According to the statistics, 1/2-2/3 precipitation occurs in June-August in Northwest China. Objective precipitation forecast is a difficult problem in NWP products interpretation at present. Precipitation in Northwest China is relatively rare event for single station. It is difficult to develop stable forecast relationships for single station. The stations are combined into regions with specific equations developed for each region. These regions are developed based on climatology and geographic similarity. These forecast equations are then applied to any station within a region. Regional forecast mode is more stable than single-station forecast mode, because precipitation sample size is increased.Seven weather divisions for summer precipitation over Northwest China are developed through Rotated Empirical Orthogonal Function method (REOF), defined by the large contours of the seven REOF models. Objective precipitation forecast is based on probability regression precipitation categorical forecast. First processing original precipitation to 0-1 corresponding categories, and then developing forecast equations of different categories and calculate each criterions. In real forecasting , the categorical precipitation will be determined through the criterion and the probability forecast of that category. Based on the daily precipitation data of  station and T639 NWP products During summer of 2007-2010 .Precipitation forecast modes over the Northwest were established. the present study concerns the area of  Northwest; however, more specifically, this covers the domain (35°-50°N, 73°-110°E), which includes a total of 260 weather stations .Precipitation during summer of 2011 was forecasted, Analysis of forecast result indicates that regional method is better than a station method, especially for moderate rain. From the factor analysis, regional model selected factor is more abundant than the single station modeling, so regional model make a better prediction. Corresponding to model direct forecast , regional forecast result is better, It has an obvious advantage on light rain forecast.