A detailed aerosol source apportionment study was performed with two sampling campaigns, during wintertime and summertime in the heavily polluted metropolitan area of Sao Paulo, Brazil. In addition to 12 h fine and coarse mode filter sampling, several real time aerosol and trace gas monitors were used. PM10 was sampled using stacked filter units that collects fine (d < 2.5 mum) and coarse (2.5 < d < 10 mum) particulate matter, providing mass, black carbon (BC) and elemental concentration for each aerosol mode. The concentration of about 20 elements was determined using the particle induce X-ray emission technique. Real time aerosol monitors provided PM10 aerosol mass (TEOM), organic and elemental carbon (Carbon Monitor 5400, R&P) and BC concentration (Aethalometer). A complex system of sources and meteorological conditions modulates the heavy air pollution of the urban area of Sao Paulo. The boundary layer height and the primary emissions by motor vehicles controls the strong pattern of diurnal cycles obtained for PM10, BC, CO, NOx, and SO2. Absolute principal factor analysis results showed a very similar source pattern between winter and summer field campaigns, despite the different locations of the sampling sites of both campaigns, pointing that there are no significant change in the main air pollution sources. The source identified as motor vehicle represented 28% and 24% of the PM2.5 for winter and summer, respectively. Resuspended soil dust accounted for 25% and 30%. The oil combustion source represented 18% and 21%. Sulfates accounts for 23% and 17% and finally industrial emissions contributed with 5% and 6% Of PM2.5, for winter and summer, respectively. The resuspended soil dust accounted for a large fraction (75-78%) of the coarse mode aerosol mass. Certainly automobile traffic and soil dust are the main air pollution sources in Sao Paulo. The sampling and analytical procedures applied in this study showed that it is possible to perform a quantitative aerosol source apportionment in a complex urban area such as Sao Paulo. (C) 2001 Elsevier Science Ltd. All rights reserved.