Assessing the PM_2.5–O₃ Correlation and Unraveling Their Drivers in Urban Environment: Insights from the Bohai Bay Region, China

Abstract

Understanding the correlation between PM_2.5 and O₃ is critical for complex air pollution control. This study comprehensively analyzed PM_2.5 and O₃ pollution characteristics, uncovered spatiotemporal variations in their correlation, and investigated the driving mechanisms of their association in Dongying, a typical petrochemical city in China’s Bohai Bay region. Results showed that PM_2.5–O₃ correlation in Dongying exhibited significant seasonal variations, spatial patterns, and concentration threshold effects from 2017 to 2023. PM_2.5 and O₃ showed strong positive correlations in summer, negative in winter, and weak positive in spring/autumn, with strongest links in western areas. The strongest positive PM_2.5–O₃ correlation occurred in summer when PM_2.5 ≤ 35 μg·m⁻³ and O₃ >160 μg·m⁻³, while the strongest negative correlation was exhibited in winter with PM_2.5 > 75 μg·m⁻³ and O₃ ≤ 100 μg·m⁻³. Meteorological conditions (T > 20 °C, RH < 30%, wind speed < 1.73 m/s, O_x > 125 μg·m⁻³) and non-sea-breeze periods enhanced the PM_2.5–O₃ positive correlation. During the four typical pollution episodes, the positive PM_2.5–O₃ correlation in summer was propelled by synchronous increases in O₃ and secondary components via shared precursors. In autumn, strong positivity resulted from secondary component–O₃ correlations (r > 0.7) and dominance of secondary formation in PM_2.5. In winter, the negative correlation stemmed from primary emissions inhibiting photochemistry. Random forest analysis showed that O_x, RH, and T drove positive PM_2.5–O₃ correlation via photochemistry in summer, whereas winter primary emissions and NO titration caused negative correlation. This study offers guidance for the collaborative PM_2.5 and O₃ control in the petrochemical cities of the Bay region.

1. Introduction

The issue of combined fine particulate matter (PM_2.5) and ozone (O₃) complex pollution has garnered significant attention from the government and various regions [1,2,3]. Both PM_2.5 and O₃ pollution pose serious health risks, with high concentrations of PM_2.5 linked to cardiovascular and respiratory diseases, while high O₃ levels can cause eye irritation, sore throat, chest tightness, and headaches [4,5]. Globally, the combined exposure to PM_2.5 and O₃ leads to 37,651 premature deaths, far surpassing the sum of deaths from PM_2.5 (18,113) and O₃ (3649) exposure alone. PM_2.5 and O₃ complex pollution exhibits synergistic effects on overall mortality rates and those specifically related to cardiovascular and respiratory systems [6]. Moreover, more than 50% of cities worldwide are affected by the PM_2.5 and O₃ complex pollution [7]. China, as a pollution hotspot, has most of its cities classified as having “high risk” to “stable” exposure levels [8]. Besides the direct health impacts, PM_2.5 and O₃ pollution also lead to significant economic burdens [9,10]. According to World Bank data [11], the economic losses due to PM_2.5 exposure in 2019 ranged from 1.7% of GDP in North America to 10.3% in South Asia, with a global total of $8.1 trillion. Selin et al. [12] predict that, by 2050, global health-related economic losses due to O₃ pollution will reach $580 billion. In addition, an OECD report [13] indicates that Asia is the region most severely affected by the health and economic impacts of air pollution. Xie et al. [14] found that, if China does not mitigate PM_2.5 and O₃ pollution, GDP losses could range from 0.09% to 2.0% by 2030. In this context, there is an urgent need to advance the coordinated management of PM_2.5 and O₃ pollution to achieve the goal of controlling complex pollution.

The interaction between PM_2.5 and O₃ is coupled through three mechanisms [15,16]: (1) Shared precursors (NO_x/VOCs) form secondary inorganic aerosols (SIAs) and secondary organic aerosols (SOAs) via gas-particle conversion processes, establishing interdependent relationships with ozone formation; (2) aerosols modulate photochemical radiation flux, thereby influencing photochemical reaction processes and ozone generation; (3) aerosols participate in heterogeneous chemical reactions, providing abundant reaction beds and complicating chemical processes. In atmospheric chemistry, emissions of chemicals are internal factors, while meteorological conditions serve as external factors [17]. The PM_2.5–O₃ correlation is influenced by two main factors. On the one hand, the interaction between PM_2.5 and O₃ formation involves PM_2.5 scattering and absorbing shortwave radiation, which reduces the amount of radiation reaching the ground, thus enhancing ground-level O₃ photochemical formation [18]. Additionally, PM_2.5 indirectly affects O₃ concentrations by altering cloud microphysical processes (e.g., cloud optical thickness, effective cloud droplet radius) [19], while heterogeneous chemical reactions occurring on its surface also influence O₃ and secondary particle concentrations. Meanwhile, O₃, being a strong oxidant, drives the gas-to-particle conversion of atmospheric precursors, like SO₂, NO_x, and VOCs, thereby accelerating new particle formation [20]. On the other hand, meteorological factors, such as temperature, precipitation, humidity, wind, cloud cover, and atmospheric stability, influence the concentrations of PM_2.5 and O₃ through photochemical and physical processes, thereby affecting their correlation [21,22]. Therefore, it is essential to clarify the interactions and correlations between PM_2.5 and O₃ to provide scientific support for the development of effective joint control strategies for both pollutants.

Current research on the PM_2.5–O₃ correlation primarily focuses on countries such as China, the United States, South Korea, and those in Southeast Asia, revealing significant regional variability in their spatiotemporal correlation. For example, the PM_2.5–O₃ correlation was predominantly negative in Los Angeles, USA during the COVID-19 pandemic [23]. In summer, strong positive correlations (r > 0.6) between PM_2.5 and O₃ were observed in Bangalore (India) and Seoul (Republic of Korea), with negative correlations observed in winter [24]. In China, PM_2.5 and O₃ generally exhibit a positive correlation in summer and a negative correlation in winter [25,26], with South China showing an overall positive correlation and North China showing an overall negative correlation [27]. Research has also explored the mechanisms influencing the PM_2.5–O₃ correlation. From a chemical perspective, Hong et al. [22] and Zhu et al. [28] analyzed PM_2.5 secondary components and found that the positive correlation in summer is attributed to the synergistic production of O₃ and secondary inorganic components in PM_2.5 (primarily sulfate, nitrate, and ammonium, collectively referred to as “SNA”) driven by atmospheric photochemical reactions. In contrast, the wintertime negative correlation arises from high emissions and stagnant meteorological conditions promoting heterogeneous formation of SNA and secondary organic carbon (SOC), while the NO titration effect significantly suppresses O₃ concentrations. On a spatiotemporal scale, Chen et al. [29] revealed that the PM_2.5–O₃ correlation shows diurnal and regional variations—lower in the morning and higher in the afternoon, with predominantly positive relationships in southern and coastal regions and negative relationships in northern and inland regions. Changes in PM_2.5 and O₃ concentrations also affect the PM_2.5–O₃ correlation. Wang et al. [30] and Qiu et al. [26] found that, in developed regions, PM_2.5 and O₃ exhibit positive correlations when PM_2.5 concentrations are below 50 μg·m⁻³, conversely turning negative above this threshold. Additionally, local circulation patterns (such as sea-breeze winds) may modulate the synergistic effects between PM_2.5 and O₃ through their transport-accumulation impacts on pollutants, but relevant research remains scarce. The relationship between PM_2.5 and O₃ varies significantly across research fields, time scales, and methodologies, leading to differing conclusions under the influence of complex topography, meteorological conditions, and pollutant emissions. Overall, the current research on the underlying causes of the PM_2.5–O₃ correlation remains insufficiently comprehensive [31,32,33].

In 2017, the national ozone non-compliance rate increased significantly, and the peak concentration rose remarkably, making this period a crucial juncture for ozone pollution research [34]. In the same year, with the implementation of China’s Air Pollution Prevention and Control Action Plan, urban PM_2.5 levels have significantly dropped, yet O₃ pollution has emerged as a new challenge, becoming a key factor affecting air quality [35]. Bay areas become global oil and gas-rich regions due to geological features, ancient marine deposits, and thermal conditions [36,37], with petrochemical cities boosting regional economies but causing air pollution and climate crises [38]. Dongying, located in Shandong Province, China, is situated in the southwest of the Bohai Bay and faces the Bohai Sea to the east. It is one of the typical coastal cities around the Bohai Bay. Its role as a major oil base in China—characterized by Shengli Oilfield resource development and a road-dominated transportation structure—has created high-intensity NO_x and VOCs emission sources [39,40], making it prone to PM_2.5 and O₃ complex pollution [41]. In Dongying, the SNA and SOA account for 31.64% to 48.30% of PM_2.5 [42], indicating that secondary pollution plays a critical role in regional air quality. Given the atmospheric complex pollution caused by the mutual influence between PM_2.5 and O₃ in Dongying in recent years, it is highly imperative to conduct a study on the correlation and underlying mechanisms between PM_2.5 and O₃ in this typical petrochemical city during the period from 2017 to 2023. This research is crucial for the synergistic control of both pollutants.

This study investigates Dongying, a representative petrochemical city in the Bohai Bay region of China. Using environmental air quality monitoring data and meteorological information from 2017 to 2023, the study analyzes the pollution characteristics of PM_2.5 and O₃ in Dongying, as well as the spatiotemporal variations in their correlation. The research also explores the impact of meteorological conditions, sea–land breeze, atmospheric oxidizing capacity, PM_2.5 and O₃ concentration levels, PM_2.5 chemical components, and precursor substances on the PM_2.5–O₃ correlation. The findings of this study provide important guidance for the targeted implementation of collaborative control measures for PM_2.5 and O₃ in petrochemical cities of the Bay region.

2. Materials and Methods

2.1. Research Region

Dongying, located in northern Shandong Province, China, borders the Bohai Sea to the east and serves as a key outlet for the Yellow River Basin. Its developed petrochemical industry results in substantial VOCs and NO_x emissions, making the city prone to atmospheric complex pollution under unfavorable meteorological dispersion conditions [43]. Between 2017 and 2023, eight national ambient air quality monitoring stations were established in Dongying (Figure 1), including four in Dongying District: Shengli Station (operational since January 2020, replacing the Xicheng Yangguang Huanbao site), Ideal City Station (operational since November 2020, replacing Gengjingcun Station), Xicheng Yangguang Huanbao Company Station (ceased operation in April 2021), and Gengjingcun Station (ceased operation in November 2020). One station was located in Hekou District (Hekou Urban Station), one in Kenli District (Minfeng Road Station), and two in the Dongying Economic and Technological Development Zone (Dongying Municipal Environmental Protection Bureau Station and Development Zone Management Committee Station). The abbreviations for these stations can be found in Table S1.

2.2. Data Sources

2.2.1. Criteria Air Pollutants

The data for conventional air pollutants used in this study were the data from the national control stations in Dongying from January 2017 to December 2023. These data included the hourly, daily, monthly and annual average values of conventional pollutants such as PM_2.5, PM₁₀, SO₂, CO, NO, NO₂ and NO_x, the monthly and annual evaluation values of PM_2.5, the hourly average value of O₃, the maximum daily 8-h average O₃ concentration (MDA8 O₃), as well as the monthly and annual evaluation values for O₃. The data were sourced from the Shandong Provincial Environmental Air Quality Data Monitoring and Management System “http://123.232.114.95:8001/ (accessed on 19 July 2024)”.

2.2.2. Super Station Observation Programs

The data used in this study were obtained from the Dongying Atmospheric Super Observation Station (hereafter referred to as “Super Station”, located at 118.59° E, 37.45° N). The hourly PM_2.5 component data and Organic Carbon/Elemental Carbon (OC/EC) data from August 2021 to July 2022, and hourly online data for VOCs and meteorological parameters (including temperature, relative humidity, atmospheric pressure, wind speed, and wind direction) were collected from May 2021 to December 2023. Details of the Super Station observation programs and associated instrumentation in Dongying are summarized in Table S2. A five-parameter compact meteorological station was employed for real-time online monitoring of wind speed, wind direction, atmospheric pressure, temperature, and humidity (Table S3).

2.2.3. Cloud Data

Hourly total cloud cover (TCC) data were acquired from ERA-5 reanalysis datasets provided by the European Centre for Medium-Range Weather Forecasts (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview, accessed on 19 July 2024) for the period from 2021 to 2022. Grid data at 118.6° E, 37.43° N (0.25° × 0.25° spatial resolution) were extracted for analysis.

2.2.4. Quality Control and Quality Assurance

Quality Control of Monitoring Instruments

Online monitoring of criteria air pollutants strictly complied with the Technical Specifications for Continuous Automated Monitoring Systems of Ambient Air Gaseous Pollutants (SO₂, NO₂, O₃, CO) Operation and Quality Control (HJ 818-2018) [44]. OC/EC analysis followed the thermal–optical transmission methodology outlined in QX/T 70-2007 [45] standards for atmospheric aerosol characterization. Inorganic ion analysis demonstrated excellent reproducibility (RSD ≤ 2%, n ≥ 7) and high internal standard recovery rates (>99%), ensuring precise quantification of ionic components in PM_2.5. VOC measurements were validated through multi-concentration standard gas calibrations (R² ≥ 0.9), routine internal verifications (deviation ≤ 10%), and daily retention time window adjustments. Quality assurance for all instrumentation is detailed in Table S2.

Data Quality Control

The data collection frequency, timing, quality assurance, and quality control for all monitoring projects adhere to the standards outlined in the following guidelines: Automatic Methods for Ambient Air Quality Monitoring (HJ/T193-2005) [46], Specifications and Test Procedures for Ambient Air Quality Continuous Monitoring System with Gas Chromatography for Volatile Organic Compounds (HJ 1010-2018) [47], Technical Specifications for Continuous Automated Monitoring of Organic carbon and Elemental Carbon in Ambient Air Particulate Matter (PM_2.5) (HJ 1327-2023) [48], Technical Specifications for Continuous Automated Monitoring of Water-Soluble Ions in Ambient Air Particulate Matter (PM_2.5) (HJ 1328-2023) [49], and Technical Specifications for Continuous Automated Monitoring of Inorganic Elements in Ambient Air Particulate Matter (PM_2.5) (HJ 1329-2023) [50].

To ensure the reliability of the research data, this study conducted quality control on the dataset and removed missing values and outliers in line with the Ambient Air Quality Standards (GB 3095-2012) [51], Technical Specification for Ambient Air Quality Evaluation (on trial) (HJ 663-2013) [52], and Technical Regulation for Air Quality Index (AQI) (on trial) (HJ 663-2012) [53].

2.3. Data Processing

2.3.1. Identification of Primary Pollutants and Statistical Analysis of Polluted Days

In accordance with the Technical Regulation on Ambient Air Quality Index (on trial) (HJ 633-2012) [53], when the AQI exceeds 50, the primary air pollutant is defined as the one with the highest Individual Air Quality Index (IAQI). Meanwhile, the Technical Specification for Ambient Air Quality Evaluation (on trial) (HJ 663-2013) [52] defines various pollution types, such as PM_2.5-P (Standalone PM_2.5 pollution), O₃-P (Standalone O₃ pollution), and PM_2.5+O₃-P (PM_2.5+O₃ complex pollution). A detailed description of polluted days can be found in Text S1, and the statistical results are presented in Table S4. Based on the variations in O₃ and PM_2.5 pollution in Dongying, this study selects different pollution episodes for case analysis across all four seasons. The selection criteria for pollution episodes are outlined in Text S1, and the results are shown in Table S5.

2.3.2. Correlation Analysis

Ambient air quality data from eight monitoring stations in Dongying spanning 2017–2023 were analyzed to compute Pearson correlation coefficients (r) between PM_2.5 and O₃ concentrations using Excel 2019. Among them, the annual, monthly, and seasonal correlations between PM_2.5 and O₃ were calculated based on the daily mean concentration of PM_2.5 and the daily maximum 8-h sliding average values of O₃, while the daily correlation was calculated based on the hourly average values of PM_2.5 and O₃. Considering the seasonal characteristics of pollutants, the data from each station were divided by season (spring: March–May; summer: June–August; autumn: September–November; winter: December–February of the following year).

2.3.3. Quantification of Primary and Secondary PM_2.5

The maximum daily concentration of 8-h moving average O₃ [MDA8] is used as the indicator for photochemical activity levels [54]. Based on this, atmospheric photochemical activity is categorized into four levels: low photochemical activity level ([MDA8] ≤ 100 μg·m⁻³), mild photochemical activity level (100 μg·m⁻³ < [MDA8] ≤ 160 μg·m⁻³), moderate photochemical activity level (160 μg·m⁻³ < [MDA8] ≤ 200 μg·m⁻³), and high photochemical activity level ([MDA8] > 200 μg·m⁻³). This study excludes data from dust storm events in Dongying and estimates secondary PM_2.5 in PM_2.5 concentrations based on photochemical activity levels. The equation is as follows:

$${\left[{\mathrm{PM}}_{2.5}\right]}_{\mathrm{pri},\mathrm{I},\mathrm{t}}={\left[\mathrm{CO}\right]}_{\mathrm{I},\mathrm{t}} \times {\left(\left[{\mathrm{PM}}_{2.5}\right]/\left[\mathrm{CO}\right]\right)}_{\mathrm{pri},\mathrm{L} },$$

(1)

$${\left[{\mathrm{PM}}_{2.5}\right]}_{\sec ,\mathrm{I},\mathrm{t}}={\left[{\mathrm{PM}}_{2.5}\right]}_{\mathrm{obs},\mathrm{I},\mathrm{t}} - {\left[{\mathrm{PM}}_{2.5}\right]}_{\mathrm{pri},\mathrm{I},\mathrm{t}},$$

(2)

where [PM_2.5]_pri represents the primary PM_2.5 concentration (μg·m⁻³), [PM_2.5]_sec refers to the secondary pollutant concentration (μg·m⁻³), and [PM_2.5]_obs is the observed PM_2.5 concentration (μg·m⁻³). ‘t’ represents the observation day (d), while the subscript L denotes low photochemical activity level, and the subscript I corresponds to low, moderate, or high photochemical activity levels (I = 1, 2, 3). For detailed calculation methods, please refer to the study by Li et al. [54].

2.3.4. Quantification of Organic Aerosol Mass Concentration

In urban areas, OC may primarily originate from SOC. To investigate the secondary transformation process of organic matter, this study further analyzed the OC/EC ratio, which is commonly used to identify the presence of secondary organic pollution. A threshold of OC/EC > 2.0 is generally accepted as an indicator of SOC formation [55]. SOC concentration was calculated using the empirical formula

$$\mathrm{SOC}={\mathrm{OC}}_{\mathrm{tot}}-\mathrm{EC}\times {\left(\mathrm{OC}/\mathrm{EC}\right)}_{\min }$$

(3)

where OC_tot represents total organic carbon, and (OC/EC)_min denotes the minimum OC/EC ratio observed over 24 h each day during the study period.

2.3.5. Objective Synoptic Classification

This study selected synoptic models most representative of pollution formation and dissipation in China’s central and eastern regions based on relevant literature [56]. To minimize the influence of local mesoscale systems, the analysis focused on synoptic-scale systems characteristics, with the surface synoptic system analyzed using the average fields from four times (02:00, 08:00, 14:00, and 20:00 local standard time).

2.3.6. Sea–Land Breeze Identification Criteria

Based on existing studies on the classification of land–sea winds [57,58,59] and the natural geographic features, such as the coastline of Dongying City, we propose the following criteria for identifying land–sea breezes in the region: winds from the SSE to NW direction (157.5° to 315°) are defined as land breezes, while winds from the NNW to SE direction (0° to 135° and 337.5° to 360°) are defined as sea breezes, as shown in Figure 1. A natural day was classified as a sea–land breeze day if three criteria were met: ① 24-h average surface wind speed < 10 m/s; ② during land breeze hours (01:00–08:00), land breeze occurrence duration ≥ 4 h and sea breeze occurrence duration ≤ 2 h; ③ during sea breeze hours (13:00–20:00), sea breeze occurrence duration ≥ 4 h and land breeze occurrence duration ≤ 2 h.

This study selected land–sea breeze days from 2022 to 2023 in Dongying. The correlations between the concentrations of PM_2.5 and O₃ during the sea–land breeze periods and non-sea–land breeze periods were comparatively analyzed, and the seasonal, diurnal and nocturnal impacts of the sea–land breeze on the correlation between PM_2.5 and O₃ were explored.

2.3.7. Random Forest

To further investigate how meteorological factors (temperature, relative humidity, atmospheric pressure, wind speed, and wind direction) and gaseous pollutants affect PM_2.5 and O₃ concentration variations (given our prior focus on seasonal trends), this study employs Breiman’s [60] random forest model to quantify the importance of contributing factors [61,62,63] (model details in Text S1). By integrating model predictions with analytical findings, this study evaluates the influence of these factors on the PM_2.5–O₃ correlation. The random forest model was developed and run using Python (Version 3.6.3) for this analysis.

3. Results

3.1. Variation Patterns in PM_2.5 and O₃ Concentration Levels

3.1.1. Spatiotemporal Distribution Characteristics of PM_2.5 and O₃

From 2017 to 2023, concentrations of SO₂, NO₂, PM₁₀, CO, and PM_2.5 in Dongying showed significant declines, leading to overall air quality improvement, whereas O₃ pollution intensified (Figure 2a). Specifically, the annual evaluation value of PM_2.5 decreased by 36%, from 55 μg·m⁻³ to 35 μg·m⁻³, while the 90th percentile of MDA8 O₃ increased from 176 μg·m⁻³ to 186 μg·m⁻³. The monthly evaluation value of PM_2.5 and O₃ displayed opposite trends (Figure 2b), with PM_2.5 concentrations peaking in winter and reaching a maximum of 74 μg·m⁻³ in January, while O₃ concentrations were highest in summer and lowest in winter. During the PM_2.5 pollution season (from December to February), daily concentrations showed a declining trend, with daytime and nighttime levels in 2023 decreasing by 23 μg·m⁻³ and 24 μg·m⁻³, respectively, compared to 2017. Meanwhile, the O₃ pollution season (from April to September) showed an increasing trend in daily concentrations, with daytime concentrations rising by 11 μg·m⁻³ and nighttime levels increasing even more significantly by 17 μg·m⁻³.

In terms of spatial distribution, PM_2.5 concentrations in Dongying from 2017 to 2023 generally followed a “high in the south and east, low in the north and west” pattern, with an overall decreasing trend (an average reduction of 23.5%). Notably, PM_2.5 concentrations were higher in the western part of the city (ranging from 55 to 76 μg·m⁻³) and lower in the central and eastern regions in winter. The overall spatial distribution of O₃ concentrations was characterized by higher levels in the northern, central, and eastern regions and lower levels in the western region, while also showed a consistent annual increase (Figure 3).

3.1.2. Pollution Regime Characterization of PM_2.5 and O₃

From 2017 to 2023, primary air pollutants in Dongying included O₃, PM_2.5, PM₁₀, and NO₂, with no occurrences of SO₂ or CO as primary pollutants (Figure S1). The annual total days with PM_2.5 and PM₁₀ as primary pollutants decreased, NO₂ increased initially before declining, while O₃ showed a year-by-year upward trend.

From 2017 to 2023, PM_2.5 pollution days (PM_2.5-P) in Dongying decreased from 73 to 34 days, with mild pollution accounting for 60.8% of PM_2.5-P, while moderate, severe, and extreme pollution represented 24.4%, 14.5%, and 0.3%, respectively (Figure S2). PM_2.5-P occurred primarily during spring, autumn, and winter, with January and December in winter being the most dominant months (Figure S2). Typical winter PM_2.5 pollution episodes were triggered by elevated precursor concentrations coupled with stagnant meteorological conditions. Analysis of two pollution phases revealed concurrent increases in PM_2.5 and O₃ concentrations (detailed in Text S2). In contrast, O₃ pollution days (O₃-P) increased from 52 to 91 days, with 80.6% mild, 18.5% moderate, and 0.9% severe pollution events (Figure S2). O₃-P occurred in spring, summer, and autumn, with June in summer showing the highest frequency (21 days) (Figure S2). Typical summer O₃ pollution episodes formed under elevated temperatures, poor dispersion, and high precursor concentrations. Two pollution phases revealed divergent PM_2.5 and O₃ trends, with atmospheric oxidizing capacity enhancing PM_2.5 formation and meteorological conditions synergistically exacerbating pollution (Text S3).

3.1.3. Complex Pollution Characterization of PM_2.5 and O₃

Between 2017 and 2023, the complex pollution days of PM_2.5 and O₃ (PM_2.5+O₃-P) in Dongying primarily occurred in March to April during spring and October in autumn, accounting for over 70% of the total. High-concentration O₃-type PM_2.5+O₃-P events were occasionally observed in June during the summer (Figure S2). The number of PM_2.5+O₃-P days peaked in 2018, gradually decreased, and began to rise again starting in 2022. The concentration range of PM_2.5 in PM_2.5+O₃-P events was between 76 and 134 μg·m⁻³ (higher in spring than autumn), while O₃ concentrations ranged from 161–256 μg·m⁻³. Four days of PM_2.5+O₃-P events occurred in the summer of 2017, indicating that the increase in O₃ concentrations during this period facilitated the formation of PM_2.5 (Figure 4). Typical PM_2.5 and O₃ complex pollution events in spring and autumn were triggered by an increase in precursor concentrations combined with stable meteorological conditions. During spring pollution events, the atmospheric oxidative capacity significantly promoted PM_2.5 formation. In autumn, the trends in PM_2.5 and O₃ concentration changes were similar (see detailed explanation in Text S4).

3.2. Spatiotemporal Analysis of PM_2.5 and O₃ Correlation Characteristics

3.2.1. Temporal Variation in PM_2.5 and O₃ Correlation

Dongying exhibited significant seasonal variations in PM_2.5–O₃ correlation from 2017 to 2023 (Figure 5 and Figure 6). Among these, summer showed the strongest positive correlation (r = 0.60), winter a negative correlation (r = −0.19), and spring/autumn weak positive correlations (r = 0.18 and 0.13, respectively). Notably, PM_2.5 and O₃ reached their highest correlation in September 2022 (r = 0.92). The PM_2.5–O₃ correlation variability was greater during the daytime than the nighttime. Notably, spring exhibited significant daytime positive correlations, while summer showed positive associations during 9:00–11:00. Nighttime PM_2.5 and O₃ were positively correlated only in summer (20:00–23:00), with negative correlations observed in other seasons.

Figure 6 revealed significant positive correlations between PM_2.5 and O₃ in spring and autumn at PM_2.5 < 50 μg·m⁻³, whereas at PM_2.5 > 60 μg·m⁻³, MDA8 O₃ concentrations decreased with increasing PM_2.5 levels. Notably, winter exhibited negative correlations across all PM_2.5. concentrations, where O₃ concentrations declined as PM_2.5 increased. Although summer showed overall synchronous increases in PM_2.5 and O₃, the rate of O₃ growth significantly slowed when PM_2.5 exceeded 60 μg·m⁻³. When PM_2.5 levels were high and O₃ levels low, the PM_2.5–O₃ correlation tended to be negative, consistent with findings of Chen et al. [29].

There was a strong positive correlation between PM_2.5 and O₃ in summer, while a negative correlation was observed in winter. In spring and autumn, a positive correlation was found when the concentration of PM_2.5 was low (PM_2.5 < 50 μg·m⁻³), and a negative correlation emerged when the concentration of PM_2.5 was high (PM_2.5 > 60 μg·m⁻³).

3.2.2. Spatial Heterogeneity of PM_2.5 and O₃ Association

From 2017 to 2023, the correlations between daily evaluated PM_2.5 and O₃ values in Dongying showed a spatially heterogeneous distribution. Annually, central stations (ICC, r = 0.72) in 2020 and western stations (XSEPC, r = 0.76) in 2021 exhibited positive correlations, while super monitoring and provincial-controlled stations generally showed negative correlations (Figure S7). Seasonally, in winter, the PM_2.5 and O₃ correlation in the northern and eastern regions were weakly negative, whereas the western region had weak positive correlation. Summer demonstrated a universally positive correlation between PM_2.5 and O₃ across all areas, with strong positive correlations observed at western stations. Western regions PM_2.5 and O₃ exhibited stronger positive correlations in spring/autumn (Figure 7). Overall, a significant synergistic effect between PM_2.5 and O₃ was observed in the western region of Dongying.

3.2.3. Pollution Level-Dependent Correlation Patterns

The daily evaluated PM_2.5 and O₃ values were classified according to Table S6. Dongying’s PM_2.5 ≥ 75 μg·m⁻³ days were more frequent in autumn and winter than spring/summer. Summer dominated O₃ ≥ 160 μg·m⁻³ days (39.5%, significantly higher than other seasons), while spring recorded the highest O₃ 100–160 μg·m⁻³ frequency. Autumn had the largest O₃ ≤ 100 μg·m⁻³ proportion (14.6% for O₃ ≥ 160 μg·m⁻³), and winter exhibited the highest O₃ ≤ 100 μg·m⁻³ percentage (95.9%).

The PM_2.5–O₃ correlation was influenced by the gradient of PM_2.5 concentrations (Figure 8a). In summer, the strongest positive correlation of the year (r = 0.43) was observed when PM_2.5 ≤ 35 μg·m⁻³, while winter exhibited negative correlations at both PM_2.5 ≤ 35 μg·m⁻³ and 75 μg·m⁻³ < PM_2.5 ≤ 200 μg·m⁻³. In autumn, the correlation was positive when PM_2.5 ≤ 35 μg·m⁻³ but it transitioned to negative correlations with progressively stronger negative trends as PM_2.5 concentrations increased. Additionally, the PM_2.5–O₃ correlation was influenced by the level of photochemical activity (Figure 8b). In summer, the strongest positive correlation (r = 0.37) occurred under moderate photochemical activity whereas, in winter, the most significant negative correlation (r = −0.34) emerged under low activity despite light activity yielding positive trends (r = 0.40). In autumn, negative correlations were predominant during low-to-moderate activity periods but, under high activity, positive correlations strengthened by 30%. In spring, positive correlations between PM_2.5 and O₃ were maintained only under light photochemical activity, with negative correlations prevailing during other periods.

In summer, the strongest positive correlation between PM_2.5 and O₃ occurred when PM_2.5 ≤ 35 μg·m⁻³ and 160 μg·m⁻³ < O₃ ≤ 200 μg·m⁻³. In winter, the strongest negative correlation emerged at 115 μg·m⁻³ < PM_2.5 ≤ 200 μg·m⁻³ and O₃ ≤ 100 μg·m⁻³. In spring, significant positive correlation was found at PM_2.5 ≤ 35 μg·m⁻³ and 100 μg·m⁻³ < O₃ ≤ 160 μg·m⁻³ but weakened as their concentrations rose. In autumn, a significant positive correlation was observed when PM_2.5 ≤ 35 μg·m⁻³ and 200 μg·m⁻³ < O₃.

3.3. Impact Analysis of Meteorological Conditions

3.3.1. Synoptic Weather Patterns

The nine weather types were categorized into three types: high-pressure field, low-pressure field, and mean pressure field (Table S7 and Figure S8). In the “standard day” scenario, the dominant weather types were mean pressure field (36%), low-pressure field (36%), and high-pressure field (29%). Low-pressure field types dominated both O₃-P (60%) and PM_2.5-P (50%) episodes, which typically featured southeasterly horizontal winds inhibited by northwestern mountainous terrain, thereby promoting O₃ formation and accumulation. Mean pressure fields accounted for 60% of PM_2.5+O₃-P events, associated with stable weather and stagnant conditions that promoted pollutant buildup (Figure 9). In Dongying, O₃ and PM_2.5 accumulations were more significantly influenced by low-pressure weather, whereas concurrent high PM_2.5–O₃ days were strongly dependent on mean pressure field conditions.

Among the three weather patterns, the correlation between daily evaluated PM_2.5 and O₃ values reached its strongest positive correlations in summer, was weaker in spring and autumn, while winter exhibited negative correlations (Figure 10a). In both summer and autumn, the most significant positive correlation between PM_2.5 and O₃ occurred under a high-pressure field (high-pressure field > low-pressure field > mean pressure field), while spring showed an opposite pattern (mean pressure > low-pressure > high-pressure). In winter, the negative correlation between PM_2.5 and O₃ followed the order: mean pressure field > high-pressure field > low-pressure field. Under spring’s mean pressure field and low-pressure field, a 1 μg/m³ increase in MDA8 O₃ concentration led to a 0.38 μg/m³ increase in PM_2.5 concentration (ΔPM_2.5/ΔO₃ = 0.38, Figure 10b,d). In summer, ΔPM_2.5/ΔO₃ was 2.69 under a high-pressure field (Figure 10c) and 2.04 under a low-pressure field (Figure 10d). In autumn, ΔPM_2.5/ΔO₃ was 0.41 under a low-pressure field (Figure 10d). In winter, under a low-pressure field, a 1 unit increase in MDA8 O₃ concentration resulted in a 0.12 μg/m³ decrease in PM_2.5 concentration (ΔPM_2.5/ΔO₃ = −0.12, Figure 10d).

The dynamic uplifting effect of the northwestern mountain range on southeastern airflows, combined with low-pressure systems, creates OPA/PPA pollution hotspots, while the mean pressure field, through thermal stability effects, becomes a key trigger for PM_2.5–O₃ complex pollution. Summer high-pressure systems promote PM_2.5–O₃ synergistic accumulation (ΔPM_2.5/ΔO₃ = 2.69) through enhanced photochemical reactions and vertical mixing, and the role of uniform pressure fields in facilitating this accumulation should not be overlooked.

3.3.2. Key Meteorological Parameters

Temperature

PM_2.5 and O₃ formation conditions varied with temperature, where high temperatures favored O₃ production and accumulation, whereas low temperatures promoted PM_2.5 formation and buildup (Figure 11a). In summer, under high temperatures, PM_2.5 and O₃ showed a strong positive correlation (r = 0.60), while in winter, under low temperatures, PM_2.5 and O₃ exhibited a negatively correlation (r = −0.19). During spring and autumn, due to temperature fluctuations, PM_2.5 and O₃ displayed a transitional weak positive correlation (r = 0.13 and 0.19, respectively). Scatter plots for all four seasons revealed a triangular distribution, where the positive correlation between PM_2.5 and O₃ strengthened significantly with increasing temperature. The correlation coefficient between PM_2.5 and O₃ increased from 0.21 to 0.56 as temperatures rose from 5–10 °C to 20–25 °C, peaking at 0.58 during high-temperature periods (>25 °C). Conversely, negative correlation deepened sharply as temperatures dropped, with the correlation coefficient between PM_2.5 and O₃ changing from −0.40 to −0.60 as temperatures fell from 0–5 °C to −10~−5 °C (Figure 11b). Qian et al. [64] found that both O₃ and PM_2.5 concentrations significantly increased with rising temperatures, with MDA8 O₃ and PM_2.5 increasing by 4.8 ppb and 1.9 μg·m⁻³ for each 1 °C temperature rise, respectively.

Studies showed [65,66] that the most significant meteorological factor influencing PM_2.5 and O₃ concentrations in northern regions was temperature, followed by wind speed, humidity, pressure, and solar radiation. Within the temperature range of 20 °C to 25 °C and above 25 °C, as temperature increased, enhanced sunlight and higher emissions of VOCs and NO_x from biogenic sources and motor vehicles [67,68] activated photochemical reactions, leading to an increase in both O₃ and PM_2.5 concentrations. Furthermore, during the summer, stratospheric O₃ intrusion [69,70] and the increased oxidizing potential of O₃ promoted the formation of secondary aerosols [54,68], further strengthening the positive correlation between O₃ and PM_2.5. When temperatures were below 5 °C, suppression of photochemical reactions reduced O₃ formation. PM_2.5, on the other hand, accumulated through processes such as the formation of particulate-phase SNA at low temperatures [71], and hygroscopic growth in high humidity (with winter peaks in water-soluble ions [72,73]). Additionally, the increase in PM₂.₅ concentration could have enhanced aerosol optical thickness [74], significantly reducing the intensity of solar radiation reaching the surface, which in turn decreased the amount of O₃ produced by NO₂ photolysis. Therefore, O₃ formation was suppressed by low temperatures while PM_2.5 concentrations continued to rise, significantly strengthening in the negative correlation of PM_2.5 and O₃.

Humidity

As shown in Figure 11c, during the summer, relative humidity (RH) ranged from 30% to 90%, with higher O₃ concentrations when the daily PM_2.5 mean concentration was below 75 μg·m⁻³. In winter, RH concentrated between 20% and 70%, and PM_2.5 and O₃ exhibited a negative correlation. As RH increased, the PM_2.5–O₃ correlation changed from positive (r = 0.09 when RH < 30%) to negative (r = −0.41 when RH > 80%) (Figure 11d). In spring, when RH < 30%, PM_2.5 and O₃ showed a significant positive correlation (r = 0.52). As RH increased, the correlation weakened. In summer, the correlation was strongest when RH > 60%. In autumn, the correlation was strongest when RH was between 30% and 50%. In winter, when RH > 80%, a significant negative correlation was observed (r = −0.79) (Figure 11e). Overall, when RH < 30%, PM_2.5 and O₃ showed a positive correlation. As RH increased, the correlation shifted from positive to negative and, when RH > 80%, a negative correlation was observed.

Under high humidity conditions (>60%), PM_2.5 and O₃ exhibited a negative correlation. On the one hand, the increased hygroscopicity of particles enhanced the contribution of sulfate and nitrate ions [75], while also boosting the liquid-phase oxidation of SO₂ and NO_x [76,77]. Additionally, high humidity promoted the reaction of VOCs with oxidants in the liquid phase to form SOA [78,79], collectively driving an increase in PM_2.5 concentration. On the other hand, the increased cloud cover associated with high humidity reduced solar radiation, weakening photochemical reactions and affecting the heterogeneous reactions on PM_2.5 surfaces. Furthermore, the liquid water content of aerosols dissolved HO₂ free radicals, promoting liquid-phase reactions (e.g., HO₂(aq) + O₃ → OH + 2O₂), significantly reducing the concentration of HO₂ in the gas phase and thereby decreasing the rate of O₃ production [80,81,82]. Under low humidity conditions (<30%), PM_2.5 and O₃ exhibited a positive correlation. Reduced water vapor decreased UV radiation absorption, increasing the photolysis rate of NO₂ and directly promoting O₃ formation. In the high temperature and low humidity haze conditions, the OH radical-driven oxidation of NO_x was enhanced, which increased the formation rate of NO₃⁻, further driving an increase in PM_2.5 concentrations [83].

Wind Speed

Wind speed influenced the concentration of atmospheric pollutants by affecting diffusion conditions. As shown in Figure 11f, with increasing daily average values of PM_2.5 and O₃, the daily wind speed gradually decreased. This indicated that wind speed significantly influenced PM_2.5 and O₃ accumulation while stable weather conditions favored their local formation and accumulation. When the daily O₃ concentration exceeded 265 μg·m⁻³, the daily average wind speed ranged from 0.71 to 2.69 m/s. Similarly, when the daily PM_2.5 concentration exceeded 100 μg·m⁻³, the wind speed ranged from 1.73 to 2.12 m/s. When both PM_2.5 exceeded 100 μg·m⁻³ and O₃ exceeded 160 μg·m⁻³, the daily wind speed was 1.73 m/s. Therefore, as the concentrations of PM_2.5 and O₃ increased, wind speed decreased, suggesting that low wind speeds associated with stable weather conditions contributed to the simultaneous increase in PM_2.5 and O₃.

Under stable meteorological conditions (wind speed < 1.73 m/s), the diffusion ability of PM_2.5 and O₃ was significantly restricted, leading to a prolonged retention time of local pollutants. NO_x and VOCs precursors accumulated continuously, efficiently generating O₃ through photochemical chain reactions. Meanwhile, O₃ and OH radicals released during these reactions oxidized SO₂ and NO_x into sulfates and nitrates, driving VOC oxidation into SOA and resulting in a simultaneous rise in PM_2.5 and O₃ concentrations.

3.3.3. Sea–Land Breeze

Between 2022 and 2023, Dongying experienced 105 sea–land breeze days, peaking in autumn (35 days) and lowest in winter (18 days), with 2023 showing an increase from the previous year (Figure 12a). During both sea–land breeze and non-sea–land breeze periods, there was a negative correlation between PM_2.5 and O₃, but the negative correlation during the sea–land breeze period was generally weaker than in the non-sea–land breeze period. The maximum attenuation occurred in spring, followed by autumn. During the non-sea–land wind period, the negative correlation between PM_2.5 and O₃ during day and night was stronger than during the sea–land breeze period, with maximal autumn gradients (Figure 12b,c). When comparing sea–land breeze (Figure 12d), in the non- sea–land breeze period, the negative correlation between PM_2.5 and O₃ was stronger during the sea breeze time (13:00–20:00) in spring/autumn/winter. During sea–land breeze periods, the sea breeze hours in summer and autumn exhibited stronger PM_2.5–O₃ correlations than land wind hours, while in spring and winter the opposite was observed. Collectively, sea–land breeze circulation significantly weakened the PM_2.5–O₃ correlation.

PM_2.5 and O₃ diurnal variations were modulated by sea–land breeze (Figure 13). PM_2.5 concentrations followed a unimodal pattern in spring/summer/autumn but a bimodal pattern in winter, with winter recording the highest peak–trough values, followed by spring, autumn, and summer. O₃ concentration showed unimodal patterns across all seasons, with summer having the highest peak–trough values, followed by spring, autumn, and winter. Summer’s O₃ peak (166 μg·m⁻³ during 13:00–16:00 sea–land breeze periods) matched Meng et al. [32] findings. During the land–sea breeze periods, the PM_2.5 peak value increased by 15–36%, while the O₃ peak value rose by 5–8%. The PM_2.5 trough value decreased by 10–20%, and the O₃ trough value decreased by 15–25%. During the land–sea breeze periods, the O₃ peak value occurred 1–2 h earlier compared to non-land–sea breeze periods (summer: land–sea breeze period 13:00–16:00, non-land–sea breeze period 15:00–18:00). The trough value occurred 2 h earlier (summer: land–sea breeze period 05:00–07:00, non-land–sea breeze period 07:00–09:00). The land–sea breeze might promote the generation and accumulation of O₃, accelerating its diffusion and removal, while also facilitating the diffusion and removal of PM_2.5 and inhibiting its formation and accumulation.

3.4. Impact Analysis of Atmospheric Oxidative Capacity

3.4.1. Seasonal Modulation of Atmospheric Oxidation Potential

As shown in Table S8, the atmospheric oxidative capacity in Dongying exhibited significant seasonal variation. In summer, the average O_x value was the highest (167 μg·m⁻³), with O₃ accounting for 90% and showing a very strong positive correlation with O_x (r = 0.99) due to photochemical reactions driving O_x formation. In autumn, O_x (129 μg·m⁻³) was still predominantly influenced by O₃ (83%, r = 0.96). In winter, O_x reached its lowest value (83 μg·m⁻³), with a reduced contribution from O₃ (75%, r = 0.65) and an increased influence from NO₂ (r = 0.96). Spring served as a transitional period (O_x = 153 μg·m⁻³), where NO₂ was the dominant factor (O₃ accounted for only 34%). Although the dominance of O₃ in winter was slightly weaker than in summer and autumn, it remained the primary contributing factor, while the influence of NO₂ was also significant. O_x and PM_2.5 showed a positive correlation across all four seasons. The strongest correlation was observed in summer (r = 0.63), followed by winter (r = 0.35). In summer, O_x mainly consisted of O₃, and the highly oxidative environment promoted the formation of secondary components in PM_2.5 [75,84,85]. In winter, the contribution of NO₂ to O_x increased, but the photochemical reactions were weaker. The contribution of O₃ to O_x was less than that in summer. High NO₂ concentrations promoted the formation of nitrates through chemical reactions [86], which in turn enhanced the formation of PM_2.5, thereby strengthening the correlation between PM_2.5 and O_x. In spring, starting from April, polluted days with ozone as the primary pollutant began to occur in Dongying, and local photochemical reactions generated O₃, leading to a simultaneous increase in both O_x and PM_2.5. In autumn, NO_x participated in photochemical reactions to generate O₃ and also directly contributed nitrates to PM_2.5 [87].

PM_2.5 and O₃ exhibited bidirectional interactions: atmospheric oxidation promoted PM_2.5 formation, while PM_2.5 accumulation attenuated O₃ generation. During the summer, O_x concentrations typically exceeded 125 μg·m⁻³, with O₃ levels ranging from 170 to 300 μg·m⁻³ and PM_2.5 concentrations ranging from 35 to 75 μg·m⁻³. During summer, daily evaluated O₃ values showed a significant positive correlation with daily evaluated PM_2.5 values, with MDA8 O₃ increases of 1 μg·m⁻³ corresponding to PM_2.5 increases of 1.99 μg·m⁻³—a ratio significantly stronger than other seasons (Figure 14). High temperatures and intense radiation favored O₃ formation and accumulation, increasing atmospheric oxidation that accelerated the rapid oxidation of gaseous pollutants (NO_x, SO₂, VOCs, etc.) into secondary particulate matter and promoted PM_2.5 generation. Consequently, PM_2.5 and O₃ exhibited a strong positive correlation. In winter, O_x concentrations ranged from 50 to 100 μg·m⁻³, and a negative correlation was observed between PM_2.5 and O₃ (ΔPM_2.5/ΔO₃ = −0.19) (Figure 14). O_x had little impact on this correlation, which was mainly due to increased primary emissions [17], suppressed photochemistry [18,88], and heterogeneous reactions [87]. In spring and autumn, O_x concentrations ranged from 75 to 150 μg·m⁻³, with a weak positive correlation observed between PM_2.5 and O₃ (Figure 14). NO_x and SO₂ formed nitrates and sulfates through heterogeneous reactions [89,90], while local photochemical reactions produced O₃. Overall, higher O_x concentrations strengthened the positive correlation between PM_2.5 and O₃, while lower concentrations weakened or reversed it. In high O_x environments, photochemistry generally dominated, leading to the co-generation of O₃ and secondary components of PM_2.5, resulting in a significant positive correlation. In low O_x environments, primary emissions prevailed, photochemical reactions were suppressed, and the NO titration effect was enhanced. These factors collectively reduced the correlation between O₃ and PM_2.5 or even caused a negative correlation.

3.4.2. Diurnal Variation in Oxidizing Capacity

The O_x concentrations in Dongying exhibited a clear diurnal pattern, with higher levels during the day and lower levels at night. Summer showed the most significant variation, while winter showed the least (Figure S9). In spring, O_x concentrations were lowest between 5:00 and 7:00 (approximately 90 μg·m⁻³), then gradually increased, peaking between 16:00 and 18:00 (around 140 μg·m⁻³). During summer, O_x concentrations were relatively low between 5:00 and 7:00 (around 80 μg·m⁻³), reaching their highest values between 14:00 and 16:00, exceeding 160 μg·m⁻³. In autumn, the O_x concentration was lower between 5:00 and 7:00 (approximately 80 μg·m⁻³), with a peak occurring between 14:00 and 16:00 (around 120 μg·m⁻³). In winter, concentrations were lowest between 5:00 and 7:00 (about 70 μg·m⁻³), reaching a peak between 16:00 and 18:00 (around 90 μg·m⁻³).

O_x-PM_2.5 correlations were significantly negative during both diurnal and nocturnal periods in spring and autumn. Summer correlations were weaker, whereas winter showed moderate negative relationships (Figure 15a,b). Spring daytime O_x-PM_2.5 correlations (r = −0.59) were slightly stronger than nighttime (r = −0.56), with weak diurnal-nocturnal covariation (r = 0.05). Summer daytime correlations (r = −0.55) were stronger than nighttime (r = −0.45). Winter diurnal and nocturnal correlations were both −0.40, indicating moderate negative trends.

3.5. Concentration-Dependent Effects of PM_2.5 and O₃

The PM_2.5 daily mean concentration was divided into three categories: PM_2.5 ≤ 35 μg·m⁻³, 35 μg·m⁻³ < PM_2.5 ≤ 75 μg·m⁻³, and PM₂.₅ > 75 μg·m⁻³. When PM_2.5 ≤ 35 μg·m⁻³, PM_2.5 and O₃ showed a positive correlation in spring, summer, and autumn, with the strongest correlation in summer (r = 0.57), while in winter, the correlation was negative. When PM_2.5 concentrations ranged between 35 and 75 μg·m⁻³, the correlation between PM_2.5 and O₃ was positive in spring, summer, and winter, and weakly negative in autumn. When PM_2.5 > 75 μg·m⁻³, the correlation between PM_2.5 and O₃ in summer was positive (r = 0.33), although the results were uncertain due to limited data, while in winter it showed a negative correlation (r = −0.20), and in spring and autumn, the correlation between PM_2.5 and O₃ was poor (Figure 8 and Figure 16a,b,e,f). Aerosols at appropriate concentrations (PM_2.5 ≤ 35 μg·m⁻³) promoted O₃ formation by enhancing solar radiation through scattering effects, thereby increasing radiation flux and intensity within the boundary layer. A moderate increase in aerosol concentration also enhanced the photolysis rate of NO_x, further promoting O₃ production [68].

The O₃ daily evaluation values were divided into three intervals: O₃ ≤ 100 μg·m⁻³, 100 μg·m⁻³ < O₃ ≤ 160 μg·m⁻³, and O₃ > 160 μg·m⁻³. When O₃ ≤ 100 μg·m⁻³, summer and winter showed good PM_2.5–O₃ correlations. In summer, PM_2.5 and O₃ showed a positive correlation (r = 0.31), in winter, they showed a negative correlation (r = −0.32), with spring and autumn exhibiting weaker correlations. When 100 μg·m⁻³ < O₃ ≤ 160 μg·m⁻³, all four seasons exhibited synergistic PM_2.5–O₃ effects, with winter showing the strongest correlation (r = 0.40) followed by spring and autumn (both r = 0.24). When O₃ > 160 μg·m⁻³, summer displayed a positive correlation between PM_2.5 and O₃ (r = 0.37), while in spring and autumn, the correlation was weakly negative (Figure 8 and Figure 16c,d,g,h).

3.6. Chemical Speciation Effects of PM_2.5

In the spring and summer, elevated O₃ concentrations (>160 μg·m⁻³) significantly enhanced atmospheric oxidizing capacity, driving the formation of secondary PM_2.5. Its concentration responded positively to O₃ levels, with the secondary PM_2.5 fraction increasing by 12.7% for every 40 μg·m⁻³ increase in MAD8 O₃ during the summer (Figure 17a,b). In the autumn, a decrease in atmospheric oxidizing capacity led to a sharp reduction in secondary PM_2.5 contribution, with the high O₃ group in autumn accounting for only 12.9% (Figure 17c). During winter, primary PM_2.5 dominated across all O₃ concentrations. As O₃ concentrations rose in spring and summer, atmospheric oxidizing capacity increased, enhancing photochemical reactions that promoted secondary aerosol formation.

The four typical episodes showed a weak positive correlation between PM_2.5 and O₃ in spring (r = 0.01), a significant positive correlation in summer (r = 0.42), a strong positive correlation in autumn (r = 0.99), and a significant negative correlation in winter (r = −0.57) (Figure 18a).

In spring, the correlation between PM_2.5 and O₃ was the weakest, with O₃ showing a positive correlation with NH₄⁺, OC, and EC, while a negative correlation was observed with NO₃⁻ and SOC (Figure 18b,c, Table S9). This might be due to the dust input from Dongying, which carried SO₄²⁻ and NH₄⁺, along with the early-phase photochemical reaction where NO_x consumption inhibited NO₃⁻ formation [91,92] (Table S9). The differentiation of secondary components ultimately led to a weak correlation between PM_2.5 and O₃ in spring. In summer, the significant positive correlation between PM_2.5 and O₃ was due to the synergistic effects of shared precursors, with SO₄²⁻, OC, SOC, and O₃ all showing positive correlations (Figure 18b,c, Table S9). Under high temperature and strong sunlight, secondary components like SO₄²⁻ and SOC shared precursors, such as SO₂, VOCs, and NO_x, with O₃, leading to coordinated growth. Additionally, NH₄NO₃ was more volatile and decomposed under high temperatures, with some NO_x preferentially participating in O₃ formation pathways, resulting in a negative correlation between NO₃⁻, NH₄⁺, and O₃. Photochemical reactions simultaneously drove both O₃ generation and the increase of PM_2.5 secondary components, strengthening their positive correlation.

In autumn, PM_2.5 and O₃ were strongly positively correlated, mainly due to the dominant contribution of secondary transformation to PM_2.5. Secondary components (SO₄²⁻, NO₃⁻, NH₄⁺, SNA, and SOC) showed a strong positive correlation with O₃ (r > 0.70) (Figure 18b,c, Table S9). The formation of secondary inorganic components, like SNA, in autumn depended on precursors such as NO_x and SO₂ [86]. The strong positive correlation between these components and O₃ was driven by high radiation, high temperatures, and unstable atmospheric conditions, which synergistically promoted VOCs and NO_x photochemical reactions. HO_x radicals enhanced atmospheric oxidation [93], leading to O₃ generation and the oxidation of SNA and SOC, causing a coordinated increase in PM_2.5 and O₃. As a result, autumn was prone to PM_2.5 and O₃ complex pollution. In winter, PM_2.5 and O₃ exhibited a significant negative correlation, with secondary components negatively correlated with O₃, while PM_2.5-sec overall had a positive correlation with O₃ (Figure 18b–d, Table S9). Winter PM_2.5 concentrations were primarily driven by primary emissions, with coal burning and biomass combustion releasing large amounts of primary OC and EC [87], which suppressed photochemical reactions (due to low temperatures and low radiation), leading to reduced O₃ formation. Low temperatures inhibited the volatility of ammonium nitrate and SOC, while stable meteorological conditions promoted heterogeneous reactions (such as N₂O₅ hydrolysis generating NO₃⁻). Only a small amount of secondary components (such as PM_2.5-sec) accumulated in stable weather, showing a weak positive correlation with O₃. Other studies have similarly found that, during winter haze in northeastern regions, OC and SNA contributed significantly to PM_2.5, with their combined contribution accounting for more than 50% of the PM_2.5 [88,92].

Overall, in spring, the input of dust (SO₄²⁻/NH₄⁺) and the consumption of NO_x in the initial stage of photochemical reactions led to the differentiation of secondary components, weakening PM_2.5–O₃ correlation. In summer, common precursors (VOCs/NO_x), boosted by high temperatures and sunlight, jointly promoted O₃ and secondary components formation, strengthening PM_2.5 and O₃ positive correlation. In autumn, secondary transformation of SNA/SOC under high oxidation conditions maximized the positive correlation, increasing compound pollution risk. In winter, primary emissions of OC and EC dominated PM_2.5, low temperature inhibited photochemistry resulting in decreased O₃, which led to a significant negative correlation between PM_2.5 and O₃.

3.7. Influence of Precursor Concentration Gradients

Between 2017 and 2023, O₃ in Dongying exhibited a significant negative correlation with other pollutants, while PM_2.5 showed a positive correlation with all pollutants except for O₃ (Figure S10a). The daily variations in pollutant concentrations were distinct (Figure S10b–f). At night, the PM_2.5–O₃ correlation was low. During the day, with increased solar radiation and human activities, O₃ concentrations rose, accelerating the oxidation of gaseous pollutants and particulate matter formation, leading to an increase in PM₂.₅ concentrations and strengthening the PM_2.5–O₃ correlation, particularly in the afternoon [28]. NO₂, PM₁₀, PM_2.5, and SO₂ exhibited higher concentrations during the morning and evening peak hours, which were associated with traffic emissions and the influence of temperature inversions [90]. O₃ concentrations showed a clear upward trend during the night, which may be related to the increased atmospheric oxidation capacity [94].

When PM_2.5 levels were ≤35 μg·m⁻³, PM_2.5 and O₃ exhibited a positive correlation in spring, summer, and autumn, with the strongest correlation observed in the summer. At this time, PM_2.5 and O₃ were positively correlated with NO₂, and weakly correlated with VOCs, indicating that, when PM_2.5 ≤ 35 μg·m⁻³, the PM_2.5–O₃ correlation was significantly influenced by NO₂. This finding was similar to the case when O₃ > 160 μg·m⁻³ in summer (Figure 19). In the summer, high temperatures and strong radiation accelerated the photolysis of NO₂, promoting O₃ formation. At the same time, the conversion of NO_x led to an increase in PM₂.₅ concentrations (NO_x + OH → HNO₃ → reacts with NH₃ to form NH₄NO₃) [95]. The low aerosol concentration weakened the atmospheric extinction effect, enhancing surface ultraviolet radiation and forming a positive feedback loop between NO₂ photolysis and O₃ generation. If the city was located in a VOC-sensitive area, high temperatures promoted the release of VOCs, but were insufficient to consume the excess NO_x. In this case, an increase in NO_x concentration directly enhanced the rate of O₃ and nitrate formation, leading to a simultaneous rise in both PM₂.₅ and O₃ levels [96]. Under conditions of high O₃ concentrations (O₃ > 160 μg·m⁻³), the photochemical reactions were already intense. NO₂ continued to supply O₃ through photolysis, enhancing atmospheric oxidative capacity and further promoting the secondary formation of PM₂.₅. Therefore, under conditions of excess NO_x, the PM_2.5–O₃ correlation mainly depended on NO₂ concentrations. In winter, PM_2.5 and O₃ exhibited a negative correlation. At this time, PM_2.5 was positively correlated with NO₂ and VOCs, while O₃ was negatively correlated with NO₂. This suggested that the PM_2.5–O₃ correlation was influenced by common precursors, NO₂ and VOCs, and this result was similar to the case when PM_2.5 > 75 μg·m⁻³ in winter (Figure 19).

When PM_2.5 levels exceeded 75 μg·m⁻³, PM_2.5 and O₃ showed a negative correlation in winter. During this period, PM_2.5 showed a positive correlation with NO₂ and VOCs, while O₃ was negatively correlated with NO₂ and VOCs. This indicated that, when PM_2.5 exceeded 75 μg·m⁻³, the relationship between PM_2.5 and O₃ was primarily influenced by NO₂ and VOCs. This result was consistent with the findings in spring, autumn, and winter when O₃ levels were ≤ 100 μg·m⁻³ (Figure 19). Heterogeneous chemistry reduced ozone concentrations by absorbing NO_x and hydroxyl radicals [97]. Under conditions of low temperature, high humidity, and excessive NO_x, VOCs reacted with NO_x free radicals to form low-volatility organic compounds (e.g., RO₂ → RONO₂) [98,99]. These reaction pathways were generally more conducive to SOA formation than reactions with O₃ [98,100], which led to an increase in PM_2.5 levels and consumed O₃ precursors. It was important to note that emission reduction measures had a dual effect. Shao et al. [97] found that, from 2006 to 2016, the decrease in PM_2.5 levels in Beijing resulted in a 37% increase in ozone formation. Li et al. [101] found that emission reductions alleviated the suppressive effect of aerosol-radiation interactions on meteorological fields and photolysis rates. Model simulations showed that, in the absence of aerosol feedback effects, emission reductions alone would lead to a decrease of 22.89 μg·m⁻³ in PM_2.5 levels and an increase of 5.43 ppb in O₃ levels.

During a typical autumn pollution event, PM_2.5 and O₃ showed a strong positive correlation (r = 0.99) with the concentrations of related pollutants, with highly synchronized fluctuations and significant overlap in peak-to-valley periods (Figure 20). Between 25 and 28 October, both reached their peaks at around 9:00 a.m. and dropped to their lowest values at around 4:00 p.m. On October 29, two peaks were observed (at 11:00 a.m. and 8:00 p.m.) along with a valley at 3:00 p.m. On 31 October, the concentration peaked at midnight, reached its highest point, and then decreased gradually after forming a secondary peak at 11:00 a.m. Pollution continued to fluctuate thereafter, with another peak observed at 4:00 a.m. on 4 and 5 November, while the concentration dropped to a valley at 4:00 p.m. on 3 and 4 November. The concentration changes in common precursor substances, such as NO₂ and non-methane hydrocarbons (NMHC), were highly synchronized with the concentration changes in PM_2.5 and O₃. The trends in NO₂ and NMHC concentrations closely matched the variations in PM_2.5 and O₃, and correlation analysis revealed a positive relationship between PM_2.5 and O₃ with NO₂ (r_O3-NO2 = 0.38, r_PM2.5-NO2 = 0.39), and between PM_2.5 and O₃ with NMHC (r_O3-NMHC = 0.50, r_PM2.5-NMHC = 0.51). NMHC, as a key component of VOCs, along with VOCs and NO₂, were crucial precursors in photochemical reactions. These substances not only contributed to the formation of O₃ through photochemistry but also reacted with substances like NH₃ to form SNA and SOC, which further elevated PM_2.5 concentrations [89,102]. Therefore, the increase in NO_x and NMHC concentrations during this pollution event may have been a key factor contributing to the high correlation between PM_2.5 and O₃.

3.8. Random Forest-Based Importance Analysis of PM_2.5 and O₃ Drivers

Using the random forest algorithm in machine learning, it was found that O_x, T, and NO₂ had significant effects on O₃ concentration, with RH also being important in summer (Figure 21a). In spring, O_x had the most significant impact on O₃ concentration, accounting for 51%, followed by T at 16% and NO₂ at 14%. In summer, O_x had the highest importance (57%), with RH (17%) and T also being significant (15%). In autumn, O_x held the highest importance (43%), followed by T at 19%, and NO₂ also played a significant role (17%). In winter, NO₂ had the highest importance (37%), followed by O_x (24%). CO (8%), PM_2.5 (6%), T (6%), and RH (6%) were also important in winter. PM₁₀ had the greatest impact on PM_2.5 concentrations across all seasons (Spring: 51%, Summer: 46%, Autumn: 43%, Winter: 44%). Following PM₁₀, CO played a significant role (Spring: 17%, Summer: 27%, Autumn: 28%, Winter: 26%). NO₂ also significantly affected PM_2.5 concentrations (Spring: 8%, Summer: 6%, Autumn: 9%, Winter: 10%). The impacts of SO₂, RH, and O_x should also be considered (Figure 21b).

The study’s mechanism diagram appeared in Figure 22. In summer, PM_2.5 and O₃ showed positive correlations driven by O_x, RH, and temperature through photochemical pathways. Under hot-humid conditions (T > 20 °C, RH > 60%) with O_x > 125 μg·m⁻³, the PM_2.5–O₃ correlation reached maximal strength. Conversely, in winter, PM_2.5 and O₃ showed negative correlations due to primary emissions and NO₂ titration, maintaining negative relationships across all PM_2.5 levels, while PM_2.5-NO₂ correlations remained positive and O₃-NO₂ correlations negative. Spring and autumn represented transitional states. In spring, PM_2.5 and O₃ weak positive correlations (r = 0.18) stemmed from dust inputs and intermittent photochemistry, particularly when T > 15 °C and RH < 30%, which enhanced the PM_2.5–O₃ correlation. In autumn, due to O_x decay and lower temperatures, NO₂ drove a positive correlations with PM_2.5 and O₃ at low PM_2.5 levels (≤35 μg·m⁻³).

4. Conclusions

(1) From 2017 to 2023, Dongying’s annual evaluation values of PM_2.5 decreased by 36% (from 55 to 35 μg·m⁻³), maintaining a spatial pattern of higher south-eastern and lower north-western concentrations. In contrast, the annual evaluation values of O₃ fluctuated upward (from 176 to 186 μg·m⁻³), with higher north-central-eastern and lower western concentrations. During the PM_2.5 pollution season (December–February), daytime and nighttime concentrations decreased by 23 and 24 μg·m⁻³, respectively, from 2017 levels while, in the O₃ pollution season (April–September), diurnal concentrations rose, with a 17 μg·m⁻³ nighttime increase surpassing the 11 μg·m⁻³ daytime rise. Between 2017 and 2023, there were 23 days of PM_2.5 and O₃ complex pollution, primarily in March, April, and October.

(2) From 2017–2023, the PM_2.5–O₃ correlation reached its maximum positive strength in summer (r = 0.60), showed negativity in winter (r = −0.19), and demonstrated weak positivity in spring (r = 0.18) and autumn (r = 0.13). Spatially, PM_2.5 and O₃ exhibited strong positive correlations in western areas, contrasting with weak negative trends in northern and eastern regions. In summer, the strongest positive correlation between PM_2.5 and O₃ occurred when PM_2.5 ≤ 35 μg·m⁻³ and 160 μg·m⁻³ < O₃ ≤ 200 μg·m⁻³, but high PM₂.₅ levels (>60 μg·m⁻³) weakened the O₃ formation rate. In winter, the strongest negative correlation emerged at 115 μg·m⁻³ < PM_2.5 ≤ 200 μg·m⁻³ and O₃ ≤ 100 μg·m⁻³. In spring, significant positive correlation was found at PM_2.5 ≤ 35 μg·m⁻³ and 100 μg·m⁻³ < O₃ ≤ 160 μg·m⁻³ but weakened as their concentrations rose. In autumn, a significant positive correlation was observed when PM_2.5 ≤ 35 μg·m⁻³ and 200 μg·m⁻³ < O₃.

(3) The mean pressure field enhanced the positive correlation of PM_2.5 and O₃, while the high-pressure field weakened it. Yet in summer, the high-pressure field had a more pronounced enhancement effect on their positive correlation. The positive correlation between PM_2.5 and O₃ emerged at temperatures ≥ 20 °C, strongest between 20 and 25 °C, while temperatures < 5 °C induced negative correlations. At high humidity (>60%), PM_2.5 and O₃ had a negative correlation, while at low humidity (<30%), PM_2.5 and O₃ showed a positive correlation. Average daily wind speeds of 1.73 m/s facilitated concurrent PM_2.5–O₃ increases. In summer, due to stronger atmospheric oxidation (O_x > 125 μg·m⁻³), PM_2.5 and O₃ exhibited the strongest positive correlation (r = 0.60); in winter, with weaker atmospheric oxidation (O_x < 100 μg·m⁻³), PM_2.5 and O₃ showed a negative correlation (r = −0.19). During sea–land breeze periods, the negative correlation between PM_2.5 and O₃ was generally weaker than during non-sea–land breeze periods, with the largest reduction in spring.

(4) When PM_2.5 ≤ 35 μg·m⁻³ and O₃ > 160 μg·m⁻³, PM_2.5 and O₃ showed a strong positive correlation in summer, with NO₂ affecting their relationship, and in spring and autumn, it was a weak positive correlation, influenced by common NO₂ and VOCs. When PM_2.5 > 75 μg·m⁻³ and O₃ ≤ 100 μg·m⁻³, PM_2.5 and O₃ had a significant negative correlation in winter, influenced by common NO₂ and VOCs. The correlation coefficients of the four typical pollution episodes in spring, summer, autumn and winter were 0.01, 0.42, 0.99 and −0.57 respectively. The positive correlation in summer was driven by synchronous increases in O₃ and secondary components via shared precursors (SO₂, VOCs, NO_x). In autumn, strong positivity resulted from secondary component–O₃ correlations (r > 0.7) and dominance of secondary formation in PM_2.5; in winter, the negative correlation was a result of primary emissions inhibiting photochemical reactions. y. The random forest results indicated that in summer, O_x, RH, and temperature drove a positive correlation between PM_2.5 and O₃ through photochemical reactions. In winter, primary emissions and NO₂ titration led to a negative correlation between PM_2.5 and O₃.

(5) Dongying should focus on PM_2.5–O₃ complex pollution in spring and autumn by strengthening collaborative emission reductions in western regions. In summer, under high-temperature and low humidity conditions with low PM_2.5 (≤35 μg·m⁻³) but excessive O₃, NO_x emissions should be prioritized for reduction to prevent moderate-severe O₃ pollution and high hourly PM_2.5 values. In autumn, when PM_2.5 > 35 μg·m⁻³ and O₃ > 160 μg·m⁻³, precise VOCs-NO_x controls should be implemented, focusing on highly reactive VOCs, like alkenes and aromatic hydrocarbons, to avoid PM_2.5–O₃ complex pollution. In winter, during cold and humid periods with high PM_2.5 (>75 μg·m⁻³), coordinated reduction of VOCs and NO_x should be enhanced to avoid moderate-severe PM_2.5 pollution.

This study’s innovation lies in overcoming the limitations of single-factor analysis by exploring the causes of the PM_2.5–O₃ correlation from meteorological, emission (common precursor), and chemical multi-factor perspectives. It also discusses the impact of sea–land breeze circulation on this correlation. Based on PM_2.5 and O₃ pollution intensity levels, the study proposes seasonal control measures for PM_2.5–O₃ complex pollution. However, this study had several limitations. (1) The analysis of the causes of the PM_2.5–O₃ correlation was based solely on observational data and statistical methods, with conclusions regarding the mechanisms of this correlation derived primarily from literature, without quantitative analysis using Observation-Based Model or Three-Dimensional Air Quality Model. (2) The correlation between PM_2.5 and O₃ at different concentration levels across seasons was not fully quantified. (3) The random forest algorithm struggled to capture the non-linear relationship between PM_2.5 and O₃ concentrations and could not explain the physical or chemical mechanisms behind the joint influence of multiple factors on their concentration variations. For future research, three improvements are planned: first, refining the concentration gradients of PM_2.5 and O₃ based on seasonal characteristics; second, using model simulations to examine the contributions of meteorological conditions, atmospheric oxidizing capacity, and precursor substances to the concentrations of PM_2.5 and O₃, and further analyzing the quantitative impacts of these factors on their correlation; third, combine random forest with other algorithms to better capture complex non-linear relationships. On this basis, precise emission reduction strategies will be proposed for PM_2.5 and ozone complex pollution in petrochemical cities in the Bay region, with a focus on zoning, seasonality, and source-based approaches.