Global Weather Emulator Forecasts at 28 km Resolution

1979-2021 training data underpins an AI-based weather emulator generating twice-daily 10-day forecasts. The model provides six atmospheric variables on 13 pressure levels and four surface variables every six hours, with global coverage at a 0.25-degree grid resolution.

Use Cases

• Train or validate AI weather models using the six atmospheric variables on 13 pressure levels as a benchmark.
• Analyze large-scale temperature and horizontal wind fields for climate pattern research.
• Compare forecast accuracy of the emulator's surface variables against operational models.
• Study atmospheric state evolution using the 6-hourly temporal resolution of the provided variables.

Strengths

• Covers a 42-year training period from 1979 to 2021 using ERA5 and operational analyses.
• Provides global geographical coverage with a consistent 0.25-degree latitude-longitude grid.
• Delivers forecasts with a 10-day lead time, updated twice daily.

Limitations

• Data is an AI emulator's output, not direct observational measurements, introducing model bias.
• Horizontal resolution of 28 km may be insufficient for localized or mesoscale weather phenomena.

Provenance

Source

Environment and Climate Change Canada (ECCC)

Collection Method

AI model trained and refined by ECCC on ECMWF's ERA5 data and operational analyses.

Time Range

Training data spans 1979-2021; forecasts are ongoing.

Freshness

Forecasts are generated twice daily.

Geography

Global