Current Issue

A BAM (5028i, Thermo Fisher Scientific, Waltham, Massachusetts, USA) was used as a beta-ray attenuation method for measuring PM_2.5 concentration. Its sampling flow rate was 16.67 L/min. The sampling time was 1 hour. Six OPCs (AQM-06, HCTM Co., Ltd., Gyeonggi-do, Republic of Korea) with 16 size channels were used with a light scattering method. Their sampling flow rate was 1.2 L/min and their sampling time varied from 6 s to 24 h. To enable comparison with the BAM, the sampling time of the OPCs was also set to 1 hour. The specifications of the BAM and OPCs used in this study are shown in Table 1. An OPC consists of a small volume chamber, a laser source, and a detector (Denker and Shklovsky, 2013). When a particle passes through a light beam, a part of the light is scattered and a detector measures it. The number of particles can be determined by the frequency of the scattered light and the size of a particle can be evaluated based on the intensity of the scattered light (Denker and Shklovsky, 2013). Through the number of particles, the mass concentration of PM at each size can be calculated with respect to its density and size, as shown in equation (1) (Colombi et al., 2013):

where C_mass is the mass concentration of PM at the particle diameter of d, C_number is the number PM at the size of d, ρ is the density of PM at the size of d, and ε is the sampling efficiency.

A BAM consists of a beta-ray source to produce beta particles, a filter holder coupled with a filter tap to collect PM, and a detector to determine the decrease in beta-ray intensity after penetrating the PM layer on the filter (Shukla and Aggarwal, 2022). The mass concentration C of PM can be calculated as shown in equation (2):

where A is the surface area of the filter (cm²), I₀ is incident beta-ray intensity (mg/cm²), I is the attenuated beta-ray intensity after passing through the PM layer (mg/cm²), μ is the attenuation coefficient (cm²/mg), Q is the air flow rate (L/min) and t is the sampling time (min).

The BAM and six OPCs were calibrated and certified by the Korea Laboratory Accreditation Scheme. The six OPCs were also calibrated in the same chamber to determine their correlation. The coefficient of determinations (r²) among OPCs was found to be >0.99 for both PM₁₀ and PM_2.5.

An underground subway station of line 7 (Sang-dong station) of Seoul Metro was selected to investigate PM_2.5 in indoor air. The instruments were set up and operated for several days before data collection from March 15 to April 15, 2021. In this period, the temperature of the station varied from 15.3-21.8°C and relative humidity was 20-70%. The station has two platforms on the B2 floor with screen doors. One BAM was placed on a platform to monitor PM_2.5. Six OPCs were installed at six sites to measure PM₁₀ and PM_2.5 in the waiting room and platforms. The experimental set-up is presented in Fig. 1. Two OPCs (i.e., OPC_1-1 and OPC_1-2) were installed in the waiting room. Two OPCs (i.e., OPC_2-1 and OPC_2-2) were set up in the platform without a BAM, and two others (i.e., OPC_2-3 and OPC_2-4) were located in the platform with a BAM. The locations of the OPCs and BAM are shown in Fig. 1. The experiment was conducted for one month.

Fig. 1. 
Experimental set-up for investigating PM_2.5 and PM₁₀ concentration at an underground subway station.

The 1h-average PM_2.5 concentrations observed by the BAM and OPCs were compared for seven consecutive days. The PM_2.5/PM₁₀ ratios were calculated to determine the relationship between PM sizes and the correlation between BAM and OPC. The distribution of particle size based on the particle number observed by OPCs was also taken into account. A Fisher’s least significant difference (LSD) procedure was used to compare the mean PM_2.5 between the BAM and OPCs at the 95.0% confidence level using MATLAB (Version 9.10.0. 1684407, MathWorks, Inc., Natick, USA). On the other hand, 24h-average PM_2.5 concentrations obtained from BAM and OPCs were also considered. All experimental results were illustrated as figures using Microsoft Office Excel 2019 (Excel 2019, Microsoft Cooperation, Redmond, Washington, USA).

The variations in1h-average PM_2.5 concentrations measured by a BAM and six OPCs in the waiting room and platforms for seven days are shown in Fig. 2. As shown in Fig. 2, PM_2.5 variations observed from the BAM had a similar pattern to those observed from six OPCs. However, the correlation between PM_2.5 concentrations measured by BAM and OPCs differed from Monday to Wednesday but were similar from Thursday to Sunday. To explain this pattern, more data are presented and discussed in the next section. In addition, the PM_2.5 concentrations in the platforms revealed unclear patterns not only between rush hours and rest hours but also between weekdays and weekends. Since the platform is connected to the train tunnel, the PM concentration in the platform is generally affected by the train’s operation, which emits abundant particles (Son et al., 2013). However, the subway station in this study had screen doors installed to separate the platform from the tunnel. It was previously reported that the PM level in a platform coupled with screen doors was significantly lower than that in the tunnel, and the correlation of PM levels in the platform and ambient air was high (i.e., 0.6-0.8) (Son et al., 2013). Moreover, many air purifiers were also employed to remove particles from the air in the current station (as shown in Fig. 1). Therefore, unclear PM_2.5 patterns between rush hour and rest hour and between weekdays and weekends were observed. In terms of the effect of relative humidity (RH) on OPCs, because all relative humidity values were less than 50% (Fig. 2, dotted line), humidity could not have a significant effect on OPCs. Besides, since the OPCs had heating units on the sampling inlet area, humidity might not be a significant problem in this study.

Fig. 2. 
Variations in hourly PM_2.5 concentrations measured by two different methods at various locations.

To clearly determine the correlation between PM_2.5 measured by BAM and OPCs, PM_2.5 concentrations observed from OPC_1-1 and OPC_1-2, OPC_2-1 and OPC_2-2, OPC_2-3 and OPC_2-4, OPC_1-1 and BAM, OPC_2-1 and BAM, and OPC_2-3 and BAM were compared using linear regressions. The results are shown in Table 2. Moreover, Fisher’s LSD procedure was employed to compare the means of different analyzers. A Box-Whisker plot was also illustrated, as shown in Fig. 3, for easy observation of the patterns.

Fig. 3. 
Distribution of 24h-average PM_2.5 concentration in a week with respect to BAM and OPCs.

As shown in Fig. 3, the means of 1h-average PM_2.5 concentrations obtained from BAM from Monday to Wednesday were significantly higher than those observed from OPCs at all sites. As shown in Table 2, two OPCs on the same floor denoted a good correlation. The correlation of BAM versus OPC_1-1 from Monday to Wednesday showed the lowest correlation (i.e., r²<0.4) due to different floor locations. However, in the same platform, the correlation between BAM and OPC_2-3 was also low from Monday to Wednesday. In contrast, BAM and OPCs showed a good correlation (i.e., r²>0.65 for different floors and r²>0.75 for the same floor) although they were located on different floors. The Fisher’s LSD results for 1h-average PM_2.5 concentrations among analyzers confirmed this significant difference between BAM and OPCs from Monday to Wednesday (Table 3).

It was previously reported that PM concentration obtained from an OPC could be different from that obtained from gravimetric method because the variation of PM composition results in a difference of its density, which is used to estimate the mass of PM based on the number of PMs (Colombi et al., 2013). Hence, the ratios of PM_2.5/PM₁₀ and the distribution of various particle sizes based on particle number were estimated to determine the relationship between PM composition and the correlation between BAM and OPC. As shown in Table 4, PM with sizes less than 2.5 μm contributed 50-80% mass of PM₁₀ from Thursday to Sunday. In contrast, this mass contribution was higher than 80% on Monday and less than 50% on Tuesday and Wednesday. In addition, when the size distribution of particles based on the particle number was investigated, the pattern of particle size from Monday to Wednesday was also different from that on the other days (Fig. 4). As shown in Fig. 4, most particles in this station were less than 0.5 μm based on particle number. The number of particles<0.5 μm observed on Monday (see red dotted circle in Fig. 4) was significantly higher than on the other days. In contrast, the number of particles<0.4 μm observed on Tuesday and Wednesday (see black dotted circle in Fig. 4) was significantly lower than on the other days. This suggests that the difference in PM_2.5 between BAM and OPCs depends on particle size.

Fig. 4. 
Size distribution of PM based on particle number obtained from six OPCs on various days.

Previous studies have reported that PM_2.5/PM₁₀ ratios varied at different underground subway stations (Table 5). Although these ratios varied depending on locations, the contribution of different levels of PM_2.5 was generally high. This indicates that when a BAM and an OPC are used to measure PM_2.5 at these stations, the correlation between BAM and OPC could be relatively high depending on the composition of PM.

Consequently, the 1h-average of PM_2.5 concentrations in an underground subway station varied with the sampling locations. Hence, a PM database observed from only one sampling site might not be representative of PM concentration at Seoul subway stations. In terms of a sampling method, PM_2.5 concentrations obtained from OPCs revealed a good correlation with those from BAM in this study when the PM₁₀ comprised 50-80% PM_2.5. However, more studies should be carried out to identify a reliable pattern of PM_2.5/PM₁₀ ratios. In addition, no OPC was installed next to the BAM to observe particle characteristics. This is a limitation of this study which should be overcome in future works.

The 24h-average PM_2.5 concentrations were investigated for one month, and their variations are depicted in Fig. 5.

Fig. 5. 
Variations of 24h-average PM_2.5 concentrations with respect to BAM and OPCs: (a) daily variations and (b) Box-Whisker plot of one month.

As shown in Fig. 5, the 24h-average PM_2.5 concentrations observed from BAM showed a similar pattern to those observed from OPCs in terms of daily variations. Moreover, the mean values of one month were similar among different measurement techniques (Fig. 5b). To clearly demonstrate this pattern, a Fisher’s LSD procedure at 95% confidence for the mean value of each technique was conducted, and linear regressions were also run. The correlation of 24h-average PM_2.5 concentrations between BAM and OPCs is shown in Fig. 6.

Fig. 6. 
Relationship of 24h-average PM_2.5 concentrations among different measurement locations and methods.

As shown in Fig. 6, although different monitoring techniques were conducted at different sampling locations in the station, 24h-average PM_2.5 concentrations revealed a very good correlation between BAM and OPCs, as well as among OPCs. The results of the Fisher’s LSD also confirmed that there was no significant difference between each pair of means because all differences were within the limit of ±9.7. The PM_2.5/PM₁₀ ratios based on the 24h-average data were also evaluated. The results are presented as a Box-Whisker plot in Fig. 7.

Fig. 7. 
Various PM_2.5/PM₁₀ ratios based on 24h-average concentrations.

As shown in Fig. 7, most PM_2.5/PM₁₀ ratios were in the range of 50-80%. This pattern was similar to that of 1h-average concentrations, which also showed a good correlation between BAM and OPCs when the PM_2.5/PM₁₀ ratios were within 50-80%. In particular, the r² coefficients between OPC and BAM found in this study were >0.9, which is similar to the correlation between OPCs and gravimetrical method found in another study (i.e., 0.85-0.97) (Colombi et al., 2013). This indicates that OPCs could be used to monitor PM_2.5 concentrations in terms of 24h-average at subway stations. Furthermore, 24h-average PM_2.5 concentrations with a good correlation among different sampling locations (i.e., waiting room vs. platform) were also found. An analyzer could be used to demonstrate 24h-average PM_2.5 concentrations and can be installed at any location in subway stations.

Since the use of 24h-avarage data could not observe the real-time status of air quality to recognize and solve indoor air quality issues in a timely manner, real-time data monitoring is suggested in addition to 1h and 24h-average data. In this case, good calibration OPCs could be used.