Drivers of a Summertime Combined High Air Pollution Event of Ozone and PM_2.5 in Taiyuan, China

Abstract

Combined air pollution of ozone and PM_2.5 often occurs in coal-based cities of China such as Taiyuan City. In this study, the Weather Research and Forecasting/Chemistry (WRF-Chem) model was employed to simulate a combined high air pollution event of ozone and PM_2.5 in Taiyuan City from 20 May to 29 May 2015,with the maximum daily 8-hour average (MDA8) for ozone exceeding 140 ppbv and PM_2.5 concentrations surpassing $200\mu \mathrm{g}{\mathrm{m}}^{-3}$. We further investigated the drivers of the combined air pollution in Taiyuan during the polluted period. The simulation results showed that the model can well simulate the combined pollution event in Taiyuan, with assessment parameters within reasonable ranges. Moreover, by analyzing the observational data and simulations, the major factors causing the PM_2.5 pollution in Taiyuan during this time period were suggested to be local emissions and pollutant transport from the North China Plain (NCP) located to the east of Taiyuan. In addition, unfavorable meteorological and geographical conditions in Taiyuan also play important roles in forming severe PM_2.5 pollution. Regarding the ozone pollution in Taiyuan, we suggest that the mechanism dominating the pollution event is that of ozone-rich air being transported to Taiyuan at high altitudes and then mixed downwards, resulting in an increase of the ground-level ozone in Taiyuan. Furthermore, we found local emissions and emissions from Taiyuan Basin and Henan Province, which are located to the south of Taiyuan, contributing significantly to the ozone pollution in Taiyuan City during this time.

1. Introduction

China has been experiencing serious air pollution problems due to massive emissions of polluting gases and aerosol particles [1]. The observed PM_2.5 (particulate matter with aerodynamic diameter ≤2.5 µm) concentration in China increased sharply from the year 2004 to the year 2007. After that, due to the implementation of a series of control strategies by the government, the PM_2.5 level in China has become stable. However, PM_2.5 pollution still occurs frequently in the autumns and winters of China [1]. PM_2.5 pollution in China has obvious diurnal and seasonal variations. The concentrations of PM_2.5 are high in the morning and at night and are low at noon, high in winter and low in summer. PM_2.5 not only affects the atmospheric visibility through light extinction but also harms human health [2]. Modeling studies, such as those reported by Chen et al. [3] and Li et al. [4], suggested that, during haze pollution events, 15–30% of the surface PM_2.5 mass is contributed by secondary organic aerosols (SOAs). However, it is important to note that measurement-based studies indicated higher fractions of SOAs (up to 67%) [5,6], with varying compositions and characteristics of SOAs. Essential sources of SOA precursors include coal combustion, transportation, catering industry emissions, biomass combustion and biogenic emissions. Chemical reactions between SOA precursors and oxidants (such as OH radicals) lead to the formation and growth of SOAs [7,8].

Current research shows that, aside from directly affecting the health of lives on Earth, aerosols can also change the concentration of ozone by providing sites for heterogeneous reactions and altering the rates of photolysis reactions, thereby impacting ozone formation [9,10]. Ma et al. [11] analyzed ground-based observations and satellite data, and they suggested the changes in the aerosol concentration and optical properties as the major factors for the increase of surface ozone.

In China, ozone pollution has become more serious, especially during the summertime (June-August). Since the 1990s, ozone concentration in China has been increasing, especially in rapidly developing city groups such as the Yangtze River Delta, the Pearl River Delta and the NCP. For example, Shanxi, a province near the NCP, is faced with severe ozone pollution in summertime. It was reported that the polluted days in the capital of Shanxi Province (i.e., Taiyuan City) in 2015 and 2018 are 35 days and 100 days, respectively [12,13]. According to the observational data from 74 Chinese cities, the mean value of the daily maximum 8-h average ozone has increased by 3%·year⁻¹ over the period 2013–2017 [9]. Thus, ozone pollution has become another major concern in China. Ozone in the troposphere is formed through combined photochemical reactions between nitrogen oxide (NO_x), carbon monoxide (CO) and volatile organic compounds (VOCs) under solar radiation. Thus, the level of ozone is highly determined by the release of these precursors. Furthermore, ozone concentration is also affected by meteorological conditions [9]. Previous studies have shown that ozone pollution events in China are usually accompanied by high temperature, low wind speed and low relative humidity conditions. The increase of ozone in the environment exerts a significant impact on human health, climate change and atmospheric chemistry [14,15]. In addition, ozone influences the atmospheric oxidation and the formation of SOA [9]. In recent years, ozone and PM_2.5 pollutions frequently occur at the same time. Jia et al. [16] analyzed the observational data of Nanjing, a city in eastern China, and found that, in hot seasons, a high ozone level in a strong oxidizing atmosphere promotes the formation of secondary particles, which may result in a positive correlation between PM_2.5 and ozone. Li et al. [4] also reported that the oxidation state of organic aerosols in the summertime of the NCP is the highest, which may be caused by strong photochemical processes with the participation of ozone.

Atmospheric aerosols and ozone play a crucial role in air quality and atmospheric chemistry. On one hand, the formation of ozone can affect many aspects, such as the formation and growth of aerosols. For example, the concentrations of O₃ and H₂O₂ determine the oxidation of SO₂ to form sulfate aerosol, and the partitioning of HNO₃ between the gas and aerosol phases governs the level of nitrate aerosol. On the other hand, aerosols can impact the ozone formation rate by altering photolysis rates [17].

Taiyuan, the capital of Shanxi Province in North China, is a major coal-based city on the Loess Plateau. It is the largest coking base in China, producing over half of the country’s charcoal [18,19]. The city has significant local emissions from coal usage and various industrial activities, leading to severe air pollution. Moreover, the geographical condition of Taiyuan City is special. As shown in Figure 1, Taiyuan is located in the northern part of the Taiyuan Basin, with an average elevation of approximately 800 m and the highest and lowest elevations of 2670 m and 760 m, respectively. Taiyuan is also adjacent to the Taihang Mountains to the east, Lvliang Mountains to the west, Yunzhong Mountain and Xizhou Mountain to the north and a river valley in the middle and the south. Thus, the terrain in the north of Taiyuan is narrow, while the terrain in the south is wide, which makes Taiyuan a trumpet-shaped terrain, characterized by a central plain surrounded by mountains on three sides, forming an overall shape resembling a trumpet. Due to the influence of Taiyuan’s trumpet-shaped terrain, the annual average wind speed in the region is low (0.3 m/s) and the frequency of calm winds is high (over 30%) [20]. Low planetary boundary layer height and low ventilation indexes (<3000 ${\mathrm{m}}^2/\mathrm{s}$) were also observed due to the special terrain of Taiyuan [21]. This unique topographical feature also makes it difficult for pollutants to disperse effectively. Previous studies [22,23] indicated that good air quality in Taiyuan is often associated with strong northerly winds, whereas southerly winds tend to transport pollutants into Taiyuan, leading to PM_2.5 concentrations above 200 $\mu \mathrm{g}{\mathrm{m}}^{-3}$. In addition, pollution events in Taiyuan can be affected by the East Asia monsoon, which mostly brings southeasterly winds during the summertime. As a result, pollutants from the Beijing–Tianjin–Hebei region can be transported to Taiyuan, causing heavy pollution [24]. In recent years, the concentration of PM_2.5 in Taiyuan has been high, and the surface ozone is also increasing, making it one of the most heavily polluted cities in China [18].

Previous studies have investigated the air pollution in Taiyuan. To name a few, Bai et al. [25] reported that heavily polluting cities in Shanxi Province include Taiyuan, Linfen, Jincheng, Changzhi, Lvliang, Yangquan and Jinzhong, and the heavily polluted areas show negligible variations between the years 2012 and 2015. Li et al. [26] indicated that coal combustion contributes highly to VOCs in the atmosphere of Taiyuan, which are precursors of ozone and secondary aerosols. Results of the positive matrix factorization model (PMF) also showed that the coking process is the largest contributor (32.56%) to the ambient VOCs in Taiyuan, followed by coal and biomass combustion (23.25%) [27]. Through analyzing the meteorological conditions of Taiyuan and using the backward trajectory, Li et al. [26] found that the regional transportation of pollutants related to coal burning and coking from the basin may aggravate the pollution in Taiyuan.

Numerical models have also been used to simulate the air pollution in Taiyuan City. Wu et al. [28] evaluated the contributions of the trans-boundary transport of NCP emissions to the air quality in northwest China during a persistent air pollution episode in May 2015. According to the results of a model that coupled air quality and meteorology, WRF-Chem, they found that the model performs well in capturing the variations of PM_2.5, ozone and NO₂. In addition, they indicated that NCP emissions contribute substantially to ozone and PM_2.5 in Shanxi Province, the downwind area of the NCP. Qi et al. [29] used WRF-Chem to simulate the air quality in 26 cities in northern China over January, April, July and October 2015, and they found that the net fluxes of PM_2.5 in Taiyuan are positive in April, July and October but negative in January. Moreover, they indicated that the overestimation of PM_2.5 in Shanxi Province by the model is due to an improper allocation of rural emissions in emission inventories, and the uncertainties in emission inventories are magnified during model computations. Bai et al. [25] employed the WRF-CAMx model to explore the spatial–temporal characteristics of air pollution in Shanxi. They found that heavily polluted areas concentrate in Taiyuan, and the effect of the emission reduction policy is not obvious there. They also pointed out possible reasons for the discrepancy of model predictions as the biases in the meteorological prediction, limitations of the special terrain and the deficiency of the model mechanism.

Despite existing studies, there is limited focus on the combined pollution of ozone and PM_2.5 pollution in coal-dependent Chinese cities like Taiyuan. Numerical simulations of combined air pollution in such cities are also scarce. This study investigates the drivers of combined ozone and PM_2.5 pollution in Taiyuan during May 2015 using the WRF-Chem model. We aim to identify key factors contributing to this pollution, enhancing understanding of air pollution in coal-based cities with unique terrain like Taiyuan, ultimately aiding in the development of effective control strategies.

In Section 2, we describe the model configurations, the gas-phase mechanism, the observational data and assessment parameters we used. In Section 3, we evaluate the model performance in capturing meteorological parameters and pollutant concentrations. Furthermore, this study analyzes the drivers of the pollution and discovers the meteorological and geographical conditions that are essential to the accumulation of pollutants in Taiyuan. Major conclusions and future work prospects are summarized in Section 4.

2. Methodology

We studied the combined pollution event of ozone and PM_2.5 in Taiyuan City and found that its peak period is from May to June. To explore the drivers of summer atmospheric combined pollution event in Taiyuan, this study selected a typical atmospheric combined pollution event in Taiyuan from 20 May to 30 May 2015 as the studied period, with the MDA8 for ozone exceeding 140 ppbv and PM_2.5 concentrations surpassing $200\mu \mathrm{g}{\mathrm{m}}^{-3}$. This event can provide a reference for the study of summertime combined pollution events in Taiyuan City. We employed the Weather Research and Forecasting/Chemistry (WRF-Chem) [30] model to simulate the pollution event. Then, the simulation results, including pollutant concentrations and meteorological parameters, were compared with the observational data, and the results were evaluated using assessment parameters. In addition, we also output the horizontal distributions of pollutants and meteorological parameters to explore the drivers of this combined pollution event.

2.1. Measurements

In this study, to evaluate the model performance, we obtained the meteorological data of Taiyuan (37.73° N, 112.56° E) from the National Climatic Data Center (https://www.ncdc.noaa.gov/, accessed on 2 March 2025). The meteorological data we used include temperature, wind speed and wind direction. To validate the simulated wind, we also decomposed the wind speed into an east–west component (u) and a north–south component (v).

In addition, concentrations of pollutants such as PM_2.5 and ozone were provided by the National Urban Air Quality Real-time Release Platform of China Environmental Monitoring Station (http://www.cnemc.cn/, accessed on 2 March 2025). In Taiyuan, there are nine air quality monitoring stations (see

Table 1. Locations of the air monitoring stations used in this study.

Station	Lon. (E)	Lat. (N)
Jiancaoping	112.52	37.88
Jianhe	112.57	37.91
Shanglan	112.43	38.01
Jinyuan	112.46	37.71
Xiaodian	112.55	37.73
Taoyuan	112.53	37.86
Wucheng	112.57	37.81
Nanzhai	112.54	37.98
Jinsheng	112.48	37.78

). We averaged the data over these stations to obtain the pollutant concentration for Taiyuan City. By pre-analyzing the variations of pollutants in Taiyuan since 2014, we found that, from 20 May to 30 May 2015, the concentrations of ozone and PM_2.5 were both high, so we chose this time period to study.

2.2. Model Description

2.2.1. Model Settings

In this research, the Weather Research and Forecasting/Chemistry (WRF-Chem) version 3.9.1 was employed, which has been developed by research institutions including NOAA, DOE/PNNL and NCAR [31]. It is a model in which air quality and meteorological parameters are coupled together. It has been proven that the WRF-Chem model can accurately predict pollutants such as ozone [30]. We defined our computational domains on the Lambert projection, and the center of the domains was 37.2° N, 108.06° E (see Figure 2). Three nested domains were set up in this research. Domain 1 covers most of the central and eastern provinces of China, nested domain 2 covers North China, and domain 3 covers almost all areas of the Shanxi Province, with a horizontal resolution of 3.3 km × 3.3 km. In present simulations, outputs of the outer domains, including the meteorological parameters and pollutant concentrations, were passed to the inner domains as the boundary conditions. Moreover, the boundary condition of the outermost domain (i.e., D01) used the default settings, because this condition only exerts a minor influence on simulations in the innermost layer (i.e., D03), due to the long distance between the innermost domain and the boundaries of the whole computational domain. The innermost domain has 186 grid points in the east–west direction and 159 grid points in the north–south direction, and 38 vertical layers. The simulated time period is from 20 May to 30 May 2015, and the simulation results were output hourly for analysis. The default initial conditions of WRF-Chem were adopted as the initial fields of the simulations. To reduce the uncertainty caused by implementing inaccurate initial conditions in the model, the first 28 h in the simulations were treated as the spin-up time. The parameterization schemes used in this study are listed in

Table 2. Options used for parameterizations of atmospheric processes in the WRF-Chem model.

Process	Option	Reference
Cloud microphysics	Morrison 2-moment	Morrison et al. [32]
Longwave radiation	Rapid Radiative Transfer Model (RRTM)	Mlawer et al. [33]
Shortwave radiation	Goddard	Chou et al. [34]
Surface layer	Monin–Obukhov scheme	Monin and Obukhov [35], Janić [36]
Land-surface physics	Noah land surface model	Chen and Dudhia [37], Ek et al. [38]
Urban surface physics	Urban canopy	Saijo et al. [39]
Planetary boundary layer	Yonsei University Scheme (YSU)	Hong et al. [40]
Cumulus parameterization	Grell 3D	Grell and Dévényi [41]

.

To accurately simulate the meteorological field, Four-Dimensional Data Assimilation (FDDA) was used to improve the model’s initialization [42]. FDDA is widely used by the air-quality community and shows a positive influence in reducing simulation bias for air-quality studies by optimizing the simulations of meteorological conditions. Applying FDDA nudging methods decreases the deviations in the simulations of meteorological parameters, especially the winds [43]. Thus, it is necessary to improve the wind field simulation using FDDA. The fifth generation reanalysis product of the European Centre for Medium Range Weather Forecasts (ERA5, 0.25° × 0.25°) [44], which combines model data and observations, was employed as the initial field to drive the FDDA.

Statewide Air Pollution Research Center (SAPRC99) [45] was used as the gas-phase mechanism in this simulation, coupled with the Model for Simulating Aerosol Interactions and Chemistry (MOSAIC),with 8 sectional aerosol bins. Chen et al. [46] conducted a comparison of RACM, SAPRC99 and SAPRC07 and found that, under heavy pollution conditions, the differences in simulation results are minor. Their study suggested the choice of the chemical mechanism might not be critical under heavy polluted conditions compared to clean conditions [46]. Given that our study focuses on heavy pollution, we opted for the SAPRC99 mechanism. The chemical mechanism is generated by the kinetic preprocessor (KPP) in the model, and equations are solved using a Rosenbrock-type solver [47,48,49].

2.2.2. Gas-Phase Chemical Mechanism

In this simulation, we adopted SAPRC99 as the gas-phase mechanism, coupled with the aerosol module MOSAIC with 8 sectional aerosol bins. In the SAPRC99 simulation, volatility basis set (VBS) is included for organic aerosol evolution [50].

SAPRC99 is a condensed mechanism based on the lumped species method. It has the ability to individually characterize the atmospheric reaction of 400 VOCs, and it can estimate the reactivity of 550 VOCs. The mechanism has been evaluated based on the results of nearly 1700 environmental laboratory experiments conducted at the University of California at Riverside, including experiments to test the predictions of ozone reactivity for more than 80 VOCs [45]. Coupled with SAPRC99, we used the Model for Simulating aerosol Interactions and Chemistry (MOSAIC) as the aerosol scheme. MOSAIC can address all major aerosol species that are important at the city, local and global scales [51]. It can also use the model or sectional approach to represent the aerosol size distribution. In this study, we used the sectional aerosol scheme of 8 bins over 39 nm–10 µm. VBS was also used for the gas-particle distribution in the aerosol scheme, which helps the model cope with the oxidation of a wide range of organic aerosol types and different VOCs in the atmosphere.

2.2.3. Emissions

Anthropogenic emissions of pollutants including NO_x, VOC, PM_2.5 and PM₁₀ were applied in the model using the 2015 emission inventory from the multi-resolution Emission Inventory for China (MEIC). These emissions are provided as different source sectors (electricity, industry, civil, transportation and agriculture) with a high spatial resolution (0.25° × 0.25°) on a monthly basis [52]. The MEIC emission inventory covers more than 700 anthropogenic emission sources in China, and it can provide a multi-scale and multi-component emission inventory that is in compliance with the air quality model. Aside from the anthropogenic emissions, we used the Model of Emissions of Gases and Aerosols from Nature (MEGAN) for biological emissions, which is calculated online based on weather and land-use data [53].

2.3. Assessment Parameters

The simulated meteorological parameters and pollutant concentrations were evaluated using the available surface observations. The assessment parameters used in this study include correlation coefficient (R), index of agreement (IOA), mean bias (MB), root mean square error (RMSE), normalized mean bias (NMB), normalized mean error (NME), mean fractional bias (MFB) and mean fractional error (MFE). Among these parameters, R, MB and RMSE were used to evaluate the simulation results of meteorological parameters, and R, IOA, NMB, NME, MFB and MFE were applied to validate the simulated pollutant concentrations. These parameters were calculated as follows [54]:

$$\mathrm{R}={\displaystyle \frac{{\sum }_{i=1}^n(P_i-\overline{P})(O_i-\overline{O})}{\sqrt{{\sum }_{i=1}^n{(P_i-\overline{P})}^2}\sqrt{{\sum }_{i=1}^n{(O_i-\overline{O})}^2}}}$$

(1)

$$\mathrm{IOA}=1-{\displaystyle \frac{{\sum }_{i=1}^n{(P_i-O_i)}^2}{{\sum }_{i=1}^n\left(\right|P_i-\overline{O}|+|O_i-\overline{O}{\left|\right)}^2}}$$

(2)

$$\mathrm{MB}={\displaystyle \frac{{\sum }_{i=1}^n(P_i-O_i)}{n}}$$

(3)

$$\mathrm{RMSE}=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n{(P_i-O_i)}^2}{n}}}$$

(4)

$$\mathrm{NMB}={\displaystyle \frac{{\sum }_{i=1}^n(P_i-O_i)}{{\sum }_{i=1}^n{O}_i}}$$

(5)

$$\mathrm{NME}={\displaystyle \frac{{\sum }_{i=1}^n\left|P_i-O_i\right|}{{\sum }_{i=1}^n{O}_i}}$$

(6)

$$\mathrm{MFB}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{\displaystyle \frac{2(P_i-O_i)}{P_i+O_i}}$$

(7)

$$\mathrm{MFE}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{\displaystyle \frac{2|P_i-O_i|}{P_i+O_i}}$$

(8)

In these formulas, n is the total number of data, P_i is the predicted value at the ith time point and O_i is the observed value at the ith time point. $\overline{P}$ and $\overline{O}$ are time-averaged values in predictions and observations, respectively.

2.4. Sensitivity Tests

To quantify the contributions of local emissions and the transportation from areas with heavy emissions (i.e., Taiyuan Basin, the NCP and Henan Province) to the air pollution in Taiyuan, we conducted sensitivity tests and calculated the percentage contribution by different processes. In these tests, we in turn switched off all anthropogenic emissions in Taiyuan City and these areas. Thus, the deviation between the results of the standard simulation and the sensitivity tests (i.e., $C_{\mathrm{standard}}-C_{\mathrm{sen}}$) represent the influence of local emissions and emissions from other regions (i.e., Taiyuan Basin, the NCP and Henan Province). In this calculation, $C_{\mathrm{standard}}$ is the simulated pollutant concentration in the standard simulation and $C_{\mathrm{sen}}$ is the pollutant concentration in the sensitivity tests. Moreover, the percentage contribution $\mathrm{P}$ was also calculated as follows:

$$\mathrm{P}={\displaystyle \frac{C_{\mathrm{standard}}-C_{\mathrm{sen}}}{C_{\mathrm{standard}}}}\times 100\%$$

(9)

3. Results and Discussions

We used WRF-Chem to simulate the pollution in Taiyuan City from 20 May to 30 May 2015. We first output the results of simulated meteorological parameters, such as temperature T at 2 m height, east–west wind speed u and north–south wind speed v at 10 m height and made a comparison with the measurements. In addition, we output the simulated ozone and PM_2.5 and compared them with data from corresponding observation stations and analyzed the temporal variations of pollutants. In order to explore the drivers of this air pollution in Taiyuan, we also analyzed the spatial distributions of pollutants (ozone and PM_2.5), meteorological parameters and emissions.

3.1. Comparison of Time Series Between Simulations and Measurements

Simulations of pollutants are strongly affected by the model accuracy in capturing meteorological parameters. For example, in this study wind plays an important role in the simulations of pollutants, because the monitoring stations are concentrated near emission sources, and the polluted area in Taiyuan is always in a small scope. If the simulated wind is largely biased from reality, a large deviation in the simulated concentrations of pollutants would occur. Thus, in order to ensure the model accuracy, we first validated the simulation results of meteorological parameters. Figure 3a gives a comparison between the simulated temperature at a height of 2 m and the observed temperature. It can be seen from the observations that, from 22 May to 27 May, the maximum of temperature in the afternoon increased day by day. The daily maximum reached a peak on 27 May, and then it began to decline. In comparison, it was found that the temperature is generally well simulated by the model, with a correlation coefficient (R) of 0.9. Overall, the gradual increase of the temperature from 20 May to 27 May, especially the peak on 27 May, which favors the ozone prediction, is well captured. In addition, the model also successfully simulates the lowest values of the temperature in the evenings, which is important for the estimation of PM_2.5.

In addition to the temperature, we also compared the simulation results of wind at a height of 10m with the measurements (see Figure 3b,c). From the observational data shown in Figure 3, the wind direction and the wind speed change consistently, and the wind speed u has a maximum 7.88 m·s⁻¹ on the night of 27 May, while the wind speed v mainly changes between −5 m·s⁻¹ and 5 m·s⁻¹. In general, the model behavior in capturing the wind speed is not as good as that in simulating the temperature. The bias in simulating the wind speed is mainly due to the special terrain of Taiyuan. Taiyuan is located in Taiyuan Basin, surrounded by three mountains; thus, it is difficult to simulate the wind using this model. However, after applying the FDDA, predictions of the wind direction and the maximum wind speed have been improved. Moreover, the maximum wind speed u on 27 May was well captured, which is important for the formation of the following severe PM_2.5 pollution, which will be discussed in a later section. In addition, according to the values of the assessment parameters (see

Table 3. Assessment parameters for meteorological predictions.

Variable	Parameter	Value
T (°C)	R	0.90
	MB	2.37
	RMSE	3.42
u (m·s−1)	R	0.40
	MB	0.37
	RMSE	1.92
v (m·s−1)	R	0.25
	MB	0.38
	RMSE	1.87

), the wind simulation is acceptable, with mean biases (MBs) of 0.37 and 0.38 for u and v, respectively. Therefore, in this study, we believe the simulations of wind are reliable.

Figure 4 shows the simulated and observed temporal variations of ozone and PM_2.5. According to the observations, ozone is low at night and reaches the maximum in the afternoon, which is consistent with the change of temperature. Moreover, the ozone maximum in the afternoon progressively increased from 23 May to 27 May under the condition of a low wind speed, peaking on the afternoon of 27 May (see Figure 4a), and then decreasing. In simulations, the overall temporal variation of ozone is consistent with that in observations (R = 0.81). Moreover, the model captured the high ozone on 27 May. From the assessment parameters (see

Table 4. Assessment parameters for ozone and PM_2.5 predictions.

Variable	Parameter	Value	Variable	Parameter	Value
O3	R	0.81	PM2.5	R	0.70
	IOA	0.88		IOA	0.81
	NMB	−0.11		NMB	0.17
	NME	0.29		NME	0.47
	MFB	−0.13		MFB	−0.01
	MFE	0.41		MFE	0.49

), the simulation of ozone is credible (e.g., |NMB| ≤ 0.15, NME ≤ 0.35). After 28 May, precipitation and a high wind, which are unfavorable for the formation and the accumulation of ozone, appeared, so that ozone dropped remarkably. The model also captured this change.

In Figure 4b, the observed PM_2.5 shows peaks in the evenings and troughs in the afternoons. A maximum of PM_2.5 on 28 May was found (see Figure 4b), following the maximum of ozone on 27 May (see Figure 4a). Similar to ozone, PM_2.5 decreased after 28 May due to the precipitation. By comparing the modeled concentration of PM_2.5 with the observations, we found that the model simulates the temporal evolution of PM_2.5 with an R of 0.7. In addition, the high PM_2.5 on the early morning of 28 May, which corresponds to the ozone peak on 27 May, is also captured by the model. In general, although the model overestimated the PM_2.5 concentration, values of assessment parameters proved that the model results for PM_2.5 are still credible. Connecting to the meteorological predictions, we suggest that the reason for the PM_2.5 overestimation by the model might be the underestimation of the wind speed in early mornings in the simulations.

In order to further evaluate the model behavior in reproducing the pollution event, this study also analyzed the simulation results at each monitoring station for individual comparisons (see Appendix A). We calculated the correlation coefficient between the simulated values and the observed values (see

Table 5. Correlation coefficient between simulated and observed values of PM_2.5 and ozone at each station.

Station	PM2.5	Ozone
Jiancaoping	0.77	0.78
Jianhe	0.84	0.76
Shanglan	0.72	0.73
Jinyuan	0.42	0.57
Xiaodian	0.51	0.62
Taoyuan	0.70	0.70
Wucheng	0.58	0.61
Nanzhai	0.83	0.84
Jinsheng	0.58	0.74

). According to the evaluation criteria of Cohen [55], $0.3\le \mathrm{R}<0.5$ indicates low correlation, $0.5\le \mathrm{R}<0.8$ indicates moderate correlation, and $\mathrm{R}\ge 0.8$ indicates high correlation. For PM_2.5 simulations, most of the stations possessed correlation coefficients greater than 0.7, indicating moderate or high correlations. For ozone simulations, most of the stations also demonstrated moderate or high correlations, with values of R greater than 0.7. Therefore, the model performed well. The appendix shows an overestimation of PM_2.5 at southern observation stations in Taiyuan (Jinyuan, Jinsheng, Xiaodian), and these three observation stations are located in rural areas. As proposed by Qi et al. [29], the improper allocation of rural emissions in the emission inventory and amplified uncertainties during model calculations could contribute to PM_2.5 biases in pollution simulations in Shanxi Province. Lu et al. [56] identified that mesoscale models often underestimate turbulent diffusivities in stable boundary layers, particularly at night, leading to an aerosol overestimation. Furthermore, Bai et al. [25] pointed out possible reasons for the differences in model predictions, including biases in meteorological forecasting, limitations in simulations for special terrains, and inadequate description by chemical mechanisms in the model. In summary, we believe that the PM_2.5 overestimation may be influenced by uncertainties in the emission inventory, underestimation of the turbulent diffusivity in the stable boundary layer, model performance in areas with complex terrains and the choice of chemical mechanisms in the model.

In previous studies, Qi et al. [29] used the WRF-Chem coupled with the CBMZ mechanism and MOSAIC 8-bin scheme to simulate the air quality of cities in North China, including Taiyuan. They also found the PM_2.5 concentration in Taiyuan was largely overestimated in simulations, which is similar to our findings. They speculated that the overestimation of PM_2.5 is due to the uncertainty in the emission inventory and the treatment of chemical reactions in the model. The MEIC inventory used in this study includes over 700 anthropogenic emission sources [52]. Because the MEIC inventory only provides monthly average data, to apply the inventory into the air quality model, factors redistributing the monthly data into different months and days as well as different altitudes are needed, which can inevitably lead to errors when the MEIC inventory is applied to individual cases. Bai et al. [25] used WRF model with extensions (CAMx) to study the variations of pollutant concentrations in Shanxi Province from 2012 to 2015, and they found that the peak value of PM_2.5 in July would be significantly underestimated by the model, while the trough value would be overestimated. Xu et al. [57] indicated that the overestimation of PM_2.5 in China by WRF-CAMx may be due to the absence of certain heterogeneous reactions in the CAMx model and the uncertainties in the emission inventory. By comparison, in this study, WRF-Chem coupled with SAPRC99 performed well in predicting concentrations of ozone and PM_2.5, with assessment parameters within reasonable ranges.

3.2. Analysis of the Drivers of the Combined Pollution Event in Taiyuan

Above, we have validated the simulation results for the temporal variations of pollutants (ozone and PM_2.5) and meteorological parameters in Taiyuan. We then output the spatial distributions of observed and simulated pollutants, meteorological parameters and emissions to clarify the role of meteorological conditions, local emissions and emissions from other regions (e.g., Taiyuan Basin, the NCP and Henan Province) in the formation of this combined air pollution in Taiyuan City.

3.2.1. Spatial Distributions of the Observed Ozone and PM_2.5

From Figure 4, we can see that, on the afternoon of May 27 and on the morning of May 28, ozone and PM_2.5 in Taiyuan reached their peaks. Thus, we first pay attention to the observed ozone and PM_2.5 concentrations in Taiyuan and surrounding cities (Yangquan, Jinzhong, Xinzhou and Lvliang) during this time period, shown in Figure 5 and Figure 6, respectively. Figure 5 shows that the monitoring stations in Taiyuan are concentrated in the east of Taiyuan, which belongs to an urban area. According to the observed ozone concentration on the afternoon of 27 May, the highest ozone in Taiyuan reached 163.33 ppbv at 16:00 (see Figure 5c), and the ozone pollution lasted until 19:00. During this period, almost all stations in Taiyuan detected high levels of ozone. In addition, stations in Jinzhong (to the south of Taiyuan) and Yangquan (to the east of Taiyuan) were also found to have serious pollution during this time. Thus, we believe that the high ozone on the afternoon of 27 May in Taiyuan might be significantly influenced by a transport of air from regions to the south or to the east of Taiyuan.

In Figure 6, the level of PM_2.5 increased remarkably from 20:00 on 27 May to 1:00 on 28 May in Taiyuan, reaching the highest value of 263 µg·m⁻³ at 1:00. This particulate matter (PM) pollution in Taiyuan lasted until 6:00 on 28 May, and analogous pollutions also occurred in Jinzhong and Yangquan during this time period, similar to the ozone pollution discussed above. Thus, PM pollutants may also be transported from regions to the south or to the east of Taiyuan. After 3:00, the PM pollutants in Jinzhong and Yangquan began to diffuse, but the PM_2.5 levels in the middle of Taiyuan were still high (see Figure 6c). This is because of a weak diffusion condition in Taiyuan due to the special geographical conditions of Taiyuan and the low wind speed during this time period (see Figure 3). Therefore, we suggest that this heavy PM_2.5 pollution in Taiyuan City might be caused by an accumulation of pollutants coming from regions to the south or to the east of Taiyuan under a weak diffusion condition.

3.2.2. Contributions of the Transport and Emissions to Air Pollution in Taiyuan

Integrating the ozone and PM_2.5 observations and the topographical characteristics of Taiyuan, we found that the combined air pollution from 27 May to 28 May in Taiyuan might be caused by the transport of pollutants, accompanied by unfavorable diffusion conditions. Therefore, we continued to output the spatial distributions of pollutants in the simulations to understand the properties of this pollution event.

Figure 7 shows the vertical profiles of PM_2.5 from 27 May to 28 May. We found that, during this pollution event, PM_2.5 peaks at the surface and its concentration decreases with height. At 23:00 on 27 May, PM_2.5 on the surface was 134 µg·m⁻³. Then, it increased to 172 µg·m⁻³ at 3:00 on 28 May, and finally reached a maximum of 178 µg·m⁻³ at 7:00. Because the increase of PM_2.5 occurred close to the surface, we then displayed the model results at the lowest layer to further analyze the drivers of the PM_2.5 pollution.

Figure 8 shows the spatial distributions of PM_2.5 and the wind during the pollution event. We can see that the high PM_2.5 in Taiyuan occurred mainly in the south of the city. The polluted area is consistent with the area, with strong emissions shown in Figure A3 in Appendix B. Thus, we suggest local primary emissions an important source of the PM_2.5 pollution in Taiyuan. Moreover, the wind direction is southeast from 21:00 on 27 May to 1:00 on 28 May in Taiyuan (see Figure 8a–d), which is conducive to the transport of pollutants from the southeast to Taiyuan. As mentioned earlier, the NCP, which is located to the east of Taiyuan, has large emissions, so the PM pollution in Taiyuan is likely to be significantly affected by the transport of pollutants from the NCP. Wu et al. [28] indicated that, during the Asia summer monsoon season (from May to September), meteorological conditions over eastern China are characterized by prevailing southwesterly–southeasterly winds. They also concluded that the PM_2.5 concentration in Taiyuan is considerably influenced by NCP emissions. This claim is further confirmed by our simulation results. Furthermore, from 21:00 on 27 May to 7:00 on 28 May (see Figure 8), Taiyuan experienced a southeasterly wind, which transported the pollutants from the NCP to Taiyuan along the Taihang Mountain channel. Afterwards, unfavorable diffusion conditions led to the accumulation of pollutants, resulting in the occurrence of this pollution, leading to a PM_2.5 increase (higher than 200 µg·m⁻³) and an expansion of area with high PM_2.5. Then, the wind speed declined and tended to zero at 7:00, which is unfavorable for the diffusion of particulate matters in Taiyuan. In addition, the complex terrain in Taiyuan also makes it difficult for the pollutants to diffuse. Thus, we suggest that the severe PM_2.5 pollution in Taiyuan is a combined result of local emissions, a trans-boundary pollutant transport from the NCP and unfavorable geographical and diffusion conditions.

To further quantify the contributions to PM_2.5 in Taiyuan by local emissions and the pollutant transport from the NCP, we performed sensitivity tests in which emissions from Taiyuan City and the NCP were switched off separately (see Appendix C). We found that emissions from the NCP are important for the severe PM_2.5 pollution occurring from 27 May to 28 May, with a contribution of 52%. In contrast, local emissions in Taiyuan contribute 43% to the peak of PM_2.5.

Figure 9 shows the contributions of the horizontal advection process (advh) and the vertical mixing process (vmix) to PM_2.5 at different heights. From Figure 9a–c, we can see that the contribution of the horizontal advection to the surface PM_2.5 in Taiyuan remains positive during this time period, and the contribution attains approximately 100 µg·m⁻³ at 5:00 on 28 May. In contrast, the contribution of the vertical mixing to the surface PM_2.5 is less than 10 µg·m⁻³ (see Figure 9d–f), which is one order smaller than that of the horizontal advection. Thus, we suggest that the horizontal advection is the major factor for the severe PM pollution at the surface in Taiyuan during this time.

We then turned to investigating the drivers of the severe ozone pollution in Taiyuan occurring on 27 May. Figure 10 shows the vertical profiles of ozone on the afternoon of 27 May. Different from PM_2.5, ozone peaks (>100 ppbv) at a height of about 2700 m early in the morning (see Figure 10a). In contrast, the surface ozone was only 55 ppbv at the same time. As time passes, the surface ozone increased due to the enhanced turbulent mixing in the afternoon, reaching 92 ppbv and 105 ppbv at 13:00 and 15:00, respectively. At 15:00, the surface ozone level is nearly equal to that at higher altitudes (see Figure 10c).

In order to reveal the factors dominating the ozone pollution in Taiyuan City, we also plotted the contributions of the horizontal advection process (advh) and the vertical mixing process (vmix) to ozone in Figure 11. We found that, from 11:00 to 15:00 on 27 May, contributions of the horizontal advection are negative on the ground but positive at a layer above 1000 m (see Figure 11a–c). In contrast, contributions of the vertical mixing are positive on the ground but negative at higher altitudes (see Figure 11e–f). Thus, the major factor causing the sharp increase of the surface ozone in Taiyuan during this time period is the vertical mixing, while the horizontal advection mainly contributes to the loss of the surface ozone. Based on these simulations, this study proposed the mechanism leading to the ozone pollution in Taiyuan during this time as follows: at high altitudes, ozone-rich air was horizontally transported to Taiyuan and then mixed downwards due to the enhanced vertical mixing, resulting in an increase of the surface ozone in Taiyuan City. From Figure 11d–f, we found the strongest downward vertical mixing occurring at an averaged height of approximately 1700 m. Thus, we continued to output the simulation results of ozone at the height of 1700 m to discover the major source regions of ozone.

Figure 12 shows the spatial distributions of ozone and the wind at the height of 1700 m on the afternoon of 27 May. From 13:00 to 16:00 (Figure 12a–d), a southerly wind prevailed in Taiyuan, which was able to carry pollutants from regions to the south of Taiyuan to this city. Moreover, we found the ozone level in regions to the south of Taiyuan larger than 140 ppbv. Thus, we suggested the ozone at higher altitudes of Taiyuan mainly come from regions to the south of Taiyuan. Regions that are located to the south of Taiyuan and possess large emissions or high ozone concentrations include Taiyuan Basin and Henan Province. Thus, we conducted a series of sensitivity tests in which we in turn switched off all anthropogenic emissions in Taiyuan, the NCP, Taiyuan Basin and Henan Province, and we calculated the percentage changes of ozone after switching off the emissions of these possible source regions. Figure A5 in Appendix C shows the contributions of emissions from different source regions to the surface ozone in Taiyuan. We found emissions from these four source regions (i.e., Taiyuan, the NCP, Taiyuan Basin and Henan Province) contributing 7%, 3%, 12% and 11% to the peak value of the surface ozone in Taiyuan, respectively. Thus, emissions from Taiyuan Basin and Henan Province were found to be relatively important for this ozone pollution in Taiyuan, while NCP emissions exerted a relatively minor influence.

Above, we have clarified the major source regions for the high ozone event occurring on 27 May in Taiyuan. However, the influence of emissions from these regions on ozone may vary with time. To further investigate this, we show the deviation in the ozone concentration between the results of the sensitivity tests and the standard simulation at different time points of 27 May (Figure 13). We found that, from 3:00 to 13:00, local emissions contributed negatively to the surface ozone (Figure 13a) and the ozone at a higher altitude (i.e., 1700 m; see Figure 13b). But the contributions became positive after 15:00. We suggest the reason for the shift in the contribution is that, in the morning, when the solar radiation is weak, local anthropogenic emissions in Taiyuan, especially the NO_x emissions, played a role in consuming ozone through the titration reaction. On the contrary, in the afternoon when the solar radiation is strong, local anthropogenic emissions would promote the ozone formation through the oxidation of VOCs. With respect to the emissions from the NCP, it is seen from Figure 13 that, in the early morning, NCP emissions contributed up to 29.33 ppbv to the surface ozone in Taiyuan. Thus, pollutants from the NCP may be transported to Taiyuan in the early morning of 27 May. However, in the afternoon, NCP emissions exerted only a minor influence on ozone in Taiyuan, no matter if they were near the ground or at a higher altitude. As we mentioned above, the prevailing wind direction in Taiyuan on the afternoon of 27 May is south, while the NCP is located to the east of Taiyuan. Thus, emissions from the NCP had only a weak impact on ozone in Taiyuan on the afternoon of 27 May. Regarding the emission from Taiyuan Basin, it can be seen in Figure 13b that its contribution to the high-altitude ozone is greater than 10 ppbv and remains stable over time. However, its contribution to the surface ozone did not show up until 7:00 (see Figure 13a). We suggest that pollutants from Taiyuan Basin were first horizontally transported to the high altitude of Taiyuan and then mixed to the ground from 7:00 in the morning, leading to the increase of the surface ozone. As for the emissions from Henan Province, they have a sustained contribution to ozone at these two height layers, with average values of 12.49 ppbv and 13.39 ppbv, respectively. Thus, we propose that pollutants from Henan Province affected the ozone at different heights uniformly in Taiyuan through the horizontal advection, along with the southerly wind.

4. Conclusions and Future Work

In this study, we used WRF-Chem to simulate a combined air pollution of ozone and PM_2.5 in a coal-based city of China, i.e., Taiyuan, from 20 May to 30 May 2015, with the MDA8 for ozone exceeding 140 ppbv and PM_2.5 concentrations surpassing $200\mu \mathrm{g}{\mathrm{m}}^{-3}$. By comparing the simulation results with measurements, we found the temporal evolutions and the peak values of pollutants well captured by the model, and the assessment parameters are also within reasonable ranges (R > 0.7), indicating that the simulation results are reliable. This example can provide a reference for the study of summertime combined pollution events in Taiyuan City. However, further research on multiple cases to verify the universality of this event is still necessary.

We also explored the main factors causing the combined air pollution in Taiyuan. We found that the PM_2.5 pollution in Taiyuan during this time period occurred mainly near the ground, and the surface PM_2.5 was greatly affected by local emissions and emissions from the NCP located to the east of Taiyuan. It was estimated that the contributions of local emissions and emissions from the NCP to the peak value of the surface PM_2.5 amount to 43% and 52%, respectively. The simulation results also showed that the horizontal advection is the dominant process for the increase of the surface PM_2.5 in Taiyuan, while the impact exerted by the vertical mixing is relatively weak. With respect to the ozone pollution, the highest ozone was found to appear at an averaged height of approximately 1700 m rather than at the surface, and the vertical mixing is the dominant process causing the downward transport of the high-ozone air and thus the elevation of ozone on the ground. We also found that the ozone at higher altitudes of Taiyuan was strongly affected by emissions from source regions to the south of Taiyuan, such as the Taiyuan Basin and Henan Province, and the contributions of emissions from these two source regions to the peak of the surface ozone in Taiyuan were estimated as 12% and 11%, respectively.

The present study still has some limitations. For example, the interaction between the two major pollutants (i.e., ozone and PM_2.5) should be studied further. To study the coupling mechanism between PM_2.5 and ozone, which is a very complex process, in more depth, simulations under different meteorological conditions would be beneficial. Additionally, sensitivity tests should be carried out by observing the contributions of specific chemical reactions and emission sources by varying the intensity of emissions. The simulations shown in the present manuscript are the start of our research, and investigations on the coupling of PM_2.5 and ozone are to be presented in our subsequent manuscripts. More processes and factors that can affect the formation of ozone and PM_2.5 pollution in Taiyuan need to be further investigated. In future studies, we plan to discover the role of many other factors, such as SOAs, in the formation of the combined pollution in Taiyuan. Furthermore, a series of sensitivity tests on emissions are to be performed in the future, which can help to guide the formulation of pollution control strategies for Taiyuan City.