A Comparison of Sampling and Analytical Methods for Assessing Occupational Exposure to Diesel Exhaust in a Railroad Work Environment

ABSTRACT

Methods of assessing occupational exposure to diesel exhaust were evaluated in a railroad work environment. The American Conference of Governmental Industrial Hygienists (ACGIH)-recommended elemental carbon and respirable combustible dust methods of sampling and analysis for assessing diesel exhaust were included in the study. A total of 215 personal and area samples were collected using both size-selective (nylon cyclone and Marple) and non-size-selective samplers. The results demonstrate that the elemental carbon method is suitable for the railroad environment and the respirable combustible dust method is not. All elemental carbon concentrations measured were below the proposed ACGIH Threshold Limit Value (TLV° R ) of 0.15 mg/m3. The concentrations of oxides of nitrogen (nitric oxide and nitrogen dioxide) were also found to be below their respective TLVs. There is no correlation between elemental carbon or respirable combustible dust and the oxides of nitrogen. The elemental carbon as fraction of total carbon is about 13 percent, except for onboard locomotives where it is about 24 percent. Comparison of elemental carbon and respirable combustible dust measurements showed consistent relationships for most sampling locations, with respirable combustible dust concentrations 12 to 53 times higher than the elemental carbon levels.

MATERIALS AND METHODS

Workplace Description and Sampling Methods

The maintenance shops at two locations (hereafter identified as Yard 1 and Yard 2) of a major Canadian railroad company were selected for the study. Two areas of potential diesel exhaust exposure were identified at each yard: the turnaround area and the heavy repair area. Both personal and area samples for various contaminants were collected at these two locations. Most of the sampling was conducted at Yard 1, and a limited number of samples were taken at Yard 2. A small number of area samples were also taken in the leading or trailing unit cabs of locomotives during several trips. These samples measured the driver’s exposure while on board locomotives.

Yard 1 contains a very large maintenance shop that houses a turnaround area measuring about 500 feet in length by 60 feet wide by 40 feet high. Locomotives are brought into this area for inspection and tuning. There are large service doors at both ends of the two tracks, and each track is equipped with an overhead exhaust canopy. The locomotive engines have to be run briefly during testing. Diesel exhaust is released into the shop during the work at the turnaround and also when the locomotives are brought into the shop. This area has the highest potential of diesel exposure. The heavy repair area, where the locomotives are repaired, measures about 200 feet by 250 feet by 35 feet, and is partially separated from the turnaround area. The heavy repair area was considered to have the next highest potential for exposure to diesel exhaust because the locomotives may or may not be running while in this area. The maintenance shop has local and general ventilation for the control of fumes.

The maintenance shop and turnaround area at Yard 2 are essentially similar to Yard 1 in dimensions and ventilation control. Yard 1 is located in central Canada and Yard 2 is located in western Canada. Personal samples were collected from machinists, electricians, hostlers, and laborers directly involved in the turnaround area, and occasionally from the supervisor of the area. Area samples were collected from turnaround and heavy repair areas. To compare sampling methods, several samplers were hung side-by-side in the turnaround and heavy repair areas and on board the locomotives. In some instances, several samplers were mounted on a tripod and placed carefully as close to the exposure area as practically possible.

The vast majority of personal and area samples were 7- to 8-hour long-term samples. The sampling duration ranged from 5 1/2 to 10 hours. A limited number of side-by-side personal samples were also collected by equipping a worker with two personal samplers to assess the relationship between EC and RCD measurements for personal sampling.

A total of 215 samples of various types were collected. The details of the sampling and analytical methods used are given in Table I. All samples were collected between April and July, essentially representing spring and summer conditions. Attempts were made to sample on cold days when doors were closed as well as on warm days when they were open. The area samples were taken at the same location in the turnaround area at six feet from the deck level with three different collecting devices (size-selective samplers): a nylon cyclone, an open face total dust sampler, and a Marple sampler. The EC method specifies the use of a sampler with a total dust open-face configuration, and in coal mines, a sampler with an impactor employing a submicron cut point is specified.

The Marple sampler used in this study meets the submicron specification of the method. It has a final cut point of 0.5 um aerodynamic diameter. The study also examined whether the cyclone sampler, with a cut point less than 10 um aerodynamic diameter, could provide results equivalent to that of the Marple sampler in the railroad environment. It was postulated that by the nature of combustion process, EC particulates from diesel exhaust are in the submicron range. Because no other significant sources of elemental carbon are present in locomotive repair facilities, or on board locomotives, it may be possible to use widely available cyclone samplers instead of a submicron sampler like the Marple.

Analytical Methods

Elemental carbon samples were analyzed by Mr. Robert Cary, Sunset Laboratory, Forest Grove, Oregon, using NIOSH Method 5040. The method measures both elemental carbon (EC) and organic carbon (OC). At the time this study was undertaken, this was the only commercial laboratory analyzing samples by this method in North America. Recently, other laboratories, including one in Canada, have developed this capability.

RCD is the organic dust that is combusted out of the respirable dust collected by using a cyclone. It is determined by the weight difference between respirable dust prior to ashing and leftover inorganic dust after ashing. The method published by CANMET was followed. The nitrogen dioxide and nitric oxide were determined by NIOSH method 6014. The RCD and NOx samples were analyzed by our own AIHA-accredited laboratory.

Quality Control

Quality control (QC) procedures were adhered to during eld sampling as well as during the analysis of samples. The field sampling QC included pre- and post-calibration of sampling devices, the use of eld blanks, and the monitoring of the sampling train during the sampling period. Analytical QC included the use of blanks, spiked samples, and other QC procedures.

Spiked samples were included with the field samples sent to Sunset Laboratory for EC and OC analysis. This was considered important because, at that time, they were the only lab offering this type of analysis commercially. The organic carbon spikes were prepared by dissolving 41.31 mg of sucrose in 100 ml of distilled water. 500 u l aliquots were spiked onto precleaned quartz lters. This amount of solution completely saturated the filter and gave a concentration of 10.18 l g of organic carbon per
square centimeter of filter area. This is the best range of deposit for organic carbon for the method (range 5–200 u g/cm2). The EC spikes were prepared by using carbon black obtained from a local carbon black manufacturing company. Carbon black was ground in a spex mixer mill and sieved to less than 10 u m. A fixed amount (0.195 mg) of carbon black was weighed on a microbalance and transferred to a 100 ml volumetric flask. 100 microlitres of aerosol OT (a wetting agent) was added to aid in the dispersion of particles and the volume was taken up to 100 ml with distilled water. The solution in the flask was ultrasonicated for two hours to disperse the particles. Ten-ml aliquots were deposited onto precleaned quartz filters using vacuum ltration. The final concentration of the elemental carbon was 11 micrograms per square centimeter of the filter area. The best range of deposit for EC is 1–15 micrograms per square centimeter. Both OC- and EC-laden filters were placed in air sampling cassettes and sent along with the field-sampled cassettes as part of the QC program. For RCD, the QC methods designed for gravimetric analysis were used, and for NOx, spiked samples were used.

Statistical Analysis

In addition to the sampling and analytical method, every sample was classified according to type of sample (personal or area), location (turnaround, heavy repair, or on board locomotive), and ventilation situation (sampling with doors open or doors closed). For each set of results, both arithmetic and geometric summary statistics were computed. Histograms were produced and inspected for distributional shape and the assessment of outliers. The relationship between sampling methods was investigated using simple least squares linear regression methods with the fitted line forced through the origin. The slope of the line is the main summary statistic that results from this analytic approach. As a check on this assumption, the models were refit allowing for an estimable intercept. Statistical significance for comparing the slope to one or the intercept to zero was declared if the corresponding p-value was less than 0.05.

RESULTS

The EC-C, RCD, NO, and NO2 results of the study are summarized in Table II with respect to their TLVs: 0.15 mg/m3 for EC-C and RCD, 31 mg/m3 for NO, and 5.6 mg/m3 for NO2. The number of non-detectable assessments, where sample analysis fell below the detection limit (DL) of the method, are also reported in Table II. The calculations (in this table only) include these samples with values imputed as one-half of the DL.

The mean values of RCDs are higher than the mean values for RDs in the Turnaround Yard 2 Area and Heavy Repair Yard 1 Area (0.05 versus 0.03, and 0.18 versus 0.16, respectively). This is due to a filter media problem encountered in the beginning of the study. It was originally thought that both RD and RCD could be sampled on pre-weighed quartz fiber filters. Because the EC samples are taken on quartz fiber filters, we felt this would aid in the method comparison by having the same filter media for all analysis. We discovered, however, that the quartz fiber filters were brittle and friable and unsuitable for gravimetric determinations. The five samples taken in Yard 2 Turnaround and three of the 14 samples taken in Yard 1 Heavy Repair area were taken on the quartz fiber filters. These filters sometimes gave higher RCD than RD values.


The results of side-by-side personal samples obtained by using two pumps on workers from various occupations are given in Table III. The results of the QC spiked organic carbon and EC samples are given in Table IV.

The results of side-by-side area samples for measuring EC taken using three different methods—cyclone (EC-C), total (ECT), and Marple (EC-M)—are shown in Figures 1 and 2, using EC-C as the comparison standard. A 45-degree line through the origin, describing a perfect one-to-one linear relationship, is superimposed. The estimated slopes (1.19 and 0.75 for the EC-T versus EC-C comparison, and the EC-M versus EC-C comparison, respectively) are both statistically significant ( p < 0.05). When the linear models were refit with estimable intercepts, neither were found to be statistically different from zero.

Unfortunately, at Yard 2, the protocol with regard to measuring EC by cyclone (EC-C) was not adhered to and in place of EC-C, eight EC-T samples were taken. To make use of the data, we used the strong linear relationship between EC-C and EC-T discussed earlier to estimate eight EC-C values.

The survey produced 62 sets of side-by-side samples suitable for statistical comparison. The respirable dust (RD) and RCD measurements were compared with the elemental carbon concentration sampled by cyclone (EC-C), treated as the “gold standard” for assessing diesel exhaust. There were 18 personal and 44 area sample sets, all approximately eight hours in duration. All personal samples were taken at the turnaround locations. Only one set was missing both the RD and RCD measurements. Of the 44 area sample sets, 21 were taken at turnaround locations, 14 were collected in the heavy repair area, and 9 sets were sampled on board locomotives.

After inspection of the summary statistics and histograms of the RCD and EC-C data, two results were clearly exceptional. One area sample in the turnaround area produced an extreme EC-C concentration of 0.025 mg/m3, considerably higher than other area samples, except those taken on board locomotives. Similarly, one personal sample in the turnaround area resulted in unusually high RD and RCD levels of 1.55 and 1.51 mg/m3, respectively. We concluded that these extreme results (more than two standard deviations above the mean) were likely due to filter contamination. Because the emphasis of this study was to compare methods, we chose to remove these two outlier samples from the subsequent statistical analysis.

A large proportion of the samples were below the limit of detection of the method (i.e., non-detectable): 42, 45, and 19 percent for RD, RCD, and EC-C, respectively. The non-detectable values were not included in the data analysis. Summary statistics for the seven location-ventilation groups are presented in Table V. Included in the table are the number of usable and non-detectable sample values, the minimum and maximum values, the arithmetic mean, the standard deviation, the geometric mean and geometric standard deviation for each type, and the location and ventilation situation.

The majority of the non-detectable assessments were recorded during sampling with the doors open, and while sampling on board locomotives. Also evident from the statistical summary are the very low concentrations which reflect the small amounts of materials collected on the filters. Because precision is related to the filter loadings, all summary statistics and statistical analysis recorded here should be interpreted with some degree of caution.

Table VI presents a summary of the slope estimates based on the linear regression analysis. Basically, this linear model parameter represents the ratio of respirable dust concentrations (RD and RCD) to the concentration of EC-C. Even though the data sets are small, it is encouraging to note the consistency of the measures within each location. Except for samples taken on board locomotives, the slope estimates varied between 12 and 53. Little difference was seen between the RD and RCD assessments.

Figure 3 displays the 11 pairs of RCD and EC-C observations for the personal samples taken in the turnaround and heavy repair area. The linear predictor clearly fits poorly, especially when the ventilation situation is considered (the black circles represent sampling in the turnaround or heavy repair areas with the doors open, and the hollow circles denote sampling with the doors closed). Although forcing the line through the origin has the effect of guaranteeing a strong positive slope, the main observation in this data set is the substantially higher concentrations of RCD relative to EC-C.

Figure 4 presents the RCD and EC-C concentrations for the 16 area sample sets taken in the turnaround and heavy repair areas. The estimated linear predictor is superimposed on the graph. Unlike the personal sampling results, there appears to be a strong linear relationship for the “doors closed” situation. The concentrations observed while the doors were open (i.e., black circles) produce a pattern more conducive to a threshold effect, albeit based almost entirely on one sample with relatively high levels of both RCD and EC-C. The area samples, taken together, produce both a strong linear and supra-linear relationship with RCD levels much higher than that of EC-C. When the models were refit allowing for an intercept, it was reassuring to find that none were found to be statistically different from zero.

The relationship between the RCD and EC-C measurements with the NO and NO2 concentrations was almost non-existent in our samples. The correlation coefficients for the various sample types, which varied from - 0.15 to + 0.20, were all not statistically different from zero (a = 0.05). A summary is presented in Table VII. The elemental carbon (EC-C) as a fraction of total carbon (TC-C) is summarized in Table VIII. A summary of elemental carbon (EC) concentrations by job category is given in Table IX. 
