In this paper, principal component analysis (PCA) and positive matrix factorization (PMF) multivariate statistical techniques were applied to evaluate the spatiotemporal variation in 13 conventional water quality parameters (TOC, TN, NO2-, NO3-, TP, SO4-2, Cl-, TSS, color, pH, temperature, DO, EC) at eight monitoring stations for a duration of a year (2016-2017) in the upstream parts of the Ergene basin (NW Turkey). The eight monitoring stations were divided into two groups (five sites for Gr-A and three sites for Gr-B) considering pollution levels of the parameters and point/non-point sources determined by field observations. The principal component analysis defined four and three latent factors explaining 87% and 89% of the total variance in Gr-A and Gr-B datasets, respectively. Component numbers defined in PCA were manually assigned to the positive matrix factorization model. PCA was seen to be an important index for defining the number of factors causing high uncertainty for PMF. The factors derived from the PMF model revealed that the dominant pollutant sources for Gr-A sites are textile and leather industry discharges, agricultural activities, domestic discharges and seasonal factors. Gr-B sites are defined as domestic discharges, agricultural fertilizers and industrial discharges. Therefore, PMF analysis for conventional water quality parameters is a consistent statistical technique for the identification of complex pollution sources.