Advances in Hydroinformatics: SimHydro 2019 - Models for Extreme Situations and Crisis Management [1st ed.] 9789811554353, 9789811554360

Citation preview

Springer Water

Philippe Gourbesville Guy Caignaert   Editors

Advances in Hydroinformatics SimHydro 2019 - Models for Extreme Situations and Crisis Management

Springer Water Series Editor Andrey Kostianoy, Russian Academy of Sciences, P. P. Shirshov Institute of Oceanology, Moscow, Russia

The book series Springer Water comprises a broad portfolio of multi- and interdisciplinary scientific books, aiming at researchers, students, and everyone interested in water-related science. The series includes peer-reviewed monographs, edited volumes, textbooks, and conference proceedings. Its volumes combine all kinds of water-related research areas, such as: the movement, distribution and quality of freshwater; water resources; the quality and pollution of water and its influence on health; the water industry including drinking water, wastewater, and desalination services and technologies; water history; as well as water management and the governmental, political, developmental, and ethical aspects of water.

More information about this series at http://www.springer.com/series/13419

Philippe Gourbesville Guy Caignaert •

Editors

Advances in Hydroinformatics SimHydro 2019 - Models for Extreme Situations and Crisis Management

123

Editors Philippe Gourbesville Université Côte d’Azur Nice, France

Guy Caignaert Hydrotechnique Society of France Paris, France

ISSN 2364-6934 ISSN 2364-8198 (electronic) Springer Water ISBN 978-981-15-5435-3 ISBN 978-981-15-5436-0 (eBook) https://doi.org/10.1007/978-981-15-5436-0 © Springer Nature Singapore Pte Ltd. 2020 This work is subject to copyright. All rights are reserved 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, express 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

Preface

With the current digital development in modern societies, hydroinformatics defined as management of information related to the water sector using ICT tools is becoming a large domain of engineering technology and sciences. Modelling and simulation are historically the points of departure for hydroinformatics and are one of the most important parts of it. Neither the SimHydro cycle of international conferences since 2010 nor the present book has the purpose or ambition to cover thematically the whole extent of the subjects. The main purpose is to concentrate on a limited number of specific areas and subjects that are not usually considered as such during most global international conferences or publications. Modelling in fluid mechanics, hydraulics and hydrology, whether using digital tools or scale models, has reached sufficient maturity to be in daily use by engineers for analysis, design and for communication. Increasingly, complex cases can be handled thanks to evermore sophisticated tools and increasingly abundant computing power and data resources. The emerging environment populated with the new generation of sensors, using cloud computing resources, producing big data, is challenging the current practices of modelling and requests innovation in methodology and concepts for real integration into the decision-making processes that are more and more requested for crisis management. At the same time, the request to integrate vulnerability and resilience dimension in various engineering approaches is becoming more and more frequent especially for environments directly exposed to major natural hazards like floods and inundations. With respect to these issues, however, a number of questions still remain open: coupling of models, data acquisition and management, uncertainties (both epistemic and random) of results supplied by models, use of 3D CFD models for complex phenomena and for large-scale problems. All these points are continuously explored and investigated by researchers, scientists and engineers. Like in all scientific domains, most recent and advanced developments have to be discussed and shared regularly in a growing community that has to face every day more challenging and complex situations. The SimHydro 2019 conference, following the four previous editions, has contributed to this objective by providing a platform for exchanges and discussions for the different actors in the water domain. v

vi

Preface

SimHydro is a permanent cycle of conferences held every 2 years, hosted by Polytech Nice Sophia and organised by the Société Hydrotechnique de France (SHF) and its partners. It aims, as the subject, at recent advances in modelling and hydroinformatics and at the participation and exchanges at European scale (it is open to all other researchers and participants but the purpose is to maintain a specific platform for the region that was a birthplace of both domains). The latest SimHydro conference was held in Sophia Antipolis, France, from 12 to 14 June 2019. The conference was jointly organised by the Société Hydrotechnique de France (SHF), the Association Française de Mécanique (AFM), the University of Nice Sophia Antipolis/Polytech Nice Sophia and with the support of the International Association for Hydro-Environment Engineering and Research (IAHR), the Environmental and Water Resources Institute (EWRI) of the American Society of Civil Engineers (ASCE) and the Canadian Society for Civil Engineering (CSCE). Several sponsors also supported the conference: EDF, CNR, ARTELIA, SETEC-HYDRATEC and ACRI Group. The conference attracted 166 delegates from 41 countries who participated in 24 sessions where 136 papers were presented. The programme was organised around twelve main themes: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

Hydro-environmental issues and extreme situations Models for extreme situations Uncertainties and data assimilation Extreme in hydraulics: how to deal with? Crisis management and models Decision support systems and models: concepts, design, challenges, implementation and operation Real-time management and models Hydraulic structures and networks: real-time operation and crisis Scale models in hydraulics and their place and complementarity in simulation concepts Modelling methods and tools for floods management 3D multiphase flows (experiments and modelling) Hydraulic machineries

Within these general themes, topics like coupling of models, data assimilation and uncertainties, urban flooding, data and uncertainties in hydraulic modelling, model efficiency and real situations, new methods for numerical models, hydraulic machinery, 3D flows in the near field of structure and models for complex phenomena have been covered. The conference, by attracting researchers, engineers and decision-makers, has promoted and facilitated the dialogue between various communities especially with a special session dedicated to catastrophe models. The purpose of catastrophe modelling is to help communities and companies anticipate the likelihood and severity of potential future catastrophes before they occur so that they can adequately prepare for their financial impact. Insurances and reinsurance companies at the worldwide scale currently develop these approaches. Catastrophe modelling combines the four components—hazard, inventory, vulnerability and

Preface

vii

loss—to aid insurers in making their decisions on what type of protection they can offer against a particular risk. Integration of hydroinformatics methods and tools in these approaches is a real challenge. Representatives from insurance and reinsurance companies have presented their approaches of extreme events and their operational implementation through international examples. Exchanges with participants have been very fruitful on crucial questions related to the crisis management during extreme flood events, the needs for operational forecasting systems, the state of the art in research and development in the domain of numerical fluid mechanics, the stakeholder’s capacity to understand results, the means for dialogue directly or indirectly between the stakeholders and the model developers and the information’s exchange between stakeholders and developers. In order to contribute to this dialogue and to provide useful references, following the successful experiences of 2012, 2014 and 2017, the organisers of SimHydro 2019 have decided to elaborate this book. This volume gathers a selection of the most significant contributions received and presented during the conference. The objective is to provide the reader with an overview of the on-going developments and the state of the art taking place in four major themes that are as follows: • • • •

Decision support systems and crisis management, Flood forecasting, Methods and models for hydrology and climate change, High performance computing and complex hydraulics applications.

Obviously, all dimensions of these themes cannot be covered in a single book. However, the editors are convinced that the contents may contribute to provide to the reader essential references for understanding the actual challenges and developments in these areas of the hydroinformatics field. This volume represents the sum of the efforts invested by the authors, members of the scientific committee and members of the organising committee. The editors are also grateful for the dedicated assistance of the reviewers who worked tirelessly behind the scene to ensure the quality of the papers. We hope this book will serve as a reference source on hydroinformatics for researchers, scientists, engineers and managers alike. Nice, France Paris, France August 2019

Philippe Gourbesville Guy Caignaert

Contents

Part I 1

2

3

4

5

6

Decision Support Systems and Crisis Management

Which Models for Decision Support Systems? Proposal for a Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Philippe Gourbesville

3

Aquavar: Decision Support System for Surface and Groundwater Management at the Catchment Scale . . . . . . . . . . . Qiang Ma, Philippe Gourbesville, and Marc Gaetano

19

Anywhere: Enhancing Emergency Management and Response to Extreme Weather and Climate Events . . . . . . . . . . . . . . . . . . . Morgan Abily, Philippe Gourbesville, Eurico De Carvalho Filho, Xavier Llort, Nicolas Rebora, Alexandre Sanchez, and Daniel Sempere-Torres Use of Anywhere Products to Assess Risky Events on Southern France and Corsica (October 2018) . . . . . . . . . . . . . . . . . . . . . . . . Eurico de Carvalho Filho, Guillaume Lahache, and Alix Roumagnac Decision-Making Support System for Crisis Operations and Logistics Aspects in Extreme Weather-Induced Events . . . . . Ivan Tesfai, Giovanni Napoli, Salvatore Ferraro, Andrea Poggioli, and Marta Speranza Operational Resilience Index Computation Tool as a Decision Support System Integrated in Eu Risks Management Platforms—Test on Biguglia Catchment, a Mediterranean Intense Precipitations Regime Prone Area . . . . . . . . . . . . . . . . . . Morgan Abily, Philippe Gourbesville, Hézouwé Amaou Tallé, Marc Gaetano, Jelena Batica, Patric Botey, and Marien Setti

29

39

53

65

ix

x

7

Contents

Realtime High Resolution Flood Hazard Mapping in Small Catchments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flavio Pignone, Lorenzo Campo, Daniele Dolia, Rocco Masi, Giacomo Fagugli, Daniele Ferrari, Simone Gabellani, Francesco Silvestro, Nicola Rebora, and Francesca Giannoni

8

Visualization of Flood Simulation with Microsoft HoloLens . . . . . Shanyu Wang, Jianrong Wang, Philippe Gourbesville, and Ludovic Andres

9

Anycare: A Serious Game to Evaluate the Potential of Impact-Based and Crowdsourced Information on Crisis Decision-Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Galateia Terti, Isabelle Ruin, Milan Kalas, Arnau Cangròs i Alonso, Tommaso Sabbatini, Ilona Lang, and Balazs Reho

10 From Catstrophe to Resilience . . . . . . . . . . . . . . . . . . . . . . . . . . . Jelena Batica and Philippe Gourbesville

79

91

103

121

11 Marine Dispersion Modelling and Expertise Tools for Accidental Radiological Contamination of French Coasts . . . . . . . Céline Duffa

135

12 A Study on Water Crisis Management Techniques by Fallout in Case of Radiation Accident Using Environmental Multimedia and Air Transport Diffusion Model . . . . . . . . . . . . . . Daemin Oh, Youngsug Kim, Sungwon Kang, Soungjong Yoo, Noriyuki Suzuki, and Yoshitaka Imaizumi

145

13 Challenges in Defining Alarm Thresholds to Improve Crisis Management Procedures: A Case Study on the French Riviera . . . Stan Nomis, Leslie Salvan, Raphaëlle Dreyfus, Franck Compagnon, and Pierre Brigode

159

14 Integrations of an Early Warning System and Business Continuity Plan for Disaster Management in a Science Park . . . . Tsun-Hua Yang, Hao-Ming Hsu, and Hong-Ming Kao

171

15 Natural Hazard Crisis Management Exercice at Metropolitis Scale: Methodolgy for Holistic Involvement of Municipalities . . . . Yannick Dorgigne, Morgan Abily, and Philippe Gourbesville

181

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Charlotte Dreger, Erik Bollaert, Olivier Stauffer, and Yves Châtelain

199

17 Development of the Similar Typhoon Search System Based on the Deep Neural Network Using Deep Learning . . . . . . . . . . . . Kohji Tanaka, Eisaku Yura, Tatsuya Yoshida, and Shigeho Maeda

213

Contents

xi

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dongeon Kim, Shie-Yui Liong, Philippe Gourbesville, and Jiandong Liu

229

19 An Integrated Approach to Water Resources and Investment Planning for Water Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damian Staszek, Dragan Savic, and Guangtao Fu

241

20 Model Improvement for Effect Evaluation of Low Impact Development Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yuting Meng, Na Li, Jing Wang, Qian Yu, and Nianqiang Zhang

257

21 Multiple Feedback Linkage in the Process of Urban River Water Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lunyan Wang, Shoukai Chen, Xiangtian Nie, and Shichao Liu

271

Part II

Flood Forecasting (21)

22 Real-Time Flood Management and Preparedness: Lessons from Floods Across the Western Japan in 2018 . . . . . . . . Daisuke Nohara, Yasuhiro Takemon, and Tetsuya Sumi

287

23 Flood Forecasting in Alpine Regions Using a Multi-model Approach: Operational Performance and Experiences After Two Years of Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . Frédéric G. Jordan, Raphael Mutzner, Alexandre Prina, and Christophe Guay

305

24 Application of an Ensemble Kalman Filter to A Semi-distributed Hydrological Flood Forecasting System in Alpine Catchments . . . Alain Foehn, Anne Schwob, Damiano Pasetto, Javier García Hernández, and Giovanni De Cesare 25 Real-Time Inundation Mapping with a 2D Hydraulic Modelling Tool Based on Adaptive Grid Refinement: The Case of the October 2015 French Riviera Flood . . . . . . . . . . . . . . . . . . Geoffroy Kirstetter, François Bourgin, Pierre Brigode, and Olivier Delestre 26 Early Warning System for Flood Warning in Campings . . . . . . . . Gonzalo Olivares, Manuel Gómez, Joan Gurrera, and Marcos Sanz-Ramos 27 An Automated Anomaly Detection Procedure for Hourly Observed Precipitation in Near-Real Time Application . . . . . . . . . Sheng-Chi Yang, Ming-Chang Wu, Hong-Ming Kao, and Tsun-Hua Yang

319

335

347

359

xii

Contents

28 Storm Water Management Model Parameter Optimization in Urban Watershed Using Sewer Level Data . . . . . . . . . . . . . . . . Oseong Lim, Young Hwan Choi, and Joong Hoon Kim 29 From Meteorological Forecasting to Floodplain Forecasting for the Protection of Populations in Urban and Peri-Urban Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sandrine Vidal and Jean Paul Ducatez

367

377

30 Feedbacks on the Deployment of and Experimental Real-Time Flood Forecasting and Crisis Management System . . . . . . . . . . . . Arnaud Koch and Armonie Cossalter

383

31 Application of Recurrent Neural Network for Inflow Prediction into Multi-purpose Dam Basin . . . . . . . . . . . . . . . . . . . . . . . . . . . Juhwan Kim, Myungky Park, Yungsuk Yoon, and Hyunho Lee

397

32 Pre-release Strategy for Flood Control in the Multi-reservoir and Rivers System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thanh Hao Nguyen, Philippe Gourbesville, Ngoc Duong Vo, and Nguyen Duc Phuoc Vo 33 Optimization of Spillway Operation for Flood Mitigation in Multi-reservoirs River System . . . . . . . . . . . . . . . . . . . . . . . . . . Thanh Hao Nguyen, Philippe Gourbesville, Ngoc Duong Vo, and Nguyen Duc Phuoc Vo 34 Flood Forecast Tool to Help Dam Management from France to Thailand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sandrine Vidal, Bertrand Richaud, Finn Hansen, and Jimmy Courtigne 35 Early-Warning System for Cyclone-Induced Wave Overtopping Aided by a Suite of Random Forest Approaches . . . . . . . . . . . . . . Jeremy Rohmer, Sophie Lecacheux, Rodrigo Pedreros, Deborah Idier, and François Bonnardot 36 Water Level Short-Term Forecasting Using Statistical Approaches: A Case Study on the Parisian Region . . . . . . . . . . . . Nicolas Cheifetz, Hugo Senetaire, Cédric Féliers, and Véronique Heim 37 Uncertainty Propagation in TELEMAC 2D Dam Failures Modelling and Downstream Hazard Potential Assessment . . . . . . Layla Assila, Matthieu Sécher, Thomas Viard, Benoît Blancher, and Cédric Goeury 38 Defining Uncertainty for a Simplified Method Dedicated to the Mapping of Extreme Floods . . . . . . . . . . . . . . . . . . . . . . . . André Paquier, Quentin Royer, Christine Poulard, and Pascal Billy

409

417

427

437

453

465

481

Contents

xiii

39 Experimental and Numerical Modelling of the Influence of Street-Block Flow Exchanges During Urban Floods . . . . . . . . . Miguel Angel Mejía-Morales, Sébastien Proust, Emmanuel Mignot, and André Paquier 40 Width Parameter Analysis on Runoff Model for Storm Water Drainage System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiancheng Tao 41 Potential Application of LID Techniques to Reduce Urban Flooding in Different Rainfall Pattern, Case Study for Quy Nhon, Vietnam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nguyen Duc Phuoc Vo, Thi Thu Tram Huynh, and Trung Dung Vo Part III

495

507

517

Methods and Models for Hydrology and Climate Change (20)

42 Possible Roles of Artificial Neural Networks in Hydraulic and Hydrological Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shie-Yui Liong, Dongeon Kim, Jiandong Liu, Philippe Gourbesville, and Ludovic Andres 43 Multi-model Approach for Reducing Uncertainties in Rainfall-Runoff Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cynthia Andraos and Wajdi Najem 44 Rainfall-Runoff Modeling to Investigate Flash Floods and Mitigation Measures in the Wadi Bili Catchment, Egypt . . . . Franziska Tügel, Abdelrahman Ali Ahmed Abdelrahman, Ilhan Özgen-Xian, Ahmed Hadidi, and Reinhard Hinkelmann 45 Flood Risk Assessment in the Tra Bong River Catchment, Vietnam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Manh Trinh Xuan and Frank Molkenthin 46 Innovative Solutions for Climate-Resilient Flood Management in the Poorer and Vulnerable Province of Leyte, The Philippines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jihyeon Park and Ilpyo Hong

529

545

559

575

593

47 Modelling Strategy of Deterministic Distributed Hydrological Model Development at Catchment Scale . . . . . . . . . . . . . . . . . . . . Qiang Ma and Philippe Gourbesville

607

48 Application of Satellite Remote Sensing Technology in River Monitoring and Governance . . . . . . . . . . . . . . . . . . . . . . Lunyan Wang, Shoukai Chen, Xiangtian Nie, and Shichao Liu

619

xiv

Contents

49 The Uncertainty in Spatial Rainfall Distribution––A Case Study for Binh Dinh Province, Vietnam . . . . . . . . . . . . . . . . . . . . . . . . . Ngoc Duong Vo, Ma Qiang, Vinh Khanh Ngo, and Philippe Gourbesville 50 Assessing Future Water Availability Under a Changing Climate in Kabul Basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Masoud Ghulami, Philippe Gourbesville, and Philippe Audra 51 A Modeling Approach for Critical Source Areas Identification and Sources Apportionment for Nitrogen Load in Yuan River Catchment, China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Qianwen He and Frank Molkenthin 52 Coupling Methods for Urban Areas Large Scale Hydraulic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Camille Duran, Thierry Lepelletier, Simon Olive, and Hubert Chièze

635

647

659

673

53 Urban Lakes: From Lack of Regard to Smart Deal? . . . . . . . . . . Olivier Fouché, Jérôme Brun, and Behzad Nasri

685

54 Reconstruction of Hydraulic Data by Machine Learning . . . . . . . Corentin J. Lapeyre, Nicolas Cazard, Pamphile T. Roy, Sophie Ricci, and Fabrice Zaoui

701

55 Assessment of the Impact of Changes in Storm Rainfall and Landscape Characteristics on the Maximum Flow of Small Rivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vitali Ilinich, Aleksey Perminov, Aleksandr Belolybcev, and Anna Naumova 56 Novel Quantification Method for Hydrograph Similarity . . . . . . . Dadiyorto Wendi, Bruno Merz, and Norbert Marwan 57 New Tools to Assess the Suitability of Physical Habitat (SPH) and the Weighted Usable Area (WUA) for Fishes . . . . . . . . . . . . . Ernest Bladé, Marcos Sanz-Ramos, Damià Vericat, and Antoni Palau-Ibars 58 Coastline Change Assessment for Quang Ngai Province Using Landsat Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Duc Tho Bui, Quang Binh Nguyen, Ngoc Duong Vo, Thi Thanh Thu Thai, and Nguyen Philippe Gourbesville 59 20 Years of Coastal Events Modelling . . . . . . . . . . . . . . . . . . . . . . Olivier Bertrand, Thibault Oudart, Eric David, Aurélie Ledissez, and Christophe Coulet

717

727

735

745

757

Contents

xv

60 Modelling NH4 + Dispersion from Wastewater of Urban Drainage to the Coastal Area of Danang City, Vietnam . . . . . . . . Phuoc Quy An Nguyen, Philippe Gourbesville, Ngoc Duong Vo, Philippe Audra, Morgan Abily, and Qiang Ma

767

61 Uncertainty Quantification of Bathymetric Effects in a Two-Layer Shallow Water Model: Case of the Gibraltar Strait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nabil El Moçayd, Alia Alghosoun, Driss Ouazar, and Mohammed Seaid

779

Part IV

High Performance Computing and Complex Hydraulics Applications

62 Vortex Siphon – From 1:1 Scale Physical Model to SPH Simulation and Prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arnaud Bart, Thibault Macherel, Giovanni De Cesare, Sean Mulligan, and Khalid Essyad

795

63 Hydrodynamics of an Innovative Discontinuous Double Breakwater, Mixed Modeling: 2D Flume Physics and 3D Digital Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christian Raffourt, Charlie Vergnet, Vasileios Afentoulis, and Philippe Bardey

809

64 Simulating the Hydrodynamics of Sewer-Inlets Using a 2D-SWE Based Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marcos Sanz-Ramos, Jackson D. Tellez Alvarez, Ernest Bladé, and Manuel Gómez-Valentín 65 Modelling of Surcharge Flow Through Grated Inlet . . . . . . . . . . . Jackson D. Tellez Alvarez, Manuel Gómez, and Beniamino Russo

821

839

66 A Predictive Data-Driven Approach Based on Reduced Order Models for the Morphodynamic Study of a Coastal Water Intake . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rem-Sophia Mouradi, Cédric Goeury, Olivier Thual, Fabrice Zaoui, and Pablo Tassi

849

67 Hybrid-Parallel Simulations and Visualisations of Real Flood and Tsunami Events Using Unstructured Meshes on High-Performance Cluster Systems . . . . . . . . . . . . . . . . . . . . . Bobby Minola Ginting, Punit Kumar Bhola, Christoph Ertl, Ralf-Peter Mundani, Markus Disse, and Ernst Rank

867

68 Analysis of the Unsteady Flow Around a Hydrofoil at Various Incidences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lei Shi, Annie-Claude Bayeul-Lainé, and Olivier Coutier-Delgosha

889

xvi

Contents

69 Analysis of High Energy Impact of a Raindrop on Water . . . . . . Mohamed Houssein Ghandour, Annie-Claude Bayeul-Lainé, and Olivier Coutier-Delgosha 70 Experiment and Numerical Analysis of a Rotating Hollow Cylinder in Free Flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yusuke Naito, Romain Montini, Hirochika Tanigawa, Jun Ishimoto, Masami Nakano, and Katsuya Hirata 71 High-Speed Computation on Fluid Forces Acting on Various Oscillating 3D Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kyohei Matsumoto, Yusuke Yamaoka, Hideki Shimohara, Hirochika Tanigawa, and Katsuya Hirata 72 Approach of Dynamic Modelling of a Hydraulic System . . . . . . . . Naly Ratolojanahary, Joel-A Gonzalez-Vieyra, Patrick Dupont, Annie-Claude Bayeul-Lainé, Christophe Sueur, Thibault Neu, and David Guyomarc’h 73 Numerical Simulations of an Innovative Water Stirring Device for Fine Sediment Release: The Case Study of the Future Trift Reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Azin Amini, Anass Chraibi, and Pedro Manso 74 Assessment of the Performance of a Protection Bell for a Micro Power Station on the Rhône . . . . . . . . . . . . . . . . . . . . Julien Schaguene, Luc Bazerque, Philippe Mauger, and Olivier Bertrand 75 Comparison Between Two Hydraulic Models (1D and 2D) of the Garonne River: Application to Uncertainty Propagations and Sensitivity Analyses of Levee Breach Parameters . . . . . . . . . . Lucie Pheulpin, Vito Bacchi, and Nathalie Bertrand

907

923

937

951

967

979

991

76 PID Controllers as Data Assimilation Tool for 1D Hydrodynamic Models of Different Complexity . . . . . . . . . . . . . . . 1009 Miloš Milašinović, Budo Zindović, Nikola Rosić, and Dušan Prodanović 77 Validation of a Semi-automatically Calibrated 1-D Open-Channel Model Against Experimental Data with Changes in Channel Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023 Adeline Visse, François-Xavier Cierco, Luc Duron, Pierre-Loïk Rothé, and Yoan Gressier

Contents

xvii

78 Analysis of River Bed Variation Based on Hydrological and Hydraulic Models: A Case Study on Hosan Stream Watershed, South Korea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039 Seung Jin Maeng, Muhammad Azam, Seung Wook Lee, and Ju Ha Hwang 79 Hybridizing Optimization Method and Artificial Neural Network for Urban Drainage System Design . . . . . . . . . . . . . . . . . 1055 Soon Ho Kwon, Donghwi Jung, and Joong Hoon Kim 80 Recycling of TBM-Excavated Materials of the Paris Basin into Technosol: A Numerical Assessment of Its Hydrological Transfer Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063 Ghada Snoussi, Yijue Yang, Imen Bel Hadj Ali, Essaieb Hamdi, and Olivier Fouché

Part I

Decision Support Systems and Crisis Management

Over the last 20 years, several new paradigms emerged for water resources planning and management. Integrated Water Resources Management (IWRM) has been recognized as an important guideline for effective and sustainable water resources management. According to UNESCO (2009), IWRM can be defined as “a step-by-step process of managing water resources in a harmonious and environmentally sustainable way by gradually uniting stakeholders and involving them in planning and decision-making processes, while accounting for evolving social demands due to such changes as population growth, rising demand for environmental conservation, changes in perspectives of the cultural and economic value of water, and climate change”. Although the concept of IWRM was already discussed in the past decades, it is not yet established how to implement IWRM concepts in water management practice. The needs for a holistic approach for water resources management were also highlighted by many actors and were expressed in the European Union Water Framework Directive (WFD) that came into force in 2000. Despite this effort, water environmental management often falls into an unstructured problem where various stakeholders are involved and multiple criteria have to be evaluated. The decision process for planning or management of water environment therefore tends to become a very complex process. Decision Support Systems (DSS) have been conceptualized and developed to support this unstructured decision making process. Considering the rapid advancement of technologies related to DSSs, the current developments are recently regarded as an iterative process rather than a single procedure. This iterative development with active participation of stakeholders is also considered to make the DSS more sustainable, because the system can be gradually improved by incorporating feedbacks from the stakeholders and end-users. This approach is seen in many recent DSS projects for water resources planning and management. A DSS therefore tends to include a combination of simple and universal models with different functions for sustainable maintenance rather than a single sophisticated model in recent years. On the other hand, advents in computer science and information technology have increased the capability of real-time water resources management. More and more data can

2

Part I: Decision Support Systems and Crisis Management

potentially be used for real-time water resources management. They include real-time observation data of the target water system, real-time water demand data and real-time meteorological and hydrological forecast data. Although these data can be considered to be very useful in real-time water resources management, it became very challenging task to handle a huge amount of data in real-time. New approaches focused on data management and data technics represent today a major axis for DSSs development. Several papers gathered within this section are addressing the concepts and the operational implementation of DSSs in various environments. A major application field for DSSs is currently the crisis management. During water related crisis, stakeholders and first responders are looking for tools able to provide an accurate overview of the current situation and also to formulate reasonable forecasts in order to optimize actions and responses. In such context, hydro informatics tools represent some of the key components of the DSSs to develop and to implement in order to answer the crisis challenges. In addition to the classical hydrological and hydraulic models, catastrophe models can be implemented within those environments. The purpose of catastrophe modelling is to help communities and companies anticipate the likelihood and severity of potential future catastrophes before they occur so that they can adequately prepare for their financial impact. Insurances and reinsurance companies at the worldwide scale currently develop these approaches. Catastrophe modelling combines the four components - hazard, inventory, vulnerability, and loss - to aid insurers in making their decisions on what type of protection they can offer against a particular risk. Integration of hydro informatics methods and tools in these approaches is a real challenge is discussed in several contributions of this section. Sophia Antipolis August 2019

Philippe Gourbesville Guy Caignaert

Chapter 1

Which Models for Decision Support Systems? Proposal for a Methodology Philippe Gourbesville

Abstract Management of water uses requests to harmonize demands and needs which are getting more and more complex and sophisticated especially with the growing urbanization. Modern cities request a larger number of services for their inhabitants and expect, at the same time, to limit investments in order to constrain the tax pressure. The need of optimization appears at various levels and request the wide spread of monitoring strategies. At the same time, urban growth mobilizes last available spaces that are frequently under the thread of natural hazards like inundations or landslides. The current situation, characterized by the fast increase of monitoring devices mainly in the urban environments, requests an integration of the modeling tools into the Information Systems (IS) that are now dedicated to the global management of urban environments and related services. Decisions Supports Systems (DSSs) that may integrated various components both for real-time monitoring and forecast through model, appear as one of the most relevant answer to the urban environment management’s expectations. The models integration is a challenging task that requests to build a global vision that ensures both technical feasibility and sustainability. As demonstrated with the AquaVar approach, several models can be orchestrated within a single environment that can address the diversity of the water related issues handled by local technical services. The models selection has to integrate the evolution of the tools and the possibility to integrate gradually new approaches and methods that are more data oriented and using the results produced from the implemented deterministic tools. Keywords Water management · Information system · DSS · Monitoring · Real-time · Models · Forecasts · Var catchment

P. Gourbesville (B) Polytech Lab, Université Cote d’Azur, Polytech Nice Sophia, 930, route des Colles, 06903 Sophia Antipolis, France e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_1

3

4

P. Gourbesville

1.1 Introduction Management of water uses requests to harmonize demands and needs which are getting more and more complex and sophisticated especially with the growing urbanization. Modern cities request a larger number of services for their inhabitants and expect, at the same time, to limit investments in order to constrain the tax pressure. The need of optimization appears at various levels and request the wide spread of monitoring strategies. At the same time, urban growth mobilizes last available spaces that are frequently under the thread of natural hazards like inundations or landslides. New urban developments appear more vulnerable and request a higher effort for risk management based on systems able to anticipate and analyze situations. The current situation, characterized by the fast increase of monitoring devices mainly in the urban environments, requests an integration of the modeling tools into the Information Systems (IS) that are now dedicated to the global management of urban environments and related services. Energy distribution, water distribution, solid wastes collection, traffic optimization are today major issues for cities that are looking for functional Decisions Supports Systems (DSSs) that may integrated the various components and operate in a sustainable perspective. The current demand is targeting classical monitoring outputs such as the real time monitoring and request forecasts based on models (analytics) and providing sufficient information for an efficient management. In addition to the analysis of the current situation by visualizing the various information sources, a frequent request is on evolution of the monitored processes in time in order to anticipate reaction and ensure an efficient management. In order to provide a real support to the decision process, several tools dedicated to the data analysis and to the simulation can be interfaced within the core part of the platform. The models used in this analytics domain start with basic statistical tools and go to complex determinist models such as those commonly used in hydroinformatics. This architecture concept for the urban information system is today commonly shared and appears as a consensus solution. If the concept of DSS is clearly understood, the integration of models is still an important issue that’s not addressed by the modelers’ community. Up to now few operational implementations have been achieved at the international scale and prototypes are just emerging. The availability of computational resources allows today looking at the deterministic models for hydrological and hydraulic issues. Obviously those tools may easily produce massive data that could be used afterward by data mining technics and stochastic models associated to AI protocols. This target architecture requests a specific methodology that describes the various steps to achieve for a successful DSS design and implementation.

1 Which Models for Decision Support Systems? …

5

1.2 Context, Needs and Methodology 1.2.1 Towards Smart Cities and Smart Water Several projections confirm that 70% of the world’s population will live in a city by 2050. Currently, around half of all urban dwellers live in cities with populations between 100,000 and 500,000 people, and almost 10% of urban dwellers live in megacities, which are defined by UN HABITAT as a city with a population of more than 10 million. As cities around the world experience this massive growth, the need to ensure sustainable expansion, efficient operation and development of high quality of life for residents becomes even greater than it is today. Within this context, the smart city concept has emerged. The term “smart cities” is trending amongst governments, urban planners and even the private sector to address the projected demands of cities in the future. Making cities smarter to support growth is emerging as a key area of focus for governments and the private sector alike. Up to 2030, cities around the world will invest US$ 108 billion in smart city infrastructure, such as smart meters and grids, energy-efficient buildings and data analytics, according to Navigant Research (https://www.navigantresearch.com/news-and-views/globalrevenue-from-smart-water-networks-projected-to-reach-72-billion-in-2025). Smart cities encompass six important sectors that need to work in unison to achieve a common goal of making a city more livable, sustainable and efficient for its residents. These sectors are smart energy, smart integration, smart public services, smart mobility, smart buildings, and smart water. Building smart cities upon the six sectors is crucial for sustainable global growth, but the financial, logistical and political challenges are enormous. The conversations about growth of smart cities have historically been dominated by large IT companies that focus on analyzing “big data” taking a top-down, software-centric approach. However, when it comes to the modernization of hundred-year-old systems like water distribution or the power grid, advanced software and networking capabilities are rarely broad enough in scope to make the necessary impact. Conversely, a bottom-up approach to smart city development is based on the belief that the rapid migration to cities will tax municipal infrastructures beyond their breaking points. The cities that succeed in transitioning to “smart” operations will be those that improve their critical systems and infrastructure at a fundamental level as well as integrate their systems through advanced technology. Lastly, smart cities will apply advanced monitoring and analytics to continuously measure and improve performance. One of a city’s most important pieces of critical infrastructure is its water system [1]. With populations in cities growing, it is inevitable that water consumption will grow as well even if the individual use will decrease. The term “smart water” points to water and wastewater infrastructure that ensures this essential resource—and the energy used to transport it—is managed effectively. A smart water system is designed to gather meaningful and actionable data about the flow, pressure and distribution of a city’s water. Further, it is critical that the consumption and forecasting of water use is accurate. A city’s water distribution and management system must be sound

6

P. Gourbesville

and viable in the long term to maintain its growth and should be equipped with the capacity to be monitored and networked with other critical systems to obtain more sophisticated and granular information on how they are performing and affecting each other. Additional efficiencies are gained when departments are able to share relevant, actionable information. One example is that the watershed management team can automatically share storm water modeling information that indicates probable flooding zones and times based on predictive precipitation intelligence. The transportation department can then reroute traffic accordingly and pre-emptively alert the population using mass notification. Water systems are often overlooked yet as critical components of energy management in smart cities, typically comprising 50% of a city’s total energy spends. Energy is the largest controllable cost in water/wastewater operations; yet optimizing treatment plants and distribution networks has often been overlooked as a source of freeing up operating funds by cash-strapped municipalities. Once facilities are optimized and designed to gather meaningful and actionable data, municipal leaders can make better and faster decisions about their operations, which can result in up to 30% energy savings and up to 15% reduction of water losses. Water loss management is becoming increasingly important as supplies are stressed by population growth or water scarcity. Many regions are experiencing record droughts, and others are depleting aquifers faster than they are being replenished. Incorporating smart water technologies allows water providers to minimize non-revenue water (NRW) by finding leaks quickly and even predicatively using real-time SCADA data and comparing that to model network simulations. Reducing NRW also allows municipalities to recover costs incurred in treatment and pumping. The reduction of NRW is a priority for cities in both developed and developing countries in order to ensure efficient service to population and sustainable use of water resources. On the wastewater side, there is a move by many water utilities—public and private—to transform wastewater treatment plants into resource recovery facilities, which includes energy. There are several examples of facilities that now produce more energy than required for their operations and sell the excess energy back to the grid. While this is not practical for all treatment plants, it is a worthy ambition for most of the major treatment sites and should be included within the implementation roadmaps or master plans at the national level. However, implementation requests to improve financial capacity of municipalities in order to implement the smart water approach and to contribute to the water security in a global way. One of the biggest obstacles to any capital-intensive project is access to funding. As cities and municipalities look to achieve smarter water, there are a number of options available to help them get started. One very effective path is through leveraging energy-saving performance contracts (ESPCs). ESPCs are a form of a public-private partnership (PPP), a financial model that capitalizes on the flexibility and resources of the private sector to pay for energy-saving capital upgrades using future energy savings. The private financial community provides the initial investment, and services are delivered by Energy Service COmpanies (ESCOs). The financier is paid from the accrued energy savings, with the ESCO guaranteeing the savings amount. An ESPC starts with an energy audit. After identifying opportunities and quantifying the potential

1 Which Models for Decision Support Systems? …

7

savings, the ESCO recommends any number of energy conservation measures, such as equipment retrofits, pumping optimization, demand monitoring and control (DSSs can be created and developed), and/or load-shedding and cogeneration which will save energy through more efficient operations.

1.2.2 Towards the Water Information System In the coming years the new technologies from the IT sector will affect the full water cycle and the management of the water related services. However, the impact of these new technologies—from sensors to Decision Support Systems (DSSs)—could be stronger and really significant if priorities are properly defined and implemented within the R&D and deployment strategies. The main driver of the strategy has to be to achieve a comprehensive architecture of an Information System (IS) dedicated to water uses and connected to others systems involved in human activities. This is the operational formulation of the smart water concept. By definition, Information Systems are implemented within an organization for the purpose of improving the effectiveness and efficiency of that organization [2]. Capabilities of the IS and characteristics of the organization, its work systems, its people, and its development and implementation methodologies together determine the extent to which that purpose is achieved. The IS is associated to an architecture which provides a formal definition of the business processes and rules, systems structure, technical framework, and product technologies for a business or organizational information system. In order to elaborate a specific IS for the management of the water cycle, a methodology is needed for identifying priorities and strategic investments to do in the ICT domain. The requested approach has to investigate all domains and provide a map of the various process taking places in the different domains of the water uses cycle. This formalization exercise, using mainly concepts and processes, is requested in order to ensure the coherence of technical choices in a holistic approach. Most of municipalities are currently engaged to this approach in an explicit or implicit way: monitoring activities are gradually introduced and allow improving the efficiency of water management, from resources to treatment operations and environment quality monitoring [3–5]. The availability of the real time monitoring systems provides a significant improvement within the management of water related services. One of the key challenges is to ensure that each specific monitoring system can integrate a wider system covering all the urban management actions. This step is highly challenging as it requests to address the legacy of each system within the target one. High financial investments can be requested and efficiency may suggest completely forgetting an existing technical solution in order to move to a more open and interoperable approach. In addition to the development of real time monitoring systems (dashboards), the need for forecasts is the following step and requests to implement modeling tools that can operate in real time too and produce realistic forecasts on the various

8

P. Gourbesville

processes that have to be managed: water consumption, pressure, flood and associated inundation [6], urban runoff, accidental pollutant behavior, etc. The models integrate an analytics domain that is added to the classical dashboard and provide the added value to the stakeholders. The shared information helps to consolidate a common approach especially for the crisis management and the optimization of the mitigation actions.

1.2.3 Methodology for Models Selection At first, the main target should be the creation and the development of a Water Information System (IS) [3] that provides the relevant resources for the services managers. The global architecture for this IS has to become explicit and a roadmap for the urbanization of this IS has to be produced by the relevant entity (most of time Municipalities and associated technical services). In most of the cases, the definition of the target IS—at the city scale—integrates existing monitoring systems in order to consolidate the current architecture and to address the legacy issues. When the global roadmap is defined and covering the forecast objectives/expectations, the design of the specific water IS can be addressed and the selection of required models can be initiated. Obviously, the consolidation of the Water IS cannot be achieved at the initial stage and it requests a continuous efforts. When the water IS roadmap is clarified with relevant objectives, the selection of models can be done based on the requested added value of the forecasts and the availability of data and computational resources. In order to maximize the efficiency of the DSS, a common format/standard for the data and for exchanges among the different tools is highly recommended and contribute to the sustainability of the Water IS. The implementation of standardized workflows ensures interoperability with the global IS that covers the various urban services. The modeling tools have to be selected for their performance to provide in the define timing the relevant forecasts: running a deterministic tool requesting a computational time larger that the process to forecast is obviously irrelevant. The modeler task is then to assess both quality of delivered results and operational implementation within the management procedures. If data and real-time monitoring systems are operating, and according to the processes to address, the key principle is to select the model that is allowing to deliver the relevant answer in the minimum of time. The deterministic models for both hydrological and hydraulic processes (surface and underground) represents a meaningful approach as relevant results can be obtained with limited data sets and assumptions that are based on physical laws. The use of these models within the operational phase will generate results that could be used afterwards as inputs data for stochastic models using AI technics such as ANN or multi agents. This last step may contribute to reduce significantly the computational efforts and the simulation time. This new performance can be very helpful for managing services (Fig. 1.1).

1 Which Models for Decision Support Systems? …

9

Fig. 1.1 Methodology for models integration within City IS

1.3 Aquavar Approach 1.3.1 Nice and Var Catchment Context The city of Nice is located on the French Riviera at the mouth of the Var catchment. The recent urban development of the fifth largest French city is currently taking place in the last available space along the Var low valley and over about 20 km of floodplain. Due to the complexity of challenges—water supply security issues from groundwater resources, inundation risk and water resources management under the perspective of climate change—the need for a DSS has been identified since the late 90’s. Unfortunately, at such time, both availability of data and technical tools (from communication protocols to modeling tools) has not permitted to engage the development of such system. However, during the last 15 years, systematic data collection on topography, climate and hydrological variables has permitted to gather a significant knowledge on the main hydrological processes within the Var catchment. Since 2014, a new approach has been engaged with the AquaVar project dedicated to the development and implementation of a first DSS able to address a wide diversity of issues: from resources management to emergency situations management [5].

10

P. Gourbesville

Fig. 1.2 Global architecture for the Nice Metroplis IS and the integration of AquaVar DSS within the Analytics domain [5]

1.3.2 Global Architecture The selected architecture for the AquaVar DSS is based on a platform elaborated over a service bus dedicated to collect and integrate field data that are related to various processes including the water services and the natural hazards. Data are formalizes through various tools such as Key Performance Indicators (KPIs), predefined alerts and directives. The synthetic dashboard allows visualizing the current situation. In addition, with the analytics components, the platform integrates deterministic modeling solutions which allow to have a full simulation of the hydrological cycle at the catchment scale, a 3D simulation of complex underground aquifer and associated relationships with 2D/3D surface flow model including pollutants exchanges. The modeling system integrated within the hypervision platform is based on 3 deterministic modeling systems (Fig. 1.2).

1.3.3 Implementation of Models For the Var low valley, the demands from the local government are targeting the water resources with the groundwater located within the low valley, the exchanges between the surface flows and the groundwater especially in case of accidental pollution and the flood events that could generate inundations and impacts on urban and commercial areas. The main requests are both for a real-time information on the current processes and on the possibility to assess a future situation through modeling tools. The models integrate the Analytics domain in the global Information System (IS) architecture

1 Which Models for Decision Support Systems? …

11

and are connected through the Service Bus to the various data sources such as water levels, discharges and water quality parameters. The hypervision interface allows to display the measurements and to interact with the modeling tools that produce the simulations. One of the key questions is obviously on the choice for the modeling tools to be integrated within the Analytics domain. In order to provide the requested diagnostics and simulations, the following modeling systems have been chosen and interconnected: • The FEFLOW modeling system, developed by DHI, for the 3D simulation of the groundwater resources simulation. In order to represent the interactions between the river and the groundwater table, the FEFLOW model is combined with a 2D surface water model; • The MIKE 21 system (DHI) is used as 2D surface water model and is connected with FEFLOW for the surface/groundwater interaction simulation. In addition, the system is used for flood events simulation and for the modeling of the morphological dynamic within the riverbed; • The MIKE SHE system (DHI) produces the hydrological data to be used as boundary conditions for FEFLOW and MIKE 21 systems (Fig. 1.3). A 3D hydraulic model based on FEFLOW modeling system has been set up over the 22 km of the Var low valley. The detailed geological structure has been integrated within the model in order to have an accurate representation of the processes [7–10]. The validation of the model has been achieved with a simulation from September 10th 2009 to February 26th 2013. Among the 24 piezometers with automatic recorder which have been set up to monitor the daily groundwater level along the valley, 6 of them have been chosen to validate the model thanks to their fully digital recording during the simulation period. Their location enables a holistic view from the upstream to the downstream (Fig. 1.4). The simulation results are shown with the measured data

Fig. 1.3 Extension of the MIKE SHE, MIKE 21 and FEFLOW models integrated within the AquaVar DSS

12

P. Gourbesville

Fig. 1.4 Piezometers used for FEFLOW model validation

Fig. 1.5 Comparison between simulated (FEFLOW model) and recorded groundwater levels

in Fig. 1.5. The results demonstrate that the model is able to represent the dynamics of the groundwater flow by considering direct water recharge, river-aquifer exchange as well as the groundwater extraction. Consequently, the model can be used as a groundwater management tool and integrated within the hypervision platform. A similar approach has been carried out with MIKE 21 FM regarding the free surface flows simulation and the morphological dynamic. The simulation of the bed evolution has been carried out with Sand Transport module in MIKE 21 FM that calculates the sediment transport capacity, the initial rates of bed level changes and the morphological changes for non-cohesive sediment due to currents. The sediment transport computation is based on hydrodynamics conditions and sediment properties. In order to obtain an efficient MIKE 21 FM model, several meshes have been created to simulate the same flood event (3rd October 2015 to 6th October 2015). The built model with a 10 m resolution combining triangular and quadrangular elements has demonstrated efficiency and well reproduced observed values. High-resolution mesh has been implemented in order to represent properly the hydraulic structures and their effects (Fig. 1.6). For the hydrological modeling, a similar approach has been implemented with MIKE SHE over the full catchment. The validation has been carried out over a period of 3 years after the validation of the numerical grid to use for the surface runoff estimation. Good results have been also obtained with this deterministic approach that provides the input data for FEFLOW and MIKE 21 systems. The 3 modeling systems are currently integrated within the AquaVar engine that is deployed with the Information System operated by Nice Côte d’Azur Metropolis services.

1 Which Models for Decision Support Systems? …

13

Fig. 1.6 Various mesh sizes tested within the 2D hydraulic model (MIKE 21)

1.3.4 AquaVar Orchestration One selected approach for the AquaVar DSS is the use of common modeling software as non-interactive services. Modeling systems like Mike SHE, Mike 21 or FEFLOW are commonly used on the desktop computer as highly interactive applications where the user can take advantage of the numerous visualization features available. Conversely, in the AquaVar DSS, these models are used in batch mode and are viewed as modules managed by a program named the orchestrator. The AquaVar engine (Fig. 1.7) automates the management of the modeling services by coordinating the exchange of data through their interactions. The engine consists in the following modules: • Simulation engines: a simulation engine is a wrapper around specific simulation software like Mike SHE, Mike 21 or FEFLOW. The wrapper makes it easy to add a new simulation engine with no change in the architecture; • Configuration modules: each simulation engine relies on a corresponding configuration module to automatically set up the simulation parameters. The configuration module is also able to perform data format conversion when necessary; • Scheduler: the scheduler allows running automatically the simulation engines in the background at regular intervals. The scheduler uses a table similar to a Unix crontab which can be set up by the user;

14

P. Gourbesville

Fig. 1.7 AquaVar orchestrator that combines the various modeling tools

• Data acquisition module: this module fetches the input data from a data store which itself collects live data from sensors. This module implements a common interface which allows easy connection to new data sources; • Data delivery module: this module pushes the result files from simulation engines into a database. Similar to the previous one, this module implements a common interface that allows delivering the results in different ways (database, plain file system, etc.); • Orchestrator: the orchestrator manages all the other modules and can be parameterized by the user through a web-based user interface. The user can set all the simulation parameters including the time step for the simulation engines. The administrator is also able to stop or restart a simulation engine. The data acquisition and data delivery modules as well as the orchestrator communicate with external applications through RESTful (Representational State Transfer) interfaces [11]. RESTful Application Program Interfaces (APIs) rely on the HTTP protocol and provide an architecture style suitable for networked applications. This architecture allows to transparently access the AquaVar engine through a web-based user interface or alternatively as a web-service. The current version of the AquaVar engine is written in Java 1.8 [12] and the RESTful APIs are powered by Jersey [13] (Fig. 1.8).

1 Which Models for Decision Support Systems? …

15

Fig. 1.8 AquaVar web interface allowing accessing to the various models and real time data sets on the Var catchment

1.4 Conclusions The development of DSSs is a current trend within the process of digitalization of the water services. The continuous increase of available sensors allows today developing efficient real-time monitoring systems that provide a first service to the technical services in charge of the water resources and associated services. Due to the growing complexity of tasks to achieve, numerous operators more and more frequently express the needs for anticipation based on reliable forecasts. The new challenge for the hydroinformatic community is to provide the relevant modeling tools that can be integrated within the framework of the real-time DSSs. Obviously, a generic and standard approach cannot be duplicated to all cases and a detailed procedure has to be defined for each case. A relevant initial analysis has to be achieved

16

P. Gourbesville

trough the various business processes that have to be simulated. In frequent cases, a good anticipation requests an accurate description of the physical processes involved. In this specific case, the deterministic models may offer a relevant approach due to their robustness and their capacity to produce results within a physically based range. The implementation of these models within DSSs requests to develop a strategy of their orchestration and their exploitation in real-time. After a first stage where the deterministic tools have produce a sufficient quantity of data, stochastic models can be introduced and associated to the initial modeling environment. This combination allows improving forecasting performances and reducing the computation time. The suggested methodology has been successful used for the AquaVar DSS. The application over the Var catchment has demonstrated all the interest to combine various deterministic tools that where not initially design for this cooperation. The orchestration principles can be easily duplicated and can be reused with other modeling environment according to the water issues that have to be addressed. Acknowledgements This research is currently developed within the AquaVar project with the support of Metropole Nice Côte d’Azur, Agence de l’Eau Rhone Mediterranée, Nice Sophia Antipolis University, Conseil Départemental 06 and Météo France. The work benefited from the data provided by the Métropole Nice Côte d’Azur, Conseil Départemental 06, Météo France and H2EA.

References 1. Gourbesville P (2019) Smart water solutions for water security: from concept to operational implementation. In: UNESCO, water security and the sustainable development goals, UNESCO, Paris, pp 47–67. https://unesdoc.unesco.org/ark:/48223/pf0000367904.locale=en 2. Silver M, Markus M, Mathis Beath C (1995) The information technology interaction model: a foundation for the MBA core course, MIS Q 19(3). Special issue on IS curricula and pedagogy (Sep., 1995), pp. 361–390. ISSN 1937–4771 3. Gourbesville P (2011) ICT for water efficiency. In: Environmental monitoring. Intech 4. Gourbesville P, Batica J, Tigli JY, Lavirotte S, Rey G, Raju DK (2012) Flood warning systems and ubiquitous computing. La Houille Blanche 6:11–16 5. Gourbesville P, Du M, Zavattero E, Ma Q, Gaetano M (2018) Decision support system architecture for real-time water management. In: Gourbesville P, Cunge J, Caignaert G (eds) Advances in Hydroinformatics. Springer, Singapore, pp 259–272 6. Demir I, Krajewski WF (2013) Towards an integrated flood information system: centralized data access, analysis, and visualization. Environ Model Softw 50:77–84 7. Potot C, Féraud G, Schärer U, Barats A, Durrieu G, Le Poupon C, Travi Y, Simler R (2012) Groundwater and river baseline quality using major, trace elements, organic carbon and Sr– Pb–O isotopes in a Mediterranean catchment: the case of the Lower Var Valley (south-eastern France). J Hydrol 472:126–147 8. Moulin M (2009) Nappe de la basse vallée du Var (Alpes-Maritimes), suivis 2006 quantité et qualité. BRGM, France 9. Guglielmi Y (1993) Hydrogéologie des aquifères Plio-Quaternaires de la basse vallée du Var, PhD thesis, Université d’Avignon et des Pays du Vaucluse, France 10. Guglielmi Y, Mudry J (1996) Estimation of spatial and temporal variability of recharge fluxes to an alluvial aquifer in a fore land area by water chemistry and isotopes. Groundwater 34(6):1017– 1023

1 Which Models for Decision Support Systems? … 11. RESTful. https://en.wikipedia.org/wiki/Representational_state_transfer 12. Java 1.8. http://www.oracle.com/technetwork/java/javase/overview/index.html 13. Jersey. https://jersey.github.io/

17

Chapter 2

Aquavar: Decision Support System for Surface and Groundwater Management at the Catchment Scale Qiang Ma, Philippe Gourbesville, and Marc Gaetano

Abstract Due to the impacts of global warming or climate changes, the hazard caused by extreme weather event becomes more frequent and serious. At same time, the last available place such as the floodplain has been strongly encroached by the growing of urbanization, which could lead more citizens to be exposed to flood risks. In current situation, the flood caused by extreme rainfall event could be characterized with shorter response time and higher flood damages. To efficiently manage this kind of flood hazard and effectively reduce the damage cost, the Decision Support System (DSS) applied in urban management has been requested to be able to produce comprehensive view of current situation in real time and further provide accurate forecast as faster and possible. Benefited from the progress of informatics and monitoring techniques, the fast increase of monitoring devices lets the real time data collection become more feasible. And the development of modelling system of hydrology and hydraulic has reach a higher level during last decades, which nowadays is bale to integrated assess the existing catchment water system and further forecast the incoming situation. Integration of those new techniques into DSS, the design of the DSS architecture should be reorganized to make the system become more operational and functional. This paper presents a generic operational DSS approach in order to address the management of natural hazards (floods & draughts) in a sophisticated urban environment and provide both real time assessment and forecast on a Web-based information platform. The proposed approach is illustrated with one model integrated real time DSS application (AquaVar DSS) on Var catchment (2800 km2 ) located at French Riviera. Three deterministic distributed model applications of hydrology, hydraulic and groundwater has been integrated into the modelling system at analytic part of the DSS and linked with a Web-based user interface to provide sufficient information for real time risk management in the area. The Q. Ma · P. Gourbesville (B) · M. Gaetano Polytech Lab, Université Cote d’Azur, Polytech Nice Sophia, 930, route des Colles, 06903 Sophia Antipolis, France e-mail: [email protected] Q. Ma e-mail: [email protected] M. Gaetano e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_2

19

20

Q. Ma et al.

integrated representation provided by AquaVar DSS has confirmed the feasibility of applying this DSS approach for dealing with extreme hazards. And similar approach could also be implemented in other managements of urban environment and related services such as energy distribution or water distribution. Keywords DSS · Real time modelling system · Deterministic distributed model applications · Var catchment

2.1 Introduction Decision Support System (DSS) has been approbated as one of most efficient tools to systematically consider different opinions and demands from various stakeholders then further propose alternative measures that maximize overall satisfactions. The history of DSS progress can be traced back to late 1950s when business companies started to recognize the effective contributions of new techniques on optimizing decision making processes. More theoretical and technological researches in DSS have been implemented from 1960s [1, 21]. The conventional DSS concept has been gradually developed by considering categories of management activities and descriptions of decision making process [5]. With advancement in concept of DSS for specific objectives, its applications in water resource management have been developed in parallel. [14] have made the first attempt of DSS application in water resource management. Follows the box-and-line structure clarified by [21], one DSS has been developed to optimize irrigation plans at central Missouri, USA. The first DSS application in water resource management designed with clear system architecture is built by [10] for alleviating flood problem in Como Lake, Italy. After that, more DSS applications for water resource management have published across the world until 1990s [2, 13, 19, 22]. The Var catchment locates on French Riviera is the largest catchment at French Mediterranean region. The city of Nice (fifth largest in France) is located at the mouth of this catchment. And its recent urban development is currently taking place along the floodplain of the Low Var Valley (last 22 km at the downstream of Var catchment). The challenges of water resource management in this region show higher complexity including security issues of water supply from groundwater resource, inundation risk and environment management under perspective of climate change. Since 1990s, the city of Nice has identified the need of a functional DSS to integrated manage all complicated water-related problems in this area. Unfortunately, at that time, neither the data shortage problem nor the technology limitation has seriously obstructed the DSS development. After 15 years’ systematic data collection on topography, meteorological and hydrological variables, which permits to gather significant knowledge and understanding on major hydrological processes within the catchment, in 2014, a new approach is engaged with project of AquaVar aims to develop and implement the first DSS to address a wide diversity of issues in the Var catchment: from resources management to emergency situations management [8, 9].

2 Aquavar: Decision Support System for Surface …

21

2.2 Context of Var Catchment With significant elevation variation from 0 m up to 3100 m above sea level, the Var catchment (2800 km2 ) is characterized with its steep slope distributed in middle and upper parts of the catchment. The Var river originates from spring at the south mountain pass and flows through a distance nearly 122 km to reach the catchment outlet between city of Nice and Saint Laurent du Var. There are three main tributaries contributed streamflow to the Var river: Estéron, Vésubie and Tinée. More than 90% of streams in this catchment are characterized as typical mountain streams with “V” shaped cross-sections formed by natural erosion (Fig. 2.1). The meteorological condition of this basin is typical Mediterranean climate with hot dry summers and cool wet winters. The annual precipitation in this catchment is around 815 mm, mainly concentrated in 65–80 days over the year. The catchment surface runoff is contributed by instance precipitation, snow melting flow and exchange flow between rivers and aquifers. Two flood periods can be identified in the catchment: spring floods caused by rainfall event combined with snow melting from summits of Alps Mountain and winter floods according to extreme precipitation event covering wide areas. With the runoff records at the Napoléon III Bridge (1985–2014) located at the

Fig. 2.1 Var catchment at French Mediterranean region

22

Q. Ma et al.

outlet of catchment, the annual average discharge of Var catchment is around 50 m3 /s, while the highest observed instantaneous discharge at flood peak can reach 3750 m3 /s (flood 1994). The downstream part of Var catchment (last 22 km), so called the Lower Var Valley, has connected the mountainous area and the Mediterranean Sea, contains rich groundwater resources [17]. The shallow aquifers in the Lower Var Valley are strongly interacted not only with the rivers but also with the conglomerate bedrock underneath the alluvium [12]. The groundwater in the unconfined alluvial aquifers in the Lower Var Valley is the main resource for supporting the social activities in this region including drinking water supply for around 600,000 inhabitants, industry, and agricultural and domestic consumption [15]. The annual groundwater extraction at public pumping stations is nearly 50 million m3 [11]. Since the beginning of 19th century, human activities have been increasing in the Lower Var Valley. On one hand, the growing of urbanization requests more constructed areas reclaimed from river floodplain. And on another hand, with more social activities being implemented in this area, the extraction of groundwater resource is evidently increased. Regarding the frequent interaction between streamflow and groundwater flow, when the river morphology has been strongly reshaped by artificial embankments from 600–1000 m width to 150–280 m, the raise of flow velocity directly leads to the enhancement of river bed erosion and further reduces the groundwater table. It makes the pumping activities become more and more difficult. The most serious groundwater shortage has been recorded at 1967 when the groundwater table was 8 m below its static level. After that, reducing the river bed erosions starts gaining more attention and has been considered as the premier effective measure to maintain the groundwater table in this area. At end of 1980s, 11 weirs were constructed at different cross-sections of the river to reduce the erosion impacts and raise the groundwater level in unconfined aquifers. Today, many industrial and agricultural zones are located at upstream of the Low Var Valley, while the urban area and some main pumping stations are at the outlet (Fig. 2.2). Recognizing the challenges in groundwater management are in aspects of both quantity and quality, the local water service has an urgent demand of a DSS for comprehensively representing major physical processes related to groundwater flow. Moreover, the cities located at lower Var valley are regularly affected by serious flood hazard. The levee system along the river is supposed to defense the flood with return period up to 75 years. However, as result of the river morphological dynamic, their protection level at many places along the river has been strongly reduced. This situation requests more careful management in order to ensure a better understanding of flood mechanism and process and higher security for exposed persons and goods. Therefore, considering those complicated demands in many aspects, one most reasonable and feasible solution is to generate a functional DSS to produce sufficient real time information and forecasts for supporting the decision making process and maximizing the satisfactions of overall stakeholders.

2 Aquavar: Decision Support System for Surface …

23

Fig. 2.2 Artificial structures and social activities in lower Var valley

2.3 Aquavar DSS The AquaVar DSS (Fig. 2.3) [7–9] is based on a platform elaborated over a service bus dedicated to collect and integrate real time information including measurements, forecasts and data related to various processes such as water services and natural hazards. Data are formalized through various standard tools such as Key Performance Indicators (KPIs), predefined alerts and directives and transferred among different

24

Q. Ma et al.

Fig. 2.3 Designed architecture of AquaVar DSS

parts of the system. In the “Analytic” part of AquaVar DSS, three deterministic models have been implemented to produce a comprehensive view of all major physical processes in the catchment water cycle then further support decision making process by answering questions defined from “Operations center”. The simplified synthetic dashboard applied in online platforms at “Visualization” part allows all stakeholders, who even don’t have strong related background and knowledge, easily understanding the on-going situations and in-coming phenomenon. The main demands of local government in the Lower Var Valley are targeting on the groundwater and flood managements. The requests are both for a real-time information of current processes and of the possibility to assess a future situation through modelling simulations. Models integrated in “Analytic” parts are set up and updated by the data collected and transferred through service bus. One of the key issues in setting up this DSS is the model selection. In order to provide sufficient information to well represent the physical processes at any locations in the catchment, three deterministic distributed models have been applied in the analytic system (Fig. 2.4): • The MIKE SHE model system, developed by DHI, for simulating all major hydrological processes covered whole Var catchment area. Based on real time data collected through service bus, the model simulation is aimed to produce accurate boundaries for lunching more detail simulations in lower Var valley. • The MIKE 21FM model system is applied as 2D high resolution surface flow simulation in connected with FeFlow for representing the flow exchange between

2 Aquavar: Decision Support System for Surface …

25

Fig. 2.4 Integrated three deterministic distributed models in AquaVar DSS

rivers and shallow aquifers. In addition, this model is also used independently for simulating the flood hazards and morphological dynamic of riverbed. • The FeFlow model system applied in AquaVar DSS is responsible to provide 3D view of groundwater resources at lower Var valley. With detail geological structure described in the model, its simulations are able to produce accurate representation of all underground processes. Moreover, by coupling with 2D surface hydraulic model (MIKE 21FM), the interactions between river and groundwater table can be also well simulated.

2.4 Results The modelling process starts from MIKE SHE simulation of the catchment hydrological cycle. Based on high resolution simulation within daily and hourly time interval respectively, the MKE SHE applications in the Var catchment are able to provide precise descriptions of most of major hydrological and hydrogeological characteristics at any places in the catchment and further to support the applications of MIKE 21FM and FeFlow to have more detail simulation of the Lower Var Valley. The 3D groundwater model of FeFlow could be highlighted with detail geological structure data of the Lower Var Valley. The simulation has been validated from September 2009 to February 2013, which covered one serious flood hazard happened at November 2011 and one drought event recorded at summer 2012. There are 24 piezometers with automatic recorder for monitoring daily groundwater table. From the upstream to the downstream, 6 piezometers have been applied for validated the simulation (Fig. 2.5). The results demonstrate that the model system is able to represent the dynamics of

26

Fig. 2.5 FeFlow application in lower Var valley

Fig. 2.6 Example of AquaVar web interface

Q. Ma et al.

2 Aquavar: Decision Support System for Surface …

27

the groundwater flow by considering direct water recharge, river-aquifer exchange as well as the groundwater extraction. Moreover, the designed DSS is implemented as one online service, which allows all the users visiting the web-site to obtain the detail simulation results at each mesh cell (Fig. 2.6).

2.5 Conclusion Benefited from advancements of new techniques such as real time monitoring, Information System (IS) and hydro-informatics tools, the DSS can produce sufficient and accurate information for optimizing the decision making process then maximizing the satisfactions of overall stakeholders. However, to achieve this objective, the DSS architecture has to be designed for enlarging the effects of new technological applications. This paper has proposed a functional DSS architecture based on the interoperability of various models and integrated in online platforms allowing efficiently managing massive data and producing real time simulation and forecasts. The current approach has been well implemented within the AquaVar project in the Var catchment at the French Riviera. Three deterministic distributed models have been integrated in the analytic part of DSS to provide a comprehensive view of all major physical processes in the catchment water cycle. The AquaVar DSS is designed as one web service allows all stakeholders participating the management activities. The achievements of AquaVar DSS demonstrate both the efficiency of the approach and the interests from the management point of view. Moreover, the developed concept for the DSS could also be extended to various catchments and for other objectives. Acknowledgements This research is currently developed within the AquaVar project with the support of Metropole Nice Côte d’Azur, Agence de l’Eau Rhone Mediterranéen, Nice Sophia Antipolis University, Conseil Départemental 06 and Meteo France. The work benefited from the data provided by the Metropole Nice Côte d’Azur, Conseil Départemental 06, Meteo France and H2EA.

References 1. Anthony RN (1965) Planning and control systems: a framework for analysis. Harvard University Graduate School of Business Administration, Cambridge 2. Davis JR, Nanninga PM, Biggins J, Laut P (1991) Prototype decision support system for analysing impact of catchment policies. J Water Resour Plann Manag 117(4):399–414 3. EU (2000) Directive 200/60/EC of the european parliament and of the council of 23 October 2000 establishing a framework for Community action in the field of water policy, Official Journal of the European Communities, L327, 22.12.2000, pp 1–73 4. Global Water Partnership (2000) Integrated Water Resources Management, GWP Technical Advisory Committee Background Papers, No 4, 67 pp 5. Gorry GA, Scott Morton MS (1971) A framework for management information systems. Sloan Manag Rev 13(1):50–70 1989

28

Q. Ma et al.

6. Gourbesville P (2008) ICT for water efficiency. In: Ekundayo EO (ed) Environment monitoring. InTech, pp 411–426 7. Gourbesville P, Du M, Zavattero E, Ma Q (2016) DSS architecture for water uses management. Procedia Eng 154:928–935. 8. Gourbesville P, Gaetano M, Ma Q (2018) AquaVar: real time models for underground and surface waters management at catchment scale. EPiC Ser Eng 3:836–843 9. Gourbesville P, Du M, Zavattero E, Ma Q, Gaetano M (2018) Decision support system architecture for real time water management. In: Advances in Hydroinformatics, SimHydro 2017 – choosing the right model in applied hydraulics, pp 259–272 10. Guariso G, Rinaldi S, Soncini-Sessa R (1986) The management of Lake Como: a multiobjective analysis. Water Resour Res 22(2):109–120 11. Gugliemi Y (1993) Hydrogéologie des aquifères Plio-Quaternaires de la basse vallée du Var, PhD thesis, Université d’Avignon et des Pays du Vaucluse, France 12. Gugliemi Y, Mudry J (1996) Estimation of spatial and temporal variability of recharge fluxes to an alluvial aquifer in a fore land area by water chemistry and isotopes. Groundwater 34(6):1017– 1023 13. Haastrup H, Maniezzo V, Marrarelli M, Mazzeo RF, Mendes I, Paruccini M (1998) A decision support system for urban waste management. Eur J Oper Res 109(1998):330–341 14. Hashemi F, Decker W (1969) Using climatic information and weather forecast for decisions in economizing irrigation water. Agric Meteorol 6(4):245–257 15. Moulin M (2009) Nappe de la basse vallée du Var (Aples-Maritimes), suivis 2006 quantité et qualité. BRGM, France 16. Nohara D, Gourbesville P, Ma Q (2018) Towards development of efficient decision support system for integrated water resources planning and management, DPRI Annuals, No 61B (2018) 17. Potot C, Féraud G, Schärer U, Barats A, Durrieu G, Le Poupon C, Travi Y, Simler R (2012) Groundwater and river baseline quality using major, trace elements, organic carbon and Sr– Pb–O isotopes in a Mediterranean catchment: the case of the Lower Var Valley (south-eastern France). J Hydrol 472:126–147 18. Rahaman MM, Varis O (2005) Integrated water resources management: evolution, prospects and future challenges, sustainability: science. Pract Pol 1(1):15–21 19. Rizzoli AE, Young WJ (1997) Delivering environmental decision support systems: software tools and techniques. Environ Model Softw 12(2–3):237–249 20. Serrat-Capdevila A, Valdes JB, Gupta, HV (2011) Decision support systems in water resources planning and management: stakeholder participation and the sustainable path to science-based decision making. In: Jao C (ed) Efficient decision support systems – practice and challenges from current to future. InTech 21. Simon HA (1960) The new science of management decision. Harper Brothers, New York 22. Stansbury J, Woldt W, Bogardi I, Bleed A (1991) Decision support system for water transfer evaluation. Water Resour Res 27(4):99.443–99.451 23. UNESCO (2009) IWRM Guidelines at River Basin Level – Part 1 Principle, 24 pp. http:// unesdoc.unesco.org/0018/001864/1864e.pdf

Chapter 3

Anywhere: Enhancing Emergency Management and Response to Extreme Weather and Climate Events Morgan Abily, Philippe Gourbesville, Eurico De Carvalho Filho, Xavier Llort, Nicolas Rebora, Alexandre Sanchez, and Daniel Sempere-Torres Abstract ANYWHERE project aims to produce a platform to empower exposed responder institutions and citizens to enhance their anticipation and pro-active capacity of response to face extreme and high-impact weather and climate events. This is achieved through the operational implementation of cutting-edge innovative technology as the best way to enhance citizen’s protection and saving lives. ANYWHERE has implemented a Pan-European multi-hazard platform providing a better identification of the expected weather-induced impacts and their location in time and space before they occur. This platform is composed of a unique Multi-Hazard Early Warning System (MH-EWS) that supports a fast analysis and anticipation of risks prior the event occurrence, an improved coordination of emergency reactions in the field and help to raise the self-preparedness of the population at risk. The A4EU platform was designed to be adapted to provide early warning products and locally customizable decision support services proactively targeted to the needs and requirements of M. Abily · P. Gourbesville (B) Polytech Lab, Polytech Nice Sophia, Université Côte d’Azur, 930 route des Colles, Sophia Antipolis 0693, France e-mail: [email protected] M. Abily e-mail: [email protected] E. De Carvalho Filho Predict Services, Castelnau le Lez, France e-mail: [email protected] X. Llort Hyds, Edifici ParcUPC K2M, Jordi Girona 1-3, Barcelona, Spain e-mail: [email protected] N. Rebora CIMA Research Foundation, Savona, Italy A. Sanchez · D. Sempere-Torres Universitat Politècnica de Catalunya/CRAHI, Barcelona, Spain e-mail: [email protected] D. Sempere-Torres e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_3

29

30

M. Abily et al.

the regional and local authorities, as well as public and private operators of critical infrastructures and networks. It has been implemented and demonstrated in 7 pilot sites to validate the prototype and was successfully transferred to the real operation. The ongoing market uptake is ensured by the cooperation with a SME and Industry Collaborative Network, covering a wide range of sectors and stakeholders in Europe, and ultimately worldwide. Keywords Extreme weather · Climate events · Risks management · Platform · H2020 research and innovation action

3.1 Introduction Extreme Weather and Climate events are the cause of a number of hazards affecting our society through their impacts on the outdoor exposed activities and assets, and when interacting with exposed and vulnerable human and natural systems they can lead to disasters. According to the Global Assessment Report on Disaster Risk Reduction [1], economic losses from disasters such as earthquakes, flooding, storm surges, wind storms, cyclones and tsunamis are now reaching an average of 250–300 billion USD each year, and two thirds of them are due to extreme weather hazards such as floodings, storm surges, and windstorms. Moreover, the Intergovernmental Panel on Climate Change [2], stresses that extreme events have increased in frequency or magnitude (specially flash floods and related debris flows and landslides, storm surges, droughts, wildfires and heat waves), and, in parallel, populations and assets at risk have also increased [3]. In this context, developing tools to support decision makers in real-time coordination of the emergency management operations is crucial to face the challenge of extreme weather and climate events. These tools need to capitalize on the advances in observation systems and in forecasting models able to anticipate the phenomena triggering these events and their impacts. The objective of ANYWHERE is to empower EU exposed responder institutions and citizens, to enhance their anticipation and pro-active capacity of response to face extreme and high-impact weather and climate events. This was achieved during the three years stretch of the ANYWHERE project, through the operational implementation of cutting-edge innovative technology as the best way to enhance citizen’s protection and saving lives. ANYWHERE has implemented a Pan-European multihazard platform providing a better identification of the expected weather-induced impacts and their location in time and space before they occur. This platform support is based on a dual pan EU forecast/custom regional impact specification (resp. MHEWS & A4EU platforms, see Sect. 2) allowing a fast analysis and anticipation of risks prior the event occurrence, an improved coordination of emergency reactions in the field and help to raise the self-preparedness of the population at risk. This significant step-ahead in the improvement of the pro-active capacity to provide adequate emergency responses is achievable capitalizing on the advanced forecasting methodologies and impact models made available by previous research and

3 Anywhere: Enhancing Emergency Management and Response …

31

technology development projects, maximizing the uptake of their innovative potential not fully exploited up to now. In the next Sect. 2, an overview of ANYWHERE project is given through an introduction to its processes and explanation of pilot sites bottom up customisation approach. In following Sect. 3, ANYWHERE outcomes and products are detailed. Last Sect. 4 provides conclusive aspects and introduced the next chapters of this book which will elaborate in details on several successful aspects of ANYWHERE.

3.2 Anywhere: Process, Pilot Sites and Overview The core of the ANYWHERE project is the integration of a number of meteorological forecasts and advanced algorithms to transform them into impact forecasting inside the MH-EWS pan-European platform. Moreover, although forecasting becomes then available, it is necessary to translate a regional view into Impacts on precise locations, to focus the efforts to protect vulnerable critical points and infrastructures (schools, transportation networks and collective centres, among others), saving lives in the last term. Introducing a new paradigm for Emergency Management, the A4EU Platform allows control centres and first responders to go beyond meteorological forecast towards impact forecast with a new level of accuracy, helping to prepare before the events and increasing their efficiency during the emergencies.

3.2.1 Selected Approach The proposed strategy was to develop the integration platform MH-EWS and, based on cornerstone platform, to initially develop six (that ended up in seven) operational customized prototypes of the A4EU platform. The scenarios for the pilot tests have been chosen on the basis of their diversity (geographic size, orography, population density, most common meteorological risks), and the A4EU prototypes have been developed following a co-creation philosophy between technological companies and first responders under a bottom-up approach (driven by the end-user) and with shared common requirements: • Enable the integration of forecasts and hazard impact indicators provided by the MH-EWS to serve a pan-European platform (A4EU) with local layers of realtime data including risk and vulnerability analysis, open and commercial data, local sensor data, high resolution radar observations, crowd sourced data and social media information, for monitoring and decision-support during emergency management; • Propose adapted solutions to the local requirements, being able to build on or to interface with the present legacy systems through standardized interface layers

32

M. Abily et al.

Fig. 3.1 General overview of ANYWHERE development and strategy

and through offering specific “plug-ins” and toolkits to support integration and sharing of data; • Be able to cope the local and high-resolution information available at the responder/user site, and to interconnect them with the MH-EWS to allow it to run specific high-resolution versions of the forecasting algorithms using this local information on demand; • Be able to provide the products of the MH-EWS to any emergency management command centre to support decision making during weather-induced emergencies “anywhere” in Europe. To that end, the lessons learnt in the implementation and demonstration in the pilot sites will be used to ensure by construction that the A4EU platform and its prototypes developed will be able to be implemented lately in any other emergency control centre in Europe, providing efficient support to the decision makers during weather induced emergencies independently of their location. The initially six defined pilot sites that implemented the A4EU prototype in their respective emergency command centres represented a specific region. During the project execution and, due to the capability of the platform to be easily extended to any other place in Europe, a new pilot site was developed (the seventh) with a wider scope covering for the first time a whole country (Spain). Moreover, the ANYWHERE system includes a set of “built-in” decision support tools able to be used operationally to support decision making during weather induced emergencies, as well as to collect data from the field and social networks.

3 Anywhere: Enhancing Emergency Management and Response …

33

The Fig. 3.1 provide a global overview of the ANYWHERE project development roadmap which conducted to the elaboration of the ANYWHERE system. The products and services generated by the MH-EWS are the key to provide anticipation and to support more effective response during emergency events. ANYWHERE will offer these products and services to three typologies of users through three commercialization/distribution schemes: • Public Protection and Disaster Relief organizations (PPDR), and specifically emergency control and command centres (usually regional Civil Protections, but also municipal Civil Protections in some cases) coordinating first responders and issuing alerts and warnings during emergencies. The purpose is to support decision making during weather-induced emergencies. The way to connect with them is through the A4DEMOS prototypes developed in WP4. This falls into the B2G (Business to Government) market classification that is fully described in the Business Plans (WP7); • End-Users: i.e. Citizens and enterprises with activities located in risk areas. For them ANYWHERE offers a variety of self-protection tools in WP5 providing services to support the implementation of self-protection risk plans pre-defined by the End-Users. This falls into the B2B2C (Business to Business to Consumer) market classification that is detailed in the Business Plans (WP7). • Third-Parties: Implementing Partners (associated enterprises) that collaborate with the ANYWHERE consortium in the commercialization of MH-EWS products in additional markets. The MH-EWS (WP3) will prepare and define the protocols to connect with them in order to reach additional markets. This falls into the B2B (Business to Business) market classification that is be detailed on the Business Plans (WP7). The Fig. 3.2 illustrates the interactions between the components of ANYWHERE system and the position of A4EU within these interactions between the MH-EWS and its foreseen users. It details the interactions between the products and services provided by MH-EWS pan-European platform and the three ways to address the three typologies of users, B2G, B2B and B2B2C. Moreover, it is worth noting that all the developers of A4EU customized version in the project pilots have designed their prototypes in a way that allow them to be implemented lately in any other emergency control centre in Europe (ensured by construction) with low adaptation efforts. This capacity to use any of the developed prototypes to supply the MH-EWS forecasts and impact assessment products to any PPDR User (internal and external to the project), is seen as an opportunity to provide different options to the project to promote a successful market uptake of the ANYWHERE developments in the B2G Market sector which implementation and roll-out strategy will be fully detailed on the Business Plans (WP7). Thus, all the 7 prototypes are integrated the same ANYWHERE MH-EWS products and capabilities to cross them with local layer of information providing support to the decision-makers during weather induced emergencies but allowing different versions able to customize the user experience and the interaction with the decision procedures tailored to the needs of the pilot site operational end-users. The lessons

34

M. Abily et al.

Fig. 3.2 General schematization of ANYWHERE, including the interactions between the products and services provided by MH-EWS pan-European platform and the three ways to address the three typologies of users: B2G, B2B and B2B2C

learnt in the implementation and demonstration of these pilot sites will be used to gather experience and feedback to better design the B2G model for Market Uptake.

3.3 Anywhere Outcomes and Products 3.3.1 Pilots Success Stories October 2018 marked the start of a new stage in the implementation of the platform for Decision Support in Emergency Management Operation Services (A4EU) and its deployment on the control centres of the Public Protection and Disaster Relief (PPDR) organisations involved in the ANYWHERE project. After the completion of the training activities to qualify the operational staff in the use of the tool, now the platform will become operational during a 1-year demonstration test period in six Pilot Sites around Europe. In this stage, the teams belonging to the PPDR organisations of the Pilot Sites of Liguria (Italy), Catalonia (Spain), Bern Canton (Swiss Alps), South Savo (Finland), Rogaland (Norway) and Corsica (France) have deeply test operationally their local versions of the A4EU tool on their daily basis. On the other hand, all along this period, feedback is collected from end-users in order to help developers improving the system and ensuring that the needs of operational teams are met. The customisation capabilities of the platform enabled implementing setting up an additional prototype in a singular wider-scope pilot site covering the whole country of Spain, being deployed in the CENEM (Spanish National Emergency Centre).

3 Anywhere: Enhancing Emergency Management and Response …

35

In the first place, operational staff has tested the weather-induced hazard forecasting products and algorithms included the ANYWHERE’s Multi-Hazard Early Warning System (MH-EWS) and check their usefulness in covering their operational needs as well as in providing the impact estimation. It was also analysed the convenience of the data representation in the platform for a suitable support in the decision-making process, including the correct integration of the different data sources available. Furthermore, it is foreseen that the performance of the impact forecasting products and its accuracy will be evaluated through the comparison of the forecasts made by the models with the occurrence and/or with respect to the emergency actions triggered. All the impressions gathered and provided by practitioners about the usability and adequacy of the current available functionalities and tools allow to point out the benefits of the A4EU in terms of reliability/usability compared to other legacy tools already available in each Pilot Site. The feedback collected together with interaction to be held between all the involved parties (emergency managers and first responders, scientists, technical experts, IT developers, etc.) during the 1-year demonstration period contributes to a powerful pre-commercial support tool in the decision-making process during weather-induced emergencies.

3.3.2 Catalogue and Way Forward The Core development and outcome that feeds the ANYWHERE platforms and 3rd party integrations is the “ANYWHERE Multi-Hazard Early Warning System”. It integrates on a single source more than 300 forecasting algorithms, including snowfall, forest fires, flash floods, droughts, storm surges, heath impacts derived from heat waves and more. Using ANYWHERE, a Solution Developer or Service Provider can harness the full power of over a decade of forecasting science, now available to be integrated into Industrial and Civilian applications that can be exploited commercially. Available for institutions, government users, industries and developers, the MH-EWS makes possible to select and use any combination of products included in the catalogue, to be added into Consultancy Solutions or 3rd party developments with the easiness of dealing with a single supplier and technical support. These product’s are available at http://anywhere-h2020.eu/catalogue/ and listed by type of hazard as illustrated in Fig. 3.3. Although forecasting is available, it is necessary to translate a regional view into Impacts on precise locations, to focus the efforts to protect people at inhabited areas and vulnerable critical points. Introducing a new paradigm for emergency management, the A4EU Platform allows control centres and first responders to go beyond meteorological forecast towards impact forecast with a new level of accuracy helping to prepare before the events and increasing their efficiency during the Emergencies.

36

M. Abily et al.

Fig. 3.3 ANYWHERE catalogue available at: http://anywhere-h2020.eu/catalogue/

This new level of accuracy helps first responder to increase their Efficiency during the Emergencies. This information will help directly Citizens to raise their SelfProtection capabilities, including specific tools for the most vulnerable locations and also support the Emergency Response Control Centers and First Responders to increase their Efficiency during Emergencies.

3.4 Conclusions Introducing a new paradigm for emergency management, ANYWHERE system through the use of the MH-EWS in connection to the A4EU platform allows control centres and first responders to go beyond meteorological forecast towards impact

3 Anywhere: Enhancing Emergency Management and Response …

37

forecast with a new level of accuracy helping them to prepare before the events occurrence. This new level of accuracy helps first responder to increase their efficiency during the emergencies. This information will help directly citizens to raise their selfprotection capabilities, including specific tools for the most vulnerable locations and also supporting the emergency response control centres. For civil protection and first responders, the A4EU platform has been developed and implemented on seven pilot sites within Europe. An array of tools has been developed, each aimed to a particular user type: civil protection organizations, population and solution providers. The A4EU Platform introduces a new paradigm, increasing the granularity of the available forecasts, and locating the critical points, like schools, train stations, hospitals, and population on areas at risk. This accuracy allows the control centres to focus on locations instead of vast regions, helping them to warn and prepare the Society and First Responders before, during and after the extreme events, maximizing their capabilities. Acknowledgements This project has received funding from the European Union’s Horizon 2020 research and innovation program (H2020-DRS-1-2015) under the grant agreement no. 700099.

References 1. Global Assessment Report on Disaster Risk Reduction 2015 (UNISDR 2015). http://go.nature. com/DmbZkA 2. IPCC (2014) Climate change 2014: synthesis report. contribution of working groups I, II and III to the fifth assessment report of the intergovernmental panel on climate change. In: Pachauri RK, Meyer, LA (eds) Core Writing Team, IPCC, Geneva, Switzerland, 151 pp 3. Quevauviller P, Barceló D, Beniston M, Djordjevic S, Harding RJ, Iglesias A, Ludwigg R, Navarra A, Navarro Ortega A, Mark O, Roson R, Sempere D, Stoffel M, van Lanen HAJ, Werner M (2012) Integration of research advances in modelling and monitoring in support of WFD river basin management planning in the context of climate change. Sci Total Environ 440:167–177

Chapter 4

Use of Anywhere Products to Assess Risky Events on Southern France and Corsica (October 2018) Eurico de Carvalho Filho, Guillaume Lahache, and Alix Roumagnac

Abstract This last October was scene for some unusual and extreme weather events on the Mediterranean basin. Liguria, Catalonia, Mallorca, Corsica, Aude and Var were some of the many regions affected by heavy rains, floods, storm surge and gusty winds. More than damages on goods and infrastructure, these risky events claimed the life of several people and grave a deep scar on the soul of citizens, mayors and safety officers. The European Commission has been financing some projects to enhance early warning systems, in order to cope with potentially dangerous events, such as ANYWHERE. On this panorama, Predict Services was invited to implement a platform to test several data arising from that project for Northern Corsica firemen department in order to assist it during its safety and logistics decision-making. The platform was made available on early October; thus, Predict Services on-call engineer team was able to closely monitor some of the aforementioned events. The present paper presents the results and the return of experience of using these data on the events of Southern Corsica and Var departments (October 10th to 11th ), Aude department (October 14th to 16th ) and Haute Corsica (October 16th to 17th ). In these events, the main products related to rainfall and flash floods nowcast were used and provided pertinent information both on geographical location and chronology, which could be translated into savings. Keywords Flood events · Nowcast · Early warning system

E. de Carvalho Filho (B) · G. Lahache · A. Roumagnac 20 rue Didier Daurat, 34170 Castelnau le Lez, France e-mail: [email protected] G. Lahache e-mail: [email protected] A. Roumagnac e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_4

39

40

E. de Carvalho Filho et al.

4.1 Introduction Weather-related hazards are among the most impacting natural phenomena, thus being able to anticipate the hydrometeorological risks (floods, storms, etc.) has become one of the biggest issues for many socioeconomic sectors. Around the world, several severe weather events such as thunderstorms, floods and tropical cyclones happen every year provoking human losses, costing damages and activity disruptions. In order to reduce the vulnerability, it is imperative to make systems more prepared, stronger and better equipped to face extreme events, increasing its preparedness its rapidity to re-establish its pre-crisis working. Aware of this, H2020 EU ANYWHERE Project aimed to improve overall ability to minimize the impact of major hydrometeorological events by the creation of pilot platforms gathering several models and algorithms as well as several sources of real time data. PREDICT Services (hereinafter called PREDICT) was thus invited to create a platform dedicated to Northern Corsica firefighters and civil protection department (SIS2B) called EU4Cor. The partnership was created in order to provide a decision-making support and a crisis management assistance to cope with major flood events. Based on risk expertise, 24/7 watch and timely warning dissemination, PREDICT assistance and ANYWHERE data contribute to improve the resilience, helping to improve activity safety towards citizens, their goods, infrastructure and to reduce economic vulnerability. For more than 10 years, PREDICT has followed municipalities in mainland France and overseas territories, as well as companies, and insurance companies involved on establishing a comprehensive risk reduction/prevention programs and real-time crisis management. This service was developed based an integrated approach within preparedness, early warning system and feedback. The first step to elaborate an effective system is to assess and cross hazard and vulnerability. PREDICT has initiated the transfer of its methods and technologies to meet the requests expressed internationally. Since its debut with a project in Haiti (2009), PREDICT has been abreast of major risks managers abroad, transmitting its knowledge, its methods and tools e.g. audit of existing systems and identification of areas for improvement, vulnerability analysis, improvement of crisis management strategies, implementation of warning systems, crisis management training, exercises. Internationalization has been notably made in the context of projects with the Inter-American Development Bank (for the project in Haiti) but also the World Bank (for the projects in Djibouti, Morocco, Buenos Aires, Argentina, and Senegal) for the improvement of early-warning systems and reinforcement of capabilities. This process has been highlighted since 2016 with the implementation of preventive solutions for 350 industrial risk managers on more than 35 countries for the automotive supplier, oil and gas company on Nigeria, Indian express air and integrated transportation and distribution company as well as assisting the city of Rio de Janeiro during the Summer Olympic Games in 2016.

4 Use of Anywhere Products to Assess Risky Events …

41

4.2 Flood Risk in Mediterranean France Around the world, flood is the most common hazard and also the deadliest natural disaster. In average, almost 10,000 people died in the world due to flooding which represents 45.4% of deaths due to natural disasters [5]. In France, a Preliminary Assessment of Flood Risks (or EPRI in French) performed by its government in 2012, found that nearly 1 of 4 French and 1 of 3 jobs are potentially exposed to the damage caused by floods. The annual amount of flood damages is estimated from e 650 million to e 800 million, which represents an important part of all costs related to natural disasters [9]. Increasingly frequent floods can put the economy of an entire country suddenly at risk. Nîmes (1988), Vaison-la-Romaine (1992), Aude (1999), Sommières (2002), Côte d’Azur (2015), Aude (2018) are a few examples that are engraved on French people’s memories. Nowadays, the Floods Directive issued in October 23rd, 2011 is the ultimate regulation in Europe over the subject. The whole European risk strategy relies on the concepts of vulnerability reduction and land resilience improvements. Thus, understand these concepts are vital to comprehend the flood risk management strategy. According to the Larousse dictionary, risk is “the probability of a fact, an event considered a harmful or injurious”. Unambiguously, the risk combines two parameters: the consequences caused by an adverse event and the probability of occurrence of this event. Therefore, “the risk is a concept that can be interpreted in different ways by different people” [3]. However, in the field of natural disasters closely related to floods, the risk is defined as “the hazard equivalent to the disorderly element, the issue that corresponds to the threatened element and the vulnerability to quantity the severity of the accident” [7]. As it is possible to remark, there is only a risk whether the hazard happens in an area with critical issues. For example, the flood risk only exists if the overflown water threatens people, their properties and economic activities. The Fig. 4.1 next illustrates that overlaying concept. Generally, the major risk is characterized by many victims, a significant cost of damage, and impacts on the environment. The vulnerability, thus, measures its consequences. The characterization of the concept of vulnerability is though more difficult. Yet, according to the Larousse Dictionary, the etymology of the word vulnerability refers to the notion of “being weak and therefore subject to be attacked by an outside agent”. Thus, in the context of natural hazards, this concept could be translated to the presence of issues in a hazard area. The concept varies according to the study area. Social vulnerability “Depends on the preparation of a society to handle with the crisis as well as its behavior during the event”. It also explains that a weak hazard can provoke dramatic consequences and that hazards of equal intensity could cause unequal magnitude damage in two different groups of people” [10]. Silva [8] defines vulnerability as “the degree of resistance or susceptibility of a socioeconomic system in relation to the impact of natural hazards and technological

42

E. de Carvalho Filho et al.

Fig. 4.1 Risk overlaying schema [7]

or environmental disasters”. It is determined as the degree of risk awareness, behavior, and health. For people, this concept can be enlarged to the ability to seek shelter, living conditions, infrastructure and public policies related to disaster management. The level of risk exposure makes a person more or less vulnerable. The vulnerability degree also varies with the characteristics of flood (flash flood or river flood, wide or localized, light or serious). For [2], there is a tied correlation between the vulnerability and the degree of resilience of a city. They explained that the more resilient is an entity (cities, territories, companies, people…) less vulnerable it would be. According to [1], the resilience is the ability of a system to anticipate and adapt to change in order to maintain control and security. Furthermore, for a company, it is the ability to restart the activity that was usual before an event, or a building to return to its usual role. For a person it’s the ease of getting his or her health back. A resilient city is where natural disasters are minimized, and infrastructure services are organized in compliance complying with standards and safety codes. Additionally, [1] believe that the city must also to have a public administrator to ensure sustainable urbanization and where there is the participation of people in decisions and planning supporting the reduction of natural disasters. To put it in a nutshell, in order to reduce the vulnerability, it is imperative to make systems more prepared, stronger and better equipped to face an extreme event, increasing its preparedness to the phenomenon and therefore its rapidity to re-establish its pre-crisis working.

4 Use of Anywhere Products to Assess Risky Events …

43

Fig. 4.2 French departments mainly touched by meteorological events October 2018

4.3 October 2018 Events in Southern France October 2018 was marked by several severe weather events in the French Mediterranean basin. Four of them were chosen to perform a Return of Experience (Rex): • • • •

Floods in Corse du Sud department (colored in green on the Fig. 4.2) Floods in the Var department (colored in blue on the Fig. 4.2) Floods in Aude department (colored in red on the Fig. 4.2) and Floods in Haute Corse department (colored in yellow on the Fig. 4.2).

4.3.1 Floods in Corse Du Sud Department (October 10th ) This episode prompted the passage of several departments on the south and southeast of France in orange vigilance for “Rain and flood” and “Storms” from Météo-France. The first vigilances were triggered on October 7th at 04 pm and continued on these areas bordering the Mediterranean until the end of the day of October 11th . This event was characterized by an active low-pressure system that crossed the country fairly from north to south on the weekend of October 6th to 7th before coming to position over Catalonia (Spain). This system caused a rainy-stormy rise at the beginning of the week on the Southeastern coast before moving towards France, then evacuating towards Italy during the day of October 11th . An intense rainy-stormy episode remained stuck on the sector of Porto Vecchio on October 10th by noon. Runoffs and floods were observed, notably on the Route of Muratello, and in several neighborhoods of Porto Vecchio, a kindergarten had to be evacuated and many homes were deprived of electricity. Other weaker thunderstorms have concerned the east coast of the island in the afternoon without associated consequences. The total accumulation over 24 h surpasses 150 mm in urbanized areas as it may be seen on the Fig. 4.3.

44

E. de Carvalho Filho et al.

Fig. 4.3 24 h accumulation rainfall over Porto Vecchio from October 10th 11 am to October 11th 11 am (Source PREDICT Observer)

During this event, 2 (two) products from ANYWHERE catalogue were used: OPERA’s 1-h rainfall accumulation and Flash flood hazard and impact assessment algorithm (FF-EWS). This last one, showed by noon an important flash flood reaction on the first hours of the afternoon, when event has intensified, and the first impacts were observed. Despite a good timing, the products presented equally impacts on flash floods over the entire southeastern coast of the island. As stated earlier, only Porto Vecchio and surrounding has some flash floods impacts. The discrepancy between model and reality relies on the fact that Porto Vecchio and surrounding areas have more issues exposed to floods than the north of Corse du Sud department, besides Porto Vecchio area presents a flatter relief with rivers and streams capable of overflowing into the plain, as it can be seen on the Fig. 4.4, that compares the issues, vulnerabilities and reliefs of Porto Vecchio and Conca and Zonza areas. Furthermore, north of Corse du Sud still keeps its dense vegetation, while Porto Vecchio is urbanized with considerably reduced infiltration areas.

4 Use of Anywhere Products to Assess Risky Events …

45

4.3.2 Floods in the Var Department (October 10th and 11th ) In a matter of fact, the same event triggered both the Floods in southern Corsica previously presented and the floods on the Var department. The region was firstly concerned by a succession of stormy rains since October 7th until October 11th . The coastline of the Var was the most touched that have produced a significant rainfall accumulation, particularly near the Gulf of Saint-Tropez as well as on the Toulon region. The Fig. 4.5 below shows the cumulative rainfall recorded over the entire region during the period. It can be noticed that during the episode, a large part of the region records accumulations greater than 100 mm while the south of the Var was the most affected with accumulations greater than 200 mm, and even locally greater than 300 mm. These large accumulations led to locally important and damaging floods on the east Var watercourses: Preconil, Garonnette, Bouillonnet, Argens, and Giscle. Other rivers have experienced smaller floods such as Aille, Endre or Mole. A first stormy passage concerned this sector on October 7th by the evening, particularly touching the Var coastline and Alpes-Maritimes. The municipalities of Cannes, Mandelieu Napoule and the Sainte-Maxime area are the most concerned with locally important runoff phenomena. Rainy-storm activity continued in this area during the following day and the first reactions of streams were observed without reaching the first overflows stages. These rains saturated the soils that were once again affected by a new rainy-storm front on October 10th . On the next day, heavy rainy storms come back from the Mediterranean and gradually took all the Mediterranean departments including the Var in the late afternoon. These rains persisted until the night on these same sectors especially on the east and

Fig. 4.4 A posteriori gravity impact chart of Corsica (left) and the comparison of Conca and Zonza region (upright) and Porto Vecchio region (downright) (Source PREDICT Analyzer)

46

E. de Carvalho Filho et al.

south of Var (sector of the gulf of Saint-Tropez). The cumulated rainfall observed led to floods locally important and flood particularly damaging the area of Sainte Maxime with a flood of Preconil, Argens and the Garonnette, as it can be visualized on the Fig. 4.6 below. During this event, the same 2 (two) products from ANYWHERE catalogue were used (OPERA’s 1-h rainfall accumulation and Flash flood hazard and impact assessment algorithm) besides it was used past data on OPERA’s 24-hour rainfall accumulation.

Fig. 4.5 120 h accumulation rainfall over Provence-Alpes-Côte d’Azur region from October 7th 04 am to October 12th 04 am (Source PREDICT Observer)

Fig. 4.6 Location and reaction Argens and Préconil rivers during the event. The watercourse in purple is the Garonnette that does not count with a gauging station (Sources Vigicrues and CENEAU)

4 Use of Anywhere Products to Assess Risky Events …

47

On both passages, the models showed a reaction of the watercourses on the region of Sainte Maxime, but not that important as the floods observed on La Garonnette river, which was the main responsible for the damages on the area during the period. In order to understand the important river flood on La Garonnette, the rainfall 2 days before should be taken into consideration. Between October 7th and the 8th , around 120 mm were responsible for the soil’s saturation. The following 200 mm, fallen between October 10th and 11th , triggered the flood. In order to better meet the expectations on this precise event, the river warning product should take a larger period to calculate the saturation of soils or integrate in situ data.

4.3.3 Floods in Aude Department (October 14th to 16th ) This event occurred following the cold air that crossed the Iberian Peninsula from the North Atlantic. Its movement was fairly eastwards over Spain and then the Mediterranean (influenced by the remains of ex-Hurricane Leslie having affected the Portugal 24 h earlier), bringing a flow of South/Southeast from the Mediterranean. This humid flow has generated rainy-stormy climbs, particularly on the Aude. This phenomenon

Fig. 4.7 36 h accumulation rainfall over western Mediterranean arc from October 14th 12 am to October 16th 00 am (Source PREDICT Observer)

48

E. de Carvalho Filho et al.

has induced a large accumulation in a short time generating major floods in the department. The precipitations of October 14th and 15th arrived after a rainy antecedent occurred the previous week. Slight floods and runoff were punctually observed. This new episode took the form of stationary rain-storm lines going up from the Mediterranean in the night from October 14th to 15th in generating strong accumulations, followed by precipitation of weaker intensities in the morning of October 15th . In parallel, the strong sea (swells related to South-Southeast flow) generates marine submersion phenomena and a significant downstream constraint to the flow of river. Essentially, the regions upstream the watersheds of the Aude, Orbieu and their tributaries as well as the reliefs of the Black Mountain who recorded the highest accumulations (sometimes greater than 200 mm, see 300 mm over the period), as it can be seen on the Fig. 4.7 next.

Fig. 4.8 Output of the Flash flood hazard and impact assessment algorithm northwestern of Carcassonne, on the ridge (dash red line) between the basins of Tarn and Aude river (blue line) at October 15th 04 am UTC or 06 am France

4 Use of Anywhere Products to Assess Risky Events …

49

Similiter, this event was followed using the same 2 (two) products from ANYWHERE catalogue: OPERA’s 1-h rainfall accumulation and Flash flood hazard and impact assessment algorithm. However, the timing on this peculiar event was not satisfactory. The main rivers have presented warnings by 02 am, around the time PREDICT had already raised the Safety Level to the highest. Just for information, awareness messages related to watercourses reaction was send by PREDICT on-call engineers between 11 pm and midnight, 2 to 3 h prior to the first signs of floods on the models, showing once more the necessity of a trained eye behind the screen to ensure the appropriate emergency response level. Nevertheless, the river warning product brought the attention to a few small watercourses reacting, on the Tarn river basin, on the other side of the ridgeline, as it can be seen on the Fig. 4.8. Although the scale of the event was difficult to foreseen, the potential damage could be identified a night before and surveyed throughout the event. Despite heavy impacts, the action of the crisis management actors who have worked to safeguard the territories and the inhabitants has made possible to ensure many people and avoided losses. An idea of the impacts of this event could be seen on the Fig. 4.9, next.

Fig. 4.9 Flooded zone by Aude River and some of its consequence on Raissac d’Aude and Puichéric (Source PREDICT Analyzer)

50

E. de Carvalho Filho et al.

4.3.4 Floods in Haute Corse Department (October 16th and 17th ) Contrarily to the previous presented events, this last one is more local and triggered by locally heavy rains arising from the south by October 16th at the morning and continued until the morning of the following day. The more intense rainy-stormy passage was observed over the region between Penta de Casina and San Nicolao, by the afternoon of October 16th when a lined rainy band remained stuck. Runoffs and river floods were observed, notably alongside the Fiume Alto River and Alesani River. Other weaker thunderstorms have concerned the east coast of the department on the following hours but without any associated consequences. The total accumulation over 24 h surpasses 250 mm on the heights, a sparse populated area, it can be seen on the Fig. 4.10.

Fig. 4.10 24 h accumulation rainfall over eastern Corsica from October 16th 00 am to October 17th 00 am (Source: PREDICT Observer)

4 Use of Anywhere Products to Assess Risky Events …

51

Fig. 4.11 Output of the Flash flood hazard and impact assessment algorithm eastern Corsica at October 16th 03 pm UTC or 04 pm France and the reaction of Fiume Alto River with indication of 10 years and 50 years return period flood peaks (Source Vigicrues)

During this event, only the Flash flood hazard and impact assessment algorithm (FF-EWS) was used. This product showed by 15 am UTC (16 am local time) a strong reaction of watercourses on the sector of Penta de Casina. Which it was corroborate by the reaction of the levels on Fiume Alto River that counts with an online data gauging station. Fiume Alto, Alesani River and other minor rivers presented orange levels, meaning that the 50 years return period flood should be exceeded. Providentially, the gauging station at Fiume Alto River possessed a good time series and provides a reliable statistics on flood peaks, thus the surpassing of 50 years stage could be confirmed and it is presented on the Fig. 4.11 next. As it can be recognized, the product presented a pertinent timing and a coherent impact forecast for this event. Despite the exhibited on the Fig. 4.11, only the Fiume Alto River poses a concern for the Civil Protection during this event, therefore the impact foreseen for the other watercourses were overestimated. The discrepancy between model and reality relies on the fact that the product has not taken into consideration the presence of the relatively recent Alesani dam, upstream the plain, whose lac of about 49 ha is enough to cushion the peak flood. Furthermore, Fiume Alto river basin also presents a little more issues exposed to floods than the other rivers in the region.

4.4 Conclusions Regarding the 4 (four) meteorological events followed by PREDICT using the products arising from the ANYWHERE catalogue, it can be concluded that the products presented a good reaction time regarding the foreseen impact. These products can afore a precious time in crisis management and anticipation. And even though the impact forecast presented a few discrepancies due to validity period, they certainly could help risk specialists to draw up quickly the impacted zones and the type of expected impact.

52

E. de Carvalho Filho et al.

During these events and the following ones, PREDICT could help its users: crisis management actors to safeguard the territories and inhabitants ensuring the safety of many people thus avoiding heavy losses. This is mainly based on its on-call engineers that have constantly analyzed the outputs of models and algorithms and passed concise information via a comprehensive communication channel integrated on the platform, doubled with e-mail, SMS and, in some cases, phone calls with the main responsible for safety. The constant presence of risk specialists to analyze data and the output of the models and algorithms showed once more vital to the good understanding of hydrometeorological events, their chronology, and the safety level they required in order to save lives, goods and infrastructures. It is more and more evident that companies in charge of early warning system activation and data analysis such as PREDICT are the vector to ensure not only the improvement of resilience but also the diminishment of weather-related impacts, which already cost a lot to society. Anticipation and communication are vital to adapt the behavior in advance, aiding Civil Protections, mayors, industrials and citizens to reduce the possible hydrometeorological impact and to increase resilience. By providing users with easy access to multi-hazard information such as the implementation of ANYWHERE data into PREDICT platforms will allow First Responders, safety teams and individuals to better anticipate hazards and improve their crisis management.

References 1. de Amorim MF, Quelhas OLG, da Motta ALTS (2014) A resiliência das cidades frente a chuvas torrenciais: estudo de caso do plano de contingência da cidade do Rio de Janeiro. Sociedade & Natureza 26(3):519–534 2. Balica SF, Douben N, Wright NG (2009) Flood vulnerability indices at varying spatial scales. Water Sci Technol 60(10):2571 3. Cameron IT, Raman R (2005) Process systems risk management. 1st ed. Elsevier, San Diego, 615 p. Process systems engineering 4. Ceneau (2019) Ceneau, March 2019 https://ceneau.com/ 5. Guha-Sapir D, Santos I, Borde A (ed) (2013) The economic impacts of natural disasters. Oxford University Press, Oxford, New York 326 p 6. Ministère de la Transition écologique et solidaire (2019). VIGICRUES (March 2018). https:// www.vigicrues.gouv.fr/ 7. Pageon JL (2008) Méthodologie d’évaluation de la vulnérabilité d’une MRC face aux ressources essentielles. École Polytechnique 8. Silva EA (2012) Bacia da Baía de Guanabara: características geoambientais, formação e ecossistemas. Editora Interciência, Rio de Janeiro. 405 p 9. Techniques de l’ingénieur (2013) Prévention du risque inondation. - Techniques de l’Ingénieur at https://www.techniques-ingenieur.fr/base-documentaire/environnement-securite-th5/risque s-naturels-etimpacts-industriels-42828210/prevention-du-risque-inondation-tba251/ 10. Veyret Y, Reghezza M (2005) Aléas et risques dans l’analyse géographique. In: Annales des mines, vol 40, pp 61–69

Chapter 5

Decision-Making Support System for Crisis Operations and Logistics Aspects in Extreme Weather-Induced Events Ivan Tesfai, Giovanni Napoli, Salvatore Ferraro, Andrea Poggioli, and Marta Speranza Abstract Extreme weather and climate events are causing several hazards affecting our society, and when interacting with exposed and vulnerable human and natural systems they can lead to disasters. Critical infrastructures such as transportation (roads, rails, etc.) can be affected and their capacities collapsed in critical points. In these situations, logistics and distribution tasks are considered in many EU countries among the priorities services both for commercial entities and for emergency operators. Early-warning measures and decision-making tools should enable to improve protection measures and, in case of catastrophic situations, improving the coordination of rescue operations. In the framework of ANYWHERE project, the aim of the current work was to design and test software solutions for decision-making able to support the emergency operators and transportation stakeholders affected by weather disruptive events, improving preparedness and planning capabilities. A platform was developed to support the implementation of actions related to emergency plan effectively presented with step-by-step operative checklists without affecting nor changing procedures and legacy systems in place. Furthermore, it also implements weather forecast and impact estimation on the local road network of interest through a dedicated simulation service for transport modeling.

I. Tesfai (B) · G. Napoli · S. Ferraro · A. Poggioli · M. Speranza RINA, Genoa, Italy e-mail: [email protected] G. Napoli e-mail: [email protected] S. Ferraro e-mail: [email protected] A. Poggioli e-mail: [email protected] M. Speranza e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_5

53

54

I. Tesfai et al.

The outcome are indications useful for emergency operators to have a wider awareness of the current situations, understand the weather impact on traffic and viability but also for logistics-related companies (i.e. food distribution) and transportation stakeholders enabling the identification of best route (i.e. alternative road, multimodal path) between two locations in such weather disruptive situations. Keywords Decision support systems · Emergency procedures · Crisis management · Weather impact · Logistics · Transportation

5.1 Introduction Day by day extreme weather and climate events are causing several hazards affecting our society, and when interacting with exposed and vulnerable human and natural systems they can lead to disasters. Critical infrastructures such as transportation (roads, rails, etc.) can be affected and their capacities collapsed in critical points. According to the Global Risks Report 2018 [1], among technological, environmental, geopolitical, economic risks, the highest ones in terms of likelihood are: extreme weather events, natural disasters, cyberattacks, and failure of climate—change mitigation and adaptation. Looking at natural disasters, in the period 1998–2017, disaster-hit countries reported direct economic losses of US$2,908 billion of which climate-related disasters accounted for US$2,245 billion or 77% of the total [2]. In 2017 only, losses from natural disasters worldwide [3] were close to US$200 billion worldwide, caused by around 700 events. Around 20% of these losses were re-insured. In EU alone, considering natural disasters since 1980, EU Member States [4] have lost over e360 billion in weather and climate extreme events (e.g. flash floods and storms, forest fires and earthquakes). In 2016, the economic costs are close to e10 billion in damages. According to various scenarios, these numbers are likely to increase in the future, due to the effects of climate changes, causing severe social and economic impacts to our infrastructure, communities, society. As example it is estimated that at EU level damages to transport infrastructure due to extreme precipitation induced by climate change could increase by 50% by 2040 to reach around e930 million/year. This would require e590 million of additional annual spending compared to a scenario without climate change [5]. Since financial allocations for maintenance of transport infrastructure can be volatile [6] depending on the financial/political context and overall operational needs, it can be argued that without a resilient approach and proper management strategies, we would not be able to tackle this challenge properly. Early-warning measures and decision-making tools are solutions following this need to enable the improvement of protection measures and, in case of catastrophic situations, the coordination of emergency management operations. Furthermore, tasks related to logistics and distribution are considered in many EU countries among the priorities services both for commercial entities and for emergency operators.

5 Decision-Making Support System for Crisis …

55

In the framework of ANYWHERE project, the aim of the current work was to design and test software solutions for decision-making able to support the emergency operators and transportation stakeholders affected by weather disruptive events, improving preparedness and planning capabilities.

5.2 Problem Statement During snow storm or flash flood as severe weather events, the roads in an urban or metropolitan area can be directly affected with a collapse of their capacity (as reported in Fig. 5.1). Due to the occurrence of these conditions, the traffic of vehicles can be suffered of a performances decrease (i.e. advance of vehicles queues with the increasing of travel time and with the speed reduction), with a high impact on public and private transport services (e.g. bus operation, goods distribution) or on eventual emergency and rescue services (as reported in Fig. 5.2). Emergency management involves the implementation of actions immediately prior to, during, and immediately after an emergency/disaster in order to ensure that the effects are minimized. Any decision is taken on an evaluation of status and effects, better the comprehension, more effective the actions. In this context, decisional capabilities can be enhanced by the application of a transport simulation model as support of the decision-making during preparation and planning phase. In order to limit the impact of weather-induced events, the purpose of the current research activity foresees the provision of a system in support to in place emergency procedures enriching the situation awareness and planning capabilities on

Fig. 5.1 Examples of reduced capacity of roads caused by Snow Storm

Fig. 5.2 Examples of effects on public and private transport services

56

I. Tesfai et al.

transportation-related tasks (e.g. identification of alternative itineraries, logistic and operation re-planning, reorganization of vehicles fleet, etc.). Such system aims driving the user in the situation comprehension and decision taken via a questionnaire, preloaded rules and traffic forecast model for logistics support. This is based on the fact that weather induced emergencies have usually a major impact on logistics. The foremost objective is to support the Emergency Operator in coordinating and operating logistics efforts to delivery assistance to emergency response team also by using the latest weather forecast made available by the MH-EWS (Multi-Hazard Early Warning System), developed in the framework of the ANYWHERE project. In particular, the major requirements identified by emergency operators1 could be summarized as follow: • Pre-conditions: • need to implement the procedures already in place in the emergency centre; • prior every user must receive a short training before the use of any support system; • The main characteristics needed for the support system itself are: • web based and not platform and technology dependent for users; • able to integrate different operators assignment (several console can work on the same data and easy to scale); • easy to update and configure; • accessible from different terminals in terms of both location (different placement) and technology; • able to track the decisions and actions path for post-evaluation and assessment. Based on weather forecasts information, in compliance to the logistics user needs (e.g. Origin/Destination) and the road network, the system should be able to simulate and evaluate different transport scenarios and logistic configurations, analysis of the demand assignment and likely the modal diversion, considering a multi and intermodal supply. In order to perform such evaluation, the impact of weather on transportation network need to be assessed and considered within the definition of transportation scenarios and road network requested by the operator. In the following chapter a literature review has been performed with the objective to identify how weather conditions affects mobility and viability in terms of capacity reduction of traffic flow.

1 Dedicated

meetings with emergency operators have been conducted in the framework of ANYWHERE project where requirements have been collected and analysed to build realistic use cases and define the solution design.

5 Decision-Making Support System for Crisis …

57

5.2.1 Weather Impacts According to [7], weather events such as rain, snow, sleet, fog, high winds, and flooding reduce roadway capacity. These events can cause slick pavement, lower traffic speeds, increase speed variability, affect traffic volume, increase delay, escalate crash risk, disrupt access to roads (e.g. lane obstruction, pavement buckling) and damage road infrastructure (e.g. traffic control devices). As congestion increases in urban areas, weather events will have even greater effects on arterial operations [7], on arterial routes, adverse weather can have an impact on the effectiveness of traffic signal timing plans designed for normal conditions. Signal timing plan parameters used in clear, dry conditions may not be optimal during adverse weather. All forms of precipitation can impact arterial traffic operations. Light rainfall or snowfall after extended periods with no precipitation can cause slippery pavement when water mixes with oil, prompting drivers to lower speeds. In addition to reducing vehicle traction, heavy rain or snow can reduce visibility distance and cause drivers to increase headway, decrease acceleration rates, and further reduce speeds. When fog, drifting snow, or blowing dust reduce visibility some drivers reduce speed more than others, increasing speed variance. On the table below the impacts of various weather events on roadways, and traffic flow are reported (Table 5.1). Capacity reductions can be caused by lane submersion due to flooding and by lane obstruction due to snow accumulation and wind-blown debris. Road closures and access restrictions due to hazardous conditions (e.g., large trucks in high winds) also decrease roadway capacity. Table 5.1 Weather impacts on roads and traffic [7] Road weather variables

Roadway impacts

Traffic flow impacts

Precipitation (type, rate, start/end times)

• Visibility distance • Pavement friction • Lane obstruction

• • • •

Roadway capacity Traffic speed Travel time delay Crash risk

Pavement condition (dry, wet, snowy, icy, temperature)

• Pavement friction • Infrastructure damage

• • • •

Roadway capacity Traffic speed Travel time delay Crash risk

Fog

• Visibility distance

• • • •

Traffic speed Speed variance Travel time delay Crash risk

Wind speed

• Visibility distance (due to blowing snow, dust) • Lane obstruction (due to wind-blown snow, debris)

• Traffic speed • Travel time delay • Crash risk

Extreme temperatures & lightning

• Infrastructure damage

• N/A

58

I. Tesfai et al.

Table 5.2 Freeway traffic flow reductions due to weather Weather conditions

Freeway traffic flow reductions Average speed %

Free-flow speed %

Volume %

Capacity %

Light Rain/Snow

3–13

2–13

5–10

4–11

Heavy Rain

3–16

6–17

14

10–30

Heavy Snow

5–40

5–64

30–44

12–27

Low Visibility

10–12

12

In accord to [8], on signalized arterial routes, speed reductions can range from 10– 25% on wet pavement and from 30–40% with snowy or slushy pavement. Average arterial traffic volumes can decrease by 15–30% depending on road weather conditions and time of day. Saturation flow rate reductions can range from 2–21%. Travel time delay on arterials can increase by 11–50% and start-up delay can increase by 5–50% depending on severity of the weather event. Differently on freeways [9], the traffic flow reductions due to different weather conditions can be summarized as indicated in the following table (Table 5.2).

5.3 Solution Identification and Development Based on the requirements elicited by emergency operators and transportation stakeholders, and the problem stated, the solution was identified in the development of a web-platform able to support the emergency operators in the implementation of actions listed in their owned emergency procedures-in-use that are presented systematically. The actions presented are on the base of current and actual situation on site, effectively shown with step-by-step operative checklists without affecting nor changing procedures and legacy systems in place. In order to increase awareness on scenarios affected by weather disruptive events, the platform also implements the latest weather forecast and impact estimation on the local road network of interest through a dedicated simulation service for transport modeling named MTCP2 —Macroscale Transport Chain Planner. By crosscorrelating weather forecasts (level of snow, precipitation, etc.) provided by MHEWS and a representative model of the road network, the simulation service is able to estimate the extension of the affected area and in particular of the transport network capacity. The overall architecture of such decision support system is showed in the following picture (Fig. 5.3). The scenario assessment is performed by answering to questions and performing actions while the MTCP interacts with the MH-EWS remote services to retrieve

2 Solution

developed and owned by RINA Consulting S.p.A.

5 Decision-Making Support System for Crisis …

59

MH-EWS DATA BASE

List of informaƟon to acquire

Gateway Assessment of the scenario MTCP DATA BASE

Status Update

List of acƟon to implement

AcƟons to implement in the current status MTCP

Fig. 5.3 Decision support system conceptual model

information about the weather conditions to build an accurate evaluation of the future traffic fluxes. In such way, the platform aims driving the user in the situation comprehension and decision taken via quick and direct questionnaire, pre-loaded rules and action prioritization. Moreover, it integrates a weather-based traffic simulation engine by using the transport network capacity as self-preparedness enabler, since improving visibility of transport network capacity means better planning, training, etc., guaranteeing flexibility over multiple users and providing operational benefits.

5.3.1 Management of Actions and Emergency Procedures According to the user requirements, a web-platform for decision support need to guide the emergency coordination staff in the execution of the actions listed in the emergency procedures that are encapsulated in a software tool and presented in a step-by-step checklist configuration. The tool foresees two main profiles, one dedicated to administrative activities— used in the configuration phase and supervision of the activities—and a second dedicated to operations activities—both used by the emergency staff in the emergency centre to coordinate the activities on field. The administrative profile allows the administrator in: • Creation/modification of users and configuration of permission; • Creation/modification of the scenarios;

60

I. Tesfai et al.

Fig. 5.4 Decision support system access page (administrators)

• Implementation of the question and action composing the checklist (by using “tree based” approach); • Monitoring of the checklist progress; • Access the Log file automatically created during software use for post processing activities, such as operation assessment (Fig. 5.4). The operation profile allows the user in the control center: • To proceed with the checklist: questions and actions are presented on the same page for an easy interaction. The questions are characterized by a text and a set of predefined possible answers. The answer has trigger that activates the proper set of actions according to preloaded procedures. The action brings to new questions as well; • To cooperate with others operators by delegating some assessment and actions to others (and receiving from others); • To communicate with others with different channels, namely: internal messaging system (provided by the application server), by email (via client mail), by Skype (call through the application); • To browse the local transport and traffic map updated according to the weather forecast. By setting the link between the answers and the action, the user can implement the procedure into the software. To help the control of the links among question and action, a tree presentation of the link has been developed as shown in Fig. 5.5 (bottom). To create a new link it is enough to drag a node on another one operating directly from the tree representation shown. Each node in the tree represents an action to perform or a question to answer to, while the colour of the lines connecting nodes

5 Decision-Making Support System for Crisis …

61

Fig. 5.5 Question & Action page (top—checklist view, down—tree-based view)

indicates what actions or questions will be further available on the basis of the answer given or the status applied to the precedent node they are attached to.

5.3.2 Traffic Simulation of Weather-Based Transportation Scenarios By the platform, it is possible to access to a page named: “Local Transport and Traffic Map”. This page embeds an OpenStreetMap on which a layer representing the MTCP traffic forecasts is drawn. As mentioned on Sect. 5.3, MTCP responds to the requests of information made by the user about the weather impacts on roadways, traffic flow and operations, providing traffic scenarios through by a local area network via GEOJson formatted data. Such scenarios are held into MTCP database and involved in the logistic decisions to achieve. In the following a detail of the process workflow within the decision support system related to transportation simulation (Fig. 5.6). From this page the users can perform the following actions: • zoom in/out; • retrieve from MTCP a new forecast within [1..48] hours; • delete the currently drawn layer;

62

I. Tesfai et al.

Fig. 5.6 Process workflow for weather-based transportation scenario simulation

• centre the map on a particular street; • set manually the capacity of the arc. In the following picture, a specific weather-based transportation scenario is depicted with nodes and arc represented. The colour of the lines connecting the arcs depicts a measure of the “fluxes of traffic intensity” and can assume “values” from green to black: a black coloured arc represents a traffic flux very busy and, so, likely unavailable for people. The opposite for a green arc. By left clicking on a particular arc, a pop-up dialog box will appear showing some information about the selected arc and giving the user the possibility to change the “capacity” of that arc in order to force the MTCP to consider this new value while building the next traffic forecast (Fig. 5.7).

5.4 Conclusions The current work presented a decision support platform built on the requirements of emergency management operators in the framework of ANYWHERE project. The platform is developed for supporting decision making to operators and managers of a crisis management room for preparedness and operations purposes, providing also an enhanced awareness of weather impact on transport network. It is developed to be technically easy to operate for users being web-based and not platform and technology dependent. It is able to integrate multiple and different operators assignment, to

5 Decision-Making Support System for Crisis …

63

Fig. 5.7 Transportation scenario induced by rain conditions and depicting fluxes of traffic intensity

track decisions and actions taken, to provide emergency functionalities such as quick calls and messaging and to display transport scenarios, for freight and passenger, in urban and national context induced by weather forecast. Such decision support system has been implemented for trials and feedback purposes in the municipality of Genoa/Department of Civil Protection, to be used in the emergency operation room supporting alert management. According to the problem stated, the Decision Support System has been implemented in the Operation Room to support the Coordinator in the supervision of the operations. In particular, the platform has been configured with the procedures for “Yellow alert” management and it has been made available to the operators of the municipality via access to a remote server. The choice to test the solution to alert management is based on the following considerations: • possibility to formalize the activities and so delegate them among different operators, thus, it requires coordination; • place requiring official actions. Any lack in the official actions could bring to critical impacts; • decision making carried out in a chaotic environment because it coincides with the starting of several processes and “replacement” of personnel. The preliminary evaluation on real environment showed that the Decision Support System developed performs all the expected functionalities and its flexibility in implementing potentially any kind of procedure is promising. Moreover, being web based, it is easy to introduce in the operational room. The municipalities and emergency managers, that drawn the software development since the project beginning, are now asked to test the systems in their risk management processes. The next activities will mainly deal with tuning of the

64

I. Tesfai et al.

configuration and technical improvement associated to experience in the operative environment. Acknowledgements The research leading to these results has received funding from the European Union’s H2020 Programme under the topic of potential of current and new measures and technologies to respond to extreme weather and climate events under grant agreement no. 700099. Acknowledge the support of Municipality of Genoa/Department of Civil Protection for the operational feedback and valuable experience kindly provided.

References 1. 2. 3. 4. 5. 6.

reports.weforum.org/global-risks-2018/ www.unisdr.org/we/inform/publications/61119 www.munichre.com/en/reinsurance/business/non-life/natcatservice/index.html https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52017DC0773 peseta.jrc.ec.europa.eu/transport.html www.cece.eu/news/eu-news-eu-infrastructure-new-publication-by-european-commissionraisesquestions-on-maintenance-funding 7. Goodwin L, Pisano P Weather-responsive traffic signal control. https://ops.fhwa.dot.gov/ weather/resources/publications/fhwa/ite04sprwxrespsigcon.doc 8. Goodwin L (2002) “Weather Impacts on Arterial Traffic Flow”, prepared for the FHWA road weather management program. http://www.ops.fhwa.dot.gov/weather/best_practices/ ArterialImpactPaper.pdf 9. Agarwal M, Maze TH, Souleyrette R (2005) Impact of weather on urban freeway traffic flow characteristics and facility capacity, Mid-Continent Transportation Research Symposium, Iowa State University

Chapter 6

Operational Resilience Index Computation Tool as a Decision Support System Integrated in Eu Risks Management Platforms—Test on Biguglia Catchment, a Mediterranean Intense Precipitations Regime Prone Area Morgan Abily, Philippe Gourbesville, Hézouwé Amaou Tallé, Marc Gaetano, Jelena Batica, Patric Botey, and Marien Setti Abstract A Decision Support tool, relying on resilience assessment approach has been developed to reinforce decision makers for climate related emergency management operations. The method proposes the calculation of an Operational Resilience Index (ORI) at city building/block scale. The ORI tool is developed based on a previous innovative research line, which was allowing computing flood resilience index for mid-term urban planning, adapted here to specificities of emergency management operations. The adaptations in the method for ORI tool, allows to use the existing concept of specific resilience, where the anthropized system is characterized in categories of urban functions and urban services. Dependencies between urban functions and services according to the impact of a given hazard is directly parameterized through a web-based GUI. The adapted new method takes into consideration for the ORI computation: multi-hazard, integration of critical infrastructures as well as social events occurrences in the adjustment parameters of the ORI computation. The requirements for the tool uses are: (i) a training of the end-user and (ii) a specific formatting for the input data. The ORI tool has been implemented and tested by end-users in the framework of a H2020 Research and Innovation project- ANYWHERE- in webservice based Pan-European multi hazard platforms DSS. Tests performed presented here, are those related to flood hazard initiated by Mediterranean intense rainfall events, which were performed on the 182 km2 Biguglia catchment in Corsica island M. Abily · P. Gourbesville (B) · H. A. Tallé · M. Gaetano · J. Batica Polytech Lab, Polytech Nice-Sophia, Université Côte d’Azur, 930 route des colles, 06903 Sophia Antipolis, France e-mail: [email protected] P. Botey · M. Setti Service d’Incendie et Secours de la Haute Corse (S.I.S.2B), Lieu-Dit Casetta, 20600 Furiani, Corse, France © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_6

65

66

M. Abily et al.

(France). A six hours long rainfall runoff event cumulating 165 mm has been simulated over a 20,000,000 cells grid using a full 2D SWE based modeling approach on a HPC structure, to generate 5 m resolution flood hazard maps, used as inputs for the ORI computation. The other inputs of the ORI method are OGC standards compliant format GIS information: building, networks (electricity, transport, communication, water supply) local critical infrastructure and social events, preformatted for the ORI computation process. This paper details respectively: ORI tool method, ORI tool interface and API development for integration within the ANYWHERE Pan-European platform and an application illustration through flood hazard map computation scenario and method. Then, a summary of the user guide is presented, explaining the needs and prerequisites for end-users to use the tool and understand the outputs. Lastly, the main advantages of this ORI computation tools and the principal limits raised by end-users (here, SIS2B fire brigade command and control center) for improvements are exposed and analyzed. Keywords Resilience · High impact climatic events · Flood hazard · Emergency management services · ANYWHERE project · H2020

6.1 Introduction In the framework of H2020 research and innovation action, the ANYWHERE project develops Global situation awareness platform/toolkit referred as A4EU (ANYWHERE for Europe). A4EU as a web-service based platform has been customized and tested by civil protection services and first responders on multiple pilot sites in Europe during the course of the project. In total, six different pilot locations have been selected for test (Catalonia—Spain, Corsica—France, Liguria—Italy, Bern Canton—Switzerland, Rogaland—Norway, and South Savo—Finland). The objective of A4EU platforms is to provide various tools for situation awareness in case of catastrophic situations, supporting the coordination of rescue operations at every phase of a crisis management cycle. For each pilot, a civil protection organization has been included from day one into the A4EU custom version development process as future end-user of the platform. Each pilot, had a specific DSS tool applied to its customized version of A4EU. Among the DSS tools, a specific tool for resilience assessment—Operational Resilience Index (ORI) assessment tool—has been developed, customized and tested for A4EU custom version for the Corsica pilot so called A4CORSICA. Objective of this DSS tool it to enable in the preparedness phase of a weather induced crisis event scenario, to help civil protection services, to anticipate the resilience of the urban system and to consequently draw up and test operational measures effectiveness on the urban system resilience in order to maximize the effects of crisis management operation. Resilience as a concept, is getting applied in different sectors related to crisis management, including climate related risk management [1, 2]. On one hand, overall

6 Operational Resilience Index Computation Tool …

67

resilience is very difficult to define. On the other hand, specified resilience [3, 4] concept is clearly defined and suitable for a more direct application: the specified resilience is defined as the resilience of what (e.g. flood, drought, etc.; in our case high impact climatic event), to what, and up to what level. For practical applications of resilience assessment such as for ORI A4EU platform context, it is decided to focus on specified resilience concept. Thus, Operational Resilience concept is classified as a specific resilience type which applies for high climatic events on urban system. Operational resilience gathers a set of different data and applies a methodology, which was first developed specifically for flood through Flood Resilience Index (FRI) computation [5]. Operational resilience allows end-users, to adapt to changing patterns. This concept applies directly to networks, organizations, supply chains, critical infrastructure and even to the population. Operational resilience computation takes into consideration planning, integrating, executing and governing that will ensure that end-users will: • identify and mitigate operational risks—eliminating disruptions before they occur; • prepare for responding to events—response is controlled (prepared in advance); • recover and restore—have an acceptable time frame. Assessment of this concept is already applied to different frameworks considering flood hazard [5, 6]. Setting the scene for operational resilience toward multi-hazard aspects brings new lights within the existing framework. The developed method and tool perform a mapping of an Operational Resilience Index (ORI) that is providing operational resilience level of an entity. Operational Resilience Assessment tool/method performs a mapping of the ORI which provides operational resilience level of an entity, at building scale. Required inputs are (i) building map, (ii) hazard maps, (iii) networks data (e.g. electricity, transportation, water, communication). A classification of the input data into proper urban function and a briefing on the method is necessary before use of the tool/method (2-3 h of training). Objective of the research line presented in this paper to present the adaptation of the FRI method to perform the ORI computation, presenting ORI methodology along with its operational implementation, resulting from tests operated on A4EU DSS for a flood hazard case in a Mediterranean catchment: Biguglia (Corsica, France). The paper presents in the next Sect. 6.2, the concept of operational resilience assessment. Section 6.3 develops its technical implementation. Section 6.4 elaborates on the practical implementation and guide line for ORI computation tool uses. Section 6.5 introduces the Biguglia catchment tests and in a final Sect. 6.6 conclusions are drown up.

68

M. Abily et al.

6.2 Concept of Operational Resilience Assessment Specified operational resilience assessment for different systems is applicable to urban systems, taking into account all physical and non-structural (e.g. planning maturity level) system components that are recognized within operations as introduced in the framework. Firstly, the urban system can be defined has having a set of function [7]. Secondly, the urban system also has the property of providing services (e.g. transport, electricity, etc.) [8], which can be seen as fluxes allowing functionalities between the functions of the system. The urban functions and services can be listed and they apply to structural level (e.g. buildings) of the system. In this approach, the urban system is consequently defined as a combination of functions and services mapped at building scale. Thus, operational resilience assessment tool provides web-based application developed for evaluation of ORI analyzing the urban area with existing elements and networks. In the approach, a given area is mapped to its components and networks (fluxes). In addition, the dependencies between different components and networks are defined and included within the methodology. The tool provides analysis of a complex urban system taking into account all its components, connections and dependencies within the defined framework. The ORI for the concerned high impact weather event is calculated at micro- (building) or meso- (district) scale and mapped over A4EU platform versions, ORI tool being included as a module where the results are displayed for decision support enhancement. To use the ORI computation approach, there are four different stages to be addressed: (i)

System analysis—the urban system with all its physical components is mapped and scaled to its attributes. (ii) Characterization of system disruption—definition of risk, risk assessment, setting the scenarios for analysis and with respect to event, integrates different scales for analysis and defines disaster parameters. (iii) Operational resilience assessment—evaluation of Operational Resilience Index for all defined scenarios. (iv) Decision-making stage—based on evaluated ORI (not described in this paper). These four stages main aspects are presented hereafter. (i) System analysis goes through the system description and the mapping phases. System Description Phase. Regarding the system description, the urban system is defined has having nine main functions listed here: 1. Housing; 2. Education; 3. Food; 4. Working; 5. Safety; 6. Governance; 7. Health; 8. Leisure; 9. Religion. Then, the methodology allows user to adapt maturity levels ranking to local context. Ranking maturity level of urban functions toward a natural hazard falls in line with following description from [9]: Level 1 = Informal: Individual action with no coordination Level 2 = Basic: Individual actions with limited coordination

6 Operational Resilience Index Computation Tool …

69

Level 3 = Initial: Limited implementation and coordination on local level Level 4 = Coordinated: Fully integration and partial implementation Level 5 = Integrated: Fully integration and implementation The main services provided by the urban system are following: 1. Energy; 2. Water supply; 3. Communication; 4. Transport. These urban services are all connected to the urban functions. In a normal situation, dependencies of urban functions toward the services can be assessed by the user of the approach. Thus the dependencies to each UF toward the different urban services can be ranked by expert/user from strong (1) to weak (5). Mapping Description Phase. The structural level at which the method is applied is at the building scale. There are 11 different types of building typology which are described in Table 6.2. This typology is defined by user for each entity representing a building through GIS use, when preparing the data. Other information (networks data such as electricity, transportation, water, communication) are mapped as vector during this step either directly inside the building attribute table or through upload of other GIS vector layers as theoretically represented in Fig. 6.1. (ii) Characterization of system disruption; This aspect consists in the definition of risk, risk assessment, setting the scenarios for analysis with respect to a given type of event. This integrates definition of disaster parameters. The assessment of the risk, e.g. for the flood risk assessment, is done by:

Fig. 6.1 Sketch example of vectorial information used for urban functions, services and flood map to be uploaded for ORI computation

70

M. Abily et al.

• Assessing different thresholds of flood levels for a given flood hazard. Scenario, are selected by the user and will be introduced as flood maps: it can be using flood map of, historical data, legal document (e.g. PPRI maps), or results of simulated scenarios. • Allowing user to parameterize/weight every urban function dependency to any urban services. This is a subjective, but expert based system, to define the vulnerability of an urban function according to its maturity level and toward its urban services dependencies and availability effects. For instance the function ‘education’ might be weighted as of prime importance (e.g. during week-day), with a high resilience maturity level due to existence of an action plan in schools in case of flood incident, but weighted as less important if event’s scenario occurs during school closing period (night, week end, vacation). Similarly, the inter dependencies of the urban services due to cascading effects can be weighted. As an example here for the UF ‘education’ again, depending on the type of considered flood event (flash flood or long term flood event) if education facility is used as a shelter, the service water supply might be weighted differently depending on the situation (short or long lasting event). (iii) ORI computation The computation of ORI is done through the evaluation process described in [5] where reader can find the details.

6.3 Technical Implementation The architecture for ORI integration into the A4EU (A4CORSICA custom platform version) is described in Fig. 6.2. The user can access in a secure way to ORI computation tool from A4CORSICA. Two ways to process to the computation are available for user: • It can be all through the A4CORSICA interface (customized by Predict services). This will allow loading through the interface the Building and Hazard layers and to compute and display the ORI results. However this computation process requires that the user has already established, e.g. during the tool training period, the parameterization for the computation. • A second way to process to the computation is by clicking on the advanced computation button on A4CORSICA interface. This will open up a new interface through a web browser, allowing user to fine tune the ORI computation as described in the next section, notably by updating networks information, tuning the parameterization, and allowing to load the critical infrastructures, including them into the resilience computation process. In that case the results are displayed on the web browser interface and the advanced settings function allow as well to download the ORI computation results under shape file format to see results in GIS or Web GIS tools.

6 Operational Resilience Index Computation Tool …

71

Fig. 6.2 Architecture of the use case of ORI computation tool inside the A4CORSICA custom version of A4EU

The overview of the generic technical specifications used for the development of ORI computation tool is summarized in Table 6.1.

6.4 Practical Implementation and Guide Line As mentioned, Operational Resilience Assessment tool/method performs a mapping of an Operational Resilience Index (ORI) that is providing operational resilience level of an entity, at building scale. Required inputs are (i) building map, (ii) hazard maps, (iii) networks data (e.g. electricity, transportation, water, communication). A classification of the input data into proper urban function and a briefing on the method is necessary before use of the tool/method (2–3 h of training). A global overview of the interface can be presented as follow: when connected to A4CORSICA interface, the user can click on a specific ORI tab. There, the user can upload files for ORI computation, launch ORI computation and eventually visualize ORI results on A4CORSICA WEB-GIS interface (Fig. 6.3). Moreover, in the A4DEMOS ORI tab, an advanced settings button allows to connect to the advanced

72

M. Abily et al.

Table 6.1 Specification and format use-case of ORI implementation in A4CORSICA version of A4EU User

List of users: First responder (Fire brigade, civil protection, police brigade), urban planner, command control centre & emergency management services (managers and planners).

Technologies

WebGIS; Java and own UNS computation algorithm

API and communication protocol for data exchange

REST API

Input/Output (with data format)

Inputs format: Shape file (follow defacto ESRI standards) • for building layer (with a specific formatting of attribute table), • for networks (electricity, transportation, communication and water supply) • for critical infrastructure & social event. • for the hazard map Outputs: KML Encoding standard & ASCII raster for the ORI output map (follow OGC standards)

Fig. 6.3 schematic chart of the process to perform use of the functionalities of ORI computation tool interface (left) and overview of the advanced setting interface (right)

interface (developed by UNS) to upload network layer to take them into consideration in computation and to parameterize maturity level of urban function and of the urban services dependencies. Illustration of the advanced setting interface can be found through: http://ori-webinterface.polytech.unice.fr/index.php and will not be described here. However, operational guidance summary is presented in this section to prepare the inputs: (i) building map, (ii) hazard maps, (iii) networks data; and regarding preparation of: (iv) the advanced computation setting, and (v) the result visualization. Operational procedure, especially for the pre-formatting of the inputs (building, hazard and network shapefile) is detailed here to support the user in having a written template on how to operate and to use the ORI computation tool. Moreover, during the training session, the user receives templates of the different input files as reference.

6 Operational Resilience Index Computation Tool …

73

Table 6.2 listing of the Urban Function (UF) as integer inside the building shapefile Housing (individual or collective)

1

Education (for local and non-local education services)

2

Food (area for food storage)

3

Working (areas for industry and areas for non-industrial activities)

4

Safety (areas established for location of police, fire brigade and rescue services)

5

Governance (Administrative bodies)

6

Health (hospitals on local and non-local level)

7

Leisure and tourism (on local and non-local level)

8

Religion (churches and cemeteries)

9

Mix/unknown

10

Critical infrastructure

11

(i) Building Formatting of the Building attribute table is as follow: ID: long integer; UF (urban function): integer; ELECTRICITY: string; LINE: string. Presetting of the UF (field in the building shape file should follow classification described in Table 6.2. The electricity and public transport relationship at building scale are here specified respectively in the electricity field and line field. Here, the information concerns the internal dependencies at building scale toward these two specific networks. Presetting of the ELECTRICITY field sees the user put the reference of the electric transformer providing electricity to each building in this column for a given building ID. Presetting of the LINE field should see the user put the references of the public transport lines separated with comma. (ii) Hazards Formatting of the hazard attribute table is as follow: ID (long integer); HAZARD (integer). For the HAZARD field, Hazard values classes (in case of flood hazard) are defined as follow: Class 0: 0 m; Class 1:0.1–0.25 m; Class 2: 0.26–0.5 m; Class 3: 0.51–0.75 m; Class 4: 0.76–1 m; Class 5:1.1–1.5 m; Class 6: > 1.51 m. (iii) Networks Example of one of the Network information is given here (Electricity). Formatting of the attribute table of the electricity network attribute table is:

74

M. Abily et al.

ID (long integer); NAME (string); NAMELV2 (string). In the NAME column, the first level of electric transformers providing electricity to buildings should be listed. Then, upper levels of electric transformers providing energy to this first level of transformers can be put in front of the corresponding transformers in column NAMELV2. In case if there is more than two upper levels, the user should add a new string type column, with name: “NAMELVn”, where n should be the number of the level of the electric transformer. Once the shape files (i, ii, iii) are prepared, the process through the interface is following: uploading of the files through the interface panel. (iv) Detail are provided here for the advanced setting computation. To parameterize the Urban function maturity levels and the dependencies to urban services for each urban functions. Maturity levels evaluation guideline is following. Level 1 = Informal: Individual action with no coordination Level 2 = Basic: Individual actions with limited coordination Level 3 = Initial: Limited implementation and coordination on local level Level 4 = Coordinated: Fully integration and partial implementation Level 5 = Integrated: Fully integration and implementation Dependencies to urban services are ranked from strong (1) to weak (5). This process is done using the “Parameterize ORI computation” button (Fig. 6.3) or has been done and saved before during the training session. (v) The results will be visualized on the interface. ORI for mapping: min value is 1, max value is 5. Four classes of colors for mapping are represented in this interface. 1–1.999 = RED 2–2.999 = ORANGE 3–3.999 = YELLOW 4–5.000 = GREEN The ORI results can be downloaded byclicking on the “Export” button of the panel. A zip file will provide a KML and Shape file format of ORI results. As a last recommendation, User is informed during the training and it is recalled here that weights are given to the urban function. These weights depends on local expertise, temporal specificities of the hazard notably. They are fixed in the computation method during the training itself as follow: 1, 2 Very low to low importance 3 Medium importance 4, 5 Medium high to high importance

6 Operational Resilience Index Computation Tool …

75

6.5 Application: Example of Biguglia Catchment 6.5.1 Biguglia Catchment, Flood Scenario and Hydraulic Model The Biguglia pound catchment drains a 182 km2 area (Fig. 6.4). This area is mostly urbanized in its downstream, and thus flatter part, where agricultural activities are also concentrated in the vicinity of the pound itself. This urbanized area has multiple classical land use types of activities (housing, working, commercial and leisure areas) [10, 11]. Main infrastructures are located along a North/South axis where the main communication road and train line are located, thus perpendicularly to the catchment main slope (East/West) which include from North to South (non exhaustive list) main road, train rail, football stadium, commercial malls and airport. Thus meanwhile this area has important economical and ecological interest, it has suffered from several natural hazard in the recent years such as wild fire, coastal storm surge and rainfall induced flood event. As a consequence, the Fire and rescue services (S.I.S.2.B) have selected this particular area to test the Operational resilience Index within the A4EU customized version, so called A4CORSICA. To perform the test, a for this case Polytech Lab team computed the hazard map for the SIS2B. A six hours long, rainfall runoff event cumulating 165 mm has been simulated over a 20,000,000 cells structured grid using a full 2D Shallow Water Equation (SWE) based modeling approach on a HPC structure over 124 22 CPU and took six hours of computation. The code, FullSWOF_2D is a well documented code [12, 13] mostly used by research community using finite volume method on structured grid, but any industrial code allowing to produce a flood map respecting the state of the art of flood map production, and in that specific case allowing to use a computing cluster, can be used. Results of computation have generated a 5 m resolution flood hazard maps, used as inputs for the ORI computation. The simulated event has been designed with a magnitude comparable to the one that affected the area the 24 Nov. 2016, where 79 municipalities have been listed as affected by a natural disaster by the French government after the event. The other inputs for ORI computation (buildings and networks) for the Biguglia catchment where public data provided by pilot site end users: Corsica fire and rescue services (SIS2B). They were formatted according to the process described in previous section.

6.5.2 Technical and Qualitative Results Through Users Feed Back Technically speaking, the implementation of the ORI tool inside A4CORSICA is fully operational for the different functionalities:

76

M. Abily et al.

Fig. 6.4 Visualization of the Biguglia catchment with the flood map of computed maximum water elevation during the event (top); focus on the 5 m resolution flood map of maximum water deep on a given area (middle); and visualization of ORI computed at building scale for the given area (bottom)

6 Operational Resilience Index Computation Tool …

• • • •

77

upload of input data; advanced settings to parameterize ORI computation; computation of ORI; visualize and export results.

Figure 6.4, shows a focus on the ORI computed for the Biguglia catchment on a district size area which is composed of a mix of urban functions (mostly housing and commercial). The ORI was also computed for a wildfire event (event of the 26.07.2017) which occurred in the area (not presented here). The first feedbacks from the end-users (SIS2B), received after the test period phase were mostly regarding: • request for improvement of the result visualization (upgraded since then); • interest to upgrade the computation method by including a critical infrastructure layer (ongoing development). Lastly, if advantage of the method and the tool has been enhanced as well as the swiftness of the ORI computation, it has to be stressed out that the sound use of it requires not only a training period, but also when using it the first time, a formatting of the data which could require a one to two working days. Unfortunately, this process of data formatting cannot be automatically done due to the diversity of the potential sources. This however allows the end-use which formats the data, to double check the validity of the data sets and to have a control on the classification of this data which impacts the afterward results of the computed ORI.

6.6 Conclusion ORI -the tool/method performs a mapping of an Operational Resilience Index (ORI) that is providing operational resilience level of an entity (building scale). Required inputs are (i) hazard maps, (ii) building maps, (iii) networks data (electricity, transportation, water, communication). A classification of the input data into proper urban function and a briefing on the method is necessary before the use of the tool/method (1–2 h of training). Acknowledgements This project has received funding from the European Union’s Horizon 2020 research and innovation program (H2020-DRS-1-2015) under the grant agreement no. 700099. Photogrammetric and photo-interpreted dataset used for this study have been kindly provided by Nice Côte d’Azur Metropolis for research purpose. This work was granted access to the HPC and vizualization resources of the “Centre de Calcul Interactif” hosted by University Nice Sophia Antipolis.

78

M. Abily et al.

References 1. Vis M, Klijn F, De Bruijn KM, Van Buuren M (2003) Resilience strategies for flood risk management in the Netherlands. Int J River Basin Manag 1(1):33–40 2. Tiernan A, Drennan L, Nalau J, Onyango E, Morrissey L, Mackey B (2019) A review of themes in disaster resilience literature and international practice since 2012. Pol Des Pract 2(1):53–74 3. Carpenter S, Walker B, Andries JM, Abel N (2001) From metaphor to measurement: resilience of what to what? Ecosystems 4(8):765–781 4. Folke C, Carpentier S, Walker B, Scheffer M, Chapin T, Rockstrom J (2010) Resilience thinking: integrating resilience, adaptability and transformability. Ecol Soc 15(4):20 http://www. ecologyandsociety.org/vol15/iss4/art20/ 5. Batica J (2015) Methodology for flood resilience asessment in urban environments and mitigation strategy development (Doctoral dissertation, Université Nice Sophia Antipolis) 6. Ler LG (2018) Résilience et smart water management: stratégies de mise en œuvre pour des villes intéligentes (Doctoral dissertation, Université Côte d’Azur) 7. Wallace D, Wallace R (2008) Urban systems during disasters: factors for resilience. Ecology and Society 13(1):18 8. McPhearson T, Andersson E, Elmqvist T, Frantzeskaki N (2015) Resilience of and through urban ecosystem services. Ecosyst Serv 12:152–156 9. Batica J, Gourbesville P (2016) Resilience in flood risk managment—a new communication tool. Proceedia Eng 154:811–817 10. BRGM (2010) Etude des interactions entre les eaux souterraines, les eaux de surface et l’étang de Biguglia—Rapport final 11. Garel E, Huneau F, Jaunat J, Celle-Jeanton H, Santoni S, Garrido M, Pasqualini V (2016) Groundwater role on the ecological preservation of the Biguglia lagoon (Corsica, France). In: 35th international geological congress, Cape Town, South Africa 12. Delestre O (2010) Simulation du ruissellement d’eau de pluie sur des surfaces agricoles (Doctoral dissertation, Université d’Orléans) 13. Delestre O, Cordier S, Darboux F, Du M, James F, Laguerre C, Lucas C, Planchon O (2014) FullSWOF: a software for overland flow simulation. In: Gourbesville P, Cunge J, Caignaert G (eds) Advances in hydroinformatics. Springer, Singapore, pp 221–231

Chapter 7

Realtime High Resolution Flood Hazard Mapping in Small Catchments Flavio Pignone, Lorenzo Campo, Daniele Dolia, Rocco Masi, Giacomo Fagugli, Daniele Ferrari, Simone Gabellani, Francesco Silvestro, Nicola Rebora, and Francesca Giannoni Abstract All over the world a lot of cities are located in flood-prone areas and millions of people are exposed to inundation risk. To cope with that the civil protection organism demands efficient tools able to predict flooding risk and the associated ground effects with the goal of social safety. For this reason, the flood forecasting systems on specific river sections, also called hydro-meteorological chains, has become very useful due to the ability to use rainfall observations and predictions to provide in advance a quantitative evaluation of possible ground effects in term of discharge and peak flow. Recently, thanks to the advancement in computing performances, the hydrometeorological chains can be completed with the hydraulic models able to compute bi-dimensional flooding maps. Unfortunately, the use in real time is still a challenging task both in large basins, for the extension of the potentially flooded areas, and in small-medium basin for the temporal scale of the phenomena. To overpass the problem, an abacus of flood scenarios has been used to create in real time inundation scenarios useful for civil protection and authorities to evaluate flood risk. The abacus is created with a full bi-dimensional hydraulic model able to assess flood extension and flow depth maps accordingly to different statistical quantiles of discharge. An algorithm able to join the different return period scenarios has been created to define the maximum flood extensions expected given the flow forecast or the observations of the hydrometric level converted in discharge for some river sections. The algorithm has been tested on past events on Liguria region where there are several small and very small catchments ( 10000 km2 ) the propagation of a flooding wave along the main river can be observed and evaluated by monitoring, while when the basin considered are small or very small (Areas < 100 km2 ) the prediction is the unique way to have enough time to react [1–32]. Looking under this perspective is missing the so called “last mile” that consist in a prediction of specific flood scenarios based on the hydro-meteorological chain input, that take into account all the possible interaction of a possible flood with the surrounding. To complete the hydro-meteorological chain is necessary to couple it with a hydraulic model that simulate a flood under specific input. Thanks to the computational development of the last years it is now possible to use hydraulic modelling for civil protection purposes. Unfortunately, the hydraulic modelling at high resolution needs computational time to be run and this is contrast with the goal of having a real-time scenario. To overpass the problem was developed an algorithm that create different scenarios associated to different return time of the flood, starting from a limited number of hydraulic simulations at fine spatio-temporal scale. The process allows to generate an abacus of flood scenario associated to discharge values along the hydrography. In this way is possible to add an important component to complete the hydro-meteorological forecasting chain. Within this paper is described how the hydro-meteorological forecasting chain coupled with a small scale hydraulic model works as well as the operational advantages of its use by the civil protection. At the moment, the

7 Realtime High Resolution Flood Hazard Mapping in Small Catchments

81

Fig. 7.1 Hydro-meteorological chain adopted in Liguria region based on the coupling between meteorological model, hydrogeological model and hydraulic model

model is operational in the Liguria region and adopted within the ANYWHERE H2020 framework on the Ligurian pilot site and within a regional POR project (http://anywhere-h2020.eu/ and https://www.regione.liguria.it/homepage/ fondieuropei/por-fesr-2014-2020.html) (Fig. 7.1).

7.2 System Framework The proposed method allows the user to obtain, in real-time, possible predictions of the flood scenario associated to the hydro-meteorological chain outputs. These scenarios report an indication of the areas prone flood generated by a specific prediction both in terms of possible extension and water depth. The procedure was done following these steps: 1. Hydraulic modelling of specific domain using as input a distribution of discharge or return time 2. Production of abacus to link a scenario to the associated discharge 3. Algorithm of composition of real-time prediction scenario The completed chain is now operational within the ANYWHERE project on the Ligurian pilot site and within the Civil protection of Liguria region after a regional project, to predict and monitoring the flood risk. In the following are described all the steps followed to implement the method.

82

F. Pignone et al.

7.2.1 Hydraulic Modelling The first step is to produce several maps, scenarios, of flooding associated to a defined discharge. The work was done with the application of the suite TELEMAC2D imposing as boundary condition different volume and peak of discharge related to know return period. In this way was possible to create a first abacus of scenarios at high spatio-temporal resolution that take into account high-resolution information related to bridges, houses, levees and local details important for a reliable hydraulic simulation.

7.2.2 Scenario Definition In the second phase, the modelled scenarios provided the hydraulic model are used as starting point to create the scenarios abacus with a relevant number of return period. The evaluation is done with an algorithm of interpolation that take into account the volume and the extension as main variables to be interpolated. As final product, an abacus of scenarios with different volumes, extensions and water depths is generated. In detail, the abacus is generated following these steps: • Definition of a grow condition that force the water depth at return time Ti+1 bigger than the return time Ti. • Evaluation of area and volume of the modelled scenarios • Creation of a function of area and volume with a linear interpolation Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) • Creation of a curve “virtual water depth – Volume” pixel by pixel • Definition of the scenarios interpolated (Fig. 7.2).

7.2.3 Realtime Prediction Scenario The final step to obtain a real-time scenario is the coupling of the results of hydrological model with the generated scenarios abacus produced. Once is given specific section where the prediction of the discharge is known, through the abacus it is possible to refer the prediction to the closest scenario generated. More are the control points along the river more precise can be the final real-time map that consists in a patchwork of different scenarios on specific domain. Within this final step is necessary some step to prepare the data as: • Subdivide the hydrography in different features related to a specific section where the discharges are predicted by the model. • Define the domain of every single feature designed, based on the topography of the area (which cells drain in the channel defined in the feature).

7 Realtime High Resolution Flood Hazard Mapping in Small Catchments

83

Fig. 7.2 Input to the creation of scenarios abacus

Fig. 7.3 Definition of the domain of the single point of interested based on the hydrography and the pointer map

• Couple the results of the hydrogeological model with the scenarios abacus and extract for every subdomain which scenario is the closest to the prediction. • Cut the scenario on every subdomain and compose the final real-time map (Fig. 7.3).

84

F. Pignone et al.

7.3 Test of the Procedure After the creation of the procedure and the abacus all the algorithm was tested on some previous event occurred on Liguria region. The test was conducted to verify that the interpolation of the scenarios was correct and the water depth and extension varying correctly respect to the return time period. To evaluate the success of the procedure are verified: • Locally on single pixel: the criteria is to take random pixel and verify if the water depth grows with return period (Fig. 7.4) • Scenario scale: the criteria is to define a percental “Growing index” that gives at a larger scale if generally the water depth grows with return time. The percentage is evaluated pixel by pixel assessing if the water depth of return time Ti is smaller than return time Ti+1, obviously when the values are different from 0. In the best way, all the map has values equal to 100%. When there are other values a specific analysis was done to verify if there is presence of hydraulic singularity or clear error (Fig. 7.5). The results shown that at least at 85% the pixel grow respect to the return time. In all other cases, there are presence of hydraulic singularity due to manmade structure or behavior of the dynamic of the flow (Fig. 7.6). A final representation of the “Growing index” that summaries the information is the creation of the Probability Density Function (Fig. 7.7).

Fig. 7.4 Return period - curve water depth

7 Realtime High Resolution Flood Hazard Mapping in Small Catchments

85

Fig. 7.5 “Growing index” evaluated on a test case on Liguria region

7.4 Back Analysis of Chiavari Flood of November 13th , 2014 To understand the effectiveness and efficiency of the model used it was decided to analyze the flood occurred in the eastern region of the Ligurian Region between 9 and 13 November 2014. The meteorological event that has taken place this area was born following the presence of a large nuisance from the west. The great humidity advection has generated significant rainfall throughout the region. Heavy rains together with high soil moisture condition (50–60% for basin of Entella River) caused numerous floods in the Entella basin causing two casualties. During this event, the Pianesi rain gauge recorded 274.8 mm in 4 days, similarly, the Caperana rain gauge recorded 234.6 mm in 4 days. The maximum water depth was observed on November 10th, on this day considerable increases were observed for the Entella, Graveglia and Lavagna rivers (Table 7.1). From the knowledge of water dept in the river section it was possible to define the maximum discharge for all area of competence (Fig. 7.8).

86

F. Pignone et al.

Fig. 7.6 Zoom of the “Growing index” evaluated on a test case on Liguria region. The picture shown which pixels have lower value and where are located. After a specific analysis was identified a hydraulic singularity related to a change of river bed slope

Fig. 7.7 Probability Density Function of the Growing index. It is shown how the scenario generally increase the water depth with the incrementation of the return time

7 Realtime High Resolution Flood Hazard Mapping in Small Catchments

87

Table 7.1 Measurements on hydrogauges along the Entella basin on 13th November 2014 Hydrometric station

Maximum water depth

Schedule

Date

Level increase

Lavagna-Carasco

8.33 m

10:00 pm

November 10, 2014

7.00 m

Graveglia-Caminata

2.41 m

11:30 pm

November 10, 2014

1.81 m

Entella-Panesi

6.39 m

10:30 pm

November 10, 2014

7.23 m

Fig. 7.8 Maximum discharge for Area of competence for Entella basin

Subsequently it was possible to define by the algorithm the composition of the flooded areas for different return period. In this case it’s possible define the different return period for different sub-basin (Fig. 7.9).

88

F. Pignone et al.

Fig. 7.9 Flood hazard map for Entella river on 11, November 2014

7.5 Conclusions The procedure described allows the user to obtain in real-time a flooding scenario as an output of the hydrometeorological forecasting chain that should be used to predict the area interested by an approaching event and take into account all necessary measure to face it. The main aspect of the methodology is coupling a discharge scenario with the associated flooding scenario, so once a discharge prediction or observation is available can be in real-time know in advance what will happen in a close future. As written in the previous paragraphs the method of running several hydraulic modellings and interpolating the scenario generated to build the abacus seems validated according to the idea that increasing the return time means increment the water depth. At the moment, a validation respect to past event is ongoing. Nowadays, the procedure is operational within the H2020 ANYWHERE project in the Ligurian pilot site. In this framework, the municipality of Genoa uses a webGIS platform (A4LIG) where are available predictions, observations and the abacus implementation all over the domain with a specific application on one of the main city rivers (Bisagno river). Moreover, the abacus is also used by the Ligurian Civil Protection on several catchments within the Ligurian territory, to help the user to predict the scenario useful to issue a regional civil protection warning.

7 Realtime High Resolution Flood Hazard Mapping in Small Catchments

89

In conclusion, the system proposed and now operational represent a valid tradeoff between computational time and accuracy of the predictions it is able to provide. The system generates in output a clear scenario of flooding in terms of extension and water depth. These systems allow the civil protection operators to take in advance important measures to face an event and reduce the impacts in terms of number of human losses and damages.

References 1. Brussolo E, Von Hardenberg J, Ferraris L, Rebora N, Provenzale A (2008) Verification of quantitative precipitation forecasts via stochastic downscaling. J Hydrometeorol 9(5):1084– 1094 2. Brussolo E, von Hardenberg J, Rebora N (2009) Stochastic versus dynamical downscaling of ensemble precipitation forecasts. J Hydrometeorol 10(4):1051–1061 3. Ferraris L, Reale O, Turato B (2001) Synoptic and hydrological analysis of a flood event. Phys Chem Earth Part B 26(9):655–661 4. Ferraris L, Rudari R, Siccardi F (2002) The uncertainty in the prediction of flash floods in the northern Mediterranean environment. J Hydrometeorol 3(6):714–727 5. Ferraris L, Gabellani S, Rebora N, Provenzale A (2003) A comparison of stochastic models for spatial rainfall downscaling. Water Resour Res 39(12), 1368. http://doi.org/10.1029/ 2003WR002504 6. Gabellani S, Giannoni F, Parodi A, Rudari R, Taramasso AC, Roth G (2005) Applicability of a forecasting chain in a different morphological environment in Italy. Adv Geosci 2:131–134 7. Gabellani S, Boni G, Ferraris L, Von Hardenberg J, Provenzale A (2007) Propagation of uncertainty from rainfall to runoff: a case study with a stochastic rainfall generator. Adv Water Resour 30(10):2061–2071 8. Gabellani S, Silvestro F, Rudari R, Boni G (2008) General calibration methodology for a combined Horton-SCS infiltration scheme in flash flood modeling. Nat Hazards Earth Syst Sci 8(6):1317–1327 9. Ghizzoni T, Giannoni F, Roth G, Rudari R (2007) The role of observation uncertainty in the calibration of hydrologic rainfall-runoff models. Adv Geosci 12:33–38 10. Giannoni F, Roth G, Rudari R (2003) Can the behaviour of different basins be described by the same model’s parameter set? A geomorphologic framework. Phys Chem Earth Parts A/B/C 28(6–7):289–295 11. Hally A, Caumont O, Garrote L, Richard E, Weerts A, Delogu F, Ivkovi´c M (2015) Hydrometeorological multi-model ensemble simulations of the 4 November 2011 flash flood event in Genoa, Italy, in the framework of the DRIHM project. Nat Hazards Earth Syst Sci 15(3):537–555 12. Hardenberg JV, Ferraris L, Rebora N, Provenzale A (2007) Meteorological uncertainty and rainfall downscaling. Nonlinear Process Geophys 14(3):193–199 13. Laiolo P, Gabellani S, Rebora N, Rudari R, Ferraris L, Ratto S, Cauduro M (2014) Validation of the Flood-PROOFS probabilistic forecasting system. Hydrol Process 28(9):3466–3481 14. Metta S, von Hardenberg J, Ferraris L, Rebora N, Provenzale A (2009) Precipitation nowcasting by a spectral-based nonlinear stochastic model. J Hydrometeorol 10(5):1285–1297 15. Molini L, Parodi A, Siccardi F (2009) Dealing with uncertainty: an analysis of the severe weather events over Italy in 2006. Nat Hazards Earth Syst Sci 9(6):1775–1786 16. Molini L, Parodi A, Rebora N, Siccardi F (2006) Assessing uncertainty in radar measurements on simplified meteorological scenarios. Adv Geosci 7:141–146 17. Parodi A, Kranzlmüller D, Clematis A, Danovaro E, Galizia A, Garrote L, Siccardi F (2017) DRIHM (2US): an e-science environment for hydrometeorological research on high-impact weather events. Bull Am Meteor Soc 98(10):2149–2166

90

F. Pignone et al.

18. Poletti ML, Parodi A, Turato B (2017) Severe hydrometeorological events in Liguria region: calibration and validation of a meteorological indices-based forecasting operational tool. Meteorol Appl 24(4):560–570 19. Rebora N, Ferraris L, von Hardenberg J, Provenzale A (2006) RainFARM: rainfall downscaling by a filtered autoregressive model. J Hydrometeorol 7(4):724–738 20. Rebora N, Ferraris L (2006) The structure of convective rain cells at mid-latitudes. Adv Geosci 7:31–35 21. Rebora N, Ferraris L, Von Hardenberg J, Provenzale A (2006) Rainfall downscaling and flood forecasting: a case study in the Mediterranean area. Nat Hazards Earth Syst Sci 6(4):611–619 22. Rebora N, Molini L, Casella E, Comellas A, Fiori E, Pignone F, Parodi A (2013) Extreme rainfall in the Mediterranean: What can we learn from observations? J Hydrometeorol 14(3):906–922 23. Siccardi F, Boni G, Ferraris L, Rudari R (2005) A hydrometeorological approach for probabilistic flood forecast. J Geophys Res Atmos 110(D5):1–9 24. Siccardi F (2012) Prevision, Prediction and Perception. OECD High level Risk Forum, 13–14 December 2012, Paris 25. Silvestro F, Rebora N, Ferraris L (2011) Quantitative flood forecasting on small-and mediumsized basins: a probabilistic approach for operational purposes. J Hydrometeorol 12(6):1432– 1446 26. Silvestro F, Gabellani S, Giannoni F, Parodi A, Rebora N, Rudari R, Siccardi F (2012) A hydrological analysis of the 4 November 2011 event in Genoa. Nat Hazards Earth Syst Sci 12(9):2743–2752 27. Silvestro F, Rebora N (2012) Ensemble nowcasting of river discharge by using radar data: operational issues on small and medium size basins. In: Red book for WRAH symposium 2012, vol 351, pp 508–513 28. Silvestro F, Rebora N (2012) Operational verification of a framework for the probabilistic nowcasting of river discharge in small and medium size basins. Nat Hazards Earth Syst Sci 12(3):763–776 29. Silvestro F, Rebora N, Cummings G, Ferraris L (2017) Experiences of dealing with flash floods using an ensemble hydrological nowcasting chain: implications of communication, accessibility and distribution of the results. J Flood Risk Manag 10(4):446–462 30. Silvestro F, Rebora N, Giannoni F, Cavallo A, Ferraris L (2016) The flash flood of the Bisagno Creek on 9th October 2014: an “unfortunate” combination of spatial and temporal scales. J Hydrol 541:50–62 31. Silvestro F, Rebora N, Rossi L, Dolia D, Gabellani S, Pignone F, Masciulli C (2016) What if the 25 October 2011 event that struck Cinque Terre (Liguria) had happened in Genoa, Italy? Flooding scenarios, hazard mapping and damage estimation. Nat Hazards Earth Syst Sci 16(8):1737–1753 32. Taramasso AC, Gabellani S, Parodi A (2005) An operational flash-flood forecasting chain applied to the test cases of the EU project HYDROPTIMET. Nat Hazards Earth Syst Sci 5(5):703–710 33. UNISDR (2015) The human cost of weather related disaster. https://www.unisdr.org/files/ 46796_cop21weatherdisastersreport2015.pdf

Chapter 8

Visualization of Flood Simulation with Microsoft HoloLens Shanyu Wang, Jianrong Wang, Philippe Gourbesville, and Ludovic Andres

Abstract It is necessary for decision makers and stakeholders to have the necessary information and tools needed to make sound decisions for mitigation plans for flooding. However, most of the decision makers and stakeholders are not technicallytrained, it is often difficult for them to envisage the flood impacts based on the reports of the flood damages analysis. Thus, in order for them to visualize the impacts of the floods, a virtual reality tool is developed where the user is transported to the flooding location where they are able to access the flood damage and the benefits of mitigation in first person perspective. The visualization of the flood simulations using virtual reality transports users into a simulated world and transforms watching the screen into a living experience. Compared with the original plane drawing, screen watching, and sand table model, exhibition with virtual reality will no longer be limited by time or space. It also provides more comprehensive information with lower cost and better viewing experience. The scenes that are presented in this work are produces using the AR device of Microsoft HoloLens. They have the following functions: (1) exhibition of the city model, (2) interactions to manipulate the model, flood simulation. The city model was firstly designed using CAD (Computer Aided Design) software. The exported model files are then imported into the Unity3D, where IDE (Integrated Development Environment) is used to design the virtual scenes. Flood simulation data are also imported into Unity3D which are fed into the program to compute the level of the flood at every location in the scene. S. Wang (B) · J. Wang Tianjin University, Tianjin, China e-mail: [email protected] J. Wang e-mail: [email protected] P. Gourbesville · L. Andres Polytech Lab, University Nice Sophia Antipolis, 930 route des Colles, 06903 Sophia Antipolis, France e-mail: [email protected] L. Andres e-mail: [email protected] L. Andres Métropole Nice Cote d’Azur, 06000 Nice, France © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_8

91

92

S. Wang et al.

Keywords Disaster simulation · Holographic visualization · User experience · Human-computer interaction · Flood simulation

8.1 Introduction Simulation of building models and disasters is often carried out in disaster prediction system. For example, urban flood forecasting requires first designing urban building models, as well as water simulations. This process needs to be constantly revised to gradually improve the model. With the development of computers, there are more convenient tools, such as the widely used AutoCAD software. However, the software still belongs to the category of two-dimensional design mode, and it cannot achieve the three-dimensional display effect. The further development of various technologies in the computer field has led to the emergence of 3D modeling software and its use in the field of architectural design. After the model is designed, because the disaster is not reproducible, the predicted results are often displayed in the form of virtual animation video or sand table model. As a result, the audience cannot be placed inside the city for an experience, and it is difficult to understand the details of the damage. The application of virtual reality technology in the field of flood simulation prediction has many important meanings, which are embodied in the following aspects. First, virtual reality technology can present the entire architectural design in a holographic image, freeing the limitations of simple detection through computer displays. The review of the model is more stereoscopic, and it is important to see more details that cannot be displayed on the screen, which is of great significance for improving the accuracy of the urban model. On the other hand, for the audience, you can be in the city model and move freely, directly to observe the internal details that the sand table and video information cannot be displayed. Second, interacting with operations in virtual reality is very simple. Simple finger clicks, drags, etc., can achieve the model’s rotation, movement, scaling and other transformations. Compared to the complexity of design software operations, designers can use virtual reality technology to more easily manipulate the model for multi-angle observation. According to the above analysis, the realization of virtual reality technology in disaster prediction has great significance. Simply passing the message through the screen is no longer sufficient for today’s needs. Omni-directional information, simplified operations, and freedom from time and space constraints, virtual reality technology can fully meet these needs. The virtual reality application designed and implemented in this paper can comprehensively and stereoscopically display urban flood simulation anytime and anywhere. With the application we produced in this paper, the user observes and operates the model through virtual reality (Fig. 8.1), which has a better experience, smooth operation, and richer information.

8 Visualization of Flood Simulation with Microsoft HoloLens

93

Fig. 8.1 User and experimental scene

8.2 VR and BIM A large number of VR (Virtual Reality) products have appeared on the market. These products implement the basic application of virtual reality for different platforms, but the quality is still uneven. Among them are several outstanding and outstanding products. Including Oculus Rift VR helmet, HTC VR helmet, Samsung mobile VR helmet, Sony Morpheus helmet, Microsoft HoloLens AR helmet and more. Among these devices, the most popular are the Oculus and HoloLens products. The development and evaluation of a trauma decision simulation system were carried out using the Oculus device by Harrington et al. [1]. Foerster et al. [2] conducted a neuropsychological assessment of visual processing capabilities. Munafo, Diedrick and Stoffregen [3] found that the Oculus Rift helmet is prone to dizziness of the user and the degree of influence on men and women is different. HoloLens was launched later than the Oculus Rift, but its launch also received a lot of attention. Online [4] and Gottmer [5] introduced the characteristics of HoloLens as a mixed reality device under the Windows10 platform. Silva [6] studied the Holographic imaging principle of HoloLens. Noor [7] analyzed the impact of HoloLens’s most highly disruptive products on mechanical engineering. Choudhury [8] describes the use of holographic objects in HoloLens. Chen et al. [9] combined with the Skype function, realized the remote cooperation scheme of HoloLens. Ghosh and Norala [10] and Hanna et al. [11] used HoloLens to conduct related research on pathology and one-to-one imaging. These efforts have proven revolutionary innovation in the visualization of VR technology across industries. BIM (Building Information Modeling) refers to the construction of a threedimensional model of a building based on various relevant data in a construction project, and it simulates the information of the building through simulation technology. BIM was proposed from the National 3D-4D-BIM program launched by the General Services Administration of the United States in 2003. A series of BIM guides have also been released. By the end of 2008, relevant BIM specifications and standards were proposed. In 2010, Japan’s Ministry of Land, Infrastructure, Transport and Tourism also began the implementation of BIM technology. As of now, BIM technology has been widely used throughout Japan.

94

S. Wang et al.

The combination of virtual reality technology and BIM can give full play to their respective advantages and use VR as a tool to better achieve the goal of BIM. Goulding, Rahimian and Wang [12] conducted research on the application of virtual reality-based BIM in facility management. Ganapathi Subramanian [13] implements a virtual reality-based cloud platform BIM integration AEC project. By extending the vertical view of the BIM software, Xue et al. [14] implement a BIM virtual reality system. The work done in this paper belongs to the BIM visualization software in BIM software, which is responsible for visualizing the built BIM model. The commonly used BIM visualizations include 3DSMax, SketchUp, etc., which are program software running on a computer, and can also be used to build models. Through the operation of the mouse and keyboard, freely rotate, zoom, and move on the screen to view the details of the model. In this paper, through the virtual reality technology, the BIM model is removed from the constraints of the screen, and the “real” placement into the world can improve the accuracy of the model and the fit of the design.

8.3 Model Preprocessing Choosing the right model and pre-processing the details of the model is the first priority before the model is presented. The Unity 3D engine is able to put 3D models into the scene, script them, and then generate Windows Store projects to further generate holographic virtual reality projects that can run on HoloLens devices through Visual Studio. The implementation of this article is also based on the Unity3D engine. Therefore, the selected building model data needs to be imported into the engine, and at the same time ensure the completeness of the data, and there is no missing information. Currently, Unity3D supports up to ten types of model formats that can be imported, but it does not support all the attributes of each external model. The specific parameters are shown in Table 8.1. As can be seen from Table 8.1, Unity3D supports different models of different data formats. For building information models, complex grid and material map information are often included. Therefore, data formats such as 3ds and obj that do not support material information import cannot be used. In addition to importing building model files, Unity3D also supports the design creation of models. In this case, we can directly add 3D objects to the scene, set up maps and other information, without the need to import. The City model used in this topic is the.skp type created by SketchUp. The urban model designed in SketchUp is shown in Fig. 8.2. In SketchUp Pro, we need to use the function of exporting 3D models, select the FBX type whose output type is Unity3D, and ensure that all planes are divided into triangles, and the planes on both sides are exported and the texture map is exported. The completed output FBX file can be imported into Unity3D. The Solidarity Create Material folder saves the material information in the model and the texture folder to save the texture used by

8 Visualization of Flood Simulation with Microsoft HoloLens

95

Table 8.1 Parameters of different model types supported by Unity Format

Software

.mb.mal

Maya

.max

3D Studio

.jasl

Cheetah 3D

.cedl2

Cinema 4D

.blendl

Blender

.dae

Autodest FBX

.xl

XSI

.skp

SketchUp Pro

.3ds

3D Studio

.obj

Wavefront

Grid √

Material √

Animation √

Skeleton √

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√ √

Fig. 8.2 City model in SketchUp

the model. Finally, a pre-made file is generated, which can be imported into the scene for use. After importing the model into Unity, we need to create a scene (Scene) for rendering the model. A scene is a view through which scenes that a user can see in a program are edited and arranged. Camera: When creating a new scene, Unity automatically places a Camera object as the main camera in the scene. When the program is deployed to the HoloLens device, the helmet becomes the MainCamera in the scene, changing the specific value of the Transform component as the helmet moves and rotates. Unlike the Camera

96

S. Wang et al.

Fig. 8.3 City model scene in Unity3D

created by default, cameras for the Holographic program require some settings. First, Clear Flags needs to be set to Solid Color, and the Background color is set to black. Then, set the Cape parameter of the Clipping Planes to a minimum value of 0.01, so that the distance between the camera’s near plane and the helmet is as small as possible, simulating the actual situation of the human eye. Illumination: Adds appropriate parallel light to the scene to illuminate the model in the scene. It should be noted that the illumination should not be too bright, otherwise the super bright scene will coincide with the real world after deployment to HoloLens, which will make the model difficult to see. Model: Due to the limitation of the computing power of the HoloLens device, when the visible field of the helmet contains too many objects, too many objects overlap, or the texture information of the texture is too complicated, it will cause the visual effect to be stuck. Such problems often occur when observing the entirety of a complex model. Therefore, the hierarchical display processing can be performed. The created scene is shown in Fig. 8.3.

8.4 Interaction The application completed in this paper is used to display the completed urban model in various dimensions, showing the overall appearance and internal details of the building, and adding the function of simulating flood. Therefore, after user analysis, the interactive content of the demand can be derived.

8 Visualization of Flood Simulation with Microsoft HoloLens

97

(1) Move: The user needs to move the holographic image of the city model in three dimensions. The input is the distance of movement in three directions, and the result is that the holographic object moves by a corresponding distance. (2) Rotate: Rotate the model. Since the rotated object is a city model, the rotation axis can be fixed to an axis from the bottom up. The input is the angle that needs to be rotated about the axis of rotation, resulting in a rotation of the holographic object. (3) Zoom: Zoom in or out on the model. The input is the ratio that needs to be enlarged or reduced, and the result is a proportional scaling of the model. (4) Enter: Enter the virtual city model. The input is an incoming command, and the result is that the user’s field of view enters a specified location in the model. (5) Walk: When walking inside the city, walk around the building. The input is the direction of walking and the distance traveled, with the result that the field of view moves a specified distance in the specified direction. As an augmented reality helmet, HoloLens allows users to interact with devices through Gaze, Gesture, and Voice. Gaze is the primary form of HoloLens input and is the primary means of targeting in mixed reality. It can be compared to moving the cursor in the screen with a mouse in a normal computer. Gaze selects specific objects in the holographic image by moving the camera position. Gaze can indicate what the user is paying attention to, and then let the developer decide the user’s current intention. Gesture operations allow the user to use both hands to complete actions on a mixed reality object. For HoloLens, Gesture input makes the user’s interaction with the entire body more natural. Since the gesture operation does not give an exact position in space, the simplified way of placing the hand directly into the HoloLens field of view and interacting with the content immediately allows the user to work directly without wearing the accessory. Through the analysis of the interaction requirements, combined with the interaction mode supported by HoloLens, the specific definitions shown in Table 8.2 can be made for the above five interactions.

8.5 Results and Discussion The demo screenshots in this section are the result of real operation of the user wearing the HoloLens device. First, after the program runs, the city model is presented in the form of virtual reality, as shown in Fig. 8.4. Note that, the blue plane is a simple simulation of the water surface. Together with the city model, a menu listing different types of operation is also presented. By selecting the button of rotate, move and scale, and performing the operations, the city model will change accordingly. As shown in Fig. 8.5. One of the most appealing features of virtual reality is that we can change the view into the model and walk around as if in the real world. Figure 8.6 shows the result after getting inside the city and walking around.

98

S. Wang et al.

Table 8.2 Definition of interaction Name

HoloLens Input

Specific operation

Result

Move

Manipulation

After the Tap action, move freely for a distance

The object moves. The direction and distance of movement are the same as the movement of the hand

Rotate

Navigation

After the Tap action, move horizontally for a distance

The object rotates. The axis of rotation is the vertical z-axis, and the angle of rotation is proportional to the distance the hand moves

Zoom

Navigation

After the Tap action, move vertically for a distance

The object is enlarged or reduced. When the Navigation action is up, it is zoomed in, and down is zooming out. The degree of scaling is proportional to the distance traveled

Enter

Double Tap

Two consecutive taps

The object moves and zooms out. The distance moved is the distance between the original position of the object and the user, and the direction is to move to the user. Scale to a specified size set by the program

Walk

Hold

Tap and hold the gesture

The object moves. The moving direction is the opposite direction of the user’s head orientation, and the moving distance is proportional to the time of the hold action

Fig. 8.4 City model and menu for operation type selection

We use a simple blue plane simulating the water surface. It is settled below the city model at first. After clicking the up button, the plane will slowly rise to a specified position, simulating the spread of floods to the city. When the plane stops rising, we may zoom out a little bit and observe how the city is damaged by the flood (Figure 8.7).

8 Visualization of Flood Simulation with Microsoft HoloLens

Fig. 8.5 Model after rotating, moving and scaling

Fig. 8.6 Walking inside the city model

99

100

S. Wang et al.

Fig. 8.7 Flood simulation with a simple plane

8.6 Conclusion For the use of the HoloLens virtual helmet, the City Model Display application that runs under the HoloLens platform has been developed. The operational model function was designed to enable user interaction and simple simulation of flooding. The entire application completes the virtual presentation and interaction of the urban model, realizing the practical application of virtual reality technology in this field. At a first stage, the processing of the model is done entirely by hand. It is also necessary to create a new scene for the processed model and add it to the program. If we can realize the complexity of the model by a program and automatically create a new scene, we can further reduce the repetitive work. In a second step, we used a simple blue plane to complete the simulation of water. The simulation of the flood propagation process is not deeply detailed or specific to represent the aspect of real flooding waters. In future work plans, particle systems will be used to complete the spread simulation of water flow and the calculation of water levels at different locations in the city. Finally, HoloLens’s perspective could be seen as too narrow and computing power needs to be improved. Currently, HoloLens has a field of view of 30°, equipped with Intel Atom x5-Z8100 processor, 2 GB of memory, 64 GB of storage space, 900 MB of application memory, and Win10 32-bit system. In the future, some improvements can be suggested in order to correct device defects or try to use other virtual reality devices to develop this application for better user experience. Acknowledgements This research is currently under the cooperation between Tianjin University and Polytech lab. The work benefited from the data provided by the Metropole Nice Côte d’Azur, Polytech Nice Sophia.

8 Visualization of Flood Simulation with Microsoft HoloLens

101

References 1. Harrington CM, Kavanagh DO, Quinlan JF et al (2017) Development and evaluation of a trauma decision-making simulator in Oculus virtual reality. Am J Surg 215(1):42–47 2. Foerster RM, Poth CH, Behler C et al (2016) Using the virtual reality device Oculus Rift for neuropsychological assessment of visual processing capabilities. Sci Rep 6(37016):37016 3. Munafo J, Diedrick M, Stoffregen TA (2016) The virtual reality head-mounted display Oculus Rift induces motion sickness and is sexist in its effects. Exp Brain Res 1–13 4. Online H (2015) HoloLens: Augmented-Reality-Brille für Windows 10. Heise Zeitschriften Verlag 5. Gottmer ML (2015) Merging reality and virtuality with microsoft HoloLens. Faculty of Humanities theses (2015) 6. Silva L (2016) Computação Holográfica com Microsoft Hololens. Tendências e Técnicas em Realidade Virtual e Aumentada 7. Noor AK (2016) The HoloLens revolution. Mech Eng Mag Sel Art 138(10):30–35 8. Choudhury D (2017) No Hologram from HoloLens. Commun ACM 60(3):11 9. Chen H, Lee A S, Swift M, et al (2015) 3D Collaboration method over HoloLens™ and Skype™ end points. In: International Workshop on Immersive Media Experiences. ACM, pp. 27–30 10. Ghosh A, Nirala AK (2014) One to one imagery using single HoloLens configuration. Opt Rev 21(6):765–768 11. Hanna MG, Ahmed I, Prajpati S et al (2017) Multiple use cases for Microsoft HoloLens in pathology. Lab Invest 97:396A–397A 12. Goulding JS, Rahimian FP, Wang X (2014) Virtual reality-based cloud BIM platform for integrated AEC projects. Electr J Inf Technol Constr 19(19):308–325 13. Ganapathi Subramanian A (2012) Immersive virtual reality system using BIM application with extended vertical field of, view 14. Xue XM, Wang F, Wei-Quan YE, et al (2016) Research on the application of BIM in facility management based on virtual reality. Comput Knowl Technol 2016(2X):257–259

Chapter 9

Anycare: A Serious Game to Evaluate the Potential of Impact-Based and Crowdsourced Information on Crisis Decision-Making Galateia Terti, Isabelle Ruin, Milan Kalas, Arnau Cangròs i Alonso, Tommaso Sabbatini, Ilona Lang, and Balazs Reho Abstract Extreme weather and climate events challenge weather forecasting and emergency response operations and are often related to high social, environmental and economic impacts worldwide. Effective disaster risk management relies not only on the accuracy and precision of official hazard predictions and related warnings issued by forecasters but also on how those are communicated to and interpreted by end-users to support informed decision-making on allocating human and material resources before and during the crisis. Recent decision-support tools promote the elaboration of multi-hazard ‘impact-based’ or ‘risk-based’ forecasts that translate meteorological and hydrological hazards and related cascading effects into sector- and location- specific impact estimations as the core to improve responder’s and public’s understanding and coping capacity to those risks. To take a first step towards exploring this hypothesis, we propose a new role-playing experiment that engages participants in the decision-making process at different levels of the G. Terti · I. Ruin (B) Univ. Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, 38000 Grenoble, France e-mail: [email protected] G. Terti e-mail: [email protected] M. Kalas · T. Sabbatini · B. Reho KAJO s.r.o., Zilinsky, Slovakia e-mail: [email protected] T. Sabbatini e-mail: [email protected] B. Reho e-mail: [email protected] A. Cangròs i Alonso Catalan Water Agency, ACA, 08036 Barcelona, Catalonia, Spain e-mail: [email protected] I. Lang Finnish Meteorological Institute, P.O. Box 503, 00101 Helsinki, Finland e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_9

103

104

G. Terti et al.

weather-related emergency system. ANYCaRE serious game experiment explores the value of modern impact-based weather forecasts on the decision-making process to (i) issue warnings and manage the official emergency response under uncertainty and (ii) communicate and trigger protective actions at different levels of the warning system. Flood/flash flood or strong wind game simulations seek to reproduce realistic uncertainties and dilemmas embedded in the real-time forecasting-warning processes. A tabletop version of the game was first tested in scientific workshops in Finland, France and Spain where European researchers, developers, forecasters and civil protection representatives helped refine the concept. An improved version was then implemented with undergraduate University students in France and with stakeholders involved in the management of hazardous weather emergencies in Finland. First results indicate that (i) multi-model developments and crowdsourcing tools increase the level of confidence in the decision-making under time pressure, and (ii) facilitates interdisciplinary cooperation and argumentation on emergency response in a fun and interactive manner. ANYCaRE tabletop version appears as a valuable learning tool to enhance participants’ understanding of the complexities and challenges met by various actors in weather-related emergency management. Keywords Role-playing simulation · Emergency decision-making · Forecasting product evaluation · Flood hazard

9.1 Introduction Advances in weather forecasts and warnings are very important to increase preparedness and response capacity of governments, local authorities, economic sectors and the public to extreme weather and climate events. Although necessary, modern technological improvements such as increases in accuracy and lead-time of the hydro-meteorological forecasts are not sufficient to guarantee reduction of fatalities and economic disruptions. Contemporary research indicates the importance of integrating social vulnerability and behavioural processes in the forecast-warning system to better capture and respond to life-threatening situations and catastrophic scenes emerging from the conjunction of the hazard and social vulnerabilities that evolve in space and time [3, 19, 25, 26]. The World Meteorological Organization (WMO) acknowledges the need to pass from ‘what a hazard will be’ to ‘what a hazard will do’; highlighting that despite having accurate and/or precise forecasts available, we still observe inadequate responses and major losses in weather events worldwide [31]. In Europe, recent examples of devastating weather events include the 2003 European heat wave from which France suffered the worst losses with almost 15,000 deaths from August 1 through August 20 [7], the 15–16 June 2010 flash flood event in the Var Department in France that caused the loss of 26 people [20], and the catastrophic fires in the forests of central Portugal that killed 64 people and destroyed more than 480 houses on 17 June 2017 [12].

9 Anycare: A Serious Game to Evaluate …

105

Building upon the Sendai Framework for Disaster Risk Reduction 2015–2030, the WMO suggests the elaboration of multi-hazard ‘impact-based’ or ‘risk-based’ forecasts that translate meteorological and hydrological hazards and related cascading effects into sector- and location- specific impact estimations as the core to improve responders’ and public’s understanding and coping capacity to those risks [11]. The European Environment Agency (EEA) recognizes similar needs and calls for more comprehensive information systems towards integrated risk disaster management across Europe [22]. “EnhANcing emergency management and response to extreme WeatHER and climate Events” (ANYWHERE) project is an innovating action that aims at developing and implementing a pan-European decision-support platform integrating cutting-edge forecasting technology. The project elaborates existing forecasting and nowcasting algorithms as well as impact assessment routines and crowdsourcing data to propose new informational products. The general shift to impact-specific forecast-warning systems relies on the hypotheses that dynamic (near)-real-time impact information (e.g., potentially affected population and critical infrastructure, economic damages) can support emergency services to: – Locate spatially and temporally critical spots for intervention and therefore, better allocate available resources to protect lives and livelihoods. – Communicate more targeted warnings and emergency guidance messages to help the public understanding how certain hazards may affect their life, livelihood and property leading to appropriate self-preparedness and self-protective actions. To take a first step towards testing these hypotheses, we propose a role-playing approach that engages participants in the decision-making process at different levels of the weather-related warning system (from hazard detection to citizen response). Role-playing games (RPGs) are the virtual simulation of real-world events especially designed to educate, inform and train the players for the purpose of solving a specific problem [1, 24]. Through realistic “what if ” scenarios the players are invited to assess different information describing an imminent risk situation and to decide collectively what protective actions, if any, are needed. An important advantage of the simulation approach is its dynamic nature that allows participants to experiment “real time” decisions and experience potential changes in the outcome over time [16]. Also this “learning by doing” process—a fundamental principle in experiential learning theory [10]—takes place in an informal setting without real consequences. Therefore, the literature recognizes such educational or training games as motivating experiential learning tools that go beyond traditional passive learning approaches often applied in conferences and seminars [4, 21]. In the field of disaster risk management, role-playing has been successfully used to increase public awareness as well as to promote preparedness and prevention of losses. Examples of applications include role-playing simulations on flood management on cultural heritage for university students [8], online serious games on natural disasters (e.g., tsunami, earthquakes) for children [Pereira, Prada and Paiva [17], and board serious games on resilience to geological hazardous events for school students or adult stakeholders [13].

106

G. Terti et al.

In this work, we develop ANYWHERE Crisis and Risk Experiment (ANYCaRE) to obtain conclusions on “if “and “how” an improved multi-model and multi-hazard output, including information on (i) impact assessments and maps and (ii) live data on exposure and vulnerability derived from social media and crowdsourced information, can support the decision chain in European warning systems towards better responses. ANYCaRE is first designed as a tabletop role-playing game (or pen-and-paper roleplaying game) for adults in which participants act their role through speech1 while sitting in a comfortable setting [2]. Especially, we benefit from researchers, developers, potential users and other stakeholders from different European agencies meeting frequently to define needs, capabilities and limitations in the frame of ANYWHERE project to play the game and explore uncertainties and dilemmas embedded in the real-time warning and emergency response processes. This paper presents the first implementation of ANYCaRE in the emergency management of flood and flash flood events. First, we describe the experiment structure and set up under the role-playing methodology. The next section explains how the game was applied and tested in two different occasions: (i) the ANYWHERE’s 2nd annual workshop in Helsinki (Finland) and (ii) in the context of a meeting involving members of French municipal councils in Le Puy en Velay (France). Then, we present the results from this testing and interpretations in terms of (i) impact-based and social media information added value on the management of the official emergency response and crisis communication in flash flood events and (ii) gaming utility. Finally, we provide insights in the gaming experience and its potentialities or limitations in the investigation of weather-related crisis decision-making are apposed. Further advancements and extensions of ANYCaRE are also discussed and opportunities for future applications are highlighted2 .

9.2 Role-Playing Gaming Methodology: Designing Anycare 9.2.1 General Concept and Hypotheses ANYCaRE role-playing game primarily aims at testing the value of impactbased and real-time social media products in the improvement emergency managers’ ability to proceed to relevant actions towards the protection of public and property in the area of their responsibility. Participants are invited to play specific characters of the decision-making chain in an interactive and collaborative storytelling related to a weather hazard (here flooding) in a European context. 1 Tabletop

RPG is the original form of RPG that is conducted through discussion between the players. Pen and paper or tables are not strictly necessary for the game. The terms pen-and-paper and tabletop are rather used to distinguish this format of RPG from other formats like Live Action Role-playing (LARP) in which participants act their characters physically as well. 2 An extended version of this serious game development and applications has just been published in Natural Hazard and Earth Science System [27, 28].

9 Anycare: A Serious Game to Evaluate …

107

According to Bowman [1], the role-playing method: (i) ‘enhances a group’s sense of communal cohesiveness by providing narrative enactment within a ritual framework’, (ii) ‘encourages complex problem-solving and provides participants with the opportunity to learn an extensive array of skills through the enactment of scenarios’, and (iii) ‘offers participants a safe place to enact alternate personas through a process known as identity alteration’. Therefore, except for an evaluation tool, the role-playing gaming simulations proposed here serve also as a communication tool to (i) enhance participant’s understanding of the weather-related decision-making complexities (e.g., forecasting/warnings, official emergency actions self-protection); (ii) facilitate collaboration and coordination between the participants who have distinct field of expertise and belong to various national or local agencies/authorities across Europe. The roles to be played and the potential decisions/actions to be chosen by the players are pre-defined based on qualitative evidence gathered during European workshops that took place in March and April 2017 [14] and in previous research [18]. The game is built based on the hypotheses that dynamic real-time impact information (e.g. potentially affected population and critical infrastructure, economic damage) can support emergency services to: – Spatially and temporally locate critical spots for intervention and, therefore, better allocate available resources to protect lives and livelihoods; – To better communicate the warnings and emergency messages by providing answers to the questions of “what?”, “where?”, “when?”, “why?”, and “how to respond?” [9, 15]; – To enforce the public’s/targeted-users’ (self-) protective actions for efficient emergency response.

9.2.2 Experiment Set up Following the principles of tabletop role-playing, the game begins with the description of the setting or storyline by the Game Master or Moderator (GM) who leads the storytelling during the game [6]. Then, each player is provided with a certain role defining his responsibilities during the game. Players get a few minutes to become familiar with their role and introduce it to the rest of the group before the main game simulations start. The game-designers’ team acts as observers of the playing process and based on their observations they facilitate the post-experiment debriefing.

9.2.2.1

Storyline

Inspired by European case studies we introduce “Anywhere City”; an imaginary agglomeration including three distinct areas: A, B and C, located on the slopes and at the foot of highlands drained by two fast-reaction rivers (Fig. 9.1).

108

G. Terti et al.

Fig. 9.1 Presentation and brief description of the three areas considered in the virtual ‘Anywhere City’ of ANYCaRE. Each area includes attributes for special consideration in flood emergency decision-making (e.g., camping, schools, dangerous intersections) representing critical points present in ANYWHERE’s pilot sites

Area A is characterized by relatively steep tree-covered slopes drained by a small basin (e.g. few hundreds km2 ) known for its fast response to precipitation. A suburb of about 1,000 residents and one school is settled on the slopes. There are no permanent settlements or critical infrastructure in flood prone zone but one campground located in the forest close to the riverbed (within the 10year return period flood prone zone). Area B, is composed of both highlands and lowlands drained by a river basin of about 3,000 km2 . The densely populated (e.g., 100,000 citizens) urban area is located in the lower part of the basin. It includes the majority of schools, hospitals and other public services. About 30% of the residential, commercial areas and public services are located in the 20-year return period flood prone zone. Finally, area C is typical lowland with a large floodplain located in the lower part of a larger river basin (up to 4,000 km2 ). There are no permanent settlements in C but the area surrounds the main bridge of Anywhere City, calibrated to resist a 50-year flood. The area is characterized by seasonal agricultural activity and a recreation place where the annual festival of Anywhere City named “AnyDay” is taking place. The game takes place at the beginning of fall and starts on a Monday, five days before the AnyDay festival takes place with outside activities across the river and a big concert with famous singers close to the bridge area (area C in Fig. 9.1). The peak of the festival is planned for Saturday when participants are expected to reach the number of 10,000. Public officials are checking out the weather to ensure that Anywhere City’s traditional festival can happen in the best safety conditions so that participants can enjoy next weekend camping and celebrating in the region.

9 Anycare: A Serious Game to Evaluate …

9.2.2.2

109

Roles

ANYCaRE is designed to simulate one or more of the three main levels in the warning-system decision chain (i.e., 3 groups of roles to be played in the game): (i) Level 1: Weather Forecasters; (ii) Level 2: Emergency managers/Authorities in charge of civil protection; (iii) Level 3: General public and targeted users (private companies). Depending on the role group (e.g., forecasters, emergency managers, and citizens), we propose a series of realistic constrains and targets and we invite players to choose some action(s) from a given list of options. Every group should be preferably played with 10 to 15 players. Each player is attributed a certain sub-role (e.g., expert hydrologist, mayor, first responder) with the objective of either contributing to the collective decision making (Level 1 and 2) or deciding individually what is the best protective option for the type of role that is played (Level 3). The collective or individual decisions to be made refer to selecting a warning or emergency response activity based on the available information and related uncertainties. Following this first step, the players have to select (among some pre-established options) the best way to communicate those decisions to the targeted public. For example, forecasters group (Level 1) need to interpret the hazard model outputs to choose the level of warning to be issued and communicated to the emergency managers (Level 2) and the general public (Level 3). Then, emergency managers evaluate the situation and decide what to do based on the forecasters’ inputs and their own assessment of the level of exposure, potentially supported by impact-based products and crowd-sourced information. The members of the general public may decide for their own self-protective actions based on their personal constraints by considering or ignoring the information communicated by Level 1 and/or Level 2. Although interaction between the various decision levels is an interesting component in ANYCaRE, as mentioned above, the game is designed in a way that independent game sessions can be also played for each role group, separately. To do so, the GM or another auxiliary non-player person provides relevant information required as inputs for the decision-making in the group. As a beta testing, this study focuses on Level 2 exploring decision-making in a virtual Emergency Operation Center. The role of the emergency management group is to keep the population safe and ensure smooth execution of everyday life activities in Anywhere City while managing a given budget. In case of weather uncertainty, this is a challenging task including a set of dilemmas. Deciding to alert the population and push people to stop their daily activity to take protective measures in areas not hit by flooding might create unhappiness and loss of confidence in public authorities. People are looking forward to the AnyDay festival; potential cancelation requires careful consideration to preserve people’s wellness without risking their security. We assume that the organization of the festival itself represents an expensive investment of the municipality (e.g., 20,000 tokens). Every decision taken by the emergency management group has a cost either in terms of economic value (token), human safety or wellness. The objective of the group is to undertake emergency activities avoiding “costly” decisions that might prove to be unnecessary at the end of the

110

G. Terti et al.

game. Distinct decisions can be made for one or more of the three areas in Anywhere City (area A, B or C in Fig. 9.1).

9.2.2.3

Gaming Simulation

The period of concern in the game are the five days of the week (from Monday to Friday) preceding the day of the festival. The initial conditions in the city are defined by assigning to the group a certain amount of (i) citizens’ safety credits; (ii) citizens’ wellness credits; (iii) tokens (budget). On Monday and Tuesday, the GM provides and comments medium-range deterministic precipitation forecasts produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) and hydrological forecasts in order to get the players familiar with the products and slowly put them in the context. Each of the three following days represents one round of the game for which collective decisions are requested from the players. By using three rounds we allow the players to experience evolving hydro-meteorological facets and test different decision-support tools, which give more and more accurate information as it gets closer to the event occurrence. Each round is composed of two trials, one where only hydro-meteorological forecasts are available and the second where impact-based and crowdsourced information are provided. Each game day (round) the players receive updated probabilistic forecasts for precipitation (mm) and river discharge (m3 /s and the corresponding return period) as well as contextual information for each of the areas A, B, and C. These data mainly refer to flood early warning products released by the European Flood Awareness System (EFAS); the first operational European system monitoring and forecasting floods across Europe [23, 29]. As a second trial the players get new ANYWHERE products3 including improved probabilistic impact-based forecasts, risk assessments and ground observations from social media, and are given the opportunity to rethink and modify their decision if necessary. Collective decisions are recorded by filling up one worksheet (i.e., selecting from the proposed decision-reporting list) for the all group in each game day. The group reports its choices in the relevant “TRIAL” column on the worksheet at the end of each trial. To replicate time restrictions and pressures realized in real-world flash flood crisis decision-making, the players are given a limited time to provide responses in each trial (e.g., 8 min). Therefore, each round lasts about 16 min (or less if the players are fast to take decisions on a given game day) including the time that the GM presents a summary of the hydrometeorological situation to the group.

3A

catalog of products to be integrated in ANYWHERE Platform is available at http://anywhereh2020.eu/catalogue/.

9 Anycare: A Serious Game to Evaluate …

111

9.3 Game Testing: European and French Experiments The first implementation of ANYCaRE was carried out in the frame of ANYWHERE’s 2nd workshop in Helsinki (September, 2017). This workshop gathered 161 attendees (90 project partners belonging to the consortium and 71 from external organizations) including researchers, forecasters, civil protection and representatives of related companies in Europe to follow up ANYWHERE’s innovations and contribution to the response-to-weather-extremes era. A game session was organized with a group of sixteen players to compose the virtual Emergency Operation Center of Anywhere City (Fig. 9.2a). Among the players there were PhD students and researchers in weather-related hazards, developers and modelers, emergency managers and operational forecasters. Therefore, participants’ experience in weatherinduced risk modeling and management ranged from low (i.e., 10 years) providing a genuine diversity to be reflected in the game. A second experiment held in Le Puy en Velay (France) gathered 11 people mostly mayors or members of municipal councils and 3 territorial agents. Two of them had an experience of flash flood management in their own municipality. Each player was given a specific sub-role to act as representative of one of the following institutions: (i) hydro-meteorological services; (ii) first responder services, (iii) municipality; (iv) school services; (v) road services. Every player, therefore, had certain interests and aspects to address in the decision-making (e.g., weather

Fig. 9.2 Role-playing session of ANYCaRE in Helsinki (20 September 2017). Players’ a debate during the simulation, and b post-experiment debriefing post-it notes

112

G. Terti et al.

warnings, evacuation of places, festival organization, school closure, road closures). One additional person was designated to act as the leader of the Emergency Operation room. Based on the discussion and advices from the others, he/she had to come up with a collective decision to be implemented and communicated. The given list of options enabled the group to decide to: – Perform no action. The emergency group just follows the weather updates and keeps monitoring the situation. – Activate the Emergency Operation Center to coordinate the rescue services and operational forces. The group might take precautionary measures such as flood protection measures in specific areas, anticipation of school pick-up or school cancelation (area B), cancelation of the festival planned for the weekend (area C). – Activate the emergency plan to trigger evacuation of vulnerable places (e.g., campsites in area A) and/or special needs buildings (e.g., schools in area B) and close intersections (e.g., bridge in area C) prone to flooding (or already flooded). The group might take emergency measures such as evacuating and sheltering the population in one or more of the city areas. Based on the selected emergency activity, the group further agreed if they would provide some generalized advice for safety to the public (e.g., “If inside, move to higher floors”,) or if they would proceed to more detailed emergency orders in specific area(s) of Anywhere City (e.g., “Evacuate immediately”). To trigger debate rather than winning spirit in this test game, no specific penalties were assigned to the options listed in the worksheet. Instead, in the first round the GM highlighted the three-fold common goal of the emergency managers to: (i) insure citizens’ safety and prevent loss of life; (ii) prevent disturbances in social life, which make people unhappy and reduce wellness; (iii) maintain their budget (minimize their expenses for protective measures compared to the actual needs). In both experiments, the game included three rounds simulating decisions from Wednesday to Friday in ANYCaRE scenario (Fig. 9.3). In each round, the players received area-specific information so that they could make distinct safety choices adapted to the predicted hazard in each area of Anywhere City. Table 9.1 shows the total list of input products elaborated in each trial within the three game rounds. For each round, the players would have two trials of decision. The first one is based on existing hydro-meteorological products. The second one proposes additional decision-support tools including high-resolution precipitation maps, and flooding probabilistic forecasts. Sometimes, those products would also include the Rapid Impact Assessment layer that combines event-based hazard maps with exposure information to assess several categories of impacts such as affected population and damages [5]. Direct economic losses are computed combining the “Coordination of information on the environment” (Corine) map with flood hazard variables (i.e., flood extent and depths) and a set of damage functions derived for European countries. In addition to that, the extension of urban and agricultural areas affected is computed using the Corine Land Cover [5]. Last information presented to the players included artificial Tweets and posts from a hypothetical crowdsourcing system especially drawn for the Helsinki experiment.

9 Anycare: A Serious Game to Evaluate …

113

Fig. 9.3 Schematic summary of the gaming procedure (steps I to III) and the information provided to the players of ANYCaRE. Each of the three game rounds played in the experiment corresponds to a weekday before the AnyDay Festival that, according to the storyline, is held on Saturday. In the second trial of each round, the players receive additional products including high-resolution forecasts and impact-based inputs

In each round, the leader of the Emergency Operation room invited the role-players to express their opinion on the situation and decide what to do to insure citizens’ security given each actor’s constraints. Their highest challenge was to decide if they would cancel the festival planned for the weekend in area C or they would maintain the event and probably set up flood protection measures. Withdrawing such a big event, for which people had been prepared and were looking for, obviously would cause upheaval and would reduce their wellness. It this case, the municipality would also pay cancelation fees and other expenses with the ultimate goal to fend people from risk.

114

G. Terti et al.

Table 9.1 Products provided as input to ANYCaRE players in each round and trial played in the test experiments. The column «source of info» indicates which roles get to receive and comment the information to the rest of EOC group. Existing data sources are noted with acronyms: ECMWF (European Centre for Medium-Range Weather Forecasts), ERICHA (European Rainfall-InduCed Hazard Assessment), DWD (Deutscher Wetterdienst), COSMO (COnsortium for Small-scale MOdelling). The data presented virtual information adapted for the needs of the game

9.4 Game Results In the debriefing phase after the simulation, the game observers invited the players to provide feedback on both the experiment set up and the forecasting products presented in the game. The participants commented on either positive remarks or potential improvements related to those two aspects. Short post-it notes were gathered by the game observers and were placed in the corresponding category on a board to open the discussion (Fig. 9.3b).

9 Anycare: A Serious Game to Evaluate …

115

9.4.1 ANYWHERE Products Increase the Level of Confidence in the Chosen Decisions At the beginning of each round the GMs described the existing weather-related conditions and explained if and why the group decision taken in the previous game day was relevant or not to ensure safety with the minimum losses (i.e., wellness, budget). It appeared that the players did a relatively good job already from Trial 1 in the first two rounds by evacuating, for example, scouts from the campsite before the occurrence of severe flash flooding in area A, and anticipating school pick-up time when flash flood threatened area B. Supplementary radar-based high-resolution rainfall accumulation forecasts and bias-corrected discharge forecasts provided in Trial 2 did not necessarily lead to better decisions for emergency response. Though, this result may be subject to specific skills of the players acted as forecasters as well as the overall experience of the participants in this session. In Le Puy en Velay experiment, players were much less experienced with the hydro-meteorological products provided, decisions between trial 1 and 2 changed mainly thanks to products based on social media posts. Indeed, in the debriefing some players mentioned that detailed meteorological data are not trivial to non forecasting-specialists: ‘Difficult information for operators in emergency centres’; ‘some of the products are very difficult to interpret if you haven’t seen them before’. Yet, hydro-meteorological forecasting was observed to dominate the decision-making and related discussion compared to impact estimations. Further examination is required to conclude if the limited use of the new impact-based tools in the game was due to lack of players’ familiarity with the new products opposed to previously seen operational tools. Other potential explanations to be explored may include: (i) inadequate understanding on how to handle the new impact information during crisis; (ii) absence of trust in the new developments; (iii) incapability of the adopted visualizations to convey helpful information. On the other hand, in both Helsinki (Finland) test experiment and in Le Puy (France) the players seemed to largely rely on impact observations assumed to be reported through comments and pictures on social media. This was obviously the case in the third and last round of Helsinki experiment where the players changed totally their emergency decision after receiving a crowdsourced image showing the bridge blockage with wood and debris. In Helsinki, the payers shifted from “no action” to the set up of flood protection measures in area C, and finally, the cancelation of the festival in Trial 2. In Le Puy, the group finally decided in their second trial to evacuate the scouts from the campsite in round 2 after receiving a tweet from scouts requiring help, and to evacuate population from zone C where social media pictures showed the bridge jammed with debris in round 3. All the players agreed that online crowdsourcing tools might be a great provider of ground facts necessary to enhance situational awareness of authorities especially in cases of high hydro-meteorological uncertainties and forecasting failures. Future developments were suggested to geolocate the social media content and present that on a map to help emergency responders to clearly identify places where urgent action is needed.

116

G. Terti et al.

Although still under development, the examples of impact-based forecasting products included in Helsinki were found to reduce the overall uncertainty in the decisionmaking process. The players acknowledged that the variety of new products increased their sureness about specific emergency activities to be chosen and communicated in particular areas. In all the three rounds they rated their confidence to maximum when they passed to Trial 2. In Le Puy, even though the players did not think the new information increased their level of confidence in their decisions, they thought the information provided was very useful to better target their actions. Communication to the public was only a small part of the players’ duties in Helsinki experiment. To simplify the beta testing of the game the players did not have to compose an emergency message by themselves but they rather got some simplified examples to choose from. Therefore, choosing the relevant official emergency response and deciding on whether that should be communicated in one or more areas in Anywhere City was considered as the main responsibility of the emergency group. When the hypothetical crowdsourcing system indicated the bridge blockage in round 3, the emergency group commanded to inform the public immediately for the imminent risk. The players opted for specific guidance to the public indicating to stay away from the bridge and the festival area.

9.4.2 Is ANYCaRE Useful? During the feedback session after the Helsinki experiment, ANYCaRE players found that role-playing was a good approach to bring people from different institutions together to share knowledge and experience while examining new forecasting products potentialities. They also expressed that it was ‘very interesting to try roles different from the usual day-by-day experience’. In both Helsinki and Le Puy experiments and similarly to previous game studies, the personality of the player was observed to be an important factor during the game; with the more extrovert and talkative players to dominate the decision-making and the shyer ones to participate much less in the debate [13]. Shy players apposed their arguments though, when they felt the need to defend a specific priority related to their role. Also some players had difficulty to stay focused on their own role and tempted to lead decisions outside the domain of the responsibility they represented in the game. For instance, in Le Puy experiment, it was the case of some players who were mayors in the real life and who had been attributed a more technical role in the game. To improve this aspect, very detailed responsibilities and constrains are suggested to be assigned to each role at the beginning of the game allowing the players to set up a common strategy with the other persons playing the same role. The test experiments were considered as successful since the game was found ‘to be representative of the reality of flood crisis management’ by players who had the experience of such situation, it was ‘very fun’ and ‘clearly demonstrated the benefit of certain ANYWHERE data’. It is important for participants to feel that the game experience was playful, the rules were understandable and the learning procedure

9 Anycare: A Serious Game to Evaluate …

117

was attractive [4, 30]. The players recognized ANYCaRE scenario as very realistic and presented a strong commitment to the storytelling. In the debriefing session, they mentioned that the simulation was motivating and ‘kept the stress’, and the whole set up was a ‘very representative group exercise’. That made participants to appreciate the game and recommend it for play in future events. Players in Helsinki introduced further aptitudes of ANYCaRE; emphasizing educational scopes such as coaching emergency services in order to sharpen their emergency agility and alertness before the crisis strikes: ‘The game can be used in my organization to test the emergency response’. Although the current level of complexity in ANYCaRE made the game to be stimulating for the players, it was observed that the given hydrometeorological information were sometimes difficult to digest in a short time from participants without forecasting expertise. Obviously, more weight should be given in the presentation and explanation of the elaborated data before the simulation starts. The game designers consider various ways to ease the gaming process in next game sittings: (i) distribute demanding roles according to the real-life competencies and players’ background (e.g., give forecasting responsibilities to players with forecast experience); (ii) introduce Level 1-simulations in which the forecasters’ group will work separately to prepare and deliver forecasts to the emergency group. Smaller groups with up to 13 players are also perceived as more controllable and efficient to prompt player’s engagement in the game world.

9.5 Discussion and Conclusion This paper presents a role-playing game designed to investigate crisis decisionmaking in weather-related risks. ANYCaRE allows us to explore how decisionmakers and stakeholders interact with scientific and operational outputs. The main target of ANYWHERE European project is to provide civil protection authorities and regional institutions across Europe with a supportive tool to better anticipate and respond to extreme and high-impact weather and climate events. Therefore, Level 2 (i.e., emergency group) was considered as the most relevant one to be simulated in ANYCaRE’s first implementation. First results from test experiments show that the game aroused the interest and enthusiasm of participants and offered to the players a protected environment to try-out emergency actions without facing true risk for human life. This participatory technique set a playful and collaborative atmosphere between the project partners and stakeholders generating debate on appropriate emergency decisions. The players affirmed that the simulation adequately reflected situations found in the real world and facilitated their involvement in the storytelling. At the final debriefing step, participants were encouraged to exchange knowledge, thoughts and insights on their capability or difficulty to decide and communicate their action based on the available information and given constrains. In a broad view, the players complied well with the scenario requests. The gaming process was diagnosed as having limitations related to the time needed for the players to understand the concept and get ready to play

118

G. Terti et al.

as well as the processing of complex data probably unfamiliar to some participants. In this sense, a well-trained game master is necessary to clearly introduce material appropriate to the players and guide them through the game. The biggest challenge identified in the test game is to engage players equally and empower their active presence in the decision-making. Flooding was selected as a prototype case since it is a common hazard under consideration in the four pilot sites of ANYWHERE (i.e., Canton of Bern in Switzerland, Catalonia in Spain, South Savo in Finland, Genoa in Italy). Though, applications to other weather-induced risks such as wildfires were encouraged by the participants in Helsinki and are considered for the future. Rather than a single tabletop role-playing, we vision ANYCaRE as a broad experiment campaign that will encompass various versions of games. A series of future expansions is considered to: (i) adjust scenarios to other weather hazards including multi-hazard cases and complex cascading effects commonly challenging European cities; (ii) test additional models and technological innovations (e.g., crowdsourcing tools, dialog systems, internet-based apps) to be developed as decision-support products; (iii) establish other formats of serious gaming such as online or board games to attract different audience (e.g., stakeholders, general public, pupils) and subsequently, enlarge the amount and variety of feedbacks. ANYCaRE is developed in accordance with ANYWHERE’s releases to provide a feedback loop for further improvements to developers. Although the experiment is basically created to value the contribution of new weather-related decision-support outputs, role-playing or serious gaming offers a promising setting to address multiple communicational or educational needs. Game implementations of ANYCaRE may be used in the future to train experts and civil protection services on the use of modern decision-support tools. Furthermore, experiments may engage either experts or general public to draw more detailed conclusions on the effectiveness of forecast visualizations and delivered warning and emergency messages (i.e., content, structure and format) in terms of comprehension and mobilization of action. Such knowledge is prerequisite for the anticipation of effective crisis communication strategies and relevant emergency responses to prevailing weather threats. Acknowledgements This project has received funding from the European Union’s Horizon 2020 research and innovation programme (H2020-DRS-1-2015) under grant agreement No 700099.

References 1. Bowman SL (2010) The functions of role-playing games: how participants create community, solve problems and explore identity. McFarland & Co, Jefferson 2. Cover JAG (2005) tabletop role-playing games: perspectives from narrative, game, and rhetorical theory. J Chem Inf Model 53(9):1689–1699. https://doi.org/10.1017/ CBO9781107415324.004

9 Anycare: A Serious Game to Evaluate …

119

3. Creutin JD, Borga M, Lutoff C, Scolobig A, Ruin I, Créton-Cazanave L (2009) Catchment dynamics and social response during flash floods: the potential of radar rainfall monitoring for warning procedures. Meteorol Appl 16(1):115–125. https://doi.org/10.1002/met.128 4. Dieleman H, Huisingh D (2006) Games by which to learn and teach about sustainable development: exploring the relevance of games and experiential learning for sustainability. J Clean Prod 14(9–11):837–847. https://doi.org/10.1016/j.jclepro.2005.11.031 5. Dottori F, Kalas M, Salamon P, Bianchi A, Alfieri L, Feyen L (2017) An operational procedure for rapid flood risk assessment in Europe. Nat Hazards Earth Syst Sci 17(7):1111 6. Heinsoo R, Collins A, Wyatt J (2008) Dungeons & dragons player’s handbook: arcane, divine, and martial heroes: roleplaying game core rules. Wizards of the Coast 7. Hémon D, Jougla E (2004) The heat wave in France in August 2003, Epidemiology and Public Health/Revue d’Epidémiologie et de. Santé Publique 52(1):3–5 8. Huyakorn P, Denpaiboon C, Kanegae H, Management ACF (2012) Role-play gaming simulation for flood management on cultural heritage: a case study of Ayutthaya Historic City. World Acad Sci 6(3):111–116 9. Jacks E, Davidson J, Wai HG (2010) Guidelines on early warning systems and application of nowcasting and warning operations. World Meteorological Association, p 22 10. Kolb DA (1984) Experiential learning: experience as the source of learning and development. Prentice Hall, Inc, pp 20–38 (1984). https://doi.org/10.1016/b978-0-7506-7223-8.50017-4 11. Luther J, Hainsworth A, Tang X, Harding J, Torres J, Fanchiotti M (2017) World Meteorological Organization (WMO)—concerted international efforts for advancing multi-hazard early warning systems. In: Sassa K, Mikoš M, Yin Y (eds) WLF 2017: advancing culture of living with landslides. Springer, Cham, pp 129–141 12. Mayer R, Sprenger R, Tait M, Macfarlane K, Wyse P (2017) Portugal’s biggest wildfire: ‘we all thought we were going to die’—video, The Guardian. https://www.theguardian.com/world/ video/2017/sep/19/portugals-biggest-wildfire-we-all-thought-we-were-going-to-die-video. Accessed 30 Sept 2017 13. Mossoux S, Delcamp A, Poppe S, Michellier C, Canters F, Kervyn M (2016) Hazagora: Will you survive the next disaster?-a serious game to raise awareness about geohazards and disaster risk reduction. Nat Hazards Earth Syst Sci 16(1):135–147. https://doi.org/10.5194/nhess-16135-2016 14. Müller A, Gebhardt O, Kuhlicke C, Capone F, Terti G, Ruin I, Gebhardt O (2017) Deliverable 1.2: Report on needs and requirements from the users Including advises on how to address ethical, legal and environmental issues 15. NRC (ed) (2013) Public response to alerts and warnings using social media: report of a workshop on current knowledge and research gaps. National Academy of Sciences, Washington, D.C. 16. Pasin F, Giroux H (2011) The impact of a simulation game on operations management education. Comput Educ. Pergamon 57(1):1240–1254. https://doi.org/10.1016/j.compedu.2010.12.006 17. Pereira G, Prada R, Paiva A (2014) Disaster prevention social awareness—the Stop Disasters! case study. In: Proceedings of VS-GAMES—6th international conference on games and virtual worlds for serious applications. IEEE, pp 115–122. https://doi.org/10.1109/vs-games.2014. 7012155 18. Ruin I, Creutin J-D, Anquetin S, Gruntfest E, Lutoff C (2009) Human vulnerability to flash floods: addressing physical exposure and behavioral questions. Res Pract, Flood Risk Manag, pp 1005–1012 19. Ruin I, Creutin JD, Anquetin S, Lutoff C (2008) Human exposure to flash floods—Relation between flood parameters and human vulnerability during a storm of September 2002 in Southern France. J Hydrol 361(1–2):199–213. https://doi.org/10.1016/j.jhydrol.2008.07.044 20. Ruin I et al (2014) Social and hydrological responses to extreme precipitations: an interdisciplinary strategy for postflood investigation. Weather Clim Soc 6(1):135–153. https://doi.org/10. 1175/WCAS-D-13-00009.1 21. Salas E, Wildman JL, Piccolo RF (2009) Using simulation-based training to enhance. Acad Manag Learn Educ 8(4):559–573

120

G. Terti et al.

22. Saurí D, Milego R, Aleix C, Ripoll A, Kleeschulte S (2003) Mapping the impact of recent natural disasters and technological accidents in Europe. Environmen. Copenhagen, Denmark: European Environment Agency, Office for official publications of the European Communities 23. Smith P, Pappenberger F, Wetterhall FJT, Krzeminski B, Salamon P, Muraro D, Kalas M, Baugh C (2016) On the operational implementation of the European Flood Awareness System (EFAS), Flood Forecasting: A Global Perspective 24. Susi T, Johannesson M, Backlund P (2007) Serious games—an overview. Elearning 73(10):28 http://doi.org/10.1.1.105.7828 25. Terti G, Ruin I, Anquetin S, Gourley JJ (2015) Dynamic vulnerability factors for impactbased flash flood prediction. Nat Hazards 79(3):1481–1497. https://doi.org/10.1007/s11069015-1910-8 26. Terti G, Ruin I, Anquetin S, Gourley JJ (2017) A situation-based analysis of flash flood fatalities in the United States. Bull Am Meteorol Soc 98(2):333–345 27. Terti G, Ruin I, Gourley JJ, Kirstetter P, Flaming Z, Blanchet J, Arthur Z, Anquetin S (2019) Towards probabilistic prediction of flash flood human impacts. Risk Anal 39(1):140–161. https://doi.org/10.1111/risa.12921 28. Terti G, Ruin I, Kalas M, Láng I, Cangròs i Alonso A, Sabbatini T, Lorini V (2019) ANYCaRE: a role-playing game to investigate crisis decision-making and communication challenges in weather-related hazards. Nat Hazards Earth Syst Sci 19:507–533. https://doi.org/10.5194/ nhess-19-507-2019 29. Thielen J, Bartholmes J, Ramos M-H, de Roo A (2008) The European flood alert system—Part 1: concept and development. Hydrol Earth Syst Sci Dis 5(1):257–287. https://doi.org/10.5194/ hessd-5-257-2008 30. Turkay S, Adinolf S (2012) What do players (think they) learn in games? Procedia Soc Behav Sci 46:3345–3349. https://doi.org/10.1016/j.sbspro.2012.06.064 31. WMO (2015) WMO Guidelines on Multi-Hazard Impact-based Forecast and Warning Services, p. 23

Chapter 10

From Catstrophe to Resilience Jelena Batica and Philippe Gourbesville

Abstract How do we cope with catastrophe events using existing models? Are we going a step forward and focus on sustainable solutions taking into account a more holistic approach and resilience concept? Natural disasters are characterized by different patterns in recent years. Urban communities have developed assets too vulnerable to disasters and now they are about to have high damage. There is a need for a new framework that takes into consideration models, crisis management, new holistic concepts along with social components. The concept of resilience to natural disasters, preferably resilience to floods will be the main subject of this paper. Keywords Natural · Disasters · Floods · Management · Natural based solutions · Urban systems

Nomenclature FRI DRR UNISDR EM-DAT

- Flood Resilience Index - Disaster Risk reduction - United Nations’ International Strategy for Disaster Reduction - International Disaster Data Base

J. Batica (B) · P. Gourbesville Université Cote d’Azur, Polytech Lab, Polytech Nice Sophia, 930, route des Colles, 06903 Sophia Antipolis, France e-mail: [email protected] P. Gourbesville e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_10

121

122

J. Batica and P. Gourbesville

10.1 Introduction As a major adverse event, resulting from natural processes, e.g. floods present a natural disaster. Measuring the number of disasters per type for the period from 1998 to 2017, floods have more than 43.4% and 3143 event in total (EM-DAT). Commonly used term disaster and catastrophe draw attention to its meaning. The disaster is caused by hazards (natural or technological) while catastrophe is a disaster with tragic outcome of a personal or public situation. In general these two terms are used equally, so in the paper, the disaster will be the adopted term. Analysing the different concepts related to risks focusing on how to manage pre and post event, what actions to take into account? Disasters become more complex nowadays. The impacts that disasters pose to society along with existing fragility and vulnerability of people and built environment are bigger. Reducing disaster risks, Disaster Risk Reduction approaches propose systematic analysis of hazard reduction, lessened vulnerability of people and built environment, and control exposure. Paper analyses components of disaster risk management and resilience as one of the key concept. How the resilience is seen and how the methodology is developed to be applicable to urban systems taking into accounts both local community and built environment. The method is applied on nine case studies and the results are comparable focusing on different dimensions.

10.2 Disaster Risk Management By managing disasters the understanding is the key: what are the components of the risk, the drivers of the risk and key concepts. Floods in urban systems cause disruption on many levels of functioning and ask for fast reaction. With disaster, especially in urban areas into the play comes a risk. It is often recognized as a consequence of interaction between three main elements: hazard, vulnerability and exposure. It is very important to highlight that all three elements need to be into play in order to have risk. Natural events (here we are focused on floods) are hazards when they have potential to threaten people and urban systems. Hazards are creating disasters. Clarity about disaster classification is crucial. Classification is therefore proposed to disaggregate groups, types and sub-types. Clearly, the analysis becomes more precise and provides better focus on particular disaster. There are two generic disaster groups: natural and technological. Floods, as natural disaster, are listed as main type of hydrological hazards [2]. The disaster sub-types are listed with respect to different flood types. As one of focal point in this paper, floods that occur nowadays differ to its patterns, mainly duration and frequency. Exposure element comes into play when people, infrastructure, tangible assets are in hazard prone area. If there is no exposure in flood rone area, there is no risk.

10 From Catstrophe to Resilience

123

The extent to which exposed people and economic assets are at risk is generally determined how vulnerable they are. Drives for exposure is concentration of people and assets usually in cities. This is driven by rapid urbanization, population growth and economic development. Hazard prone areas such as coastal or riparian areas are often places with high concentration of people and tangible assets. This is increasing risk to dangerous levels. Vulnerability presents a pre-event characteristic of a system that has a potential to harm. Vulnerability is in a function of exposure or sensitivity of a system to disturbance. This is explained through answer on question who or what is at risk? Vulnerability defines the conditions determined with physical, social, economic, or environmental factors or processes which are increasing the weakness of community to the impact of hazard [8]. Specific definition of vulnerability to flood events is given in Balica [1] as the extent to which a system is susceptible to floods due to exposure, a perturbation, in conjunction with its ability (or inability) to cope, recover, or basically adapt. Vulnerability is a human dimension of disasters and it the result of the range of the economic, social, cultural, institutional, political and psychological factors that shape people lives and the environment that they are [9]. Vulnerability is very complex and relates to different factors: physical, social, economical and environmental. It is also depending on cultural, political, institutional, natural dimension, everything that is creating their living environment. There are many risk drivers. The common one is climate change, environmental degradation, economic development, existing poverty and inequality and non proper governance. There are many ways how climate change is increasing disaster risk. By increasing global temperature the agricultural sector is directly hit. The sea level rise is dictating increase hazards in low land areas. Also, change in geographic distribution of weather related hazards. Decreasing in resilience is also triggered by climate change and directly hits the poorer communities and developing countries. Ecological degradation is both driver for disaster risk and its consequence. It is occurring when environment is not able to meet social and ecological needs. Weak governance stands as one key driver for disaster risk. This element is often linked with other risk drivers such as poverty, poor planned urban development and globalized economic development. The key concepts of disaster risk are defined through: capacity, deterministic and probabilistic risk, loses (both direct and indirect), disaster risk reduction and risk management and resilience. Capacity refers to all strengths attributes and resources available within the community, organization to manage reduce disaster risk and increase resilience. Within the concept of deterministic and probabilistic risk the analysis is focused on single or multiple scenarios. Probabilistic risk is simulating future disasters based on scientific evidence. These risk assessment as a result is solving the problem posed by the limits of historical data. The deterministic risk models are treating the event as finite. Within this approach the scenarios are modelled with known input values and out values are observed. Disaster risk management focuses on managing risk not disaster. Focusing on multi sector approach, people-centred, building resilience, increasing capacity, creating the culture of prevention and resilience the disaster risk management includes

124

J. Batica and P. Gourbesville

the huge palette of strategies. The available strategies are focused on avoidance and construction of new risks, addressing risks. The effective disaster risk reduction is combination of top-down institutional changes and strategies and bottom-up local and community based approaches.

10.3 Resilience Framework Resilience is a one of the main concepts of disaster risks. As a term, is often left open to debate and doesn’t have a general or consensual definition although it is increasingly used in integrated urban drainage management, (Ashley et al. 2012), [5– 8, 11]. The diverse interpretations of resilience reflect the complexity of this concept and made it ‘difficult’ in implementation of integrated urban drainage management. From a general perspective, resilience represents the capacity of an urban system or community exposed to hazard to adapt by resisting or changing in order to reach an acceptable level of functioning, organization and structure [10]. By narrowing down the resilience the specified resilience comes to focus. This resilience deals with following principle: “of what to what” [3]. It can be defined by identifying what system attributes are to be resilient, and to what kind of disturbances. Specified resilience in the context of Integrated Urban Drainage Management (IUDM) has often been defined in a restricted sense to express the ability of the whole system to recover from the reaction of flood waves [5–7]. Suitable definition for resilience adopted for the research in this paper is proposed by the United Nations’ International Strategy for Disaster Reduction (UNISDR). In the context of urban flooding, resilience can be defined as follows: “Resilience is the capacity of a system, community or society potentially exposed to hazards to adapt, by resisting or changing in order to reach and maintain an acceptable level of functioning and structure. This is determined by the degree to which the social system is capable of organising itself to increase this capacity for learning from past disasters for better future protection and to improve risk reduction measures” [10]. From this wide sense the concept provides a perhaps more suitable background framework to develop and assess integrated approaches to flood risk management. Resilience is therefore specified here in respect to the broader social–ecological context as the capacity of the system to absorb flood waves in annual variability, and to reorganize while undergoing change in flood wave frequency and severity in the long term, so as to enable it to function normally. The resilience approach is aiming to prevent the urban system to move from an undesirable state from which it is not possible to recover from flood impact to functional state. These preventions go in following directions (Fig. 10.1): 1. Adjusting the thresholds of a system with respect to changes in response to flood waves; 2. Defining the level to which system is capable of self organizing;

10 From Catstrophe to Resilience

carrying capacity Maximum tolerable damage

125

vulnerability

resilience

Measure and assess carrying capacity of an urban system

Fig. 10.1 Carrying capacity, vulnerability and resilience

Fig. 10.2 Defined elements for flood risk management cycle with transition to resilience concept - 5R concept [1]

3. Define the level to which the system is able to build and increase capacity for learning and adaptation. This defines resilience thinking, a different point of view for guiding and organizing urban systems. Defined terminology of vulnerability and resilience stands as an important element in the analysis of urban areas and their existing flood risks but there should be a distinction between the flood vulnerability and resilience of people on one side and the urban structure on the other side. The resilient urban systems and urban communities should have ability to accept, resist, recover and learn from the events. Capacity of urban systems and communities is improved in each part of the flood risk management cycle. It covers actions related to preparedness, response and recovery. Within this research the five elements of flood risk management are developed: Reflect, Relief, Resist, Response, and Recovery. These new elements explain as follow (Figs. 10.2, 10.3 and 10.4): Relief phase

– Defined as ‘a buffer’ where existing structures and urban functions accept floodwater (green areas, different playgrounds, etc) using its own capacities. Implementation of physical, technical, non-structural and procedural measures relates to the concept “living with floods”.

126

J. Batica and P. Gourbesville

Fig. 10.3 Mapping of the city according to urban functions and services

Fig. 10.4 FRI diagram assessment for the different scales

Resist phase

– Reduction of flood risk if possible. This is in direct connection with existing threshold capacity. Under this phase, the activities relay on the existing structural (done before flood event). Limiting flood damage and easing recovery by planning and building adaptation, infrastructure, surfaces and economic activity relate to the concept of resistance.

10 From Catstrophe to Resilience

127

Response phase – Focus is on measures taken during the flood related to crisis management. At this time, flood impact is reduced by implementation of physical, technical, non-structural and procedural measures relates to the concept “living with floods”. Recovery phase – Take into account activities that provide support by developing capacity building in communities enable to cope with the impacts after flooding events, cleaning and rebuilding. Reflect phase – Here, the focus is on increasing awareness and adaptive capacity, learning from past event and/or preparation for an uncertain future. These activities enhance awareness and engagement in all aspects of flood risk. It is influencing the managing at the policy level (politicians/decision makers), professionals (of the involved authorities and elsewhere) and at the public participation (people, companies, developers, insurance companies) Presented elements introduce resilience concept into flood risk management cycle. The elements are structured into different phases starting from striking the flood through whole duration of event including post flood period. The last element is very important since it focuses on: learning from event, rising awareness, alerting and engaging decision makers and key stakeholders.

10.3.1 Flood Resilience Assessment Assessment focuses on analysis of urban system. The system components are mapped with respect to different function in the system. Also the system is scaled into: parcel, block, district and city (the whole system). The methodology focuses on mapping the components of urban system for improving analysis and focus on critical elements. Mapping of urban system is needed in order to achieve a balanced study of flood resilience. In addition, scaling of urban system allows being able to recognize main urban patterns. Mapping and scaling of urban system brakes down the structure of urban pattern into physical components and system requirements [4]. Functional analysis of system will allow evaluating flood resilience of each system element. Physical components of the systems are urban functions and services: buildings, streets, parks, water distribution network, shops, industrial buildings, electricity network, religion areas, etc. Some of them represent assets that the city needs to have in order to perform while others provide connections between different system components. Urban functions of a city define physical components that urban system needs to provide as fundamental needs to residents. The physical component of a city has spatial extension and the expression is through units (m2 ). There are nine main urban functions, which urban system needs to have in order to fulfil requirements related to integral need provided to residents. The urban functions are listed as follows: • houses (individual or collective), • educational areas (for local and non-local education services),

128

J. Batica and P. Gourbesville

• food (area for food storage), • work areas (areas for industry and areas for non-industrial activities), • areas established for location of police, fire brigade and rescue services (on local level), health areas (hospitals on local and non-local level), • areas for leisure and tourism (on local and non-local level) and • areas for religion activities (churches and cemeteries). The city services give connectivity between physical components. Services in the city gives functionality to urban features (e.g. the function of a house is to provide space for living).

10.3.2 Flood Resilience Index (FRI) The need for index introduction comes from the intent to have a compact presentation of resilience. The index is represented by number and describes the level of flood resilience in analysed area and for certain flood characteristics with value from 0–5. This way of presentation comes from considering resilience as a characteristic by definition and represents ability to accept a disturbance up to some level. This ability is defined up to the level where the system is able to organize itself and preserve the structure and function. Reflected in urban systems this means that resilience is defined up to the level that urban structure and urban community are able to accept disturbance, preserve the ‘level of functioning’, organize and recover from it. To measure flood resilience in urban spaces, the index takes into account several indicators based on the notions of the five R’s of resilience regarding flood management: Reflect, Relief, Resist, Response, and Recovery. There is a different process of index evaluation based on scale dependency. For the small scales: (i) parcel/building and (ii) block scale, index evaluation focuses on the urban function. In the process of calculating the FRI, external and internal requirement are evaluated. The method for macro scale: (i) district and (ii) city scale takes into account whole system through five dimensions.

10.3.2.1

Micro Scale Assessment

The evaluation on micro scale takes into account urban functions and services as main elements. The evaluation of FRI for property scale where focus is on urban function and its structure and level of functioning during flooding conditions presents a union of all external and internal requirements presented in table below (Table 10.1). Each mapped property is evaluated and presented with flood resilience level.

10 From Catstrophe to Resilience

129

Table 10.1 FRI evaluation at property scale Requirements for urban function

Availability level (0–5)

FRI (property scale)

EXTERNAL SERVICES (re )

0, 1, 2, 3, 4, 5

F R Ibuilding =

Energy

0, 1, 2, 3, 4, 5

Water

0, 1, 2, 3, 4, 5

Waste

0, 1, 2, 3, 4, 5

Communication

0, 1, 2, 3, 4, 5

Transport

0, 1, 2, 3, 4, 5

5

i=1 rei ×wi+

3

i=1 rii ×wi

8

INTERNAL SERVICES (ri ) Food availability

0, 1, 2, 3, 4, 5

Occupation of urban function

0, 1, 2, 3, 4, 5

Access to urban function

0, 1, 2, 3, 4, 5

Where re is an external service ri is an internal service wi is assigned weight

Fig. 10.5 Schematic presentation of FRI evaluation of city/district scale

10.3.2.2

Macro Scale Assessment

The assessment of Flood Resilience Index on the parcel and block scale is focused on the building (urban function) while for the bigger scale (city/district) the evaluation of Flood Resilience Index is done through five dimensions (natural, physical, economical, social and institutional). These five dimensions describe the physical and social attributes of urban system. One of the main objective criteria was to evaluate urban community and its ability to accept certain disturbance and recover from it. The scheme on Fig. 10.5 presents the flow path for evaluation of FRI for macro scale. The calculation is done using the developed matrix with 91 indicators that are corresponding to different element of flood risk management cycle following already developed 5R concept. Evaluation continues with allocation each indicator to particular dimension (natural, physical, economic, social and institutional).

130

J. Batica and P. Gourbesville

Fig. 10.6 Schematic presentation of FRI evaluation of city/district scale

The overall resilience assessment is developed to fit the urban systems and provide better guidance to disaster risk reduction. The methodology is developed for flood hazard and takes into account different scales and different levels of data (Figs. 10.6 and 10.7).

10.4 Examples In this section the results from different case studies will be presented. In eight years log period, method has been applied on nine cases study areas located in Europe and Asia. Focusing on macro scales the results are eligible for comparison and analysis. They offer key stakeholders the perspective on natural, physical, institutional, economical, and social dimensions. This scale is chosen because of data availability (Table 10.2). Figure above shows the FRI values calculated for different dimensions. The values show the level of flood resilience in dimensions based on case study analysis. The calculation is done using developed matrix with 91 indicators that were evaluated for particular case study taking into account natural, social, economic, institutional and physical settings.

10 From Catstrophe to Resilience

131

Natural

5 4 3 2

Physical

Social Ayutthaya

1

Barcelona

0

Beijing Chatellailon-Plage Genoa Hamburg Nice Taipei Economic

Institutional

Fig. 10.7 Radar chart presentation of FRI values for nine case studies

Table 10.2 Case studies with calculated FRI for macro scales Case study

Country

Dimensions

FRI

Natural

Social

Economic

Institutional

Physical

Ayutthaya

Thailand

3.00

2.76

3.57

1.75

3.44

2.90

Barcelona

Spain

3.50

3.25

3.81

3.31

3.62

3.49

Beijing

China

2.50

2.55

3.00

1.83

2.72

2.52 3.04

Chatellailon-Plage

France

2.50

3.65

3.29

3.62

2.12

Genoa

Italy

1.88

3.39

2.38

2.00

1.76

2.28

Hamburg

Germany

4.00

3.87

3.96

3.71

4.20

3.95

Nice

France

3.50

3.17

3.65

3.70

3.25

3.45

Taipei

Taiwan

3.50

3.36

1.70

3.61

3.06

3.05

Presented tool is practical for different stakeholders where they have a possibility to ‘play’ with the different indicators exploring how to increase resilience. Also, there is a possibility to create different scenarios with different strategies and make comparison.

10.5 Conclusions The Flood Resilience Index (FRI) represents a tool for stakeholders and decision makers. The concept brings a new philosophy to urban systems, ‘living with floods’.

132

J. Batica and P. Gourbesville

Presented approach transforms the existing structure of urban system and creates a system that is accepting the water with minimal damages, system that is able to recover in a minimum time frame and system that is able to have a certain level of functioning during the flood. The resilience of a system could be improved by using diverse regulations such as institutional, urban planning and design, architectural design, public participation, financial stimulation, etc. The importance is in the possibility to use experience from flood resilience urban systems and avoid huge flood damages and dysfunction. The developing urban systems can find a good practice and good paths towards flood resiliency without reaching a low level of functioning. As presented the framework is applicable on different institutional setting in different urban systems. Future development of presented methodology focuses on application to different hazards such as fire, landslides, droughts as well as technological risk. The framework is potentially applicable to any urban system on any geographic scale. Connections and dependences between main city elements and natural hazards (in this case urban flooding process) have to be defined. With its implementation, social, economical, political and cultural relations within city will be more visible and better established. The approach should uncover the role of physical components of urban system and population in relation to urban flooding processes. A further strategy focuses on simulation of community losses and recovery measures. As a major challenge that faces urban systems nowadays, the research on resilience prioritizes in following years. Acknowledgements The research work presented here has received funding from the European Union Horizon 2020 Programme under Grant Agreement n° 776866 for the research project RECONECT (Regenarating ECOsystems with Naturebased solutions for hydro-meteorological risk rEduCTion). The research work presented here has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under Grant agreement n° 603663 for the research project PEARL (Preparing for Extreme And Rare events in coastaL regions). The research and its conclusions reflect only the views of the authors and the European Union is not liable for any use that may be made of the information contained herein. Research on the CORFU (Collaborative research on flood resilience in urban areas) project was funded by the European Commission through Framework Programme 7, Grant Number 244047.

References 1. Batica J, Gourbesville P, Hu F-Yu (2013) Methodology for flood resilience index. In: International Conference on Flood Resilience Experiences in Asia and Europe—ICFR, Exeter, United Kingdom 2. Below R, et al (2011) Annual disaster statistical review 2010. Centre for Research on the Epidemiology of Disasters 3. Carpenter S, Walker B, Anderies JM, Abel N (2001) From metaphor to measurement: resilience of what to what? Ecosystems 4:765–781 4. Daniell K, et al (2005) Integrated urban system modelling: methodology and case study using multi-agent systems

10 From Catstrophe to Resilience

133

5. De Bruijn KM (2004) Resilience and flood risk management. Water Pol 2004(6):53–66 6. Klein RJT, Smit MJ, Goosen H, Hulsbergen CH (1998) Resilience and vulnerability: coastal dynamics or dutch dikes? Geograph J 164(3):259–268 7. Sayers P, Gouldby BP, Simm JD, Meadowcroft I, Hall J (2003) Risk, Performance and uncertainty in flood and coastal defence—a review. R&D Technical report FD2302/TR1 8. Sendzimir J, Magnuszewski P, Flachner Z, Balogh P, Molnar G, Sarvari A, Nagy Z (2007) Assessing the resilience of a river management regime: informal learning in a shadow network in the Tisza River Basin. Ecol Soc 13(1):11. http://www.ecologyandsociety.org/vol13/iss1/ art11/ 9. Twigg J (2004) Disaster risk reduction: mitigation and preparedness in development and emergency programming. Overseas Development Institute (ODI) 10. UN/ISDR (2004) United nations, international strategy for disaster reduction. http://www. unisdr.org/we/inform/terminology 11. Vis M, Klijn F, de Bruijn KM, van Buuren M (2003) Resilience strategies for flood risk management in the Netherlands. Int J River Basin Manag 2003/1:33–40. vom Februar 1962, School authority, 1962 (in German)

Chapter 11

Marine Dispersion Modelling and Expertise Tools for Accidental Radiological Contamination of French Coasts Céline Duffa Abstract IRSN develop tools to manage any marine contamination of French coastal areas. In case of radioactive accidental marine contamination, we should be able to evaluate and anticipate the post-accidental situation: contaminated areas localisation and contamination levels, and possible consequences. Many sites should be considered for potential source terms into the sea: Coastal Nuclear Facilities, Military Harbours as homeports of nuclear powered vessels, and different river mouths that could be contaminated by any accidental release from a nuclear power plants situated upstream. The modelling tool, STERNE (Simulation du Transport et du transfert d’Eléments Radioactifs dans l’environNEment marin), simulates radionuclide dispersion and contamination of water and marine species, incorporating spatial and temporal processes. To operate it, 3D hydrodynamic data should be provided routinely to IRSN crisis center. Different possible radiological source terms can be taken into account: direct liquid releases, atmospheric depositions or river inputs of radionuclides. STERNE calculates radionuclides transport using advection and diffusion equations offline from hydrodynamic calculation. Radioecological model based on dynamic transfer equation to evaluate concentrations in marine organisms is also implemented. Needed radioecological parameters (concentration factors and single or multicomponent biological half-lives) have been compiled for some important radionuclides and for generic marine species (fishes, molluscs, crustaceans). Dispersion and transfer calculations are carried out simultaneously on a 3D grid. Results, available as netcdf files can be represented on maps, with possibility to follows temporal and spatial evolution. Post-treatment and representation are then possible. In parallel, marine environment stakes and characteristics are compiled for the different sites, identifying potential stakes for human protection like aquaculture areas, beaches, or industrial water intakes, and ecological richness. This information will be used to facilitate decision-making during an emergency and could serve as a basis for post-accident sampling strategies leading to realistic environmental impact assessment. C. Duffa (B) Institut de Radioprotection et Sûreté Nucléaire (IRSN), PSE-ENV/SRTE/LRTA, Centre de Cadarache, 131150 St Paul Lez Durance Cedex, France e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_11

135

136

C. Duffa

Keywords Crisis management · Accidental radioactive release · Marine contamination · Dispersion model

11.1 Introduction The nuclear accident at Fukushima in 2011 resulted in the largest ever accidental release of artificial radionuclides in coastal waters. Radioactive material from Fukushima nuclear power plant entered the marine environment directly through weakly contaminated radioactive liquid release and indirectly from the deposition onto the ocean surface of material released to the atmosphere and dispersed over the ocean (Fig. 11.1). Once in the water column, considering their own physico-chemical properties, behaviour of radionuclides is largely governed by ocean processes (transport and dispersion, transfer to sediments and living organisms, see Fig. 11.2, [10]). In Japan, following the accident, numerous seawater samples measurements were carried out in the ocean by the government and agencies. The large amount of data provided a relatively good understanding of the spatial and temporal distribution in the ocean of contamination. This huge effort of sampling was necessary as no marine

Fig. 11.1 Fate of radionuclides in the marine environment (adapted from Hervé Bouilly, IRSN)

11 Marine Dispersion Modelling and Expertise Tools …

137

Fig. 11.2 Localization of the French nuclear installations

dispersion modelling results were provided in the first crisis period. The first modelling results were provided about 10 days after the accident by Toulouse University1 at the request of the International Atomic Energy Agency (IAEA). After that, many other models have been implemented, with different resolutions and using different forcings [7, 8]. Their agreement levels with measurement results were essentially dependent of the good representation of the local hydrodynamics [6]. The French Institute for Radiological Protection and Nuclear Safety (IRSN) has for many years now equipped its emergency response center with operational tools to assist experts in the assessment of potential risks to local populations and terrestrial environments in the event of accidental release of radionuclides to the atmosphere. These tools were used in particular in the case of the Fukushima accident, among other things to simulate the short and long-range atmospheric dispersion of released radionuclides [4, 5]. Atmospheric dispersion computer codes are combined with

1 http://sirocco.omp.obs-mip.fr/outils/Symphonie/Produits/Japan/SymphoniePreviJapan.htm

138

C. Duffa

computational modules designed to predict exposure levels and activity concentrations in different environmental compartments. This calculation system is operational and can give estimations of radiological consequences in the first hours for any atmospheric accidental release in France. The French coastline is equipped with 5 nuclear power plants (Gravelines—Six 900 MW reactors, Penly—two 1300 MW reactors, Paluel—four 1300 MW reactors, Flamanville—two 1300 MW reactors and future EPR, Blayais—four 900 MW reactors), 4 military harbors that receive nuclear powered vessels (Cherbourg, Brest, Ilelongue, Toulon) et de fuel reprocessing plant Orano La Hague. Mouths of four main French Rivers (Seine, Loire, Gironde and Rhône River) are potential source terms to the sea in case of radiological contamination of the upstream water (Fig. 11.2). Therefore, Chanel, Atlantic and Mediterranean French coastal waters should be directly contaminated in case of any accidental radioactive release. Considering this, it appears essential to take the risk of contamination of French marine waters into account, and to improve our knowledge and modelling capabilities as existing for atmospheric dispersion. For this reason, to ensure optimal preparedness in the event of a nuclear emergency affecting the marine environment, IRSN is implementing complementary tools for assessing the evolution and consequences of radioactive marine contamination event considering both risks of atmospheric deposition and direct release due to any accidental situation. In case of any accidental situation, IRSN expertise should require a fast representation of marine dispersion of radionuclides to: (a) provide an estimates of expected activity concentrations in seawater, to ensure the protection of populations directly or indirectly exposed to contaminated environments, (b) provide estimates of expected activity concentrations in fishing or aquaculture products, (c) provide contamination distribution maps and evolution to adapt sampling strategies. Complementary to this model need and to facilitate risk assessment and decision making, detailed information regarding the site-specific issues (ecological, economic and health-related interests) have to be collected. The approach adopted to meet these objectives is twofold: – Development of a computer code to simulate the dispersion of radionuclides in seawater and their transfer to aquatic organisms; – Preparation of site-specific data sheets for each coastal area identified as particularly vulnerable in terms of exposure to an accidental release of radionuclides (coastal nuclear installations, river mouths, military ports).

11.2 STERNE The STERNE simulation tool (“Simulation du Transport et du transfert d’Eléments Radioactifs dans l’environNEment marin”) is designed to provide a fast assessment of the radiological impact of any accidental release affecting the aquatic environment. STERNE is intended to simulate radionuclide dispersion in seawater and to

11 Marine Dispersion Modelling and Expertise Tools …

139

Fig. 11.3 Schematic diagram of STERNE implementation principle

calculate expected activity concentrations in different biological compartments. The results can be used both for predicting the evolution of contamination events and for dose assessment purposes (via post-processing tools). When cross-referenced with contextual data, these results can be used to define measures prohibiting swimming, fishing or other site-specific activities, and also to provide guidance for sampling strategies during emergency response and post-accident phases. The implementation principle of the STERNE simulation tool is shown schematically in Fig. 11.3. The basic principle is the same as for atmospheric dispersion calculations currently performed at IRSN’s emergency response centre, with source terms (radionuclide input) and meteorological data used as input data. For marine dispersion calculations, source terms and hydrodynamic data are fed directly into the simulation tool.

11.2.1 Inputs Hydrodynamic data to be provided to STERNE in a 3D Cartesian grid include: – Water flux cumulated between two save time steps, which correspond to a 3D table of current speeds multiplied by each grid cross section. – Sea surface elevation. – Vertical diffusion. These data are provided in a netcdf file for the area and time period that include the accidental scenario. Source term corresponds to the input of radionuclides defined either by a direct release into seawater or by atmospheric deposition onto the sea surface. This can be also a combination of different inputs. Direct release into seawater in defined by its spatial coordinates and depth. Input data are flux of each individual radionuclide on a specified time space (x becquerels between two dates). This input data are provided to STERNE through a text file. Atmospheric deposition onto seawater corresponds to the output of our atmopheric dispersion modelling. These netcdf files include values of deposition flux of individual radionuclide (Bq/m2 /s) on a the atmospheric model specific cartesion grid. STERNE is able to interpolate these data onto the grid of the hydrodynamic file.

140

C. Duffa

11.2.2 Radionuclide Dispersion Calculation STERNE, developed in FORTRAN95, calculates the horizontal and vertical transport of dissolved radionuclides taking into account their radioactive decrease defined by their half-life. Eulerian dispersion is calculated using a classical tracer advectiondiffusion approach (Eq. 11.1). ∂C − → − →→ = Div(K ∇ C) − ∇ (− u C) ∂t

(11.1)

Where: C is the radionuclide concentration K is the turbulent diffusion tensor − → u is the advection current t is the time elapsed Activity concentrations are then calculated offline from the hydrodynamic at each grid point of a chosen sub-modelled area (Fig. 11.4). In this sub-area, the original 3D grid provided for the hydrodynamic data is kept for dispersion calculations. Calculation time step values are user-defined based on an acceptable compromise between calculation time and numerical stability. It depends on the mesh size and maximum sea current velocity for the area considered. For example, if we use a 1.2 km resolution hydrodynamic input for the Northern Mediterranean Sea, the time step is set to 50 s; for a higher resolution hydrodynamic model (Toulon area, 100 m grid size), the time step has to be set to 15 s.

Fig. 11.4 Example of a sub-hydrodynamic MARS3D (MARS3D hydrodynamic data provided by IFREMER—PREVIMER in 2014) modelled area used for STERNE dispersion calculations

11 Marine Dispersion Modelling and Expertise Tools …

141

11.2.3 Transfer to Biota Calculation STERNE calculates also activities in specified living organisms using biokinetic radioecological approach at each time step of the dispersion calculation as follow: dCo (t) = ki · Cw (t) − (λ p + ko ) · Co (t) dt ki = (ko + λ p ).C F Co Cw ki ko λp CF

(11.2)

is the activity concentration in the organism (Bq kg−1 fresh weight) is the activity concentration in seawater (Bq l−1 ) is the uptake or accumulation rate constant (d−1 ) is the elimination or depuration rate constant (d−1 ) is the physical decay constant (d−1 ) is the concentration factor for the studied organism (l kg−1 f.w.)

In STERNE code, this equation is computed as following (Eq. 11.3) for a time step i (see [2] for details): Co(i) = a × Co(i−1) + FC × (1 − a) × Cw(i)

(11.3)

Where i is the time increment. With a = e−T (ko +λ p ) and T = t(i) − t(i − 1) the constant time step duration. In order to refine this one compartment dynamic model and adapt it to postaccident conditions (for which various studies report multiple depuration rate constants), the STERNE simulation tool allows users to combine two independent compartments as A.Co1+B.Co2. A and B values are between 0 and 1 and A + B = 1. Each compartment Co1 and Co2 has its own biological half-life (Tb). All required radioecological parameters (concentration factors and single or multicomponent biological half-lives) are compiled from literature [1, 3, 9] for main radionuclides and generic marine species (fish, molluscs, crustaceans, algae). These default values can be change by the user if necessary. For biota, calculations are not done in each grid mesh. Surface, bottom, or vertical mean seawater activity is used for this calculation (depending on the living area of the considered organisms).

11.2.4 Outputs Seawater volume activities are calculated for each mesh of the defined area and for each time step. Results of activities in seawater (Bq/m3 ) and marine organisms (Bq/kg) are available for chosen radionuclides and organisms in a netcdf file with an output time step defined by the user (generally 1–6 h) (Fig. 11.5).

142

C. Duffa

Fig. 11.5 Example of the calculated 137 Cs activities in surface water for a random release in the French Atlantic coastal area

It is also possible to generate a text file with time evolution of radionuclides activities for a specific point (defined by the user with x, y, z coordinates). Figure 11.6 presents an example of this kind of result obtained with Fukushima dispersion model.

11.3 Local Descriptions For each vulnerable site, an inventory is drawn up, listing all information required for preliminary analysis of sanitary and environmental issues near the release point. This includes: – Descriptions of dominant sea currents as a function of meteorological or tidal conditions, allowing for rapid identification of vulnerable areas and assessment of corresponding time frames. This information can provide guidance for the implementation of initial population protection actions (prohibition of swimming, fishing or other site-specific activities, suspension of water intake and port operating activities).

11 Marine Dispersion Modelling and Expertise Tools …

143

Fig. 11.6 Example of STERNE calculation results for a specified point nearby Fukushima nuclear power station

– Maps showing local site-specific interests, including identification of coastal occupation or activity areas for effective implementation of population protection measures. These maps must also show areas of economic interest (fishing, aquaculture and associated activities, industrial activities requiring water intake, sea therapy). – Maps of ecological issues, like protected areas or major ecological richness.

11.4 Conclusions IRSN has developed a specific dispersion and transfer to biota modelling tool, able to be operated offline from the hydrodynamic models. Considering the intended use as an emergency response tool, only dissolve radionuclides are modeled and a simple biokinetic model is used. In case of any marine radioactive contamination, fast results can be expected as hydrodynamic forecast data should be already available in IRSN crisis center. As atmospheric data are provided by Météo-France, hydrodynamic data should be provided by the SHOM through an on-going collaboration project. In case of marine contamination, STERNE modelling tool and all the information ready to use for any vulnerable coastal site will help experts to characterize the radiological state of the marine environment and to propose adapted sampling stategies.

144

C. Duffa

Acknowledgements The IRSN thanks IFREMER and UBO (Université de Brest Occidental) for their collaboration to develop STERNE code. The operational implementation of STERNE in IRSN Crisis Center is now done in collaboration with SHOM who will provide hydrodynamic forecasts for French coastal areas and go on this project and associated research with IRSN.

References 1. IAEA (2004) Technical report series No 422, sediment distribution coefficients and concentration factors for biota in the marine environment 2. Fievet B, Plet D (2003) Estimating biological half-live of radionuclides in marine compartments from environmental time-series measurements. J Environ Radioact 65:91–107 3. Gomez LS, Marietta MG, Jackson DW (1991) Compilation of selected marine radioecological data for the formerly utilized sites remedial action program: summaries of available radioecological concentration factors and biological half-lives, SANDIA report, SAND89-1585 4. Korsakissok I, Mathieu A, Didier D (2013) Atmospheric dispersion and ground deposition induced by the Fukushima Nuclear power plant accident: a local-scale simulation and sensitivity study. Atmos Environ 70:267–279 5. Mathieu A, Korsakissok I, Quélo D, Groëll J, Tombette M, Didier D, Quentric E, Saunier O, Benoît J-P, Isnard O (2011) Atmospheric dispersion and deposition of radionuclides from the Fukushima Daiichi nuclear power plant accident. Elements 8(3). 1811-5209/12/00080195$2.50 https://doi.org/10.2113/gselements.8.3.195 6. Periáñez R, Bezhenar R, Brovchenko I, Duffa C, Iosjpe M, Jung KT, Kobayashi T, Lamego F, Maderich V, Min BI, Nies H, Osvath I, Outola I, Psaltaki M, Suh KS, deWith G (2016) Modelling of marine radionuclide dispersion in IAEA MODARIA program: lessons learnt from the Baltic Sea and Fukushima scenarios. Sci Total Environ 569–570:594–602 7. Science-Council-of-Japan (2014) A review of the model comparison of transportation and deposition of radioactive materials released to the environment as a result of the Tokyo Electric Power Company’s Fukushima Daiichi Nuclear Power Plant accident. Science Council of Japan, In: Sectional committee on nuclear accident committee on comprehensive synthetic engineering, S.C.o.J. (Ed.), 111 pp 8. UNSCEAR (2014) Sources, effects and Risks of Ionizing Radiation—UNSCEAR 2013 Report to the General Assembly with Scientific Annexes, Volume 1, United Nation 9. Vives i Batlle J, Wilson RC, McDonald P (2007) Allometric methodology for the calculation of biokinetic parameters for marine biota. Sci Total Environ 388:256–269 10. Vives i Batlle J, Aoyama M, Bradshaw C, Brown J, Buesseler KO, Casacuberta N, Christl M, Duffa C, Impens NREN, Iosjpe M, Masqué P, Nishikawa J (2018) Marine radioecology after the Fukushima Dai-ichi nuclear accident: are we better positioned to understand the impact of radionuclides in marine ecosystems? Sci Total Environ 618:80–92

Chapter 12

A Study on Water Crisis Management Techniques by Fallout in Case of Radiation Accident Using Environmental Multimedia and Air Transport Diffusion Model Daemin Oh, Youngsug Kim, Sungwon Kang, Soungjong Yoo, Noriyuki Suzuki, and Yoshitaka Imaizumi Abstract Northeast Asia has a high density area of nuclear power plants and recognizes the need for crisis management from radioactive accidents due to actual nuclear power plant. When a radioactive exposure accident occurs, the cesium in the radionuclide has a long half-life and is released into the atmosphere, and it can move through the wind. We need to be able to analyze the effects of deposition by fallout on the water system in our country, along with transport and diffusion from the atmosphere in order to accurately predict radioactive materials. In addition, we can be decided pollution levels in watersheds and rivers by using an environmental multimedia model about the phenomena of transport through air, watershed and rivers. Therefore, it can be predicted and controlled with the whole process of radioactive accident using the result of this study as a crisis management technique. Keywords Nuclear accident · Cesium · Fallout · River · Crisis management

D. Oh (B) · Y. Kim · S. Kang · S. Yoo KICT, 283 Goyangdae-ro, Ilsanseo-gu, Goyang, Gyeonggi, Korea e-mail: [email protected] Y. Kim e-mail: [email protected] S. Kang e-mail: [email protected] S. Yoo e-mail: [email protected] N. Suzuki · Y. Imaizumi National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8056, Japan e-mail: [email protected] Y. Imaizumi e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_12

145

146

D. Oh et al.

12.1 Introduction Northeast Asia has a high density area of nuclear power plants and recognizes the need for crisis management from radioactive accidents due to actual nuclear power plant. Especially, this area have a case of disasters which is Fukushima nuclear power plant accident in Japan. In recent years, the risk of natural disasters such as climate changes and earthquakes is increasing, we have to start crisis management from radioactive accidents. The occurrence of a nuclear accident has a great impact on the surrounding area even in the event of a single accident. Radiation accidents not only pollute the air, but also pollute the land and crops in the surrounding area, polluting rivers and oceans [1]. If a nuclear power plant accident occurs in nearby country, the other countries will be easily contaminated due to the inflow of radioactive materials. Therefore, it is necessary to crisis management from the accident of radiation. After the Fukushima Daiichi Nuclear Power Plant accident (FDNPP), the researchers conducted a study on the risk management of radioactive contamination. Also, Some published researches, that it was revealed that major part of Cs-137 had deposited to forest area, most of Cs-137 strongly attached to soil surface after the deposition, and therefore Cs-137 would slowly run off from the forest area [2]. When a radiation exposure occurs, the radionuclide has a long half-life and is released into the atmosphere, and it can move through the wind. We need to be able to analyze the effects of deposition by fallout on the water system in our country, along with transport and diffusion from the atmosphere in order to accurately predict radioactive materials. In addition, we can be decided pollution levels in watersheds and rivers by using an environmental multimedia model about the phenomena of transport between air, watershed and rivers. Therefore, it can be predicted and controlled with the whole process of radioactive accident using the result of this study as a crisis management technique.

12.2 Materials and Methods 12.2.1 Study Area The study area is Northeast Asia, including China, Korea, and Japan. The area is located on the west wind direction area, and it is the target of Tenwan (34.70 ° N, 119.48 ° E) and Pangzasan (30.44 ° N, 120.96 ° E) in China. Areas affected by radiation accidents include all areas of Korea, adjacent coastal areas and islands.

12 A Study on Water Crisis Management Techniques …

147

12.2.2 Model In this study, HYSPLIT, an atmospheric pollutant transport model, and G-CIEMs, a multimedia fate model, were simultaneously applied to simulate the transport diffusion phenomenon of radioactive materials due to nuclear accidents in neighboring countries. HYSPLIT modeling was carried out using NOAA’s global data assimilation system (GDAS) weather data and IAEA’s Chernobyl nuclear accident source values for transport diffusion modeling of radioactive materials. The changes of Cs137 activity due to the movement of radioactive materials and the amount of fallout deposits that can flow into the domestic river water system were analyzed and used as input values of G-CIEMs.

12.2.2.1

Model Description

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model is a Lagrangian particle and puff model that is used by air quality researchers and forecasters to model the transport of pollutants using 3-dimensional gridded meteorological fields. HYSPLIT is capable of tracing the forward and backward trajectories of the target material and calculating the dry and wet deposition amount. In addition, HYSPLIT can model the diffusion behavior of the radioactive material and the deposition of fallout through the types of radioactive materials, half-life, and radioactive accident information. This allows prediction of the concentration and exposure of radioactive materials in the atmosphere and soil [3, 4]. The basic equations used in this model are 3D particle calculations, velocity calculations of parallel turbulence, and calculation formulas for atmospheric concentrations of the model. In particular, this equation is considered for advection and parallel turbulence (Eq. 1), and the atmospheric concentration can be expressed by dividing the total mass of the 3D particles by time divided by the lattice volume (Eq. 2). Pfinal (t + t) = Pmean (t + t) + U  (t + t) t

(1)

c = m(xyz)−1

(2)

where U ‘(t + t) is the recent time; U’ (t) is the time at the past specific time; R is the autocorrelation coefficient; c is the concentration; x, y and z are the distance of each grid. Also, The model is based on G-CIEMs (Grid-Catchment Integrated Multimedia Modeling system), which was developed for assessing the risk exposure by chemical pollutants. The model components and conditions that were used this model’s default value. However, we try to change several of conditions related to the fate of Cs-137 supporting condition from Japan’s reference data. The G-CIEMs model formulation is based on the fugacity formulation by Mackay [6]. The environmental compartments of the G-CIEMS-Multi model are air, river, sediments of rivers, soil

148

D. Oh et al.

with separate land-use categories, sediments. [9–12]The model can be formulated by a linear differential equation as follows: df/dt = Af + E

(3)

where f is the vector of fugacity variable; A is the matrix consisting of D values (Mackay); E is the vector of emission term; and t is time. A dynamic (level IV) solution can be obtained by solving the previous differential equation [5].

12.2.2.2

Model Setup

The weather data of HYSPLIT model used temperature, wind direction, wind speed, ground pressure of GDAS data provided by NOAA/ARL. The area of the modeling was constructed with a grid of 0.04 ° N and 0.06 ° E spacing in latitude and longitude for latitude 33.2 ° N, −38.56 ° N and longitude 125.1 ° E, −129.9 ° E including the Korean peninsula. It was also assumed that radioactive material was emitted at 1.05 × 1015 Bq/hr at a height of 50 m above ground level in a nuclear accident. The aerodynamic mean diameter of Cs-137 was 0.4 µm, particle density was 1.9 g/cc, dry deposition rate was 0.002 m/s, And the cleaning coefficient under the cloud was 5.0 × 10−5 s−1 . The analytical altitude was set at 200 m above the ground and at the surface for analysis of Cs-137 concentrations in the atmosphere and on the surface, and was analyzed up to 12 days at 24 h intervals. [7, 8] The G-CIEMs model used weather data for each city, weather, wind speed, wind speed, and precipitation data provided by the Korea Meteorological Agency and used topographical data such as water and watersheds generated through ARC-GIS. Also, the sediment concentration of Cs-137 calculated by HYSPLIT model was converted by Kriging method of Arc-GIS program. The results were used as input data for the G-CIEMs model. In order to simulate the diffusion of Cs-137 between multimedia through G-CIEMs model, we converted and improved the grid code system for G-CIEMs atmospheric data and terrain data according to domestic environment. The atmospheric grid is 5 km × 5 km in the atmospheric grid of the study area, and a total of 10,720 grid lines are constructed. In order to apply G-CIEMs in Korea, the input parameter interface was improved. The physicochemical properties of cesium were molecular weight, melting point, density, heat of evaporation, and partition coefficient. Krigging method was used to input meteorological data to each grid. Geographic data input of G-CIEMs was classified into seven categories using land cover map of 5 m spatial resolution provided by Ministry of Environment. Hydrological data are input by using river length, river slope and river flow as main input factors and using GIS and river maintenance plan data. The river code system of G-CIEMs used the method including river and lake, and the main factor was input by assigning the code number to Korean river basin. We also used the Series-parameter function to construct a daily fallout deposition scenario. As shown below in Fig. 12.1 and Table 12.1, all input parameters are input to each grid to create a MDB (Microsoft Access Database), and the input MDB is simulated by putting it into the interface

12 A Study on Water Crisis Management Techniques …

(a)

(b)

149

(c)

Fig. 12.1 Model setup by environmental multimedia G-CIEMs (a) and Meteorological Data by Krigging method (b) Create input structure of river by middle basin (c) land cover map Table 12.1 Input parameter for C-CIEMs Items

Input condition

Phys-chem properties

Cs-137 references dataset (Molecular weight, Partition coefficient)

Geographic data

Land use type, Average size of soil polygon grid (GIS)

Meteorological data

Temperature, Wind velocity and direction, Vapor pressure, Precipitation

Hydrological data

River information (Length, Width), River flow

12.2.3 Deposition Scenario by Radioactive Fallout The simulation cases were classified according to the weather of each nuclear power plant in case of a virtual nuclear accident. Each case was classified according to the wind direction, wind speed, temperature, and rainfall, factors that affect the transport and diffusion of radioactive materials, in order to predict the domestic inflow of radioactive materials. In case of the wind direction which has the greatest influence on the inflow of radioactive materials, the case of winds of westerly winds was selected for the nuclear power plants located on the east coast of China. In the case of wind speed, wind velocity less than 0.5 m/s, wind velocity of 0.5−5 m/ s, which is considered to be middle wind, and wind velocity of 5 m/s. And the cases were classified into 10, 10–20, and 20 °C, which indicate low temperature and low temperature, respectively. In the case of rainfall, which has a great effect on the deposition of radioactive materials, cases of radioactive materials were classified by the presence or absence of rainfall, and a total of 64 weather cases were selected. The type of nuclear power plant accidents applied to this study assumes that 8.5 × 1016 Bq of radiation source value such as Chernobyl accident is released. The location of the accident was targeted at Chinese nuclear power plants adjacent to Korea, and the release rate of radioactive materials was 1.05 × 1015 Bq/hr, and the total calculation time was 168 h (7 days). We used the physicochemical characteristics of Cs-137 and deposition data by fallout from a HYSPLIT. As shown below in Fig. 12.2, a fallout

150

D. Oh et al.

Fig. 12.2 Schematic diagram of deposition scenario by radioactive fallout

sedimentation scenario can be created based on the residence time and weather conditions of radioactive clouds entering Korea. (wind direction, wind speed, and precipitation) Through, total of 20 deposition scenarios were made by nuclear power plants accident.

12.2.4 Analysis of Water Crisis Management by Radioactive Fallout Effect The study from the surrounding environment, use the HYSPLIT model and integrated environmental multi-media model (G-CIEMs) to predict the fallout phenomena that could affect the domestic environment when a nuclear accident. It is possible to predict the time and concentration of the radioactive material flowing into the atmosphere through the air. Using the results of the model results, we can also observe daily changes in radioactive materials in the air, soil, and streams of rivers and dams. In addition, daily model results are converted to shp files, which helps to easily identify the distribution of spatial contamination effects due to radioactive fallout. It is possible to manage the concentration change of each stream and to analyze the diffusion of media between media and another media so that the total amount of radioactive materials can be controlled through mass balance analysis.

12.3 Results and Discussions 12.3.1 Transfer of Radioactive Materials Through Atmospheric Diffusion 12.3.1.1

Prediction of Radioactive Transport Diffusion

We assume that the accident occurs at the Tenwan nuclear power plant and the Pangzasan nuclear power plant, and classify the wind direction, wind speed, temperature, and precipitation occurring in the accident area in the accident home. The following Fig. 12.3 and Fig. 12.4 shows simulation results that have a large effect on the Korea due to the accident of the target nuclear power plant.

12 A Study on Water Crisis Management Techniques …

151

(a) D+1

(b) D+2

(c) D+3

(d) D+4

(e) D+5

(f) D+6

Fig. 12.3 HYSPLIT Simulation Results in case of Tenwan nuclear accident

(a) D+1

(b) D+2

(c) D+3

(d) D+4

(e) D+5

(f) D+6

Fig. 12.4 HYSPLIT Simulation Results in case of Pangzasan nuclear accident

152

D. Oh et al.

In the case of the Tenwan nuclear power plant accident located in the western part of Korea, it can be confirmed that the radioactive material moves inland in a few hours due to the wind of the westerly wind after the nuclear accident on March 29. In addition, the radioactive material has a wide concentration range of 1,000– 10,000 Bq/m2 . In the case of the Pangzasan Nuclear Power Plant located south-west of Korea, after the nuclear accident on June 8, it can be seen that the radioactive material is covered by the southwest wind over 10,000 Bq/m2 in the whole region of Korea in a few hours.

12.3.1.2

Influence Analysis by Radiation Diffusion

The results of analysis of the first arrival time by fallout deposition in each case are shown in Fig 12.5. As a result of the analysis, radioactive materials were inflowed within 5.4 h from Tenwan nuclear power plant, 5.2 h in the Han River basin, 5.1 h in the Nakdong river basin, 5.1 h in the Geum river basin, 5.1 h in the Yeongsan river basin and 5.4 h in the Seomjin river basin. In addition, the radioactive material was found to be introduced within about 7.4 h in the event of a Pangzasan nuclear power plant accident. The following Fig. 12.6 shows the analysis of the time spent influencing the radioactive materials that are transported and diffused to the country during the nuclear accident in Korea. The average time to stay in Korea for each case was 93.6 h for Tenwan nuclear power plant accident and 79.4 h for Pangzasan nuclear power plant accident.

(a) Tenwan

Fig. 12.5 First arrival time by fallout deposition

(b) Pangzasan

12 A Study on Water Crisis Management Techniques …

(a) Tenwan

153

(b) Pangzasan

Fig. 12.6 Retention time by radioactive cloud

12.3.2 Impact Analysis of Radioactive Material Fallout As a result of simulation of HYSPLIT, the atmospheric transport model, the selected cases were not affected on the Korea. A total of 20 simulated cases were determined to simulate the G-CIEMs, a radioactive material dropout diffusion model.

12.3.2.1

Analysis of the Effects of River by Radioactive Fallout

The Tenwan nuclear power plant needs to be withdrawn from the Geum River, the Youngsan River, and the Nakdong River to serious stages. As shown below in Fig. 12.7, the Pangzasan Nuclear Power Plant is expected to be at a serious stage in the Youngsan River and Nakdong River in some areas.

12.3.2.2

Analysis of Water Impacts by Radioactive Fallout

It is the result of analyzing the radioactivity concentration change in the river according to the main point due to radiation exposure accident. The main branches of each water system are used as water sources. The Han River basin is designated as Amsa, the Geum River Basin as Jungri, the Yeongsan River Basin as Sungcheonbo, and the Nakdong River Basin as Maeri. In the case of the Tenwan Nuclear Power Plant (February 22), the Han River basin was injected within 3.3 days when a radiation exposure occurred and a total of one radioactive cloud was injected during the forecast period. Especially, the concentration of radioactivity influenced by winds and air currents showed a large variation among the major points in the country as shown below in Fig. 12.8. In addition, high river radioactivity concentrations of more than 10,000 Bq/L were predicted in the

154

D. Oh et al.

(a) D+1 (Tenwan)

(b) D+7 (Tenwan)

(c) D+1 (Pangzasan)

(d) D+7 (Pangzasan)

Fig. 12.7 Identify the distribution of spatial contamination effect by fallout (Tenwan: April 22, Pangzasan: July 1)

Han River, Geum River, and Nakdong River basins where the atmospheric deposition concentration was high. The Youngsan River basin, which had relatively few radioactive clouds, exceeded 3,000 Bq/L four days after the radiation accident. In the case of the Pangzasan Nuclear Power Plant (July 1), the Han River basin (Amsa) was introduced within 1.8 days of the radiation exposure incident, and showed a total of one radioactive cloud flow during the forecast period as shown in Fig. 12.9. Especially, the concentration of radioactivity influenced by winds and air currents showed a large variation among the major points in the country. In Han

12 A Study on Water Crisis Management Techniques …

155

(a) Han River Basin (Amsa)

(b) Geum River Basin (Jungri)

(c) Yeongsan River Basin (Sungcheonbo)

(d) Nakdong River Basin (Maeri)

Fig. 12.8 Variation of Radioactivity Concentration in river (in case of Tenwan)

River and Yeongsan River basins, river water concentration was influenced by water source, so the management of water source was in urgent need.

12.3.3 Water Crisis Management Through Mass Balance Analysis In order to prepare the mass balance of radioactive materials, we used the results of major radioactive exposure cases for each nuclear power plant to analyze the movement of radioactive materials, their concentration and distribution in each medium. The Fig. 12.10 is a conceptual diagram of the mass balance of the transport of radioactive materials between media and another media. Air was characterized by rapid passing of radioactive clouds from the accident source according to the weather conditions, showing a high radioactivity concentration, while a small amount of fallout deposition (atmospheric, river). Soil flows into the atmosphere due to radioactive fallout and most of it is not lost in the soil, so the amount of soils flowing into the river is very small. In the rivers, radioactive fallout deposits from the atmosphere and outflows into the rivers in the watershed, and they are dissolved in the rivers or settled down to the river bed, and part of them are drained to the river estuary.

156

D. Oh et al.

(a) Han River Basin (Amsa)

(b) Geum River Basin (Jungri)

(c) Yeongsan River Basin (Sungcheonbo)

(d) Nakdong River Basin (Maeri)

Fig. 12.9 Variation of Radioactivity Concentration in river (in case of Pangzasan)

(a) Tenwan

(b) Pangzasan

Fig. 12.10 Mass balance analysis through media transfer

12.4 Conclusions In this study, the influx of radioactivity into the water system in China occurred within 5.1–7.4 h due to the wind direction and velocity. In addition, the radioactivity inflow into domestic air was slightly different due to the weathering and geomorphic characteristics of the radioactive retention time, and the maximum radioactive deposition concentration was 100,000 Bq/m2 . Air was characterized by rapid passing of radioactive clouds from the accident source according to the weather conditions, showing a high radioactivity concentration, while a small amount of fallout deposition (atmospheric, river). Soil flows into the atmosphere due to radioactive fallout and most of it is not lost in the soil, so the amount of soils flowing into the river is very small. Also, a radioactive material is

12 A Study on Water Crisis Management Techniques …

157

attached silt in watershed and it is runoff during rainfall. The radioactive material deposited in the river was easily dissolved and showed a characteristic that the stream rapidly flowed out. Therefore, it can be predicted and controlled with the whole process of radioactive accident using the result of this study as a crisis management technique. Acknowledgements This Work Was Supported by The National Research Council of Science & Technology (NST) Grant By The Korea Government (MSIT) (No. CAP-15-07-KICT).

References 1. Oh DM, Kim YS, Kang SW (2016) A study on the estimate method for radioactive materials runoff by fallout in basin using grid-catchment integrated environmental modeling system. In: KSWW & journal of KSWE conference, pp 259–260 2. Imaizumi Y, Kuroda K, Morino Y, Hayashi S, Suzuki N (2016) Temporal-spatial fate of cesium137 in terrestrial environment around Fukushima Daiichi Nuclear Power Plant: a challenge for daily simulation. In: SETAC Europe 26th annual meeting 3. Rolph GD, Ngan F, Draxler RR (2014) Modeling the fallout from stabilized nuclear clouds using the HYSPLIT atmospheric dispersion model. J Environ Radioact 136:41–55 4. Kim CH, Song CK (2003) Lagrangian particle dispersion modeling intercomparison: Internal versus foreign modeling results on the nuclear spill event. J Korean Soc Atmos Environ 19(3):249–261 5. Suzuki N, Murasawa K, Sakurai T, Nansai K, Matsuhashi K, Moriguchi Y, Tanabe K, Nakasugi O, Morita M (2004) Geo-referenced multimedia environmental fate model (G-CIEMS). Model formulation and comparison to the generic model and monitoring approaches. Environ Sci Technol 38:5682–5693 6. Mackay D (2001) Multimedia environmental fate model: the fugacity approach, 2nd ed. Lewis Publishers 7. Korsakissok I, Mathieu A, Didier D (2013) Atmospheric dispersion and ground deposition induced by the Fukushima Nuclear power plant accident: a local-scale simulation and sensitivity study. Atmos Environ 70:267–279 8. Stohl A, Seibert P, Wotawa G, Arnold D, Burkhart JF, Eckhardt S, Yasunari TJ (2012) Xenon133 and caesium-137 releases into the atmosphere from the Fukushima Dai-ichi nuclear power plant: determination of the source term, atmospheric dispersion, and deposition. Atmos Chem Phys 12(5):2313–2343 9. Cowan CE, Mackay D, Feijtel TCJ, van de Meent D, DiGuardo A, Davis N, Mackay N (eds) (1995) The multi-media fate model: a vital tool for predicting the fate of chemicals. SETAC Press, Penascola 10. Van de Meent D (1993) SimpleBox: a generic multi-media fate evaluation model; RIVM Report 6727200001. Institute of Public Health and the Environment 11. McKone TE (1993) CalTOX, a multimedia total-exposure model for hazardous wastes sites. Part II: the dynamic multimedia transport and transformation model; Report UCRL-CR-11456 Part II (1993). Lawrence Livermore National Laboratory 12. Organization for Economic Cooperation and Development (2002) Report of the OECD/UNEP workshop on the use of multimedia models for estimating overall environmental persistence and long-range transport in the context of PBTS/POPS assessment. OECD environment, health and safety publications: series on testing and assessment No 36

Chapter 13

Challenges in Defining Alarm Thresholds to Improve Crisis Management Procedures: A Case Study on the French Riviera Stan Nomis, Leslie Salvan, Raphaëlle Dreyfus, Franck Compagnon, and Pierre Brigode Abstract When dealing with crisis management in the context of river floods, clear and straightforward procedures must be established. SMIAGE (Syndicat Mixte Inondation, Aménagement et Gestion de l’Eau maralpin) is a public structure dealing with flood, river and water management in the AlpesMaritimes department in the SouthEast of France. This structure was created to gather the department operational forces to consider water management from the river basin’s point of view, rather than from the administrative limits point of view. Indeed, SMIAGE was created after the deadly event that occurred on the October 3rd , 2015, when the needs of a wider consideration of flood events were demonstrated on the French Riviera. Moreover, in the Mediterranean context, flash floods must be considered. These events are widespread and poorly understood. One of the main missions of this entity is to assist municipalities by operating a flood warning system, especially for the Siagne catchment. Everywhere else on the SMIAGE’s territory, the objective is to maintain dikes and other hydraulic structures and to be able to help municipalities to get accurate and clear information while a flood event is happening and afterwards. While crisis procedures are implemented already, warning thresholds are mainly defined by historical experience and local knowledge. Water depths generating road flooding are often well-known by municipalities. However, intense urbanization has changed catchments’ response and vulnerable areas’ location. Meanwhile, hydrometric and rain gauges together with radar data are available. A reflection is undertaken about a large-scale modelling system. Data of different types are examined with several modelling philosophies. First results show that large-scale modelling can help defining warning water depths at selected river sections. Event management procedures can be refined and improved thus reducing false and missed alarms. S. Nomis · L. Salvan (B) · R. Dreyfus · F. Compagnon, SMIAGE, Nice, France e-mail: [email protected] S. Nomis · P. Brigode Polytech Nice Sophia, Université Nice Sophia Antipolis, Nice, France P. Brigode Université Côte d’Azur, CNRS, OCA, IRD, Géoazur, Nice, France © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_13

159

160

S. Nomis et al.

Keywords Alarm thresholds · Crisis management procedures · Large-scale modelling · Open-source modelling · Public stake-holder

13.1 Introduction The catastrophic floods of October 3rd , 2015 and their disastrous consequences have shown the need to organize the territory of the Alpes-Maritimes in the light of these natural hazards. The principle of creating a Public Territorial Basin Establishment (EPTB) was adopted to pool skills and concentrate resources to meet the challenges of river management and flood prevention. The SMIAGE, was created on January 1st , 2017 to carry out these missions to gather skills and concentrate resources to meet the challenges of river management, water resources and flood prevention at the basin scale. Thus, one of the missions of SMIAGE is flood protection. Indeed, out of the 11,000 km of streams of the supervised territory, approximately 50 km of watercourse are diked. The rivers of the French Riviera generally have a torrential regime. The associated floods can be intense. In addition, many dikes systems date back to the 19th century. Therefore, SMIAGE wishes to develop tools to assist municipalities to ensure the protection of the population in parallel with reinforcement work. The SMIAGE’s area of action is extensive and therefore requires adapted tools combining simple methods with field work by agents. On the one hand, hydraulic modelling allows SMIAGE to define alert thresholds that make it possible to associate a streamflow value with a level of danger and thus inform and help concerned municipalities as best as possible. On the other hand, a permanent monitoring by agents makes it possible to identify malfunctions to restore a good flow in the stream network. This article presents SMIAGE’s ambitions to improve the definition of its alert thresholds for the management of hydraulic structures, including dike systems. The objective is to couple hydrological and hydraulic modelling to establish scenarios that will serve as a decision-making aid. This work will have to be applied to both gauged and ungauged catchments. The main characteristics of the study area will be presented in Part 2. Part 3 will present the tools available to SMIAGE for crisis management. In Part 4, the concepts of the method for defining alert thresholds will be presented. Finally, Part 5 will present the first conclusions regarding our approach.

13.2 The French Riviera Characteristics The French Riviera is a territory located between the Alps and the Mediterranean Sea. This means that hydrological catchments are generally impacted by both mountainous and marine influences. In particular, the Var river is one of the most powerful of the French rivers with very high solid transport. Upstream parts of catchments are characterized by high slopes and torrential regime, while downstream parts are

13 Challenges in Defining Alarm Thresholds …

161

heavily urbanized. On top of that, there are many small coastal catchments which have a very short hydrological response. A high number of them have a covered and canalized downstream part in cities. Overall, the territory is very diverse and operational services have to handle this diversity and to adapt to different watercourse flow conditions, different catchment behaviors. In addition, the area of action of SMIAGE is large. Indeed, this territory extends over more than 5300 km2 with 11 000 km of rivers (Fig. 13.1).

Fig. 13.1 SMIAGE territory extent

162

S. Nomis et al.

13.3 Available Tools 13.3.1 Hydrometeorological and Hydraulic Tools to Provide Forecast and Support 13.3.1.1

Hydrometeorological Forecasting

SMIAGE structure is organized to be able to deal with flood issues 24/7 all along the year to warn municipalities in case of a flood event which may cause damages, or which could represent a threat for populations. Moreover, hydraulic structures such as dikes and reservoirs are under the SMIAGE’s management. Therefore, there is a 24 h-on-call system (Fig. 13.2). The weather watchman is the person in charge of the hydrometeorological watch, thanks to several tools, such as weather forecasts from Meteo-France, hydrometric stations all over the different catchments, public hydrometric stations (Vigicrues, https://www.vigicrues.gouv.fr/) for the Var river, and X-band weather radar image and forecast developed by Novimet and called “RAINPOL”. From another point of view, hydraulic modelling is a great tool which is necessary to enable this 24 h watch. Indeed, using hydrometric stations is very interesting for flood forecasting, but this is completely useless when you do not know when you have to warn municipalities. Warning thresholds need to be defined according to the experience and to hydrological and hydraulic modelling. The objective is to set alarms at suitable levels of both rainfall and water level or discharge so that the SMIAGE has enough time to organize and react to warn local authorities and enable municipalities to warn and protect their populations.

Fig. 13.2 Organization of the SMIAGE 24 h-on-call system

13 Challenges in Defining Alarm Thresholds …

163

Fig. 13.3 Alert threshold graph for flood event management

There are limits in this process for areas which are not gauged, for example. Also, because SMIAGE is a developing structure, not all the territory hydrological network is modelled yet. Moreover, a supervision tool is developing to gather all the local data. For now, not all the needed data are accessible. Worse, some hydrometric stations need to be scaled or adapted to new flow conditions, after renovation works for instance. These are the reasons why a great development work is undertaken in order to improve supervision tools and processes.

13.3.1.2

Alert Threshold’s Definition

As explained previously, one of the most important steps of the supervision development lies in the alert threshold’s definition for flood management and for hydraulic structures supervision. The objective is to have alert thresholds from two different points of view: weather forecasts (which might be obtained from external sources for the next 24, 12, 6 or 1 h(s)) and instant river state at supervised location (which is obtained by internal measuring stations network). This approach is presented in a graphical form to allow a simple understanding for 24 h-agents. Figure 13.3 shows the alert thresholds graph presented using discharge thresholds (y-axis) and 24-h precipitation forecasts (x-axis). This is used for example for dike monitoring. This first version makes it possible to determine the crisis management state (standby, pre-alarm, alarm 1, alarm 2) based either on a weather forecast or on a flow value at the dike. For instance, precipitation forecasts will be used the day before to mobilize teams for the next day or to set a night-team in case of a forecasted extreme event. Then, during the flood event, instant flow rate thresholds takes over. First, Q1 value is a protection value that allows the patrol agents to be ready to react. Q2 value is taken close below the protection flow rate of the dike. The idea is to set a threshold

164

S. Nomis et al.

1 h before the peak flow of the design hydrograph. This value will make it possible to anticipate overflow at the dike.

13.4 An Approach to Improve Alert Thresholds 13.4.1 Hydrological Modelling to Generate Scenarios 13.4.1.1

Objectives

In Sect. 13.3.1, the tools available to SMIAGE were presented. However, these tools operate independently and therefore require separate expertise. Therefore, SMIAGE wishes to articulate the use of these tools to provide better results and improve crisis management. Thus, hydrological modelling will link hydrometeorological forecasting with the results of hydraulic models. The ambitions of SMIAGE are based on the use of rainfall forecasts to determine, via hydrological modelling, a flow rate in a given river. In parallel with the RAINPOL service, Novimet is developing its hydrological model based on the Irstea GR method (https://webgr.irstea.fr/modeles/?lang=en). SMIAGE will have access to the results of this modelling. However, the SMIAGE aims to create its own hydrological model to obtain additional information. The SMIAGE model will be used to define scenarios. Indeed, by using remarkable rainfall values associated with different soil conditions, different cases will be established. For example, modelling a 150 mm rainfall over a watershed with three possibilities: the soil in the watershed is saturated, conditions are normal, and the soil is very dry. Thus, with one rainfall, hydrological modelling gives three flow values. These flows will be compared with the values obtained during the hydraulic modelling carried out at the dikes. This comparison will make it possible to assess if the dike will perform its protective role properly. Then, we will use as input data a DEM and a rain presented on a spatial grid. An intersection between the rain grid and the contours of a watershed will determine the proportion of gross rain in the watershed. Then the proportion of net rainfall in the watershed will be determined using the SCS-CN.

13.4.1.2

SCS-CN Production Model

In hydrological modelling, the production function divides gross precipitation into the net precipitation that will participate in the direct flow during a rainy event. The SCS-CN approach [7] is one of the oldest and most widely used production functions in spatial modelling. It was proposed by the Soil Conservation Service (SCS) based on experimental results from small American agricultural watersheds

13 Challenges in Defining Alarm Thresholds …

165

to link net rainfall accumulation to gross rainfall accumulation. The model is based on several assumptions: – a certain amount of rain is intercepted (noted Ia), – the soil has a maximum storage capacity (noted S), – the runoff coefficient evolves according to a proportionality relationship. The model is expressed in the following form: Pn =

(P − Ia )2 P − Ia + S

Where Pn refers to the accumulation of net rainfall [mm], P to the rain accumulation [mm], S to the maximum storage capacity [mm] and Ia interception capability [mm]. In addition, the following relationship between Ia and S is generally accepted: Ia = 0.2S Finally, the Curve Number CN parameter, is defined from the capacity S: S=

2540 − 254 CN

which allows us to consider that the SCS-CN production function depends only on an CN parameter varying from 30 to 100. The CN parameter reflects the flow propensity of a watershed. Its estimation must consider the main factors involved: nature and land use, soil moisture conditions at the time of the event. The estimation of this parameter is subject to many uncertainties. In this project, we will use predetermination tables for CN values proposed by the USDA that consider soil type and land use type. CN values will be intersected with the DEM to obtain a CN value per DEM pixel. We will use a formulation that allows us to adjust the CN values according to soil moisture conditions.

13.4.1.3

The Unit Hydrograph as a Transfer Function

In hydrological modelling, the transfer function simulates the transformation of a net rainfall into a hydrograph at the outlet of the studied watershed. The unitary hydrograph is a concept widely used to simulate in a global way the effects of the different water paths within a watershed. The unit hydrograph is the response of a watershed to a rainfall impulse. This is the function of distributing travel times within the watershed. The unit hydrograph will allow to calculate a hydrograph at the outlet of a watershed from a hyetograph. Tools for processing DEMs for hydrology make it possible to automatically create drainage trees. The distribution of travel times will be done using a drainage tree and a hypothesis on transfer rates.

166

13.4.1.4

S. Nomis et al.

Determination of Flow at the Outlet of a Watershed

The unit hydrograph allows to calculate a hydrograph at the outlet of a watershed from a hyetogram. For each elementary surface area of the catchment area, the net rainfall contributions of the different time steps reach the basin outlet with a delay corresponding to the travel time. For the travel time, a speed on the hillsides and in rivers will be defined as a parameter. Thus, the flow Q(t) corresponds to the convolution product between the net rain Pn and the unit hydrograph HU:  Q(t) =

τ

Pn (t − τ )HU (τ )dτ 0

By forcing the position of the outlet near a dike it will be possible to determine the flow rate at the structure.

13.4.2 Hydraulic Modelling to Determine Alert Thresholds 13.4.2.1

Objectives

The scope of SMIAGE’s action is huge. Thus, hydraulic modelling requires a lot of work hours. Indeed, of the 50 km of dikes, 38 km are classified dikes. In addition, information on past modelling is often difficult to collect. SMIAGE needs a tool to automate hydraulic modelling to the maximum. Indeed, the definition of alert thresholds must be applied to each dike. Moreover, the different specificities of the Côte d’Azur make the creation of a 1D hydraulic model a time-consuming task. The Cartino tool was chosen. Its first use was in the context of the application of the European Flood Directive, where it was used to map floodplains resulting from extreme events [4, 5]. The use of CARTINO will allow SMIAGE to build its own hydraulic model. The interest is to facilitate the process of determining the alert thresholds in the right of the structures. In addition, it will be easier to update the information for the construction of alert threshold graphs later.

13.4.2.2

An Automatic 1D Hydraulic Modelling Tool

During Le Bihan’s thesis [2], Cartino was adapted to generate catalogues of flood maps for a wide range of streamflows of small rivers subjected to flash flooding. Subsequently, Martin [3], as part of an internship at IFSTTAR (Institut Français des Sciences et Technologies des Transports, de l’Aménagement et des Réseaux), recoded this tool under R [6] to automatize it. Thus, the preparation of data and the production of the flood map catalogue no longer requires manual intervention. Then, the Cartino R method was evaluated in different sectors with generally satisfactory results [1]. This Cartino R version is still under development at IFSTTAR.

13 Challenges in Defining Alarm Thresholds …

167

Cartino R is a tool that allows to produce catalogues of floodable area maps automatically for streamflows of different return periods and from a high-resolution DEM and a hydrographic network. To do this, it automatically positions cross section on the hydrographic network and extracts the geometry of these cross sections from the DEM. Then, it builds and launches a HEC-RAS 1D steady-state hydraulic model to evaluate the longitudinal profiles of the water line. Finally, post-treatment allows the water levels to be interpolated on the DEM to delimit the flooded area. For SMIAGE, the aim is not to produce flood maps. However, the use of a part of the Cartino tool will allow to extract cross sections and to initialize a 1D hydraulic simulation in steady state. These initial results can be edited under HEC-RAS to determine the alert thresholds for dike management. The steps of the Cartino routine adapted to the needs of SMIAGE are presented in Fig. 13.4. Cartino uses a DEM, a hydrographic network, flow information and a roughness value as input. Initially, cross sections will be positioned in fixed space steps along the hydrographic network. Then, the created cross sections are extracted

Fig. 13.4 simplified diagram of CARTINO’s operation adapted to the needs of the SMIAGE according to [1]

168

S. Nomis et al.

on the DEM (1). A first hydraulic calculation is performed to determine the mirror width on each initially positioned cross profile and a linear interpolation curve of the widths along the section is created (2). This curve allows the cross profiles to be repositioned (3). Once this step is completed, a test is performed to identify (4) and reorient the intersecting profiles (5). If the cross profiles are not wide enough, then a treatment is carried out to enlarge them (8). Finally, the HEC-RAS results are recovered (9).

13.5 Conclusion This article presents the progress of SMIAGE for river management as well as its ambitions to set and improve alert thresholds for hydraulic structures’ supervision. The objective is to adjust alert thresholds next to hydraulic structures to be able to supervise a wide territory. On one side, the hydrometeorological tools available are used to define areas at risk during flood events. Then, hydrological conditions of the concerned catchment are determined. On the other side, hydrological forecasting is combined with streamflow to determine the potential alert level of the structure. The next steps of the work are: – – – –

The development of a wide hydrological model; The evaluation of the hydrological model’s results; The construction of a global and simplified hydraulic model (with Cartino); The definition of a coupling method enabling a disconnected use of both types of models.

Finally, the success of this project will allow SMIAGE to anticipate extreme phenomena and thus allow for better crisis management. Saving time is important for SMIAGE because many catchments have a short reaction time, which implies the use of tools that are easy and quick to set up.

References 1. Hocini N (2018) Développement de la méthode CARTINO pour la cartographie du risque inondation lié aux crues soudaines 2. Le Bihan G (2016) Distributed flash flood forecasts based on regional hydrological models: towards the forecast of flood possible impacts and damages (Theses). Université Bretagne Loire 3. Martin F (2017) Cartographie du risque inondation lié aux crues soudaines : Simulation des inondations d’octobre 2015 sur la Côte d’Azur. IFSTTAR 4. Pons F (2014) Cartographie des surfaces inondables extrêmes pour la directive inondation : cas de la Nartuby; Flood hazard maps for extreme event scenario: the study of Nartuby river 5. Pons F, Alquier M, Roux I (2018) Semi-automatic Maps for 2015 French Riviera Floods, pp 497–513. https://doi.org/10.1007/978-981-10-7218-5_35 6. The R Core Team (2018) R: A language and environment for statistical computing. [WWW Document]. R Found. Stat. Comput. https://www.r-project.org/

13 Challenges in Defining Alarm Thresholds …

169

7. USDA (1956) Soil Conservation Service. National Engineering Handbook; Supplement A, Section 4, Hydrology; Soil Conservation Service: Washington, DC, USA

Chapter 14

Integrations of an Early Warning System and Business Continuity Plan for Disaster Management in a Science Park Tsun-Hua Yang, Hao-Ming Hsu, and Hong-Ming Kao

Abstract Extreme weather events such as typhoons and torrential rain can cause loss of life, damage of property and business. Recent research and events showed that flood damage to industrial areas not only cause damage to high-tech instruments and products but also lead to the supply chain disruption. The later one has a significant impact on the global market. Therefore, this study proposes a system that integrates weather and flood forecasts with a concept of business continuity plan (BCP) to minimize the disruption of business operation due to floods. BCP consists of a series of actions of preparedness, information analysis, response, recovery and maintenance, providing constructive suggestions to decision-makers. Given information from a purpose-driven design flood routing model and the BCP, the decision makers can react to the disaster accordingly. This study selected a science park in central Taiwan as the study area. A custom-made BCP and associated decision support system (DSS) were established for analyzing information and retrieving relevant suggestions of actions based on the factories within the science park. The administration office of the park along with the factories conducted several flood drills and the results showed that the system improves the emergency preparedness and responses. The science park can strengthen their resilience to disasters through the system. The system and concept can be easily adopted and applied to other industrial parks because of its effectiveness and performance. Keywords Flood routing · Business continuity plan · Decision support system · Disaster management · Resilience T.-H. Yang (B) · H.-M. Hsu Department of Civil Engineering, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan (R.O.C.) e-mail: [email protected] H.-M. Hsu e-mail: [email protected] H.-M. Kao Disaster Prevention and Water Environment Research Center, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan (R.O.C.) e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_14

171

172

T.-H. Yang et al.

14.1 Introduction Extreme weather events such as typhoons and torrential rain can cause loss of life, damage of property and business. Recent research and events showed that flood damage to industrial areas not only cause damage to high-tech instruments and products but also lead to the supply chain disruption. Moreover, the supply chain disruption can cause a significant negative impact on the global market. [1–3] Some studies reviewed the risk in supply chain and then revealed the significance of business continuity plan (BCP). [4, 5] BCP is the plan designed to minimise the disruption of the business operation. A variety of research described the effect of BCP for different types of business. [6–8] Therefore, this study introduced the idea of BCP to a science park in order to enhance the disaster resilience of enterprises and organisations. This study also proposed a system that integrates real-time observations and weather and flood forecasts with a concept of BCP. This study focused on application of BCP and proposed a decision support system (DSS) to mitigate the impact of disasters in a science park. We will introduce the idea of BCP briefly first in Sect. 14.2, and then describe the study site in Sect. 14.3. Then in Sect. 14.4, we will elaborate the structure of the whole system and provide an example of its application. Finally, the conclusion will be presented in Sect. 14.5.

14.2 Business Continuity Plan Business continuity plan (BCP) is essential to an organisation, and it is designed to reduce the negative impact and to return to normal business operations. BCP consists of a series of actions to make sure that the organisation can maintain their operation above or at least at the minimum level during the disruption and restore their condition to the ordinary condition within the maximum tolerable period of disruption. The concept is shown as Fig. 14.1. Every organisation may have their own BCP according to their situation. The process of BCP design was standardised as ISO22301 [9], organisations can follow the process to develop an appropriate BCP. The basic steps can be described as below. First, an organisation should review the whole operation and identify the risk. Then they need to analyse the business impact according to the risk assessment and develop their response strategies to mitigate the negative effect and to restore from the crisis. Next, they have to test the strategies to ensure the effectiveness, and then train the personnel on the basis of the strategies. Finally, they must maintain and regularly update the BCP.

14 Integrations of an Early Warning System and Business Continuity …

173

Fig. 14.1 Illustration of Recovery Process (with and without BCP)

14.3 Study Area This study selected Huwei Science Park, a science park in Huwei Township, Taiwan, and the surrounding area as the study area, as shown in Fig. 14.2, and the red line is the boundary of the park. The science park is situated in central Taiwan with a total area of about 97 hectares, with key industries such as optoelectronic and biotechnological manufactories. This area slopes gently from the north towards the south, with a slope of 0.001. Outside the science park, there are full of farmlands and spare lands. The sky-blue line at the north is New Huwei Creek, and the other thinner sky-blue lines represent a tributary of the creek, Hsinchuangzhi drainage. There are dykes with a height of 2–3 m alongside the creek, but on the contrary, there is no such protection along the drainage. The sewer system of the park is designed to collect all the precipitation within the whole science park. Then the water of the entire sewer system will be gathered together into 2 detention ponds, represented as navy blue areas in the figure, at the lowest part of the park. The water will only be released through sluice gates into Hsinchuangzhi drainage when the water level of the pond is higher than that of the drainage. Seven water gauges (marked as triangle in Fig. 14.2) are in the sewer system. There are three at intersection for observing the water levels in storm drainages and four at the detention ponds and Hsinchuangzhi drainage for monitoring the interaction between the inside and the outside of the sewer system. Besides, there are additional four water gauges at bridges for river flood warning (the marked as rectangular in Fig. 14.2). There are two at the north for New Huwei Creek and two at the south for Hsinchuangzhi drainage.

174

T.-H. Yang et al.

Fig. 14.2 Study Area - Huwei Science Park and the surrounding area

14.4 Methodology and Application In this study, a custom-made BCP and a decision support system (DSS) were established for analysing information and retrieving relevant suggestions of actions based on factories of different kinds in the science park. The process is shown in Fig. 14.3. A system, named DayuSWS, would collect all the on-site monitoring data, optimize weather forecasts, and provide future 6-h inundation simulation results. Then the DSS would gather all the information from the DayuSWS and from the government open data platform (https://data.gov.tw/) for analysing. The DSS would search and retrieve relevant suggestions of actions from the database of BCP. The users would be notified of the suggestions automatically and they could also check the information online manually. The details of DayuSWS, the BCP and the database, the DSS and the application will be elaborated in the following sections.

14.4.1 Smart Water System A cyber-physical smart flood early warning system for Huwei science park, called Dayu Smart Water System (DayuSWS) has been implemented for emergency

14 Integrations of an Early Warning System and Business Continuity …

175

Fig. 14.3 Flow Chart of the BCP Operation

response [10]. The system utilises the on-site observations (i.e., water levels and the rainfall) and rainfall forecasts as the inputs of the model to provide 6-h inundation forecasts. The system employs Storm Water Management Model (SWMM) [11] to build up a simple 1D-1D (streets-sewer) dual drainage model for fast flood modelling inside the science park. The underground pipes, the detention ponds of the sewer system, and the roads were connected together by the manholes in the model. The system considers observed water levels and the combination of the rainfall observations and forecasts as model inputs. The outfall at the junction of the larger detention pond and Hsinchuangzhi drainage is the only outlet. A sluice is there and it is governed by the difference of water levels. Then, the system will simulate flood routing and locate the possible inundated areas, providing information to the local firms and the authorities for emergency responses. In addition, the system will also show the location of potential for inundation outside the science park. For checking the effect of inundation on transport, three checkpoints (the pink circles in Fig. 14.2) were selected on the artery, Yongxing Road, where Taiwan High Speed Rail also passes by.

176

T.-H. Yang et al.

14.4.2 Business Continuity Plan and the Database Business Continuity Plan (BCP) consists of a series of actions of preparedness, information analysis, response, recovery and maintenance, providing constructive suggestions to decision-makers. The BCP for Huwei Science Park is to maintain the operation of public facilities and ensure the safety of the personnel and the building. On the contrary, the BCP for the firms inside the science park is to protect the equipment and the machine, to maintain the logistics, to alleviate the interruption of production and so on. In this study, we mainly focused on the BCP for the park. The BCP were created through interviewing and discussing with the response teams and the decision-makers, based on several different scenarios. These scenarios included regular maintenance, extreme heavy rain, torrential rain, strong wind, typhoon warning, potential for inundation, power supply abnormality, water supply abnormality, traffic interruption, communication interruption and the other abnormalities. Then the plans were well-categorised according to the scenarios and were stored into a database. We divided these scenarios into five different phases: general phase, operation of disaster response centre, operation of joint disaster response organisation, response phase and recovery phase, with more than 80 series of actions. This study utilized SQLite software to establish a database for organising the BCP. The database contained all the scenarios, the detailed actions of plan and the competent authorities. The scenarios and the corresponding detailed measures are linked together by introducing certain threshold conditions. The competent authorities automatically receive associated measure recommendations when the thresholds are met. With these relationships, the users can easily find the corresponding series of actions from the scenarios layer by layer.

14.4.3 Decision Support System The DSS not only regularly collects all the on-site observations and the model predictions from DayuSWS, but include information such as weather forecasts (rainfall forecast and typhoon forecast) and government announcements (typhoon warning and typhoon day off) from Government Open Data Platform. The Python language based DSS would automatically analyze all the information and retrieve relevant strategies from BCP database with specific criteria according to the threshold conditions of each series of actions: • forecasted accumulated rainfall: – – – –

more than 80 mm within the next 3 h or not more than 150 mm within the next 3 h or not more than 200 mm within the next 24 h or not more than 350 mm within the next 24 h or not

14 Integrations of an Early Warning System and Business Continuity …

177

• water level detected at the key locations (7 in the park and 4 at the bridges): – higher than the warning levels or not • model prediction from DayuSWS (in the park and on the artery): – inundated (water depth higher than 0.25 m) or not • typhoon warning announcement from Central Weather Bureau: – sea warning announced or not – land warning announced or not • typhoon day off: – day off called or not • power supply: – abnormal or not • communication: – abnormal or not • transport: – abnormal or not • other relevant notification updated manually. The contact information of every competent authority was recorded in the DSS, and therefore, the information could be sent to them automatically via mobile applications, email, text messages and so on. To increase the efficiency of emergency response, this study sends information through a communication app named Line which is the most popular communication app in Taiwan. Furthermore, the webpage for display all the detailed information was constructed by Django, a Python-based web framework. The webpage updates the information every hour using the most available observations and forecasts. The DSS considers three danger levels: green for general condition, red for warning condition and blue for recovery condition. Users can easily notice the priority from the colour of each suggestion of action.

14.4.4 Application Huwei Science Park and the firms inside the park have practised together and improved the resilience to disasters by means of repeated exercises. With the help of the DSS, the response teams and the competent authorities would automatically receive the optimum suggestions of action from the BCP for disaster mitigation and prevention in a timely manner. For instance, if the system receives a rainfall forecast that will be more than 150 mm rainfall in 3 h, the suggestions of emergency response

178

T.-H. Yang et al.

will be displayed on the webpage and sent to the response teams and the competent authorities via Line automatically. Based on the BCP for the park, the DSS will inform the director-general, Environment and Labour Affair Division (for safety and environment protection), Construction Management Division (for the inspection, maintenance and repair of public facilities) and Personnel Office and advise them to organise an emergency operation centre (EOC). The EOC is a temporary emergency response team, which consists of 6 members: Environment and Labour Affair Division, Construction Management Division, Public Property Management Division, Land Development Division, Security Guards and Personnel Office. The team should be organized in a designated time frame and all the members will be notified of their missions by the DSS system. For example, the member of the Environment and Labour Affair Division will be informed to communicate with the factories and enquire about damage, and to schedule emergency shuttle service bus for personnel in the science park and associated areas.

14.5 Conclusion Given information from a purpose-driven design flood routing model and the BCP, the decision makers can react to the disaster accordingly. The administration office of the park along with the factories conducted several flood drills and the results showed that the system improves the emergency preparedness and responses. The science park can strengthen their resilience to disasters through the system. The system and concept can be easily adopted and applied to other industrial parks because of its effectiveness and performance. Acknowledgements The research was part of plan (grant no. 106f600441) funded by Central Taiwan Science Park, Taiwan. This research is done in collaboration of Huwei Science Park, Central Taiwan Science Park, SGS in Taiwan, Taiwan Secom Company Limited, National Center for HighPerformance Computing, Taiwan Typhoon and Flood Research Institute and National Chiao Tung University. We would like to express our gratitude to their support.

References 1. Craighead CW, Blackhurst J, Rungtusanathan MJ, Handfield RB (2007) The severity of supply chain disruptions: design characteristics and mitigation capabilities. Decis Sci 38(1):131–156 2. Ando M, Kimura F (2012) How did the Japanese exports respond to two crises in the international production networks? The global financial crisis and the Great East Japan Earthquake. Asian Econ J 26(3):261–287 3. Komori D, Nakamura S, Kiguchi M, Nishijima A, Yamazaki D, Suzuki S, Kawasaki A, Oki K, Oki T (2012) Characteristics of the 2011 chao phraya river flood in Central Thailand. Hydrol Res Lett 6:41–46 4. Finch P (2004) Supply chain risk management. Supp Chain Manag Int J 9(2):183–196

14 Integrations of an Early Warning System and Business Continuity …

179

5. Pathak S, Ahmad MM (2016) Flood recovery capacities of the manufacturing SMEs from floods: a case study in Pathumthani Province, Thailand. Int J Dis Risk Reduct 18:197–205 6. Morgan SJ, Rackham RA, Penny S, Lawson JR, Walsh RJ, Ismay SL (2015) Business continuity in blood services: two case studies from events with potentially catastrophic effect on the national provision of blood components. Vox Sang 108(2):151–159 7. Unnikrishnan L, Nidhin SP, Chitranshi J, Kaur R (2016) Importance of robust contingency plan for IT companies. Int J Eng Manag Res 6(1):440–443 8. Matsushita N, Hideshima E, Taniguchi H (2017) The Mitigation effect of BCP on financial damage—an empirical study of the non-manufacturing industries in the great east Japan earthquake. J JSCE 5:78–86 9. International Standards Organisation (2012) Societal security—business continuity management systems – Requirements. ISO 22301:2012, London, British Standards Institution 10. Yang TH, Yang SC, Kao HM et al (2018) Smart water. Cyber-physical-system-based smart water system to prevent flood hazards. Smart Water, 3(1): 1–13 11. Rossman LA (2015) Storm water management model. User’s manual Ver 5.1, U. S. Environmental Protection Agency

Chapter 15

Natural Hazard Crisis Management Exercice at Metropolitis Scale: Methodolgy for Holistic Involvement of Municipalities Yannick Dorgigne, Morgan Abily, and Philippe Gourbesville Abstract For the French municipalities, crisis management system involves by law, two main responsible and decision-making actors who are the city Mayor, and the state representative at the county level, the Préfet [1]. Nonetheless, since the development of the French Metropolis entity system—cluster of geographically close municipalities aiming mutual management of sets of selected transversal urban functions—many competencies that were handled by the municipalities in case of crisis are now transferred to the newly created Metropolis entity. Municipal Council Safeguarding Plans (PCS) are established at municipalities level whereas, the Metropolis entity has to handle coordination and support of its application—without having the full legal responsibility which remains in metropolis Mayor hands-, and with limited means, whereas at the same time, solicitations at Metropolis level increase. PCS are the key strategic procedure in case of extreme climatic alerts and catastrophic events. The presented paper focuses on how, after having provided support to setup the PCS at municipal levels, the Metropolis of Nice Cote d’Azur design a crisis management exercise aiming to test basis and good reflexes of PCS application as well as to test their coordination. The exercise for PCS activation was based on intense rainfall and flooding event scenarios. This exercise took place on the 3rd of October 2018, echoing to the 3rd of October 2015 flood events in the French Rivera [2]. This paper focuses on the development—methodological and organizational aspects—of a crisis management exercise project dedicated to trained main stakeholders and raised awareness. The step-by-step elaboration process of the exercise is presented. Organizational, technical means such as scenario and numerical models elaboration and multi-objective matrixes are detailed. Main feedback enhanced in this paper Y. Dorgigne (B) Service Risques Majeurs Direction de la Prévention et de la Gestion des Risques, Metropole Nice Cote d’Azur, 06364 Nice Cedex 4, France e-mail: [email protected] M. Abily · P. Gourbesville Polytech Lab, Université Côte d’Azur, Polytech Nice Sophia, 930 route des colles, 06903 Sophia Antipolis, France e-mail: [email protected] P. Gourbesville e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_15

181

182

Y. Dorgigne et al.

are the results focused on the necessary equilibrium to achieve in order to adapt such type of exercises to different nature and organizational levels divergent objectives and political constraints shared—or not—by the Metropolis municipalities and by the Metropolis itself. The proposed and implemented method is synthesized as guidelines and can be used as template for management crisis exercise design. Keywords Crisis management · PCS · Civil protection · Drill exercise design · Guidelines

Glossary COD DirAnim DPGR MNCA PAPI CCP PCM PCS ROE

Departmental Operational Centre Animation Direction Risk Prevention and Management Department Metropolis Nice Cote d’Azur Flood Prevention and Action Program Communal Command Post Metropolitan Coordination Station Council Safeguarding Plans Return Of Experience

15.1 Introduction One of the most effective tools for validation of the operational, efficient and intuitive nature of a Municipal Backup Plan dedicated to crisis management is to carry out a real size exercise. Well prepared, adapted to the points to be analysed, an exercise makes it possible to highlight many points regarding the strength and weaknesses of the Backup Plan. The organization of the various actors (actors are here defined as people or entities that participate in the organization or participate to the exercise such as local first responders’ teams, technical services from municipalities, emergency services, representatives from municipal authorities and central government, NGOs, etc.) and their interactions are complex during a crisis. The various interactions must be clearly articulated in order to ensure efficiency of the Backup Plan application. The implementation of an exercise brings first an assessment of the community that is responsible for the development of the action plan dedicated to crisis management. In the French regulation framework, these action plans are defined at the municipal scale and are known under the name of Plan Communaux de Sauvegarde (PCS). One of the main interests of an exercise is to assess the effectiveness and pragmatism of a PCS. Indeed, if the plan is not properly rooted, shared and consensual, the crisis site will have no chance of being clearly laid out and all the planned management procedure

15 Natural Hazard Crisis Management Exercice …

183

will be flawed or even fully irrelevant. Another major interest for an exercise is to test the communication protocols between the different actors. In a concrete way, if for simple crisis situations the problem can be solved with local resources, most of the time complex and major events require coordinated actions of very diverse actors who are operating with varied modes and who may have concurrent objectives. A consistent action plan should integrate all dimensions and lead to a shared consensus accepted by all. Therefore, the communication issue among all the stakeholders is a major component within the crisis management process. Finally, an exercise offers the possibility to test the planned emergency organization under very unlikely situations and cases allowing to create or recreate increasing constraints. The idea is there to push the actors to their limits by simulating situations that they will probably never have to manage due to the magnitude of constraints. However, this practice contributes to, develop actors’ adaptive capacity by using the available tools and also by developing innovation solutions that could match the extreme crisis situations. The approach focuses on risks assessment regarding inadequate communication protocol, demotivation in front of the tasks to be accomplished, etc. The Métropole Nice Côte d’Azur (NCA) is the metropolis, an intercommunal structure, centred on the city of Nice and federate 49 municipalities. It is located in the Alpes-Maritimes department, in the Provence-Alpes-Côte d’Azur region, southeastern France (Fig. 15.1). The total population is about 544, 000 in 2014, of which 390,000 for Nice city itself. If the tourist image of the French Riviera is well known at the international level, the territory the metropolis is characterized by very serious threads related to natural hazards: flash floods, landslides, inundations, wild fires, earthquakes are some the identified phenomena that take place regularly within the area. During the last years, several major events like the 2015 flooding event (reference) have targeted the urban environment and have led to major impacts on population and goods. In order to prevent such extreme situations, NCA has established an operational intercommunal organization dedicated to the crisis management. In order to validate and improve efficiency of the designed procedures, the use of simulation exercises has been promoted and accepted by all NCA municipalities. Within this context, a full-scale stage test was carried out on October 3rd , 2018. The defined scenario was on a scale beyond municipal territories in order to have actors facing unusual situations and to assess the different levels of crisis management. Voluntarily destabilizing, the intercommunal scale exercise tended to bring constructive elements to a new methodology developed elsewhere. It also made it possible to assess Communal Safeguard Plans (PCSs) at different scales, different maturity levels and different aims.

15.2 Genesis of the Exercise and Network of Participants The preparation and implementation of the exercise was only made possible by the trust and working habits gradually woven between all the actors. Initially based on a program bringing together 16 players—person or entity who participated to the

184

Y. Dorgigne et al.

Fig. 15.1 Extension of Nice Côte d’Azur Metropolis, France

exercise—from the lower valley of the Var river, as part of the Flood Prevention and Development Plan (PAPI Var), contracted in 2009—PAPI 1 agreement signed on 07/24/2009 [3]—a technical network of actors in charge of risk management in their municipality has decided to share experiences by relying on the new means brought by the Nice Côte d’Azur Metropolitan Area (MNCA). The services of the Metropolis have been found legitimate to coordinate the network activities since they carry transversal actions in particular on the realization and the homogenization of the PCS (action initiated in 2014—action 2.1 PAPI Var 1). However, the network put in place reflection and actions which went far beyond the framework of the PAPI to encompass broader issues related to risk management, from prevention to alert, from action to recovery, and this, through unanimous consensus and brought naturally. It should be noted that the municipalities that makes up the MNCA participating in the exercise are of very different sizes and means. The smallest municipality playing in the exercise has 49 inhabitants while the city centre of Nice gathers around 390,000 inhabitants [4]. Obviously, the means are not comparable, nor the issues: small municipalities have limited technical and financial resources compared to larger municipalities. What brings these communities together are the means transferred

15 Natural Hazard Crisis Management Exercice …

185

to the agglomeration (metropolis entity) and the growing interdependence on various themes which linked them such as communication networks, means of rescue and intervention, tools for monitoring floods or other phenomena, etc. [5]. However, the culture of risk, an essential factor for effective management of potentially crisis situations, is generally found more anchored in small municipalities compare to larger agglomerations. Lack of specific crisis management resources and closer direct observation and contact to the environment and of its variations and excesses makes the populations of the small communes expect less from the public authorities. Observed tendency in these remote and small communities is to put themselves in safety and to adapt their attitudes toward natural risk. This technical network, based entirely on the voluntary commitment of participants, enriches all participants. The smallest municipalities can benefit from pooling and lower costs as well as the skills of specialized technicians from more structured municipalities. The more structured can benefit from the common sense, the pragmatism and the support (paradoxically) of the more modest municipalities. The concept of project team therefore deserves to be studied with particular attention, because it allows enriching team competences by individual experience and strengthen efficiency by sharing knowledge. For crisis management coordination action in general and for the exercise preparation phase, meetings the place almost every quarter on various and varied subjects, always with a view to mutual enrichment, learning from each other in its differences, since in times of crisis get to know each other better is to be more efficient and united. As announced above, the members of the network represent municipalities of very different sizes and means. Moreover, those municipalities have very contrasting geographic situations such as coastal environment, hilly environment or mountainous area environment in the upstream part of the territory. They are consequently exposed to very different types of natural hazard. Furthermore, the hazards that can affect them are also quite disparate depending on the location, if we exclude the widely shared seismic risk as well as the risk of forest fire. However, it seems that a common base stands out. Indeed, if avalanches are located in mountains, marine submersion phenomena on the coast and technological risks in urban and peri-urban areas, extreme weather phenomena affect the entire territory, with variations depending on geographic areas. For instance, when only taking into account the flood hazard, the geographical and physical situation of MNCA territory can cover almost all the types of flood hazard with various possible concomitant combinations being possible (flash flood, river flood, storm surge, ground water table rising, etc.). A first point of convergence stands out for the metropolitan network: a generalized increase in extreme weather phenomena, slightly variable depending on the sector, but often generalized and impacting at the same time, but to varying degrees, the entire territory of the metropolis [6]. However, the means of anticipation and monitoring of these phenomena, as well as the reaction and action to reduce their impact, must be shared by all the actors, members of the network.

186

Y. Dorgigne et al.

15.3 Implementation of the Exercise In 2016, during one of the pluriannual network’s technical meetings, one of the network member offered to work on the feasibility of a crisis management exercise. This proposal immediately received general approval of the network. From this starting point, the Metropolitan services have been responsible, within the framework of their assistance and advice missions, to work on this crisis management exercise feasibility, and to elaborate and propose a roadmap.

15.3.1 Validation of the Partnership Principle To initiate a process of participation for the partners, the starting step was for the network to set-up the objectives of the exercise. The global objectives chosen by the partners were to test during the exercise following points: 1. Alerting process of actors who have to come in CCP; 2. Arming of Communal Command Posts (CCPs); 3. Communicating between the Communal Command Stations and the Metropolitan Coordination Station; 4. Communicating between the Communal Command Posts and the Departmental Operational Centre; 5. Using of a common weather forecast tool; 6. Communicating between the Communal Command Posts and the metropolitan subdivisions. As soon as the global objectives were shared, it was proposed to all the members of the network to clearly state whether or not they wished to participate in the technical working group that was going to set up the concrete course of the exercise. The only prerequisite for group participation was that on D-Day the people directly involved in the crisis exercise elaboration could not play. They still had the roles of observers or facilitators.

15.3.2 Roles’ Distribution Within the framework of its advisory and assistance missions, the Department of Risk Prevention and Management (DPGR) of the MNCA continued to fulfil its transversal role of animation and steering in both exercise elaboration and exercise playing phases. This role was essential to ensure the widest possible portage. This is in a way the common point of all the participating structures, not only the municipalities but also the external partners involved in the exercise such as academics or state representatives). Identifying a pilot who federates around him a shared consent is the first cornerstone of the exercise.

15 Natural Hazard Crisis Management Exercice …

187

15.3.3 Preparation for the Exercise A small technical working group brought together some representatives of municipalities of different sizes, intermunicipal and academics (Polytech Sophia Antipolis). This group’s mission was to deepen the organization of the exercise and define the scripts. Updates, in order not to keep alive interest in the exercise toward the future participants and observers, were made every 6 months. The exercise had to adapt from the start to the various levels of involvement of the participants. In fact, not everyone had the same level of experience in taking risk into account and it took an exercise commensurate with their ambitions. In addition, it was necessary to keep in mind the testing of the method in the background (while avoiding that the partners do not focus on it but rather participate in its elaboration by building the response collectively with for strength the differences of each one, but which overlap with the main principles. Three progressive levels have been proposed: • Level 1 = Common base: 6 municipalities Arming the CCP PCC and PCM communication Coordination between PCC and PCM • Level 2: General staff exercise (specific scenarios by municipality) 4 municipalities (Players A, B, C and E) • Level 3: Staff exercise + real action (s) in the field 2 municipalities (Players D and F).

15.3.4 Observation and Assessment Procedure In order to verify the impact of the exercise on the different organizations and analyse the relevance of these, it was necessary to find an evaluation grid allowing to verify the relevance of the PCS according to three axes: their operational, functional and intuitive character (Fig. 15.2). First, a definition had to be defined for each of the criteria, so that all the assessors had a common and comparable approach. In a second step, the analysis grid was developed in order to assess the operational, functional and intuitive characteristics of each of PCSs. This grid sought to list and evaluate all the important constituent elements of the PCS. It was designed to help identify the orientations of the Communal Plan for the Protection of the municipality observed along the three axes. Two months were necessary to list as exhaustively as possible all the criteria to be analysed, then to propose a weighting of the three axes tested

188

Fig. 15.2 Example of the evaluation grid used (1st page)

Y. Dorgigne et al.

15 Natural Hazard Crisis Management Exercice …

189

Fig. 15.3 Performance of a CCP according to defined criteria Fig. 15.4 Overall performance of a CCP axis by axis

for each of the criteria. Illustration of these outcomes for anonymised CCP is given as examples in Figs. 15.3 and 15.4. Finally, the grid was tested on several fictitious cases to verify that it could adapt to different PCSs without rough results. Weights were also tested, and confronted with various crisis managers to ensure that there was no unrealistic effect of the assessment. Training then made it possible to detect students who would be entrusted with the assessments of the various players. In fact, each participant was assigned at least two totally external assessors, who made themselves invisible during the exercise, in order to assess, according to the grid, the communal command posts. In total, there

190

Y. Dorgigne et al.

were 27 observers allowing us to analyse and use as many informed evaluation sheets. Subsequently, the collected results were processed in order to better analyze the PCS and suggest areas for improvement. Analyzes are still in progress. Individualized refunds are planned, municipality-by-municipality.

15.3.5 Scripts Used for the Exercise For the second and third levels of the exercise, a script specific to each municipality was developed (Fig. 15.5). The preparation working group set out to integrate injections that correspond as closely as possible to the points that the municipalities wanted to test. An animation department (DirAnim) has been set up. DirAnim allowed contacting the players during the exercise in order to send them the information they needed to process. In return, when the players wanted to contact actors in crisis management (state services, Météo France, water managers, electricity, gas, etc.), They contacted DirAnim using a directory dedicated to the ‘exercise. 12 facilitators were mobilized (Fig. 15.6). To make the scenarios more realistic, it was decided to carry out maps making allowing to simulate the temporal evolution of the water level on the two rivers affected by the floods and generating inundations. These maps were also sent to the Communal Command Posts (PCC) affected by the scenarios, every 30 min. To make the exercise realistic, for some of the crisis management actors, videos of simulations

Fig. 15.5 Extract from the script

15 Natural Hazard Crisis Management Exercice …

191

Players

Total of inputs

Hours

Inputs / hour

% (Players inputs/total of inputs)

Comments

1st level players

7

2

3,5

7%

All wish to play exercise at staff level

A

19

3h30

5,4/h

19%

Could have managed more inputs

B

19

3h30

5,4/h

19%

C

30 + maps

5h

6,0/h

30%

D

22 + maps

6h

3,7/h

22%

E

18+ maps

5h

3,6/h

18%

F

26 + maps

6h

4,3/h

26%

Note that there are some inputs to add for specific cartographic (map) (1 input every half hour )

Total number of inputs within the script: 99 including 7 maps

Fig. 15.6 Summary table of inputs

where produce. These videos were reproducing views of simulated river flood from a municipal control camera perspective and the use of this Augmented Reality setup was a first for this type of crisis event exercise at the municipality (Figs. 15.7 and 15.8).

Fig. 15.7 Example of fictitious flood mapping used within scenario

192

Y. Dorgigne et al.

Fig. 15.8 Example of a fictitious flood mapping used within scenario

These maps were generated by the water department of Polytech Sophia Antipolis that modelled fictitious inundation scenarios, inspired by the flood prevention plan of the lower Var valley for one of them [7]. These flood simulations were shared to the players of different levels using: videos, leaflets on the field and at some specific points Augmented Reality glasses and tablets were setup and used to better understand spatial extension of inundated areas. The idea was to make the exercise as realistic as possible by using these numerical modelling tools and digital devices for understanding the various inputs.

15.4 Achievement on Exercice Day A schedule has been drawn up to monitor the completion of the exercise on the 3rd of October 2018 (Tables 15.1 and 15.2). Common sense is not always shared, and assistantship very often undermines the effectiveness of optimal crisis management. Indeed, it can be implicitly noted that as soon as a system is on a “wait-and-see” mode, it loses its functionality as the analysis of the evaluations of the different CCPs has already demonstrated.

15 Natural Hazard Crisis Management Exercice …

193

Table 15.1 Exercise general planning Time

Script

Expected reaction

Real reaction

To -15 min

Message to the watchman in the participating municipalities

Alert and arming of CCPs Convocation of members

Good reactions

To

Beginning of the schedule

Physical arming of CCPs Contact with the Prefecture (COD) and the PCM 1st assessment of the situation

Beginning at To + 40 due to a technical problem in DirAnim Physical arming and first contacts ok 1st assessment more difficult

To + 25 min

Start of entries

To + 30 min

First send of mapping of the flood model for 2nd and 3rd level

End of PCCs’ arming Minimal objective: alert and convocation of members of the CCPs, information of COD and PCM

Every PCCs armed

To + 40 min

Real beginning of the script

1st responses to script

Unfolding of the script, city by city

To + 1 h

First end of exercise for 1st level

To + 2 h

Last end of exercise for 1st level

To + 2h30

Presence of the Deputy Prefet on the field End of exercise for 1st level

Press point (TV, web, written press, etc.)

Protocol and press sequence

To + 3h30

First end of the exercise for 2nd level

To + 5 h

Last end of the exercise for 2nd level First end of the exercise for 3rd level

End of exercise for 2nd level

On time

To + 6 h

Last end of the exercise for the 3rd level

End of exercise for 3rd level

On time

D + 30

/

ROE with animators and observers

Concluded

D + 90

/

ROE with all players together

Concluded

D + 6 months and more

/

Beginning of the ROE with all player on by one

Ongoing

Protocol sequence Speech by the Mayor, the Deputy head and the state’s representative

194

Y. Dorgigne et al.

Table 15.2 Summary table of general involvement in the exercise and roles’ distribution Role

Players (entities)

Animators

Observers

Breastplates

General pilots

Total

Workforce

12 municipalities + COD (about 130 physical players involved)

12

27

30

3

About 200 players in the exercise (not including media and protocol)

15.5 Lessons Learn Several feedbacks under the form of characterized Return of Experience notes (ROE) have been carried out with different stakeholders such as ROE with the municipalities, with the DirAnim actors, evaluators and with the partners. While all were generally satisfied with the process and the results, interesting issues appeared for highly valuable for continuous and future improvements.

15.5.1 Hot ROE As soon as the exercise was completed, hot ROEs were planned in each CCP. Thus, the first lessons revolved around analyses, constructive criticism and proposals essentially focused on the organization of the exercise and the functioning or the lack of tools. Hot ROE is useful but only concerns elements of immediate apprehension: – – – –

good or bad functioning of the means of communication, inconsistencies in the scripts and their sequences, interest of the injections, insufficient or too large a number of injections depending on the municipality, delay in the start of the exercise which postponed the course and forced to shorten certain parts, – protocol difficulties, – logistical improvement points in the very organization of the exercise (meeting of actors, assessors and breastplates …). An analysis requiring a minimum of hindsight is therefore not possible. Indeed, this ROE, although important to improve certain tools for example, does not allow an analysis on the answers that the players were able to find and even less the relevance of their organization. It is of course necessary to list the ideas emitted hot, but not to postpone at the risk of moving away from a more relevant interest. They are generally

15 Natural Hazard Crisis Management Exercice …

195

very concrete and of little pragmatic scope. In addition, fatigue and stress can distort exchanges and give counterproductive results (misunderstanding, misunderstanding, blockages, etc.). This hot ROE is therefore essential, but its lessons should not be analyzed until much later, with a rested head. Better to allow a little time for the tensions to calm down to have a “warm” ROE, without waiting for too much time to have passed (different from cold ROE).

15.5.2 Cold ROE Scheduled ROEs one month after the exercise allowed for a more focused analysis of the weaknesses and strengths of the various organizations, as well as areas for improvement. To facilitate the analyses, the cold ROEs were done at different levels: – At the level of the participating municipalities in a collegial manner; – At the level of the students who played the roles of assessors, animators and breastplates (those involved on the site where the field part of the exercise was played); – At the level of the services of the Metropolis; – With the participating services of the State in charge of crisis management; – Then individualized ROE allowing going into more detail without gene or selfcensorship were organized and are still in progress. Indeed, the first lesson of this exercise is the great variability of approaches, objectives and strategies (assumed or not), depending on many factors and not necessarily shared at first, by listing in a non-exhaustive manner: – – – – – –

Size of the municipality; Geographical characteristics of the municipality; Communication infrastructures (roads, accesses); Experience in crisis management; Political will; Skills and competences of the actors.

In addition to the analysis of the application of the method to the PCS of communities with unequal means and different sizes, the interaction between these different scales brings very useful lessons.

15.6 Analysis and Perspectives The participating municipalities unanimously accepted the exercise and saw the interest to commit themselves to this training action. The participants who played at the first level even wanted to engage a new exercise where they could play at staff exercise level with a script dedicated to validate their organizations and actions.

196

Y. Dorgigne et al.

During the exercise, the municipalities met all the basic objectives planed during the preparation phase: – – – –

Arming of the PCC within a reasonable time; Interacting smoothly and efficiently with other municipalities; Producing first situation assessment in all CCPs; Deactivation of PCCs was carried out properly.

For the municipalities that had to interact according to the scripts, the reactions were good. 95% of the inputs were played and resulted in treatment. These actions allowed players to optimize their organization their PCS for critical weather conditions and to identify axis for improvement. The main improvement axis focused on the communication protocol between state services—or state managed services— and the municipalities. In fact, during the exercise, few communications were established, and limited interactions took place. Of course, these interactions are limited by definition during an exercise. However, several potential difficulties have been identified: – – – –

Loss of contact means telephone numbers or email addresses for key actors; Wrong contacts information; Failure to reach state services for the municipalities; Failure to report efficiently the first messages from the municipalities to state services that have been therefore de facto excluded from this part; – Misunderstanding during the initial briefings of players from the state services. In real cases, the same communication difficulties are frequently observed and, consequently, coordination, between the municipalities and the state services is deeply affected. There is a break within the process and between the actors who can be described as “field actors” in the sense that they are in direct contact with the phenomena, and the external actors who often do not, paradoxically, have a clear vision on problems to be solved. The loss of references related to concrete realities often detracts from the effectiveness of support from reinforcements when they should allow a welcome help. This situation happens at least in the early hours of the crisis, when there is still a local structure able of handling the phenomena. One of the solutions, like implemented by Quebec organization [8], would be to keep a local management, as long as it is possible, with the provision of reinforcement means, including state ones, to the local manager (the mayor, Director of Operations in his town). To ensure efficiency, when the phenomena are affecting several territories and administrative entities, the coordination of operations and resources does not necessarily have to move to the higher level such as the state level. The implemented exercise demonstrated all its interest by first allowing the actors to get to know each other better as well as to train and get use to the crisis management processes and means. When the protagonists are used to working together, communication is desecrated, improved and the hindsight needed for an efficient management is facilitated. During a real event, there can appear an additional stress when discovering interlocutors when everything is destabilizing. Training exercise with its preparation and actuation aims to reduce this eventuality. In addition, it allowed,

15 Natural Hazard Crisis Management Exercice …

197

within the communities, to raise awareness of actors who had little or no knowledge in crisis management. Their participations in the ROEs have contributed to improve the PCS by bringing constructive external views. This approach increases both functionality of the PCS, but also functionality and intuitiveness. Indeed, effective crisis management must allow approaching a new equilibrium as soon as possible, acceptable to the largest number of actors. Relying on pre-existing operating modes has, at least, the interest to reduce and minimize the destabilisation process affecting an organisation. Producing a clear and operational action plan will contribute to increase resilience. The use of an evaluation grid adapted to the territory of the exercise has enabled a real in-depth analysis of the organizations setup and efficiency. The results of this analysis are still in progress. Nevertheless, the first interesting outputs for optimizing organizations are emerging, and often without any prohibitive financial commitments. These tracks most often lead to integration as early as possible of all the actors: a crisis organization must not be cut off from the so-called daily organization of systems.

References 1. République Française (2004) Loi de modernisation de la sécurité civile, Journal officiel de la République Française 2. Dorgigné Y, Abily M, Salvan L, Gourbesville P (2018) Creation and life of an operational crisis management center in Nice Metropolis: consolidation of flood events handling using feedbacks following the 3rd October flood event, In Advances in Hydroinformatics. Springer, Singapore, pp. 483–496 3. République Française (2009 & 2011). Programme d’Actions de Prévention des Inondation du Var 1—24 juillet 2009/Programme d’Actions de Prévention des Inondation du Var 2—18 avril 2011. Recueil des actes administratifs de la Préfecture des Alpes-Maritimes 4. Métropole Nice Côte d’Azur (2019) Site officiel de la Métropole Nice Côte d’Azur. www. nicecotedazur.org/la-metropole/l-institution/les-chiffres-clés 5. République Française (2018) Arrêté portant modification des statuts de la métropole Nice Côte d’Azur—7 novembre 2018—Actes administratifs de la préfecture des Alpes-Maritimes 6. Préfecture des Alpes Maritimes (2016) Dossier Départemental des Risques Majeurs des Alpes Maritimes—Cyprès 7. République Française (2011) Plan de Prévention du risque inondation de la basse vallée du Var, 18 avril 2011. Recueil des actes administratifs de la Préfecture des Alpes-Maritimes 8. Gouvernement du Québec (2018) Règlement sur les procédures d’alerte et de mobilisation et les moyens de secours minimaux pour protéger la sécurité des personnes et des biens en cas de sinistre—Gazette officielle du Québec

Chapter 16

Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland) Charlotte Dreger, Erik Bollaert, Olivier Stauffer, and Yves Châtelain

Abstract The Tinière torrent is located in the western part of Switzerland and has its exutory in the Lake of Geneva. In 2006 and 2007, major flood events generated deposition of 10,000 m3 of solid material along the canalized part and more than 100,000 m3 in the upstream ravines. These deposits have caused the inundation of urbanized areas as well as erosion of torrent banks upstream. Also, several bridges have been destroyed. Detailed hazard mapping has shown significant risk of damage to the federal highway A9 as well as to the main railway and cantonal highway crossing Villeneuve for flood events between 100 and 300 years. Hence, three check dams have been constructed in the upstream part of the torrent. The retention volumes created by these dams aim at retaining the major part of the solid material that is being transferred during major flood events. However, during construction of the check dams, in July 2013, a major flood event has occurred at the construction site. A series of debris flows filled up the volume behind the most upstream located dam (~5,000 m3 ). This structure was not yet finished at the time of the event, only the concrete core of the dam was put into place. The event has caused damage to the structure but has saved Villeneuve from potential damage. In the following, the 2013 flood event and its crisis management are presented, together with the behavior of the partially constructed check dam and its positive impact on downstream safety. Keywords Check dam · Debris flows · Extreme flood during construction · Dam behavior · Hazard mapping

C. Dreger (B) · E. Bollaert AquaVision Engineering, Vaud, Switzerland e-mail: [email protected] E. Bollaert e-mail: [email protected] O. Stauffer · Y. Châtelain Etat de Vaud, Vaud, Switzerland e-mail: [email protected] Y. Châtelain e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_16

199

200

C. Dreger et al.

16.1 Introduction The Tinière torrent is located in the western part of Switzerland and has its exutory in the Lac Léman. The torrent has a catchment area of about 11 km2 and has been canalized in the past over its most downstream 1.2 km, because passing through the heavily urbanized village of Villeneuve. The torrent has average slopes of 35% all upstream and of 11% in its most downstream part (Fig. 16.1). The catchment area contains 19 ravines (Figs. 16.2 and 16.3). All of these ravines are able to generate significant sediment transport as well as debris flows during flood events. In July 2006 and August 2007, the Tinière torrent has been particularly active, with several major flood events generating the displacement and deposition of about 10,000 m3 of solid material along the canalized part of the torrent and 50,000 m3 from the ravines upstream into the main bed of the Tinière torrent. These deposits have caused the inundation of part of the urbanized area close to the Lac Léman as well as significant local erosion of the torrent banks more upstream. Also, several bridges along the upper part of the torrent have been completely destroyed, in particular along the Plan Cudrey ravine. Following these extreme events, a series of emergency measures have been taken, such as the rehabilitation of eroded weirs, banks and thalweg, the protection of torrent banks, the reconstruction of destroyed bridges, etc. Second, it has been decided to perform a detailed study of the geo-morphological behavior of the whole catchment area [1]. Detailed analytical and numerical modeling of both hydraulics and sediment transport and/or debris flow behavior have been performed to define appropriate countermeasures to protect the urbanized part of the Tinière torrent in Villeneuve (Fig. 16.1) [2].

Fig. 16.1 Debris flow deposits along the Plan Cudrey ravine

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

201

Fig. 16.2 Plan view of the catchment area of the Tinière torrent, showing the upstream and middle reaches, the downstream dejection cone towards the Lac Léman (canalized reach), and the three check dams constructed Middle reach

Downstream reach

Upstream reach

Elevation [m]

1300 1200

FLO-2D model beginnig

1400

Cudray upstream

1600 1500

Freeway

Cantonal highway

1700

Cudray dowstream

1800

1100 1000

Canalized segment

900 800 700 600 500 400 300 200

Average slope: 11%

100

Average slope: 16%

Average slope: 33%

0

0

1000

2000

3000

4000

5000

6000

Distance to Lake Geneva [m]

Fig. 16.3 Longitudinal profile of the Tinière torrent, showing the upstream and middle reaches, as well as the canalized part towards the Lac Léman

202

C. Dreger et al.

16.2 Hydrology Flood events have been defined as a function of flood return period, with typical values of 29 m3 /s for the 100-year flood event and 59 m3 /s for the extreme (300year) flood event. The catchment area can be divided into 18 sub-basins (Fig. 16.2). However, only 5 of them have a clear waterway, although they do not have a constant source of water. Consequently, discharges are directly related to rainfall events and snowmelt. Its highest elevation is 2042 m and the lowest elevation is 372.30 m (lake level). It is possible to divide the torrent in three main reaches with average slopes of 11, 16 and 33%. These three reaches have the following features: • Downstream reach: Dejection cone corresponding to the canalized part of the torrent crossing the village of Villeneuve. Main infrastructures present are the Freeway A9, the Cantonal Road, the railway and a series of local bridges. The slopes are between 7 and 10%. • Middle reach: Starts at the upstream end of the canalized part and continues until 4 km away from the lake. At km 2.2 is located Plan Cudrey, a small village adjacent to the Tinière torrent, as well as the junction with the Plancudray ravine. • Upstream reach: Natural part of the torrent with an average slope of 33% and a lot of important ravines that join the Tinière and are able to transfer large amounts of solid material to the torrent.

16.3 Sediment Transport 16.3.1 Historical Events Historical events along the Tinière torrent have been described in a study performed by [3], which has mentioned 25 significant flood events in the Tinière basin between 1707 and 1995. The gravity of flooding in the downstream reach depends on how much material can be triggered from the tributary ravines. Unfortunately, neither discharge nor material volume measurements are available. [3] presents a flood-intensity determination criterion and shows that the July 2006 event would be classified with an intensity level of 4 on a 1 to 5 intensity scale. In addition, the return period was estimated to be no more than a 30-year-flood, which may correspond to a relatively frequent flow or mud flood event. From this study, it appears that floods with solid material transport tend to occur in groups of 2 to 3, with an interval of only a few days or weeks, like observed in 1846, 1927, or 1932. It seems that the solid material transferred by the ravines into the Tinière torrent by debris flows cannot be completely transported downstream in a single event. Hence, a series of subsequent events displaces towards downstream what is left. Accordingly, it is possible to establish a sediment activity as follows:

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

203

Fig. 16.4 Volumes of solid material deposited along the Tinière torrent following the July 2006 extreme event [6]

• Trigger event: debris flow from rills and gullies, resulting in deposition of alluvium material in the middle reach and only a percentage of it is carried to the lake. • Subsequent events: sediment-laden or clear water inflows able to displace and transfer available solid material towards the canalized part of the torrent. Regularly observed flood events resulted in the past in the construction of a large series of thalweg stabilizing weirs. (Fig. 16.4)

16.3.2 Sediment Transport Capacity The theoretical sediment transport capacity has been computed based on the Smart & Jaeggi and Meyer-Peter & Muller formulae. Some of the results are shown at Table 16.1 and, for the downstream canalized part of the torrent, indicate transportable volumes of about 65,000 m3 for the 30-year flood event, about 70,000 m3 for the 100year event, and finally about 140,000 m3 for the extreme flood event. Corresponding volumetric sediment concentrations are around max. 10%. It can be seen also that, in the upper reaches, the theoretical transport capacity is much higher. This tends to state that, during extreme events, a lot of solid material will deposit along the thalweg due to these slope changes. The event of 2006 has confirmed this hypothesis. Based on field measurements, bed grain size curves have a d50 between 1.2 and 5.1 cm and a d90 between 6.5 and 22.5 cm (Figs. 16.5 and 16.6).

204

C. Dreger et al.

Table 16.1 Theoretical debris flow capacity along the Tinière torrent for different flood events and reaches

Slope D50 [cm]

Critical discharge 3

[m /s,m]

Cv Tognacca 1999

Takahashi 1991

0.06 0.06 0.06

0.08 0.08 0.08

0.16 0.16 0.16

0.23 0.23 0.23

0.50 0.50 0.50

0.71 0.71 0.71

0.60 0.60 0.60

0.86 0.86 0.86

9%

1.2 5.1 2.5

0.09 0.76 0.27

1.2 5.1 2.5

0.05 0.40 0.14

1.2 5.1 2.5

0.02 0.17 0.06

1.2 5.1 2.5

0.01 0.08 0.03

16%

33%

63%

Event HQ30 HQ100 EHQ

Solid volume [m3] Togn. Taka. 145' 125 225' 410 185' 300 151' 391 235' 142 193' 300 305' 325 474' 236 389' 800

Water volume [m3]

Average Conc.

775' 051

0.20

808' 600

0.20

1'630' 600

0.20

16.3.3 Debris Flow Capacity Second, based on the observation of small debris flow events near the canalized part of the torrent, the theoretical debris flow transport capacity has also been computed based on the [5] and [7] formulae. The results are shown at Table 16.1, which presents the critical discharges, sediment concentrations and transportable volumes calculated based on [5] and [7] for different reaches: canalized, middle, upstream and, Plancudray ravine. The presented volumes are valid for the middle reach, showing a transport capacity between 150,000 and 230,000 m3 for the 30-and 100-year events. This is in reasonable agreement with the results for sediment transport of Table 16.1 for the same average slope of 16%.

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

205

Fig. 16.5 Location of 2061 and 1977 check dams and available sediment deposition volumes on the Tinière torrent

Fig. 16.6 Plan view and longitudinal section of 2061 check dam

16.3.4 Available Solid Material The solid material that is available over the whole catchment area has been defined based on an extensive field study of all ravines and of the Tinière torrent itself [6]. Then, an occurrence probability was assigned for every flood. Figure 16.7 illustrates the volumes and their given probabilities placed in the torrent profile. It can be seen that, if all ravines inject solids into the Tinière torrent, several hundred thousands of m3 of material can theoretically be supplied to the system. In practice, however, a probability study has shown that, on the average, max. 3–4 ravines are triggered during an extreme event, with an average solid volume of about 24,000 m3 per ravine.

206

C. Dreger et al.

Fig. 16.7 Location of 1000 check dam and available sediment deposition volumes on the Tinière torrent

During the July 7th 2006 event, three ravines were triggered by rainfall in the Tinière, with more than 60,000 m3 of total volume, which validates this hypothesis.

16.3.5 Plancudrey An important sediment source is the ravine Plancudrey, located at 2,200 m from the lake. This torrential system was activated in 2005 and 2007, generating an important debris flow, and bed erosion as a consequence. This ravine varies its slope from 72% to 63% at the junction with the Tinière. Unfortunately, there are not any water/sediment measures available, only a field observation has suggested that an available maximum volume of 100,000 m3 can be triggered by a 100-year flood.

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

207

16.3.6 Summary of Sediment Yield By comparison of the transport capacities and the supplyable solid material, a choice has been made of the quantities to be injected in the numerical computations. The results are presented at Table 16.1. For a 100-year flood event, the minimum and maximum possible material volumes that might be triggered and transported through the Tinière torrent are estimated between 70,000 m3 and 195,000 m3 . For the extreme flood event, comparable volumes are between 140,000 and 390,000 m3 .

16.4 Check Dam Mitigation Measures Three check dams have been constructed following the series of events between 2006 and 2007. The main objective of these dams is to retain as much debris material as possible during extreme events, but to allow regular sediment transport to be transferred downstream during normal floods, i.e. for return periods of a few years only (morphogenic events).

16.4.1 Location of Check Dams The location of the three check dams is presented in Fig. 16.2. Two dams are located about 2 km upstream of the village of Villeneuve, in the Plan Cudrey area, and one dam is situated just upstream of the village, at Cave des Rois. Hence, the dams are situated in the downstream part of the catchment area, allowing to be efficient in retaining debris material coming from the upstream ravines.

16.4.2 2061 Dam The first check dam is located just upstream of the confluence between the Tiniere torrent and the Plan Cudrey ravine. It aims at retaining debris flows coming from upstream of Plan Cudrey. As shown in Fig. 16.6, the dam has a height of 10 m and a crest length of 40 m, and is equipped with a central overflow spillway section of 10 m width and a main flow gallery through the main body of 20 m of length, 2.5 m of height and 5 m of width. The overflow section is equipped with a rock block protection layer. The bottom of the gallery is equipped with abrasion resistant granite stones. The upstream entrance of the gallery has a vertical trash rack made of 27 cm diameter steel cylinders placed one meter apart from each other (5 in total, spacing of ~ 50 cm). The retention volume is estimated at about 5,000 m3 .

208

C. Dreger et al.

Fig. 16.8 Cross-section of 1000 check dam

16.4.3 1977 Dam The second check dam is located some 100 m downstream of the first one (Fig. 16.5) and aims at retaining debris flow deposits coming solely from the Plan Cudrey ravine. This ravine is the most important of all ravines with an available stock of debris volume estimated at about 100,000 m3 . The dam has the same structure and sediment retention volume as the 2061 dam.

16.4.4 1000 Dam The third check dam is located just upstream of Villeneuve, near Cave des Rois. The dam and its retention volume are presented in Figs. 16.7 and 16.8. The dam consists of a debris trash rack placed upon a granite stone layer in between two vertical concrete walls. The dam has a width of 10 m and a height of 3 m. The trash rack is made of 16 steel cylinders of 19 cm diameter placed 50 cm apart from each other (spacing of ~ 30 cm). The upstream retention volume has been created by a significant widening by excavation of the left river bank. The volume is estimated at about 4,000 m3 in total.

16.4.5 Hazard Map The hazard map obtained after construction of the three check dams is illustrated in Fig. 16.9. Almost the entire region of Villeneuve is now protected and situated in “residual” danger (dashed surface), i.e. only generating acceptable damage during an extreme flood event. The areas containing intermediate or high danger (dark grey) only appear inside the main course of the Tiniere torrent.

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

209

Fig. 16.9 Hazard map of Villeneuve region following construction of the check dams

16.5 The July 2013 Debris Flow During construction of the first (i.e. most upstream) check dam, in July 2013, a major debris flow event has occurred at the construction site. A series of severe debris flows filled up the volume behind the 2,061 dam (~ 5,000 m3 ). This structure was not yet finished at the time of the event, only the concrete part of the central overflow section of the dam was put into place and anchored into the rock. Figure 16.10 shows the blockage of the gallery and the uncontrolled overflow over the non-finished and unprotected main body of the dam. The return period of the event is estimated at only about 20–30 years, nevertheless because of the non-finished state of the dam the event has caused damage to the structure. At the same time, it has saved Villeneuve from another major debris flow event with potential damage to urban areas, as well as allowed to prove the debris flow concept put into place by the check dams. The total volume of deposited sediment and debris in the Tiniere torrent was estimated at 12,000 m3 , as such about half of this volume was retained upstream by the first check dam under construction.

16.6 The July 2014 Flood Event Near the end of construction of the three check dams, in July 2014, a flood event has occurred, mainly involving sediment transport. This event partially filled up the volume behind both the 1,977 and 2,061 dams (~ 5,000 m3 ), and completely

210

C. Dreger et al.

Fig. 16.10 Photos of the 2013 debris flow that occurred at the dam site during construction phase

blocked the trash rack at the 1,000 dam by woody material (Fig. 16.11). The return period of the event is estimated at about 5 years [8]. The three retention structures were able to retain most of the transported sediments, as well as almost the entire volume of wooden debris. This has prevented downstream bridges from clogging and inundations in the village. Also, a real-time intervention of the local construction

Fig. 16.11 Photo of the 2014 wooden debris that clogged the 1,000 check dam

16 Check Dam Behavior Under Extreme Circumstances at Villeneuve (Switzerland)

211

Fig. 16.12 Photo of the 2014 sediment deposits in between the 1,977 and the 2,061 check dams

company, during the whole night, has allowed preventing the gallery of the upstream check dam to be clogged (Fig. 16.12).

16.7 Conclusions This article presents the main aspects of a flood and debris flow protection project of Villeneuve, situated on the Tinière torrent in western Switzerland. Based on extensive 1D and 2D numerical modeling of flood and mud flow events, three check dams have been constructed on the main course of the torrent, allowing to retain up to 14,000 m3 of sediments and debris material during flood events. The concept and principal dimensions of the dams are presented, together with their behavior during two distinct flow events that occurred during dam construction: a major debris flow in July 2013, mainly retained by the concrete structure of the most upstream check dam, and a major flood event involving sediment transport, managed by the ensemble of three dams. These two events have proved the efficiency and usefulness of the check dams before end of construction. Acknowledgements The Authors would like to acknowledge R. Gex from INGEX Sàrl in Bex, Switzerland, as well as M. E. Morard from BEB SA in Aigle, Switzerland, for their valuable participation in the study and the construction follow-up of the check dams.

212

C. Dreger et al.

References 1. AquaVision Engineering Ltd. (2007) Debris flow mitigation measures in the Tinière torrent. Internal Report, Ecublens, Switzerland (in French) 2. BWG (2001). Flood Control at Rivers and Streams. Guidelines. Federal office for the environment. http://www.bafu.admin.ch/publikationen/index.html 3. Grot M (2000) Historique des crues de la region de Villeneuve. In: Tinière P, Eau-Froide D, Hegg, Ch. Von der Mühll, D. (eds.) Beiträge zur Geomorphologie. Eidg. Forschungsanstalt WSL, Birmensdorf, pp. 45–54 4. O’Brien JS (2006) FLO-2D – User Manual. Version 2006.01. P.O. Box 66. Nutrioso, AZ 85932. United States of America 5. Tognacca C (1999) Beitrag zur Untersuchung der Entstehungsmechanismen voy Murgängen. Laboratory of hydrauilcs, Hydrology and Glaciology. ETH Zürich. Mitteilung 164 (in german) 6. Ravot E et al. (2006) Couplage entre les processus de pente et le système hydrographique de la Tinière, Villeneuve. UNIL- Faculté de Géosciences et de l’Environnement. CH-1015 Lausanne 7. Takahashi T (1991) Debris flow. IAHR Monograph Series. Rottendam A.A. Balkema 8. Sodelo (2007) Analyse hydrologique de la Tinière. Estimation des crues de la Tinière

Chapter 17

Development of the Similar Typhoon Search System Based on the Deep Neural Network Using Deep Learning Kohji Tanaka, Eisaku Yura, Tatsuya Yoshida, and Shigeho Maeda

Abstract We present a method for improving the search accuracy of the similar typhoon research system by leveraging deep neural networks, which are an extension of artificial neural networks. To the search engine, we apply three parameters of typhoons: course, temporal central pressure, and speed. We show that these parameters can improve accuracy when searching for past typhoons having characteristics similar to the target. Furthermore, the accuracy of functions designed to support the expected disaster prevention actions and flood fighting services was assessed based on the results from the search system. Keywords Deep learning · Deep neural network · Disaster prevention support system · Similar typhoon researching system

17.1 Introduction In recent years, the Japan Meteorological Agency and the Ministry of Land, Infrastructure, Transport, and Tourism (MLIT) have made positive efforts to disclose information to support disaster prevention and mitigation measures. The basin of the Kumano River in the Kii Peninsula, the subject of this study, has undergone K. Tanaka (B) Department of Civil Engineering and Urban Design, Faculty of Engineering, Osaka Institute of Technology, 5-16-1, Ohmiya, Asahi-ku, Osaka 535-8585, Japan e-mail: [email protected] E. Yura Division of Water Resources and Disaster Manegement, C.T.I. Engineering, Co., Ltd., 1-6-7, Dosho-machi, Chuo-ku, Osaka 541-0045, Japan e-mail: [email protected] T. Yoshida · S. Maeda Kinan River and Road Administer, Kinki Regional Bureau, Ministry of Land, Instracture, Transport and Tourism, 142, Nakamaro, Tanabe, Wakayama 646-0003, Japan e-mail: [email protected] S. Maeda e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_17

213

214

K. Tanaka et al.

much flood damage caused by typhoons, particularly the torrential rains brought by Typhoon No. 12 in 2011. Therefore, MLIT and related local governments have promoted cooperation with one another, in terms of information sharing, to ensure preliminary actions [1, 2] for disaster prevention are taken, according to given timelines. The Kii Mountains, spreading into the upriver Kumano River basin, are one of Japan’s heaviest rainfall regions. Orographic rainfall generated by wet southeast air currents from typhoons blowing against the Kii Mountains causes heavy downpours. When a typhoon passes across the west side of the Kumano River, heavy rains can be expected in the Kumano River area because of the southeast air currents that tend to flow into the peninsula. Thus, the Kinan Office of River and National Highways, which manages rivers, pays close attention to typhoon tracks, taking actions based on river stage forecasting information from the Kumano River Flood Forecast System [3], as well as information on similar typhoons and information to assist flood-fighting activities, which are provided by the Kumano River Flood Risk Management Information System [4] (i.e., “current system”) developed in 2014. A disaster action plan (i.e., timeline) based on the foreseeable path of a disaster is implemented five days before the arrival of a typhoon. To defend against the approaching typhoon, related local governments, the prefecture, the meteorological observatory, and the Kinan Office of River and National Highway collaborate and share information on the current conditions and confirm the details and frameworks of measures to be taken, during a teleconference, based on the timeline1). The Kinan Office of River and National Highway forecasts future river stages during this meeting. The current system [4], however, addresses only the information of typhoons that cause severe damage to the Kumano River. Therefore, they face difficulties in providing information appropriate for the size of an approaching typhoon when advising local governments on disaster prevention.

17.2 Operational Problems of the Similar Typhoon Search System Figure 17.1 shows the structure of the Kumano River Flood Risk Management System. In this system, past typhoons having similar characteristics as the current typhoon are analyzed, using the information of past typhoon tracks. Then, information designed to support flood-fighting frameworks is provided, based on the hydrographs of the searched typhoons. A hearing with staff members of the Office about the current system was recently conducted, revealing the following two problems. (i) The similar typhoon search function using typhoon tracks retrieves only those having caused scale flooding, which is impractical (see Fig. 17.2). (ii) Because the system does not allow registration of hydrographs of typhoons with small- and medium-scale flooding, it cannot be used to accurately forecast river stages when a typhoon is approaching.

17 Development of the Similar Typhoon Search System...

215

Fig. 17.1 System structure of Kumano River Flood Risk Management Information System

Fig. 17.2 Search results for similar typhoons using current system

Problem (i) results from the fact that only past typhoons with large-scale flooding, having caused damage to the Kumano River, have been incorporated into the similar typhoon search function of the current system. A possible solution for (i) would be a similar typhoon search system that incorporates typhoons with small- and mediumscale flooding. However, as with this system, ANNs are not capable of developing accurate learning models when trained to learn only the tracks of multiple typhoons as search objects [4]. For this reason, learning models have been developed using DNNs with multiple parameters, such as atmospheric pressure, speed, and track.

216

K. Tanaka et al.

Fig. 17.3 Model structure of ANN

Output layer

Input layer

Typhoon number

Typhoon track

Intermediate layer

The database is extended as a solution for (ii). Thus, a database of hyetographs and hydrographs of typhoons, including ones with small- and medium-scale flooding, was used with data of large-scale flooding for development.

17.3 Development of DNN Models Using Deep Learning 17.3.1 Overview of DNN The similar typhoon search function of the current system uses three-layer hierarchical neural network models [5] with one hidden layer, as shown in Fig. 17.3. Deep learning typically requires training a neural network with two or more hidden layers. Deep learning also requires extracting elements that are common, to some extent, from massive amounts of input data to reduce the number of focused dimensions and to give more weight (i.e., priority) to the process that determines those common elements. The DNNs used in the study [6] are hierarchical neural network models with two hidden layers, as shown in Fig. 17.4.

17.3.2 Learning Method of the Neural Network A neural network [6] is a mathematical model that imitates the mechanism of neurons, as shown in Fig. 17.5. It responds to input data and provides output matching data. N

s = Σ wn x n n=0

(17.1)

17 Development of the Similar Typhoon Search System... Input layer

Fig. 17.4 Model structure of DNN

217 Output layer

Typhoon number

Typhoon track + central atmospheric pressure + speed

Intermediate layer

Fig. 17.5 Diagram of elements making up a neural network

y = sigmoid(s)

(17.2)

Input to units are indicated as x0 , x1 , x2 · · · , xN ; y indicates output; w0 , w1 , w2 , · · · , wN indicate synaptic weights; and w0 indicates a threshold value, another synaptic weight. The error back propagation method was used for the learning method of the neural network in this study. The error back propagation method is a typical learning method for feed-forward neural networks, which makes minor rectifications to synaptic weights each time training data is given. As Fig. 17.6 shows, when a neural network with network output, bi, and target output, ti(i = 1, 2, · · · , M), are given, it rectifies the synaptic weights within the network, based on the principle of the gradient method, so that the evaluation scale, E, shown as Formula (17.3), is minimized. E=

M    bi − t 2  i i=1

(17.3)

218

K. Tanaka et al.

Fig. 17.6 Three-layer feed-forward neural network

17.3.3 Development of a Learning Model Using DNN The mesh information at 2 × 2°, shown in Fig. 17.7, was prepared to obtain typhoon track information (i.e., track, atmospheric pressure, and speed) for input. Typhoon tracks were arranged as 0,1 information, indicating the time-based locations of a typhoon. To make learning more efficient, central atmospheric pressures were standardized so that 1.0 to 0.0 covers 900 ha to 1,000 ha. The speed of typhoons was arranged as standardized information of 0.0 to 1.0 by dividing traveling speeds of typhoons, calculated based on their time-series locations, by the 95% upper limit value (mean value + 2σ) of the speeds of all typhoons fed. Typhoon information was obtained online [7]. Track information of 181 typhoons, from 1951 to 2016 (66 years), mainly covering those that passed within a 300-km radius of the Kumano River, was learned. DNN was consisted in Fig. 17.8. Typhoon courses, central pressure and typhoon speed were inputted at the 450 elements in input layer. Two Intermediate layers were consisted of 200 elements and 50 elements, which have a function to cluster the typhoon data. We made this DNN learned 181 typhoon datum shown in Fig. 17.8.

17 Development of the Similar Typhoon Search System...

219

Fig. 17.7 Mesh information developed from typhoon track information

17.4 Extension of Database The database of the similar typhoon search system was extended. Data, including typhoons with small- and medium-scale flooding, as well as those with large-scale flooding, were developed. The dates that the typhoons formed and dissipated, timeseries locations, central atmospheric pressures, and speeds were registered to the typhoon track information database. Moreover, for typhoons with rainfall amounts analyzed by Radar-AMeDAS and/or data from rainfall observatories, the flow rate of each was calculated using a distributed flow model based on a 250-m mesh covering the Kumano River basin. Then, river stage hydrographs were calculated using a river channel single-dimension unsteady flow calculation model for direct managed downstream sections. Additionally, because a special emergency project for the control of severe river disasters was implemented in the Kumano River in the wake of Typhoon No. 12 in 2011, the latest cross-sections of the river channel were considered for calculation. Therefore, the rainfall and river stage hydrographs of 140 typhoons, out of the 181 whose tracks were registered, were developed and registered to the database.

220

K. Tanaka et al.

Input layer

Typhoon courses

Output layer Intermediate layer

Central pressure Typhoon speed

200 elements. 50 elements.

181 elements. 450 elements. Fig. 17.8 Constitution of learned artificial neural network

17.5 Improvement of Search Performance Using Dnn Models 17.5.1 Verification of Similar Typhoon Search Performance The similar typhoon search function was verified using a DNN learning model. The input values of the 181 typhoons used for learning were applied to a developed model to calculate search values (i.e., output value). The output search values ranged from 0.0 to 1.0. The total of all search values was 1.0. The largest search value came atop search results. The two learning models, the ANN used for the current system and the DNN used in the study, were compared. The input conditions for each model were as follows: (1) track; (2) track and atmospheric pressure; and (3) track, atmospheric pressure, and

17 Development of the Similar Typhoon Search System...

221

Table 17.1 Results of learning model verification No.

Case name

➀

ANN

Hit ratio (%)*a

Search value*b

track

16

➁

track, atmospheric pressure

12

0.11

➂

track, atmospheric pressure, speed

7

0.05

100

0.79

➃

DNN

track

0.15

➄

track, atmospheric pressure

100

0.89

➅

track, atmospheric pressure, speed

100

0.93

*a Hit ratio is number of times that an input typhoon was correctly retrieved/ number of input typhoons (181) *b The mean value of the top searches test typhoons

speed. In the evaluation of search performance, hit ratios were defined as indicators showing whether a model retrieved the same typhoon as an input typhoon. Table 17.1 shows the verification results of the learning models. The search results revealed the following. The hit ratios of all three ANN cases were low. Additionally, the hit ratio tended to be lower when more parameters were fed. This shows that the ability of a model with one hidden layer is limited when it comes to learning with multiple parameters. Thus, this type of model cannot satisfy the needs of the system. The hit ratios of all three DNN cases were 100%. Also, when atmospheric pressures and speeds were fed as parameters, the search value of the similar typhoon corresponding to an input typhoon increased. Thus, we estimate that multiple hidden layers enabled greater recognition in learning of multiple characteristics of a typhoon, improving search performance (see Fig. 17.9). From these results, we verified that the model using the DNN with three parameters (i.e., track, atmospheric pressure and speed) has advantages as a learning model for similar typhoon search.

17.5.2 Verification of Search Performance for Similar Typhoons Test typhoons were included in the learning data for the verification described in the preceding section. Hence, we checked the type of similar typhoon that was retrieved when the input values of test typhoons were excluded from input values for learning. The results are shown in Fig. 17.10, which also shows that the model retrieved typhoons with similar tracks, even when test typhoons were excluded. Additionally, Table 17.2 compares the central atmospheric pressures and cumulative rainfall amounts of the main typhoons in 2015 and 2016 to those of the similar typhoons retrieved. The table shows that the model retrieves typhoons with a similar tendency. For example, the track of Typhoon No. 18 in 2016 missed the Kumano

222

K. Tanaka et al.

1.0 0.9

the first of the search value

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

199020 199021 199028 199114 199117 199203 199209 199210 199304 199313 199314 199320 199426 199606 199612 199707 199708 199709 199719 199805 199807 199808 199810 199916 200111 200115 200206 200207 200213 200221 200304 200310 200404 200406 200410 200411 200416 200421 200422 200423 200507 200511 200607 200704 200709 200813 200909 200918

0.0

Year / Typhoon No.

ANN course

ANN course, pressure

ANN course, pressure, speed

DNN course

DNN course, pressure

DNN course, pressure, speed

Fig. 17.9 Comparison between search values of the top searched typhoons of each learning

Test typhoon (201511) Similar typhoon

Test typhoon (201618) Similar typhoon

(197010)

(201509)

Fig. 17.10 Results of similar typhoon search without input values of test typhoons

17 Development of the Similar Typhoon Search System...

223

Table 17.2 Comparison of specifications of similar typhoons searched by DNN model Test typhoon

Similar typhoon

Year/No.

Central pressure*a

cumulative precipitation*b

Year/No.

Central pressure*a

cumulative precipitation*b

2015 11

965 ha

506

1970 10

980

–

2015 15

945 ha

170

1993 13

955

130

2016 10

965 ha

170

1966 04

965

–

2016 12

1,004 ha

70

1962 13

980

–

2016 16

980 ha

100

1998 10

965

140

2016 18

965 ha

20

2015 09

975

12

*a: Central atmospheric pressure when the typhoon made a landfall or approached most closely to the Kumano River *b : Cumulative rainfall amount in the upstream of the Kumano River Aiga point. No rainfall database for typhoons with “-”

River, causing a small amount of rainfall in the basin. A similar typhoon to this (i.e., Typhoon No. 9, 2015) produced similar rainfall. The performance of similar typhoon searches incorporated by the DNN was evaluated based on the above results. It identified even small typhoons accurately, enabling broader searches than the current system, while retrieving more similar typhoons. Thus, we conclude that the DNN model is practical.

17.6 Development and Operation of Similar Typhoon Search System 17.6.1 Development of a System that Considers Preliminary Disaster Prevention Actions, According to the Timeline To forecast the river stages caused by an approaching typhoon, times are synchronized using the hydrographs of the similar typhoon. In the current system, time synchronization is attempted, based on judgment from the current location of an approaching typhoon and the location of the similar typhoon. However, this method does not clarify the relation between the locations of a typhoon and the time points of the hydrograph. The possibility of misjudging the peak time of a river stage was recognized through its actual operation. Thus, as shown in Fig. 17.11, the function was improved so that it automatically synchronizes the time of an approaching typhoon with that of a similar typhoon, based on the closest forecasted approach to the Japanese islands, among all forecasted typhoon tracks. Additionally, a function of manual time synchronization was installed for minor adjustments of synchronized times.

224

K. Tanaka et al. Narukawa Precipitation(mm/hr) Time synchronization time

Precipitation(mm/hr)

Takaoka Current time

Water surface level

Current location of typhoon

Current time

Water surface level

Time synchronization point (closest approach point to Japan Islands)

Fig. 17.11 Time synchronization between an approaching typhoon and a similar typhoon with hydrograph after synchronization

17.6.2 Development of Function to Support Flood-Fighting Frameworks The manager of rivers at the Kinan Office of River and National Highway is required to provide relevant local governments with a forecast of future river stages and an outlook on flood protection warnings and forecasts during collaborative timeline meetings held at the onset of a typhoon. We developed a function to output information required for the meetings (e.g. similar typhoon, hydrograph after time synchronization, estimated time and date of flood protection warnings and flood forecast) on a sheet of paper, as shown in Fig. 17.12. Additionally, a screen providing support information for flood-fighting frameworks (Fig. 17.13) was developed as a function to support disaster prevention actions of the Kinan Office of River and National Highway at the onset of a typhoon. This function provides a list of disaster prevention actions to be taken by each team/group according to each timeline level for the Office, as judged by the Meteorological Agency’s typhoon track forecast and/or retrieved similar typhoons. This function also allows the Office to enter the details of actual actions and save them as a record, enabling information sharing with involved parties. Although we improved this function to facilitate data entry, work still needs to be done via actual operation for better utilization.

17 Development of the Similar Typhoon Search System...

Forecasted dates and times of flood protection warning and flood forecast

Hydrograph after time synchronization Similar typhoon track

Fig. 17.12 Output screen of information to be provided

Timeline level of office

Disaster prevention actions to be taken by each team/group at each timeline level Brank Part to input chronology

Fig. 17.13 Flood fighting frameworks support information screen

225

226

K. Tanaka et al.

17.6.3 Utilization of the System During Timeline Cooperated Meetings To verify the similar typhoon search function, typhoons like Typhoon No. 11, in July 2015, were searched with the system. Figure 17.14 shows the overview of the typhoon and shows its hydrograph. Additionally, the retrieved information of the similar typhoon, one to three days before landfall, is shown in Figs. 17.15 and 17.16, reflecting typhoon tracks and hydrographs after time synchronization.

Fig. 17.14 Similar typhoon to typhoon No. 11, July 2015 (3 days before landfall) and its rainfall amount and water stage (after synchronization)

Fig. 17.15 Similar typhoon to typhoon No. 11, July 2015 (2 days before landfall) and its rainfall amount and water stage (after synchronization)

Fig. 17.16 Similar typhoon to typhoon No. 11, July 2015 (1 day before landfall) and its rainfall amount and water stage (after synchronization)

17 Development of the Similar Typhoon Search System...

227

According to those data, three days before landfall (i.e., July 14), the river stage at the Takaoka point was forecasted to reach the evacuation stage by July 17, around 1 p.m. Then, two days before landfall, it was forecasted to reach the stage on July 17, around 7 a.m., which is about 12 h behind the actual time of occurrence (i.e., July 16, around 8 p.m.). However, given that the system could forecast a situation where evacuation would be required two days ahead of time, it can be well-used to forecast future river stages with an aim to support flood-fighting frameworks. Disaster prevention actions can be taken in a smooth manner if the central government provides local governments with river stage forecasts during the cooperated timeline meetings. When the forecasted track, central atmospheric pressure or speed of a typhoon changes, there is a possibility that the similar typhoon search function will retrieve different typhoons. However, the gap of a forecasted track typically becomes smaller as the typhoon approaches the Japanese islands. Thus, the same typhoon is more likely to be retrieved, even with the progress of time. Additionally, the similar typhoon search function is assumed to be valid until about one or two days prior to landfall. Henceforth, the river stage forecast information is assumed to be obtained from the Kumano River Flood Forecast System3), which estimates river stages within 6 h of the current time.

17.7 Conclusion In this study, we attempted to improve the accuracy of the similar typhoon search function that searches for a typhoon similar to an approaching one using DNN models trained by deep learning. Then, we developed a system that forecasts the river stages of the Kumano River from a few days before typhoon landfall, based on the hydrograph of the selected similar typhoon. This helps decision-making officials take preliminary disaster prevention actions according to timelines. The DNNs used for the similar typhoon track search were extended from existing ANNs. Initially, a Convolutional Neural Network4) (CNN), which is effective for image analysis, was used as a deep learning method for the similar typhoon search. However, a learning model with sufficient accuracy was not developed. Therefore, whereas CNNs are good at recognizing spatial information patterns, they may be weak at recognizing and optimizing the patterns of limited or temporal information, such as typhoon tracks. Our future plan is to improve the system so that it addresses frontal heavy rain as well as typhoons. Aiming at more sophisticated flood forecasting, we will incorporate parameters, such as meteorological charts and pressure patterns, into a search function to enable rainfall search by factor or incorporate rainfalls and river stages of the past into deep learning as parameters.

228

K. Tanaka et al.

References 1. Katou S (2015) Effort of the Kumano River Timeline. Disaster Prevention and Maintenance Section No. 9, Collected Papers of Kinki Regional Development Bureau Research Conference 2. Ohashi K (2016) Disaster Prevention Action Plan Using Similar Typhoon Search System. Disaster Prevention and Maintenance Section No. 11, Collected Papers of Kinki Regional Development Bureau Research Conference 3. Sasaki M, Tsujikura H, Tanaka K, Shirahase T, Fujii M (2014) Development of River Stage Forecast Method for Channel Sections Subject to Degradation of Estuary Sandbar. Collection of River Engineering Papers, vol. 20, 223–228 4. Tanaka K, Yura E, Sasaki M, Shirahase T, Shimokawa A, Kato S (2015) Development of the Similar Typhoon Search System Supporting the Disaster Prevention and Mitigation Action Plan Based on the Results of Typhoon Course. Collection of River Engineering Papers, vol. 21, pp. 443–448 5. Kumazawa I (1998) Learning and Neural Networks. Electronic Information Communication Engineering series, Korikita Publishing Co., Ltd 6. Okatani T (2015) Deep Learning. Machine Learning Professional Series, Kodansha Ltd 7. Kitamoto A National Institute of Informatics, Digital Typhoon. http://agora.ex.nii.ac.jp/digitaltyphoon/

Chapter 18

An Innovative DEM Improvement Technique for Highly Dense Urban Cities Dongeon Kim, Shie-Yui Liong, Philippe Gourbesville, and Jiandong Liu

Abstract This paper presents an innovative approach to derive an improved Digital Elevation Model (DEM) using multispectral imagery and Artificial Neural Network (ANN). The DEM is crucial in land and water management which reflects the actual topographic characteristic on earth surface. However, a high accuracy DEM is very difficult to acquire because it is often very costly and is treated as confidential. DEM from Shuttle Radar Topography Mission (SRTM) has been improved using multispectral imagery of Sentinel 2 and the ANN with its strength of pattern recognition in big data processing. SRTM is widely used in the area where the high accuracy DEM is not available as it is easily accessible to the public with no cost. However, its accuracy is limited due to its coarse resolution (≈30 m) and sensor limitations. Sentinel 2 provides the 13 spectral band spans from the visible and the near infrared to the short wave infrared at different resolutions ranging from 10 to 60 m. Sentinel 2 produces different reflectance values in different land-uses. These two remote sensing data are used in ANN as input data. The ANN is trained with reference DEM which has a high accuracy level and different weights are calculated to reduce the error between the elevation of SRTM and reference DEM. The trained ANN is applied to a different place to evaluate the performance. The improved SRTM presents clearer images with higher resolution than the original SRTM with 6 to 26% lower Root Mean Square Error (RMSE). The paper should be of interest to readers in the areas of remote sensing, artificial intelligence and D. Kim (B) · S.-Y. Liong · J. Liu Tropical Marine Science Institute, National University of Singapore, 18 Kent Ridge Road, Singapore, Singapore e-mail: [email protected] S.-Y. Liong e-mail: [email protected] J. Liu e-mail: [email protected] D. Kim · P. Gourbesville Polytech Nice, University of Nice Sophia Antipolis, 930 Route des Colles, Biot, France e-mail: [email protected] S.-Y. Liong · J. Liu Willis Towers Watson, 51 Lime Street, London, UK © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_18

229

230

D. Kim et al.

land/water management, especially for the policymakers who require land surface simulation with higher accuracy of topography. Keywords Artificial Neural Network · Remote sensing · Improved SRTM

18.1 Introduction DEM is crucial in land and water management planning which reflects the actual topographic characteristics of the surface. The Shuttle Radar Topography Mission (SRTM) is publicly accessible provided at no cost, however, its accuracy is limited with Root Mean Square Error (RMSE) of approx. 8 m in Singapore’s dense urban areas [1, 2]. Therefore, in the areas where high accuracy DEM is not available, additional studies are required to obtain reliable topography. In developing countries, unavailability of high-resolution topography data is the prime limitation for simulating hydrodynamic models [3]. [2] has developed an improved DEM scheme using remote sensing data and ANN technique to correct the SRTM DEM in the forested area of Singapore. They trained ANN using 11 bands of Landsat-7 data to eliminate the error caused by canopy level and reduced RMSE between reference DEM and SRTM up to 50%. This scheme is further developed in this research to apply in dense urban cities. The multispectral imagery of Landsat-7 is replaced by Sentinel-2, and 1-arcsecond resolution of SRTM used instead of 3-arcsecond resolution. In ANN model setup, it requires 3 types of data mentioned above; multispectral imagery, the DEM to be improved (SRTM in this study), and a reference DEM (high accuracy elevation). These data are input for the ANN for training, and later for validation. Once the performance of the trained ANN is acceptable, it can be applied to areas where their SRTM DEMs are to be improved. Figure 18.1 illustrates the schematic diagram of DEM improvement methodology. Multispectral imagery is produced by the sensors which measure the reflected energy within several specific bands/sections of the electromagnetic spectrum. It can be defined as “acquisition of images in hundreds of contiguous, registered, spectral bands such that for each pixel a radiance spectrum can be derived” [4]. The multispectral sensors have 3 to 10 different measurements of the band in each pixel of the images. Various earth observation satellites are being used to capture the images of the earth. Such satellites are called imaging satellites which are normally operated by the commercial companies and governments around the world [5]. Over the years many countries have launched different satellites to acquire the images of the Earth. Sentinel 2 is an earth observation mission which was developed by the European Space Agency (ESA) as a part of Copernicus Programme to perform terrestrial observations in support of services such as forest monitoring, land cover changes detection, and natural disaster management [6]. The multispectral imagery can be used for land use classification, for seasonal monitoring, agricultural and environmental applications ([7–9]). Using different reflectance values from different land use types, the area can be classified by clustering and machine learning methods. [3] analysed the different

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities

231

Fig. 18.1 Schematic Diagram of DEM Improvement methodology

reflectance of Sentinel 2 with different land uses. The reflectance of SWIR bands (Bands 6-8) in forest areas is higher than that in urban areas; on the other hand, the reflectance of NIR bands (Bands 2-5) in urban area is higher than that in forest area. These different characteristics at each band help to classify land use in ANN as input nodes. These characteristics have been fully utilized for this study to generate the improved SRTM using both multispectral imagery and ANN. A DEM can be used to depict the terrain of the earth and is an organised array of the numbers which represent the elevations of spatial distributions above an arbitrary datum. The primary principle of a DEM is to describe the elevations of various points in a given area in digital format. The term DEM is usually applied to land surface topography, but it is a general term which is used to depict the spatial patterns of various surfaces e.g. surface water, ground surface, canopy, etc. Digital Surface Model (DSM) and Digital Terrain Model (DTM) are the two other terms which are frequently used for the ground terrain. DTM is referred as to the Earth terrain i.e. bare ground while DSM includes objects on ground like the buildings and trees, as shown in Fig. 18.2. DEM is a crucial input for numerical water modelling. [11] assessed the DTM for hydrogeomorphological modelling in small Mediterranean catchments, including SRTM, ASTER and LiDAR datasets. The RMSE results of the vertical accuracy show that SRTM and ASTER have differences of 6.98 and 16.10 m respectively over the study areas due to systematic distortions and coarse horizontal resolution. The authors concluded that these limitations should be carefully considered when applying the data for numerical modelling and this research utilised SRTM DEM as it has less errors than ASTER DEM. [12] developed a runoff modelling using very high resolution DSM. Two types of DSM data (SRTM and ASTER DEM) were used as topography data in the model and compared to LiDAR data only, and the combination of photogrammetric and

232

D. Kim et al.

Fig. 18.2 Digital surface model and digital terrain model [10]

LiDAR. Both data were able to capture the main buildings; but small buildings were not captured by LiDAR data. This resulted in significant differences in the flood map outputs. The authors recommended that fine-tuning topographic data is necessary for high resolution flood modelling. Artificial Intelligence (AI) is the recreation of human intelligence processes by machines, especially computer systems. These processes include learning (the acquisition of information and rules for using the information), reasoning (using the rules to reach approximate or definite conclusions), and self-correction [13]. ANN is one of the machine learning systems to achieve AI. ANNs apply mathematical learning algorithms which are simulated by properties of the biological neural networks. ANNs are loosely based on biological neural networks in such a way that they are implemented as a system of interconnected processing elements, sometimes called nodes, which are functionally analogous to biological neurons. The connections between distinct nodes have numerical values, called weights, and systematic altering of these values will give the ability to approximate the desired function [14]. The ANN is formed in three layers: input layer, hidden layer and output layer. The input layer has input neurons that transfer information via synapses to the hidden layer, and similarly the hidden layer transfers this information to the output layer via additional synapses. The synapses store values referred to as weights that help them to control the input and output to different layers. Figure 18.3 shows the schematic diagram of ANN.

18.2 Methodology 18.2.1 Data Pre-processing Since all of the remote sensing data have different resolutions (i.e. SRTM 30 m; Sentinel 2 10–60 m; surveyed DEM 1 m), all input layers need to be standardized to a common resolution through resampling method as shown in Fig. 18.4. In this study, 10 m resolution is chosen for the assessment of the developed methodology.

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities

233

Fig. 18.3 The schematic diagram of ANN layers [15]

Fig. 18.4 Standardization of different resolutions from different sources

All remote sensing data were processed using ArcGIS desktop software developed by Environmental System Research Institute (ESRI). ArcGIS is a Geographic Information System (GIS) for working with maps and geographic information. It is used for compiling geographic data, analysing mapped information, and managing geographic information in a database [16]. All raster layers from its original data set were standardised into common resolution and matched cell alignment (extent and

234

D. Kim et al.

origin of each cell). The raster format of data was extracted by point type of shapefile and the attribute table of the shapefile can generate the table which is numeric inputs to the ANN.

18.2.2 Artificial Neural Network Setup Matlab Neural Network Toolbox was used for developing DEM improvement scheme in this study. It provides a neural network to generalize nonlinear relationships between inputs and outputs using feedforward networks. The feedforward neural network is the first and simplest type of ANN devised [17]. It contains multiple neurons (nodes) arranged in layers and all these have the connections. The ANN algorithm is divided into 3 steps: (1) Training/Learning: The network processes the input and compares its resulting outputs against the target layer. The errors are then propagated back through the system to adjust the weights. (2) Validation: This is used to measure the performance of the network generalization, and to halt training when generalization stops improving. (3) Testing: This has no effect on training and provides an independent measure of network performance after the training. The network is trained with Levenberg-Marquardt (LM) backpropagation algorithm [18–20]. This method is a standard technique for solving nonlinear least squares problems to fit a curves by minimizing the sum of the square of the errors between input and output nodes. The training is continued until the validation error ceased to decrease; the trained ANN is then applied to test data set. In this study the data set was randomly divided into 70% for training, 15% to validate the network to stop training before the overfitting and 15% for independent testing of network generalization. Table 18.1 shows the example of input, target and output layers in ANN. Table 18.1 Input, Target and Output layers in ANN (example) Input Layer

Target Layer

Output Layer

B02

B03

B04

B05

B06

B07

B08

B8A

SRTM (m)

Reference (m)

Improved SRTM (m)

0.0898

0.0884

0.0604

0.0922

0.2329

0.3059

0.2901

0.3296

26

23.95

0.089

0.0865

0.0608

0.0928

0.2408

0.3137

0.2787

0.3416

27

24.32

To be calculated

0.0835

0.0749

0.0454

0.0866

0.2321

0.2976

0.2733

0.325

27

30.24

0.0933

0.0927

0.0715

0.0892

0.2442

0.3112

0.2762

0.3503

27

25.55

0.0879

0.0797

0.0517

0.0676

0.2178

0.272

0.2666

0.308

26

25.08

0.0856

0.0761

0.0496

0.0834

0.2172

0.2736

0.2367

0.3096

25

29.17

0.0944

0.0901

0.0695

0.0976

0.2332

0.2924

0.2684

0.3247

26

23.39

0.0884

0.0856

0.0515

0.0915

0.2488

0.3174

0.2898

0.3507

28

22.93

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities

235

18.3 Results and Discussion This section proofs the concept of improved DEM development scheme mentioned in the earlier section. DEMs in Nice (France) and Singapore are taken into consideration. The improved DEMs will be compared with the high resolution (1 m) and high accuracy (40 cm) DEM in Nice. The data is obtained from Nice Côte d’Azur Metropolis. TanDEM-X in Singapore is used to evaluate their performances. TanDEM-X is from German Aerospace Center (DLR) and it is a global DEM which has a spatial resolution of 0.4 arc-second (≈12 m) with 2–4 m in relative vertical accuracy [21]. The ANN model is trained and validated in dense urban areas in Nice (France). The training area has an area of 12.0 km2 while the test area 5.2 km2 . Figure 18.5 shows the satellite image of the training (box with blue comb pattern) and test (box with red comb pattern) areas. The areas are mainly urbanized with buildings, mild slopes with elevation ranging from 0 to 200 m. The ANN is trained in the training area with 1 m reference DEM data used in the target layer. The weights are calculated from the training to reduce the errors between reference DEM and original SRTM. Once the validation is satisfied in the training process, the trained ANN is then applied to the test area to improve SRTM (iSRTM). The performances are evaluated using the reference DEM. Figure 18.6 shows the comparison of elevation maps of various DEMs. Figure 18.6 (a) is a satellite image of test area depicting the land shapes; Fig. 18.6 (b) is the area from 1 m reference DEM; Fig. 18.6 (c) is the area from the original SRTM with 30 m resolution; Fig. 18.6 (d)

Fig. 18.5 Training and Test areas in Nice: Dense urban areas

236

D. Kim et al.

Fig. 18.6 Performance of improved SRTM in dense urban areas in Nice, France

is the area resulting from iSRTM DEM with 10 m resolution. The reference DEM shows the most clear land shapes (i.e. buildings and roads); iSRTM DEM shows clearer land shapes in the developed areas than the original SRTM and it matches the reference DEM categorically much more. The significant improvements are reflected in the scatter plots and the RMSE (Fig. 18.7) as well. The RMSE of iSRTM DEM reduces from 8.36 to 7.82 m (a 6.5% reduction). The trained ANN in Nice, France is validated in Singapore. The interest is to investigate the quality of DEM generated by an ANN, trained in an area with a variety of patterns (e.g. flat and open surfaces, densely urban areas with low and high rise buildings) when it is applied in a faraway area within the aforementioned patterns learned. This is important to know whether one has no other choice but to totally rely on SRTM DEM. Figure 18.8 shows the comparison of elevation maps of various DEMs. In Singapore, TanDEM-X is used as a reference DEM. The improvements are reflected in the scatter plots and RMSE (Fig. 18.9) as well. The RMSE of iSRTM trained in Nice is reduced from 4.23–3.12 m (a 26.2% reduction). The SRTM DEM can still be significantly improved with ANN trained in a faraway dense urban area where high-quality ground truth data are available.

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities

237

Fig. 18.7 Scatter Plots and RMSE/Correlation comparisons between SRTM and iSRTM in dense urban area in Nice, France

Fig. 18.8 Performance of improved SRTM in dense urban areas in Singapore

Earlier research conducted by [2] showed the RMSE reduction in the forested area of Singapore was around 50%. However, the performance of the urban area in this research showed 6–26% of RMSE reduction. This can be explained as the urban area is more complicated than the forested area as the urban area contains buildings and narrow roads. Although the resolutions are downscaled from 30 to 10 m as Sentinel-2 have 10 m resolution in 2-, 3-, 4- and 8-band, this could not represent the

238

D. Kim et al.

Fig. 18.9 Scatter Plots and RMSE/Correlation comparisons between SRTM and iSRTM in dense urban area in Singapore

complexity of urban characteristics. However, the generated improved SRTM DEM is still clearer than its original SRTM and the trained ANN can be applied to faraway countries using the pattern learnt from training area. It is useful to consider different ANN structures and more remote sensing data which can contain information of urban to increase the performance.

18.4 Conclusions This study presented a DEM improvement scheme using multispectral imagery and Artificial Neural Network. A publicly available DEM, SRTM (30 m resolution), was under consideration. The very high resolution (1 m) and high accuracy (40 cm) DEM of Nice (France), made available by Nice Côte d’Azur Metropolis (France), was used as the target data in the training of an Artificial Neural Network (ANN) while the input data were from SRTM DEM and Sentinel-2. Validation of the DEM generated by the trained ANN on other areas in Nice and Singapore showed better DEM than the original SRTM DEM, in terms of lower RMSE and visual clarity. The improvement is on dense urban cities, a lower RMSE by about 6–26%. It is of great interest to test the DEM quality of ANN, trained with various land-use patterns and characteristics when it is applied in a faraway country within the range of the same land-use pattern and characteristics. The test in Singapore’s performance showed its applicability with SRTM DEM improvement. It would be interesting to consider other ANN structures to improve performance. Also, utilizing more remote sensing data which contains urban characteristics such as buildings, roads information so the complexity of urban characteristics can be clearly classified in ANN.

18 An Innovative DEM Improvement Technique for Highly Dense Urban Cities

239

These findings provide researchers, engineers, city planners and policymakers in their work to assess more accurate topography, for examples, flood simulations, and analysis. Acknowledgements We are very grateful to Willis Towers Watson (UK), Nice Côte d’Azur Metropolis (France) and German Aero Space Centre (DLR) for providing the data and for making this study possible.

References 1. Kim D-E, Gourbesville P, Liong S-Y (2019) Overcoming data scarcity in flood hazard assessment using remote sensing and artificial neural network. Smart Water 4(1):2 2. Wendi D, Liong S-Y, Sun Y, Doan CD (2016) An innovative approach to improve SRTM DEM using multispectral imagery and artificial neural network. J. Adv. Model. Earth Syst. 8(2):691–702 3. Kim D., Sun Y., Wendi D., Jiang Z., Liong S.-Y., Gourbesville P. (2018) Flood modelling framework for Kuching City, Malaysia: overcoming the lack of data, in Advances in Hydroinformatics. 2018, Springer. p. 559–568 4. Goetz AF, Vane G, Solomon JE, Rock BN (1985) Imaging spectrometry for earth remote sensing. Science 228(4704):1147–1153 5. Entwistle N, Heritage G, Milan D (2018) Recent remote sensing applications for hydro and morphodynamic monitoring and modelling. Earth Surf. Proc. Land 43(10):2283–2291 6. Drusch M, Del Bello U, Carlier S, Colin O, Fernandez V, Gascon F, Hoersch B, Isola C, Laberinti P, Martimort P (2012) Sentinel-2: ESA’s optical high-resolution mission for GMES operational services. Remote Sens. Environ. 120:25–36 7. Andersen OB, Woodworth PL, Flather RA (1995) Intercomparison of recent ocean tide models. J. Geophys. Res. Oceans 100(C12):25261–25282 8. Moody D.I., Brumby S.P., Rowland J.C., Altmann G.L. (2014) Land cover classification in multispectral imagery using clustering of sparse approximations over learned feature dictionaries, vol. 8, no. 19, pp. 1–19. SPIE 9. Ashish D, McClendon RW, Hoogenboom G (2009) Land-use classification of multispectral aerial images using artificial neural networks. Int. J. Remote Sens. 30(8):1989–2004 10. Asharyanto H., Soeksmantono B., Wikantika K. (2015) Three Dimensional City Building Modelling With Lidar Data (Case Study: Ciwaruga, Bandung)„ in INA-Rxiv 11. Graf L, Moreno-de-las-Heras M, Ruiz M, Calsamiglia A, García-Comendador J, Fortesa J, López-Tarazón J, Estrany J (2018) Accuracy assessment of digital terrain model dataset sources for hydrogeomorphological modelling in small mediterranean catchments. Remote Sens. 10(12):2014 12. Abily M., Delestre O., Amossé L., Bertrand N., Richet Y., Duluc C.-M., Gourbesville P., Navaro P. (2015) Uncertainty related to high resolution topographic data use for flood event modeling over urban areas: toward a sensitivity analysis approach. ESAIM: Proceedings and Surveys, 48, 385–399 13. Axelberg P. (2007) On tracing flicker sources and classification of voltage disturbances. Department of Signals and Systems, Chalmers University of Technology 14. Gurney K. (2014) An introduction to neural networks. https://www.inf.ed.ac.uk/teaching/ courses/nlu/assets/reading/Gurney_et_al.pdf. CRC press 15. Haykin S. (1994) Neural Networks: A Comprehensive Foundation. Prentice Hall PTR, 768 16. ESRI, Environmental Systems Research Institute (2018) 17. Schmidhuber J (2015) Deep learning in neural networks: an overview. Neural Netw. 61:85–117

240

D. Kim et al.

18. Levenberg K (1944) A method for the solution of certain non-linear problems in least squares. Q. Appl. Math. 2(2):164–168 19. Marquardt DW (1963) An algorithm for least-squares estimation of nonlinear parameters. J. Soc. Ind. Appl. Math. 11(2):431–441 20. Madsen R.A., Hunt T.E., Higley L.G. (2004) Alfalfa: Simulated Clover Leaf Weevil Injury and Alfalfa Yield and Quality 21. Wessel B., Fritz T., Busche T., Rizzoli P., Krieger G. (2016) TanDEM-X Ground Segment DEM Products Specification Document, DLR

Chapter 19

An Integrated Approach to Water Resources and Investment Planning for Water Utilities Damian Staszek, Dragan Savic, and Guangtao Fu

Abstract Traditionally, water companies in England and Wales use supply-demand modelling methods as decision-making tools for water resources evaluation and investment planning. Companies use separate models for water resources planning and for water resources investment-planning. This approach entails calculating deployable output in a water resources model, for an assumed level of service. That deployable output calculation is exported to an investment model. There is no twoway integration between the water resources model and water resources investment model in the traditional approach, hence any investments do not affect deployable output estimated first in the former model. This paper proposes a new integrated approach—where water resources modelling and investment modelling are integrated, by way of a single modelling tool. The aim of this paper is to compare and contrast traditional water resources and investment planning with the proposed new integrated approach. The new integrated model will allow for the impact of investment choices on the supply and demand position at any point over the planning horizon. The new tool is run for the baseline scenario and for climate change supply scenarios. The results from the integrated model are compared with the results presented by Bristol Water in their Water Resources Management Plan. The level of service metric is calculated in the new model for both baseline and climate change supply scenarios. Finally, this paper proposes further improvement of the integrated model, including potential multi-objective optimisation and new approach to represent uncertainty. Keywords Water resources modelling · Supply-demand problem · Investment planning · Optimisation

D. Staszek (B) · D. Savic · G. Fu College of Engineering, Mathematics and Physical Sciences, University of Exeter, Streatham Campus, North Park Road, Exeter EX4 4QF, UK e-mail: [email protected] D. Savic KWR Water Cycle Research Institute, Nieuwegein, The Netherlands © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_19

241

242

D. Staszek et al.

19.1 Introduction The Water Industry in England and Wales is privately-owned (in England) or operates as not-for-profit company (for Welsh Water) and is regulated by Department for Environment, Food and Rural Affairs (Defra) and Welsh Government set the overall water and sewerage policy framework in England (Defra) and Wales (Welsh Government). The Water Services Regulation Authority (Ofwat) is responsible for economic regulation. The Environment Agency (for England) and the Natural Resources Wales agency (for Wales) are the environmental regulators. The remaining regulators are Drinking Water Inspectorate (water quality regulator), Consumer Council for Water (consumers’ representation) and Natural England (advisor on the natural environment). Water companies in England and Wales are required to submit Water Resources Management Plans (WRMPs) and Business Plans to their regulators every 5 years. WRMPs contribute to Business Plan submissions, as investments selected during the WRMP evaluation are included in the business planning assessments submitted in Business Plans. WRMPs are submitted to the Defra for their review. The Environment Agency and Ofwat are statutory consultees during the WRMP submission process. These two regulators are required to advise Defra on their finding regarding the water resource strategies being proposed by companies. Any investments proposed in WRMPs can be funded through Ofwat’s price review process. Regulators require from water companies to consult their customers on various issues, including: (1) the long-term planning strategies to set out the standard of service that customers can expect to receive on acceptable costs (or willingness to pay for improvement in services), and (2) what their priorities are regarding water resources and long are -term environmental planning. Levels of service are one of the measures describing a water company’s commitments to its customers [1]. Levels of service describe the average frequency of water supplies restriction that companies are planning to apply to their household and non-household customers [2]. There are four water-usage restriction categories reflected in WRMP evaluation: (1) voluntary reductions, encouraged by a publicity campaign, (2) hosepipe restrictions, (3) drought orders restricting non-essential use of water, and (4) drought orders imposing standpipe usage or rota cuts [1]. Separate supply and demand forecasts are produced by the companies, and then decision-making tools are used to decide on future strategies. When forecasting future demand, various scenarios for population growth need to be considered. The future water availability should also represent supply variability due to the predicted impact of climate change [2]. There is a large body of literature concerning the decision-making methods (DMMs) used in water resource planning, and regulators also produce recommendations for the tools and methods which should be used [3, 4]. Traditionally, companies in England and Wales use supply-demand modelling methods as decision-making

19 An Integrated Approach to Water Resources …

243

tools for water resource evaluation and investment-planning. In these methods, supply and demand are treated as single values for each year over the planning horizon—companies use separate models for water resource planning and for water resource investment planning. This approach entails calculating deployable output in a water resources model, for an assumed level of service, under chosen drought scenarios. That deployable output calculation is then used as an input to an investment model. There is no two-way interaction between the water resources model and the investment model in the traditional approach. Hence, any investments do not affect the deployable output estimated in the water resources model. The majority of investment models used by water companies in England and Wales run on yearly data. This paper proposes a new integrated approach for water resources and investment modelling, which will be compared with the traditional approach. The new integrated model allows for the impact of investment choices on supply and demand measures to be accounted for in the resource availability at any point over the planning horizon. The same input data is used as in the traditional approach, i.e., all baseline supply options including water export-import options, all demand predictions for household and non-household customers, outage data and headroom components. As with the traditional approach, we use the headroom component to account for uncertainty affecting supply and demand over the planning horizon. The new tool, which encapsulates the proposed methodology, runs on monthly data as it allows for the representation of varying supply-demand levels over a given year. A case study of Bristol Water is presented in this paper, and results from the new modelling approach are compared and contrasted with those presented by Bristol Water in the WRMP, which the company submitted for regulatory review in 2018.

19.2 Methodology 19.2.1 Introduction Traditional supply-demand modelling, used by water companies during the most recent (and indeed previous) WRMPs, consist of the building blocks shown in Fig. 19.1. Most companies use a similar modelling approach as this means it will be consistent with water resource planning requirements [1, 2]. The consistent approach also enables relatively easy completion of the planning table templates which need to be submitted to their regulators. Deployable Output (DO) can be defined as the output of a commissioned source or group of sources or of bulk supply as constrained by environment, licence, pumping plant and/or well/aquifer properties, water transfer or water quality issues [1]. Companies first calculate the deployable output for their assumed level of service. Various water resources tools are used to estimate the DO (Aquator, MISER) [6]. Frequently, simplified versions of water resources tools are used to make computational

244

D. Staszek et al.

Fig. 19.1 Building blocks of WRMP model optimisation

DO estimations faster and for a high number of scenarios, including DO estimation under various climate change scenarios [7, 8]. Outage allowance represents the risk of temporary or short term losses of supply [2]. Water Available For Use (WAFU) is a final forecast that combines estimations of DO and its future changes, outage and all water transfers. Demand forecasts include household and non-household demand, leakage, and water taken unbilled any future changes to demand. Demand and supply forecasts are compared to identify periods with water deficit. Target headroom is a buffer between supply and demand designed to cater for specified uncertainties [1] and it is included on the demand side of the supply-demand equation. If supply does not meet demand over planning horizon then companies use one of the decision support tools to select optimal investment strategies. The optimization may need to be rerun if the decision maker decides to change the proposed level of service.

19.2.2 Water Resources Model The proposed Water resources model is built in the R environment and runs on monthly data for 25 years planning horizon (2019–2044). Abstraction limits and compensation requirements are included in the model. The model uses future flow data, source licence constraints and demand data, and for each month returns the reservoir level (the Mendip Reservoir Group modelled as a single reservoir). This reservoir level is compared with control curves in order to determine whether the

19 An Integrated Approach to Water Resources …

245

level of service has been breached [5, 6]. However, the integrated model is general in nature, such that any number of reservoirs can be modelled.

19.2.3 Investment Model and Optimisation Tool The investment part of the integrated tool considers the following options in order to mitigate the water deficit at any point over the planning horizon: demandmanagement and efficiency options, supply schemes, water transfers and drought options. Each investment option is evaluated based on its net present value (NPV) calculated using capital, operational and carbon costs. There is an assumption that if an intervention is selected it will be introduced in April of the given year, as the UK financial year runs from April to March. The integrated model selects the investment programme with scheduling of the most cost beneficial options that will guarantee no water deficit at any point over the planning horizon. The objective function minimizes the total costs of the programme. All supply options selected by the optimiser increase the total WAFU and all chosen demand management or efficiency options decrease the distribution input (defined as the sum of demand, leakage and target headroom) figures in the water resources part of the integrated model. The updated forecast for water availability and water usage are employed to calculate the actual level of service post optimisation. The integrated model uses a nsga2R, R Based Non-dominated Sorting Genetic Algorithm, to provide Pareto optimum solutions for the optimisation [7, 8].

19.3 Case Study 19.3.1 Introduction The case study of Bristol Water’s DMMs used for their draft 2019 WRMP [5, 9] is presented and compared with the integrated model in this section. Bristol Water is a water-only company located in South West England. It serves an area of approximately 2,400 square kilometers, with 1.19 million people in Bristol and surrounding areas. The company, after consultation with its customers, has proposed the level of service specified in Table 19.1 [5]. The company models their supply and demand over 25 years, based on annual data. Their deployable output is estimated using MISER [5, 10] and Excel-based Mass Balance water resource models.

246 Table 19.1 Level of Service (2019) proposed by Bristol Water

D. Staszek et al. Drought Action

Years on average

Temporary use ban (TUBs)

1 in 15

Drought Order—Non-essential use ban

1 in 33

Emergency Drought Order—Partial supply or rota-cuts

1 in 100

19.3.2 Water Resources Model For our integrated model, we have replicated the Bristol Water’s Mass Balance Model that is based on monthly data. The water resources part of the integrated model includes following water sources: groundwater, abstraction from the surface water sources (Sharpness Canal, River Axe and Chew Magna) and the Mendip Reservoir Group. Groundwater resources are assumed to be stable over the planning horizon. Water from the Sharpness Canal, which is the biggest source of the raw water for the Bristol Water system, is licence-constrained. It is assumed that permitted abstraction levels will not change over the planning horizon. The same assumptions used in the WRMP submission have been used in the integrated model: (1) Bristol Water’s operational decisions relate to the volume of raw water abstracted from the Mendip Reservoir Group; (2) Sharpness Canal water is available up to the licenced volume, but this source needs more expensive treatment than the raw water from the Mendip Reservoir Group; and (3) only the Mendip Group Reservoirs will be affected by climate change. We used the company’s control curves for the Mendip Reservoir Group to define whether if in any month the rules for the introduction of TUBs are triggered/ the assumed level of service is breached. These control curves are described in the company’s Drought Management Plan 2018 [6].

19.3.3 Supply Scenarios The UK climate projection (UKCP09), developed by the UK Climate Impact Programme (UKCIP), is a major source for climate change data and has been used by all companies in England and Wales for their 2019 WRMPs. UKCP09 data requires hydrological modelling, hence a rainfall-runoff model, such as HYSIM [11] is needed, whose use is outside the scope of this study. UKCP09 has been used by the “Future Flows Climate Programme [12] to produce climate change projections for river flows. These scenarios are transient flow projections, hence rainfall-runoff modelling is not needed. They can be used to generate future flows directly derived from the historical record for all 11 climate change scenarios. We have employed the rolling flow factor method to produce factors for each year for the future planning horizon [13]. As a baseline, historical inflows for the River

19 An Integrated Approach to Water Resources …

247

Axe, Mendip Group and Chew Magna for the years 1961–1990 were used. Monthly averages from the baseline were compared with monthly averages for each 30 year future flow period. For instance, to generate flow factors for 2030, a future flow from “future Flow Climate Programme” for the period from 2015–2044 is compared with the baseline 1961–1990. Then, to generate the flows factor for 2040, the future flow period from 2025–2054 is again compared with the baseline. In the next step, historical flow records were resampled on a 3 monthly seasonal basis. In the final step, the historical resampled river flows are perturbed by flow factors generated during the stage where future flows projections are compared with the baseline. The full details of this approach can be found in [13, 14].

19.3.4 Demand Scenarios We used Bristol Water’s demand forecast that was derived using population figures and a per capita consumption (PCC) forecast as a proxy of per household consumption (PHC), with an assumed household occupancy ratio. PCC figures for each year were derived based on micro-component analysis, where water consumption for each water device is forecasted separately [5]. The following water devices have been considered: WC flushing, shower use, bath use, tap use, dishwasher use, washing machine use, water softener use, external use and miscellaneous use [5]. For each microcomponent, a linear model has been developed with PHC as an explanatory variable. As there is a significant difference for water use between metered and unmetered households, separate micro-component models were derived for unmeasured households, existing metered households, voluntary metered households (optants), new build metered households and selective (or compulsory) metered households [5]. Due to the complexity of demand forecast calculations, there are various sources of uncertainty including around occupancy ratio, population growth figures and assumptions for each micro-component water use. In their previous WRMP submitted in 2014, companies tended to consider just three scenarios for populations growth (Low, Principal, and High) affecting the final demand forecast [10]. For WRMP19, companies were not required to submit varying demand forecasts, but they were expected to test their plans against various resilience measures—hence they are now tending to test against forecasted population increase in terms of assumed growth rate by the final year [2].

19.3.5 Intervention Strategies Bristol Water considers the following intervention strategies: options to reduce water consumption, options to reduce water losses, and options to provide additional water resources. Demand management and efficiency options are preferable as stated by regulators [15]. The full list of intervention options based on the information available

248

D. Staszek et al.

from draft WRMP [5], company’s Statement of Response to Public Consultation [9] and Water Resources Market Information dataset [16] is presented in Table 19.2.

19.3.6 Optimisation Tool In the presented integrated model we employed nsga2R optimisation package [7, 8] and the following algorithm parameters (derived following testing of numerous combinations for optimisation) were used (Table 19.3):

19.3.7 Scenario Testing Companies are expected to test their WRMPs against a range of stresses [2, 15]. For instance, Bristol Water has tested its proposed WRMP-related Business Plan investments/measures against the following stresses: • changes to the Water Framework Directive (WFD) [2] or Sustainability Assessment Impacts on its sources—reducing the amount which is expected to be licensed; • changes to the demand forecast due to changes in per capita consumption; • changes in demand due to new industrial/commercial demand; and • changes to climate change impact assumptions on its source yields. Sustainability changes, after consultation with the Environment Agency or Natural Resources Wales, may need to be required to “achieve an efficient, sustainable and secure supply of water that protects the environment effectively” [2]. Also, the WFD objectives to support the environmental objectives in the river basin management plans may impact the company’s supply forecast [2]. For Bristol Water’s WRMP19, it was not necessary to provide various demand forecasts—instead, testing against demand increases was provided [5, 9].

19.4 Results We have compared and contrasted Bristol Water’s results with the results achieved using the integrated model. For the purpose of this paper, only single objective modelling was proposed to be comparative with the modelling approach and results presented by Bristol Water. The submission by Bristol Water stated that at this stage the more sophisticated multi-objective optimisation is not required [5]. Bristol Water preoptimisation baseline supply-demand position over the planning horizon is shown in Fig. 19.2.

19 An Integrated Approach to Water Resources …

249

Table 19.2 Bristol Water WRMP options [9, 10] Option Name

Earliest option start (Year)

Years to fully implement (Years)

Yield (Ml/d)

Capital Costs (£M/year)

Operating Costs (£M/year)

Promotion of Water Efficiency to customersEnhanced water efficiency communications campaign

2020

5

0.08

0.00

1.00

Promotion of Water Efficiency to customersEducation programme on water efficiency on different key stages

2020

5

0.08

0.00

0.10

Promotion of Water Efficiency to customers. Household water efficiency devices installation programme

2020

5

0.27

0.62

0.00

Selective metering of domestic customers based on high consumption e.g. sprinkler use and/or zones of high demand

2020

5

0.57

0.56

0.00

Enhanced promotion of free water meters to unmeasured households

2020

5

0.57

0.56

0.00

(continued)

250

D. Staszek et al.

Table 19.2 (continued) Option Name

Earliest option start (Year)

Years to fully implement (Years)

Yield (Ml/d)

Capital Costs (£M/year)

Operating Costs (£M/year)

Installation of rainwater harvesting in new build households and non-households

2020

5

0.03

0.45

0.00

Increase performance of existing sources (Charterhouse) to increase deployable output to near licensed volume

2020

1

1.70

7.53

0.64

Increase performance of existing sources (Forum) to increase deployable output

2020

1

2.64

7.44

0.10

Catchment Management of the Mendip Lakes (Chew, Blagdon and Cheddar) to manage outage risk from algal blooms

2020

0

0.39

0.00

0.50

Alderley WTW (increased production)

2020

2

2.00

0.52

0.08

Cheddar WTW (increased production)

2020

2

4.00

13.27

0.09

Reduced leakage from raw water mains

2020

5

5.50

1.05

0.00

(continued)

19 An Integrated Approach to Water Resources …

251

Table 19.2 (continued) Option Name

Earliest option start (Year)

Years to fully implement (Years)

Yield (Ml/d)

Capital Costs (£M/year)

Operating Costs (£M/year)

New water sources within Bristol Water CAMS area for the location Middle River Avon at Bathford

2020

2

1.40

11.63

0.31

New water sources within Bristol Water CAMS area for the location Bristol Frome at Frenchay

2020

2

1.10

2.93

0.14

Cheddar Reservoir Standard WRMP14 design

2020

3

16.00

46.67

0.00

Purchase water from third parties from water companies

2020

3

10.00

13.77

0.58

Bring Honeyhurst source back into supply

2020

1

2.40

4.80

0.02

Active Leakage Control, Leakage Reduction to 36.50

2020

0

2.80

0.05

0.19

Active Leakage Control, Leakage Reduction from 36.5 to 35

2029

0

1.50

0.03

0.08

252 Table 19.3 Optimisation parameters for nsga2R optimisation

D. Staszek et al. Parameter

Value

Size of population

200

Number of generations

200

Crossover probability

0.7

Mutation probability

0.2

Fig. 19.2 Bristol Water Supply Demand Balance position [5]

The increase of the total WAFU in 2025/26, both for baseline and final planning simulations, is due to the reduction of water export to the neighbouring company. There is an expected water deficit from 2035/36 hence the company needs to implement some mitigation strategies. The final supply demand balance position selected by Bristol Water is presented in Fig. 19.3. The company achieved a positive supply-demand balance position for final planning after: (1) reducing leakage from 39.33 Ml/d in 2020/21 to 36 Ml/d in 2029/30 and further to 35Ml/d in 2034/35, (2) reducing export to neighbouring company from 11.37 Ml/d to 4.37 Ml/d after 2024/25, and (3) raw water leakage reduction by 5.5 Ml/d from 2040/41. The rapid leakage reduction was required by Ofwat, hence results presented in Fig. 19.3 may not be optimal for cost-benefit calculation [16]. The raw water transfer reduction was agreed with the neighbouring water company and for the revised draft WRMP is included within the baseline, hence it is not considered during the optimisation process [9]. The scenario testing implies that more challenging supply-demand position than the presented for the baseline needs to be considered. As a result, Bristol Water needs to implement more mitigation strategies including drought management option,

19 An Integrated Approach to Water Resources …

253

Fig. 19.3 Final Supply-Demand Balance positions from: Bristol Water WRMP [5, 9] and the integrated model

increase production from Forum Water Treatment Works and raw water leakage reduction. For the comparison purposes we run the model for the baseline Bristol Water’s supply-demand scenario, hence raw water export is already included within the baseline. The results of that optimisation using the integrated model returns the portfolio of the following schemes: (1) leakage reduction up to leakage level at 36.5 Ml/d from 2034/35 to 2038/69, (2) further leakage reduction up to 35Ml/d from 2039/40, and also (3) raw water leakage reduction (by 5.5Ml/d) from 2040/41. Bristol Water implements leakage saving options from year 2020/21. These savings affect the customer’s section of the water distribution pipe, hence they reduce PCC. The company is required to report their rounded PCC figure to an integer. The rounded PCC value is used to derive the distribution input figure, causing a difference of 0.8 Ml/d in the distribution input calculation for year 2020/21. The total WAFU figure from the integrated model is the same as reported by Bristol Water in their WRMP19 [5, 9] as the presented model selected the same supply options with the same timing for their implementations. The preferred supply-demand position after using our integrated approach is presented in Fig. 19.3. The main difference is a much smaller water surplus over the planning horizon for presented integrated approach when compared with Bristol Water’s supply-demand problem optimisation. The leakage reduction schemes are delayed till the time when water deficit first time appeared. Bristol Water chose to reduce leakage at the beginning of the planning horizon because of the included customers’ willingness to pay

254 Table 19.4 A number of simulations results with reservoir level below control curve

D. Staszek et al. Reservoir Level below Control Curve

No of simulations

none

17

1

17

2

34

3

20

4

7

5

3

6 Total

1 99

(WTP) preference to reduced leakage. As Bristol Water employed their cost optimization model in order to select the most cost beneficial options, hence options with costs adjusted by customer WTP values were preferable. In the traditional two stages approach (first DO estimation within water resources model then selection of the investment strategies using the investment model), DO stays constant over the planning horizon. Water companies are also required to report their actual level of service [15]. As DO does not vary over time, the actual level of service needs to be assessed against the minimum value of WAFU. If for particular year WAFU exceeds the distribution input then the actual level of service is greater than the target level of service. In the case of simulation methods, and for our proposed integrated model, the actual level of service can be represented in probabilistic terms. To assess the actual level of service for the system with proposed options, we run the integrated model with these options affecting supply-demand positions for each year over the planning horizon. The model was run 99 times for demand-supply simulations and these results are presented in Table 19.4. As the simulation was run for 25 years and Bristol Water assumed level of service is 1 in 15, the majority of simulations show that the reservoir level is below the control curves up to 3 times over the 25 years period. The integrated model runs on monthly data hence simulation results shows the effect of varying supply and demand positions on the actual level of service each month. The two stages approach—first delivering a DO figure from the water resources model for an assumed levels of service, and then using this DO in the investment model, does not test on how many occasions the proposed levels of service has been breached. The presented integrated model allows to calculate the number of levels of service breaches over the planning horizon for each supply simulation.

19.5 Conclusions and Further Work Based on the case study results the following conclusions can be drawn about the integrated model:

19 An Integrated Approach to Water Resources …

255

• it can provide valuable information in terms of investment strategies and intervention timing. In the investigated case study, the leakage reduction schemes were selected by the integrated model later than presented in Bristol Water WRMP19. When only the cost saving objective is considered, it could potentially bring some monetary savings to the company. • it can provide a better understanding of the proposed and actual levels of service under uncertainty. Bristol Water derived their DO for their assumed levels of service and this measure has not been recalculated post-optimisation for the final planning scenario. The integrated method guarantees that any investment selected by the optimiser affects WAFU and DI figures. For the integrated approach the post-investment values both for WAFU and DI are used to calculate the actual level of service. And, • it consistently selects the same investment options as the optimisation tool used by Bristol Water. As those options provide the best cost-benefit ratio in terms of NPV cost per yield achieved, this demonstrates that the integrated approach correctly integrates economic and water resource parameters into the decisionmaking process. Further improvements of the model could provide a tool that can support decisionmaking for more extreme demand-supply positions. Currently, all considered uncertainties are represented by the headroom approach. Some well-understood uncertainties could be excluded from the headroom assessment, and they should then be represented as uncertainties against a specific asset or process. Modelling a supplydemand balance problem with directly-linked uncertainties, against a particular asset or process, could potentially improve understanding of the effect of assets or process management under that uncertainty on decision-making. As the proposed model is multi-objective, performance indicators other than minimal total costs can be also considered. For instance, Bristol Water captures the resilience metric against each investment option. The integrated model could optimise for two objectives: minimal total costs and maximum total resilience metric and achieved results could be compared with Bristol Water’s results for single-objective optimisation. This is a further direction for our research.

References 1. Environment Agency, Ofwat, Defra and the Welsh Government, Water resources planning guideline (2012). The technical methods and instruction 2. Environment Agency & Natural Resources Wales, Final Water Resources Planning Guideline (2016) 3. UKWIR, WRMP 2019 Methods -Decision Making Process: Guidance (2016) 4. UKWIR, WRMP 2019 Methods –Decision Making Process: Risk Based Planning (2016) 5. Bristol Water, Draft Water Resources Management Plan. Bristol Water (2017) 6. Bristol Water, Drought Plan (2016) 7. Deb K. (2001) Multi-objective optimization using evolutionary algorithms/Kalyanmoy Deb. Wiley, Chichester

256

D. Staszek et al.

8. Deb K, Pratap A, Agarwal S, Meyarivan T (2002) A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Trans. Evol. Comput. 6(2):182–197 9. Bristol Water, Draft Water Resources Management Plan 2019. Statement of response to public consultation (2018) 10. Bristol Water, Water Resources Management Plan 2014 (2014) 11. Water Resource Associate Ltd HYSIM User Manual (2006) 12. Prudhomme C, Dadson S, Morris D, Williamson J, Goodsell G, Crooks S, et al. (2012) Future flows climate data. NERC Environmental Information Data Centre 13. Roach T, Kapelan Z, Ledbetter R (2015) Comparison of info-gap and robust optimisation methods for integrated water resource management under severe uncertainty. Procedia Eng. 119:874–883 14. Ledbetter R, Prudhomme C, Arnell N (2012) A method for incorporating climate variability in climate change impact assessments: sensitivity of river flows in the Eden catchment to precipitation scenarios. Clim. Change 113(3–4):803–823 15. Ofwat (2017) Consulting on our methodology for the 2019 price review 16. Environment Agency (2017) Leakage in WRMPs, June 2017

Chapter 20

Model Improvement for Effect Evaluation of Low Impact Development Measures Yuting Meng, Na Li, Jing Wang, Qian Yu, and Nianqiang Zhang

Abstract Low impact development (LID) practices, such as bioretention and green roof, aim to increase infiltration and retention to manage urban flood. Flood simulation model is an effective tool for scheme comparison at the design stage. In addition, it is also a good method to evaluate the effects of LID after construction. The Flood Risk Analysis Software (FRAS), developed independently by China Institute of Water Resources and Hydropower Research (IWHR), is integrated software that can simulate the whole flood process, mainly including 1D-2D coupling hydraulic model, hydrological model and drainage model. And FRAS also has accurate and reasonable structure routines for urban flood simulation because of the coupling calculation of different models. In order to reflect the influences of LID measures on runoff generation and confluence more accurately, the following improvements were made based on the original software: (1) The SCS method is added to the original rainfall-runoff model due to its simple but sensitive parameter, which can comprehensively reflect initial soil moisture, soil type and land use type of different LID measures. Additionally, it is coupled with the 2D surface hydraulic model in real-time to precisely describe the infiltration process. (2) According to the main functions and characteristics of different LID practices, three typical measures, i.e., green roof, bioretention and porous pavement, are set as special land use types, Y. Meng · N. Li (B) · J. Wang · Q. Yu · N. Zhang China Institute of Water Resources and Hydropower Research, A-1 Fuxing Road, Haidian District, Beijing, People’s Republic of China e-mail: [email protected] Y. Meng e-mail: [email protected] J. Wang e-mail: [email protected] Q. Yu e-mail: [email protected] N. Zhang e-mail: [email protected] Research Center on Flood and Drought Disaster Reduction of the Ministry of Water Resources, A-1 Fuxing Road, Haidian District, Beijing, People’s Republic of China © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_20

257

258

Y. Meng et al.

respectively. The parameters and calculation methods are adjusted accordingly in the model. After these improvements, the model can have more precise simulation about the increased infiltration and flood retention of LID measures. Keywords Low impact development · Model improvement · Effect assessment · SCS method

20.1 Introduction The traditional development mode has rapidly increased the imperious area and decreased the infiltration rate in urban area, accompany with the increase of runoff volume and the peak discharge, as well as the advance of peak time [2, 13, 19]. Therefore, the damage of rainstorm to the city has also been increased. The LID mode occurred in 1990s aims to use the scattered and small-scale measures to manage the city runoff [1], as well as to realize the win-win result between economy and environment protection. The LID measures can be both used for the reconstruction of old town and the new construction of new region [1, 7]. In order to evaluate the effects of LID measures on urban runoff, many scholars and institutions have carried out a large amount of research on LID measures in the aspects of site design, laboratory studies and model evaluation [1, 7]. Among this, the model evaluation is a kind of scientific and economic method, which has been used a lot for effects evaluation under various situations. There are lots of numerical models and softwares, such as SWMM, MIKE, SUSTAIN, InfoWorks and Sewer GEMS, are widely applied to effects evaluation of LID measures [4, 11, 15, 21]. Most numerical models are designed to simulate the flood problems in watershed or urban area respectively in the beginning, but adding the LID module for model function improvement [3]. Since their specific characteristics, these softwares or models aresuitable for different spatial scale and have different kind of advantages [21]. For example, the SWMM model is capable to simulate the individual LID measures at small region and also the effects of combined LID measures at city scale [3], which has precise attribute series concerned with all kinds of component. However, data input in SWMM is very inconvenient and inaccurate since the DEM data can’t be directly imported to the model and the software cannot automatically extract the surface parameters. And SUSTAIN model, which is developed based on GIS platform and combined with scheme optimization module, is more accurate at small scale and has more comprehensive functions. The Flood Risk Analysis Software (FRAS), developed independently by China Institute of Water Resources and Hydropower Research (IWHR), is integrated software that can simulate the whole flood process, mainly including 1D-2D coupling hydraulic model [5, 14], hydrological model and drainage model. This software has already used to simulate the urban flood through GIS-based analysis interface in China for many years. The 2D hydraulic model inside is used to simulate the surface flow and flow in wide rivers and the 1D hydraulic model is for flow in narrow

20 Model Improvement for Effect Evaluation of Low Impact Development …

259

rivers and the flow spreading along streets. And the 1D hydraulic model inside is coupled with the 2D hydraulic model by calculating the flood along the passage and the flow exchange between the passage and the grids on its both sides. Besides, hydrological model simulates the process that precipitation transfers to net rain and drainage model reflect the function of drainage pipe network, pump station and other drainage facilities. The drainage model contains underground reservoir model, the equivalent network model and network model. In addition, the hydrological and drainage model are both coupled with the 1D and 2D hydraulic model on the special passages and grids respectively. So the models are closely connected and real-time synchronous simulation is realized. Therefore, FRAS has accurate and reasonable structure routines for urban flood simulation because of the coupling calculation of different models. However, the LID measures are special kinds of facilities that can increase the infiltration and have water retention effect, which cannot be reflected in the original models. Consequently, this study has improved the models in the software for effect evaluation of LID measures, so as to be better applied for new urban situation.

20.2 Model Improvement According to numerous studies, the effect on runoff of LID measures is related to various aspects, such as the scale of measures, topography of the study area and different climatic conditions, which will lead to a wide range of effect [6, 8, 16, 17]. In order to reflect the influences of LID measures on runoff generation and confluence more accurately under different conditions, the following improvements were made based on the original software: (1) The SCS method is added to the original rainfallrunoff model and it is coupled with the 2D surface hydraulic model in real-time. (2) The parameters and calculation methods are adjusted in calculation units where the LID measures are located.

20.2.1 Hydrological Model 20.2.1.1

Comparison of Different Hydrological Calculation Method

The hydrological model in RARS simulates the process that precipitation transfers to net rain. In order to reflect the infiltration and water retention effects of LID measures, the chosen hydrology model should be able to describe the changes of soil property and land use type after implementing the LID measures. In this study, four commonly used urban hydrology models: runoff coefficient model, Horton formula method [12], Green-Ampt formula method [10] and SCS-CN curve method [18] are compared in Table 20.1 to choose the most suitable one.

260

Y. Meng et al.

Table 20.1 Comparison of typical surface runoff calculation methods Method

Theory

Advantage

Disadvantage

Runoff coefficient method

Empirical runoff coefficients of different underlay surfaces are used

Only one parameter, simple for calculation and data collection

(1) Cannot describe the runoff process precisely (2) LID measure have wide range of effect because of the related condition, runoff coefficient of LID measure is difficult to define

Horton formula method [12]

The infiltration rate decreases exponentially during rainfall, Horton formula is a relatively accurate empirical formula of infiltration rate changing with time

(1) Well represent the change of soil infiltration capacity over time in reality (2) Suitable for watershed runoff calculation

Hard to collect accurate infiltration rate of different LID measures in specific area

Green-Ampt formula method (Green and Ampt 9)

A physical model to describes the water movement in unsaturated soil layer based on capillary theory

(1) Simple formula form (2) Parameters have definite physical meaning (3) Accurate calculation

(1) Detailed and accurate soil data is required (2) Hard to collect accurate infiltration rate of different LID measures in specific area

SCS-CN curve method [18]

An empirical model based on the rainfall runoff data from small watershed of different areas in the United States

(1) Few parameters & simple calculation procedure (2) Easy to get the required data to define the parameters (3) Sensitive parameter to reflect effects of initial soil moisture, soil type and land use on runoff generation

(1) Curve number is empirical number with no specific physical meaning (2) Cannot describe the real physical process precisely

From comparison results, the infiltration parameters of Horton formula and GreenAmpt formula are difficult to collect and the empirical runoff coefficients of LID measures under different situation are hard to calibration. However, the Curve Number (CN) can be easily obtained from The CN Value Retrieval Table [20] based on different initial soil moisture, soil type and land use type. Moreover, the CN value is sensitive to various conditions and can reflect the change of hydrology characteristics after implementing the LID measures. Therefore, the SCS method is chosen

20 Model Improvement for Effect Evaluation of Low Impact Development …

261

in this study when the rationality of relevant parameters, simulation precision and convenience are all taken into consideration. In addition, this method is improved to describe the details of infiltration process and is combined with the 2D hydraulic model in real time.

20.2.1.2

Principle and Application of SCS-CN Curve Method

The SCS-CN method is proposed by the former US Department of Agriculture Soil Conservation Service. This method is an empirical model based on the rainfallrunoff data from small watershed of different areas in the United States. The relation between rainfall and runoff is expressed as following: F Q = S P − Ia

(20.1)

Where F is the latter loss of surface runoff (the infiltration), S is the maximum potential water storage capacity of soil, I a is the initial loss before the surface runoff (including vegetation retention, depression filling and soil retention of surface layer, etc.), P is the rainfall and Q is the surface runoff. The improvement in this study is mainly to apply the SCS-CN method in each time step to obtain the actual infiltration rate for runoff generation calculation based on the former Flood Simulation Numerical Model. In reality situation, P and F are both accumulated value and the S will change with time through the whole rainfall-runoff process. Therefore, the potential infiltration capacity till current time of each computing element will be obtained according to Eq. (20.1) in every time step. Besides, the initial rainfall loss is mainly reflected in the depression filling during urban hydrology process, which can be described by gridded terrain in the 2D hydraulic model. So the Eq. (20.1) is transferred to Eq. (20.2), where the I a is ignored and the infiltration is regarded as the main rainfall loss. And the potential infiltration rate related to soil is calculated according to the actual accumulated infiltration and the potential infiltration capacity from Eq. (20.2). In addition, the actual infiltration rate is the smaller one between the potential infiltration rate and the actual maximum infiltration rate, which is related to the water within each computing element. The calculation formula of infiltration rate of each time step is shown in Eq. (20.3):  2  PT PT −  PT + ST

(20.2)

F potential,T −DT )/DT, (PT +HT −DT )/DT )

(20.3)

F potential,T = f T = min((Fr eal,T −





 the infiltration rate, F potential,T is the potential cumulative infiltration Where f T is till time T-DT, which is till time T, Fr eal,T −DT is the real cumulative infiltration  accumulated from real infiltration obtained in the model, PT is the cumulative rainfall till time T, DT is the time step, S T is the maximum potential water storage

262

Y. Meng et al.

Fig. 20.1 Coupling principle of SCS method and hydraulic model

capacity of soil at time T, P T is the rainfall within the current time step and H T-DT is the inundation of last calculation moment. The net rainfall obtained within current time step will be used as source term in the hydraulic model to calculate the runoff process. And the real-time coupling procedure is shown in Fig. 20.1. According to the actual infiltration procedure, the parameters are adjusted as following when applying the SCS-CN method to each calculation element: (1) Due to the wide range of the initial maximum potential water storage capacity of soil S 0 , the curve number (CN) representing different soil-vegetation combinations is used to describe the S 0 . The relationship between S 0 and CN is expressed as:  S0 = 25.4

1000 − 10 CN

 (20.4)

Where CN is a comprehensive parameter reflecting the surface runoff generation capacity, and the value of CN is usually between 30 and 98 in reality. (2) Besides the actual rainfall, the inundation within the calculation unit will also influence the infiltration. Therefore, the maximum potential infiltration volume in Eq. (20.3) is set as the sum of the actual rainfall in the current time step and the inundation in previous time step.

20 Model Improvement for Effect Evaluation of Low Impact Development …

263

(3) The maximum  potential water storage capacity of soil ST and the real cumulative infiltration Freal,T will change with time. Therefore, the two parameters should update to be used as the input value of the next time step after each time step in the model.

20.2.2 Simulation of LID Measures There have been a large number of studies about individual LID measures till now, many of which are about the bioretention, green roof and porous pavement. So the three typical measures are also chosen in this study to improve the simulation of LID measures in FRAS. The parameters of the grid or road or calculation method are adjusted accordingly in the model.

20.2.2.1

Adjustment for the Parameters

The main function of LID measures involved in the research is to increase infiltration and flood retention, so the influenced parameters in the model are the elevation, roughness and CN values of grid and road passage. The parameters of calculation unit is a weighted average number of different parts inside, which will be changed after the adjustment to the parameter values of the area with LID measures in the grid and road. (1) CN value The CN value mainly influences the hydrology process to reflect the effect of increased infiltration after implementing LID measures. The number is obtained from CN value retrieval table given in the National Engineering Handbook (USDA NRCS, [20] according to the type of soil, the type of land use and the antecedent moisture conditions (AMC). The soil type of different LID measures is identified as following: ➀ Green roof: The type of substrate layer, ➁ Porous pavement: The type of porous base course, ➂ Bioretention: The type of planting soil. And the weighted average CN values of the grid or road are calculated as shown in Eq. (20.5):  C Nweighted =

Ai ∗ C Ni  Ai

(20.5)

Where CN i is the CN value of each kind of land use within the calculation unit and Ai is the area of each land use within the calculation unit. (2) Elevation The adjustment of elevation is applied to the bioretention to reflect the water retention ability by reducing the elevation of region with measures. And the reduction

264

Y. Meng et al.

value is consistent with the depth of the water storage layer in the surface. The elevation ZB of the calculation unit containing bioretention is calculated as Eq. (20.6): Z B = Z Boriginal −

Hbio ∗ Abio Asum

(20.6)

Where ZBoriginal is the original elevation of calculation unit, Asum is the total area of calculation unit, H bio is the water storage depth of bioretention and Abio is the area of bioretention within the calculation unit. (3) Roughness mainly affects the confluence process in hydraulic calculation, which is influenced by the type of surface land use. The increased roughness can reflect the effect of measures on flood detention. The calculation method of weighted average roughness of grid or road can refer to the calculation of CN value. (4) The maximum water storage is a new parameter added to the original parameter series of grid. This parameter reflects the limited infiltration capacity of green roof due to its thinner soil layer than other measures. Consequently, the green roof area will not infiltrate when the cumulative infiltration exceeds the water storage capacity during the runoff generation calculation.

20.2.2.2

Improvement of Calculation Method About the Runoff in Green Roof Region

The grids with green roof are divided into three regions: green roof area, non-greenroof housing area and non-housing area. The housing area in the grid is considered as impervious region (no infiltration) for the calculation of the original hydrologicalhydraulic model. And the inundated water is concentrated in non-housing area. However, the increased infiltration and water retention effects of green roof influence the runoff generation procedure of grid. For this reason, the green roof area and nonhousing area of the grid are calculated respectively in the model according to the process in Fig. 20.1 for infiltration. In the non-housing area, maximum potential infiltration volume in Eq. (20.3) is the sum of rainfall and grid inundation volume. But the maximum potential infiltration volume in the green roof area is just the rainfall and the green roof area will not infiltrate when the cumulative infiltration is bigger than the maximum water storage value. In addition, the non-green-roof housing areas do not infiltrate, where the net rainfall is equal to the original rainfall. The net rainfall was redistributed in the non-housing areas using the area weighting method after obtaining the net rainfall in these three areas. Then the net rainfall of grid Pnet,grid will be used as source term to couple with the hydraulic model. Pnet,grid is calculated by Eq. (20.7): Pnet,grid =

Ano−house ∗ Pnet,no−house + A g f ∗ Pnet,g f + Ano - green roo f ∗ Prain A grid (20.7)

20 Model Improvement for Effect Evaluation of Low Impact Development …

265

Where Ano−house is the area of non-housing region, Pnet,no−house is the net rainfall of non-housing region, Ag f is the area of green roof region, Pnet,g f is the net rainfall of green roof region, Ano−gr een r oo f is the area of no-green-roof housing region, Prain is the original rainfall, and Agrid is the area of whole grid.

20.3 Model Application In order to test the rationality of improved model, Square West Ditch basin, belong to Jinan City, Shang Dong province, China, is chosen as pilot area for scheme simulation. The basin is located in the central urban area with flat terrain and high urbanization. The area of this watershed is 5.66 km2 , 16% of which is the residential area. The simulation schemes are divided into two parts: ➀ Simulation of a typical historical flood at 16th August 2016 in Jinan. ➁Simulation of different scale of LID measures under 24 h design storm with 10-year return period of Jinan. The study area is divided into 4133 irregular grids with an average area of 1500 m2 (38 × 38 m). As the width of Square West Ditch does not reach the average size of the grid, all rivers are set as special river passages. In addition, all roads are set as special road passages. All other attributes concerned are extracted in the software based on the collected data. 9 lower boundaries for outflow (1 from river, 1 from road and 5 from ordinary cells) are set in the model since the higher terrain in the southern of basin, discharge process of which are calculated according to Manning formula. And the initial water depth of Square West Ditch is set referring to its annual average water depth.

20.3.1 Simulation of Typical Historical Flood The typical historical flood at 16th August 2016 in Jinan has accumulated precipitation between 55.8–98.3 mm in different rainfall stations. And the rainfall is concentrated at 4:00 to 11:00 am. The total simulation time in the model is set as 13 h with the 3 s time step. According to Hydrological Yearbook of Jinan in 2016, the measured data of hydrological station in Yuhan road was obtained. The results in Table 20.2 show that the absolute difference between measured maximum water depth and simulated one is less than 10 cm, representing that the overall accuracy is Table 20.2 Comparison results about maximum water depth of road Station name

Shengwei second dormitory

Location

Yuhan road

Maximum water depth (m) Measured

Simulated

Absolute difference (m)

0.10

0.12

0.02

266

Y. Meng et al.

Fig. 20.2 Distribution of simulated grid maximum water depth for the rainstorm in 16th August 2016

good. In addition, the simulated inundation distribution (shown in Fig. 20.2) is consistent with the characteristics of the terrain distribution. Therefore, all the results reflect that the improved model and chosen parameters are suitable for this area to simulate urban flood.

20.3.2 Simulation of Different Scale of LID Measures Three individual LID measures, bioretention, green roof and porous pavement, are setting respectively in the basin for the effect comparison under different scale. 30, 50, 70 and 100% of the settable area of different LID measures are selected to implement, with the increase order from upstream to downstream successively. The settable areas of green roof, porous pavement and sunken green belt are 0.958, 1.988 and 0.821 km2 respectively in this basin. And the scenario without LID measures is also simulated. Values of the region with measures (shown in Table 20.3) are set according to the common material of the three measures and The CN value Retrieval Table.

20 Model Improvement for Effect Evaluation of Low Impact Development …

267

Table 20.3 Relevant model attribute of LID measures in the pilot area LID measures

Elevation

Roughness

CN

Bioretention

Reduce 20 cm

0.06

61

Porous pavement

No change

0.035

66

Green roof

No change

0.07

61

Maximum water storage

27.7 mm

Fig. 20.3 Surface runoff hydrograph under scenarios with different size of green roof

Fig. 20.4 Surface runoff hydrograph under scenarios with different size of bioretention

In this study, the hydrograph of Square West Ditch basin is chosen to compare the effects of individual LID measures under different scale (Figs. 20.3, 20.4 and 20.5). The results indicate that all measures can decrease the total volume of surface runoff and the reduced volume will increase with the increase of measure’s size. And the simulation results also show that this model can distinguish the effect of different LID measures. This is mainly because the parameters are set according to specific kind of LID measures,

268

Y. Meng et al.

Fig. 20.5 Surface runoff hydrograph under scenarios with different size of porous pavement

20.4 Conclusion This study developed the original FRAS software by adding the improved SCS method and LID simulation method to the original model. The improved SCS method can comprehensively reflect the effect of initial soil moisture, soil type and land use type of different LID measures during the whole runoff generation process. In addition, the characteristics of three kinds of typical LID measures, i.e., green roof, bioretention and porous pavement, are reasonably generalized by adjusting parameters and calculation method in the model. After these improvements, the intergrated model can reflect the influences of LID measures on runoff generation and confluence process more accurately. In order to test the rationality of improved model, Square West Ditch basin is chosen as pilot area for scheme simulation. The simulation results under different scenarios reflect that the improved model is suitable to simulate urban flood and can distinguish the effect of different LID measures under different conditions. Therefore, the model can have more precise simulation about the increased infiltration and flood retention of LID measures. Besides, this software will provide better technical support for scheme comparison and effect evaluations of LID practices under different spatial scales and conditions.

References 1. Ahiablame LM, Engel BA, Chaubey I (2012) Effectiveness of low impact development practices: literature review and suggestions for future research. Water Air Soil Pollut 223(7):4253–4273 2. Booth DB, Hartley D, Jackson R (2010) Forest cover, impervious-surface area, and the mitigation of stormwater impacts. JAWRA J Am Water Resour Assoc 38(3):835–845 3. Cai LH (2016) Introduction of hydrological and hydraulic models for “sponge city”. Digit Landsc Archit 02:33–43

20 Model Improvement for Effect Evaluation of Low Impact Development …

269

4. Cheng T et al (2017) Rainfall-runoff simulations for Xinglong sponge city pilot area of Jinan. J Hydroelectr Eng 36(06):4–14 5. Cheng XT (2009) Urban flood prediction and its risk analysis in the coastal areas of China. China Water Power Press, Beijing 6. Davis AP (2008) Field performance of bioretention: hydrology impacts. J Hydrol Eng 13(2):90– 95 7. Dietz ME (2007) Low impact development practices: a review of current research and recommendations for future directions. Water Air Soil Pollut 186(1–4):351–363 8. Dreelin EA, Fowler L, Ronald C (2006) A test of porous pavement effectiveness on clay soils during natural storm events. Water Res 40(4):799–805 9. Green WH, Ampt GA (1911) Studies on soil phyics. J Agr Sci 4(1):1–24 10. Green WH, Ampt GA (2015) Studies on soil physics part i - the flow of air and water through soils. Int J Nonlinear Sci Numer Simul 4(7–8):1–24 11. Guan M, Sillanpää N, Koivusalo H (2015) Assessment of LID practices for restoring predevelopment runoff regime in an urbanized catchment in southern Finland. Water Sci Technol 71(10):1485–1491 12. Rui XF (2013) The discovery and development of runoff formation models. Adv Sci Technol Water Resour 33(1):1–6 13. Leopold L (1968) Hydrology for urban planning—A guidebook on the hydrologic effects of urban land use. US Department of the Interior US Geological Survey Circular, p 554 14. Liu SK, Song YS, Cheng XT, et al (1999) Risk analysis and disaster reduction countermeasures for flood plain area and flood detention area of Yellow River. Yellow River Water Resources Press 15. Paul S, Calmbacher J, Manguerra H, Wellborn J (2011) Evaluating implementation of LID/BMP in storm water system using EPA SUSTAIN. Proc Water Environ Fed 2011(2):590–601 16. Roehr D, Kong YW (2013) Runoff reduction effects of green roofs in Vancouver, BC, Kelowna, BC, and Shanghai, P.R. China. Can Water Resour J 35(1):53–68 17. Rushton BT (2001) Low-impact parking lot design reduces runoff and pollutant loads. J Water Resour Plan Manag 127(3):172–179 18. Soil Conservation Service (1972) National engineering handbook, section 4: hydrology. US Department of Agriculture Soil Conservation Service, Washington D C 19. Suriya S, Mudgal BV (2012) Impact of urbanization on flooding: the Thirusoolam sub watershed – a case study. J Hydrol 412(1):210–219 20. USDA NRCS (2004) National Engineering Handbook [EB/OL]. Title 210-VI. Part 630. In: Hydrologic Soil-Cover Complexes. U.S. Department of Agriculture, Natural Resources Conservation Service, Washington DC 21. Zhang CS, et al (2018) Overview of typical storm flood models applied to sponge city construction. Water Purif Technol 37(08):51–55+60

Chapter 21

Multiple Feedback Linkage in the Process of Urban River Water Treatment Lunyan Wang, Shoukai Chen, Xiangtian Nie, and Shichao Liu

Abstract According to the measurement data of hydrology and water quality, the decision support system in the multiple feedback linkage system is used for calculation, combined with the judgment and decision of the expert system in the water treatment expert assistant decision analysis system, the precise control of the Internet of Things equipment in the water treatment project were realized. an intelligent water body management system was established that combined human intelligence with machine intelligence to realize the process of water body treatment and monitoring technology to collect, analyze and process data, to achieve more efficient water treatment and more accurate water body monitoring. Use the comprehensive water-related data collected to provide decision support for regional industry development. Keywords Multiple feedback linkage · Internet of Things · Intelligent · Water treatment

L. Wang (B) · S. Chen · X. Nie School of Water Conservancy, North China University of Water Resources and Electric Power, No. 136 Jinshui East Road, Zhengzhou, China e-mail: [email protected] S. Chen e-mail: [email protected] X. Nie e-mail: [email protected] Henan Key Laboratory of Water Environment Simulation and Treatment, No. 136 Jinshui East Road, Zhengzhou, China L. Wang · S. Chen Water Science Institute Building, Academician Workstation of Water Environment Governance and Ecological Restoration in Henan Province, No.11 Weiwu Road, Zhengzhou, China L. Wang · S. Chen · S. Liu Water Science Institute Building, Innovation Strategic Alliance of Water-Soil Environment Collaborative Treatment Industry Technology, No.11 Weiwu Road, Zhengzhou, China e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2020 P. Gourbesville and G. Caignaert (eds.), Advances in Hydroinformatics, Springer Water, https://doi.org/10.1007/978-981-15-5436-0_21

271

272

L. Wang et al.

21.1 Introduction The sustainable use of water resources is extremely important guarantee for sustainable economic and social development [1]. With the rapid development of social economy, water resources have become the most important resource that restricts China’s economic and social development, and are essential elements for ecology and environment [2]. At this stage, water resources are facing serious threats, and the water ecological conservation space is severely squeezed [3]. Unreasonable development mode and human activities have caused competition with rivers and lakes, water ecological conservation space such as Water conservation area, river and lake marsh area, flood storage and stagnation area have been seriously encroached, resulting in significant changes in rivers, lakes, water and other cycling conditions, the lakes and river tails have shrunk, and the water ecological spatial pattern has been squeezed and destroyed [4, 5]. At present, there are many kinds of river ecological restoration technologies [6], including: in the river improvement, combined with flood management, The concept of “giving space to the river” is implemented through the construction of spillway and the lowering of flood plain elevation.; river continuity recovery: vertical connectivity and horizontal communication between rivers and floodplain areas, as well as construction of low dams and setting of fish passes; dikes for demolition or retreat; Meandering restoration of the river; ecological protection of river bank slopes; reconstruction of river deep troughs and shoal sequences; creation of wetland characteristics in floodplain; enhancement of habitats in rivers (such as shelters, shading, diversion facilities, etc.); construction of hydrophilic facilities; utilization of dredged rivers; development and application of porous and permeable revetment materials and structures, and engineering construction techniques [7–9]. In addition, in combination with river ecological restoration planning and design, some planning design models and methods have also been proposed [10]. In the dambuilding rivers, in order to improve the ecosystem status of the downstream rivers, research and demonstration of reservoir optimization scheduling methods have been carried out in some countries, and some results have been obtained, such as river ecological water demand assessment technology, the impact of flood process on fish reproduction natural hydrological process simulations, etc [11, 12]. This paper realizes automatic water monitoring and water restoration by constructing intelligent water treatment system, and carries out independent analysis and prediction based on feedback data to improve water treatment efficiency. The Internet of Things technology controlled by the multiple feedback linkage system enhances and maintains the self-purification ability of the water, restores and maintains the natural connectivity and fluidity of the water system, thereby constructing a virtuous cycle urban water system and repairing the quality of the urban natural ecological environment [13].

21 Multiple Feedback Linkage in the Process …

273

21.2 Expert Assistant Decision-Making Analysis System Based on Urban River Water Treatment The expert assistant decision-making analysis system for water treatment is based on advanced technologies such as big data analysis, machine learning, intelligence, and networking. It collects various types of water data and uses big data analysis technology, water knowledge base and expert knowledge base to assist decision-making and other technologies. Learning to adapt to changes in water, learning new knowledge independently, updating the knowledge base, automatically providing convenient water treatment solutions for the system, and inputting the treatment scheme to the water treatment field operation through the network, collecting feedback data of treatment effects, and implementing treatment data. Realized the intelligent whole process of collection, treatment planning, field treatment operation, and feedback on treatment effects, and an intelligent system for water treatment with simulated human expert thinking is established to achieve efficient and precise operation of water treatment, thereby improving the efficiency of water treatment and reducing cost.

21.2.1 System Framework Analysis The expert assistant decision-making analysis system for water treatment is based on the water big data analysis system, and the solution to the whole water environment data feedback problem is given according to data analysis results in time. The system uses big data, machine learning and other platform technologies to intelligently manage the entire process, from water analysis and diagnosis, treatment plan generation, program review and prediction, treatment field operation, data collection from treatment effect, to the final treatment effect tracking and evaluation. It mainly includes: water dynamic data collection, water analysis and evaluation, water treatment program management, treatment field operation management, treatment effect tracking and evaluation, water knowledge base management, etc. The big data water treatment expert decision-making base system, It can also be a subsystem of an independent system. Applying it to urban water treatment and maintenance can achieve an effective combination of human intelligence and machine intelligence (Figs. 21.1, 21.2 and 21.3).

21.2.2 System Function Design The water treatment expert assistant decision-making system is divided into functional modules: water knowledge base management subsystem, water environment

274

L. Wang et al.

Water environment model management Water quality assessment and prediction Water environment capacity Analysis parameter setting Water quality exceeded diagnosis Water balance analysis

Self-learning generation scheme Program process editor modification Governance program process audit Governance process parameter configuration Governance effect forecast Governance program process release

Environmental

Water environment problem of analysis results analysis diagnosis and evaluation

Governance effect data collection Governance effect monitoring Governance effect analysis and evaluation Self-learning update knowledge base Self-learning update governance plan model

Program execution effect

Water environment management plan

Governance effect tracking and evaluation

Water quality data Water pollution data

Dynamic data collection

Water Environment Big Data Processing Center

Water environment expert database management

Data collection filter configuration

Basic environment

Water environment knowledge base management

Hydrological data

Intelligent expert assistant decisionmaking platform

Hadoop+Hbase Big Data System

Distributed environment

Data analysis mining tool

Governance plan and process input

Machine learning and artificial intelligence environment

Governance data feedback

Governance process site implementation

Optimal management plan and process display

Intelligent governance operations Process requirements

Water environment treatment site

Process requirements

Governance implementation feedback

Working equipment PLC acquisition Actual job data report

Fungus delivery

spray

Actual spray volume collection

operators

Fig. 21.1 Framework of water environment treatment intelligent expert assistant decision-making system

model, self-learning system of water treatment in Pingyu county, China, and water expert database. (1) Water knowledge base management subsystem The water knowledge base management subsystem through the process of obtaining, creating, sharing, integrating, recording, accessing and updating the information and knowledge of rivers and lakes in the whole treatment system to achieve the goal of continuous self-renewal, and feedback the updated information to the knowledge system at the same time. By repeating the above process, it can not only expand

21 Multiple Feedback Linkage in the Process …

275

Water environment data collection Water quality data

Water pollution data

Water environment standard data

Hydrological data

Data collection

Water environment big data

Data to be analyzed Water environment analysis and evaluation

Model library

Model library management

Water quality assessment, prediction, balance analysis and capacity calculation

Water environmental problem diagnosis

Problem to be solved Governance program management Water Environment Expert Knowledge Base

Self-learning automatic generation scheme

Program modification, review and performance forecasting

Program release

Governance programs and processes Governance field work Program receiving display

Intelligent governance operations

Actual governance feedback

Governance outcome Governance effect tracking and evaluation Expert knowledge base management

Self-learning update expert knowledge base

Effect analysis

Performance data collection

Need to continue to govern Fig. 21.2 Water treatment automation

the width of the original information of the system, but also help enterprises make correct decisions to adapt to the changes of the market. (2) Decision model of water treatment The water environment treatment decision model is the core of the entire big data decision support system. Through the model and algorithm, the system automatically performs verification calculation based on the collected data and selects the treatment

276

L. Wang et al. Regional water quality data collection

Regional hydrological data collection

Regional environmental data collection

Meteorological data collection

Water balance judgment

Water environment decision model

No

System decision result

Yes

Emergency process

Machine learning, correcting model parameters

Expert review

Linked execution

Flood control and drought resistance

Ineffective

Performance feedback

Expert analysis

Good effect

Fig. 21.3 Implementation of expert assistant decision-making system for water treatment

scheme. In the initial stage, experts are required to judge. The operation management personnel conducts flood control and drainage and water treatment according to the prediction and early warning data. In the later stage, the system carries out intelligent learning and judgment (machine learning) according to the manual processing scheme (historical response operation), thereby realizing the automatic and intelligent development of the system. (3) Self-learning system of urban water treatment The self-learning system of urban water treatment is developed based on the current Internet background and big data technology. The system uses big data mining technology to continuously train the water environment model derived by traditional laboratories, so that the laboratory model can accurately match the actual situation of the project and the decision model can obtain reasonable and effective environmental governance decision data. At present, linear regression (Linear Regression) analysis method is often used for system self-study, and statistics and prediction are performed while performing a large amount of data analysis. (4) Water expert database subsystem The Water Expert Library Subsystem is a system for experts who are familiar with the project of water system or have experience in water treatment. The expert database system includes basic data management, expert management, expert notification,

21 Multiple Feedback Linkage in the Process …

277

system configuration management, and report statistics. Experts must have certain qualifications when they are selected into the expert database. If an individual applies, the unit recommends, the peer expert recommends and so on the way obtains the warehousing qualifications, and the individual application, the unit recommends the letter and the peer expert recommends should archive for future reference, as well as the process and results of the selection of experts. The system will create a file for each selected expert, detailing the specific situation and expertise of the expert.

21.3 Construction of Multiple Feedback Linkage System The purpose of the multiple feedback linkage system is to control the IoT equipment (gate pump, water quality analysis, in situ repair, etc.) based on the measurement results of hydrological water quality, calculated by the decision support system, and the judgment and decision of the expert system, so that the equipment can be opened automatically or the agent can be sprayed automatically. According to the difference of specific objectives and business processes, the feed-back linkage system is divided into: sluice pump linkage control, water quality equipment linkage control, in situ repair system linkage control.

21.3.1 Sluice Pump Linkage Control In the automatic control system of the sluice pump, the uppermost layer is the dispatching platform as the core node of the communication network. On the one hand, the real-time monitoring of each water quantity is unified with the real-time control technology of the sluice pump group, and various dispatching schemes and strategies based on mathematical models can be implemented. The real-time control system PLC of the sluice pump station in the field equipment layer is connected to the communication network through the Ethernet interface. The local monitoring center workstation and the dispatching platform workstation all realize the control of the sluice pump through the communication network. In the water quantity monitoring system, the local information of the sluice pump will be fed into the PLC. The communication mode adopts “duplex communication”: it means that the on-site monitoring computer can interact with the remote client. The sluice local control application RS232 protocol communicates with the frontend PLC, and sends corresponding control commands, which can realize real-time control of the sluice, remotely set the sluice opening degree, control the sluice to rise, fall, stop. At the same time, the monitoring interface and the operation status of all sluices are displayed on the software, and the video image of the corresponding sluice can be viewed. Compared with the traditional manual control in the past, this function improves the accuracy of the sluice control, and can instantly obtain accurate

278

L. Wang et al.

data such as the sluice opening degree, improve the safety and visibility of the sluice control, and save a large part of the labor cost. The local control unit uses PLC to complete the logic control of the sluice. The PLC first judges the existing conditions from the existing parameters of the sluice, such as the sluice position, the mechanical upper/lower limit position, the electronic upper/lower limit position, the button input, whether the sluice has a running mechanical fault alarm, etc. Then through the control command and logic operation to make the sluice complete the operation. In the software, the logic interlock and reverse delay in the rising and falling of the sluice are considered to prevent the mechanical and electrical shock of the sluice and valve. Whether the sluice height at the right and left reaches the requirements of deviation correction and whether the error is exceeded in the rising and falling are all realized through the ladder diagram software programming in the PLC (Figs. 21.4, 21.5 and 21.6). Big data expert decision system

Comprehensive feed linkage system

VPN private network

Results of enforcement

Local control cabinet

Local control cabinet

Gate valve

Gate valve

Fig. 21.4 Schematic diagram of gate pump remote linkage control service

Local control cabinet

Gate valve

21 Multiple Feedback Linkage in the Process …

279

Switching power supply

Photoelectric encoder

Button indicator Monitoring center host

Control method selection Gate opening gauge

Emergency stop 485

A/D conversion

Intermediate relay

Water level gauge DI AI DO Communication High pressure

Communication module

PLC programmable controller

upper limit

Air switch

380V

Contactor

Lower limit

Contactor

Therma l relay

Motor and brake

Fig. 21.5 Schematic diagram of gate pump control technology

Analog acquisition module

Water level indicator

Water level gauge before and after the gate

PLC programmable controller

Gate opening gauge

Absolute Absolute photoele photoelec ctric tric encoder encoder

485 communication module

Series Low Voltage Power Distribution and Protection Devices

Motors, solenoid valves and other equipment that require control

Fig. 21.6 Schematic diagram of the local control cabinet of the sluice pump

Based on the sluice position signal from the photoelectric sluice opening sensor installed at both ends of the sluice, the sluice measuring and controlling instrument can judge the actual position and state of the sluice, such as whether the sluice has reached the designated position, whether the left and right are out of alignment, whether the upper/lower limit of the sluice is in operation, etc. Then, according to the panel operation setting and button operation or monitoring the central sluice monitoring computer operation instruction, the PLC is given instructions for running, in-position, stop, automatic deviation correction and limit parking. These functions of the PLC are implemented by hardware and program software. All PLC form a

280

L. Wang et al.

communication system of a master and a multi-slave structure through the RS-485 bus communication interface and the monitoring center gate monitoring computer, and complete the acquisition and control commands issued by the upper computer. The monitoring center computer directly monitors the state of the sluice and valve, and controls the sluice and valve to the set opening degree through the application software. The measuring and controlling instrument in the local control cabinet and PLC directly control the sluice and valve, and respond to the control command and acquisition command of the upper computer, constitute a distributed system, and jointly complete the monitoring of the gate. (1) Local control cabinet control function i. In the local manual mode, the PLC button can be used to raise and lower the gate to the designated position. ii. Remote manual control: The operator inputs the required sluice opening degree or clicks the rising and falling commands in the upper computer interface, and issues an instruction to the main control PLC. The PLC responds to the upper computer command to realize the sluice operation control. iii. Remote automatic control: The operator sets relevant data such as water flow and sluice opening according to the situation. The software performs comprehensive calculation on the data, and transmits control commands to the PLC in real time. The PLC automatically performs the corresponding actions. (2) Local control cabinet function i. Priority: Manual Control > Remote Manual > Remote Automatic Control. ii. Protection: It has multiple protection functions such as mechanical limit, digital electronic limit and mode lock to ensure high reliability of sluice operation. iii. The parameters such as sluice opening degree, sluice lifting, sluice landing status, and stroke setting value can be displayed and set on site. iv. Communication: the monitoring center can modify the setting parameters remotely and issue control commands to make the sluice run to the designated position.

21.3.2 Water Quality Equipment Linkage Control The schematic diagram of linkage control of water quality monitoring equipment is shown in Fig. 21.7. The specific steps are as follows: (1) The big data platform judges whether it is necessary to collect water quality based on various data. (2) The platform gives the collection frequency and reporting method of different water quality collection devices, and the collection device makes corresponding collection actions according to the platform instructions. (3) The collection device uploads the collection result through wired and wireless methods, and the platform makes corresponding feedback after receiving the collection result.

21 Multiple Feedback Linkage in the Process …

281

Fig. 21.7 Schematic diagram of water quality equipment linkage

(4) The platform uses the acquisition results as input to other system actions.

21.3.3 In-Situ Repair System Linkage Control The PGPR in situ ecological restoration system can perform in situ repair under the condition of avoiding disturbance and damage to the water. Considering that the area to be repaired is large, the power required is large, and the amount of the sprayed agent

282

L. Wang et al.

is large. Therefore, the cost of long-term water maintenance should be controlled, and the optimal amount of spray is obtained through calculation and analysis by the decision system. The amount of pollution consumed by in situ ecological restoration technology within 12 months is greater than the value of water pollution load minus water environmental capacity. Therefore after 12 months of in situ ecological restoration. So after 12 months of in situ ecological restoration, the major indexes of the water, the COD, ammonia Table 21.1 The consumption rate of ecological restoration agents under different water quality conditions Concentration of contaminant (mg/L) COD

NH3-N

TP

Consumption rate of ecological restoration agent v (mg/L/hr)

>40

>2

>0.4

0.0071

>40

>2

0.3–0.4

0.0066 0.0065

>40

>2

2

>0.4

0.0062

30–40

>2

0.3–0.4

0.0060

30–40

>2

0.4

0.0061

2

0.3–0.4

0.0059

2

40

1.5–2

>0.4

0.0061

>40

1.5–2

0.3–0.4

0.0059

>40

1.5–2

0.4

0.0058

30–40

1.5–2

0.3–0.4

0.0055

30–40

1.5–2