eddy-permitting model by Bryan et al. (1999)
lends support to the use of non-uniform mixing
coefficients as proposed by Visbeck et al. (1997),
and discussed also by Treguier et al. (1997). It
is clear that the statistical output of the recent
generation of eddy ‘resolving’ models will be of
tremendeous importance to further test and refine
the parameterization of eddy effects in coarse-grid
climate models.
2.2.4.3 Thermohaline circulation: Response to
perturbations in deep water formation
There are two principal issues related to the
response of large-scale ocean flows to high-latitude
buoyancy forcing that have to be taken into
account in all model studies of thermohaline circulation. One is the time scale for a near-dynamic
equilibration of ocean currents, as discussed in
Section 2.2.3. The effective adjustment within
O(10–20 years) of major aspects of the general circulation in ocean basins (an exception to this time
scale is the ACC (Antarctic Circumpolar Current)
transport, which appears to take several more
decades to equilibrate; see Chapter 4.6 for a discussion of ACC dynamics), justifies the use of
high-resolution models with decadal-scale integration spans. An obvious appeal of relatively short
model integrations is that the deep water mass
properties remain close to climatology while the
three-dimensional flow field has undergone
dynamical adjustment to a balanced state and
developed a considerable fraction of the observed
energetic scales, facilitating their use in investigations of dynamical mechanisms that would not be
possible using observations alone. Hence beyond
the ‘classical’ application of eddy-resolving models,
i.e. investigations into the role of mesoscale flow
features as discussed in the previous section,
these models have increasingly been utilized for
a quantification of large-scale aspects of the
thermohaline circulation such as the global, threedimensional pathways of deep water and its
conversion to surface water (e.g. Semtner and
Chervin, 1992; Döös and Coward, 1997; Bleck,
1998), or to obtain insight into the role of conspicuous phenomena such as exchanges with marginal
seas and between deep ocean basins divided by
ridges and gaps.
The second, more problematic issue is that the
simulation of important aspects of the ‘mean’ circulation critically depends on the representation of
small-scale ocean processes that are sub-grid-scale
even for ‘eddy-resolving’ models. Accordingly,
there is a strong sensitivity of model solutions to a
host of model factors, including numerical formulation, resolution, and parameterization of subgrid-scale physics, whose relative effects have only
partially been unravelled to date. Model sensitivity
studies over the last decade have focused mainly
on the circulation of the North Atlantic. Since
the subarctic Atlantic is the region where a substantial fraction of the world ocean’s deep water
is formed, the ability to reproduce the essential
features of that water mass conversion and the
associated thermohaline overturning obviously
represents a fundamental aspect of a model’s performance. Systematic model studies have especially
been conducted in the framework of the CME, i.e.
using a domain bounded at 15°S and 65°N, with
boundary conditions aiming to mimic the water
mass transformations outside the domain (Bryan
and Holland, 1989).
A particular aspect of this model configuration
was that the complex processes involved in the
water mass transformations in the Nordic Seas,
and the exchange of water across the sill of the
Denmark Strait, were not part of the simulation.
(The Iceland–Scotland Ridge was part of the CME
domain; effects of the Faeroe Bank Channel
throughflow were investigated in Redler and
Böning, 1997.) Instead, the effect of the Denmark
Strait outflow on the hydrographic properties
south of the sill was included by a damping of T
and S towards prescribed data, allowing an assessment of its role on the large-scale circulation. A
suite of case studies using both medium-resolution
(1°) and eddy-permitting CME model versions
revealed the local density profile in the outflow
regime to be a key factor for both the maximum
strength and the vertical structure of the meridional overturning circulation in the North Atlantic
(Döscher et al., 1994; Böning and Bryan, 1996). In
the example shown in Figs 2.2.10a and b, a specification of conditions that aim at capturing the
conditions of the narrow core of Denmark Strait
Overflow Water (DSOW), i.e. with a minimum
temperature of about 1°C, leads to an enhanced
transport and downward shift in the southward
flowing (North Atlantic Deep Water, NADW)
branch compared with a model case using conditions as given by Levitus (1982) (i.e. with a
minimum temperature of about 3°C. (Note,
2.2 Modelling of Thermohaline and Wind-Driven Circulation
73
Böning and Semtner
lends support to the use of non-uniform mixing
coefficients as proposed by Visbeck et al. (1997),
and discussed also by Treguier et al. (1997). It
is clear that the statistical output of the recent
generation of eddy ‘resolving’ models will be of
tremendeous importance to further test and refine
the parameterization of eddy effects in coarse-grid
climate models.
2.2.4.3 Thermohaline circulation: Response to
perturbations in deep water formation
There are two principal issues related to the
response of large-scale ocean flows to high-latitude
buoyancy forcing that have to be taken into
account in all model studies of thermohaline circulation. One is the time scale for a near-dynamic
equilibration of ocean currents, as discussed in
Section 2.2.3. The effective adjustment within
O(10–20 years) of major aspects of the general circulation in ocean basins (an exception to this time
scale is the ACC (Antarctic Circumpolar Current)
transport, which appears to take several more
decades to equilibrate; see Chapter 4.6 for a discussion of ACC dynamics), justifies the use of
high-resolution models with decadal-scale integration spans. An obvious appeal of relatively short
model integrations is that the deep water mass
properties remain close to climatology while the
three-dimensional flow field has undergone
dynamical adjustment to a balanced state and
developed a considerable fraction of the observed
energetic scales, facilitating their use in investigations of dynamical mechanisms that would not be
possible using observations alone. Hence beyond
the ‘classical’ application of eddy-resolving models,
i.e. investigations into the role of mesoscale flow
features as discussed in the previous section,
these models have increasingly been utilized for
a quantification of large-scale aspects of the
thermohaline circulation such as the global, threedimensional pathways of deep water and its
conversion to surface water (e.g. Semtner and
Chervin, 1992; Döös and Coward, 1997; Bleck,
1998), or to obtain insight into the role of conspicuous phenomena such as exchanges with marginal
seas and between deep ocean basins divided by
ridges and gaps.
The second, more problematic issue is that the
simulation of important aspects of the ‘mean’ circulation critically depends on the representation of
small-scale ocean processes that are sub-grid-scale
even for ‘eddy-resolving’ models. Accordingly,
there is a strong sensitivity of model solutions to a
host of model factors, including numerical formulation, resolution, and parameterization of subgrid-scale physics, whose relative effects have only
partially been unravelled to date. Model sensitivity
studies over the last decade have focused mainly
on the circulation of the North Atlantic. Since
the subarctic Atlantic is the region where a substantial fraction of the world ocean’s deep water
is formed, the ability to reproduce the essential
features of that water mass conversion and the
associated thermohaline overturning obviously
represents a fundamental aspect of a model’s performance. Systematic model studies have especially
been conducted in the framework of the CME, i.e.
using a domain bounded at 15°S and 65°N, with
boundary conditions aiming to mimic the water
mass transformations outside the domain (Bryan
and Holland, 1989).
A particular aspect of this model configuration
was that the complex processes involved in the
water mass transformations in the Nordic Seas,
and the exchange of water across the sill of the
Denmark Strait, were not part of the simulation.
(The Iceland–Scotland Ridge was part of the CME
domain; effects of the Faeroe Bank Channel
throughflow were investigated in Redler and
Böning, 1997.) Instead, the effect of the Denmark
Strait outflow on the hydrographic properties
south of the sill was included by a damping of T
and S towards prescribed data, allowing an assessment of its role on the large-scale circulation. A
suite of case studies using both medium-resolution
(1°) and eddy-permitting CME model versions
revealed the local density profile in the outflow
regime to be a key factor for both the maximum
strength and the vertical structure of the meridional overturning circulation in the North Atlantic
(Döscher et al., 1994; Böning and Bryan, 1996). In
the example shown in Figs 2.2.10a and b, a specification of conditions that aim at capturing the
conditions of the narrow core of Denmark Strait
Overflow Water (DSOW), i.e. with a minimum
temperature of about 1°C, leads to an enhanced
transport and downward shift in the southward
flowing (North Atlantic Deep Water, NADW)
branch compared with a model case using conditions as given by Levitus (1982) (i.e. with a
minimum temperature of about 3°C. (Note,
2.2 Modelling of Thermohaline and Wind-Driven Circulation
73
Böning and Semtner