MODELLING AND FORECASTING DEEP OCEAN AND NEAR SURFACE MESOSCALE EDDIES: HINDCASTING AND FORECASTING WITH, AND COUPLING A SURFACE BOUNDARY LAYER MODEL TO, THE HARVARD QUASIGEOSTROPHIC MODEL

Hindcasting and forecasting mid-ocean mesoscale flows with a quasigeostrophic open ocean model is studied. The Harvard quasigestrophic open ocean model has been calibrated and documented by others. Here, these studies are synthesized and computationally introduced errors in the integration of the quasigeostrophic equations are attributed. Combining the studies leads to some additional conclusions concerning the boundary conditions and the number of vertical levels. Issues related to the application of the model to data assimilation are focused upon. We note that the hindcast studies to be reported in the following chapter support the calibration results carried out on idealized flows.;A surface boundary layer model is coupled to the quasigeostrophic model to provide a means for process studies of the near surface and as a data assimilation tool. A set of mesoscale boundary layer equations is derived. The result of this derivation is a set of equations which resemble the Ekman equations but include an arbitrary bulk parameterization of mixing within the surface boundary layer, horizontal and vertical advection by the velocity fields of the interior flow, and an additional contribution to the divergence within the boundary layer. The equations are implemented numerically and calibration experiments for idealized flows are described. A Gulf Stream initialization is presented to demonstrate the applicability of the model to realistic ocean flows.;The initialization methodology is developed for a data set (from the POLY-MODE Synoptic Dynamics Experiment) which consists of both temperature and velocity measurements and spans just over one year. A series of thirty standard hindcasts is presented. Comparing the hindcast fields with the data indicates that the primary reason for a hindcast which disagrees with the analyzed fields is a lack of data. It is suggested that the model acts as an interpolator and that the hindcast fields may be better estimates than the statistically interpolated fields. Additional simulations with higher resolution and without topography are performed and demonstrate the importance of topography and marginal benefits associated with higher resolution.