2.2 Modelling of Thermohaline and Wind-Driven Circulation
65
Böning and Semtner
years, model simulations have increasingly made
use of synoptic analyses (both operational fields
and re-analysis products) of global meteorological
data produced by the major weather forecasting
centres.
The improvements in wind stress fields by
meteorological analyses have derived rather directly
from higher atmospheric resolution and, in the case
of re-analyses, from the application of a consistent
algorithm over several decades. Improvements in
the fluxes of heat and moisture have been slow and
uneven; thus, there has been a continuing need for
ocean modellers to scrutinize and sometimes correct the atmospheric fields available for computing
these fluxes. Another important difference between
the wind and thermohaline forcing functions of
ocean circulation models concerns the physics of
the air–sea exchange: due to the local feedback of
the ocean on the heat and freshwater fluxes, the
formulation of the thermohaline forcing must
include a parameterization that accounts for this
retroaction. For a more comprehensive discussion
of both the accuracy of available air–sea flux estimates and the definition of consistent parameterizations, the reader is referred to Barnier (1998).
The distribution of large-scale, seasonal and
intraseasonal sea level variability and its relation
to changes in ocean circulation, were elucidated in
coarse resolution (1° by 2°) global model experiments (Chao and Fu, 1995; Fukumori et al.,
1998). Steric changes of sea level, primarily due to
seasonal changes in short-wave radiation, were in
general smaller than the wind-driven variability,
except for the mid-latitudes around 30°N and
30°S (see also Stammer, 1997a). Wind-driven
changes were largely baroclinic in the tropics, and
dominated by the annual harmonic. Large-scale
sea-level variability at higher latitudes was associated with barotropic motions, and dominated by
high frequencies that reflect the general frequency
spectra of the wind stress curl. The reliability of
the coarse resolution simulations was demonstrated by a significant correlation with altimetric
data (spatially smoothed to suppress mesoscale
energy) (Chao and Fu, 1995), and qualitative similarity of frequency spectra with altimetric and tide
gauge measurements (Fukumori et al., 1998).
The simulation of local ocean variability is complicated due to the presence of mesoscale eddies
and potential effects of small-scale topographic
features. Insight into the realism of the forced
upper-ocean variability in a high-resolution, global
model has been obtained by comparing its sea
surface height variations with both tide gauge data
and altimetric data (Tokmakian, 1996). The model
is the 1/4° Parallel Ocean Climate Model (POCM)
that had been built on the near-global model of
Semtner and Chervin (1988, 1992). It is forced
here with a time series of European Centre for
Medium Range Weather Forecasting (ECMWF)
fields from 1979 to the present. Figure 2.2.3 compares simulated time series of Sea Surface Height
(SSH) with a suite of tide gauges from Pacific
coastal and island sites. There is a qualitative
agreement for many of the stations. However,
median correlation for all gauges after removal of
the seasonal cycle is only 0.4, indicating that
much of the variance is not related to the local
atmospheric forcing. There are, however, regional
differences; e.g. correlation is higher (0.7) for areas
dominated by ENSO signals. There is an indication
that ENSO signals at middle to high latitudes along
the coast of the Americas are more attenuated
in the simulation than in reality – this may indicate
that the model grid is still too coarse to allow the
model’s biharmonic damping coefficients to be
small enough not to interfere with propagation of
Kelvin and shelf waves. The long-term information
at individual sites is complemented by the temporally more restricted (5 years) but global perspective from altimeter data. It indicates highest
correlations between the model’s SSH variations
and the observed fields for the tropical oceans and
throughout those regions of the (northern) hemisphere outside of the Kuroshio and Gulf Stream
extensions, which are dominated by internally
generated, mesoscale variability. The correlations
in the southern hemisphere are generally lower,
possibly an indication of less accurate wind forcing
fields due to fewer observations (Tokmakian, 1998).
A comparison of POCM results and (TOPEX/
POSEIDON (T/P) variability using Hovmöller
diagrams for various latitudes in the Pacific also
showed excellent tracking of phenomena, especially
in the lower latitudes (Stammer, 1997a). Global
comparisons of 1/4° and 1/6° models against T/P
have been made to show large-scale anomaly patterns on intraseasonal, annual and interannual time
scales (Stammer et al., 1996; Fu and Smith, 1996).
As in the case of the coarse-resolution model simulations mentioned above, there was a generally
good agreement with respect to the large-scale
65
Böning and Semtner
years, model simulations have increasingly made
use of synoptic analyses (both operational fields
and re-analysis products) of global meteorological
data produced by the major weather forecasting
centres.
The improvements in wind stress fields by
meteorological analyses have derived rather directly
from higher atmospheric resolution and, in the case
of re-analyses, from the application of a consistent
algorithm over several decades. Improvements in
the fluxes of heat and moisture have been slow and
uneven; thus, there has been a continuing need for
ocean modellers to scrutinize and sometimes correct the atmospheric fields available for computing
these fluxes. Another important difference between
the wind and thermohaline forcing functions of
ocean circulation models concerns the physics of
the air–sea exchange: due to the local feedback of
the ocean on the heat and freshwater fluxes, the
formulation of the thermohaline forcing must
include a parameterization that accounts for this
retroaction. For a more comprehensive discussion
of both the accuracy of available air–sea flux estimates and the definition of consistent parameterizations, the reader is referred to Barnier (1998).
The distribution of large-scale, seasonal and
intraseasonal sea level variability and its relation
to changes in ocean circulation, were elucidated in
coarse resolution (1° by 2°) global model experiments (Chao and Fu, 1995; Fukumori et al.,
1998). Steric changes of sea level, primarily due to
seasonal changes in short-wave radiation, were in
general smaller than the wind-driven variability,
except for the mid-latitudes around 30°N and
30°S (see also Stammer, 1997a). Wind-driven
changes were largely baroclinic in the tropics, and
dominated by the annual harmonic. Large-scale
sea-level variability at higher latitudes was associated with barotropic motions, and dominated by
high frequencies that reflect the general frequency
spectra of the wind stress curl. The reliability of
the coarse resolution simulations was demonstrated by a significant correlation with altimetric
data (spatially smoothed to suppress mesoscale
energy) (Chao and Fu, 1995), and qualitative similarity of frequency spectra with altimetric and tide
gauge measurements (Fukumori et al., 1998).
The simulation of local ocean variability is complicated due to the presence of mesoscale eddies
and potential effects of small-scale topographic
features. Insight into the realism of the forced
upper-ocean variability in a high-resolution, global
model has been obtained by comparing its sea
surface height variations with both tide gauge data
and altimetric data (Tokmakian, 1996). The model
is the 1/4° Parallel Ocean Climate Model (POCM)
that had been built on the near-global model of
Semtner and Chervin (1988, 1992). It is forced
here with a time series of European Centre for
Medium Range Weather Forecasting (ECMWF)
fields from 1979 to the present. Figure 2.2.3 compares simulated time series of Sea Surface Height
(SSH) with a suite of tide gauges from Pacific
coastal and island sites. There is a qualitative
agreement for many of the stations. However,
median correlation for all gauges after removal of
the seasonal cycle is only 0.4, indicating that
much of the variance is not related to the local
atmospheric forcing. There are, however, regional
differences; e.g. correlation is higher (0.7) for areas
dominated by ENSO signals. There is an indication
that ENSO signals at middle to high latitudes along
the coast of the Americas are more attenuated
in the simulation than in reality – this may indicate
that the model grid is still too coarse to allow the
model’s biharmonic damping coefficients to be
small enough not to interfere with propagation of
Kelvin and shelf waves. The long-term information
at individual sites is complemented by the temporally more restricted (5 years) but global perspective from altimeter data. It indicates highest
correlations between the model’s SSH variations
and the observed fields for the tropical oceans and
throughout those regions of the (northern) hemisphere outside of the Kuroshio and Gulf Stream
extensions, which are dominated by internally
generated, mesoscale variability. The correlations
in the southern hemisphere are generally lower,
possibly an indication of less accurate wind forcing
fields due to fewer observations (Tokmakian, 1998).
A comparison of POCM results and (TOPEX/
POSEIDON (T/P) variability using Hovmöller
diagrams for various latitudes in the Pacific also
showed excellent tracking of phenomena, especially
in the lower latitudes (Stammer, 1997a). Global
comparisons of 1/4° and 1/6° models against T/P
have been made to show large-scale anomaly patterns on intraseasonal, annual and interannual time
scales (Stammer et al., 1996; Fu and Smith, 1996).
As in the case of the coarse-resolution model simulations mentioned above, there was a generally
good agreement with respect to the large-scale