Ocean Multi-Scale Data Assimilation Using Scale-Dependent Vertical Covariances

With the SWOT satellite providing unparalleled high-density altimetry data, data assimilation systems must adapt to make full use of the abundance of new observations to model previously under-sampled ocean features. Compared to traditional 3-dimensional variational (3DVAR) data assimilation, the use of multiple horizontal correlation lengths in multi-scale 3DVAR has been shown to provide better model skill when assimilating high density data. We aim to extend this idea into the vertical regime by determining a set of covariances that describe vertical physics associated with mesoscale and sub-mesoscale features for a multi-scale, multi-physics assimilation system.

We use a spatial filter to separate 4 years of ocean model forecasts into features larger and smaller than 250 km. We calculate the vertical covariances within model profiles at these independent scales as inputs to the Improved Synthetic Ocean Profile (ISOP) system, which converts remotely-measured sea surface height (SSH) altimetry observations into synthetic ocean profiles ready for assimilation. This new scale-dependent ISOP system is used in conjunction with established MS-3DVAR technique. We present results by simulating SWOT observations for Observing System Simulation Experiments (OSSEs) to quantify how effectively the new assimilation method causes the model to evolve towards a known truth.