OS31D-1030:
A New Approach for 3D Ocean Reconstruction from Limited Observations

Wednesday, 17 December 2014
Xiao Xiao, New York University, New York, NY, United States
Abstract:
Satellites can measure ocean surface height and temperature with sufficient spatial and temporal resolution to capture mesoscale features across the globe. Measurements of the ocean's interior, however, remain sparse and irregular, thus the dynamical inference of subsurface flows is necessary to interpret surface measurements. The most common (and accurate) approach is to incorporate surface measurements into a data-assimilating forward ocean model, but this approach is expensive and slow, and thus completely impractical for time-critical needs, such as offering guidance to ship-based observational campaigns. Two recently-developed approaches have made use of the apparent partial consistency of upper ocean dynamics with quasigeostrophic flows that take into account surface buoyancy gradients (i.e. the "surface quasigeostrophic" (SQG) model) to "reconstruct" the interior flow from knowledge of surface height and buoyancy.
Here we improve on these methods in three ways: (1) we adopt a modal decomposition that represents the surface and interior dynamics in an efficient way, allowing the separation of surface energy from total energy; (2) we make use of instantaneous vertical profile observations (e.g. from ARGO data) to improve the reconstruction of eddy variables at depth; and (3) we use advanced statistical methods to choose the optimal modes for the reconstruction. The method is tested using a series of high horizontal and vertical resolution quasigeostrophic simulation, with a wide range of surface buoyancy and interior potential vorticity gradient combinations. In addtion, we apply the method to output from a very high resolution primitive equation simulation of a forced and dissipated baroclinic front in a channel. Our new method is systematically compared to the existing methods as well. Its advantages and limitations will be discussed.