been observed (see Toole and McDougall, Chapter 5.2) for the large-scale circulation has only
just begun to be explored (e.g. Webb and
Suginohara, Chapter 4.2; Marotzke, 1997; Hasumi
and Suginohara, 1999a).
Since mesoscale eddies cannot at present be
resolved in the ocean components of coupled models, their effects on the large-scale circulation must
be parameterized. Considerable advances have
been made in this area during the WOCE period
with the development of a parameterization (Gent
and McWilliams, 1990) that has two desirable
properties: first, it mimics the release of available
potential energy by baroclinic instability, and second, it removes grid-scale enstrophy in a way that
does not introduce spurious diapycnal mixing. The
latter property has been shown to be particularly
important in allowing coarse-resolution models to
maintain a realistically sharp thermocline and in
improving meridional heat transports in both
hemispheres (Danabasoglu et al., 1994; Böning
et al., 1995). For model resolution of order 1°, a
further refinement has been found useful. In order
to simulate eddy transports across the Antarctic
Circumpolar Current, eddy thickness diffusivities
of order 2000 m
2 s
91 are appropriate (Treguier
et al., 1997; Visbeck et al., 1997). However, the
use of such a large value is less appropriate in
other parts of the ocean, and leads to weakening
of narrow current systems such as the North
Atlantic Current and Kuroshio. In the HadCM3
model, the parameterization of Visbeck et al.
(1997) was therefore used, in which the thickness
diffusivity is determined as a function of the local
baroclinic structure (Gordon et al., 2000). This
issue appears less important in coarse-resolution
models, where the structure of the narrow currents
is presumably determined more by the larger eddy
viscosity required.
2.3.2.4 Sea ice
A variety of sea ice models has been used in coupled models. Models range from simply assuming
ice cover of fixed properties at ocean gridpoints
where the SST falls below freezing point (Guilyardi
and Madec, 1997), through single or multilayer thermodynamic-only models (Murphy and
Mitchell, 1995; Barthelet et al., 1998; Voss et al.,
1998), to models with simple ice dynamics (e.g.
Manabe et al., 1991; Johns et al., 1997b; Gordon
et al., 2000), and more sophisticated dynamics with
non-trivial ice rheologies (e.g. Roeckner et al.,
1996; Gordon and O’Farrell, 1997; Weatherly
et al., 1998; Washington et al., 2000).
2.3.2.5 Coupling of component models
The atmosphere, ocean and sea ice are coupled
through exchanges of heat, water, momentum and
turbulent kinetic energy at their interfaces. Coupled models are typically built from component
submodels that may have different resolution,
time-steps, etc. The models are run quasi-independently, but exchange information at regular intervals (e.g. once per day). Care is needed to ensure
that conservation properties are maintained, particularly in models that are used to study climate
on decadal or longer time scales.
Surface exchanges must also reflect the different
boundary layer processes going on over land, sea
and sea ice. This is a particular issue when the
atmosphere and ocean model horizontal grids and
coastlines do not match. In this case it is possible,
for example, that part of an atmosphere grid box
can cover part open ocean, part sea ice and part
land. Separate surface exchanges must be calculated for each component. If one component
model has much finer resolution than another (e.g.
the ocean is run at finer resolution than the atmosphere), the surface exchanges must be calculated at
the finer resolution, since the non-linearity of the
bulk formulae that are used to calculate surface
fluxes means that averaging the ocean surface
properties onto the atmospheric grid would not
provide the correct mean surface flux. Examples
of software solutions to these problems are the
NCAR flux coupler (Boville and Gent, 1998) and
the OASIS (Ocean Atmosphere Sea Ice Soil) coupler
(Cassou et al., 1998).
2.3.2.6 Closing the freshwater budget
Water vapour is lost from the ocean to the atmosphere by evaporation and sublimation from water
and sea ice/snow surfaces. Water is resupplied to
the ocean through direct precipitation onto the
ocean and sea ice surfaces, through river runoff
and through iceberg calving from the Greenland
and Antarctic ice sheets. The runoff and icebergs
are a residual of the many processes contributing
to the land surface water budget, including accumulation and evaporation of soil moisture, evapotranspiration by vegetation, and accumulation of
snow on forest canopies and ice sheets. All coupled
SECTION 2 OBSERVATIONS AND MODELS
82
just begun to be explored (e.g. Webb and
Suginohara, Chapter 4.2; Marotzke, 1997; Hasumi
and Suginohara, 1999a).
Since mesoscale eddies cannot at present be
resolved in the ocean components of coupled models, their effects on the large-scale circulation must
be parameterized. Considerable advances have
been made in this area during the WOCE period
with the development of a parameterization (Gent
and McWilliams, 1990) that has two desirable
properties: first, it mimics the release of available
potential energy by baroclinic instability, and second, it removes grid-scale enstrophy in a way that
does not introduce spurious diapycnal mixing. The
latter property has been shown to be particularly
important in allowing coarse-resolution models to
maintain a realistically sharp thermocline and in
improving meridional heat transports in both
hemispheres (Danabasoglu et al., 1994; Böning
et al., 1995). For model resolution of order 1°, a
further refinement has been found useful. In order
to simulate eddy transports across the Antarctic
Circumpolar Current, eddy thickness diffusivities
of order 2000 m
2 s
91 are appropriate (Treguier
et al., 1997; Visbeck et al., 1997). However, the
use of such a large value is less appropriate in
other parts of the ocean, and leads to weakening
of narrow current systems such as the North
Atlantic Current and Kuroshio. In the HadCM3
model, the parameterization of Visbeck et al.
(1997) was therefore used, in which the thickness
diffusivity is determined as a function of the local
baroclinic structure (Gordon et al., 2000). This
issue appears less important in coarse-resolution
models, where the structure of the narrow currents
is presumably determined more by the larger eddy
viscosity required.
2.3.2.4 Sea ice
A variety of sea ice models has been used in coupled models. Models range from simply assuming
ice cover of fixed properties at ocean gridpoints
where the SST falls below freezing point (Guilyardi
and Madec, 1997), through single or multilayer thermodynamic-only models (Murphy and
Mitchell, 1995; Barthelet et al., 1998; Voss et al.,
1998), to models with simple ice dynamics (e.g.
Manabe et al., 1991; Johns et al., 1997b; Gordon
et al., 2000), and more sophisticated dynamics with
non-trivial ice rheologies (e.g. Roeckner et al.,
1996; Gordon and O’Farrell, 1997; Weatherly
et al., 1998; Washington et al., 2000).
2.3.2.5 Coupling of component models
The atmosphere, ocean and sea ice are coupled
through exchanges of heat, water, momentum and
turbulent kinetic energy at their interfaces. Coupled models are typically built from component
submodels that may have different resolution,
time-steps, etc. The models are run quasi-independently, but exchange information at regular intervals (e.g. once per day). Care is needed to ensure
that conservation properties are maintained, particularly in models that are used to study climate
on decadal or longer time scales.
Surface exchanges must also reflect the different
boundary layer processes going on over land, sea
and sea ice. This is a particular issue when the
atmosphere and ocean model horizontal grids and
coastlines do not match. In this case it is possible,
for example, that part of an atmosphere grid box
can cover part open ocean, part sea ice and part
land. Separate surface exchanges must be calculated for each component. If one component
model has much finer resolution than another (e.g.
the ocean is run at finer resolution than the atmosphere), the surface exchanges must be calculated at
the finer resolution, since the non-linearity of the
bulk formulae that are used to calculate surface
fluxes means that averaging the ocean surface
properties onto the atmospheric grid would not
provide the correct mean surface flux. Examples
of software solutions to these problems are the
NCAR flux coupler (Boville and Gent, 1998) and
the OASIS (Ocean Atmosphere Sea Ice Soil) coupler
(Cassou et al., 1998).
2.3.2.6 Closing the freshwater budget
Water vapour is lost from the ocean to the atmosphere by evaporation and sublimation from water
and sea ice/snow surfaces. Water is resupplied to
the ocean through direct precipitation onto the
ocean and sea ice surfaces, through river runoff
and through iceberg calving from the Greenland
and Antarctic ice sheets. The runoff and icebergs
are a residual of the many processes contributing
to the land surface water budget, including accumulation and evaporation of soil moisture, evapotranspiration by vegetation, and accumulation of
snow on forest canopies and ice sheets. All coupled
SECTION 2 OBSERVATIONS AND MODELS
82