The Advanced Regional Prediction System model is used to simulate a precipitation case occurred from 18 to 22 September 2010 in the western North China complex topography. A set of sensitivity tests is also conducted to study the impact of cloud analysis on precipitation and temperature prediction. The results show： (1) The ARPS model for complex terrain has a very good adaptability. The comparison of ARPS prediction with local observation shows that ARPS performs very well both in start/stop time and spatial distribution of precipitation in two forecast process (24 h and 48 h); the precipitation of 24 h forecast corresponds with local observations，and the 48 h forecast tends to be bigger than local observations. (2) The ARPS model can also simulate temperature changes accurately, and the average absolute temperature error of interpolated 5 day forecast for six stations is only 2.69 ℃.（3） Cloud analysis can significantly improve the model initial field， especially the water substance, so that the bigger rainfall forecast at the strong precipitation center improved obviously, which indicates that the cloud analysis contributes to the success of precipitation forecasts, but the impact on temperature forecast is inconspicuous.