Precipitation Estimation With NWP Model and Generative Diffusion Model

Abstract Recent advancements in state‐of‐the‐art generative deep‐learning models, particularly diffusion models, have significantly enhanced the capability to produce realistic and diverse synthetic images and videos. These advancements have had a profound impact on fields such as computer vision an...

Full description

Saved in:
Bibliographic Details
Main Authors: Haolin Liu, Jimmy C. H. Fung, Alexis K. H. Lau, Zhenning Li
Format: Article
Language:English
Published: Wiley 2025-04-01
Series:Geophysical Research Letters
Online Access:https://doi.org/10.1029/2024GL110625
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Recent advancements in state‐of‐the‐art generative deep‐learning models, particularly diffusion models, have significantly enhanced the capability to produce realistic and diverse synthetic images and videos. These advancements have had a profound impact on fields such as computer vision and natural language processing. In this study, we leverage this cutting‐edge generative model to refine Numerical Weather Prediction (NWP) precipitation outputs. By conditioning the generative model with fundamental meteorological variables simulated by the Weather Research and Forecasting model, we aim to reproduce the high‐resolution satellite precipitation product, specifically CMORPH. Benefiting from the superior ability of generative diffusion models to learn the distribution of target data, these models excel in providing detailed and accurate precipitation estimations over the raw NWP outputs and traditional predictive models. With this presented pipeline, we provide valuable insights and practical tools for refining precipitation forecasting while preserving its extremities and variability thus better guiding decision making regarding weather dependent activities.
ISSN:0094-8276
1944-8007