Mapping of flood inundated areas using earth observation data and cloud computing platform

Abstract

Floods pose a significant threat, not only impacting human lives and property, but also causing extensive damage to critical infrastructure. Rapid flood impact assessment methods require comprehensive, updated and accurate flood maps to deliver the dependable information for effective disaster risk management. In this study, the flood-inundated areas were mapped and monitored in Punjab (India) using dual-polarized (VH and VV) Sentinel-1 (A and B) images of July–August 2023 and the impact of flood on cropped area was assessed using Geographic Information System (GIS). The satellite images were processed in Google Earth Engine (GEE) platform to extract the spatial extent of flood-inundated areas. The flooded areas were validated with limited ground truth. The district wise zonal statistics was derived to find the flood inundated areas. Results showed that the backscattering coefficient (σ0) of the flooded area ranged from −4 to −20 dB for VH polarization and from −4 to −30 dB for VV polarization. These threshold limits were used to extract the flood pixels followed by masking of permanent water bodies. The water extent reached its peak from 11 to July 17, 2023 (first flood period) and again from 14 to August 17, 2023 (Second flood period) in Punjab with 5.52 % of the area inundated during first flood period and 2.65 % of the cropped area inundated during the second flood period. The overall accuracy was 91.7 % with Kappa Coefficient of 0.83 in VV polarization and 86.2 % with Kappa Coefficient of 0.73 in VH polarization. Paddy was mostly inundated by flooding in the areas with lower elevation, high runoff and along the river breeches. These results suggest that spatio-temporal mapping of flood inundation is important for policy makers towards flood risk preparedness, raising awareness, designing the mitigation strategies and implementation of flood management planning to protect nature and human life.