specified or diffusive meridional transports (e.g.
Rahmstorf and Willebrand, 1995; Wang et al.,
1999) to more complex energy and moisture balance models, sometimes including simple parameterizations of clouds and land surface processes
(e.g. Fanning and Weaver, 1996). Sometimes such
models are coupled to simplified (e.g. zonally averaged) ocean models rather than full ocean GCMs
(Schmittner and Stocker, 1999; Petoukhov et al.,
2000). These simplified model configurations have
the advantage of allowing long integrations, or
multiple sensitivity studies, to be run efficiently in
order to understand particular processes at work
in the climate system. However, the simplifications
inherent in such models mean that they must be
used with caution for quantitative simulation and
prediction. See Section 2.3.6.2 for an example of a
problem where uncertainty over the magnitudes of
a number of competing processes produces significant uncertainty in modelled climate response,
even among complex GCMs. Also see Stocker and
Schmittner (1997) and Rahmstorf and Ganopolski
(1999) for examples of how the ability to perform
many runs of the simplified models allows them to
be used to explore the consequences of such uncertainties in a way that would be impractical with
the GCMs themselves. In this sense the simplified
climate models are complementary to GCMs.
2.3.2.2 Land surface
The land surface plays an important role in the
global budgets of heat and water. Elements of the
system that are usually represented in some way in
coupled models include soil, permafrost, vegetation, snow cover and land ice (ice sheets and glaciers), and model climate simulations are sensitive to
details of these parameterizations (Crossley et al.,
2000). Some features, for example ice sheet extents
and vegetation types, are typically fixed, but models
are being developed in which the vegetation
responds to climate variations, allowing for additional feedbacks that may be important on decadal
time scales (Wang and Eltahir, 2000; Cox et al.,
2000). The role of land surface elements in the
fresh water budget is discussed in Section 2.3.2.6.
2.3.2.3 Ocean
A variety of ocean models have been used in coupled modelling. The z-coordinate models remain
the most popular type, including the Bryan–Cox/
MOM (Manabe et al., 1991), LSG (Voss et al.,
1998), HOPE (Wolff et al., 1997) and OPA
(Guilyardi and Madec, 1997) models. However,
the quasi-isopycnic OPYC model has also been
used (Roeckner et al., 1996) (see the Appendix for
definitions of the acronyms for institutions and
models used in this chapter). The computational
cost of long runs restricts the resolution used, so
that most coupled models have horizontal ocean
resolution in the range 1–4°. The finest ocean
resolution used to date in a global, coupled model
is slightly less than 1° (Washington et al., 2000),
although some models use enhanced resolution in
the tropics to resolve the equatorial waveguide
(Roeckner et al., 1996; Guilyardi and Madec,
1997; Barthelet et al., 1998).
Typical models have 20–30 levels or layers in
the vertical. In some cases an attempt is made to
resolve the surface boundary layer and to parameterize near-surface turbulent mixing processes
using the K-profile method (Large et al., 1994;
Gent et al., 1998), turbulent kinetic energy
schemes (Blanke and Delecluse, 1993; Barthelet
et al., 1998) and hybrid schemes (Johns et al.,
1997b). In other cases a bulk mixed-layer model
(Roeckner et al., 1996), or no specific near-surface
mixing parameterization at all (Manabe et al.,
1991), is used. The latter may be justified as much
of the thermohaline structure of the thermocline
and deep ocean is determined by deep winter
convection; however, summer Sea Surface Temperatures (SSTs) are sensitive to shallow wind- and
shear-induced mixing.
Diapycnal mixing is known to be of fundamental importance in the meridional overturning
circulation. In climate models, it is usually parameterized through vertical or diapycnal diffusion,
but the strength and structure of the overturning
are known to be highly sensitive to the value
chosen for the diffusivity (Bryan, 1987). In practice, most coupled models impose a fixed profile of
vertical diffusivity below the surface mixed layer,
with deep ocean values close to the ‘canonical’
value of 1.010
94 m
2 s
91 (e.g. Johns et al., 1997b).
In the various coupled models that incorporate the
OPA ocean model (Guilyardi and Madec, 1997;
Barthelet et al., 1998), diffusivities are calculated
using a turbulent kinetic energy closure scheme,
leading to considerable horizontal as well as vertical variations of diffusivity (Blanke and Delecluse,
1993); however, the impact and importance of the
large horizontal variations of mixing that have
2.3 Coupled Ocean–Atmosphere Models
81
Wood and Bryan
Rahmstorf and Willebrand, 1995; Wang et al.,
1999) to more complex energy and moisture balance models, sometimes including simple parameterizations of clouds and land surface processes
(e.g. Fanning and Weaver, 1996). Sometimes such
models are coupled to simplified (e.g. zonally averaged) ocean models rather than full ocean GCMs
(Schmittner and Stocker, 1999; Petoukhov et al.,
2000). These simplified model configurations have
the advantage of allowing long integrations, or
multiple sensitivity studies, to be run efficiently in
order to understand particular processes at work
in the climate system. However, the simplifications
inherent in such models mean that they must be
used with caution for quantitative simulation and
prediction. See Section 2.3.6.2 for an example of a
problem where uncertainty over the magnitudes of
a number of competing processes produces significant uncertainty in modelled climate response,
even among complex GCMs. Also see Stocker and
Schmittner (1997) and Rahmstorf and Ganopolski
(1999) for examples of how the ability to perform
many runs of the simplified models allows them to
be used to explore the consequences of such uncertainties in a way that would be impractical with
the GCMs themselves. In this sense the simplified
climate models are complementary to GCMs.
2.3.2.2 Land surface
The land surface plays an important role in the
global budgets of heat and water. Elements of the
system that are usually represented in some way in
coupled models include soil, permafrost, vegetation, snow cover and land ice (ice sheets and glaciers), and model climate simulations are sensitive to
details of these parameterizations (Crossley et al.,
2000). Some features, for example ice sheet extents
and vegetation types, are typically fixed, but models
are being developed in which the vegetation
responds to climate variations, allowing for additional feedbacks that may be important on decadal
time scales (Wang and Eltahir, 2000; Cox et al.,
2000). The role of land surface elements in the
fresh water budget is discussed in Section 2.3.2.6.
2.3.2.3 Ocean
A variety of ocean models have been used in coupled modelling. The z-coordinate models remain
the most popular type, including the Bryan–Cox/
MOM (Manabe et al., 1991), LSG (Voss et al.,
1998), HOPE (Wolff et al., 1997) and OPA
(Guilyardi and Madec, 1997) models. However,
the quasi-isopycnic OPYC model has also been
used (Roeckner et al., 1996) (see the Appendix for
definitions of the acronyms for institutions and
models used in this chapter). The computational
cost of long runs restricts the resolution used, so
that most coupled models have horizontal ocean
resolution in the range 1–4°. The finest ocean
resolution used to date in a global, coupled model
is slightly less than 1° (Washington et al., 2000),
although some models use enhanced resolution in
the tropics to resolve the equatorial waveguide
(Roeckner et al., 1996; Guilyardi and Madec,
1997; Barthelet et al., 1998).
Typical models have 20–30 levels or layers in
the vertical. In some cases an attempt is made to
resolve the surface boundary layer and to parameterize near-surface turbulent mixing processes
using the K-profile method (Large et al., 1994;
Gent et al., 1998), turbulent kinetic energy
schemes (Blanke and Delecluse, 1993; Barthelet
et al., 1998) and hybrid schemes (Johns et al.,
1997b). In other cases a bulk mixed-layer model
(Roeckner et al., 1996), or no specific near-surface
mixing parameterization at all (Manabe et al.,
1991), is used. The latter may be justified as much
of the thermohaline structure of the thermocline
and deep ocean is determined by deep winter
convection; however, summer Sea Surface Temperatures (SSTs) are sensitive to shallow wind- and
shear-induced mixing.
Diapycnal mixing is known to be of fundamental importance in the meridional overturning
circulation. In climate models, it is usually parameterized through vertical or diapycnal diffusion,
but the strength and structure of the overturning
are known to be highly sensitive to the value
chosen for the diffusivity (Bryan, 1987). In practice, most coupled models impose a fixed profile of
vertical diffusivity below the surface mixed layer,
with deep ocean values close to the ‘canonical’
value of 1.010
94 m
2 s
91 (e.g. Johns et al., 1997b).
In the various coupled models that incorporate the
OPA ocean model (Guilyardi and Madec, 1997;
Barthelet et al., 1998), diffusivities are calculated
using a turbulent kinetic energy closure scheme,
leading to considerable horizontal as well as vertical variations of diffusivity (Blanke and Delecluse,
1993); however, the impact and importance of the
large horizontal variations of mixing that have
2.3 Coupled Ocean–Atmosphere Models
81
Wood and Bryan