Implementation of Seasonal Forecast SEAS5

#newfcsystem

Description of upgrade

The fifth generation of the ECMWF seasonal forecasting system, in short SEAS5, will be introduced in the autumn of 2017, replacing System 4, which was released in 2011. SEAS5 includes updated versions of the atmospheric (IFS) and interactive ocean (NEMO) models and adds the interactive sea ice model LIM2. The IFS uses a new grid and horizontal resolution has been increased (details below). Ocean horizontal and vertical resolution have also been increased.  Ocean and land initial conditions have been updated, and the re-forecast ensemble size has been increased from 15 to 25. While re-forecasts span 1981 to 2016, the re-forecast period used to calibrate the forecasts when creating products will use the more recent period 1993 to 2016. SEAS5 highlights include a marked improvement in SST drift, especially in the tropical Pacific, and improvements in the prediction skill of Arctic sea ice.

The page will be updated as required. It was last changed on 05.07.2017

For a record of changes made to this page please refer to  Document versions.

Further information and advice regarding the upgrade can be obtained from User Support.

Timetable for implementation

The planned timetable for the implementation of SEAS5 is as follows:

The timetable represents current expectations and may change in light of actual progress made.

Current Status

The Beta testing phase of SEAS5 has started. SEAS5 re-forecast and forecast data are both available in MARS.

Summary description of SEAS5

With few exceptions, SEAS5 is configured to be identical to the extended part of the ENS, as implemented with IFS Cycle 43r1. In particular, the same resolution is used for both the monthly and seasonal forecast ranges. However, while the ENS will be regularly upgraded over the next few years with each new cycle of the IFS, the SEAS5 will remain fixed at IFS Cycle 43r1. Previous seasonal forecast systems have had a lifetime of about 5 years, and although future upgrades of SEAS may happen more often there is not yet a planned date for the introduction of SEAS6.

The table below summarizes the upgrades in model components and initialization for SEAS5.

Meteorological content of SEAS5

Atmosphere model

The IFS has been upgraded from cycle 36r4 to 43r1. In this time many changes have been made to model physics (surface drag, boundary layer entrainment, shallow convection, vertical diffusion, orographic drag, surface climate fields, CO2/CH4 climatologies, convective detrainment, cloud scheme physical processes, introduction of a lake model, radiation-surface interactions, solar zenith angle calculations, freezing rain physics, cloud erosion, CAMS O3 climatology, boundary layer cloud) and to model numerics (revised semi-Lagrangian extrapolation, revised interpolation of moist variables, cubic truncation and octahedral grid, semi-lagrangian departure point iterations, spectral viscosity). Further details can be found in the documentation of the evolution of the IFS.

The configuration of the IFS used in SEAS5 is almost identical to that used in ENS, but there are a small number of differences. In order to allow a reasonable evolution of the QBO in the seasonal forecasts, the amplitude of the tropical non-orographic gravity wave drag forcing has been re-tuned to account for the enhanced level of resolved wave activity in the model stratosphere - the new high resolution cubic grid is much more active than the previous low resolution linear grid. The tropospheric sulphate aerosol climatology is not the default fixed version, but the time varying CMIP5 climatology as used in ERA5. Volcanic stratospheric sulphate aerosol continues to be treated as in System 4 - a simple specification of an initial value which then decays towards a climatological value. This is important to allow a representation of any future major volcanic eruptions in the operational forecasts. Although the methodology is the same, the GISS dataset used has been updated to the most recent version, which has a higher background level of stratospheric aerosol than before. Compared to System 4, note that the ozone is no longer radiatively interactive in the model - the default Cycle 43r1 climatology is used instead, just as in ENS.  Ozone is thus included as an essentially passive variable (it has a very minor influence on the calculation of the vertical distribution of volcanic aerosol), and is calculated from a new ozone scheme.

Initial conditions for the IFS largely come from ERA-Interim for the re-forecasts and then from the operational analysis from 1 Jan 2017 onwards. The land surface initial state for the re-forecasts uses the same technique as in System 4 (an offline run of a matching version of the land surface model at the required resolution, forced by ERA-Interim). Unlike System 4, no GPCP correction was made to the ERA-Interim forcing because this has been shown to be unhelpful, especially over Europe. The real-time land surface is taken from the operational analysis. As in System 4, a limiter is used to prevent the real-time land surface values taking unrealistic values relative to those used in the re-forecasts. This is necessary as a safety constraint - the operational analysis is at a much higher resolution, and interpolating this to the lower resolution needed by SEAS5 can result in locally large differences compared to initial conditions prepared directly at the lower resolution. Future changes in the operational analysis may also introduce further incompatibilities in the land initial conditions, and the limiter will act as a safety constraint in these situations, too. As in System 4, ozone is initialized from climatology, although in SEAS5 the climatology is derived from a constrained run of the ozone model.

Ocean model: higher resolution and wave interaction

As for System 4, SEAS5 uses the NEMO (Nucleus for European Modelling of the Ocean) ocean model (Madec 2008), but with upgrades in the model version, ocean physics and resolution. The resolution has been increased from 1 degree and 42 layers in SEAS4 to 0.25 degrees and 75 layers in SEAS5. The vertical resolution is particularly refined in the uppermost part of the ocean, with an increase in the number of levels in the first 50 metres from 5 to 18. The increase in horizontal resolution improves the representation of sharp fronts and ocean transports. The vertical resolution increase means that the diurnal cycle of sea-surface temperatures (SST) is much better captured, with a 1-metre top level in the new configuration compared to the previous 10-metre top level.

The ocean model configuration is based on that developed by the DRAKKAR for the NEMO version V3.4 (SEAS4 used V3.2), and contains upgrades introduced at ECMWF regarding aspects of ocean-surface wave interaction (Breivik et al. 2015). These aspects include a momentum flux estimated from the dissipation term (accounting for the intensity of breaking waves) ; the surface boundary condition of the turbulent kinetic energy equation, which now account for the energy flux from breaking waves (Craig and Banner 1994); and the Coriolis-Stokes forcing term is introduced in the momentum equation

Prognostic Sea Ice model

The interactive sea-ice model LIM2, the Louvain-la-Neuve Sea Ice Model developed at the Belgian Université Catholique de Louvain, has been implemented in SEAS5. LIM2 is part of the NEMO modelling framework also used at ECMWF to model the ocean. This is the first time that the ECMWF seasonal system includes a prognostic sea-ice model. In SEAS4, the sea-ice was prescribed by sampling the recent history of sea-ice occurrences. The prognostic sea-ice model allows sea-ice cover to respond to changes in the atmosphere and ocean states, and intends to capture inter-annual variability and trends in the sea-ice over. Therefore SEAS5 should be able to provide seasonal outlooks of sea-ice cover, which is a product of interest for users.

Ocean Initial conditions: ensemble of ocean reanalyses

SEAS5 ocean and sea-ice initial conditions are provided by the new ocean analysis and reanalysis ensemble (ORAS5). ORAS5 uses the same ocean model and sea-ice as the couple forecasts in SEAS5, and consists on 5 ensemble members covering the period 1975 to the present. Compared to its predecessor ORAS4 (Balmaseda et al 2013), used to initialized SEAS4, ORAS5 has higher resolution, updated data assimilation and observational data sets, and most importantly, provides sea-ice initial conditions by assimilating sea-ice concentration.

ORAS5 has been developed based on Ocean Reanalysis Pilot 5 (ORAP5) (Zuo et al. 2015; Tietsche et al. 2015), using updated observational data sets. The ocean in-situ temperature and salinity comes from the recent quality-controlled EN4 (Good et al. 2013), which has higher vertical resolution and fuller spatial coverage than the previous version EN3 - used in ORAP5 and ORAS4. The altimeter sea-level data have also been updated to the latest version (DUACS2014) from CMEMS (Copernicus Marine Environmental Monitoring Services). The underlying SST analysis before 2008 has also been changed in ORAS5: it comes from the HadISST2 dataset, the same used in the currently ERA5 reanalysis. From 2008 onwards the SST and sea-ice are given by the OSTIA product, which is also used in the ECMWF NWP operational analysis.

The perturbation scheme used to generate the ensemble of reanalyses has also been modified. The scheme consists of two distinct elements: perturbations to the assimilated observations, both profiles and surface, and perturbations to the surface forcing fields. For details see Zuo et al 2017, Technical Memorandum 795).

Meteorological impact of SEAS5

Pre-implementation testing showed that the model version planned for SEAS5 gives substantial improvements in SST bias in the tropical Pacific, with consequent improvements in ENSO forecast skill.  A more comprehensive assessment of the meteorological impact will be given based on analysis of the SEAS5 re-forecasts once they are completed. Assessments of mid-latitude forecast skill are subject to considerable sampling error, and the limited sample sizes available from pre-implementation testing are not considered reliable predictors of the eventual outcome of the SEAS5 re-forecast dataset.

The graphical products for SEAS5 show anomalies with respect to a 1993-2016 base period, whereas System 4 used a 1981-2010 base period. Due to global warming, the use of a more recent base period results in significant changes to the plotted anomalies, particularly for temperature and geopotential height. Thus although the actual (calibrated) temperature forecast may be very similar, the plotted anomaly can look quite different. Users are encouraged to calibrate the forecasts according to their own needs by using the re-forecast data available in MARS. Re-forecasts cover 1981-2016 to maximize the data available for skill estimation and calibration.

Technical details of SEAS5

Changes to GRIB encoding

Model identifiers

The GRIB model identifiers (generating process identification number) for SEAS5 will be changed as follows:

GRIB 2 data

SEAS5 model level data is coded in GRIB 2, compared to GRIB 1 for System 4. The change to GRIB 2 for model level data was introduced in IFS Cycle 37r2 in 2011. See under User impact below for more information.

Use of the octahedral reduced Gaussian grid

SEAS5 introduces a new form of the reduced Gaussian grid, the octahedral grid, for the atmospheric (mostly) single level data.  See Introducing the octahedral reduced Gaussian grid for further details.

New model output parameters and changes to existing parameters

Surface parameters added

Several new parameters have been added related to the lake model, 6h instantaneous stresses have been added to help user applications such as wave modelling, and top incoming solar radiation has been added with a 24 hour output frequency. Note also that surface and sub-surface runoff are now fully available (in System 4, these parameters were available for the real-time forecasts, but only a limited number of the re-forecasts).

Surface fields output at additional timesteps

Some surface fields previously archived every 24 hours are now archived every 6 hours. As well as increased availability of synoptic data, this also changes the monthly means for the instantaneous fields such as SST and soil temperature. These are now calculated from 6 hourly data, and no longer have the diurnal cycle aliased into them.  (For accumulated fields, the monthly mean rate of accumulation is independent of whether synoptic data is written at 6 or 24h intervals).

Model level data

SEAS5 model level data are now produced to 6 months instead of 5 months for System 4.

Ocean waves parameters added

All fields are output every 24 hours

Impact on users

Software

EMOSLIB version 443, GRIB API version 1.17.0 or ecCodes version 2.0.0 are the minimum versions recommended to manipulate all SEAS5 fields.

Increased field sizes

The size of the fields archived have increased by a factor 2 to 4.. When retrieving data via MARS or dissemination, if no spectral truncation or grid resolutions are specified, fields are provided at the model resolution.

In particular, users should be aware of the increase in memory and CPU time needed to process the increased resolution fields and adjust their programs and batch scripts appropriately.

Availability of SEAS5 Data

Re-forecast data in MARS

The SEAS5 re-forecast data can be accessed in MARS from:

• Daily data (stream MMSF) - example for May 2001
• Monthly mean data (stream MSMM) - example for 2001

Only registered users with access to MARS will be able to access the re-forecast data sets.

At present the re-forecast data for May, June, July, August and November start dates have been released. Other re-forecast data will be accessible and validated by ECMWF as they are produced.

Test SEAS5 Forecast data in MARS

SEAS5 test data are available in MARS with experiment version 9001 (MARS keyword EXPVER=9001) starting from 1 May 2017 at  00UTC.

The data can be accessed in MARS from:

• Atmospheric forecasts (class=od, stream=mmsf, expver=9001, system=5, method=1)
•  Wave forecasts (class=od, stream=wasf, expver=9001, system=5, method=1)
• Atmospheric monthly means (class=od, stream=msmm, expver=9001)
• Wave monthly means (class=od, stream=swmm, expver=9001)

• Seasonal forecast monthly anomalies (class=od, stream=mmsa, expver=9001)

Only registered users with access to MARS with permission to access real-time data will be able to access these test data sets.

The SEAS5 Forecast beta release data are intended for testing technical aspects and should not be used for operational forecasting.  Please report any problems you find with this data to User Support.

Test real-time SEAS5 data in dissemination

SEAS5 data will become available in dissemination before we reach the release candidate testing phase. More details will be provided nearer the time.

Parallel run of System 4 and SEAS5

SEAS5 has run for May 2017. We will continue running SEAS5 for June and July and aim at running SEAS5 in near real-time from August onwards. SEAS5 data will then become available a few days after the System 4 data have been released.

On the switch-over date, presently expected to be for the November run, SEAS5 will become the operational system. System 4 will continue to be run as a near real-time system, for a limited period of time. Data from System 4 will be available from MARS and through dissemination for those interested. We intend to stop running System 4 before March 2018. Users are advised to make sure that all their applications are switched to using SEAS5 data as soon as possible after the implementation.

Graphical display of SEAS5 test forecasts

Graphical products from the seasonal forecast system are displayed on the Forecast Charts pages of the ECMWF website. We will make SEAS5 graphical products available on the ECMWF website with the release candidate testing.

References

• Madec G (2008) NEMO ocean engine. Tech. rep., Institut Pierre-Simon Laplace (IPSL), URL https://www.nemo-ocean.eu/doc/

• Good, S. A., M. J. M., and R. N. A., 2013: EN4: Quality controlled ocean temperature and salinity profiles and monthly objective analyses with uncertainty estimates. J. Phys. Oceanogr.,118, 6704–6716, doi:10.1002/2013JC009067.

• Fichefet T,Maqueda MAM (1997) Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. Journal of Geophysical Research 102(C6):12,609–12,646, DOI 10.1029/97JC00480,URL http://doi.wiley.com/10.1029/97JC00480

• Balmaseda MA, Mogensen K, Weaver AT (2013) Evaluation of the ECMWF ocean reanalysis system ORAS4. Quarterly Journal of the Royal Meteorological Society 139(674):1132{1161

• Steffen Tietsche, Magdalena a. Balmaseda, Hao Zuo, and Kristian Mogensen. Arctic sea ice in the global eddy-permitting ocean reanalysis ORAP5. Climate Dynamics, jun 2015. ISSN 0930-7575. doi: 10.1007/s00382-015-2673-3. URL http://link.springer.com/10.1007/s00382-015-2673-3.

• Hao Zuo, Magdalena A Balmaseda, and Kristian Mogensen. The new eddy-permitting ORAP5 ocean reanalysis: description, evaluation and uncertainties in climate signals. Climate Dynamics, 2015. doi: 10.1007/s00382-015-2675-1. URL http://link.springer.com/article/10.1007/s00382-015-2675-1.

• Zuo, H., M.A. Balmaseda, E. Boisseson and S. Hirahara, 2017: A new ensemble generation method for ocean reanalyses. ECMWF Technical Memorandum 795.

• Zuo et al 2018 (in preparation)

Document versions