High-Resolution Dynamic Flood Susceptibility Mapping Across the Mediterranean Region

Effective disaster prevention necessitates the production of high-resolution flood susceptibility maps (FSM) that accurately identify potential flood-prone areas. Conventional FSMs, however, provide static representations that overlook the inherent dynamicity of flood susceptibility, which is influenced by temporal variations, precipitation intensities, and other factors. Additionally, traditional FSMs often lack the high-resolution climate data required for precise risk assessment. To address these limitations, we propose a novel dynamic FSM approach that incorporates temporal variations and high-resolution climate data.

Our approach employs the Random Forest machine learning algorithm, trained on a comprehensive dataset of flooded and non-flooded areas (Global Flood Database v1). The algorithm considers seven critical factors influencing flooding events: elevation, slope, land cover, proximity to rivers, drainage density, soil moisture, and precipitation. This approach enables the generation of high-resolution (1 km) dynamic FSMs for the Mediterranean region, under varying seasonal conditions, precipitation intensities, and post-drought scenarios.

To assess and compare the model's performance, we employed both training and testing datasets, conducting evaluations using various metrics. The study results demonstrate the superior performance of the Random Forest model, establishing its efficacy as a robust tool for mapping dynamic flood susceptibility. The accuracy and reliability of the results obtained through this approach offer crucial insights for mitigating flood-related risks and enhancing disaster management strategies. This study is an integral part of the Open-Earth-Monitor Cyberinfrastructure (OEMC) project. As our next step, we aim to expand the application of our dynamic flood susceptibility mapping methodology to cover the European region.