40
Kirk Bryan
horizontal mixing across tilted isopycnal surfaces could reverse the predicted upwelling in the main thermocline in the GFDL ocean model. Following up on Rooth
and Willebrand’s ideas, and an earlier paper by Solomon (1971), Redi (1982) formulated a parametrization in which the mixing takes place along density surfaces
in a Cartesian coordinate model. Now this idea is best known as one element in a
more comprehensive lateral mixing parametrization by Gent and McWilliams (1990).
Designing the correct numerical implementation of this parametrization proved to be
much more difficult, and the details have only recently been worked out (Griffies et al.,
1998).
At Wally Broeker’s suggestion we brought two of his students to Princeton,
Jorge Sarmiento and Robbie Toggweiler. This allowed us to explore the application
of our ocean circulation models to geochemistry. The new field rapidly became a
flourishing enterprise. In spite of the limited resolution of the models used, the early
simulations of Sarmiento (1983) and Toggweiler et al. (1989a,b) offered promising correspondence with geochemical data and the best estimates at the time of the
expected ocean sequestration of anthropogenic carbon dioxide. In the early 1980s
the Max Planck Institute in Hamburg also became interested in ocean geochemical
modeling. We developed a friendly two-way rivalry in this area. Ocean geochemical
modeling grew with the participation of many more laboratories both in the United
States and abroad.
Early versions of the coupled ocean–atmosphere models contained a model of
passive sea-ice for the polar areas, but we were under no illusions as to its completeness. In the 1970s Bill Hibler became a frequent visitor to GFDL and we collaborated
in coupling the ocean circulation model with his model of ice dynamics (Hibler and
Bryan, 1984). The early tests of the coupled ocean and active sea-ice were very encouraging. They illustrated in a quantitative way how ocean currents were responsible
for the observed position of the sea boundary in the subpolar North Atlantic. Gradually active sea-ice models are becoming a standard component of climate models,
but it is very difficult to realize their full potential because the ocean’s synoptic scales
are so extremely small in the Arctic.
The GFDL model is based on Cartesian coordinates. In the 1980s two other
ocean circulation model architectures became widely accepted. In the “sigma” models
the vertical coordinate is normalized by the total depth. Pioneering development
on this type of model was carried out by Blumberg and Mellor (1987) and it has
become a nearly standard model for coastal and near shore applications. Bleck and
Boudra(1981) developed the first ocean circulation model based on hybrid vertical
coordinates. A Lagrangian vertical coordinate is used for the main thermocline, based
on isopycnal surfaces, while a Cartesian coordinate is used for the mixed layer. Both
architectures have unique advantages. The “sigma” model has a much better treatment
of complex bottom topography, but has the disadvantage of inherent errors in the
calculation of horizontal pressure gradients over steep bottom slopes. Since mixing
and advection in the main thermocline largely take place along rather than across
isopycnal surfaces, isopycnal surfaces appear to be the most “natural” coordinate
Kirk Bryan
horizontal mixing across tilted isopycnal surfaces could reverse the predicted upwelling in the main thermocline in the GFDL ocean model. Following up on Rooth
and Willebrand’s ideas, and an earlier paper by Solomon (1971), Redi (1982) formulated a parametrization in which the mixing takes place along density surfaces
in a Cartesian coordinate model. Now this idea is best known as one element in a
more comprehensive lateral mixing parametrization by Gent and McWilliams (1990).
Designing the correct numerical implementation of this parametrization proved to be
much more difficult, and the details have only recently been worked out (Griffies et al.,
1998).
At Wally Broeker’s suggestion we brought two of his students to Princeton,
Jorge Sarmiento and Robbie Toggweiler. This allowed us to explore the application
of our ocean circulation models to geochemistry. The new field rapidly became a
flourishing enterprise. In spite of the limited resolution of the models used, the early
simulations of Sarmiento (1983) and Toggweiler et al. (1989a,b) offered promising correspondence with geochemical data and the best estimates at the time of the
expected ocean sequestration of anthropogenic carbon dioxide. In the early 1980s
the Max Planck Institute in Hamburg also became interested in ocean geochemical
modeling. We developed a friendly two-way rivalry in this area. Ocean geochemical
modeling grew with the participation of many more laboratories both in the United
States and abroad.
Early versions of the coupled ocean–atmosphere models contained a model of
passive sea-ice for the polar areas, but we were under no illusions as to its completeness. In the 1970s Bill Hibler became a frequent visitor to GFDL and we collaborated
in coupling the ocean circulation model with his model of ice dynamics (Hibler and
Bryan, 1984). The early tests of the coupled ocean and active sea-ice were very encouraging. They illustrated in a quantitative way how ocean currents were responsible
for the observed position of the sea boundary in the subpolar North Atlantic. Gradually active sea-ice models are becoming a standard component of climate models,
but it is very difficult to realize their full potential because the ocean’s synoptic scales
are so extremely small in the Arctic.
The GFDL model is based on Cartesian coordinates. In the 1980s two other
ocean circulation model architectures became widely accepted. In the “sigma” models
the vertical coordinate is normalized by the total depth. Pioneering development
on this type of model was carried out by Blumberg and Mellor (1987) and it has
become a nearly standard model for coastal and near shore applications. Bleck and
Boudra(1981) developed the first ocean circulation model based on hybrid vertical
coordinates. A Lagrangian vertical coordinate is used for the main thermocline, based
on isopycnal surfaces, while a Cartesian coordinate is used for the mixed layer. Both
architectures have unique advantages. The “sigma” model has a much better treatment
of complex bottom topography, but has the disadvantage of inherent errors in the
calculation of horizontal pressure gradients over steep bottom slopes. Since mixing
and advection in the main thermocline largely take place along rather than across
isopycnal surfaces, isopycnal surfaces appear to be the most “natural” coordinate