DiffScale: Continuous Downscaling and Bias Correction of Subseasonal Wind Speed Forecasts using Diffusion Models
Maximilian SpringenbergApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Noelia OteroApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Yuxin XueApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 G · 2025
Read DiffScale: Continuous Downscaling and Bias Correction of Subseasonal Wind Speed Forecasts using Diffusion Models by Maximilian SpringenbergApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Noelia OteroApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Yuxin XueApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 G (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-2503-23893.txt.
Description
Maximilian Springenberg and co-authors sharpen long-range wind forecasts with a diffusion model, the kind of generative system behind image makers. Forecasts two weeks to two months ahead matter for wind power planning, but they come out coarse and biased. Their model, DiffScale, redraws a forecast at finer detail, guided by weather patterns, and it can work at any zoom factor or lead time without retraining.
Instead of predicting step by step, the model estimates the odds of the forecast, which keeps it fast and flexible. In tests, it upgraded European weather center forecasts toward reference data and beat the baselines through the third week. Correcting the forecast's errors is part of the same pass. The authors pitch DiffScale as a versatile tool for the energy sector, where a better wind guess weeks ahead can move money and keep grids balanced.
Catalog record
| CREATOR | Maximilian SpringenbergApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Noelia OteroApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 Germany, Yuxin XueApplied Machine Learning GroupFraunhofer Heinrich-Hertz InstituteBerlin, 10587 G |
|---|---|
| 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/2503-23893