Section 2.1.2.1 Medium Range Ensemble forecasts

Medium Range Ensemble forecasts

The medium range ensemble forecast suite provides a range of possible future weather states in the medium range.  This allows investigation of the detail and uncertainty in the forecast during the first 15 days.

The medium range ensemble (ENS):

• • is run four times daily:
    • Day0 to Day15 based on 00UTC and 12UTC data
    • Day0 to Day6 based on 06UTC and 18UTC data. Results from these runs are:
      • used in Early cut off analysis (SCDA) and Long window data analysis (LWDA).
      • disseminated T+0 to T+72
  • has horizontal resolution of 9 Km and vertical resolution of 137 model levels.  These values are the same as those of HRES (High Resolution in Cy48 and earlier).
  • has land-atmosphere and ocean-atmosphere coupling.
  • has 50 ensemble members plus an unperturbed ensemble control member.
  • the Ensemble Control Forecast (ex-HRES) output runs on the schedule of HRES in Cy48 and earlier.   It runs before the medium range ensemble starts.  It is labelled HRES for the convenience of users but the name will be withdrawn in a future update cycle.
  • the Ensemble Control Forecast (ex-HRES) output runs on the schedule of HRES in Cy48 and earlier.   It runs before the medium range ensemble starts.  It is labelled HRES for the convenience of users but the name will be withdrawn in a future update cycle.
  • the dissemination schedule is given in Section 3.1:
    • Ensemble Control Forecast (ex-HRES) schedules and products.
    • Medium range schedules and products.  

The ensemble consists of one unperturbed control member and 50 perturbed members.  These are similar to the control member but differ only that initial states and model physics have been perturbed.  This allows exploration of the currently understood range of uncertainty in the observations and the model.   A set of medium range re-forecasts provide a basis for the medium range model climate (M-Climate).

The medium range 15 day ensemble forecasts run with data analysed from the short cut off data analysis (SCDA).  

When averaged over many forecasts each of the perturbed members has lower skill than the control member.  Nevertheless any individual forecast might show a higher skill.  A range of ensemble products show information in different ways and appropriate for different uses.  These include probabilities, clusters of alternative scenarios, and tropical cyclone and extra-tropical cyclone products.  Certain important products are based on, or directly show, anomalies relative to the M-climate.  Particularly important are meteograms and the Extreme Forecast Index and Shift of Tails (EFI/SOT).   The ENS products and chart presentations show where and when weather may deviate from the average for the location and time of year, and if such an anomaly could be extreme.

Ensemble forecasts are run coupled to the Wave model (ECWAM) and the Dynamic Ocean model (NEMO) which includes modelling of sea ice (LIM2).  Atmosphere-ocean coupling occurs every hour.  This high-frequency coupling generally has a positive impact on the forecast development of synoptic-scale systems (e.g. tropical cyclones).

With effect from Cy49r1, HRES and Ensemble Control are scientifically, structurally and computationally identical with 9km resolution and their products are known as Ensemble Control Forecast (ex-HRES).

Additional Sources of Information

(Note: In older material there may be references to issues that have subsequently been addressed)

• View the ECMWF outline information on the rational behind probabilistic weather forecasts.
• View the ECMWF eLearning module on the Ensemble Forecasting Introduction - Sources of Forecast Uncertainty.

(FUG Associated with Cy49r1)