Mohammad Omary

This work describes a methodology for modeling the impact of traffic-generated air pollutants in an urban area. This methodology presented here utilizes road network geometry, traffic volume, temporal allocation factors, fleet mixes, and emission factors to provide critical modeling inputs. These inputs, assembled from a variety of sources, are combined with meteorological inputs to generate link-based emissions for use in dispersion modeling to estimate pollutant concentration levels due to traffic. A case study implementing this methodology for a large health study is presented, including a sensitivity analysis of the modeling results reinforcing the importance of model inputs and identify those having greater relative impact, such as fleet mix. In addition, an example use of local measurements of fleet activity to supplement model inputs is described, and its impacts to the model outputs are discussed. We conclude that with detailed model inputs supported by local traffic measurements and meteorology, it is possible to capture the spatial and temporal patterns needed to accurately estimate exposure from traffic-related pollutants.

Exposure studies rely on detailed characterization of air quality, either from sparsely located routine ambient monitors or from central monitoring sites that may lack spatial representativeness. Alternatively, some studies use models of various complexities to characterize local-scale air quality, but often with poor representation of background concentrations. A hybrid approach that addresses this drawback combines a regional-scale model to provide background concentrations and a local-scale model to assess impacts of local sources. However, this approach may double-count sources in the study regions. To address these limitations, we carefully define the background concentration as the concentration that would be measured if local sources were not present, and to estimate these background concentrations we developed a novel technique that combines space-time ordinary kriging (STOK) of observations with outputs from a detailed chemistry-transport model with local sources zeroed out. We applied this technique to support an exposure study in Detroit, Michigan, for several pollutants (including NO_x and PM_2.5), and evaluated the estimated hybrid concentrations (calculated by combining the background estimates that addresses this issue of double counting with local-scale dispersion model estimates) using observations. Our results demonstrate the strength of this approach specifically by eliminating the problem of double-counting reported in previous hybrid modeling approaches leading to improved estimates of background concentrations, and further highlight the relative importance of NO_x vs. PM_2.5 in their relative contributions to total concentrations. While a key limitation of this approach is the requirement for another detailed model simulation to avoid double-counting, STOK improves the overall characterization of background concentrations at very fine spatial scales.