State and parameter update of a hydrodynamic-phytoplankton model using ensemble Kalman filter

Kalman filter has been successfully used in assimilating observations into the existing models, and has been continually adjusted for its wider use. In this study, one of the Kalman filter techniques (ensemble Kalman filter) was used to assimilate measured data into a spatial hydrodynamic-phytoplankton model for predicting dynamics of phytoplankton biomass in Lake Taihu. In order to investigate the effects of the initial conditions (chlorophyll a) and the model parameter on the model fit, we carried out three simulations with different update strategies of parameter and variable using ensemble Kalman filter. Two simulations updated both of model parameter and state variable once and twice a week, respectively. Another simulation updated the state variable once a week, respectively. The simulation results show that the model fit was improved when the state variable (chlorophyll a) was updated by measured data in a shorter term, and was slightly improved with time-varying parameter. In this respect, good estimates of initial chlorophyll a conditions are critical to achieve good predictions of chlorophyll a dynamics in Lake Taihu. This study demonstrates the success of the ensemble Kalman filter technique in improving models' predictive skill, and implementing time-varying parameters for ecological models