This study investigates a method using multi-temporal, multi-polarization ENVISAT (Environmental Satellite) Advanced Synthetic Aperture Radar (ASAR) data for mapping an agricultural area in a sub-tropical cloud-prone and rainy area of Pearl River Delta in south China. A total of six
scenes of HH (radio waves transmitted and received in horizontal polarization) and VV (radio waves transmitted and received in vertical polarization) polarization ASAR data acquired from March to November 2006 were used for land cover classification. Meanwhile, four field surveys of 320 test
sites were carried out simultaneously with ASAR image acquisition in May, July, September and November 2006. A decision tree classifier is used to classify seven main types of land cover features including sugarcane, banana fields, lotus ponds, fish ponds, mangrove wetland, seawater and buildings.
As a result, a classification map of Nansha Island was generated with overall accuracy of 80% and a kappa coefficient of 77%. The results show that the multi-temporal and multi-polarization ASAR images can have good performance in separating the basic land cover categories in a sub-tropical
cloud-prone and rainy area. The decision tree classifier is also approved to work efficiently on satellite SAR images with good classification accuracy. The analysis to get the best combination of radar scenes for the decision tree also proves that multi-temporal radar backscatter information
received in the crop growth period is important in improving classification accuracy.
Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Room 615, Esther Lee Building, Shatin, Hong Kong,Department of Atmospheric and Oceanic Science, University of Maryland, College Park, MD, USA 2:
Department of Atmospheric and Oceanic Science, University of Maryland, College Park, MD, USA 3:
School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou, PR China