Forest cover classification and change detection analysis using Alos Palsar Mosaic Data

The major setback of mapping forest cover using optical imagery for tropical countries is persistent cloud cover, which is also the case for Leyte Island, Philippines. Thus, Synthetic Aperture Radar (SAR) data acquired by the Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) sensor system was used to produce forest cover and forest cover change for years 2007 and 2010. Coconut palm, forest and non-forest/agriculture were the target classes classified to further explore separating coconut palm from forest, which is usually difficult for the said study area. Three different supervised classification algorithms were tested namely: Maximum Likelihood (MLC), Neural Network (NNC) and Support Vector Machine (SVM). The overall accuracies of the forest cover maps for 2010 using MLC, NNC and SVM were 75.63% (κ=0.63), 89.45% (κ=0.84) and 88.27% (κ=0.82), respectively.