Weather Lab

Weather Lab is an interactive website developed by Google DeepMind and Google Research for exploring AI weather forecasts. It provides a map-based interface for visualizing global forecast layers, tracking tropical cyclones with predicted and observed tracks, and downloading experimental cyclone forecast data.

Weather Lab

Global forecasts

Weather Lab visualizes gridded global weather layers from the following models:

Model Variables Time Step Forecast Type Horizon
WeatherNext 3 2m temperature, total precipitation, 10m wind speed, sea level pressure 1 hour Ensemble (64 members) — displays ensemble mean 15 days (6-hourly) / 48 hours (interim hourly)
WeatherNext 2 (FGN) 2m temperature, total precipitation, 10m wind speed, sea level pressure 6 hours Ensemble (64 members) — displays ensemble mean 15 days
Global MetNet Precipitation 15 min Deterministic 12 hours

Cyclone tracking

Weather Lab tracks tropical cyclones using WeatherNext Cyclones (Google's AI cyclone model) alongside ECMWF's ENS and HRES as independent baselines. Observed ground-truth tracks are sourced from NHC/JTWC TCVitals (real-time) and IBTrACS (historical). You can compare model predictions, view ensemble member tracks, and inspect deep-dive charts for individual storms.

For more on the research behind WeatherNext Cyclones, see the Nature publication and the blog post.

Download cyclone forecasts

You can download experimental cyclone track predictions from Weather Lab in CSV or ATCF format. Downloads include predicted tracks (position, intensity, timing), ensemble member tracks, and paired observed tracks for verification. Data is available from approximately 2023 to present.

Licensing

Data Type Definition License
Real-Time Data Relates to a time less than 1 hour ago and the future GDM Real-Time Weather Forecasting Experimental Data Terms of Use
Historical Data Relates to a time 1 hour ago or more Creative Commons Attribution 4.0 International (CC BY 4.0)

When real-time data relates to a time 1 hour ago or more, the license automatically changes to CC BY 4.0. For more details, see Terms of Service and Disclaimers.

Access global forecast data

For programmatic access to global forecast datasets, see the forecast data guides for BigQuery, Earth Engine, and Google Cloud Storage.