Weak constraint four-dimensional variational data assimilation in a model of the California Current System
A new approach is explored for computing estimates of the error covariance associated with the intrinsic errors of a numerical forecast model in regions characterized by upwelling and downwelling. The approach used is based on a combination of strong constraint data assimilation, twin model experiments, linear inverse modeling, and Bayesian hierarchical modeling. The resulting model error covariance estimates Q are applied to a model of the California Current System using weak constraint four-dimensional variational (4D-Var) data assimilation to compute estimates of the ocean circulation. The results of this study show that the estimates of Q derived following our approach lead to demonstrable improvements in the model circulation estimates and isolate regions where model errors are likely to be important and that have been independently identified in the same model in previously published work.