Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa [1 ed.] 9781617615122, 9781617610035

Citation preview

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved. Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011. ProQuest

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved. Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

AFRICAN POLITICAL, ECONOMIC, AND SECURITY ISSUES

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

RECENT ADVANCES IN REMOTE SENSING AND GIS IN SUB-SAHARA AFRICA

No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in rendering legal, medical or any other professional services. Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

AFRICAN POLITICAL, ECONOMIC, AND SECURITY ISSUES Additional books in this series can be found on Nova’s website under the Series tab.

Additional E-books in this series can be found on Nova’s website under the E-book tab.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

COMPUTER SCIENCE, TECHNOLOGY AND APPLICATIONS Additional books in this series can be found on Nova’s website under the Series tab.

Additional E-books in this series can be found on Nova’s website under the E-book tab.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

AFRICAN POLITICAL, ECONOMIC, AND SECURITY ISSUES

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

RECENT ADVANCES IN REMOTE SENSING AND GIS IN SUB-SAHARA AFRICA

COURAGE KAMUSOKO CHARLES NDEGWA MUNDIA AND

YUJI MURAYAMA EDITORS

Nova Science Publishers, Inc. New York

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011 by Nova Science Publishers, Inc. All rights reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical photocopying, recording or otherwise without the written permission of the Publisher. For permission to use material from this book please contact us: Telephone 631-231-7269; Fax 631-231-8175 Web Site: http://www.novapublishers.com NOTICE TO THE READER The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained in this book. The Publisher shall not be liable for any special, consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or reliance upon, this material.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Independent verification should be sought for any data, advice or recommendations contained in this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage to persons or property arising from any methods, products, instructions, ideas or otherwise contained in this publication. This publication is designed to provide accurate and authoritative information with regard to the subject matter covered herein. It is sold with the clear understanding that the Publisher is not engaged in rendering legal or any other professional services. If legal or any other expert assistance is required, the services of a competent person should be sought. FROM A DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS. Additional color graphics may be available in the e-book version of this book. LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA Recent advances in remote sensing and GIS in Sub-Sahara Africa / editors, Courage Kamusoko, Charles Ndegwa Mundia, Yuji Murayama. p. cm. Includes index.

ISBN:  (eBook)

1. Remote sensing--Africa, Sub-Saharan. 2. Geographic information systems--Africa, Sub-Saharan. 3. Land use--Africa, Sub-Saharan. 4. Africa, Sub-Saharan--Geography. I. Kamusoko, Courage. II. Mundia, Charles Ndegwa. III. Murayama, Yuji, 1953G70.5.A357R43 2010 621.36'780967--dc22

2010028504

Published by Nova Science Publishers, Inc. † New York Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

CONTENTS   Foreword

vii 

Preface

ix 

About the Editors

xi 

About the Contributors

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Chapter 1

Chapter 2

Chapter 3

Chapter 4

Introduction: Usefulness of Spatial Analysis with Remote Sensing and GIS in Sub-Sahara Africa Yuji Murayama  Modeling Spatial Processes of Urban Growth in an African City: A Case Study of Nairobi Charles N. Mundia, Masamu Aniya and Yuji Murayama  Markov-Cellular Automata Approach for Modelling Land Use/Cover Changes in an African Rural Landscape: A Case Study in the Bindura District, Zimbabwe Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro  Urban Land Use Change and Landscape Fragmentation in Lagos, Nigeria Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

xiii  1 

5 

23 

47 

vi Chapter 5

Chapter 6

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Chapter 7

Contents A Multistage-Artificial Neural Network and Expert System (MANNES) Approach for Classifying Urban Built-Up Areas Based on ALOS Data: The Case of Harare, Zimbabwe Courage Kamusoko and Munyaradzi Manjoro  Spatial-Temporal Patterns and Driving Forces of Land Use/Cover Changes in an African Wildlife Sanctuary Charles N. Mundia and Yuji Murayama  Exploratory Woodland Fragmentation Analysis Based on Mathematical Morphology and Landscape Metrics Courage Kamusoko and Enos Chikati 

71 

91 

113 

Chapter 8

Exploratory Land Use/Cover Change Analysis in a Municipality in Kenya Using Markov Chain Model 129  Charles N. Mundia, Kenneth M. Mubea and Moses K. Gachari 

Chapter 9

Open Source Web Mapping Technologies: Can Developing Countries Leverage and Tap the Potential they Portend? David N. Kuria and Douglas E. Musiega 

Chapter 10

Chapter 11

147 

SDI Requirements for Transport Planning: A Case Study of Kampala, Uganda Mazzi Lydia K. Ndandiko and Anthony Gidudu 

175 

Remote Sensing and GIS: Current Status and Future Prospects in Sub-Sahara Africa Charles N. Mundia and Moses K. Gachari 

199 

Index

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

209 

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

FOREWORD The two editors of this monumental book, Dr. Charles Mundia Ndegwa and Dr. Courage Kamusoko, are graduates of the University of Tsukuba, Japan, where they specialized in application of remote sensing and geographic information science (GIS) to address socio-economic issues in Africa for the graduate work. This is probably the second book (after Adeniyi, 1999) written by those who were born and raised in the sub-Saharan region that focused on impending serious problems of socio-economic conditions in the sub-Saharan countries, utilizing remote sensing and GIS techniques. They have used field data that could have been possibly collected by only those who know the regions and countries by heart. In Africa, the use of remote sensing data and analyzing techniques such as GIS is essential for any studies that involve an extensive area, because often than not topographic maps at a scale of 1:50,000 or larger are not available for detailed mapping on the ground. In addition basic socio-economic and physical data, such as census data, environmental data, and infrastructure data are lacking or not kept updated for modeling analyses. Conventional aerial photographs are sometimes available, normally taken by the former colonizing country before independence; however, they are often not updated after the independence. For this reason, the launch of Landsat in 1972 is particularly significant for the African countries, because if Landsat data is available, we can go back to the 1970s as the base year for change analysis as shown by papers in this book. The sub-Saharan countries have been severely affected by recent global warming that has been primarily caused by the developed countries. Ensuing droughts and famine have driven many rural people to urban areas, making the living condition of the cities much worse that was already bad. The socio-

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

viii

Foreword

economic modeling based on such changes in the past, coupled with GIS techniques using information derived from remote sensing is very powerful to provide future scenarios for effective planning to alleviate burden and undesired changes in the urban areas. I am very positive that this book will stimulate further research using remote sensing and GIS with modeling in developing countries and I do hope that policy makers in African countries will recognize the potential of these studies for the planning of sustainable development. November 30, 2009 Masamu Aniya Professor Emeritus University of Tsukuba, Japan

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

REFERENCE Adeniyi, P. O. (ed.). (1999). Geoinformation Technology Applications for Resource and Environmental Management in Africa. Wura-Kay Prints, Nigeria.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

PREFACE This book focuses on impending serious problems of socio-economic conditions in the sub-Saharan countries utilizing remote sensing and GIS techniques. In Africa, the use of remote sensing data and analyzing techniques such as GIS is essential for any studies that involve an extensive area because more often than not, topographic maps at a scale of 1:50,000 or larger are not available for detailed mapping on the ground. In addition, basic socioeconomic and physical data, such as census data, environmental data, and infrastructure data, are lacking or not kept updated for modeling analyses. GIS and remote sensing technologies play a critical role in sustainable natural resource management and land use planning, especially in Sub-Sahara Africa where areas of interest are large and existing geospatial data sets are either outdated or non-existent. More recently, the African Monitoring of the Environment for Sustainable Development (AMESD) initiative financed by the European Union has embarked on a programme to extend the use of remote sensing data for environmental applications to ensure long-term sustainable development in the region. With contributions from both the academia and research institutes, the proposed book will cover some of the most recent advances in GIS and remote sensing, with emphasis on techniques and applications. The range of techniques that will be covered includes artificial neural networks, texture classification, cellular automata and Markov-cellular automata modeling. The range of applications will include: land use/cover mapping and modeling; advances in web mapping; and the development of spatial data infrastructure (SDI) for managing transport infrastructure. “Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa” will be valuable to both researchers and students of GIS and remote sensing. This book will present results and analyses from the “Advanced Land

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

x

Courage Kamusoko, Charles Ndegwa Mundia and Yuji Murayama

Observing Satellite” (ALOS), a remote sensing satellite sensor launched by the Japan Aerospace Exploration Agency (JAXA) in 2006 as well as recent developments and novel applications of existing GIS and remote sensing techniques. Most chapters report results from case studies, while one chapter focuses on a broad overview of the current status and future prospects of remote sensing and GIS in Sub-Sahara Africa.

Courage Kamusoko Asia Air Survey Co., Ltd, 2-2 Manpukuji-1, Asao-ku, Kawasaki, Kanagawa, 215-0004 Japan

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Charles N. Mundia Jomo Kenyatta University of Agriculture and Technology, Department of Geomatic Engineering and Geospatial Information Science, P.O. Box 62000-00200 Nairobi, Kenya Yuji Murayama Division of Spatial Information Science, Graduate School of Life and Environmental Sciences, University of Tsukuba,1-1-1 Tennodai, Tsukuba, Ibaraki 305-8572 Japan.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

ABOUT THE EDITORS

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Courage Kamusoko is researcher at Asia Air Survey Co., Ltd, Japan. He was formerly an assistant professor at the University of Tsukuba, Japan. His research interests include remote sensing and GIS, land use/cover change analysis and modelling, and environmental and urban geography. Charles Ndegwa Mundia is a lecturer in the Department of Geomatic Engineering and Geospatial Information Science at Jomo Kenyatta University of Agriculture and Technology in Kenya. His research interests include remote sensing, urban land use/cover change analysis and modelling, and environmental impact assessment. Yuji Murayama is a professor at the Graduate School of Life and Environmental Sciences at the University of Tsukuba, Japan. His research interests include GIS, spatial analysis, and urban /transport geography.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved. Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

ABOUT THE CONTRIBUTORS Aniya, Masamu, Graduate School of Life and Environmental Sciences, University of Tsukuba,1-1-1 Tennodai, Tsukuba, Ibaraki 305-8572 Japan. Braimoh, Ademola K., Hokkaido University, Japan.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Chikati, Enos, University of South Africa (UNISA), South Africa. Gachari, Moses K., Department of Geomatic Engineering and Geospatial Information Science, Jomo Kenyatta University of Agriculture and Technology, Nairobi, Kenya. Gidudu, Anthony, Department of Surveying, Makerere University, Uganda. Kamusoko, Courage, Asia Air Survey Co., Ltd, 2-2 Manpukuji-1, Asao-ku, Kawasaki, Kanagawa, 215-0004 Japan. Kayondo, Mazzi L., Department of Surveying, Makerere University, Uganda. Kuria, David N., Department of Geomatic Engineering and Geospatial Information Science, Jomo Kenyatta University of Agriculture and Technology, Nairobi, Kenya Munyaradzi Manjoro, Geosciences Department, Nelson Mandela Metropolitan University, South Africa. Marcotullio, Peter J., United Nations University Institute of Advanced Studies, International Organizations Center, Pacifico Yokohama, 2208502, Japan.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

xiv Courage Kamusoko, Charles Ndegwa Mundia and Yuji Murayama Mubea, Kenneth W., ESRI Eastern Africa, Nairobi, Kenya. P.O. Box.619900300, Nairobi, Kenya. Mundia, Charles N., Department of Geomatic Engineering and Geospatial Information Science, Jomo Kenyatta University of Agriculture and Technology, Nairobi, Kenya. Murayama, Yuji, Division of Spatial Information Science, Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Ibaraki, Japan. Musiega, Douglas E., Department of Geomatic Engineering and Geospatial Information Science, Jomo Kenyatta University of Agriculture and Technology, Nairobi, Kenya.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Onishi, Takashi, Research Center for Advanced Science and Technology, University of Tokyo, Komaba, Japan.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

In: Recent Advances in Remote Sensing … ISBN: 978-1-61761-003-5 Editors: C. Kamusoko, C. Mundia et al. © 2011 Nova Science Publishers, Inc.

Chapter 1

INTRODUCTION: USEFULNESS OF SPATIAL ANALYSIS WITH REMOTE SENSING AND GIS IN SUB-SAHARA AFRICA

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Yuji Murayama 1.1. INTRODUCTION In Sub-Sahara Africa, changes in the land use/cover have been accelerating since the end of the 20th century. This has been caused mainly by various human activities rather than natural impact, including cutting trees, opening of plantations, intensive farming, construction of plants by rapid industrialization, and so forth. Today, with the explosive population growth, rural-to-urban migration is a dominant phenomenon in this region. The living environment of poor people in large cities is being worsened by the rapid urbanization. Though this will definitely bring serious social problems in the near future, systematic academic research is still lacking in the region. To monitor land use/cover changes and simulate future landscape change scenarios in Sub-Sahara Africa, spatial information technologies that can visualize and analyze the changing earth surface are very useful. In particular, Geographical Information Systems (GIS), Remote Sensing (RS), and Global Positioning System (GPS) are effective tools for monitoring the environmental consequences at a regional scale. Thus it is a great merit for us to use these

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

2

Yuji Murayama

advanced tools and techniques with digital spatial data. Furthermore, today, we can access high quality and timely acquired satellite images at a low cost.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

1.2. SUMMARY OF BOOK CHAPTERS In the following chapters, we would like to focus and demonstrate the usefulness of the spatial information technologies in Sub-Sahara Africa through various empirical studies employing RS, GIS, and GPS. This book consists of 11 chapters including this introduction (Chapter 1). The contents of each chapter are summarized as follows. Chapter 2 analyzes the land use/cover changes and constructs an urban growth scenario for Nairobi using dynamic Cellular Automata (CA). The model is calibrated using multistage Monte Carlo simulation to derive a 30 year prediction until 2030. An unsustainable sprawled urban growth in Nairobi is forecasted from the analysis. The study shows an application of CA which is very useful for urban modeling and could be an effective weapon to foresee the spatial consequences of poor planning policies in many African cities. Chapter 3 deals with recent landscape transformation in rural Africa. Taking Zimbabwe as a case study, this chapter simulates future land use/cover changes employing a Markov-cellular Automata (MCA) model. Transition potential maps are generated from biophysical and socioeconomic data through a GIS-based multi-criteria evaluation (MCE) procedure. The simulated land use/cover pattern for 2030 indicates that if the current land use/cover trends continue without holistic sustainable development policies, severe land degradation will occur, with potential threats to rural sustainability. Urgent efforts are required to be initiated to reduce deforestation in Zimbabwe. The simulated future land use/cover maps derived from this analysis can serve as an early warning of the future effects, particularly in other rural areas in Sub-Sahara Africa. Chapter 4 discusses the landscape fragmentation process in Lagos through the spatial analysis of satellite images over time. Logistic regression and landscape metrics techniques are used to find the probability of residential land development. This study shows that the continuation of low density sprawl will increase the cost of infrastructure to link the suburbanized areas to the main city. Furthermore, the authors argue that sprawling in areas with relatively low population leads to an increase in physical fragmentation of non-urban land. It is concluded that, due to the abundance of economic opportunities in Lagos compared to other Nigerian cities, the conversion of

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Introduction

3

agricultural and natural landscapes to residential uses most likely continue at the urban fringes. Improvement of land administration is indispensable to cope with the rapid population growth. Chapter 5 evaluates the usefulness of a multistage-artificial neural network and expert system (MANNES) approach for classifying built-up areas based on the Advanced Land Observing Satellite (ALOS) spectral and textural data. Results reveal that classification accuracy can be improved by minimizing spectral confusion between built-up areas and non-built-up areas such as open vacant plots and agriculture fields. This chapter demonstrates a cost-effective way for visualizing urban land use/cover using medium resolution satellite images in sub-Saharan cities such as Harare, Zimbabwe. Chapter 6 analyzes long-term landscape transition and wildlife population dynamics in Masai Mara ecosystem in southwestern Kenya composed of the National Reserve and the adjoining group ranches. The area holds spectacular concentration of wildlife and is home to the iconic Masai pastoralists and their livestock. In this study, multi-temporal satellite images, together with physical and social economic data are employed in a GIS post-classification analysis to discuss outcomes of different land use practices and policies. Result of this analysis indicates rapid land use/cover conversions and a drastic decline for a wide range of wildlife species. In addition, substantial losses in forest cover and habitat fragmentation are observed. Land use policy, agricultural expansion, and mushrooming tourism activities are the major driving forces of land use changes. It is extremely urgent that all necessary measures are taken in order to maintain a balance between wildlife conservation and economic development in the ecosystem region. Chapter 7 explores the effectiveness of mathematical morphology and landscape metrics, for assessing woodland fragmentation in order to provide detailed information on the status and trends of the health of woodlands and biodiversity. Remote sensing and GIS analyses revealed that woodland areas declined, while non-woodland increased between 1973 and 2000. A dominant transformation from woodland to non-woodland indicates a high rate of deforestation. Furthermore, the woodland areas become more fragmented as reflected in the decrease of core woodland areas and the edge density on one hand, and an increase in woodland patches on the other. Chapter 8 examines land use/cover changes classified from multispectral Landsat images for 1973, 1988, and 2000 in a Kenyan municipality. A Markov chain model is employed to simulate land use/cover changes for Nakuru Municipality up to 2015. Results indicate that there has been a notable uneven urban growth and substantial loss in forest land showing an unstable changing

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

4

Yuji Murayama

process. The projected land use/cover for 2015 reveals substantial increase in urban and agricultural land uses. Chapter 9 focuses on the availability of the open source web mapping system in developing countries. The websites developed from this work demonstrate that web mapping technologies can be effectively used to serve GIS content and functionality to personnel in various departments possessing different GIS skills using intuitive interfaces. The authors argue that open source web mapping technologies are useful to open their domestic economies to the wider international community through web mapping applications targeting tourism, agriculture, natural resource exploitation potential, environment protections, etc. Taking Kampala as a case study, Chapter 10 discusses the importance of Spatial Data Infrastructures (SDI) in developing countries. The authors investigate the nature of available spatial data in Kampala with guidelines to meet SDI requirements applied to transport planning and management. The study demonstrates that government needs to overcome a number of difficult problems including duplication of efforts in data collection, political interference in decision making, unskilled staff in geospatial data management, absence of formalized polices for data exchange, limited resources for metadata compilation, and lack of institutional collaboration. It is emphasized that construction of SDI is indispensable for regional and urban planning in Sub-Sahara Africa. Finally, chapter 11 discusses the usefulness of spatial information technologies for sustainable development in Sub-Sahara Africa. The authors state that with the expanding power supply and increasing Internet connectivity, spatial information technologies are moving forward from basic data capture to more of data processing, given the increased recognition of the role of spatial analysis for improving socio-economic development in SubSahara Africa. Moreover, remote sensing and GIS data as well as the advances in geospatial techniques have the capacity to provide timely and accurate information on natural and human-dominated landscapes critical for supporting environmental protection and ecological planning policies in SubSahara Africa.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

In: Recent Advances in Remote Sensing … ISBN: 978-1-61761-003-5 Editors: C. Kamusoko, C. Mundia et al. © 2011 Nova Science Publishers, Inc.

Chapter 2

MODELING SPATIAL PROCESSES OF URBAN GROWTH IN AN AFRICAN CITY: A CASE STUDY OF NAIROBI

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Charles N. Mundia, Masamu Aniya and Yuji Murayama ABSTRACT The objective of this chapter was to study the dynamics of land use/cover changes and simulate future urban expansion, in order to address the need for urban management tools that can guide sustainable urban planning policies. Cellular automata that integrates biophysical factors with dynamic spatial modeling, was used for this study. The model was calibrated and tested using time series of urbanized areas derived from land use/cover maps, produced from multi-spectral satellite imageries, and future urban growth projected out to 2030. The model accuracy assessment results showed high accuracies, indicating that the simulated results were realistic and accurate, thereby confirming the effectiveness of the model. Results show that the model is useful for urban modeling and an effective tool to foresee the spatial consequences of poor planning policies in the context of many cities in Sub-Sahara Africa. The forecast for Nairobi shows an unsustainable sprawled urban growth. The results show that urban simulations can represent a useful approach to an understanding of the consequences of current planning policies or their incompleteness.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

6

Charles N. Mundia, Masamu Aniya and Yuji Murayama

Keywords: Land use/cover changes, urban growth, cellular automata model, African cities, Nairobi

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

2.1. INTRODUCTION Most of the world’s populations live in urban areas where the main socioeconomic and environmental processes that affect human societies take place. The rapid growth in urban population entails rapid growth in size and number of urban places. In major cities in Africa, urban population is increasing at a much faster rate than in the rest of the world. The World Bank report shows that by 2020 Africa will have 11 mega-cities and almost 3,000 cities with populations of more than 20,000, an increase of almost 300 per cent from 1990. Understanding urban growth and changes in these rapidly changing environments is critical to city planners and resource managers (Knox, 1993, Turner et al., 1993). The estimation of future impacts of existing spatial plans and policies on land-use development, and the consideration of alternative planning and policy scenarios for impact minimization, are of particular interest. The consequences of inaccurate planning in major cities of developing countries are also of interest to other stakeholders, such as those involved in research studies and policy-making processes related to sustainable development (Cohen, 2004). Nairobi city serves simultaneously as the national and regional hub for economic activities. The city is expanding rapidly with its population currently standing at 3.5 m spilling into the adjoining towns. The environmental and social consequences of a growing population in a loosely planned urban system could be dramatic, mainly when urban areas experience tremendous growth in a short period of time (Cohen, 2006). The rapid urbanization process if unchecked will continue in the years and decades to come with serious consequences (Torrey, 1998). For example, Nairobi city is now already faced with serious urban management problems, and thousands of its residents live in informal settlements (Mundia and Aniya, 2007). A growing number of urban residents are finding shelter in sub-standard housing in informal settlements with severe sanitation problems. Deterioration of the conditions is also manifested in decaying infrastructure, poor management of waste, lack of proper sanitation facilities, poor drainage systems among other problems. Remote sensing techniques have shown their value in mapping urban dynamics and as data sources for the analysis and modeling of urban growth and land use change (Clarke et al., 2002). Remote Sensing provides spatially

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Modeling Spatial Processes of Urban Growth in an African City

7

consistent data sets that cover large areas with both high spatial detail and high temporal frequency. These kinds of data sets are necessary for land use/cover analysis, which is an essential element of ecological studies. As urbanization occurs, changes in land use/cover accelerate and land making up the natural resource base such as forests and agricultural land are replaced, leading to fragmentation and land degradation (Mundia and Aniya, 2005). The study of land use/cover changes is essential not only for land use management but also in detecting environmental change and in formulating sustainable development strategies (Barnsley and Barr, 1997). Accurate information on land use changes is needed for documenting growth, making policy decisions and improving land-use planning (Jacobson 2001). Information concerning land use changes is also required for predictive modeling (Epstein et al., 2002). Model building using Cellular Automata (CA) has gained attention for its utility in predicting spatial patterns of urban development and in the investigations on planning regimes and land use patterns (Silva and Clarke, 2002). CA belongs to the family of discrete, connectionist techniques that are used to investigate fundamental principles of dynamics, evolution, and selforganization (White and Engelen, 1993). Essentially, a cellular automaton model is composed of a finite set of grid cells, the current state of each cell, a set of transition rules for the cells, and the neighbourhood of the cell. In a strict cellular automaton the rules must be uniform and must apply to every cell, state, and neighbourhood. Every change in state must be local implying that there is no action at a distance (Batty and Xie, 1997). There are many renditions of CA, but the current ones that have applicability to urban systems follow Conway’s logic (Batty and Xie, 1994). Previous studies (Wagner, 1997; Oreskes et al., 1994; Openshaw, 1979; and Wu, 1989), which have used traditional modeling approaches based solely on demographic trends, have failed to account for contemporary urban growth. These studies have not taken into account urban growth through time, recent urban evolutions and collective social preferences for different styles of living (Couclelis, 1997). These elements have a key impact on how urban land uses are organised. Urban modeling approaches such as CA, which are sensitive to diversity in these factors, can be more successful when simulating various shapes and intensities of urban growth (Silva and Clarke, 2002; Batty and Xie, 1994; and Pikajankowski et al., 1997). Cellular Automata (CA) is useful for capturing and simulating the complexity inherent in dynamic systems such as major urban areas. The power of Cellular Automata comes from the ease with which simple preconditions,

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

8

Charles N. Mundia, Masamu Aniya and Yuji Murayama

distributions, rules, and actions can lead to extra-ordinary complexity (Batty and Howes, 2001). The model described in this chapter takes care of the shortcomings of the traditional modeling approaches and incorporates complexity theory to capture continuous change in urban environment. As a planning tool, this CA urban modeling is interactive and can be visualized and quantified to play an important role especially in investigating the consequences of the African urban settlements to avoid environmental and social consequences as a result of the rapid spatial expansion. This chapter focuses on land use/cover changes and urbanization in an African city. Unlike European or American cities, African cities are unique in many aspects. Some of the serious issues in these cities include the increasing numbers of immigrants from rural areas. Because majority of these immigrants are poor, urban expansion is taking place against a background of stagnant living conditions. This often leads to mushrooming of unplanned, chaotic squatter settlements often of high densities. Most of these cities, show characteristic patterns of urban sprawl where urban development evolve around nexus of the main transportation routes, with urban growth tending to grow in sectors emanating from city centres. The objective of this chapter is therefore to analyze the land use/cover changes and produce an urban growth model for the city of Nairobi using CA dynamic spatial model. The model was calibrated using multistage Monte Carlo method and a 30 year prediction simulation was run until 2030. Our model predicts future land use development under existing policies and examines other scenarios under different policies in order to assess their effect on future land use development.

2.2. DATA AND METHODOLOGY Nairobi is one of the fastest urbanizing cities in sub-Saharan Africa (Obudho, 1997), and was found typical and ideal for applying the Clarke urban growth model’s flexibility to adapt and evolve over time depending on the changing characteristics of the city. Nairobi is the capital city of Kenya and the urban pattern of the city and its environs are characterized by intense urban pressures, especially along the main highways, and through the development of sub urban areas. The population densities in Nairobi vary widely throughout the city ranging from 500 people per square kilometer in high income areas to 75 000 people per square kilometer in slum areas. Urbanization of Nairobi city is as a result of a number of different factors but the principal reason is the

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Modeling Spatial Processes of Urban Growth in an African City

9

explosive population increase as a result of rural-urban migration. The population increase is causing geographic expansion of urban areas through annexations, transformations and reclassification of neighboring rural areas into urban settlements. Existing urban growth models focus mainly on cities in the industrialized world where socio-economic and other data are readily and widely available. However, it is in the developing cities where most of the urban growth will occur in the next decades (United Nations, 2004). Problems regarding data availability and accuracy in cities in developing countries make their studies difficult, making it necessary to develop models applicable to such cities. In many cases, socio-economic data in these cities can be unreliable or nonexistent making urbanization forecasts difficult. Recognizing these problems, our model utilizes data that are widely available and routinely collected (e.g., satellite data). The Clarke Cellular Automata Urban Growth Model (Clarke and Gaydos, 1998) was modified and calibrated to produce urban growth simulations for the city of Nairobi. The framework adopted for the land use/cover change analysis and urban expansion modeling comprised GIS and CA modules. The GIS module allowed GIS analyses to determine suitability factors, model constraints and land use/cover change statistics while the CA module was useful for model calibration and for applying transition change rules. The urban growth simulation used CA, terrain mapping and land use/cover modeling to address urban growth. Modeling of Nairobi city utilized a number of input data layers: “slope”, “land use/cover”, “areas excluded from development”, “urban areas”, “road network” and “hillshade”. Table 1 summarizes the characteristics of satellite data used for land use/cover change analysis, while Table 2 outlines the sources, descriptions and resolutions of the data used for modeling. Various types of urban growths were simulated. These included: spontaneous growth, new spreading centres growth, edge growth, and road-influenced growth. The growth types were applied sequentially during each growth year and were controlled through the interactions of growth parameters which describe an individual growth characteristic and when combined with other characteristics describe several different growth processes. Multi-temporal Landsat images for 1976, 1988, and 2000 (Table 1) were used in a post classification analysis with GIS to map land use/cover changes. Original data sets were at 30m resolution for TM and ETM, and 60m for the MSS imagery. These were resampled to a common resolution of 100m while maintaining the spatial extent of the study area. The urban extents for the various years were then extracted from the land use/cover maps.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Charles N. Mundia, Masamu Aniya and Yuji Murayama

10

Table 1. Characteristics of satellite data used for modeling urban growth in Nairobi city Data type Landsat MSS Landsat TM Landsat TM+ Landast TM+

Satellite Landsat 2 Landsat 4 Landsat 7 Landsat 7

Resolution 79/120 30/120 15/30/60 15/30/60

Acquisition data 11 February 1976 17 October 1988 30 January 1995 21 February 2000

Table 2. Sources, description and resolution of data used for modeling urban growth Data Layer

Source

Description

Resolution

Urban Extent

Land use map Road map

Land use/cover map for 1976, 88, 95, 2000 Classified roads for 1976, 1988 Derived from TIM

30m

Roads

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Slope Exclusion Hillshade Population GDP

1:50,000 topo map 1:50,000 topo map 1:50,000 topo maps Population census Economic survey

Vector coverage of protected areas Derived from contours Population census for 1979, 89, 2000 Household surveys for 1979, 89, 99

N/A 30m N/A 30m N/A N/A

Two time periods for transportation were prepared for 1976 and 1988 from topographical maps. Slope layer and the layer of all “areas excluded from development” were generated from topographical maps. In addition to ruralurban migration which is an important factor, these natural environment factors were considered because land use, topography and accessibility are important factors especially in the context of African cities, as compared to urban expansion in industrialized world where urbanization is influenced mainly by social, economic and demographic forces. All input files were rasterized at a 100m resolution to the spatial extent of the study area. Using

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Modeling Spatial Processes of Urban Growth in an African City

11

these data, model calibration as described below, was carried out to derive parameters for simulating urban growth.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

2.3. CALIBRATING CELLULAR AUTOMATA MODEL Model calibration was achieved through a brute force Monte Carlo calibration method. This method determines, given a starting image of urban extent, a set of initial control parameters that lead to a model run that best fits the observed known data. The method steps through the coefficient space in a complete, regular and irreducible manner (White et al., 1997). By running the model, a set of control parameters were refined in the sequential calibration phases (coarse, fine, and final calibrations). Between the calibration phases, attempts were made to extract the values that best matched the five coefficient factors that controlled the behavior of Nairobi city: diffusion (overall scatter of growth), breed (likelihood of new settlements being generated), spread (growth outward and inward from existing spreading centers), slope resistance (flat more preferred), and road gravity (attraction of urbanization to roads and diffusion of urbanization along roads). Coefficients combinations resulted in a number of metric measures. These metric measures were coefficients of determination of fit between actual and predicted values either for the pattern (such as number of pixels, number of edges, number of clusters), for spatial metrics such as shape measures, or for specific targets, such as the correspondence of land use and closeness to the final urban pixel count. The calibration using Monte Carlo simulations computed the averages across multiple runs to ensure robustness of the solutions. This made it possible to adapt the model to existing characteristics for Nairobi city throughout the various stages of calibration by using different spatial resolutions and the sequential multistage optimization of the coefficient that controlled the system. To examine the role of spatial resolution on model calibration and outputs, calibrations were performed at different fixed resolutions. The quarter calibration was performed using only the quarter resolution data (216 x 186). The other two calibrations were the half (432 x 372) and the full (864 x 743) calibrations. By narrowing both the spatial scale and the range of parameters in the three calibration sequences, it was possible to close in on the parameter set that best simulates the urban growth for Nairobi city. These parameters were then used to determine the coefficient values that best allow the model to predict the future urban growth of Nairobi city. Results from the three phases

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

12

Charles N. Mundia, Masamu Aniya and Yuji Murayama

of calibration (Coarse, Fine, and Final calibrations) indicated successive improvement in the parameters that control the behaviour of the urban expansion. After the coarse calibration, the resulting calibration values narrowed and became more sensitive to the local conditions within the city. The comparison of the modelled final “population” (number of urban pixels) and the urbanization of the control years give a high summary correlation of 0.95 and a compare statistic of 0.78 making it reasonable to say that the prediction of the model based on the initial seed year of the present urban pattern was quite accurate. These correlation values also suggest that the calibration phases adopted for Nairobi allowed the model to simulate urban growth with a high degree of fit. The shape and form of urbanization seem also to confirm that calibration adjusted the values to reflect the actual characteristics of Nairobi city. Table 3 gives a summary of the resulting parameters and the best ten coefficient sets from the final calibration. The calibration results for Nairobi city indicate that the Spread coefficient is the highest followed by the Road Gravity coefficient. The Slope coefficient was ranked fourth suggesting that slope has minimal influence on the urbanization process of Nairobi city. The resulting coefficients suggest that the urbanization of Nairobi city has tended to occur from the main nucleus (spread coefficient at 98) and along the main roads network (road gravity coefficient at 75) with a little influence of the local terrain (slope coefficient at 5). The calibration results for Nairobi city indicate that the Spread coefficient is the highest followed by the Road Gravity coefficient. The Slope coefficient was ranked fourth suggesting that slope has minimal influence on the urbanization process of Nairobi city. Table 3. Parameters and coefficient sets from final model calibration A. Initial Parameters for the Final Calibration

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011. ProQuest Ebook

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

B. Coefficients Sets from Final Calibration Model runs 6525 6527 6535

LeeSallee

Compare

Clusters

Pop

Edges

Diffusion

Breed

Spread

Slope

0.09026 0.09026 0.09026

0.1101 0.1101 0.1101

0.3246 0.3246 0.3246

0.9740 0.9740 0.9740

0.5680 0.5680 0.5680

5 5 5

20 20 20

92 92 92

4 4 4

Road gravity 40 45 40

6544 6553 6536 6562 6345 226 235

0.09026 0.09026 0.09026 0.09026 0.09026 0.09026 09026

0.1101 0.1101 0.1101 0.1101 0.1101 0.1101 0.1101

0.3246 0.3246 0.3246 0.3246 0.3246 0.3246 0.3246

0.9740 0.9740 0.9740 0.9740 0.9740 0.9740 0.9740

0.5680 0.5680 0.5680 0.5680 0.5680 0.5680 0.5680

5 5 5 5 5 5 5

20 20 20 20 20 20 20

92 92 92 2 92 92 92

3 4 1 5 4 4 5

40 40 45 40 45 35 40

14

Charles N. Mundia, Masamu Aniya and Yuji Murayama

2.4. ACCURACY ASSESSMENT

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

The goal of model calibration as outlined was to derive a set of values for the growth patterns that could effectively simulate growth during the time period, 1976-2000. To be certain that the calibration coefficients obtained were accurate, the simulated growths for 1995 and 2000 were compared with the actual growth obtained through satellite images by several least squares regression statistics. Figure 1 and Table 4 summarizes the spatial accuracy assessment for 1995 and 2000. The overall spatial accuracy was high at 80% for 1995 and 86% for 2000. Errors of omission (producer’s accuracy) and commission (user’s accuracy) for the urban class for both years suggest that the prediction of the location of urbanized pixels was reasonably accurate. After confirming the accuracy of our calibration, the set of coefficients derived during calibration were used to predict future patterns of urbanization.

Figure 1. Accuracy assessment by comparing actual situation (satellite mapping) on the left hand side and with simulated results on the right hand side.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Modeling Spatial Processes of Urban Growth in an African City

15

Our model used the following inputs: urban extent for initialization, an initial transportation network with provision for incorporation of future networks, an excluded layer, and a slope and hillshade layer. This allowed us to model current trends scenario that reflects the policies that are currently in place. Our model made the following assumptions: (1) that there would be continuation of economic development at the rate of 4.3%; (2) that the population increase would continue growing at the rate of 4.2% per annum; (3) topographical slope beyond 21% would inhibit urban growth; (4) existing urbanization would encourage peripheral growth, and (5) road network would be a primary correlation to urban growth.

2.5. RESULTS AND DISCUSSION

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

The post-classification comparison approach was employed for detection of land use/cover changes, by comparing independently-produced classified land use/cover maps. This method provided descriptive information on the nature of changes that occurred. The classified land use/cover maps for Nairobi in 1976, 1988 and 2000 are shown in Figure 2.

Figure 2. Land use/cover maps of Nairobi for 1976, 1988 and 2000.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

16

Charles N. Mundia, Masamu Aniya and Yuji Murayama

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Table 4. Spatial Accuracy Assessment

Table 5. Statistics derived for the land use/cover changes in Nairobi city Year Land use/cover Urban/Built up Areas Agricultural Areas Decidous Forests Bushland Mixed Rangeland Shrub/Brush Range Open Transitional Areas Land use/cover Water Total

1976

1988

Area (km2) 13.99

% 1.9

5.8

Area (km2) 61.25

49.83

6.9

57.83

8.1

87.78

12.3

100.15 154.48 357.32 25.22 6.92

14.0 22.3 50.1 3.5 0.9

29.09 101.49 340.62 64.19 77.96

4.1 14.2 47.7 8.9 10.9

23.56 95.98 237.63 170.78 32.72

3.3 13.5 33.3 23.9 4.6

Area (km2) 0.50 713.41

%

Area (km2) 1.09 713.44

%

Area (km2) 3.77 713.45

%

0.1 100.0

%

2000 %

Area (km2) 41.18

0.2 100.0

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

8.6

0.5 100.0

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Modeling Spatial Processes of Urban Growth in an African City

17

The spatial distributions of each of the classes were extracted from each of the land use/cover maps by means of GIS functions. The statistics derived from land use/cover change analysis are summarized in Table 5. The urban/built-up areas increased from 14 km2 in 1976 to 61 km2 in 2000. Agricultural fields occupied 49 km2 in 1976 and have increased substantially to 88 km2 in 2000. Forested lands have, however, decreased substantially from 100 km2 in 1976 to a mere 23 km2 in 2000, a record loss of 77 km2. The rangelands, consisting of mixed rangeland and shrub/brush rangeland have decreased from 357 km2 in 1976 to 237 km2 in 2000. The rangelands have given way mainly to the expanding agriculture and urban sprawl. The results from the model prediction output are illustrated in Figure 3 and indicate a significant amount of growth from 2000 to 2030 including the occupation of large tracts of non urbanized areas. The city is predicted to urbanize substantially by 2030. The predicted results show dramatic growth, with built up areas dominating the urban landscape. Our results also showed that by translating urban planning policies into exclusion probabilities, it would be possible to obtain different scenarios of future urban growth based on a variety of assumptions. Figure 4 shows such a scenario where managed growth reflecting a more stringent protection policy and targeted toward limited growth and natural resource protection is emphasized.

Figure 3. Simulated urban growth for Nairobi city.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

18

Charles N. Mundia, Masamu Aniya and Yuji Murayama

Figure 4. Simulated managed growth scenario with focused resource protection.

This scenario would ensure environmental conservation and protection of the national park and existing forests. The results of the scenario predictions show higher dispersed development patterns for the current trends as compared to the managed growth scenario.

2.6. SUMMARY AND CONCLUSIONS Major cities in Africa are experiencing rapid urban expansion leading to sprawl and undesirable environmental and socio consequences. We used Nairobi city to analyse the dynamics of land use/cover changes between 1976 and 2000 and to produce an urban growth model using Cellular Automata spatial modeling. To achieve this, we utilised multi-spectral Landsat images for 1976, 1988, 1995, and 2000 to map land use/cover changes. The information and the findings obtained from the land use/cover analysis were

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Modeling Spatial Processes of Urban Growth in an African City

19

employed for modeling the urban growth of Nairobi city. The urban growth model integrated some biophysical factors with Cellular Automata spatial modeling. The results show fast spatial expansion with simulated urban land taking up most of the available land within the city and the surrounding areas. These spatial results obtained from the simulations for 2030 raises some pertinent questions and indicate that Nairobi city is expanding without strategic planning. The consequences for the quality of life of the people of Nairobi might be severe if serious measures are not adopted and applied in time. There is a very urgent need to take urban development and planning measures to manage Nairobi city as anticipated in 2030, taking into consideration the existing demographics, economic and social constraints. Dynamic modeling by Cellular Automata has proven to be an effective tool for exploring different regional futures for the purposes of developing regional approaches to land use management and to foresee the spatial consequences of loose planning policies in the context of African cities. Nairobi, a rapidly urbanizing city in Africa has experienced rapid growth in terms of population and spatial extent. The sprawling urban growth has led to the removal of natural vegetation and is threatening the environment and the natural resources. If the current trends of growth continue and the predictions hold true, future urban growth will mainly take the pattern of urban sprawl. This has several significant economic, environmental, and social implications for policy-making and urban planning. Economically, there will be an increase of pressure on urban infrastructure. The wider spreading residential, commercial, industrial, and service centres will require more intra-urban roads, bridges, water and sewer lines as well as other civil services such as fire stations, schools and hospitals. Balancing this need for urban growth with the efficient use of resources is a daunting issue for policy makers and planners especially given the limited resources in African cities. Environmentally, this process inevitably involves altering or destroying natural environments, building barriers to natural processes, and altering natural geo-chemical cycles through pollutant disposal, as well as many other problems. Preliminary analyses of the simulated urban area suggest the continued decline of forest land, cropland, and further landscape fragmentation at least up to 2030. Even if policies and regulations are implemented to protect some areas from development, they can not guarantee that these protected areas will not be polluted and degraded.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

20

Charles N. Mundia, Masamu Aniya and Yuji Murayama

Proactive and preventive environmental considerations will be a crucial component for effective land use management. Forecast tools are needed in order to assess the dangerous future consequences of current urban growth trends and to help in understanding their dynamics and in controlling their growth. A realistic modeling system can be used for developing regional approaches to land use management. Because of the ability to simulate complex behaviour of urban systems, cellular automata represent a viable approach for regional modeling. This research has explored the suitability of utilizing CA for regional planning applications. The urban growth model was found to be useful for demands for regional modeling and an effective tool to foresee the spatial consequences of planning policies. Alternative scenario development and the ability to visualize and analyze the model outcome spatially can be very useful for planning purposes, especially for coming up with different planning strategies. This research, in this sense, has provided a model that can provide quantitative, visual, spatial, and temporal information for policy makers, planners, environmentalists, and developers. Some assumptions about growth processes prevent the model from being able to capture a wider range of growth patterns and processes. This model assumed stable political and economic conditions up to 2030. The continuation of the population growth rate at the current rate 4.2 % per annum and economic growth rate at 4.3 % were assumed. Such assumptions may not be realistic in the long term because the economy and population are bound to change. In addition, a more detailed thematic categorization of the urban environment, which would possibly be more suitable for simulating urban growth, was not included. Despite these assumptions and considerations, the model was found to be a useful tool for assessing the impact of alternative scenarios and succeeded in simulating urban growth from 2000 to 2030. The model can be used to study different planning strategies, to measure the spatial consequences of policy decision and, more importantly, to study future land use dynamics. The capacity of the model to reproduce the actual urban shape through a bottom-up approach is remarkable and will provide planners with more powerful tools for urban and regional scenarios generation. The results obtained encourage further improvement of the model by incorporating other variables such as human behaviours, tax, income, and environmental aspects. Such an integrated model can be a more powerful tool for urban and regional scenarios generation.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Modeling Spatial Processes of Urban Growth in an African City

21

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

REFERENCES Barnsley, M.J., and Barr, S.L (1997). A graph based structural pattern recognition system to infer urban land use from fine spatial resolution land cover data. Computer, Environment and Urban systems, 21(4), 209-225. Battty, M., and Xie, Y. (1994). Modeling inside GIS Part 1: Model structures, exploratory spatial data analysis, and integration. International Journal of Geographic Information Systems, 8, 291-307. Battty, M., and Xie, Y. (1997). Possible urban automata. Environment and Planning B, 24, 175-192. Batty, M., and Howes, D. (2001). Predicting temporal patterns in urban development from remote imagery. In J.P. Donny, M. J. Barnsley, and P.A Longley (Eds.), Remote Sensing and Urban analysis (pp. 185-204). London: Taylor and Francis. Clarke, K. C., and Gaydos, L. (1998). Loose coupling a cellular automata model and GIS: long-term growth prediction for San Francisco and Washington/Baltimore. International Journal of Geographical Information Science 12: 699-714. Clarke, K.C., Parks, B.O., and Crane, M.P. (2002). Geographic information systems and environmental modeling. New Jersey: Prentice Hall. Cohen, B. (2006). Urban growth in Developing countries: Current trends, future projections, and key challenges for sustainability. Technology in Society 28: 63-80. Cohen, B. (2004). Urban growth in Developing countries: A review of current trends and a caution regarding existing forecasts. World Development 32: 23-51. Couclelis, H. (1997). From cellular automata to urban models: new principles for model development and implementation. Environment and Planning B 24: 165-174. Epstein, J., Payne, K., and Kramer, E. (2002(. Techniques of mapping suburban sprawl. Photogrammetric Engineering and Remote Sensing, 59, 991-996. Jacobson, L. (2001). Lawsuit accuses small business administration of promoting sprawl. Planning, 67(1), 28-47. Knox, P. L. (1993). The restless urban landscape. Englewood Cliffs, NJ: Prentice-Hall. Mundia, C.N., and Aniya, M. (2007). Modeling and predicting urban growth of Nairobi city using cellular automata with Geographical Information Systems. Geographical Review of Japan 12: 777-788.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

22

Charles N. Mundia, Masamu Aniya and Yuji Murayama

Mundia, C.N., and Aniya, M. (2005). Analysis of land use changes and urban expansion of Nairobi city using remote sensing and GIS. International Journal of Remote Sensing 26:2831-2849. Obudho, R.A. (1997). National capital and regional hub. In The urban challenge in Africa: Growth and management of its large cities, ed. C. Rakodi, 292-334. Tokyo: United Nations University press. Oreskes, N., Shrader-Frechette, K., and Belitz, K. (1994). Verification, validation, and confirmation of numerical models in the earth sciences. Sciences 263, 641-644. Openshaw, S. (1979). A methodology for using models for planning purposes. Environment and Planning A, 11:879-896. Pijankowski, B. C., Long, D. T., Gage, S. H., and Cooper, W.E. (1997). A Land Command syntac and an application for Michigan’s Saginaw Bay watershed, http://www.Ncgia.ucsb.edu/conf/landuse97/ accessed October 2004. Silva, E.A., and Clarke, K.C. (2002). Calibration of the SLEUTH urban growth model for Lisbon and Porto, Spain. Computers, Environment and Urban Systems 26: 525-552. Torrey, B. (1998). We need more research on the impact of rapid urban growth. The Chronicle of Higher Education 45: B6. Turner, B.L., Moss, R. H., Skole, D. L. (1993). Relating land use and global land cover change: A proposal for an IGBP-HDP core project. International-Geosphere-Biosphere program, IGBP report No. 24, HDP report No: 5. Stockholm: Royal Swedish Academy of Sciences United Nations. (2004). World urbanization prospects, 2003. New York: United Nations. Wagner, D.F. (1997). Cellular automata and geographic information systems. Environment and Planning B,24, 219-234. White, R., and Engelen, G. (1993). Cellular automata and fractal urban form: a cellular modeling approach to the evolution of urban land use patterns. White R., Engelen, G., and Uljee, I. (1997). The use of constrained cellular automata for high resolution modeling of urban land use dynamics. Environment and Planning B 24: 323-343. Wu, F. (1989). Simulating urban encroachment on rural land with fuzzy-logic controlled cellular automata in a GIS. Journal of Environment Management 53: 293-308.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

In: Recent Advances in Remote Sensing … ISBN: 978-1-61761-003-5 Editors: C. Kamusoko, C. Mundia et al. © 2011 Nova Science Publishers, Inc.

Chapter 3

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

MARKOV-CELLULAR AUTOMATA APPROACH FOR MODELLING LAND USE/COVER CHANGES IN AN AFRICAN RURAL LANDSCAPE: A CASE STUDY IN THE BINDURA DISTRICT, ZIMBABWE Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro ABSTRACT Land use/cover changes are continuously transforming key aspects of the Earth’s ecological systems, which support human needs. Therefore, an understanding of land use/cover changes is fundamental for sustainable agriculture and forestry management, particularly in rural areas in sub-Sahara Africa that are threatened by land degradation. Taking Zimbabwe as a case study, future land use/cover changes were simulated based on a Markov-cellular automata (MCA) model that integrates the Markovian chain and cellular automata (CA) models. Transition potential maps were generated from biophysical and socioeconomic data using a GIS-based multicriteria evaluation (MCE) procedure. Dynamic adjustments of transition probabilities and transition potential map thresholds were implemented in the MCA model through a multi-objective land allocation (MOLA) procedure. Based on the 2000

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

24

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro calibration scenario, the MCA model projected a continuing downward trend in woodland areas and an upward trend in bareland areas in 2030.

Keywords: Land use/cover changes; Markov chain; cellular automata (CA); multicriteria evaluation (MCE); multi-objective land allocation (MOLA); Markov-cellular automata (MCA); Zimbabwe

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

3.1. INTRODUCTION Recently, land use/cover changes have attracted increasing attention from decision-makers, as rapid population growth is continuously exerting more pressure on ecological systems, particularly in developing countries (Etter et al., 2006). In sub-Sahara Africa, two-thirds of the population living in the rural areas depends on agriculture and other natural resources such as timber and firewood for their economic and social needs (Gambiza et al., 2000; Campbell et al., 2000). Although the region is endowed with a wide range of natural resources, poverty and land degradation are threatening the sustainability of agriculture and forestry with adverse effects on the livelihoods of the rural citizens (Pinstrup-Andersen et al., 1997; Abalu and Hassan 1998; Rockstrom et al., 2004; Kamusoko and Aniya, 2007). According to FAO (2003), southern Africa accounted for the highest rate of deforestation in sub-Sahara Africa during the 1990s. Forest areas decreased from 199.4 million hectares to 183.1 million hectares, while 40-60% of the soils were either eroded or degraded (UNEP 1991; FAO, 2003). In order to attain sustainable development, it is necessary to prioritize environmental research, particularly the assessment and analysis of land use/cover changes. Furthermore, simulated land use/cover changes are needed in order to evaluate likely or alternative future landscape change scenarios (Torrens, 2006). This requires robust and spatially explicit land use/cover change models, which can capture both biophysical and socioeconomic factors (Fang et al., 2005). Such an approach falls in line with the community-based sustainable development approaches, fundamental in ameliorating rural livelihoods. The purpose of this chapter is to simulate future land use/cover changes in the Masembura and Musana communal areas of the Bindura district, Zimbabwe based on the Markov-cellular automata (MCA) model that combines Markov chain analysis and cellular automata (CA) models. The

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

25

chapter is organized into five sections. The first section focuses briefly on the basics of the Markov chain analysis, CA and MCA models. The study area is briefly described in the second section, while the third section focuses on the data used and the MCA framework. The fourth section presents the MCA modelling framework results, and discusses the strengths and limitations of the model. Finally, the summary and conclusions, as well as the research contributions and recommendations are highlighted.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

3.2. BACKGROUND Land use/cover change models are reproducible tools that are important for analysing and simulating land use/cover changes in order to make informed decisions (Costanza and Ruth, 1998; Verburg et al., 2004). The rationale for developing land use/cover models is to: (1) understand the driving forces and dynamics of land use/cover changes; (2) understand the future economic and environmental implications of the current land use/cover changes; and (3) serve as a means of projecting the impact of policy changes on the current trajectory of land use/cover changes (Pijanoswski et al., 2002; Eastman et al., 2005). While literature review reveals a plethora of land use/cover models based on different modelling techniques and traditions (Verbug et al., 2004), we focus briefly only on the Markov chain, CA and Markov-CA models in sections 3.2.1, 3.2.2 and 3.2.3, respectively.

3.2.1. Markov Chain Analysis A Markov chain analysis is a stochastic model based on transition probabilities that describes a process that move in a sequence of steps through a set of states (Wu et al., 2006). Markov chain analyses have been widely used to model land use/cover changes (Drewett, 1969; Bourne, 1971; Bell, 1974; Bell and Honojosa, 1977; Robinson, 1978; Jahan, 1986; Muller and Middleton, 1994). Stationary and first-order Markov chain have usually been assumed except in a few studies where the stationarity or the order of Markov chains was tested (Bell, 1974; Robinson, 1978). Markov chain models for land use/cover change analysis have several assumptions and limitations. Firstly, land use/cover changes are considered as a stochastic process, where the transition probabilities are stationary and land use/cover classes are in different states of the Markov chain (Wu et al., 2006).

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

26

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Because land use/cover change is a result of a complex dynamics of socioeconomic, political and biophysical factors over time, it would be difficult to expect stationarity in land use/cover data. However, it might be practical to regard land use/cover change to be reasonably stationary if the time span is not too long (Weng, 2002). Secondly, land use/cover modelling based on the Markov chain model that handle stationary processes may not be appropriate for incorporating human activities because transition probabilities among landscape states are not constant (Boerner et al., 1996; Weng, 2002). Finally, a stochastic Markov chain model does not consider spatial knowledge of distribution within each class and transition probabilities are not constant among landscape states (Boerner et al., 1996). While the Markov chain models have some limitations, they are relatively easy to derive (or infer) from land use/cover data. In addition, the Markov model does not require deep insight into the mechanisms of change, but it can help to indicate areas where such insight would be valuable and hence act as both a guide and stimulant for further research. This chapter assumed land use/cover changes as a finite first-order Markov chain with stationary transition probabilities because the time span of the study is not too long (Wu et al., 2006).

3.2.2. Cellular Automata Cellular Automata (CA) models were originally conceived by Ulam and Von Neumann in the 1940s in order to understand the behaviour of complex systems. Generally, CA models consist of a regular grid of cells, each of which can be in one of a finite number of possible states, updated synchronously in discrete time steps according to local interaction and transition rules (Messina and Walsh, 2001). Transition rules for the typical CA model depend on the state of a cell, the state of its surrounding cells, the physical characteristics of the cell, and the weights associated with the neighbourhood context of the cell (Wolfram, 1984). These weights and neighbourhood conditions are determined from empirical analyses of land use/cover changes. The CA model works by; (a) simulating the present by extrapolating from the past using the land use/cover map time-series, (b) validating the simulations via a time-series of previous remote sensing imagery and the ground truth data, and (c) allowing the model to iterate to any other selected date (Messina and Walsh, 2001). The rural landscape can be considered as a dynamic and complex system, characterised by processes and relationships that are non-linear. For example,

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

27

land use/cover is viewed as a mosaic of discrete states, where changes are multidirectional, instead of “ordered” and “unidirectional” (Li and Reynolds, 1997). In addition, the complex structure of social inequalities that emerges from non-linear interactions of households, land tenure system, and land uses at local scales (Walsh et al., 2006) indicate that rural landscapes are dynamic and complex. In this context, CA models are appropriate for capturing the land use/cover dynamics that conventional mathematical systems cannot appropriately explain (Gar-On-Yen, 2009) because of their simplicity, transparency, strong potential for dynamic spatial simulation, and innovative bottom-up approach (Cheng and Maser, 2004).

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

3.2.3. Markov-Cellular Automata Model The Markov-Cellular Automata (MCA) model, which comprises the Markov chain and cellular automata models, is robust for modeling land use/cover changes because geographic information system (GIS) and remote sensing data can be integrated (Silvertown et al., 1992; Li and Reynolds, 1997). According to Wang and Zhang (2001), biophysical and socioeconomic data can be used to; (1) define initial conditions, (2) parameterize the MCA model, (3) calculate transition probabilities, and (4) determine the neighborhood rules with transition potential maps. In the MCA model, the Markov chain process controls temporal dynamics among the land use/cover classes based on transition probabilities, while the spatial dynamics are controlled by local rules determined either by the cellular automata spatial filter or transition potential maps (Silverton et al., 1992; Eastman et al., 2005). Although the potential of the MCA model has been recognized by Li and Reynolds (1997), few studies have combined biophysical and socioeconomic data for simulating land use/cover changes (Myint and Wang, 2006).

3.3. STUDY AREA The Masembura and Musana communal areas are located in the Bindura district to the northeast of the capital city, Harare (17o 20’-17o 00’ S, 31o 05’31o 35’ E) in the Mashonaland Central province of Zimbabwe (Figure 1). The study area covers an area of approximately 525 km2. The altitude varies from 940 m to 1440 m above sea level, falling into the middle veldt. The highest

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

28

Courage Kam musoko, Masaamu Aniya andd Munyaradzi Manjoro

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

temperaatures usually occur in the second half of o October or early Novem mber with an average maxiimum temperaature in the rannge of 26 oC-335 oC.

Figure 1. Study area; Masembura M and Musana M commuunal areas, origiinal ASTER VN NIR M 2005) in fallse color displayy (bands 3, 1, annd 1 - RGB). image (aacquired on 5 May

Thee study area receives r a meaan annual rainnfall ranging from 700 mm m to 1000 mm and is distriibuted from mid-October m too mid-March. Fin ne grained archhaelian granodiorites with pockets of doolerite and gneeiss are pred dominant in the study areea. The underrlying geologgy has a markked influencce on the soills in study areea, which aree mostly sandyy fersialitic soils with in nherent low fertility f and low l water hoolding capacitty (Nyamapfeene, 1991). Masembura M a Musana communal areas are dominaated by Miom and mbo woodlan nds, and mostt predominanttly bushland with w canopy 28-80%. 2 Musana commun nal areas prresent more cultivation and mixed rangelands thhan woodlan nds. The major economic activity in thhe study area is mostly semisubsisteence agricultuure, with major crops such as maize, groundnuts, and cotton, as well as vegetables in i those areaas with irriggation. Howevver, b the unreliaable producttion of the maajor rain-fed crops is usuaally affected by rainfall patterns, partticularly the laate onset of thhe rainy seasoon. Accordingg to

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

29

the 2002 population census, the population density is approximately 47 inhabitants per km2 (Central Statistical Office 2004). The study area was thus selected because of its range of biophysical and socioeconomic characteristics dominated by semi-subsistence farming, which is typical in most of Zimbabwe’s rural landscape.

3.4. METHODOLOGY

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

3.4.1. Data For simulating land use/cover changes, we used a total of 13 data layers grouped into two categories namely ward/statistical boundaries and biophysical data, and socioeconomic data. The description of the data is following. The ward boundaries were digitized from the Bindura district administrative map using ArcGIS 9.0 software (ESRI, 2004). Land use/cover maps for 1973, 1989 and 2000 (Figure 2) were extracted from the Bindura district land use/cover maps that were classified from Landsat data based on a hybrid supervised/unsupervised classification approach coupled with GIS (Kamusoko and Aniya, 2007). In addition, a land use/cover map for 2005 was produced from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data (Figure 2). Overall land use/cover classification accuracy levels for the four dates range from 86% to 90%, with Kappa indexes of agreement ranging from 0.83 to 0.88. Elevation was derived from the Bindura district digital elevation model (DEM), obtained from the Forestry Company of Zimbabwe. Other biophysical data such as distance to the town centre, and distance to rivers were digitized from the 1:50 000 topographic maps (Department of the Survey General, Zimbabwe). The DEM, “distance to town centre”, and “distance to rivers” were converted to 30-m raster. Socioeconomic data was collected from household surveys in 2004 (pilot household survey) and 2005 (actual household survey), respectively (Kamusoko and Aniya, 2009). Socioeconomic data collected included “population density”, “area under the cultivation of maize”, “area under the cultivation of groundnuts”, “fuelwood consumption”, “distance travelled to fetch fuelwood”, “total yield of maize produced”, “total yield of groundnuts produced” “livestock density”,

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

30

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

and “quantity of fertilisers used”. Finally, the socioeconomic variables were rasterized and converted into ArcGIS grid format at a 30-m spatial resolution.

Figure 2. Land use/cover maps of Masembura and Musana communal areas (with a ward boundary overlay): (A) 1973; (B) 1989; (C) 2000; and (D) 2005.

3.4.2. Markov-Cellular Automata Modelling Framework The following procedures that were performed in IDRISI Kilimanjaro software in order to implement the MCA model; (1) computation of land use/cover transition potential maps based on multicriteria evaluation (MCE) procedure, (2) computation of transition probabilities using Markov chain analysis, and (3) spatial allocation of simulated land use/cover probabilities based on multi-objective land allocation (MOLA) and a CA spatial filter.

Computation of Land Use/Cover Transition Potential Maps Firstly, the different range and measurement units of the biophysical and socioeconomic factors were transformed into comparable transition potential values using the IDRISI standardization algorithm to compute transition

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

31

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

potential maps that represent the likelihood or the probability that land would change from one land use/cover class to another (Eastman, 2003). Then, an Analytic Hierarchy Process (AHP) weight derivation tool was used to compute factor weights based on preference factor information (Saaty, 1977), derived from questionnaires administered to three agricultural extension officers and ten ward councilors during the 2005 field survey. Subsequently, Consistency Ratio (CR) that shows the probability to which the preference factor ratings were randomly assigned was calculated. In this study, we obtained a satisfactory CR of 0.09 (Saaty and Vargas, 2001). Thus, to compute the “1989 agriculture transition potential map”, agriculture change map from 1973 to 1989, and biophysical and socioeconomic data for 1989 were combined based on a weighted linear combination algorithm (Eastman et al., 1995) using weights derived from the AHP procedure (Table 1). For the “1989 woodland transition potential map”, the woodland change map from 1973-1989, and biophysical and socioeconomic data were used. The same biophysical and socioeconomic data were used for computing the “mixed rangeland and bareland transition potential maps”. Table1. Weights derived from the Analytic Hierarchy Process (AHP) procedure Factors Distance travelled to fetch fuelwood

Weight 0.13

Fuelwood consumption Population density Elevation Quantity of fertilisers used Area under the cultivation of groundnuts Area under the cultivation of maize Total yield of maize produced Total yield of groundnuts produced Distance to rivers Livestock density Distance to town centre Total

0.12 0.12 0.11 0.09 0.09 0.08 0.08 0.07 0.06 0.04 0.01 1.00

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

32

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

Since the area of water in the study is small, transition potential to water was not computed. With the same procedure, we computed the 2000 land use/cover transition potential maps, using biophysical and socioeconomic data for 2000.

Computation of Transition Probabilities Using Markov Chain Analysis The Markov chain analysis was used to compute transition probabilities based on the Landsat derived land use/cover maps for 1973, 1989 and 2000 (see Figure 2). Two transition matrices were constructed from the cross tabulation of the land use/cover maps (that is, the 1973-1989 and 1989-2000 land use/cover maps). The time intervals used for calibration were 16 and 11 89

years for the 1973-1989 transition matrix ( M 73 ), and the 1989-2000 transition 00

matrix ( M 89 ), respectively. In order to account for differences in the lengths

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

of the two time periods (16 years for the 1973-1989 matrix and 11 years matrix for the 1989-2000 matrix), transition probabilities were normalized to annual time steps as demonstrated by Pastor et al. (1993).

Spatial Allocation of Simulated Land Use/Cover Probabilities Three datasets, (I) the 1989 land use/cover base map, (II) the 1989 transition potential maps, and (III) the 1973-1989 transition area matrix, were integrated using MOLA and CA spatial filter in order to simulate the 2000 land use/cover map. Cellular automata (CA) iterations were specified as 16 because of the 16-year difference between 1973 and 1989. With each cellular automata pass, each land use/cover transition potential map is re-weighted as a result of the 5 x 5 contiguity filter, which determines the location of the simulated land use/cover class (Pontius and Malanson, 2005). Once re-weighted, the revised transition potential maps are then run through MOLA to allocate 1/16 of the required land use/cover in the first run, and 2/16 the second run, and so forth, until the full allocation of land for each land use/cover class is obtained (Myint and Wang, 2006). MOLA procedure resolves land allocation conflicts by allocating the cell to the objective for which its weighted transition potential is highest based on a minimum distance to ideal point rule using the weighted ranks (Houet and Hubert-Moy, 2006). The transition area matrix derived from the Markov chain analysis determines how much land is allocated to a class over the n-year period (Myint and Wang, 2006).

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

33

At the end of each iteration, a new land use/cover map is generated by overlaying all results of the MOLA procedure. For the simulation of the 2005 land use/cover map, a similar procedure described for the 2000 simulated map was carried out, specifying 11 cellular automata iterations based on the (1) 2000 land use/cover base map, (2) the 2000 transition potential maps, and (3) the 1989-2000 transition area matrix.

3.5. RESULTS AND DISCUSSION 3.5.1. Analysis of Land Use/Cover Changes and Transition Probabilities

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Results show that agriculture, woodland and mixed rangeland areas were the dominant land use/cover classes in the study area (Figures 2 and 3). From 1973 to 2000, agriculture areas increased from 211 km2 to 346.4 km2, while in 2005 they slightly decreased to 339 km2. However, woodland areas decreased significantly from 182.7 km2 to 35.4 km2 between 1973 and 2005.

Figure 3. Areas of land use/cover classes in the study area.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

34

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

During the 1973 to 1989 period, mixed rangeland areas decreased from 116.9 km2 to 99.8 km2, whereas between 1989 and 2005 they increased slightly from 99.8 km2 to 109.5 km2. Generally, bareland areas increased from 12.2 km2 to 38.8 km2, while water areas slightly changed over the study period. The major land use/cover changes were mainly from woodland and mixed rangeland areas to agriculture areas. Tables 2 and 3 show the land use/cover transition probabilities and transition area matrix for the 1973-1989 and 1989-2000 periods, calculated on the basis of the frequency distribution of the observations. The diagonal of the transition probability represents the selfreplacement probabilities, that is the probability of a land use/cover class remaining the same (shown in bold in tables 2 and 3), whereas the off-diagonal values indicate the probability of a change occurring from one land use/cover class to another. While there is a 16 year time lag for the 1973-1989 matrix and 11 year time lag for the 1989-2000 matrix, table 4 shows that the differences in the normalized land use/cover transition probabilities for the two time periods are significantly small. Therefore, the land use/cover transition probabilities derived from the two time periods can be used as an input in the MCA model since they indicate the direction and magnitude of land use/cover process (Weng, 2002). Table 2(a). Land use/cover transition probabilities, 1973-1989 1989 1973 Agriculture Woodland Mixed rangeland Agriculture 0.81 0.03 0.11 Woodland 0.47 0.21 0.24 Mixed rangeland 0.62 0.11 0.16 Bareland 0.12 0.10 0.06

Bareland 0.05 0.08 0.11 0.72

Table 2(b). Land use/cover transition area matrix (in km2), 1973-1989

1973 Agriculture Woodland Mixed rangeland Bareland

Agriculture 264.39 18.12 53.37 2.06

1989 Woodland 4.17 30.63 16.42 2.83

Mixed rangeland 29.21 28.32 21.19 1.06

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Bareland 7.20 2.67 8.46 31.59

Markov-Cellular Automata Approach for Modelling Land Use …

35

Table 3(a). Land use/cover transition probabilities, 1989-2000

1989 Agriculture Woodland Mixed rangeland Bareland

Agriculture 0.86 0.24 0.53 0.21

2000 Woodland 0.02 0.35 0.08 0.09

Mixed rangeland 0.11 0.36 0.35 0.12

Bareland 0.01 0.04 0.03 0.59

Table 3(b). Land use/cover transition area matrix (in km2), 1989-2000

1989 Agriculture Woodland Mixed rangeland Bareland

Agriculture 298.86 10.48 53.56 6.47

2000 Woodland 5.89 15.27 8.05 2.70

Mixed rangeland 37.23 15.65 35.54 3.63

Bareland 3.92 1.70 3.01 18.48

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Table 4(a). Normalized land use/cover transition probabilities, 1973-1989

1973 Agriculture Woodland Mixed rangeland Bareland

Agriculture 0.988 0.039 0.059 0.008

1989 Woodland 0.002 0.939 0.007 0.007

Mixed rangeland 0.007 0.017 0.927 0.004

Bareland 0.003 0.005 0.007 0.982

Table 4(b). Normalized land use/cover transition probabilities, 1989-2000

1989 Agriculture Woodland Mixed rangeland Bareland

Agriculture 0.987 0.025 0.066 0.021

2000 Woodland 0.002 0.932 0.008 0.009

Mixed rangeland 0.011 0.040 0.923 0.012

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Bareland 0.001 0.004 0.003 0.959

36

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

3.5.2. Validation of the Markov-Cellular Automata Model

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

For model validation, we compared the simulated land use/cover maps for 2000 and 2005 with the actual satellite-derived land use/cover maps based on the Kappa statistic. The MCA’s overall simulation success is 69% and 83% for 2000 and 2005, respectively. Based on Figure 5, agriculture, woodland and mixed rangeland classes in the simulated land use/cover map for 2000 are relatively similar to the corresponding classes in the actual land use/cover map for 2000, while the bareland class is poorly simulated. The best agreement is shown in the mixed rangeland class, where the actual class is 100.5 km2, while the corresponding simulated class is 109.6 km2 (Figure 4). Analysis of the 2005 results indicate that agriculture, woodland, and mixed rangeland classes in the simulated land use/cover map are relatively close to the corresponding classes in the actual land use/cover map, while the bareland class is poorly simulated (Figure 5). The best agreement is shown in the woodland class, where the actual class is 35.4 km2 and the corresponding simulated class is 36.2 km2 (Figure 6). Analysis of the simulated land use/cover maps in 2000 and 2005 reveals that the Markov-cellular automata model generally underpredicted the location of the bareland class.

Figure 4. Simulated versus actual land use/cover maps in 2000.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

37

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Figure 5. Simulated versus actual land use/cover classes in 2000.

This is partly explained by the poor calibration of bareland transition potential maps due to the unavailability of spatial data such as the area of extended or abandoned crop fields. Furthermore, the MCA model employs the contiguity rule, with which to simulate the growth of a land use/cover class near the existing similar land use/cover class (Pontius and Malanson, 2005). In this chapter, the MCA’s contiguity rules applies a 5 x 5 spatial filter to the transition potential maps with strong weighting towards predicting new agriculture areas near mixed rangeland and woodland edges. Since many of the nearby pixels belong to the land use/cover class such as agriculture or mixed rangeland, high transition potential is maintained, resulting in moderate to good simulation of the agriculture, mixed rangelands and woodland classes. This suggests that the model’s simulation accuracy increases with the proportion of a given land use/cover class relative to others. Conversely, if few of the nearby pixels belong to a land use/cover class such as bareland, then the transition potential is down-weighted, which could possibly result in the poor simulation of that class. It is also important to consider the influence of the time step on the location of the simulated land use/cover class when the contiguity rule is applied because the definition of edge is updated at every iteration of the time step (Pontius and Malanson, 2005). For example, if the extrapolation has small time steps, then the Markov-cellular automata model can simulate incremental growth at the locations with high transition potential

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

38

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

values, because smaller time steps lead to more iterations and hence more frequent updates of the spatial dependency.

3.5.3. Simulated Future Land Use/Cover Changes

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Based on the success of the models for 2000 and 2005, we simulated future land use/cover maps for 2010, 2020 and 2030 (Figure 6), using the 2000 land use/cover base map, the 1989-2000 transition area matrix and the 2000 transition potential maps. The Markov-cellular automata model simulations predicted that woodland areas could decrease from 8.0% to 4.0% in the study area, while mixed rangeland could increase slightly from 19.2% to 19.6%. Bareland areas could also increase from 6.4% to 12%. Conversely, agriculture areas could slightly decrease from 66.0% to 65.3%. The simulated future land use/cover changes have significant environmental and socioeconomic implications for sustainable rural land use planning in the study area.

Figure 6. Simulated future land use/cover maps: (A) 2010, (B) 2020, and (C) 2030.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

39

Taking into consideration the high population density and overcrowding in the communal areas, the simulated future land use/cover changes indicates increasing pressure on land and woodland resources. For instance, the continuing decline in woodland areas on one hand and the increase in bareland areas on the other hand imply severe land degradation in the future, which potentially threatens rural livelihoods in the communal areas since woodlands provide life-sustaining products such as food and fuelwood. The MCA model employed in this study area was able to advance previous research (e.g., Paeglow, 2005) by incorporating additional socioeconomic factors such as “population density”, “distance travelled to fetch fuelwood”, “fuelwood consumption”, “area under the maize of cultivation”, “area under the cultivation of groundnuts”, “total yield of maize produced” to list but a few. Since the simulated future land use/cover change maps takes into account significant socioeconomic factors, deforested areas are identified, which can thus be prioritized for immediate policy interventions.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

3.6. SUMMARY AND CONCLUSIONS The MCA model that combines the Markov chain analysis and CA models simulated successfully land use/cover changes in the Musana and Masembura communal areas in Zimbabwe. The model’s overall simulation success was 69% for the 2000 simulated land use/cover map and 83% for the 2005 simulated land use/cover map. Statistical analyses revealed that while agriculture, woodland and mixed rangelands are relatively well simulated, the bareland class was poorly simulated due to lack of input spatial data. Based on the 2000 calibration scenario, the MCA model simulated future land use/cover changes up to 2030, projecting the decrease in woodland areas and an increase in bareland areas. The simulated land use/cover map for 2030 indicated that if the current land use/cover trends continue without holistic sustainable development policies involving the participation of all the stakeholders in the study area, severe land degradation will occur, with potential threats to rural sustainability. This chapter represents an important contribution to land use/cover modeling as shown by the integration of biophysical and socioeconomic data into a spatially explicit MCA land use/cover simulation model, which to our knowledge has not been attempted before in a African rural landscape. Equally important, the MCA model applied in this study area incorporated local

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

40

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

knowledge in the simulation of land use/cover changes through a GIS-based multicriteria decision support system. In light of land degradation problems in Zimbabwe, the simulated future land use/cover maps produced in this study provide a strategic guide to rural land use planning, especially efforts to reduce deforestation. Furthermore, the simulated future land use/cover maps can serve as an early warning system of the future effects of land use/cover changes, particularly in other rural areas in Sub-Sahara Africa, which are experiencing similar land use/cover changes. Although the model has successfully simulated future land use/cover changes based on biophysical and socioeconomic factors, policies that influence the behavior of local farmers have not been considered. Thus, future study should attempt to include policy related factors in the simulation of future land use/cover changes.

ACKNOWLEDGMENTS

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

We sincerely thank the department of Agriculture Research and Extension (AREX), the ward councilors, and all the participants of the 2004 and 2005 household surveys in the Bindura district, Zimbabwe.

REFERENCES Abalu, B., and Hassan, R. (1998). Agricultural productivity and natural resource use in southern Africa. Food Policy, 23 (6), 477-490. Amissah-Arthur, A., and Miller, R. B. (2002). Remote sensing applications in African agriculture and natural resources: Highlighting and managing the stress of increasing population pressure. Advances in Space Research, 30 (11), 2411-2421. Arkin, H., and Colton, R. (1963). Table for Statistics. New York: Barnes and Noble Publication. Barreteau, O., and Bousquet, F. (2000). SHADOC: A multi-agent model to tackle viability of irrigated systems. Annals of Operations Research, 94, 139-162. Bell, E. J. (1975). Stochastic analysis of urban development. Environment and Planning A, 7, 35-39.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Markov-Cellular Automata Approach for Modelling Land Use …

41

Bell, E. J., and Hinojosa, R. C. (1977). Markov analysis of land use change: continuous time and stationary processes. Socio-Economic Planning Sciences, 11, 13-17. Berger, T. (2001). Agent-based spatial models applied to agriculture: A simulation tool for technology diffusion. Resource use changes and policy analysis. Agricultural Economics, 25, 245-260. Boerner, R. E .J., DeMers, M. N., Simpson, J. W., Artigas, F. J., Silva, A., and Berns, L. A. (1996). Markov Models of inertia and dynamic on two contiguous Ohio landscapes. Geographical Analysis, 28, 56-66. Bourne, L. S. (1971). Physical adjustment processes and land use succession: a review and central city example. Economic Geography, 47, 1-15. Bousquet, F., Bakam, I., Proton, H., and Le Page, C. (1998). Cormas: common-pool resources and multi agent systems. Lecture Notes in Artificial Intelligence, 1416, 826-837. Campbell, B. M., Costanza, R., and van den Belt, M. (2000). Land use options in dry tropical woodland ecosystems in Zimbabwe: Introduction, overview and synthesis. Ecological Economics, 33, 341-351. Cheng, J., and Masser, I. (2004). Understanding spatial and temporal processes of urban growth: cellular automata modelling. Environment and Planning B: Planning and Design, 31, 167-194. Chenje, M., Sola, L., and Palecny, D (Eds). (1998). The State of Zimbabwe’s Environment 1998. Ministry of Mines, Environment and Tourism, Harare, Zimbabwe, 509 pp. Chimhowu, A. and Hulme, D. (2006). Livelihood dynamics in planned and spontaneous resettlement in Zimbabwe: Converging and vulnerable. World Development, 34, 728-750. Davidson, O., Halsnaes, K., Huq, S., Kok, M., Metz, B., Sokona, Y., and Verhagen, J. (2003). The development and climate nexus: the case of subSaharan Africa. Climate Policy, 351, S97-S113. Drewett, J. R. (1969). A stochastic model of the land conversion process. Regional Studies, 3, 269-280. Eastman, J. R., Jin, W., Kyem, P. A. K., and Toledano, J. (1995). Raster procedures for multi-criteria/multi-objective decisions. Photogrammetric Engineering and Remote Sensing, 61 (5), 539-547. Eastman, J. R. (2003). Idrisi Kilimanjaro, Guide to GIS and Image Processing. Clark University Edition, 328 pp. Eastman, J. R., Solorzano, L. A., and Van Fossen, M. E. (2005). Transition potential modeling for land-cover change. In D. J. Maguire, M, Batty, and

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

42

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

M. F. Goodchild (Eds.), GIS, Spatial Analysis, and Modeling (pp. 357385). California, ESRI Press. ESRI. (2004). ArcView GIS: The Geographic Information Systems for Everyone. Redlands, CA, USA, 572 pp. Etter, A., McAlphine, C., Wilson, K., Phinn, S and Possingham, H. (2006). Regional patterns of agricultural land use and deforestation in Colombia. Agriculture, Ecosystems & Environment, 114, 369-386. Evans, T. P., and Kelley, H. (2004). Multi-scale analysis of a household level agent-based model of land cover change. Journal of Environmental Management, 72 (1-2), 57-72. Fang, S., Gertner, G. Z., Sun, Z., and Anderson, A. A. (2005). The impact of interactions in spatial simulation of the dynamics of urban sprawl. Landscape and Urban Planning, 73, 294-306. FAO, 2003. Forestry Outlook Study for Africa: regional report- opportunities and challenges towards 2020. African Development Bank, European Commission and the Food and Agriculture Organization of the United Nations, Rome. Flamm, R. O., and Turner, M.G. (1994). Alternative model formulations for a stochastic simulation of landscape change. Landscape Ecology, 9, 37-46. Gambiza, J., Bond, W., Frost, P. G. H., and Higgins, S. (2000). Special section: Land use options in dry tropical woodland ecosystem in Zimbabwe. A simulation model of miombo woodland dynamics under different management regimes. Ecological Economics, 33, 353-368. Gar-On Yeh, A and Li, X. (2009). Cellular Automata, and GIS for Urban Planning. In M. Madden (Ed.), Manual of Geographic Information Systems (pp. 591-619). USA: ASPRS. Hamandawana, H., Nkambwe, M., Chanda, R., and Eckardt, F. (2005). Population driven changes in land use in zimbabwe’s Gutu district of Masvingo province: some lessons from recent history. Applied Geography, 25, 248-270. Houet, T., and Hubert-Moy, L. (2006). Modelling and projecting land-use and land-cover changes with a cellular automaton in considering landscape trajectories: an improvement for simulation of plausible future states. EARSel eProceedings, 5 (1), 63-76. Jahan, S. (1986). The determination of stability and similarity of Markovian land use change processes: a theoretical and empirical analysis. SocioEconomic Planning Science, 20, 243-251.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Markov-Cellular Automata Approach for Modelling Land Use …

43

Kamusoko, C., and Aniya, M. (2007). Land use/cover change and landscape fragmentation analysis in the Bindura district, Zimbabwe. Land Degradation and Development, 18, 221-233. Lahiff, E. (2003). The politics of land reform in southern Africa. Sustainable Livelihoods in Southern Africa Research Paper 19, Institute of Development Studies, Brighton. Li, H., and Reynolds, J. F. (1997). Modeling effects of spatial pattern, drought, and grazing on rates of rangeland degradation: a combined Markov and cellular automaton approach. In: D. A. Quattrochi, and M. F. Goodchild (Eds.), Scale in Remote Sensing and GIS (pp. 211-230). Boca Raton, Florida: Lewis Publishers. McCusker, B. and Ramudzuli, M. (2007). Apartheid spatial engineering and land use change in Mankweng, South Africa: 1963-2001. The Geographical Journal, 173, 56-74. Messina, J., and Walsh, S. (2001). 2.5D morphogenesis: modeling landuse and landcover dynamics in the Ecuadorian Amazon. Plant Ecology, 156, 7588. Muller, M. R., and Middleton, J. (1994). A Markov model of land use change dynamics in the Niagara Region, Ontario, Canada. Landscape Ecology, 9, 151-157. Munro, W. A. (2003). Ecological ‘crisis’ and resource management policy in Zimbabwe’s communal lands. In T. J. Basset, and D. Crummey (Eds.), African Savannas. Global Narratives and Local Knowledge of Environmental Change, (pp. 179-204). Portsmouth NH, UK: Heinemann. Myint, S. W, and Wang, L. (2006). Multicriteria decision approach for land use land cover change using Markov chain analysis and a cellular automata approach. Canadian Journal of Remote Sensing, 32 (6), 390404. Nyamapfene, K. (1991). Soils of Zimbabwe. Harare, Zimbabwe: Nehanda Publishers, 179 pp. Paegelow, M., and Olmedo, M. T. C. (2005). Possibilities and limits of prospective GIS land cover modelling - a compared case study: Garrotxes (France) and Alta Alpujarra Granadina (Spain). International Journal of Geographic Information Science, 19 (6), 697-722. Pankhurst, D., 2000. Unravelling reconciliation and justice? Land and the potential for conflict in Nambia. Peace & Change, 25, 239-254. Parker, D. C., Manson, S. M., Janssen, M. A., Hoffman, M., and Deadman, P. (2003). Multi-agent systems for the simulation of land-use and land-cover

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

44

Courage Kamusoko, Masamu Aniya and Munyaradzi Manjoro

change: A review. Annals of the Association of American Geographers, 93 (2), 314-337. Pastor, J., Bonde, J., Johnston, C., and Naiman, R. J. (1993). Markovian analysis of the spatially dependent dynamics of beaver ponds. In R.H. Gardner (ed.) Predicting Spatial Effects in Ecological Systems. Lectures on Mathematics in the Life Sciences, (Vol. 23, pp. 5-27). American Mathematical Society, Providence, RI. Paudel, G. S., and Thapa, G. B. (2004). Impact of social, institutional and ecological factors on land management practices in mountain watersheds of Nepal. Applied Geography, 24, 35-55. Pinstrup-Andersen, P., Pandya-Lorch, R., and Babu, S. (1997). A 2020 vision for food, agriculture, and the environments in southern Africa. In L. Haddad (Ed.), Achieving Food Security in Southern Africa. Washington, D. C: International Food Policy Research Institute (IFPRI). Pontius Jr, R. G., and Malanson, J. (2005). Comparison of the structure and accuracy of two land change models. International Journal of Geographical Information Science, 19 (2), 243-265. Robinson, V. B. (1978). Information theory and sequences of land use: an application. Professional Geographer, 30, 174-179. Rockstrom, J., Folke, C., Gordon, L., Hatibu, N., Jwitt, G., Penning de Vries, F., Rwehumbiza, F., Sally, H., Savenije, H., and Schulze, R. (2004). A watershed approach to upgrade rainfed agriculture in water scarce region through water system innovations: an integrated research initiative on water for food and rural livelihoods in balance with ecosystem functions. Physics and Chemistry of the Earth, Parts A/B/C, 29, 1109-1118. Saaty, T. L. (1977). A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology, 15, 234-281. Saaty, T. L., and Vargas, L. G. (2001). Models, Methods, Concepts and Application of AnalyticHierarchy Process.Boston, USA: Kluwer Academic Publishers, 333 pp. Silverton, J., Holtier, S., Johnson, J., and Dale, P. (1992). Cellular automaton models of interspecific competition for space-the effect of pattern on process. Journal of Ecology, 80, 527-534. Torrens, P. M. (2006). Simulating sprawl. Annals of the Association of American Geographers, 96 (2), 248-275. Turner, M. G. (1987). Spatial simulation of landscape changes in Georgia: a comparison of 3 transition models. Landscape Ecology, 1, 29-36. Veldkamp, A., Verbug, P. H., Kok, K., de Koning, G. H. J., Priess, J., and Bergsma, A. R. (2001). The need for scale sensitive approaches in

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Markov-Cellular Automata Approach for Modelling Land Use …

45

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

spatially explicit land use change modelling. Environmental Modeling and Assessment, 6, 111-121. Walsh, S. J., Entwilse, B., rindfuss, R. R and Pages, P. H. (2006). Spatial simulation modelling land use/land cover change scenarios in northern Thailand: a cellular automata approach. Journal of Land Use Science, 1 (1), 5-28. Wang, Y., and Zhang, X. (2001). A dynamic modelling approach to simulate socioeconomic effects on landscape changes. Ecological Modelling, 140, 141-162. Weng, Q. (2002). Land use change analysis in the Zhujiang Delta of China using satellite remote sensing, GIS and stochastic modelling. Journal of Environmental Management, 64, 273-284. Wood, E. C., Tappan, G. G., and Hadj, A. (2004). Understanding the drivers of agricultural land use change in south-central Senegal. Journal of Arid Environments, 59, 565-582. Wolfram, S. (1984). Cellular automata as models of complexity. Nature, 311, 419-424. Wu, Q., Li, H., Wang, R., Paulusen, J., He, Y., Wang, M., and Wang, Z. (2006). Monitoring and predicting land use changes in Beijing using remote sensing and GIS. Landscape and Urban Planning, 78 (4) 322-333.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved. Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

In: Recent Advances in Remote Sensing … ISBN: 978-1-61761-003-5 Editors: C. Kamusoko, C. Mundia et al. © 2011 Nova Science Publishers, Inc.

Chapter 4

URBAN LAND USE CHANGE AND LANDSCAPE FRAGMENTATION IN LAGOS, NIGERIA

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio ABSTRACT The objective of this study was to identify the factors associated with residential land development in regions with different population densities in Lagos, Nigeria. Land use changes were mapped from satellite images, while logistic regression was used to model the probability of residential land development. Landscape metrics were further used to assess human impacts on the landscape. Residential land conversion within the low population density region occurred on the highest elevations, in areas with the highest proportion of rural land and change in population potential, and at the farthest distances from the central business district and designated industrial centers. Conversely, new residential land development within the high population density region occurred at the farthest distance from major roads, waterworks and protected forests. Higher rates of development in low to medium density regions suggest that the dominant process of land change is extensification of residential areas, leading to an increase in physical fragmentation of the landscape. These patterns have significant policy implications for urban environmental management in Lagos.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

48

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio

Keywords: Residential development, land use change, Lagos, population density, pattern metrics

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

4.1. INTRODUCTION Africa is currently undergoing an urban transition at an unprecedented scale and pace. With an estimated population growth rate of 5%, the proportion of Africa’s urban residents doubles every 15 years (UN, 2002). Urbanization in Africa is characterized by a high population momentum, ruralurban migration, and the appropriation and re-classification of land around the periphery of urban areas (Cohen, 2004). However, Africa’s urban transition is occurring within the context of a vulnerable economic base exposed to the vagaries and pressures of global competition (Kessides, 2005). Managing urban growth is one of the most important sustainable development challenges of the region in the 21st century (Rakodi, 1997). Even though urban areas currently occupy about 3% of the earth’s surface, the physical transformation of landscapes associated with the process of urbanization has important consequences for global environmental change (McGranahan et al., forthcoming). Urbanization sometimes occurs on prime agricultural land and forests, thereby leading to the loss of the services provided by these ecosystems. The conversion of rural to urban land also leads to an increase in impervious surface, sedimentation and fragmentation of landscapes (Pickett et al., 1991), while air and water pollution is also associated with increasing levels of urbanization. Changes in urban land use patterns reflect the response of land users to a number of institutional, economic, social and biophysical factors affecting transactions in land and the physical process of construction of buildings. Many of these factors interact at different organizational levels, and in a dynamic way to produce complex patterns of urban land use. An understanding of the driving factors of urban expansion is therefore important for the development of sustainable urban management policies. Lagos city serves simultaneously as the national and regional focal point of economic growth for Nigeria and West Africa. It functioned as the administrative capital of Nigeria since the colonial era in 1914 through independence in 1960 and up till 1990 when the capital was moved to Abuja in central Nigeria. Lagos provides an important linkage to the different economies of Nigerian localities through infrastructures such as roads, dense telecommunications, airports, seaports and markets. It also provides

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Urban Land Use Change and Landscape Fragmentation …

49

employment opportunities to a large population with varying skills in the industrial and commercial sectors. As such it is a rapidly growing city. The metropolitan area during this period grew from 70 km2 to 1,068 km2 engulfing neighboring towns and villages. The city also provides integration to the larger global economy through investment and trade. In the mid-1980s it accounted for 62 percent of gross industrial output and 61 percent of the total national industrial value added (Abiodun, 1997). As the city with the largest seaport and the premier national and international airport in Nigeria, it offers minimum transportation costs for imported inputs from the port to factories. Notwithstanding the importance of the city as an important industrial center, land use change associated with the city’s growth is largely increases in residential areas (Braimoh and Onishi, 2007). An important challenge facing the city is high levels of poverty. The estimated poverty level of 70% (Ministry of Economic Planning and Budget, 2004) makes it one of the poorest of the world’s largest cities. There is also the evidence of infrastructure decay, inadequate housing, slums and environmental degradation. Urban renewal projects were planned in the 1990s, but due to inadequate funds, the scope of these projects was narrowed significantly (Olanrewaju, 2001). The provision of basic infrastructure remains a critical urban development challenge (Abiodun, 1997). Despite the importance of Lagos in the socioeconomic development of Nigeria and for the West African sub-region, there is limited understanding of the factors responsible for the development of its spatial pattern of land use. Such a dearth of knowledge hinders the development of policies for sustainable land use planning and management of the city. In particular, there are few studies on the spatial change in the city’s land use that could provide information for management. The objectives of this chapter are to • •

Identify the factors responsible for residential land use change; and Assess the ecological consequences of residential expansion in Lagos, Nigeria.

To do so, we integrate land use information derived from satellite images with other spatial datasets to model residential development. The next and second section of the chapter presents study area information. The third section presents the data and methods. The fourth section includes the results of the analysis and a discussion. We summarize and conclude in the final section.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

50

Ademola K. Braimoh, B Takkashi Onishi annd Peter J. Maarcotullio

4.2. STUDY ARE EA

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Thee study area off approximateely 2910km2 liies between laatitudes 6º 26’and 6º 50’ N and stretchees between lonngitudes 3º 099’ and 3º 46’ E (Figure 1). It I is characteerized by a wet w equatoriall climate withh mean annuaal rainfall aboove 1800mm m, mean monnthly temperatture of 30ºC, and relative humidity h rangging from 80 0% to 100%. The terrain is relatively flat f with a mean m elevationn of about 24m. 2 This predisposes the coastal plain to occasionall flooding durring intense rainfall. Laggos is compprised of deppositional lanndform featurres: ds, barrier islannds, beaches, low-lying tidaal flats and esttuaries. wetland

Figure 1. Map of Nigeriia showing Laggos State (top). Extent E of the stuudy area is show wn ottom. at the bo

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Urban Land Use Change and Landscape Fragmentation …

51

The wetland environment is comprised of alluvial and yellowish soils, mangrove and freshwater swamps and some tropical hardwoods. Lagos originated as fishing and farming settlement in the 17th Century. The arrival of the British made it to grow in importance as a center of commerce and the administrative capital of Nigeria. Despite the movement of the capital of Nigeria to Abuja in 1991, Lagos remains the dominant center for nonagricultural production, distribution and business services. Currently, Lagos State is one of the 36 autonomous states within the Federal Republic of Nigeria. It is subdivided into 20 local government administrative units, and its population is estimated to grow annually at the rate of 5.6% (UN, 2002). A plurality of land tenure and administrative rules and laws (that is, customary and statutory) exist in Lagos. The customary sector is characterized by heritable usufructary rights. Control over the use of the land is vested in the traditional ruler who is responsible for allocating unused land and adjudicating in land disputes (Adedipe et al., 1997). The statutory land tenure system derives from two sources: the common Law of England based on the Principles of Equity and Statues of General Application, and Local Legislations in Nigeria. The customary tenure was the major form of landholding up to the 1970s. It enabled low income earners access to land for building houses without resorting to illegal occupation (Rakodi, 1997). However, the process by which land was held under the customary sector was associated with increasing commercialization and inequity in land distribution that tended to favor influential groups (Udo, 1990). The Federal Government of Nigeria therefore established the Land Use Act in 1978 to curb speculation in urban land and rationalize usufructary rights. State governors control land in urban areas, whereas local governments control rural lands. Statutory rights of occupancy are granted by the State for a specific period subject to rental payments payable to the State. Customary rights of occupancy are granted for agricultural or residential purposes on the condition that there are no statutory rights on the land. The Land Use Act appears not to have been able to solve the problem of equity in land distribution. Problems associated with the Act include the inefficiency of Land Use and Allocation committees in issuing certificates of occupancy, inconsistencies in implementing the legislation, and bureaucratic, administrative requirements that tend to favor civil servants and wealthy individuals (Okolocha, 1993). Thus, the Land Use Act has not been able to practically replace customary land allocation system especially outside the metropolitan part of Lagos. Rather, the Land Use Act has improved government access to land, and largely benefited high net-worth individuals

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

52

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio

and corporate organizations with top-level connections with government officials. Rapid increase in population has increased demand for housing in Lagos. Over 90% of housing in Metropolitan Lagos is provided by the private sector (Abiodun, 1997). The difficulty for the low income earners to access land for residential development, and the high cost of housing construction partly account for the widening gap between supply and demand for housing. This has in turn escalated the costs of rented apartments. High rent burden and the weak purchasing power of the low income earners result in overcrowding, slums and substandard housing in several parts of Lagos. Information on the land use change and how to manage it effectively is therefore urgently needed.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

4.3. METHODS Land use maps for 1984 and 2000 derived from Landsat Thematic Mapper (TM) images were used. Radiometric correction was carried out by normalizing the radiance values of the 1984 image to those of 2000. Next, a tasseled cap orthogonal transformation was performed on Bands 1-5 and 7 to condition the image data prior to classification (Richards 1995). Last, a variance spatial texture algorithm was applied to the Landsat datasets as inputs to the classification (Anys et al., 1994). Land use/cover classification was carried out using the spectral angle mapping technique (Kruse et al., 1993). Four land use/cover classes were distinguished from the images: residential, industrial/commercial, non-urban (farmlands and forests) and water (Figure 2). Residential and non-urban land areas in 1984 and 2000 were extracted from the maps using relational operators within a geographical information system (GIS). One feature of Lagos is the uneven distribution of population within its local governments, with those within the metropolitan area exhibiting the highest population densities. In this study, we hypothesize that the determinants of residential development might vary among areas with different population densities. Thus, population density surfaces were mapped by linking population census data (Nigeria Population Commission, 1997) with the map of local governments. Three population density classes (regions) were distinguished: low population density region (LPD) with population densities below 1000 persons/km2, medium population density region (MPD) with population densities between 1352 persons/km2 and 5794 persons/km2; and high population density region (HPD) comprising places with population density values ranging from 10,049 persons/km2 to 48,135 persons/km2.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Urban Landd Use Changee and Landscappe Fragmentattion …

53

Figure 2. Land use mapps for 1984 and 2000.

Thee population density d map was w clipped with w the land use u maps so that t both dattasets have thee same spatiall extent. Oth her spatially explicit e indepeendent variablles were creatted within a GIS. G These in nclude travel times, t physicaal (Euclidean) distances from m protected arreas and watter bodies, maanufacturing value v added pootentials, popuulation potenttials and neiighborhood inndices (Tablee 1). Travel times were computed at an estimateed average motor m vehiclee speed of 60 km hr-1, and using slope gradientt, land use in 1984 and roaad types as friictional layerss (Environmenntal Systemss Research Institute, I 20001). Manufactturing value added potenntial measurees the distribuution of the monetary m valuee of non-agricultural econom mic activity of productiion/service centers c acrosss the landsccape. Populattion potentiaal is a measure of populatioon pressure foor a given areaa. It is a meassure of an activity a space of people’s movement m andd impact on land, rather thhan merely a residential presence p of people in a givenn area (Verbuurg et al., 20044).

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011. ProQuest Ebook

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Table 1. Independent variable lista All data

Variable Name/Description

Abbreviation

Mean

Elevation (m) Distance from water (m) Travel time to Lagos Island, the major commercial centre (minutes) Travel time to industrial centers (minutes) Travel time to major roads (minutes) Change in manufacturing and services Value Added between 1984 and 2000 (106 1995 US Dollars) Population potential in 1984 Change in population potential between 1984 and 2000

ALTITUDE WATER CBD

Low Population density region

Medium Population density region

High Population density region

Mean

Mean

Mean

24 6665 107

Standard Deviation 23 6532 32

39 c 6214 a 104 c

21 b 6535 a 75 a

9a 7535 a 85 b

INCENT

63

44

43 b

23 a

28 a

ROADS

56

33

43 a

42 a

46 b

INCOME

19.8

11.8

20.2 a

19.9 a

19.2 a

POP84 POPCHG

6404 23554

8046 28029

6240 a 27353 b

6490 a 18176 a

11400 b 20752 a

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011. ProQuest Ebook

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

All data

Mean

67

Standard Deviation 44

Medium Population density region Mean

68 a

64 a

65 a

N2_FOR

65

41

80 b

75 b

55 a

N3_FOR

58

42

59 a

56 a

55 a

N1_RES

15

30

15 a

19 a

17 a

N2_RES

15

28

15 a

22 b

35 c

N3_RES

8

19

8.4 a

10.9 a

7.5 a

Variable Name/Description

Abbreviation

Mean

Frequency of non-urban land within 270m neighborhood (%) Frequency of non-urban land within 630m neighborhood (%) Frequency of non-urban land within 1260m neighborhood (%) Frequency of residential land within 270m neighborhood (%) Frequency of residential land within 630m neighborhood (%) Frequency of residential land within 1260m neighborhood (%)

N1_FOR

Low Population density region

High Population density region Mean

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011. ProQuest Ebook

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

Table 1. (Continued) All data

a

Low Population density region

Medium Population density region

Mean

Mean

6

Standard Deviation 22

High Population density region Mean

5.8 a

6.4 a

5.4 a

N2_IND

6

21

0.7 a

0.6 a

4.1 b

N3_IND

9

16

9a

9a

11a

WATERWORKS

13026

8337

12410 a

11967 a

13854 a

CONSERVATION

18221

8382

17697 a

17521 a

18417 a

Variable Name/Description

Abbreviation

Mean

Frequency of commercial/ industrial land within 270m neighborhood (%) Frequency of commercial/industrial land within 630m neighborhood (%) Frequency of commercial/industrial land within 1260m neighborhood (%) Distance from waterworks (m) Distance from protected forest (m)

N1_IND

Means with the same letter along the same row are not significantly different among the population density regions (P < 0.05) using the Duncan multiple range test.

Urban Land Use Change and Landscape Fragmentation …

57

Finally, neighborhood indices account for the possible effects of spatial interaction on residential land use decisions (Irwin and Geoghegan 2001). Spatial interaction (also referred to as spatial externality) implies that residential land use development in a particular location depends on the neighborhood characteristics of that location. They were computed from the 1984 land use maps consisting of four land classes and expressed as frequencies. Logistic regression was used to model the probability of residential land use change in each population density region as a function of the variables listed in Table 1. The dependent variable for the logistic regression was a binary presence or absence event where y = 1 means a given pixel was converted to residential land use between 1984 and 2000, and y = 0, otherwise. The logistic regression model is of the form

⎡ p ( y = 1⏐X ) ⎤ n ln ⎢ ⎥ = β 0 + ∑i =1 β i x i + e ⎢⎣1 − p ( y = 1⏐X ) ⎥⎦

(1)

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

where p(y=1| X ) is the probability y takes the value 1, given the vector of independent variables X , βs are model parameters (logit coefficients) to be estimated, and e is the residual error. The quantity

p( y = 1⏐X )

is referred 1 − p( y = 1⏐X ) to as the odds-ratio. After back transformation, the result of the regression may be expressed in terms of conditional probability as: βˆ + ∑ n βˆ x

e 0 i =1 i i  p ( y = 1⏐X ) = βˆ + ∑ n βˆ x 1 + e 0 i =1 i i

(2)

where the hat notation is used to indicate estimated values. Analyses were carried out at a spatial resolution of 90m for the three population density regions. The scale of analyses was dictated by the spatial resolution of the digital elevation model. Prior to performing the logistic regression, we standardized the independent variables to zero mean and unit standard deviation. This ensures that the parameters of the regression models are free from the disparity and scales of measurements of the independent variables. The results of logistic regression can either be interpreted with the logit coefficients or the odds ratio (Hosmer and Lemeshow 2000). The higher

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

58

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio

the absolute value of the logit coefficient, the higher is the contribution of the independent variable in explaining the probability of residential land use change. Odds ratio on the other hand measures changes in the probability of residential development given a unit increase in a given independent variable. β

When β < 0, the odds ratio, e < 1 ; implying that the probability of β residential land use change is decreased. When β > 0 , e > 1 implying the

probability of residential development is increased. When β = 0 , e = 1 implying the probability of residential development is unchanged. The ability of the models to correctly assign the probability of change on the landscape was assessed using the relative operating characteristic, ROC (Pontius and Schneider 2001). The ROC compares observed values of the dependent variable, (that is, the binary data) over the whole range of predicted probabilities. It aggregates into a single index of agreement, the ability of the model to predict the probability of observing residential land use change at various locations on the landscape. The ROC varies from 0.5 for a model that assigns the probability of change at random to 1 for a model that perfectly assigns change. Finally, changes in landscape pattern due to urban development were evaluated using landscape pattern metrics (Turner et al., 2001). Landscape metrics are indices used to quantify landscape pattern. They reflect the ability of the landscape to support ecosystems functions and maintain biodiversity which may not be directly observable. For the analysis, landscapes were defined by the area within the three population density regions. The assumption is that differences in demographic structure and other land use determinants might result in different patterns of urban and non-urban land uses. Three metrics were computed at the class level from the land use/cover maps to assess human impact on the landscape between 1984 and 2000. These are the

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

β

•

•

Mean Patch Size (MPS), which is the average patch size (that is, extent of a contiguous area) measured in hectares. If patches of a given land cover e.g. forest decreases in time, MPS will correspondingly decrease, and could lead to a reduction in species diversity (Forman and Godron, 1986). Landscape Shape Index (LSI), which is a standardized measure of total edge, that is, perimeter to area ratio. It measures the degree of pattern fragmentation. It gives an indication of how much urban land

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Urban Land Use Change and Landscape Fragmentation …

•

59

has developed (and correspondingly, the amount of farmland and forests lost), as well as the degree of irregularity of the landscape due to human impact. LSI is free of units. The higher the value the less aggregated the forms of land use and the more the evidence of human impact on the landscape. Higher values indicate potential loss of ecological functions. Interspersion Juxtaposition Index (IJI), which is a measure of the landscape configuration. It indicates the extent to which patches of different land uses/covers are interspersed. It is a function of the interaction of shape and class. Measured in percent, lower values characterize landscapes in which the patch types are poorly interspersed (i.e., disproportionate distribution of patch type adjacencies) (McGarigal and Marks, 1995). Lower IJI indicates potentially higher risks for organisms to access the needed resources for their growth and development.

4.4. RESULTS AND DISCUSSIONS Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

4.4.1. Quantitative Data on Land Use Change The extent of residential and non-urban land use in 1984 and 2000, and the changes over the 16-year period are shown in Figure 3. Non-urban land represents the proportion of land that can be potentially converted to urban uses. LPD experienced the highest absolute decrease of 16000 ha from about 174000 ha in 1984 to more than 157000 ha in 2000 (Figure 3a). MPD experienced the next highest absolute decrease of about 9000ha over the 16year period, whereas HPD experienced the least (1200 ha). The yearly rates of decrease in non-urban land were 0.6%, 3.2%, and 2.1% for LPD, MPD and HPD, respectively. The overall decrease from about 194000 ha to about 168000 ha implies a net annual loss of 0.8% in non-urban land between 1984 and 2000. There was an increase in residential land use from over 18000 ha to about 35000 ha in the LPD, from more than 17700ha to about 26000 ha in the MPD, and from more than 6600 ha to about 7600 ha in the HPD (Figure 3b). Thus, the annual rates of increase in residential land development were 5.6%, 2.9% and 0.9% for low, medium and high population density regions, respectively. This shows that places with the highest population densities are not necessarily those characterized with the highest rates of residential land development.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

60

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio

250000 200000 1984

ha

150000

2000 100000

Decrease

50000 0 Low

Medium

High

All data

Population density categories

a) Non-urban 80000 70000

ha

60000 50000

1984

40000

2000

30000

Increase

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

20000 10000 0 Low

Medium

High

All data

Population density categories

b) Residential Figures 3. Changes in land use within the three different population density regions.

New developers tend to avoid regions with high population densities. The overall increase from about 45000 ha in 1984 to over 70000 ha in 2000 implies an annual increase of 3.6% in residential land expansion. This rate is much below the population growth rate of 5.6% and provides evidence that housing is grossly inadequate to cater for the increasing population. As mentioned, this problem is related to a number of social and institutional factors such as the skewed distribution of private land, the high costs of undeveloped land as well as housing construction, and weak planning regulations.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

Urban Land Use Change and Landscape Fragmentation …

61

Copyright © 2011. Nova Science Publishers, Incorporated. All rights reserved.

4.4.2. Characterization of New Residential Land Use Change Descriptive statistics of parcels within the three population density regions converted to residential land uses between 1984 and 2000 are given in Table 2. Elevation shows a significant difference among the three population density regions (P < 0.05). Low population density areas are on the highest elevations (mean = 39m), whereas the high population density areas are on the least (mean = 9m). The mean distance of new residential areas from waterbodies are not significant among the population density regions. The mean travel time to Lagos Island, the Central Business District (CBD) is higher than 60 minutes, the estimated average travel time to work in Lagos (United Nations Centre for Human Settlements, 1998). Furthermore, the mean travel time to CBD is significantly different among the density regions. New residential areas within the low population density regions are the farthest from the CBD (mean = 104 minutes). This may be due to two reasons. Reason one is the tendency of the cost of undeveloped land to decrease as one moves away from the CBD. The second reason is the less stringent or sometimes lack of land use control outside the metropolitan area. The mean travel time to designated industrial centers for low population density region is also the highest (43 minutes), and is significantly different from that of medium and high population density regions (23 minutes and 28 minutes, respectively). New residential developments within the high population density region occur at the longest distance from major roads (46 minutes). This average travel time is significantly different from those of low and medium population density regions. Thus, a very high proportion of total residential land conversion (24, 635 ha out of 25, 545 ha = 96%) occurred at an average distance of about 43 minutes from major roads. Both formally developed estates and informal settlements that lack secure tenure occur along the major road networks. The latter often occurs in unsafe environmental conditions lacking the basic amenities for living. Manufacturing value added is highest for the low population density regions and lowest for the high population density region. The differences are however not significant. The population potential for LPD and MPD were not significantly different in 1984, but significantly lower than that of HPD. The LPD however experienced significantly higher change in population potential relative to other population density regions between 1984 and 2000. Values of the neighborhood indices at the three distances are similar across population density regions. There are no significant differences in the means of all neighborhood indices at 270m and 1260m neighborhood distances.

Recent Advances in Remote Sensing and GIS in Sub-Sahara Africa, Nova Science Publishers, Incorporated, 2011.

62

Ademola K. Braimoh, Takashi Onishi and Peter J. Marcotullio Table 2. Logistic Regression results for the three population density regions a) Low population density region (LPD)

Variables POPCHG ALTITUDE N2_FOR N2_IND INCENT CBD ROADS INTERCEPT

Standard Error ( β )

Significance Probability 0.229 0.054