Abstract
We have implemented a four-dimensional variational data assimilation system (4D-Var) for a biogeochemical model to improve a result of 3D simulation of water quality in a coastal sea, and we conducted two twin-experiments with 6 vertical profiles of synthetic observations to verify the system. Experiment 1 that controls initial conditions of temperature, salinity, and eleven biogeochemical state variables reduces only root mean square errors (RMSEs) of temperature, salinity, chlorophyll and DO that are observed. Experiment 2 that controls the initial conditions and thirty-seven biogeochemical parameters reduces all RMSEs of the controlled variables. In addition, assimilation run became stable with 12-hour assimilation windows by imposing a condition to filter the observations.
