SAR based approach for rapid flood impact assessment on agriculture and human exposure

This study presents an innovative framework for rapid flood impact assessment on agriculture using Sentinel-1 Synthetic Aperture Radar (SAR) data at 10-meter resolution, integrated with Google Earth Engine (GEE). In September 2019, an exceptional flood event, characterized as a 32-year return period flood, impacted the northeastern provinces of Thailand, including Yasothon, Amnat Charoen, Ubon Ratchathani, Si Sa Ket, Roi Et, and Kalasin. This event, precipitated by the convergence of Tropical Storm PODUL and Tropical Depression KAJIKI, led to substantial human and economic losses. The methodology employs: (i) flood mapping with an enhanced Kittler-Illingworth thresholding technique incorporating terrain correction to mitigate slope-dependent backscatter variations; (ii) crop classification using a novel Dual Polarization SAR Vegetation Index (DpSVI); and (iii) robust validation against field data and optical imagery. The analysis encompassed flood extent determination, flooded agricultural area quantification, crop type distribution (Sugarcane, Cassava, Maize, and Rubber), assessment of crop type area affected by flooding, and evaluation of farmers’ exposure to flooding from September 1st to September 30th, 2019. Results indicate that 3890.4 km 2 (8.4% of the study area) was flooded, with Yasothon experiencing the highest agricultural impact (19.6% of its agricultural area). The framework demonstrates high accuracy (Overall Accuracy: 82.6%, Kappa: 0.94) and temporal generalizability through validation against 2018 and 2020 flood events. This approach enhances disaster response by providing reliable, cloud-independent flood and crop impact assessments, supporting targeted recovery strategies. • Used Sentinel-1 SAR data to map 2019 floods in Northeast Thailand’s six-province. • Addressed rapid flood impact assessment on agriculture, focusing on crop losses. • Developed novel Dual Pol SAR Vegetation Index (DpSVI) for crop type classification. • Quantified 826.07 km 2 flooded agricultural land, Yasothon hardest hit at 19.63%. • Google Earth Engine enabled scalable, cloud-free analysis, proposing monitoring systems.