Greg Hakim University of Washington
Ensemble Kalman filters (EnKFs) are a potential candidate approach for contributing to the Analysis of Record. This approach acknowledges the fundamentally probabilistic property of state estimation. By using this probabilistic information, EnKFs simplify the problem and offer the potential for overcoming formidable challenges to mesoscale data assimilation. One key challenge involves the background error covariance matrix, which spreads observation influence to all state variables and locations remote from an observation in a dynamically consistent way. On the mesocale, these covariance relationships vary strongly in space and time, rendering questionable the conventional assumption of fixed background covariance. Illustrative examples of these flow-dependent covariances will be discussed, as will current work at the University of Washington toward an operational mesoscale EnkF.