To gain detailed insight into the gaseous sources of PM2.5, three kinds of positive matrix factorization were developed and applied to hourly elemental and gaseous data (PMF-GAS). According to the results of PMF-GAS1, which relied on gaseous markers to enhance source discrimination, crustal dust, vehicular exhaust, coal combustion, sulfate and nitrate (CD1, VE1, CC1, SUL1 and NIT1) directly contributed 20%, 17%, 15%, 34% and 14% to the PM2.5, respectively. Using PMF-GAS2, which summed species found in PM2.5, and their corresponding precursors to estimate the overall influence of primary sources, the total contribution from CD2, VE2 and CC2 to PM2.5 was quantified as 30%, 29% and 41%, respectively. Finally, PMF-GAS3 quantified the contribution of oxidation (OXI3; 27%) based on the relationships between the species and their precursors. NH4+ was found mostly in CD2 and CD3 but also in SUL1 and NIT1, implying that crustal dust can provide alkaline conditions for the formation of ammonium salts. High correlations were found between corresponding source contributions, but CD1 showed relatively weak links with CD2 and CD3 due to NH4+ being present in different factors. A fully populated map of classical bootstrap (BS) runs implies the high stability of these PMF-GAS results.