Repository navigation
Bris
Last update: September 8, 2026
Bris is an experimental AI-based model under development. We are looking for feedback from interested users.
Overview • Available data streams • Files • Parameters • Known data issues • References • More information
Bris is MET Norway's operational data-driven weather prediction model. Unlike physics-based models such as MEPS, Bris is a neural network trained on historical gridded NWP analyses. By learning the transitions from one analysis to the next, Bris can predict the weather forward in time. The operational version of Bris closely resembles the work presented in Nordhagen et al. (2025) and Ingstad et al. (2026). Bris is a global model with a stretched grid, where the grid has high resolution (2.5 km) over our Nordic and Arctic focus areas and lower resolution (31 km) elsewhere.
Due to its computational efficiency, one advantage of Bris is that it is affordable to provide an ensemble of seamless high-resolution (2.5 km) forecasts 10 days into the future.
Data from Bris is available in our real-time stream, which contains the most recently produced (last 3 days) forecasts. Data in this folder is produced redundantly in two separate datahalls to ensure high availability.
At the moment, we have not set up archiving of operational runs. This will be added on the future. We also have plans to rerun Bris for an extended period in the past, allowing users to see it's predictive capability. These will be available in a rerun archive.
At the moment, we have made available Bris output for the Nordic domain. This is the same grid as the MEPS domain, except that a band of 50 grid points is removed around the whole domain. The files are available through https://thredds.met.no/thredds/catalog/metno.html under the folder called Other products/Bris. See the Data access page for information on how to download data.
Two types of files are produced:
| Filename | Update freq | Delay | Members | Lead times |
|---|---|---|---|---|
| bris_ens_2_5km_nordic_<timestamp>.nc | 3h | 2h 10m | 16 | Hourly to 90h |
| bris_ens_2_5km_6h_nordic_<timestamp>.nc | 12h | 7h 30m | 30 | 6-hourly to 258h |
where timestamp = {year}{month}{day}T{hour}Z (UTC or Zulu Time Zone), e.g. 20200601T12Z; Year is a 4-digit number (e.g. 2022); month is a 2-digit number (e.g. 02 for February); and day is a 2-digit number
An example of a HTTPS URL is: https://thredds.met.no/thredds/fileServer/brisnordiclatest/bris_ens_2_5km_nordic_20260908T12Z.nc.
An example OpenDAP URL is: https://thredds.met.no/thredds/dodsC/brisnordiclatest/bris_ens_2_5km_nordic_20260908T12Z.nc.
The hourly model runs for 00Z are initialized using 00Z MEPS and 18Z IFS forecasts. The 6h model runs for 00Z are initialized using 00Z MEPS and 00Z IFS analyses.
Bris is a full 3D atmospheric model, which includes surface parameters as well as parameters on pressure levels. At the moment, we do not output all parameters in the model, focusing instead of parameters frequently requested by end users. These are described below.
| NetCDF name | Description | Units |
|---|---|---|
| air_pressure_at_sea_level | Instantaneous air pressure reduced to sea level | pa |
| air_temperature_0m | Instantaneous temperature at 0 m | K |
| air_temperature_2m | Instantaneous temperature at 2 m | K |
| altitude | Altitude above sea-level of each grid cell | m |
| cloud_area_fraction | Fraction of sky covered by cloud | 1 |
| dew_point_temperature_2m | Instantaneous dew point temperature at 2 m | K |
| high_type_cloud_area_fraction | Cloud cover higher than 5000m above the ground | 1 |
| integral_of_surface_downwelling_ shortwave_flux_in_air_wrt_time* |
The time integral of incoming solar radiation at the surface over the hour leading up to the timestamp. The units are therefore W/m^2 s | W s m⁻² |
| integral_of_surface_downwelling_ longwave_flux_in_air_wrt_time* |
The time integral of long wave radiation at the surface over the hour leading up to the timestamp. The units are therefore W/m^2 s | W s m⁻² |
| land_area_fraction | Fraction of each grid cell that is land (not sea) | 1 |
| low_type_cloud_area_fraction | Cloud cover lower than 2000m above the ground | 1 |
| medium_type_cloud_area_fraction | Cloud cover between 2000 and 5000m above the ground | 1 |
| precipitation_amount_acc | Precipitation accumulated from the beginning of the forecast. | mm |
| relative_humidity_2m | Instantaneous relative humidity at 2 m | 1 |
| surface_air_pressure | Instantaneous air pressure at surface | pa |
| x_wind_10m | 10 minute average x-component of wind speed at 10m | m/s |
| y_wind_10m | 10 minute average y-component of wind speed at 10m | m/s |
(*) "integral_of_Y_wrt_X" means int Y dX. The data variable should have an axis for X specifying the limits of the integral as bounds. "wrt" means with respect to. The surface called "surface" means the lower boundary of the atmosphere. "shortwave" means shortwave radiation. Downwelling radiation is radiation from above. It does not mean "net downward". Surface downwelling shortwave is the sum of direct and diffuse solar radiation incident on the surface, and is sometimes called "global radiation". When thought of as being incident on a surface, a radiative flux is sometimes called "irradiance". In addition, it is identical with the quantity measured by a cosine-collector light-meter and sometimes called "vector irradiance". In accordance with common usage in geophysical disciplines, "flux" implies per unit area, called "flux density" in physics.
M. S. Ingstad, M. C. A. Clare, O. Ersland, V. Gahlen, H. H. Haugen, O. Miralles, E. M. Nordhagen, T. N. Nipen, I. A. Seierstad, J. B. Bremnes, M. Maier-Gerber, Z. B. Bouallègue, H. Cook, C. Lessig, G. Mertes, C. O'Brien, F. Pinault, A. P. Nemesio, and M. Chantry, 2026: HourGlass: A Probabilistic Data-Driven Temporal Downscaler for Global and Regional Weather Forecasting. Preprint on arXiv, arXiv:2607.11457.
E. M. Nordhagen, H. H. Haugen, A. F. S. Salihi, M. S. Ingstad, T. N. Nipen, I. A. Seierstad, I.-L. Frogner, M. C. A. Clare, S. Lang, M. Chantry, P. Dueben, and J. Kristiansen, 2025: High-Resolution Probabilistic Data-Driven Weather Modeling with a Stretched Grid. Preprint on arXiv, arXiv:2511.23043.
You may find more information in the open issues or closed issues. If you want to contact us for something related to the dataset, we encourage you to open an issue. If you have any questions regarding the Bris datasets please use this email address: bris[at]met[dot]no.
Copyright Norwegian Meteorological Institute