Ho‐Tang Liao1, Yu‐Chi Lai1, Hsing Jasmine Chao2, Chang‐Fu Wu This email address is being protected from spambots. You need JavaScript enabled to view it.1,3 1 Institute of Environmental and Occupational Health Sciences, College of Public Health, National Taiwan University, Taipei 10055, Taiwan
2 School of Public Health, College of Public Health, Taipei Medical University, Taipei 11031, Taiwan
3 Department of Public Health, College of Public Health, National Taiwan University, Taipei 10055, Taiwan
Received:
October 25, 2022
Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.
Revised:
December 27, 2022
Accepted:
January 3, 2023
Download Citation:
||https://doi.org/10.4209/aaqr.220361
Liao, H.T., Lai, Y.C., Chao, H.J., Wu, C.F. (2023). Vertical Characteristics of Potential PM2.5 Sources in the Urban Environment. Aerosol Air Qual. Res. 23, 220361. https://doi.org/10.4209/aaqr.220361
Cite this article:
Exposure to urban air pollution, particularly fine particulate matter (PM2.5), is known to be harmful to human health. Source apportionment of urban PM2.5 provides information to develop effective control strategies, thus reducing the exposure concentrations and health risks. However, this is a challenging task in metropolitan areas where people live in high‐rise buildings. To understand the vertical characteristics of air pollution sources in urban areas, a total of 114 vertically stratified PM2.5 samples were collected at six heights (1st, 7th, 10th, 13th, 17th, and 20th floors) of one building during the period between 19 October and 22 December 2020. Absorbance, 16 trace elements, 8 water‐soluble ions, and water‐soluble organic carbon on Teflon‐membrane filters were measured. Positive Matrix Factorization was utilized to achieve the source apportionment analysis. Six source factors, including biomass burning/industry, traffic related, secondary aerosol, soil dust, contaminated road dust, and sea salt, were retrieved. During the sampling period, the major contributor to PM2.5 was secondary aerosol (28.8%), followed by biomass burning/industry (24.4%) and traffic related (13.3%). It should be noted that road traffic emissions (traffic related and contaminated road dust) accounted for 24.7%, making them the second largest contributor to PM2.5. Contributions of road traffic emissions significantly declined with height (29.3%–21.4%), which was in line with the findings in previous studies, and could explain the vertical variation of PM2.5 identified in this study. These findings help estimate the realistic exposure at different residential heights, consequently facilitating control strategy development.HIGHLIGHTS
ABSTRACT
Keywords:
Source apportionment, Fine particulate matter, Positive matrix factorization, Water soluble organic carbon, Vertical distribution
In 2018, the urban population was more than 50% in the world and accounted for 78% in more developed regions, including Taiwan (United Nations, 2018). In addition to many benefits of urban life, rapid urbanization also results in some negative impacts such as poor air quality. Exposure to air pollution has been associated with numerous adverse health effects and caused approximately seven million premature deaths worldwide in 2016 (WHO, 2021). These effects, including cardiopulmonary disease and cancer, were closely associated with fine particulate matter (PM2.5) that can penetrate into and be deposited in the lung (Kim et al., 2015; Lu et al., 2015; IARC, 2016). To develop effective control strategies, identifying urban PM2.5 sources and quantifying their contributions to the exposure concentrations and health risks are warranted. The multivariate Positive Matrix Factorization (PMF) solution is a useful tool for apportioning sources of PM2.5 (Hopke et al., 2020; Schneider et al., 2022; Silva et al., 2022). The source apportionment analysis is carried out by solving the chemical mass balance equation. Given the speciated concentration data xij and the measurement uncertainty uij, PMF simultaneously estimates factor contribution gik and factor profile fkj during the iterative process to obtain a minimum of objective function Q (Paatero and Tapper, 1994; Hopke, 2016). The vertical distribution of source contributions is a critical issue in the urban area where people live in high‐rise buildings (Zauli Sajani et al., 2018; Chen et al., 2020). To date, only a few source apportionment studies have been conducted focusing on the vertical distribution of contribution estimates of PM2.5 sources at different heights (Wu et al., 2015; Wang et al., 2016; Liao et al., 2020, 2021). In addition, there was great variation in the altitudes investigated due to different sampling strategies among studies. For example, two studies were conducted at two or four heights at a 225‐m‐high meteorological tower (Wu et al., 2015; Wang et al., 2016), which is much higher than the surrounding residential buildings. The other two source apportionment studies in Taiwan collected PM2.5 samples at three floors from buildings lower than 40 m and concluded that traffic related emissions showed decreasing trends with increasing height (Liao et al., 2020, 2021). Nonetheless, the proportion of high-rise buildings is getting increased in urban areas. To better understand the vertical variation of urban residential exposure to ambient PM2.5, this study was conducted at one building that has more than twenty floors (approximately 70 m) to collect vertically stratified samples at six heights. PMF was applied to the measured PM2.5 components to explore the vertical characteristics of PM2.5 sources in metropolitan areas, thus providing more information in designing pollution control strategies. In Taiwan, more than three quarters of the population live in urban areas. Taipei metropolis, which has the highest population density in Taiwan and numerous high‐rise buildings, was chosen to explore the vertical characteristics of urban PM2.5 sources. Vertically stratified samples were simultaneously obtained at six floor‐levels (1st, 7th, 10th, 13th, 17th, and 20th floors) of one building that has balconies facing a major road. The sampler inlets were set at approximately 1.5 m, 20.1 m, 30.9 m, 41.7 m, 52.5 m, and 63.3 m above ground level, respectively. Adjacent to the sampling sites is a low building, which is 20 meters away. In addition, another building across from the sampling sites is 40 meters away. Therefore, the street configuration is considered having a minor effect on the environmental conditions of the sampling sites. Ambient PM2.5 samples were collected on pre‐weighed 37 mm Teflon‐membrane filters (Pall Corporation, Ann Arbor, MI, USA) using multiple Harvard Impactors (Air Diagnostics and engineering, Inc., Harrison, ME, USA) on the balconies. The 24‐h integrated filter samples (10:30–10:30 local time) were collected twice a week (every Monday and Thursday) during the period between 19 October and 22 December 2020. The sampled Teflon‐membrane filters were re‐weighed using a microbalance (UMX2, Mettler‐Toledo International Inc., Greifensee, Switzerland) in a temperature‐ and humidity‐controlled chamber (21–25°C and 30–40%), followed by measurements of absorbance using a Smoke Stain Reflectometer (Diffusion System Ltd, London, UK). Concentrations of 16 trace elements (Mg, Al, Si, S, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Ba, and Pb) on the filters were quantified using a non‐destructive energy‐dispersive X‐ray fluorescence method (Epsilon 4, Malvern Panalytical Ltd., Almelo, Netherlands). Subsequently, aqueous extracts of the filters were analyzed for water‐soluble organic carbon (WSOC) and eight water‐soluble inorganic ions (Cl–, NO3–, SO42–, Na+, NH4+, K+, Mg2+, and Ca2+) using a total organic carbon analyzer (Aurora 1030W TOC analyzer, OI Analytical, College Station, TX, USA) and ion chromatography (model DX‐120, DIONEX, Sunnyvale, CA, USA), respectively. Lab and field blank samples were collected and analyzed to detect contamination in sample handling. The method detection limit (MDL) was determined as the triple standard deviation calculated from replicate analyses of lab blank samples (elements) or the lowest concentration of the calibration curve (WSOC and ions). In each batch of filter samples one control sample with known amount of certified standards was measured for validation of data quality. The U.S. EPA’s PMF program (the latest version 5.0) was utilized to achieve the source apportionment analysis (Norris et al., 2014). As shown in Eq. (1), the iterative run simultaneously estimates factor contribution gik and factor profile fkj and converges after obtaining the best‐fit Q‐value (Norris et al., 2014): where n and m represent the number of samples and species, eij denotes the matrix of residuals between measurements and predicted values, and p is the number of factors. The measurement uncertainty uij, which is essential input in the modeling process, was calculated as (Liao et al., 2017): where MDLj represents the species‐specific method detection limit. The BDL value (xij < MDLj) was considered less reliable and thus was set at MDL/2. The corresponding uncertainty was assigned as MDL × 5/6 to down‐weight the BDL value. The signal‐to‐noise ratio (S/N) calculation in the PMF 5.0 software helps evaluate data quality of each species before running the model. Species having S/N smaller than 0.5 were excluded, while those with S/N ranging from 0.5 to 1 were down‐weighted. In addition, species containing more than 70% of BDL values were excluded. The optimal number of factors was estimated by the maximum individual column mean (IM) and the maximum individual column standard deviation (IS) accompanied by the interpretability of the retrieved profiles, which was examined by the mass fractions and explained variations [EVkj in Eq. (3)] of marker species (Lee et al., 1999; Liao et al., 2020): To achieve the total mass apportionment, PM2.5 mass was specified as the “Total Variable” in the PMF 5.0 software to eliminate the need for the post‐hoc regression process (Norris et al., 2014). Subsequently, the uncertainty of the ‘Total Variable’ was down‐weighted to avoid potentially influencing the model results. The Wilcoxon signed rank test with Bonferroni correction was conducted in the statistical analysis to examine whether the source contribution estimates were different between floor levels. The SAS 9.3 software was used to perform the test and the result with a p‐value < 0.05 was proposed statistically significant. A total of 114 samples (6 floors × 19 days) were collected. Among the measured species, Al, Ti, and Ba were excluded due to their poor data quality. In addition, the elements and their corresponding ions (Mg/Mg2+, S/SO42–, K/K+, and Ca/Ca2+) were well correlated (r > 0.84). To avoid double counting, the species with lower S/N (Mg, SO42–, K+, and Ca2+) were excluded. Table 1 shows the 20 variables included in the final model. PM2.5 mass ranged from 2.06 to 15.73 µg m–3, with an average concentration of 7.47 µg m–3, throughout the study period. WSOC was the most abundant component in PM2.5 at all sampling sites, followed by S and NO3–. Recently, WSOC has become a topic of health concern because of its cytotoxicity caused by reactive oxygen species (Daher et al., 2012; Wang et al., 2018; Jin et al., 2020). To provide sufficient sample size for receptor modeling, the data was pooled across all floors. A previous study has demonstrated that similar source profiles were retrieved using either the individual datasets or the pooled dataset (Xie et al., 2012). Based on the judging criteria that have been mentioned earlier, the 6‐factor solution was considered the best‐fit result. The six factors are shown in Fig. 1 and interpreted as follows. Factor 1 is identified by the abundance of S, K, NO3–, NH4+, and WSOC, accompanied by considerable EVs of K, Cr, Zn, and NH4+. K is a tracer of biomass burning activities (Cheng et al., 2009; Gugamsetty et al., 2012), whereas S, NO3–, and NH4+ could be generated from secondary formation of biomass burning emissions (Song et al., 2005; Thepnuan et al., 2019). Cr and Zn could be associated with industrial emissions (Dai et al., 2015; Lane et al., 2020). Factor 2, characterized by Abs, Cu, Zn, and WSOC, is interpreted as traffic related emissions. Absorbance and WSOC, as surrogates of elemental carbon (EC) and organic carbon (OC), are correlated with traffic emissions in the atmosphere (Jin et al., 2016; Qi et al., 2016; Wen et al., 2018). Cu and Zn can be emitted from the abrasion of brakes and tires (Gugamsetty et al., 2012; Pio et al., 2013; Ponco Wardoyo and Dharmawan, 2019). Zn is also found in lubricating oil used in motor vehicles (Huang et al., 1994; Todorović et al., 2020). Factor 3 is identified by both high loadings and high EVs of S and NH4+, accompanied by moderate EVs of V, Na+, and Mg2+. S and NH4+ are the major components of secondary aerosol (Yin et al., 2018; Ghosh et al., 2019). V is a known indicator of oil combustion, which is generally from ship emissions after the phase‐out of oil boilers in Taipei City (Pandolfi et al., 2011; Hsu et al., 2017; Liao and Wu, 2020). The presence of Na+ and Mg2+ and absence of Cl– indicated chloride depletion in aged sea salt (Tang et al., 1997; Sudheer et al., 2014; Adachi and Buseck, 2015). Therefore, Factor 3 is characterized as secondary aerosol that could be from local accumulation and regional transport, accompanied by marine and shipping aerosol. Factor 4 is recognized by the enrichment of Si, which is one of the major crustal elements, and could be interpreted as soil dust (Kim et al., 2003; Wimolwattanapun et al., 2011; Gugamsetty et al., 2012). Factor 5 is characterized by the high EV of Mn and moderate EVs of Abs, Fe, Ni, Pb, and Cl–, which could be emitted from industrial sources. There were few industries but several road construction sites around the sampling building. Therefore, emissions from the construction equipment and road materials are possible sources of these species, which can be deposited on the road and easily re‐suspended by traffic (Adachi and Tainosho, 2004; Carrero et al., 2013; Zhang et al., 2014). Factor 6 is characterized by high EVs of Cl–, Na+, and Mg2+, which are typical markers of sea salt (Taiwo et al., 2014; Carnelos et al., 2019). As shown in Fig. 2, the major contributor to PM2.5 during the sampling period was secondary aerosol (28.8%), followed by biomass burning/industry (24.4%) and traffic related (13.3%) sources. It should be noted that road traffic emissions (traffic related and contaminated road dust) accounted for 24.7%, making them the second largest contributor to PM2.5. Above findings were in line with previous studies in Taipei, where secondary aerosol and road traffic accounted for more than half of the contributions to PM2.5 (Ho et al., 2018; Liao et al., 2020; Liao and Wu, 2020). With regard to the contributors to WSOC, road traffic emissions accounted for 47.8%, followed by biomass burning/industry (24.0%). As shown in Table 1, the vertical distribution patterns were different among variables. Most species, including PM2.5 mass, showed the highest concentrations at the two lowest floor levels (1st and 7th), whereas S and K had the greatest concentrations at the 20th floor. Several species, including Si, S, Ca, V, Ni, NH4+, Mg2+, and WSOC, showed comparable (difference < 5%) concentrations between the 1st and 20th floors. It should be noted that WSOC exhibited a decreasing trend from the 1st to the 13th floor. The PM2.5 mass concentrations generally declined with height, except for that above the 17th floor. PM2.5 mass, Abs, Fe, Cu, Zn, and WSOC showed significant differences between lower (1st or 7th) and higher (13th, 17th, or 20th) floor levels. However, no statistically significant difference was found between the 17th and 20th floors. The distribution patterns of PM2.5 and traffic‐related components, such as Abs (the surrogate of EC) and Fe, were similar to those revealed by Zauli Sajani et al. (2018) during the cold season. Fig. 3 shows the vertical distribution patterns of the source contribution estimates by exhibiting the ratio of contribution at each floor to that at the 1st floor. The significant differences in source contribution estimates between floor‐levels were found for traffic related and contaminated road dust. In general, traffic related contributions declined with height, showing significant differences between lower (1st or 7th) and higher (13th, 17th, or 20th) floor levels. Although the contribution at the 20th floor was slightly higher than that at the 17th floor, no statistically significant difference was found between these two floors. Regarding the decreasing trend for the contributions from contaminated road dust, statistically significant difference was found between the 1st and 20th floors, potentially supporting that this factor was likely from local emissions. In contrast, secondary aerosol, the largest contributor to PM2.5, slightly increased with height although without any significant difference between floor levels. The increasing trend indicated the partial influence of regional transport (Wu et al., 2015). The other three source factors did not exhibit specific pattern of vertical variation and showed no significant differences in contributions between floor levels. The comparable contributions among floor levels might reveal multiple source origins (Liao et al., 2021). For example, soil dust could originate from local sources and also be transported from distant areas. Contributions of road traffic emissions significantly declined with height (29.3%–21.4%), which was in conformity with the findings in previous studies (Moeinaddini et al., 2014; Wu et al., 2015; Wang et al., 2016; Liao et al., 2020), and could explain the vertical variation of PM2.5 identified in this study. The above results indicated considerable contributions from local ground‐level emissions in the study area. Since road traffic emission is known to have a specific temporal pattern, closing windows during rush hour could prevent exposure to high levels of air pollutants. These results help improve our knowledge about the vertical characteristics of PM2.5 sources. It should be noted that long‐term exposure patterns to PM2.5 warrant further investigation since this study was conducted in a short period of time (19 days within two months). In this study we explored the vertical characteristics of potential PM2.5 sources apportioned by the PMF modeling. In addition to secondary aerosol, road traffic emissions (traffic related and contaminated road dust) represented the second largest contributor to PM2.5 during the study period. Contributions of road traffic emissions significantly declined with height, which were similar to the results of previous studies that collected PM2.5 samples at three floors from buildings lower than 40 m. In the present study, a broader vertical distribution pattern (six floor‐levels) up to 63 m was investigated. The non‐significant difference between the 17th and 20th floors suggested that the declining trend was limited at a certain height. With regard to the other source factors, they showed non‐significant vertical variation. Consequently, the vertical variation of PM2.5 identified in this study could be explained by road traffic emissions. Our findings suggested that enhanced traffic emission control policies could be beneficial in reducing PM2.5 exposure for residents living in apartment buildings in metropolitan areas. This study was supported in part by research grants from the Ministry of Science and Technology of Taiwan (MOST 106‐2221‐E‐002‐021‐MY3, 109‐2221‐E‐002‐057‐MY3, and 110‐2811‐E‐002‐502‐MY3) and the “National Taiwan University Higher Education Sprout Project (NTU‐111L8810)” within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.1 INTRODUCTION
2 METHODS
2.1 Data Collection
2.2 Data Analysis
3 RESULTS AND DISCUSSION
3.1 Source ApportionmentFig. 1. Factor profiles retrieved from the 6‐factor solution of the PMF model run.
Fig. 2. Source contribution estimates to PM2.5 mass at the sampling site during the study period.
3.2 Vertical DistributionFig. 3. Vertical distribution patterns of source contribution estimates to PM2.5 mass at the sampling site during the study period.
4 CONCLUSIONS
ACKNOWLEDGMENTS
DISCLAIMER
REFERENCES