Change Detection over Tropical Peat Swamp Forests in Indonesia Using SAR Time-Series Data

Monitoring forest ecosystems requires both, continuous time-series datasets and a comprehensive method for analyzing the same. Synthetic Aperture RADAR (SAR) data aids the processing by maintaining an uninterrupted time-series dataset. The backscatter derived from SAR data is closely related to the changes taking place in the targeted area. Therefore the changes and disturbances taking place in the forests results in corresponding variations in the backscatter values. These temporal signatures of backscatter can be used for monitoring forests with statistical change detection algorithms. In this study, Cumulative Sums of Change (CuSum) and Sequential Omnibus algorithms are used with Sentinel-1 SAR data for deforestation mapping in Central Kalimantan, Indonesia. CuSum showed an overall better performance than the sequential omnibus algorithm. It was observed that the drop in backscatter on account of deforestation was significant in case of both polarization channels. However, cross-pol channel showed better results with an average overall accuracy of 0.74.