Abstract: To examine effects of radar data assimilation and increasing horizontal re solution on the shortrange numerical weather prediction of precipitation, comp a rative numerical experiments are conducted for a Huabei torrential rainfall even t using the Advanced Regional Prediction System (ARPS) and its Data Analysis Sys tem (ADAS). The experiments differ in five different horizontal gridspacings o f 18, 15, 9, 6 and 3 km and two initial fields with/without initial radar data as similation. Results show that, when the radar data are not used in the initial f ields (i.e., using the conventional observation data only), increasing horizonta l resolution can improve the shortrange prediction within six hours with bette r representation of the frontal structure and higher scores of rainfall predictio n, especially heavy rain prediction. The initial radar data assimilation is abl e to significantly improve the precipitation location and intensity prediction w ithin the first six hours such that the 6 h precipitation prediction of the ARPS at 18 km resolution with initial radar data assimilation is better than that of the APRS at 3 km resolution without initial radar data assimilation. This sugge sts that the radar data assimilation is more effective in improving the shortrange precipitation prediction than increasing the horizontal resolution al one. Un der the condition of initial radar data assimilation, experimental results of di fferent resolutions show that differences of prediction results among among diff erent resolution schemes are not evident for ordinary precipitation (≤40 mm/( 6 h)) ,but prediction of higher resolution experiment scheme for extraordinary precip itation (>40 mm/(6 h))is better than other schemes. However, when the resolution increases from 6 km to 3 km, there is no evident improvement in precipita tion pr ediction whether the initial radar data assimilation is performed or not. Therefore, the effect of increasing resolution is limited, and combining an adequate h igher resolution with the initial radar data assimilation might be an eff ective way to improve the shortrange prediction of precipitation.