Inverse Modeling Of The Ocean And Atmosphere PDF Download

Inverse modeling activities in oceanography have recently been intensified, aided by the oncoming observational data stream of WOCE and the advance of computer power. However, interpretations of inverse model results from climatological hydrographic data are far from simple. This thesis examines the behavior of an inverse model in the WOCE CME (Community Modeling Effort) results where the physics and the parameter values are known. The ultimate hypotheses to be tested are whether the inferred circulations from a climatological hydrographic data set (where limited time means and spatial smoothing are usually used) represent the climatological ocean general circulations, and what the inferred "diffusion" coefficients really are. The inverse model is first tested in a non-eddy resolving numerical GCM ocean. Numerical/scale analyses are used to test whether the inverse model properly represents the GCM ocean. Experiments show how biased answers could result from an incorrect model, and how a correct model must produce the right answers. When the inverse model is applied to the time-mean hydrographic data of an eddy-resolving GCM ocean in the fine grid resolution of the GCM, the estimated horizontal circulation is statistically consistent with the EGCM time means in both patterns and values. Although the flow patterns are similar, the uncertainties for the GCM time means and the inverse model estimates are different. The former are very large, such that the GCM time-mean circulation has no significance in the deep ocean. The latter are much smaller, and with them the estimated circulations are well defined. This is consistent with the concept that ocean motions are very energetic, while variations of tracers (temperature, salinity) are low frequency. The inverse model succeeded in extracting the ocean general circulation from the "climatological" hydrographic data. The estimated vertical velocities are also statistically indistinguishable from the GCM time means. However, significant differences between the estimated "diffusion" coefficients and the EGCM eddy diffusion coefficients are found at certain locations. These discrepancies are attributed to the differences in physics of the inverse model and the EGCM ocean. The "diffusion" coefficients from the inversion parameterize not only the eddy fluxes, but also (part of) the temporal variation and biharmonic terms which are not explicitly included in the inverse model. Given the essentially red spectrum of the ocean, it makes sense to look for smooth solutions. Aliasing due to subsampling on a coarse grid and the effects of spatial smoothing are addressed in the last part of this thesis. It is shown that this aliasing could be greatly reduced by spatial smoothing. The estimated horizontal circulation from the spatially smoothed time-mean EGCM hydrographic data with a coarse grid resolution (2.4° longitude by 2.0° latitude) is generally consistent with the spatially smoothed EGCM time means. Significant differences only occur at some grid points at great depths, where the GCM circulations are very weak. The conclusions of this study are different from some previous studies. These discrepancies are explained in the concluding chapter. Finally, it should be pointed out that the issue of properly representing a GCM ocean by an inverse model is not identical to the issue of represent ing the real ocean by the same inverse model, since the GCM ocean is not identical to the real ocean. Numerical calculations show that both the non-eddy resolving and the eddy-resolving GCM oceans used in this work are evolving towards a statistical equilibrium. In the real ocean, the importance of temporal variation terms in the property conservation equations should also be analyzed when a steady mverse model is applied to a limited time-mean (the climatological) data set.