SOME OCEAN MODEL FUNDAMENTALS
21
vertical coordinates. Another approach, starting from an isopycnal layered model, has been taken by the Hybrid Coordinate Ocean Model
(HYCOM) of Bleck, 2002. HYCOM is arguably the most mature of the
generalized vertical coordinate models.
From an abstract perspective, it is a minor point that different modelers use the same code, since in principle all that matters should be the
continuum equations which are discretized. This perspective has, unfortunately, not been realized in practice. Differences in fundamentals
of the formulation and/or numerical methods often serve to make the
simulations quite distinct, even when in principle they should be nearly
identical. Details do matter, especially when considering long time scale
climate studies where small differences have years to magnify.
An argument against merging model development efforts is that there
is creative strength in diversity, and so there should remain many ocean
codes. A middle ground is argued here, whereby we maintain the framework for independent creative work and innovation, yet little effort is
wasted developing redundant software and/or trying to compare different model outputs using disparate conventions. To further emphasize
this point, we stress that the problems of ocean climate and operational
oceanography are vast and complex, thus requiring tremendous human
and computational resources. This situation calls for merging certain
efforts to optimize available resources. Furthermore, linking modelers
together to use a reduced set of code environments does not squelch creativity nor does it lead to less diversity in algorithmic approaches. Instead, environments ideally can provide modelers with common starting
points from which to investigate different methodologies, parameterizations, and the like.
The proposal for model environments is therefore analogous to use of a
few spoken/written languages (e.g., english, french) to communicate and
formulate arguments, or a few computer languages (e.g., Fortran, C++)
to translate numerical equations into computer code. Focusing on a few
ocean model environments, rather than many ocean models, can lead to
enhanced collaboration by removing awkward and frustrating barriers
that exist between the presently wide suite of model codes. Ultimately,
such will (it is hoped!) lead to better and more reproducible simulations,
thus facilitating the maturation of ocean modelling into a more robust
and respectable scientific discipline.
1.2
Some fundamental questions
It is possible to categorize nearly every question about ocean modelling into three classes.
21
vertical coordinates. Another approach, starting from an isopycnal layered model, has been taken by the Hybrid Coordinate Ocean Model
(HYCOM) of Bleck, 2002. HYCOM is arguably the most mature of the
generalized vertical coordinate models.
From an abstract perspective, it is a minor point that different modelers use the same code, since in principle all that matters should be the
continuum equations which are discretized. This perspective has, unfortunately, not been realized in practice. Differences in fundamentals
of the formulation and/or numerical methods often serve to make the
simulations quite distinct, even when in principle they should be nearly
identical. Details do matter, especially when considering long time scale
climate studies where small differences have years to magnify.
An argument against merging model development efforts is that there
is creative strength in diversity, and so there should remain many ocean
codes. A middle ground is argued here, whereby we maintain the framework for independent creative work and innovation, yet little effort is
wasted developing redundant software and/or trying to compare different model outputs using disparate conventions. To further emphasize
this point, we stress that the problems of ocean climate and operational
oceanography are vast and complex, thus requiring tremendous human
and computational resources. This situation calls for merging certain
efforts to optimize available resources. Furthermore, linking modelers
together to use a reduced set of code environments does not squelch creativity nor does it lead to less diversity in algorithmic approaches. Instead, environments ideally can provide modelers with common starting
points from which to investigate different methodologies, parameterizations, and the like.
The proposal for model environments is therefore analogous to use of a
few spoken/written languages (e.g., english, french) to communicate and
formulate arguments, or a few computer languages (e.g., Fortran, C++)
to translate numerical equations into computer code. Focusing on a few
ocean model environments, rather than many ocean models, can lead to
enhanced collaboration by removing awkward and frustrating barriers
that exist between the presently wide suite of model codes. Ultimately,
such will (it is hoped!) lead to better and more reproducible simulations,
thus facilitating the maturation of ocean modelling into a more robust
and respectable scientific discipline.
1.2
Some fundamental questions
It is possible to categorize nearly every question about ocean modelling into three classes.