A Numerical Study of the Setting Method of System Error in Data Assimilation for Inner Bay

It is expected to reproduce the accurate current and water quality by applying data assimilation to a numerical model for the inner bay. In order for data assimilation to function effectively, it is important to set system errors appropriately. Calculation by adding noise to the boundary condition of the system is one of the methods for setting the system error. However, it is not clear how the system error in the inner bay can be properly expressed by adding noise to the system boundary conditions. In this study, calculation methods of appropriate system error in data assimilation were investigated.

The study was performed by applying the ensemble Kalman filter to the numerical model developed by the authors. In examining the setting method of a system error, twin experiments were conducted in order to focus only on the system boundary conditions and excluded the effects of noise due to system imperfections and observation noise. And in order to perform the examination efficiently, a vertical two-dimensional numerical tank simulating Ise Bay in Japan was used. The target seasons were summer, rainy season, winter, and spring/autumn. The boundary conditions for adding noise were open boundary, weather, and river inflow.

As a result of the experiments, the system noises were different depending on season. Moreover, the knowledge about how to give the noise in the boundary condition for inner bay was obtained.