2.2.5 Concluding remarks
The primary aim of the WOCE modelling effort is
the development of ocean component models for
coupled ocean–atmosphere models that can be
used to predict accurately climate change. The
focus is hence on time scales from a few years to
hundreds of years, covering both the decadal–
interdecadal natural variability in climate and the
long-term anthropogenic changes. Obviously, the
testing of models suitable for predicting future
evolution cannot be based on evaluations of equilibrium behaviours alone, but must extend also to
the system’s response to variations or perturbations in the external forcing, that is, it must aim at
the physics of processes governing ocean dynamics
at the time scales of interest.
The results of ocean modelling efforts of the past
decade show that in assessing the performance of
large-scale circulation models one needs to distinguish between different classes of phenomena:
1 Models are increasingly successful in reproducing observed aspects of large-scale circulation,
including, in particular, the major characteristics of near-surface currents and mesoscale
variability, and the atmospherically induced
(especially wind-driven) variability on intraseasonal to interannual time scales.
2 In contrast, modelling of longer-term (decadal
to secular) aspects, which include the dynamics
of the thermohaline circulation, has been found
to depend very critically on a number of local
processes that are either not sufficiently well
understood to allow suitable parameterization
(as in the case of passage throughflows) or are
(e.g. in the case of sill overflows and associated
entrainment processes) dependent on a host of
model factors that present modelling efforts
have just begun to unravel.
In retrospect, one may note that the basic
deficits of general ocean circulation models with
respect to long-term behaviours were obvious in
the mid-1980s even on the basis of the crude
climatogical data sets available, and the principal
requirements for model improvement were quite
clear even without the need to collect more
(WOCE) data; however, the goal of WOCE ‘to
collect the data to test’ models, while then premature, has led to acquisition of data sets that now
appear indispensable for the refinement of the present generation of models. Ocean model development for climate studies has to account for the fact
that the anticipated development of computational
resources will not allow resolutions finer than
about 1° in the immediate future. However, the
recent success in parameterizing the effects of
mesoscale eddies allows smaller-scale mixing
parameterizations to be tested in long simulations
without the need for eddy resolution.
The characteristics of model behaviour in the
different dynamical regimes have important implications for the utilization of ocean circulation
models for interpretation of oceanographic measurements or observation systems design. Since the
analysis of ‘mean’ or low-frequency current patterns and fluxes of mass, heat and other properties
in any single model realization has to account for
possible strong model sensitivities, the prime goal
of modelling in this regime has continued to be on
a principal understanding of dynamical mechanisms, e.g. of the role of individual phenomena in
the global circulation. In contrast, the robustness
of solutions with respect to the dynamical
response to wind-driven changes, and the increasing number of examples where model predictions
of phenomena have eventually been confirmed in
dedicated measurement programmes, strongly suggest that the practice of ocean modelling is rapidly
acquiring a more prominent role in oceanographic
studies of phenomena on intraseasonal, seasonal
and, to some extent, interannual time scales.
2.2 Modelling of Thermohaline and Wind-Driven Circulation
77
Böning and Semtner
The primary aim of the WOCE modelling effort is
the development of ocean component models for
coupled ocean–atmosphere models that can be
used to predict accurately climate change. The
focus is hence on time scales from a few years to
hundreds of years, covering both the decadal–
interdecadal natural variability in climate and the
long-term anthropogenic changes. Obviously, the
testing of models suitable for predicting future
evolution cannot be based on evaluations of equilibrium behaviours alone, but must extend also to
the system’s response to variations or perturbations in the external forcing, that is, it must aim at
the physics of processes governing ocean dynamics
at the time scales of interest.
The results of ocean modelling efforts of the past
decade show that in assessing the performance of
large-scale circulation models one needs to distinguish between different classes of phenomena:
1 Models are increasingly successful in reproducing observed aspects of large-scale circulation,
including, in particular, the major characteristics of near-surface currents and mesoscale
variability, and the atmospherically induced
(especially wind-driven) variability on intraseasonal to interannual time scales.
2 In contrast, modelling of longer-term (decadal
to secular) aspects, which include the dynamics
of the thermohaline circulation, has been found
to depend very critically on a number of local
processes that are either not sufficiently well
understood to allow suitable parameterization
(as in the case of passage throughflows) or are
(e.g. in the case of sill overflows and associated
entrainment processes) dependent on a host of
model factors that present modelling efforts
have just begun to unravel.
In retrospect, one may note that the basic
deficits of general ocean circulation models with
respect to long-term behaviours were obvious in
the mid-1980s even on the basis of the crude
climatogical data sets available, and the principal
requirements for model improvement were quite
clear even without the need to collect more
(WOCE) data; however, the goal of WOCE ‘to
collect the data to test’ models, while then premature, has led to acquisition of data sets that now
appear indispensable for the refinement of the present generation of models. Ocean model development for climate studies has to account for the fact
that the anticipated development of computational
resources will not allow resolutions finer than
about 1° in the immediate future. However, the
recent success in parameterizing the effects of
mesoscale eddies allows smaller-scale mixing
parameterizations to be tested in long simulations
without the need for eddy resolution.
The characteristics of model behaviour in the
different dynamical regimes have important implications for the utilization of ocean circulation
models for interpretation of oceanographic measurements or observation systems design. Since the
analysis of ‘mean’ or low-frequency current patterns and fluxes of mass, heat and other properties
in any single model realization has to account for
possible strong model sensitivities, the prime goal
of modelling in this regime has continued to be on
a principal understanding of dynamical mechanisms, e.g. of the role of individual phenomena in
the global circulation. In contrast, the robustness
of solutions with respect to the dynamical
response to wind-driven changes, and the increasing number of examples where model predictions
of phenomena have eventually been confirmed in
dedicated measurement programmes, strongly suggest that the practice of ocean modelling is rapidly
acquiring a more prominent role in oceanographic
studies of phenomena on intraseasonal, seasonal
and, to some extent, interannual time scales.
2.2 Modelling of Thermohaline and Wind-Driven Circulation
77
Böning and Semtner