Simultaneously assimilating multi-source observations into a three-dimensional suspended cohesive sediment transport model by the adjoint method in the Bohai Sea

The performance of the suspended cohesive sediment transport model can be improved by using data assimilation; however, only one source of observations of suspended sediment concentrations (SSCs) is assimilated in the previous studies. This study investigates the simultaneous assimilation of multi-source SSC observations, including in-situ SSC observations and GOCI-retrieved SSCs, into a three-dimensional suspended cohesive sediment transport model by the adjoint method in the Bohai Sea. The artificial SSC observations obtained by running the suspended cohesive sediment transport model are firstly assimilated in the twin experiments. When the initial surface condition obtained using GOCI-retrieved SSCs was used, the model performance after assimilating multi-source artificial SSC observations was improved than that after assimilating only artificial GOCI-retrieved SSCs or in-situ SSC observations. The real multi-source SSC observations are then assimilated in practical experiments. The experimental results indicate that the initial conditions are not only important for SSC simulations, but also significant for data assimilation. Except for the surface layer, assimilating only GOCI-retrieved SSCs can significantly improve the simulated SSCs in the middle and bottom layers. On the whole, the results of simultaneously assimilating multi-source SSC observations are just slightly closer to the SSC observations than those after assimilating only GOCI-retrieved SSCs, but the convergence of adjoint data assimilation is accelerated and the model performance in deep layers is further improved, demonstrating the effectiveness of simultaneously assimilating multi-source SSC observations.