Boundary layer parameters, such as daytime mixed layer depth and temperature, nocturnal inversion strength, and low level wind speed and direction, play a substantial role in air quality forecasting. The accuracy of these boundary layer parameters predicted by models depend, to a large degree, on the type of surface and boundary layer turbulence parameterizations used by models. This study examines the sensitivity of predicted ozone air quality to several different approaches to parameterize surface and boundary layer turbulence. An ozone episode during the Central California Ozone Study (CCOS) conducted in the Central Valley of California during the summer of 2000 was selected for the study. The predicted meteorological fields using the MM5 model with four different turbulence parameterization schemes are used to drive air quality forecasting using US EPA Models-3, and the resulted ozone forecasts were compared to near surface and aircraft observations in the Central Valley. Relatively large differences were found among different schemes in predicted surface ozone concentrations. Generally higher near surface ozone concentrations are found to be associated with TKE based schemes when simulated afternoon mixed layer depths are significantly lower. The near surface ozone distribution patterns, however, are similar across the different schemes because of the relatively small differences found in predicted flow patterns.