Modeling spatiotemporal variability of intra-urban air pollutants in Detroit: A pragmatic approach

This study combined a three-year time series of air pollutant measurements from the Michigan Air Sampling Network (MASN) with spatially detailed datasets for two two-week periods in September 2008 and June 2009. The objective was to produce monthly pollutant concentration models for the city of Detroit, Michigan, USA from January 2008 through December 2010, in support of a related epidemiological study examining adverse birth outcomes in Detroit. Two gaseous analytes, NO2 (nitrogen dioxide) and total BTEX (benzene, toluene, ethyl-benzene, and xylene), as well as two particulate matter size fractions, PM2.5 and PM10, were investigated. The September 2008 and June 2009 datasets were modeled using ordinary kriging to produce high spatial density concentration maps with 300 m by 300 m resolution across the city. A weighted average was applied to these maps to generate a series of monthly spatial models for each pollutant. Temporal variability was then incorporated by adjusting each monthly spatial model using an average bulk shift derived from MASN time series measurements for the corresponding month over the three-year study period. sulting models incorporate temporal trends while preserving neighborhood scale spatial variability. Seasonal variation was evident in NO2 models, but not readily discernable in BTEX or PM models across the three year study period. The greatest spatial and temporal variability was observed in the BTEX distributions, which are inferred to be strongly influenced by local sources. The methodology employed assumes that the interpolated monthly models adequately capture spatial variability of the air pollutants across the study area, the spatial distribution of pollutant concentrations remained consistent while their magnitude fluctuated from month to month, and that the available time series measurements reflect temporal trends across the city of Detroit throughout the three-year study period.