70
STEPHEN GRIFFIES
gorithm methodology. For example, those aiming to faithfully represent the ocean’s quasi-adiabatic interior generally prefer an isopycnal
layered model using quasi-Lagrangian methods over either terrain following or geopotential models using quasi-Eulerian methods. However,
modelling, largely due to the simplicity of representing and parameterizing air-sea and ice-sea interactions as well as the ocean mixed layer.
Hence, these models remain the dominant tool for global climate modelers, even given their well known problems with spurious mixing and
difficulties handling overflow processes (see the discussion in Griffies
et al., 2000a). Additionally, non-hydrostatic models, such as that from
Marshall et al., 1997, have traditionally used geopotential coordinates.
Indeed, there is presently no non-hydrostatic algorithm for use in the
ocean that is based on a quasi-Lagrangian algorithm. That is, all layered models for the ocean are hydrostatic. Finally, those focusing on
shallow ocean dynamics and estuaries have traditionally chosen terrain
following coordinates due to their fidelity with bottom boundary layer
processes. However, such models have only recently been employed for
global climate studies, largely due to difficulties with pressure gradient
errors (see Section 2 of Griffies et al., 2000a).
In summary, it is unlikely that modelers will arrive at one universally best vertical coordinate. Instead, vertical coordinates will remain
chosen for the particular model application in mind. Modelers may,
however, converge on an optimal algorithm methodology, especially if
quasi-Lagrangian methods can be extended to non-hydrostatic models.
In general, it is useful for model designs to evolve from being based on
a single vertical coordinate, to model environments mentioned in Section 1.3 that are flexible enough to include many vertical coordinate
algorithms.
7.
Closing remarks
It is incumbent on ocean model designers and developers to provide a
thorough and pedagogical rationalization of their codes, from the basic
equations that the model aims to integrate, to the limitations of their
subgrid scale (SGS) parameterizations. Likewise, it is essential that
model users understand elements of the model algorithms and SGS parameterizations. The sophisticated and productive use of ocean models
comes from a firm understanding of model fundamentals. It is hoped
that through more schools such as this one or those documented by
O’Brien, 1986, Chassignet and Verron, 1998, and others, as well as
books on the subject of geophysical fluid modelling such as Haltiner and
there have been decades of experience with -models for global climate
z
STEPHEN GRIFFIES
gorithm methodology. For example, those aiming to faithfully represent the ocean’s quasi-adiabatic interior generally prefer an isopycnal
layered model using quasi-Lagrangian methods over either terrain following or geopotential models using quasi-Eulerian methods. However,
modelling, largely due to the simplicity of representing and parameterizing air-sea and ice-sea interactions as well as the ocean mixed layer.
Hence, these models remain the dominant tool for global climate modelers, even given their well known problems with spurious mixing and
difficulties handling overflow processes (see the discussion in Griffies
et al., 2000a). Additionally, non-hydrostatic models, such as that from
Marshall et al., 1997, have traditionally used geopotential coordinates.
Indeed, there is presently no non-hydrostatic algorithm for use in the
ocean that is based on a quasi-Lagrangian algorithm. That is, all layered models for the ocean are hydrostatic. Finally, those focusing on
shallow ocean dynamics and estuaries have traditionally chosen terrain
following coordinates due to their fidelity with bottom boundary layer
processes. However, such models have only recently been employed for
global climate studies, largely due to difficulties with pressure gradient
errors (see Section 2 of Griffies et al., 2000a).
In summary, it is unlikely that modelers will arrive at one universally best vertical coordinate. Instead, vertical coordinates will remain
chosen for the particular model application in mind. Modelers may,
however, converge on an optimal algorithm methodology, especially if
quasi-Lagrangian methods can be extended to non-hydrostatic models.
In general, it is useful for model designs to evolve from being based on
a single vertical coordinate, to model environments mentioned in Section 1.3 that are flexible enough to include many vertical coordinate
algorithms.
7.
Closing remarks
It is incumbent on ocean model designers and developers to provide a
thorough and pedagogical rationalization of their codes, from the basic
equations that the model aims to integrate, to the limitations of their
subgrid scale (SGS) parameterizations. Likewise, it is essential that
model users understand elements of the model algorithms and SGS parameterizations. The sophisticated and productive use of ocean models
comes from a firm understanding of model fundamentals. It is hoped
that through more schools such as this one or those documented by
O’Brien, 1986, Chassignet and Verron, 1998, and others, as well as
books on the subject of geophysical fluid modelling such as Haltiner and
there have been decades of experience with -models for global climate
z