In order to develop the fine particular matter (PM2.5) air quality forecast capacity, the air quality forecast system, based on linking NOAA's North American Mesoscale (NAM) meteorological model with EPA's Community Multiscale Air Quality (CMAQ) model, was deployed in developmental mode over the continental United States during 2007. The ability of bias correction technique – Kalman Filter Predictor approach – in improving the accuracy of the PM2.5 forecasts at discrete monitoring locations is investigated through their application to archived real-time surface level PM2.5 forecasts from the NAM-CMAQ air quality forecasting modeling system and archived near real-time PM2.5 observations obtained from EPA's AIRNow measurement network. The Kalman Filter Predictor bias correction is a recursive algorithm to optimally estimate bias correction term from previous measurements and forecasts. Results show that the approach can significantly reduce forecast errors compared with original model forecasts. Detailed performance evaluation for both the raw model and the bias corrected forecasts will be presented based on seasons and also geological regions.