Probabilistic Downscaling for Flood Hazard Models

Abstract Riverine flooding poses considerable risks. Developing strategies to manage flood risks requires flood projections at decision-relevant scales, often at high spatial resolutions, and with solid uncertainty characterization. However, obtaining high-resolution flood projections can be computationally prohibitive. To address this challenge, we propose a probabilistic downscaling approach that maps low-resolution flood projections onto higher-resolution grids. The existing literature presents two distinct types of downscaling approaches: (1) probabilistic methods which are versatile and applicable across various physics-based models and (2) deterministic downscaling methods specifically tailored for flood hazard models. Both types of downscaling approaches come with their own set of mutually exclusive advantages. Here we introduce a new approach, PDFlood, that combines the advantages of existing probabilistic and flood hazard downscaling approaches: model-based uncertainty quantification and accurate point estimates from approximating physical processes. PDFlood provides (1) narrower prediction intervals and more accurate flooding probabilities compared to a state-of-the-art probabilistic downscaling approach and (2) comparably accurate point estimates and improved classification of cells as flooded versus non-flooded relative to a state-of-the-art flood hazard downscaling approach. By allowing users to consider both accurate point estimates and uncertainties, PDFlood can better inform the design of risk management strategies. While we develop PDFlood for flood hazard models, the general concepts translate to other applications such as wildfire models.