[2604.03459] Physics-Constrained Adaptive Flow Matching for Climate Downscaling

Physics > Atmospheric and Oceanic Physics arXiv:2604.03459 (physics) [Submitted on 3 Apr 2026] Title:Physics-Constrained Adaptive Flow Matching for Climate Downscaling Authors:Kevin Debeire, Aytaç Paçal, Pierre Gentine, Luis Medrano-Navarro, Nils Thuerey, Veronika Eyring View a PDF of the paper titled Physics-Constrained Adaptive Flow Matching for Climate Downscaling, by Kevin Debeire and 4 other authors View PDF HTML (experimental) Abstract:Regional climate information at kilometer scales is essential for assessing the impacts of climate change, but generating it with global climate models is too expensive due to their high computational costs. Machine learning models offer a fast alternative, yet they often violate basic physical laws and degrade when applied to climates outside of their training distribution. We present Physics-Constrained Adaptive Flow Matching (PC-AFM), a generative downscaling model that addresses both problems. Building on the Adaptive Flow Matching (AFM) model of Fotiadis et al. (2025) as our baseline, we add soft conservation constraints that keep the downscaled output consistent with the large-scale input for precipitation and humidity, and use gradient surgery via the ConFIG algorithm to prevent these constraints from interfering with the generative objective. We train the model on Central Europe climate data, evaluate it on a 10-time downscaling task (63km to 6.3km) over six variables (near-surface temperature, precipitation, specific humidity,...