EOS-04 and S-1 for Mapping the Deforestation in Sub-Himalayan Forests in India

Abstract

The tropical and sub-tropical forests in the foothills of the Himalayas are under threat of deforestation and degradation. With launch of India's EOS-04 C-band SAR, it is important to assess its capabilities to monitor these forests. This study uses multi-sensor C-band SAR data from EOS-04 and S-1 for deforestation mapping across multiple forests in India. To understand the performance of these two datasets in a controlled environment with a sufficient number of logging events within a two-year duration, the analysis is carried out in the managed forests of Haldwani, India. Change maps were generated with a statistical change detection algorithm - Cumulative Sums of Change. The VH (S-1) and HV (EOS-04) cross-pol channels were found to give comparable results with values of overall accuracy at 0.71. The kappa coefficient and F1 Score were 0.58 & 0.65 (VH S-1), and 0.64 & 0.54 (HV EOS-04), respectively. However, it was observed that the detected change areas were better demarcated using S-1 data with preserved compartment boundaries and homogeneous enclosed area. The impact of inaccurate Forest/Non-forest mask resulted in grazed grasslands classified as deforested area with S-1 data. The findings of this study highlight the impact of satellite datasets, polarization channel, and spatial resolution for deforestation mapping. The results show that EOS-04 and Sentinel-1 based change maps have comparable accuracy depending on the temporal data availability. © 2025 Elsevier B.V., All rights reserved.