Spatial Modeling of Diesel Exhaust Markers in South Seattle

Abstract

Background: South Park and Georgetown, two of Seattle's most diverse and affordable neighborhoods, contain the primary commercial traffic corridors from the Port of Seattle to interstates and state highways. Residents of these communities have expressed concern about exposure to diesel exhaust emitted by the large number of commercial trucks that pass through their neighborhoods. The aim of this project was to model the spatial distribution of diesel exhaust markers at a fine scale across these neighborhoods using measurements from a high-density air sampling campaign. Methods: Two-week average concentrations of two markers of diesel exhaust, 1-nitropyrene (1-NP) and light-absorbing carbon (LAC), were measured in summer and winter at 24 sites. Land-use regression models were built using spatial characteristics of sampling sites, including land use and road density. Mobile source emissions predictions from the CAL3QHCR dispersion model were included in spatial models. Light-scattering particle concentrations measured by a mobile monitoring platform that drove through the neighborhoods were also included as model covariates. Model predictions were generated using land-use regression equations for a grid of points 50m apart across the study area. Universal kriging was applied to these grid points to generate a raster surface of the gradient of predictions. Results: 1-NP concentrations ranged from 0.263 pg/m3 to 2.51 pg/m3 in summer and 1.11 pg/m3 to 5.71 pg/m3 in winter. LAC concentrations, measured as the absorption coefficient of collected fine particles, ranged from 4.31E-06 m-1 to 7.84E-06 m-1 in summer and 6.30E-06 m-1 to 9.42E-06 m-1 in winter. The summer 1-NP model had an R2 of 0.87 and a leave-one-out cross-validated R2 of 0.73. No prediction model of winter 1-NP was identified. The LAC models had R2 values of 0.78 and 0.79 and leave-one-out-cross-validated R2 values of 0.66 and 0.70 for August and December, respectively. Conclusions: Spatial modeling was successfully used to identify a clear gradient in concentrations of diesel exhaust markers at a fine scale within the neighborhoods of South Park and Georgetown. Spatial features that predicted diesel exhaust concentrations included dispersion model predictions, mobile monitoring results, land use, and distance to railroad tracks, roads and intersections. The existence of this gradient suggests that particularly in stagnant periods, the health and environmental impacts of diesel traffic are not evenly distributed across these neighborhoods.