Ozone Pollution in Chinese Cities: Spatiotemporal Variations and Their Relationships with Meteorological and Other Pollution Factors (2016–2020)

Abstract

With the acceleration of urbanization, ozone (O₃) pollution has become increasingly serious in many Chinese cities. This study analyzes the temporal and spatial characteristics of O₃ based on monitoring and meteorological data for 366 cities and national weather stations throughout China from 2016 to 2020. Least squares linear regression and Spearman’s correlation coefficient were computed to investigate the relationships of O₃ with various pollution factors and meteorological conditions. Global Moran’s I and the Getis–Ord index ${\mathrm{G}}_{\mathrm{i}}^{*}$ were adopted to reveal the spatial agglomeration of O₃ pollution in Chinese cities and characterize the temporal and spatial characteristics of hot and cold spots. The results show that the national proportion of cities with an annual concentration exceeding 160 μg·m⁻³ increased from 21.6% in 2016 to 50.9% in 2018 but dropped to 21.5% in 2020; these cities are concentrated mainly in Central China (CC) and East China (EC). Throughout most of China, the highest seasonal O₃ concentrations occur in summer, while the highest values in South China (SC) and Southwest China (SWC) occur in autumn and spring, respectively. The highest monthly O₃ concentration reached 200 μg·m⁻³ in North China (NC) in June, while the lowest value was 60 μg·m⁻³ in Northeast China (NEC) in December. O₃ is positively correlated with the ground surface temperature (GST) and sunshine duration (SSD) and negatively correlated with pressure (PRS) and relative humidity (RHU). Wind speed (WIN) and precipitation (PRE) were positively correlated in all regions except SC. O₃ concentrations are significantly differentiated in space: O₃ pollution is high in CC and EC and relatively low in the western and northeastern regions. The concentration of O₃ exhibits obvious agglomeration characteristics, with hot spots being concentrated mainly in NC, CC and EC.

1. Introduction

With the rapid development of industrialization and urbanization, surface ozone (O₃) concentrations are rising in cities throughout China. In addition to the natural emissions of surface ozone, biological and anthropogenic volatile organic compounds (BVOCs and AVOCs, respectively) in the atmosphere have tremendous impacts on the formation of O₃. BVOCs play a particularly important role globally, as their emission volume is estimated to be 10 times greater than that of AVOCs. Additionally, many BVOCs are characterized as highly reactive and can rapidly result in peroxy radicals, which are important for the formation of O₃ and ultrafine particles [1,2,3,4,5,6]. As a consequence, after fine particulate matter (PM_2.5), O₃ has become one of the most harmful gases to urban air quality [7]. Specifically, surface O₃ pollution can stimulate and damage human mucosal tissues (such as the respiratory system and eyes) and even damage the nervous and reproductive systems [8,9]. In addition, O₃ increases agricultural losses and accelerates the formation of pollutants such as particulate matter, which in turn affects the frequency and intensity of high-pollution weather events [10,11]. Therefore, the rising O₃ concentrations in many regions are drawing increasing concern.

Meteorological conditions further play a leading role in the diffusion, transformation and removal of atmospheric O₃ [12,13,14,15]. Thus, if effective measures are to be implemented to control and reduce O₃ concentrations, it is crucial to understand the temporal and spatial variations of O₃ and its relationship with various pollution factors and meteorological conditions [16,17,18]. Accordingly, over the past ten years, many studies have investigated these issues.

Remote sensing data, model simulation data and ground station data are all commonly used to research air pollutants. For example, Zhang, JX. et al. [19] focused on the temporal and spatial dynamics of total O₃ over China from 1981 to 2010 by using multisource remote sensing data and ground-level stations. Tan, K. C. et al. [20] used the Ozone Monitoring Instrument (OMI) on board the Aura satellite to obtain the total column ozone (TCO) and to examine the spatial and temporal distributions of atmospheric O₃ over Peninsular Malaysia in 2015. Nevertheless, although remote sensing data have the advantages of a large detection range and few restrictions with regard to the ground conditions, the obtained data are the total O₃ column concentrations, whereas the near-surface O₃ concentration is often of interest. Hence, in addition to using remote sensing data, some studies have conducted modeling to simulate and study O₃ concentrations. For example, Fu, Y. et al. [21] used the Model of Emissions of Gases and Aerosols from Nature (MEGAN) embedded within the global three-dimensional Goddard Earth Observing System Chemical Transport Model (GEOS-Chem) to simulate the concentrations of O₃ in China between 2001 and 2006. Hu, J. et al. [22] conducted a year-long (2013) air quality simulation using the Weather Research and Forecasting (WRF) model and the Community Multiscale Air Quality (CMAQ) model to provide detailed spatiotemporal information on O₃. However, although these model simulation data can solve the insufficient spatial coverage problem, the data generated by such modeling efforts are not convincing.

Surface O₃ data were very sparse before 2013. Starting in 2013, the surface monitoring network greatly expanded, and the China Ministry of Ecology and Environment began to publicly provide detailed hourly data across all of China. These data comprise information necessary for assessing the temporal and spatial patterns of air quality in China and its driving factors. As a result, several studies have utilized these monitoring data to determine the spatiotemporal characteristics of O₃ in China and its correlations with pollution and meteorological factors. For example, Zhang, Z. et al. [23] studied the temporal and spatial distributions of O₃ and the influences of meteorological factors on O₃ in the Yangtze River Delta region. Nevertheless, despite the long-term studies on O₃ in the literature, most previous studies have been limited to urban agglomerations, whereas little work has been conducted on suburban or rural areas. Moreover, Mao, J. et al. [24] studied the relationships between O₃ and meteorological factors in eastern China in 2017, but their analysis was limited to one year. In cases such as this, the temporal distribution of O₃ cannot be fully explored.

Based on the above problems, this paper uses the hourly average concentrations of six pollutants (O₃, nitrogen dioxide (NO₂), carbon monoxide (CO), inhalable particulate matter (PM₁₀), PM_2.5 and sulfur dioxide (SO₂)) in 366 cities in China and the daily averages of meteorological variables (including surface temperature, sunshine duration, pressure and relative humidity) from 839 national weather stations to study the spatiotemporal distribution of O₃ during the period of the 13th Five-Year Plan (2016–2020). On this basis, the relationships between O₃ and other air pollutants and meteorological factors are discussed, as are the effects of other air pollutants and meteorological factors on O₃, and the spatial aggregation characteristics of O₃ are analyzed.

2. Materials and Methods

2.1. Study Area

Considering geography, climate, economy and other factors, this article divides China into seven regions (Figure 1) according to the Chinese geographic regional classification scheme: Central China (CC), East China (EC), North China (NC), South China (SC), Northwest China (NWC), Southwest China (SWC), and Northeast China (NEC) [25].

2.2. Dataset

2.2.1. Air Quality Data

The data analyzed in this study include hourly O₃ concentrations recorded at China Environmental Monitoring Stations in 366 cities across China. The start and end times of the data are 1 January 2016, and 31 December 2020, respectively. The study area encompasses all of China, including all prefecture-level cities (except Hong Kong, Macao, and Taiwan).

According to the O₃ concentration specifications and standards of the Ambient Air Quality Standard (GB3095-2012), the original data were subjected to quality control according to the data validity requirements, and negative hourly O₃ concentrations and missing values in the original data were removed. The seasonal average refers to the arithmetic average of the maximum 8 h average concentrations on each day during a given season, while the annual average refers to the arithmetic average of the maximum 8 h average concentrations on each day in that calendar year.

2.2.2. Meteorological Data

The acquired meteorological data include the China Surface Climate Data Daily Value Dataset (V3.0) and Ground Cumulative Annual Value Dataset published by the China Meteorological Science Data Sharing Network (http://data.cma.cn, accessed on 30 April 2022). The meteorological elements considered include the ground surface temperature (GST), pressure (PRS), relative humidity (RHU), wind speed (WIN), sunshine duration (SSD) and precipitation (PRE), the data of which originated from 699 stations across China. The original data files were strictly controlled and inspected to exclude outliers and null values from the dataset.

2.3. Methodology

2.3.1. Evaluation Standard

The newly implemented Ambient Air Quality Evaluation Technical Specification (Trial) (HJ663-2013) in China requires that the daily evaluation should consider the maximum 8 h average O₃ concentration in each city to determine its daily compliance status, whereas the monthly evaluation should consider the maximum daily O₃ concentrations in a given month in each city (here, the 90th percentile of the 8 h moving average is used to determine monthly compliance), and the annual evaluation depends on the 90th percentile of the highest 8 h moving average value of the daily O₃ concentrations throughout the year to determine the annual compliance status.

O₃ concentration limits are associated with different standards according to different functional areas. The first-level excess concentration threshold is 100 µg·m⁻³, and the second-level excess concentration threshold is 160 µg·m⁻³. This article uniformly applies the second-class environmental function zone standard to determine whether the daily, monthly and annual average values in each city satisfy the national standards.

2.3.2. Multiple Linear Regression

In regression analysis, if there are two or more independent variables, the process is called multiple regression. In the real world, the target variable is often associated with multiple variables, so the target variable is jointly estimated by the optimal combination of multiple independent variables. Multiple linear regression is an extension of univariate linear regression. Its basic formula is as follows:

$$y=a{x}_1+b{x}_2+\dots +n{x}_m+\epsilon$$

(1)

This paper uses multiple linear regression to test the impacts of other air pollutants on O₃. A larger coefficient corresponding to a certain pollutant signifies that that pollutant can contribute more to the generation of O₃.

2.3.3. Global Moran’s I Index

Since near-ground O₃ is continuously distributed in space and because atmospheric photochemical reaction precursors in cities across the country interact with each other, this article applies global Moran’s I index [26] to test the spatial autocorrelation of O₃ on a national scale [27,28]. The index is calculated as follows:

$$\mathrm{I}=\frac{\mathrm{n}}{{\mathrm{S}}_0}\frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{ij}}{\mathrm{Z}}_{\mathrm{i}}{\mathrm{Z}}_{\mathrm{j}}}{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{Z}}_{\mathrm{i}}^2}$$

(2)

where ${\mathrm{Z}}_{\mathrm{i}}$ is the deviation between the attribute value for element i and the average value (x_i − $\overline{\mathrm{X}}$); ${\mathrm{w}}_{\mathrm{i},\mathrm{j}}$ denotes the spatial weights between elements i and j; n is the total number of elements; and

$${\mathrm{S}}_0={\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}$$

(3)

Global Moran’s I was tested for significance using the Z test, and the calculation formula is

$$\mathrm{Z}\left[\mathrm{I}\right]=\frac{\mathrm{I}-\mathrm{E}\left[\mathrm{I}\right]}{\sqrt{\mathrm{V}\left[\mathrm{I}\right]}}$$

(4)

where $\mathrm{Z}\left[I\right]$ is the Z test value of global Moran’s I; $\mathrm{E}\left[\mathrm{I}\right]$ is the mathematical expectation; and $\mathrm{V}\left[\mathrm{I}\right]$ is the variance

Generally, I is between −1.0 and 1.0. When I ∈ [−1, 0), there is a negative spatial correlation. The closer the value of I is to −1, the greater the difference in attribute values between spatial units. When I = 0, the spatial distribution is random. As I approaches 0, the attribute values between spatial units become increasingly irrelevant. When I ∈ (0, 1], the correlation is spatially positive, and values closer to 1 indicate a stronger correlation of attribute values between spatial units.

2.3.4. Getis–Ord ${\mathrm{G}}_{\mathrm{i}}^{*}$ Index

Although global Moran’s I index can reflect the spatial correlation of O₃ concentrations as a whole, it cannot reflect the sign and degree of correlation between any two cities. The Getis–Ord index ${\mathrm{G}}_{\mathrm{i}}^{*}$ [29,30] is selected for a local area to judge the spatial heterogeneity in the area by analyzing local information, revealing cold spots and hot spots of O₃ concentrations that reflect the degree of correlation between neighboring areas. This index is calculated as follows:

$${\mathrm{G}}_{\mathrm{i}}^{*}=\frac{{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}{\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{X}}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}}{\mathrm{S}\sqrt{\frac{\left[\mathrm{n}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}^2-{\left({{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}\right)}^2\right]}{\mathrm{n}-1}}}$$

(5)

where x_j is the attribute value for element j; ${\mathrm{w}}_{\mathrm{i},\mathrm{j}}$ is the spatial weight matrix between elements i and j; n is the total number of elements; and

$$\overline{\mathrm{X}}=\frac{{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{j}}}{\mathrm{n}}$$

(6)

$$\mathrm{S}=\sqrt{\frac{{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{j}}^2}{\mathrm{n}}-{\left(\overline{\mathrm{X}}\right)}^2}$$

(7)

The ${\mathrm{G}}_{\mathrm{i}}^{*}$ statistic is the Z score. If Z(${\mathrm{G}}_{\mathrm{i}}^{*}$) is significant and positive, it indicates that the value near position i is higher than the average, and the higher the Z(${\mathrm{G}}_{\mathrm{i}}^{*}$) score is, the greater the number of high-value clusters (hot spots). In contrast, if Z(${\mathrm{G}}_{\mathrm{i}}^{*}$) is significant and negative, it indicates that the value near position i is lower than the average, and the lower the Z(${\mathrm{G}}_{\mathrm{i}}^{*}$) score is, the greater the number of low-value clusters (cold spots). This article uses the Hot Spot Analysis (Getis–Ord ${\mathrm{G}}_{\mathrm{i}}^{*}$) tool for ArcGIS 10.7 software (Esri, California, United States) to analyze the annual, quarterly and monthly O₃ concentrations in 366 cities in China and to conduct an analysis of hot and cold spots.

3. Analysis of Spatiotemporal Characteristics and Influencing Factors

3.1. Analysis of Temporal and Spatial Changes

3.1.1. Temporal and Spatial Changes in the Annual Average Ozone Concentration

From 2016 to 2020, the average 90th percentile values of the 8 h average O₃ concentrations in Chinese cities were (successively) 138.7, 149.3, 151.4, 147.1 and 126.8 µg·m⁻³, and the O₃ concentrations exhibited similar variations in each year. Among the regions, due to the large number of cities in CC, EC and SC, human factors appear to have certain impacts on the O₃ concentration and caused the O₃ concentration to rise until 2019 [31]. In contrast, the number of cities in NWC is relatively small, so the O₃ concentration began to gradually decline in 2017. NC, SWC and NEC have moderate numbers of cities, so these regions show the same trend as the whole country; that is, the O₃ concentrations rose up until 2019, whereupon they started to decline.

The annual average O₃ concentrations have strong spatial characteristics (Figure 2). Over the past five years, most cities with O₃ concentrations exceeding 160 μg·m⁻³ were located in the southern part of NC and the northern parts of CC and EC, and the proportion increased from 20% in 2016 to 50% in 2018. This percentage finally declined to 20% in 2020. The cities with O₃ concentrations ranging from 100 to 160 μg·m⁻³ were evenly distributed throughout China, and the proportion remained 50% or higher. In addition, most cities with O₃ concentrations below 100 μg m⁻³ were located in NEC and SWC, and the proportion remained 5% or lower.

3.1.2. Temporal and Spatial Changes in the Seasonal Average Ozone Concentration

The spatial patterns of O₃ concentrations exhibited substantial seasonal variations (Figure 3), according to which China could be divided into three regions. The first region includes NC, NWC and NEC. The O₃ concentrations in this region were the highest in summer and the lowest in winter. The concentrations in spring and autumn were moderate, and the O₃ concentrations in spring were higher than those in autumn, especially in NEC. The second region includes CC and EC. The summer and winter O₃ concentrations in this region were comparable to those in the first region and were similar to the spring and autumn O₃ concentrations. The third area includes SC and SWC. The seasonal O₃ concentrations in this area were relatively close, but the O₃ concentrations remained the lowest in winter. However, in SC, the O₃ concentrations in autumn were the highest, while in SWC, the O₃ concentrations in spring were the highest. From north to south, the absolute difference in the O₃ concentration among the four seasons gradually decreased. Due to climatic reasons, the absolute temperature difference among the four seasons is large in the north but small in the south, which is warm year-round [32]. Hence, temperature affects the rate at which precursors are converted into O₃.

3.1.3. Temporal and Spatial Changes in the Monthly Average Ozone Concentration

The monthly O₃ concentration was spatially correlated, and the number of cities exceeding the standard in each region presented different variation characteristics (Figure 4). Due to the large number of urban agglomerations in the eastern regions, the number of cities in EC exceeding the standard in the whole year was higher than that in the western regions. The higher latitudes of the northern Chinese regions lead to an earlier onset of warming each year, so the northern regions reached the highest number of cities exceeding the standard earlier than the regions to their south. In addition, due to the large temperature difference across the northern regions throughout the year, the number of cities exceeding the standard varied greatly from month to month. For example, the number of cities in EC exceeding the standard in May was 343 times that in January. In contrast, because the latitudes and longitudes of the southern Chinese regions are relatively low, the temperature rises slowly, and the temperature difference is small throughout the year; consequently, only a small number of cities exceeded the standard among the southern regions throughout the year, and the temperature distribution was relatively uniform each month.

During January–March, the number of cities exceeding the standard was relatively small; examples of these cities include Suzhou in NC in 2016, Beihai in SC in 2017 and Guoluo in SWC in 2019. With a gradual increase in temperature, the speed with which O₃ is generated increases, and the number of cities exceeding the standard begins to increase [32]. Among the different regions, EC had a particularly high number of cities exceeding the standard. From April to September, the number of such cities accounted for approximately 50% of the national total and peaked at 343 in May, 15.5 times the number in SC. Meanwhile, the numbers of cities exceeding the standard in CC and NC reached approximately 20% of the national total and peaked at 135 (September) and 150 (June), respectively. In NEC, NWC and SWC, on the one hand, there are relatively few urban agglomerations, and on the other hand, the temperature is relatively low; as a result, few cities exceeded the standard (only 23, 9, and 3 cities, respectively, exceeded the standard in September). In addition, the changes in the number of cities exceeding the standard in SC were relatively unique; due to the low latitude, the temperature rises slowly, and the change range is small, causing the number of cities exceeding the standard to peak at 61 in September. Over the next 10–12 months, only the three southernmost regions (CC, EC and SC) contained cities with excess O₃, but their numbers were slowly decreasing. Among them, SC (the southernmost tip of China) has always maintained a high level; nevertheless, even in December, only two cities in SC, Zhuhai and Yangjiang, exceeded the standard, with concentrations of only 161 μg·m⁻³ and 163 μg·m⁻³, respectively.

3.2. Analysis of the Relationships between O₃ and Other Pollution Factors and Changes in Hourly Concentrations

The hourly concentration trends of O₃ and five other gases were compared (Figure 5). Because the change trend of O₃ is clearly related to those of these other gases, their correlations were tested by using Spearman’s correlation coefficient and the T test. Since the significance and correlation coefficients of the linear regression equations in each region are relatively similar, here, only the results for all of China are discussed.

The calculated correlation coefficient between NO₂ and O₃ is −0.753, the T value is −108.97, and the P_t value is 0, indicating that these two gases attain a significant negative correlation. Atmospheric O₃ at ground level is formed in the presence of ultraviolet (UV) light (λ < 424 nm) through the direct photolysis of NO₂ and VOCs ($\mathrm{N}{\mathrm{O}}_2+\mathrm{V}\mathrm{O}\mathrm{C}\mathrm{s}\begin{array}{c}\underset{\_}{\underset{\_}{\mathrm{U}\mathrm{V}}}\\ \end{array}{\mathrm{O}}_3$) [3,33], and sunlight accelerates the photochemical conversion of NO₂ into O₃, causing O₃ concentrations to increase and NO₂ concentrations to decrease during the day; in contrast, at night, the rate of this reaction is slow, resulting in relatively low O₃ concentrations.

CO attains a correlation coefficient of −0.791 with O₃, and the T value is −85.33, while the P_t value is 0. This indicates that these two gases also achieve a significant negative correlation. Under UV irradiation, CO reacts with oxygen to form O₃ and carbon dioxide ($\mathrm{C}\mathrm{O}+2{\mathrm{O}}_2\begin{array}{c}\underset{\_}{\underset{\_}{\mathrm{U}\mathrm{V}}}\\ \end{array}\mathrm{C}{\mathrm{O}}_2+O_3$) [34], and sunlight further increases the rate of this photochemical reaction, increasing O₃ concentrations and decreasing CO concentrations during the day; conversely, the slow photochemical reaction rate at night results in relatively low O₃ concentrations.

PM_2.5 and O₃ attain a correlation coefficient of −0.794 with a T value of −66.31 and a P_t value of 0, similarly indicating a significant negative correlation between these two pollutants. The reason for the low correlation coefficient is that PM_2.5 and O₃ exhibit a coupling mechanism. When the value of the kinetic parameter, i.e., the hydrogen superoxide (HO₂) uptake coefficient, is 0.2, the change in the heterogeneous HO₂ uptake rate exerts an important impact on O₃ [35], while a reduction in PM_2.5 leads to an increase in HO₂, which increases the number of OH radicals. These OH radicals then generate O₃ under photochemical reactions, causing the opposing trends of PM_2.5 and O₃.

In addition to utilizing the Spearman correlation coefficient to verify the effects of other air pollutants on O₃, a multiple regression analysis was also performed. A larger parameter for a certain pollutant suggests that it has a greater impact on the O₃ concentration. The expression obtained after multiple iterations of training is as follows (8):

$${\mathrm{O}}_3 \approx \left(-0.744\right)\mathrm{C}\mathrm{O}+\left(-0.38\right)\mathrm{N}{\mathrm{O}}_2+\left(0.442\right)\mathrm{P}{\mathrm{M}}_{2.5}+\left(0.017\right)\mathrm{P}{\mathrm{M}}_{10}+\left(-0.08\right)\mathrm{S}{\mathrm{O}}_2$$

(8)

The approximately equal sign is used in this formula because the overall error is not considered here. Nevertheless, the parameters of NO₂, CO and PM_2.5 are clearly larger than those of PM₁₀ and SO₂, which means that the former three air pollutants have greater impacts on O₃.

3.3. Analysis of the Relationships between O₃ and Meteorological Factors

Upon testing the significance of the correlation of certain meteorological factors with O₃, some factors attained a p value greater than 0.01, indicating that the correlation is not significant [36]. These meteorological factors are not examined here. It follows, then, that the correlations discussed below are all significant; i.e., their P values are less than 0.01. In addition, we stipulate that the correlation can be explained only when the correlation coefficient |r| > 0.3 [37].

The annual average O₃ concentrations in China exhibited significant spatial correlations with GST, PRS, RHU, WIN, SSD and PRE during 2016–2020 (Figure 6). The colored bars extending above and below the invisible axis in each panel of the figure signify positive and negative correlation coefficients, respectively. GST and SSD attain a significant positive correlation in all regions. This is because high temperatures and UV radiation speed up the rate at which reactions with molecules such as NO₂ produce O₃ [38,39]. PRS exhibits a significant negative correlation in most regions because O₃ is easily soluble in water, and an increase in PRS accelerates the rate at which O₃ dissolves in water [40,41]. RHU achieves a significant negative correlation in all regions (except NC) because RHU affects UV exposure and reduces the rate at which other air pollutants produce O₃ [42,43]. WIN attains a significant negative correlation in CC and SC but a significant positive correlation in the remaining regions. The reason for this phenomenon is that high wind speeds lift the height of the atmospheric boundary layer, causing the O₃ in the upper layer to be mixed down to the surface, resulting in an increase in the concentration of near-surface O₃ [44,45]. Finally, except for the significant negative correlation attained in SC, PRE exhibits a significant positive correlation in the other regions. This means that large amounts of PRE reduce O₃ concentrations, while small amounts of PRE increase O₃ concentrations. When there is a large amount of PRE, the cloud cover is generally greater, and the cloud layer—which absorbs UV rays—reduces the rate of O₃ generation [46,47]. In addition, O₃ dissolves into the water film in large quantities, resulting in a decrease in the O₃ concentration [48].

The seasonal average O₃ concentrations in China and the seasonal averages of GST, PRS, RHU, WIN, SSD and PRE exhibited notable spatial correlations during 2016–2020 (Figure 6). According to the data, GST is basically positively correlated with O₃ in all regions. GST exerts a greater impact during spring and autumn than during summer and winter and in the eastern and northern regions than in the western and southern regions. On the one hand, this variability is due to the large temperature changes that occur in spring and autumn, which shows a clear correlation. On the other hand, compared with the south, the north experiences greater temperature changes in the four seasons, so the correlation in the north is higher than that in the south [38,39].

In terms of PRS, because the changing of seasons has less effect on PRS, PRS is negatively correlated to a similar degree with O₃ in all seasons. PRS further increases the water solubility of O₃, causing O₃ concentrations to decrease with increasing PRS [40,49]. Likewise, RHU is negatively correlated with O₃ in all seasons. Since O₃ is easily soluble in water, O₃ concentrations gradually decrease with increasing RHU [41].

In contrast, WIN displays a positive correlation with O₃ in all regions during all four seasons of the year. Among the various regions, the influence of the wind speed in the north is obviously greater than that in the south because there are many plains in the north, and the wind speed is often higher there than in the south [42,43]. Similarly, SSD is positively correlated with O₃ throughout the year, and the correlation coefficients in the south are higher than those in the north. Because SSD directly reflects the radiation intensity, a high level of irradiation accelerates the conversion of O₃ precursors into O₃, thereby increasing the O₃ concentrations [44,45].

Finally, PRE is negatively correlated with the seasonal average O₃ concentrations in China in summer, while in the other seasons, PRE generally attains a negative correlation in the south while typically reaching a positive correlation in the north. This is because when PRE is weak, O₃ is brought from the stratosphere close to the surface, resulting in higher near-surface O₃ levels. In contrast, when PRE is heavy, the humidity further increases, causing more O₃ to dissolve in the water and thus reducing the O₃ concentrations [46,47].

4. Spatial Agglomeration Characteristics of O₃ Concentrations in Chinese Cities

4.1. Spatial Autocorrelation Test of O₃ Concentrations

The spatial autocorrelation (Moran’s I) index of the annual mean, quarterly mean and monthly mean O₃ concentrations were calculated for all 366 cities by the ArcGIS Spatial Analyst tool. Generally, if Z[I] > 1.65 or Z[I] < −1.65, the confidence is 90%; if Z[I] > 1.96 or Z[I] < −1.96, the confidence is 95%; and if Z[I] > 2.58 or Z[I] < −2.58, the confidence is 99%. When Z[I] < −2.58, the spatial distribution of the O₃ concentration is negatively correlated and scattered; conversely, when Z[I] > 2.58, the O₃ concentration is positively correlated in space (i.e., high values agglomerate in separate clusters, as do low values, forming distinct hot and cold spots).

As shown in

Table 1. Spatial autocorrelation indexes of the seasonal average O₃ concentrations in 366 Chinese cities.

Value		Spring	Summer	Autumn	Winter	Year
Moran’s I	2016	0.17	0.23	0.40	0.17	0.24
	2017	0.36	0.27	0.49	0.24	0.34
	2018	0.30	0.28	0.69	0.21	0.37
	2019	0.31	0.31	0.58	0.26	0.36
	2020	0.37	0.30	0.68	0.18	0.38
Z[I]	2016	17.92	24.71	45.39	29.4	29.36
	2017	32.73	36.54	57.21	42.66	42.29
	2018	30.40	35.12	73.13	37.20	43.96
	2019	34.75	40.00	69.15	46.55	47.61
	2020	35.99	37.58	66.87	29.64	42.52

and

Table 2. Spatial autocorrelation indexes of the monthly average O₃ concentrations in 366 Chinese cities.

Value		1	2	3	4	5	6	7	8	9	10	11	12
Moran’s I	2016	0.03	0.11	0.05	0.19	0.14	0.17	0.23	0.22	0.5	0.12	0.27	0.28
	2017	0.25	0.25	0.14	0.36	0.32	0.27	0.21	0.21	0.38	0.55	0.38	0.30
	2018	0.25	0.16	0.14	0.33	0.25	0.31	0.24	0.23	0.63	0.67	0.41	0.10
	2019	0.27	0.04	0.21	0.19	0.27	0.30	0.28	0.24	0.53	0.73	0.51	0.30
	2020	0.10	0.07	0.14	0.18	0.24	0.25	0.28	0.24	0.62	0.58	0.43	0.26
Z[I]	2016	0.03	0.11	0.05	0.19	0.14	0.17	0.23	0.22	0.5	0.12	0.27	0.28
	2017	0.25	0.25	0.14	0.36	0.32	0.27	0.21	0.21	0.38	0.55	0.38	0.30
	2018	0.25	0.16	0.14	0.33	0.25	0.31	0.24	0.23	0.63	0.67	0.41	0.10
	2019	0.27	0.04	0.21	0.19	0.27	0.30	0.28	0.24	0.53	0.73	0.51	0.30
	2020	0.10	0.07	0.14	0.18	0.24	0.25	0.28	0.24	0.62	0.58	0.43	0.26

, Moran’s I indexes are all positive, and all Z[I] values are greater than 2.58, passing the 1% test of significance and indicating that China’s urban O₃ concentrations during 2016–2020 were spatially autocorrelated. Each region is assigned a single city as the basic unit. In addition to the different characteristics between different regions, there are also differences in the natural environmental conditions of different cities within the same region, which together lead to changes in the air quality in each region. Therefore, cluster analysis was performed on each city to find the hot and cold spots (i.e., where O₃ is concentrated) to better understand the mechanisms of O₃ pollution in different regions and to better formulate targeted prevention and control policies.

4.2. Spatial Agglomeration Characteristics of O₃ Concentrations

The average Z[I] index for the past five years is 41.15, indicating that the O₃ concentrations are spatially concentrated. Figure 7 shows that hot spots are concentrated mainly in NC, southern NEC, eastern NWC, northern CC, and central and northern EC; in contrast, cold spots are concentrated mainly in northern Xinjiang, northern NEC, central and southern SWC, and SC. However, the spatial autocorrelation of O₃ is not obvious in various scattered cities throughout central NC, central and southern NWC, CC, southern EC and northern SWC.

In recent years, with the advancement of market-oriented reforms, the urbanization and industrialization levels of some cities in EC and NC have continuously risen. As some of these cities play important roles in driving economic development, in the development of these economies, automobile exhaust and industrial waste gas are continuously and inevitably emitted. Hence, the emission of man-made precursors raises the O₃ concentrations in these urban agglomerations relative to those in surrounding cities.

Meteorological conditions, especially temperature, have an important influence on the O₃ concentrations near the ground. In different seasons, changes in the solar radiation intensity lead to large seasonal differences in the concentrations of O₃, and periodic changes are evident (Figure 7). In spring, hotspot cities are distributed mainly in southern NEC, NC, northern CC and EC. The distribution of hotspot cities in summer is further expanded and ranges from the east coast to inland areas, indicating that the photochemical reactions of O₃ precursors caused by solar radiation are enhanced in summer. The extent of hotspot cities further increases in autumn and spreads to the southeast coast, with these cities concentrated mainly in southern NC, EC, CC and SC. Then, in winter, the breadth of hotspot cities shrinks sharply but continues to expand south, with these cities appearing mainly in southern SWC, southern EC and SC. The cold spot cities in spring and summer are concentrated mainly in NWC, SWC and SC; those in autumn are concentrated in NWC, SWC and NEC; and those in winter are concentrated predominantly in NEC, NC, northern CC and EC. Except for some cities in the northern parts of NWC and NEC, no area in the country features stable and continuously low O₃ concentrations.

5. Conclusions

This paper analyzes the temporal and spatial changes in urban O₃ pollutants across China from 2016 to 2020, as well as the influencing factors and spatial agglomeration characteristics of O₃.

(1)

The O₃ concentrations exhibited the same change trend year over year. However, the O₃ concentrations in major urban agglomerations continued to rise in 2019. Seasonally, O₃ has strong spatial characteristics and can be latitudinally divided into three regions from south to north. On a monthly scale, many cities exhibited O₃ concentrations exceeding the national standard (160 μg·m⁻³) from April to September, with the peak appearing in June.

(2)

Cities with heavy O₃ pollution were concentrated in the northern part of CC, southern NC and EC. Over time, the number of cities with O₃ concentrations above 160 μg·m⁻³ has gradually increased, with the highest value being discovered in 2018, accounting for 50% of the total number of cities. In terms of monthly O₃ concentrations, the number of cities with O₃ concentrations above 160 μg·m⁻³ was relatively high from April to September, and most of these cities were concentrated in CC, EC and NC.

(3)

O₃ displays significant negative correlations with NO₂, PM_2.5 and PM₁₀, and its correlation coefficient with NO₂ is the highest (with an r value of −0.399). The correlations between O₃ and various meteorological factors vary both seasonally and spatially. GST and SSD show a significant positive correlation with O₃ in all regions. PRS and RHU show significant negative correlations in most regions, while WIN and PRE show significant positive correlations in most regions. The effects of GST and PRS are greater in spring and autumn than in summer and winter. PRE is negatively correlated with the O₃ concentration in summer and negatively correlated in the other seasons. For RHU, WIN and SSD, their correlations with O₃ do not significantly differ from season to season.

(4)

The concentrations of O₃ in Chinese cities have significant spatial agglomeration characteristics. Seasonal differences in the O₃ concentration are large, and periodic changes are obvious. The hotspot cities in spring are distributed mainly in the southern part of NEC, NC, EC and northern CC. The distribution of summer hotspot cities further expands from the east coast to the inland areas. In autumn, the scope of hotspot cities expands further and spreads to the southeast coast, mainly in the southern part of NC, EC, CC and SC. The scope of winter hotspot cities shrinks sharply but continues to expand south, with cities appearing mainly in SWC, SC, EC and SC.

Due to the complex formation mechanism of O₃ pollution, many factors, such as the concentrations of precursors, emission intensity, and meteorological conditions, can affect O₃. Accordingly, the interactions between the O₃ concentration and its various influencing factors need to be further studied. Only by fully revealing the causes and sources of O₃ in different regions and quantifying the leading factors influencing the formation of O₃ can we effectively improve the prevention and control of O₃ pollution. Therefore, the state’s prevention and control of O₃ can be carried out from two aspects. On the one hand, it is necessary to treat O₃ precursors (NOx and VOCs) at the same time to reduce the generation of O₃. On the other hand, controlling meteorological changes (for example, introducing artificial rainfall in summer when the temperature is higher) can reduce the concentrations of O₃.