Spatial and temporal variability of black carbon in New York City



[1] Measurements of black carbon (BC) were made at two sites in New York City during the winter of 2004 and are compared. Spatial and temporal variability in the black carbon concentrations were explored by examining the diurnal, weekday/weekend, and day-of-week trends in black carbon concentrations at these two sites. It was observed that BC tracked PM2.5 well at the Queens site and contributed about 8% to the fine particle mass measured at this site. At the IS 52 site, however, BC did not track PM2.5 as well and accounted for a slightly higher 11% of the total fine mass. Diurnal BC behavior was observed during weekdays as well as weekends consistent with traffic patterns and atmospheric mixing conditions. However, a clear day-of-week dependence of BC concentrations was not observed. Conditional probability functions were computed using the surface wind data and the BC concentrations at the two sites and were used to identify local point sources. Associations of black carbon with gaseous pollutants such as CO, NOx, and nonmethane hydrocarbons are also explored for the Queens site in an attempt to identify the temporal variations in their emissions. It was found that the concentrations of black carbon vary in response to the interplay of source activity, meteorological conditions, and spatial factors. The high time resolution measurements were found to yield insights into the spatial scales of emissions that are not possible with longer timescale measurements such as 24 hour averages.

1. Introduction

[2] Carbonaceous aerosols are a ubiquitous component of ambient particulate matter (PM), both urban and rural, and are significant contributors to the fine aerosol burden of urban atmospheres [Gray et al., 1984]. Recent chemical speciation measurements of PM2.5 in New York City indicate that close to half the total mass is carbon based [Ito et al., 2004] and the measurements show little seasonal variations. Carbonaceous aerosol is a complex mixture, but may be broadly classified into two major components, organic carbon (OC) that is volatile and elemental carbon (EC) that is nonvolatile and strongly light absorbing. The mass related to light absorption is black carbon (BC). Even though BC and EC are often used interchangeably, and are well correlated, studies have shown that they reflect different characteristics of the particles. [Jeong et al., 2004].

[3] Current PM2.5 standards require greater emphasis on the control of primary source emissions and gas to particle conversions in the atmosphere. Developing effective strategies for the mitigation of PM necessitate determining the quantitative relationships between the sources and formation processes for the carbon component, as well as a better understanding of the atmospheric transformation, fate and transport of carbonaceous pollutant emissions. Aerosol BC is emitted by combustion processes involving carbonaceous material, is not strongly degraded under atmospheric conditions, exists mainly in the accumulation mode, and thus may be transported over long distances [Chylek et al., 1984; Hansen et al., 1984]. In urban settings, such as New York City, BC is usually associated with diesel and automotive emissions. Source tests have shown that the black carbon fraction of the total fine carbonaceous PM is higher in diesel exhaust than it is in gasoline exhaust [Hildemann et al., 1991; Watson et al., 1994; Lowenthal et al., 1994]. BC has been the subject of interest for atmospheric chemists and climatologists because of its strong capability in absorbing light, causing the extinction of solar radiation and consequent role in visibility and climate change. Direct radiative forcing by BC aerosols has been conjectured as a potential factor causing global warming [Jacobson, 2001], while the other components of fine PM, like sulfate aerosols, introduce a cooling effect.

[4] Recent attention has also been focused on BC as an important component of diesel exhaust particles (DEP), because of the potential role of diesel exhaust in the pathogenesis and exacerbation of asthma in urban settings [Pandya et al., 2002]. The association between air pollution and respiratory diseases such as asthma remain poorly understood. Since exposure is always to a mix of pollutants, it is difficult for epidemiological studies to define causal agents. In order to evaluate the hypothesis that DEP are causative agents, these species would need to be measured. In addition, spatial and temporal factors have to be considered to ensure that exposures can be gauged from ambient measurements using fixed site monitors and that sufficient temporal resolution and measurement frequency will allow the hypothesis to be tested. Spatial considerations are also important to explore health risks across diverse airsheds with different exposure microenvironments.

[5] Toward this end, a winter-intensive study was conducted in Flushing, New York in 2004 as part of the PM2.5 Technology Assessment and Characterization Study–New York (PMTACS-NY), and near real-time BC was measured. BC was continuously measured at another New York City site maintained by the New York State Department of Environmental Conservation (NYSDEC) during this period. Weekday-weekend effects, day-of-week and diurnal trends as well as relationships between particulate BC and gaseous copollutants such as CO, NOx, and nonmethane hydrocarbons (NMHCs) are explored.

2. Experimental Methods

[6] The study was conducted during the period of 12 January 2004 to 5 February 2004 at the New York Supersite at Queens College, about 14 km east of Manhattan. The Long Island Expressway (I-495) and the Van Wyck Expressway (I-678) are situated less than 2 km from the site. The companion site used in this study was the IS 52 site at 681 Kelly Street, in the South Bronx, maintained by the NYSDEC. The South Bronx has a high concentration of diesel trucks and waste transfer facilities. The Bruckner Expressway is less than a kilometer southeast of the site. The South Bronx also has large volumes of heavy vehicle traffic passing through it along several major highways (Interstates 87, 95, 278 and 895). The La Guardia Airport lies approximately midway between the two sites. The distance between the Queens Supersite and IS 52 was 9.1 km. Aethalometers (AE-20 and AE-21, Magee Scientific, Berkeley, California) was employed to acquire 5 min average BC data at these two sites. This method is based on the optical attenuation from BC in terms of the decrease in light transmission at 880 nm through a quartz fiber filter for a given volume of sampled air. The manufacturer's conversion factor of 16.6 m2 g–1 was used to calculate BC concentrations. The principle and working of the Aethalometer are described in detail elsewhere [Hansen et al., 1984]. Samples introduced into the Aethalometer were size segregated using a sharp-cut PM2.5 cyclone [Kenny and Gussman, 2000]. Since UV-BC data were available at the Queens site, the hourly data were screened for unusual local source influences such as wood smoke by examining the UV-BC data relative to the BC data [Jeong et al., 2004]. The meteorological measurements, and the gas phase measurements at the Queens site, were made by the NYSDEC monitors situated at the respective sites. The meteorological data from the NYSDEC monitors at the nearby Botanical Gardens site (located 5.4 km north-norheast of IS 52) were used for the IS 52 site, since the NYSDEC does not maintain monitors at the IS 52 site for this purpose. The hourly PM2.5 data used in this study were measured using a standard 50°C Tapered Element Oscillating Microbalance (TEOM, Thermo Electron, East Greenbush, NY, USA) at both the sites. The boundary layer mixing depths for the two sites were obtained from Stability Time series data available on the NOAA Air Resources Laboratory web server.

3. Results and Discussion

3.1. Data

[7] High time resolution measurements can yield insights into the spatial scales of pollutants that are not possible with the 24 hour averaged or even hourly averaged data. Black carbon is an especially useful observable for this purpose and can be used to estimate middle-, neighborhood-, and urban-scale contributions to a fixed monitor measurement [Watson and Chow, 2001]. Black carbon concentrations for the 25-day period at the two sites for sample durations ranging from 5 min to the entire 25-day period are summarized in Figure 1. Out of the possible 7200 5-min average values for the 25 days in this study, 6476 (90%) are valid for the Queens site and 7069 (98%) are valid for the IS 52 site. Hourly averaged BC concentrations in Queens varied from 0.23 μg/m3 to 5.87 μg/m3 with a mean of 1.01 μg/m3. BC concentrations in IS 52 ranged from 0.42 μg/m3 to 7.67 μg/m3 with a mean of 1.38 μg/m3. The BC concentrations were high enough to be measured with reasonable precision in greater than 95% of the valid hourly averages. Very few of the 5 min and hourly BC averages approached the maximum concentrations, as indicated by the distance between the corresponding 95th and 99th percentile values and the maximum values. The 6 hour and daily averages are better correlated with the maximum concentrations observed, which is to be expected since longer averaging times spread the highest concentrations. However, the disadvantage of using these longer averaging times is that they do not provide good estimates of the frequencies or magnitudes of the high concentrations to which people might be exposed. These statistics seem to indicate the need to initiate studies to assess the response of populations to shorter time resolution maximum concentrations, and to consider their efficacy as an independent exposure parameter in predicting acute health effects.

Figure 1.

BC Statistics for different averaging times at the Queens and IS 52 sites for 12 January 2004 through 5 February 2004.

[8] The regression of BC versus PM2.5 measured at the two sites, shown in Figure 2, indicates that BC tracked PM2.5 well at the Queens site and contributed about 13% to the fine particle mass measured at this site. At the IS 52 site, however, BC did not track PM2.5 as well, and accounted for slightly less than 11% of the total fine mass.

Figure 2.

Regression plots of hourly averaged BC versus PM2.5 at the (a) Queens and (b) IS 52 site for 12 January 2004 through 5 February 2004.

3.2. Site Similarity

[9] Any difference in measurements made at two sites has two components to it: the temporal correlation of the measurements, and the quantitative difference in concentration. Meteorological conditions such as wind speed and direction, mixing heights etc., the composition and relative magnitudes of the source emissions, and the variability in emissions on many different timescales would affect one or both of these components. From an epidemiological perspective, the temporal correlation is important for the strength of association, while the quantitative difference is important for the effect size. The temporal correlation and the coefficient of divergence (CD) [Wongphatarakul et al., 1998], were calculated for collocated BC and PM2.5 data from the two sites as a measure of the closeness or similarity between the two sites. The CD is defined as follows:

equation image

where xij represents the concentration for chemical component i at site j, j and k represent two sampling sites, and p is the total number of measured concentrations. The approach of the CD value toward zero or one indicates high similarity or no relevance between the two sites, respectively.

[10] The CD for PM2.5 and BC at the two sites were calculated to be 0.15 and 0.29, respectively, which seems to indicate some degree of spatial heterogeneity in the measured concentrations, although this is more prominent for BC than for PM2.5. Regression of simultaneously measured PM2.5 from the two sites yielded a correlation coefficient (r2) of 0.59 (moderate) and regression of simultaneously measured BC produced an r2 of 0.30 (weak) over the relatively short separation distance of 9 km between the two sites indicating heterogeneity in their temporal correlations as well. The weak to moderate correlations of BC and fine particle concentrations at the two sites suggest that these concentrations are less homogenous, even across short spatial scales of the order of a few kilometers than had been previously assumed. As a result, PM-effect analyses done on the basis of composite ambient time series for large spatial scales such as a city or on day-to day differences in calculated area wide 24-hour average concentration levels are likely to be misleading. Similarly, single point measurements of the ambient concentrations may lead to error in the estimation of the effects of emissions from an area like New York City to the global balance of atmospheric aerosol mass or black carbon concentrations.

3.3. Diurnal, Weekday/Weekend and Day-of-Week BC Effects

[11] The data from monitors in the same geographical area only a few kilometers apart can be substantially different. Variation on this spatial scale is generally presupposed to be driven by local mobile particle source emissions. One of the best indicators of local mobile source aerosol in urban areas is BC, associated with primary diesel and automotive emissions.

[12] Diurnal variations in BC concentrations along with the variations in meteorological conditions at the two sites for a typical weekday are shown in Figure 3. Air temperature and wind speed followed the general pattern of global irradiance with some delay as can be expected. Relative humidity was inversely related to the diurnal temperature with maxima in the early morning and minima around 3 PM. The measured diurnal variations in meteorological conditions during the sampling period are expected for this time of the year. As expected, a pronounced morning increase in BC concentrations is observed at both sites, peaking between 0700 and 0900 LT, consistent with traffic patterns. This increase can be attributed to local mobile sources combined with low mixing heights and consequent poor dispersion conditions at this time of day. The increase in BC concentrations in the morning is more evident at the Queens site while it is in the form of short-duration spikes at the IS 52 site. The winds during this time are observed to be mainly from the northwest, directly from the Van Wyck and Long Island Expressways for the Queens site, and from the junction of Interstates 87 and 95 for the IS 52 site. This peak is followed by a gradual decrease in BC concentrations at both sites through the afternoon. This might be explained in part by the growth of the mixing layer depth and more atmospheric ventilation during the afternoon. Such an increased mixed layer depth leads to better mixing of BC emitted at ground level with relatively clean air from above. In addition, the wind direction at both sites changed during this period from the northwest to a more westerly flow that might also lead to lower emissions from the aforementioned sources reaching the respective monitoring sites. An evening peak is observed at both sites consistent with the evening rush hour traffic, although as before with the morning rush hour, this rise is more evident at the Queens site with the IS 52 site seeing numerous spikes of short duration. The fact that the growth in mixing depth does not dilute this rush hour peak seems to indicate that middle-scale local sources, i.e., sources at 0.1–1 km from the receptor site [Watson and Chow, 2001], play a major role in the observed BC concentrations at this time of day, especially at the Queens monitoring site. BC concentrations from middle-scale sources would have insufficient transport time to be strongly affected by changes in mixing height. The wind speeds during this time are low, of the order of 4–5 mph that could have led to poor dilution and dispersion of pollutants emitted from ground-level sources throughout the immediate area. The BC concentrations drop off at night in Queens because of reduced traffic and removal of particles from the nocturnal mixing layer by dry deposition. At IS 52, however, the rise in observed BC concentrations continues beyond the evening rush hour and into the night, and this might be due to continuing use of the major interstate highways (I-95) by truck traffic during the low car traffic period of the night. Given that the South Bronx is downwind of Manhattan and New Jersey, both the cumulative effect of area sources as well as local traffic emissions might account for the observed rise in BC concentrations at night. These results strongly suggest that the dominant sources of BC were local mobile sources.

Figure 3.

Variation of 5-min measured BC concentrations at the Queens and IS 52 sites with various meteorological parameters.

[13] Weekday/weekend differences in ambient concentrations of pollutants have been a topic of research interest since the 1970s [e.g., Lebron, 1975; Elkus and Wilson, 1977]. Figure 4 shows the weekday/weekend variations in BC concentrations at the two sites. The weekday only plots for both Queens and IS 52 show similar morning and evening rises in BC concentrations with sharper rises evident for both sites with the weekend data removed. An early morning peak in BC concentrations is also seen between 0200 and 0300 LT. This rise may be due to the increase in heavy diesel trucks that use the nearby expressways during this time in order to avoid commuter traffic. A similar, but more pronounced early morning peak is seen at IS 52 during the weekend. However, a high standard deviation was observed and there are only a limited number of data points available for the weekends. Thus this peak may be due to a local event of a BC source passing close to the sampler. The same can be said for the observed, but much broader evening peak at this site, since trucking activities are usually less on weekends. From the weekend plot for Queens, it can be seen that there is a gradual increase in BC concentrations through the morning until late afternoon after which a gradual decrease is observed likely because of the decreased traffic volume aided by increased atmospheric ventilation.

Figure 4.

Weekday/weekend diurnal variation of BC concentrations at the Queens and IS 52 sites for the sampling period.

[14] The day-of-week trends shown in Figure 5 indicate no obvious trends in BC concentrations at both sites. Even though a decrease in BC is expected on the weekends because of less traffic, an increase is observed on Saturday at IS 52 and on Sunday at Queens. A decrease in BC is instead observed at both sites on Friday. The weekday trends show a midweek rise on Wednesday and Thursday. The lack of an observable day-of-week trend might be in part due to the fact that the data were obtained during a short winter intensive study. Long-term data would help establish a clearer day-of-week dependence of BC concentrations.

Figure 5.

Day-of-week trends in BC concentrations at the two sites for the sampling period.

3.4. Conditional Probability Function

[15] A conditional probability function (CPF) [Ashbaugh et al., 1985] was computed using the concentrations of BC coupled with wind speed and wind direction values measured at the two sites in order to identify likely locations of local point sources of BC to these two sites. The hourly wind data were matched to the 5-min high time resolution BC concentrations from the two sites. Calm winds (< 1 m/s) were excluded from this analysis. In our analysis, the CPF is defined as,

equation image

where, mΔθ is the number of occurrences from wind sector Δθ that exceeded the threshold criterion, and nΔθ is the total number of data points that fell within the same wind sector. In this study, Δθ was set at 15°. The sources are likely to be located in the directions that have high conditional probability values. The CPF plot in Figure 6 shows likely source locations when the threshold was set at the upper 25th percentile of the BC concentrations observed at each site.

Figure 6.

Five-minute resolution CPF plots for the upper 25% BC mass concentrations for the Queens and IS 52 sites.

[16] From the CPF plot for Queens, it appears that the black carbon contributions to the site are highest from the south-southeast, west and southwest corridor. From the diurnal plots, it was observed that these wind directions were associated with low average wind speeds and hence poor dispersion conditions that might have contributed in part to the high observed concentrations from these directions. The Van Wyck Expressway is upwind to the west and southwest of the Queens site, and could be the source contributing to the wider, more dispersed spike from this direction. Surprisingly, the Long Island Expressway that lies to the northwest of the Queens site in the path of the predominant wind direction during the period of the study is not seen to have an impact on the observed BC concentrations at the site. This result might be due to the fact that winds from the northwest had higher average wind speeds leading to better dilution and dispersion of pollutants from that direction, and hence minimizing its impact on the concentrations observed at the Queens site. A smoothed impact CPF plot looking at the upper 50th percentile concentrations (figure not included) reveals a dispersed regional signal from the direction of the Van Wyck, suggesting that it is a perceptible neighborhood-scale (∼1–5 km) source of BC as opposed to the middle-scale local source impacts seen in the diurnal plots. Since the nighttime BC concentrations at this site were low, contribution from regional sources to the site can be assumed to be minimal.

[17] The CPF plot for the IS 52 site indicates the major black carbon contributions to the site coming from the southwest, the east and southeast directions. The east and southeast source directions seem to pinpoint the Bruckner Expressway as the source contributing to the majority of these data occurrences in the upper 25th percentile. A peak impact CPF plot looking at the source direction of the upper 5th percentile of the observed BC concentrations (figure not included) indicates the high impact concentrations coming from the southeast, where the Bruckner Expressway is closest to the monitoring site. The signal from the southwest is a more dispersed spike pointing toward the I-87 corridor to the southwest of the site. Since higher nighttime concentrations are observed at the IS 52 site, regional-scale contributions from upwind Manhattan and Brooklyn which lie to the southwest and southeast of the IS 52 site, may have had a cumulative effect along with the local sources, to the observed source directions.

3.5. BC-Gaseous Pollutant Associations

[18] During the winter when this study was conducted, strong winds and active frontal passage strongly influenced the concentration variability at the sites with photochemical processes playing a minor role. Therefore atmospheric concentrations of gas phase pollutants like NOx, CO, and nonmethane hydrocarbons (NMHC) should be largely a function of source characteristics. According to the National Emissions Inventory (NEI) data for 1999, for the Bronx and Queens, mobile sources (on-road and off-road) contributed 73% of all NOx emissions. Mobile sources accounted for 90% of all CO emissions at these sites with gasoline vehicle exhaust contributing 96% of this total. Given that mobile sources are the dominant source of NOx and CO emissions at these sites, variations in BC should correlate with these gaseous pollutants.

[19] Figure 7 shows the associations between BC and its gaseous copollutants observed at the Queens site. The comparisons of the diurnal variation of mean NOx with BC, CO and NMHC show that NOx was more closely associated with BC than CO or NMHC. CO and NMHC are observed to have similar temporal variations but differed from those exhibited by BC and NOx. This plot shows that the early morning peak in BC is accompanied by a dip in CO and NMHC concentrations supporting the hypothesis that this peak was caused by diesel trucks operating at this early hour. The concentrations of NOx and BC exhibit similar diurnal variations, and the regression plot between BC and NOx (Figure 7b) confirms that they track each other reasonably well. BC and CO track each other weakly (Figure 7c). BC is usually most abundant in diesel exhaust, wood burning and cold start or high emitting gasoline vehicle emissions [Watson and Chow, 2002]. No wood burning episodes were observed, as evidenced by the UV-BC tracking the BC concentrations throughout the duration of the study, and hence wood burning does not appear to be a source of BC emissions. This result suggests that the sources of BC at this site were mainly diesel vehicles.

Figure 7.

(a) Diurnal covariations of mean BC, CO, NMHC and NOx concentrations; (b) regression plot of hourly averaged BC versus NOx; and (c) regression plot of BC versus CO at the Queens site for 12 January 2004 through 5 February 2004.

4. Conclusion

[20] The measurements of BC made at the Queens and IS 52 sites suggest that there is spatial and temporal variability over short spatial scales of the order of a few kilometers. BC contributed 13% and 11% to the total fine particle mass at these two sites respectively. The BC measured at these two sites was only weakly correlated, suggesting that the daily/diurnal fluctuations of the sources represented by BC cannot be well estimated for exposure assessment at larger spatial scales. Diurnal BC behavior was observed during weekdays as well as weekends consistent with traffic patterns and atmospheric mixing conditions. However, a clear day-of-week dependence of BC concentrations was not observed. Impacts from local point sources were clearly seen from the CPF plots for the two monitoring sites. The diurnal variations in BC as well as its gaseous copollutants indicate that local mobile sources are the dominant emission source of BC. BC and NOx were found to track each other well at the Queens site, suggesting the possibility of common emission sources. On the other hand, BC and CO tracked each other poorly. Since CO is mainly emitted by gasoline vehicle exhaust, the sources of BC seem to be mainly diesel in origin at this site. However, it cannot be concluded that BC is specific to diesel sources. The BC statistics obtained from these two sites suggest the need to study the importance of peak exposures on populations when predicting acute health effects. If the peak exposures are found to be important, a more dense monitoring network would be required to estimate exposure.


[21] This work was supported in part by the U.S. Environmental Protection Agency (EPA) through cooperative agreement R828060010 and Science to Achieve Results Program (STAR) grant subcontract from the University of Rochester PM and Health Center grant R827354. The acquisition of equipment used in this study was supported by the New York State Office of Science, Technology, and Academic Research (NYSTAR).