Assessing, monitoring and mapping forest resources in the Blue Nile region of Sudan using an object-based imageanalysis approach

Table of contents List of tables XIII List of figures XV Acronyms and abbreviations XVII ABSTRACT 1 Kurzfassung 3 1 Introduction 5 1.1 Background and motivation 5 1.1.1 Overview 5 1.1.2 Deforestation and land degradation 6 1.1.3 Remote sensing and forest resource assessment 7 1.2 Research hypotheses 9 1.3 Objectives 10 2 Study area 13 2.1 Overview 13 2.2 Introduction to the study area 13 2.2.1 General description 13 2.2.2 Vegetation 14 2.2.3 Climate 15 2.3 Test sites 17 2.3.1 LU/LC level 17 2.3.2 Forest stand level. 18 3 Theoretical background 19 3.1 Overview 19 3.2 Introduction 19 3.3 Object based image analysis 20 3.3.1 Image segmentation 21 3.3.2 Image objects hierarchy 23 3.3.3 Image object information extraction and feature measures 24 3.3.4 Nearest neighbour classification 28 3.3.5 Rule based classification 29 3.3.6 An application of object-based approach for mapping earth surface features 30 3.3.7 Quality assessment 37 4 Data and research approaches 39 4.1 Overview 39 4.2 Data 39 4.2.1 Earth observation data 39 4.2.2 Field survey data 42 4.3 Methodology 45 4.3.1 Atmospheric correction 45 4.3.2 Geometric correction 46 4.3.3 Image classification 47 4.3.4 Change detection 56 4.3.5 Classification accuracy assessment 57 4.3.6 Forest parameters estimation and model validation 57 5 Land use/ land cover analyses 63 5.1 General overview 63 5.2 Pixel-based classification of Aster imagery by maximum likelihood (ML) classifier 64 5.3 OBIA approaches 66 5.3.1 Segmentation result 66 5.3.2 Hierarchical classification 67 5.3.3 Object-based classification of Aster imagery by Nearest Neighbor (NN) classifier 69 5.3.4 Object-based classification of Aster imagery by rule-based (RB) method 72 5.4 OBIA vs. per pixel classification 74 5.4.1 Classification of agricultural land 76 5.4.2 Classification of residential areas 78 5.4.3 Discrimination of vegetation classes 80 5.5 Scale issue and hierarchical OBIA for mapping LU/LC 82 5.5.1 Overview 82 5.5.2 Pixel-based classification of higher resolution RapidEye scene 83 5.5.3 Object-based classification of higher resolution RapidEye scene 85 5.5.4 Overall assessment of the approaches based on the two selected scales 87 5.6 Summary 88 6 Spatiotemporal analyses and driving forces 91 6.1 General overview 91 6.2 Land use/ land cover classification and accuracy assessment 92 6.2.1 Land use/ land cover classification 92 6.2.2 Accuracy assessment based on error matrix 94 6.2.3 Accuracy assessment based on best classification result 95 6.3 Distribution of LU/LC during the periods 1990, 1999 and 2009 97 6.3.1 Trend, rate and magnitude of land use/ land cover changes 98 6.4 Spatiotemporal object-based post-classification analysis 101 6.4.1 Forest cover change dynamics 1990- 1999 103 6.4.2 Forests cover change dynamics 1999- 2009 106 6.4.3 Overall evaluation of the adopted approach in mapping forest cover change dynam-ics 111 6.5 Major driving forces of forest cover loss and restoration problems 115 6.5.1 Deterioration of forest cover in the area 115 6.5.2 Driving forces of forest cover loss 115 6.5.3 Restoration of destroyed forests 117 6.6 Summary 119 7 Object-based texture measures and forest inventory 121 7.1 General overview 121 7.2 Segmentation 122 7.3 Descriptive analysis of field data 123 7.4 Image metrics and forest inventory 125 7.4.1 Volume 125 7.4.2 Height 125 7.4.3 Density 126 7.4.4 Basal area 127 7.5 Development of the regression model to estimate Acacia nilotica stand volume 128 7.5.1 Model validation 131 7.6 Mapping forest stand volume in Alambwa Acacia nilotica pure stand 133 7.7 Summary 134 8 Conclusions, recommendations and future work 137 8.1 Conclusions 137 8.1.1 Overview 137 8.1.2 Land Use/ Land cover categorical level 137 8.1.3 Integrated framework of change analysis 138 8.1.4 Continuous forest structural attributes level 139 8.2 Recommendations 140 8.3 Limitations 141 8.4 Future work 141 References 143 Appendixes 163