Dr. Paul Mattern, UC Santa Cruz
Marine ecosystem models are widely used and have become important tools for predicting the biochemical state of the ocean and interpolating it to places where no observation are available. Yet unconstrained model simulations can quickly diverge from realistic results and much can be gained from confronting these numerical models with observations: comparison with observations can yield better estimates of biological rate parameters and the assimilation of data into models can drastically improve their results. We present several examples of ecosystem model applications related to model sensitivity, biological parameter estimation and data assimilation: two plankton parameters of a coastal model in the northwestern North Atlantic reveal a time-dependence consistent with biological seasonal cycles in that area. The parameter values have an annual periodicity that can be used to improve the forecasting abilities of the model. Results of an uncertainty analysis for a model simulating hypoxia in the northern Gulf of Mexico suggest that uncertainty in estimates of predicted hypoxia are high with respect to many physical model inputs such as river runoff. Finally, we examine if a more complex, seemingly more realistic ecosystem model formulation yields better results in a data assimilation application for the U.S. west coast.