Statistical analyses of the health effects of air pollution have increasingly used GIS-based covariates for prediction of ambient air quality in “land-use” regression models. More recently these regression models have accounted for spatial correlation structure in combining monitoring data with land-use covariates. The current paper builds on these concepts to address spatio-temporal prediction of ambient concentrations of particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5) on the basis of a model representing spatially varying seasonal trends and spatial correlation structures. Our hierarchical methodology provides a pragmatic approach that fully exploits regulatory and other supplemental monitoring data which jointly define a complex spatio-temporal monitoring design. We explain the elements of the computational approach, including estimation of smoothed empirical orthogonal functions (SEOFs) as basis functions for temporal trend, spatial (“land use”) regression by Partial Least Squares (PLS), modeling of spatio-temporal correlation structure, and generalized universal kriging prediction of ambient exposure for subjects in the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air) project. Analyses are demonstrated in detail for the South California study area of the MESA Air project using AQS monitoring data from 2000 to 2006 and supplemental MESA Air monitoring data beginning in 2005. Results of application of the modeling and estimation methodology are presented also for five other MESA Air metropolitan study areas across the country with comments on current and future research developments.


Statistical Methodology | Statistical Theory

Previous Versions

Nov 30 2009