Using spatio-temporal modeling to predict long-term exposure to black smoke at fine spatial and temporal scale

Whilst exposure to air pollution is linked to a wide range of adverse health outcomes, assessing levels of this exposure has remained a challenge. This study reports a modeling approach for the estimation of weekly levels of ambient black smoke (BS) at residential postcodes across Northeast England (2055 km2) over a 12 year period (1985–1996). A two-stage modeling strategy was developed using monitoring data on BS together with a range of covariates including data on traffic, population density, industrial activity, land cover (remote sensing), and meteorology. The first stage separates the temporal trend in BS for the region as a whole from within-region spatial variation and the second stage is a linear model which predicts BS levels at all locations in the region using spatially referenced covariate data as predictors and the regional predicted temporal trend as an offset. Traffic and land cover predictors were included in the final model, which predicted 70% of the spatio-temporal variation in BS across the study region over the study period. This modeling approach appears to provide a robust way of estimating exposure to BS at an inter-urban scale.

Research highlights
► Using remote sensing, GIS, and statistical modeling to predict air pollution levels.
► A two-stage modeling to predict spatiotemporal variation in black smoke (BS) levels.
► First stage is a temporal model and the second stage is a spatiotemporal model.
► Predicting 70% of spatiotemporal variation in BS in Northern England, 1985–96.
► This modeling approach is efficient at both inter-urban and intra-urban levels.