Ambient particulates of PM 2.5 were sampled at three sites in Kaohsiung, Taiwan, during February and March 1999. In addition, resuspended PM 2.5 collected from traffic tunnels, paved roads, fly ash of a municipal solid waste (MSW) incinerator, and seawater was obtained. All the samples were analyzed for twenty constituents, including watersoluble ions, organic carbon (OC), elemental carbon (EC), and metallic elements. In conjunction with local source profiles and the source profiles in the model library SPECIATE EPA, the receptor model based on chemical mass balance (CMB) was then applied to determine the source contributions to ambient PM 2.5 .The mean concentration of ambient PM 2.5 was 42.69-53.68 µg/m 3 for the sampling period. The abundant species in ambient PM 2.5 in the mass fraction for three sites were OC (12.7-14.2%), SO 4 2-(12.8-15.1%), NO 3 -(8.1-10.3%), NH 4 + (6.7-7.5%), and EC (5.3-8.5%). Results of CMB modeling show that major pollution sources for IMPLICATIONS It is of great importance to identify air pollution sources in the development of air quality control strategies. For about a decade in Taiwan, more efforts have been put into regulation and control of stationary sources. In order to improve the air quality in a serious and cost-effective manner, other sources, such as vehicle emissions and photochemical aerosols, must be reviewed and quantified. The CMB receptor model was used to assess the potential source contributions to ambient PM 2.5 , which is believed to be responsible for visibility deterioration in Kaohsiung. The measurements in this study show that the abundant species in ambient PM 2.5 are OC, secondary aerosols (containing SO 4 2-, NO 3 -, and NH 4 + ), and EC. Assessment with the CMB model suggests that mobile sources should be regulated and controlled more thoroughly in this area than they were previously. ambient PM 2.5 are traffic exhaust (18-54%), secondary aerosols (30-41% from SO 4 2-and NO 3 -