An air pollution prediction model system (APOPS) is developed and applied to northern Taiwan with complex terrain and local thermal circulations. It consists of a nonhydrostatic mesoscale meteoro-logical model system (MMPMS) and a gas/aerosol transport and air quality model (GATAM). The basic processes relevant to modeling the urban air pollution problems such as meteorology, dispersion, chemistry and deposition are solved at the same time on practically the same grid. The APOPS was tested on a high-ozone event in northern Taiwan on 16 November 1998. Comparison with observed surface winds shows able to predict local flow patterns such as sea/land breezes and mountain-valley wind in this high air pollution episode. The predicted surface concentrations of ozone and other pollutants are compared with measured values, and a fairly good agreement with the mean normalized biases of −6%, −11%, for one-day simulation and for daytime, respectively, is obtained for ozone. Thus, it is confirmed that the APOPS can be utilized to predict urban air quality in complex terrain area.