PM2.5 (Particles up to 2.5 µm in diameter) have a wide variety of health effects on human beings, and are found to be related to different kinds of traffic and land use factors. The land use regression (LUR) model has been used in many studies to model air pollution concentration across the world. However, the LUR model does not incorporate the temporal variation, so it can hardly be used in epidemiological studies.
To undertake analysis at a finer spatial and temporal resolution, this paper explores the monthly, daily and hourly PM concentration in Ohio and Franklin County. At the large scale of the Ohio state, PM concentrations within different seasons, months and days are explored and predicted. And at the smaller scale of Franklin County, PM concentrations at 24 hours of 10 days in 2008 are predicted and used to estimate the journey-time exposure when people are moving around Franklin County. To incorporate both the time-varying and time-invariant factors in one regression model, the Generalized Additive Model (GAM), of which the relationship between response and independent variables is cubic spline rather than linear, is used to estimate the short-term variations in air pollution. In the paper, the time-varying factors include meteorological and traffic variables while the time-invariant factors include different land use type, transportation, population density, major roads and highway information.
This paper is the first study that undertakes spatially complex and temporally dynamic time-space modeling in Ohio.