Long-range Forecasting

How forecasts are generated

The second version of the National Centers for Environmental Prediction (NCEP)'s Climate Forecast System (CFSv2) was made operational at NCEP in March 2011. It is a fully coupled model representing the interaction between the Earth's atmosphere, oceans, land and sea-ice.

Forecasts are generated using ensembles of 40 members, or 40 model runs. As four model runs are available each day, the current practice is to use the latest 10 days’ forecast to produce the ensemble.

10 days x four runs = 40 ensemble members

For example, the August and AOS forecasts were generated using model runs from the 6th to 15th of July.

The re-forecast data are used in the verification of the forecasts, as well as for creating the climatology. We use a NCEP CFSv2 re-forecast period of 1982-2010, with four runs made available for every 5th day. Our current practice is to use all available model runs from the similar 6th to 15th period of each calendar month to create an ensemble of 12 members.

3 days x four runs = 12 ensemble members

For forecast products issued in March 2021 onwards, a re-forecast reference period of 1993-2010 is used to calibrate the NCEP CFSv2 model.

For forecast products issued in May 2022 onwards, Copernicus C3S-NCEP data will be used.

The C3S-NCEP CFSv2 forecasts are generated using ensembles of 50 members, or 50 model runs. As four model runs are available each day, the current practice is to use the latest 12/13 days’ forecast to the 1st of each month to build the ensemble for the nominal start date on the first of that month.

The re-forecast data are used in the verification of the forecasts, as well as for creating the climatology. The C3S-NCEP CFSv2 re-forecast is available for 1993-2016, with five to seven start days (4 model runs per day) made available depending on the calendar start month. Our current practice is to use all available start days' runs for each calendar month to create an ensemble of 20 to 28 members. 

Further information on NCEP CFSv2 (C3S dataset) can be found in Copernicus Services link. More information on NCEP CFSv2 can also be found in Saha et al. 2014:

Saha, S., Moorthi, S., Wu, X., Wang, J., Nadiga, S., Tripp, P., Behringer, D., Hou, Y.-T., Chuang, H.Y, Iredell, M., et al. 2014. The NCEP Climate Forecast System Version 2. Journal of Climate, Volume 27, pp. 2185-2208.

ECMWF is the European Centre for Medium-Range Weather Forecasts. The fifth generation of the ECMWF seasonal forecasting system, SEAS5 was introduced in November 2017, replacing System 4. ECMWF SEAS5 includes updated versions of the atmospheric (IFS) and ocean (NEMO) models and adds the interactive sea ice model LIM2.

ECMWF SEAS5 forecasts are generated using ensembles of 51 members, or 51 model runs initiating on the first day of each month.

The re-forecast data are also used in the verification of the forecasts, as well as for creating the climatology. We use a re-forecast period of 1993-2016, with 25 runs made available on the first day of each month.

Further information on ECMWF SEAS5 (C3S dataset) can be found in Copernicus Services link.

GloSea stands for Global Seasonal Forecasting System. GloSea is the seasonal prediction system developed and run operationally at the UK Met Office. GloSea6 became operational in February 2021, replacing GloSea5. GloSea6 is an ensemble prediction system built around the high resolution version of the Met Office climate prediction model: HadGEM3-GC3.2 family atmosphere—land—ocean—sea-ice coupled climate model.

The UK Met Office GloSea6 forecasts are generated using ensembles of 50 members, or 50 model runs. As two model runs are available each day, the current practice is to use the latest 25 days’ forecast to the 1st of each month to build the ensemble for the nominal start date on the first of that month.

The re-forecast data are used in the verification of the forecasts, as well as for creating the climatology. The UK Met Office GloSea6 re-forecast is available for 1993-2016, with seven runs made available on the 1st, 9th, 17th and 25th day of each month. Our current practice is to use all available model runs for each calendar month to create an ensemble of 28 members.

4 days (1st, 9th, 17th and 25th) x seven runs = 28 ensemble members

For forecast products issued in March 2021 onwards, it is based on GloSea6. Further information on the UK Met Office GloSea (C3S dataset) can be found in Copernicus Services link.