Introduction

Fine particulate matter with an aerodynamic diameter ≤ 2.5 μm (PM2.5) is emitted from natural and anthropogenic sources, formed through atmospheric oxidative processes, and causes multiple adverse health effects. A variety of sources result in a complex chemical mixture in PM2.5. The physicochemical properties of PM2.5 facilitate penetration into the lungs where they can affect health through multiple mechanisms (US EPA 2019). Oxidative stress is one of the mechanisms that lead to inflammatory reactions in the lung cells and may disturb the immune system. Potential carcinogens such as Cr, As, and Pb are found in PM2.5. Epidemiological studies have reported DNA damage resulting from oxidative injuries caused by trace element species, such as Mn, Ni, and Pb (Kim et al. 2004; Prahalad et al. 2000), which are mostly associated with primary sources.

PM2.5 pollution is problematic in megacities on the western coast in South Korea (Seoul, Incheon, and Gwangju). While the annual average PM2.5 concentrations are declining (19.5 μg/m3 in 2020 to 19.1 μg/m3 in 2021), the country did not meet the World Health Organization’s annual air quality guideline value of 5 μg/m3 by in 2021 (IQAir 2022). The air quality in Korea typically maintains a moderate level and does not pose a significant concern during the warm seasons. When the ambient temperatures drop and energy consumption for heating in China increases, winter monsoons transport polluted air from the northwest region to the Korean peninsula, and oftentimes the stagnant atmospheric conditions hinder dispersion and induce severe haze events. Due to the generally unfavorable meteorological conditions in winter, the heating season in China plays a substantial role in the air quality in Korea. To address PM2.5 pollution in Korea, the Korean government implemented the Comprehensive Plan on Fine Dust in 2017, aiming to improve air quality and protect public health. This plan includes restricting the operation of old vehicles, limiting activities of factories and coal-fired power plants, and regulating emissions from illegal incinerations. In the Comprehensive Plan on Fine Dust for 2020 to 2024, Seoul, Incheon, and Gwangju, three large metropolitan cities on the west coast of South Korea, were identified as megacities with severe PM2.5 pollution. To reduce wintertime base concentration, the first seasonal management system policy was implemented in 2019. The seasonal management system was in effect from December to March, and in each following year, the regulations became more stringent.

Currently, positive matrix factorization (PMF) has become the most widely used source apportionment tool for PM2.5 (Hopke et al. 2020). PMF provides both qualitative and quantitative information on PM2.5 sources. However, conventional PMF may lose information from the measured concentrations due to atmospheric variations such as atmospheric chemistry and meteorological dilution. The dispersion normalized PMF (DN-PMF) has recently been introduced to reduce the effects of meteorology and has been extensively used since 2020 (Chen et al. 2022; Dai et al. 2020; Kim et al. 2022; Park et al. 2022; Song et al. 2021). For example, Dai et al. (2020) applied the DN-PMF to hourly data in China and distinguished the diel patterns of local sources; Chen et al. (2022) explored the effect of dispersion normalization on 24-h speciated samples in New York and reported that DN-PMF was able to better resolve seasonal PM2.5 source patterns. These studies support that DN-PMF can resolve sources that reflect their actual source strengths.

While an increasing number of source apportionment investigations focus on the capital area and specific industrial clusters in Korea, there is a notable gap in understanding the mixed effects of local and long-range transported pollutants along the western coast. Due to the many industrial enterprises dispersed along the western coast of Korea, the investigations of mixed effects of both local and long-range transport (LRT) sources are essential for preparing site-specific air quality management plans. Our study leveraged the simultaneously collected multi-site filter samples to characterize the sources of PM2.5 in the three megacities on the western coast of South Korea (Seoul, Incheon, and Gwangju). The DN-PMF was applied to the PM2.5 speciated data to resolve sources contributions with reduced meteorological effects. Hybrid receptor models were used to delineate the source locations affecting each megacity. As the study period includes the ongoing pandemic and wintertime air quality control periods, the impacts of pandemic-related and the efficacy of the seasonal management system policies were scientifically examined.

Materials and methods

Study areas

Seoul Metropolitan City is the capital of South Korea and is the country’s largest city with a population of 9.5 million people as of 2021. Due to its central role as the business and financial hub of South Korea, both the city and its surrounding areas are polluted by the large traffic volumes and various industrial activities. Incheon Metropolitan City, adjacent to Seoul and with a population of 2.9 million, comprises industrial complexes and busy ports. It has the second largest port in South Korea, which handled 3.3 million TEU in 2021. In addition, a large-scale coal-fired power plant (5080 MW) is located south of the city. Emissions from the port, factories, and transportation continuously deteriorate the air quality in Incheon. Lastly, Gwangju Metropolitan City, is also one of the largest cities in southwestern South Korea, with a population of 1.4 million. Multiple large-scale national industrial complexes are dispersed in the city; therefore, although not as densely populated as the other two metropolitan cities, pollution from traffic sources is relatively dominant in Gwangju, causing the city to focus its air quality control measures on vehicles. The three megacities are located downwind of China and are affected by transboundary air pollution introduced by westerly winds (Han et al. 2008; Koo et al. 2008). As a result, both local and LRT sources pose significant concerns for the air quality in Seoul, Incheon, and Gwangju; thus, their PM2.5 sources must be thoroughly investigated to develop effective mitigation measures.

Sampling site descriptions

Sampling in Seoul was conducted on the rooftop deck of the Graduate School of Public Health (37.46°N, 126.95°E) at Seoul National University. The Seoul site is located in the southern part of the city and heavy traffic, mountains, and residential areas coexist in the surrounding areas. The Incheon site was located on the rooftop (2.7 m above ground) of the National Institute of Environmental Research (37.57°N, 126.64°E), with seaports and national industrial complexes located 4 km to the west. Thus, in our study, Incheon represented a coastal metropolis with industrial complexes. Sampling in Gwangju was conducted on the roof (10 m above ground) of the 3rd building of the College of Engineering (37.18°N, 126.91°E) at Chonnam National University, which is surrounded by residential areas and expressways as well as national industrial complexes 3.3 km northwest of the site. All three study sites were situated downwind of the west coast of the Korean peninsula, where they were directly influenced by China.

Measurement of PM2.5 mass and chemical composition

In this study, the measurement campaign started from September 2020 to March 2022 for all sites. Ambient samples were collected using a three-channel low-volume air sampler with a flow rate of 16.7 L/min for 23-h (from 11:00 a.m. to 10:00 a.m. the next day). Sampling frequency was every third day from November to March and every sixth day during the remaining periods. The sampler consisted of a filter pack (URG-2000-30FG, URG, USA), cyclone (URG-2000–30-EH, URG, USA), dry gas meter, and vacuum pump for each channel. PM2.5 samples were collected on two types of Teflon filters and a Quartz filter each with 47 mm diameter. Teflon filters with a support ring (MTL, Pall Corporation, USA) were used for gravimetric and trace element measurement. Quartz filters (TISSUQUARTZ 2500-QAT-UP, Pall Corporation, USA) were prebaked at 450°C before the collection of carbonaceous species. The other type of Teflon filters (TF-1000, Pall Corporation, USA) were pretreated in anhydrous ethyl alcohol before being installed in the filter packs.

PM2.5 mass concentration was measured using a microbalance (Quintix125D, Sartorius, Germany). The Teflon filters used for the collection were stored in a desiccator and equilibrated in a controlled environment (temperature: 21 ± 1.4°C, relative humidity: 35 ± 5%) for at least 24-h before usage. All filters were weighed at least three times sequentially on a microbalance and the average values were recorded.

The carbonaceous species (OC: organic carbon; EC: elemental carbon) were measured using a carbon aerosol analyzer (Model 5 L, Sunset Laboratory Inc., USA). The analysis followed the thermal optical transmittance method of the National Institute for Occupational Safety and Health (NIOSH) 870 protocol.

For the ionic species, three anions (Cl−, NO3−, SO42−) and three cations (Na+, NH4+, K+) collected on the Teflon filters were measured by ion chromatography (ICS-1100, Thermo Fisher, USA). Filter samples were extracted in distilled water (resistivity of 18.2 MΩ \(\bullet\) cm) and then filtered using a 0.2 μm syringe. The total extraction volume for each sample was 32 mL.

The trace elements were measured using an energy-dispersive X-ray fluorescence spectrometer (EDXRF Spectrometer, Thermo Fisher, USA), and 20 trace element species (Mg, Al, Si, S, Cl, K, Ca, Ti, V, Cr, Mn, Ba, Fe, Ni, Cu, Zn, As, Se, Br, and Pb) were quantified. These trace elements were classified into crustal (Al, Si, K, Ca, Ti, Fe) and non-crustal (Mg, S, Cl, V, Cr, Mn, Ba, Ni, Cu, Zn, As, Se, Br, Pb) origins.

Mixing layer height

The modeled mixing layer heights obtained from the ERA5 during the sampling period for a target grid size of 1° × 1° were used for each site in this study. ERA5 is the fifth generation of the European Centre for Medium-Range Weather Forecasts and atmospheric re-analysis of the global climate can be used upon request. ERA5 hourly data on single levels from 1959 until the present are available on their website (https://cds.climate.copernicus.eu).

DN-PMF

The PMF model is a factor analysis model based on the least squares method that decomposes a matrix of sample data into a factor contribution matrix (\(g\)) and a factor profile matrix (\(f\)) and aims to find the solution that best minimizes residuals. It is widely used in source apportionment studies for its effective quantification and qualification of source contributions using concentration and uncertainty data. The basic PMF equation is as follows (Eq. 1):

$${x}_{ij}=\sum\nolimits_{k=1}^{p}{g}_{ik}{f}_{kj}+{e}_{ij}$$
(1)

where for a \(p\) number of independent sources, \({x}_{ij}\) is the \(j\)-th species concentration of the \(i\)-th sample, \({g}_{ik}\) is the particulate mass concentration from the \(k\)-th source contributing to the \(i\)-th sample, \({f}_{kj}\) is the mass fraction of the \(j\)-th species from the \(k\)-th source, and \({e}_{ij}\) is the residuals associated with the \(j\)-th species concentration measured in the \(i\)-th sample.

DN-PMF is an enhanced version of PMF that reduces the meteorological effects on concentrations by incorporating a ventilation coefficient (VC). VC is defined as the product of the mixing layer height (m) for period \(i\) (\({MLH}_{i}\)) and the mean wind speed (m/s) for period \(i\) (\(\overline{{u}_{i}}\)) (Eq. 2):

$${VC}_{i}={MLH}_{i}\times \overline{{u}_{i}}$$
(2)

The resulting VC is used to normalize the concentrations (\({C}_{VC,i}\)) by multiplying the measured concentrations (\({C}_{i}\)) for period \(i\) by \({VC}_{i}/{VC}_{mean}\), where \({VC}_{mean}\) is the average of period-specific VC values over the entire study period:

$${C}_{VC,i}={C}_{i}\times \frac{{VC}_{i}}{{VC}_{mean}}$$
(3)

The normalized concentrations computed by Eq. 3 and their uncertainties were used as input data for the PMF analyses by multiplying the normalization coefficient for each concentration. The resolved source contributions were scaled back to their original values by multiplying the inversed VCs for each corresponding day. Further details on DN-PMF can be found in recent literature (Dai et al. 2020). While averaging daily meteorological data cannot resolve diel patterns, it can resolve more accurate seasonal patterns (Chen et al. 2022).

The input data for the DN-PMF were created after quality assurance procedures, such as evaluating the ion balance and mass closure of each sample. Concentrations and uncertainties below the method detection limits (MDLs) were replaced with 1/2 MDL and 5/6 MDL, respectively. The uncertainties in PM2.5, carbonaceous, ionic, and trace element species were calculated separately. The uncertainty calculations for each chemical species are summarized in Table S1. By using SO42−, Cl−, and K+, we excluded double-counted species such as S, Cl, and K in the PMF analysis.

Conditional bivariate probability function (CBPF)

The conditional bivariate probability function (CBPF) model is a hybrid receptor model that combines PMF source contributions with meteorological data (wind direction and wind speed) (Uria-Tellaetxe and Carslaw 2014). The integrated data generates a polar plot that not only displays the directionality of a source but also the wind speed it was primarily affected by, denoted by color variations. The CBPF equation is as follows:

$${CBPF}_{\Delta \theta ,\Delta u}=\frac{{\left.{m}_{\Delta \theta ,\Delta u}\right|}_{C\ge x}}{{n}_{\Delta \theta ,\Delta u}}$$
(4)

where \({m}_{\Delta \theta ,\Delta u}\) is the number of samples in the wind sector \(\Delta \theta\) with wind speed interval \(\Delta u\) and a concentration \(C\) greater than the threshold value \(x\), and \({n}_{\Delta \theta ,\Delta u}\) is the total number of samples in the same wind direction-speed sector. The threshold value of the 75th percentile was used in this study. Wind speed and wind direction data for the study period were obtained from the Korea Meteorological Administration website (Seoul: Gimpo International Airport, Incheon: Incheon International Airport, Gwangju: Muan International Airport). CBPF plots were generated using Openair R package.

Joint potential source contribution function (J-PSCF)

A potential source contribution function (PSCF) model was used to estimate the possible source areas and LRT (Kim et al. 2018; Pekney et al. 2006; Zíková et al. 2016). In this study, we ran the PSCF model using the 96-h backward trajectories from the HYSPLIT 4 model. The PSCF value was computed using the following equation:

$$PSCF=\frac{{m}_{ij}}{{n}_{ij}}$$
(5)

where \({n}_{ij}\) is the number of endpoints that pass the \(ij\)-th grid cell and \({m}_{ij}\) is the number of endpoints that pass the \(ij\)-th grid cell when the source contributions were greater than the threshold value, which was set to the 75th percentile. Since the PSCF value of a grid cell with small \({n}_{ij}\) could be biased, a weighting function (\(W\)) was applied as follows:

$$W=\left\{\begin{array}{c}\begin{array}{c}1.0, \left(n>{3n}_{avg}\right)\\ 0.7, \left({1.5n}_{avg}<n\le {3n}_{avg}\right)\end{array}\\ 0.4, {(n}_{avg}<n\le {1.5n}_{avg})\\ 0.2,\left(n\le {n}_{avg}\right)\end{array}\right\}$$
(6)

In this study, the Joint-PSCF (J-PSCF) was used to analyze the potential source locations affecting multiple sites by combining the PSCF values for each receptor site using the following equation:

$${J-PSCF}_{ij}=\frac{\sum_{n=1}^{k}{({~}^{{m}_{ij}}\!\left/ \!{~}_{n}\right.)}_{n}}{\sum_{n=1}^{k}{({~}^{{n}_{ij}}\!\left/ \!{~}_{n}\right.)}_{n}}$$
(7)

where \(k\) is the number of receptor sites. As such, J-PSCF can minimize the trailing effects and overestimation of PSCF values near a receptor site.

Statistical analysis

The non-parametric Mann–Whitney U and Kruskal–Wallis tests at \(\alpha =0.05\) were used to test the significant differences between the two groups and multiple groups (such as inter-site comparisons) that did not pass the normality tests. Dunn’s method was used for post hoc analysis for the inter-site comparisons. Statistical analyses were performed using SigmaPlot (version 14.0).

Results

PM2.5 and chemical compositions

This study collected 222, 221, and 224 samples in Seoul, Incheon, and Gwangju, respectively, from September 2020 to March 2022. Intensive monitoring (every third day) was performed during the heating season (November-March) while less frequent monitoring (every sixth day) was performed during the remaining non-heating season (April–October). Here, the heating season refers to the general heating periods in North China. The average mass concentrations of PM2.5 were 24.2 ± 21.4 µg/m3, 24.6 ± 20.0 µg/m3, and 18.9 ± 16.5 µg/m3 in Seoul, Incheon, and Gwangju, respectively, all of which exceeded the annual National Ambient Air Quality Standard (NAAQS) in South Korea of 15 µg/m3. The Kruskal–Wallis one-way analysis of variance on ranks followed by Dunn’s method returned that the concentrations in Seoul and Incheon were significantly higher than in Gwangju (\(p\le 0.05\)). High concentration events (HCEs) exceeding the 24-h NAAQS of 35 µg/m3 occurred on 48, 40, and 18 days in Seoul, Incheon, and Gwangju, respectively. The Seoul and Incheon sites are within 30 km of each other, which explains the similar frequency of HCEs. HCEs mostly occurred during the heating season (Seoul: 44 days, 92%; Incheon 36 days, 90%; Gwangju: 16 days, 89%). The dates of HCEs at each site are presented in Table S2.

During the study period, PM2.5 was dominated by ionic species (Seoul: 61%; Incheon: 56%; Gwangju: 57%), followed by carbonaceous (Seoul: 23%; Incheon: 26%; Gwangju: 26%) and trace element species (Seoul: 16%; Incheon: 17%; Gwangju: 17%) (Figure S1a). During the heating season, ionic species accounted for an average of 62% of the total PM2.5 mass (Figure S1b), where nitrate was the most abundant species (Figure S2). Except for trace element species in Seoul, all chemical constituent concentrations were significantly higher during the heating season (\(p\le 0.05\)). Measured concentrations of PM2.5 mass and its chemical constituents are available in Table S3.

Source apportionment of PM2.5

PM2.5 speciated data were examined for ion balance and mass closure of each measurement to screen for outliers before creating the input data for the PMF analyses. After quality assurance, 184 samples in Seoul, 183 samples in Incheon, and 186 samples in Gwangju were used for DN-PMF source apportionment. The present study used filter samples from two monitoring projects. The second project started in July 2021, one month after the completion of the first project. The focus of these two projects was to investigate wintertime air quality so there are fewer summer samples available for the present study.

Ten PM2.5 sources in Seoul and Incheon and nine sources in Gwangju were resolved from the DN-PMF. The source profiles with DISP intervals and the temporal variation plots of the source contributions are presented in Figs. 1, 2, and 3. The number of factors for each site was chosen after examination of the scaled residual distributions, interpretability of the profiles, no displacement (DISP) swaps, good pairing of the bootstrap run contributions with the base run values, lack of mixed or split factors, and their interpretability.

Fig. 1
figure 1

Source profiles and time series plot of each source in Seoul from September 2020 to March 2022: a Secondary nitrate, b secondary sulfate, c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt, j industry. Gray bars indicate the relative concentration of each chemical species, error bars indicate the maximum and minimum displacement (DISP) intervals, filled circles indicate the explained values of each source of corresponding chemical species, and blank circles indicate the mean DISP values

Full size image
Fig. 2
figure 2

Source profiles and time series plot of each source in Incheon from September 2020 to March 2022: a Secondary nitrate, b secondary sulfate, c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt, j metal plating. Gray bars indicate the relative concentration of each chemical species, error bars indicate the maximum and minimum displacement (DISP) intervals, filled circles indicate the explained values of each source of corresponding chemical species, and blank circles indicate the mean DISP values

Full size image
Fig. 3
figure 3

Source profiles and time series plot of each source in Gwangju from September 2020 to March 2022: a Secondary nitrate, b secondary sulfate, c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt. Gray bars indicate the relative concentration of each chemical species, error bars indicate the maximum and minimum displacement (DISP) intervals, filled circles indicate the explained values of each source of corresponding chemical species, and blank circles indicate the mean DISP values

Full size image

The first source displayed high loadings and narrow DISP intervals of NO3− and NH4+ and was named secondary nitrate. The two marker species showed highly explained variations of NO3− (71%) and NH4+ (52%). Low temperatures and high relative humidity conditions accelerate the formation of secondary nitrate particles (Steinfeld 1998). On average, the secondary nitrate source accounted for 46% (10.8 μg/m3), 37% (9.06 μg/m3), and 36% (6.54 μg/m3) of PM2.5 in Seoul, Incheon, and Gwangju, respectively, and the contributions did not differ significantly from each other.

Secondary sulfate was characterized by high loadings and tight DISP intervals of SO42− and NH4+. This source explained 55% of SO42− and 25% of NH4+. Photochemical reactions in the summer promote the formation of secondary sulfate (Liu et al. 2015; Steinfeld 1998). However, the source contributions were generally higher during the cold months for all sites. This discrepancy might have resulted from the lack of summer samples or an abundance of primary sulfate sources during the heating season. The secondary sulfate source accounted for 12% (2.72 μg/m3) of PM2.5 in Seoul, 13% (3.22 μg/m3) in Incheon, and 15% (2.68 μg/m3) in Gwangju. There were no significant differences in source contributions among the three sites.

Biomass burning source features high loadings and tight DISP intervals of OC and K+. The biomass burning factor explained 21% and 59% of the OC and K+, respectively. Water-soluble potassium, which is present in biomass burning plumes, is a well-known tracer for biomass burning (Cheng et al. 2013; Echalar et al. 1995). The biomass burning source accounted for 10% (2.44 μg/m3) of the PM2.5 in Seoul, 12% (2.85 μg/m3) in Incheon, and 14% (2.62 μg/m3) in Gwangju.

The mobile source features high loadings and narrow DISP intervals for OC and EC. OC and EC are well-known tracers of traffic sources (Chow et al. 2003; El Haddad et al. 2009; Schauer et al. 2002), and their explained variations were high (OC: 31%, EC: 41%). Additionally, some Mg, Al, Si, and Ca were observed in the profiles. These crustal elements from non-exhaust roadside dust can also represent vehicular sources (Viana et al. 2008). Zn, Cu, and Fe, which arise from additives in motor oil (Zn) and abrasion of brake linings (Cu, Fe) have small DISP bounds in the source profiles of Incheon and Gwangju (Thorpe and Harrison 2008). The mobile source accounted for 9% (2.18 μg/m3) of the PM2.5 in Seoul, 12% (2.84 μg/m3) in Incheon, and 15% (2.77 μg/m3) in Gwangju. The source contributions were not significantly different between the heating and non-heating seasons demonstrating the year-round uniform contributions of mobile sources.

The soil source was characterized by crustal elements including Mg, Al, Si, Ca, Ti, Mn, and Fe (Nayebare et al. 2016; Aldabe et al. 2011; Song et al. 2006; Zhao and Hopke 2004). High explained variations and small DISP intervals were found for these species. The soil source accounted for 6% (1.48 μg/m3) of the PM2.5 in Seoul, 7% (1.60 μg/m3) in Incheon, and 3% (0.622 μg/m3) in Gwangju. Although obvious Asian Dust events were excluded from input data, the results in Seoul and Incheon still reflected some effects of LRT soil particles.

The next factor displayed high loadings and a narrow DISP interval for Cl− and was named waste incinerator. Cl− emissions are attributable to multiple sources including the combustion of polyvinylchloride plastics, and salt-rich food waste in Korea (Park et al. 2019; Yang et al. 2016). Additionally, small amounts of Pb and Zn were found in Incheon and Gwangju, which can be emitted from municipal waste incinerators (Greenberg et al. 1978a, b) and have been found in cyclone ash (Gao et al. 2002; Morishita et al. 2006; Yoo et al. 2002). Cl− explained 71% of the incinerator factor. In Incheon, species such as NH4+, K+, Pb, and Zn explained 16% on average of the source characteristics, whereas their explained values were negligible in Seoul (2%) and Gwangju (7%). The waste incinerator source accounted for 5% (1.17 μg/m3) of the PM2.5 in Seoul, 10% (2.46 μg/m3) in Incheon, and 7% (1.29 μg/m3) in Gwangju and was significantly higher during the heating season (p ≤ 0.001). This result is consistent with the conversion of particulate Cl− from HCl in a low thermal environment. Hence, the waste incinerator source contributions are not directly linked with actual emissions. Additionally, the waste incinerator source contribution was significantly higher in Incheon than in the other two cities (Kruskal–Wallis, p ≤ 0.05), whereas the differences between Seoul and Gwangju were not significant. There are nine municipal waste incinerators in Incheon and five in Seoul. While Seoul has a greater amount of generated waste of 1.8 million tons in 2021 than Incheon (1.2 million tons in the same year) (Ministry of Environment and Korea Environment Corporation 2022), the Incheon site has closer proximity to the point sources. Regardless of the total generated amount, the prevailing wind and the proximity to the emission stacks might have resulted in such discrepancies.

Coal combustion can be distinguished by its distinctive tracers such As and Pb (Gieré et al. 2006; Wang et al. 2006). Coal combustion can be found in coal-fired power plants, industrial activities such as cement-making or steel-making, and residential purposes. As and Pb explained 80% and 32%, respectively, of the coal combustion source. The coal combustion source contributions in Seoul and Gwangju were significantly higher during the non-heating season (p ≤ 0.001). Coal combustion sources accounted for 4% (0.986 μg/m3) of the PM2.5 in Seoul, 4% (0.924 μg/m3) in Incheon, and 5% (0.953 μg/m3) in Gwangju.

The oil combustion source was characterized by Ni and V, however, they were mixed with industrial sources at all sites. Ni and V are distinctive tracers of residual oil combustion from vessels (Jeong et al. 2017; Schembari et al. 2014). The oil combustion source profiles at all sites included industrial tracers including Fe, Cu, and Zn. Thus, they were named industry/oil combustion.

Cr, Mn, Fe, Cu, Zn, and Pb are tracer species of industrial emissions, in which Fe and Mn represent ferrous metallurgy and Cu, Zn, and Pb represent non-ferrous metallurgy sources (Querol et al. 2007; Swietlicki et al. 1996). The industry/oil combustion sources accounted for 2% (0.492 μg/m3) of the PM2.5 in Seoul, 2% (0.536 μg/m3) in Incheon, and 3% (0.483 μg/m3) in Gwangju. Seoul featured an additional industry source with narrow DISP intervals for Cr, Mn, Zn, and Pb, most likely indicating emissions from galvanizing activities (Begum and Hopke 2019; Dai et al. 2023). Pb is often added to Zn to enhance the reflective properties. This source contributed 9% (0.923 μg/m3) of the total mass in Seoul. In Incheon, a unique metal plating source was found, featuring large fractions with small DISP bands for Cr and Ni (Sun et al. 2017). Metal plating source contributed 2% (0.445 μg/m3) of the particles in Incheon. Additional identification of site-specific industrial sources can be attributed to the presence of larger and diverse array of industries in the capital areas.

The high presence of Na+ with a small DISP interval and some Cl− implied the influence of the marine aerosol. Fresh sea salt is reported to be found exclusively in the coarse particle fraction (Zhao and Gao 2008), and the low Cl− indicated that this source was aged sea salt. The retention time of sea salt particles in the atmosphere provides ample opportunities for chloride chemistry, such as depletion reactions with HNO3 and H2SO4, which release Cl− as HCl (Knipping and Dabdub 2003; Yao and Zhang 2012). A small portion of Cl− was observed in the source profile of Incheon, which implied mixing with fresh sea salt. This result was in good agreement with the proximity of the Incheon site to the West Sea. The aged sea salt factor explained 82% of Na+ at all sites. Aged sea salt sources accounted for 3% (0.521 μg/m3) of the PM2.5 in Seoul, 2% (0.586 μg/m3) in Incheon, and 3% (0.521 μg/m3) in Gwangju. There were no inter-site differences.

CBPF results

The source locations of each source at each site were investigated with CBPF and J-PSCF analyses. For secondary nitrate, both local and foreign source locations were explored. As a dominant source during the heating season, high CBPF values (Figs. 5a, 6a, 7a) were found only during the heating season at all sites. In Seoul, secondary nitrate increased when the southwesterly winds with wind speeds ranging from 0 to 10 m/s prevailed (Fig. 4a). Vehicles and industrial processes emit NOX precursors. Although the CBPF plots of the mobile, and industry-related sources did not necessarily match those of the secondary nitrate, the source locations coincided with the roads and industrial complexes. Large industrial complexes in Ansan City (Sihwa and Banwol Industrial Complex) were situated 20 km southwest of the sampling site in Seoul. In Incheon, high CBPF values for secondary nitrate were found in the south-southwest wind sector with wind speeds ranging from 2 to 12 m/s (Fig. 5a). Lower probabilities existed for wind sectors of ≤ 2 m/s, implying no strong local sources. These influences may be attributed to the emissions from more distant industrial complexes which are south-southwest of the Incheon sampling site, and to the road traffic nearby. Secondary nitrate was transported to the sampling site in Gwangju by strong southwesterly winds with wind speeds of ≥ 6 m/s and by gentle north-northwesterly winds with wind speeds of ≤ 2 m/s (Fig. 6a). Multiple national industrial complexes exist within 15 km of the sampling site.

Fig. 4
figure 4

CBPF results in Seoul: a Secondary nitrate, b secondary sulfate, c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt, j industry (For each source, the left box depicts the results for the heating season and the right box for the non-heating season)

Full size image
Fig. 5
figure 5

CBPF results in Incheon: a Secondary nitrate, b secondary sulfate), c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt, j metal plating (For each source, the left box depicts the results for the heating season and the right box for the non-heating season)

Full size image
Fig. 6
figure 6

CBPF results in Gwangju: a Secondary nitrate, b secondary sulfate, c biomass burning, d mobile, e soil, f waste incinerator, g coal combustion, h industry/oil combustion, i aged sea salt (For each source, the left box depicts the results for the heating season and the right box for the non-heating season)

Full size image

Secondary sulfate source contributions were dominant during the heating season in Seoul and Incheon, while it showed a notable peak during the non-heating season in Gwangju. This result is contradictory to previous results carried out in the Seoul Metropolitan City area where higher secondary sulfate contribution during the non-heating seasons was observed (Shin et al. 2022; Kim et al. 2018; Heo et al. 2009). However, the disparity in the number of summer and winter samples likely had a substantial influence on the mean values for the two seasons. The high concentration of sulfate during the heating season suggests the presence of primary sulfate emissions from residential coal combustion in China (Dai et al. 2019; 2023) affecting Seoul and Incheon. The sulfur contents in residual oil used to power marine diesel engines vary. However, ship emissions are an important source of SO2 and primary sulfate (Agrawal et al. 2010). Thus, year-round marine traffic in the West Sea would constantly supply sulfate in the atmosphere. Also, previous studies found residential coal combustion in China to be an important source of primary sulfate in winter (Dai et al. 2019; Li et al. 2019). These sulfate particles formed outside of the Korean territory could be the reason for high source contribution during the heating season, but the limited summer samples in the present study prohibit definitive conclusions.

During the heating season in Seoul, influences from the west-northwest and the south were present. Strong west-northwesterly winds with wind speeds of 10 m/s and less strong southerly winds with wind speeds ranging from 5 to 10 m/s contributed to the transportation of secondary sulfate to the sampling site (Fig. 4b). Particularly, the influence primarily from the southwest of the Seoul site was noticeable. Large-scale coal-fired power plants are densely situated along the western coastal areas of the Korean peninsula including Yeongheung, Taean, Dangjin, and Boryeong, all of which are southwest of the Seoul site. In Incheon, the sulfate contributions increased when very strong south-southwesterly winds with wind speeds of 12 m/s prevailed (Fig. 5b). A large oil refinery, a widely known source of SO2, is 65 km southwest of the Incheon site (Zhang et al. 2023; Civan et al. 2015). However, the high temperature that burns fuel would emit sulfur in gaseous pollutants. Because of the very slow oxidation of S(IV) to S(VI) in winter, shipping emissions are still likely more important to particulate sulfate. There was influence from the northwesterly winds and the southeasterly winds but at lower wind speeds of 10 m/s. Since it takes time for SO2 to convert to sulfate, sulfate transported from China and port emissions likely had a substantial impact in the case of Incheon. In Gwangju, strong west-northwesterly winds and north-northwesterly winds with wind speeds of 10 m/s, even reaching 12 m/s in some cases, transported secondary sulfate from areas distant from the sampling site (Fig. 6b). Relatively weaker southwesterly winds with wind speeds of 6 m/s also contributed to the source contribution increase.

The biomass burning source in Seoul was dominant during the heating season. The source contribution increased when the fast winds (wind speeds from 5 to 10 m/s) from the northwest, west, and southwest were dominant (Fig. 4c). Several farmlands within 30 km of the sampling site corresponded to the high-probability wind sectors in the CBPF. There are large farming areas in Gimpo-si, which is 38 km northwest of the Seoul site. In addition, furniture manufacturers are densely located on the outskirts of Gimpo-si, where waste wood materials are burnt illegally, often causing fires in the area. Local effects were also identifiable in the 0 ~ 5 m/s northwesterly wind sector where clusters of meat-cooking restaurants such as in Samgak-ji, are situated. Thus, the combination of field burning, wood waste, and cooking emissions characterize the biomass burning source in Seoul. In Incheon, the biomass burning source contribution was significantly higher during the non-heating season which was contrasting with the source contribution pattern in Seoul and Gwangju (p ≤ 0.001). The CBPF plot of the non-heating season showed that there was a high probability of the source location being northeast of the sampling site in Incheon (Fig. 5c). Several camping sites were identified in the east-northeast direction of the sampling site. A possible explanation of this can be the increased recreational activities during the second year of the pandemic, which will be discussed in the next section. In Gwangju, the biomass burning source contribution increased when the southwesterly winds with a wind speed of 10 m/s were dominant (Fig. 6c). Gwangju is a less populated megacity and is surrounded by more rural areas compared to Seoul and Incheon. Large rural areas consisting of active agricultural lands exist within 40 km southwest of the sampling site in Gwangju. Unlike Seoul and Incheon, there were no seasonal differences in Gwangju. In many Asian countries, crop residue burning after barley and wheat harvests occurs from spring to summer while burning after rice harvest occurs in autumn. Thus, a year-round field burning seems to occur in these areas. Much weaker CBPF probabilities of the increased source contribution were also present for northwesterly winds with wind speeds of ≤ 8 m/s and under stagnant atmospheric conditions with wind speeds of ≤ 2 m/s.

The source contributions of the mobile source increased when the atmosphere was stagnant with wind speeds of ≤ 4 m/s at all sites implying that the source locations were nearby (Fig. 4d). Local roads and motorways (Gangbyeon-daero and Olympic-daero) which are within 6 km constantly suffer from traffic congestion. These local impacts seem to be reflected in the CBPF plots. Slow winds with wind speeds of ≤ 4 m/s from all directions except the southwest caused increases in the mobile source contribution in Incheon (Fig. 5d). The Capital Region Second Ring Expressway crosses the sampling site from south to north, while the Incheon International Airport Expressway crosses from east to west. Traffic passing on these two expressways likely influences the Incheon sampling site. Like Incheon, the sampling site in Gwangju was very close to roads with high traffic volumes. Slow winds with wind speeds of 4 m/s or less from all around increased the source contributions in Gwangju (Fig. 6d). A major expressway, Honam expressway, and busy residential roads lie 1 km north of the sampling site.

For the soil source, Asian Dust events were excluded from the input data in the DN-PMF source apportionment. However, some soil particle transport was apparent in the CBPF plots. In Seoul, the soil source was from strong northwesterly winds with wind speeds of 10 m/s and above during the heating season, and the southeasterly winds with wind speeds ≤ 10 m/s during the non-heating season (Fig. 4e). In Seoul, snow precipitation was observed for a total of 49 days which mostly occurred from November through March. CaCl2 is widely used in Korea to prevent roads from freezing, and in some cases, a mixture of sand is sprayed on the road. These practices would provide opportunities for soil particles to be airborne after the road dries (Squizzato et al. 2018). The soil source in Incheon was transported from northwesterly winds with wind speeds ranging from 6 to 12 m/s that prevailed during the non-heating season (Fig. 5e). Suspended soil particles during cultivation in the rural areas of Gimpo-si may be the source. Much weaker probabilities in the same wind sector were also present during the heating season, also reflecting the transport of crustal dust from open soil agricultural fields. In Gwangju, high CBPF probabilities were found during the heating season. There were high probabilities in the southwest wind sector with wind speeds ranging from 2 to 10 m/s and in the north-northeast wind sector with wind speeds of 10 m/s (Fig. 6e). Each wind sector corresponds to the farmlands in the southwest and the mountains in the northeast of the sampling site.

The influences of the waste incinerator sources were more prominent during the heating season at all sites due to the poorer dispersion conditions and the limited number of summer samples. In Seoul, the source contribution increased when strong westerly winds with wind speeds of 10 m/s prevailed (Fig. 4f). Moderate CBPF values were also found in the southern wind sector with lower speeds. Several waste incineration facilities are found in Gwangmyeong, Ansan, and Bucheon, and identified by the CBPF results. In Incheon, the source contributions increased when south-southwesterly winds with wind speeds ranging from 6 to 12 m/s were dominant (Fig. 5f). Also, gentle south-southeasterly winds contributed to the elevated source contributions. The incineration facilities in Songdo and Bucheon, 20 km south and 12 km southeast of the sampling site, respectively, were identified. Being adjacent to each other, the waste incinerator sources affected both Seoul and Incheon depending on the dominant wind direction and intensity. In Gwangju, very high CBPF probabilities were found when the north and north-northwesterly winds with strong wind speeds ranging from 9 to 12 m/s prevailed, but also when the southerly winds with wind speeds of 10 m/s were dominant (Fig. 6f). There is a waste incineration facility in Jeonju-si that is 72 km northnortheast of the Gwangju site, although the directionality does not precisely align with the CBPF results. The two local waste treatment facilities in Gwangju are not equipped with incinerators. Urban areas collect, recycle, and properly dispose of municipal wastes, while this is not the case in rural areas where municipal wastes are usually burnt. Therefore, regional influences seemed to be an important factor than the local ones in Gwangju.

The coal combustion source contributions were high during the non-heating seasons at all sites. Notably high source contributions were observed from late September to December 2020 and remained insignificant throughout the rest of the study period. The seasonal average source contributions showed a 3.3–12.1-fold difference between the maximum and minimum source contributions. The average source contribution and its tracer species (As and Pb) concentrations were typically low in summer. In Seoul, high CBPF values were found under stagnant atmospheric conditions and from the western wind sector with wind speeds from 5 to 10 m/s (Fig. 4g). The CBPF hotspots of the coal combustion source were like those of the industry sources. The source locations were suggested as the industrial complexes in Guro-gu (8 km west) and in Incheon (30 km west), both comprising several metallurgical facilities. The coal combustion in Seoul likely originated from coking in steel-making processes. Similar directionality was found in Incheon as well, in which the source contributions increased when very strong west-northwesterly winds with wind speeds from 10 to 12 m/s prevailed (Fig. 5g). Additionally, the coal combustion activities affected the sampling site from the north at wind speeds of 5 m/s. Since the west-northwest of the sampling site in Incheon is the open sea, the high CBPF probability shown in that wind sector implies the impact of LRT from the eastern coastal Chinese cities. The local sources to the north may be from the coking processes, considering the locations of metal processing industries near the Incheon site. A large coal-fired power plant (Yeongheung Power Plant) is 40 km south-southwest of the sampling site in Incheon. However, the CBPF plots did not point in that direction. This result was probably due to the lack of southwesterly winds during the study period, so significant CBPF values were not found. In Gwangju, a very high probability was shown in the west-northwesterly and north-northwesterly wind sectors with strong wind speeds exceeding 8 m/s and up to 12 m/s (Fig. 6g). Like Incheon, this may be implying the influence of LRT sources. Moderate CBPF values were also found in the north-northeast wind sector with wind speeds of 6 m/s. Facilities that burn coal for various processes, such as crematories or metallurgical industries, situated in the close north-northeast direction, were possible emission sources.

Unlike previous results from our research group, the industrial source and the oil combustion were mixed as one source at all sites. For the industry/oil combustion source in Seoul, the CBPF results demonstrated high probabilities from all wind sectors during the non-heating season. A very high CBPF value in the northwesterly winds with a wind speed of 15 m/s implicated the transport of emissions from the port 30 km west of the sampling site in Seoul (Fig. 4h). The industry source, a separate source from the industry/oil combustion source, was most likely located near the west of the sampling site. The CBPF values were high when slow westerly winds with wind speeds of ≤ 5 m/s prevailed (Fig. 4j). Industrial facilities that handle galvanizing and non-ferrous metallurgy are in Geumcheon-gu, which is 8 km west of the sampling site in Seoul. The CBPF during the heating season points southwest to a hot-dip galvanizing facility in the Sihwa industrial complex, but the probability values are much smaller. In Incheon, the source contribution of the industry/oil combustion source increased under stagnant atmospheric conditions and when the fast southeasterly winds with wind speeds of ≥ 10 m/s prevailed (Fig. 5h). Many clusters of metallurgical industries were in the local areas of the Incheon site and two large national industrial complexes were found 28 km southeast of the sampling site in Incheon. The metal plating source in Incheon increased when north-northwesterly winds with wind speeds from 6 to 8 m/s affected the sampling site in Incheon (Fig. 5j). The Hagun industrial complex is 6 km northeast of the Incheon site, where several metal plating industries were active. In Gwangju, two different source locations were appearing in different seasons. During the heating season, a high probability value was found in the strong northwesterly winds with wind speeds of 12 m/s (Fig. 6h). However, southeasterly winds with wind speeds of 6 m/s greatly influenced the sampling site in Gwangju during the non-heating season. Slower westerly winds also increased the source contribution during the non-heating season at the Gwangju site. This source’s directionality was aligned with the locations of the national industrial complexes existing within 15 km west of the sampling site.

The CBPF results for the aged sea salt source pointed its source location to the nearest ocean. In Seoul, high CBPF values were found in the west, southwest, and south-southwestern wind sectors with high wind speeds of 10 m/s during the heating season (Fig. 4i). During the non-heating season, the inflow of aged sea salt particles to the Seoul sampling site from the west, but also present with the easterly winds as well. The monsoons related to the heating and non-heating seasons reflected the influences of the West Sea and the East Sea. In Incheon, the aged and fresh sea salt particles were present year-round from the northwest at high wind speeds exceeding 10 m/s (Fig. 5i). The West Sea was the likely source affecting both Seoul and Incheon. In Gwangju, the CBPF probabilities showed high values from the south and southeast wind sectors with fast wind speeds (heating season: 8 to 10 m/s, non-heating season: 6 m/s) (Fig. 6i). Considering that the prevailing winds in Gwangju were mostly from the north and northwest as depicted in the wind rose plots (Figure S6), the aged sea salt source is likely from the South Sea, and not the West Sea.

J-PSCF results

Our study investigated the potential source areas of secondary sources (secondary nitrate and secondary sulfate), biomass burning, and oil combustion sources commonly affecting all sites. The J-PSCF map (Fig. 7a) during the heating season demonstrated high PSCF values in Chinese provinces, such as Hebei, Shandong, and Jiangsu. The Beijing-Tianjin-Hebei (BTH) region is densely populated by iron and steel industries and has been reported to contribute a significant amount of NOX and SO2 emissions (Yang et al. 2019). Similarly, Shandong and Jiangsu Provinces have been reported to emit a significant amount of NOX from vehicular fleets (Song et al. 2019; Sun et al. 2016).

Fig. 7
figure 7

J-PSCF maps created for heating and non-heating seasons (heating season: a secondary nitrate, b secondary sulfate, c biomass burning, non-heating season: d secondary sulfate, e industry/oil combustion)

Full size image

For secondary sulfate, high PSCF values were found in Shanxi Province, BTH region, Shandong, and Jiangsu Provinces, which were largely identical to the potential source areas of secondary nitrate. Shanxi Province has abundant coal resources, and coal-related activities pose health risks from heavy metals and a significant amount of sulfur emissions can be inferred (Li et al. 2022; Su et al. 2021). The potential source areas of secondary sulfate also include the Yangtze River Delta (YRD) region and some parts of the East China Sea, which is well-known for busy marine vessel traffic (Bie et al. 2021; Kang et al. 2019). Intensive industrial production and traffic activities characterize the YRD region, where coal-fired power plants and industrial boilers emit SO2 gases that facilitate secondary formation (Jia et al. 2020). The PSCF map generated during the heating season included Jiaxiang City in Shandong Province, where coal-fired power plants operate to provide electricity for urban and industrial uses (Fig. 7b). Besides the use of coal in power plants, residential coal burning still exists in northern China. Although some cities like Beijing transitioned from residential coal to electricity, it is uncertain whether much of the rest of northern China stopped using coal for heating (Zhao et al. 2023). Sulfate emissions from coal combustion may have contributed to the formation and transboundary transport of secondary sulfate to the coastal areas of South Korea (Kuang et al. 2022). The PSCF map during the non-heating season indicated the potential source areas in China (Yellow Sea, Jiangsu, Zhejiang, and Fujian Provinces) and Japan (the coastal areas near Kumamoto) (Fig. 7d). The port of Shanghai (geographically located in the middle of Jiangsu and Zhejiang Provinces) on the coast of the Yellow Sea, one of the world’s busiest ports, may be responsible for intensive sulfur emissions from shipping activities (Wang et al. 2019). Heavily air-polluting enterprises exist along the southeast coastal part of China which includes cities such as Jiangsu, Anhui, and Zhejiang (Wei et al. 2014; Choi et al. 2013).

There are frequent volcanic eruptions in Kumamoto, and coincidentally a large eruption took place in Aso City, Kumamoto in October 2021. The SO2 emissions ranged from 2600 to 7500 tons per day from October 14 to October 27, 2021 (https://www.data.jma.go.jp/vois/data/fukuoka/rovdm/Asosan_rovdm/gas/gas.html). The upper 25th percentile concentration was highest in Gwangju (3.83 μg/m3) likely due to the closer distance to Japan compared to Seoul and Incheon. Abundant SO2 emissions from volcanic activities in Japan seem responsible for the substantial contributions to secondary sulfate concentrations during the non-heating season at all sites in Korea. However, large-scale volcanic eruptions are anomalous, so the potential sources of secondary sulfate found in Japan should not be generalized and should be limited to this study period only.

During the heating season, potential source areas of biomass burning source existed in Inner Mongolia, the BTH region, and Jiangsu Province (Fig. 7c). Previous studies have reported that agriculture-related open burning and biomass fuel consumption in rural areas are the main contributors to biomass burning emissions in the BTH region (Dong et al. 2022; Li et al. 2020; Zhao et al. 2022). A study conducted in Ordos City, located in the southeastern part of Inner Mongolia, found that regional biomass burning and biogenic sources accounted for 40% of the PM2.5 OC mass (Khuzestani et al. 2018). However, the Chinese government has steadily made efforts to ban open-field burning since 1999. In 2016, the Air Pollution Prevention and Control Law of the People’s Republic of China introduced heavy fines of RMB 500–1000 for straw burning. In some districts, local officials would be discharged or relegated if illegal burning occurred in their areas. Wang et al. (2022) reported decreasing annual trends in biomass burning areas in China from 2001 to 2019, and a steep drop was apparent from 2016. Therefore, Chinese open-field burning would not likely be an important source from 2020 to 2022 for Korea. Instead, biomass burning for home heating and cooking from northern China is much more attributable to the LRT sources affecting downwind areas.

The J-PSCF map of the industry/oil combustion source during the non-heating season displayed high probabilities in the East Sea and the East China Sea. Marine vessels passing the East China Sea may have contributed to the elevated PSCF values. The potential source areas in the East Sea included the land and coastal areas of Japan. In addition, industrial complexes were found in major cities and cities along the coastlines.

Discussion

Policy impacts on source contributions

Our study was conducted during the COVID-19 pandemic and included two seasonal management system policy implementation periods (the second and third year of the seasonal management system) on PM2.5. The unprecedented global pandemic, which started at the beginning of 2020, dramatically affected source emissions and human activities. Strong social distancing measures were introduced to the public to curb the spread of the virus. This intervention effort limited restaurant hours as well as the number of persons allowed in private gatherings. At the same time, the country’s efforts to reduce wintertime PM2.5 concentrations continued as the seasonal management system phased into its second year in December 2020.

The source contributions of biomass burning in the three megacities illustrated different patterns. The time variation plots for weekly and for the whole study period were created for detailed investigation (Fig. 8). The biomass burning source contributions tended to increase on the weekends at all sites, while the temporal variations differed greatly. Regardless of the pandemic, gatherings and leisure activities are likely to increase to some degree on the weekends. In Seoul, there were peaks of source contribution during the heating season of 2020 extending to March 2021 and the heating season of 2021 extending to March 2022, in which the peak during the latter period was much greater. The medians of the biomass burning source contributions during the first and second heating seasons were 2.10 μg/m3 and 3.93 μg/m3, respectively, and the source contribution from November 2021 extending to March 2022 was significantly higher than that from November 2020 to March 2021 (p ≤ 0.001). The stringent social distancing measures that prevented gatherings were eliminated starting from November 2021, indicating that the pandemic and its related policies indirectly impacted the biomass burning activities in Seoul. The lifting of these measures allowed active gatherings and revived the restaurant business, including meat-cooking restaurants. In Incheon, source contribution peaks were present not only during the heating seasons but also during the non-heating season.

Fig. 8
figure 8

Weekly variations and box plots by heating seasons of biomass burning source contributions in Seoul, Incheon, and Gwangju (Heating season in 2020: December 2020 – March 2021, Heating season in 2021: December 2021 – March 2022) (***: \(p \le 0.001\))

Full size image

As described in the previous section, the CBPF results for the non-heating season hinted at locations where there were campsites dispersed along the Han River. The pandemic changed the lifestyles of Koreans, including an increase in recreational activities according to a survey carried out in 2021 by the Korea Tourism Organization (Korea Tourism Organization 2022). Active camping activities and outdoor barbecuing may explain the distinctive source contribution peak during the non-heating season. Meanwhile, there were no significant differences in the source contributions between the heating and non-heating seasons in Gwangju. Gwangju is less populated than Seoul and Incheon and surrounded by agricultural land. Several peaks were observed during the harvesting season in Gwangju. Unlawful incinerations and field burning of crop residues are subject to control in the seasonal management system. However, despite these regulations, the source contribution in Gwangju increased on weekends, implying that agricultural wastes were still burnt, especially on weekends when public officers were off duty.

The regulations on the transportation sector are particularly strict in the seasonal management system. During the policy implementation period, the operation of Class 5 vehicles such as old diesel vehicles, is restricted on weekdays from 06:00 to 21:00. For Seoul and Incheon, the restrictions on Class 5 vehicles were implemented from the first year of the seasonal management system. In Gwangju, this regulation on the transportation sector will be implemented starting from the fifth year of the seasonal management system and thus outside of the study period for the present work. The mobile contribution in Gwangju was significantly greater than that of Incheon during the second year of the seasonal management system (Kruskal–Wallis, p ≤ 0.001). There were no significant differences between all sites during the third year of the seasonal management system. The median values of the source contribution during the non-seasonal management system and the seasonal management system period were compared to evaluate the reduction effect. The reduction amounts were 1.6%, 24.2%, and – 0.2% in Seoul, Incheon, and Gwangju, respectively (Table 1). The mobile source in Incheon had the highest median source contribution during the non-seasonal management system period and the lowest median value during the policy implementation period, showing the largest reduction rate among all sites (Figure S3). While reductions were found for Seoul and Incheon, there were none found in Gwangju during the policy implementation period. The median source contribution in Gwangju during the seasonal management system period was 2.43 µg/m3, which was 24% greater than the average of the other sites. Although the median values of the mobile source contribution were not statistically different between all sites, we were able to conclude the efficacy of the seasonal management system at least for Seoul and Incheon.

Table 1 Changes in mobile source contribution during the policy implementation period
Full size table

Conclusions

This study measured over 220 ambient PM2.5 speciated samples from September 2020 to March 2022 at three megacities on the western coast of South Korea. The average PM2.5 concentrations in Seoul (24.2 μg/m3), Incheon (24.6 μg/m3), and Gwangju (18.9 μg/m3) during the study period did not meet the WHO standard of 5 μg/m3. HCEs were frequently observed during the cold seasons when energy consumption typically increases, and dispersion becomes poorer. DN-PMF was used for the source apportionment of compositional data from Seoul, Incheon, and Gwangju to eliminate the meteorological covariances and enhance the accuracy of the resolved source strengths. Nine sources (secondary nitrate, secondary sulfate, biomass burning, mobile, soil, waste incinerator, coal combustion, industry/oil combustion, and aged sea salt) were common sources of PM2.5 in the western megacities of South Korea. Additional site-specific industrial sources in Seoul and Incheon were industry and metal plating, respectively. Despite their minuscule contributions to the total mass, these site-specific sources should be recognized in source apportionment because they will vary by different locations. All sites showed high secondary nitrate contributions and CBPF analyses implicated considerable local influence from gaseous precursors such as from industrial emissions and vehicle fleets near each sampling site. J-PSCF analyses indicated foreign impact from northeast China and Inner Mongolia on secondary nitrate, secondary sulfate, and biomass burning during the heating season, due to factory emissions and open field burning. Local biomass burning sources (meat-cooking or outdoor recreational activities) in Seoul and Incheon seemed to increase or display unexpected pattern due to pandemic-induced lifestyle changes. Weekend biomass burning peaks suggested unlawful agricultural burning in Gwangju. More effective monitoring and prevention efforts are required to target uncontrolled biomass burning in the future seasonal management system. The mobile source contributions decreased in Seoul and Incheon (and not in Gwangju) where vehicle operation restrictions were enforced during the policy period. Future studies should evaluate mobile source variations in other cities as the Korean government plans to expand vehicle restrictions during the seasonal management periods. While the current seasonal management system focuses on controlling for primary emissions, addressing secondary sources is crucial for effective air quality management. Future policies should explore the implementation of localized control measures for precursor emissions, as well as fostering intergovernmental collaboration to improve air quality in the upwind and downwind regions.