Generative artificial intelligence improves projections of climate extremes

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tie, Ruian, Zhong, Xiaohui, Shi, Zhengyu, Li, Hao, Chen, Bin, Liu, Jun, Libo, Wu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908588652363776
author Tie, Ruian
Zhong, Xiaohui
Shi, Zhengyu
Li, Hao
Chen, Bin
Liu, Jun
Libo, Wu
author_facet Tie, Ruian
Zhong, Xiaohui
Shi, Zhengyu
Li, Hao
Chen, Bin
Liu, Jun
Libo, Wu
contents Climate change is amplifying extreme events, posing escalating risks to biodiversity, human health, and food security. GCMs are essential for projecting future climate, yet their coarse resolution and high computational costs constrain their ability to represent extremes. Here, we introduce FuXi-CMIPAlign, a generative deep learning framework for downscaling CMIP outputs. The model integrates Flow Matching for generative modeling with domain adaptation via MMD loss to align feature distributions between training data and inference data, thereby mitigating input discrepancies and improving accuracy, stability, and generalization across emission scenarios. FuXi-CMIPAlign performs spatial, temporal, and multivariate downscaling, enabling more realistic simulation of compound extremes such as TCs.
format Preprint
id arxiv_https___arxiv_org_abs_2508_16396
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Generative artificial intelligence improves projections of climate extremes
Tie, Ruian
Zhong, Xiaohui
Shi, Zhengyu
Li, Hao
Chen, Bin
Liu, Jun
Libo, Wu
Atmospheric and Oceanic Physics
Artificial Intelligence
Climate change is amplifying extreme events, posing escalating risks to biodiversity, human health, and food security. GCMs are essential for projecting future climate, yet their coarse resolution and high computational costs constrain their ability to represent extremes. Here, we introduce FuXi-CMIPAlign, a generative deep learning framework for downscaling CMIP outputs. The model integrates Flow Matching for generative modeling with domain adaptation via MMD loss to align feature distributions between training data and inference data, thereby mitigating input discrepancies and improving accuracy, stability, and generalization across emission scenarios. FuXi-CMIPAlign performs spatial, temporal, and multivariate downscaling, enabling more realistic simulation of compound extremes such as TCs.
title Generative artificial intelligence improves projections of climate extremes
topic Atmospheric and Oceanic Physics
Artificial Intelligence
url https://arxiv.org/abs/2508.16396