likely to happen in the next few years. The horizontal resolution anticipated for the ocean component of the next generation of climate system
models is of the order of 1°; while further decrease
in grid size for single experiments may be possible,
this has to be balanced against the need (and associated costs) for including more components of the
climate system, and for ensembles of experiments integrated over long time spans (WOCE
International Project Office, 1999; see also Wood
and Bryan, Chapter 2.3). Obviously, the critical
issue that continues to separate coarse-resolution
and fine-resolution ocean modelling is integration
period: in order to remove transients in the deep
interior temperature and salinity fields and to
achieve a state of near-thermodynamic equilibrium
between the field variables and the surface fluxes,
global models need to be run over time spans of
several thousand years, a time scale effectively set
by the pace of the weak advective and diffusive
processes in the abyssal ocean.
There is little doubt that in order to be useful in
coupled climate studies ocean models need to
demonstrate sufficient realism in equilibrium solutions. Obviously, it is only through long-term
integrations that the effect of parameterizations of
sub-grid-scale mixing processes on water mass
structure and thermohaline circulation will become
apparent. As has been demonstrated by a number
of authors, the diffusion across isopycnal surfaces
not only governs pycnocline structure and dynamics
(Salmon and Hollerbach, 1991), but also latitudinal
property fluxes associated with the thermohaline
overturning circulation (Bryan, 1987). Accordingly,
comparison of simulated property distributions
with the mean hydrographic state of the ocean represents a prime means of assessing model mixing
parameterizations. (It should be noted, however,
that although this means of testing equilibrium
solutions utilizes the prime source of information
gathered in the history of oceanographic field work,
it is built on the implicit assumption that the
observed state of the ocean is in near-thermodynamic equilibrium: as discussed, e.g. in Chapter
7.3 (Dickson et al.), the WOCE period appears
unusual enough that this has to regarded with some
caution.) For example, some of the typical deficiencies of ocean climate models with respect to pycnocline sharpness and meridional heat transport have
been attributed to either spurious mixing associated
with the parameterization scheme (Danabasoglu
et al., 1994; England and Hirst, 1997; England and
Rahmstorf, 1999) or the chosen profiles of vertical
diffusivity (Cummins et al., 1990; Tsuijino et al.,
2000). In the last few years there has been considerable success in eliminating spurious diapycnic
mixing in z-coordinate models by adopting
improved parameterization schemes for eddy
transports, building on the method of Gent and
McWilliams (1990) (see also Section 2.2.4.2); there
have also been first attempts to assess the effect
on global circulation patterns of enhanced vertical
diffusivity over rough bathymetry (Hasumi and
Suginohara, 1999a).
However, while controlling the secular behaviour of the water mass properties and thermohaline circulation, the slow small-scale interior
mixing processes have little influence on the
response of the flow field to atmospheric forcing
anomalies at shorter time scales, a situation analogous to the spin-up of an ocean circulation model
from given hydrographic conditions. It has been
shown in both idealized (Kawase, 1987) and full
general circulation models (Döscher et al., 1994;
Gerdes and Köberle, 1995) that an initial dynamic
adjustment to thermohaline forcing is effectively
obtained on the same time scale as for a winddriven circulation (Anderson and Killworth,
1977), set by the travel time of baroclinic Rossby
waves through the basin. Specifically, changes in
the properties of Denmark Strait Overflow Water
(DSOW), in the northern Irminger Basin, are communicated rapidly towards the equator via continental shelf waves along the western boundary,
leaving a signature in both deep and upper western
boundary currents, meridional overturning and
poleward heat transport on time scales of 1–2
years; through the interior baroclinic adjustment
process it then takes 10–15 years before a new
quasi-equilibrium is established in the mid-latitude
North Atlantic (Fig. 2.2.2). During the course of
the initial, dynamic adjustment phase – that is, the
first few decades of an ocean model spin-up – the
large-scale water mass properties below the thermocline remain close to the initial conditions,
except for regions of strong currents or near the
deep water formation sites.
The differences in the governing mechanisms
and dynamical balances between a long-term equilibrium state of the system and its response to
changes in the surface fluxes have some important
implications. First, it is the large difference in the
2.2 Modelling of Thermohaline and Wind-Driven Circulation
63
Böning and Semtner
models is of the order of 1°; while further decrease
in grid size for single experiments may be possible,
this has to be balanced against the need (and associated costs) for including more components of the
climate system, and for ensembles of experiments integrated over long time spans (WOCE
International Project Office, 1999; see also Wood
and Bryan, Chapter 2.3). Obviously, the critical
issue that continues to separate coarse-resolution
and fine-resolution ocean modelling is integration
period: in order to remove transients in the deep
interior temperature and salinity fields and to
achieve a state of near-thermodynamic equilibrium
between the field variables and the surface fluxes,
global models need to be run over time spans of
several thousand years, a time scale effectively set
by the pace of the weak advective and diffusive
processes in the abyssal ocean.
There is little doubt that in order to be useful in
coupled climate studies ocean models need to
demonstrate sufficient realism in equilibrium solutions. Obviously, it is only through long-term
integrations that the effect of parameterizations of
sub-grid-scale mixing processes on water mass
structure and thermohaline circulation will become
apparent. As has been demonstrated by a number
of authors, the diffusion across isopycnal surfaces
not only governs pycnocline structure and dynamics
(Salmon and Hollerbach, 1991), but also latitudinal
property fluxes associated with the thermohaline
overturning circulation (Bryan, 1987). Accordingly,
comparison of simulated property distributions
with the mean hydrographic state of the ocean represents a prime means of assessing model mixing
parameterizations. (It should be noted, however,
that although this means of testing equilibrium
solutions utilizes the prime source of information
gathered in the history of oceanographic field work,
it is built on the implicit assumption that the
observed state of the ocean is in near-thermodynamic equilibrium: as discussed, e.g. in Chapter
7.3 (Dickson et al.), the WOCE period appears
unusual enough that this has to regarded with some
caution.) For example, some of the typical deficiencies of ocean climate models with respect to pycnocline sharpness and meridional heat transport have
been attributed to either spurious mixing associated
with the parameterization scheme (Danabasoglu
et al., 1994; England and Hirst, 1997; England and
Rahmstorf, 1999) or the chosen profiles of vertical
diffusivity (Cummins et al., 1990; Tsuijino et al.,
2000). In the last few years there has been considerable success in eliminating spurious diapycnic
mixing in z-coordinate models by adopting
improved parameterization schemes for eddy
transports, building on the method of Gent and
McWilliams (1990) (see also Section 2.2.4.2); there
have also been first attempts to assess the effect
on global circulation patterns of enhanced vertical
diffusivity over rough bathymetry (Hasumi and
Suginohara, 1999a).
However, while controlling the secular behaviour of the water mass properties and thermohaline circulation, the slow small-scale interior
mixing processes have little influence on the
response of the flow field to atmospheric forcing
anomalies at shorter time scales, a situation analogous to the spin-up of an ocean circulation model
from given hydrographic conditions. It has been
shown in both idealized (Kawase, 1987) and full
general circulation models (Döscher et al., 1994;
Gerdes and Köberle, 1995) that an initial dynamic
adjustment to thermohaline forcing is effectively
obtained on the same time scale as for a winddriven circulation (Anderson and Killworth,
1977), set by the travel time of baroclinic Rossby
waves through the basin. Specifically, changes in
the properties of Denmark Strait Overflow Water
(DSOW), in the northern Irminger Basin, are communicated rapidly towards the equator via continental shelf waves along the western boundary,
leaving a signature in both deep and upper western
boundary currents, meridional overturning and
poleward heat transport on time scales of 1–2
years; through the interior baroclinic adjustment
process it then takes 10–15 years before a new
quasi-equilibrium is established in the mid-latitude
North Atlantic (Fig. 2.2.2). During the course of
the initial, dynamic adjustment phase – that is, the
first few decades of an ocean model spin-up – the
large-scale water mass properties below the thermocline remain close to the initial conditions,
except for regions of strong currents or near the
deep water formation sites.
The differences in the governing mechanisms
and dynamical balances between a long-term equilibrium state of the system and its response to
changes in the surface fluxes have some important
implications. First, it is the large difference in the
2.2 Modelling of Thermohaline and Wind-Driven Circulation
63
Böning and Semtner