COMMUTER EXPOSURE TO AEROSOL POLLUTION ON PUBLIC TRANSPORT IN SINGAPORE 2

The ratios of transport mode to background site BG measurements in Table 4-2 provide a measure of the difference between concentrations in the different transport microenvironments and a

Chapter 4 Results

The results are presented in three parts The statistical analysis of the door-to-door journeys on each of the four modes of transport is presented first, followed by the spatial variations observed within each transport mode and the results and analysis of the dosage calculations

Concentration spikes and calculation of geometric means

The 1 s and 10 s logging intervals managed to capture the numerous short-lived concentration spikes that occurred throughout the sampling period These short-lived spikes resulted in distributions of exposure concentration that were skewed towards larger values To illustrate the frequency of these spikes, time-series of the measurements taken on 20 May 2013 are shown in Figure 4-1 Although some of these sharp peaks may be classified as extreme outliers (defined as > 3 standard deviations [SD] above the mean), they are a common occurrence and a real component of the total pollutant concentration, thus they have not been removed To account for these values without giving them too much weight, the geometric mean (GM) is used as the main descriptive statistic for each trip The GM is essentially the logarithmic mean of the data, obtained by applying a logarithmic transformation on the data, taking the arithmetic mean, then back-transforming the mean values by taking the antilog Other studies have also used the GM instead of the arithmetic mean to better describe the data (e.g Gulliver and Briggs, 2004, de Nazelle et al., 2012)

Figure 4-1: Post-processed data measured on 20 May 2013 Time-series shown include one day of measurements on all transport modes High variability and presence of spikes

are evident in all measured parameters

In the rest of this thesis, the GM represents the average pollutant concentration measured during one door-to-door trip for each mode of transport or within a particular section of a trip The arithmetic mean was only used to average

HR and V E For MRT mode, data from both trips on each day of measurement were combined for a total of 23 sets for analysis However, for calculation of dosage, only data from the first MRT mode trip of each day were used In the following

or section The boxplots presented in this chapter are also based on the GM from each trip or section

4.1 Commuter exposure on door-to-door trips

Mean values and SDs of all measurements are summarised in Table 4-1 More detailed descriptive statistics including maximum and minimum values can be found

in Appendix D.1 The ratios of transport mode to background site (BG) measurements in Table 4-2 provide a measure of the difference between concentrations in the different transport microenvironments and at the ambient level Generally, all three PM size-fractions and PN concentrations are higher for the four transport modes than at the background site, with the exception of the MRT mode This agrees with existing research on commuter exposure, which finds that measurements in the transport microenvironment are elevated compared to ambient readings (Kaur et al., 2005b; Gulliver and Briggs, 2007; de Nazelle et al., 2012) The Walk (MRT) mode exhibits the highest (lowest) mean pollutant concentrations, while

CO values in the transport environment and at BG are similar The results of the individual pollutant metrics are described in more detail in the following sections

Table 4-1: Mean (SD) of pollutant metrics from all trips for 4 transport modes and measured at the background site (N = 23 for Bus, Taxi, and MRT, N = 22 for Walk) Transport mode PM 1 (µg m -3 ) PM 2.5 (µg m -3 ) PM 10 (µg m -3 ) PN (# cm -3 )

Table 4-2: Transport mode to BG ratios from all trips for pollutant metrics measured in

Table 4-3: Results from the Kruskal-Wallis test validating that concentrations measured

on each mode of transport were significantly different from each other and the

background site H = test statistic, df = degrees of freedom

4.1.1 Particulate matter mass concentrations (PM)

The results for the three PM size-fractions measured are discussed together since the trends between transport modes are similar (Figure 4-2) The highest mean concentrations were observed during the Walk mode, with values above 35 µg m-3 for all three fractions (PM1: 37 µg m-3, PM2.5: 37 µg m-3, PM10:42 µg m-3) Maximum values (individual data points) for PM1, PM2.5 and PM10 were also observed during the Walk mode In terms of PM10, the maximum recorded was 677 µg m-3, while

PM2.5 and PM1 reached a maximum of 532 µg m-3and 529 µg m-3 respectively (see Appendix D.1) The mode with the second highest mean PM concentrations was the Bus mode, with mean PM1, PM2.5 and PM10 values of 29 µg m-3, 29 µg m-3, and 32 µg

m-3 respectively The mean concentration levels observed for the MRT and Taxi mode were very similar and close to the Bus mode, with a minor difference of 1 µg

m-3 for PM2.5 One difference between these two transport modes is the smaller spread

of data for the MRT mode, with SDs of 4 µg m-3 for the three PM size-fractions compared to 7 µg m-3 for the Taxi mode (Table 4-1)

Mean PM values were higher on the four transport modes than at the background site (Table 4-2) However, only the Walk mode exhibited a relatively high ratio of ~1.50 The other modes were approximately 1.1 – 1.2 times higher than

BG As with the mean values presented above, all three size-fractions of PM exhibit similar patterns across the different transportation modes

Figure 4-2: Boxplots of PM 1 (top), PM 2.5 (middle) and PM 10 concentrations (bottom) measured during the four transport modes and at the background site (BG) averaged across the entire dataset Boxes and thick horizontal line represent the 25 th to 75 th

0 20 40 60

Comparison with other, similar studies reveals that PM2.5 values in the present study are the low end of values found elsewhere The range observed on the various transport modes are most similar to those of recent studies in Sydney (Knibbs and de Dear, 2010) and Barcelona (de Nazelle et al., 2012), with values ranging from

20 – 35 µg m-3 (see Table 2-2) However, the variability across transport modes between studies is different One possible explanation is the ventilation setting of the vehicles, which will be explained in more detail in Chapter 5

Ratios of PM1 to PM2.5 and PM2.5 to PM10 were calculated as a measure of the proportion of fine to coarse-particles (Table 4-4) The ratios are very high, demonstrating that fine particles are the dominant component of PM in the Singapore transport microenvironment This result is consistent with the close proximity to vehicular exhaust which emits particles of < 0.1 µm diameter (Lighty et al., 2000) The ratios at BG are similarly high indicating that that most of the pollution measured

at background was of local origin, since the particles have not had time to grow to larger sizes in the atmosphere

Table 4-4: PM 1 /PM 2.5 and PM 2.5 /PM 10 ratios for each transport mode and at the

background site averaged across the entire dataset

Table 4-5: Results of multiple-comparisons tests for effect of transport mode on PM 1

Table 4-6: Same as Table 4-5 but for PM 2.5

Table 4-7: Same as Table 4-5 but for PM 10

4.1.2 Particle number concentration (PN)

The PN data showed slightly different variability across transport modes compared to

PM (Figure 4-3) The Walk mode exhibited the highest mean concentration (44,490 #

cm-3), followed by the Taxi (30,882 # cm-3) and Bus modes (28,916 # cm-3) The

experienced for the MRT mode is also fairly consistent with a small standard deviation (2,625 # cm-3) The variability observed for the Taxi mode is considerably larger than for other transport modes, with a SD of 13,799 # cm-3 Both the highest and lowest PN values recorded during the entire sampling period were also taken on the Taxi mode (See Appendix D.1)

Figure 4-3: Boxplots of PN measured on the four modes of transport and background

site (BG) For explanation of boxplot symbols see Figure 4-2

The PN concentrations observed in this study were lower than those reported for other cities Mean values ranged from 14,418 # cm-3 on the MRT mode journeys

to 44,490 # cm-3 on the Walk mode In comparison, commonly reported mean PN concentrations in the literature range from upwards of 30,000 # cm-3 (see Table 2-2)

de Nazelle et al (2012) even observed a geometric mean of 117,600 # cm-3 for car commutes in Barcelona In comparison, the present study found a mean of 30,882 #

cm-3 for the Taxi mode Only Quiros et al (2013a) found relatively lower values of

PN in a residential neighbourhood in Santa Monica, USA with concentrations inside cars as low as 3,120 # cm-3 with closed windows The low PN for MRT mode trips is

20 40 60

consistent with the mean value calculated by Knibbs et al (2011) for powered trains

electric-The Walk mode still displays the greatest difference to background levels, with a ratio of 2.0 (Table 4-2) This is followed by the Taxi (1.47), Bus (1.32), and MRT mode (0.6) modes These results imply a strong influence of vehicular emissions on UFP levels

The Kruskal-Wallis test for PN revealed a statistically significant difference amongst the four transport modes and the background site (p-value < 0.05) (Table 4-3) Unlike the results for PM, the multiple comparisons tests revealed that there was

a statistically significant difference between all transportation modes except between the Bus and Taxi modes (Table 4-8) The Taxi mode is also the only transport mode that does not reveal a statistically significant difference from the background site for

PN

Table 4-8: Results of multiple-comparisons tests for effect of mode on PN concentrations

4.1.3 Active surface area (ASA), particle-bound polycyclic aromatic

hydrocarbons (pPAH), pPAH to ASA (PC/DC) ratio, and diameter of average surface ( Dave,S)

The Taxi mode exhibited the highest mean concentrations of ASA at 140 mm2 m-3, followed by Walk (137 mm2 mm-3), Bus (133 mm2 m-3), and MRT (114 mm2 m-3)

observed an extremely high GM (266 mm m ), which might have raised the mean value The Kruskal-Wallis test revealed a statistical significant difference amongst the four modes (p-value < 0.05) (Table 4-3), and multiple-comparisons tests show that only MRT mode trips are significantly different from the other three modes (Table 4-9), which are similar as suggested by the great degree of overlap of the inter-quartile ranges (IQR) (Figure 4-4)

Figure 4-4: Boxplots of ASA measured on the four modes of transport For explanation

of boxplot symbols see Figure 4-2

Table 4-9: Results from multiple-comparisons tests for effect of mode on ASA

100 150 200 250

similar ranging from 79 – 99 ng m , but the MRT mode exhibited a considerably lower mean value of 30 ng m-3 This was confirmed by the multiple comparisons test which revealed statistically significant differences between MRT and the other three transport modes (Table 4-10) This result agrees with present understanding regarding pPAH formation through incomplete combustion (Ravindra et al., 2008) Since the MRT is powered by electricity, no major combustion sources were encountered on MRT mode trips, resulting in low pPAH values Similar to the ASA data, there was

no statistically significant difference between the Bus, Taxi and Walk modes (Table 4-10)

Figure 4-5: Boxplots of pPAHs measured on the four modes of transport For

explanation of boxplot symbols see Figure 4-2

Table 4-10: Results from multiple-comparisons tests for effect of mode on pPAHs

Only few studies have measured ASA and pPAHs in the transport microenvironment Velasco et al (2004) measured pPAHs and ASA concentrations

in different outdoor, indoor, and street environments of Mexico City, Mexico, and found mean ASA and pPAHs concentrations ranging from 122 – 702 mm2 m-3 and 17 – 582 ng m-3, respectively in the outdoor environments The mean ASA concentrations found in the present study are similar to those found in Mexico City roads with traffic lights at regular intervals and a mix of vehicle types Such road conditions are similar to those found in the present study The pPAHs values in the present study are much lower than those observed by Velasco et al (2004) (average

173 ng m-3) A possible explanation is the much lower proportion of diesel-fuelled vehicles in Singapore’s vehicular fleet (Land Transport Authority, 2014c)

The relationship between pPAHs and ASA has been found to vary depending

on the presence of nuclei mode particles (Bukowiecki et al., 2002) All measured pPAHs and ASA values were plotted against each other for each of the transport modes (Figure 4-6) The pattern of the scatter extends along the x-axis, which suggests the presence of nuclei mode particles, which are typically composed of non-photoemitting material leading to low pPAHs values (Bukowiecki et al., 2002) However, this distribution can also be explained by particles that have low surface

concentrations of PAH (Bukowiecki et al., 2002) The low r 2 values suggest a wide range of particles sampled in terms of surface chemistry suggesting that particles measured in the present study are a mix of freshly emitted and aged particles Strong relationships were not observed because complete trips encompassed both indoor and outdoor environments which are influenced by and have varying proximity to combustion sources More useful is the analysis of the pPAHs and ASA relationship

in outdoor route sections which directly experience emissions from combustion sources, which is presented in Section 4.2

The mean PC/DC ratios for the three on-road transport modes are fairly low, ranging from 0.59 – 0.74 ng mm-2 (Table 4-11) These values are slightly higher compared to those measured by Ott and Siegmann (2006) on an arterial highway in California, USA, (mean: 0.42 – 0.58 ng mm-2) and by Velasco et al (2004) on roads

in Mexico City (0.53 ng mm-2) The ratio is lowest on the underground MRT trips at 0.27 ng mm-2, which can be explained by the lack of combustion sources leading to extremely low pPAHs values

Figure 4-6: Correlation between pPAHs and ASA for the entire dataset for each

Table 4-11: Mean (SD) PC/DC ratio and D ave,S for four transport modes (N = 23 for Bus,

Taxi, and MRT, N = 22 for Walk) Transport mode PC/DC (ng mm -2 ) D ave,S (nm)

combustion sources in indoor areas

In addition to exhibiting the highest PM, PN and pPAH concentrations, the

Walk mode also exhibits the smallest D ave,S of 31.28 nm This is followed by the Bus (38.65 nm) and Taxi modes (39.30 nm) The largest average particle diameter of 50.54 nm is observed on the MRT mode; however, this value may be an underestimation due to the lack of combustion sources during MRT mode trips

(Bukowiecki et al., 2002) The D ave,S indicates the approximate age of particles Based

on the results, the Walk (MRT) mode experiences the largest (smallest) amount of freshly emitted particles

Following the work by Bukowiecki et al (2002), the PC/DC ratio and D ave,S

were also plotted against each other (Figure 4-7) Combined with PN concentrations, the distribution of points in these plots can indicate the presence of: (i) nuclei mode particles (low pPAH, low or high ASA, and high PN), (ii) accumulation mode particles (high pPAH, high ASA, and low or high PN), or (iii) accumulation mode particles that are covered in a layer of chemicals not measured by the PAS sensor (low pPAH, low or high ASA, and low PN) (Bukowiecki et al., 2002) Along with the plots in Figure 4-6, the clustering of points close to the origin in Figure 4-7 indicates the presence of a nuclei mode in the observations, suggesting the strong influence of freshly emitted particles Data for the MRT mode are also presented However, as noted above, the interpretation of these results must be done carefully since the majority of the particles are unlikely to be emitted from combustion sources

Figure 4-7: PC/DC ratio versus D ave,S for each transport mode

4.1.4 Black carbon (BC)

Like for other pollutant metrics, the Walk mode exhibits the highest mean BC concentrations while the MRT mode exhibits the lowest (Figure 4-8) The mean values for the Bus and Taxi modes are very similar (Bus: 4.416 µg m-3 and Taxi: 4.368 µg m-3) Although similar to the PM and PN data, statistical testing revealed a significant difference between BC measured on the Walk mode and the three vehicular transport modes (Table 4-12) The Bus and MRT modes do not appear to be significantly different from the Taxi mode

Figure 4-8: Boxplots of BC measured on the four modes of transport For explanation of

boxplot symbols see Figure 4-2

Table 4-12: Results from multiple-comparisons tests for effect of mode on BC

of the pollutants The highest correlation was found between BC and pPAHs on the Walk mode (Spearman correlation 0.58) Both CO and ASA were also found to be poorly correlated with BC regardless of transport mode (CO: -0.03 – 0.02 and ASA: 0.07 – 0.14), which could point to different control processes between these pollutants Additionally, the low correlations suggest that CO is not a suitable tracer for vehicle activity in Singapore All three size-fractions of PM appear to be

0 5 10 15 20

moderately well correlated with BC (0.34 – 0.46) except on the MRT mode, again likely due to the different emission sources in subways Particle number concentrations were fairly well correlated with BC on the Bus and Walk modes (0.43 and 0.41, respectively), but only moderately correlated for the MRT and Taxi modes (0.29 and 0.26, respectively) The correlations in the present study are similar to those found by de Nazelle et al (2012) (PN: 0.30, PM2.5: 0.39) They suggested that the low correlations observed are due to the presence of different vehicle types which emit different proportions of pollutants For example, BC is emitted primarily from diesel-fuelled vehicles, whilst CO and UFPs may be predominantly emitted by gasoline-fuelled vehicles (de Nazelle et al., 2012)

Table 4-13: Results of Spearman rank correlation between BC and other metrics on the

four transport modes

4.1.5 Carbon monoxide

The mean concentration levels for CO were extremely low across all modes of transport and at the background site, ranging from 0.5 ppm on the MRT mode to 1.0 ppm on the Taxi mode (Table 4-1) These values are comparable to those observed in London (Kaur et al., 2005b) and Barcelona (de Nazelle et al., 2012) (see Table 2-2) However, there is great variation in the measurements, with many outliers (Figure 4-9), and the corresponding SDs are of similar magnitude as the mean values (Table 4-1) As described in the data quality section in Chapter 3, approximately 17% of data from the CO measurer were reported as zero values and removed from analysis under the assumption that the concentrations fell below the sensor detection limits This might have introduced a stronger positive skew into the data

Mean CO concentrations experienced on the Bus, MRT and Walk modes are low compared to the background site (0.88, 0.53, and 0.69, respectively) Only the Taxi mode exhibited marginally higher CO levels, with a ratio of 1.09 These low readings across the transport modes suggest that Singapore has managed to curb CO emissions from vehicles as recommended by the 1970 WHO report for Singapore (See Chapter 1 for details)

The results from the multiple comparisons tests indicate a statistically significant difference between the MRT and the three other transport modes, as well

as between the Taxi and Walk modes (Table 4-14) There was no statistically significant difference between data measured at the background site and on the four different modes of transport

Figure 4-9: Boxplots of CO measured on the four modes of transport and at the background site For explanation of boxplot symbols see Figure 4-2

Table 4-14: Results from multiple-comparisons test for effect of mode on CO

4.2 Spatial variation within transport modes

As described in Chapter 3, the door-to-door routes for each transport mode were split into several sections for a more detailed analysis of the particle pollution to which commuters are exposed Different amounts of time were spent within the different in-vehicle, indoor, and outdoor spaces depending on the transport mode (Table 4-15) Average durations for an entire journey on each mode of transport were 14.6 (Bus), 14.3 (MRT), 18 (Taxi), and 25.8 (Walk) minutes, respectively This, combined with

0 1 2 3 4

The following sections describe the spatial variation of pollutant concentrations observed on Bus, MRT, Taxi, and Walk mode journeys in greater detail In addition to descriptive statistics, time-series plots and maps of PM2.5 and PN concentrations of one trip are presented to better visualise the spatial variability The trips most representative across all measurements for each transport mode are used as examples

Table 4-15: Mean time spent in each section for all measurements presented in minutes

and percentage of overall trip

Sections Time spent a (min) Time spent a (%)

4.2.1 Bus

The spatial variation of PM2.5 and PN concentrations for a door-to-door trip on the Bus mode on 10 June 2013 is shown in Figures 4-10 and 4-11 Clear differences are visible in different sections of the trip, although the patterns for PM2.5 and PN are slightly different There is a noticeable increase in PM2.5 concentration towards the Underpass section, however there is no corresponding rise in PN except for a large spike halfway through the Underpass (Figure 4-10) The effect of opening and closing bus doors can be observed in the increases followed by a gradual decline of PN concentrations inside the Bus Similar jumps in PM2.5 concentrations were less distinct

Figure 4-10: Time-series of PM 2.5 and PN concentrations during the Bus mode trip on 10 June 2013 Vertical dashed lines delineate the different sections of the trip Periodic

Sidewalk Mall Underpass Sidewalk/

Figure 4-11: Spatial variation of PM 2.5 (top) and PN (bottom) concentrations during the

Bus mode trip on 10 June 2013

Mean PM1, PM2.5, PM10, and PN concentrations measured in the five sections

of Bus mode trips were slightly higher than at the background site, with outdoor sections (Bus-stop and Sidewalk) exhibiting much higher mean concentrations (Figure 4-12) Except for concentrations inside the Bus, CO concentrations in the other sections were lower than at the background site as well (Figure 4-12) This agrees with the trip-averaged results described in Section 4.1

Across the various sections of the Bus mode route, the Bus section exhibits the smallest mean values for PM1, PM2.5, PM10, and ASA (Table 4-16) However, in-vehicle CO concentrations, as well as the PM2.5/PM10 and PC/DC ratios were elevated compared to the other sections (Table 4-16 and Table 4-17) This might be due to the regular introduction of freshly emitted particles into the vehicle when bus doors open and close to pick up passengers, which is illustrated in Figure 4-10

The indoor locations of the Mall and Underpass generally observed lower

values of PN and pPAHs, and correspondingly smaller (larger) PC/DC (D ave,S) (Figure 4-12 and Table 4-17) However, despite having the lowest mean PN concentrations, the Underpass exhibited the highest mean concentrations of PM1, PM2.5, and PM10 (40,

40, and 50 µg m-3, respectively) (Table 4-16) The ratio of PM2.5/PM10 is also lowest

in the Underpass, with a value of 0.80 Taken together, this indicates a larger proportion of coarse particles, and a smaller amount of UFPs than in the other sections of Bus mode trips

Figure 4-12: Boxplots of 8 pollutant metrics in different sections of the Bus mode trips For explanation of boxplot symbols see Figure 4-2 Mean background site concentrations,

where available, are indicated as dashed line on the respective panel

Bus MallUnderp

ass Bus-stop Sidewalk

The outdoor areas (Bus-stop and Sidewalk) experienced higher mean PN concentrations (51,436 and 39,895 # cm-3, respectively) and smaller D ave,S values (35.18 and 33.96 nm, respectively), likely the result of the close proximity to vehicular exhaust (Table 4-16 and Table 4-17) This is also supported by the comparatively higher concentrations of BC in these locations (Bus-stop: 8.054 µgm-3and Sidewalk: 6.603 µg m-3) The calculated D ave,S in these sections are also consistent with that for the similar outdoor Walk mode (33.68 nm) In addition, the Bus-stop observed the highest mean concentrations of ASA and pPAHs of 195 mm2

m-3 and 124 ng m-3, respectively, as well as the second highest PM10 concentrations of

43 µg m-3 Statistical testing revealed that the Bus-stop and Sidewalk areas were significantly different from the other sections in terms of PN (Table 4-19 and Appendix D.2.1)

Table 4-16: Mean (SD) of pollutant metrics for different sections of the Bus mode

journeys (N = 23) Section PM 1 (µg m -3 ) PM 2.5 (µg m -3 ) PM 10 (µg m -3 ) PN (# cm -3 )

Table 4-17: Mean PM 1 /PM 2.5, PM 2.5 /PM 10, PC/DC ratios and D ave,S in different sections of

Bus mode journeys (N = 23) Section PM 1 /PM 2.5 PM 2.5 /PM 10 PC/DC (ng mm -2 ) D ave,S (nm)

Table 4-18: Results of Spearman rank correlation between BC and other metrics in the

different sections of Bus mode trips

Similar to the analysis for entire trips (see Section 4.1.3), the pPAH and ASA

in the outdoor areas of Bus mode trips were plotted against each other The results are inconclusive with only one or two days of measurement that exhibited a more distinct

presence (r 2 > 0.5) of either a nuclei mode (horizontal to negative slope) or an accumulation mode (positive slope) (Figures 4-13 and 4-14) For the Bus-stop section, this lack of a distinct pattern may, in part, be due to the small number of sampling points Most Sidewalk measurements exhibit a horizontal to negative slope, indicating the presence of a nuclei mode (Figure 4-14) However, the generally poor

r 2 values suggest the presence of particles of different ages which have already gone through growing processes and chemical reactions that do not contribute UFPs in the nuclei mode

Figure 4-13: pPAHs and ASA data collected at Bus-stop sections plotted against each

other for each day of sampling Linear regression lines and the r 2 of the relationship are

Figure 4-14: pPAHs and ASA data collected at Sidewalk sections during Bus mode trips

plotted against each other for each day of sampling Linear regression lines and the r 2 of

the relationship are also plotted

4.2.2 MRT

The MRT mode trips were split into five sections as well (Figure 4-15) Since most of the travelling was done underground, only a short amount of time (< 1 minute) was spent outdoors on the Sidewalk where the start and end points of the sampling route were located (Figure 4-15) This meant that there were no GPS readings for MRT trips, thus the spatial variation for MRT mode was not plotted Distinctions between sections of the MRT journey are not as well-defined in terms of PM2.5 There is a gradual rise in PM2.5 concentrations further indoors, reaching a maximum inside the Train However, there is no similar pattern in the PN time-series, although concentrations inside the Train are higher than at the Platform section This implies a different group of sources or processes controlling the amount of particles for MRT mode trips

Figure 4-15: Time-series of PM 2.5 (top) and PN (bottom) concentrations during the MRT mode trip on 20 May 2013 Vertical dashed lines delineate the different sections of the

Except for the Station areas, mean PM1, PM2.5, and PM10 measured in all sections of the MRT mode are slightly higher than at the background site (Figure 4-16) In particular, PM concentrations appear to increase deeper inside the subway system, which agrees with the findings of Kim et al (2008) in the Seoul Metropolitan Subway However, with the exception of the Sidewalk, mean PN in different sections

of MRT journeys were well below background concentrations (Figure 4-16), contributing to the low MRT mode to background concentrations ratio of 0.6 presented above (Table 4-2) Mean PN concentrations observed at the Platform even dropped below 10,000 # cm-3 Concentrations in the Station and the Train are similarly low with mean readings just above 10,000 # cm-3 The relatively small SD further indicates that PN concentrations are consistently low in these spaces (Table 4-20) All spaces of the MRT mode journey exhibited mean CO concentrations that fell below background levels (Figure 4-16) However, the spread of data is extremely large, with standard deviations even larger than mean values (Table 4-16)

Unlike the Bus mode, the highest mean concentrations of PM1, PM2.5, and

PM10 were observed inside the vehicle (i.e Train) However, the second lowest mean

PN concentrations were also observed in this location (Table 4-20), indicating a higher proportion of larger particles and less UFPs In addition to the low PN concentrations, the Platform section also showed the lowest mean concentrations of ASA, pPAHs, BC, and CO (Table 4-20) The Station section exhibited similarly low values as the Platform section except for ASA concentrations, leading to a relatively

smaller D ave,S (Table 4-21) All sections of the MRT mode exhibited lower PC/DC

and larger D ave,S compared to the sections of Bus mode journeys, although as

mentioned, D ave,S values may be an underestimation due to the lack of combustion sources (Table 4-21)

Figure 4-16: Boxplots of the 8 pollutant metrics measured in the different sections of MRT mode trips For explanation of boxplot symbols see Figure 4-2 Mean background site concentrations where available are indicated as a dashed line on the respective

Train Mall Platform Station Sidewalk

Mean concentrations of BC were relatively low in all four indoor and vehicle sections of MRT journeys (ranging from 2.204 – 3.355 µg m-3) Only the outdoor Sidewalk section exhibited higher a concentration of 5.996 µg m-3 Additionally, all pollutant variables exhibited poor correlation with BC in all sections except the outdoor areas This further indicates that freshly emitted particles from vehicular combustion is not the major source of particle pollution encountered on MRT mode trips.

in-Only the Station and Train spaces exhibited a significant difference from the other sections in terms of PM1, PM2.5, and PM10 (see Appendix D.2.2) For PN, the Station and Train were the only two sections that were not significantly different The Sidewalk also appeared to be significantly different from the Platform, Station, and Train sections in terms of BC, again possibly due to the absence of exhaust emissions

in the indoor spaces

Table 4-20: Mean (SD) of measured pollutant metrics in different sections of MRT mode

Table 4-21: Mean PM 1 /PM 2.5 , PM 2.5 /PM 10, PC/DC ratios and D ave,S in different sections of

Table 4-22: Results of Spearman rank correlation between BC and other metrics in the

different sections of MRT mode trips

4.2.3 Taxi

The spatial variation within Taxi mode trips is clearly illustrated in Figure 4-17 and Figure 4-18 Both PM2.5 and PN move in tandem, decreasing inside the Mall and Taxi, and rising at Sidewalk and Taxi-stand (Figure 4-17), which points towards a common emission source This is also observed in the maps, with the Taxi-stand areas near the start and end points exhibiting higher PM2.5 and PN concentrations (Figure 4-18) Such clear distinctions between sections are not observed for all pollutants measured, although statistical testing revealed significant differences between spaces for all pollutant variables (Table 4-27) Similar to the Bus and Train, PM2.5 and PN concentrations decrease rapidly within the first 500 m of travel inside the Taxi (Figure 4-18) However, PM2.5 concentrations inside the Taxi exhibited short-lived spikes that are not observed in the PN measurements (Figure 4-17)

Figure 4-18: Spatial variation in PM 2.5 (top) and PN (bottom) concentrations during the

Taxi mode journey on 20 May 2013

Except for the in-vehicle section, all other parts of Taxi mode trips experienced higher mean concentrations of PM1, PM2.5, PM10 and PN than at the background site Similar to the Bus mode trips, outdoor areas in particular (Taxi-stand and Sidewalk) observed much higher concentration than the background site (Figure 4-19) These two outdoor areas also exhibited higher concentrations for the other variables (Table 4-24) Carbon monoxide concentrations are lower in the other sections of Taxi mode trips compared to inside the vehicle Again, CO concentrations exhibit a very large spread, with standard deviations on the same order as mean values

The Taxi sections exhibited mean CO concentrations of 1.6 ppm, which is higher than observed inside the Bus and Train sections Despite higher CO levels, the lowest mean concentrations of PM1, PM2.5, PM10, PN, ASA, and BC were found inside the Taxi (Table 4-24) The Bus and Taxi sections appear similar in terms of pollutant concentrations except for PN, which was found to be much lower in taxis than buses (18,901 # cm-3 and 27,775 # cm-3 respectively) The lower PN and ASA

values also led to a relatively larger D ave,S compared to the other sections (Table 4-25)

Figure 4-19: Boxplots of the 8 pollutant metrics measured in the different sections of Taxi mode journeys For explanation of boxplot symbols see Figure 4-2 Mean background site concentrations where available are indicated as a dashed line on the

Taxi Mall Taxi-stand Sidewalk

Table 4-24: Mean (SD) of pollutant metrics for different sections of Taxi mode journeys Section PM 1 (µg m -3 ) PM 2.5 (µg m -3 ) PM 10 (µg m -3 ) PN (# cm -3 )

Table 4-25: Mean PM 1 /PM 2.5 , PM 2.5 /PM 10, PC/DC ratios and D ave,S for different sections

of Taxi mode journeys

Table 4-26: Results of Spearman rank correlation between BC and other metrics in the

different sections of Taxi mode trips