Global 20km Gridded Probabilistic Weather Forecasts from Met Office IMPROVER

Met Office IMPROVER Blended Probabilistic Forecast system provides global weather predictions on a 20km resolution grid. The data includes 31 parameters, such as precipitation, wind, and cloud cover, with hourly forecasts up to 120 hours. This calibrated ensemble forecast is available on AWS under a CC BY-SA license.

Use Cases

• Training ML models for extreme weather event prediction based on probabilistic exceedance thresholds.
• Analyzing global climate patterns using the 31 available weather parameters on a consistent 20km grid.
• Calibrating local forecast models by downscaling the global gridded probability data.
• Conducting risk assessment for sectors like aviation and shipping based on wind and visibility probabilities.

Strengths

• Global spatial coverage with a consistent approximately 20-kilometer grid resolution.
• Includes 31 distinct weather parameters, from precipitation to UV index.
• Forecasts are probabilistically calibrated and verified using spread-skill and reliability checks.
• Provides hourly timesteps for the first 120 hours of the forecast period.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• Timestep availability varies significantly for some parameters, requiring checks of external documentation.

Provenance

Source

Met Office

Collection Method

Derived from operational Numerical Weather Prediction ensembles and nowcasts, blended and calibrated using the IMPROVER pipeline.

Time Range

Forecast lead times from 0 to 192 hours, with varying timestep frequency.

Freshness

Update frequency is unknown; the description notes parameters and timesteps may change after April 2026.

Geography

Global coverage.