An examination of a multi-scale three-dimensional variational data assimilation scheme in the Kuroshio Extension using the naval coastal ocean model

• We apply an MS3DVAR data assimilation in the Kuroshio Extension.
• First implementation of MS3DVAR in western boundary current region.
• MS3DVAR is capable of handling multiple scales of variability.
• MS3DVAR produces lower analysis errors when compared with traditional 3DVAR.

The implementation of a multi-scale three dimensional variational (MS3DVAR) data assimilation scheme for use with the Navy Coastal Ocean Model (NCOM) in the Kuroshio Extension western boundary current region is presented here. This work leverages on Li et al. (2013), who initially developed this method. MS3DVAR data assimilation allows for the effective assimilation of both spatially coarse and dense collections of observations. Traditional 3DVAR produces an inherent filtering of dynamical features smaller than the decorrelation length. The MS3DVAR allows for a scale selective background error covariance capable of handling a wider range of ocean scales. Here the MS3DVAR is examined in an energetic coastal regime using simulated and real observations. The results show that the MS3DVAR reduces analysis errors when compared to a traditional 3DVAR scheme. Forecast errors appear to be similar for both systems and are most likely due to the coarse resolution of the surface forceing being applied.