Modeling Ocean Circulation
37
et al., 1977), Bert combined the best features of the UCLA ocean model and the
GFDL model and distributed the new code (Semtner, 1974) . In Princeton we adopted
Semtner’s code, which was clearly superior to the one we were using. A little later Joe
Smagorinsky came to me about a request he had received for a copy of our code, and
told me that it was against lab policy to share code. Naturally he was a little surprised
when I told him that at present we were not using our own code, but the UCLA code!
Once the precedent was set by Bert Semtner, code distribution continued through
the years. Cox (1984) distributed regular code updates through e-mail, which was
just coming into use. A great boon to modelers were the global climatological data
sets (Levitus, 1982; Hellerman and Rosenstein,1983). After Cox’s untimely death in
1989, Ron Pacanowski (Pacanowski et al., 1991) took over the task of updating and
distributing what became known as the GFDL Modular Ocean Model (MOM). A
detailed description of the structure and physics of this type of model is given in a
recent book by Griffies (2004).
The 1970s was the decade of MODE, a U.S. program to measure mesoscale,
ocean eddies. This program spawned some very important ocean modeling by Peter
Rhines, William Holland, and others. Models of ocean circulation on a planetary scale
were not in fashion because critics pointed out that there was not enough data available
to verify such a model, and a model which could not resolve mesoscale eddies would
not be valid in any case. To counter the prevailing thinking, Larry Lewis and I carried
out a calculation which we called “A Water Mass Model of the World Ocean” (Bryan
and Lewis, 1979). We felt that the main features of the temperature and salinity of the
World Ocean were reasonably well known, even if the details of the ocean circulation
were not. The goal was to make a more detailed calculation of the evolution of the main
thermocline than had been possible for Takano et al. (NAS,1975), resolving the
long time scales of adjustment of the main thermocline. The great difficulty of the
calculation was the centuries of “spin up” time required to get a balanced state of
the ocean. We used the method of grid refinement pioneered by Cox’s Indian Ocean
study. The final solution was on a 2
◦
× 2
◦ latitude longitude grid. This resolution was
no better than that used by Cox (NAS,1975) in his earlier global model, but Cox’s case
was an essentially diagnostic calculation, using observed data as input. The World
Ocean forward model did not have the resolution to give a good simulation of western
boundary currents, but the Bryan and Lewis calculation did model the main features
of the global temperature and salinity fields, and gave a remarkable simulation of the
formation of Antarctic Intermediate Water in the Southern Hemisphere.
After this World Ocean calculation the way seemed clear to couple the ocean
and atmospheric models together to make a realistic global climate model. Manabe
and I had already made several different calculations in idealized geometries, which
tested ocean–atmosphere coupling, and even the incorporation of sea ice. However,
simulating the present global climate in a coupled model turned out to be much more
difficult than we expected. Without the constraint of fixed upper boundary conditions,
the ocean component of the coupled model did many unexpected things. The deep
water circulation in the model seemed to bear no resemblance to what was expected.
37
et al., 1977), Bert combined the best features of the UCLA ocean model and the
GFDL model and distributed the new code (Semtner, 1974) . In Princeton we adopted
Semtner’s code, which was clearly superior to the one we were using. A little later Joe
Smagorinsky came to me about a request he had received for a copy of our code, and
told me that it was against lab policy to share code. Naturally he was a little surprised
when I told him that at present we were not using our own code, but the UCLA code!
Once the precedent was set by Bert Semtner, code distribution continued through
the years. Cox (1984) distributed regular code updates through e-mail, which was
just coming into use. A great boon to modelers were the global climatological data
sets (Levitus, 1982; Hellerman and Rosenstein,1983). After Cox’s untimely death in
1989, Ron Pacanowski (Pacanowski et al., 1991) took over the task of updating and
distributing what became known as the GFDL Modular Ocean Model (MOM). A
detailed description of the structure and physics of this type of model is given in a
recent book by Griffies (2004).
The 1970s was the decade of MODE, a U.S. program to measure mesoscale,
ocean eddies. This program spawned some very important ocean modeling by Peter
Rhines, William Holland, and others. Models of ocean circulation on a planetary scale
were not in fashion because critics pointed out that there was not enough data available
to verify such a model, and a model which could not resolve mesoscale eddies would
not be valid in any case. To counter the prevailing thinking, Larry Lewis and I carried
out a calculation which we called “A Water Mass Model of the World Ocean” (Bryan
and Lewis, 1979). We felt that the main features of the temperature and salinity of the
World Ocean were reasonably well known, even if the details of the ocean circulation
were not. The goal was to make a more detailed calculation of the evolution of the main
thermocline than had been possible for Takano et al. (NAS,1975), resolving the
long time scales of adjustment of the main thermocline. The great difficulty of the
calculation was the centuries of “spin up” time required to get a balanced state of
the ocean. We used the method of grid refinement pioneered by Cox’s Indian Ocean
study. The final solution was on a 2
◦
× 2
◦ latitude longitude grid. This resolution was
no better than that used by Cox (NAS,1975) in his earlier global model, but Cox’s case
was an essentially diagnostic calculation, using observed data as input. The World
Ocean forward model did not have the resolution to give a good simulation of western
boundary currents, but the Bryan and Lewis calculation did model the main features
of the global temperature and salinity fields, and gave a remarkable simulation of the
formation of Antarctic Intermediate Water in the Southern Hemisphere.
After this World Ocean calculation the way seemed clear to couple the ocean
and atmospheric models together to make a realistic global climate model. Manabe
and I had already made several different calculations in idealized geometries, which
tested ocean–atmosphere coupling, and even the incorporation of sea ice. However,
simulating the present global climate in a coupled model turned out to be much more
difficult than we expected. Without the constraint of fixed upper boundary conditions,
the ocean component of the coupled model did many unexpected things. The deep
water circulation in the model seemed to bear no resemblance to what was expected.