Met Office Global 10 km forecasts in GribStream

GribStream now includes the Met Office Global Deterministic 10 km forecast. The first release focuses on six surface fields: air temperature, dew point, relative humidity, mean-sea-level pressure, and 10 metre wind speed and direction.

That gives you another global forecast to query alongside GFS and ECMWF IFS, using the same coordinates, time filters and JSON or CSV responses. No new request format to learn.

Global coverage, not just the UK

The Met Office Global Deterministic model is a global configuration of the Unified Model, the Met Office's physics-based weather and climate modelling system. Its output covers the whole world on a 2,560 by 1,920 latitude–longitude grid, at approximately 10 km resolution.

For teams already comparing global forecasts, it adds a different forecasting system to the same workflow. You can examine differences in the timing of a temperature change, the passage of a pressure system, or the strength and direction of surface winds. Compare matching initialization times and lead times, and verify against observations: another model is useful evidence, not a guarantee of a better forecast at every location.

What is included first

The initial GribStream dataset is metofficeglobal10km.

Two details matter when comparing these values with other models. The Met Office screen-level fields are at 1.5 metres, not 2 metres. And relative humidity is a fraction: a returned value of 0.75 means 75%. Temperature is in kelvin; subtract 273.15 for Celsius. These field definitions and units follow the Met Office data sheet.

This first collection does not include precipitation, cloud, radiation or pressure-level fields. It also does not include the UKV regional model or the MOGREPS ensembles. The global deterministic forecast provides one model trajectory, not ensemble probabilities.

Four cycles daily, with longer forecasts twice a day

The source publishes 00, 06, 12 and 18 UTC cycles. The 00 and 12 UTC forecasts reach seven days, with hourly steps through +54 hours, three-hour steps from +57 to +144, and six-hour steps from +150 to +168. Forecast reference time is the model's starting time, not the moment it becomes downloadable: the source registry describes typical availability three to six hours later.

The 06 and 18 UTC runs are shorter. Our inspected source objects include the six selected fields through +69 hours, although the published summaries give different short-cycle endpoints. The model page records that distinction; applications should use the records actually available for their selected cycle.

A familiar point-forecast request

This example selects the 6 September 2026 00 UTC cycle at +54 hours for London. It uses the same /runs interface as other GribStream models. Set GRIBSTREAM_TOKEN to your API token before running it. These dates identify a verified historical cycle; choose a more recent cycle for current forecasts, allowing for the source publication delay.

curl --request POST \
  'https://gribstream.com/api/v2/metofficeglobal10km/runs' \
  --header "Authorization: Bearer $GRIBSTREAM_TOKEN" \
  --header 'Content-Type: application/json' \
  --header 'Accept: application/json' \
  --data '{
    "forecastedFrom": "2026-09-06T00:00:00Z",
    "forecastedUntil": "2026-09-06T01:00:00Z",
    "minHorizon": 54,
    "maxHorizon": 54,
    "coordinates": [
      {"lat": 51.5074, "lon": -0.1278, "name": "London"}
    ],
    "variables": [
      {"name": "air_temperature", "level": "1.5 m above ground", "alias": "temperature_k"},
      {"name": "wind_speed", "level": "10 m above ground", "alias": "wind_ms"}
    ]
  }'

Use /runs to compare specific forecast cycles. Use /timeseries when you want to select by valid time instead. The parameter catalog lists the supported selectors, and the API documentation explains the shared query options.

History and attribution

Indexed history covers forecast runs from 21 January 2026 at 12 UTC onward, for all six fields, subject to the upstream rolling two-year retention window. That starting cycle is the first supported by the current reader; older objects use a different layout. The model inventory describes the coverage. The initial historical backfill is complete. Our 8 September audit covered 920 cycles and confirmed that 52 source files were missing upstream, in the 6 July 12 UTC and 7 July 06 UTC cycles. Individual fields or forecast steps can therefore be absent. Source gaps and expiry still apply; this is not a permanent archive.

British Crown copyright 2023–2025, the Met Office. The source data is available under Creative Commons Attribution–ShareAlike 4.0. Reuse should retain appropriate attribution and follow the license's terms for adaptations. GribStream is independent of, and is not endorsed by, the Met Office. The provider describes this open-data feed as unsupported and advises against relying on it for critical business uses.

Which Met Office fields would be most useful in your workflow next? Tell us whether you need more global surface variables, upper-air fields, UKV detail or ensemble guidance.