A total of 190 grab water samples were collected from 19 rivers along the water conveyance system of the Middle Route of China's interbasin South to North Water Transfer Project (MRSNWTP). Multivariate statistics including principal component/factor analysis (PCA/FA), analysis of variance (ANOVA), and cluster analysis (CA) were employed to assess water quality, and the receptor model of factor analysis-multiple linear regression (FA-MLR) was used for source identification/apportionment of pollutants from natural processes and anthropogenic activities to river waters. Our results revealed that river waters were primarily polluted by CODMn, BOD, NH4+-N, TN, TP, and Cd with remarkably spatio-temporal variability, and there were increasing industrial effluents in rivers northward. FA/PCA identified four classes of water quality parameters, i.e., mineral composition, toxic metals, nutrients, and organic pollutants. CA classified the selective 19 rivers into three groups reflecting their varying water pollution levels of moderated pollution, high pollution, and very high pollution. The FA-MLR receptor modeling revealed predominantly anthropogenic inputs to river solutes in Beijing and Tianjin, i.e., 77% of nitrogen and 90% of phosphorus from industry, and 80% of CODMn, from domestics. This study is critical for water allocation and division in the water-receiving areas using the existing rivers for MRSNWTP. (C) 2011 Elsevier B.V. All rights reserved.