A Comprehensive Framework for Multisource Spatiotemporal Fusion of Optical and SAR Images for Flood Mapping in a Cloud-Based Environment Greetta Pinheiro (*a), Sonajharia Minz(b)
a) Assistant Professor, Faculty of School of Computer Science Engineering & Technology, Bennett University, India
*greetz.pinheiro[at]gmail.com
b) Professor, Faculty of School of Computer and System Sciences, Jawaharlal Nehru University, India
Abstract
Floods are increasing in frequency and severity due to climate change, posing a major challenge to sustainable development. While satellite remote sensing is a powerful tool for disaster monitoring, relying on single sensors presents significant performance bottlenecks. Optical imagery is often obscured by the cloud cover and heavy rainfall typical of flood events, whereas Synthetic Aperture Radar (SAR) data, despite its all-weather capability, can suffer from over-classification of permanent water bodies, leading to inaccurate flood extent maps. This research proposes a novel and scalable framework for high-accuracy flood mapping by leveraging the complementary strengths of Sentinel-1 SAR and Sentinel-2 optical imagery within the Google Earth Engine (GEE) cloud platform. Our automated, two-stage methodology first employs a synergistic combination of a SAR change detection index and Edge Otsu segmentation on Sentinel-1 data for rapid initial flood detection. This is followed by a critical refinement stage that integrates the JRC Global Surface Water dataset to accurately mask permanent water bodies, thereby correcting over-classification errors and DEM-based slope and elevation masks to eliminate false positives. The entire workflow is designed for seamless implementation in GEE and subsequent integration into QGIS for local-level analysis and decision support. This framework provides an efficient, accessible, and precise tool for policymakers and first responders, contributing directly to the UN Sustainable Development Goal 13.
