Regional data assimilation is conducted for a coastal estuary using the ensemble Kalman filter, real observation data from Ise Bay, Japan, and a simulation model called the Ise Bay Simulator. The applicability and robustness of the method are then examined. We also analyze the relationship between the boundary conditions, which add perturbations and the data assimilation results of water temperature and salinity. A method of creating an ensemble by perturbing three boundary conditions (atmospheric forcing, lateral boundary conditions, river discharge forcing) is then proposed. In situ water temperature and salinity profiles observed at fixed points are assimilated daily. The proposed assimilation method provides stable data assimilation without unnatural values for water temperature and salinity throughout the year. Further, applying a perturbation to the three boundary conditions does not lead to filter divergence, thus indicating good applicability and robustness. Applying a perturbation to the three boundary conditions does not degenerate the ensemble spread. According to a sensitivity experiment, perturbing the atmospheric boundary conditions of air temperature and wind speed increases the ensemble spread of water temperature, especially near the surface layer. Wind speed has the greatest influence on the magnitude of the salinity ensemble spread, and its dominance depends on location. Perturbation of lateral boundary conditions increases the ensemble spread of water temperature and salinity at all water depths near the bay mouth, and the observations are effectively assimilated. Perturbation of river discharge forcing successfully assimilates water temperature and salinity near the estuary.