|
Abstract:
|
For development of effective air pollution control strategies, it is crucial to identify major sources and estimate how much each source contributes to pollution. Multivariate receptor modeling aims to address these questions by decomposing ambient concentrations of multiple air pollutants into components associated with sources. With the EPA Speciation Trends Network, extensive multivariate air pollution data obtained from multiple monitoring sites are becoming available. Although considerable research has been conducted on modeling the multivariate space-time data in other contexts, there has been little research on multivariate receptor modeling for the multi-site multi-pollutant data. We present a Bayesian spatial multivariate receptor modeling approach that can incorporate spatial correlations in the multi-site multi-pollutant data into estimation of source profiles and contributions, while simultaneously dealing with model uncertainty. The proposed method can also provide the uncertainty estimates of predicted source contributions at any location. The method is applied to ambient concentrations of 17 VOCs measured at 9 monitoring sites in Harris County, TX, during 2000-2005.
|
