Source apportionment of PM2.5 at multiple Northwest U.S. sites: Assessing regional winter wood smoke impacts from residential wood combustion

Abstract

Wood smoke from residential wood combustion is a significant source of elevated PM2.5 in many communities across the Northwest U.S. Accurate representation of residential wood combustion in source-oriented regional scale air quality models is challenging because of multiple uncertainties. As an alternative to source-oriented source apportionment, this work provides, through receptor-oriented source apportionment, an assessment of winter residential wood combustion impacts at multiple Northwest U.S. locations. Source apportionment was performed on chemically speciated PM2.5 from 19 monitoring sites using the Positive Matrix Factorization (PMF) receptor model. Each site was modeled independently, but a common data preparation and modeling protocol was used so that results were as comparable as possible across sites. Model solutions had from 4 to 8 PMF factors, depending on the site. PMF factors at each site were associated with a source classification (e.g., primary wood smoke), a dominant chemical composition (e.g., ammonium nitrate), or were some mixture. 15 different sources or chemical compositions were identified as contributing to PM2.5 across the 19 sites. The 6 most common were; aged wood smoke and secondary organic carbon, motor vehicles, primary wood smoke, ammonium nitrate, ammonium sulfate, and fugitive dust. Wood smoke was identified at every site, with both aged and primary wood smoke identified at most sites. Wood smoke contributions to PM2.5 were averaged for the two winter months of December and January, the months when wood smoke in the Northwest U.S. is mainly from residential wood combustion. The total contribution of residential wood combustion, that from primary plus aged smoke, ranged from 11.4% to 92.7% of average December and January PM2.5 depending on the site, with the highest percent contributions occurring in smaller towns that have fewer expected sources of winter PM2.5. Receptor modeling at multiple sites, such as that conducted in this work, provided some significant advantages over modeling a single or small number of sites. Analysis at multiple sites allowed common factor chemical compositions to be identified, making it easier to evaluate when a PMF factor at a particular site represents a mix of sources versus a single source. The identification of similar PMF factors across multiple sites also allowed average chemical profiles to be established for the 6 the most commonly identified PM2.5 sources or compositions in this study. These average profiles have the potential to be used as source profile inputs in future Chemical Mass Balance receptor modeling, when a limited number of samples may restrict the ability to conduct PMF receptor modeling, or when the availability of local source profiles is limited. Receptor modeling results spanning a range of community sizes and source compositions, as in this study, could be used to evaluate and improve the representation of wood smoke and other specific sources in source-oriented regional scale air quality models by providing an independent source impact assessment.