Flood Mapping Methodologies in Google Earth Engine Using Optical and Radar Data: A Comparative Study

Abstract

Floods are among the most severe and frequent natural disasters, impacting numerous countries worldwide. This study investigates flood mapping methodologies utilizing Google Earth Engine (GEE) with Sentinel-1, Sentinel-2, and Landsat data, focusing on the January 2021 Tetouan flood in Morocco. Three approaches were assessed: Sentinel-1 thresholding and NDWI (Normalized Difference Water Index) methods applied to Sentinel-2 and Landsat imagery. The analysis revealed flooded areas of 891 hectares (Sentinel-1), 814 hectares (Sentinel-2), and 1237 hectares (Landsat), validated against ArcGIS (Geographic Information System) results estimating 900 hectares. Sentinel-1 demonstrated superior accuracy with only a 9-hectare deviation and proved effective under cloudy conditions. Sentinel-2 provided a balance between spatial resolution and error levels, with moderate commission and omission errors. Landsat detected the largest flood extent but exhibited a slight overestimation. The study emphasizes the advantages of GEE’s cloud-based platform, which significantly reduced processing time, facilitating rapid flood extent mapping. This scalability and efficiency make GEE an invaluable tool for disaster management. The results underline the potential of these methodologies for accurate and timely flood monitoring, enabling informed decision-making in resilience planning and emergency response. Such advancements are critical for mitigating the impacts of flooding and supporting sustainable disaster management strategies in vulnerable regions worldwide.