Particulate matter levels and comfort conditions in the trains and platforms of the Athens underground metro

: A study of indoor environmental quality inside the old (naturally ventilated) and new (air-conditioned) train cabins and platforms of four main stations of the Athens subway system (Attiko Metro), took place in different two-day measurements from June to August 2012. Portable instrumentation provided continuous measurements of particulate matter (PM 10 , PM 2.5 and PM 1 ), carbon dioxide (CO 2 ) along with temperature (T) and absolute humidity (AH). PM concentrations were significantly higher on the underground platforms of the network from 3 to 10 times, as compared to outdoor measurements. In particular, mean PM 1 , PM 2.5 and PM 10 concentrations at the deeper and most crowded station of Syntagma reached 18.7, 88.1 and 320.8 μg m −3 respectively. On the contrary, the ground level, open station of the Airport, showed values comparable to the outdoor (2, 6.4 and 34.4 μg m −3 , respectively). All PM fractions were lower than the platforms inside the old and new train cabins while the air conditioned trains experienced lower particulate pollution levels. More specifically, mean PM 1 , PM 2.5 and PM 10 concentrations were 5.5, 16.8 and 58.3 μg m −3 , respectively in new cabins while in the old they reached 10.3, 47.5 and 238.8 μg m −3 . The PM 2.5 /PM 10 and PM 1 /PM 2.5 ratios did not exceed 0.33 on both platforms and trains verifying the dominance of crustal coarse particles originating from the train and ground materials. As expected CO 2 levels were higher inside the trains as compared to the platforms and in some cases surpassed the 1000 ppm limit during the hottest days of the experimental campaign. Temperature and humidity remained relatively stable on the platforms, whereas measurements inside the cabins fluctuated, depending on the type of train and track locations. Correlations between measured PM along the routes to and from the Airport indicated covariance of concentrations along train cabins of the same direction.


Introduction
The modern way of life has lead commuters to spend 8% of a working day inside public transport means such as the underground systems in order to reduce travelling times and avoid traffic congestion [1,2]. The behavior of pollutants in the particular microenvironment is of great interest because of the large number of passengers present, the dependence on mechanical ventilation, the indoor pollutant sources and the exchange rates with the frequently polluted urban air. This diversified microenvironment observed on platforms and train cabins of underground metro networks presents higher levels of particulates as has been observed by numerous studies conducted globally [3]. The association between elevated levels of particles, particularly the finer fractions (2.5 μm or less), with adverse health effects has been well documented [4][5][6][7][8].
According to [9], that monitored the PM levels at the Taipei underground system platforms and cabins, the residence time of people has increased. More specifically it is important to know the exposure times of employees of the underground system as well as of vulnerable population groups (elderly and/or children) [10], since [11] found that spending 2h in the metro system of London per day would increase personal 24-h exposure of PM 2.5 by 17 μg m −3 .
In this respect, measurements in the subway of Helsinki demonstrated that concentration levels of PM in the underground environment were 3 to 4 times higher compared to the corresponding external [1]. This behaviour seemed to be more pronounced during the summer period as cases of discomfort for passengers and employees especially in non air conditioned cabins were reported. It should be noted that in the subway system of Beijing (where there is a mechanical ventilation system and natural openings are closed) internal concentration of particles was mainly influenced by external sources [12]. In the Buenos Aires subway, one of the oldest in the world, concentrations of total suspended particles (TSP) were measured at different stations, both underground and ground level. They were up to 3 times higher in the underground stations, while chemical analyses demonstrated the presence of iron (Fe) and copper (Cu) originating from ground excavations and zinc (Zn) that is associated with vehicle traffic [13]. In the study of [14], for the assessment of exposure to fine particulate matter (PM 2.5 ) in relation to the means of transport used (metro, bicycle, bus and car) in the city of London, 3 to 10 times higher concentration levels were recorded in the subway which was the most burdened means of transfer. Moreover, fluctuations of exposure to particles showed strong seasonality (higher in summer). The researchers also stressed the fact that the chemical composition of fine particles in the underground is different from that of the road surface as it consists of iron (Fe) and silicon (Si). Similarly, at the Stockholm subway, passenger exposure levels to PM were 5 to 10 times higher than those measured at the most crowded streets [10]. In line with the reports above, PM 10 and PM 2.5 consisted of iron (Fe), manganese (Mn) and copper (Cu) which probably are produced by construction materials. Regarding the chemical composition of particulates, [15] reported that the underground suspended particles are more toxic than those of the surface due to the presence of iron (Fe). Similar studies with extensive chemical analyses were carried out in the subways of Budapest [16], Milan [17], Zurich [18], Mexico City [19], Seoul [20][21][22], Istanbul [23] and Barcelona [24][25][26].
High concentrations of suspended particles have also been measured in other subway systems all over the world, as in Rome [27] where PM 10 and PM 2.5 levels were 3.5 times higher in the platforms and tunnels compared to the local roads, Prague [28], Berlin [29], New York [30,31], Montreal [32], Paris [33] and Sao Paulo [34]. Experimental measurements in newer underground rail networks such as that of Los Angeles (1993 operating year) showed that levels of PM 10 and PM 2.5 were 2.5 and 2.9 times higher than those of the outside environment. Measurements were also performed at the ground level part of the rail, which proved that the concentrations of PM 10 and PM 2.5 were strongly correlated with the external values of particles but at lower levels than those of the underground network [35]. Generally most studies concluded that the levels of suspended particles were associated with the ventilation system, the frequency of routes, the construction features of the trains and the maintenance of the lines [26]. In light of the above, efforts are being made to improve the air quality in burdened underground systems, such as in the Hong Kong Motor Rail Transport system where the active ventilation and the instalment of full height glass platform screen doors across the stations have improved the air quality making it less polluted than bus, train cabins and bus stations [36].
The Athens underground network (Attiko Metro) is one of the newest in the world and commenced operating in 2000. Previous short measurements of TVOCs, PM 10 , ΡΜ 2.5 , ΡΜ 1 , T and RH that were performed in train cabins in order to develop a fuzzy inference system to assess air quality, identified elevated pollutant levels although measurements were limited to draw conclusions on the spatial and temporal variations of the pollutants [37]. Moreover, [26], measured PM 2.5 concentrations and performed chemical analyses of the samples collected in three European subway systems including Athens on selected platform stations and outdoor environments as well as some measurements during travel times. The Athens measurements took place at a suburban area station that is close to the ground for a route travelled and showed that PM 2.5 concentrations were higher in the platform and during the train travel time than in the outdoor air, while the train frequency affected their levels. Concentrations were lower at night-time when the station was closed. The chemical analyses, in agreement with previous researchers indicated that higher metal concentrations were found on the subway platforms compared to ambient air. Fe was the most abundant element, followed by other metals originating from rails, wheels and brakes (e.g. Ba, Cu, Mn, Zn etc.).
Within that frame, the main aim of this work is to present results from the experimental campaign that took place in the platforms and train cabins of the Athens underground network (Attiko Metro) from June 27 to August 9, 2012. More specifically, the concentrations of PM 10 , PM 2.5 and PM 1 , CO 2 , as well as temperature, relative humidity and the number of passengers were monitored during six two-day periods from 6:30 am to 7:00 pm. The scope of the experimental campaigns was to: a) Identify the air quality status and indoor environmental conditions within the old (naturally ventilated) and new (air-conditioned) train cabins while travelling across the whole length of Line 3, b) monitor the air quality on four main platforms along the train route located at different depths below ground (Syntagma, Egaleo and D. Plakentias) and at ground level (Airport), c) examine the influence of outdoor environmental conditions on the indoor quality.

Description of Athens
The Greater Athens Area (including the Athens basin) features a complex topography (covering an area of 450 km 2 with approximate 4 million inhabitants), surrounded by mountains to the east, north and west and the sea to the southwest. The Thriassion Plain is located to the west of the Athens basin (mainly an industrial zone) and the Mesogia Plain is located to the east, a rural and suburban rapidly developing area due mainly to infrastructure works such as new highways and the Athens International Airport. The topographical features in combination with the local pollution sources (traffic, central heating, industries, shipping) and the prevailing weather conditions (sea-breeze cells, strong temperature inversions, calm wind) that impede ventilation, lead to pollution episodes in the summer [38,39].
Athens, lies at the southeastern-most part of the mainland of Greece and enjoys a prolonged warm and dry period during the year with July and August being the hottest and driest months. The normal value of the summer (JJA) daily maximum temperature (Tmax) at Athens (NOA) is 31.6 °C while the 90th/95th percentiles correspond to 35.3 °C/36.3 °C, respectively [40]. During the monitoring period three consecutive heat waves occurred with maximum temperatures higher than 37 °C for more than three consecutive days, i.e. from 9 to 17/7 (Tmax = 40.5 °C), from 28/7 to 1/8 (Tmax = 38.3 °C) and from 6 to 10/8 (Tmax = 40.9 °C), making this a very hot summer with poor comfort conditions.

Experimental methodology and instrumentation
The Athens subway system (Attiko Metro) is the only underground network in Greece. In 2012 it consisted of 3 lines with 54 stations in total [41]. On a daily basis, an average of 614,000 commuters uses it. Line 1 has been operating since 1869 and its largest part is over ground, while the modern lines 2 and 3 commenced operation in 2000 and are almost entirely underground. Line 3, where the measurements took place, had at the time 20 stations covering a distance of 37.7 km across the Athens basin from the Athens International Airport to the northeast to Egaleo at the southwest. The only terrestrial part of the line is between the stations of D.Plakentias and the Airport. In Figure 1, the Attiko Metro network is presented. Two types of train were in service on lines 2 and 3 in the subway: a) First generation (manufactured in 1999) naturally ventilated through windows, non-air-conditioned and commuting strictly on the underground part of the network (from Egaleo to D. Plakentias) and b) Second generation (manufactured in 2003) air-conditioned, travelling with windows closed and covering the whole Line 3 route (Egaleo to the Airport). Both types of train consisted of 7 wagons (cabins). The frequency of the routes depended on the time of the day. During the rush hours the frequency of train service was 3 minutes and for the rest of the day it was 10 minutes.
The experimental instrumentation consisted of portable continuous recording equipment. More specifically, Tinytag Plus 2 thermo-hygrometers were used for T and RH measurements, Turnkey Osiris and Lighthouse Handheld 3016 continuous monitors of mass particulate pollution (PM 10 , PM 2.5 and PM 1 ) and CO 2 concentration monitors IAQ RAE and MultiRAE IR. All parameters were measured at 10 seconds intervals and quality assurance of instrumentation was achieved by intercalibration measurements. It should be noted that hourly values of temperature, relative humidity, wind speed and gaseous pollutants were obtained from the National Observatory of Athens  In-cabin, measurements were taken in the middle -car of the moving trains and at 1 m from the ground (approximately the breathing height). As for stations, measurements were taken on the platform's end, 5 minutes before the departure of each train and for 5 minutes after its arrival to the station. The selected platforms were Egaleo (suburban traffic-approximately 20 m below ground), Syntagma (centre traffic-approximately 200 m below ground), Doukissis Plakentias (suburban traffic-approximately 50m below ground) and the Airport (ground level) (Figure 1).
At the end of the experimental campaign, raw data were examined qualitatively and quantitatively and were analysed firstly with the aid of boxplot diagrams. The median, minimum, maximum, first and third quartile (25% and 75%) of every distribution is indicated simultaneously along with the dispersion and the existence of outliers. The distance between the two quartiles is denoted as the interquartile range (IQR) while the median refers to a perpendicular equal to the width of the box. From each lateral side, a line is extended from the maximum to the minimum value on condition that they do not exceed the IQR by 1.5 times. These lines are labeled as whiskers. The ±1.5 IQR interval is called inner fence and ±3 IQR is the outer fence. Values outside the inner fence are considered suspected outliers while those outside the outer fence, extremes. Figure 2 describes in detail the structure of a boxplot.
The relative humidity data (RH) were converted to absolute humidity (AH) as that is a conservative property, independent from temperature. Absolute humidity was calculated by the following formula proposed by [45]: where, T is the temperature in °C and RH is relative humidity in %. AH values are measured in g H2O m −3 . This may be attributed to the in-tunnel particulate sources (sparks of rubbing when trains commute on rails, brakes of trains, ventilation system, construction and maintenance works), the frequency of train arrivals as well as the number of commuters, that directly affect the platform microenvironment (which is open) and secondarily the cabins (that mostly travel with closed windows and are confined spaces). In both cases, the majority of the recorded values exceeded the respective 24 hr-average exposure limits of PM 2.5 (25 μg m −3 ) and PM 10 (50 μg m −3 ). A plethora of PM outliers was also observed, which for the case of platforms are related with the piston effect-as the train travels the confined air is forced to move along the tunnel. Behind the moving vehicle suction is created and the air is forced to flow into the tunnel. This movement of air by the train is similar to the operation of a mechanical piston and affects the recording of the instruments located on a platform, causing instant increase of PM levels. Outliers in cabins are observed due to frequent movement of the passengers inside and/or the occasional window-opening.
On the other hand, the average CO 2 concentration in cabins reaches 796 ppm and that of the platforms does not exceed 598 ppm. This result makes sense, as numerous passengers huddle in the limited space of the train cabin and a large amount of CO 2 is released by exhalation. For that reason many outliers are observed in cabins and very few on platforms.
In general the temperature levels were elevated on all platforms as compared to the trains. More specifically, the average temperature on the platforms was 30.7 °C while in the cabins it did not exceed 28.8 °C. This may be attributed to the fact that the new generation trains are air-conditioned.

Station platforms
Regarding the station platforms, Tables 1 and 2 present the overall environmental quality status and Figures 3 and 4 the boxplot diagrams. In order to ensure that differences between various stations are statistically significant, a Kruskal-Wallis (non-parametric) test (p < 0.05) was applied since the measured data are highly skewed (as can be seen from Figures 3 and 4). Each case includes four grouping variables (station platforms) and one parameter. For all cases the p-value computed was very close to zero, demonstrating a statistically significant difference between pollutants among the stations. .000 .000 .000 .000 The most congested microenvironment was the platform of Syntagma, the central and deeper underground station. The average concentrations of PM 1 , PM 2.5 and PM 10 Figure 3c shows that PM 10 concentrations exceeded the 24-hour limit at all platforms, except in the Airport platform. In Figure 3b, it is observed that only Syntagma station presents PM 2.5 concentrations over the annual limit of exposure. Moreover, a significant number of outliers (1.5 and 3 times the inter-quartile range) was observed. The appearance of these values is associated with the arrival of trains at the stations and the movement of passengers, where the particles re-suspend and the instrumentation records momentary peaks.
The highest carbon dioxide levels were recorded at the Airport platform. The average concentration was 791 ppm with a maximum of 1156 ppm. Similar concentrations were recorded at the Egaleo station (771 ppm) a relatively small but crowed platform, especially in the morning and evening rush hours. The majority of outlier values are observed at Syntagma station owing to the large numbers of passengers. D.Plakentias station exhibited the lowest mean concentration (516 ppm), (Figure 3d).
Elevated temperature levels were observed at D.Plakentias station (average value 32.3 °C). The same platform also presented the highest maximum temperature compared to all others (34.8 °C). This may be attributed to the fact that the platform is quite close to the ground and because a big part of the roof is made of glass thus allowing the solar radiation to enter. The lowest average temperature was recorded at the Airport platform (28.4 °C), which was excepted since it is open.
Absolute humidity varies for each station. The highest concentration was found in Syntagma (15.4 g m −3 ), followed by Egaleo (13.6 g m −3 ), the Airport (12.6 g m −3 ) and D.Plakentias (11.1 g m −3 ). Concentrations of humidity are influenced by the operation of the ventilation system and the presence of people.

Old and new train cabins
The analysis of results shows that old train cabins are more burdened than the new ones. A Mann-Whitney (non-parametric) test (p < 0.05) has been applied to investigate if the differences of values are statistically significant. New and old trains were examined in pair-wise comparisons for every parameter (PM, CO 2 , T, and AH). The results indicate that the two types of trains have statistically significant differences as the p-value of all parameters approaches zero. Table 3 summarizes the results. The particulate concentration range in the two types of train cabins is quite large. Measurements in old cabins indicated average PM 1 , PM 2.5 and PM 10  The open windows in a moving cabin in an underground tunnel will lead to increased PM concentrations even though the commuter perceives the illusion of natural ventilation. All PM fractions appear numerous outlier values because of the re-suspension phenomenon caused by the movement of passengers within the cabin, the regular door opening and the open windows. It is also noted that in the new trains travelling to the Airport, higher particle concentrations were recorded compared to the opposite direction (i.e. towards Egaleo), most probably due to a greater number of passengers on these routes.
The mean concentration of carbon dioxide in the new (air-conditioned) cabins reached 826 ppm and was higher than that of the old ones (684 ppm), although they do not differ much when examining the 50% median. In closed air-conditioned cabins, the air renewal rate is lower thus leading to higher CO 2 levels from exhalation, while natural ventilation through windows helps keep CO 2 at lower levels. Figure 4 shows the behavior of airborne particles and carbon dioxide inside old and new trains for all experimental days. One may also observe that even though the median and quartile values for all PM fractions are lower for the new trains, the outliers are more and in some cases exceed the values measured in the old trains especially the finer PM fractions. This may be attributed to the frequent door opening at the stations and the performance of the ventilation system that can filter coarser particles more effectively.
The average recorded temperature in the new trains reached 28.4 °C while the concentration of the mean absolute humidity was found to be 13.7 g m −3 . These values, as expected, are lower than those of the old trains (31.5 °C and 17.6 g m −3 ) since the air conditioners contribute to the air temperature reduction and to dehumidification (Table 4). This behavior is due to the non-continuous operation of the air conditioning, the quick change in the number of passengers in the limited space of the cabin and the influence of the terrestrial segment of the metro (Figure 6).   A statistical comparison of PM 10 at the central station of Syntagma and within the train cabins is presented in Table 5. Because of the highly skewed data distributions, Spearman's rank correlation coefficient (rho) has been applied. The results vary, depending on the location in the subway. The primary observation is that within the trains, routes of the same direction showed very good correlations, while those of antisense are strongly negative, which demonstrates the covariance of PM 10 . The Syntagma platform values did not show any significant correlation with those measured inside the train cabins indicating that the cabin microenvironment is relatively independent from the tunnel besides the frequent door and window opening.
Moreover, another correlation that was examined was the CO 2 concentrations with the number of passengers between the Syntagma platform and the train cabins. As a product of exhalation, CO 2 is inextricably associated with the human presence in indoor environments. A linear regression of the two variables is presented in Figures 7 and 8. Due to the remarkable skewness and kyrtosis of the raw data, a logarithmic transformation was implemented to the CO 2 data.
Inside the trains, CO 2 varies depending on the number of passengers. The coefficient of determination (R 2 ) is equal to 0.205 while the correlation coefficient (R) was found to be 0.452. The p-value of the ANOVA (parametric) test (p < 0.05) approached zero, indicating a high level of significance. It is clear that, 20.5% of the variability of CO 2 concentrations can be accounted to the number of passengers and that there is a moderate correlation between the two variables. In the limited space of a cabin, CO 2 is slightly influenced by the presence of commuters. The air flowing through the open windows may reduce concentrations and affect correlations. On the contrary, CO 2 concentrations at the platform do not seem to be influenced by the commuters. The station is large enough to create a microenvironment without major fluctuations of CO 2 . Characteristically, R 2 and R are 0.066 and 0.258 respectively, indicating that only 6.6% of the variability of CO 2 is explained by the changes in passengers' numbers, while the correlation between the two variables was found to be relatively weak. Table 5. Spearman correlations of PM 10 between routes of trains and the platform of Syntagma. Every route is denoted with the letter ''R''. Routes with odd number represent one direction and with even number the reverse. Notice the strong correlations for routes of the same direction. PM

Comparison with outdoor measurements
In order to assess the impact of the outdoor environment on the indoor air quality, PM 10 concentrations from stations of the Athens Air Quality Monitoring Network were compared with data from the four platforms and cabins for each experimental day. Line 3 was divided into three sectors depending on the underground depth and location with respect to the ground. Sector 1 covers the densely populated area of west and central Athens (stations Egaleo to Panormou). Sector 2 is the suburban area between to the northeast (stations Katehaki to D.Plakentias). Sector 3 includes the ground level part travelling through the eastern suburbs outside the Athens basin (stations Pallini to Airport). The nearest stations are Aristotelous, Ag.Paraskevi and Spata, respectively (Table 6). Results show that during all experimental days the concentration of PM 10 in Syntagma is higher by 4 to 7 times with respect to the busy local street of Aristotelous, and constantly above 50 μg m −3 . Subsequently, the same behavior is observed at D.Plakentias station with air pollution levels exceeding almost three times those of the outdoor environment. At the Airport station the PM 10 concentrations are quite smaller, compared to the other platforms but still higher than those of the Spata station. Furthermore, with the exception of the Airport station platform, no significant correlation was observed between indoor and outdoor PM 10 levels, indicating that they are not affected by the air pollution changes that occur outside, (Table 7).
Regarding the in-cabin data, in Sector 1 a passenger taking the route Egaleo-Panormou (or vice versa) will be exposed to PM 10 concentrations almost three times higher compared to someone who moves on Aristotelous street. A similar behavior but at lower concentration levels was found in Sector 2, where for the Katehaki -D.Plakentias route the values were higher than these of the suburban station of Ag. Paraskevi. On 25 and 26 of July measurements were taken in old cabins only, which explains the significantly higher values. Sector 3 is the biggest part of the route to the Airport and it can be seen that PM 10 levels remained lower than the external, approximately for half experimental days. Finally, as expected, Table 9 demonstrates that no significant covariance of PM 10 among trains and the local meteorological stations is observed.
In figures 9a and b one may observe the PM 10 and PM 2.5 fluctuations within the new train travelling the route from Egaleo to the Airport, while measurements are taken across the Syntagma platform. It is observed that in the deeper parts of the route the concentrations are higher (from LST 9:20 am to 9:41 am approximately) and unsteady, while moving towards the Sectors 2 and 3 they drop and finally remain almost constant until it reached the Airport station and they peak again for a short while. It is also observed that PM concentrations are significantly higher and vary on the platform while the train arrivals further increase their values.

Concluding remarks
The monitoring of indoor air quality in the Athens Metro took place during the summer of 2012 and lasted for 12 days, between June 27th and August 9th. Continuous measurements of PM 1 , PM 2.5 , PM 10 , CO 2 , T and RH were taken with the aid of portable instrumentation between 6:30 am and 7:00 pm. The experimental methodology included simultaneous measurements across the full length of the platforms of Egaleo, Syntagma, D.Plakentias and Airport, as well as measurements in old (naturally ventilated) and new (air-conditioned) train cabins of Line 3 (Egaleo-Airport). Line 3 has an underground segment from Egaleo to D.Plakentias and a terrestrial from there to the Airport.
All PM fractions exhibited higher concentrations in the underground platforms as compared to old and new cabins and the outdoor air. Syntagma, the most crowded and deep station platform presented the most polluted environment, followed by Egaleo and D.Plakentias that are closer to the ground. The frequent movement of passengers on platforms along with the frequency of train services led to the re-suspension of particulates emitted from sources such as train brakes and excavation material. On the other hand PM concentrations in the old and new trains were lower because of the confined space that limits passenger movements and the closed (air-conditioned) windows that isolate the in-cabin air from the polluted tunnel air. Sudden peaks (characterized as outliers) appeared during embarkation-disembarkation of passengers at each station. PM 10 levels on the Airport were similar to the outdoor ones since it is a ground, open air space.
New train cabins that are air-conditioned presented the lowest (although still burdened) PM levels since the windows on the coaches were closed. However, CO 2 concentrations were elevated as compared to the platforms and old cabins because of the recirculation of air and the number of passengers on board. CO 2 measured on the platforms was unaffected by the number of people present. The temperature in the trains was lower than on platforms. Absolute humidity was slightly elevated in the cabins. Both variables remained almost constant on the platforms in contrast with the trains where they fluctuated depending on the position of the train (terrestrial or underground) and the number of commuters.
Continuous measurements of all variables in the moving trains showed that PM levels are higher in the underground parts of the route and reduce significantly at the terrestrial part. However, statistical comparisons between the outdoor and indoor measurements showed that the underground environment is not affected by the outdoor conditions. Moreover, the biggest part of PM consists of coarse particles that originate from the excavation and construction works, the train materials and the re-suspension. The external particulate pollution had no significant influence on the underground platforms of Syntagma and Doukissis Plakentias and did not affect the interior environment of the cabins. However, strong correlations were observed with the terrestrial platform of the Airport.