The ECMWF monthly forecasting system has two components:
- The real-time forecasting system
- The back-statistics needed to create a model climatology to calibrate the real-time forecasting system.
The real-time forecasting system
The real-time ENS/monthly forecasting system is a 51-member ensemble of 46-day coupled ocean-atmosphere integrations. The first 10 days are performed with a TL639L91 resolution forced. After day 10, the model has a resolution of TL319L91. The extension of ENS to 46 days is performed every Thursday and Monday. Before January 2008, the monthly forecasting system was a separate system from EPS and before 12 May 2015 the forecast length was 32 days. The first operational real-time monthly forecast was realized on Thursday, 7 October 2004.
Atmospheric component: IFS with the same cycle as the high resolution forecast:
- CY24R3 from 27/03/2002 to 24/04/2000 (included) : Monthly forecasts every two weeks (Wed) - 12-year hindcasts
- CY25R1 since 24/04/2002 (included)
- CY25R4 since 29/01/2003 (included)
- CY25R5 (IBM) from 12/03/2003 (included)
- CY26R1 from 07/05/2003 (included)
- CY26R3 from 08/10/2003 (included)
- CY28R1 from 10/03/2004 (included) : Monthly forecasts once a week (thursdays)
- CY28R3 since 07/10/2004 (included)
- CY29R1 since 07/04/2005 (included)
- CY29R2 since 30/06/2005 (included) + new sea-ice treatment*
- CY30R1 since 02/02/2006 (included) + change of resolution to 62 vertical levels
- CY31R1 since 14/09/2006 (included)
- CY31R2 since 14/12/2006 (included)
- CY32R2 since 07/06/2007 (included)
- CY32R3 since 08/11/2007 (included)
- CY32R3V since 13/03/2008 (included) : Switch to VAREPS/Monthly - 18-year hindcasts, 18-year hindcasts
- CY33R1 since 05/06/2008 (included)
- CY35R1 since 30/09/2008 (included)
- CY35R2 since 12/03/2009 (included)
- CY35R3 since 10/09/2009 (included)
- CY36R1 since 28/01/2010 (included) + change of horizontal resolution to T639/T319
- CY36R2 since 24/06/2010 (included) : EDA
- CY36R4 since 11/11/2010 (included)
- CY37R2 since 18/05/2011 (included) (GRIB2 for model levels) : Monthly Forecasts twice a week since 10/10/2011 (Mondays + Thursdays)
- CY37R3 since 17/11/2011 (included): IFS is coupled to NEMO instead of HOPE. Oceanic component: NEMO with a zonal resolution of about 1 degree (same as in SYSTEM 4). Coupling: OASIS (from CERFACS, France). The atmospheric fluxes of momentum, heat and fresh water are passed to the ocean every 3 hours. In exchange, the ocean surface temperature (SST) is passed to the atmosphere.
- CY38R1 since 21/06/2012 (included): 20 years of hindcasts (instead of 18 years) + new soil reanalysis in the hindcasts.
- CY38R2 since 27/06/2013 (included):
- CY40R1 since 21/11/2013 (included): change of vertical resolution to 91 vertical levels, single executable for coupled ocean-atmosphere coupling, ocean-atmosphere coupling from day 0, sea ice cover relaxed to past 5-year climatology
- CY41R1 since 12/05/2015 (included): Extended-range forecasts extended to 46 days (instead of 32 days), re-forecasts produced twice a week (instead of once a week ) with 11 ensemble members (instead of 5 members).
After 10 days of coupled integrations, the model drift begins to be significant. It displays similar patterns to seasonal forecasting after 6 months of integrations, but with less amplitude. The strategy for dealing with model drift is straightforward. We initialize the ocean, atmosphere and land surface to be as close to reality as possible, and calculate the forward evolution of the system as best we can using numerical approximations of the laws of physics. No "artificial" terms are introduced to try to reduce the drift of the model and no steps are taken to remove or reduce any imbalances in the coupled model initial state: we simply couple the models together and start to integrate forward. The effect of the drift on the model calculations is estimated from previous integrations of the model in previous years (the back-statistics). The drift is removed from the model solution during the post-processing.
An additional motivation for creating a model climatology is that after about 10 days of forecasts, the spread of the ensemble is very large (see, for instance, forecast plumes). Therefore, the probability distribution function (pdf) of the model climatology needs to be evaluated, in order to detect any significant difference between the ensemble distribution of the real-time forecast and climatology.
In the present system, the climatology (back-statistics) is a 11-member ensemble of 46-day EPS integrations, starting on the same day and month as each real time forecast for each of the past 20 years. For instance, if the first starting date of the real-time forecast is 27 March 2013. The corresponding climatology is a 11-member ensemble starting on 27 March 2012, 27 March 2011, ..., 27 March 1993. The 11-member ensemble is thus integrated with 20 different starting dates. This represents a total of 220 integrations . Real-time forecasts are calibrated using a 1- week widow of re-forecasts, which represents a total of 660 (3 start dates x 20 years x 11 members) re-forecast integrations.
The back statistics are created once a week and are ready 3 weeks before the real-time forecasting suite starts. The real-time monthly forecasts produced on Thursdays are calibrated using only the back-statistics associated to the date of the real-time forecasts. The real-time monthly forecasts produced on Mondays are calibrated using the back statistics associated to the previous and next Thursdays, and a weight averaging is applied.
* new sea-ice treatment: before 30 June 2005, the sea-ice cover was computed from the SSTs produced by the ocean model. Since 30 June 2005, the sea-ice cover is persisted from the atmospheric initial conditions till day 10, then relaxed towards climatology. After day 30, the sea-ice cover is the climatological sea-ice cover (from ERA40).