Integrating PM25 Observations, Model Estimates and Satellite Signals for the Eastern United States by Projection onto Latent Structures

Detailed, time-varying spatial fields of air contaminant concentrations are valuable to public health professionals seeking to identify relationships between human health and ambient air quality, and policy makers interested in assessing compliance with air quality regulations. In this paper PM25 fields are created from a linear model that predicts PM25 at unmonitored grid points from observed PM25 concentrations, CMAQ model outputs, and satellite estimates of aerosol optical density. The dimensionality of the input data set is first reduced using projection onto latent structures. Parameters of the linear model are mapped to the CMAQ model domain, permitting estimation of PM25 at unmonitored sites.