community members were aware of the workshop, the RC was a voluntary effort that anyone could participate in. The sample size of the survey data is not large, and there is concern that analyzing the data separately for each ward may reduce the ability to detect the effect of literacy. Therefore, the heterogeneity of the data by ward is not discussed, and the data are integrated and analyzed. To clarify the effect of participating in the RC, only cases that never participated in the RC and cases that participated in the first and second workshops are included in the analyzed cases. Hereafter, the treatment group refers to cases that participated in the RC, and the control group refers to cases that did not participate.

360

T. Sugahara et al.

Table 2 Overview of the survey and timeframe for completion Area

Ward

Number of targets

Number of responses

Response rate (%)

Test period

Itoshima City

Ihara

260

156

60.0

4.2020–8.2020

Omaru

60

34

56.7

12.2020–2.2021

Takasu

268

129

48.1

2.2021–4.2021

Onojo City

Suenaga

125

55

44.0

5.2021–7.2021

Inokuchi

730

310

42.5

5.2020–7.2020

Otoganadai

580

166

28.6

4.2020––6.2020

Otogana

1010

370

36.6

12.2020–2.2021

Otoganahigashi

800

222

27.8

12.2020–2.2021

Kamabuta

1370

615

44.9

12.2020–2.2021

3.2 Effectiveness Measurement Method Using Propensity Score Analysis Selection Bias in Workshops. When assessing the effectiveness of a workshop attended by those who voluntarily participate, caution should be exercised in selecting the analytical method. This is because those who participate in RC workshops tend to have an interest in disaster prevention, and this interest is believed to influence both participation in RC and disaster risk reduction literacy. This situation is known as confounding, which can bias the results of the analysis between RC participants and non-participants. To address this issue, a randomized control trial (RCT) is usually conducted, where residents are randomly assigned to participate or not, and the differences between groups are examined to demonstrate the causal effects of workshop participation on disaster risk reduction literacy. However, it is not feasible to randomly assign participants in this study. Therefore, propensity score analysis was used to create a situation that approximates an RCT through statistical manipulation. Analytical Method. Propensity score analysis is used to adjust for bias between the treatment and control groups and estimate causal effects. The technique creates conditions similar to RCT, where other conditions besides the treatment are kept as similar as possible. To calculate the propensity score, a regression model predicting participation in RC is created, with confounding factors such as age, gender, and experience with disaster as independent variables. The resulting predictive value is used to match samples with similar propensity scores between the treatment and control groups. The effect of RC is then estimated by comparing the two groups, with unmatched samples excluded from the analysis.

Estimating the Effects of Community Disaster Management Plan …

361

3.3 Variables Used in the Analysis Treatment Variables and Covariates. The treatment variable for this study is defined as participation or non-participation in the RC workshops, with the first and second workshops being used as a binary dummy variable. Covariates are the variables that have an effect on both RC participation and disaster risk reduction literacy. Basic personal attributes such as gender, age, job, and presence of family members in need of assistance, experiences related to disasters such as membership in disaster prevention organizations and past disaster experience, as well as housing characteristics such as number of floors and housing structure were included as covariates. Propensity scores were calculated using a logistic regression model, with age as the only quantitative variable and other categorical variables as dummy variables entered into the model. Outcomes. Establishing items to measure disaster risk reduction literacy involves defining and selecting items that assess participant’s understanding of threats, proactive measures, and confidence in responding to disasters. Previous studies by Kawami [8], Matukawa [9], and Fujimoto [10] have used items for this purpose. In this study, the items listed in Table 3 were chosen as measures of disaster risk reduction literacy. The questionnaire asks participants to rate their agreement with these items on a 5point scale ranging from “1. Not applicable at all” to “5. Very applicable.” To score the responses, multiple correspondence analysis (MCA) was used to perform dimensionality reduction on the multidimensional qualitative data. This method, similar to principal component analysis, scores the response trends rather than assuming equal scores for each category. The R software “FactoMineR” was used for the MCA analysis. Table 4 presents the category scores obtained from the MCA analysis.

3.4 Analysis of Results Calculate Propensity Scores and Balance Check Covariates. First, a logistic regression analysis is performed using covariates as independent variables and participation in RC as the dependent variable, and propensity scores are calculated. The propensity scores are calculated and matched using R software. Next, the propensity score matching method is used to pair close scores of the treatment and control groups to check the balance of covariates (Table 5). The number of samples extracted as pairs by matching is 133. When adjusted by propensity score, the covariates are considered adjusted if their standardized difference is less than 0.1 [11]. Table 5 presents that many covariates are below 0.1, indicating that they are balanced compared to uncorrected. In particular, the treatment group has more members of disaster prevention organizations and more people who have experienced a disaster before the correction, which is expected to reduce the effect of confounding.

362

T. Sugahara et al.

Table 3 Elements of disaster risk reduction literacy and applicable questions Literacy elements

No

Questionnaire

Understanding of threats

1

Strong interest in each disaster (flood, landslide, earthquake)

2

Regularly obtains information related to disaster prevention

3

I know what disasters have occurred in my area in the past

4

Hazard maps show the danger areas in the living area

5

Have sufficient knowledge about disasters and countermeasures

6

Know the landslide warning area around the und home

7

Decide with your family how to contact you in case of a disaster

Proactive measures

Confidence in acting during disasters

8

Have an emergency carryout bag prepared

9

Usually stockpile drinking water, emergency food, etc

10

Furniture is secured in preparation for disasters

11

Know evacuation sites and routes in case of disaster

12

I often participate in local disaster drills

13

Able to act quickly and effectively when disaster strikes

14

Assuming what to do when disaster strikes

15

Disaster message boards and message dials can be used

16

When disaster strikes, choose to evacuate on your own, even if your neighbors have not

17

Unable to decide for themselves when to evacuate

18

Lack of confidence in their ability to determine when to evacuate

19

When you evacuate, you can call out to your neighbors to evacuate

Estimation of Causal Effects by Propensity Score Analysis. After adjusting for covariates using the matching method, the effect of participation in RC on disaster risk reduction literacy was analyzed. First, the effect of correcting for covariates by propensity score was checked. Therefore, the differences in literacy between the treatment and control groups were discussed in the uncorrected case and the corrected case (Table 6). Before correction, the t-distribution test showed that understanding of threats and confidence in acting during disasters differed significantly at a significance

Estimating the Effects of Community Disaster Management Plan …

363

Table 4 Category scores for each question item Understanding of threats

Before scoring

Not applicable at all Does not apply

After scoring Q.1

Q.2

Q.3

Q.4

Q.5

Q.6

1

− 1.49

− 1.21

− 0.61

− 1.08

− 1.32

− 1.11

2

− 0.93

− 0.84

− 0.36

− 0.69

− 0.75

− 0.81

Neither

3

− 0.42

− 0.16

0.01

− 0.22

− 0.04

− 0.19

Apply

4

0.13

0.53

0.39

0.50

0.89

0.36

Very applicable

5

1.09

1.86

1.62

1.59

2.46

1.60

Proactive measures

Before scoring

After scoring Q.7

Q.8

Q.9

Q.10

Q.11

Q.12

Not applicable at all

1

− 1.72

− 1.09

− 1.58

− 1.63

− 2.11

− 0.74

Does not apply

2

− 0.49

0.24

− 0.37

− 0.54

− 1.01

0.12

Neither

3

0.30

0.51

0.30

0.24

0.08

0.28

Apply

4

0.58

0.52

0.56

0.39

0.18

0.33

Very applicable

5

0.61

0.72

0.56

0.54

0.23

0.42

Q.18

Q.19

Confidence in Before scoring After scoring acting during Q.13 Q.14 disasters

Q.15

Q.16

Q.17

Not applicable at all

1

− 1.27 − 0.90 − 0.41 − 1.08 − 1.19 − 0.46 − 0.74

Does not apply

2

− 0.73 − 0.61 − 0.31 − 0.97 − 1.02 − 0.40 − 0.73

Neither

3

− 0.17 − 0.22 − 0.10 − 0.56 − 0.57 − 0.26 − 0.36

Apply

4

0.96

0.70

0.44

0.19

0.33

0.27

0.12

Very applicable

5

2.70

2.18

1.66

1.46

1.64

1.28

1.32

level of less than 1%, while proactive measures showed no significant difference. After correction, understanding of threats and confidence in acting during disasters remained significantly different according to the t-distribution test at a significance level of less than 1%. Discussion. In the analysis of this study, the covariates are not fully balanced, including some occupations. However, we believe that the confounding effect is minimal compared to cases where the covariates are biased. Propensity score analysis revealed that participation in a risk mapping workshop facilitated “understanding of threats” and “confidence in acting during disasters.” The process of creating a risk map involves discussions about risk in the living area not only among residents but also with city hall and experts, leading to increased recognition of the spatial distribution of disaster risks among residents. As a result, differences in “understanding

364

T. Sugahara et al.

Table 5 Balance check of covariates Balance before correction

Balance after matching

Treatment

Control

Std. diff.

Treatment

Control

Std. diff.

Male

109

474

0.78

92

87

0.01

Female

41

778

–

41

46

–

20–40 y/ o

0

4

–

0

0

–

40–60 y/ o

6

239

–

6

6

–

60–80 y/ o

31

481

–

29

37

–

80–y/o

113

528

–

98

90

Average

65.68

55.83

0.85

65.3

64.6

0.06

Clerk

6

127

0.31

6

8

0.08

Public servant

4

46

0.06

4

3

0.05

Profession

6

119

0.28

6

5

0.04

Primary industry

17

14

0.32

10

8

0.05

Construction

10

44

0.13

7

10

0.09

Service industry

8

157

0.32

8

7

0.03

Production

5

28

0.06

5

5

0

Transportation

5

31

0.05

5

8

0.13

Housewife

28

377

0.29

28

23

0.1

Retired

43

146

0.38

38

38

0

Unemployed

0

9

0.09

0

0

0

Student

0

1

0.03

0

0

0

Other

18

153

–

16

18

–

Belonging

41

54

0.52

28

31

0.05

Belonged

30

65

0.37

26

28

0.04

Never

79

1133

–

–

–

–

People who Yes need support No

26

192

0.05

21

22

0.02

124

1060

–

112

111

–

Housing structure

Wooden

120

967

0.07

106

104

0.04

R/C structure

10

88

0.01

8

10

0.06

Gender Age

Job

Disaster prevention organization

Items

Other

20

197

–

–

–

–

Disaster experience

More than once

33

203

0.14

27

30

0.05

Never

117

1049

–

106

103

–

Number

150

1252

133

133

Estimating the Effects of Community Disaster Management Plan …

365

Table 6 Effects on disaster risk reduction disaster prevention literacy Uncorrected Treatment average

After matching Control average

t-ratio

Treatment average

Control average

t-ratio

Understanding of 0.46 threats

− 0.05

8.62**

0.41

0.06

4.11**

Proactive measures

0.10

− 0.01

1.90

0.08

0.05

0.47

Confidence in acting during disasters

0.38

− 0.05

7.26*

0.36

0.06

3.45*

**

p < 0.01, * p < 0.05

of threats” and “confidence in acting during disasters” were observed between the treatment and control groups. Thus, the risk mapping workshop, which promotes residents’ risk perception and describes temporary evacuation sites and evacuation routes, is an effective activity. However, no effect was confirmed for proactive measures, as it is assumed that these elements may be effective through the creation of a timeline. It remains to be seen whether the creation of a timeline will have an effect on disaster risk reduction literacy. In the present analysis, correction by propensity score did not change the results of the t-distribution test. However, the literacy scores of the control group increased by approximately 0.1 points from pre-correction to post-correction, indicating that the differences in literacy scores between groups were overestimated before the correction. This suggests that even if a simple between-group comparison confirms a difference in literacy between the treatment and control groups, it is not possible to determine whether this is solely due to participation in the RC.

4 Conclusion In this study, we utilized the CAUSE model as a reference to design risk communication (RC) methods, including the development of risk maps and timelines, and applied them in nine wards located in Fukuoka prefecture. Our primary focus was on workshops conducted to facilitate the development of risk maps among RC participants, and we aimed to investigate whether participation in these workshops could improve disaster risk reduction literacy. To address potential confounding effects, we employed statistical correction techniques using propensity scores. Propensity score matching revealed significant differences between the treatment group (workshop participants) and the control group in terms of two components of literacy: understanding of threats and confidence in disaster response. These results confirmed that RC participation positively impacts these two factors. However, we did not observe significant changes in proactive measures. While our study focused on the effects of

366

T. Sugahara et al.

risk maps on literacy, important issues remain to be addressed, particularly in investigating the effects of timelines. Therefore, it is imperative to continue discussing and developing strategies to improve disaster risk reduction literacy and prevent damage in the event of a disaster. Furthermore, further research is warranted to consider measures that promote CDMP, as well as enhance disaster risk reduction literacy in urban areas, including the city center of Fukuoka prefecture, which was not included in our present study. In conclusion, our findings highlight the potential benefits of participation in risk mapping workshops as part of an RC intervention in improving understanding of threats and confidence in disaster response. However, further research is needed to explore the effects of timelines and proactive measures, as well as to consider measures to promote CDMP and improve disaster risk reduction literacy in urban areas. Acknowledgements This work was supported by JST SPRING, Grant Number JPMJSP2136.

References 1. Cabinet Office Japan.: White Paper on Disaster Management 2015, Japan (2015). 2. Katada T et al (2007) Desirable utilization of flood hazard maps for risk. Proc JSCE Div D 63(4):498–508 3. Atsushi Tanaka et al (2011) A survey for residents’ behavior affected by a torrential rain: a case study from the town of Sayo in Hyogo prefecture in 2009, research survey reports in information studies, interfaculty initiative in information studies, The University of Tokyo, vol 63. The University of Tokyo, Japan, pp 49–99 4. Ota T, Matuno I (2016) BOUSAI literacy. Morikita Publishing, Japan 5. Matuo I (2016) Timeline: Japan’s disaster preparedness will change. The Daily Engineering & Construction News, Japan 6. Rowan KE (1994) Why rules for risk communication are not enough: a problem-solving approach to risk communication. Risk Anal 14(3) 7. Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70(1):41–55 8. Kawami F et al (2016) A Study of nonlinear interaction effects of disaster risk reduction literacy with seismic hazard risk, physical and human damages perception on risk avoidance behavior: report of 2015 Hyogo prefecture survey on preparedness. J Soc Saf Sci 29:135–142 9. Matsukawa A et al (2019) Impact evaluation by propensity score analysis of inclusive disaster drill. J Soc Saf Sci 35:279–286 10. Fujimoto S et al (2019) Theoretical model construction and empirical verification of protective action decision making during disaster: structural equation modeling of the social survey on landslide disasters in Oita prefecture. J Soc Saf Sci 35:305–315 11. Nakamura N (2009) Data science in R Vol2: methods for multidimensional data analysis. KYORITSU SHUPPAN CO., LTD, Japan

Research on Methods for Determining and Understanding the Soundness of Retaining Walls Using Image Analysis Kenki Owada, Anurag Sahare, Kazuya Itoh, and Naoaki Suemasa

Abstract In high-density urban areas, retaining walls have been used for retaining earth when a housing lot is built in sloping terrain by cutting and filling earth. Although standards for residential retaining walls are set by the Building Lots Development Regulation Law, many retaining walls built before the law came into effect remain in unqualified conditions. It has been pointed out that when a large earthquake occurs, aging retaining walls can be damaged, affecting evacuation and disaster relief activities, restoration of residential areas, and reconstruction of daily life. The 2016 Kumamoto earthquake collapsed retaining walls and broke down into gutters in residential areas mainly in the Kumamoto metropolitan area and the Aso region. When a housing lot gets damaged from a disaster, a retaining wall needs to be restored before any of the damaged house. Therefore, it is important to promote reinforcement of existing aged retaining walls, which will lead to the resilience of cities, smooth rescue, and cost reduction for restoration. In this study, we suggest a method to assess vulnerability of a retaining wall in a simple way, aiming to develop an application for judging the danger level from images by utilizing machine learning. We used CNN as a deep learning method, which is an effective method for image recognition, to identify a type of retaining walls. The number, condition of drainage holes and the method of identifying cracks using median filters were also investigated. Keywords Housing retaining walls · Drainage holes · Crack detection

1 Introduction In urban areas with high land-use density, many retaining walls were built before the enactment of the Building Lots Development Regulation Law. In the event of a large earthquake, many aging retaining walls will be damaged, which will affect evacuation and disaster relief activities, restoration of housing lots, and reconstruction of people’s lives [1]. The Kumamoto earthquake that occurred on April 14, 2016 caused K. Owada (B) · A. Sahare · K. Itoh · N. Suemasa Tokyo City University, 1-28-1 Tamadutsumi Setagaya, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_34

367

368

K. Owada et al.

damage to more than 180,000 houses and caused large-scale landslides, mainly in the Kumamoto metropolitan area and the Aso region, as well as led to collapse of multiple retaining walls [2]. In the event of a housing disaster, such as the collapse or sliding collapse of a housing retaining wall, it is necessary to restore the retaining wall prior to the reconstruction of any of the residential houses. Therefore, it is important to develop a method for seismic diagnosis of existing aged retaining walls and to promote preventive measures by reinforcing them. The purpose of this study is to understand the seismic resilience of aged retaining walls. In this report, we use a convolutional neural network (CNN) and image feature extraction to identify the type of retaining wall. In addition, a method for identifying the number and location of drainage holes was studied using an existing application. For the detection of deterioration due to an earthquake loading such as cracks, we attempted to use noise reduction using median filters and Grad-CAM on the aged retaining wall images.

2 Retaining Wall Type Identification with CNN 2.1 Outline of the Study CNN is used for identifying the different type of retaining walls in this study. First, images of retaining walls that can be used as a main data were collected and divided into three types: concrete retaining walls, concrete block retaining walls, and masonry retaining walls (including dry masonry retaining walls), and then, the images are processed. A CNN model is then created, and the images are examined to check the accuracy of the separation mechanism.

2.2 Collected Images and Image Processing The images used in this study were taken with a smartphone, and 20 images were taken for each type of wall. The object distance was set at 1–2 m so that no obstacles other than the retaining walls were included. The images were processed using a simple segmenting image method, thresholding, in which binary images are created from a grayscale image. Thresholding replaces each pixel with a white (or black) pixel if it is larger than an assigned threshold value and with a black pixel (or white) otherwise. Case 1 is a process in which pixels less than the fixed values become white, and Case 2 is a process in which pixels greater than the fixed values become white. In this study, five threshold values: 117, 122, 127, 132, and 137 were assigned and applied for each image to increase the amount of main original data. 100 pieces of original data were examined for each image of retaining wall. Figure 1 shows an example of original data processed with a threshold value of 127.

Research on Methods for Determining and Understanding …

(original image)

(Case1)

369

(Case2)

Fig. 1 Examples of created original data

2.3 Building Machine Learning Environments and Models A model to identify the type of retaining wall was built using Conv2D, a twodimensional convolutional layer [3]. The convolution layer was a 3 × 3 filter with 16 layers, and the ReLU function was applied as the activation function. The ReLU function was also used as the activation function for all the coupling layers. The pooling layer outputs the largest value in a “2 × 2” region of the input image. A softmax function was used as the activation function for the output layer, and a predicted probability of each output was calculated so that the sum is 1.

2.4 Study, Results and Discussions In examination, 90% of the prepared data is used for investigation and 10% for validation. In addition, loss function values can be checked to verify the accuracy of the model. A loss function value is a deviation between estimated and true output values from the model. Here, one training dataset was iterated about 30 times. Figures 2 and 3 show the training accuracy results for Case 1 and Case 2, and Figs. 4 and 5 show the loss function values for Case 1 and Case 2, respectively. Accuracy represents the training accuracy, loss represents the loss function values, and Epoch represents the number of times the training data was investigated. The graphs show that the accuracy of both Case 1 and 2 increases as the number of training sessions increases. The final loss function values were 1.03 × 10–7 in Case 1 and 6.99 × 10–7 in Case 2. Since the loss function value is preferred to be smaller than 1, it can be read that both Case 1 and 2 are accurate enough. However, in the latter part, accuracy becomes 1, which indicates that overlearning had occurred. The reason for the overlearning may be the lack of original data for the CNN model.

370

K. Owada et al.

Fig. 2 Case 1 learning accuracy

Fig. 3 Case 2 learning accuracy

3 Retaining Wall Type Identification from Feature Extraction 3.1 Outline of the Study We examine whether it is possible to identify a type of retaining wall by extracting features from the images of a retaining wall. The images used for the study are the 60 original images collected in 2–1. and arbitrary images collected from the Internet. The procedure is to extract the features of an arbitrary image and select an image from the 60 original images which has the most similar features. If a same type of

Research on Methods for Determining and Understanding …

371

Fig. 4 Case 1 loss function value

Fig. 5 Case 2 loss function value

retaining wall is correctly selected from similar images, it can be said that the type of retaining wall can be determined from the image features.

3.2 Results of Study Figure 6 shows a result example of the verification. The verified image is a masonry retaining wall, and a similar masonry retaining wall is correctly selected by the extracted features. Also, in other types of images, walls were properly selected.

372

K. Owada et al.

Fig. 6 Feature extraction results

Therefore, it is confirmed that the extraction of image features is effective in identifying a type of retaining wall and is more reliable.

4 Extracting and Grasping Holes 4.1 Outline of the Study Two types of retaining walls are used: concrete retaining walls and concrete block retaining walls. We examine whether the number and location of drainage holes can be determined by two methods: template matching and the existing application “Count Things [4]”. The images were taken with a smartphone at a distance of 1–2 m from target retaining walls.

4.2 Grasping with Template Matching Template matching is a method for finding the location of a template image in an image. In this study, the normalized correlation coefficient was adopted, and a program was created by referring to the literature [5] so that the areas similar to the template image can be enclosed with a red frame. Here, the degree of similarity can be changed for each time, and a cropped image of a drainage hole in a concrete retaining wall is used as the template image. Figure 7 shows how the template image was extracted. The program tested various retaining walls, and it was found that changing the degree of similarity allowed the position of the drainage hole to be enclosed in a red box, but irrelevant locations were detected.

Research on Methods for Determining and Understanding …

373

Fig. 7 Creating template images

4.3 Grasping with “Count Things” The application “Count Things” is mainly used in business and research to count the number of objects. On the smartphone application “Count Things”, the option, “metal tubes”, was selected based on the appearance of the target drainage hole, a retaining wall image photographed with its camera function was selected, and the number of halls was counted. As a result of the verification, the location and the number of drainage holes were correctly identified on the whole as shown in Fig. 8. However, as shown in Fig. 9, it was not possible to detect drainage holes containing objects such as stones or empty cans inside. In addition, drainage holes with small diameters were likely to be overlooked because they are not designed specifically for retaining walls.

5 Crack Detection 5.1 Outline of the Study An attempt is made to detect cracks in the images of both aging concrete and concrete block retaining walls. Cracks are extracted by creating a smoothed image using a median filter and subtracting the difference between the original image and the smoothed image, based on the prior studies [6]. Then, a morphological transformation is performed to highlight the crack lines for easily detecting them.

374

K. Owada et al.

Fig. 8 Successes at Count Things

Fig. 9 Failures in Count Things

5.2 Background Subtraction Processing Morphological processing was performed to enhance crack lines using background subtraction processing. Morphological processing consists of two processes: shrinking and expanding that can remove noises around or within objects. In this study, we used the opening process that expands after shrinking and the closing

Research on Methods for Determining and Understanding …

375

process that shrinks after expanding. The results showed that the opening process reduced noises with leaving the crack lines intact, while the closing process did not show any significant change in both the concrete retaining wall and the concrete block retaining wall.

5.3 Crack Line Detection Crack lines were detected and colored in the image using a program that detects straight lines. The minimum length of the lines that were likely to be straight and the length between the points of the lines that were likely to be the same were determined according to the crack width. The images were then superimposed on the original images for comparison to investigate the degree to which cracks have been detected. Figures 10 and 11 show the results of crack line detection and coloring for each retaining wall image. Thick crack lines were detected in both types of retaining walls. However, joints were also detected and colored in red in the concrete block retaining wall. Fig. 10 Crack line detection result (concrete retaining wall)

376

K. Owada et al.

Fig. 11 Crack line detection result (concrete block retaining wall)

6 Anomaly Detection Using Grad-CAM 6.1 Outline of the Study Grad-CAM is a visualization method used for visualizing image classification by CNN [7]. This paper examines the possibility to detect anomalies in images of degraded retaining murals using this method. First, a retaining wall image is preprocessed by resizing it to 224 × 224 and normalizing it. After that, the models are acquired. In this study, five models (AlexNet, vgg, resnet, densenet, and squeezenet) were acquired to compare the Grad-GAM. We also set up a layer to obtain the gradient of each model and a function to superimpose the gradient on the original image. Change in detection was also examined depending on the degree of gradient. In addition, the results were verified using Grad-CAM++ as well.

6.2 Results of the Study Some of the results using Grad-CAM and Grad-CAM++ are shown in Fig. 12. The results from first case show that cracks and leaks are red when the gradient is high in all models. When the slope was reduced to detect the deteriorated areas more clearly in second case, AlexNet was the most successful in color-coding the deteriorated areas. Therefore, it was found that it is effective to change the degree of slope according to the degree of deterioration.

Research on Methods for Determining and Understanding …

377

Fig. 12 Results with Grad-CAM

7 Conclusions The lack of training data was a problem in identifying a type of retaining wall using CNN. However, it was possible to discriminate among concrete retaining walls, concrete block retaining walls, and masonry retaining walls based on extraction of the features in the images. For identifying drainage holes, it is necessary to learn and detect circular pipes using “Count Things” and to improve the system to be specific to retaining walls. It was found that cracks can be extracted to some extent by background subtraction. However, in the detection of crack lines, many false detections occurred because the noises were not eliminated enough or the joint lines between blocks could not be distinguished from crack lines. Therefore, changing the filter type or removing joint lines beforehand could be effective to extract crack lines only. It was also confirmed that the Grad-CAM can be used to color-code deteriorated areas.

References 1. Japanese Ministry of Land. https://www.mlit.go.jp/tec/gjutu/kaihatu/pdf/r1/190711_04jizen. pdf. Last Accessed 10 Oct 2022

378

K. Owada et al.

2. Hashimoto Takao KF, Matsushita S (2018) The residential land damage analysis by the 2016 Kumamoto Earthquake. Jpn Soc Civ Eng AI (Struct Earthquake Eng) 74(4):I_522-I_533 3. Child Programmer Homepage. https://child-programmer.com/ai/cnn-originaldataset-sample code/. Last Accessed 14 July 2021 4. Count Things Homepage: https://countthings.com/ja/, last accessed 2022/01/29. 5. Copyright Homepage. https://pystyle.info/opencv-template-matching/#outline_6. Last Accessed 29 Jan 2022 6. Yusuke Fujita F, Hideaki Nakamura S, Yoshihiko Hamamoto T (2012) Classification of crack widths in concrete structures by image processing. Ann Proc Jpn Concr Inst 34(1):1792–1797 7. Yurui Deep Learning Homepage (2022). https://www.yurui-deep-learning.com/2021/02/08/ grad-campytorch-google-colab/. Last Accessed 9 Sep 2022

Education for Sustainable Development Goals

A Decision Support Tool for the Sustainability Rating Index for the Maintenance of Low-Volume Rural Roads in India Raji Reddy Myakala and S. Shankar

Abstract Rural road maintenance with minimal resources in developing nations like India seems complicated. Deferred rural road maintenance affects society and the environment, which may influence maintenance expenses. A sustainability rating tool seems to be a great approach to compiling, describing, and assessing these effects. In any case, by considering users and neighborhood demands, low-volume rural road maintenance (LVRRM) can provide better services, including increased connectivity and safety. For distinct projects to be compared uniformly, it is essential to consider evaluating the sustainability of LVRRM. Examining the significance and advantages of sustainable LVRRM only informs the sector that it is essential and viable. The rating (evaluation) tool can be used for evaluation purposes and has other benefits. The primary goal of this research is to create a tool for the best creation of long-term maintenance plans LVRRM. As a result, the rural road sustainability index for maintenance (RRSIM), a novel sustainability assessment instrument, is created and may be utilized to assess the sustainability of LVRRM. The quadruple bottom line (QBL), which stands for environment, economy, society, and material, will be covered by this instrument, which also examines eight sustainability assessment systems. The tool tries to minimize green house gas (GHG) emissions resulting from the administration of maintenance solutions while maximizing the network’s long-term efficacy, given a maintenance budget. Keywords Road assessment · Rural road maintenance · Sustainability indicators · Sustainability rating tool · Sustainable maintenance programs

R. R. Myakala (B) · S. Shankar Department of Civil Engineering, Transportation Division, National Institute of Technology Warangal, Warangal, Telangana, India e-mail: [email protected] S. Shankar e-mail: [email protected] S. Shankar Department of Civil Engineering, Transportation Division, National Institute of Technology, 506004 Warangal, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_35

381

382

R. R. Myakala and S. Shankar

1 Introduction A nation’s economy, industry, culture, and social fabric development depends heavily on its well-maintained and large road system [1]. In developing nations like India, which make up about 80% of the total road network, the importance of low-traffic rural roads is quite significant [2]. These roads receive little traffic; therefore, maintenance is neglected and not routinely performed [3]. However, scientific planning is necessary to maintain such a vast system of rural roads to make the greatest use of the funding sources and expedite repairing damaged roads [4]. Any repair or maintenance work on the road seems to be either corrective or preventive. However, these advantages would be significantly reduced if these resources were not properly maintained [5]. Therefore, low-volume rural road system’s timely maintenance is crucial in attempt to gain from newly generated assets. The order in which maintenance tasks are prioritized relies on a number of variables, including the current state of the road—that is, the deterioration’s quality and quantity, the rate at which deterioration is accelerating, the significance of the various sections, etc. As a result, it can be challenging to decide which tasks in a road network should be prioritized for maintenance. Therefore, the creation of a sustainability rating tool has been crucial for maintenance tasks to be conducted in a low-volume road system. A self-evaluation technique designed for civil infrastructure works, such as I-LAST, Green roads, INVEST, GreenLites, and BE2ST-InHighways, has been frequently included in the rating framework. Software tools for planning and predicting have been included in the second class, which has a much broader scope. Several tools have already seen extensive use, including the Scottish transport appraisal guidance (STAG) and the UK Department of Transport Analysis (Web TAG). These tools have been used to (i) define guidelines for evaluating possibilities, (ii) offer expert advice for transportation works, and (iii) exhibit best practices. In order to increase road sustainability effectiveness predicated on the specified criteria, instruments in this method use standards to provide data on best practices and processes associated with optimal road projects [6]. However, there are some limitations, like higher processing time. Therefore, this research work is done. The four categories of LVRRM-related problems noted above can be discussed regarding sustainability as QBL.

2 Development of Rural Road Sustainability Index for Maintenance (RRSIM) The RRSIM, a newly introduced rating instrument, assesses the viability of rural road maintenance projects. This tool will include matters linked to the implications under the QBL. RRSIM’s objectives are to: Establish a link between the theory and the practice of sustainable LVRRM; a simple index where users may seek sustainability initiatives and contrast the sustainability of various initiatives or projects; create a

A Decision Support Tool for the Sustainability Rating Index …

383

certification system for awards based on the extent of sustainability attained; tracking and measuring the progress of sustainability targets; and inform the public about sustainable maintenance procedures. The five main essential steps used to create RRSIM are described as follows: • • • • •

Establish the rating categories Identify each category’s rating indicator Determine each category’s priority Identify each indicator’s point and Identify the extent of certification.

2.1 Rating Categories Determination The selection of the rating categories seems to be the initial stage in creating a sustainability assessment tool. Each category concentrates on particular sustainability metrics. The categories chosen for RRSIM and rating indicators are shown in Table 1.

2.2 Category Priority Determination Each RRSIM category’s weighting was established using the analytical hierarchy process (AHP). The AHP seems to be a strategy for rating many factors and making judgments that Thomas Saaty created in the late 1970s. The AHP was widely utilized across numerous industries and has become a recognized industry standard [7, 8]. The AHP divides the decision issue into a number of sub-issues hierarchy, determines the importance of each sub-issue using the expertise of experts or consumers, and then assigns quantitative weights to the entire hierarchy. A representative team, which included the following individuals, was asked to comment on the relative importance of each category: Employees from transportation agencies and maintenance companies; civil engineering instructors and students who had taken at least a single class on a sustainable layout; pedestrians, cyclists, and motorists whom rural road maintenance projects have traversed; working crew people (includes technicians as well as engineers); and residents and employees in an area where the country road has been kept up. Expert choice, an AHP programme, evaluated the group’s feedback to determine the sequence in which the eight categories should be prioritized. The ultimate priorities have been determined by calculating and normalizing the median of the various outcomes. According to the findings, safety (0.285) came in first, followed by management (0.152), environment (0.146), technique (0.114), material (0.109), social (0.088), economy (0.082), and innovation (0.054); the findings were shown in Table 3.

384

R. R. Myakala and S. Shankar

Table 1 Rural road sustainability index for maintenance rating indicators Category

Indicator

Note

Environment

Forest area

When maintenance efforts affect a forest area, it should be repaired and preserved as much as feasible

Atmosphere

Atmospheric pollution must be taken into consideration

Weather

According to the climate and weather, maintenance projects must be scheduled and carried out

Energy

Consideration should be given to the energy consumed by moving vehicles

Social

Empowerment Service

Enhancing public service quality and availability, among other things

Public management

Technique

Economy

Cultural heritage

The rural road system’s cultural components should be taken into account

Value addition

While reducing costs, maintain the features and quality of the same

Livestock

Quantitative social indicators concentrate on rural road maintenance

Material production

It is essential to consider the asphalt plant’s sustainability

Standard procedure

For reliable outcomes, a standard maintenance process should be used

Smoothness adjustment

A suitable strategy should be used to eliminate glaring flaws between managed and nearby low-volume roadways

Maintenance techniques

If a specific maintenance strategy is chosen from a range of available ones, justifications must be provided

Repair and disturbance

Repair any harm done to the nearby infrastructure

Market supply–demand

Keeping roads in good condition increases market supply–demand

Transportation

Transportation infrastructure development in relation to economic growth

Supply chain

The availability of maintenance materials is directly impacted by changes in interest rates and prices, which in turn affects the efficiency of the entire supply chain

Financial investment

The amount of money invested in road maintenance projects will increase or decrease (continued)

A Decision Support Tool for the Sustainability Rating Index …

385

Table 1 (continued) Category

Indicator

Note

Material

Flexible road

The proportion of the concrete mix should be appropriate

Quality certification

Road construction materials should be of an exceptional quality

Recycle materials

Surplus or waste resources should have been recycled when possible

Rigid road

Materials used for the preparing mix should be of higher quality

Local materials

Local suppliers should be used to purchase construction materials as feasible

Budget plan

To calculate and keep track of project costs, a budget plan seems required

Maintenance schedule

The implementation and completion of maintenance tasks, as well as effectiveness monitoring, must be coordinated in advance

Quality management

For the successful delivery of the project, the process and performance standards must be ensured

Work zone management

The requirements of the traffic, working team, and neighbors should all be taken into account when managing the work zone

Project team

The project team must be informed about everything about the project

Project record

It must be simple to access information on both past and present maintenance actions

Project interaction

To prevent conflict between various road projects in the identical area, communication should be carried out effectively

Construction Safety

Consideration should be given to safety issues involving the construction team and machinery

Glare control

Implementing appropriate techniques will help reduce glare from vehicle headlights and worksite illumination

Traffic control

Planning and implementing adequate and acceptable traffic control is necessary

Traffic marking

If traffic markings were disturbed, they should be put back in place

Sustainability representative

It is advised that someone with experience in sustainable infrastructures take part in the project

Management

Safety

Innovation

(continued)

386

R. R. Myakala and S. Shankar

Table 1 (continued) Category

Indicator

Note

Creative ideas

Any innovative methods that can increase the project’s capacity for sustainability must be taken into account

Certified sustainable rural road

If the current rural road has been approved by a sustainable transportation/rural road programme, bonus points can be gained

2.3 Points Distribution Determination Most of the current sustainability rating methods for rural roads assign points under each indication (indicator weighting) based on the rating tool’s creators’ or users’ subjective judgement. RRSIM would employ an objective weighting mechanism to assign points to each indicator based on statistics. First, 200 points seem to be the maximum obtained via RRSIM. As per the category priorities, the potential points under each category have been proportionally calculated, as indicated in Table 2. Adilabad, Nirmal, Hyderabad, Nizamabad, Karimnagar, Jagityal, Ranga Reddy, and Warangal district departments of transportation (DOTs) have been examined for their policies in order to determine the points assigned to each indicator. It is thought that these district DOTs either have the most acceptable sustainability practices or can significantly increase the sustainability of rural roads throughout Telangana. Below seems to be a process for calculating the points under every criterion by looking at district DOT procedures. 1. Gather the instructions for rural road maintenance construction from the district DOT Web site. 2. Examine the manuals (more than 300 have been gathered and examined) to see which chapters, sections, and manuals address RRSIM indicators. 3. Include the district DOT chapters, sections, and manuals under the relevant indication as “practices.“ There have been three stages to show the number of chapters, sections, and manuals included (Level 1 indicates a few; Level 2 indicates some; Level 3 indicates many). 4. Ascertain whether the “practices” can be used for rural road maintenance procedures. Three coefficients represent this preparedness: 3 for ready, 2 for somewhat ready, 1 for not ready, and 0 for not discussed. 5. To determine the district’s DOT’s practice score, multiply the “level” by the “coefficient” (the score is between 0 and 9). 6. Depending on the QBL, an indicator entails several sectors; the mean practice score for each indication derived from the practices of the eight districts’ DOTs has been then somewhat modified. 7. The RRSIM points for each indicator make up the modified practice score; the point distribution’s outcomes are shown in Table 2.

A Decision Support Tool for the Sustainability Rating Index …

387

Table 2 Rural road sustainability index for maintenance point distribution Category

Indicator

Category priority

QBL sectors involved

Initial distribution of points

Modified distribution of points

Mean

Environment

Forest area

29

Social

1

5

5

4.87

Atmosphere

3

8

9

7.12

Weather

2

3

6

2.52

Energy

4

11

9

9.5

Empowerment

3

3

4

2.75

Service

19

2

4

5

3.25

Public management

3

2

2

1.12

Cultural heritage

4

2

3

2.25

Value addition

1

4

4

3.72

Livestock Technique

Economy

Material

Management

1

5

1

4.87

3

8

6

6.52

Standard procedure

2

10

8

8.5

Smoothness adjustment

2

3

4

2.75

Maintenance techniques

4

0

3

0.25

Repair and disturbance

1

1

2

1

2

3

2

3.25

Transportation

3

2

5

2.12

Supply chain

2

4

4

3.87

Financial investment

2

5

4

4.75

Material production

Market supply–demand

Flexible road

23

15

1

6

5

5.62

Quality certification

3

3

2

3.25

Recycle materials

3

5

6

5.12

Rigid road

4

4

3

3.5

Local materials

3

3

4

3.25

2

4

6

4

3

3

2

2.72

Budget plan Maintenance schedule

20

30

(continued)

388

R. R. Myakala and S. Shankar

Table 2 (continued) Category

Safety

Innovation

Indicator

QBL sectors involved

Initial distribution of points

Modified distribution of points

Mean

Quality management

4

5

6

5.37

Workzone management

4

4

4

4.25

Project team

4

6

3

5.87

Project record

4

5

5

5.25

Project interaction

3

2

4

1.87

1

19

17

11.62

Glare control

1

5

7

4.82

Traffic control

3

25

23

16.75

Traffic marking

2

6

9

5.75

4

3

4

2.75

Creative ideas

3

5

3

5.52

Certified sustainable rural road

2

0

1

0.35

Construction Safety

Sustainability represent

Category priority

56

8

Table 3 Certification levels of RRSIM Anticipated scoring rate (%)

Indicators quantity 10–19

20–29

30–39

40–60

0–30

Not sustainable

*

**

**

31–40

*

**

**

***

41–50

**

***

***

****

51–60

***

***

***

****

Note: * is single star rating, ** Double Star Rating, *** Triple Star Rating, **** Four Start Rating, and so on

2.4 Certification Level Determination A score known as the road sustainability index (RSI) would be given to a rural road maintenance project once it has been assessed using the RRSIM paradigm to represent its sustainability accomplishments. The total points received by projects or initiatives under each criterion are calculated using the current sustainability rating methods for transportation or low-volume roads as the foundation for certification. Nevertheless, as each project has a history and each project’s scale may vary considerably, a

A Decision Support Tool for the Sustainability Rating Index …

389

low-volume rural road maintenance project will not meet all the criteria for RRSIM. Therefore, by considering both the indicators considered and the points received, the RSI should represent the assessment under the applicable indicators. After assessment and RSI computation, a project must be given a reasonable certification level to show the sustainability accomplishments of various rural road maintenance projects. According to the anticipated scoring rate and indicator quantity in Table 3, there have been four certification levels. A project would receive one to four PSIM stars depending on its RSI.

3 An Instance of the Assessment Process (Telangana State Rural Road) A frequently carried out low-volume road maintenance work in Telangana state— namely, Ranga Reddy District major district road (MDR)—has been chosen as a case example to demonstrate the evaluation procedure to test the RRSIM applicability on a real rural road maintenance work. The following are representations of common RRSIM assessment report contents (i.e., assessment findings): The real and reachable RSI, as well as the certification (if relevant); a description of each indicator in detail; a description of strengths (sustainable accomplishments) and weaknesses (possible improvements); and a description of the actions that, if appropriate, merit innovation points. The research project details were collected from the Roads and Buildings (R&B) Department and were designed to conduct a thorough investigation and sustainability analysis of the rural road which the R&B, Telangana State, had built in July 2019. As a test portion for this investigation, a recently built rural road measuring 2 km in length between Devarampally and Shankarpally (21.180 km to 19.175 km) on a significant district road (MDR) in Ranga Reddy District of Telangana state was selected. The RRSIM indicators used in this study are listed in Table 4 along with their accompanying justifications. The project’s RSI was 26/25.5% since it received an overall of 32 points under 26 RRSIM indications. This rating indicates that, according to RRSIM, the project for maintaining rural roads was not sustainable.

4 Conclusion For developing nations to grow, prosper, and progress economically, reliable and secure transportation infrastructure is essential. Systems of roads and highways play a crucial role in fostering and sustaining economic growth and a livable standard. Sustainability is seen to be a fantastic option for the current country roads. In addition to variable levels of commitment, scarce financial capabilities, and an absence of knowledge about sustainability ideas and how to tackle them considering the country-particular features, the issues in developing nations like India include many

390

R. R. Myakala and S. Shankar

Table 4 Rural road sustainability index for maintenance evaluation outcomes for an MDR in Ranga Reddy district Category Environment

Social

Indicator

Actual points

Maximum points

Forest area

1

4

Atmosphere

0

4

Weather

2

5

Energy

1

3

Service

2

2

Public management

0

1

Cultural heritage

0

1

Value addition

1

3

Technique

Material production

0

2

Standard procedure

1

6

Economy

Transportation

3

4

Supply chain

1

2

Material

Management

Safety

Financial investment

1

3

Flexible road

0

2

Quality certification

0

5

Rigid road

0

2

Local materials

0

4

Budget plan

0

2

Maintenance schedule

2

2

Quality management

2

5

Project record

2

5

Construction safety

4

6

Traffic control

3

5

Traffic marking

4

6

Sustainability representative

0

3

Certified sustainable rural road

0

2

Total points

30

89

Earned percentage

26.5

Innovation

other distinctive features. In order to analyse rural road maintenance sustainability, it was analyzed; in this study with an emphasis on road construction, and a description of sustainable rural road maintenance has been given. The rating tool, known as RRSIM, comprises 38 indicators, eight categories, and 200 possible points to help users assess any rural road maintenance project’s sustainability. Compared to other rating tools, RRSIM uses distinct methods to weigh each rating indication and establish a certification level. It was decided to use road maintenance work in Telangana’s Ranga Reddy District to illustrate the RRSIM assessment procedure and findings.

A Decision Support Tool for the Sustainability Rating Index …

391

The outcomes showed that the road maintenance project’s application of RRSIM has been successful. Even though the RRSIM assessment determined that the lowvolume road maintenance work chosen was not sustainable, some recommendations have been made to increase the sustainability of comparable rural road maintenance works, including periodic maintenance and posting a notice of the work’s completion date during road construction. Comparable projects will obtain a higher RSI after these recommendations are implemented in the future, increasing their sustainability. RRSIM identifies possibilities for enhancing the road maintenance project’s sustainability and offers advice on what to consider.

References 1. Joni HH, Alwan IA, Naji GA (2020) Utilizing artificial intelligence to collect pavement surface condition data. Eng Technol 38(1A):74–82 2. Tawalare A, Vasudeva Raju K (2016) Pavement performance Index for Indian rural roads. Perspect 8:447–451 3. Labi S, Faiz A, Saeed TU, Alabi BNT, Woldemariam W (2019) Connectivity, accessibility, and mobility relationships in the context of low-volume road networks. Transp Res Rec J Transp Res Board 2673(12):717–727 4. Torres-Machi C, Pellicer E, Yepes V, Chamorro A (2017) Towards a sustainable optimization of pavement maintenance programs under budgetary restrictions. J Clean Prod 148:90–102 5. Agarwal P, Singh AP (2010) Some strategies for sustainable maintenance of rural roads in India. Int J Adv Eng Technol 1(3):304–311 6. Bueno PC, Vassallo JM, Cheung K (2015) Sustainability assessment of transport infrastructure projects: a review of existing tools and methods. Transp Rev 35(5):622–649 7. Jawad D (2013) Sustainable transport rating tool via traffic impact studies. J Traffic Logist Eng 1(1):30–35 8. Triantaphyllou E, Mann SH (1995) Using the analytic hierarchy process for decision making in engineering applications: some challenges. Int J Ind Eng Appl Pract 2(1):35–44

Educational Journeys: Student Perception of School Life in Disaster Recovery Contexts Emily Gall Orillon

Abstract This article presents data collection and analysis from an interpretive phenomenological analysis (IPA) approach qualitative study conducted on student and former student perception of post-disaster or conflict-impacted education and school life set in diverse settings around the globe; post 3/11 triple disaster Japan, Hurricane Irma affected Saint Martin, post-2010 earthquake Haiti, and the resettled refugee context in France. It contextualizes the findings and recommendations for future practice for application to the current Education in Emergencies landscape. Student voices highlight the crucial role of social–emotional learning programs, student-led innovation in response to post-disaster education and resilience-building initiatives for dealing with the loss of school community, adapting to new environments, and handling hardship. They tell the story of academic gaps and the financial burden on school life after disasters or conflicts and light the way forward through their examples of leveraging resilience and tales of the long-term influence of disasteraffected educational journeys on life trajectories. The United Nations 2030 Agenda for Sustainable Development and its Sustainable Development Goals (SDGs) are presented as an essential global framework for engagement by students, teachers, and other education stakeholders as they tackle global challenges related to disaster recovery, rehabilitation, and reconstruction. Keywords Post-disaster education · Resilience · SDGs

1 Introduction In 2019, the researcher embarked on an exploration of the critical field of Education in Emergencies (EiE), a term and field encountered as a Japanese language educator working on post-3–11-2011 tsunami student collaborations in Japan. The initial goal was to understand EiE in multiple crisis-impacted contexts around the globe. The research sought to identify educational barriers and pathways for students displaced E. G. Orillon (B) Helping Hands Noramise, USA and Fukuoka International School, Fukuoka, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 H. Hazarika et al. (eds.), Geo-Sustainnovation for Resilient Society, Lecture Notes in Civil Engineering 446, https://doi.org/10.1007/978-981-99-9219-5_36

393

394

E. G. Orillon

by the Fukushima nuclear accident in Japan, Hurricane Irma in Saint Martin, the post-2010 earthquake in Haiti, and the war in Syria as a refugee in France. Although each context differs geographically, politically, culturally, and economically, all have the potential to provide global knowledge to help guide the way to securing education for students affected by disaster or conflict in both current and future EiE settings. This study views the post-crisis educational journey through the lens of Bronfenbrenner’s (1979) ecological systems theory, a framework designed for understanding child development utilized here for viewing the influence of environmental factors on a student’s school life after disaster or conflict through five domains (Fig. 1); the microsystem; the individual student and their immediate family and school setting, the mesosystem; connections between peers, teachers/schools, and other systems of direct social interaction, the exosystem; the indirect impact of local government and institutions, the macrosystem; national/international entities, political/economic factors, values and norms, and the chronosystem: the passage of time, a crucial element when looking at the perception of educational journeys that often exceed two decades. The research question is: How do students impacted by disaster or conflict make sense of their post-crisis educational journey?

Chronosystem: Changes and Experiences Over Time

Macrosystem:Discrimina.on, Ideologies/Inclusion/Economics

Exosystem: Gov. Agencies, Boards of Ed./Healthcare

Mesosystem: Interac.ons between Family/Peers/ Teachers Microsysytem: Family, New School Peers, Neighborhood, Housing

Student in Post-Crisis Context

Fig. 1 Bronfenbrenner’s ecological systems model (1979) adapted for post-crisis education

Educational Journeys: Student Perception of School Life in Disaster …

395

2 A Glance at Existing Literature 2.1 Disaster- or Conflict-Impacted Learning Disaster- or conflict-affected students can spend years and even decades living in displacement and face barriers to education throughout their entire educational journey. They are at risk for discrimination in host schools due to religious, cultural, ethnic and racial differences, and outsider status/stigma [8, 10]. Burde et al. (2017) emphasize that despite obstacles, teachers in the field of EiE believe that schools keep students safe and provide meaningful activity in times of upheaval and, in some cases, even save lives. However, Dryden-Peterson (2016) argues that students learning in displacement come to the classroom with an educational past often not validated in their new schools. This body of research shows the critical need for education initiatives at all stages of a disaster- or conflict-impacted educational journey. Impact of Trauma. Forced out of school environments by disaster or conflict, students face a myriad of obstacles. They deal with disrupted lives and divided families and often make their journeys alone without adult support [15]. Disaster-impacted youth may face trauma-related social and emotional challenges in their new learning environments [10]. Sullivan and Simonson (2016) assert that addressing disasterrelated student trauma requires professional skills and training and that teachers with disaster-/conflict-affected students need professional development and training in recognizing signs of trauma, experience with the mental health referral process, and skills to provide a safe and inclusive environment for students. Stigmatization and Economic Stress. Research by Oda (2016), Tsujiuchi et al. (2016) and Vue (2021) looks at stigma and bias and categorizes the trickle-down of societal discrimination as an indirect effect of disasters and conflicts. Oda’s (2016) work revolves around what he refers to as place name stigma, a phenomenon that connects disaster survivors to bias through association with disaster sites. Tsujiuchi et al. (2016) found that Fukushima IDP students expressed emotional stress related to their evacuation experience, fear of discrimination, and radiation stigma in their new school communities. Vue (2021) examines the role silence plays in young Hmong American refugee life and asserts that there is a silence brought on by an intergenerational reluctance to discuss past traumatic events, a sense of being silenced by emotions such as fear and shame, as well as a social silence enforced by the dominant culture on the refugee’s heritage voice of identity. This body of research introduces a burden shift from people in power to disaster and conflict survivors and considers the role of the media in spreading negative misinformation.

2.2 Policy and Programs The United Nations 2030 Agenda for Sustainable Development and its SDGs [17] are set in place to achieve universal transformative and human rights-based sustainable

396

E. G. Orillon

development. The education-focused target, SDG4, is an international and nationstate response to the worldwide need for equitable education for all [14]. SDG4 differs from earlier UN education frameworks in that it aims for quality education rather than mere access and promotes worldwide collaboration to achieve global equitable learning opportunity for all [15]. Classroom Programs and Practices. A growing body of literature that focuses on the core competencies identified by the Collaborative for Academic, Social, and Emotional Learning (CASEL) [3], self-awareness, self-management, social awareness, relationship management, and responsible decision-making, suggests that schools play a critical role in guiding a sense of belonging and emotional recovery for students in protracted crisis [4, 9, 12]. Trauma, discrimination, and stigma emerge in the literature as categories of exploration for this research project as well as the need for schools as safe academic and social–emotional spaces for disaster- or conflict-affected students.

3 Methodology This study utilizes interpretive phenomenological analysis (IPA) qualitative methods to explore post-crisis educational journeys directly from the source of those touched by the post-crisis experience during their school-age years. IPA is centered around the detailed exploration of individual human experience and gives voice to unheard stories one case at a time before moving to the general [11]. Before launching this study, approval was gained through Northeastern University’s Institutional Review Board (IRB), including steps for human research protection and international guidelines. The young adult participants of this study were asked to look back over their school-age years and recount their educational experiences from when disaster or conflict hit to the present. The participants for this study (Table 1) were young adult students or former students affected by disaster or conflict during their school-age years.

4 Findings and Analysis The following section presents transcript analysis based on three superordinate and seven subthemes (Table 2).

4.1 Theme One: The Need for Emotional Support Loss of Former School Community. Multiple participants shared the abrupt loss of their hometown and school community as a significant factor that weighed heavily

Educational Journeys: Student Perception of School Life in Disaster …

397

Table 1 Participant profiles Mia

Mia grew up near the Fukushima Daichi Nuclear Power Plant in Fukushima prefecture. On March 11, 2011, she was at her high school when a magnitude 9.1 earthquake hit off the coast of the Tohouku region of Japan. The Fukushima nuclear power plant explosion on the following day led to her long-term internal displacement in another prefecture

Beatrice Beatrice is a proud Port-au-Princienne whose life drastically changed when a devastating earthquake struck Haiti in 2010 during her final year of high school, and she was forced to move to another department of Haiti. Due to her mother’s loss of income, she could no longer afford her private school fees and had to attend an unfamiliar school far away from her home Peterson Peterson grew up in Port-au-Prince, Haiti facing financial hardships that made it difficult for his parents to pay his school fees. Despite tough times, he cherished going to school and playing soccer with his friends. The January 12, 2010 earthquake disrupted his already financially vulnerable middle school life, making his return to school nearly impossible Kahled

Kahled grew up outside of Aleppo, Syria. He enrolled in a five-year civil engineering undergraduate program with plans to pursue a master’s degree. The onset of the 2011 Syrian civil war disrupted Kahled’s education, and his university was under attack. Despite these challenges, Kahled completed his undergraduate degree in 2013 and was accepted into a master’s program. Due to the danger, he was forced to flee his country

Remy

Remy is from Sandy Ground, a Caribbean immigrant community on the French side of Saint Martin. When Hurricane Irma struck in 2017, Remy was struggling to find his way in high school. In the aftermath of the disaster, he faced even greater obstacles, as he tried to navigate a damaged school system with teachers who themselves were dealing with crisis

Table 2 Superordinate themes and sub-themes Need for emotional support

Obstacles in post-crisis school life

The way forward

Dealing with loss of former school community

Perception of academic gaps and barriers

Leveraging resilience

Adapting to new school environment

Post-crisis financial impact on school life

Long-term influence on life trajectories

Handling Hardship

along their post-crisis journey. In many cases, the swift and urgent exit from school during crisis was an ambiguous farewell that disrupted strong friendships and other pillars of identity, such as the stability of familiar teachers. Kahled described the option to return home as “a closed door”, a burden that accompanied him along his journey away from Syria, through Turkey, Greece, and finally into France, and became inextricably emphasized during his time living in a refugee camp. When Kahled added, “I used to be a very good student…,” we are reminded that this loss goes beyond brick-and-mortar structures to include a student’s mindset along

398

E. G. Orillon

their learning trajectory. Mia shared Kahled’s status of protracted displacement compounded with the indefinite inability to return. A high school student on the cusp of entering her final year, Mia left her school late on the night that disaster hit her region of Japan and never returned. Throughout her interview, she repeatedly spoke of her unexpected loss of her hometown and themes of isolation. Similarly, Beatrice was in her pivotal final year of high school when the earthquake struck Haiti. She shared, “I lost the ability to go to school and prepare for the national final exams.” Data from these participants points to a profound disruption brought on by the loss of school community. Adapting to New School Environment. Navigating new school environments is a challenge for crisis-affected students worldwide, and for Kahled, a young adult refugee in France, this included a new language, culture, and education system. He recalled the support he received from a compassionate professor, “He went out of his way to provide language support,” emphasizing that even at the higher education level, students benefit from deliberate support for their specific educational needs. After Hurricane Irma disrupted his education, Remy perceived his exchanges with teachers as particularly negative and remembered being told that he would not make it. Remy’s account highlights the need for teacher-preparedness and well-being initiatives after a disaster in school settings where teachers also deal with hardship. In Mia’s case, it was not until she encountered a teacher at her new school who asked her to talk about what she experienced in Fukushima that she began to find a way to cope with her feelings of loss. Beatrice also found comfort in the support of a teacher who helped her cope with the loss of her home and school community in Port-auPrince. Beatrice elaborated, “This teacher listened to me…”. In disaster-impacted settings, students like Beatrice often need to relocate to a new school soon after disaster. Her account of having access to trained teachers and professionals indicates a significant factor in her coping process. Handling Hardship. When asked to share their school life memories soon after crisis, many participants spoke of the need to emotionally deal with crisis-related separation, loss of family, community fatalities, and repeated displacement. Peterson in Haiti, Remy in Saint Martin, and Kahled along his journey from Syria to France all endured prolonged disruption to their education. All three participants spoke about this post-crisis period away from school as a time of uncertainty and frustration. Peterson remembered coping with the devastating loss of his father in isolation. When asked about his life during this painful period, Peterson hesitated, switched from French to Haitian Creole, and then shared his memories of his post-earthquake, “I wondered if things would have been different if my father was here?” His delay in returning to school resulted in a lost opportunity for emotional support disseminated by school psychologists placed in many schools after the earthquake in Haiti.

Educational Journeys: Student Perception of School Life in Disaster …

399

4.2 Theme Two: Obstacles in Post-Crisis School Life Academic Gaps and Barriers. All participants shared a perception of academic gaps and barriers in their post-crisis-impacted academic life. Rigid education systems or requirements often perpetuated this. Mia shared, “I had to do all the regular transfer requirements, …”. She mentioned that each prefecture handled disaster evacuees differently, with some waiving the rigorous new school application process. Still, in Mia’s case, she was subjected to stressful interviews and tests at a time when she was coping with loss and adjustment. In her new school in Haiti, Beatrice also faced academic hurdles caused by the pressure of a rigid systemic process, that of the challenging national exam. This culminating Haitian exam requires a great deal of dedication in the best of times. Beatrice entered this stressful period in an unfamiliar environment, internally displaced after the earthquake. For Beatrice, education was central in her life. Fortunately, her new school offered academic support to help her prepare for this important exam. As a refugee student crossing several borders during his seven-year post-crisis educational journey from Syria, through Turkey and Greece, and finally into France, Kahled faced a multitude of academic hurdles linked to language barriers, credential incompatibility across borders, and documentation requirements. Post-crisis Financial Impact. All participants shared an emotional toll of financial loss as a factor in their post-crisis education. Mia explained that when her family left Fukushima, they packed only what they could fit in their car, unaware they would start over as internally displaced long-term evacuees and never return to their hometown. Beatrice faced the loss of her mother’s business in Port-au-Prince and the resulting inability to pay her private school tuition. When asked to talk about his memories of things that he found helpful during the post-crisis period, Peterson shared that financial support was his most significant factor. Peterson recalled, “I could finally go back to school.”

4.3 Theme Three: The Way Forward Leveraging Resilience. Participants experienced adverse conditions due to disaster or conflict and spoke about finding ways to adapt. They displayed life skills related to resilience, a trait developed in response to the burden of crisis and utilized as a tool to manage adversity. Beatrice shared that she loved school before the 2010 earthquake struck Haiti and still loved it after the disaster. Finding her academic grounding required self-determination. “I said to myself…all life is not over…” She shared a perception of a need to persevere through hardship in order to achieve positive outcomes in her post-disaster school life and broader future. “It was the first time that I could see that I could survive a natural disaster and all the aftermath,” she elaborated. Beatrice utilized this attitude of resilience to reset herself on a course from surviving to thriving. Kahled’s post-crisis educational journey extended over multiple

400

E. G. Orillon

years, calling on him to overcome one hardship after another. When he found out he would be sent to France, he was frustrated due to his existing hard-earned English language skills. Once in France though, Kahled got to work on fulfilling his plan to become a civil engineer. Kahled explained, “I got accepted into a program and said okay, let’s do it!” Kahled’s optimism led to his ability to adapt to the unexpected and continue on his course to complete his studies. Mia’s displacement after evacuating from Fukushima lasted longer than expected due to lingering radiation levels in her hometown. She showed personal motivation and strength as she found support in her new school environment and endurance to move forward. Long-term Influence on Life Trajectories. All participants spoke of the ways their post-disaster experience influenced their current situations. As participants continued forward along their long-term paths, their experiences, coping strategies, and resiliency impacted their lives. As a staff member in a children’s welfare center in an area affected by the 2011 disaster in Japan, Mia understands the importance of providing an opportunity to “be heard” and is working on a plan to collaborate with an NGO to form a storytelling project for disaster-affected youth. Beatrice’s account of her feelings of support from a teacher and school psychologist who looked out for her emotional well-being were followed by statements of strength and positivity directly connected to the memory of this teacher. She stated, “It was thanks to this teacher that I went into psychology.” Beatrice now works at a Port-au-Prince school, supporting students with their academic and emotional struggles.

5 Discussion and Recommendations for Practice This study’s findings align with existing EiE literature and point to the critical need for school-based interventions for students who have experienced disaster or conflict. Viewing student experience across all of Bronfenbrenner’s (1979) ecological systems emphasizes the various roles of students, educators, education stakeholders, and civil societies worldwide and points to a close link to the United Nations 2030 Agenda for Sustainable Development and its SDGs (Fig. 2). The SDGs that address poverty, hunger, health, gender, clean water, employment, sustainable communities, equality, and peace all converge on education with students and teachers situated at the center of the efforts to achieve these global goals. Organizations and schools securing quality and equitable education (SDG4) for students in post-crisis contexts can use this study’s findings. Participant accounts of educational journeys brought to light their needs, obstacles, and the invaluable hidden gems gleaned from the ongoing link between disaster impact and education. The scope of this research has implications and relevance for education stakeholders at local, national, and international levels. What follows are recommendations for practice aligned with the SDGs.

Educational Journeys: Student Perception of School Life in Disaster …

401

Fig. 2 17 goals for sustainable development [17]

5.1 Safe and Inclusive Spaces for Storytelling: SDG Alignment: 3–5, 9, 10, 16 Create small group opportunities for safe interactions for students to share stories of their experiences and emotions. Schools and universities can designate time for small group interactions where students take part in trust and relationship-building activities. These small group gatherings can incorporate opportunities for storytelling, a strategy revealed in this study as having a positive impact in crisis contexts. Digital technologies can be integrated into this initiative to create a variety of opportunities for students to share their experiences. Some students may prefer an alternative mode of expression through art, music, or physical movement.

5.2 Wellbeing Leadership Team: SDG Alignment 2 -5, 8 -10, 16, 17 Form a Wellbeing Leadership Team (WLT) comprised of teachers and students across all grade levels to serve as a school’s think-tank on student and teacher SEL and well-being in post-disaster settings or schools with disaster- or conflict-impacted student representation. SEL curriculum, safe space development, and projects to foster emotional well-being can be created or integrated into existing initiatives and activities. The WLT can organize individual student–teacher partnerships for frequent one-on-one well-being check-ins for vulnerable students. Lastly, with the knowledge that teachers can also be impacted by crisis and overwhelmed by student needs, a school’s WLT can promote teacher well-being through a teacher-to-teacher buddy system.

5.3 School Garden: SDG Alignment 2 -5, 8 -13, 15–17 A school garden project in a post-disaster context can be developed and utilized as a multi-faceted initiative to promote student agency in sustainable development,

402

E. G. Orillon

nutrition and food sovereignty, health and well-being, cultural preservation, climate resilience, peer relationships, youth leadership, and economic empowerment. [6].

5.4 Post-Crisis Education Teacher Training: SDG Alignment 3–5, 8, 10, 16 Facilitate post-crisis education teacher training that includes strategies for identifying degrees of trauma and indicators to recognize when a student needs more support than what can be offered at school. The WLT can contextualize existing resources, such as those provided through the Inter-Agency Network for Education in Emergencies (INEE) [7].

5.5 Community Coalition: SDG Alignment 1–6, 8, 10 -13, 16, 17 Form a school-based Community Coalition similar to the WLT but with a broader scope of focus that extends to academic needs, family resources related to food, water and health, and family support systems. The coalition can be comprised of teachers, parents, student, and school leadership and other community stakeholders. The coalition can use student and family voice to identify areas of need specific to disaster- or conflict-affected students, find solutions, and streamline information to the proper channels for implementation.

5.6 Resilience Curriculum Framed by the SDGs: SDG Alignment 1–17 Frontload students’ education by incorporating crisis resilience development into the curriculum, promoting essential recovery skills before, during and after disasters. Strengthen the capacity for building resilience by teaching coping strategies, fostering a positive mindset, and establishing partnerships with community role models. Empower students through a resilience curriculum centered on the SDGs. This enables students to understand their role in achieving these goals within their community, country, and globally. Framing resilience curriculum on the SDGs can help students develop a sense of agency as they collaborate to initiate student-led grassroots projects and research that address the global goals.

Educational Journeys: Student Perception of School Life in Disaster …

403

6 Conclusion This study sought to understand how students from crisis-affected communities perceive their post-crisis school life in Japan, Saint Martin, Haiti, and the refugee context in France. The nexus between post-crisis education and student engagement with The UN 2030 Agenda for Sustainable Development and the SDGs emerged during this investigation as a solid foundation for building disaster recovery, rehabilitation, and reconstruction. The participants’ narratives reflected an urgent need to persevere in their crisis-affected school lives. They found ways to turn all they had endured into hope and optimism. Stories shared in this study were embedded with self-driven solutions and situated the students themselves as their own resilient heroes.

References 1. Bronfenbrenner U (1979) The ecology of human development: experiments by nature and design. Harvard University Press, Cambridge 2. Burde D, Kapit A, Wahl R, Guven O, Skarpeteig M (2017) Education in emergencies: a review of theory and research. Rev Educ Res 87(3):619–658 3. CASEL Homepage. https://casel.org/guide/. Last Accessed 23 July 2023 4. Cho H, Wang X, Christ T (2019) Social-emotional learning of refugee English language learners in early elementary grades: teachers’ perspectives. J Res Child Educ 33(1):40–55 5. Dryden-Peterson S (2016) Refugee education in countries of first asylum: breaking open the black box of pre-resettlement experiences. Theory Res Educ 14(2):131–148 6. HHN Homepage. http://noramise.org. Last Accessed 24 July 2023 7. INEE Homepage. https://inee.org. Last Accessed 29 July 2023 8. Oda T (2016) Place stigmatization through geographic miscommunication: unravelling the Fukushima Disaster 38 9. Powell T, Bui T (2016) Supporting social and emotional skills after a disaster: findings from a mixed methods study. Sch Ment Heal 8(1):106–119 10. Ruf B (2013) Educating traumatized children. Lindisfarne Books, M. Saar 11. Smith J, Flowers P, Larkin M (2012) Interpretive phenomenological analysis: theory, method and research. Sage Publishing, Thousand Oaks 12. Sullivan A, Simonson G (2016) A systematic review of school-based social-emotional interventions for refugee and war-traumatized youth. Rev Educ Res 86(2):503–530 13. Tsujiuchi T, Yamaguchi M, Masuda K, Tsuchida M, Inomata T, Kumano H, Mollica R (2016) High prevalence of posttraumatic stress symptoms in relation to social factors in affected population one year after the Fukushima nuclear disaster. PLOS ONE 11(3) 14. United Nations SDG4 Homepage. http://un.org/sustainabledevelopment/education/. Last Accessed 23 July 2023 15. UNHCR Homepage. https://www.unhcr.org. Last Accessed 23 July 2023 16. Vue R (2021) Trauma and resilience in the lives and education of Hmong American students: forging pedagogies of remembrance with critical refugee discourse. Race Ethn Educ 24(2):282– 301 17. United Nations Department of Economic and Social Affairs, Sustainable Development Homepage. https://sdgs.un.org/goals. Last Accessed 23 July 2023