[2604.00897] Super-Resolving Coarse-Resolution Weather Forecasts With Flow Matching

Computer Science > Machine Learning arXiv:2604.00897 (cs) [Submitted on 1 Apr 2026] Title:Super-Resolving Coarse-Resolution Weather Forecasts With Flow Matching Authors:Aymeric Delefosse, Anastase Charantonis, Dominique Béréziat View a PDF of the paper titled Super-Resolving Coarse-Resolution Weather Forecasts With Flow Matching, by Aymeric Delefosse and 2 other authors View PDF HTML (experimental) 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 consist...