The authors evaluate a three-dimensional variational (3dVAR) system for the South China Sea (SCS) in this study. The assimilation method applied in the system takes into consideration error correlation along each ground track and uses recursive lter for optimiza- tion. Data from three R/V cruises during the spring and summer of 1998 and the summer of 2000 are used to evaluate the system. The root-mean-square error and bias are reduced signi cantly and when the altimeter data are assimilated, the distribution of the error is much closer to the Gaussian distribution. Precipitation and river discharge in the southwestern SCS are reproduced, and the variability of sea surface height is e ciently transferred to the subsurface. The 3dVAR system performs well for each of the three cruises, suggesting that it is steady for routine usage. The authors evaluate a three-dimensional variational (3dVAR) system for the South China Sea (SCS) in this study. The assimilation method applied in the system takes into consideration error correlation along each ground track and uses recursive lter for optimiza- tion. Data from three R / V cruises during the spring and summer of 1998 and the summer of 2000 are used to evaluate the system. The root-mean-square error and bias are reduced signi cantly and when the altimeter data are assimilated, the distribution of the error is much closer to the Gaussian distribution. Precipitation and river discharge in the southwestern SCS are reproduced, and the variability of sea surface height is e ciently transferred to the subsurface. The 3dVAR system performs well for each of the three cruises, suggesting that it is steady for routine usage.