ocean. Terrain-following (or sigma) coordinates
(Haidvogel et al., 1991; Ezer and Mellor, 1997; de
Miranda et al., 1999) attempt to avoid spurious
effects associated with stepwise representation of
bathymetry. A systematic evaluation of the diverse
models available today, including an assessment of
the robustness of solution features and an identification of model idiosyncracies, has only begun in
recent years. Following a few basin-scale model–
model comparisons in relatively coarse resolution
(Chassignet et al., 1996; Roberts et al., 1996;
Marsh et al., 1996), a systematic investigation of
the performance of higher-resolution models (at an
average grid size of 1/5°) based on the different
vertical coordinate treatments was made in the
DYNAMO project (DYNAMO Group, 1997;
Willebrand et al., 2001).
The bulk of global model simulations and
basin-scale model sensitivity and intercomparison
studies have been performed on horizontal grids
that still fall short in the simulation of important
mesoscale current features at higher latitudes.
Typical deficits in these ‘eddy-permitting’ models
include the intensity and spectral distribution
of eddy variability (Beckmann et al., 1994a,b;
McClean et al., 1997; Maltrud et al., 1998) and
the regional distortions in intense western boundary currents such as in the pathway of the Gulf
Stream near Cape Hatteras (Dengg et al., 1996;
Chao et al., 1996) and the Grand Banks off
Newfoundland (Stammer et al., 1996; Willebrand
et al., 2001). Recently, a few very ambitious North
Atlantic simulations using grid refinements to
1/12° at the equator, based on both isopycnic
(E. Chassignet, personal communication) and
geopotential coordinate models (Smith et al.,
2000), have begun to examine systematically the
impact of resolving the oceanic flow spectrum up
to the first mode deformation radius over the
whole North Atlantic; an illustration of the richness in flow structures emerging here is given in
Fig. 2.2.1 (see Plate 2.2.1, p. 76).
By now, the reader may be concerned about the
diversity of methods used to characterize model
grid size. Various grid metrics have included the
approximate square grid size in spherical models
at middle or high latitudes, the latitudinal average
grid size in square-grid Mercator models, and the
equatorial grid size in square-grid Mercator models. It helps to invoke a specific metric, namely the
average over valid ocean gridpoints of the effective
grid size in km, expressed in degrees using a conversion factor of 110 km per °. Under this method,
which was proposed by R. Smith, the various
spherical coordinate models are close to their
original designations, the ‘1/4°’ and ‘1/6°’ global
models in Stammer et al. (1996) and Maltrud et al.
(1998) become about 1/3° and 1/5°, respectively;
and the ‘1/12°’ and ‘1/10°’ North Atlantic models
of Chassignet (personal communication) and
Smith et al. (2000) become approximately 1/17°
and 1/14°, respectively.
Despite the remarkable progress in the computational domains realized in recent years as a result
of the increase in computing capabilities available
to the ocean modelling community, it has to be
emphasized that the gap between those models
developed and used primarily for oceanographic
applications, and those used as components in
climate system models, has not yet been closed.
Common to all high-resolution model studies of
ocean circulation, up to the present, has been a
limitation of integration periods to, typically,
20–30 years (only a few eddy-permitting model
experiments have been run for periods of 50–100
years). The nature of the dynamical balances realized in such model solutions and what this implies
for their interpretation, for example in comparison
with actual oceanic measurements, will be discussed in the next section.
2.2.3 Basic model design considerations:
equilibrium versus non-equilibrium
solutions
Since it is impossible to embrace, in a single model
calculation, the vast range of space and time scales
governing a natural, geophysical fluid system, any
model construction involves a number of critical
choices. The most basic choice in ocean general
circulation modelling concerns integration time,
and had led to a rather sharp distinction of two
model categories: models aiming at solutions in
thermodynamic equilibrium that require a compromise in spatial resolution; and models aiming
at an accurate depiction of the actually observed,
energetic flow scales that require a compromise in
the adjustment to a full equilibrium state.
Despite significant strides in computer capabilities, the distinction between the two main lines
characterizing ocean model development over the
last couple of decades has not yet faded, nor is
SECTION 2 OBSERVATIONS AND MODELS
62
(Haidvogel et al., 1991; Ezer and Mellor, 1997; de
Miranda et al., 1999) attempt to avoid spurious
effects associated with stepwise representation of
bathymetry. A systematic evaluation of the diverse
models available today, including an assessment of
the robustness of solution features and an identification of model idiosyncracies, has only begun in
recent years. Following a few basin-scale model–
model comparisons in relatively coarse resolution
(Chassignet et al., 1996; Roberts et al., 1996;
Marsh et al., 1996), a systematic investigation of
the performance of higher-resolution models (at an
average grid size of 1/5°) based on the different
vertical coordinate treatments was made in the
DYNAMO project (DYNAMO Group, 1997;
Willebrand et al., 2001).
The bulk of global model simulations and
basin-scale model sensitivity and intercomparison
studies have been performed on horizontal grids
that still fall short in the simulation of important
mesoscale current features at higher latitudes.
Typical deficits in these ‘eddy-permitting’ models
include the intensity and spectral distribution
of eddy variability (Beckmann et al., 1994a,b;
McClean et al., 1997; Maltrud et al., 1998) and
the regional distortions in intense western boundary currents such as in the pathway of the Gulf
Stream near Cape Hatteras (Dengg et al., 1996;
Chao et al., 1996) and the Grand Banks off
Newfoundland (Stammer et al., 1996; Willebrand
et al., 2001). Recently, a few very ambitious North
Atlantic simulations using grid refinements to
1/12° at the equator, based on both isopycnic
(E. Chassignet, personal communication) and
geopotential coordinate models (Smith et al.,
2000), have begun to examine systematically the
impact of resolving the oceanic flow spectrum up
to the first mode deformation radius over the
whole North Atlantic; an illustration of the richness in flow structures emerging here is given in
Fig. 2.2.1 (see Plate 2.2.1, p. 76).
By now, the reader may be concerned about the
diversity of methods used to characterize model
grid size. Various grid metrics have included the
approximate square grid size in spherical models
at middle or high latitudes, the latitudinal average
grid size in square-grid Mercator models, and the
equatorial grid size in square-grid Mercator models. It helps to invoke a specific metric, namely the
average over valid ocean gridpoints of the effective
grid size in km, expressed in degrees using a conversion factor of 110 km per °. Under this method,
which was proposed by R. Smith, the various
spherical coordinate models are close to their
original designations, the ‘1/4°’ and ‘1/6°’ global
models in Stammer et al. (1996) and Maltrud et al.
(1998) become about 1/3° and 1/5°, respectively;
and the ‘1/12°’ and ‘1/10°’ North Atlantic models
of Chassignet (personal communication) and
Smith et al. (2000) become approximately 1/17°
and 1/14°, respectively.
Despite the remarkable progress in the computational domains realized in recent years as a result
of the increase in computing capabilities available
to the ocean modelling community, it has to be
emphasized that the gap between those models
developed and used primarily for oceanographic
applications, and those used as components in
climate system models, has not yet been closed.
Common to all high-resolution model studies of
ocean circulation, up to the present, has been a
limitation of integration periods to, typically,
20–30 years (only a few eddy-permitting model
experiments have been run for periods of 50–100
years). The nature of the dynamical balances realized in such model solutions and what this implies
for their interpretation, for example in comparison
with actual oceanic measurements, will be discussed in the next section.
2.2.3 Basic model design considerations:
equilibrium versus non-equilibrium
solutions
Since it is impossible to embrace, in a single model
calculation, the vast range of space and time scales
governing a natural, geophysical fluid system, any
model construction involves a number of critical
choices. The most basic choice in ocean general
circulation modelling concerns integration time,
and had led to a rather sharp distinction of two
model categories: models aiming at solutions in
thermodynamic equilibrium that require a compromise in spatial resolution; and models aiming
at an accurate depiction of the actually observed,
energetic flow scales that require a compromise in
the adjustment to a full equilibrium state.
Despite significant strides in computer capabilities, the distinction between the two main lines
characterizing ocean model development over the
last couple of decades has not yet faded, nor is
SECTION 2 OBSERVATIONS AND MODELS
62