Abstract: An uncomplicated four-dimensional data assimilation scheme is established, following the conditions and characteristics of tropical limited area NWP in Guangzhou Regional Meteorological Center. Parallel tests have been done to analyze in detail the impact of the scheme on the analysis and forecast. It is shown that, by the use of the scheme, the analysis is more equilibrated and approximately consistent with the model atmosphere, and the spin-up time of the model is greatly shortened and the amplitude of adjustment in the initial stage of model integration reduced , improving the forecast, The improvement is much more enormous in data-sparse area than in data-dense area. Forecasts of both height field and wind field are improved satisfactorily with the 4-D assimilation, and so are the evolutions of both mass field and total kinetic energy during the integration. By treating data-sparse areas south of 20°N differently from data-dense areas north of it, in the assimilation, rainfall forecast can be obviously improved as well.