This table shows the centres that provide data to this project together with the latest configuration of their systems. Follow the link of each Data Provider for specific model description.

 NOTES

(1) For each real-time forecast start date, the associated hindcast data is produced and made available through the CDS at least one month in advance. The details about data availability can be found in "Summary of available data" page.

(2) The production schedules of forecasting system with lagged start dates don't prescribe how to build an ensemble for a specific nominal start date. The choices currently in use for the C3S products for those forecasting systems can be found within the information about the concept of nominal start date.

(3) Despite being produced in a lagged mode, the data from Météo-France forecasting systems is currently encoded and provided in CDS as if all the members were initialized on the 1st.

(4) Due to the flexibility of the Met Office forecasting system, forecast failures on a given date are not usually recovered by re-running the missed forecasts at a later date, but by running more members with initial conditions of the day of recovery. Example: An incident affected the 22 August 2017 forecast so no members are available for that date. Instead, 4 members were initialised on 23 August 2017.

(5) Details of the complete list of start dates are provided in the description of CFSv2 system, section 6. Other relevant information.

(6) Details of the complete list of start dates are provided in the description of CPS3 system, section 6. Other relevant information.

(7) ECCC contribution, CanSIPSv2.1, is a multi-system composed of two independent forecasting systems.

Versioning of forecasting systems contributing to C3S

All C3S seasonal forecast data served through the CDS is currently held at ECMWF MARS archive, where data is stored in GRIB format. Due to that fact, the different forecasting systems of a given producing centre are labelled using the MARS-specific keyword "system" contained in each GRIB header as an integer number. For any data retrieval, the CDS will require that "system" keyword to be specified.

The use of the "system" keyword to match real-time forecasts with their relevant hindcasts is a critical element for a correct use of the C3S seasonal forecast data.

Details on how that label is assigned can be found in the following sections.

Systems producing their hindcasts as a fixed dataset

For these forecasting systems, a unique value of the keyword "system" is assigned trying to map as close as possible the version numbering used by the forecasting centres. Specifically:

(*) This model_id tag is used internally in some parts of the CDS infrastructure to label the seasonal forecast data. In particular data which is archived on netCDF format will use this labelling. Currently this only applies to ocean variables contained in the seasonal-monthly-ocean dataset.
(a) Due to technical reasons, a new version of ECMWF SEAS5 data has been created using the keyword system=51. See details on Announcements (date 12/October/2022)

Systems producing their hindcasts on-the-fly

Systems following this approach generate a new set of hindcasts close to the production of the real-time forecast for which those hindcasts are relevant. This production schedule approach enables these forecasting systems to implement updates and changes more frequently. But often the lifetime of a given version is longer than a year, and that would result in hindcast data for the same start dates being produced with exactly the same version of the forecasting system twice (or more times, if that version's lifetime ends up being longer than 2 years). To enable the distinction between hindcasts for the same hindcast start dates produced in those situations some form of sequential labelling would be needed.

Before March 2021

Following the data practices in EUROSIP, the predecessor of C3S seasonal forecast activity, the 'system' keyword was used as a numeric sequential label without any meaning related to the underlying forecasting system. In that sense, a given version of the forecasting system could span a few different values of the 'system' keyword as listed below.

The only producing centre following this approach in this period was the UK MetOffice, and the increments on the value of 'system' were introduced on the May nominal start date each year. Specifically:

(*) This model_id tag is used internally in some parts of the CDS infrastructure to label the seasonal forecast data. In particular data which is archived on netCDF format will use this labelling. Currently this only applies to ocean variables contained in the seasonal-monthly-ocean dataset.

After March 2021

With the start of operations of MetOffice GloSea6 forecasting system, a new numeric labelling paradigm has been introduced with two main aims: on one hand, to keep the incremental labelling when needed, and on the other hand, make more transparent and user-friendly the mapping between different versions of the forecasting system and that 'system' keyword.

Essentially, the value of the 'system' keyword for hindcast data produced on-the-fly (and their related real-time forecasts) would be assigned with the following approach: a three-digit number with the first digit being the version of the forecasting system; the second digit will be assigned to those changes in the forecasting system constituting minor revisions; and finally the last digit will start in 0 and will be updated incrementally each year if a new hindcast is going to be produced with the same version of the forecasting system (i.e. when the first two digits have not changed).

The following table lists the values of 'system' keyword used in the CDS for on-the-fly hindcasts systems:

(*) This model_id tag is used internally in some parts of the CDS infrastructure to label the seasonal forecast data. In particular data which is archived on netCDF format will use this labelling. Currently this only applies to ocean variables contained in the seasonal-monthly-ocean dataset.