IPSL-AID: Generative Diffusion Models for Climate Downscaling from Global to Regional Scales

2026-03-23

arXiv.org

AI Summary

This study addresses the limitations of global climate models, which—due to their coarse resolution (150–200 km)—struggle to accurately represent key regional-scale climate processes, thereby hindering fine-grained climate decision-making. For the first time, the authors introduce a denoising diffusion probabilistic model (DDPM) for global-to-regional climate downscaling. Leveraging ERA5 reanalysis data and incorporating spatiotemporal contextual information, the method generates high-resolution (0.25°) fields of temperature, wind speed, and precipitation from coarse inputs, while enabling multi-scenario probabilistic outputs and rigorous uncertainty quantification. The approach faithfully reconstructs critical statistical characteristics—including extreme event frequencies, spatial structures, and power spectra—offering a novel downscaling paradigm that achieves high fidelity, physical consistency, and explicit uncertainty representation.

Abstract

Effective adaptation and mitigation strategies for climate change require high-resolution projections to inform strategic decision-making. Conventional global climate models, which typically operate at resolutions of 150 to 200 kilometers, lack the capacity to represent essential regional processes. IPSL-AID is a global to regional downscaling tool based on a denoising diffusion probabilistic model designed to address this limitation. Trained on ERA5 reanalysis data, it generates 0.25 degree resolution fields for temperature, wind, and precipitation using coarse inputs and their spatiotemporal context. It also models probability distributions of fine-scale features to produce plausible scenarios for uncertainty quantification. The model accurately reconstructs statistical distributions, including extreme events, power spectra, and spatial structures. This work highlights the potential of generative diffusion models for efficient climate downscaling with uncertainty

Problem

• climate downscaling
• high-resolution projections
• regional climate processes
• global climate models
• uncertainty quantification

Innovation

• diffusion models
• climate downscaling
• uncertainty quantification
• generative modeling
• high-resolution climate projections