Multivariate statistical approach was used to identify the source apportionment of 25 semi-volatile compounds (SVOCs) in surface water in Nanjing reach of Yangtze River, China. Surface water samples were collected from six sites in January 2007. Concentrations of SVOCs were determined by gas chromatography equipped with a mass spectrometry detector (CC-MS). Cluster analysis (CA) and principal component analysis-multiple linear regression (PCA-MLR) were used for evaluation of contamination status, distribution and source apportionment. The results of CA showed that the six sites in this study could be divided into two significant groups, i.e. low and high pollutant groups. PCA-MLR results suggested that four source types were identified, explaining about 84.5% of the total variance. Of four sources, industrial and domestic sewage sources contributed 43.5-64.8% of various SVOCs in six sites, precipitance and runoff sources 4.22-42.87%, shipping sources 4.22-23.49%, and other miscellaneous sources 1.05-8.97%. These results indicated that point source pollutants affected the water quality of this region. High correlation coefficient (R-2 = 0.973) between the measured and predicted concentrations of SVOCs suggested the applicability of the PCA-MLR for estimation of source contribution to the SVOCs in water. (C) 2009 Elsevier B.V. All rights reserved.