Elemental and organic carbon exposure in highway tollbooths: A study of Taiwanese toll station workers

ABSTRACT

The carbon composition of fine particles (PM2.5) from traffic exhausts may play a role in adverse health effects. The objective of this study was to assess the concentrations of elemental and organic carbon in PM2.5 in traffic exhausts from different types of vehicles in the booths of Taiwanese toll station workers and estimate the relations between traffic density and carbon concentrations.  Tollbooth indoor monitoring samples were collected for 10 days to assess the 8 h integrated PM2.5 concentration. Particle samples were analyzed for the content of total carbon, and elemental, and organic carbon. The mean carbon concentrations in the bus and truck lanes were [total: 167.7 μg/m3 (SD 79.8 μg/m3); elemental: 131.7 (66.2); organic: 36.0 (25.8)], substantially higher compared with the car lanes with cash payment [39.2 (29.5); 20.2 (19.5); 19.2 (14.6)] and the car lanes with ticket payment [34.1 (26.1); 15.8 (17.6); 18.5 (12.2)].  The increase in elemental carbon concentration per vehicle in the bus and truck lane was 14 and 9 times greater than that of car lanes of ticket payment and car lanes of cash payment. The mass fraction of carbonaceous species in PM2.5 accounted for 54% in bus and truck lanes, whereas the corresponding figure was 30–31% for car lanes. Elemental carbon is an important component of diesel exhaust. Workers in toll stations are exposed to high levels of both elemental and organic carbon.

METHODS

Study location

The selected toll station was located on the First Highway, 10 km south from Taipei City. According to the Bureau of Highway Records, this toll station has the highest traffic density among all toll stations in Taiwan. There are 20 tollbooths, 10 used to collect the toll from traffic from Taipei City to Taipei County (from north to south), and the remaining 10 booths collect toll from the opposite traffic flow. Of the 20 booths, 2 to 3 in each direction were designed to accommodate bus and truck traffic, and the other 7 to 8 were for passenger cars. In addition, some of the toll lanes were set up to accommodate drivers who use prepaid tickets and other lanes accommodated those who pay with cash. All tollbooths are of similar size (L ×W×H= 1.5 m ×1.0 m× 2.1 m) and have a door (W×H=75 cm× 190 cm) opening towards the vehicle lane, and one window installed on the opposite side (L ×W= 65 cm× 30 cm). During the field sampling period the only window was closed and no ventilation was used in any of the tollbooths. Tollbooth workers were leaning out during toll collection. The employer provided a cotton mask (Chen-Hua, Taipei, Taiwan) to all toll station workers. According to an interview, 38 (80.9%) always wore mask on-duty, 4 (8.9%) sometimes, and 1 (2.1%) never.

Sampling strategy

Exposure assessment was conducted by a personal PM2.5 monitor in the booth with an inlet installed at a height of 150 cm besides the tollbooth worker (i.e., near breathing zone). This tollbooth indoor monitoring provided shift and workplace specific 8-hour average total, elemental, and organic carbon concentrations. The data collection took place from January 4 to 14, 2001. We selected representative lanes to measure the PM2.5 concentrations in the tollbooths at different types of traffic flow during the three shifts. We chose 2nd, 4th, 5th, 6th, 15th, 16th, 17th, and 19th lanes to collect air samples.

Particle samples were collected on quartz fiber filters (Tissuquartz ™, SKC, Houston, Texas, USA) using an active personal sampler operated at a flow rate of 4 L/min. The sampler (URG-2000-25A, University Research Glassware Corp., Carrboro, North Carolina, USA) had an inlet nozzle and a greased impactor plate that eliminated particles with an aerodynamic diameter greater than 2.5 µm from the air stream before collection on the filter. Filters were separated within the filter-pack by Teflon o-rings and Teflon coated stainless steel screens. All cartridge materials were cleaned prior to assembly and field sampling to ensure minimal background contamination. All air-sampling pumps were calibrated before and after each time of sampling with Minibuck soap bubble airflow calibrator. The sampling flow rate was maintained within ± 5% of 4 L/min. Ten percent of all collected samples were submitted for analysis as field blanks. Gravimetric analysis was conducted in a constant humidity, 60% relative humidity, weighing chamber using a microbalance (Mettler-Toledo, MT5, Greifensee, Switzerland) with 1-µg reading. The microbalance was calibrated prior to each day's use with manufacturer-supplied traceable calibration weights.  All filters were conditioned for at least 24 h before weighing. In addition, all filters were passed over a static neutralizer (Allfield, Taipei, Taiwan) to reduce filter electrostatic charge that could interfere with accurate gravimetric analysis. The mean field blank weight change was 0.4 µg (n = 40; SD, 2.09 µg) for the filters. The detection limit, defined as 3 times the standard deviation of field blanks divided by the sampled volume of 1.92 m3, was 3.27 µg/m3. The detection limit for fine particles was sufficiently low (3.2 μg/m3) compared to the range of mass concentrations (22.7– 599.9 μg/m3).

EC and OC analysis

The carbon species in air samples were determined using the methods of Huang eta at. (2003). Briefly, particle samples collected on quartz fiber filter were analyzed for the content of total carbon and elemental carbon by a Total Organic Carbon Analyzer coupled with a Solid Sample Module (Shimadzu, Tokyo, Japan). After sampling, each filter was weighed to determine the particle mass concentration and then cut into two halves. These two halves were weighted separately. The weight proportion of each half filter was calculated and used to adjust the mass of carbonaceous measurements.  One half was placed on a sample boat and burned at 950 °C in the combustion chamber of the Solid Sample Module to determine the mass of total carbon. The combustion products, water vapor and carbon dioxide, were led to a drain separator where the water was trapped. Carbon dioxide was transferred into the Total Organic Carbon Analyzer to determine the concentration of CO2 by nondispersive infrared radiation according to Lambert-Beer's law. The mass of total carbon in each particle sample was calculated from a calibration curve established using standard solutions of glucose. The other half of filter was conditioned at 350 °C for 45 min and purged with helium gas to remove organic carbon and then placed in a sample boat and analyzed for the remaining elemental carbon content (EC) by the same procedure. The difference between the mass of total and elemental carbon was the mass of organic carbon. The total carbon analyzer was calibrated by four glucose standards ranging from 0.02 to 0.16 mg. The r square of the calibration curve was greater than 0.997. The relative standard deviations were 9% and 5% for elemental carbon and total carbon. The detection limit of carbon was 1.69 μg. Given the detection limits of both fine particulate carbon measurements, the accuracy of the exposure measurements can be deemed to be good.


Traffic counting and meteorological data

Hourly traffic data were obtained from the Tollbooth Administration, which maintained an hourly record of the total vehicle counts passing through each tollbooth.

We obtained meteorological data from the local station, maintained by the Central Weather Bureau and located 10 km from the toll station. 

Statistical methods

First, the average exposure concentrations according to the type of traffic lanes were described. The distribution of 8hour average concentrations of total, elemental, and organic carbon did not meet the normality by means of Shapiro–Wilk and Shapiro–Francia tests. Therefore we compared the 8-hour average concentrations of total, elemental, and organic carbon between different lanes using Wilcoxon rank sum test. Second, we estimated the relations between traffic density and total, elemental, and organic carbon concentrations by calculating the unit change in the mean concentration per 100 vehicles per hour in different lanes. Third, we assessed the determinants of total, elemental, and organic carbon levels using the method of generalised estimating equations (GEE), which accounted for correlated successive measurements of carbonaceous species concentrations measured on the same lanes. The effect of ignoring correlations underestimates standard errors (SEs) for the same lanes effect. The higher the correlation, the more SEs are underestimated. We applied GEE to investigate the independent effects of different determinants of carbonaceous species. The statistical analyses were performed with the STATA 8 version statistical package.

RESULTS

Meteorological and traffic density data

Table 1 shows the meteorological data during the study period provided by the Central Weather Bureau. During the period of sampling, wind velocity, temperature, and humidity ranged from 1.8 to 5.6 m/s, 17.5 °C to 20.6 °C and 83.6% to 94.0%, respectively. The average traffic density in the car lanes with ticket payment, 661 per hour (SD 221), was significantly higher compared with 390 vehicles per hour (SD 105) in the cash payment line (P < 0.001) and with 274 vehicles per hour (SD 85) in the truck and bus lane (P < 0.001). The traffic density was also significantly higher in cash payment lane, compared with truck and bus lane (P < 0.001).


Total, elemental, and organic carbon concentrations and traffic density in different lanes

Table 2 shows the distribution of total, elemental, and organic carbon, and PM2.5 concentrations and traffic density (vehicles per hour) according to the three types of lanes. The mean total carbon concentrations were over 4-fold higher in the bus and truck lanes (mean 167.7, SD 79.8 µg/m3, n = 59 shifts) compared with the two types of car lanes, with cash payment (mean 39.2, SD 29.5, n = 54 shifts; Wilcoxon rank sum test: P b 0.001) and with tickets (mean 34.1, SD 26.1, n = 108 shifts; Wilcoxon rank sum test: P < 0.001).  The mean total carbon concentration in the car lanes with cash payment was similar with that of the car lanes with ticket payment (Wilcoxon rank sum test: P = 0.35). As we reported previously , the levels of PM2.5 in the bus and truck lanes (308 µg/m3) were nearly 3fold higher than those in the two car lanes, with cash payment (116 µg/m3) and ticket payment (110 µg/m3).

In particular, the mean elemental carbon concentrations were substantially, over 6-fold higher, in truck and bus lanes (mean 131.7 µg/m3, SD 66.2 µg/m3, n=59 shifts) compared with the two types of car lanes, with cash payment (mean 20.2, SD 19.5, n=54 shifts; Wilcoxon rank sum test: Pb 0.001) and with tickets (mean 15.8, SD 17.6, n=108 shifts; Wilcoxon rank sum test: P< 0.001). 

The difference in the mean organic carbon concentrations was only 2-fold higher in the bus and truck lanes (mean 36.0 µg/m3, SD 25.8 µg/m3, n = 59 shifts) compared with the two types of car lanes, with cash payment (mean 19.2, SD 14.6, n = 54 shifts; Wilcoxon rank sum test: P b 0.001) and with tickets (mean 18.5, SD 12.2, n = 108 shifts; Wilcoxon rank sum test: P < 0.001).

Carbonaceous composition of PM2.5 according to the source of emissions 

On average, the mass fraction of carbonaceous species in PM2.5 accounted for 54% in the bus and truck lanes, whereas the corresponding figure was 31% for the car lanes with cash payment and 30% for the car lanes with ticket payment. These figures also show that elemental carbon dominated the carbonaceous species in bus and truck emissions. However, organic carbon and elemental carbon were evenly distributed among the carbonaceous species on car emissions (both in cash and ticket payment lanes).

Relation between traffic density and total, elemental, and organic carbon concentrations

We studied the relations between the traffic density and the mean total, elemental, and organic carbon concentrations for different types of vehicles by calculating the unit change in the mean concentrations per 100 vehicles per hour in different lanes (Table 3).

The increase in the mean total carbon concentration was 4.8 µg/m3 per 100 vehicles/h (95% CI 2.27–7.4) (n = 108) for car lanes with ticket payment, 10.0 μg/m3 per 100 vehicles/h (95% CI 3.2–16.5) (n = 54) for car lanes with cash payment and 54.9 μg/ m3 per 100 vehicles/h (95% CI 36.5–73.4) (n = 59) for bus and truck lanes. The mean change in total carbon concentration per 100 vehicles/h was 11 times greater in bus and truck lanes compared with car lanes with ticket payment.

The mean change in elemental carbon per 100 vehicle/h was 14 times higher in bus and truck lanes (mean 47.8 μg/m3 per 100 vehicles/h, 95% CI 32.2–63.4, n =59) compared with car lanes with ticket payment (3.4 µg/m3 per 100 vehicles/h, 95% CI 1.6–5.2, n =108). The cash payment in car lanes increased the levels almost 2-fold (5.5 μg/m3 per 100 vehicles/h, 95% CI 1.3–9.6, n =54).

The differences in organic carbon concentrations between the lanes were smaller than those in elemental carbon, with a mean change of 1.6 µg/m3 per 100 vehicles/h (95% CI 0.57–2.60) (n = 108) for car lanes with ticket payment, 4.5 μg/m3 per 100 vehicles/h (95% CI 0.95–8.02) (n = 54) for car lanes with cash payment and 6.9 μg/m3 per 100 vehicles/h (95% CI 0.04–13.66) (n = 59) for bus and truck lanes.

Figs. 1 and 2 illustrate the relations between concentrations of elemental and organic carbon and traffic density in the three different types of lanes. For both elemental and organic carbon, the slope is always the greatest for the bus and truck lanes and smallest for the car lanes with ticket payment. The slope for the bus and truck lanes is steeper for elemental carbon compared with organic carbon indicating that elemental carbon levels are more driven by traffic density than organic carbon levels. 

Determinants of carbonaceous species concentrations

We applied the method of generalised estimating equations to relate the log-transformed total, elemental, and organic carbon concentrations to the corresponding traffic density, wind speed, temperature, and relative humidity for each of the three types of lanes (Table 4). The results showed that none of the three environmental factors (i.e., wind speed, humidity and temperature) had a significant effect on carbonaceous species concentrations (all with P-valueN 0.05). Bus and truck lane was the strongest determinant of carbonaceous species concentrations. Traffic density without specification of the type of vehicles did not provide much additional information after including the types of lanes in the model.


