In light of growing health concerns over short-duration exposure to pollution, this paper analyses fine particulates collected on a minute-by-minute basis inside a car. A time-series modelling approach is adapted to study the effects of various interventions (speed, traffic conditions, in-vehicle environment, time-of-day, etc.) while controlling for problems of autocorrelation. We also statistically detect and control for peak-exposure levels attributed to specific events such as following a smoky vehicle. Univariate time-series models in which only the previous PM2.5 levels are used explain over 75% of the variance. Multivariate time-series models show that vent position, air-conditioning status, time-of-day, en-route traffic conditions, and travel speed are all significant factors in the explanation of PM2.5 exposure levels. In addition, the multivariate model performs better than the univariate time-series model with lower unexplained variance and lower residual variation. (A) Reprinted with permission from Elsevier.
Abstract