Implementing multi-scale ocean sea surface temperature data assimilation

Numerical ocean circulation models diverge from reality due to non-deterministic processes. Present global and regional, operational ocean data assimilation systems, designed to correct the forecast trajectories, primarily constrain mesoscale variability governed by dynamics associated with the Earth’s rotation. The data assimilation methods discussed in this presentation aim to extend the constraints to scales of surface variability smaller than the mesoscale. The approach is to apply two scales of correction for sea surface temperature (SST). This is achieved in two three-dimensional variational (3DVAR) analyses each producing temperature, salinity, and velocity corrections to the forecast. The first correction utilized present operational data assimilation methods and data streams to correct the mesoscale variability. The scales of variability that are smaller generally than the Rossby radius are filtered out of the innovations for the first correction analysis, since these smaller variations do not represent the mesoscale. The second correction ingests only satellite SST data but utilizes the smaller scales of variability. We find that the present SST data streams provide information on scales much smaller than the Rossby radius decorrelation used in the traditional analysis. As a result, the doubly corrected forecast is more accurate at smaller scales of variability, compared to traditional methods. We describe a regional numerical modelling domain for the Western Atlantic with 1 km resolution and 100 levels that covers the domain form 78⁰W to 54.6⁰W and from 34⁰N to 47.8⁰N. The surface forcing is from the Navy Global Environmental (atmospheric) Model (NAVGEM) and open boundaries from Earth Prediction Capability (ESPC). Preliminary results suggest the new prediction capability will provide more accurate forecast for the near surface ocean boundary layer.