311
spheric model (in turn forced with climatological sea surface temperatures,
SSTs). The fluxes of heat and freshwater on the other hand are based on a
combination of (i) observed climatology (Oberhuber, 1988), (ii) a bulk flux
parameterization and (iii) additional relaxation towards observed SSTs (averaged for the years 1979-1988) and surface salinity (Levitus, 1982). When
the restoring boundary conditions are replaced by the fluxes computed in
the coupled model, a drift towards a new model state is taking place. This
drift is partly due to inconsistencies between the dynamical fluxes of the
coupled system and the mixture of different forcing mechanisms employed
in the spin up phase. The drift is also occurring due to model errors such
as deficiencies in the simulation of clouds leading to errors in the amount
of heat reaching the ocean surface. Other uncertainties are related to the
hydrological cycle over the oceans and difficulties to reproduce the salinity
distribution due to different problems with sharp ocean currents and transport of sea ice. Major reductions have taken place in reducing the systematic errors in recent models and remaining problems are mainly confined
to higher latitudes. (Latif et al. 1994). The technique to circumvent the
problem with climate drift has been to employ a so called flux adjustment
(e.g. Sausen et al., 1988), whereby the calculated fluxes are corrected with
the fluxes occurring in the uncoupled state as derived from climatology. In
this study an approach has been applied whereby only the mean annual
drift has been adjusted. The importance of leaving the annual cycle free
is particularly valid where for example SST undergoes major changes independent of the annual heat flux, such as during El Nino (Neelin et al.,
1992).
In the current model an alternative approach (Bacher and Oberhuber,
1995) is applied. It differs from the traditionally one in two major aspects,
namely (i) the flux adjustment is computed by gradual updating during
a 100 year spin up run of the coupled model and (ii) only the annual
mean of heat and fresh water is corrected while other quantities including
wind stress are generated by the coupled model without any corrections
altogether. After the 100 years spin up with variable adjustment, the model
has been integrated for another 100 years with the flux adjustment fixed
in time (Fig. 9). A relatively small secular drift was found with a cooling
trend (all layers) of about O.l°C, a reduction of the upper ocean salinity
by ca. 0.2 psu and a salinity increase with about the same rate in the deep
ocean.
Fig. 10 shows the evolution of the globally averaged surface air temper-
spheric model (in turn forced with climatological sea surface temperatures,
SSTs). The fluxes of heat and freshwater on the other hand are based on a
combination of (i) observed climatology (Oberhuber, 1988), (ii) a bulk flux
parameterization and (iii) additional relaxation towards observed SSTs (averaged for the years 1979-1988) and surface salinity (Levitus, 1982). When
the restoring boundary conditions are replaced by the fluxes computed in
the coupled model, a drift towards a new model state is taking place. This
drift is partly due to inconsistencies between the dynamical fluxes of the
coupled system and the mixture of different forcing mechanisms employed
in the spin up phase. The drift is also occurring due to model errors such
as deficiencies in the simulation of clouds leading to errors in the amount
of heat reaching the ocean surface. Other uncertainties are related to the
hydrological cycle over the oceans and difficulties to reproduce the salinity
distribution due to different problems with sharp ocean currents and transport of sea ice. Major reductions have taken place in reducing the systematic errors in recent models and remaining problems are mainly confined
to higher latitudes. (Latif et al. 1994). The technique to circumvent the
problem with climate drift has been to employ a so called flux adjustment
(e.g. Sausen et al., 1988), whereby the calculated fluxes are corrected with
the fluxes occurring in the uncoupled state as derived from climatology. In
this study an approach has been applied whereby only the mean annual
drift has been adjusted. The importance of leaving the annual cycle free
is particularly valid where for example SST undergoes major changes independent of the annual heat flux, such as during El Nino (Neelin et al.,
1992).
In the current model an alternative approach (Bacher and Oberhuber,
1995) is applied. It differs from the traditionally one in two major aspects,
namely (i) the flux adjustment is computed by gradual updating during
a 100 year spin up run of the coupled model and (ii) only the annual
mean of heat and fresh water is corrected while other quantities including
wind stress are generated by the coupled model without any corrections
altogether. After the 100 years spin up with variable adjustment, the model
has been integrated for another 100 years with the flux adjustment fixed
in time (Fig. 9). A relatively small secular drift was found with a cooling
trend (all layers) of about O.l°C, a reduction of the upper ocean salinity
by ca. 0.2 psu and a salinity increase with about the same rate in the deep
ocean.
Fig. 10 shows the evolution of the globally averaged surface air temper-