Integration of ALOS PALSAR and Landsat Data for Land Cover and Forest Mapping in Northern Tanzania

Land cover and forest mapping supports decision makers in the course of making informed decisions for implementation of sustainable conservation and management plans of the forest resources and environmental monitoring. This research examines the value of integrating of ALOS PALSAR and Landsat data for improved forest and land cover mapping in Northern Tanzania. A separate and joint processing of surface reflectance, backscattering and derivatives (i.e., Normalized Different Vegetation Index (NDVI), Principal Component Analysis (PCA), Radar Forest Deforestation Index (RFDI), quotient bands, polarimetric features and Grey Level Co-Occurrence Matrix (GLCM) textures) were executed using Support Vector Machine (SVM) classifier. The classification accuracy was assessed using a confusion matrix, where Overall classification Accuracy (OA), Kappa Coefficient (KC), Producer’s Accuracy (PA), User’s Accuracy (UA) and F1 score index were computed. A two sample t-statistics was utilized to evaluate the influence of different data categories on the classification accuracy. Landsat surface reflectance and derivatives show an overall classification accuracy (OA = 86%). ALOS PALSAR backscattering could not differentiate the land cover classes efficiently (OA = 59%). However, combination of backscattering, and derivatives could differentiate the land cover classes properly (OA = 71%). The attained results suggest that integration of backscattering and derivative has potential of utilization for mapping of land cover in tropical environment. Integration of backscattering, surface reflectance and their derivative increase the accuracy (OA = 97%). Therefore it can be concluded that integration of ALOS PALSAR and optical data improve the accuracies of land cover and forest mapping and hence suitable for environmental monitoring.