32
Kirk Bryan
model, or any model for that matter. Doug Lilly was an excellent mentor, introducing
me to the literature, and his own notation, which he had developed for analyzing
conservation properties of numerical models. Another research group with the same
aim, but with much less in the way of resources, started up at about the same time
at the University of California in Los Angeles, headed by Yale Mintz. There was
a continual flow of ideas between the two groups. The numerical methods of Akia
Arakawa (1966) at UCLA were key to the success of the first attempts to model ocean
circulation at the Weather Bureau.
It is hard today to imagine how primitive even the most advanced computers
of the early 1960s were. A large room full of equipment was far less capable than
the simplest modern laptop. The two-dimensional flow of an idealized wind-driven
ocean model appeared to be a tractable problem. Linear models had been developed
by Stommel and Munk and there was a good deal of speculation about the role of
inertial terms in the equations of motion, which could not be easily studied by analytic
methods. From a modern perspective the two-dimensional wind-driven ocean model
is only of academic interest, but the successful extension of the model into a nonlinear
domain (Bryan, 1963) excited a lot of interest. The next step, an extension to threedimensions, was certainly premature considering the limited computing resources
available. We were well aware of the difficulty of resolving the long time-scales,
and wide spectrum of spatial scales of the ocean circulation. The prudent strategy
would have been to hold off for a few years. However, that was not the philosophy
in the laboratory. Joe “Smag” was an incurable optimist, and an early disciple of
“Moore’s Law,” which nearly 40 years ago projected that the number of transistors
on a computer chip would double every 2 years. Global models of the ocean were
needed to construct global models of air–sea interaction, and we were going to build
them, regardless of the limitations of existing computers.
At about this time Michael Cox (Figure 3.3) joined Smag’s group. He was first
hired as a computer operator. He then became an accomplished programmer, and
eventually a leading ocean modeler. Mike Cox had extraordinary technical ability
along with a great intellectual curiosity. Quite a few hours of the week he would be
in the little library of the laboratory, and would constantly bring to my attention new
references for the work we were doing. Together, we developed a three-dimensional
model based on the primitive equations and incorporating Akia Arakawa’s ideas
in the numerical design (Bryan, 1969). This model with later contributions from
many people is the remote ancestor of many of the ocean circulation models used
in ocean–climate and ocean–geochemistry applications today (Griffies, 2004). Since
Mike Cox and I could not hope to resolve all the scales of the ocean circulation,
we used the newly minted and fashionable ideas of geophysical fluid dynamics to
formulate “oceanlike” problems that were more feasible, but still allowed us to move
forward (Bryan and Cox, 1967, 1968a,b). At the end of the 1960s other groups
began to develop ocean circulation models. In the former Soviet Union Artem Sarkisyan developed a geostrophic model for the diagnosis of ocean circulation. Gunter
Fischer and Jim O’Brien were active at the National Center for Atmospheric
Kirk Bryan
model, or any model for that matter. Doug Lilly was an excellent mentor, introducing
me to the literature, and his own notation, which he had developed for analyzing
conservation properties of numerical models. Another research group with the same
aim, but with much less in the way of resources, started up at about the same time
at the University of California in Los Angeles, headed by Yale Mintz. There was
a continual flow of ideas between the two groups. The numerical methods of Akia
Arakawa (1966) at UCLA were key to the success of the first attempts to model ocean
circulation at the Weather Bureau.
It is hard today to imagine how primitive even the most advanced computers
of the early 1960s were. A large room full of equipment was far less capable than
the simplest modern laptop. The two-dimensional flow of an idealized wind-driven
ocean model appeared to be a tractable problem. Linear models had been developed
by Stommel and Munk and there was a good deal of speculation about the role of
inertial terms in the equations of motion, which could not be easily studied by analytic
methods. From a modern perspective the two-dimensional wind-driven ocean model
is only of academic interest, but the successful extension of the model into a nonlinear
domain (Bryan, 1963) excited a lot of interest. The next step, an extension to threedimensions, was certainly premature considering the limited computing resources
available. We were well aware of the difficulty of resolving the long time-scales,
and wide spectrum of spatial scales of the ocean circulation. The prudent strategy
would have been to hold off for a few years. However, that was not the philosophy
in the laboratory. Joe “Smag” was an incurable optimist, and an early disciple of
“Moore’s Law,” which nearly 40 years ago projected that the number of transistors
on a computer chip would double every 2 years. Global models of the ocean were
needed to construct global models of air–sea interaction, and we were going to build
them, regardless of the limitations of existing computers.
At about this time Michael Cox (Figure 3.3) joined Smag’s group. He was first
hired as a computer operator. He then became an accomplished programmer, and
eventually a leading ocean modeler. Mike Cox had extraordinary technical ability
along with a great intellectual curiosity. Quite a few hours of the week he would be
in the little library of the laboratory, and would constantly bring to my attention new
references for the work we were doing. Together, we developed a three-dimensional
model based on the primitive equations and incorporating Akia Arakawa’s ideas
in the numerical design (Bryan, 1969). This model with later contributions from
many people is the remote ancestor of many of the ocean circulation models used
in ocean–climate and ocean–geochemistry applications today (Griffies, 2004). Since
Mike Cox and I could not hope to resolve all the scales of the ocean circulation,
we used the newly minted and fashionable ideas of geophysical fluid dynamics to
formulate “oceanlike” problems that were more feasible, but still allowed us to move
forward (Bryan and Cox, 1967, 1968a,b). At the end of the 1960s other groups
began to develop ocean circulation models. In the former Soviet Union Artem Sarkisyan developed a geostrophic model for the diagnosis of ocean circulation. Gunter
Fischer and Jim O’Brien were active at the National Center for Atmospheric