A dynamic process convolution approach to modeling ambient particulate matter concentrations

Abstract

Elevated levels of particulate matter (PM) in the ambient air have been shown to be associated with certain adverse human health effects. As a result, monitoring networks that track PM levels have been established across the United States. Some of the older monitors measure PM less than 10 µm in diameter (PM₁₀), while the newer monitors track PM levels less than 2.5 µm in diameter (PM_2.5); it is now believed that this fine component of PM is more likely to be related to the negative health effects associated with PM. We propose a bivariate dynamic process convolution model for PM_2.5 and PM₁₀ concentrations. Our aim is to extract information about PM_2.5 from PM₁₀ monitor readings using a latent variable approach and to provide better space-time interpolations of PM_2.5 concentrations compared to interpolations made using only PM_2.5 monitoring information. We illustrate the approach using PM_2.5 and PM₁₀ readings taken across the state of Ohio in 2000. Copyright © 2007 John Wiley & Sons, Ltd.