ENACT Common setup for analyses

ENACT ocean analyses: spinup and relaxation

A common strategy for spinup, and for relaxation to sub-surface and surface fields during spinup and during the assimilation phase, was required for the ENACT ocean analyses.

 This is a summary of the current design.

Spinup and relaxation issues for ENACT ocean analyses

Before starting an ocean analysis sequence, with prescribed surface fluxes and with or without assimilation of ocean observations, an appropriate initial (time t_start_analysis) OGCM state is needed that is an estimate of the actual ocean state at that time. This can be obtained by starting the OGCM from some other ocean state (e.g. climatology, or an 'analogue' analysis) at time t_start_spinup and 'spinning up' with prescribed forcing and no assimilation to t_start_analysis.

During spinup, and during the analysis phase, some form of relaxation to prescribed temperature T and salinity S fields may be used to:

1. Reduce subsurface drift
2. Impose observed SST

For ENACT, the ocean analyses have two main and simultaneous purposes:

1. Obtain estimates of the ocean state extending over several years (up to 40 years)
2. Provide initial ocean conditions for seasonal to multi-annual range forecasts

Spinup times

The agreed spinup strategy for ENACT is:

5 years climatological ERA40 forcing + 4 years actual ERA40 forcing

Stream 1: t_start_analysis = 1jan1987

• Start on 1jan1978 from Levitus98 January climatology
• Use ERA-40 climatological surface fluxes (1971-2000 daily climatology) up to end 1982
• Use daily ERA-40 surface fluxes from 1jan1983 to 31dec1986

(NB 1987 start selected for stream 1 as this is when Geosat altimeter data are available.)

Stream 2: t_start_analysis = 1jan1962

• Start on 1jan1953 from Levitus98 January climatology
• Use ERA-40 climatological surface fluxes (1971-2000 daily climatology) up to end 1957
• Use daily ERA-40 surface fluxes from 1jan1958 to 31dec1961

 (NB 1958 start selected for stream 2 to match the ERA40 atmospheric re-analysis which is available Sept 1957 to Aug 2002.)

Spinup forcing fields: ERA40 surface flux climatology

Daily climatology is being provided by ECMWF

• To be based on the 0-24hour average fields from ERA40 (as for the daily fields already supplied), EXCEPT for precipitation minus evaporation (PminusE).
• Climatology based on 1971-2000 ERA40 data
• Supplied on 1.125 x 1.125 lat-long grid

Note on PminusE:

In ERA40 there is excess rainfall in the tropics, in the first few hours (and days) of each forecast, particularly over the oceans, when satellite data are used in the atmospheric analysis. A latitude-dependent scaling factor has been applied to P data. (The E data do not need adjustment.) The scaling factor depends on the period considered. For further info, see Troccoli A. and Kallberg P. (2004), Precipitation correction in the ERA-40 reanalysis, ERA-40 Project Report Series, 13.

ECMWF provided daily corrected PminusE ERA40 climatology (ECMWF also provided uncorrected P fields and E fields separately)

ECMWF also provided corresponding corrected PminusE fields along with the other actual [non-climatological] ERA40 fields

ECMWF also provided SST data on the same datasite:

• Daily values on 1 x 1 lat-long grid(see below)
• Including 1971-2000 climatology based on HadISST climatology

Surface T relaxation

Relaxation to gridded observed SST and sea-ice analyses is used in the spinup and ocean analysis stages with a rate equivalent to 200W/m2/degC (uniform over globe)

• Monthly HadISST data up to 1982 (as in ERA40)
• Weekly Reynolds oi_v2 1982 on (from date first available).

Both are available on a 1 x 1 lat-long grid.

Methods may differ as to how the heat gain/loss provided by relaxation is distributed in the upper ocean. Some ocean analysts may add it in the top model level, some may distribute it over several levels. This choice is related to the types of mixing scheme used, and as such is regarded as part of the various systems. However, systems using the same type of OGCM (e.g. OPA) should use the same vertical distribution choice if possible, to allow better comparisons.

Time interpolation:

Simple linear interpolation between monthly (and to a lesser extent weekly) average values leads to a decrease in variability, and (for example) the monthly average of the interpolated field does not match that of the original field. To avoid this effect, a modified interpolation (using quadratic forms) that conserves the original averages would be preferable to calculate and provide daily fields from the monthly HadISST SST data. However, this has not been implemented in ENACT.

For the weekly Reynolds data, simple linear interpolation has been used to provide daily values

Sea-ice:

As SST/sea-ice analyses provide ice concentration estimates, the location of sea-ice is not clearly defined in these datasets. As sea-ice is not a high-importance topic for ENACT, we propose using a simple temperature threshold to label each 1 x 1 gridbox as ice-covered or ice-free. (A threshold can be chosen to give a reasonable approximation to observed ice extent: the Met Office currently uses this approach. For both HadISST and Reynolds oi_v2 the threshold is -1degC: i.e. ocean gridboxes with T < -1 are labelled as ice-covered.)

Spatial resolution issue:

The OGCMs have higher (tropical) resolution than the SST analyses, so (for example) the analyses may underestimate the E Pac cold tongue SST that the OGCMs are trying to resolve. It will be up to the individual systems to use their standard methods to interpolate forcing and SST data to their OGCM grid

Subsurface relaxation:

Some form of weak subsurface relaxation to climatology is desirable, to control model drift but not suppress interannual/decadal variability. Alberto Troccoli has carried out tests using various subsurface relaxation timescales.

For ENACT:

• Weak global subsurface relaxation to Levitus98 gridded T and S monthly climatology, smoothed with a 3-month running mean
• The smoothed climatology (monthly fields on Levitus grid, in NetCDF format) is available at ECMWF along with the forcing fields
• Relaxation with a 3-year timescale at all vertical levels and all lat-long locations
• Standard linear time interpolation

Note: the subsurface relaxation effectively also provides weak relaxation to Levitus T and S at the top ocean model level. For T this is dominated by the much stronger relaxation to SST described above. For S this is effectively a weak sea surface salinity relaxation.

Note: the OPA OGCM contains some in-built local variations, such as decreased relaxation near coasts. These variations are regarded as part of the particular ocean system, that should be common to all the OPA models in ENACT.