A bayesian multivariate receptor model for estimating source contributions to particulate matter pollution using national databases

Amber J. Hackstadt, Roger D. Peng

Research output: Contribution to journalArticlepeer-review

Abstract

Time series studies have suggested that air pollution can negatively impact health. These studies have typically focused on the total mass of fine particulate matter air pollution or the individual chemical constituents that contribute to it, and not source-specific contributions to air pollution. Source-specific contribution estimates are useful from a regulatory standpoint by allowing regulators to focus limited resources on reducing emissions from sources that are major contributors to air pollution and are also desired when estimating source-specific health effects. However, researchers often lack direct observations of the emissions at the source level. We propose a Bayesian multivariate receptor model to infer information about source contributions from ambient air pollution measurements. The proposed model incorporates information from national databases containing data on both the composition of source emissions and the amount of emissions from known sources of air pollution. The proposed model is used to perform source apportionment analyses for two distinct locations in the U.S.A. (Boston, Massachusetts and Phoenix, Arizona). Our results mirror previous source apportionment analyses that did not utilize the information from national databases and provide additional information about uncertainty that is relevant to the estimation of health effects.

Original languageEnglish (US)
Pages (from-to)513-527
Number of pages15
JournalEnvironmetrics
Volume25
Issue number7
DOIs
StatePublished - Nov 1 2014

Keywords

  • Air pollution
  • Chemical speciation network
  • Multivariate receptor model
  • National emissions inventory
  • Source apportionment
  • Speciate

ASJC Scopus subject areas

  • Statistics and Probability
  • Ecological Modeling

Fingerprint Dive into the research topics of 'A bayesian multivariate receptor model for estimating source contributions to particulate matter pollution using national databases'. Together they form a unique fingerprint.

Cite this