A hybrid method coupling physical process-driven model with generative deep learning for probabilistic flood forecasting

doi:10.1016/j.jhydrol.2026.135319

A hybrid method coupling physical process-driven model with generative deep learning for probabilistic flood forecasting

2026-03-16

0

PRE

AI

DOI：10.1016/j.jhydrol.2026.135319

Original

Original request for help

Save

Abstract

En 中文

• A probabilistic flood forecasting framework that couples the process-driven model with a data-driven model is developed to improve the accuracy and reliability. • A conditional diffusion model is developed for implicitly learning the complex conditional distribution of forecast errors and estimating the range of error impact. • The hybrid models that use generative deep learning model components exhibit better performance compared to traditional models.

Keywords:

probabilistic flood forecasting

process-driven model

generative deep learning

conditional diffusion model

hybrid modeling

A hybrid method coupling physical process-driven model with generative deep learning for probabilistic flood forecasting

Abstract

Journal

Organization