Abstract: Using radiosonde data during sounding intensive observation day by day in June 2013, regular observation data, prediction products simulated by WRF model, and automatic station data in south- west, effect of sounding intensive observation data on numerical weather prediction has been analysed by adopting numeric simulation, statistic method and weather analysis. The results show that radiosonde data assimilation increased TS score value, decreased precipitation prediction empty radio and improved prediction quality of 6h accumulated precipitation of moderate rain and heavy rain in total. Radiosonde data assimilation mainly made potential height increase and cyclone circumfluence bate over precipitation areas, and finally affected precipitation location and precipitation intensity. Radiosonde data assimilation decreased root- mean- square error of analysis fields and prediction fields of meteorology elements. The improvement of upper air elements prediction of 0 ~ 24h were obvious, especially 0 ~ 6h prediction. Radiosonde data assimilation had positive effects on surface air elements prediction during almost whole integral time.