In this study, 1376 samples for 12 water quality parameters were obtained from 50 monitoring sites across Beijing City's drinking water supply watershed from 2004 to 2010. Multivariate statistical techniques and a Geographic Information System (GIS) were used to explore the spatial - temporal variations of water quality and investigate the primary pollution sources. Moran's I analysis indicated that spatial autocorrelation had little impact on the explanation of spatial variation of water quality. Kruskal-Wallis test and coefficients of variation showed that total nitrogen, nitrate, nitrite, and total phosphorus had significant temporal and spatial differences. The factor analysis indicated that the pollution degree was mainly attributed to agricultural activities, rural domestic sewage, and mining tailings runoff. Before and after rainy season, river conditions were influenced by mixed source pollution (agricultural overland flow, mining overland flow, and rural domestic sewage). During rainy season, non-point source pollution (agricultural runoff) was the only dominate cause of water pollution. Although the importance of parameters to river conditions changed along with seasons, total nitrogen and nitrate always had the greatest impact on water quality variation and furthermore their concentrations exceeded lowest surface water quality standard. Thus, suggestions regarding design and allocation of Best Management Practices (BMPs) for reducing nitrogen pollutant yields from the drinking water supply watershed of Beijing are presented.