From Global to Local: Efficient Regional Weather Downscaling with Global Weather Foundation Model

2026-07-03

AI Summary

This work addresses the challenge of regional weather forecasting, which demands both fine-scale detail and consistency with global dynamical constraints. Traditional approaches are computationally prohibitive, while existing learning-based methods often overlook physical and statistical mismatches across scales. The authors propose an efficient downscaling framework built upon a pretrained global weather foundation model, introducing lightweight multi-scale prediction heads in its latent space. This design enables high-resolution regional forecasts without retraining the backbone model. Notably, it is the first to leverage the latent space of a pretrained weather model for downscaling, circumventing the physical inconsistencies inherent in conventional super-resolution techniques and supporting resolution enhancements up to 100-fold. Experiments demonstrate superior performance over traditional numerical weather prediction on most metrics, substantially reduced computational cost, and effectively mitigated systematic biases as validated against ground observations.

Abstract

Accurate regional weather prediction requires resolving fine-scale structure while remaining consistent with global dynamics. Traditional limited area models rely on computationally expensive simulations, while many learning-based approaches frame the problem as super-resolution, overlooking statistical and physical mismatches across scales. We propose a foundation-model-driven downscaling framework that learns regional refinements of global forecasts by augmenting a pretrained weather model backbone with lightweight, multi-scale prediction heads operating directly in its latent space. Despite being trained on substantially coarser inputs, the pretrained backbone supports regional adaptation at resolutions corresponding to a two-order-of-magnitude increase in grid-cell resolution, without the need for retraining. The proposed approach uses regional numerical simulations as training targets and is evaluated not only against gridded datasets but also against ground-based weather station observations, enabling analysis of systematic biases between global reanalysis, regional simulations, and in-situ weather station observations. Our experiments show improved accuracy in comparison to NWP on most of the metrics at the fraction of computational cost. Moreover, we observe that building on a latent space of globally pre-trained weather foundation model offers better downscaling capabilities than the standard image-based super-resolution approaches.

Problem

• weather downscaling
• regional weather prediction
• global-to-local consistency
• multi-scale mismatch
• computational efficiency

Innovation

• weather foundation model
• regional downscaling
• latent-space refinement
• multi-scale prediction heads
• physics-consistent super-resolution