• No results found

Spatial Variability in NO 2 Concentrations

Legislative Requirements

NO 2 monitors; unshaded states have none.

2.5. Exposure Issues

2.5.3. Spatial Variability in NO 2 Concentrations

2.5.3.1. Variability of NO

2

Concentrations across Ambient Monitoring Sites

Summary statistics for the spatial variability in several urban areas across the U.S. are shown in Table 2.5-1. Data were obtained from EPA’s Air Quality System (AQS). These areas were chosen because they are major urban areas with at least five monitors operating from 2003 to 2005. Values in parentheses indicate the number of monitoring sites in that particular city. The second column shows the 3-year mean concentration across all monitors and the range in these means at individual monitor sites. Metrics for characterizing spatial variability include the use of Pearson correlation coefficients (r; column 3), the 90th percentile of the absolute difference in concentrations (column 4), and coefficient of

divergence (COD; column 5). The ranges represent results of pairwise monitor comparisons.

These three metrics are calculated based on measurements of daily avg concentrations at individual site pairs. The COD provides an indication of the variability across the monitoring sites in each city and is defined in Equation 2.5-6, as follows

(2.5-6) where Xij and Xik represent observed concentrations averaged over some measurement averaging period

(hourly, daily, etc.), for measurement period i at site j and site k, and p is the number of observations. A COD of 0 indicates there are no differences between concentrations at paired sites (spatial homogeneity), while a COD approaching 1 indicates extreme spatial heterogeneity.

Table 2.5-1. Spatial variability of NO2 in selected U.S. urban areas. URBAN AREA (# OF MONITORS) 3-YEAR MEAN CONCENTRATION (RANGE) PEARSON CORRELATION COEFFICIENT 90TH PERCENTILE DIFFERENCE BETWEEN MONITORS COEFFICIENT OF DIVERGENCE New York, NY (5) 29 ppb (25–37) 0.77–0.90 7–19 0.08–0.23 Atlanta, GA (5) 11 ppb (5–16) 0.22–0.89 7–24 0.15–0.59 Chicago, IL (7) 22 ppb (6–30) -0.05–0.83 10–39 0.13–0.66 Houston, TX (7) 13 ppb (7–18) 0.31–0.80 6–20 0.13–0.47 Los Angeles, CA (14) 25 ppb (14–33) 0.01–0.90 8–32 0.08–0.51 Riverside, CA (9) 21 ppb (5–32) 0.03–0.84 10–40 0.14–0.70

The same statistics shown in Table 2.5-1 have been used to describe the spatial variability of PM2.5 (U.S. Environmental Protection Agency, 2004; Pinto et al., 2004) and O3 (U.S. Environmental Protection Agency, 2006a).

Because of relative sparseness in data coverage for NO2, spatial variability in all cities considered for PM2.5 and O3 could not be considered here. Thus, the number of cities included here is much smaller than for either O3 (24 urban areas) or PM2.5 (27 urban areas). For urban areas with monitors for all three pollutants, data may have been collected at different locations, with different responses to local sources. For example, concentrations of NO2 collected near traffic will be highest in an urban area, but

concentrations of O3 will tend to be lowest there because of titration by NO forming NO2. However, some general observations can still be made. Mean concentrations of NO2 at individual monitoring sites are not as highly variable as for O3 but are more highly variable than PM2.5. Lower bounds on intersite correlation coefficients for PM2.5 and for O3 tend to be much higher than for NO2 in the same areas shown in Table 2.5-1. CODs for PM2.5 are much lower than for O3, whereas CODs for NO2 tend to be the largest among these three pollutants. The greater spatial variability for NO2 compared to O3 and PM2.5 could lead to larger exposure error in epidemiologic studies.

2.5.3.2. Small-Scale Horizontal Variability

Large gradients in NO2 concentrations near roadways have been observed in several studies, and NO2 concentrations have been found to be correlated (or inversely correlated) with distance from roadway, traffic volume, season, road length, open space, and population density (Bignal et al., 2007; Cape et al., 2004; Gauderman et al., 2005; Gilbert et al., 2007; Maruo et al., 2003; Monn et al., 1997; Pleijel et al., 2004; Roorda-Knape et al., 1998; 1999; Singer et al., 2004). A sample gradient is shown in Figure 2.5-2.

Singer et al. (2004) found a strong gradient for concentrations downwind of freeways within the first 230 m. An exponential decay model (e.g., Cape et al., 2004) has been fit to near-road concentration data to estimate NO2 concentration as a function of distance from the roadway. Gilbert et al. (2007) found that associations remained robust when sites within 200 m of roadways were removed from the analysis, indicating that traffic influences concentrations as far as 2000 to 3000 m from roadways. Small-scale spatial variations in NO2 concentrations are more pronounced during spring and summer seasons due to meteorology and increased photochemical activity (Monn, 2001).

Source: Singer et al. (2004). Figure 2.5-2. NO2 and NOX concentrations normalized to ambient values, plotted as a function of downwind

distance from the freeway. Symbols indicate freeway closest to each monitor.

Localized effects of roadway sources lead to variability in NO2 concentrations that is not captured by the regulatory monitoring network. This variation affects population-level exposure estimates and adds exposure error to time-series epidemiologic studies relying on ambient concentrations as indicators of exposure. Elevated concentrations near roadways also increase exposure of anyone residing, working, or attending school in the vicinity. As discussed in Chapter 4, these elevated concentrations found near roadways may lead to increased vulnerability among those exposed to high near-roadway concentrations of NO2.

2.5.3.3. Small-Scale Vertical Variability

Inlets to instruments for monitoring gas-phase criteria pollutants can be located from 3 to 15 m above ground level (CFR, 2002). Depending on the pollutant, there can be a positive, negative, or no vertical gradient from the surface to the monitor inlet. Positive gradients (i.e., concentrations increase with height) result when pollutants are formed over large areas by atmospheric photochemical reactions

pollutants emitted near the surface. Pollutants that are emitted by sources at or just above ground level show negative vertical gradients. Pollutants with area sources (widely dispersed surface sources) and that have minimal deposition velocities show little or no vertical gradient. Restrepo et al. (2004) compared data for criteria pollutants collected at fixed monitoring sites at 15 m above the surface on a school rooftop to those measured by a van whose inlet was 4 m above the surface at monitoring sites in the South Bronx during two sampling periods in November and December 2001. They found that CO, SO2, and NO2 showed negative vertical gradients, whereas O3 showed a positive vertical gradient and PM2.5 showed no significant vertical gradient. As shown in Figure 2.5-3, NO2 mixing ratios obtained at 4 m (mean ~74 ppb) were about a factor of 2.5 higher than at 15 m (mean ~30 ppb). Because tail pipe emissions occur at lower heights, NO2 values could have been much higher nearer to the surface and the underestimation of NO2 values by monitoring at 15 m even larger. Restrepo et al. (2004) noted that the use of the NO2 data obtained by the stationary monitors underestimates human exposures to NO2 in the South Bronx. This situation is not unique to the South Bronx and could arise in other large urban areas in the U.S. with similar settings. This adds another dimension to the exposure assessment, namely, the exposure of pedestrians who spend time walking in these street canyons, and urban residents, who have windows opening onto these canyons. These groups may experience high exposures to near-road concentrations of the same magnitude as exposures that occur on or adjacent to arterial and interstate roadways.

Source: Restrepo et al. (2004). Figure 2.5-3. NO2 concentrations measured at 4 m (Van) and at 15 m at NY Department of Environmental

The magnitude of the vertical gradient of NO2 in street canyons depends strongly on the

configuration of the buildings forming the canyons and the meteorological conditions; in particular, static stability in the lower planetary boundary layer, local wind direction and speed, and differential solar heating all affect turbulence in street canyons. These meteorological factors also help determine the relative importance of turbulence induced by traffic, in addition to traffic volume and speed. Detailed descriptions of the effects for many of these factors are available only from complex numerical models such as large eddy simulations and very fine grid resolution computational fluid dynamics (CFD) models. A semi-empirical integral model with simplifying assumptions has shown reasonable correlation to measured NO2 concentrations over moderate time scales (1 month) (Berkowicz et al., 2008), while other studies have applied such models to urban neighborhoods to estimate traffic emissions and evaluate the representativeness of air quality monitoring data (Ghenu et al., 2008; Mensink and Cosemans, 2008; Vardoulakis et al., 2005). By constructing simplified geometries, investigators can obtain good agreement between the performance of integral and CFD models; however, generalization and quantitative

application of these results to complex urban situations, even at the same location at different times, is difficult due to multi-scale variability in meteorological conditions, traffic composition and flow, building geometry, street dimensions, street canyon aspect ratios, and building packing density (Di Sabatino et al., 2007).

Weak associations might be found between concentrations at ambient monitors and other outdoor locations and between concentrations in indoor microenvironments and personal exposures in part because of the spatial (horizontal and vertical) variability in NO2. This variability is itself location- and time-dependent, and can lead to either over- or underestimates of exposure, depending on the siting of monitors and location of the exposed population. NO2 ambient monitors may be less representative of community or personal exposures than are ambient monitors for O3 or PM2.5 for their respective

exposures. This conclusion is based on a comparison of metrics of spatial variability for O3 or PM2.5 used in the last PM AQCD (U.S. Environmental Protection Agency, 2004) and O3 AQCD (U.S. Environmental Protection Agency, 2006a), indicating generally lower correlations and larger relative spreads in

concentrations than for O3 or PM2.5. As mentioned earlier, there are far fewer monitors for NO2 than for O3 or PM2.5, making estimation of the spatial variability in NO2 levels more difficult.