Parameter estimation in a three-dimensional marine ecosystem model using the adjoint technique

An adjoint data assimilation method is applied to estimate the poorly known ecological parameters in a coupled physical–biological model of the Bohai Sea and the Yellow Sea (BYS). The biological model includes five state variables: dissolved inorganic nitrogen, dissolved inorganic phosphate, phytoplankton, zooplankton and detritus. For the selection of control variables, a new sensitivity analysis method is introduced. This method is based upon the gradient equations of cost function with respect to model parameters. By analyzing the linear combination of Lagrange multiplies in these equations, an unrelated parameter set is selected. In twin experiments, the feasibility of estimated control variables is tested. The results indicate phytoplankton data are the strongest constraints during the parameter estimation process. Climatology daily mean chlorophyll concentrations derived from Modis-Aqua and SeaWiFS are also assimilated into the model to simulate the seasonal cycle of the distributions of phytoplankton. The results indicate a two-bloom characteristic in the BYS, a major bloom at spring and a minor bloom at fall. The simulated horizontal distributions of surface phytoplankton are consistent with the biological roles of seasonal stratification. The concentrations of the nutrients at the surface are lowest in spring, after the phytoplankton bloom. The highest concentrations occur in winter.