Air pollution retention within a complex of urban street canyons: A two-city comparison

Abstract

Abstract Uncharacterized microscale spatial and temporal variability in urban air pollutant concentration dynamics may contribute uncertainty or bias to epidemiological model results. In this study, a method for quantifying this variability is presented. Urban buildings were treated as a matrix of bluff bodies to estimate the retention of air pollution in the street canyons downstream of the buildings. This method was based primarily on bluff body theoretical work that derived functional relationships between nondimensional contaminant residence time ( H ) within a wake and the following fluid properties of the air: 1) Reynolds Number (Re), 2) street canyon height ( D ) to width ( W ) aspect ratio ( D / W ), and 3) turbulence intensity, defined as the square root of turbulence kinetic energy ( k ) divided by the freestream wind speed ( U ). Empirical relationships between these variables were built from sulfur hexafluoride (SF 6 ) concentration and meteorological data collected during the Midtown Manhattan 2005 (MID05) Study held in August, 2005 in Manhattan, NY, along with geographical information system (GIS) data describing the building topography. Results were then compared with results from a similar previous analysis using data collected during the Joint Urban 2003 (JU2003) study in Oklahoma City, OK. For the MID05 data, Re ranged from 1.65 × 10 6 to 7.74 × 10 7 , with a median of 1.13 × 10 7 . The range of Re was consistent with earlier observations from the JU2003 study, although the measured winds tended to be more turbulent (median k = 2.2 m 2 s −2 ) compared with JU2003 (median k = 0.45 m 2 s −2 ). Values for H ranged from 7.2 to 1186, with a median H of 80.9. The distribution of H was substantially wider for MID05 than for JU2003, with model estimates exceeding observations of H by an order of magnitude for single obstacle wind tunnel studies with Re ∼ 10 4 . Inverse relationships were validated between H and Re and between H and D / W for the MID05 data and for a pooled data analysis from the MID05 and JU2003 studies. The model of H vs. Re for pooled MID05 and JU2003 data provided a good fit overall but produced a positively biased estimate of the Oklahoma City model results. The model of H vs. D / W for pooled MID05 and JU2003 data did not provide a good fit, suggesting that the building topographies of the two cities are too different to produce a reasonable comparison. These inter-study comparisons suggest that the topographic relationships may contain underlying site-specific features that would require elucidation prior to generalizing to other urban sites. Overall, results from this work present a foundational method for generating estimates of H based on readily available sources of data such as building coordinates and dimensions and meteorological parameters.