Improving tropical deforestation detection by fusing multiple SAR change measures

The paper studies tropical deforestation detection in Riau province, Indonesia with L-band ALOS PALSAR and Cband ENVISAT ASAR data. Multiple change measures as SAR image intensity, texture, and temporal variations of ScanSAR time series are extracted and employed for deforestation detection. These measures are then combined for improving the detection with subsequent performance evaluation by comparing with the World Wildlife Fund´s land cover maps as reference data. When applied on the FBD scene overlaid by both the PALSAR and ASAR data, the detection rate achieves up to 83.2%(at a false alarm rate of 20%) with a significant improvement of over 18% compared with the detection based on the single measure.