The identification of PM sources and their contribution to measured PM concentrations is crucial for the environmental policy making since the findings be able to conduce to the development of relevant legislation in order to achieve effective air quality manage. In this study, the sources of PM10 at three receptors with different characteristics, within the Western Macedonia (WM) in NW Greece, were investigated: S1 in the center of Kozani, a medium sized city located at the southern edge of the industrial axis of WM, where urban activities and traffic density occur. S2 in the city of Ptolemaida, a medium sized city located in the centre of industrial zone, and S3 in the village of Eratyra, a rural residential district outside of the industrial area. For this purpose the multivariate Positive Matrix Factorization (EPA PMF 5.0) receptor model was applied on elemental data. Specifically, PM10 samples, obtained by filtration during 1-year sampling campaign, were analyzed by ICP-MS instrument. Twenty five elements were detected at quantifiable concentrations in the investigated PM samples. For the particle samples obtained in areas within and proximal from mining and power station operations, a six-factor model gave source profiles that attributed to be vehicle exhaust, road dust, soil dust, coal combustion, oil combustion and biomass burning. Furthermore, at the background site, the major contributors were biomass burning, soil dust and oil burning while no distant transport from industrial axis was recorded.