The implementation of a 4-dimensional variational data assimilation approach for regional interdisciplinary ocean modeling

We report on the implementation of a data assimilation system appropriate for coupled biogeochemical and physical circulation models that builds on the existing 4-dimensional variational (4D-Var) approach presently available within the Regional Ocean Modeling System (ROMS). The method, originally developed within meteorology, accounts for the different error statistics of biogeochemical variables by application of a logarithm transform to biogeochemical fields while preserving Gaussian error statistics for physical fields. The assimilation system is fully coupled dynamically, with observations of biological variables able to influence physical model fields and physical observations contributing to model ecosystem state estimates. The approach has been implemented within ROMS and investigated in both model twin experiments and realistic configurations of the California Current System in which satellite-derived surface chlorophyll estimates along with more traditional physical variables (sea surface height, temperature and salinity) are assimilated. This presentation will focus the relative skill of Gaussian vs lognormal 4D-Var estimates, relative impact of physical data assimilation in isolation, biological data assimilation in isolation, and coupled physical and biological data assimilation. We will also discuss the performance of the coupled asssimilation system in a realistic 1-year configuration, demonstrating the practical viability of this approach to coupled physical and biogeochemical ocean observing systems.