Generating ensembles of spatially-coherent in-situ forecasts using flow matching
David Landry 1, Claire Monteleoni 1,2 and Anastase Charantonis 1 · 2025
Read Generating ensembles of spatially-coherent in-situ forecasts using flow matching by David Landry 1, Claire Monteleoni 1,2 and Anastase Charantonis 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-03463.txt.
Description
David Landry, Claire Monteleoni, and Anastase Charantonis generate ensembles of station weather forecasts that hang together across space, in a 2025 paper. Postprocessing fixes raw model forecasts at real observation sites. Older fixes treat each station alone, scrambling patterns between neighbors.
Their method, FMAP, learns with a flow matching generative model. A spatial attention backbone keeps nearby stations consistent. One training run serves all forecast lead times. From a few numerical runs it can spin out many plausible forecasts cheaply. On a European benchmark it improved single station accuracy too. Tests covered temperature and wind gusts up to five days ahead.
Catalog record
| CREATOR | David Landry 1, Claire Monteleoni 1,2 and Anastase Charantonis 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-03463