Abstract: The main purpose of this research is to grasp effectively the spatial and temporal forecast laws of the short-term quantitative precipitation forecast（QPF） products of the ECMWF model in Xinjiang,and to provide an effective basis for the objective model correction algorithm. In this study,technical methods such as frequency statistics,non-parametric testing methods,cumulative distribution functions,and equitable threat score were used to analyze the forecast performance of 12 ～36 h QPF products from September 2016 to December 2019 based on 151 meteorological observation stations in Xinjiang. The results showed that:QPF products of the ECMWF exists a big gap with SPO data,there are obvious false forecast when 3 h-QPF below 2.4 mm,and obvious false alarms when 24 h-QPF below11.0 mm. 3 h-QPF above 16.5 mm and 24 h-QPF above 52.5 mm have no forecasting ability. The medium level rainfall forecast of QPF has the best effect. The precipitation above the heavy rain in Xinjiang mainly occurs in the mountains. The 24 h-QPF product has more reference value than other regions for heavy precipitation forecasts in the middle of Tianshan Mountains and the Ili River Valley.The 3 h-QPF frequency of small,medium,and heavy rain for 8 times in a day has the opposite situation with SPO in time trend. QPF and SPO above heavy rain are a non-linear and uncertain correspondence.It is recommended to establish a false negative model first,and then use the frequency matching method would be more effective.