396
an equilibrium atmospheric model. Latif and Villwock (1990) also used hybrid models in their
studies. They developed two HCMs both using the same oceanic GCM (that of Latif (1987))
but with different model atmospheres. The first atmospheric model was purely statistical with
no explicit dynamics. The second one used the linear atmosphere described by Opsteegh and
Murea.u (1984). Latif and Fliigel (1991) conducted a large ensemble of predictions experiments
with a simplified coupled ocean-atmosphere model consisting of the Latif (1987) OGCM coupled
to the same simple atmospheric feedback model as in Latif and Villwock (1990). Here the
feedback of the atmosphere on the ocean was described by a linear relation between the stress
anomaly at each grid point and the local SST anomaly field. The set of linear coefficients
was determined empirically by a regresion analysis using SST anomalies from the uncoupled
ocean model simulation with observed wind stress for the 25-year period 1961-85. Bamett et
0.1. (1992) also used the Latif (1987) OGCM and atmospheric statistical model based on EOF
(Empirical Orthogonal Function) analysis of 19 years (1967-85) of monthly SST and surface
wind stress data from the tropical Pacific. Graham et o.l. (1992) used a shallow-water model
coupled to an atmosphere represented by a simple field to field relation. Syu et 0.1. (1993)
considered a version of GFDL Modular Ocean Model (Paca.nowski, Dixon and Rosati, 1991)
based on the Cox (1984) model and a linear relationship between wind stress anomalies and
SST anomalies using the SVD (Singular Value Descomposition) technique, FSU wind data and
COADS (Comprehensive Ocean-Atmosphere Data Set) SST for a period of 21-years. Balmaseda
(1993) forced an intermediate 2 ~ ocean model (Anderson) for the 31-year period (1961-1991).
The SST field obtained was then used to construct a linear and non-local empirical atmosphere
by means of an EOF technique which was then subsequently coupled to the same ocean model.
The aim of this paper is to describe the 33-year forced reference experiment from which model
SSTs for this period wil be extracted. These SSTs together with FSU wind data for the same
period will be used to build the empirical atmosphere that will constitute the atmospheric
component in the HCM. The forced experiment will also serve as reference for the coupled results.
The atmospheric model will be constructed following the Balmaseda (1993) approach, giving the
advantage of a full OGCM in the forced and coupled experiments instead an intermediate ocean
model. With regard to the Latif and Villwork (1990) and the Latif and Fliigel (1991) statistical
models this approach differs in the non-local nature of the linear relation deduced between
SSTs and wind stresses. Also a longer time period is considered. Bamett et al. (1992) take
observations of SST and wind stress over the 19-year period 1967-85 to construct their statistical
atmosphere and not the SST obtained from a forced ocean experience.
2 The Ocean Model
The model used in this study is the General Circulation Model developed at LODYC known as
OPA (Delecluse et al., 1993). This model solves the primitive equations for the ocean with the
incompressibility, Boussinesq and hydrostatic approximations. The rigid lid approximation is
also made. It was adapted for use in the tropical Pacific (40 0 S-48°N, 120°E-700W) by Dandin
(1993) and tested in the TOGA-NEG intercomparison (Stockdale et 0.1., 1993).
2.1 Basic equations.
The set of primitive equations solved by the model is given by three conservation laws (momentum, temperature and salinity) which prognostic ally cOlllpute the horizontal velocity, the
potential temperature T and the salinity S. Three diagnostic equations allow the computation
of the vertical velocity W, pressure P and density p.
an equilibrium atmospheric model. Latif and Villwock (1990) also used hybrid models in their
studies. They developed two HCMs both using the same oceanic GCM (that of Latif (1987))
but with different model atmospheres. The first atmospheric model was purely statistical with
no explicit dynamics. The second one used the linear atmosphere described by Opsteegh and
Murea.u (1984). Latif and Fliigel (1991) conducted a large ensemble of predictions experiments
with a simplified coupled ocean-atmosphere model consisting of the Latif (1987) OGCM coupled
to the same simple atmospheric feedback model as in Latif and Villwock (1990). Here the
feedback of the atmosphere on the ocean was described by a linear relation between the stress
anomaly at each grid point and the local SST anomaly field. The set of linear coefficients
was determined empirically by a regresion analysis using SST anomalies from the uncoupled
ocean model simulation with observed wind stress for the 25-year period 1961-85. Bamett et
0.1. (1992) also used the Latif (1987) OGCM and atmospheric statistical model based on EOF
(Empirical Orthogonal Function) analysis of 19 years (1967-85) of monthly SST and surface
wind stress data from the tropical Pacific. Graham et o.l. (1992) used a shallow-water model
coupled to an atmosphere represented by a simple field to field relation. Syu et 0.1. (1993)
considered a version of GFDL Modular Ocean Model (Paca.nowski, Dixon and Rosati, 1991)
based on the Cox (1984) model and a linear relationship between wind stress anomalies and
SST anomalies using the SVD (Singular Value Descomposition) technique, FSU wind data and
COADS (Comprehensive Ocean-Atmosphere Data Set) SST for a period of 21-years. Balmaseda
(1993) forced an intermediate 2 ~ ocean model (Anderson) for the 31-year period (1961-1991).
The SST field obtained was then used to construct a linear and non-local empirical atmosphere
by means of an EOF technique which was then subsequently coupled to the same ocean model.
The aim of this paper is to describe the 33-year forced reference experiment from which model
SSTs for this period wil be extracted. These SSTs together with FSU wind data for the same
period will be used to build the empirical atmosphere that will constitute the atmospheric
component in the HCM. The forced experiment will also serve as reference for the coupled results.
The atmospheric model will be constructed following the Balmaseda (1993) approach, giving the
advantage of a full OGCM in the forced and coupled experiments instead an intermediate ocean
model. With regard to the Latif and Villwork (1990) and the Latif and Fliigel (1991) statistical
models this approach differs in the non-local nature of the linear relation deduced between
SSTs and wind stresses. Also a longer time period is considered. Bamett et al. (1992) take
observations of SST and wind stress over the 19-year period 1967-85 to construct their statistical
atmosphere and not the SST obtained from a forced ocean experience.
2 The Ocean Model
The model used in this study is the General Circulation Model developed at LODYC known as
OPA (Delecluse et al., 1993). This model solves the primitive equations for the ocean with the
incompressibility, Boussinesq and hydrostatic approximations. The rigid lid approximation is
also made. It was adapted for use in the tropical Pacific (40 0 S-48°N, 120°E-700W) by Dandin
(1993) and tested in the TOGA-NEG intercomparison (Stockdale et 0.1., 1993).
2.1 Basic equations.
The set of primitive equations solved by the model is given by three conservation laws (momentum, temperature and salinity) which prognostic ally cOlllpute the horizontal velocity, the
potential temperature T and the salinity S. Three diagnostic equations allow the computation
of the vertical velocity W, pressure P and density p.