PIFF: A Physics-Informed Generative Flow Model for Real-Time Flood Depth Mapping

TL;DR

PIFF is a novel physics-informed generative flow model that provides real-time flood depth mapping by integrating hydrodynamic priors, rainfall data, and topography, significantly improving flood prediction efficiency and accuracy.

Contribution

This paper introduces PIFF, a flow-based neural network that combines physics-informed constraints with deep learning for near real-time flood mapping, a novel approach in flood modeling.

Findings

PIFF accurately predicts flood depths across diverse rainfall scenarios.

The model outperforms traditional methods in speed and reliability.

Effective in a real-world case study in Tainan, Taiwan.

Abstract

Flood mapping is crucial for assessing and mitigating flood impacts, yet traditional methods like numerical modeling and aerial photography face limitations in efficiency and reliability. To address these challenges, we propose PIFF, a physics-informed, flow-based generative neural network for near real-time flood depth estimation. Built on an image-to-image generative framework, it efficiently maps Digital Elevation Models (DEM) to flood depth predictions. The model is conditioned on a simplified inundation model (SPM) that embeds hydrodynamic priors into the training process. Additionally, a transformer-based rainfall encoder captures temporal dependencies in precipitation. Integrating physics-informed constraints with data-driven learning, PIFF captures the causal relationships between rainfall, topography, SPM, and flooding, replacing costly simulations with accurate, real-time flood maps. Using a 26 km study area in Tainan, Taiwan, with 182 rainfall scenarios ranging from 24 mm to 720 mm over 24 hours, our results demonstrate that PIFF offers an effective, data-driven alternative for flood prediction and response.