Parameter Optimization of a Box and a Three-Dimensional Ocean Carbon Cycle Model Using the Adjoint Method: What Can We Learn?

A data assimilation technique is used to optimize marine ecosystem parameters that constraining carbon fluxes in a three-dimensional ocean carbon cycle model. An identical twin experiment is able to produce an a posteriori parameter set with significantly reduced model-data misfit. Error analysis using the Hessian matrix is applied to provide detailed information on the sensitivity of different parameters. Parameters associated with zooplankton grazing and phytoplankton growth reveal to be most sensitive with respect to surface chlorophyll concentration. In order to analyze the sensitivity of these parameters in more detail, biweekly BATS data are compared with a box model as well as with three-dimensional model. Interestingly, when different ocean regions are assimilated separately based on different physical and biogeochemical conditions, regional variation of ecosystem parameters significantly reduces the model-data misfit. Our results also show that certain regions need to have improved ecosystem model formulations in order to reduce the model data-misfit further.