Improving Pluvial Flood Simulations with Multi-source DEM Super-Resolution

Abstract. Due to the limited availability of high-resolution topographic data, accurate flood simulation remains a significant challenge in many flood-prone regions, particularly in developing countries and in urban domains. While publicly available Digital Elevation Model (DEM) datasets are increasingly accessible, their spatial resolution is often insufficient for reflecting fine-scaled elevation details, which hinders the ability to simulate effectively pluvial floods in built environments. To address this issue, we implemented a deep learning-based method, which efficiently enhances the spatial resolution of DEM data, and quantified the improvement in flood simulation. The method employs a tailored multi-source input module, enabling it to effectively integrate and learn from diverse data sources. By utilizing publicly open global datasets, low-resolution DEM datasets (such as the 30 m SRTM) in conjunction with high-resolution multispectral imagery (e.g., Sentinel-2A), our approach allows to produce a super-resolution DEM, which exhibits superior performance compared to conventional methods in reconstructing 10 m DEM data based on 30 m DEM data and 10 m multispectral satellite images. Such superior performance translates, when applied to pluvial flood simulations, into significantly improved accuracy of floodwater depth and inundation area predictions compared to existing alternatives. This study underscores the practical value of machine-learning techniques that leverage publicly available global datasets to generate DEMs that allow enhancing flood simulations.