Varda-single-1.0: deterministic data-driven weather forecasting at 1 km resolution over Switzerland's complex topography

@misc{pennino2026vardasingle10,
  title = {{Varda-single-1.0: deterministic data-driven weather forecasting at 1 km resolution over Switzerland's complex topography}},
  author = {Alberto Pennino and Francesco Zanetta and Michele Cattaneo and Claire Merker and Radi Radev and Jonas Bhend and Louis Frey and Hugues de Laroussilhe and Ophélia Miralles and Carlos Osuna and Daniele Nerini and Andreas Pauling and Daniel Hupp and Ulrich Hamann and Mary McGlohon and Marti Bosch and Luca Lanzilao and Marco Arpagaus and Lukas Jansing and Daniel Leuenberger and Mark A. Liniger and Katrin Ehlert and Matthew Chantry and Håvard Homleid Haugen and Gert Mertes and Ana Prieto Nemesio and Mario Santa Cruz and Jasper Wijnands and Gabriel Moldovan and Harrison Cook and Oliver Fuhrer},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.01835},
  eprint = {2610.01835},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.01835},
  note = {CC BY 4.0},
}

TY  - UNPB
TI  - Varda-single-1.0: deterministic data-driven weather forecasting at 1 km resolution over Switzerland's complex topography
AU  - Alberto Pennino
AU  - Francesco Zanetta
AU  - Michele Cattaneo
AU  - Claire Merker
AU  - Radi Radev
AU  - Jonas Bhend
AU  - Louis Frey
AU  - Hugues de Laroussilhe
AU  - Ophélia Miralles
AU  - Carlos Osuna
AU  - Daniele Nerini
AU  - Andreas Pauling
AU  - Daniel Hupp
AU  - Ulrich Hamann
AU  - Mary McGlohon
AU  - Marti Bosch
AU  - Luca Lanzilao
AU  - Marco Arpagaus
AU  - Lukas Jansing
AU  - Daniel Leuenberger
AU  - Mark A. Liniger
AU  - Katrin Ehlert
AU  - Matthew Chantry
AU  - Håvard Homleid Haugen
AU  - Gert Mertes
AU  - Ana Prieto Nemesio
AU  - Mario Santa Cruz
AU  - Jasper Wijnands
AU  - Gabriel Moldovan
AU  - Harrison Cook
AU  - Oliver Fuhrer
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2610.01835
UR  - https://arxiv.org/abs/2610.01835
AB  - We present Varda-single-1.0, a medium-range data-driven weather prediction system built for the Alpine domain. It provides hourly deterministic regional forecasts on a mesh of 1 km resolution and global forecasts on a 31 km mesh. The system comprises two independently trained stretched-grid Graph Transformer models with encoder-processor-decoder architecture, developed in the Anemoi framework: a 6-hourly autoregressive forecaster and a temporal downscaler reconstructing hourly forecasts between the forecaster's steps. Its training curriculum includes pre-training on ERA5 reanalysis data, followed by training on a 20-year kilometre-scale regional reanalysis, and finally fine-tuning on operational kilometre-scale analyses. Verified over one year against operational analyses and surface station observations, Varda-single is competitive with or improves on MeteoSwiss' operational numerical weather prediction baselines for most headline scores and variables. It broadly matches the skill of the high-resolution 1 km ICON-CH1-EPS control at lead times up to +33 h and generally outperforms the 2 km ICON-CH2-EPS control at lead times up to +120 h. Despite competitive aggregate scores, Varda-single underestimates some local wind maxima and produces overly smooth convective precipitation fields, consistent with the smoothing associated with squared-error training. To gain insight into the model's behaviour, we investigate three case studies beyond the aggregated headline scores, and find particular weaknesses in Varda-single's representation of local winds over complex terrain. Varda-single represents an important step in the development of high-resolution ML forecasting over complex terrain, in complementing the operational regional numerical weather prediction models of MeteoSwiss with data-driven models and in providing a pretrained model for researchers and user-specific applications.
ER  -