The following project has been adapted by M. Pyrina, 2026 from the following:

Generative diffusion-based downscaling for climate

Robbie A. Watt & Laura A. Mansfield https://arxiv.org/abs/2404.17752 using the implementation by T. Karras et al. (https://arxiv.org/abs/2206.00364) and code addapted from https://github.com/NVlabs/edm.

This repo contains code to go alongside "Joint Bias Correction and Downscaling of Subseasonal Forecasts via Diffusion Models" (2026) preprint. In this work, we apply a diffusion based model (DM) to a downscaling and bias correction task using ECMWF subseasonal temperature hindcast data and gridded temperature observations from MeteoSwiss over Switzerland.

File structure

• src_mean: contains code used to train, validate, and test the DM model using the ensemble mean (used in the manuscript)
• src_mems: contains code used to train, validate, and test the DM model using the ensemble members (not used in the manuscript, only partly tested)

Data

We are using ECMWF subseasonal hindcast data and gridded observational data from MeteoSwiss.

Dependencies

python>=3.9, torch, tensorboard, xarray, netcdf4, cartopy, matplotlib, scipy, numpy

Citation of original code

@misc{watt2024generative,
      title={Generative Diffusion-based Downscaling for Climate}, 
      author={Robbie A. Watt and Laura A. Mansfield},
      year={2024},
      eprint={2404.17752},
      archivePrefix={arXiv},
      primaryClass={physics.ao-ph}
}

Current work

@misc{pyrina2026_ch_downscaling,
      title={Joint Bias Correction and Downscaling of Subseasonal Forecasts via Diffusion Models}, 
      author={M. Pyrina, A. Imamovic, D. Büeler, C. Spirig, D. I. V. Domeisen},
      year={2026},
      eprint=xxx,
      archivePrefix={arXiv},
      primaryClass=xxx
}