MODELLING DYNAMICS OF URBAN SPATIAL GROWTH USING REMOTE SENSING AND GEOGRAPHICAL INFORMATION SYSTEM Thesis submitted to Andhra University in partial fulfillment of the requirements for the award of Master of Technology in Remote Sensing and Geographical Information System ANDHRA UNIVERSITY By L.R.B.Jitendrudu Supervised by Mr.Sandeep Maithani Scientist/Engineer ‘SE’ Human Settlement Analysis Division IIRS Dehradun Indian Institute of Remote Sensing National Remote Sensing Agency Dept. Of Space, Govt. of India Dehradun-248001, India Certificate This is to certify that the dissertation titled “Modelling Dynamics of Urban Spatial Growth using Remote Sensing and Geographical Information System” is a bonafide record of work carried out by L.R.B.Jitendrudu submitted in the partial fulfillment of Master of Technology Degree in Remote Sensing and Geographical Information system. (Mr.B.S.Sokhi) (Mr.Sandeep Maithani) Head, HUSAD Scientist/Engineer ‘SE’, HUSAD IIRS IIRS Dehradun Dehradun Page CONTENTS No Acknowledgement Abstract List of Figures List of Tables Part A Research Introduction Chapter I Introduction 1.1 Prelude 1 1.1.1 Census Definition of Urban Places 1 1.1.2 Urban Growth Trend in India 2 1.1.3 The Million Cities 2 1.2 Research Objective 3 1.3 Research Methodology 3 1.4 Materials Available 4 1.5 Scope of Study 5 1.6 Structure of Document 5 Chapter II Study Area 2.1 Description of Study Area 6 2.1.1 Origin and History 6 2.1.2 Physiography 8 2.1.3 Climate 9 2.1.4 Major Functions of the City 9 2.2 Rationale of Study Area Identification 11 2.3 Demography 12 2.3.1 Population Statistics 12 2.3.1 Population Density 13 Part B Literature Review Chapter III Urban Theory and Modelling 3.1 Introduction 14 3.2 Urban Growth 14 3.3 Urban Systems 15 3.3.1 Cities as Systems and Their Complexity 15 3.3.2 Projection of Complexity in Urban Growth 18 3.4 Characteristics of Factors of Urban Growth Dynamics 21 3.4.1 Factors of Urban Growth Dynamics at Work 23 3.5 Modelling Urban Environment 24 3.6 Urban models 26 Chapter IV Artificial Neural Networks 4.1 Neural Network Foundation and Definition 29 4.2 Benefits of Neural Networks 30 4.3 Neuron Model 31 4.4 Network Architecture 33 4.5 Learning Process 33 4.5.1 Learning Tasks 35 4.6 Towards Supervised Learning 36 4.7 Single Layered Perceptron 37 4.7.1 Perceptron Learning Algorithm 37 4.7.2 Least Mean Squared Learning Algorithm 38 4.8 Multi Layered Back Propagation Learning Algorithms 39 4.9 Computational Paradigm: Back Propagation Algorithm 41 4.9.1 Inside Back Propagation Neural Networks 42 4.9.2 Training Algorithms 43 4.10 Practical Approach 46 Part C Urban Growth Modelling Chapter V Urban Growth Model Design and Development 5.1 Introduction 56 5.2 Database Preparation 56 5.2.1 Pre-processing of Satellite Data and Maps 56 5.2.2 The GIS Data base for Site Attribute and ANN Training Data 56 5.2.2.1 Variables Driving the Urban Growth of Dehradun 59 5.2.3 Exclusionary Zones 62 5.3 Land Cover Classification 62 5.4 ANN based Calibration of Model 66 5.4.1 Data Preparation for Neural Networks Training 70 5.4.2 Design of Neural Network Architecture 71 5.4.3 Training Neural Network 72 5.5 simulation of Network Response on Model data 77 5.6 Conversion of ANN results to GIS 77 5.7 Temporal Indexing 77 5.7.1 Population Projection and Allotment of Cells 78 Chapter VI Model Accuracy Assessment and Results 6.1 Introduction 81 6.2 Urban Growth Model: Accuracy Assessment Methods 81 6.3 Percentage Match Metric 81 6.4 Urban Growth Distribution Metric 81 6.5 Model A: Six Factors of Urban Growth 83 6.6 Model B: Five Factors of Urban Growth 85 6.7 Simulating and Prediction of Urban Growth in Year 2009 87 Chapter VII Conclusion and Recommendations 7.1 Conclusions 89 7.2 Recommendations 90 References 92 Appendix A 96 Appendix B 97 Appendix C 100 Appendix D 102 Appendix E 113 Acknowledgement I thank Dr. V. K. Dadhwal, Dean, IIRS, for kindly allowing me to pursue my M.Tech. and utilize the lab-setup of IIRS throughout the study period. I express my gratitude and regards for Mr.B.S.Sokhi, Head, Human Settlement Analysis Division, IIRS, who has been inspiring and the instrument for the completion of the entire project. I avail this opportunity to extend a very sincere gratitude to Mr.Sandeep Maithani for giving me an opportunity to work on a challenging topic and guiding me very astutely throughout the entire project. I thank Dr. A.K.Mishra (Program Coordinator, M.Tech in Remote Sensing and GIS) for extending all possible help during my entire project work. Sincerely remain obliged to Mr.B.D.Bharat and Ms.Sadhana Jain, faculty members of Human Settlement Analysis Division, IIRS, Dehradun, for their encouragement during the entire study period. I thank all functionaries of IIRS for making the stay a memoir. I thank Sandeepan Karmakar, Biren Baishya, Dr.Velu Murgan, Mr.Praveen Thakur, Dr.Prashant, Ujjwal Sur, Pushpa Dash and Sreyasi Maiti for much appreciated help and fruitful discussions. I should express my thanks to Shiva, Amit, Devu, Chotu, Raju and other children in IIRS for giving me enjoyable hard games to sweat out the work sickness. I can never forget to mention my friends Rishabh Phagre, N.L.Arjun, Mayank Gera, Himanshu Joshi, Mona De, Swati Ghosh, Debarati Roy, Vedula Pavan, Venkat Rao, Vinayak Sexana, Amit Layak, Ghanshyam Chandorkar, Umesh Sharma, Akshay Mani, Abhishek Mani, Ranbir Singh and R.V.Srikanth for always being there to lift my morale. I am as always grateful to my parents and humbly thank all my family, Dad, Amma, Amlu and Alexe for their perseverance, unstinted support and encouragement throughout the study period. Abstract The research work presents a framework for integrating artificial neural networks and geographical information systems that is implemented for modeling of urban growth and forecasting the land development in the future. Artificial neural networks find a wide range of applications in the modelling of complexity fields such as cybernetics, physics, economics, computer science and many more. Urban systems have also been identified as a non-linear complex system. With the development of high end computational and visualization environments paved by the advancement computers in frontiers of geographical information systems and remote sensing image processing, modelling complex non-linear urban systems (an amalgamation of various physical and stochastic factors) and exploration of new techniques has come to age. Dehradun city has been identified as the study area of this research for parameterization of the model owing to its tremendous growth since becoming an interim Capital of newly formed Uttaranchal State. The model explores the physical factors of urban growth in Dehradun such as transportation network, present development and topographical characteristics. The ANN’s are used to learn the patterns of development in the study area. While GIS is used to develop the spatial and predictor drivers and perform spatial analysis on the results. Remote sensing data is employed in providing the calibration data for the model in the form of temporal datasets that allow land use land cover mapping and change detection. The validation of the model has been performed by designing spatial metrics namely, percentage match metric and urban growth dispersion metric. The factors and model that performed well on simulation has been used to forecast the future urban growth, keeping the nature of influence by decision making complexity afield. Keywords: Urban Growth; Land use change; Artificial neural networks; Geographical Information system; Urban complex system; Simulation. List of Figures Figure 1.1 Population growth of India Figure 1.2 Methodology flowchart Figure 2.1 Study Area Figure 2.2 Master Plan 1982-2001 Figure 4.1 Human Brain Figure 4.2 Human Neuron Cells Figure 4.3 Neuron Model Figure 4.4 Sigmoid Activation Function Figure 4.5 (a, b, c) Network Architecture Figure 4.6 Perceptron Figure 4.7 Error Surface Figure 4.8 Over fitting and Under fitting of Training Data Figure 4.9 Bias Figure 4.10 Generalization of ANN Figure 5.1 IRS 1D LISSIII, 2001 Figure 5.2 IRS P6 LISSIII, 2005 Figure 5.3 Transportation Network Figure 5.4 Digital Elevation Model Figure 5.5 Landcover Map-2001 Figure 5.6 Landcover Map-2005 Figure 5.7 Wardwise Population Density Map Figure 5.8 Model ANN Design Figure 5.9 Training Parameters Figure 5.10 ANN Performance Curve Figure 6.1 Simulated Landcover-2005, Model A Figure 6.2 Simulated Landcover-2005, Model B Figure 6.3 Simulated Landcover-2009, Model B Figure appendixB.1 Landcover Map-2001 Figure appendixB.2 Landcover Map-2005 List of Tables Table 2.1 Decadal Population Growth of Dehradun, 1901-2001 Table 4.1 Differences Between Supervised and Unsupervised Learning Table 4.2 Notation Employed in the BPN Algorithm Discussion Table 4.3 Heuristics of Number of Hidden Nodes Table 4.4 Heuristics of Number of Training Samples Table 5.1 Vector Database Table 5.2 Factors Driving Urban Growth of Dehradun Table 5.3 Description of Land Cover Classes Table 5.4 (a, b) Transverse Divergence Matrix, LISS III-2001&2005 Table 5.5 Accuracy Assessment for Binary Encoded Land Cover Classification Table 5.6 Classification Codes Table 5.7 Target Codes Table 5.8 Training Experiments Table 5.9 Population Projections Table 6.1 Percentage Match Metric Classes Table 6.2 Shannon's Entropy Calculation Summary Table 6.3 Summary of Driving Variables, Model A Table 6.4 Percentage Match Metric, Model A Table 6.5 Summary of Driving Variables, Model B Table 6.6 Percentage Match Metric, Model B 1.1 Prelude India stands close second to China in terms of population numbers, 1,080,264,388 in figures, that is 16% of the world population as in 2005 (source: world population clock) and still growing at 1.4% growth rate.