Characterization of Occupational Exposure to Air Contaminants in Modern Tunnelling Operations

ABSTRACT 

Objectives:

Personal air measurements of aerosols and gases among tunnel construction workers were performed as part of a 11-day follow-up study on the relationship between exposure to aerosols and gases and cardiovascular and respiratory effects.

Methods:

Ninety tunnel construction workers employed at 11 available construction sites participated in the exposure study. The workers were divided into seven job groups according to tasks performed. Exposure measurements were carried out on 2 consecutive working days prior to the day of health examination. Summary statistics were computed using maximum likelihood estimation (MLE), and the procedure NLMIXED and LIFEREG in SAS was used to perform MLE for repeated measures data subject to left censoring and for calculation of within- and between-worker variance components.

Results:

The geometric mean (GM) air concentrations for the thoracic mass aerosol sub-fraction, α-quartz, oil mist, organic carbon (OC), and elemental carbon (EC) for all workers were 561, 63, 210, 146, and 35.2 μg m−3, respectively. Statistical differences of air concentrations between job groups were observed for all contaminants, except for OC, EC, and ammonia (P > 0.05). The shaft drillers, injection workers, and shotcreting operators were exposed to the highest GM levels of thoracic dust (7061, 1087, and 865 μg m−3, respectively). The shaft drillers and the support workers were exposed to the highest GM levels of α-quartz (GM = 844 and 118 μg m−3, respectively). Overall, the exposure to nitrogen dioxide and ammonia was low (GM = 120 and 251 μg m−3, respectively).

Conclusions:

Findings from this study show significant differences between job groups with shaft drilling as the highest exposed job to air concentrations for all measured contaminants. Technical interventions in this job should be implemented to reduce exposure levels. Overall, diesel exhaust air concentrations seem to be lower than previously assessed (as EC).

METHODS

Work characteristics

These tunnel construction workers work 12 days consecutively and are then off for 9 days. A typical work shift lasts 10–12 h and includes two breaks of 30 min each. Tunnel construction workers are engaged in rock drilling, charging of explosives, and various support- and finishing work. Occupational job groups included in this study are described in Table 1. Briefly, the excavation process starts off with drilling and charging of explosives. After blasting, the rock is loaded and trans- ported out of the tunnel using dump trucks. Finally, removing of loose rocks using a scalar and various types of rock support is carried out. Rock support includes, e.g. fastening of unsafe rock with steel bolts and sealing of the rock by spraying wet concrete onto the excavated surface. Other important tasks during excavation are mounting ventilation ducts, maintenance and repair of machines, and installation of electrical power supply. If the risk of water leakage into the tunnel is considered high, injection workers carry out rock consolidation with micro concrete to prevent leakage. Up to 25 tons of concrete may be injected during one work shift in holes that can be 22 m deep and cover the entire excavated profile. All tunnels investigated in this study had forced ventilation systems using fans and ventilation ductings to dilute aerosols and gases for workers in all areas of the tunnel. Excavation of the shaft followed the same sequence as for tunnels, however, instead of using an underground drilling rig, pneumatic handheld equipment for rock drilling and a raise climber were used. The only ventilation in the shaft was from pressurized air used to power the drills.

Study design

All tunnel construction workers (n = 91) employed at 11 available tunnel construction sites located at different parts of Norway were invited for this study in 2010–2011. Participation was voluntarily. One worker decided not to participate. Health effects assessments were performed shortly before the work shift on the first day back on site after 9 days off. After 11 days of work, the medical tests were performed again at the same time of the day.

The workers were stratified into job groups according to tasks performed. Job groups included in this study were drill and blast workers, drill and blast mechanics (a subgroup of the drill and blast workers), support workers, loaders (a subgroup of support workers), injection workers, shotcreting operators, and shaft drillers. Personal air measurements were carried out on 2 consecutive working days prior to the day of the second health examination. Each worker was sampled twice to assess between-worker (BW) and within-worker (WW) variabilities. Thoracic dust, elemental carbon (EC), organic carbon (OC), α-quartz, and NO2 were measured in all workers. Oil mist, oil vapour, and NH3 were measured in a subsample of workers from all job groups (N = 57), except shotcreting operators and injection workers. All samples were collected in the breathing zone outside personal protective respirators. To ensure that the exposure level was effectively measured, an attempt was made to sample for at least 7–8 h. Exceptions to this were when workers did particular jobs (e.g. shaft drilling and shotcreting) because overloading of the sampling filters was anticipated. In these situations, the sampling time was reduced. Alternatively, when work activities were less than anticipated due to irregularities of the work operations and unplanned delays, sampling equipment was carried for longer durations.

The sampling time varied between 270 and 855 min [arithmetic mean (AM) = 569 min].

Sampling methods

The thoracic aerosol mass sub-fraction was collected with a thoracic cyclone (BGI GK 2.69 sampler, BGI Inc., MA, USA) operated at a flow rate of 1.6 l min−1. The thoracic fraction is defined by a penetration curve of the total aerosol with a 50% cut-off at an aerodynamic diameter of 10 µm and geometric standard deviation (GSD) of 1.5 (CEN, 1993). Filters were polyvinyl chloride membrane with pore size 5 μm (PVC502500, Millipore Corporation, MA, USA) and a sodium iodide impregnated gas filter pad (cellulose support pad) placed after the aerosol filter by inserting an extra ring into the standard three-part 37-mm aerosol filter cassette (Millipore, MA, USA). The impregnated filter was inserted into the filter cassette to simultaneously collect NO2 (Hovland et al., 2012).

EC and OC containing particles were collected on pre-cleaned quartz filters (Pallflex Tissue quartz 2500QAT-UP, Pall Corporation, Port Washington, NY, USA) using a 37-mm standard, three-part aero- sol filter cassette (Millipore, MA, USA). The sampling flow rate was 2.0 l min−1. Filter cassettes were purchased from Sunset Laboratory Inc. (Tigard, OR, USA). The personal air sampling pumps employed were in-house constructed at the National Institute of Occupational Health (Oslo, Norway) (NIOH) and operated at constant air flow rates through the filters.

Using a combination of 37-mm standard, three- part aerosol filter cassette (Millipore, MA, USA) equipped with a glass filter (No. 1820-037, Whatman GF, Madistone, UK) on top of a cellulose acetate fil- ter (AAWP03700, Whatman GF, Madistone, UK) and tubes containing 150 mg  charcoal  (100 mg  in front section) (No. 226-01, SKC, Blandford Forum, Dorset, UK) mounted in series during sampling, oil mist and oil vapour were collected, respectively (Galea et al., 2012). Personal air sampling pumps (type 224- PCTX4, SKC, Eighty Four, PA, USA) were used to collect samples within a maximum time window of 2 h at an air flow rate of 1.4 l min−1.

Calibrated rotameters were used both at the beginning and at the end of each sampling period to measure the air flow rates through the filters. Samples with >10% decrease in air flow rate over the sampling period were rejected.

Air concentrations of NO2 and NH3 were measured with portable direct-reading electrochemical sensors with data logging built into the instruments (PAC7000 Dräger AK, Lübeck, Germany). An averaging period of one reading every 30 s as logging interval was selected. The limit of detection (LOD) for NO2 and NH3 were 376 and 1393 µg m−3, respectively. The response factors of the electrochemical sensors were calibrated when the data were downloaded (i.e. about every month) with certified calibration gases (Yara Praxair ASA, Oslo, Norway).

Gravimetric measurement of aerosol mass

The masses of thoracic cyclone filters were measured gravimetrically using a Sartorius AG, MC 210p labora- tory micro balance (Göttingen, Germany). The LOD (3 × standard deviation of field blank filters) was 3 µg (31 µg m−3 based on 8 h of sampling at a flow rate of 1.6 l min−1). The gravimetric measurements were per- formed in a climate-controlled room with continuous measurement of the temperature (20 ± 1°C) and relative humidity (40 ± 2%). To ensure comparable and accurate weighing conditions, all air filters were acclimatized for at least 5 days in the climate-controlled room before weighing. Static charge was eliminated prior to weighing the filter at all occasions using a 210Po Staticmaster® (NRD LLC, NY, USA).

Field blanks consisted of sampling cassettes loaded with filters, which were taken to the field together with the sample cassettes but were never opened (i.e. kept sealed). One field blank for every 10 particulate samples, with at least 1 blank per day, was included. All blanks were analysed gravimetrically for thoracic dust and were below the LOD.

Chemical analysis

All reagents and water used for chemical analysis were of analytical quality.
The impregnated filter pads were placed in poly- propylene tubes (Prod. No. 62.554.001, Sarstedt AG & Co, Germany) and leached in deionized water. Bromide was added as internal standard to all sample solutions. The concentrations of nitrite (NO −) and nitrate (NO −) were determined by ion chromatography using a Dionex DX-500 ion chromatograph (Dionex, Sunnyvale, CA, USA) (Hovland et al., 2012). The LOD for NO2 was 20 µg (21 µg m−3 based on an 8-h sampling period at a flow rate of 1.6 l min−1).

Oil mist was determined using a model Spectrum 100 Fourier transform infrared spectrophotometer (Perkin Elmer, Waltham, MA, USA), and oil vapour was determined using a Perkin Elmer Autosystem XL gas chromatograph and a flame ionization detector (Galea et al., 2012). The LOD for oil mist and oil vapour was 0.05 and 0.1 mg m−3, respectively, based on a 2-h sampling period at a flow rate of 1.4 l min−1.

EC and OC were determined by Sunset Laboratory Inc. (Tigard, OR, USA) using an OCEC Dual- Optical Analyzer according to NIOSH Method 5040 (NIOSH, 2003). The LOD of the method is ~2 ng m−3 and 2 µg m−3 based on an 8-h sampling period at a flow rate of 2.0 l min−1 collected on a 37-mm filter with a 1.5 cm2 punch from the sample filter for EC and OC, respectively. The accuracy of this method for measuring total carbon was ascertained in this measurement campaign by analysing a known quantity of carbon in the form of sucrose.

The α-quartz content in the thoracic mass sub-fraction was measured by X-ray diffraction spectrometry, applying Philips PW1729 X-ray generator, Philips 1710 diffractometer control, Philips PW2253/20 X-ray tube, and Philips PW1050 goniometer. α-Quartz content was determined according to NIOSH Method 7500 (NIOSH, 2003). The LOD (3 × standard deviation of field blank filters) was 10 µg (13.0 µg m−3 based on an 8-h sampling period at a flow rate of 1.6 l min−1).

Data analysis

The frequency distribution was examined visually using probability plots and indicated that a log-normal distribution provided a better fit to the exposure data. The data were therefore ln-transformed before statistical analysis. The measured air concentrations were used without further adjustments. Air concentrations were summarized by geometric means (GM), GSD, minimum concentrations (Min), and maximum concentrations (Max) using maximum likelihood estimation (MLE). AM was estimated from the expression EXP[lnGM + 0.5 lnGSD2] (Seixas et al., 1988). The SAS procedure NLMIXED was used to perform MLE for repeated measures data subject to left censoring and for calculation of WW and BW variance components for all contaminants except for NH3 where the SAS procedure LIFEREG was used because there was no repeated measurements ( Jin et al., 2011). To perform MLE computations, the SAS program uses numerical values above the LOD, information on the proportion of data below the LOD, and a mathematical formula for an assumed distribution of the data.

To evaluate the significance of fixed effects on the BW and WW variance components, values of the variance components obtained from the mixed-effect models were compared to values obtained from the random-effects model without fixed effects. The percentage of the total variance explained by the fixed effect (job group) was calculated by subtracting the sum of the WW and BW variances from the total variance (random effects only).

The likelihood ratio test was used to compare the models with random effects to those having both random- and fixed effects.

Correlations between exposure variables were evaluated using Spearman’s correlation coefficient.

Statistical analyses were carried out with SPSS 21.0 (SPSS Inc, Chicago, IL, USA) and SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).

RESULTS

A total of 90 tunnel construction workers carried personal sampling equipment in the exposure study, and all workers were monitored twice. Few workers reported use of personal protective respirators, except shotcreting operators who used filtering half mask respirators with P3 filters for filtering solid and liquid particles (3MTM) part of the sampling time. In total, 6 samples of α-quartz and 20 samples of EC and OC were discarded because of technical failures.

In total, 79 personal full-shift samples of NO2 using direct-reading instruments (DRI) were evaluated. Only 8 of these measurements had time-weighted average (TWA) above the LOD of 376 μg m−3 (results not shown). The median NO2 TWA concentration
 
of samples above LOD was 565 μg m−3 (range: 376– 1317 μg m−3) (results not shown). However, in 17 of the DRI measurements, maximum observed peak value incidents for a 30-s averaging period of NO2 were detected (>3764 μg m−3).

Table 2 gives an overview of air concentrations by contaminant. The GM air concentrations for the thoracic mass aerosol sub-fraction, α-quartz, oil mist, OC, EC, NO2 (filter), and NH3 for all workers were 561, 63, 210, 146, 35.2, 120, and 251 μg m−3, respectively (Table 2). Statistical differences of air concentrations between job groups were observed for all contaminants, except for OC, EC, and NH3 (P > 0.05). On average, OC accounted for 76% of the total carbon measured, and total carbon accounted for 49% of the thoracic aerosol mass (results not shown). Also, statistical differences of air concentrations of α-quartz between construction sites were observed (P < 0.05). The AM percent of α-quartz in the thoracic mass aero- sol sub-fraction ranged from 3 to 40% between sites (results not shown).

Tables 3 and 4 give an overview of air concentrations by job group. The shaft drillers, injection workers, and shotcreting operators were exposed to the highest air concentrations of thoracic aerosol mass (GM = 7061, 1087, and 865 μg m−3, respectively). The shaft drillers and the support workers were exposed to the highest concentrations of α-quartz (GM = 844 and 118 μg m−3, respectively). Shotcreting operators and drill and blast workers were the highest exposed workers to NH3 (GM = 2927 and 2857 μg m−3, respectively). The highest levels of NH3 were found during
loading of crushed rock into dump trucks using an excavator following detonation of the explosive. An example is shown in Fig. 1.

In models with only random effects, the BW component was higher than the WW component for all contaminants (Table 5). When job group was added as fixed effect, the BW variance components were reduced for all contaminants (13–59%). Job group explained between 7 and 57% of the total variance (Table 5).

Table 6 shows the correlation matrix between exposure variables. No correlation coefficient exceeded 0.6. The highest correlations were between air concentrations of OC and oil mist and between EC and NO2 [rSpearman = 0.56 and 0.60, respectively (P < 0.0001)].
