Sources affecting surface sediment samples at the western Bay of lzmit were investigated using a relatively new type of factor analysis: positive matrix factorisation (PMF). PMF uses estimates of the error in data to provide optimum data point scaling and applies the non-negativity constraints to the factors. The PMF results presented in this study are compared with the previously published factor analysis (FA) results to determine whether similar sources can be apportioned. A multi-element data set, generated by analysing sediment samples by inductively coupled plasma-atomic emission spectrometry (ICP-AES), was used in the comparison. To identify the sewage factor with the FA, dummy variables were also added to data set. Three, four and five sources were resolved with the FA without dummy variables (FAW), FA with dummy variables (FAD) and with the PMF, respectively. Among the resolved factors, three sources were common in all approaches: the iron and steel industry, the paint industry and crustal sources. The sewage factor was identified with the FAD and PMF. The PMF also resolved the motor vehicle factor. In this particular study. PMF allowed us to separate the anthropogenic factor obtained in the two factors, one representing the discharges from industries and the other representing discharges from sewage. The PMF method resolved 97% of the sediment mass concentrations and showed a significantly higher source resolution than the previously derived FA model. (C) 2012 Elsevier B.V. All rights reserved.