Super-Resolving Coarse-Resolution Weather Forecasts With Flow Matching

2026-04-01

AI Summary

High-resolution numerical weather prediction remains challenging due to the high computational cost of balancing accuracy and efficiency. This work proposes a modular framework that decouples coarse-resolution forecasting from generative super-resolution, introducing flow matching—a technique previously unexplored in meteorological downscaling—as a post-processing step. By training a residual-based stochastic inverse problem model on reanalysis data, the method reconstructs physically plausible small-scale variability while preserving large-scale structures inherent in the coarse forecasts. Evaluated at 0.25° resolution, the approach achieves probabilistic forecast skill comparable to operational ensemble systems, yet with substantially lower training costs than end-to-end high-resolution models.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Multimodal Learning
• Reasoning under Uncertainty: Probabilistic Programming

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Machine learning-based weather forecasting models now surpass state-of-the-art numerical weather prediction systems, but training and operating these models at high spatial resolution remains computationally expensive. We present a modular framework that decouples forecasting from spatial resolution by applying learned generative super-resolution as a post-processing step to coarse-resolution forecast trajectories. We formulate super-resolution as a stochastic inverse problem, using a residual formulation to preserve large-scale structure while reconstructing unresolved variability. The model is trained with flow matching exclusively on reanalysis data and is applied to global medium-range forecasts. We evaluate (i) design consistency by re-coarsening super-resolved forecasts and comparing them to the original coarse trajectories, and (ii) high-resolution forecast quality using standard ensemble verification metrics and spectral diagnostics. Results show that super-resolution preserves large-scale structure and variance after re-coarsening, introduces physically consistent small-scale variability, and achieves competitive probabilistic forecast skill at 0.25° resolution relative to an operational ensemble baseline, while requiring only a modest additional training cost compared with end-to-end high-resolution forecasting.

Problem

• super-resolution
• weather forecasting
• spatial resolution
• computational cost
• stochastic inverse problem

Innovation

• flow matching
• generative super-resolution
• stochastic inverse problem
• modular forecasting
• reanalysis-based training