Assessing the Value of BGC Argo Profile Observations for Ocean Biogeochemical Data Assimilation in a Model of the Gulf of Mexico

Biogeochemical models are useful tools for improving understanding of ocean processes and for prediction. However, their results contain uncertainties arising from conceptual and numerical errors, and uncertainties in model initialization, external forcing, and parameters. Data assimilation methods, which dynamically incorporate observations into models, can provide improved estimates compared to just models or observations in isolation, and can lead to improvements in predictive skill. Biogeochemical data assimilation falls into two categories: parameter optimization and state estimation; both critically depend on appropriate observations. Satellite data of ocean color have been the major source for biogeochemical data assimilation. They provide synoptic observations at a relatively high temporal and spatial resolution, but are limited to the near surface of the ocean. BGC Argo increasingly provides subsurface information of chlorophyll and other parameters. Here, we assess the value that these profiles add beyond satellite chlorophyll observations for biogeochemical data assimilation. We conducted parameter optimization experiments to determine poorly known biogeochemical parameters and state estimates using a deterministic Ensemble Kalman Filter (DEnKF) to sequentially update the model state. Results show that parameters optimized with respect to satellite data cannot reproduce subsurface distributions unless profiling observations (chlorophyll, phytoplankton, and POC) are used. The state updates show that assimilating profiling observations can further reduce the bias between model results and observations.