In order to model the eddies accurately with regard to time and location in the South China Sea, an incremental 4D-Var data assimilation system is performed with the Regional Ocean Modeling System(ROMS). The observations used in the assimilation cycles include the following: 1-day grided, 0.25°sla from AVISO, 1-day gridded, 0.1°resolution SST from POES AVHRR, and surface and subsurface hydrographic measurements of temperature and salinity collected with ARGO float from UK Met-Office EN4 project dataset. The assimilated model results are compared to forward model results and independent observations. The results show that 4D-Var data assimilation makes an improvement in predicting the mesoscale eddies in the region of interest even though the parameters of 4D-Var may be not optimal and need more effort to be adjusted.