Landcover classification using ERS SAR/INSAR data on coastal region of Central Sumatra

In this study, we present a classification method using both optical and SAR data in order to perform landcover classification. We investigate the use of a vector of information, composed of a serie of signatures derived from both SAR and INSAR data. First, different relevant parameters are derived from ERS-SAR data using multitemporal and interferometric analysis. Optical data is then used to define a training set in to order to perform a supervised classification. Our test site is located in a tropical area, in the coastal region of Central Sumatra, Indonesia. This site has a wide variety of landcover types. The region has been undergoing rapid deforestation with the logging of commercially exploitable timber and the conversion of forest to agricultural land. Our results demonstrate that ERS-1/2 tandem data is more suitable than the 35 days repeat pass in discriminating various landcover types. The results of classification using these techniques are compared with existing landuse maps and information derived from SPOT and LANDSAT data.