Personal Exposure to PM2.5 in the Various Microenvironments as a Traveler in the Southeast Asian Countries

Corresponding Author: Semih Ozler Department of Civil and Environmental Engineering, University of WisconsinMadison, 53706, Madison, WI, USA Email: ozler@wisc.edu Abstract: Air pollution has become a pressing issue in today’s society because of its significant effects on humans, animals, plants, air quality, climate and the wider environment. Most urban areas are associated with one or more air pollutants which are emitted from local or regional pollution sources including vehicle exhausts, fossil fuels using in energy production, emissions from industrial and mining activities, agricultural and construction operations, household usage of chemicals and materials and natural causes. Most personal exposure studies are focused on local environments and shortterm periods. Previous controlled experiments and studies were done in a small number of designated areas in cities. Our research study used timebased activity data; 3 main and 17 sub-microenvironments were applied over 37 days-long research while traveling through Southeast Asian countries. In this study, personal exposure of PM2.5 for a traveler was monitored using an assembled low-cost monitor with Plantower PMS 3003 PM2.5 sensor which has a light-scattering principle. All time-based activity data was recorded with a smartphone whenever microenvironments changed during the study period. The goal of this study was to understand more about the personal exposure to PM2.5 related air pollution in the global travel environment as a traveler and to understand how an individual’s activity and location impact PM2.5 exposure. According to the results from the Southeast Asia study, the personal PM2.5 exposure varied in the categorized microenvironments. Port/Station (outdoor) and Café/Pub/Restaurant (indoor-outdoor) were the most polluted microenvironments with 32.8 and 29.6 μg/m3 1-h mean PM2.5 concentration, respectively. Market/Shopping Mall (indoor), Street (outdoor) and Cable Car/Metro/Tram (vehicle) were also concerning microenvironments with 19.3, 19.3 and 18.9 μg/m3 1-h mean PM2.5 concentrations, respectively. Passenger Car microenvironment had the lowest 1-h mean PM2.5 concentration of 2.3 μg/m 3 which agrees with some other studies on transportation microenvironments in the literature.


Introduction
Within the last two decades, urban air pollution from large cities has been recognized as one of the world's most concerning environmental issues (Koçak et al., 2011;Yusuf and Resosudarmo, 2009). Most developing countries, following a general trend, have experienced rapid growth including urbanization, urban sprawl, vehicle ownership and industrialization which is reflected in pollutant emissions to the atmosphere. Urbanization not only affects the nearby landscape, air quality, regional climate and ecosystems of the polluting cities but also affects areas downwind of these regions despite recent improvements in air pollution science, air pollution control technologies and public awareness (Molina et al., 2004). Air pollution remains a massive challenge due to continuing uncontrolled emission discharges, inadequate environmental regulations, poor governmental control mechanisms, individual consumption behaviors and rebound effects of technological improvements for many city dwellers and travelers (Akimoto, 2003).
Moreover, the World Health Organization (WHO) states that 92% of the world's population lives in cities and metropolises where PM 2.5 concentration levels exceed the WHO suggested PM 2.5 concentration limits which are 10 µg/m 3 annual mean and 25 µg/m 3 24-h mean (2017). Those limits are suggested as guidance to reduce the health impacts of air pollution, but they are not legally binding policies. Many countries don't have any regulations related to PM 2.5 . Some countries regulate PM 2.5 with higher concentration values than WHO's guidelines; this makes it difficult for some cities to comply with the WHO's ambient air quality guidelines (Molina et al., 2004;Sharma et al., 2013). Reported measurements from those cities far exceed WHO's guidelines, resulting in millions of premature deaths (Lim et al., 2012). As seen in the dataset from WHO, there is a large variation in PM 2.5 exposure across different countries and continents. The highest annual average urban PM 2.5 concentration is 127 µg/m 3 in Saudi Arabia and the lowest average urban PM 2.5 concentration is 5 µg/m 3 in New Zealand and Brunei Darussalam. In addition to that, only 19 countries comply with WHO's suggested annual mean PM 2.5 concentration limits. The remaining 160 countries exceed those levels, ranging from 11 to over 100 µg/m 3 (2017). It is not surprising that PM 2.5 related air pollution is a global environmental health problem that affects people worldwide, but lowincome and middle-income countries experience this burden more than high-income countries.
Moreover, a new conceptual framework about Earth systems, planetary boundaries, was developed by a group of environmental scientists in 2009 (Rockström et al., 2009). They identified nine planetary boundaries that must not be surpassed in order to maintain global sustainability. It is believed that exceeding one or more planetary boundaries may be deleterious or even catastrophic due to the risk of triggering non-linear, abrupt environmental change from continental to planetary-scale systems (Rockström et al., 2009;Steffen et al., 2015). Atmospheric aerosol loading is one of the most important subjects in the planetary boundaries report because of the influence of aerosols (suspended particles in the atmosphere) on climate systems and their adverse acute and chronic effects on human health (Knibbs and de Dear, 2010;Rockström et al., 2009;Steffen et al., 2015). Although it is hard to determine the exact effects and mechanisms of atmospheric aerosol loading due to the complex composition of particles and gases that are involved in the aerosol production, their atmospheric chemistry and the spatial and temporal dynamics of pollutants.
Long-term exposure of PM 2.5 has been associated with allergies, respiratory diseases, cardiovascular problems and lung cancer; these health outcomes result in increased mortality, morbidity and hospital admissions (Kingham et al., 2013;Wang et al., 2011). The Global Burden of Disease study showed the significant role of air pollution as a global player, placing it among the top ten risks factors faced by human beings (Lim et al., 2012). According to the summary report of WHO's "Burden of Disease", 4.3 million deaths globally were attributable to household air pollution and 3 million deaths were attributable to ambient air pollution in 2012 (2012). According to the Institute for Health Metrics and Evaluation (IHME)'s dataset, it was estimated that outdoor air pollution resulted in 4.2 million deaths in 2016; this represents an increase from 3.4 million in 1990. Overall, the majority of pollution-related deaths are in South Asia, East Asia and Southeast Asia which alone accounted for nearly 3 million deaths in 2016 (OWiD, 2018). Moreover, as stated in Giannadaki's et al. (2016) study, the global premature mortality by PM 2.5 was estimated at 3.15 million/year in 2010. These studies show that PM 2.5 is a global actor and responsible for millions of premature deaths. As stated in Steinle's et al. (2015) study, controlling air pollution not only directly reduces adverse health effects but also increases general well-being, quality of life, public health and can have positive impacts on the environment in those countries and regions.
Monitoring air pollutants and measuring their ambient concentration is important because regulatory decisions and policies by governmental agencies are made based on those data. As a worldwide application, ambient air pollution concentrations for various air pollutants are measured using instruments such as Fixed Monitoring Stations (FMSs). Although these instruments are reliable and the reliability of the measured data is ensured by applying standard procedures for calibration, data collection and post-processing, these instruments are designed for research purposes (Kumar et al., 2015). In addition, they are generally not user-friendly and require experienced users to operate them due to their complexity. Moreover, they are expensive and a significant amount of investment is required to maintain and regularly calibrate them. Often, small quantities of FMSs are used at designated sampling points (Kumar et al., 2015) which leads limited spatial coverage in many places (Gao et al., 2015). In many cases, these monitoring stations are placed in secured areas with lower human interaction. Additionally, it is preferred to locate FMSs away from roadsides, crowded city centers, industrial zones and major traffic congestion areas where local ambient air quality is affected by various emissions from diverse sources. This distribution method/preference often fails to represent the localized air quality data, local air pollution trends and the actual exposure levels of air pollutants (Kumar et al., 2011, Wang et al., 2011Gao et al., 2015). The current networks fail to capture spatiotemporal variations of air pollutants (Gao et al., 2015) and as a result, PM 2.5 exposure may be significantly underestimated (Meng et al., 2012). A study in New York showed that the real-time exposure to PM 2.5 can be between 20-200% higher than the readings from local FMSs (Wang et al., 2011). Similar studies that examined personal exposure in different transportation microenvironments found that individuals may be exposed to much higher ambient PM 2.5 concentrations than local FMSs indicated (de Nazelle et al., 2012;Int Panis et al., 2010;Kaur et al., 2005;2007).
A new and innovative method of measurement, lowcost monitors, may present a possible solution to the problems with FMSs. Technological improvement and continued development of low-cost sensing technology have made low-cost monitors less expensive, easier to deploy, operate and manage than traditional systems. Low-cost monitors can be deployed in significant numbers for the same price as a single FMS. By doing so, it is easy to create monitoring networks for designated areas or even cities, calculate emission inventories of various pollutants, measure real-time short-term and long-term exposure of users in some designated areas, detect pollution hotspots and high concentration zones and with the help of this more accurate data, focus on air pollution mitigation strategies (Steinle et al., 2013;Gao et al., 2015). These monitors can be used by individuals or researchers to understand real personal exposure values in different microenvironments.
Low-cost monitor networks especially help to provide a more detailed picture of indoor air quality and reliable data about indoor air quality levels, which is quite important since people spend most of their time in indoor environments (Steinle et al., 2015). Some low-cost PM sensors have been evaluated for their performance in designated field and laboratory studies that found these PM sensors were promising for the future applications and useful as monitoring equipment both indoors and outdoors (Austin et al., 2015;Chen et al., 2017;Gao et al., 2015;Holstius et al., 2014;Kelly et al., 2017;Marques et al., 2018;Wang et al., 2015). Moreover, some projects exist that are dedicated to measuring ambient air concentrations of pollutants in the urban areas with low-cost monitors. Some such initiatives are the OpenSenseMap project (2017) which is mainly used by European users, OpenAQ (2017) and Purple Air platform (2017) which have active users both in Europe and in the United States. Air quality data collected from low-cost monitors that are connected to the internet is shared online for the users, public and researchers.
Many particulate matter exposure studies are done using sensors in local environments to address the exposure of selected study groups like pedestrians, city dwellers, commuters, car drivers and cyclists in different transportation environments. Often, findings are different both between and within studies due to different urban settings, cities' vehicle fleet, regional traffic configuration and congestion level, local emission profiles, ambient pollutant concentrations and meteorological conditions (de Nazelle et al., 2012;Int Panis et al., 2010;Kaur et al., 2005;2007;Kingham et al.. 2013;Knibbs and de Dear, 2010;Steinle et al.. 2015;Strak et al.. 2010;Thai et al., 2008;Wang et al., 2011). In contrast, this study's main framework will focus on a single traveler, including his daily activity and the total exposure of PM 2.5 in 3 main microenvironments and 17 sub-microenvironments. Travelers are an important group that tend to have high personal exposure levels not only to PM 2.5 but also other pollutants. They can be a concerned group of people because the aims of traveling including exploring new destinations, seeing unfamiliar places, discovering new cultures and traditions, meeting new people and trying new foods and drinks in the relatively short amount of time, lead to prominent levels of body activity. High body activity results in a large increase in breathing frequency and tidal volume which influences their inhaled dose of air pollutants and increased deposition on the target's lung tissues (Int Panis et al., 2010;de Nazelle et al., 2012). It is well-explained in another study that higher concentrations of PM 2.5 don't negate the health benefits of physical activity, but they do decrease the advantages of increased physical activity (Tainio et al., 2015). In this research approach, we believe that understanding more about personal exposure to PM 2.5 related air pollution is not only important for global travelers but also the local populations.

Study Area and Travel Plan
During the study period, Thailand, Cambodia, Singapore, Taiwan, Hong Kong, Macau, Qatar and the United States (17 cities: Bangkok, Ayutthaya, Phetchaburi, Hua Hin, Pattaya, Sattahip, Aranyaprathet, Poipet, Siem Reap, Phnom Penh, Singapore, Taipei, Hong Kong, Macau, Doha, Chicago and Madison) were visited in 37 days. A summarized Southeast Asia travel plan is shown in Fig. 1. There was no pre-designated travel plan or time distribution in the countries in order to mimic the unpredictable travel choices of a solo traveler. 18 days were spent in Thailand, 6 days in Cambodia, 3 days each in Singapore and Taiwan, 6 days in Hong Kong (nearly half of the day in Macau) and the rest of the activity data was recorded on the way back to Madison, WI, US through Doha, Qatar and Chicago, IL, US. The travel dates and times spent in each country, in addition to maximum, minimum and mean recorded PM 2.5 concentrations during those times are shown in Table 1. Not all recordings are well-represented for all the listed countries due to the limited travel and recording time in cities such as Doha, Qatar. It is believed that others such as Thailand, Cambodia, Singapore, Taiwan and Hong Kong were well represented because a relatively long amount of time was spent in the various microenvironments.

Microenvironments Approach
A microenvironment was described as a small space in which human contact with a pollutant takes place and which could be treated as a well-characterized, relatively homogenous locations with respect to pollutant concentrations for a specified time in Özkaynak's et al. (2008). Moreover, the microenvironments approach is commonly used in personal exposure and air quality studies to understand the effects of different variables on ambient PM 2.5 concentration. Indoor, outdoor and vehicle environments generally have different air pollution statistics and dynamics even though indoor and vehicle microenvironments are affected directly by outdoor microenvironments. In this study, 3 mainmicroenvironments and 17 sub-microenvironments were categorized and listed in Fig. 2. This approach was utilized in this study to understand the effects of different microenvironments on personal exposure of PM 2.5 while traveling in different countries.     During the study, the monitor was actively recording PM 2.5 concentrations for about 49,693 min (around 828.22 h/34.51 days). The non-recording time was about 2,991 min (around 49.85 h/2.08 days), due to file copying from a micro-SD card into the computer, changing the drained battery, closing the monitor because of the security checks at the airport, security concerns during international flights and some other unforeseen issues. According to the activity log data, around 59% of the total recording time was spent indoors, 31% outdoors for travel activities and 10% in vehicles for transportation purposes. Table 2 shows a more detailed picture of the total time spent in 17 submicroenvironments. 32% and 27% of the total recorded time was spent in Hostel/Hotel/Spa category and House category, respectively. It was obvious that the main PM 2.5 exposure occurred in indoor environments; although some of the recorded logs are related to outdoor PM 2.5 concentration in the House category. This is consistent with a previous literature study with a similar result (Steinle et al., 2015). 11% and 7% of the total recorded time were related to high-level outdoor activities such as walking, hiking and cycling in the Street and Attraction categories respectively. 6% of the total recorded time was also spent in both indoor and outdoor-based activities in Café/Pub/Restaurant category. The total remaining recorded time was spent in other various microenvironment categories which are listed in Table 2.

The Study Participant/Traveler
The study participant/traveler was trained to use the monitor before beginning to travel. He had the related Health, Safety and Environment training and gained proper information on how to use the designated research equipment during his travel. He was also provided with a written operational manual. He was informed to choose his daily activity and trip plan randomly, but he was requested to record all microenvironment changes and their times into his smartphone. A set of data which is related to the 3 main and 17 sub-microenvironments and the time series of the microenvironmental change were gathered by using the notes from his smartphone. Temperature, humidity, 1-min averaged PM 2.5 concentrations and time information were gathered from the research equipment itself.

Research Equipment
During the study period, an assembled low-cost monitor was used. It had a Plantower PMS 3003 PM 2.5 sensor which was evaluated in controlled laboratory and ambient environmental conditions in the previous studies and presented a reliable performance (Kelly et al., 2017;SCAQMD-AQ-SPEC, 2017a;2017b;2017c). The Plantower PMS 3003 uses the light-scattering principle for particle counting. A fan is employed to draw air through a measurement chamber. 90° scattered light is detected by a photo-diode detector and the data is converted to a mass distribution (µg/m 3 ). The laser wavelength was estimated to be around 650±10 nm. According to the manufacturer, it detects PM in the range of 0.3 µm to 10 µm, it has a 10s response time and its counting efficiencies are 50% @ 0.3 µg/m 3 and 98% @> 0.5 µg/m 3 (Kelly et al., 2017). The assembled low-cost monitor and its parts are shown in Fig. 3. The teensy microcontroller, clock, humidity and temperature sensors were the main parts of the research equipment. Other than these main system parts, some secondary parts included the protection case, voltage regulator, barrel jack power cable, LED indicator, micro-SD card and battery. The research equipment is user-friendly and easy to operate. There is no on/off button or screen that can make things complicated for users. When it is plugged into the energy source, the LED light indicator starts blinking and then continues to blink until it is unplugged from the energy source. It weighs 0.3 kg without the battery and its dimensions are 15.5×8.5×4.5 cm which makes the instrument easy to carry.
A 30 L hiking backpack with side pockets was used to carry the instrument with its inlet facing up for proper air uptake into the sensor's chamber. The backpack was generally worn in the outdoor microenvironments and it was kept close to the bearer by placing it on a table, chair, seat, or the bearer's lap when it wasn't being worn in some of the vehicle and indoor microenvironments (Fig. 4). The assembled low-cost monitor can run approximately 55 h with used 10400mAh battery, but this running time is slightly affected by environmental conditions such as temperature. During the study period, two batteries were used. Either every day or two days in a row the used one was replaced with fully charged one. Moreover, an 8 GB micro-SD card was used for proper data storage or management.

Data Collection, Management and Processing
The data was collected for 37 days from June 24, 2017 to July 30, 2017 by using the assembled low-cost monitor. Once the data was moved from the micro-SD card to the computer as a .txt file, the unnecessary data including startup information of the monitor for each start was deleted and then the cleaned data was converted into a .cvs file to input microenvironments and countries data manually with using recorded notes data from the smartphone. After inputting, processing and checking the data carefully, the final datasets were prepared. A spreadsheet application Microsoft Excel, a word processor Microsoft Word, a programming language for statistical computing and graphics RStudio and a simple raster graphics editor Microsoft Paint were used to create PM 2.5 related figures and tables for this study.

Previous Performance Evaluations of Plantower PMS 3003 Low-cost Sensor
A previous study focused on assessing the performance of a Plantower PMS 1003 sensor against one of the Federal Reference Method (FRM), two of the Federal Equivalent Methods (FEM) (TEOM and Sharp) and a research-grade monitor (GRIMM) under ambient PM 2.5 concentrations in Salt Lake City, Utah during winter. They compared 1-h averaged data from the measurements and they found high PM 2.5 correlations with FEMs (R 2 = 0.82-0.92) and the research-grade instrument (0.83-0.93) (Kelly et al., 2017). Another study focused on using a Plantower PMS 3003 sensor and an Unmanned Aerial Vehicle (UAV) to gather environmental data including temperature, humidity and PM in Nan province of Thailand. Before starting their field research, they tested the Plantower PMS 3003 sensor against a FEM (TEOM). Their 1-h averaged data showed low PM 2.5 correlations with TEOM (R 2 = 0.66) (Chunitiphisan et al., 2018).
Another study from the South Coast Air Quality Management District (SCAQMD) -Air Quality Sensor Performance Evaluation Center (AQ-SPEC) focused on field-testing commercially available low-cost monitors such as the Laser Egg Sensor (Plantower PMS 3003), the PurpleAir (Plantower PMS 1003) and the PurpleAir PA-II (Plantower PMS 5003) (2017a, b, c). They collocated these monitors next to a FEM (BAM) and a researchgrade monitor (GRIMM) under ambient environmental conditions in Southern California and compared 1-h averaged data from the measurements. They reported low PM 2.5 correlations for the Plantower PMS 3003 with a BAM (R 2 = 0.57), high PM 2.5 correlations for the Plantower PMS 1003 with a GRIMM (R 2 = 0.91) and medium PM 2.5 correlations for the Plantower PMS 1003 with a BAM (R 2 = 0.77) and high PM 2.5 correlations for the Plantower PMS 5003 with a GRIMM (R 2 = 0.93) and a BAM (R 2 = 0.86) (2017a, b, c).
A recent study campaign was designed to evaluate the performance of Plantower PMS 3003 sensors under ambient PM 2.5 concentrations. During the first stage of their study, they deployed the Plantower PMS 3003 sensors to low PM concentration suburban regions of Durham, North Carolina. At the first site, those sensors were compared to an E-BAM and at the second site, they were compared to FEMs (Sharp and TAPI T640). They reported low PM 2.5 correlations with the E-BAM (R 2 = 0.41) at the first site, both low PM 2.5 correlations with the SHARP (R 2 = 0.28) and medium PM 2.5 correlations with the TAPI T640 (R 2 = 0.70) at the second site. Then they Inlet Inlet deployed those sensors to high concentration urban locations of Kanpur, India to compare their results with the E-BAM. They reported low PM 2.5 correlations for Plantower PMS 3003 sensors with the E-BAM (R 2 = 0.61) during monsoon season and medium PM 2.5 correlations (R 2 = 0.78) during the post-monsoon season (Zheng et al., 2018). Figure 5 gives brief information about the total measurements from the monitor for PM 2.5 exposures on selected days. All the concentration values that were used for graphing were 1-min averaged PM 2.5 concentration. 3 days (June 25, 2017-in Thailand, July 2, 2017-in Thailand and July 8, 2017 were selected to show prominent levels of PM 2.5 exposure while engaging in travel activities. The highest recorded concentration was 1,142 µg/m³ in an Outdoor -Café/Pub/Restaurant microenvironment while being exposed to tobacco smoke of the cigarette users around the pub. The second highest was 525 µg/m³ in an Indoor -Café/Pub/Restaurant microenvironment with exposure to hookah and tobacco smoke from the cigarette and hookah users in the café. This recorded value was lower than the first one even though they had a similar pollution source. The differences may have been related to distance from the pollution source, dilution effects of the air and wind, the position and placement of the monitor and the average ambient concentration of PM 2.5 in those microenvironments. The third highest PM 2.5 concentration value was 492 µg/m³ in an Outdoor -Attraction microenvironment while being exposed to smoke from burning candles and incenses around the temple. As stated in Int Panis's and Gao's paper, the health effects of short bursts of high exposure to PM 2.5 in contrast to chronic exposure are not well understood (Int Panis et al., 2010;Gao et al., 2015).  (2017) and the Southeast Asia study result. As can be seen from the table, all recordings from the study are lower than the countries' yearly statistics from the given sources. Although PM 2.5 concentration values from the Southeast Asia study are very similar to the United States and Singapore and relatively similar to Cambodia, they are not well matched with WHO's country statistics shown in Table 3. Moreover, the same pattern is observed when the Southeast Asia study's outdoor 1-hour mean PM 2.5 concentrations are compared to WHO's country statistics except in the United States, Singapore and Qatar where concentrations are lower for the United States, higher in Singapore and there is no outdoor data for Qatar. Figure 6 shows WHO's suggested 24-h mean PM 2.5 concentration limit as a green line which at 25 µg/m 3 , WHO's annual mean ambient PM 2.5 concentration data for each country as a red triangle and 1-h mean PM 2.5 concentrations from the Southeast Asia study in the traveled countries as a blue square which was shown in Table 3. The probable reasons for low PM 2.5 concentration values compared to WHO's annual mean statistics are having only indoor measurements at the airport in Doha, Qatar, relatively good ambient air quality due to monsoon season which affects the air quality positively, other meteorological factors such as temperature, humidity, precipitation, wind speed and wind direction which yielded lower PM 2.5 concentration values in Thailand, Taiwan, Hong Kong and Macau at that time of the year and at last but not least, it may be related to having low performance of the monitor during those measurement periods as compared to FMSs.    were the only available countries that OpenAQ has lowcost monitor users and PM 2.5 recordings. According to data from those low-cost sensors and the assembled monitor which was used in our study, the results matched well for the United States and Taiwan datasets. In addition, Fig. 8 informs us that the Southeast Asia study's 1-h mean PM 2.5 concentrations for 3 microenvironments are less than WHO's 24-h mean PM 2.5 exposure limits. Indoor, outdoor and vehicle microenvironments' PM 2.5 concentrations are 9.8, 16.1 and 13.5 µg/m 3 , respectively. Figure 9 illustrates the Southeast Asia study's 1-h mean PM 2.5 concentrations for 17 microenvironments. The microenvironments approach is important for understanding the effects of environments on personal PM 2.5 exposure. Despite spending a relatively small amount of time in microenvironments that have high pollutant concentrations, they may be a significant predictor of 24-h mean personal exposure (Strak et al., 2010;de Nazelle et al., 2012). One study in Flanders, Belgium showed that spending 6% of volunteers' time in the different travel modes resulted in 21% of the total personal exposures and 30% of the total inhaled dose in a single day (Dons et al., 2012). The average PM 2.5 concentrations for the 17 microenvironments varied between 2.3 and 32.8 µg/m 3 . According to the findings of this study, the most problematic/polluted microenvironments are Port/Station (outdoor) and Café/Pub/Restaurant (indoor-outdoor) with 32.8 and 29.6 µg/m 3 , respectively. The main reasons for measuring high concentrations of PM 2.5 in these microenvironments were proximity to tailpipe emissions from transportation sourced combustion engines such as cars, buses, boats, ferries and light-duty vehicles and emissions from food cooking/baking activities. Market/Shopping Mall (indoor), Street (outdoor) and Cable Car/Metro/Tram (vehicle) were the second most concerning microenvironments after measuring 19.3, 19.3 and 18.9 µg/m 3 , respectively. Interestingly, the Passenger Car (vehicle) microenvironment had the lowest PM 2.5 concentration with 2.3 µg/m 3 . This both agrees and disagrees with some studies on transportation microenvironments in the literature (Knibbs and de Dear, 2010;Wang et al., 2011;Kaur et al., 2005;de Nazelle et al., 2012). During the transportation, all the windows were closed and all of the air passed through the Air Conditioning (AC) and it is highly likely that the used passenger car had an effective AC system for particle removal which resulted in low PM 2.5 concentrations and exposure compared to other microenvironments. As can be seen from the results, PM 2.5 concentration measurements (µg/m 3 ) are given in 1-h mean format but WHO's guideline value is in the 24-h mean. It seems an inappropriate way to compare 1-h personal exposure values to WHO's 24-h ambient guideline value directly, due to the limited data of personal exposure for each microenvironments that were listed in this study and shortterm measurement in some of the microenvironments during the study period, it is the only applicable way to compare measured concentrations. By comparing the measurement results in the listed microenvironments to WHO's guideline value, each microenvironment can be categorized as polluted, concerned or not-concerning. This gives a broad idea and helps for understanding the effects and dynamics of these microenvironments on presented PM 2.5 concentration levels. Moreover, it is generally true that long-term and short-term measurements may be correlated with each other in some cases and studies, so we believe that applying this approach to this study is acceptable.

Results from Southeast Asia Study
Moreover, multiplying 1-h mean PM 2.5 concentrations by the time spent in each microenvironment gives PM 2.5 exposure in µg-hour/m 3 (in Table 2). It helps us to understand which microenvironment is dominant during our study. Unsurprisingly, the Hostel/Hotel/Spa microenvironment had the highest exposure due to having the highest time spent, 32%, even though it has a relatively low-level 1-h mean PM 2.5 concentration. The Street microenvironment was the second highest PM 2.5 exposure, having a relatively high 1-h mean PM 2.5 concentration, although less time was spent there, 11%. Interestingly, the Café/Pub/Restaurant and House microenvironments had the third highest PM 2.5 exposure, 1,449.4 µg-hour/m 3 and 1,426.6 µg-hour/m 3 , respectively. The Café/Pub/Restaurant microenvironment was one of the most polluted environments and just 6% of the recording time was spent there. The House microenvironment had a relatively low 1-h mean PM 2.5 concentration, but one of the highest recording times, 27%. It is obvious that personal exposure has a strong correlation with ambient PM 2.5 concentrations and time spent in the different microenvironments, both determine the daily exposure of PM 2.5 for an individual person.

Discussion
As noted in previous studies, the low-cost sensors need more improvement and further studies to understand their technological limitations and to overcome their drawbacks. They are not as accurate or as consistent as FMSs and can be affected by meteorological conditions such as temperature, humidity, precipitation, wind speed and wind direction, seasonal variations, the chemical composition of PM 2.5 and other factors (Gao et al., 2015;Wang et al., 2015). They have relatively short operational time because of their fragility under diverse and extreme environmental conditions (Kumar et al., 2015). This issue leads to another environmental concern, the additional burden of electronic-waste management because of the low-cost monitors' short-term operational time. Furthermore, maintenance, data management, analysis, post-processing and visualization costs can exceed the cost of the actual monitor itself (Kumar et al., 2015). Low-cost sensing introduces opportunistic sensing under the banner of citizen science which can cause another challenge in addition to present ones. PM 2.5 and other pollutants' concentration values have been collecting and posting online. Some concerns include managing the enormous amount of data collected by the monitor networks, transparency of the recorded concentrations and organizations and access possibilities to them as users, citizens, scientists and government officials. In addition, without giving proper information and training to users, concentration measurements may cause unnecessary public concern or complacency about real air pollution levels. Lastly, they are also vulnerable to manipulation which may create unrealistic datasets that cause public pressure on environmental agencies with or without purpose.
Limitations of our study include working with just one traveler, using one monitor to sample the PM 2.5 concentration, not having a perfect design of the lowcost monitor, potential airflow perturbations when the monitor was used in the side pocket of the backpack, a lack of any co-location study with traditional monitoring stations to compare the PM 2.5 readings with a robust reference instruments and also the effect of environmental variables (weather patterns, chemical compositions of PM).

Conclusion
Low-cost sensing provides advanced spatial and temporal variability exposure datasets which help to improve the characterization of air pollutants, air pollution levels, exposure patterns and increase the quality of exposure models via representative personal exposure studies. These datasets allow environmental health scientists, epidemiologists and allergists to have current knowledge of air pollution levels and trends in order to advise their patients accordingly (White et al., 2012). They can advise their patients to minimize their outdoor activities during days with high pollution levels to avoid harmful effects and recommend them to review their indoor activities and pollution sources to manage their indoor exposure levels. Our measurements showed that some microenvironments are concerning due to higher PM 2.5 concentrations. Sensitive people should spend less time in these microenvironments in order to limit their PM 2.5 exposure. Moreover, by sharing air pollution and quality results with the public via mobile phone applications or the internet, sensitive people, especially those already at risk, can plan their outdoor activities, check their residential areas' and workplaces' air quality and may choose new residential locations or workplaces which are away from localized air pollution sources. Individuals may also make decisions about which countries/cities to travel based on pollution data (Bernstein et al., 2004). An environmentally sensitive tourism approach, defined by choosing less polluted, more environmentally friendly and sustainable developed countries/cities will create economic pressure on countries/cities with higher PM 2.5 pollution levels not to lose their share of the tourism industry. Furthermore, patients, doctors, patient advocates, engineers and environmentalists who support clean air movements can push governments and environmental agencies to implement stricter and more science-based air pollution standards and regulations.
Several aspects of this study can be addressed in future work such as: encouraging researchers to set up personal exposure studies, raising awareness to the effects of microenvironments and people's activities on potential exposure, extending measurements to a longer time frame, using more low-cost monitors and working with more travelers to have more spatial coverage, increasing the study area to other countries and continents, considering dose calculation for individual travelers and examining particle composition and toxicity of PM 2.5 .