Quantifying PM2.5 source contributions for the San Joaquin Valley with multivariate receptor models

Abstract

UNMIX and Positive Matrix Factorization (PMF) solutions to the Chemical Mass Balance (CMB) equations were applied to chemically speciated PM2.5 measurements from 23 sites in California's San Joaquin Valley to estimate source contributions. Six and seven factors were determined by UNMIX for the low_PM2.5 period (February to October) and high_PM2.5 period (November to January), respectively. PMF resolved eightfactors for each period that corresponded with the UNMIX factors in chemical profiles and time series. These factors are attributed to marine sea salt, fugitive dust, agriculture-dairy, cooking, secondary aerosol, motor vehicle, and residential wood combustion (RWC) emissions, with secondary aerosol and RWC accounting for over 70% of PM2.5 mass during the high_PM2.5 period. A zinc factor was only resolved by PMF. The contribution from motor vehicles was between 10 and 25% with higher percentages occurring in summer. The PMF model was further evaluated by examining (1) site-specific residuals between the measured and calculated concentrations, (2) comparability of motor vehicle and RWC factors against source profiles obtained from recent emission tests, (3) edges in bi-plots of key indicator species, and (4) spatiotemporal variations of the factors' strengths. These evaluations support the compliance with model assumptions and give a higher confidence level to source apportionment results for the high_PM2.5 period.