Decouples weather forecasting from spatial resolution by using Flow Matching to super-resolve coarse trajectories as a post-processing step.
April 2, 2026
Original Paper
Super-Resolving Coarse-Resolution Weather Forecasts With Flow Matching
arXiv · 2604.00897
The Takeaway
High-resolution weather models are computationally ruinous to train. This modular framework allows for training cheap, coarse-resolution forecasters while achieving state-of-the-art 0.25° resolution skill, drastically lowering the barrier for climate modeling.
From the 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 f