Towards an integrated observation and modeling system in the New York Bight using variational methods. Part II : repressenter-based observing strategy evaluation

Abstract

As part of an effort to build an integrated observation and modeling system for the New York Bight, this study explores observing system design using a representer-based method. The Representer of a single observation describes the covariance between the observed quantity and ocean state at all locations at any time. It is related closely to the influence of the observation on control variable correction in a 4D Variational data assimilation system. We prove that these properties hold for the combination of representers that is associated with an arithmetic function of model variables or a group of observations. The representer-based method is used here to identify which of a set of proposed tracks for an autonomous coastal glider is better for predicting horizontal salt flux within the Hudson Shelf Valley in a 2-day forecast period. Twin experiments confirm the result. The system is also used to compare different observation strategies. We show that a glider that traverses a regular transect influences a larger area than a continuously profiling mooring, but the mooring carries stronger influence at the observation location. The representer analysis shows how the information provided by observations extends toward the dynamically upstream and how increasing the duration of the analysis window captures more dynamical connections and expands the area of influence of the observations in data assimilation. Overall, the study demonstrates that the representer methodology can quantitatively contrast different observational strategies and determine spatial patterns and temporal extent of the influence of observations, both of which are helpful for evaluating the design of observation networks.