Effect of Sensor-and Field-Based Parameters on Deforestation Mapping with Time-Series C-Band and L-Band SAR Data

Abstract

Deforestation mapping is an important application for keeping an account of carbon loss and sequestration. To continuously monitor forested areas, it is important to have a consistent time-series data set that is fulfilled by Synthetic Aperture Radar (SAR). Based on the temporal signature of the SAR backscatter, forests can be monitored, and abrupt changes in the backscatter can be studied and investigated. The wavelength of the SAR signals largely governs the interactions with the targets. In this study, we analyze the performance of the open-source C-band and L-band SAR datasets from Sentinel-1 (S-1) and ALOS-2/PALSAR-2 ScanSAR. We use a statistical change detection algorithm, Cumulative Sums of Change (CuSUM) to identify the point of change in the temporal datasets. An increase in L-band copol backscatter was observed after deforestation in peat swamp forests, attributed to enhanced double bounce scattering. In contrast, deforestation of mangroves resulted in a decrease in backscatter. Using the C-band SAR data, the VH polarization provides accurate change maps with an average overall accuracy of 0.76 (Kalimantan) and 0.8 (Mondah). The results show that the Dual-Pol Radar Vegetation Index (DPRVI) provides accurate and homogeneous change maps with L-band SAR data with an average overall accuracy of 0.58 (Kalimantan) and 0.62 (Mondah). © 2025 IEICE Electronics Society.