Skip to content
DraconiticResearch
Search Sources API
226,742 studies31 sourcesUpdated 36 min ago
PreprintCC BYarXiv (Cornell University)

WeatherNext 3: Increasing resolution and performance of global weather models with raw observations

13 more authors
  1. 1Google Research ,
  2. 2Google

Abstract

State-of-the-art AI weather models have shown impressive medium-range forecast skill and computational efficiency, but suffer two key shortcomings: their forecasts have lower spatial and temporal resolution than the best physics-based models and they are exclusively initialized with and trained on analysis data. As a result, they cannot directly make use of observations, and any biases in the analysis are inherited by the forecast. WeatherNext 3 addresses these shortcomings and establishes a new state-of-the-art for probabilistic medium-range forecasting skill. First, WeatherNext 3 generates new forecasts every hour (rather than every 6 hours like traditional global models) by ingesting low-latency geostationary satellite data. Second, WeatherNext 3's temporal and spatial resolution are on par with physics-based global models, with hourly time steps and 0.1 degree resolution for single-level variables, including solar radiation and cloud cover. Third, WeatherNext 3 moves beyond traditional analysis variables by learning to predict satellite-derived precipitation estimates, as well as tropical cyclone and station observations. Modelling sparse station data allows WeatherNext 3 to make 2m temperature and dewpoint predictions at any location and time, conditioned on local geographical features, with substantially lower error than competing global models, even when evaluated against unseen stations. Together, WeatherNext 3's capabilities move operational AI-based weather forecasting beyond emulating the traditionally distinct stages of data assimilation, forecasting and post-processing, which helps to further push the frontier of performance and granularity for global weather prediction.

Related

  1. On the Effective Resolution of AI Weather Prediction Models
    Tobias Selz, Anna Allén, Wessel P. Bruinsma +3
    Artificial Intelligence for the Earth SystemsPeer-reviewedLicence not stated6 citedAtmospheric science & meteorologyOpenAlex
  2. AIFS-DOP: End-to-End Medium-Range Weather Prediction from Observations Alone with Machine Learning
    Ewan Pinnington, Peter Lean, Mihai Alexe +8
    arXiv (Cornell University)PreprintCC BY-NC-NDFull text8 citedAtmospheric science & meteorologyOpenAlex
  3. Learning joint probabilistic weather forecasts from station observations alone
    Chaeyeon Yi, Yun Am Seo
    arXivPreprintCC BY 4.0Full textAtmospheric science & meteorologyStatisticsarXiv
  4. Detecting Weather Fronts from Satellite Observations: What Does Reanalysis Add?
    Ioannis Dravilas, Miruna Stoicescu, Roope Tervo +5
    arXivPreprintarXiv licenceFull textAtmospheric science & meteorologyarXiv
  5. Improving global precipitation forecasts with an AI weather model trained on satellite observations
    Julian F. Schmitt, Bertrand Delorme, Robert C. King +4
    arXivPreprintCC BY 4.0Full textAtmospheric science & meteorologyarXiv
BibTeX
@misc{rasp2026weathernext,
  title = {{WeatherNext 3: Increasing resolution and performance of global weather models with raw observations}},
  author = {Stephan Rasp and Boris Babenko and Dominic A. Masters and Andrew El-Kadi and Samier Merchant and Guy Shalev and Ilan Price and Fred Zyda and Rémi Lam and Sasha Shysheya and Matthew Willson and Stratis Markou and Shreya Agrawal and Suhani Vora and Mohammed Alewi Hassen and Sunny Mak and Tom R. Andersson and M. M. Bela and Akib Uddin and Nofar Peled Levi and Ben Gaiarin and Ferran Alet and Aaron Bell and Peter Battaglia and Álvaro Sánchez‐González},
  howpublished = {arXiv (Cornell University)},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.03582},
  url = {https://arxiv.org/abs/2609.03582},
  note = {CC BY},
}
RIS
TY  - UNPB
TI  - WeatherNext 3: Increasing resolution and performance of global weather models with raw observations
AU  - Stephan Rasp
AU  - Boris Babenko
AU  - Dominic A. Masters
AU  - Andrew El-Kadi
AU  - Samier Merchant
AU  - Guy Shalev
AU  - Ilan Price
AU  - Fred Zyda
AU  - Rémi Lam
AU  - Sasha Shysheya
AU  - Matthew Willson
AU  - Stratis Markou
AU  - Shreya Agrawal
AU  - Suhani Vora
AU  - Mohammed Alewi Hassen
AU  - Sunny Mak
AU  - Tom R. Andersson
AU  - M. M. Bela
AU  - Akib Uddin
AU  - Nofar Peled Levi
AU  - Ben Gaiarin
AU  - Ferran Alet
AU  - Aaron Bell
AU  - Peter Battaglia
AU  - Álvaro Sánchez‐González
PY  - 2026
DA  - 2026/09/03
PB  - arXiv (Cornell University)
DO  - 10.48550/arxiv.2609.03582
UR  - https://arxiv.org/abs/2609.03582
AB  - State-of-the-art AI weather models have shown impressive medium-range forecast skill and computational efficiency, but suffer two key shortcomings: their forecasts have lower spatial and temporal resolution than the best physics-based models and they are exclusively initialized with and trained on analysis data. As a result, they cannot directly make use of observations, and any biases in the analysis are inherited by the forecast. WeatherNext 3 addresses these shortcomings and establishes a new state-of-the-art for probabilistic medium-range forecasting skill. First, WeatherNext 3 generates new forecasts every hour (rather than every 6 hours like traditional global models) by ingesting low-latency geostationary satellite data. Second, WeatherNext 3's temporal and spatial resolution are on par with physics-based global models, with hourly time steps and 0.1 degree resolution for single-level variables, including solar radiation and cloud cover. Third, WeatherNext 3 moves beyond traditional analysis variables by learning to predict satellite-derived precipitation estimates, as well as tropical cyclone and station observations. Modelling sparse station data allows WeatherNext 3 to make 2m temperature and dewpoint predictions at any location and time, conditioned on local geographical features, with substantially lower error than competing global models, even when evaluated against unseen stations. Together, WeatherNext 3's capabilities move operational AI-based weather forecasting beyond emulating the traditionally distinct stages of data assimilation, forecasting and post-processing, which helps to further push the frontier of performance and granularity for global weather prediction.
SN  - 2331-8422
ER  -