COMMUTER EXPOSURE TO AEROSOL POLLUTION ON PUBLIC TRANSPORT IN SINGAPORE 1

COMMUTER EXPOSURE TO AEROSOL POLLUTION ON PUBLIC TRANSPORT IN SINGAPORE TAN SOK HUANG (B.Soc.Sci. (Hons), NUS) A THESIS SUBMITTED FOR THE DEGREE OF MASTER OF SOCIAL SCIENCES DEPARTMENT OF GEOGRAPHY NATIONAL UNIVERSITY OF SINGAPORE 2014 DECLARATION I hereby declare that this thesis is my original work and it has been written by me in its entirety. I have duly acknowledged all the sources of information which have been used in the thesis. This thesis has also not been submitted for any degree in any university previously. _________________________ Tan Sok Huang 30 December 2014 ii Acknowledgements To my advisors, family and friends, This thesis would not have been possible without your kind support, guidance and patience. Thank you so much for helping me and putting up with me during the past few years. iii Table of Contents Acknowledgements iii Table of Contents . iv Abstract vi List of Tables .viii List of Figures . xii List of Abbreviations xv Chapter 1. Introduction . 1.1 Human exposure to air pollution . 1.2 Singapore’s air quality 1.3 Objectives . 1.4 Thesis outline Chapter 2. 2.1 Estimating exposure 10 2.1.1 2.2 Literature Review 10 Measuring exposure to particle pollution 11 Particle pollution in the transport microenvironment . 16 2.2.1 Transport emissions 17 2.2.2 Spatial and temporal distribution of particles . 19 2.3 Personal exposure in the transport microenvironment 23 2.3.1 Summary of past results 26 Chapter 3. Methods . 33 3.1 Measurement Area and Study Period 33 3.2 Instrumentation . 37 3.3 Sampling design 43 3.3.1 Sampling route 43 3.3.2 Background measurement site 49 3.3.3 Instrument set-up and sampling procedures 50 3.4 Data quality control . 53 3.4.1 Pre-sampling procedures . 53 3.4.2 Data post-processing . 53 Chapter 4. Results . 57 4.1 Commuter exposure on door-to-door trips 59 4.1.1 Particulate matter mass concentrations (PM) 61 4.1.2 Particle number concentration (PN) 64 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) . 66 iv 4.1.4 Black carbon (BC) 72 4.1.5 Carbon monoxide 75 4.2 Spatial variation within transport modes 76 4.2.1 Bus 78 4.2.2 MRT 87 4.2.3 Taxi . 92 4.2.4 Walk 101 4.3 Dosage 105 4.3.1 Ventilation rates 105 4.3.2 Dosage results . 107 Chapter 5. Discussion . 111 5.1 Comparison across overall trips and background site . 111 5.2 Spatial variation of pollutant concentrations 115 5.2.1 Bus-stops and Taxi-stands 118 5.2.2 In-vehicle concentrations 119 5.3 Dosage 122 Chapter 6. Conclusion 126 6.1 Summary of key findings 126 6.2 Final notes and suggestions for future research 129 References . 132 Appendix A DustTrak Calibration 141 Appendix B Description of a trip on each transport mode 147 Appendix C Anderson-Darling Test for Normality . 149 Appendix D Supplementary results . 154 v Abstract Brief periods of exposure to high concentrations of air pollution may have significant health impacts. In cities, a large proportion of exposure to airborne pollutants, in particular, particulate matter, is likely experienced during daily commuting trips due to the proximity to a number of pollution sources (vehicular traffic, industry, construction sites, etc). A better understanding of the variability in pollutant concentrations across available transport modes is important for commuters and authorities. Unfortunately, personal exposure to particle pollution in the transport microenvironment of Singapore to date has not been well documented. The present study analyses exposure concentrations of particulate matter (PM10, PM2.5), particle number (PN), black carbon (BC), carbon monoxide (CO), particle-bound polycyclic aromatic hydrocarbons (pPAH), and active surface area (SA) measured along a selected route in the commercial shopping district of Singapore. Portable instruments capable of real-time monitoring were used during door-to-door trips on three different modes of public transport (bus, taxi, MRT) and walking. Simultaneous measurements of PM (various sizes), PN, and CO were taken at a local park to capture the background concentrations. In addition to exposure concentrations, inhaled dose (dosage) was also estimated. Except for CO, exposure concentrations of all pollutant metrics were highest during walking, and lowest on MRT trips. Mean PM2.5 concentrations observed during bus, MRT, taxi and walking modes were 1.17, 1.09, 1.11 and 1.50 times higher than at the background site. PN exhibited a similar trend except for the MRT mode which showed lower average values than at the background site (ratio of 0.6). In-vehicle concentrations for buses and taxis were also lower than those found in the literature, which may be attributed to differences in local driving behaviour, fleet composition and age, and ambient pollution. Such differences also highlight the vi importance of monitoring pollutant exposure under local conditions. After taking into account the effect of inhalation and travel duration in the calculation of dosage, differences between transport modes increased by a factor of two. Mean dosages of PM2.5 and PN on the Walk mode were 2.5 – times higher than that experienced on the other three transport modes. vii List of Tables Table 1-1: Air quality standards for US-EPA, WHO and NEA and annual air quality in Singapore in 2012 for the criteria pollutants. Singapore’s air quality in 2013 is provided in parentheses for comparison purposes. Table 2-1: Summary of transport modes and metrics measured for selected studies 25 Table 2-2: Summary of PM2.5, UFP and CO results for the studies listed in Table 2-1. 27 Table 2-3: Comparison of PM2.5 exposure concentrations with inhaled dose from two of the studies listed in Table 2-1 31 Table 3-1: Mean ambient T and RH for the entire Singapore island and 24 h-averaged PSI based on PM10 and PM2.5 concentrations reported at 16:00 h for the Central region of Singapore during the entire fieldwork campaign. (Data from NEA website) 34 Table 3-2: Sensors employed in the present study and their measurement characteristics. Sources: EcoChem Analytics (2005), Langan Products (2006), Onset Computer Corporation (2011), Polar Electro Oy (2013), TSI Incorporated (2007, 2009, 2010) . . 39 Table 3-3: Characteristics of the indoor and outdoor spaces associated with each transport mode. Values of T, RH and time spent in each section are means of 23 transects. 47 Table 3-4: Total number (N) of trips sampled for each mode of transport and the number of trips used in the final analysis after discarding measurements affected by transboundary pollution, precipitation or technical problems. 54 Table 3-5: Fraction of total data used for analysis after quality control including removal of suspicious data based on field notes and zero values. 55 Table 4-1: Mean (SD) of pollutant metrics from all trips for transport modes and measured at the background site. (N = 23 for Bus, Taxi, and MRT, N = 22 for Walk) . 59 Table 4-2: Transport mode to BG ratios from all trips for pollutant metrics measured in both environments. . 60 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 . 60 Table 4-4: PM1/PM2.5 and PM2.5/PM10 ratios for each transport mode and at the background site averaged across the entire dataset. . 63 Table 4-5: Results of multiple-comparisons tests for effect of transport mode on PM1 concentrations . 64 Table 4-6: Same as Table 4-5 but for PM2.5. . 64 Table 4-7: Same as Table 4-5 but for PM10. . 64 Table 4-8: Results of multiple-comparisons tests for effect of mode on PN concentrations . 66 viii Table 4-9: Results from multiple-comparisons tests for effect of mode on ASA concentrations . 67 Table 4-10: Results from multiple-comparisons tests for effect of mode on pPAHs concentrations . 68 Table 4-11: Mean (SD) PC/DC ratio and Dave,S for four transport modes. (N = 23 for Bus, Taxi, and MRT, N = 22 for Walk) . 71 Table 4-12: Results from multiple-comparisons tests for effect of mode on BC concentrations . 73 Table 4-13: Results of Spearman rank correlation between BC and other metrics on the four transport modes. 74 Table 4-14: Results from multiple-comparisons test for effect of mode on CO concentrations . 76 Table 4-15: Mean time spent in each section for all measurements presented in minutes and percentage of overall trip. 77 Table 4-16: Mean (SD) of pollutant metrics for different sections of the Bus mode journeys. (N = 23) 82 Table 4-17: Mean PM1/PM2.5, PM2.5/PM10, PC/DC ratios and Dave,S in different sections of Bus mode journeys. (N = 23) . 83 Table 4-18: Results of Spearman rank correlation between BC and other metrics in the different sections of Bus mode trips. 83 Table 4-19: Results of the Kruskal-Wallis test for effects of the different sections on pollutant concentrations during Bus mode trips. H = test statistic, df = degrees of freedom. 83 Table 4-20: Mean (SD) of measured pollutant metrics in different sections of MRT mode journeys. . 90 Table 4-21: Mean PM1/PM2.5, PM2.5/PM10, PC/DC ratios and Dave,S in different sections of MRT mode journeys . 91 Table 4-22: Results of Spearman rank correlation between BC and other metrics in the different sections of MRT mode trips . 91 Table 4-23: Results of Kruskal-Wallis test for effect of the different sections on pollutant concentrations during MRT mode journeys. H = test statistic, df = degrees of freedom 91 Table 4-24: Mean (SD) of pollutant metrics for different sections of Taxi mode journeys. . 96 Table 4-25: Mean PM1/PM2.5, PM2.5/PM10, PC/DC ratios and Dave,S for different sections of Taxi mode journeys. . 96 Table 4-26: Results of Spearman rank correlation between BC and other metrics in the different sections of Taxi mode trips. . 96 Table 4-27: Results of Kruskal-Wallis test for effect of the different sections on pollutant concentrations during Taxi mode journeys. H = test statistic, df = degrees of freedom. . 97 Table 4-28: List of taxi models sampled. The vehicles’ age was obtained from the drivers. 100 ix Table 4-29: Maximum, minimum and mean HR and VE for different sections of the four transport modes for all measurements (N = 23 for Bus, MRT, and Taxi, N = 22 for Walk). 106 Table 4-30: Inhaled dose by mode and section for PM2.5 and PN based on all data. 109 Table 4-31: Ratios of PM2.5, PN, BC, and pPAH concentrations and inhaled dose between Bus, MRT, and Taxi modes and Walk mode. 110 Table D-1: Descriptive statistics of measurements on the four modes of transportation and at BG. Maximum and minimum are the highest and lowest datapoints recorded throughout the sampling. GM is the geometric mean of all measured data-points. Mean (SD) is the arithmetic mean (standard deviation) of the GM for each trip 154 Table D-2: Descriptive statistics of measurements in different sections within Bus mode trips. See Table D-1 caption for details. . 155 Table D-3: Same as Table D-2 but for MRT mode. See Table D-1 caption for details. 157 Table D-4: Same as Table D-2 but for Taxi mode trips. See Table D-1 caption for details. 158 Table D-5: Results from multiple-comparisons tests for effect of different sections on PM1 concentrations on Bus mode trips . 160 Table D-6: Same as D-5 but for PM2.5. . 160 Table D-7: Same as D-5 but for PM10 . 160 Table D-8: Same as D-5 but for PN. . 160 Table D-9: Same as D-5 but for ASA. 160 Table D-10: Same as D-5 but for pPAHs. 161 Table D-11: Same as D-5 but for BC 161 Table D-12: Same as D-5 but for CO. 161 Table D-13: Results of multiple-comparisons tests for effect of different sections on PM1 concentrations on MRT mode trips . 161 Table D-14: Same as D-13 but for PM2.5. . 162 Table D-15: Same as D-13 but for PM10 . 162 Table D-16: Same as D-13 but for PN. . 162 Table D-17: Same as D-13 but for ASA. 162 Table D-18: Same as D-13 but for pPAHs. 162 Table D-19: Same as D-13 but for BC 163 Table D-20: Same as D-13 but for CO. 163 Table D-21: Results from multiple-comparisons tests for effect of different sections on PM1 concentrations on Taxi mode trips. 163 Table D-22: Same as D-21 but for PM2.5. . 163 Table D-23: Same as D-21 but for PM10 . 163 Table D-24: Same as D-21 but for PN. . 164 Table D-25: Same as D-21 but for ASA. 164 x Active surface area concentration The Diffusion Charging Sensor (DC 2000CE, EcoChem) measures the ASA of particles. The DC uses a corona discharge to produce negative oxygen ions. These negatively charged ions diffuse through the carrier gas and transfer the electrical charge to the surface of the particles they come into contact with. Thus only the part of the particle surface accessible by direct diffusion, the ASA, will be measured (Bukowiecki et al., 2002). These electrically charged particles are collected onto an electrically insulated particle filter, and the resultant electric charge is converted to a voltage signal (EcoChem Analytics, 2005). Particle-bound polycyclic aromatic hydrocarbons concentration The Photoelectric Aerosol Sensor (PAS 2000CE, EcoChem) measures the mass concentration of pPAH in the air based on the photoelectric ionization of particles. The PAS employs a pulsed UV light with a wavelength of 207 nm to ionize the PAH adsorbed onto the surfaces of particles while gas molecules and other aerosols remain neutral (Burtscher, 1992). The positively-charged PAH are collected on a filter connected to an electrometer which produces a current reading. Although other chemicals and materials may also exhibit strong photoemission, the PAS responds only to pPAH due to the particular UV light wavelength used (Dhaniyala et al., 2011). A limitation of the PAS is that it only measures the total concentration of pPAH and does not provide details on the PAH speciation, which requires other techniques. Black carbon mass concentration The miniature black carbon monitor or microaethalometer (AE51, Aethlabs) samples and analyses the mass concentration of BC particles in the air. It is essentially a small version of an aethalometer, which measures the attenuation of light transmitted through the particles that are continuously collected onto a filter (TSI Incorporated, 2010). This incremental change in light passing through the filter during a sample period (ti to ti+1) is converted to a BC concentration through numerical factors specific 41 to each instrument (Hagler et al., 2011). One known problem is that at high logging intervals or when measuring air parcels with low BC values, the incremental change in light may be affected by instrumental or optical noise, leading to unchanging or negative values in the dataset. The Optimized Noise Reduction Averaging (ONA) algorithm was developed to resolve this issue (Hagler et al., 2011), and was applied for this thesis. A second correction was also applied to account for the decreasing instrument sensitivity associated with increased filter load as more particles get deposited on the filter. Concentrations were adjusted using the empirical relationship derived by Kirchstetter and Novakov (2007) based on the instrument-reported attenuation coefficient. Carbon monoxide The Langan CO Measurer (T15n, Langan Products Inc.) is a passive gas sampler measuring CO in ppm. The instrument combines an electrochemical sensor manufactured by City Technology, Ltd., and a data-logger manufactured by Onset Computer Company in a small portable package (Langan Products, 2006). The CO readings were corrected with temperature from the HOBO loggers according to procedures recommended by the instrument manufacturer (Langan Products, 2006). Heart rate The heart rate monitor (RCX3, Polar Electro Oy) used in this study consists of a wristwatch that doubles as a receiver and a strap to be worn around the individual’s chest with an attached transmitter. The chest strap is an electrode that detects the heart’s electrical signals and transmits the data as beats per minute (bpm) (Polar Electro Oy, 2013). This electrode has to be wet to be activated. Although some of the modes of transport are less active than others, there were no issues of dry skin which led to poor electrode contact. 42 In addition to the calibrations and corrections necessary to ensure instrument accuracy, there are some general issues with using these instruments for such a study. In general, most sensors are sensitive to mechanical shock or vibration, which is a concern for mobile measurements. Some of the issues particular to each sensor have been described above. Section 3.4 in the latter part of this chapter describes the steps taken to ensure good data quality. 3.3 Sampling design 3.3.1 Sampling route To capture the spatial and temporal variations on door-to-door commutes for the four transportation modes, a 1.5 km route between two popular shopping malls in the Orchard Road district was selected. Each trip on a particular transport mode had common start and end points, beginning and ending at the entrance of the two shopping malls (Figure 3-4). The sampling route encompasses bus-stops and three MRT stations in total, including the start and end points. The routes for all four modes of transport were nearly identical, except for the MRT mode and the Taxi mode. Both the Bus and Walk mode were located along the main thoroughfare (labelled Main route in Figure 3-4) The MRT is located underground with stations linked to the basement levels of the shopping malls (not labelled on map), and the Taxi mode measurements involved travel on secondary streets due to regulations on pick-up and drop-off locations (Figure 3-4). 43 Figure 3-4: Route selected to evaluate exposure concentrations on four common transport modes within the commercial Orchard Rd district, Singapore. Bus and Walk mode trips were taken along the main route (dashed line), and Taxi mode trips included travel along secondary roads (dotted line). MRT mode trips were entire underground and not pictured. Also shown is the location of the background measurement site. (Source: Google Maps). Simultaneous measurements on multiple transport modes were not possible due to equipment constraints. Thus a shorter route was selected to enable consecutive measurements on all four transport modes within a 2-hour period. The short distance and time period could also better accommodate the battery life of the instruments. Although not representing the typical daily commute between home and workplace, the selected route encompasses the variety of microenvironments within the transportation network that commuters will encounter. This is also a common travelling route that visitors to the area may take. Thus the present measurements are expected to be representative of the pollution levels that commuters experience. 44 Sections within transport modes To investigate the spatial variation of pollutant exposure in the transport microenvironment, the door-to-door journeys for each transport mode were split into separate sections for analysis. The journey on each of the transport modes involved travelling through a series of in-vehicle, indoor, and outdoor spaces some of which are unique to that mode (Figure 3-5 and Figure 3-6). As described in the previous chapter, such spaces are affected by unique emission sources and conditions that can lead to distinct pollutant concentrations. Table 3-3 provides a brief description of the indoor and outdoor spaces associated with each transport mode, as well as their typical temperature and humidity, and time spent within each section. There are minor variations in actual distance travelled each day due to individual behaviour. Similarly, the travel time for each mode varied each day due to different waiting times. The variation in length of time spent for each trip has important implications regarding the duration of exposure and eventual dosage. 45 Figure 3-5: Indoor spaces covered when using specific transport modes in this study: (a) mall, (b) MRT platform, (c) MRT station , and (d) underpass. Figure 3-6: Outdoor spaces covered when using specific transport modes in this study: (a) bus-stop, (b) Taxi-stand , (c) sidewalk near the start point of the Walk mode and (d) sidewalk near the end point of the Walk mode. 46 Table 3-3: Characteristics of the indoor and outdoor spaces associated with each transport mode. Values of T, RH and time spent in each section are means of 23 transects. Sections T (°C) RH (%) Time spent (min) Description Bus In-vehicle Bus 28.0 38.0 6.6 Air-conditioned bus cabin Mall 28.4 58.4 0.9 Air-conditioned indoor area, connected to the MRT Station, Sidewalk and Underpass sections of the route Underpass 28.4 45.9 2.3 Air-conditioned underground passage with an escalator directly to street level (Figure 3-5b); ongoing construction work throughout the sampling period Bus-stop 28.6 66.3 1.2 Sheltered area along the Sidewalk (Figure 3-6a) Sidewalk 27.7 65.1 3.6 Paved pedestrians walking area (Figure 3-2a & b) Indoor Outdoor MRT In-vehicle 27.4 42.9 4.1 Air-conditioned train carriage running on underground tracks Mall 26.8 47.9 1.5 Same as Mall above Platform 27.8 45.3 2.6 Underground air-conditioned area located one floor below the Station (Figure 3-5b) Station 28.2 49.9 4.6 Underground air-conditioned area accessible from the Mall (Figure 3-5c); part of the area is taken up by stalls selling snacks 28.5 65.7 MRT Indoor Outdoor Sidewalk 1.54 Same as Sidewalk above Taxi In-vehicle Taxi 29.4 31.5 7.6 Air-conditioned car cabin, recirculation 30.6 45.1 0.9 Same as Mall above Sidewalk 29.1 56.5 4.2 Same as Sidewalk above Taxi-stand 30.3 61.1 5.3 Sheltered area along the Sidewalk next to the road (Figure 3-6b) Indoor Mall Outdoor Walk Outdoor Sidewalk 30.9 59.0 24.9 Paved pedestrian walking area (Figures 31b, 3-5c & d) 47 The major difference between the in-vehicle, indoor, and outdoor sections is in relative humidity (RH) (Table 3-3). Air-conditioned in-vehicle and indoor spaces exhibited lower RH than outdoor areas. For the three vehicular transport modes, invehicle RH (31.5% – 42.9%) was lower than at shopping malls (45.1% – 58.4%). Relative humidity may be an important factor controlling the growth of particles. However, understanding this relationship is beyond the scope of the present study. In-vehicle spaces were defined from the point of entry onto the vehicle until exit, even if the vehicle doors were still open during those periods of time. Since all vehicles used in the present study were part of the public transport system, ventilation settings could not be controlled. This meant the samples are representative of the range of situations experienced by commuters. All vehicles were fully air conditioned and windows were closed throughout the trip. The majority of buses and taxis used in this study were diesel-fuelled, while the MRT trains are powered by electricity. For taxis, all drivers had set the ventilation to recirculate. Taxi drivers were also interviewed to find out the age of the vehicle. Bus drivers were not interviewed, thus there is no information on ventilation settings, bus model or age. There are no physical barriers obstructing airflow between different indoor and outdoor sections. However, clear variations in particle concentration were observed between adjacent spaces. For example, the MRT station is integrated into the basement of the shopping mall. The only barriers between these areas are m high gantries between the Mall and Station, and the Platform is accessible from the Station level by a flight of stairs (Figure 3-5). Similarly, there are no barriers between the Sidewalk and Bus-stops and Taxi-stands (Figure 3-6). Although these spaces appear to share a common ventilation system (or lack thereof for the outdoor locations), there appeared to be significant changes in pollutant concentrations in tandem with the movement across different spaces. A set of rules was developed to delineate these sections from each other. Transitions between Mall, Station and 48 Platform sections were taken to occur the moment the volunteers passed through the gantry (Mall–Station) and walked exactly half-way down the stairs (Station– Platform). Changes in the floor tiles were used to differentiate between the Mall and Underpass sections in the Bus modes. For outdoor locations, the Bus-stop and Taxistand sections are denoted to start from the moment volunteers join the queue for a bus or taxi respectively. 3.3.2 Background measurement site Data were collected at a background location to distinguish the pollution associated with the immediate surroundings and emission sources along the selected route from the urban pollution at ambient level (i.e. district scale). Previous studies have typically chosen rooftops or local parks as their background site (e.g. Kaur et al., 2005, Apte et al., 2011), or obtained data from local air quality monitoring networks (e.g. Dons et al., 2012). The 24-h rolling average concentrations of PM10, PM2.5, and CO reported by the local environmental agency were not used here because of the difference in the averaging periods. Fort Canning Park, an 18 public park located to the southeast of the transect route on slightly elevated land (approximately 35 m above street level) was selected as background site (Figure 3-4). Measurements of PM1 PM2.5, PM10, PN and CO were conducted in an open field located at least 200 m from the nearest main road (Figure 3-7). At that distance, pollutant concentrations due to traffic emissions would have decreased to levels close to ambient levels (Zhu et al., 2002). Thus the concentrations measured at this location are expected to be a good indicator of ambient air quality. The higher elevation of the park is assumed to have a negligible effect on the measurements. Only PM, PN and CO measurements were taken at this background location. 49 Figure 3-7: Field at Fort Canning Park used as the background site for comparison purposes of pollution data collected along the selected route and ambient pollution levels. Instruments were placed on a table on the cement platform in the center of the photograph. 3.3.3 Instrument set-up and sampling procedures Due to the sensitive nature of the instruments, procedures were put in place to avoid handling errors during the measurements. Sampling was carried out by the author and volunteers. All volunteers were trained on proper handling of instruments and had the opportunity to practice carrying the instruments prior to fieldwork. Background site set-up One set of the DustTrak, CPC, Langan CO monitor and HOBO sensor was arranged at the background site at m above ground level (Figure 3-8a). Care was taken to ensure inlets and sensors pointed in the same direction. One volunteer was stationed at the site to take notes of any anomalous activity that could affect the data (e.g. cigarette smoking by park users) since the area is a public park. 50 Mobile measurement set-up All mobile measurements were carried out by a team of two (the author and one volunteer). For portability, a backpack was used to house the instruments. The DC, PAS, and Microaethlometer were placed in the backpack, with sampling lines arranged to sample the typical breathing zone of adults, a region of 30 cm around the nose and mouth. The CO sensor was attached to the straps of the backpack, with the gas cell facing forward at shoulder height. The DustTrak and CPC were carried by hand and held up to roughly chest-level (Figure 3-8b). One person would carry the backpack containing instruments and take field notes whilst the second person wore the heart rate monitor and carried the DustTrak and CPC. Both researchers walked abreast to ensure that instruments were sampling the same parcel of air. Events that might be related to elevated concentrations or anomalous data were recorded during the measurement. Figure 3-8: Instrument set-up (a) at background site and (b) during measurements on different transport modes. Sensors measuring ASA, pPAHs and BC were placed in a backpack with sampling lines arranged to sample the typical breathing zone of adults. 51 Transport modes sampling procedure Measurements were carried out during afternoon rush hours (16:00 to 19:00 h) on weekdays. Although the ambient concentration of pollutants can change during this period as a consequence of the local meteorology, in particular mixing height, the concentrations at ground level of pollutants are not severely affected. Salmond and McKendry (2009) noted that the influence of the mixing height on ground level pollution tends to be negligible due to the proximity to the emission sources. Hence for the purposes of this study, weather and traffic conditions were assumed to be constant during this 3-hour period. Measurements were only carried out during dry weather conditions as rainfall can strongly influence the measurements and even damage sensors. Sampling after rain events was also avoided as the DustTrak does not perform well in high relative humidity, which leads to overestimation of PM concentrations. One set of measurements consisted of a door-to-door journey on each of the four transport modes (Bus, Taxi, MRT and Walking) sampled consecutively one after the other. The Taxi, Bus, and Walk mode transects were sampled in the same direction from the start to end point, in that order (Figure 3-4). On reaching the end point, researchers travelled (by MRT) back to the start point for the next trip. This meant that there were two MRT trips per day of sampling. Although the MRT trips are taken in the reverse direction of the other transport modes (i.e. from end point to start point), the entire trip is located underground. Thus the direction of travel is assumed to have a negligible effect on the measurements. There were two bus-stops and one MRT station between boarding and disembarking the bus and train, respectively. One of the main conclusions drawn from previous studies is that variability in exposure concentrations are highly dependent on individual choices (or availability of choices) of the commuter. Thus, for the Bus and Taxi modes, the first available 52 vehicle was boarded regardless of the bus or car model in order to capture a wider range of vehicle types. This could better represent the range of exposure concentrations that commuters may experience. More detailed travel descriptions for each mode of transport are provided in Appendix B. 3.4 Data quality control 3.4.1 Pre-sampling procedures Prior to each measurement day, all instruments were synchronized to a computer clock in the laboratory. This ensured that the time stamp was consistent across all instruments. The volunteer in charge of taking notes during the measurements also synchronized their wrist-watch to the same computer clock. Instruments with removable parts were dismantled and re-assembled for each day of sampling. This included the Microaethlometer filter which was changed daily. For the CPC, the alcohol reservoir was recharged overnight after each measurement. Zero calibration procedures for the CPCs and DustTraks were carried out according to manufacturer instructions before each set of measurements upon arrival at the background site. Instruments were left to log data for 10 at the background site prior to the actual sampling. All instruments were placed side-by-side with inlets close together during these parallel measurement periods. This data was later used to correct the instruments at the background site to those used for the mobile measurements. 3.4.2 Data post-processing After discarding the seven days of measurements affected by severe transboundary pollution, as well as trips affected by precipitation or technical problems with the instrumentation, more than 20 complete trips were sampled for each of the transport 53 mode. The total number of trips sampled during the complete study and the final number of trips used in the analysis are listed in Table 3-4. Table 3-4: Total number (N) of trips sampled for each mode of transport and the number of trips used in the final analysis after discarding measurements affected by transboundary pollution, precipitation or technical problems. Mode N trips N trips used in analysis Bus 30 23 MRT 59 45 Taxi 30 23 Walk 29 22 The data post-processing started with the application of correction equations to instruments with known issues such as the DustTrak and Microaethlometer. The DustTrak data were corrected for relative humidity following Ramachandran et al. (2003) and calibrated according to a gravimetric calibration done prior to the study (Appendix A). Black carbon data from the Microaethlometer was corrected using the ONA method available on the manufacturer website (wwww.aethlabs.com). Carbon monoxide data was also corrected with temperature according to procedures recommended by the manufacturer. Although instrument clocks were synchronized prior to measurements, each sensor has a slightly different response time which creates a lag in the logged data that needs to be accounted for. These time lags were computed for each instrument using cross-correlations against the DustTrak data to achieve better synchronicity across all instruments. Lag times ranged from to 15 s. Data from instruments used at the background site were also adjusted to the instruments used for the transect measurements. A ratio of the background to transect instruments for each day was calculated from data collected during the parallel measurement prior to the actual sampling. This ratio was then applied to the rest of the background data sampled on that day. The ratios for DustTrak, CPC and CO 54 Monitor ranged from 1.0 – 1.3, 0.9 – 1.0, and 0.07 – 0.7 respectively. Although the range of ratios for CO monitors appears unduly large, this is also partly due to the very low values recorded. Thus small but rapidly fluctuating measurements would appear significantly different. Only data from the background site that coincides with simultaneous transect measurements was used in the analysis. Following the corrections described above, suspicious data for individual instruments were removed based on notes taken during the sampling. As mentioned previously, some sensors were sensitive to movement and occasionally displayed an error status for a few seconds during sampling. These erroneous data were removed as part of the data quality checks. The DC and CO monitors also recorded zero value at times, which were removed from the dataset under the assumption that the concentrations present simply fell below the detection limit of the sensor. Table 3-5 lists the percentage of remaining data for each instrument after this step of data quality control. For heart rate measurements, 23 samples collected from 10 different volunteers are used for the dosage calculations. Table 3-5: Fraction of total data used for analysis after quality control including removal of suspicious data based on field notes and zero values. Instrument Data (%) DustTrak (transect) 100.0 DustTrak (background) 99.9 CPC (transect) 98.4 CPC (background) 89.8 Microaethlometer 89.7 PAS 100.0 DC 99.3 CO Measurer (transect) 82.7 CO Measurer (background) 87.2 HRM 100.0 55 Data post-processing and analysis were carried out with code written in Basic designed specifically for this study, Microsoft Excel (MS Office 2010), statistical software Minitab (Release 14.1), and statistical software package R (version 3.1.0). Statistical tests were carried out at 95% confidence level to identify significant differences between concentrations experienced across the different modes of transport and within different sections. The Anderson-Darling test for normality was used (see Appendix C for results). The results from the test suggest that data for the different pollutant variables are not normally distributed, thus the non-parametric Kruskal-Wallis Test and Mann-Whitney U Test with Bonferroni correction were applied. 56 [...]... O3 Figure 1- 1: Map of air quality monitoring stations (yellow triangles) across Singapore island (NEA, 2 013 ) In addition to general air quality monitoring, NEA also plays a strong regulatory role by controlling emissions at the source (National Environmental Agency, 2 013 a) This is enforced through inspections on industrial premises and monitoring stack emissions directly Vehicular emissions are also... be broken into the following research questions: 1 What are the levels of aerosol pollution that commuters are exposed to when travelling via different modes of public transport and walking in Singapore? 2 How do the aerosol concentrations during door -to- door trips on each mode of transport compare against each other? 3 What is the spatial variation of pollutant concentrations within the transport. .. dispersion and removal processes Shi et al (19 99) made measurements in Birmingham, UK, which suggest that higher wind speeds contribute differently to mass concentration than number concentrations Charron and Harrison (2005) found that median PM2.5 concentrations decreased from 25 to 18 µg m-3 as wind speed increased from below 1 to > 9 m s -1 within an urban canyon in London, UK However, the concentrations... the air pollution situation in Singapore (Cleary, 19 70) A few months after the assessment, a campaign against smoky motor vehicles was launched based on recommendations in the report This increased public awareness about air pollution issues in general and led to the formation of the Anti -Pollution Unit This unit has since evolved into the present-day National Environment Agency (NEA) which is in charge... has been declining in recent years, stabilizing the vehicle population at just under 1 million (Figure 1- 2) Figure 1- 2: Singapore s vehicle population from 2002 to 2 013 (Data from Land Transport Authority, 2 014 ) As mentioned above, despite the increased recognition of the impacts of particle pollution from traffic on human health, personal exposure to PM has not been well documented in Singapore Past... al., 2 012 ) Such particles are known as secondary particles Strategies to control traffic volume have been shown to be effective in controlling pollution at street level In Los Angeles, USA closing a section of a major freeway led to substantial reductions in PN, PM2.5 and BC (Quiros et al., 2 013 b) 17 Stringent traffic controls in Beijing resulted in decreases in ambient concentrations of up to 50%... 10 1 Figure 4-23: Time series of PM2.5 (top) and PN (middle) concentrations, and PM2.5/PM10 ratio (bottom) during the Walk mode trip on 20 May 2 013 10 2 Figure 4-24: Spatial variation in PM2.5 (top) and PN (bottom) concentrations during the Walk mode journey on 20 May 2 013 Traffic light symbols denote traffic junctions 10 3 Figure 4-25: Photograph of construction area on walk... amounts of hydrocarbons due to inefficient combustion (Kittelson et al., 20 01) Using a mobile emissions laboratory (MEL), Kittelson et al (2004) found on- road PN ranging between 10 4 to 10 6 # cm-3 when driving at free-flowing speeds Previously, the MEL recorded lower PN values (10 3 to 10 5 # cm-3) during a traffic jam where the average speed was < 32 km h -1, (Kittelson et al., 20 01) These findings were replicated... Balasubramanian, 2008), but none has looked at street level exposure The present study aims to fill this lack of 3 information for the transport microenvironment by measuring the personal exposure to aerosols of commuters on different modes of public transport including walking 1. 2 Singapore s air quality This section describes the local air quality management to put the present study within the context of the actual... found to increase with increasing wind speed, suggesting that 20 particle re-suspension is an important process contributing to PM10 concentrations (Charron and Harrison, 2005) Buonanno et al (2 011 ) also reported that higher wind speeds intensified the difference between the windward and leeward sides of the urban canyon vortex flow Total PN has been found to decrease with increased distance downwind . background site 11 1 5.2 Spatial variation of pollutant concentrations 11 5 5.2 .1 Bus-stops and Taxi-stands 11 8 5.2.2 In- vehicle concentrations 11 9 5.3 Dosage 12 2 Chapter 6. Conclusion 12 6 6 .1 Summary. Review 10 2 .1 Estimating exposure 10 2 .1. 1 Measuring exposure to particle pollution 11 2.2 Particle pollution in the transport microenvironment 16 2.2 .1 Transport emissions 17 2.2.2 Spatial and. Abbreviations xv Chapter 1. Introduction 1 1. 1 Human exposure to air pollution 1 1. 2 Singapore s air quality 4 1. 3 Objectives 8 1. 4 Thesis outline 9 Chapter 2. Literature Review 10 2 .1 Estimating