OCEAN MODELS
103
Meddies and Agulhas eddies). The effects of such eddies are highly nonlocal, and no parameterization has yet been proposed for such processes.
6.
Conclusion
Parameterizations and resolution are the two fundamental characteristics of an ocean model. By choosing them, we actually pick up the
“ocean” we try to model. We have reviewed different parameterizations, based on the example of the forecast models listed in tables 1 and
2. What emerges from this review is a rather unsatisfactory state of
affairs. Some paramerizations are well grounded in physics (like convection) and have been evaluated by comparison with more complete
models (non-hydrostatic in this case). Even then, though, we find that
some features are not completely agreed upon among modellers (like the
Prandtl number) and modifying them has a strong effect on the solution
of low resolution models. But this is the best situation. Generally, the
parameterizations do not have sound physical basis, have not been fully
evaluated against laboratory experiments or more complete models, and
there are strong numerical constraints limiting the choices of modellers.
The boundaries of the ocean, at the bottom and at the surface, are
places where progress needs to be made. Regarding the bottom, the
main problem is the representation of flow-topography interactions in
z-coordinate ocean models. This is an issue of numerics rather than
parameterization, which is discussed in the chapter by Griffies. The
main effect of staircase topography in z-coordinate models, which is
not completely alleviated by using a partial step representation, is the
existence of large and noisy vertical velocities which often contaminate
the upper layers (especially on the continental slopes). This is a big
obstacle to the use of such models for biogeochemistry. Hybrid models
like HYCOM may be better candidates for such applications, although it
is not clear that numerical factors affecting the communication between
the surface z layer and the interior isopcynic layers will not prove an
even bigger obstacle. It is quite surprising that although σ coordinate
models are extensively used for regional and coastal modelling, no larger
scale σ configurations have been built for demonstration purposes, either
for climate prediction or forecasting.
The representation of the surface layers in ocean models is perhaps
the point where progress is the most likely in the coming years. Today,
model solutions in the mixed layer critically depends on the parameterizations. This dependency may decrease as we resolve more physical
processes. It is possible to do so with existing parameterizations simply
by increasing the vertical resolution (to about 1 m) and using higher
103
Meddies and Agulhas eddies). The effects of such eddies are highly nonlocal, and no parameterization has yet been proposed for such processes.
6.
Conclusion
Parameterizations and resolution are the two fundamental characteristics of an ocean model. By choosing them, we actually pick up the
“ocean” we try to model. We have reviewed different parameterizations, based on the example of the forecast models listed in tables 1 and
2. What emerges from this review is a rather unsatisfactory state of
affairs. Some paramerizations are well grounded in physics (like convection) and have been evaluated by comparison with more complete
models (non-hydrostatic in this case). Even then, though, we find that
some features are not completely agreed upon among modellers (like the
Prandtl number) and modifying them has a strong effect on the solution
of low resolution models. But this is the best situation. Generally, the
parameterizations do not have sound physical basis, have not been fully
evaluated against laboratory experiments or more complete models, and
there are strong numerical constraints limiting the choices of modellers.
The boundaries of the ocean, at the bottom and at the surface, are
places where progress needs to be made. Regarding the bottom, the
main problem is the representation of flow-topography interactions in
z-coordinate ocean models. This is an issue of numerics rather than
parameterization, which is discussed in the chapter by Griffies. The
main effect of staircase topography in z-coordinate models, which is
not completely alleviated by using a partial step representation, is the
existence of large and noisy vertical velocities which often contaminate
the upper layers (especially on the continental slopes). This is a big
obstacle to the use of such models for biogeochemistry. Hybrid models
like HYCOM may be better candidates for such applications, although it
is not clear that numerical factors affecting the communication between
the surface z layer and the interior isopcynic layers will not prove an
even bigger obstacle. It is quite surprising that although σ coordinate
models are extensively used for regional and coastal modelling, no larger
scale σ configurations have been built for demonstration purposes, either
for climate prediction or forecasting.
The representation of the surface layers in ocean models is perhaps
the point where progress is the most likely in the coming years. Today,
model solutions in the mixed layer critically depends on the parameterizations. This dependency may decrease as we resolve more physical
processes. It is possible to do so with existing parameterizations simply
by increasing the vertical resolution (to about 1 m) and using higher