BOSTON, Nov. 17, 2026 /PRNewswire/ -- AI weather models can generate global forecasts in minutes, but they still need an estimate of the current atmosphere for initialization. Many rely on conventional data assimilation systems that typically update the starting picture every six hours.
Tomorrow.io's Sky 1 uses machine-learning data assimilation to update initial conditions hourly, incorporating fresh observations from our constellation of 15 microwave sounders. These instruments sense atmospheric temperature and moisture, including through clouds, adding observations over oceans and other sparsely monitored regions.
In 28 test cases, adding Tomorrow.io's observations reduced initial-state water-vapor error by approximately 5–8% compared with Sky 1 without them, using ERA5, a reconstruction of past weather, as the reference. These improved initial conditions also benefited forecasts from leading AI weather models. Across 112 forecasts from Aurora 1.5 initialized by Sky 1, the observations lowered average near-surface temperature error at weather stations at every evaluated lead time through five days. Forecast gains were modest and strongest in the first two days. A similar pattern was observed in Google DeepMind's WeatherNext 2 with the sounders assimilated. Started from Sky 1 instead of raw GFS data, Aurora 1.5 gained about a day of lead time on near-surface temperature, scored against ERA5.
Through Sky 1, the growing fleet and improved instrument calibration are being evaluated across leading AI weather models, including their impact on hurricane tracks, atmospheric river landfalls and severe weather outbreaks. By providing updated initial conditions every hour, Sky 1 enables these fast-running models to bring newly observed atmospheric changes into the next forecast sooner.
Resources
- The full piece, with the method behind every number: Tomorrow.io gives Google DeepMind's WeatherNext 2 and Microsoft's Aurora 1.5 a better start
- Images and figures: on request, through the media contact below
- Interviews with Shimon Elkabetz, CEO, and Forest Cannon, chief weather officer: through the media contact below
Data available upon request.
About Tomorrow.io
Tomorrow.io is the world's leading weather intelligence platform, helping organizations and individuals predict, manage, and adapt to climate and weather risks in real time. Powered by next-generation satellite technology, proprietary AI, and global atmospheric models, Tomorrow.io delivers actionable weather insights to enterprises, governments, and communities worldwide.
Media Contact
[email protected]
SOURCE Tomorrow.io

Share this article