1. Forecast system version

Identifier code: SEAS5

First operational forecast run: 1 November 2017

2. Configuration of the forecast model

Is the model coupled to an ocean model?   Yes

Coupling frequency: 1 hour

2.1 Atmosphere and land surface

ModelIFS Cycle 43r1
Horizontal resolution and grid

Dynamics:TCO319 cubic octahedral grid

Physics: O320 Gaussian grid (36 km)

Atmosphere vertical resolutionL91
Top of atmosphere0.01 hPa (approx. 80 km)
Soil levels (layers)

4

Layer 1   : 0 - 7 cm
Layer 2   : 7 – 28 cm 
Layer 3   : 28 – 100 cm
Layer 4   : 100 – 289 cm 

Time step20 minutes

Detailed documentation: IFS cycle 43r1 documentation

2.2 Ocean and cryosphere

Ocean model

NEMO v3.4

Horizontal resolutionORCA 0.25
Vertical resolutionL75
Time step1 hour
Sea ice modelLIM2
Sea ice model resolutionORCA 0.25
Sea ice model levelsN/A
Wave modelECMWF wave model
Wave model resolution0.5 degrees

Detailed documentation: NEMO documentation and IFS cycle 43r1 documentation

3. Boundary conditions - climate forcings

Most forcings are those used by default in CY43r1 NWP configurations such as HRES or ENS, and are described in more detail in the IFS 43r1 documentation. Some options used by ERA5 were also adopted for SEAS5, notably the time-variation of tropospheric sulphate aerosol and the solar forcing. Only the specification of volcanic aerosol was specific to SEAS5, and this was designed to allow a real-time response to any major volcanic eruption.

Greenhouse gasesCMIP5 historical values of CO2, CH4, N2O, CFC11 and CFC12 to 2005, then RCP3-PD (otherwise known as RCP2.6 - see van Vuuren et al. 2011) values from 2006 onwards. Used in combination with a zonal mean seasonal cycle to give space and time varying fields.
OzoneRadiation scheme sees a seasonally varying but otherwise fixed climatological ozone field. The standard Cy43r1 climatology used is based on the CAMSiRA reanalysis of ozone.
Tropospheric aerosolsSulphate aerosol has decadally varying values, calculated (using CMIP5 historical emissions to 2005 followed by RCP4.5 emissions) by the NCAR CAM3.5 model and processed for use by the IFS (see Hersbach et al., 2013). Other aerosols (dust, salt, organic matter, black carbon) have a fixed seasonally-varying climatology from Tegen et al. The cloud scheme uses two fixed values representing continental/maritime air, with no time variation.
Volcanic aerosolsBased on vertically integrated AODs from the 2012 update by GISS (https://data.giss.nasa.gov/modelforce/strataer/). Damped persistence from 3 initial values (NH, tropics, SH), with a fixed vertical profile and optical properties representing sulphate aerosol with a fixed particle size distribution. Initial values for real-time forecasts generally set to a climatological background, but would be set to best estimated values after any large eruptions. No such eruptions have occurred so far.
Solar forcingTime-variation of total solar insolation (TSI), specified as CMIP5 annual mean values up to 2008, then a repeat 13 year cycle. No variation in solar spectrum. See Hersbach et al., 2013.

Detailed documentation: IFS cycle 43r1 documentation and SEAS5 user guide

4. Initialization and initial condition (IC) perturbations

4.1 Atmosphere and land


HindcastForecast
Atmosphere initialization
ERA-InterimECMWF operations
Atmosphere IC perturbationsEnsemble data assimilation and leading singular vectors applied to upper air variablesEnsemble data assimilation and leading singular vectors applied to upper air variables

Land Initialization

43r1 land surface model driven by ERA-Interim (like ERA-Interim land)ECMWF operations
Land IC perturbationsEnsemble data assimilation applied to some land fieldsEnsemble data assimilation applied to some land fields
Soil moisture initialization43r1 land surface model driven by ERA-Interim (like ERA-Interim land)ECMWF operations
Snow initialization43r1 land surface model driven by ERA-Interim (like ERA-Interim land)ECMWF operations
Unperturbed control forecast?YesYes

Data assimilation method for control analysis: 4D Var (atmosphere) and 3DVAR (ocean/sea-ice)

Horizontal and vertical resolution of perturbations:  T42L91 SVs+ T399L137 EDA perturbations

Perturbations in +/- pairs: Yes

Detailed documentation: IFS cycle 43r1 documentation and SEAS5 user guide


4.2 Ocean and cryosphere


HindcastForecast
Ocean initializationORAS5ORTA5
Ocean IC perturbations Yes - generated through perturbations to assimilated observations and surface forcingYes - generated through perturbations to assimilated observations and surface forcing
Unperturbed control forecast?NoNo

Detailed documentation: ECMWF ocean reanalysis documentation and SEAS5 user guide

 

5. Model Uncertainties perturbations:

Model dynamics perturbationsNo
Model physics perturbations3-level SPPT and SPBS

If there is a control forecast, is it perturbed?

Yes

Detailed documentation: IFS cycle 43r1 documentation

6. Forecast system and hindcasts

Note, the ECMWF seasonal forecasts cover two time ranges: the long range (LR) forecasts out to 7 months and annual range (AR) forecasts out to 13 months (the latter not available from C3S). The model used for these forecasts is identical, but they have different numbers of forecast members.

Forecast frequency

monthly (LR)

quarterly (AR)

Forecast ensemble size

51 (LR)

15 (AR)

Hindcast years36 (1981-2016)
Hindcast ensemble size

25 (LR)

15 (AR)

On-the-fly or static hindcast set?static

7. Other relevant information

Horizontal Interpolation

The IFS is a spectral model, and upper air fields such as temperature, geopotential are originally output and archived as spherical harmonic coefficients, while grid-point fields exist on an O320 reduced gaussian grid. The C3S archive of SEAS5 consists of all fields pre-interpolated to a standard 1 degree lat/long grid.

SEAS5 contribution to C3S is split in two slightly different datasets, due to changes in available interpolation software at ECMWF. Before the move of ECMWF HPC infrastructure to Bologna in late 2022, the data is labelled as system=5. For consistency, the reforecast dataset was also re-processed with the new interpolation software, and the reprocessed data was given the label SEAS5.1  (system=51). The underlying model integrations and data are identical, only the interpolation has changed.

system=5  interpolation

Data was interpolated using the spectral transform and grid interpolation options embedded within the ECMWF MARS software as it existed when SEAS5 was created (More details here).

system=51 interpolation

For single level grid-point fields a bespoke grid-box-average interpolation is used from the native O320 reduced gaussian grid (~36 km resolution) to the C3S 1x1-degree grid. Variables in pressure levels, stored as spherical harmonic coefficients in the SEAS5 archive, are interpolated using MARS default MIR interpolation which first transforms the spherical harmonics coefficients to a full Gaussian grid equivalent to 1-degree resolution, before interpolating to the regular latitude-longitude 1x1-degree grid. Fields coming from the ocean component (available in the seasonal-monthly-ocean CDS dataset) are first interpolated from the original ORCA025 grid (1/4 degree resolution) to a regular 0.25x-0.25-degree grid using 4 nearest neighbours inverse distance weighting interpolation (Shepard method), and they are eventually interpolated to the C3S 1x1-degree grid using a grid-box-average interpolation.

Vertical interpolation

The IFS uses sigma coordinates in the vertical, meaning that near-surface model levels follow the orography. When spectral fields are transformed to pressure levels, such as temperature at 850 hPa, the resulting fields are output in spectral space and thus must be spatially complete. Interpolation from the IFS sigma coordinates to pressure levels must be done in grid-point space, and to allow the required globally complete fields uses complex extrapolation formulae for temperature below the surface of any orography. A similar approach is used when calculating global mean sea-level pressure fields.  The detailed formulae used by the IFS are contained in the FULLPOS documentation by Karim Yessad, maintained at CNRM.http://www.umr-cnrm.fr/gmapdoc/IMG/pdf/ykfpos43.pdf.


8. Where to find more information

ECMWF seasonal forecast documentation page