Title: Global ocean reanalysis at CERFACS
1Global ocean reanalysis at CERFACS
- A. T. Weaver and N. Daget
- CERFACS, Toulouse
CLIVAR/GODAE Ocean Reanalysis Workshop, ECMWF, 31
Aug. 1 Sept.
2CERFACS reanalysis system specifications
- Ocean model
- OPA version 8.2 (Madec et al. 1998)
- Global, free surface configuration (ORCA grid)
- 2 deg. zonal resolution, variable meridional
resolution with equatorial refinement (1/2 deg.
at equator) - 31 vertical levels with 16 in the upper 200m
- Surface forcing fields
- ERA-40 with bias-corrected precipitation
(Troccoli and Kallberg 2004) - Delayed adjustment of P-ER based on global mean
sea level drift (adjustment timescale 1 day) - Relaxation fields
- Strong relaxation (-200Wm-2K-1) to gridded
Reynolds SST - Weak relaxation (3-year timescale) to subsurface
temperature and salinity from gridded Levitus 98
- Part of the common experimental design in the
ENACT project - (see www.ecmwf.int/research/EU_projects/ENACT
and Davey et al., CLIVAR Exchanges, Vol. 11, No.
3, 2006)
CLIVAR/GODAE Ocean reanalysis workshop, ECMWF, 31
Aug. 1 Sept.
3CERFACS reanalysis system specifications
- Assimilation algorithm
- Based on the OPAVAR source code as described in
various references Weaver et al. (2003), MWR
Vialard et al. (2003), MWR Ricci et al. (2005),
MWR Weaver et al. (2006), QJRMS - Assimilated observations and observation error
specification - T and S data profiles from the quality-controlled
ENACT/ENSEMBLES data-set, version EN2v1a (Ingleby
and Huddleston 2006, JMR) - Profiles vertically thinned (max. 5 measurements
per model layer) - Cut-off depth at 1000m cut-off latitude at
65oN/S - No data assimilated in closed or in some
semi-enclosed seas - Observation errors assumed uncorrelated
- Observation error variance profiles specified
from the globally-averaged estimates of Ingleby
and Huddleston (2006) increased near coastlines.
CLIVAR/GODAE Ocean reanalysis workshop, ECMWF, 31
Aug. 1 Sept.
4CERFACS reanalysis system specifications
- Version of OPAVAR used for the reanalysis
experiments - Multivariate 3D-Var
- Incremental formulation
- Innovation computed with the full model on the
outer loop - Increment computed by quadratic CG minimization
in the inner loop - 2 outer iterations are used, with 10 30 inner
iterations for each outer iter. - Control variables are temperature and the
unbalanced part of salinity - Multivariate constraints based on background T-S
conservation, hydrostatic balance, dynamic
height, geostrophy are used to compute the
balanced parts of S, u, v and SSH - Background error statistics for T and
unbalanced S - These variables are assumed to be mutually
uncorrelated - Horizontal and vertical (univariate) correlations
are assumed to be approximately Gaussian and are
modelled using a diffusion equation - Zonal/meridional length scales are
stretched/shrunk near the equator - Vertical correlation length scales are made
resolution-dependent - T and unbalanced S variances are parameterized in
terms of the background vertical T gradient and
background vertical mixing coefficient,
respectively - Cycled using a 10-day assimilation window and
Incremental Analysis Updates (Bloom et al. 1996,
MWR)
CLIVAR/GODAE Ocean reanalysis workshop, ECMWF, 31
Aug. 1 Sept.
5Summary of CERFACS ocean reanalysis experiments
- Exp. CFCS2 CERFACS Control (ENACT) Stream 2
- OPA-ORCA2, ERA-40, no data assimilation,
1962-2001 - Exp. CFAS2 CERFACS 3D-Var Assimilation (ENACT)
Stream 2 - OPA-ORCA2, ERA-40, 3D-Var of T and S profiles,
1962-2001 - Exp. CFASA CERFACS Assimilation (ENACT) Stream
A - OPA-ORCA2, ERA-40, upgraded 3D-Var of T and S
profiles,1993-2001 - Contributions made to intercomparison Items 3,
4, 5, 6 and 7 - 5-year spin-up prior to start of each
reanalysis stream
CLIVAR/GODAE Ocean reanalysis workshop, ECMWF, 31
Aug. 1 Sept.
6Future
- ENSEMBLES project production of ensembles of
variational analyses - Climate forecast initialization
- Ocean reanalysis
- Online background-error estimation
- Collaboration with ECMWF and MERCATOR on
development of future versions of OPAVAR
(NEMOVAR).
CLIVAR/GODAE Ocean reanalysis workshop, ECMWF, 31
Aug. 1 Sept.