Forest Discrimination Analysis of Combined Landsat and ALOS-PALSAR Datasets

The joint processing of remote sensing data acquired from sensors operating at different wavelengths has the potential to significantly improve the operation of global forest mapping and monitoring systems. This paper presents an analysis of the forest discrimination properties of Landsat TM and ALOS-PALSAR data when considered as a combined source of information. This study is carried out over a test site in north-eastern Tasmania, Australia. Canonical variate analysis, a directed discriminant technique, is used to investigate the separability of a number of training sites, which are subsequently used to define spectral classes as input to maximum likelihood classification. An accuracy assessment of the classification results is provided on the basis of independent ground validation data, for the Landsat, PALSAR, and combined SAR–optical data. The experimental results demonstrate that: 1) considering the SAR and optical sensors jointly provides a better forest classification than either used independently, 2) the HV polarisation provides most of the forest/non-forest discrimination in the SAR data, and 3) the respective contribution of each of the Landsat and PALSAR bands to the separation of different types of forest and nonforest land covers varies significantly.