A Strategy for Estimating Cross-Interface Covariance Localization for an Ensemble Coupled Data Assimilation System

We will describe a strategy for estimating the cross-interface covariance localization that will be used for the “strongly-coupled” CESM-DART ocean-atmosphere ensemble coupled data assimilation system being developed at the National Center for Atmospheric Research.  The method is based on the development of a time-constant, geographically varying “correlation attenuation factor” that can be used to limit (and protect) the influence of observations native to the ocean (atmosphere) on the state of the atmosphere (ocean).  The method is based in part on ideas originally presented in Anderson (2012), suggesting that the “attenuation factor” should be chosen to minimize the expected RMS error between the population and the sample correlations used in the ensemble Kalman filter update.   New here is the reformulation of this idea with an informed prior on the population correlation that takes into account the time-nonstationarity.  This prior is based on 5 years of 6-hr forecasts from the CESM-DART “weakly-coupled” integration.