Leveraging Crowd-sourced Social Sensing for Enhancing Remote Sensing-Based Real time Flood Mapping. Case Study of The Nile Nyabarongo Lower Level I, Rwanda

Abstract

Flooding is a common natural catastrophe that can devastate communities by inflicting structural damage, loss of life, interruption of key services, and environmental degradation. In May 2023, Rwanda experienced intense flooding with notable socio-economic impacts, coinciding with peak rainfall in the study area as indicated by CHIRPS 5-day satellite data. This study assessed the flood extent, socioeconomic infrastructures and utilities at flood risk in the Nile Nyabarongo Lower Level I Catchment by integrating Sentinel-1 Synthetic Aperture Radar (SAR) data processed in Google Earth Engine with crowdsourced social sensing images. Flood detection employed an image differencing technique with Otsu’s automatic thresholding, while social media images were filtered, geotagged, and overlaid with SAR-derived maps. The results showed an inundated area of approximately 21.34 km², with VV polarization slightly outperforming VH, and strong agreement with UNOSAT reference data. Twelve unique geotagged images and four manually geolocated ones confirmed the SAR findings. Exposure analysis revealed that several critical infrastructures, including roads, water treatment plants, other utilities and socio-economic facilities, were at risk. The study demonstrates that combining satellite imagery with crowdsourced data strengthens real-time flood mapping and delivered capabilities of socio sensing platforms in real time flood mapping, particularly in data-scarce regions.