Source apportionment and dynamics of PM_2.5 across regions during and after the coronavirus 2019 pandemic

Abstract

The coronavirus disease 2019 (COVID-19) pandemic was reported to improve air quality. However, the effects of lockdown on PM₂.₅ and its chemical components based on source-specific characteristics remain unclear. This study evaluated and compared the components of PM₂.₅ across three target sites—Seoul, Ansan–Siheung, and Wonju—based on source distribution patterns during the pandemic and post-pandemic periods. The results revealed that total ion concentrations decreased by 7–19% at the target sites, while trace element concentrations increased by 19–29% in the post-pandemic period. Carbon concentrations showed region-specific trends. Components associated with secondary formation, such as SO₄²⁻, NO₃⁻, and NH₄⁺, decreased, whereas natural-origin components, such as Cl⁻, Na⁺, Mg²⁺, and Ca²⁺, increased, indicating enhanced dust resuspension. Furthermore, a positive matrix factorization (PMF) analysis demonstrated a predominance of primary emission sources and resuspended dust in the post-pandemic period, with a decline in contributions from secondary formation sources. These regional variations were primarily influenced by the extent of socioeconomic activities, with Seoul exhibiting a higher contribution from primary emissions and suspended dust. This study emphasizes the importance of establishing region-specific air quality improvement policies by scientifically identifying the impacts of anthropogenic activity on PM₂.₅ components and their sources.

Introduction

The first case of coronavirus 2019 (COVID-19) in the Republic of Korea was reported on January 20, 2020. The number of cases increased radically after a regional outbreak on February 18, 2020, when members of a religious group gathered in Daegu Metropolitan City. On February 29, 2020, the Korean government officially proposed a “social distancing” policy to prevent the spread of COVID-19 which came into effect after April 19, 2020. The enforcement of this policy resulted in the mandatory closure of public facilities (e.g., libraries, museums, and swimming pools), suspension of operations of facilities vulnerable to infection (e.g., religious facilities, indoor sports facilities, and entertainment facilities), and encouragement of work-from-home culture (the proportion of employees working-from-home/remote working reaching 32.3% in August 2021). This revolutionary action markedly restricted most social and industrial activities^1,2,3,4,5,6. All restrictions were finally lifted on April 18, 2022⁷. By 2024, people returned to their normal life, as in pre-COVID era.

During these 2-year and 1-month period of “social distancing” in the Republic of Korea, the routine daily activities of most Koreans and industrial activities in general were fiercely restricted. Several studies have reported the positive impacts of this blanket restriction policy on air quality^8,9,10,11,12. In previous studies, it has been reported that air quality improved significantly during the COVID-19 lockdown period, with PM_2.5concentrations decreasing by 12% on average in the 50 most polluted capital cities worldwide¹³. Bao and Zhang¹⁴ analyzed the impact of reduced human mobility on ambient air quality in 44 northern cities from January to March 2020, reporting an average decrease of 7.8% in the AQI, along with reductions of 5.93% in the PM_2.5. In addition, studies from global cities such as New York, Los Angeles, Rome, and Zaragoza demonstrated significant reductions in PM_2.5concentrations¹⁵, with decreases of up to 32% in New York and 58% in Zaragoza compared to the same periods in previous years. These improvements were mainly due to decreased vehicle traffic and reduced industrial operations, highlighting the significant impact of mobility restrictions and economic slowdowns on air quality. However, most studies focused on reporting the effects of air quality improvement regarding PM_2.5, PM₁₀, NO₂, CO, and SO₂, while only a few studies^13,16 have addressed the changes in the chemical components of PM_2.5, focusing primarily on health impacts. Thus, studies that comprehensively consider emission changes per source and thoroughly investigate the real causes of air quality improvements are still lacking.

PM_2.5—particulate matter with an aerodynamic diameter of ≤ 2.5 μm, has a larger surface area per unit volume than PM₁₀ with an aerodynamic diameter of 10 μm. This characteristic of PM_2.5leads to the adsorption of diverse hazardous air pollutants (HAPs), such as heavy metals^17,18. These adsorbed HAPs are sourced from anthropogenic activities along with varying chemical components and composition ratios depending on the distribution of nearby sources^19,20,21.

Social lockdowns owing to COVID-19 are expected to have significantly impacted the chemical composition of PM_2.5 due to changes in emissions from different sources. Previous studies have confirmed varying distribution patterns of PM_2.5components during the COVID-19 lockdown^10,22, witnessing decreases in NO₃^- and black carbon as well as increases in SO₄^2-, indicating the dominance of stationary emission sources. Conversely, Ma et al.²³ reported cases where the concentration of NO₃^- increased after the lockdown. Notably, the COVID-19 pandemic significantly altered the operational characteristics of the major air pollutant emission sources. An in-depth analysis of the impact of these unprecedented events on air quality is fundamental to framing air pollution mitigation policies in the future.

This study aimed to investigate the changes in the distribution of PM_2.5 and its chemical components (ions, carbons, and trace elements) following the implementation of the social distancing policy during COVID-19. Additionally, we identified the impacts of this scenario through the source apportionment of PM_2.5. Achieving these goals required dividing the research period into the pandemic (social distancing) and post-pandemic (back-to-normal) periods. Additionally, the selected target sites with varying source distribution characteristics (Seoul, Wonju, and Ansan–Siheung) were compared and analyzed for air quality changes. We applied the positive matrix factorization (PMF) model of the United States Environmental Protection Agency (EPA) to analyze the changes in the contributions of primary and secondary pollutants during these periods and identified the patterns of emission source recovery after the lifting of the lockdown.

Methods and materials

Overview of the COVID-19 outbreak in South Korea and PM_2.5 measurement

The research period was classified into pandemic and post-pandemic periods based on the “social distancing” policy, which was effective until April 17, 2022. After which, the lockdown was lifted entirely when the COVID-19 situation normalized in South Korea (Table 1).

Table 1 Timeline of COVID-19 policy phases in South Korea and PM_2.5 sampling periods.

Sampling sites

We collected PM_2.5 samples from the three locations: Seoul, a representative urban area in the Republic of Korea; Wonju, a city with a high congregation of old agricultural-based industrial sites; and Ansan–Siheung, two cities with two national industrial complexes (Banwol and Sihwa Industrial Complexes) and a complex of various industries, including machinery, petrochemicals, and metal processing. Sampling was conducted between August 1, 2020 and March 31, 2024 (Fig. 1, Table S1).

Fig. 1

Sampling sites of the study area. Maps were generated using Python and contextily with ESRI World Imagery basemaps (© Esri, Maxar, Earthstar Geographics, and the GIS User Community). Attribution is automatically included in the image.

Seoul

Seoul, the capital of the Republic of Korea, had a population of 9.4 million as of 2023, making it the most densely populated area in the country (15,533 persons/km²)²⁴. The data from Statistics Korea reveal that the main source of PM_2.5in Seoul pre-COVID-19 outbreak (2017–2019) was suspended dust (40.3%, 1,333,611 kg), followed by mobile sources, such as heavy construction equipment (37.4%, 1,238,959 kg)²⁵. Additionally, as of 2023, the total number of registered vehicles in Seoul was 3.19 million vehicles, accounting for approximately 12% of the nationwide total of 25.95 million vehicles, making it the second highest among all regions²⁶. Accordingly, five sites with high population mobility and heavy vehicle traffic were selected as the sampling sites.

Ansan–Siheung

Ansan–Siheung is located to the southwest of Seoul near the West Sea. Ansan has approximately 660,000 residents, whereas Siheung has 470,000, bringing the total population to 1.01 million. The population density is 4,252 and 3,438 persons/km²in Ansan and Siheung, respectively, and both cities are adjacent to the metropolitan area. Ansan–Siheung is located near two national industrial complexes, Banwol and Sihwa, which host several industrial activities, such as manufacturing, petrochemical, machinery manufacturing, and metal processing. These two industrial complexes encompass 18,670 businesses²⁷. In 2015, these industrial complexes faced multiple civil complaints from neighbors concerning pollutants and odors related to production activities²⁸. As of 2021, the number of employees in the complexes were approximately 240,000 and their daily commutes contribute significantly to vehicle traffic²⁷. Additionally, residential complexes are located north of the Banwol and Sihwa industrial complexes, while a green space is toward the southwest.

Wonju

Wonju is located in Gangwon Province, east of Seoul, and had an approximate population of 360,000 as of March 2023²⁹. With a population density of 415 persons/km², it is a small city, approximately 1/40 th the size of Seoul. Wonju has industrial complexes that developed between the 1970 s and early 2000 s in Munmak, Donghwa, and Usan-dong, along with numerous old business sites. As of 2023, operational businesses in Wonju included small-scale manufacturers, such as metal processing companies and medical device manufacturers³⁰. Furthermore, as public transportation is not as developed as in the metropolitan area, the proportion of small private vehicles is high (68.5%). The impact of mobile pollution sources on air quality is estimated to be significant due to a highway adjacent to the measurement points³¹. Wonju has witnessed continuous large-scale construction projects for regional development. Geographically, it is characterized by a basin terrain surrounded by mountains, leading to atmospheric stagnation and a high frequency of windless conditions that further contribute to elevated PM_2.5³². Considering these diverse pollution sources and topographical features, which may affect the concentrations of atmospheric PM_2.5, we selected five measurement points near industrial complexes and the highway to characterize PM_2.5 pollution in Wonju.

Sampling method

We collected PM_2.5 samples twice a week (Mondays and Wednesdays) during the winter (December–February) and spring (March–May) seasons, when high concentrations of PM_2.5are typically observed in the Republic of Korea³³, and once a week (Mondays) during the summer (June–August) and fall (September–November) seasons. Each sampling session, starting at 8.00 am, lasted 24 h. A PM_2.5 low-volume air sampler (PMS 204, APM Co., Ltd., Korea, Bucheon), design based on the US EPA’s FRM PM_2.5, was utilized to collect samples at a flow rate of 16.7 L/min (Table S2). For sample collection, a Teflon filter (PTFE 2.0 μm, Ø47 mm) was used to obtain the concentrations of ionic species and trace elements, while quartz fiber filters (Pured Quarts fiber, Ø47 mm) were used for carbon analysis. During sampling, a total of 5,189 Teflon filters were collected: 1,878 in Seoul, 1,878 in Ansan–Siheung, and 1,433 in Wonju. Similarly, 5,226 quartz fiber filters were collected: 1,867 in Seoul, 1,888 in Ansan–Siheung, and 1,471 in Wonju. The filter papers with PM_2.5 samples were transported to the laboratory by placing them in a refrigerant-containing cooler to minimize losses, maintaining a temperature below 4℃ for analysis. Additionally, we used a buffered transport device composed of materials that prevent static electricity and shaking to reduce the losses of PM_2.5, collected on filter papers from shaking or impact during transport.

The collected PM_2.5 samples, were analyzed for eight ions (Cl^-, NO₃^-, SO₄^2-, Na⁺, NH₄⁺, K⁺, Mg²⁺, and Ca²⁺), two carbons (Organic carbon and Elemental carbon), and 19 trace elements (Al, Ti, V, Mn, Fe, Ni, Co, Cu, Zn, As, Sr, Mo, Cd, Ba, Pb, P, S, Cr, and Si). The analytical methods and quality control of the components in PM_2.5 are presented in the Supplementary Material(Table S3-S6).

Source distribution and contribution assessment

PMF is a statistical technique used to analyze the main sources of atmospheric PM_2.5 and estimate the contribution of each pollution source to the overall PM_2.5³⁴. This method has been employed in several studies worldwide. PMF is applied to apportion source configuration profiles and their associated PM_2.5, based on observed values and uncertainties without any prior information. In this study, we utilized the US EPA’s PMF version 5.0³⁵ for all PMF analyses.

Before conducting PMF analysis, we identified the main sources using principal component analysis (PCA), a technique widely used to reduce data dimensionality and extract latent factors to determine the relationship between measured variables³⁶. We applied varimax rotation in the PCA, and the number of factors was set to cumulative variance > 70% and eigenvalue > 1.0^37,38. Before performing the PCA, we used the Z-score to preprocess and normalize the data. The number of factors calculated through PCA (± 2) was attributed to the number of sources in the PMF analysis³⁹.

Generally, PMF analysis utilizes NO₃^-, SO₄^2-, and NH₄⁺ to assess the contributions from secondary inorganic aerosol. However, secondary products in the atmosphere also include inorganic substances; except for the secondary organic carbon (SOC), which is formed through complex reactions between volatile organic compounds (VOCs) in the atmosphere and airborne radicals or O₃⁴⁰. A complex computational process utilizing VOC observational data is required to estimate the contribution of PM_2.5to SOC formation^41,42,43. The EC tracer methodology was applied to estimate SOC using EC and OC concentrations, facilitating the additional calculation of primary organic carbon (POC) and SOC. POC and SOC were added as variables to the PMF analysis to assess their combined contributions to SOC, and are calculated using Eqs. (1) and (2), respectively⁴⁴.

$$\:\text{P}\text{O}\text{C}\:=(\frac{\text{O}\text{C}}{\text{E}\text{C}}{)}_{\text{p}\text{r}\text{i}}\times\:\text{E}\text{C}\:+\text{O}\text{C}\text{n}\text{c}$$

(1)

$$\:\text{S}\text{O}\text{C}\:=\:\text{O}\text{C}-\text{P}\text{O}\text{C}$$

(2)

where EC, which is dominated by primary emissions, is considered a tracer of POC; $\:{\text{O}\text{C}}_{\text{n}\text{c}}$ refers to very small or negligible biological OC concentrations. In urban environments, $\:{\text{O}\text{C}}_{\text{n}\text{c}}$is typically assumed to be zero, as the contribution of biogenic OC is minimal compared to combustion-related emissions⁴⁵. Therefore, $\:{\text{O}\text{C}}_{\text{n}\text{c}}$ is set to zero to minimize uncertainty in calculations. $\:(\frac{\text{O}\text{C}}{\text{E}\text{C}}{)}_{\text{p}\text{r}\text{i}}$is the lowest value of OC/EC calculated for each season through Deming regression. Unlike ordinary least squares regression analysis, which minimizes the perpendicular distance to a straight line of the cost function and considers only the error of the Y value in the analysis, Deming regression estimates the relationship between two variables by considering the errors of both X and Y⁴⁶. $\:(\frac{\text{O}\text{C}}{\text{E}\text{C}}{)}_{\text{p}\text{r}\text{i}}$ was determined using Deming regression (X-EC, Y-OC) by selecting the bottom 15% of OC/EC values for each season and pandemic period based on previous studies (Table 2). To ensure the robustness of the selected threshold, we performed a sensitivity analysis by comparing the results using alternative thresholds of 10–20% (Table S7). The analysis indicated that the 15% threshold provided the most consistent and stable estimates of $\:(\frac{\text{O}\text{C}}{\text{E}\text{C}}{)}_{\text{p}\text{r}\text{i}}$, especially considering seasonal variability and emission characteristics. Therefore, we concluded that the 15% threshold was the most appropriate choice for our dataset.

Table 2 Deming regression results for $\:(\frac{\text{O}\text{C}}{\text{E}\text{C}}{)}_{\text{p}\text{r}\text{i}}$ in different seasons.

Input data description in PMF

For PMF analysis, species with a signal/noise ratio (S/N ratio) of < 0.2 were classified as “Bad” and excluded from the analysis. Additionally, species with an S/N ratio between 0.2 and 2.0 were categorized as “Weak”, which increased analysis uncertainty. Species with an S/N ratio of ≥ 2.0 were classified as “Strong”, indicating reliable data quality. Additionally, species that did not satisfy source distribution were reclassified into their appropriate categories during analysis. Input data for the PMF analysis were utilized based on the Polissar et al.⁴⁷protocol and PMF 5.0 Guideline of the US EPA³⁵. First, data below the method detection limit (MDL) were set to MDL/2 and uncertainty (Unc) was calculated using Eq. (3), whereas for concentration above the MDL, Unc was calculated using Eq. (4)⁴⁸.

$$\:\text{U}\text{n}\text{c}\:=\:\frac{5}{6}\times\:\text{M}\text{D}\text{L}$$

(3)

$$\:\text{U}\text{n}\text{c}\:=\:\sqrt{(\text{E}\text{F}\times\:\text{C}{)}^{2}\times\:(0.5\times\:\text{M}\text{D}\text{L}{)}^{2}}$$

(4)

where $\:\text{E}\text{F}\:$ refers to the error fraction; and $\:\text{C}$is the concentration of the individual chemical species. EF was set to 10% based on previous studies^49,50,51,52and extra modeling uncertainty in the analysis was set to 20%⁵³. The final PMF analysis results, calculated using these inputs, were used to determine pollution source distributions, primarily based on the US EPA’s profile database (USA-SPACIATE V5.3)⁵⁴ and relevant study.

Statistics analysis

All statistical analyses were conducted using the SPSS software ver. 23. The differences in concentrations of PM_2.5 and its chemical components between the pandemic and post-pandemic periods were determined at a significance level of p < 0.05, using the Wilcoxon signed rank test.

Results and discussion

Distribution of PM_2.5 and its chemical components

PM_2.5 concentration analyses in the three selected regions (Table 3), revealed that Seoul and Ansan–Siheung had higher concentrations post-pandemic than during the pandemic. In contrast, Wonju showed higher concentrations during the pandemic than during the post-pandemic period. However, none of these differences were significant (p> 0.05). Globally, according to Rodriguez-Urrego⁵⁵, most cities with usually high levels of PM_2.5 reported a decrease of approximately 12% during the lockdown. However, this study found no significant difference in PM_2.5 levels in the pandemic and post-pandemic period.

Table 3 PM_2.5 concentrations during the pandemic and post-pandemic periods (µg/m³).

Notably, we documented variation in the distribution and composition ratios of ions, carbons, and trace elements in PM_2.5 during the pandemic and post-pandemic periods (Fig. 2a). First, total ion concentrations were lower during the post-pandemic period by 12.0%, 7.2%, and 19.1% in Seoul (10.33 vs. 11.65 µg/m³), Ansan–Siheung (13.08 vs. 13.93 µg/m³), and Wonju (8.77 vs. 10.26 µg/m³), respectively. In Seoul and Ansan–Siheung, total carbon (OC + EC) concentrations were higher during the post-pandemic period by approximately 6.4% and 4.7%, respectively (Seoul: 5.91 vs. 5.62 µg/m³; Ansan–Siheung: 7.45 vs. 7.13 µg/m³). However, in Wonju, total carbon concentration decreased by 15.6% during the post-pandemic period (6.31 µg/m³) compared to that of the pandemic period (7.21 µg/m³). Conversely, trace elements increased by approximately 28.7% in Seoul and Ansan–Siheung and 11.9% in Wonju during the post-pandemic period (Seoul: 2.31 vs. 1.79 µg/m³; Ansan–Siheung: 3.02 vs. 2.31 µg/m³; Wonju-1.89 vs. 1.66 µg/m³).

Fig. 2

Concentrations of ions, carbons, and elements during lockdown and recovery phase at different sampling sites: (a) Concentrations of ions, carbons, and elements (µg/m³) and (b) percentage of ions, carbons, and elements (%).

The composition ratios of each chemical component in PM_2.5 (Fig. 2b) indicated that ions accounted for the largest proportion (50–60%) of PM_2.5. They decreased by approximately 3–5% in all the three regions during the post-pandemic period compared with that of the pandemic period. Notably, this decrease was most pronounced in Seoul, where the ionic component decreased by approximately 5.44%. Unlike the ionic components, the composition ratio of carbon increased in all three regions, but was most significant in Seoul, where the composition ratio increased from approximately 29.50% during the pandemic to 31.86% post-pandemic. Finally, the composition ratio of trace elements increased by approximately 2–3% in all three regions, and the difference (3.1%) between the composition ratios during the pandemic (9.37%) and post-pandemic (12.46%) periods was the largest in Seoul.

In other words, all three regions demonstrated differences among the concentration and composition ratios of chemical components in PM_2.5 during both periods, suggesting that the air quality was regionally impacted by the pandemic. Among the three regions, Seoul showed the most significant differences in both concentrations and composition ratios, indicating that such large cities were markedly affected by lockdown restrictions. This may be due to the halting of external activities. Identifying the causes of these differences could provide a strong premise for establishing policies to effectively reduce PM_2.5 sources in the future.

The distribution patterns of individual components of PM_2.5 during social restrictions revealed notable trends (Table S8). Among the ionic components, the concentrations of NO₃^-, SO₄^2-, and NH₄⁺ (components of the SNA components [NH₄)₂SO₄, NH₄NO₃, and NH₄HSO₄]) were lower post-pandemic in all three regions. Previous studies have shown that meteorological conditions, such as humidity and temperature, play a significant role in the formation of secondary inorganic aerosols (SIA)⁵⁶. This decrease may be attributed not only to the continuous reduction in domestic PM_2.5but also to changes in meteorological conditions (such as temperature and humidity) and reductions in precursor emissions (NOₓ and SO₂) during the post-pandemic period. Previous studies have shown that decreased NOₓ and SO₂ emissions, coupled with changes in atmospheric conditions, can significantly reduce the formation of SIA, including SNA⁵⁷. Therefore, the observed decrease in SNA concentrations after the pandemic could be the result of the combined effects of reduced precursor emissions and altered atmospheric conditions. It is expected that a more comprehensive understanding of SNA variations can be achieved through further studies that simultaneously consider various meteorological factors, emission sources, and atmospheric pollutants.

However, the concentrations of Cl^-, Na⁺, Mg²⁺, and Ca²⁺were higher during the post-pandemic period in all three regions. Na⁺, Mg²⁺, and Ca²⁺ are elements of natural origin or dust from construction activities^58,59. Since the restrictions halted the construction and vehicular movements, road dust settled and resuspension frequency decreased, reducing these ionic species. However, resumption of activities post-pandemic raised these ions concentration. These findings indicate that the sources affecting the individual ionic components functioned differently during the pandemic and post-pandemic periods, necessitating further investigation through source analysis.

Among the carbon components, the concentration of EC was lower in the post-pandemic (0.47 ± 0.25 µg/m³) than in the pandemic (0.65 ± 0.32 µg/m³) periods in all three regions, with Wonju exhibiting the largest drop (27.7%). In contrast, OC concentrations were higher in Seoul (6.0%) and Ansan–Siheung (4.6%) in the post-pandemic (Seoul: 5.31 ± 2.60 µg/m³; Ansan–Siheung: 6.78 ± 3.40 µg/m³) than in the pandemic (Seoul: 4.99 ± 2.12 µg/m³; Ansan–Siheung: 6.47 ± 3.13/m³) periods. However, in Wonju, OC levels decreased by approximately 12.2% during the post-pandemic period (5.76 ± 2.66 µg/m³) compared to the pandemic period (6.56 ± 3.21 µg/m³).

Fu and Chen⁶⁰reported that EC is produced from the incomplete combustion of fossil fuels or industrial activities (a primary source). In contrast, OC is emitted directly from the primary sources and secondary atmospheric reactions (gas-to-particle conversion)⁶¹. Owing to these characteristics, the OC/EC ratio has been used in various studies^62,63, where a higher OC/EC indicates the elevated influence of secondary reactions compared to that of the primary emission sources⁶⁴.

In this study, the OC/EC ratio (Fig. 3) increased during the post-pandemic period in all three regions, with Wonju showing the most pronounced increase—about 1.5 times higher than during the pandemic. This trend implies that although the overall carbon concentration decreased in Wonju, the relative contribution of OC compared to EC increased significantly. The decrease in EC, a marker of primary combustion emissions, is likely attributable not to a net decrease in total primary activity or a substantial increase in SOC, but rather to changes in emission characteristics (e.g., shifts in vehicle fuel types or reduced combustion intensity). The increased OC/EC ratio, therefore, likely reflects both a compositional shift within primary emissions and enhanced secondary organic aerosol formation.

Fig. 3

Concentrations of organic carbon (OC) and elemental carbon (EC) (µg/m³): During the pandemic (a) in Seoul, (b) in Ansan–Siheung, and (c) in Wonju, and post-pandemic (d) in Seoul, (e) in Ansan–Siheung, and (f) in Wonju.

Regarding trace elements, the concentration of S, which accounted for the largest mass fraction, increased significantly during the post-pandemic period in all regions. Most other trace elements exhibited higher concentrations in the post-pandemic period than in the pandemic period. However, the concentrations of Cu, Zn, P, Mn, and Ti in Seoul, Cu and P in Ansan–Siheung, and V, Fe, Cu, Zn, Pb, and Cr in Wonju were high during the pandemic period. These variations suggest that differences in the operation of sources in each region may have influenced the composition ratio of PM_2.5. According to Wang et al.¹⁵, most trace elements decreased during the lockdown period, where Cr, Mn, Fe, and Zn decreased by approximately 37–60%, due to the overall decrease in the operational capacity of manufacturing and processing plants at the corresponding target sites. These changes (e.g., production and number of employees) were mostly affected by the measures taken during the pandemic, influencing the composition ratio of chemical species in the atmosphere.

Source distribution and contributions

PCA

Pre-PMF analysis, the PCA was conducted to identify the approximate number of sources (Table 4). All three regions were estimated to have eight main sources regardless of their pandemic status. Examining the PC1, the components of PC1 in Seoul remained the same, whereas those in Ansan–Siheung and Wonju differed, indicating that the sources of the main components may be different. The PCA results are provided in the Supplementary Material.

Table 4 Summary of principal component analysis results by area and period (Compounds of PC1 is major contributing compounds (Loading ≥0.5)).

PMF analysis

Information on the input data and a summary of the source distribution are depicted in the Supplementary Material. First, the source distribution results for Seoul (Fig. 4, Table S9-S12) revealed six major sources during the pandemic (R² = 0.91) and seven sources in the post-pandemic period (R² = 0.92). The contribution of these sources related to secondary formations (e.g., SOC formation, secondary nitrate, and secondary sulfate) accounted for approximately 72.5% of total PM_2.5, but decreased to 51.3% during the post-pandemic period.

Fig. 4

Contribution of each source obtained through positive matrix factorization (a) during pandemic and (b) post-pandemic in Seoul.

For sources related to primary emissions and natural sources (e.g., yellow dust, Asian dust, and suspended dust), combustion (2.2%), soil dust (12.5%), and biomass burning (10.8%) accounted for 27.5% of total PM_2.5. In contrast, during the post-pandemic period, aged sea salt (3.2%), complex emissions from combustion (11.2%), road dust (10.8%), Asian dust (17.0%), and industrial combustion (6.5%) were the primary sources, accounting for 48.7% of total PM_2.5. This indicates that the contribution of primary emissions to PM_2.5 concentrations increased during the post-pandemic period and number of sources became diverse. This suggests the dominance of primary emissions in the post-pandemic period. Conversely, the relative contribution of secondary pollutants increased during the pandemic period owing to a decrease in primary emissions, whereas it decreased during the post-pandemic period due to an increase in primary emissions as economic and social activities resumed. However, considering the higher OC/EC ratios in the post-pandemic period, the quantitative levels of secondary pollutants may have increased during the post-pandemic period. This finding is attributable to the increase in primary emissions (especially transportation activities) due to the resumption of social activities and concentration of precursors contributing to secondary formations and secondary pollutants. However, the highest emission rate from primary sources suppressed the share of secondary formations. The contribution of secondary nitrate, the largest contributor in both the pandemic and post-pandemic periods, was 39.5% in the pandemic period but decreased to approximately 10–29.5% in the post-pandemic period. PMF results confirmed that the concentration of secondary nitrate in PM_2.5 was lower during the post-pandemic period (6.91 µg/m³) than during the pandemic period (8.9 µg/m³). Briefly, the absolute emission and relative contribution of secondary nitrate in PM_2.5 decreased simultaneously, suggesting that the production of secondary nitrate from NO_x during the pandemic was high despite the resumption of socioeconomic activities post-pandemic. However, this result contradicts the findings of Kim et al.⁶⁵, who reported that public and private transportation traffic in Seoul decreased by approximately 47.0% compared with that pre-pandemic levels. Additionally, Seoul Metropolitan Government traffic data indicates that traffic volume increased in 2023 (approximately 10,265,000 vehicles/d) compared to the period of 2020–2022 (approximately 10,070,000 vehicles/d)^66,67,68. According to the Ministry of Land, Infrastructure, and Transport, the number of registered electric vehicles (EVs) demonstrated a 4.5-fold increase in 2024 (606,610) in comparison to its level in 2020 (139,962) owing to expanded policy support for EVs. This change in vehicle fuel type aligning with government policies during the pandemic contributed to improved air quality, with a reduction in secondary nitrate production.

Source distribution analysis in the pandemic and post-pandemic periods in Ansan–Siheung, categorized these sources into seven groups in both periods (R² = 0.94, R² = 0.93, respectively; Fig. 5a, b, Table S13-S16). These sources contribution to secondary formations (SOC formation, secondary nitrate, and secondary sulfate) were confirmed. During the pandemic, sources related to secondary formations accounted for approximately 61.5% of total PM_2.5, whereas it only reached about 51.9% during the post-pandemic period. An analysis of sources related to primary emissions and natural sources (e.g., emissions from industrial activities, aged sea salt, yellow dust, and biomass burning) revealed key differences between these periods. During the pandemic, the main sources were industrial process emissions (8.3%), combustion (8.4%), primary emissions (9.8%), and aged sea salt (12.0%), accounting for 38.5% of total PM_2.5. However, during the post-pandemic period, the main sources were biomass burning (10.6%), Asian dust (8.1%), primary emission (7.2%), combustion (9.2%), as well as industrial vehicle and combustion (13.1%), accounting for 48.1% of total PM_2.5. These findings indicate the increase in the types of primary sources (suspended dust) post-pandemic. In contrast, the contribution rate of secondary nitrate, which was the largest contributor in both periods, was 35.1% in the pandemic period and 27.5% in the post-pandemic period, with a decrease of approximately 7.6%. According to the data from the Gyeonggi-do Transportation Information Center⁶⁵, the annual traffic volume in Ansan–Siheung decreased post-pandemic. The average traffic volume per lane (e.g., expressways, general roads, state-funded local highways, and local highways) was approximately 85,188 vehicles/d in 2020–2022, further declining to 75,945 vehicles/d in 2023. These findings contradict the increase in traffic volume in Seoul. However, considering the lower traffic volume in Ansan–Siheung post-pandemic, it is assumed that the lower NO_x emissions from reduced road traffic contributed to the decreased secondary nitrate levels. Thus, the lower contribution of secondary nitrate in Ansan–Siheung in the post-pandemic period is likely due to a decrease in traffic volume and changes in vehicle fuel types. These changes are important considerations for future air quality improvement policies.

Fig. 5

Contribution of each source obtained through positive matrix factorization (a) during pandemic and (b) post-pandemic in Ansan–Siheung.

Source distribution analysis for both periods in Wonju (Table S17-S20), categorized them into eight source groups (R² = 0.91, R² = 0.91; Fig. 6a, b). Source contribution to the secondary formations (SOC formation, secondary nitrate, and secondary sulfate) accounted for approximately 66.1% of total PM_2.5 during the pandemic, whereas they accounted for approximately 45.3% post-pandemic, marking a decrease of 20.8%. An analysis of sources related to primary or natural emissions (e.g., emissions from industrial activities, aged sea salt, yellow dust, and biomass burning) revealed that the main sources during the pandemic were resuspended dust (6.2%), combustion I (4.8%), coal combustion (5.3%), combustion II (5.8%), industrial processes (9.0%), and Asian dust (2.6%), accounting for 33.9% of total PM_2.5. In contrast, post-pandemic, combustion (1.77%), aged sea salt (2.1%), industrial process (5.6%), resuspended dust (12.1%), Asian dust (6.1%), and primary emission (21.3%) were the major sources, accounting for 54.7% of total PM_2.5. Such findings align with the results from Seoul and Ansan–Siheung: during the post-pandemic, the contributions of primary emission or suspended dust to PM_2.5 concentrations increased compared to the pandemic period, and the types of primary sources became diverse, confirming the increases in primary emission or suspended dust during the post-pandemic period. However, unlike Seoul and Ansan–Siheung, Wonju exhibited relatively higher contributions from sources related to primary emissions or suspended dust than those related to secondary emissions post-pandemic. This suggests that in Wonju sources related to primary emission and suspended dust predominated during the post-pandemic.

Fig. 6

Contribution of each source obtained through positive matrix factorization (a) during pandemic and (b) post-pandemic in Wonju.

However, the contribution rate of secondary nitrate, identified as the largest contributor during the pandemic, was 40.1% during the pandemic period but decreased to 21.3% during the post-pandemic period, representing a decrease of approximately 18.8%. According to the Wonju Statistical Year Book⁵⁸, the number of vehicles registered in Wonju has gradually been increasing with 179,251 in 2020, 185,530 in 2021, and 192,437 in 2022. These results are similar to those of Seoul, suggesting that changes in vehicle fuel types coinciding with the pandemic contributed to this similarity. In addition, the expansion of policy support for EVs and corresponding increase in EV registrations during the same period may have contributed to the reduction in airborne NO_x emissions. However, the number of registered vehicles alone may not sufficiently represent the overall increase in primary emissions observed post-pandemic in Wonju. While vehicle numbers indeed increased, changes in vehicle fuel types, particularly the increased registration of electric vehicles (EVs), likely mitigated the emission impacts typically associated with higher vehicle numbers. Additionally, other primary emission sources such as industrial operations and residential heating could also have significantly contributed to the observed rise in primary emissions. Therefore, a more comprehensive analysis encompassing various emission sources is recommended to fully understand the factors driving changes in primary emissions in Wonju.

The results of the PMF analysis (Fig. 7) indicated that, compared to the concentration of secondary formations, the concentrations of primary emissions and suspended dust increased during the post-pandemic period compared with that of the pandemic period. Among the three regions, Wonju showed the greatest decrease (31.6%) in sources related to secondary formations, from 15.49 µg/m³ in the pandemic period to 10.60 µg/m³ in the post-pandemic period. In contrast, Seoul experienced a notable increase (49.8%) in primary emissions and suspended dust from 5.73 µg/m³ in the pandemic period to 11.42 µg/m³ in the post-pandemic period.

Fig. 7

Concentrations of secondary formation, primary emission and suspended dust using positive matrix factorization analysis (µg/m³).

These findings suggest that Wonju experienced the greatest decrease in the sources related to secondary formations among the three regions post-pandemic, suggesting that factors such as decrease in precursor emissions (e.g., NO_x and VOCs), changes in meteorological conditions, and delays in the recovery of economic activities during the post-pandemic period may have contributed to regional differences. Thus, Wonju needs a strategic approach to maintain this declining trend of secondary products and restrain the increase of primary emissions. In particular, the efficient management of PM_2.5 concentrations entails the reduction of industrial and heating emissions, and strengthening the air quality management plan. Contrarily, in large cities such as Seoul where the main source of pollution is transportation, the recovery of socioeconomic activities post-pandemic, directly led to an increase in primary sources. In Seoul, future efforts to control the reduction of primary sources (e.g., combustion activities) and dust (e.g., roads and yellow dust) may decrease PM_2.5 concentrations.

This study provides important insights into the impact of unprecedented social changes on air quality and found that the resumption of socioeconomic activities during the post-pandemic period increased primary emissions and suspended dust in Seoul and Ansan–Siheung. Evaluation of source characteristics implied that regionally tailored policies are required in the formulation of post-pandemic air quality management strategies and emphasized the need to comprehensively consider changes in sources corresponding with changes in socioeconomic activities. This study demonstrated that transportation and industrial emission management in Seoul and Ansan–Siheung, as well as heating and industrial emission reduction in Wonju can become important strategies to improve the air quality. Therefore, the findings of this study may help future researchers to advance their study in the direction of developing pollution reduction policies and improving urban air quality, particularly when circumstances have long-term impacts, such as pandemics.

Conclusion

This study aimed to generate fundamental data on air quality status during the pandemic to aid formulation of air quality improvement policy in future, including qualitative and quantitative analysis of PM_2.5 and its chemical components, particularly its source-specific effects based on regional differences in the lockdown period. This study highlighted significant differences in PM_2.5 components and the composition ratios during both pandemic and post-pandemic periods in the three regions. The total ion concentrations decreased in all three sites, whereas the trace element concentrations increased in the post-pandemic period. Carbon concentrations increased in Seoul and Ansan–Siheung but decreased in Wonju. As for individual components, concentrations of the SNA series (NO₃⁻, SO₄²⁻, and NH₄⁺) were lower in the post-pandemic period, whereas Cl⁻, Na⁺, Mg²⁺, and Ca²⁺ increased in the same period, indicating a remarkable dust resuspension from natural sources and construction activities. All cities experienced a decrease in EC, but OC slightly increased in Seoul and Ansan–Siheung but decreased in Wonju, leading to an increase in the OC/EC ratio, particularly in Wonju. Regarding trace elements, most elements, primarily S, depicted an increase in the post-pandemic period. Trace elements with relatively high concentrations during the pandemic differed by region, likely reflecting changes in emission sources during these periods. These variations were likely attributed to changes in emission sources with primary emissions and dust suspension contributing mostly in the post-pandemic period, while secondary emissions characterized the PM_2.5 levels in the pandemic period. The regional variations in PM_2.5 were marked by the extent of socioeconomic activities in the post-pandemic period with Seoul and Ansan–Siheung showing greater contributions from primary emissions and dust resuspension. Especially in Seoul, the concentrations of pollutants increased highlighting the importance of managing traffic emissions and road dust. However, Wonju witnessed significant reduction in PM_2.5 sources related to secondary formations in the post-pandemic period indicating reduction of precursor emissions (NO_x and VOCs), changes in meteorological conditions, and delayed recovery of the economic activities within the region.

This study was successful in scientifically evaluating the impact of the forced suspension of anthropogenic activities due to the pandemic on PM_2.5-related air quality. These results are expected to serve as a basis for framing future air pollution reduction policies and encouraging urban environmental management practices. Furthermore, this study confirmed that strong sociopolitical actions can impose strong constraints on anthropogenic activities and alter the pattern of distribution of pollutants regionally. Thus, the findings necessitate regionally tailored responses when establishing future air quality management policies.

Therefore, to improve air quality management based on the distinct characteristics observed in each region, we propose the following region-specific policy directions: In Seoul, where traffic-related emissions and road dust were dominant in the post-pandemic period, implementation of low-emission zones, strengthened vehicle emission regulations, and road dust suppression programs would be effective. In Ansan–Siheung, with high industrial activity and increased elemental pollution, industrial emission controls, transition to cleaner fuels, and real-time emission monitoring systems should be prioritized. In contrast, Wonju, which showed a relative delay in economic recovery and reduction in secondary formation sources, may benefit from residential heating emission regulation and energy efficiency incentives to address seasonal combustion-related sources. These tailored policy suggestions are expected to support more effective and sustainable air quality improvement in each region.