Use of SAR and Optical Satellite Data for Land Use and Land Cover Classification in the Songkhla Lake Basin, Thailand

摘要

Land use/land cover classification is a mainstream application in remote sensing. Nevertheless, depicting accurately the land use/land cover spatial distribution in humid tropical areas is a challenging task, attributed mainly to the complex biophysical environment typically encountered in the tropics and technical limitations of earth observation data. In this paper, we used a medium resolution C-band, VV/VH dual polarized Sentinel 1A image and a multispectral Landsat 8 image to classify the Songkhla Lake Basin, Thailand in main land use/land cover classes, namely forest, paddy field, lake and plantations. The Maximum Likelihood supervised and ISO Cluster unsupervised algorithms have been applied. The results showed that all investigated approaches provided accurate estimations, with the most accurate derived from the Sentinel 1A image with ISO Cluster unsupervised classification method with overall accuracy 0.87% and kappa coefficient 0.89.