Is replacing missing values of PM2.5 constituents with estimates using machine learning better for source apportionment than exclusion or median replacement?

East Asian countries have been conducting source apportionment of fine particulate matter (PM2.5) by applying positive matrix factorization (PMF) to hourly constituent concentrations. However, some of the constituent data from the supersites in South Korea was missing due to instrument maintenance and calibration. Conventional preprocessing of missing values, such as exclusion or median replacement, causes biases in the estimated source contributions by changing the PMF input. Machine learning (ML) can estimate the missing values by training on constituent data, meteorological data, and gaseous pollutants. Complete data from the Seoul Supersite in 2018 was taken, and a random 20% was set as missing. PMF was performed by replacing missing values with estimates. Percent errors of the source contributions were calculated compared to those estimated from complete data. Missing values were estimated using a random forest analysis. Estimation accuracy (r2) was as high as 0.874 for missing carbon species and low at 0.631 when ionic species and trace elements were missing. For the seven highest contributing sources, replacing the missing values of carbon species with estimates minimized the percent errors to 2.0% on average. However, replacing the missing values of the other chemical species with estimates increased the percent errors to more than 9.7% on average. Percent errors were maximal at 37% on average when missing values of ionic species and trace elements were replaced with estimates. Missing values, except for carbon species, need to be excluded. This approach reduced the percent errors to 7.4% on average, which was lower than those due to median replacement. Our results show that reducing the biases in source apportionment is possible by replacing the missing values of carbon species with estimates. To improve the biases due to missing values of the other chemical species, the estimation accuracy of the ML needs to be improved.

Analysis of PM2.5 inorganic and organic constituents to resolve contributing sources in Seoul, South Korea and Beijing, China and their possible associations with cytokine IL-8.

China and South Korea are the most polluted countries in East Asia due to significant urbanization and extensive industrial activities. As neighboring countries, collaborative management plans to maximize public health in both countries can be helpful in reducing transboundary air pollution. To support such planning, PM2.5 inorganic and organic species were determined in simultaneously collected PM2.5 integrated filters. The resulting data were used as inputs to positive matrix factorization, which identified nine sources at the ambient air monitoring sites in both sites. Secondary nitrate, secondary sulfate/oil combustion, soil, mobile, incinerator, biomass burning, and secondary organic carbon (SOC) were found to be sources at both sampling sites. Industry I and II were only identified in Seoul, whereas combustion and road dust sources were only identified in Beijing. A subset of samples was selected for exposure assessment. The expression levels of IL-8 were significantly higher in Beijing (167.7 pg/mL) than in Seoul (72.7 pg/mL). The associations between the PM2.5 chemical constituents and its contributing sources with PM2.5-induced inflammatory cytokine (interleukin-8, IL-8) levels in human bronchial epithelial cells were investigated. For Seoul, the soil followed by the secondary nitrate and the biomass burning showed increase with IL-8 production. However, for the Beijing, the secondary nitrate exhibited the highest association with IL-8 production and SOC and biomass burning showed modest increase with IL-8. As one of the highest contributing sources in both cities, secondary nitrate showed an association with IL-8 production. The soil source having the strongest association with IL-8 production was found only for Seoul, whereas SOC showed a modest association only for Beijing. This study can provide the scientific basis for identifying the sources to be prioritized for control to provide effective mitigation of particulate air pollution in each city and thereby improve public health.

Factors affecting recent PM2.5 concentrations in China and South Korea from 2016 to 2020.

This study used observational data and a chemical transport model to investigate the contributions of several factors to the recent change in air quality in China and South Korea from 2016 to 2020. We focused on observational data analysis, which could reflect the annual trend of emission reduction and adjust existing emission amounts to apply it into a chemical transport model. The observation data showed that the particulate matter (PM2.5) concentrations during winter 2020 decreased by -23.4 % (-14.68 µg/m3) and - 19.5 % (-5.73 µg/m3) in China and South Korea respectively, compared with that during winter 2016. Meteorological changes, the existing national plan for a long-term emission reduction target, and unexpected events (i.e., Coronavirus disease 2019 (COVID-19) in China and South Korea and the newly introduced special winter countermeasures in South Korea from 2020) are considered major factors that may affect the recent change in air quality. The impact of different meteorological conditions on PM2.5 concentrations was assessed by conducting model simulations by fixing the emission amounts; the results indicated changes of +7.6 % (+4.77 µg/m3) and + 9.7 % (+2.87 µg/m3) in China and South Korea, respectively, during winter 2020 compared to that during winter 2016. Due to the existing and pre-defined long-term emission control policies implemented in both countries, PM2.5 concentration significantly decreased from winter 2016-2020 in China (-26.0 %; -16.32 µg/m3) and South Korea (-9.1 %; -2.69 µg/m3). The unexpected COVID-19 outbreak caused the PM2.5 concentrations in China to decrease during winter 2020 by another -5.0 % (-3.13 µg/m3). In South Korea, the winter season special reduction policy, which was introduced and implemented in winter 2020, and the COVID-19 pandemic may have contributed to -19.5 % (-5.92 µg/m3) decrease in PM2.5 concentrations.

Local and transboundary impacts of PM2.5 sources identified in Seoul during the early stage of the COVID-19 outbreak.

Countries in Northeast Asia have been regulating PM2.5 sources and studying their local and transboundary origins since PM2.5 causes severe impacts on public health and economic losses. However, the separation of local and transboundary impacts is not fully realized because it is impossible to change air pollutant emissions from multiple countries experimentally. Exceptionally, the early stage of the COVID-19 outbreak (January-March 2020) provided a cross-country experiment to separate each impact of PM2.5 sources identified in Seoul, a downwind area of China. We evaluated the contributions of PM2.5 sources compared to 2019 using dispersion normalized positive matrix factorization (DN-PMF) during three meteorological episodes. Episodes 1 and 2 revealed transboundary impacts and were related to reduced anthropogenic emissions and accumulated primary pollutants in Northeast China. Anthropogenic emissions, except for the residential sector, decreased, but primary air pollutants accumulated by residential coal combustion enhanced secondary aerosol formation. Thus, the contributions of sulfate and secondary nitrate increased in Seoul during episode 1 but then decreased maximally with other primary sources (biomass burning, district heating and incineration, industrial sources, and oil combustion) during episode 2 under meteorological conditions favorable to long-range transport. Local impact was demonstrated by atmospheric stagnation during episode 3. Meteorological condition unfavorable to local dispersion elevated the contributions of mobile and coal combustion and further contributed to PM2.5 high concentration events (HCE). Our study separates the local and transboundary impacts and highlights that cooperations in Northeast Asia on secondary aerosol formation and management of local sources are necessary.

α-Fe2O3 nanoparticles and hazardous air pollutants release during cooking using cast iron wok in a commercial Chinese restaurant.

Long-term exposure to fine particles (PM2.5), ultrafine particles (UFPs), and volatile organic compounds (VOCs) emissions from cooking has been linked to adverse human health effects. Here, we measured the real-time number size distribution of particles emitted when cooking two served food in Chinese restaurants and estimated the emission rate of UFPs and PM2.5. Experiments were conducted under a control hood, and both online measurement and offline analysis of PM2.5 were carried out. The measured emission rates of PM2.5 generated from deep-frying and grilling were 0.68 ± 0.11 mg/min and 1.58 ± 0.25 mg/min, respectively. Moreover, the UFPs emission rate of deep-frying (4.3 × 109 #/min) is three times higher than that of grilling (1.4 × 109 #/min). Additionally, the PM2.5 emission of deep-frying was comprised of a considerable amount of α-Fe2O3 (5.7% of PM2.5 total mass), which is more toxic than other iron oxide species. A total of six carcinogenic HAPs were detected, among which formaldehyde, acrolein, and acetaldehyde were found to exceed the inhalation reference concentration (RfC) for both cooking methods. These findings can contribute to future evaluation of single particle and HAPs emission from cooking to better support toxicity assessment.

Health effects of PM2.5 constituents and source contributions in major metropolitan cities, South Korea.

Ambient PM2.5 is one of the major risk factors for human health, and is not fully explained solely by mass concentration. We examined the short-term associations of cause-specific mortality (i.e., all-cause, cardiovascular, and respiratory mortality) with the 15 chemical constituents and sources of PM2.5 in four metropolitan cities of South Korea during 2014-2018. We found transition metals consistently showed significant associations with all-cause mortality, while the effects of other constituents varied across the cities and for cause of death. Carbonaceous components strongly affected the all-cause, cardiovascular, and respiratory mortality in Daejeon. Secondary inorganic aerosols, SO42- and NH4+, showed significant associations with respiratory mortality in Gwangju. We also found the sources from which species closely linked to mortality generally increased the relative mortality risks. Heavy metal markers from soil or industrial sources were significantly associated with mortality in all cities. However, several sources influenced mortality despite their marker species not being significantly associated with it. Secondary nitrate and secondary sulfate sources were linked to mortality in DJ. This could be attributed to the deep inland location, which might have facilitated formation of secondary inorganic aerosols. In addition, primary sources including mobile and coal combustion seemed to have acute impacts on respiratory mortality in Gwangju. Our findings suggest the necessity of positive matrix factorization (PMF)-based approaches for evaluating health effects of PM2.5 while considering the spatial heterogeneity in the compositions and source contributions of PM2.5.

Source apportionment of PM2.5 in Seoul, South Korea and Beijing, China using dispersion normalized PMF.

East Asian countries experience severe air pollution owing to their rapid development and urbanization induced by substantial economic activities. South Korea and China are among the most polluted East Asian countries with high mass concentrations of PM2.5. Although the occurrence of transboundary air pollution among neighboring countries has been recognized for a long time, studies involving simultaneous ground-based PM2.5 monitoring and source apportionment in South Korea and China have not been conducted to date. This study performed simultaneous daily ground-based monitoring of PM2.5 in Seoul and Beijing from January to December 2019. The mass concentrations of PM2.5 and its major chemical components were analyzed simultaneously during 2019. Positive matrix factorization (PMF) as well as dispersion normalized PMF (DN-PMF) were utilized for the source apportionment of ambient PM2.5 at the two sites. 23 h average ventilation coefficients were applied for daily PM2.5 chemical constituents' data. Nine sources were identified at both sites. While secondary nitrate, secondary sulfate, mobile, oil combustion, biomass burning, soil, and aged sea salt were commonly found at both sites, industry/coal combustion and incinerator were identified only at Seoul and incinerator/industry and coal combustion were identified only at Beijing. Reduction of the meteorological influences were found in DN-PMF compare to C-PMF but the effects of DN on mobile source were reduced by averaging over the 23 h sampling period. The DN-PMF results showed that Secondary nitrate (Seoul: 25.5%; Beijing: 31.7%) and secondary sulfate (Seoul: 20.5%; Beijing: 17.6%) were most dominant contributors to PM2.5 at both sites. Decreasing secondary sulfate contributions and increasing secondary nitrate contributions were observed at both sites.

Fossil-driven secondary inorganic PM2.5 enhancement in the North China Plain: Evidence from carbon and nitrogen isotopes.

Measuring isotopic ratios in aerosol particles is a powerful tool for identifying major sources, particularly in separating fossil from non-fossil sources and investigating aerosol formation processes. We measured the radiocarbon, stable carbon, and stable nitrogen isotopic composition of PM2.5 in Beijing (BJ) and Changdao (CD) in the North China Plain (NCP) from May to mid-June 2016. The mean PM2.5 concentrations were 48.6 ± 28.2 µg m-3 and 71.2 ± 29.0 µg m-3 in BJ and CD, respectively, with a high contribution (∼66%) from secondary inorganic aerosol (SIA; NO3-, NH4+, and SO42-). The mean δ13C of total carbon (TC) and δ15N of total nitrogen (TN) values differed significantly between the two sites (p-value of <0.001): -25.1 ± 0.3‰ in BJ and -24.5 ± 0.4‰ in CD and 10.6 ± 1.8‰ in BJ and 5.0 ± 3.1‰ in CD, respectively. In BJ, the average δ15N (NH4+) and δ15N (NO3-) values were 12.9 ± 2.3‰ and 5.2 ± 3.5‰, respectively. The ionic molar ratios and isotopic ratios suggest that NO3- in BJ was formed through the phase-equilibrium reaction of NH4NO3 under sufficient NH3 (g) conditions, promoted by fossil-derived NH3 (g) transported with southerly winds. In BJ, fossil fuel sources comprised 52 ± 7% of TC and 45 ± 28% of NH4+ on average, estimated from radiocarbon (14C) analysis and the δ15N and isotope mixing model, respectively. These multiple-isotopic composition results emphasize that PM2.5 enhancement is derived from fossil sources, in which vehicle emissions are a key contributor. The impact of the coal source was sporadically noticeable. Under regional influences, the fossil fuel-driven SIA led to the PM2.5 enhancements. Our findings demonstrate that the multiple-isotope approach is highly advantageous to elucidate the key sources and limiting factors of secondary inorganic PM2.5 aerosols.