Please see the respective known issues table below for guidance and workarounds.

Amended data errors

Note: Data has been amended in the archive so all retrievals made after the completion of the recovery process (see table below) are not affected by the issues.

Fixed issues number

Issue typeWrong DataRecovery period
(some data could still be wrong)
Recovery Finished
(all the data has been fixed)
E1.WEB-API retrievals
  • Monthly statistics of wind speed (param=207) from the MetOffice seasonal forecast system (origin=egrr):
    • stream=msmm, type=em/hcmean/fcmean/fcmin/fcmax/fcstd
    • stream=mmsa, type=em/fcmean
  • Start dates: 2017-09-01, 2017-10-01 and 2017-11-01

between 2017/11/24 at 11:00 GMT and 2017/11/28 at 12:00 GMT

after 2017/11/28 at 12:00 GMT

E2.

CDS retrievals
(forms and API) & WebAPI retrievals

  • Pressure variables data from GCFS2.0-v20171123 (DWD GCFS2.0) data
    • Affects data from monthly and daily CDS catalogue entries for start date 2011-06-01 and member=13
    • All variables (including 'single-level') have been substituted during the recovery process

between 2019/03/12 at 11:30 GMT and 2019/03/12 at 12:05 GMT

after 2019/03/12 at 12:05 GMT

E4.b

CDS retrievals
(forms and API)

Monthly statistics and anomalies data are temporarily not available for the following systems and dates:
NOTE that high-frequency data in the "daily" CDS catalogue entries are not affected

UK MetOffice (system=14): November hindcasts (1993-2016)

between 2019/10/25 at 11:00 GMT and 2019/10/25 at 12:00 GMTafter 2019/10/25 at 12:00 GMT
E3.a

CDS retrievals
(forms and API) & WebAPI retrievals

Snowfall amount from CMCC-CM2-v20160423 (CMCC SPSv3) data:
  • Affects all data from monthly and daily CDS catalogue entries
    • Start dates:
      • all November start dates for hindcasts (1993-2016)

between 2019/11/07 at 11:00 GMT and 2019/11/07 at 17:00 GMT

after 2019/11/07 at 17:00 GMT

E3.b

CDS retrievals

(forms and API) & WebAPI retrievals

Snowfall amount from CMCC-CM2-v20160423 (CMCC SPSv3) data:
  • Affects all data from monthly and daily CDS catalogue entries
    • Start dates:
      • all December and January start dates for hindcasts (1993-2016)
between 2019/11/25 at 11:00 GMT and 2019/11/25 at 14:00 GMTafter 2019/11/25 at 14:00 GMT
E3.c

CDS retrievals

(forms and API) & WebAPI retrievals

Snowfall amount from CMCC-CM2-v20160423 (CMCC SPSv3) data:

  • Affects all data from monthly and daily CDS catalogue entries
    • Start dates:
      • November 2018 real-time forecast.
  • It also affects monthly anomalies CDS catalogue entry
    • Start dates: 2018-11-01
between 2019/11/27 at 17:00 GMT and 2019/11/27 at 18:00 GMTafter 2019/11/27 at 18:00 GMT
E3.d

CDS retrievals

(forms and API) & WebAPI retrievals

Snowfall amount from CMCC-CM2-v20160423 (CMCC SPSv3) data:

  • Affects monthly anomalies CDS catalogue entry
    • Start dates: 2018-12-01,  2019-01-01
between 2019/11/28 at 16:20 GMT and 2019/11/29 at 13:00 GMTafter 2019/11/29 at 13:00 GMT
E4.a1

CDS retrievals
(forms and API)

Monthly statistics and anomalies data are temporarily not available for the following systems and dates:
NOTE that high-frequency data in the "daily" CDS catalogue entries are not affected

NCEP CFSv2: November hindcasts (1993-2016)


after 2019/12/13 at 12:00 GMT
E4.a2

CDS retrievals

(forms and API)

Monthly statistics and anomalies data are temporarily not available for the following systems and dates:
NOTE that high-frequency data in the "daily" CDS catalogue entries are not affected

NCEP CFSv2: October  hindcasts (1993-2016) and real-time forecasts for October and November 2019


after 2020/01/08 at 00:00 GMT
E7.CDS retrievals

Missing members from CFSv2-v20110310 (NCEP CFSv2) data.

It affects all data available in the high-frequency (daily/subdaily) datasets:

Members with start time at 06h,12h and 18h are not being retrieved from the archive by the CDS and just members with start time at 00h are returned

between 2020/05/15 at 15:45 GMT
and 2020/05/15 at 16:30 GMT

after 2020/05/15 at 16:30 GMT

E7b.CDS retrievals

Missing members from CFSv2-v20110310 (NCEP CFSv2) data.

It affects all data available in the high-frequency (daily/subdaily) datasets:

Members with start time at 06h,12h and 18h are not being retrieved from the archive by the CDS and just members with start time at 00h are returned.

This issue affects retrievals done after the release of the October 2020 forecast, which happened at 12UTC on the 13th of October 2020.


after 2020/10/25 at 16:15 GMT
E9.CDS retrievals

2m dewpoint temperature from System8-v20210101 (Méteo-France System8):

It affects:

Sporadic values of 1.e+20 in individual grid points on the 6-hourly original data. Note this affected a very reduced amount of grid points and time steps.

This issue affects retrievals done after the release of the October hindcasts which happened on the 16th of September of 2021.

between 2021/10/12 at 12:00 GMT and 2021/10/12 at 15:30 GMT.

after 2021/10/12 at 15:30 GMT

Not fixed data errors

The issues described in the table below have not been fixed yet

Not fixed issues number

Issue typeWrong DataRecovery period
(some data could still be wrong)
Recovery Finished
(all the data has been fixed)
E5a.

CDS retrievals
(forms and API)

Pressure level variables from CMCC-CM2-v20160423 (CMCC SPSv3) data.

It affects all hindcast data from the following start dates and ensemble members:

March: 2013 (member 34)
April: 2013 (members 20 and 33)
May: 2013 (members 23, 29 and 32)
June: 2013 (member 26)
July: 1993 (members 11 and 12); 2013 (member 31)
August: 1994 (member 9); 2013 (members 21, 22, 25, 27 and 28)
September: 1998 (member 19)

All variables (pressure levels and single level)from CMCC-CM2-v20160423 (CMCC SPSv3) data.

It affects all hindcast data from the following start dates and ensemble members:

March: 2013 (member 33)
July: 1996 (member 9)

archived (wrong) data won't be overwritten
CMCC SPSv3 system has been superseded by SPSv3.5

archived (wrong) data won't be overwritten
CMCC SPSv3 system has been superseded by SPSv3.5

E5b.

CDS retrievals
(forms and API)

Real-time forecast data from CMCC-CM2-v20160423 (CMCC SPSv3) data.

It affects anomalies calculated using the wrong data specified in E5.a, specifically:

Pressure variables: March to September 2019
All variables: March 2019, July 2019

archived (wrong) data won't be overwritten

archived (wrong) data won't be overwritten
E6.

CDS retrievals
(forms and API)

Surface solar radiation variables from CFSv2-v20110310 (NCEP CFSv2) data.

It affects:

  • forecast data for nominal start dates October, November, December 2019 and January, February 2020
  • hindcast data (1993-2016) for nominal start dates October, November, December, January and February

Two variables - "surface net solar radiation" and "surface solar radiation downwards" - have been swapped with each other. This error affects the daily data and corresponding monthly statistics and anomalies.

data with a fix for the affected start dates will be provided
(date to be determined)

data with a fix for the affected start dates will be provided
(date to be determined)

E8.CDS retrievals
(forms and API)

Total cloud cover from JMA cps2-v20150526 (JMA/MRI-CPS2) data.

It affects all available data for that variable (daily data, monthly data and anomalies) both for real-time forecasts and hindcast dates

Due to a bug during the data ingestion to the C3S archive the values of this parameter have been divided by 100.

archived (wrong) data won't be overwritten
JMA CPS2 will be superseded by CPS3 in February 2022
archived (wrong) data won't be overwritten
JMA CPS2 will be superseded by CPS3 in February 2022
E9.CDS retrievals
(forms and API)

Sea surface temperature and land surface temperature (called 'soil temperature level 1' in the CDS) from CMCC-CM2-v20191201 (CMCC SPSv3.5) data.

It affects the following data for those two variables:

  • forecast data (daily data, monthly data and anomalies) for nominal start dates March, April, May, June and July 2022

According to CMCC report about the issue:

The problem only affects the points at the borders between the three realms (sea ice, ocean and land), and is due to a change in the land-sea or sea ice-sea mask interpolation algorithm, which was implemented in the forecasts (starting from March).
Therefore, the anomalies of the above mentioned variables show unrealistic patterns near the sea-land or sea-sea ice boundaries, because, in these grid points, the new interpolation algorithm has made the forecasts, from March onwards, inconsistent with the hindcasts.
Starting from the August 2022 real-time forecast, it will be restored the procedure used in the hindcasts.

archived (wrong) data won't be overwrittenarchived (wrong) data won't be overwritten

Data and metadata deficiencies

Known issue numberCentreForecasts
or
Hindcasts
Start dates(1)Description
D1.

Met Office
(all systems)

BothAll

Near-surface temperatures (6-hourly and daily minimum and maximum) and dewpoint temperature are encoded as being calculated at a height of 2m.
Those parameters are calculated at 1.5 meters height in Met Office seasonal forecast systems

D2.

Met Office
(system=12)

Both


Hindcasts: May-Dec and January
Forecasts: Sep2017 to Jan2018

Global grid is incomplete (just 180 latitude points, between 90S and 89N)
Requests with MARS-based interpolations (i.e. using "grid=NN/NN") won't work unless a suitable regional subselection is requested (with "area=LL/LL/LL/LL")

Note that additionally, in this grid all points are set at integer values of latitude/longitude

D3.

All parameters on pressure levels are not globally complete, having "missing value indicators" in points below the orography.

D4.Météo-France
(system=5)
ForecastsAll

There is an issue with the postprocessing used at origin of the lagged start dates.

Only members #27 to #51 are affected having data with a time shift of 7 days (168 hours). For instance, data with date/time labelled as YYYYMM08 00h would in fact correspond to YYYYMM01 00h.

D5.CMCC
(system=3)
HindcastsYears between 1993 and 1999

There are anomalous values of SST (and associated parameters, e.g. fluxes, and vertical oceanic temperature profiles) appearing in the first hours of a very limited number of ensemble members of forecasts.

The issue is caused by a bug in the oceanic perturbation generation tools used at origin to produce the ocean initial conditions.  The errors/perturbations appear initially as excessively high SST values, confined to the reported hindcast years, in a limited number of grid points all in the tropical Pacific. They are very quickly dissipated by the coupled model physics and dynamics, affecting, therefore, only the first 24 hours of the forecasts at most.

D6.CMCC
(system=3)
BothAll

There are some variables affected by a spin-up issue. This affects just some ensemble members depending on the initial conditions each member started from. Unfortunately, as the selection of the forecast members is done randomly from a bigger pool of initial conditions, there is no metadata describing this in the current service based on GRIB files stored in ECMWF's MARS archive.

D7.

DWD
(system=2)

HindcastsAll 2017 start datesFor this forecasting system, year 2017 start dates have been computed as hindcasts and not real-time forecasts.
D8.CMCC
(system=3)
BothAllFor this forecasting system, the evaporation field is masked over the ocean (only land points are available)
D9.Met Office
(system=12,13)
BothAllThe fractional land-sea mask doesn't match the masked (land-only or sea-only) fields. For the files affected, the land-sea mask provided was produced as the result of a 1-degree interpolation of the binary land-sea mask in the native model resolution

Data gaps

Note: (1) See description of nominal start dates here

Known issue numberCentreForecasts
or
Hindcasts
Start dates(1)Description
G1.

Met Office
(system=12)

Forecasts

20180102
20180103
20180104
20180105

No data is available.
G2.

Met Office
(system=12,13)

ForecastsNominal start dates previous to November 2018Monthly statistics and anomalies are just available for 50 members - specifically, the 50 members closest and previous or equal to the 1st of the nominal start month.
G3.

MetOffice
(system=12)


ForecastsNominal start dates from February to April 2018Parameter "Soil temperature at level 1" (GRIB/MARS paramId=139) is not available.
HindcastsNominal start dates from February to April

MetOffice
(system=13)

ForecastsNominal start dates from May to December 2018
HindcastsNominal start dates from May to December
G4.MetOffice
(all systems)
HindcastsNominal start date January 1993January 1993 data is not available.
MetOffice system produces hindcasts from January 1993 to December 2016, and due to the way nominal start dates are assigned, January 1993 would require data from December 1992 which are not produced by the MetOffice system.


G5.MetOffice
(system=600)
Forecasts20210222
20210223

No data is available.
An issue with assimilation of sea-ice data was reported by the MetOffice affecting forecasts initialised on 22 and 23 February 2021. Due to the potential impact of that wrong sea-ice initialisation throughout the forecast, those start dates haven't been processed by C3S.

The GloSea6 forecasts initialised on 22 and 23 February were initialised with low Arctic sea ice extents. This resulted from corrupt data used in the derivation of ocean initial conditions. The sea ice largely recovers in these forecasts, but is biased low through the whole 6-month forecast compared to neighbouring days.

Initial sea-ice conditions for different GloSea6 February 2021 start dates:

Evolution of sea-ice fraction forecasts for different GloSea6 February 2021 start dates:

G6.JMA
(system=3)
Forecasts20220102 to
20220111
No data is available.
JMA CPS3 started its operational production at JMA on the 12th January 2022
G7.MetOffice
(system=601)
Hindcasts20030417

Due to a technical issue with the acquisition at ECMWF of MetOffice data, member 106 was incomplete.

To fix this in the archive the data from one member from system 600 (same forecasting system) for this start date was used to substitute member 106


Data irregularities

Known issue numberCentreForecast or hindcastDescription
I1.MetOffice
(all systems)
Forecasts

MetOffice real-time component is designed to produce a constant number of members each start date (see Description of the C3S seasonal multi-system).

In those rare occasions when the MetOffice is not able to produce some of the expected members for a given start date, the affected members will be missing. Instead of re-running precisely those missing members, the MetOffice system schedules the run on a later start date of as many additional members as they were missing.

The following table records all the occurrences of this irregularity by listing the dates when some members are missing (and how many of them) and the dates when additional members were produced.

SystemMissing membersAdditional members 

Start date


Start date
12201708222201708232
13201806211201806281
13201810102201810112
600202102021202102041
600202106082202106141
202106151
600202111212202111232




This document has been produced in the context of the Copernicus Climate Change Service (C3S).

The activities leading to these results have been contracted by the European Centre for Medium-Range Weather Forecasts, operator of C3S on behalf of the European Union (Delegation agreement signed on 11/11/2014). All information in this document is provided "as is" and no guarantee or warranty is given that the information is fit for any particular purpose.

The users thereof use the information at their sole risk and liability. For the avoidance of all doubt, the European Commission and the European Centre for Medium-Range Weather Forecasts have no liability in respect of this document, which is merely representing the author's view.