Data Resolution and Model Complexity in Pluvial Flood Risk Assessment: Insights and Trade-Offs

Abstract

The increasing severity and frequency of urban pluvial floods has prompted numerous research studies on pluvial flood risk assessment. However, the literature reveals significant variability in the data resolutions and model types employed. In this study, we aim to provide quantitative insights on the data and modelling requirements for pluvial flood risk practitioners and scientists. We analyse combinations of various data resolutions, ranging from 2 to 50 m, and model complexities, including cellular automata (CA), 2D hydrological models, and coupled 2D–1D models. We evaluate each combination's performance against a benchmark, assessing accuracy in flood mapping, hazard, damage, and risk quantification. Our findings indicate that data resolution affects accuracy more than model complexity. Lower data resolutions lead to underpredictions in flood depth, which subsequently affect hazard, damage, and risk assessment. Thus, we recommend using a digital elevation model (DEM) resolution between 2 and 5 m for accurate flood modelling. With respect to model complexity, we show that combining CA models with high-resolution data (e.g., 2 m DEM) offers substantial computational time savings while maintaining high F-statistic values (between 0.76 and 0.78). Additionally, we found that all models were able to estimate relative residential damage with errors between +10% to −20% when combined with DEMs with 2 and 5 m resolution, while lower DEM resolutions were associated to higher errors (above −30%). The results presented in this study, and their discussion, serve as a guide in selecting appropriate models and data resolutions for pluvial flood risk modelling, and in understanding the loss in accuracy related to multiple model–data combinations.