Analysis of Aerosol Composition and Assessment of Tunnel Washing Performance within a Mass Rapid Transit System in Taiwan

The metro system is the main mode of transportation in Taipei City. The air quality of metro stations is crucial for passengers. This study investigated particle size distribution and its elemental composition and assessed tunnel washing performance in the Taipei Rapid Transit System (TRTS). A 24-hour particle sampling process was performed in the tunnels, platforms, and entrances and exits to measure particulate matter (PM)2.5 and PM10 concentrations in an underground metro station and to analyze PM metal components. PM10 and PM2.5 concentrations decreased sequentially from the tunnels to the platforms and then to the entrances and exits to ambient environment in the metro station. The main metal components of suspended particulates in the TRTS mainly included iron, barium, copper, manganese, magnesium, aluminum, chromium, zinc, nickel, and lead. The total PM10 and PM2.5 metal proportions were 33.9%–24.7% and 32.9%– 22.8%, respectively. Furthermore, the effectiveness of tunnel washing in reducing the PM concentration was investigated. Monitoring results showed an increase in PM10 and PM2.5 concentrations after tunnel washing on the first day. The PM concentration started to decrease from the second day. Furthermore, 3.5 months after tunnel washing, the PM10 concentration decreased by 45.9%, and 2 months after tunnel washing, the PM2.5 concentration decreased by 71.3%. The mechanism of the continuous reduction in the PM10 concentration after cleaning is probably related to the porous material of tunnel walls, which may provide a deposition sink for aerosol particles, as well as the filter effect of the air conditioning system. This is the first study to use full-section tunnel washing to reduce PM exposure at mass rapid transit (MRT) stations. Although full-section tunnel washing reduced the PM concentration in the metro station, accessorial technology, such as the air conditioning system or platform design, may help in reducing the exposure of MRT passengers.


INTRODUCTION
Mass rapid transit (MRT) systems are the main modes of transportation in metropolitan regions.For example, the London Metro began operation as early as in 1863.In response to the high demand for urban transportation, other metropolitan regions established MRT systems in rapid succession.In densely populated metropolitan regions, MRT systems often adopt an underground orbital construction design because of regional land use restrictions.The quality and characteristics of the air in underground stations are considerably different from those of the ambient air above the ground.The daily volume of passenger flow through metro stations is in millions (e.g., 8 million people in the Beijing Metro, 7 million people in the Seoul Metro, 4 million people in the New York City subway, and 3 million people in the London Metro).Therefore, the designs of air conditioning systems, as well as fire, earthquake, and flood resistance systems, of metro stations are extremely crucial in ensuring the safety and health of passengers.
In the first European metro stations (underground), such as the Stockholm Metro, the particulate matter (PM) concentrations were reported to be several times higher than those in surrounding environments (Johansson and Johansson, 2003).This finding elicited the interest of scholars in variations in PM concentrations in MRT systems.A PM concentration higher than that in surrounding environments has been reported in underground metro stations or stations of other MRT modes in metropolitan regions such as Helsinki (Aarnio et al., 2005), London (Seaton et al., 2005), Rome (Ripanucci et al., 2006), Beijing (Li et al., 2006), Budapest (Salma et al., 2007), Prague (Braniš, 2006), Seoul (Kim et al., 2008), Buenos Aires (Murruni et al., 2009), Paris (Raut et al., 2009), Sydney (Knibbs and de Dear, 2010), New York (Wang and Gao, 2011), Los Angeles (Kam et al., 2011), Mexico City (Mugica-Álvarez et al., 2012), Barcelona (Querol et al., 2012), and Milan (Colombi et al., 2013).In addition to MRT stations having PM concentrations higher than those in surrounding environments, their aerosol compositions completely differ from those in surrounding environments, in which PM is mainly composed of fine aerosols (PM 2.5 ) and coarse aerosols (PM 10 -PM 2.5 ).Fine aerosols are generated by industrial processes and the burning of petrochemical fuels and wood (Buerki et al., 1989;Allen et al., 2001;Espinosa et al., 2001;Lin et al., 2008;Srivastava et al., 2008;Wang et al., 2008), whereas coarse aerosols originate from sea salt, soil, and street dust.The main components of aerosols include oxidized minerals, salts, carbon compounds, and crust elements.However, PM collected from MRT underground stations have been reported to exhibit a higher metal content than that collected from surrounding environments, with its metal components mainly including iron (Fe, the primary metal component), manganese (Mn), chromium (Cr), copper (Cu), barium (Ba), zinc (Zn), and nickel (Ni) (Aarnio et al., 2005;Murruni et al., 2009;Nieuwenhuijsen et al., 2007;Salma et al., 2009;Querol et al., 2012;Martins et al., 2016).
PM in MRT stations contains high concentrations of metals, which are generated by friction in mechanical systems, such as that between rails and wheels, that between collector shoes and conductive rails, and that in train braking systems (Johansson and Johansson, 2003;Sundh et al., 2009;Jung et al., 2010, Loxham et al., 2013).Then, the generated PM is blown into station platform areas and halls through the piston effect of the train (Moreno et al., 2014).When passengers inhale high doses of metal particles, their alveolar cells may experience adverse symptoms such as cytotoxic and inflammatory potential and oxidative stress (Seaton et al., 2005;Karlsson et al., 2005;Karlsson et al., 2006;Karlsson et al., 200;Gustavsson et al., 2008).
The objectives of the present study were as follows: (1) to understand the variations in PM concentrations and in PM metal compositions in tunnels, platforms, and entrances and exits; (2) to compare these PM metal compositions with those of MRT equipment such as tracks, vehicles, and collector shoes; (3) to compare PM concentrations with those in the surrounding environment through sampling and simultaneous analysis of PM in tunnels, platforms, and entrances and exits; and (4) to design a trial to assess whether full-section tunnel cleaning effectively reduces passengers' exposure to PM concentrations in MRT stations.

Particulate Concentration and Composition Analysis Site Description
Supplementary Fig. S1 shows Metro Station A (underground) in the Taipei MRT System.This metro station was selected for PM sample collection and composition analysis.This four-story underground station with doubledeck platforms on floors 2 and 4 was first opened to traffic in December 1998.At the ground level, the station has two entrances and exits for passengers.The daily volume of passenger flow is approximately 48,000 (based on statistical data from January 2015).Tunnel, platform, and ground-level entrance and exit sites were selected for the 24-hour PM 2.5 and PM 10 sampling processes.In this study, one-time sampling for the on-site measurement of air quality was simultaneously performed for the tunnel, platform, and ground-level entrances and exits at a height of 1.5 m away from any shelter to simulate the breathing zone of passengers.The sampling site of the tunnel was set at 10 m before the train enters the station.The sampling site of the platform was set in the center, and those of entrances and exits were set near the door.In addition, the data of ambient PM concentrations were collected from the nearest air quality monitoring station, 1.7 km from station A.
The weights of the sampled microaerosols were measured on a scale (resolution: ± 1 μg) before and after measurement.The total mass concentration of the microaerosols was analyzed.

Tunnel Washing And Particulate Reduction Site Description
The effect of tunnel washing on PM concentration reduction was investigated.Station B of the Taipei MRT System was selected for study.Station B is a three-story underground station with an island platform and four aboveground entrances and exits.This station was opened to traffic in November 1999.The entire line of this system is underground, and the daily volume of passenger flow is approximately 17,000 (statistical data from January 2015).PM measurements were performed on the station platform.

Sample Collection and Analysis
A continuous particulate monitor (MET ONE Inc., BAM1020, Oregon, USA) was used to measure variations in PM concentrations, in accordance with the EPA-TW NIEA A206.10C method, with detection ranges from 1 to 10,000 μg m -3 for PM 10 and PM 2.5 .The monitor was set on the platform of Station B before cleaning to record hourly real-time measurements and to monitor variations in PM 10 and PM 2.5 concentrations.The measurement mode was divided into short-and long-term monitoring modes on the basis of the measurement duration.The short-term mode involved continuous hourly monitoring, which started 1 day before cleaning and ended 2 days after cleaning (a total of 72 hours).The long-term mode involved continuous 24-hour monitoring for 1 day before cleaning, 1 day after cleaning, 1 month after cleaning, 2 months after cleaning, and 3.5 months after cleaning.The PM concentration during operating hours was calculated on each measurement day.
A high-pressure cleaning vehicle (China Steel Corporation Inc., RC48-335, Kaohsiung, TW) (Supplementary Fig. S2) was used for tunnel washing.Seven nozzles, each with a pressure range from 50 to 195 bars, were arranged facing the track surface; the cleaning coverage of the water sprays overlapped to more effectively clean the tunnel walls.Fourteen nozzles were arranged facing the tunnel wall with adjustable spray modes of column or mist, and the pressure of each nozzle ranged from 50 to 195 bars.The water tank capacity of the cleaning vehicle was 4,000 L. A water tank platform vehicle with a capacity of 30,000 L was attached to the cleaning vehicle to enable cleaning for a duration of up to 80 minutes.The cleaning vehicle was driven at 20 km h -1 during the process of tunnel washing, and the cleaning range included the tunnel section between Station B and adjacent stations.

Data Analysis
This study applied only descriptive statistics to demonstrate the PM concentrations at different sites or different times.We did not use other statistical methods for estimation or analysis.

Chemical Species of PM 2.5 and PM 10 in Different Locations
Table 1 shows the average PM 10 and PM 2.5 concentrations as well as the range of these concentrations in different subway systems around the world.The PM 10 and PM 2.5 concentrations varied widely in different locations.As mentioned, 24-hour sampling of PM 10 and PM 2.5 was performed in the tunnel, platform, and ground-level entrances and exits in Station A, and the samples collected on the filter paper were analyzed through ICP-MS; the corresponding results are shown in Tables 2 and 3 as well as in Figs. 1  and 2. The results of the PM 10 analysis revealed that the total PM 10 concentrations in the tunnels, platforms, and ground-level entrances and exits were 423.00, 227.00, and 86.00 μg m -3 , respectively, with metals accounting for 33.9%, 24.6%, and 24.7% of the particulate content in these respective areas.The proportions of Fe within the metal content were the highest at 93.1%, 91.6%, and 92.4%, respectively, of the total metal concentration.The other measured metal components, arranged in descending order in terms of concentration, were Ba, Cu, Mn, Mg, Al, Cr, Zn, Ni, and Pb.Moreover, the total PM 2.5 concentrations in the tunnel, platforms, and ground-level entrances and exits were 168.00, 85.00, and 50.00 μg m -3 , respectively, with metals accounting for 32.9%, 36.6%, and 22.8% of the particulate content, respectively.The proportions of Fe within the metal content were the highest at 92.6%, 93.2%, and 92.9%, respectively, of the total metal concentration.The ranking of the other measured metal components was consistent with that observed in the PM 10 analysis.In this study, the distribution of the aerosol composition observed in subways is similar to that reported in other studies (Kang et al., 2008;Kim et al., 2010;Jung et al., 2012;Eom et al., 2013).As shown in Tables 3 and 4, the PM concentration decreased from the tunnel to the platforms and then to the entrances and exits.The total proportions of metals in PM 10 and PM 2.5 found in the tunnel were approximately 33.9% and 32.9%, respectively, whereas the total proportions of metals in PM 10 and PM 2.5 observed at the entrances and exits decreased to 24.7% and 22.8%, respectively.Moreover, the PM 2.5 :PM 10 ratio increased from 0.4 in the tunnel to 0.58 at the entrances and exits.This increase indicates that at locations close to the ground level, the underground air was mixed with the surrounding air, approaching the composition ratio of this air.However, the PM concentration at entrances and exits remained higher than that of ambient air near the station on the same day, indicating that PM from inside the station escapes into the surrounding atmosphere through air conditioning, thus producing piston and diffusion effects.
Previous studies (Nieuwenhuijsen et al., 2007;Sundh et al., 2009;Jung et al., 2010;Loxham et al., 2013) have compared PM concentrations as well as the elemental composition of PM and its determinants in an MRT system.These studies have suggested that the sources of PM may include wheels, ventilation, air conditioning, and braking systems; however, they did not provide evidence to explain the difference in PM metal compositions.
The track equipment of the Taipei MRT System uses tracks of UIC60 specification (Hung et al., 2008), as required by the International Union of Rail (UIC).UIC860 series 900A rails are used in the general track area, and UIC860-grade 1100 rails are used in the curved areas.According to the UIC specification regarding the chemical composition of steel rails, in addition to a Mn proportion of 0.8%-1.3%, a 1100 rail should contain 0.8%-1.3%Cr.These composition ratios are close to the results of the present study.In this study, the Mn:Fe ratios for PM 10 and PM 2.5 in the tunnel were 0.017 and 0.016, respectively, and the Cr:Fe ratios for PM 10 and PM 2.5 in the tunnel were 0.0005 and 0.0004, respectively.The sources of Fe, Mn, and Cr in the air of the MRT system can be deduced because UIC860-grade 1100 wear-resistant rails are used for the curved tracks (a small proportion of all tracks).The steel wheels are Class-B wheels, as recommended by the Association of American Railroads.In addition to Fe, the rails contain Mn, Ni, Cr, Mo, V, Cu, Al, Ti, and niobium (Nb) in the proportions of 0.6%-0.9%,0.25% (max), 0.25% (max), 0.1% (max), 0.04% (max), 0.35% (max), 0.06% (max), 0.03% (max), and 0.05% (max), respectively.Furthermore, a conductive track is a composite track with an Al skeleton, conductive stainless steel surface (containing 10%-30% Cr), and collector shoes produced by Mersen using metal-impregnated graphite (containing 22% Cu).PM produced by the friction engendered by train movement has similar metal components and proportions to those obtained in the present study.Therefore, PM in the tunnels and platforms is mainly generated by contact friction between the wheels and the conductive rails during train movement.Tsai et al. (2008) investigated PM 10 and PM 2.5 concentrations in the atmosphere of Taipei and found that PM 10 and PM 2.5 concentrations ranged from 28.40 to 34.20 μg m -3 and from 14.84 to 22.22 μg m -3 , respectively.In PM 10 , the mass Fe, Mn, Pb, As, Cd, and Cr concentrations were 242.54, 7.54, 6.44, 1.03, 0.31, and 6.22 ng m -3 , respectively.In PM 2.5 , the mass Fe, Mn, Pb, As, Cd, and Cr concentrations were 138.73, 5.55, 6.52, 0.77, 0.25, and 4.24 μg m -3 , respectively.Moreover, the metal contents of PM 10 and PM 2.5 were approximately 0.9% and 0.88%, respectively, of the total PM mass.A comparison of metal contents found in the present study and those in other MRT systems (Chillrud et al., 2004;Aarnio et al., 2005;Seaton et al., 2005;Salma et al., 2007) indicated that the main metal components accounted for more than 30% of the PM 10 and PM 2.5 mass concentrations.In addition, the PM 10 and PM 2.5 concentrations in the tunnel were 10 times more than those in the surrounding environment.This result shows that the composition, characteristics, source, and hazard composition of PM in MRT systems are considerably different from those in the surrounding environment.

Cleaning Performance within the MRT
The measurement results obtained from different sites in the metro stations indicated that PM produced by trains during the course of operation is blown into the platform area through the piston effect, which is the main source of PM pollution in MRT stations.Even at the exits of the MRT stations, the PM 10 (86 µm m -3 ) and PM 2.5 (50 µm m -3 ) concentrations remained higher than those in the surrounding environment.Therefore, the reduction or removal of PM in the tunnel can improve the air quality of MRT stations.Previous studies (Hu et al., 2006;Feng et al., 2012;Guo et al., 2016) have reported that the time and amount of rainfall had different scavenging rates (up to 80%) for PM of different sizes in the surrounding air, because many soluble salts are found in the surrounding air.Therefore, we evaluated whether the use of water columns to wash the inner walls of the tunnels in an MRT system can reduce the PM concentration.
To the best of our knowledge, no relevant literature on the effectiveness of washing the inner walls of tunnels using water columns is available.In the current study, a cleaning operation was performed in the tunnel between Station B and the adjacent stations of the Taipei MRT system.Subseqently, short-(3 days) and long-term (approximately 3.5 months) measurements of PM 10 and PM 2.5 mass concentrations at Station B were obtained in Table 5.During this time, the regular train schedule was maintained, and volume changes in the passenger flow through the station were less than 5%.The tunnel cleaning vehicle and the scenes before, during, and after tunnel washing are shown in Supplementary Fig. 2. The shortterm continuous measurement results (Fig. 3) showed that the PM 10 concentration increased rapidly to 1,000 μg m -3 during the cleaning process, which is possibly because of water vapor and re-suspended dust; however, it could not be distinguished by the air monitor used in this study.The slight increase in the PM 2.5 concentration may be due to the fact that the cleaning process produces large particulates because of the attachment of water and particles.However, after the first day of cleaning, the monitoring results showed an increase in both PM 10 and PM 2.5 concentrations.The average PM 10 and PM 2.5 concentrations (during operation) on the first day after cleaning increased to 145.0% and 152.3%, respectively, compared with those prior to cleaning, and the maximum concentrations also increased to 133.4% and 140.1%, respectively.These results may be caused by the sewage containing residual particles, which remained at the bottom of the tunnel after cleaning.Thus, 1 day after tunnel washing, PM was blown back into the station platform area due to the piston effect of train operation following the evaporation of sewage.Two days after tunnel washing, the average PM 10 and PM 2.5 concentrations decreased by 81.4% and 78.0%, respectively, compared with those prior to cleaning (during operation hours), and the maximum concentrations significantly decreased by 67.2% and 67.0%, respectively.The constant decrease in PM concentrations 2 days after tunnel washing may be attributed to 1) the removal of accumulated dust in the tunnel, 2) the filter effect of the air conditioning system, and 3) the dilution by fresh outside air.Therefore, a trend in PM reduction was observed.
To determine the duration of the cleaning effect and the change in PM concentrations after cleaning, a 3.5-month follow-up monitoring of PM concentrations in Station B after tunnel washing was performed.As shown in Fig. 4, the average PM 10 concentration after 3.5 months was 111.2% higher than that on the day after cleaning, and the average PM 2.5 concentration after 3.5 months was significantly higher than that on the day prior to cleaning.One month   after cleaning, the average PM 10 and PM 2.5 concentrations decreased by 79.2% and 71.3%, respectively, compared with those on the day prior to cleaning.Two months after cleaning, the average PM 10 concentration decreased by 45.9% compared with that on the day prior to cleaning, whereas the average PM 2.5 concentration on the day prior to cleaning was 78.1%.At 3.5 months after cleaning, the average PM 10 concentration reached 46.1%, whereas the average PM 2.5 concentration remained at 76.8%.The mechanism of the continuous reduction in the PM 10 concentration after cleaning is probably related to the porous material of the tunnel wall, which may provide a deposition sink for aerosol particles, as well as to the filter effect of the air conditioning system.The results show that tunnel washing significantly reduced the PM 10 concentration; specifically, the short-term reduction in the PM 10 mass concentration reached 20%, and the PM 10 mass concentration during 2 to 3.5 months after cleaning remained at 50% of the concentration on the day prior to cleaning.However, the mechanism of the continuous reduction in the PM 10 concentration after cleaning requires further study because this is probably related to the porous material of the tunnel wall, which may provide sites for aerosol particle adsorption after cleaning, or to the filter effect of the air conditioning system.However, the short-and long-term PM 2.5 concentrations were reduced by only approximately 21% and 29%, respectively.These results indicate that tunnel washing had a lower effect on fine aerosols with smaller particle sizes, such as PM 2.5 , than on PM 10 .The continuous reduction in the PM concentrations after tunnel washing was possibly caused by PM reduction factors such as the ventilation and air filtering effects of the air conditioning system, the dilution effect resulting from the introduction of fresh outside air, and the redeposition (adsorption) effect of the clean concrete tunnel walls.Further investigation is required to clarify the contribution of these effects to the sustained decline in the PM concentration.The effectiveness assessment of PM reduction in MRT stations showed that the use of water columns to clean tunnels can reduce PM concentrations.
The tunnel cleaning test results suggest that the ringtype high-pressure water column cleaning method is suitable for shield tunnels, because the water columns can thoroughly clean tunnels with circular sections.However, before cleaning, facilities (lighting, communications and signaling apparatus, and power supply units) within the tunnel must be confirmed to be waterproof to prevent system failure or anomalies caused by the cleaning process.After cleaning, the sewage must then be properly discharged, because the residual sewage accumulated on tunnel bottoms can easily cause secondary dust pollution upon resuspension.
In this study, the number of sampling points for analysis was limited to different locations in the MRT station to observe the decreasing trend of PM concentrations at different sites on the same day of measurement.Therefore, additional measurements must be performed to assess the average PM concentrations or seasonal variations in the stations.Moreover, only one short-term test and one longterm test were conducted to assess tunnel washing due to operational and logistical limitations.Therefore, the actual PM reduction effect may depend on the air conditioning system, track length, and PM accumulation.

CONCLUSIONS
The results of this study reveal that PM in metro stations contains a high proportion of metal components, of which Fe is the main component along with Mn, Cr, Cu, and other metals.The PM 10 and PM 2.5 concentrations decreased from the tunnels to the platforms and then to the entrances and exits.The main metal components of PM in the Taipei Rapid Transit System are Fe (21.2%-34.1%),Mn (0.3%-0.5%),Cu (0.3%-0.6%), and Ba (0.3%-0.7%); these metal components are also present in the MRT track, wheel, and conductive shoe material.
The reduction effect of MRT tunnel washing on PM 10 and PM 2.5 concentrations was also evaluated.The shortterm monitoring results show that the PM 10 concentration in the platform area increased rapidly during the cleaning process and even reached twice the maximum precleaning value, whereas the PM 2.5 concentration did not increase significantly during the cleaning process and reached only 87.3% of the maximum value before cleaning.However, on the second day, the average PM 10 and PM 2.5 concentrations decreased significantly.Compared with the concentrations on the day prior to cleaning, the average PM 10 and PM 2.5 concentrations were determined to decrease by 81.4% and 78.8%, and the maximum concentrations decreased by 67.2% and 67.0%, respectively.Long-term monitoring showed that the PM 10 concentration continued to decrease after tunnel washing, and the minimum PM 10 concentration on the day prior to cleaning was 45.9%.The PM 10 concentration started to slowly increase 2-3.5 months after tunnel washing.The minimum PM 2.5 concentration after cleaning was 71.3% on the day prior to cleaning, and this increased again 1-2 months after cleaning.

ACKNOWLEDGMENTS
This study was supported by the Taipei Rapid Transit System.The interpretation and conclusions contained herein do not represent those of the Taipei Rapid Transit System.

DISCLAIMER
The accountability of the research matter articulated in this journal is entirely of the authors concerned.The view expressed in the research articles in this journal does not essentially correspond to the views of the publisher or editor.The publisher or editor of the journal is not liable for errors or any consequences arising from the exercise of information contained in it.

Table 1 .
Average PM 10 and PM 2.5 concentrations and concentration ranges in different subway systems and locations.

Table 2 .
PM 10 concentrations and compositions at different metro station sites.

Table 4 .
Comparison of the fine:course ratios of PM (unit: µg m −3 ).Outdoor data obtained from the nearest air monitoring station of TW-EPA at the same time were selected.