28
STEPHEN GRIFFIES
What processes are represented explicitly, and what are the important ones to parameterize? This is one of the most critical and difficult
questions of ocean model design and use. The lectures by Anne Marie
Treguier from this school summarizes many of the issues. She notes
that the choice of model resolution and parameterization prejudices the
simulation so much so that they effectively determine the "ocean" to
be simulated. Discussions in Chassignet and Verron, 1998 thoroughly
survey various aspects of the parameterization problem. This book is
from a 1998 school on ocean modelling and parameterization. Many of
the issues raised there are still unresolved today. Finally, Griffies, 2004
has much to say about some of the common parameterizations used in
ocean climate models.
Numerical methods are necessary to transform the continuum equations into accurate and efficient discrete equations for stepping the ocean
forward in time. There are many methods of use for doing this task.
Should they be based on finite volume methods? Such methods
are becoming more common in ocean modelling. They provide the
numericist with a useful means to take the continuum equations
and cast them onto a finite grid.
What sorts of time stepping schemes are appropriate, and what
properties are essential to maintain? Will the ubiquitous leap-frog
methods2 be supplanted by methods that avoid the problematic
time splitting mode? Chapter 12 of Griffies, 2004 provides a discussion of these points, and argues for the use of a time staggered
method, similar to that discussed by Adcroft and Campin, 2004
and used in the Hallberg Isopycnal Model (Hallberg, 1997) and
Modular Ocean Model version 4 (Griffies et al., 2004).
Should the numerical equations maintain a discrete analog to conservation of energy, tracer, potential vorticity, and potential enstrophy satisfied by the ideal continuum equations? For long term
climate simulations, tracer conservation is critical. What about
the other conserved quantities?
What are the essential features needed for the numerical tracer
advection operator? Should it maintain positivity of the tracer
field? Can such advection operators, which are nonlinear, be easily realized in their adjoint form as required for 4D variational
2As noted in Griffies et al., 2000a, the majority of ocean models supported for large-scale
oceanography continue to use the leapfrog discretization of the time tendency.
STEPHEN GRIFFIES
What processes are represented explicitly, and what are the important ones to parameterize? This is one of the most critical and difficult
questions of ocean model design and use. The lectures by Anne Marie
Treguier from this school summarizes many of the issues. She notes
that the choice of model resolution and parameterization prejudices the
simulation so much so that they effectively determine the "ocean" to
be simulated. Discussions in Chassignet and Verron, 1998 thoroughly
survey various aspects of the parameterization problem. This book is
from a 1998 school on ocean modelling and parameterization. Many of
the issues raised there are still unresolved today. Finally, Griffies, 2004
has much to say about some of the common parameterizations used in
ocean climate models.
Numerical methods are necessary to transform the continuum equations into accurate and efficient discrete equations for stepping the ocean
forward in time. There are many methods of use for doing this task.
Should they be based on finite volume methods? Such methods
are becoming more common in ocean modelling. They provide the
numericist with a useful means to take the continuum equations
and cast them onto a finite grid.
What sorts of time stepping schemes are appropriate, and what
properties are essential to maintain? Will the ubiquitous leap-frog
methods2 be supplanted by methods that avoid the problematic
time splitting mode? Chapter 12 of Griffies, 2004 provides a discussion of these points, and argues for the use of a time staggered
method, similar to that discussed by Adcroft and Campin, 2004
and used in the Hallberg Isopycnal Model (Hallberg, 1997) and
Modular Ocean Model version 4 (Griffies et al., 2004).
Should the numerical equations maintain a discrete analog to conservation of energy, tracer, potential vorticity, and potential enstrophy satisfied by the ideal continuum equations? For long term
climate simulations, tracer conservation is critical. What about
the other conserved quantities?
What are the essential features needed for the numerical tracer
advection operator? Should it maintain positivity of the tracer
field? Can such advection operators, which are nonlinear, be easily realized in their adjoint form as required for 4D variational
2As noted in Griffies et al., 2000a, the majority of ocean models supported for large-scale
oceanography continue to use the leapfrog discretization of the time tendency.