Yuya Kawakami, Daniel Cayan, Dongyu Liu +2 more
Pluvial flooding is the kind that comes straight from rainfall, when water arrives faster than the ground and drains can take it, rather than from a river bursting or the sea coming in. It accounts for 45% of National Flood Insurance Program claims in the United States, and it is the hardest of the three to predict.
Part of the difficulty is scale. Existing approaches are either coarse, or restricted to one region, or built on process-based models too computationally heavy to run daily across a country. That leaves a gap precisely where the decisions are: everywhere, every day.
DELUGE is a multimodal deep learning framework aimed at that gap, predicting daily pluvial flood damage at roughly one kilometre resolution nationally. The emphasis on interpretable conditioning matters for the audience, because insurers and emergency planners have to justify decisions, and an unexplainable number is hard to act on however accurate it is.
Pluvial (rainfall-driven) flooding accounts for 45% of National Flood Insurance Program (NFIP) claims in the United States and is harder to predict than its riverine and coastal counterparts, with existing approaches limited to coarse resolution, regional domains, or computationally intensive process-based models unsuitable for daily continental-scale use. We present DELUGE, a multimodal deep learning framework for daily pluvial flood damage prediction at ~1 km resolution…
Neural spectroscopy of AlphaFold2 reveals encoded protein conformational landscapes
arXiv (cs.CL) · July 17, 2026Controlling Implicit Shortcut Reliance in L2 Spoken English Auto-markers
arXiv (cs.LG) · July 17, 2026Pick-to-Learn Calibration of an MPC Policy for an Origin-to-Destination Flight Problem
arXiv (cs.LG) · July 17, 2026Physics-Based Deep Spatiotemporal Hyperlocal Radar Nowcasting with a Multi-Variable U-Net for High-Resolution Precipitation Forecasting