Find out about us: our forecasts

The following forecasts are produced operationally at ECMWF:

Medium-range: global atmospheric model coupled to ocean wave model

• Forecasts to ten days from 00 and 12 UTC at 16 km resolution and 91 levels.
• 51-member ensemble forecasts to 15 days from 00 and 12 UTC at 31 km resolution to day 10 and 62 km to day 15 and 62 levels.
• Global Ocean forecasts to ten days from 00 and 12 UTC at 28 km resolution; European waters wave forecast to five days from 00 and 12 UTC at 10 km resolution.

Products:  http://www.ecmwf.int/products/forecasts/d/charts/medium/
Documentation: (below for summary)

Extended range (monthly): atmosphere-ocean coupled model

Products: http://www.ecmwf.int/products/forecasts/d/charts/mofc_multi/forecast/
Documentation: http://www.ecmwf.int/research/monthly_forecasting/Documentation.html

Long range (seasonal): atmosphere-ocean coupled model

The seasonal forecast is run once per month as a 51-member ensemble with a forecast range of seven months. The initial conditions are for the 1st day of the month; the forecast is released on 8th. The atmospheric resolution is 80 km, on 91 levels; the ocean model has horizontally-varying resolution (0.3° to 1.4°) and 42 levels.

Products: http://www.ecmwf.int/products/forecasts/d/charts/seasonal/forecast/seasonal_range_forecast/
Documentation:  http://www.ecmwf.int/products/forecasts/seasonal/documentation/

MEDIUM AND EXTENDED RANGE FORECASTING SYSTEM (4-32 DAYS)

Data assimilation

The operational data assimilation is four-dimensional variational multi-variate analysis (4D-Var) for wind, temperature, humidity, surface pressure and ozone. Increment minimisation (inner loop) is run in three stages, one at T159 (125 km) followed by two more at T255 (80 km) resolution on 91 model levels. The operational configuration comprises two 6-hour 4D-Var analyses, centred at 00 and 12 UTC, from which the main deterministic and EPS forecast are initialised. In addition there is a continuous 12-hour 4D-Var analysis cycle (observations from 09–21 UTC and 21–09 UTC) that runs several hours after observation time to allow the maximum possible number of observations to be used. Short forecasts from these analyses are used as the background fields for the main 00 and 12 UTC 6-hour assimilations.

Surface parameters: Sea surface temperature analysis from the UK Met Office and sea ice analysis from the EUMETSAT Ocean and Sea Ice SAF; soil moisture (Optimum Interpolation); snow depth; screen level temperature and humidity.

Model

All components of the ECMWF data assimilation and forecasting system use the Integrated Forecasting System (IFS) computer model.

Specification

• Variables (recalculated at each time-step): Wind, temperature, humidity, cloud fraction and water/ ice content, ozone content, pressure (at surface grid-points).
• Numerical scheme: Semi-Lagrangian, semi- implicit time-stepping formulation.
• Horizontal grid: wind and temperature are held as spectral fields. The grid for computation of physical processes is a reduced, linear Gaussian grid.
• Vertical grid: hybrid levels.
• Physics: Orography (terrain height and sub-grid-scale characteristics); four surface and sub-surface levels (allowing for vegetation cover, gravitational drainage, capillarity exchange, surface and sub-surface runoff); stratiform and convective precipitation, snow-fall, snow-cover and snow melt; radiation (incoming short-wave and out-going long-wave); sub-grid-scale orographic drag, gravity waves and blocking effects; non-orographic gravity wave effects; evaporation, sensible and latent heat flux; simplified ozone chemistry.

Ensemble Prediction System (EPS)

The EPS uses the same forecast model as used for the deterministic forecast but at lower resolution, and runs twice per day as the high-resolution system. As in the high-resolution system, initial SST anomalies are persisted for the first 10 days of the forecast. From day 10 onwards, for forecasts started from 00 UTC analyses, the atmospheric model is coupled to a 29 level ocean model that is also used for seasonal forecasts.

The ensemble comprises one control forecast (starting from the un-perturbed operational analysis interpolated to the EPS resolution) and 50 members starting from perturbed initial conditions. Initial perturbations are from a combination of low-resolution (T42L62) initial singular vectors (SV) and perturbations from an ensemble of data assimilations (EDA). 50 SV are used in the extra-tropics, plus up to 30 SV in the vicinity of Tropical Cyclones. The EDA consists of an ensemble of ten independent lower-resolution (T399L91, corresponding to 50km) 4D-Var assimilations that differ from the operational analysis by perturbing observations and sea surface temperature fields and by using the SPPT scheme (see below) in the non-linear integrations.

The EPS also includes a stochastic scheme to simulate the effect of model error: random perturbations of the physical tendencies (SPPT scheme) are applied to the perturbed forecasts.

The EPS re-forecast suite

Some of the EPS products (e.g. the Extreme Forecast Index and the weekly anomaly forecasts of the 32-day extended EPS) are constructed by comparing the EPS forecast with the model climate. The EPS model climate is estimated using the EPS re-forecast suite, which includes a 5-member ensemble run for 32-days with initial conditions every Thursday at 00UTC for the past 18-years. The EPS re-forecast suite uses the same model cycle and resolution as the operational EPS. Its un-perturbed initial conditions are defined by ERA-interim analyses, and its initial conditions are constructed using initial-time SVs computed from the ERA-interim analysis. Since there is no EDA for the past years (the EDA started running in January 2010), the EDA-component of the EPS re-forecast suite is generated using the EDA of the current year. The EPS re-forecast suite uses the same SPPT scheme as the operational EPS to simulate model uncertainty.

Ocean waves

An ocean wave model is integrated (two-way coupling) in all operational configurations of the ECMWF atmospheric model (IFS). The model is based on the WAM cycle 4 model, with various enhancements as documented in ECMWF Tech. Memo. No. 478 and No. 509 (available from the ECMWF website http://www.ecmwf.int/publications). The wave model resolution is 30 km for the deterministic forecast; 55 km for the EPS. In addition, a limited area European shelf version is run at 10 km resolution to 5 days.

http://www.ecmwf.int/products/forecasts/wavecharts/