Polycyclic aromatic hydrocarbons (PAHs) and particulate matter (PM) are co-pollutants emitted as by-products of combustion processes. Convincing evidence exists for PAHs as a primary toxic component of fine PM (PM2.5). Because PM2.5 is listed by the US EPA as a "Criteria Pollutant," it is monitored regularly at sites nationwide. In contrast, very limited data is available on measured ambient air concentrations of PAHs. However, between 1999-2001, ambient air concentrations of PM2.5 and benzo(a)pyrene (BaP) are available for California locations. We use multivariate linear regression models (MLRMs) to predict ambient air levels of BaP in four air basins based on reported PM2.5 concentrations and spatial, temporal and meteorological variables as variates. We obtain an R2 ranging from 0.57-0.72 among these basins. Significant variables (p<0.05) include the average daily PM2.5 concentration, wind speed, temperature and relative humidity, and the coastal distance as well as season, and holiday or weekend. Combining the data from all sites and using only these variables to estimate ambient BaP levels, we obtain an R2 of 0.55. These R2-values, combined with analysis of the residual error and cross validation using the PRESS-statistic, demonstrate the potential of our method to estimate reported outdoor air PAH exposure levels in metropolitan regions. These MLRMs provide a first step towards relating outdoor ambient PM2.5 and PAH concentrations for epidemiological studies when PAH measurements are unavailable, or limited in spatial coverage, based on publicly available meteorological and PM2.5 data.