between the equilibrium climates with normal
and above-normal atmospheric CO 2 concentrations. Incorporation of ocean circulation dynamics
in model simulations of anthropogenic increase of
atmospheric trace gases was first accomplished in
the 1980s, initially by studies of the climate system
response to a step function increase in idealized
(Bryan et al., 1982; Bryan and Spelman, 1985) and
more realistic domains (Schlesinger et al., 1985),
followed by first investigations of gradual increase
scenarios (Washington and Meehl, 1989).
While coarse-resolution models of the global
ocean circulation became a standard for coupled,
climate system studies, their typical application
in oceanic research was not that of realistic simulation, but directed primarily at investigations of
principal mechanisms. For example, a host of
global model studies helped to elucidate how wind
and buoyancy fluxes at the sea surface conspire to
maintain the ocean’s water-mass properties and
tracer distributions, and how these distributions
depend on model assumptions and parameters
(Bryan and Lewis, 1979; Cox, 1989; Toggweiler
et al., 1989; England and Hirst, 1997).
Following the intensive field programmes in
the 1970s (POLYGON, MODE, POLYMODE),
much of the interest in ocean model development
and application shifted towards the dynamics of
mesoscale current features and variability that
became recognized as the prime signal in direct current measurements and satellite observations. The
subsequent development of eddy-resolving models
of ocean circulation, pioneered by Holland and Lin
(1975), Semtner and Mintz (1977), and Robinson
et al. (1977), progressed effectively separate from
the ocean climate models. The heavy demands of
computing power necessary to represent mesoscale
variability in large-scale models often required
compromises in potentially important physics, e.g.
by adopting simplified (adiabatic, quasigeostrophic)
dynamical equations, and by limiting the computational domain to idealized basins.
Some significant strides in ocean model development in the second half of the 1980s can be identified that helped to blur the distinction between
the two, hitherto non-interactive lines of ocean
modelling. A pioneering step towards explicit representation of mesoscale processes in a wind- and
buoyancy-driven, geophysically sized ocean basin
was taken by Cox (1985): a grid size of 1/3° in latitudinal and 2/5° in longitudinal direction permitted
a resolution of the first mode Rossby radius in the
subtropics, and thus inclusion of parts of the eddy
spectrum. Shortly thereafter, the investigation of
the role of mesoscale processes in the thermohaline
circulation of realistic ocean basins, under realistic
atmospheric forcing, was taken up by a collection
of ambitious efforts. The model domains considered in these first planetary-sized, ‘eddy-permitting’
calculations (i.e. with grid sizes enabling representation of the upper range of mesoscale eddies)
included the Southern Ocean (FRAM Group,
1991), the North and Equatorial Atlantic (Bryan
and Holland, 1989), and the World Ocean without
the Arctic (Semtner and Chervin, 1988, 1992).
All of the computations induced a considerable
number of studies, often in collaboration with
observational oceanographers, probing the models’ validity across a host of phenomena. The
model configurations gradually evolved in subsequent years, building on the experience from the
initial experiments and capitalizing on advance
in computing capabilities, numerical methods,
and forcing data sets. Especially for the Atlantic
configuration, a large number of experiments
performed by several groups as a collective ‘Community Modelling Effort’ (CME) added to an
exploration of the influence of forcing functions,
resolution and sub-grid-scale parameterization (for
an overview, see Böning and Bryan, 1996). Developments in global models by three groups included
substantial reductions in horizontal grid sizes:
to a (latitudinal) average of 1/4° (Stammer et al.,
1996), 1/6° (Maltrud et al., 1998), and 1/4° and
1/8° (Webb et al., 1997; Saunders et al., 1999).
An important contribution towards the model
refinements, eventually realized during the last
several years, was the formulation of numerical
model concepts and algorithms that provided
alternatives to the early paradigm of Bryan (1969).
The model formulation based on coordinate surfaces coinciding with geopotential levels proved to
be a convenient and versatile concept for a wide
range of applications, but had to deal with some
inherent difficulties associated with a staircase representation of topography and spurious diapycnic
mixing. New model developments had in particular focused on a different representation of the vertical coordinate and, associated with that, bottom
topography. Isopycnic coordinates (Bleck et al.,
1992; Oberhuber, 1993) appear optimally suited
for a representation of advective transports in the
2.2 Modelling of Thermohaline and Wind-Driven Circulation
61
Böning and Semtner
and above-normal atmospheric CO 2 concentrations. Incorporation of ocean circulation dynamics
in model simulations of anthropogenic increase of
atmospheric trace gases was first accomplished in
the 1980s, initially by studies of the climate system
response to a step function increase in idealized
(Bryan et al., 1982; Bryan and Spelman, 1985) and
more realistic domains (Schlesinger et al., 1985),
followed by first investigations of gradual increase
scenarios (Washington and Meehl, 1989).
While coarse-resolution models of the global
ocean circulation became a standard for coupled,
climate system studies, their typical application
in oceanic research was not that of realistic simulation, but directed primarily at investigations of
principal mechanisms. For example, a host of
global model studies helped to elucidate how wind
and buoyancy fluxes at the sea surface conspire to
maintain the ocean’s water-mass properties and
tracer distributions, and how these distributions
depend on model assumptions and parameters
(Bryan and Lewis, 1979; Cox, 1989; Toggweiler
et al., 1989; England and Hirst, 1997).
Following the intensive field programmes in
the 1970s (POLYGON, MODE, POLYMODE),
much of the interest in ocean model development
and application shifted towards the dynamics of
mesoscale current features and variability that
became recognized as the prime signal in direct current measurements and satellite observations. The
subsequent development of eddy-resolving models
of ocean circulation, pioneered by Holland and Lin
(1975), Semtner and Mintz (1977), and Robinson
et al. (1977), progressed effectively separate from
the ocean climate models. The heavy demands of
computing power necessary to represent mesoscale
variability in large-scale models often required
compromises in potentially important physics, e.g.
by adopting simplified (adiabatic, quasigeostrophic)
dynamical equations, and by limiting the computational domain to idealized basins.
Some significant strides in ocean model development in the second half of the 1980s can be identified that helped to blur the distinction between
the two, hitherto non-interactive lines of ocean
modelling. A pioneering step towards explicit representation of mesoscale processes in a wind- and
buoyancy-driven, geophysically sized ocean basin
was taken by Cox (1985): a grid size of 1/3° in latitudinal and 2/5° in longitudinal direction permitted
a resolution of the first mode Rossby radius in the
subtropics, and thus inclusion of parts of the eddy
spectrum. Shortly thereafter, the investigation of
the role of mesoscale processes in the thermohaline
circulation of realistic ocean basins, under realistic
atmospheric forcing, was taken up by a collection
of ambitious efforts. The model domains considered in these first planetary-sized, ‘eddy-permitting’
calculations (i.e. with grid sizes enabling representation of the upper range of mesoscale eddies)
included the Southern Ocean (FRAM Group,
1991), the North and Equatorial Atlantic (Bryan
and Holland, 1989), and the World Ocean without
the Arctic (Semtner and Chervin, 1988, 1992).
All of the computations induced a considerable
number of studies, often in collaboration with
observational oceanographers, probing the models’ validity across a host of phenomena. The
model configurations gradually evolved in subsequent years, building on the experience from the
initial experiments and capitalizing on advance
in computing capabilities, numerical methods,
and forcing data sets. Especially for the Atlantic
configuration, a large number of experiments
performed by several groups as a collective ‘Community Modelling Effort’ (CME) added to an
exploration of the influence of forcing functions,
resolution and sub-grid-scale parameterization (for
an overview, see Böning and Bryan, 1996). Developments in global models by three groups included
substantial reductions in horizontal grid sizes:
to a (latitudinal) average of 1/4° (Stammer et al.,
1996), 1/6° (Maltrud et al., 1998), and 1/4° and
1/8° (Webb et al., 1997; Saunders et al., 1999).
An important contribution towards the model
refinements, eventually realized during the last
several years, was the formulation of numerical
model concepts and algorithms that provided
alternatives to the early paradigm of Bryan (1969).
The model formulation based on coordinate surfaces coinciding with geopotential levels proved to
be a convenient and versatile concept for a wide
range of applications, but had to deal with some
inherent difficulties associated with a staircase representation of topography and spurious diapycnic
mixing. New model developments had in particular focused on a different representation of the vertical coordinate and, associated with that, bottom
topography. Isopycnic coordinates (Bleck et al.,
1992; Oberhuber, 1993) appear optimally suited
for a representation of advective transports in the
2.2 Modelling of Thermohaline and Wind-Driven Circulation
61
Böning and Semtner