A data assimilation technique applied to estimate parameters for the NEMURO marine ecosystem model

We have applied a data assimilation technique to determine biological parameters in the PICES (North Pacific Marine Science Organization) proto type lower trophic level model (NEMURO). North Pacific Ecosystem Model for Understanding Regional Oceanography (NEMURO) has about 80 biological parameters and 11 initial values. We used a sensitivity analysis to choose eight parameters which mostly impacted the simulated values of interest. These parameters were selected as control variables for the data assimilation. Using an adjoint method, we assimilated biological and chemical data from Stn.A7 (off Hokkaido, Japan) into the model. Twin experiments were conducted to determine whether the data constrain those eight control variables. Model output, using optimum parameter values determined by the assimilation, agreed with the data better than those obtained with the first guess parameter values. But some problems still remain even in the simulations using the optimum parameters: namely a large discrepancy is seen between the simulation and data for small zooplankton and the simulated bloom of large phytoplankton that is too large.