Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The details of each forecasting system correspond to the values of the "system" keyword (as used both in the CDS API and on the CDS interactive download forms). A full description of the use of that keyword and its correspondence with the different forecast systems names is available from the "Description of the C3S seasonal multi-system" page.

Nominal Start DatesC3S multi-system operational components ('system' keyword values)
ECMWFMétéo-FranceUK MetOfficeDWDCMCCNCEPJMAECCCBOM
September 2017 - October 20174512





November 2017 - April 20185512 (1)





May 2018 - October 20185513





November 2018 - April 2019561323



May 2019 - September 20195614 (2)23



October 2019 - April 20205714232


May 2020 - September 20205715232


October 2020571523522

November 2020 - February 20215715213522

March 2021 - April 202157600213522

May 2021 - June 202157600213522

1 and 2


July 2021 - November 202158600213522

1 and 2


December 2021 - January 202258600213522

2 and 3


February 202258600213523

2 and 3


March 2022 - October 202258601213523

2 and 3


November 2022 - Februrary 202351 (3)8601213523

2 and 3


March 2023(4) - February 2024518602213523

2 and 3


March 2024 - June 2024518603213523

2 and 3


July 2024518603
3523

 


August 2024 - February 2025518603213523

4 and 5


March 2025518604213523

4 and 5


April 2025518604223522

4 and 5












Legend:

bold fontforecast systems introduced for the first time at the beginning of the nominal start dates range
non-bold fontforecast systems that were excluded from the multi-system at that time

Notes:

(1) From February 2018: upgrade on data provision: changes in set of parameters available, see column noted with 'C3S-0.1 netcdf' in this link. Data deficiencies fixed, see known issues D2 and D3.
(2) Data deficiencies fixed, see known issue D9.
(3) Due to technical reasons, a new version of ECMWF SEAS5 data has been created using the keyword system=51. 

Expand
titleSee additional information about ECMWF system=51

Starting from November 2022 a new version of ECMWF SEAS5 output has been introduced in C3S. This new version (v5.1) has been labelled with the keyword system=51 

  • What are the changes in this new version of ECMWF SEAS5 data?
    • A different interpolation tool has been used to produce the 1-degree grid, both for hindcasts and real-time forecasts.
    • A new 1-degree grid, now consistent with the grid available from the other participating forecast systems. Specifically, it will have grid-points centered at half-degree latitude/longitude, and hence it has 180 values in the latitude coordinate (from -89.5 to 89.5) instead of the current 181 values (from -90 to 90)
    • A couple of additions to complete the list of available variables: Top solar incoming radiation and 1000hPa level for the 5 variables available at pressure levels. Additionally, on top of the values of runoff, for SEAS5.1 both sub-surface and surface runoff will be also available.


  • What are the implications for users of ECMWF SEAS5 data? 
    • (warning) From November 2022, only system=51 real-time forecast data will be available from the CDS
    • (warning) Users will need to use system=51 hindcast data alongside the new real-time forecasts.
      Data from the current version (system=5 ) should not be used, so those users with previously downloaded hindcast data will need to download the new version.
  • Why have these changes been introduced?
    • When ECMWF SEAS5 was released in 2017, the C3S 1-degree version was created with the default interpolation tool (emoslib ) available from MARS. With the migration of ECMWF computing infrastructure to its new Data Centre in Bologna, this tool has been discontinued. That means new real-time forecasts produced after the migration can't be produced anymore in a way consistent with the available SEAS5 hindcast in the CDS. After careful evaluation of the impact of the interpolation, it was decided to create a new version of the hindcast data. This has been labelled as SEAS5.1 and encoded with keyword system=51.
    • With this new version of SEAS5 hindcast, a couple of additional improvements were also included:
      • A new 1-degree grid consistent with the one provided for other forecast systems (see details above)
      • A couple of additions to the list of variables (see details above).

(4) From March 2023 a dataset with ocean variables is available from the CDS catalogue (see details on availability here).



Hindcasts

Nominal start dateECMWFMétéo-FranceMet Office(a)DWDCMCCNCEPJMAECCCBOMNominal start date
System 4
(system=4)
SEAS5
(system=5)
SEAS5 
(system=51)(b)
System 5 (system=5)System 6 (system=6)System7 (system=7)System8 (system=8)GloSea5-GC2
(system=12)
GloSea5-GC2
(system=13)
GloSea5-GC2-LI
(system=14)
GloSea5-GC2-LI
(system=15)
GloSea6
(system=600)

GloSea6 (system=601)

GloSea6
(system=602)

GloSea6
(system=603)

GloSea6 (system=604)GCFS2.0 (system=2)GCFS2.1 (system=21)GCFS2.2 (system=22)

SPSv3  (system=3)

SPSv3.5 (system=35)CFSv2 (system=2)CPS2 (system=2)CPS3 (system=3)GEM-NEMO (system=1)CanCM4i (system=2)GEM5-NEMO (system=3)CanESM5.1p1bc (system=4)GEM5.2-NEMO (system=5)

ACCESS-S2

(system=2)

January(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(c)(tick)(c)(tick)(c)(tick)(c)(tick)(c)(tick)(c)(tick) (c)(tick) (c)(error)  (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick) (tick) (error) January
February(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick) (tick) (error) February
March(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick)(tick) (tick) (tick)(tick)(tick) (tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick) (tick) (tick) March
April(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick)(tick) (tick) (tick)(tick)(tick) (tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick)  (tick)  (tick) April
May(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) 
(error)
(tick)  (tick)(tick)
(error)
(tick)  (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)  
(error)
(tick)  
(error)
(tick) May
June(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)
(error)
(tick)  (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)
(error)
(tick)  (tick)  (error) June
July(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(error) July
August(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (tick)(error) August
September(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (tick) (error) September
October(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (tick) (error) October
November(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)(tick)(error) (tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (tick) (error) November
December(error)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick)(tick) (error) (tick)
(tick)(error) (tick)(tick)(tick)(tick)(tick)(error)(tick)(tick)(tick) (tick) (error) December


(a)For the Met Office contribution, due to the production of hindcasts "on-the-fly", the CDS keyword 'system' does not have the same meaning as for the other contributors. More details here.
(b) 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)
(c) Hindcast data for January 1993 is not available from MetOffice forecast systems (see "known issue" G4 on the data gaps section)

...