Evidence for Coordinated Control of PM_2.5 and O₃: Long-Term Observational Study in a Typical City of Central Plains Urban Agglomeration

Abstract

Fine particulate matter (PM_2.5) and Ozone (O₃) pollution have emerged as the primary environmental challenges in China in recent years. Following the implementation of the Air Pollution Prevention and Control Action Plan, a substantial decline in PM_2.5 concentrations was observed, while O₃ concentrations exhibited an increasing trend across the country. Here, we investigated the long-term trend of O₃ from 2015 to 2022 in Xinxiang City, a typical city within the Central Plains urban agglomeration. Our findings indicate that the hourly average O₃ increased by 3.41 μg m⁻³ yr⁻¹, with the trend characterized by two distinct phases (Phase I, 2015–2018; Phase II, 2019–2022). Interestingly, the increasing rate of O₃ concentration in Phase I (7.89 μg m⁻³) was notably higher than that in Phase II (2.89 μg m⁻³). The Random Forest (RF) model was employed to identify the key factors influencing O₃ concentrations during the two phases. The significant dropping of PM_2.5 in Phase I could be responsible for the O₃ increase. In Phase II, the reductions in nitrogen dioxide (NO₂) and unfavorable meteorological conditions were the major drivers of the continued increase in O₃. The Observation-Based Model (OBM) was developed to further explore the role of PM_2.5 in O₃ formation. Our results suggest that PM_2.5 can influence O₃ concentrations and the chemical sensitivity regime through heterogeneous reactions and changes in photolysis rates. In addition, the relatively high concentration of PM_2.5 in Xinxiang City in recent years underscores its significant role in O₃ formation. Future efforts should focus on the joint control of PM_2.5 and O₃ to improve air quality in the Central Plains urban agglomeration.

1. Introduction

Tropospheric Ozone (O₃) is a typical secondary gaseous pollutant and the third most significant greenhouse gas (IPCC, 2021). It has a profound impact on human health, ecosystem stability, and vegetation productivity [1]. In recent years, O₃ pollution has emerged as a major environmental issue in the urban areas of China. Observational data indicate that ground-level O₃ concentrations have been rising nationwide [2]. For instance, Wang et al. reported that the maximum daily 8 h average (MDA8) O₃ level increased by 2.6 μg m⁻³ yr⁻¹ in the warm season (April–September) from 2013 to 2020 [3]. This upward trend in O₃ concentrations was similarly observed in many megacities in China, such as Beijing [4], Shanghai [5], Sichuan Basin [6], and other cities [7,8,9]. However, the long-term trend of O₃ concentrations in the Central Plains urban agglomeration remains relatively deficient at present.

In the troposphere, O₃ is formed through complex radical chain reactions involving the oxidation of volatile organic compounds (VOCs) in the presence of nitrogen oxides (NO_x = NO₂ + NO) under sunlight [10]. The rising trend in O₃ concentration is influenced by a variety of factors, including increased global O₃ background concentrations, the changes in meteorological conditions, and shifts in chemical regime due to various regulations affecting NO_x and VOCs emissions [11,12]. Although meteorological conditions and emission changes have been dominant drivers in recent O₃ increases, their contributions have varied across different periods. Liu et al. [2] revealed that the impact of anthropogenic emissions on the O₃ rise from 2017 to 2020 (1.2 μg m⁻³) was much lower than that during 2013–2017 (5.2 μg m⁻³) in China. In addition, the O₃ concentration can be highly sensitive to the meteorological conditions in the given phases and periods. For instance, the meteorological conditions in May 2020 led to a significant increase of O₃ by 26.8 μg m⁻³ compared to May 2019 in the Sichuan Basin [6]. Factors such as temperature, relative humidity, radiation intensity, wind speed, and wind direction were regarded as the main factors affecting O₃ formation [13,14,15,16,17]. However, the key meteorological factors vary across different regions. According to Weng et al. [18], surface solar radiation is a primary determinant of O₃ fluctuations in the Yangtze River Delta (YRD) and Sichuan Basin, while temperature is identified as the most important meteorological variable in the Beijing–Tianjin–Hebei (BTH) region.

Aerosols exert a complex influence on the O₃ production rate through heterogeneous reactions, alterations in photolysis rates, and modifications to the boundary layer [19]. The “aerosol inhibited” regime in O₃ formation, where heterogeneous reactions on aerosol particles predominantly lead to HO₂ loss, has been identified through chemical transport modeling [20]. The enhancement of HO₂ due to the dropping of aerosols has been recognized as a key driver for the increasing summertime O₃ concentration in the North China Plain from 2013 to 2017 [21,22]. Furthermore, a study by Shao et al. [23] revealed that O₃ formation in Beijing increased by 37% from 2006 to 2016 following a reduction in PM_2.5 levels. Consequently, the reduction in PM_2.5 concentrations could offset the effectiveness of traditional O₃ precursor (VOC and NO_x) control strategies under the “aerosol inhibited” photochemical O₃ regime [3]. Hence, understanding O₃ formation mechanisms and identifying the key factors are crucial for accurately managing O₃ pollution, not only in China but also globally.

Machine learning techniques, such as artificial neural networks, random forest (RF), and the convolutional neural network, have been widely used in atmospheric research [24,25,26,27,28]. Among these methods, RF is employed to account for the nonlinear interactions between different input parameters without assuming any specific relationships [29]. Numerous studies [4,18,27,29,30] have demonstrated the efficacy of the RF model in predicting O₃ levels and identifying primary factors influencing O₃ formation. However, the interpretability of results from the RF model is limited due to its “black box” nature. As a complementary method, the observation-based model (OBM) coupled with the Master Chemical Mechanism (MCM) serves as an effective tool for investigating atmospheric photochemistry mechanisms. The MCM has been widely used to investigate in situ O₃ formation processes and the sources of radicals [31,32,33,34,35,36]. However, OBM-MCM relies heavily on detailed observation data and is limited in its ability to conduct long-term and large-scale O₃ pollution research.

Xinxiang City, located in the northern region of Henan Province, is a rapidly developing city within the Central Plains urban agglomeration. As a member of the “2 + 26” city cluster, which serves as a major air pollution transmission channel in the Beijing–Tianjin–Hebei region, Xinxiang suffered the severe haze pollution. In recent years, the exacerbation of O₃ pollution has emerged as a critical environmental challenge. However, the quantitative relationship between reductions in PM_2.5 concentrations and concurrent increases in O₃ remains unclear. To investigate the relationship between PM_2.5 and O₃, this study proposes a multi-temporal analytical framework integrating RF and OBM.

By integrating long-term continuous monitoring data (2015–2022) with short-term intensive high-density observations, this study aims to quantify long-term key drivers and elucidate the underlying mechanisms in O₃ pollution in Xinxiang City. Firstly, the long-term trend and seasonal variation of O₃ during this period were explored by using hourly observations of O₃ collected from the national monitoring network. Subsequently, the RF model was employed to investigate the factors influencing O₃ levels and assign importance rankings to these factors. Finally, the OBM was utilized for illustrating the mechanism underlying the identified influencing factors in O₃ formation. The results of this work are expected to provide insights beneficial for controlling O₃ pollution in cities within the Central Plains urban agglomeration.

2. Materials and Methods

2.1. Data Sources

Xinxiang City has been equipped with four state-operated air quality automatic monitoring stations since 2015, which are strategically positioned primarily within the urban area (Figure 1). Hourly concentrations of air pollutants (including O₃, NO₂, CO, SO₂, PM_2.5, and PM₁₀) from these four sites were obtained from the China National Environmental Monitoring Centre (http://www.cnemc.cn/, accessed on 16 May 2024), covering the period from 1 January 2015 to 31 December 2022. The pollutants data were normalized based on the change of atmospheric conditions before (273.15 K, 1 atm) and after (298.15 K, 1 atm) September 2018.

The meteorological data from 1 January 2015 to 31 December 2022 were obtained from the ERA5 datasets of the European Centre for Medium-Range Weather Forecasts (ECMWF) (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview, accessed on 19 May 2024). The research area (34°55′–35°50′ E; 113°30′–115°01′ N) covered the entire Xinxiang City. The hourly resolution of significant meteorological variables involving the O₃ formation mechanism with a spatial resolution of 0.25° × 0.25° was utilized in our study, including a 10 m u-component of wind, 10 m v-component of wind, 2 m dewpoint temperature, 2 m temperature, boundary layer height, surface net solar radiation, surface pressure, total cloud cover, and total precipitation. The detailed information of these variables can be found in

Table 1. The main information of the nine meteorological variables.

Abbreviations	Names of Variable	Unit
U10	10 m u-component of wind	m·s−1
V10	10 m v-component of wind	m·s−1
D2m	2 m dewpoint temperature	K
T2m	2 m temperature	K
BLH	Boundary layer height	m
SSR	Surface net solar radiation	J·m−2
SP	Surface pressure	Pa
TCC	Total cloud cover	Dimensionless
TP	Total precipitation	cm

.

The field measurement campaign was also conducted from 1 June to 31 June in 2021. The sampling site was located at the Xinxiang Municipal Party School (35.29° N, 113.93° E), a typical urban area. The gaseous pollutants, including O₃, NO₂, NO, SO₂, CO, and NMVOCs, were measured in our study. The Model 42i, Model 48i, Model 43i, and Model 49i (Thermo Fisher Scientific, Waltham, MA, USA) were used for online measurements of NO_x (NO₂, NO), SO₂, CO, and O₃. The hourly NMVOCs concentrations, including alkanes, alkenes, alkynes, aromatics, and oxygenated compounds were measured by GC-FID/MS (TH-300B, Wuhan Tianhong Environmental Protection Industry Co., Ltd., Wuhan, China).

2.2. Random Forest Model

The RF model is an ensemble learning algorithm with high accuracy and a strong ability to avoid overfitting. Here, the RF model was developed to predict the concentrations of O₃ and identify critical variables in O₃ formation. The performance of RF depends on hyperparameters. Details of all parameters tuned for the RF model are presented in Table S1. The randomForest package for the R software (version 4.2.3) is used for analyses and validation processes in our study.

For the RF model, the in situ observation pollutants concentrations and meteorology factors were selected as input variables. Due to a lack of long-term hourly observation, VOCs were excluded from the input parameters in this study. According to previous studies [37,38,39,40], the variability of surface O₃ was well-explained by the ML algorithm with meteorological information alone, particularly in the VOC-limited regime. Like many other urban areas in China, O₃ production in Xinxiang City is generally in the VOC-limited regime. Therefore, it is reasonable to simulate O₃ using a supervised RF model without considering the VOC concentration.

The datasets were randomly divided into training and testing subsets at a ratio of 7:3. The fivefold cross-validation method was used to evaluate the performance of the RF model [27] (Figure S1). The relative importance of the input variables was ranked by calculated variable importance scores, represented as the aggregated increase in the mean squared errors (%IncMSE). The mean squared errors were calculated by the RF model by randomly assigning values to each input variable. The variables with a higher importance score (%IncMSE) had a more significant impact on O₃ formation.

2.3. Observation-Based Model

OBM incorporated with MCM v3.3.1 was built to investigate the chemical mechanism of how PM_2.5 affects the formation of O₃. The detailed description of the gas-phase chemical processes by the MCM displays that it was involved in methane and 142 non-methane VOCs [41]. To establish a direct relationship between PM_2.5 concentrations and O₃ formation, the OBM considered the heterogeneous reactions and variations in photolysis rates. The aerosol optical depth (AOD) could be calculated by the PM_2.5 concentration [23,42] using Equation (1):

$${\displaystyle \frac{\mathrm{A}\mathrm{O}\mathrm{D}}{\mathrm{H}}}={\mathrm{P}\mathrm{M}}_{2.5}\times \mathrm{K}\times \mathrm{f}\left(\mathrm{R}\mathrm{H}\right)\times {10}^{-6}$$

(1)

where H represents the atmosphere boundary layer height; f(RH) denotes the hygroscopic growth factor, which is determined by relative humidity (RH), and K is the given parameter.

The calculated AOD was used to quantify the hourly photolysis rates of NO₂ (JNO₂) [43,44], thus establishing a direct link between PM_2.5 concentration and photolysis rates (see details in the Supplemental Information). The photolysis rates (J_i) of other species were calculated by the solar zenith angle (SZA) and built-in parameters (L_i, M_i, and N_i) [45]; see Equation (2):

$$J_i={\mathrm{L}}_i\times \mathrm{c}\mathrm{o}\mathrm{s}\left(SZA\right)\times {\mathrm{M}}_i\times \mathrm{e}\mathrm{x}\mathrm{p}\left(–{\mathrm{N}}_i\times \mathrm{s}\mathrm{e}\mathrm{c}\left(SZA\right)\right)$$

(2)

The photolysis rates would be further scaled according to the calculated photolysis rates of NO₂ (JNO₂) based on the PM_2.5 concentration.

The heterogeneous reaction of HO₂ was assumed to be the first order reaction [21,46], and the reaction constant (k) could be calculated by Equation (3):

$$\mathrm{k}={-\left({\displaystyle \frac{\mathrm{r}}{{\mathrm{D}}_{\mathrm{g}}}}+{\displaystyle \frac{4}{{\mathsf{\gamma }\mathrm{H}\mathrm{O}}_2}}\times {v\mathrm{H}\mathrm{O}}_2\right)}^{-1}\times {\mathrm{S}}_{\mathrm{a}\mathrm{e}\mathrm{r}\mathrm{o}}$$

(3)

where r, Dg, and vHO₂ were the surface-weighted particle radius, gas phase diffusion coefficient, and mean molecular speed of HO₂, respectively. The relevant values of these parameters were selected according our previous study [33]. γHO₂ was the uptake coefficient of HO₂ on aerosols, ranging from 0.02 to 0.2. The O₃ concentration under different γHO₂ was tested by OBM (Figure S1). In our study, the maximum γHO₂ value of 0.2 was adopted to magnify the effect by the model according to Shao et al.’s study [23]. S_aero was the aerosol surface concentration, which is calculated by the PM_2.5 concentration (further details are provided in the Supplemental Information).

The observed and calculated data, including pollutant concentrations (CO, SO₂, NO_x (NO, NO₂), and NMVOCs) and meteorological factors (relative humidity, temperature, pressure, and the photolysis rates in related species) were subjected to the model constraints. The time resolution of the input parameters was averaged or interpolated to 1 h.

2.4. Model Evaluation

The mean bias (MB), root mean squared error (RMSE), and index of agreement (IOA) were used to assess the model (RF and OBM) performance based on the observed (O_i) and simulated (S_i) hourly O₃ values according to the following equations:

$$\mathrm{M}\mathrm{B}={\displaystyle \frac{{\sum }_{i=1}^N(S_i-O_i)}{N}}$$

(4)

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^N{(S_i-O_i)}^2}{N}}}$$

(5)

$$\mathrm{I}\mathrm{O}\mathrm{A}=1-{\displaystyle \frac{{\sum }_{i=1}^N{\left(O_i-S_i\right)}^2}{{\sum }_{i=1}^N{\left(\left|O_i-\overline{O}\right|+\left|S_i-\overline{O}\right|\right)}^2}}$$

(6)

where $\overline{O}$ is the mean concentration of the observed O₃.

3. Results and Discussion

3.1. O₃ Pollution Profiles

3.1.1. Long-Term Trend of O₃ and Related Pollutants

The year variations of 1 h O₃ concentrations and related pollutants (NO₂ and PM_2.5) in Xinxiang City are presented in Figure 2. The O₃ concentration exhibited an increasing trend from 2015 to 2022, with an average growth rate of 3.41 μg m⁻³ yr⁻¹. The similar upward trends in O₃ concentrations over the past 1–2 decades have been observed in other Chinese urban areas, such as Beijing [47], Shanghai [48], the Sichuan Basin [6], the Pearl River Delta [49], and various other Chinese urban sites [2,7]. In contrast, the concentrations of PM_2.5 and NO₂ showed significant declines from 2015 to 2022. This reduction is attributed to the stringent implementation of clean air policies in China, including the Air Pollution Prevention and Control Action Plan (2013–2017) and the Three-Year Action Plan for Winning the Blue Sky Defense Battle (2018–2020) [3]. The former plan focused primarily on reducing particulate matter, while the latter emphasized the coordinated control of NO_x and VOCs, with a targeted 10% reduction in VOC emissions [50].

The increasing O₃ trend could be further separated into two phases (Phase I, 2015–2018; Phase II, 2019–2022) based on the different increasing rate. During Phase I, the average 1-hourly O₃ concentration increased at a rate of 7.89 μg m⁻³ yr⁻¹. The average annual concentration of PM_2.5 was at a high level, and had a significant decrease (from 85.54 to 56.70 μg m⁻³). However, no significant changes were observed in NO₂ concentration during Phase I. In contrast, during Phase II, the increase rate of O₃ was 2.76 μg m⁻³ yr⁻¹, which was much smaller than that in Phase I. The concentration of PM_2.5 was also at the high level (approximately 50 μg m⁻³), although it experienced a relatively smaller decrease compared to Phase I. By contrast, the concentration of NO₂ had an obvious decreasing tendency in Phase II.

NO₂ was the important precursor in O₃ formation through the “NO_x cycle”, exhibiting a non-linear relationship with O₃ formation. Under the VOC-limited conditions, which were thought to prevail in urban China, decreasing NO_x would increase O₃, while under NO_x-limited conditions, reducing NO_x could decrease O₃ concentrations [22]. The effect of PM_2.5 on O₃ formation was mainly by changing photolysis rates and heterogeneous chemical processes [23], with its influence heavily dependent on the level of the PM_2.5 concentration. In Xinxiang City, O₃ formation was under VOC-limited regimes alongside a high PM_2.5 concentration. In the condition, reductions in both NO_x and PM_2.5 can lead to increased O₃ production. Hence, the decline in PM_2.5 concentration could be a primary factor driving the rise in O₃ in Phase I. The result was consistent with the results that a reduction of PM_2.5 stimulated O₃ production over the 2013–2017 periods in the North China Plain. The impact of PM_2.5 controls on O₃ formation likely weakened in Phase II due to the relatively minor reduction in PM_2.5 concentrations. The unbalanced changing of the precursor concentration (VOC and NO₂) might be the main reason for O₃ increasing in Phase II.

3.1.2. Seasonal Variation of O₃ Pollution

The seasonal variation of O₃ during 2015–2022 is shown in Figure 3. O₃ concentrations exhibit pronounced seasonal patterns, peaking during the summer, and remaining at relatively lower levels in the winter. The rise in temperature and solar radiation intensity plays a critical role in photochemical formation of O₃ in summer [51]. Enhanced photochemical production and the rapid cycling of RO_x radicals (OH + HO₂ + RO + RO₂) typically overcome the radical and NO titration in summer [52]. Consequently, the potential health hazards associated with O₃ exposure are particularly significant during the warm season. According to the updated WHO Global Air Quality Guidelines (AQGs) from September 2021, the recommended peak season O₃ concentration is lower than 60 μg m⁻³. However, the average concentration of O₃ in summer and spring exceed the recommended threshold from 2015 to 2022. In autumn and winter, a steady increase in O₃ concentration has been observed since 2018, with levels exceeding 60 μg m⁻³ in autumn 2022. The extension of the O₃ pollution season from the warm season is a nationwide phenomenon in China [53]. The rapid rise in O₃ levels outside of the summer season can enhance atmospheric oxidative capacity, potentially leading to the increased formation of secondary PM_2.5, including nitrate, sulfate, and organic components.

3.2. Identifying Key Factors Using RF Models

The RF model was employed to predict O₃ concentrations for both Phase I and Phase II. As shown in Figure S3, during the training phase, the model explained 83% and 86% of the measured O₃ for Phase I and Phase II, respectively. The RMSE was 9.55 and 7.96 μg m⁻³ for Phase I and Phase II, respectively. The performances of the testing dataset in RF model for the two phases are shown in Figure 4. For both phases, the values of MB were minor, and the values of R² and IOA were close to 1. The slope and intercept values were 0.77 and 14.25 for Phase I and 0.79 and 13.49 for Phase II. It is noteworthy that the RF model tended to underestimate and overestimate O₃ concentrations at relatively high and low values, resulting in relatively higher RMSE for both phases. This discrepancy can be attributed to the RF model’s tendency to exhibit larger biases in predicting extreme values due to the absence of certain O₃ precursor data, such as VOC [29,54]. Nevertheless, the RF model could successful reproduce O₃ concentration using the selected factors.

As shown in Figure 4, the top 5 factors for Phase I were NO₂, SSR, PM_2.5, T2m, and V10. The photolysis of NO₂ produced an oxygen atom, and O₃ was then produced from the combination of the oxygen atom and O₂ [1]. Hence, NO₂ and SSR were the notably influential factors in O₃ formation. PM_2.5 was identified as the third most significant factor contributing to O₃ formation in Phase I. The reduction of PM_2.5 during this phase may elevate O₃ concentrations through modulations in atmospheric heterogeneous reaction kinetics, solar radiation-driven photolysis efficiencies, and planetary boundary layer transport dynamics [20,22]. Temperature was also an important factor influencing O₃ formation in Phase I. Chemical kinetics rates involved in O₃ production increased with the increase of temperature [55]. Additionally, the VOC emissions, including biogenic emission rates and anthropogenic emissions (such as solvent evaporation), may be enhanced in hot weather [56,57].

In Phase II, NO₂ and SSR remained prominent factors, indicating the local formation of O₃. Other high-ranking variables were predominantly meteorology-related, including U10, D2m, T2m, and V10. Unlike in Phase I, PM_2.5 was less important due to the relatively smaller change in the concentrations in Phase II. O₃ enhancement due to PM_2.5 dropping significantly depends on the current level of PM_2.5 concentration and its decline magnitude [22,23]. The decreased amplitude and the level of PM_2.5 concentrations were smaller in Phase II, resulting in less importance of PM_2.5 in O₃ formation. Although PM_2.5 ranked seventh in Phase II, its %IncMSE value was close to the high-ranking meteorology-related factors—higher than BLH and CO. In addition, the concentration of PM_2.5 was at a high level (about 51 μg m⁻³ in 2022), exceeding Class I limit values of the National Ambient Air Quality Standard (NAAQS) (35 μg m⁻³). Hence, PM_2.5 remains a significant factor in O₃ formation in recent years for Xinxiang City and also for the other cities with a high PM_2.5 concentration.

3.3. Role of PM_2.5 in O₃ Formation

From 19 June to 25 June 2021, the average O₃ concentration (128.63 μg m⁻³) was in excess of the CNAAQS 1 h mass-based standards of 120 μg m⁻³. The period was identified as being traceable to an O₃ episode. During the episode, a high level of O₃ concentration (up to 257 μg m⁻³) was observed. The concentrations of SO₂, NO, NO₂, and CO were 13.60, 3.64, 32.46 μg m⁻³, and 0.50 mg m⁻³ on average. The PM_2.5 concentration was at a relatively low level, with 25.00 μg m⁻³ being the average. The average mixing ratios of 35 NMVOCs are summarized in Table S2. The other information about the O₃ episode is also introduced in the Supplementary Materials.

The identified O₃ episode (19 June to 25 June 2021) was used by OBM for a simulation study. The comparison of observed and simulated O₃ during the identified O₃ episode is shown in Figure S4. The model accurately captured the diurnal profile of O₃, demonstrating satisfactory performance. The average concentration of observed and simulated O₃ during the episode was 128.09 and 128.63 μg m⁻³, respectively, with a high R² value of 0.96. In addition, the MB, RMSE, and IOA were 0.82 μg m⁻³, 8.08 μg m⁻³, and 0.97, respectively, further validating the model’s capability to reproduce the variations of O₃ effectively and enabling its use for subsequent analysis.

As mentioned in Section 2.3, the heterogeneous reactions and changing of photolysis rates linked to PM_2.5 were incorporated into the model. Hence, we conducted experiments to assess how variations in PM_2.5 concentration affected O₃ levels. During the episode, the concentration of PM_2.5 was relatively low, with about 25 μg m⁻³ on average. To illustrate concrete situations of pollution, O₃ concentration was simulated by OBM with a series of PM_2.5 concentrations (0–3 times PM_2.5 concentration). The diurnal profile of O₃ concentration under different PM_2.5 concentrations is shown in Figure S5. The O₃ concentration rose with the dropping of the PM_2.5 concentration. The maximum disparity in O₃ concentration under different PM_2.5 concentrations reached up to 32.46 μg m⁻³ (Figure S6), indicating the significant impact of PM_2.5 on O₃ formation. In addition, the rangeabilities of O₃ under difference PM_2.5 concentrations was higher in the daytime and lower at nighttime. The reduction of HO₂ by heterogeneous loss in PM_2.5 was the major mechanism at nighttime. The decrease in PM_2.5 could lead to an increasing HO₂ concentration due to less HO₂ heterogeneous loss on the ambient aerosol [22]. The NO titration effect on O₃ could be offset by an elevated HO₂ concentration [11]. During the daytime, enhanced photolysis rates resulting from decreased PM_2.5 concentration further facilitated O₃ formation. Both mechanisms played significant roles in O₃ formation during the daytime.

The Empirical Kinetic Modeling Approach (EKMA) diagram can categorize O₃ formation into either “NO_x limited” or “VOC limited” regime [20], providing a basis for effective O₃ pollution control policies. To investigate the impact of PM_2.5 on O₃ pollution control strategies, EKMA curves were constructed by OBM under both 0.0 × PM_2.5 and 3.0 × PM_2.5 scenarios (Figure 5). The EKMA curve was changed under different concentrations of PM_2.5, with the slope of the ridgeline (VOC/NO_x) increasing from 8.36 under 0.0 × PM_2.5 scenarios to 11.48 under 3.0 × PM_2.5 scenarios. The result meant that the O₃ formation regime tended to “NO_x limited” with the dropping of PM_2.5 concentration. PM_2.5 has a great impact on the O₃ sensitivity regime, thereby affecting the production rate of surface O₃. The aerosol chemistry and photochemistry were the main mechanism for the shift of O₃ chemical regimes under different PM_2.5 concentrations [58]. As previously discussed, the concentration of the HO₂ concentration increases as the level of PM_2.5 declines, accelerating the RO_x cycle (OH→RO→RO₂→HO₂→OH) with peroxyl-radical self-reactions predominating under these conditions. Therefore, when formulating policies for VOC and NO_x emission reductions to control O₃ pollution, it is crucial to pay more attention to changes in PM_2.5 concentration [44].

3.4. Limitations

In the present study, a comprehensive dataset from field measurements was employed as a model constraint. The dataset included key parameters, such as the concentrations of reactive species, mixing layer height, and photolysis frequencies. Additionally, the state-of-the-art gas chemistry mechanism (MCM) was used in the OBM. The uncertainties of the OBM were mainly determined by the complexities of atmospheric “Haze Chemistry” [59]. Multiple heterogeneous reactions coexisted on the aerosol surfaces [60,61]. Therefore, some heterogeneous reaction might have an impact on O₃ production, such as the heterogeneous formation of HONO and HNO₃ [62] and heterogeneous loss of O₃ [63]. However, the heterogeneous reactions mechanism was unrevealed, with a big range of heterogeneous uptake coefficients [11,22]. Hence, only the heterogeneous reaction of HO₂ on aerosols surfaces, the paramount heterogeneous reaction impacting O₃ formation, was considered in our study.

4. Conclusions

Clean air actions have been implemented by the Chinese government to improve the severe air pollution issue since 2013. However, the increasing trend of O₃ has been inconsistent with the decline of PM_2.5 in China. In Xinxiang City, the O₃ concentration increased by the rate of 3.41 μg m⁻³ yr⁻¹ from 2015 to 2022. This increase can be divided into two phases: Phase I (2015–2018) saw a high rate of increase (7.89 μg m⁻³), while Phase II (2019–2022) experienced a lower rate (2.89 μg m⁻³). The O₃ pollution from warm seasons should be paid more attention, due to the steady increasing O₃ concentration in autumn and winter since 2018. The developed RF model effectively simulated O₃ concentrations, identifying NO₂ and surface net solar radiation as primary factors in O₃ formation for both phases. In Phase I, PM_2.5 ranked third in O₃ formation, while in Phase II, PM_2.5 remained a significant factor due to its persistently high concentration in Xinxiang City. The OBM incorporated into MCM was used to explore how PM_2.5 influences O₃ formation. The O₃ concentration was raised with the dropping of PM_2.5 by the process of the heterogeneous reaction and photolysis rates. The O₃ formation regime tended to “NO_x limited” with the dropping of the PM_2.5 concentration. Neglecting the role of PM_2.5 in O₃ formation could have adverse effects on O₃ pollution control policies. Further research into heterogeneous uptake coefficients would be beneficial in reducing the uncertainties associated with heterogeneous reactions in real atmospheric aerosols. Our results provide powerful evidence for on-going coordinated control of O₃ and PM_2.5 in a typical city of the Central Plains urban agglomeration.