Regional tropical deforestation detection using ALOS PALSAR 50 m mosaics in Riau province, Indonesia

Abstract

ALOS PALSAR has provided a five-year archive of systematic global tropical forest coverage, released as slope-corrected and ortho-rectified 50 m mosaic products by JAXA, which are of great help to monitor forest changes. Based on the mosaics, a map of deforestation between the years 2007 and 2008 is built for Riau province, Indonesia, achieving a detection rate of over 86% at a false alarm rate of 20%, calculated by treating independent World Wildlife Fund (WWF) 2007 and 2008 maps as a reference. Non-forest is discriminated from forest by thresholding the HH/HV ratio in the PALSAR data. Deforestation is then estimated by detecting non-forest in the 2008 synthetic aperture radar (SAR) image (acquired close in time to the WWF 2008 reference map) within the region identified as natural forest in the WWF 2007 reference map. This reduces the errors caused by comparing the change detection between the bi-temporal SAR images with the WWF reference maps that are temporally mismatched with the SAR data, leading to more realistic estimates of the ability of PALSAR to detect deforestation than was previously reported.