Unraveling Spatiotemporal Synergistic Features of PM2.5–O₃ and Systematic Management Policy Based on Multiple Scenario-Driven Factor Analysis in the Changsha–Zhuzhou–Xiangtan Urban Agglomeration, Central China

Abstract

Fine particulate matter (PM2.5) and ozone (O₃) are the key factors restricting the continuous improvement of air quality in the Changsha–Zhuzhou–Xiangtan urban agglomeration (CZT). Due to the potential correlation between variations in urban PM2.5–O₃ concentration, the analysis of its composite pollution characteristics is helpful in formulating accurate and thorough air control policies. Based on the long-term concentration change in PM2.5 and O₃, this study analyzed the features and synergistic factors of PM2.5–O₃ pollution in the CZT by using spatial autocorrelation and a linear driving model of PM2.5–O₃. The results showed that from 2017 to 2023, under the current Chinese atmospheric environment standard, the CZT saw four combined pollution days. However, if the daily limit values were viewed in line with Grade II of the WHO transition period (O₃: 120 μg/m³, PM2.5: 50 μg/m³), the combined pollution days would reach 111. The concentration of O₃ in Zhuzhou and Xiangtan was about 10 μg/m³ lower than that in Changsha. Lower SO₂ levels in Changsha might influence the partitioning of OH radicals and reactive nitrogen species, potentially affecting local O₃ formation efficiency. NO₂ and meteorological conditions jointly influence the co-variation in PM2.5 and O₃, with NO₂ playing a more significant role in PM2.5 formation. The long-term time series and daily concentrations of PM2.5 and O₃ in the CZT showed opposing values, but there were short-term synergistic events on the scale of daily concentrations, and the time period was typically 3–10 days. Low humidity and strong sunlight may cause antagonistic events in which the concentration of O₃ rises rapidly. Under static and stable weather conditions with low wind speed, no rainfall and moderate humidity, the concentration of PM2.5 and O₃ rose alternately on sunny and cloudy days, demonstrating synergistic growth.

1. Introduction

Now, the fine particulate matter (PM2.5) and ozone (O₃), which have negative impacts on urban human health and environment, are the two most significant air pollutants not only in China, but also in other countries. PM2.5 refers to particulate matter with an aerodynamic diameter of less than 2.5 μm. Its anthropogenic sources are divided into primary emissions, mainly consisting of dust, black carbon, and heavy metal particles, and secondary sources, in which organic compounds (VOCs), nitrogen oxides (NOx), and sulfur dioxide (SO₂) undergo chemical transformation [1]. Tropospheric O₃, with NOx and VOCs as its main precursors, is generated through photochemical reactions under conditions of high temperature and strong light radiation [2]. The strong oxidizing property of O₃ further accelerates precursor reactions, potentially aggravating O₃ pollution [3]. PM2.5 and O₃ cause adverse effects in the respiratory tract, cardiovascular system, and lungs, thereby increasing the risk of premature mortality [4]. Due to shared precursors and complex interaction pathways, PM2.5 and O₃ pollution are closely interrelated [5]. Following the implementation of policies such as the “Action Plan for the Prevention and Control of Air Pollution” and the “Three-Year Action Plan to Win the Blue Sky Defense Battle” in recent years, the typical air pollutants (SO₂, NO₂, and PM₁₀) have been effectively controlled (Table S1), and further, PM2.5 concentrations have declined but still one third of the cities exceeding the standards; however, O₃ pollution has become increasingly prominent (Table S1). In 2022, O₃ exceeded the national Grade II MDA8 standard (GB 3095–2012) on 47.9% of days nationwide [6], and compound pollution contributed to an excess mortality risk of 0.69%, particularly in major urban agglomerations [5,7]. Consequently, the “Action Plan for the Continuous Improvement of Air Quality”, issued in 2023 under the framework of the 14th Five-Year Plan, emphasizes coordinated control of PM2.5 and O₃ and calls for strengthened joint prevention strategies in key regions, along with improved criteria for activating heavy pollution weather warnings.

To date, there have been many achievements in research on PM2.5 and O₃ compound air pollution, with the main research directions including distribution characteristics, influencing factors and the conditions for the occurrence of compound pollution. PM2.5 pollution in China has mainly been concentrated in winter, while O₃ pollution is mainly concentrated in summer, and the main polluted areas for the two do not completely overlap [8]. PM2.5–O₃ compound pollution zones were concentrated in the Yangtze River mid-lower reaches urban agglomeration, the North China Plain, and the Chengdu–Chongqing urban agglomeration, reaching a peak between 2017 and 2019. PM2.5 or O₃-dominant pollution zones often existed around these compound pollution zones [9,10]. There are still no consistent conclusions regarding the occurrence conditions and influencing factors of compound pollution in different regions. For example, regarding the concentration correlation between PM2.5 and O₃, high levels of O₃ in the Beijing–Tianjin–Hebei region might lead to an increase in the proportion of secondary PM2.5 [11], but in the Xi’an area, high concentrations of atmospheric particulate matter effectively reduced the photolysis rate, resulting in a decrease in O₃ concentration [12]. In the source apportionment of compound pollution, PM2.5 sources were relatively clear, and it was easy to analyze the causes of pollution. In contrast, O₃ was mostly secondary-generated and its emission sources were less obvious than those of PM2.5 [13]. NO_x, SO₂, and VOCs were the main precursors affecting compound pollution. Among them, O₃ concentration changes were significantly affected by meteorological and transmission disturbances, which reduced the influence of precursors [14].

Meteorological conditions are a crucial factor influencing PM2.5–O₃ compound pollution. Top-down models have often been used to identify the meteorological drivers of compound pollution. In most cases, the meteorological conditions driving PM2.5 and O₃ concentrations differed significantly [15,16]. O₃ concentration was positively correlated with temperature and solar radiation, and surface temperature or solar radiation were direct meteorological factors promoting O₃ generation [17,18]. Many meteorological factors have been found to be related to PM2.5, such as temperature, relative humidity, wind speed, and precipitation. This might be because the influence of meteorological conditions on PM2.5 is mainly through diffusion, transmission and deposition [19]. The driving force behind compound pollution might require various meteorological factors. Dai et al. [20] examined meteorological drivers of O₃ and PM2.5 compound pollution in the Yangtze River Delta region from 2013 to 2019, and reported a positive correlation between the two pollutants from April to October, with compound episodes largely influenced by wind speed, surface temperature, and relative humidity. In addition, the sensitivity of atmospheric pollutants to meteorological factors varied in different seasons and regions, and the occurrence conditions of compound pollution might have specificity [21,22]. In summary, the principles of concentration changes in PM2.5 and O₃ gradually became clear, but due to the complexity of influencing factors, the characteristics of the concentration linkage changes in the two, and especially the occurrence conditions of compound pollution, urgently need specific analysis based on local conditions.

The Yangtze River Economic Belt is a significant national strategic development area in China, with over 40% of the country’s population and GDP. Within it, the CZT is located in the middle reaches of the Yangtze River and consists of three cities: Changsha, Zhuzhou, and Xiangtan. It is the region with the highest intensity of economic activity in Hunan Province [23]. As a pioneer in the construction of national eco-civilization urban agglomerations among the six provinces of central China and a key force in the region’s rise, the CZT is of strategic importance. Based on statistics of typical air pollutants over the past nine years (Table S2), the research on PM2.5–O₃ in this region is crucial for coordinated control of compound pollution. It also provides a basis for regional joint prevention in the CZT and the urban agglomerations in the middle reaches of the Yangtze River. This study selected the CZT as the research area and, based on pollution and meteorological data from 2017 to 2023, used methods such as spatial autocorrelation and driver analysis to analyze the pollution characteristics and the factors contributing to the synergy of PM2.5 and O₃ in the CZT, with the aim of providing scientific and effective support for the formulation of coordinated governance strategies.

2. Materials and Methods

2.1. Regional Overview and Data Sources

The Changsha–Zhuzhou–Xiangtan urban agglomeration (CZT) is the core area of economic development and urbanization in Hunan Province, consisting of the three cities of Changsha, Zhuzhou and Xiangtan (Figure 1). The urban agglomerations are distributed in a “T” shape, with distinct seasonality and abundant rainfall. The area has a typical subtropical monsoon climate. The main built-up areas of the city are located on both sides of the Xiangjiang River in the northern part of the urban agglomeration. The built-up areas are mostly surrounded by farmland, while the eastern and southern parts are mainly mountainous [24]. The CZT has a resident population of over 17 million and an annual GDP exceeding 2 trillion CNY. The industrial structure in Changsha is relatively diversified, while Xiangtan and Zhuzhou function as sub-regional centers with a stronger emphasis on industrial production, which has led to differences in emission sources [25]. Research data on air pollutants for the national ecological environment (https://www.mee.gov.cn/hjzl/dqhj/ accessed on 10 July 2025) is issued by the Changsha–Zhuzhou–Xiangtan air quality monitoring stations. The data were measured hourly, and the time range is from 2017 to 2023. Except for Shaoshan Station, which is a natural environment control point, the rest of the monitoring stations are all located in built-up areas of the city and its surroundings. The heights of the air monitoring stations are consistent with the average human breathing height. Meteorological data were derived from the ground-station meteorological datasets published by NASA (https://www.nasa.gov/ accessed on 10 July 2025) and the China Meteorological Administration (https://data.cma.cn/ accessed on 10 July 2025). The meteorological datasets were measured on a daily scale. Selected pollutant data included PM2.5, O₃, SO₂ and NO₂; meteorological factors included air pressure (pre, kPa), sunshine duration (The monitoring data were only available up to 2020) (sunt, h), dew point (dewp, °C), temperature (airt, °C), 6 h rainfall (rain, mm), wind direction (windd, º), and wind speed (winds, m/s). In addition, the dew point difference (DPD (t-dewp)) was used to characterize the relative humidity. The larger the DPD value is, the drier the air becomes and the lower the relative humidity is.

2.2. Summary of Pollutant Concentration Standards

The concentration standards for PM2.5 and O₃, as well as the discrimination of compound pollution days, are generally called “Ambient Air Quality Standards” (GB 3095–2012) [6]. The daily concentration of PM2.5 is obtained by averaging the hourly PM2.5 data from national monitoring stations. The daily concentration of O₃ is the maximum 8 h moving average concentration of O₃ (MDA8O₃) within one day at the national monitoring station. In urban areas, the daily concentration limit values of O₃ and PM2.5 are 160 μg/m³ and 75 μg/m³, respectively. When the concentration of MDA8 O₃ on a certain day is higher than 160 μg/m³ and the average daily concentration of PM2.5 is higher than 75 μg/m³, it is determined to be a compound pollution day. To evaluate the gap between current air quality conditions and the second-stage control targets, O₃ concentrations over longer time scales (monthly and annual) were calculated as the average of daily maximum 8 h average (MDA8) O₃ values. This study referred to the World Health Organization (WHO) Global Air Quality Guidelines (2021) [26], including the first-stage peak limit for MDA8 O₃ (100 μg/m³) and the second-stage daily guideline values for MDA8 O₃ (120 μg/m³) and PM2.5 (50 μg/m³), which were adopted as pollutant control benchmarks.

2.3. Research Method

2.3.1. Spatial Auto-Correlation Analysis

The spatial agglomeration of PM2.5 and O₃ in the CZT was evaluated using the global and local Moran’s I indices [27,28]. The assessment was conducted using ArcGIS 10.2 (Esri Inc., Redlands, CA, USA). The calculation formula of global Moran’s I is as follows:

$$\mathrm{I}={\displaystyle \frac{\mathrm{n}}{{\mathrm{S}}_0}}\times {\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}})}{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}}$$

(1)

where x_i and x_j represent the pollutant concentrations of grids (points) i and j; n is the number of grids (points); $\overline{\mathrm{x}}$ is the average value of pollutant concentration; and w_ij is a spatial weight matrix. Due to the long transmission distance of PM2.5 and O₃ in the atmosphere, it is constructed using the Queen second-order adjacency rule. S₀ is the sum of the weights of ownership. I ∈ [−1, 1], I > 0 indicates positive correlation, I < 0 indicates negative correlation, and I = 0 indicates no correlation. The larger I is, the stronger the spatial autocorrelation is. The significance of I was tested using the two-sided Z-statistic test, and its expression is:

$$\mathrm{Z}\left(\mathrm{I}\right)=\left[\mathrm{I}-\mathrm{E}\left(\mathrm{I}\right)\right]/\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{I}\right)}$$

(2)

where E (I) and Var (I) are, respectively, the mathematical expectation and variance of I.

The local Moran’s I is used to measure the spatial aggregation degree of each grid (point) and the surrounding grids (points). Its calculation formula is as follows:

$${\mathrm{I}}_{\mathrm{i}}={\mathrm{Z}}_{\mathrm{i}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{Z}}_{\mathrm{j}}$$

(3)

where Z_i and Z_j are the pollutant concentrations on grids (points) i and j, respectively. Finally, the grids (points) are divided into hotspots, cold points and insignificant points according to the significance level and the symbol of Z (I).

2.3.2. PM2.5–O₃ Multiple Linear Driving Model

In order to explore the co-variation law of PM2.5 and O₃, it was first necessary to identify the driving factors that may affect the co-variation. From the perspective of precursors, NO₂ is a common precursor of PM2.5 and O₃, meaning that a certain concentration of NO₂ participating in different reaction types might lead to opposite concentration changes in PM2.5 and O₃. When SO₂ generates PM2.5, it promotes the reaction of NO₂ generating PM2.5, and therefore could also indirectly affect the concentration of O₃. Meteorological conditions can affect the diffusion, deposition and transmission of PM2.5. These meteorological conditions are closely related to the photochemical reactions that generate O₃. Therefore, the driving model selected typical factors that might affect the concentrations of PM2.5 and O₃, such as precipitation, wind speed, light intensity, etc. All the indicators are daily averages. This study employed the Variance Inflation Factor (VIF) to test for multicollinearity among the various indicators, and excluded the variables with a VIF value greater than 10. The pollution precursors in the model were NO₂ and SO₂. The meteorological factors in the model included temperature (airt), dew point difference (dpd), atmospheric pressure (pre), wind direction (windd), wind speed (winds), rainfall (rain) and sunshine duration (sunt). Due to the unavailability of observational VOCs data, the model did not incorporate VOCs as a driving factor.

The driving model adopts multiple linear regression. As the model with the highest frequency of use in existing studies, multiple linear regression could be used to analyze the influencing factors and the degree of influence [29,30]. Model calculations were performed using SPSS Statistics 26 (IBM Corp., Armonk, NY, USA). The model is as follows:

$$\mathrm{P}=\mathsf{\beta }0+{\sum }_{\mathit{\alpha }=1}^{\mathit{k}}{\mathsf{\beta }}_{\mathsf{\alpha }}{\mathrm{I}}_{\mathsf{\alpha }}+\mathsf{\mu }$$

(4)

where $\mathrm{P}$ is the pollutant concentration (PM2.5 or O₃), ${\mathrm{I}}_{\mathsf{\alpha }}$ is the α-th driving factor (including precursors and meteorological variables), ${\mathsf{\beta }}_0$ is the intercept term, ${\mathsf{\beta }}_{\mathsf{\alpha }}$ are the regression coefficients, $\mathrm{k}$ is the number of predictors, and $\mathsf{\mu }$ is the random error term.

To test the effect of the penetration model, the coefficient of determination (R²) and the significance test (t value) were used to determine the degree of model fitting and the significance of the coefficient [31,32]. R², also known as goodness of fit, was used in multiple linear regression to measure the strength of the relationship between the observed values and the predicted values (calculated values) of the dependent variable. The calculation formula is

$${\mathrm{R}}^2=1-{\displaystyle \frac{\sum {({\mathrm{y}}_{\mathrm{i}}-({\mathrm{y}}_{\mathrm{i}})°)}^2}{\sum {({\mathrm{y}}_{\mathrm{i}}-\overline{\mathrm{y}})}^2}}$$

(5)

where y_i is the i-th observed value, $\left({\mathrm{y}}_{\mathrm{i}}\right)°$ is the i-th predicted value, and $\overline{\mathrm{y}}$ is the average. In the multiple linear regression model, the closer R² is to 1, the better the fitting effect is. The t-statistic was used to test the statistical significance of each regression coefficient. The calculation formula is

$${\mathrm{t}}_{{\mathrm{b}}_{\mathrm{j}}}={\displaystyle \frac{{\mathrm{b}}_{\mathrm{j}}}{{\mathrm{s}}_{{\mathrm{b}}_{\mathrm{j}}}}}$$

(6)

where b_j is the estimated regression coefficient of the independent variable x_j, and ${\mathrm{s}}_{{\mathrm{b}}_{\mathrm{j}}}$ is the regression standard error of b_j. The t-statistic tests the null hypothesis that b_j = 0. A larger absolute value of $t$ corresponds to a smaller p-value (sig.), indicating stronger statistical evidence that the coefficient differs from zero. In this study, coefficients with p-values less than 0.05 were considered statistically significant at the 95% confidence level.

3. Results and Discussion

3.1. Analysis of Time Series of PM2.5–O₃ Concentration

The daily distribution of PM2.5 and O₃ concentrations in the cities of Changsha, Zhuzhou and Xiangtan from 2017 to 2023 is shown in Figure 2. Area I represents the compound pollution days under the Grade II standard of the Ambient Air Quality Standards (GB 3095–2012) [6], and Areas II and IV, respectively, represent the PM2.5 or O₃ pollution days under the Grade II standard of GB 3095–2012. Zone III indicates the relatively clean days, when neither PM2.5 nor O₃ exceeded the Grade II standard of GB 3095–2012. The number of relatively clean days was the largest (2114 days); the number of days with PM2.5 pollution (291 days) was higher than days with O₃ pollution (138 days). There were only four days of compound pollution. The severity of PM2.5 pollution in the CZT was higher than that of O₃, and the probability of compound pollution was relatively low. However, if the daily limit values of the WHO transition period Phase 2 were taken as the target, the proportions of Zones I, II and IV would expand. The daily rate of increase in O₃ pollution would be higher than that of PM2.5 pollution, and the daily rate of increase in compound pollution would be the highest, reaching 111 days. Therefore, starting with stricter O₃ control measures may help manage potential compound pollution and further improve the quality of the atmospheric environment.

Figure 3 illustrates both the long-term variation (2017–2024) and the seasonal diurnal patterns of PM2.5 and O₃ concentrations in the CZT region. From the interannual perspective (Figure 3A), O₃ exhibited pronounced seasonal cycles with summer maxima, while PM2.5 showed higher concentrations during winter. The seasonal amplitude of O₃ was generally larger than that of PM2.5. Notably, O₃ levels in the summer of 2020 were comparatively lower than in adjacent years. Although no large-scale lockdown measures were implemented in the region, short-term localized mobility restrictions during the COVID-19 period may have influenced traffic-related emissions. However, this reduction did not substantially alter the overall seasonal pattern. From the diurnal perspective (Figure 3B), clear hourly variations were observed across seasons. In spring, summer, and autumn, O₃ concentrations increased from morning to afternoon, peaking around 15:00, while PM2.5 displayed an opposite trend. This pattern reflects enhanced photochemical activity under strong solar radiation, which favors O₃ formation at midday [33,34]. In winter, the diurnal variation in O₃ was weaker, and PM2.5 remained relatively stable throughout the day. In spring, summer, and autumn, the maximum 8 h seasonal mean O₃ approached or exceeded the WHO guideline value (100 μg/m³), indicating potential health risks.

To further examine the temporal relationship between PM2.5 and O₃ at the monthly scale, Figure 4 presents the changes in the correlation between the average daily concentrations of PM2.5 and O₃ each month from 2017 to 2023. The red part represents a positive correlation, the blue part represents a negative correlation, and the white part represents a low correlation. June to August generally exhibited positive daily correlations between PM2.5 and O₃, indicating frequent short-term synergistic behavior. In contrast, April, May, September, and October showed more variable correlations, reflecting the coexistence of synergistic and antagonistic events. It should be noted that a positive (or negative) daily correlation within a given month does not necessarily imply that the monthly mean concentrations of PM2.5 and O₃ vary in the same direction. This suggests that synergistic or antagonistic interactions often occur over relatively short time scales (several days) and may not substantially influence the overall monthly average concentration trends.

3.2. Spatial Analysis of PM2.5–O₃ Concentration

Since most of the monitoring stations are located in built-up areas of cities, the spatial analysis areas mainly focus on those. Figure 5 shows the mean values of PM2.5 and O₃ at the monitoring stations. The concentration difference between the highest and lowest value stations of PM2.5 and O₃ is about 10 μg/m³. The areas with high concentrations of PM2.5 are all located in the three central cities, with clustered economic activities and heavy-industry facilities. The concentration in the new urban area and the urban–rural fringe area is relatively low, with the lowest concentration at the control point, Shaoshan Station. The distribution of O₃ is significantly different from that of PM2.5, showing a trend of being higher in the north and lower in the south. In some built-up areas of Changsha City, the concentration of O₃ was significantly higher than that in the built-up areas of Zhuzhou and Xiangtan. Furthermore, the control point of Shaoshan Station was not the location with the lowest O₃ concentration. This may indicate that Changsha has atmospheric conditions more favorable for O₃ formation. The spatial correlation results showed that O₃ has higher spatial aggregation than PM2.5, and there were obvious hot areas (Changsha) and cold areas (Xiangtan, Zhuzhou).

In addition to the control stations, the PM2.5 hotspots were located in the surrounding centers of the built-up areas of the three cities. In view of the significant differences in the distribution of O₃ and PM2.5, the meteorological conditions in the three cities were relatively uniform. This study analyzes the distribution of O₃ from the perspective of precursor emissions. In the public data set from the China Environmental Protection Administration, NO₂ is a common precursor of PM2.5 and O₃, and SO₂ is one of the secondary precursors of PM2.5 [35,36]. The NO₂ pollution concentrations in the CZT were similar (about 30 μg/m³), but the SO₂ concentration in Changsha (7.9 μg/m³) was lower than those in Zhuzhou (10.7 μg/m³) and Xiangtan (11.2 μg/m³). This might be because there are more industrial emission sources in Xiangtan and Zhuzhou. Previous studies have shown that under some conditions, NO₂ could act as an oxidant to convert dissolved SO₂ into sulfate, enhancing secondary PM2.5 [37,38]. Altered reaction pathways of NO₂ may alter the availability of reactive nitrogen for photochemical reactions, and SO₂ can consume OH radicals [39]. They can consequently influence O₃ formation. Similarly, due to the lack of SO₂ emission sources at the Shaoshan control point, the O₃ concentration was higher than that in the urban areas of Zhuzhou and Xiangtan. To sum up, the PM2.5 hotspots in the CZT were urban built-up areas and heavy-industry clusters, while the O₃ cold spots were heavy-industry clusters. Controlling NO_x emissions is currently an effective measure to reduce both PM2.5 and O₃.

3.3. Analysis of Synergistic Changes in PM2.5–O₃ Pollution

To explore the mechanisms underlying short-term synergistic or antagonistic events between PM2.5 and O₃, this study examines two aspects: common precursors and meteorological factors. First, the driving factors of PM2.5 and O₃ in the CZT region are analyzed using daily average pollutant and meteorological data. Typical synergistic and antagonistic events are then investigated based on the model results. The regression results are presented in

Table 1. Driving factor stand error (in parentheses) analysis of PM2.5 and O₃ in CZT.

Pollutants	R2	NO2	SO2	Airt	DPD	Windd	Winds	Rain	Sunt
O3	0.65	0.08 (0.07)	−0.39 * (0.15)	2.48 * (0.12)	4.47* (0.36)	0.06 * (0.01)	−0.20 * (0.07)	0.01 (0.04)	2.24 * (0.29)
PM2.5	0.50	1.14 * (0.07)	0.83 * (0.13)	−0.5 * (0.1)	−0.84 * (0.33)	0.05 * (0.01)	0.20 * (0.04)	−0.21 * (0.04)	−0.16 (0.26)

* Indicates that the coefficient is significant when the confidence interval is 95%. Units of regression coefficients correspond to the change in pollutant concentration (μg/m³) per unit change in each predictor variable. All predictor units are described in Section 2.1. Sunshine duration (sunt, h), temperature (airt, °C), 6 h rainfall (rain, mm), wind direction (windd, º), wind speed (winds, m/s), dew point difference (DPD (t-dewp)).

. The R² of the O₃ model is higher than that of the PM2.5 model, indicating that O₃ variability is more strongly explained by the selected driving factors. PM2.5 is also influenced by direct primary emissions (e.g., black carbon and heavy metal particles), which do not directly contribute to O₃ formation and were therefore not included in the synergistic driving framework [40]. The regression coefficients indicate the direction and statistical significance of each driving factor. For PM2.5, precursor concentrations and air humidity show significant positive effects, whereas rainfall and air temperature exhibit significant negative effects. For O₃, meteorological variables exert stronger influences than precursors, as reflected in the significant positive effects of high temperature, longer sunshine duration, and low humidity. Wind speed shows significant but opposite effects on PM2.5 and O₃, suggesting that wind may enhance pollutant transport for PM2.5 while promoting the diffusion and dilution of O₃. Considering the seasonal pollution characteristics in the CZT, although synergistic or antagonistic events may occur in winter and summer, compound pollution episodes are less likely to develop under those conditions. Therefore, the analysis focuses on spring and autumn, when atmospheric conditions are more unstable and conducive to compound pollution. Based on the model results, representative antagonistic events (May 2017; April 2021) and synergistic events (October 2017; April 2022) were further examined to clarify the underlying mechanisms.

Figure 6 shows the changes in monitoring values of typical precursors and meteorological conditions for representative synergistic and antagonistic events in the month. First of all, the daily temperatures during the four months were all around 20 °C, which helps minimize seasonal background effects of PM2.5 and O₃. The concentration fluctuation of SO₂ was relatively low, and NO₂ showed positive variation trends during certain synergistic periods. Given the relatively small fluctuation of SO₂, its contribution to sulfate formation during these events is likely limited. Therefore, the main influencing factor of synergistic/antagonistic events was meteorological conditions. Antagonistic events normally last for relatively short periods—often only 2 to 3 days. For example, from the 27th to the 30th of April 2021, the precipitation and the air humidity significantly decreased. From the 11th to the 12th of May 2017, there was short-term heavy rainfall, but still long hours of bright sunshine, which could enhance photochemical reactions and promote O₃ formation. In October 2017, starting from the 17th, PM2.5 and O₃ experienced a synergistic event that lasted for a total of half a month, resulting in the occurrence of two compound pollution days. Firstly, from the 17th to the 27th, NO₂ concentration continued to rise, contributing to the precursor. In addition, during the period of coordinated rise, the wind speed decreased, the humidity remained in an intermediate state, and there was no rainfall, creating relatively stagnant atmospheric conditions. These meteorological conditions are similar to those described by Dai et al. [20] and were consistent with previous observation [41]. Under such conditions, enhanced solar radiation strengthened photochemical reactions. NO₂ photolysis contributes to O₃ formation. Meanwhile, NO₂ can be further oxidized to nitric acid, which subsequently forms nitrate (NO₃⁻), an important component of secondary inorganic PM2.5 [39]. However, NO₂ is not a significant driver of O₃ in the regression model, indicating that O₃ variability was more strongly controlled by photochemical conditions. During this period, on sunny days, the increase rate of O₃ was faster, and on cloudy days, the increase rate of PM2.5 was even faster. In addition to meteorological variability, the alternating enhancement may also be related to oxidation-driven secondary aerosol formation processes. Since O₃ can act as an oxidant to promote the formation of secondary organic aerosols, the antagonistic effect may also be related to this type of reaction [42]. The calm and stable weather ended on the 29 October. Strong winds caused the spread of O₃ and PM2.5, and their concentrations decreased simultaneously. The synergistic events did not always make P2.5–O₃ increase simultaneously. In April 2022, the fluctuation trends of PM2.5, O₃ and NO₂ were similar, and concentration changes were relatively stable. The concentration decreases around 15 and 30 April were both caused by the cold wave, accompanied by strong winds and rainfall.

In conclusion, the ascending phase of the synergistic event required an increase in the concentration of precursors, relatively constant humidity, and alternating weather conditions that were not prone to diffusion with sunny and cloudy days. In fact, most synergistic episodes lasted between 3 and 10 days, after which then they would be suppressed by new meteorological conditions before resulting in compound pollution. In order to decrease the concentrations of PM2.5 and O₃ simultaneously, strong winds are needed for diffusion or continuous rainfall, which makes the diffusion rate or sedimentation rate greater than the formation rate. Continuous rainfall can also suppress the conditions of photochemical reactions. With the continuous reduction in pollutant emissions in recent years, even if there were synergistic events, it would be difficult for them to cause compound pollution. On the contrary, as O₃ could undergo relatively large concentration changes in a relatively short period of time, antagonistic events might further exacerbate O₃ pollution. Therefore, under the current standards, the main work constraint is the high O₃ pollution level caused by antagonistic events. If the standards are further constrained, it will become necessary to further address the compound pollution caused by synergistic events.

4. Conclusions and Policy Recommendations

In terms of the long-term series and daily concentration changes in PM2.5 and O₃ in the CZT, the pollution trends were opposed. However, when measured by the average daily concentration, there were short-term synergistic events, with the time period mainly ranging from 3 to 10 days, although in extreme cases this could reach half a month. High-concentration areas of PM2.5 were concentrated in the built-up areas of various cities. The concentration of O₃ in Zhuzhou and Xiangtan was lower than that in Changsha, with a concentration difference of about 10 μg/m³. The comparatively lower SO₂ concentration in Changsha may influence the partitioning of nitrogen species between ozone formation and secondary particle production. Regarding the synergistic drivers of pollution, NO₂ was the main factor for PM2.5, while dew point difference and temperature were the main factors affecting O₃. During the spring and autumn seasons, against the backdrop of stable weather with low wind speeds (<3 m/s), no rainfall and moderate humidity (dew point difference between 4 and 6 °C), the alternation of sunny and cloudy days (sunshine duration between 1 and 9 h) caused the concentrations of PM2.5 and O₃ to rise alternately, forming a synergistic growth event.

Therefore, when aiming to reduce the combined pollution of PM2.5–O₃, controlling NO_x emissions may represent a potentially effective mitigation strategy. A substantial portion of urban NOx emissions originate from motor vehicles. Based on the meteorological condition results presented in this article, a more advanced algorithm for large-scale predictions of collaborative events can be developed. Temporary traffic control measures for vehicles, such as odd–even license plate restrictions, could be considered during predicted high-risk periods. O₃ is highly sensitive to precursors and weather conditions. The lack of a comprehensive VOC monitoring network limits current collaborative research and management for PM2.5–O₃. Future studies would benefit from improved VOC estimation methods such as developing and utilizing methods for indirectly estimating VOCs through satellite remote sensing. In addition, monitoring data for one of the driving factors of O₃, sunshine duration, is missing after 2020. It may be possible to replace these data with correlation indicators such as cloud coverage. To further improve air quality, the establishment of seasonal O₃ concentration management targets may further enhance control effectiveness. The summer special governance against O₃ can significantly reduce the annual average pollution level. The precise management strategy of PM2.5–O₃ for the region may contribute to achieving the WHO Stage II targets in the long term. In the future, research will be expanded to regions with different climate and economic backgrounds to assess its broad applicability.