*3.2. Spatial Variation Trend of PM2.5*

To determine the spatial distribution characteristics of PM2.5 concentrations in the study areas, we calculated the Global Moran's I during 2015–2019. As shown in Table 2, with *p*-values < 0.01 and Z-score > 2.58, the Global Moran's I was acceptable. From 2015 to 2019, the PM2.5 concentrations in the study areas showed a significant positive spatial correlation, which indicated that the diffusion of PM2.5 concentrations between cities was not random, and rather showed similar spatial connections and tended to aggregate. This spatial correlation has been gradually increasing since 2016. To better exhibit the agglomeration characteristics of the study area, we drew a Moran scatter diagram, as shown in Figure 4. Most cities are concentrated in the first and third quadrants, and only a few cities appear in the second and fourth quadrants which indicate that PM2.5 pollution in the study areas presented obvious "high–high" and "low–low" agglomeration. This spatial characteristic is caused by the unbalanced economic development in the earlier period. With the sustainable development of the economy and the transformation of urban planning and layout, it would change.



To clearly determine the high and low concentration areas of PM2.5 pollution, we drew a Getis-Ord Gi\* statistical graph for the study area during 2015–2019, as shown in Figure 5. On the whole, the cold spots in the study area were mainly distributed in the north of Shanxi and Hebei Provinces, and the eastern coastal areas of Shandong Province and the hot spots were mainly concentrated in the junction area of Hebei, Shandong, and Henan Provinces. In terms of temporal change, the cold spots gradually shifted from the northwest to the north of the study area, while those in the eastern coastal region of Shandong Province were composed of YT, QD, and WH with no change. Additionally, the hot spot moved to the southwest gradually from 2015 to 2019. This moving of the PM2.5 pollution center does not mean that the air quality in hot spots city were getting worse. In fact, almost all cities had been experiencing PM2.5 pollution alleviation at different levels. The PM2.5 concentration in some cities, such as SJZ, JN, and DZ, decreased sharply from hot spots to insignificant spots; some cities, such as JY, LYY, and PDS, declined slowly from insignificant spots to hot spots. This conversion of hot and cold spots is essentially determined by the transformation of the local industrial structure and the implementation of environmental protection policies. In fact, the upgrading and relocation of heavily polluting enterprises in the Beijing–Hebei–Tianjin region may also be one of the reasons for the moving of the pollution centroid. XT, HD, LC, AY, KF, PY, HB, XX, and other cities had always been hot spot cities during 2015–2019, indicating that the pollution in these cities was relatively serious and that control measures still needed to be taken for reducing the PM2.5 pollution risk level.

**Figure 3.** Probability density functions of each province during 2015–2019.

**Figure 4.** Moran scatter diagram from 2015 to 2019. (**a**) 2015; (**b**) 2016; (**c**) 2017; (**d**) 2018; (**e**) 2019.

**Figure 5.** Cold–hot spot diagram of PM2.5 concentration from 2015 to 2019.

#### *3.3. Analysis of Socioeconomic Influence Factors*

Different socioeconomic indicators reflect different human activities, which could affect the spatial and temporal heterogeneity of PM2.5 concentrations to various degrees. In this study, we used a spatial lag model (SLM) to determine the impact of various socioeconomic factors on PM2.5 concentrations. To ensure the data conformed to the normal distribution, a logarithmic transformation was performed on the socioeconomic data and PM2.5 concentrations before using SLM. Table 3 shows the quantified results of the SLM model from 2015 to 2019.


**Table 3.** Results of spatial lag model.

\*\*: Significant at 0.01 levels; \*: significant at 0.05 levels.

The spatial lag model introduced the spatial effect coefficient *ρ* to characterize the influence of PM2.5 levels from the surrounding areas on the local area. From 2015 to 2019, there was a positive relationship between PM2.5 concentration in local and surrounding regions, indicating that local PM2.5 levels were significantly influenced by surrounding areas. This is consistent with the "high–high" and "low–low" agglomeration characteristics of PM2.5 concentrations in the study area. Local PM2.5 pollution was not only related to local pollutant emissions but was also affected by pollution transport from other regions. Dong et al. [23] studied the pollution transmission contribution in the Beijing–Tianjin– Hebei region and the results showed 32.5% to 68.4% contribution of PM2.5 transmission in 2017. Local emission sources remain important contributors to the Beijing–Tianjin–Hebei region but the interactions between cities are also strong.

GDP represents the local economic development level. Except for 2016, GDP showed a significant negative correlation with the PM2.5 level, indicating that economic development had a certain inhibitory effect on PM2.5 pollution in the study area. As an economy grows, local investment in air pollution control will also increase. In addition, a relatively developed economy is conducive to effective integration and utilization of resources, affecting the local industrial structure and urban layout. Dong et al. [24] found that economic development and industrial upgrading were the main driving forces for haze pollution improvement in China's regions, while the transportation industry and construction industry were the two major sources of PM2.5 pollution. This is consistent with our findings, but other studies have shown different results. Yan, Kong, Jiang, Huang, and Ye [13] observed that the impacts of economic development on PM2.5 pollution varied with the degree of development. Economic development can alleviate PM2.5 pollution in developed areas, while it can promote PM2.5 pollution in underdeveloped areas. As noted by the theory of the Environmental Kuznets Curve (EKC), a later stage of urbanization is ultimately conducive to alleviating the pollution caused by the early stage of urbanization, and there is a threshold of an inflection point in the middle. Wang et al. [25] explained this in detail and obtained similar results to us.

Over 2015–2019, POP and PM2.5 levels showed a positive correlation, passing the significance test, indicating that population growth contributed to the formation of urban PM2.5 pollution. The increase in the population size resulted in growing demands for employment, housing, transportation, and energy consumption; thus, promoting the emission of pollutants. Han et al. [26] analyzed the relationship between population variations and PM2.5 levels, and the results showed that there was a positive trend between population and PM2.5 in most cities in China and that the contribution rate of megacities was 5.40 ± 4.80 μg/m<sup>3</sup> per million people. However, there was also a negative trend between population size and PM2.5 in some regions [13], because megacities with dense populations help to integrate resources and improve the utilization efficiency of urban infrastructure and natural resources, thus reducing PM2.5 pollution.

UP refers to the proportion of the urban population in the total population, which is usually used to represent the level of urbanization. The results of Table 3 indicate that UP had a positive impact on PM2.5 pollution in 2015 and 2017, but did not pass the significance test in other years. The growth or aggregation of an urban population usually leads to an increase in automobiles, housing and energy consumption, industrial production, and construction activities, which would have an impact on the increase in PM2.5 concentrations. Relevant studies [27] showed that the relationship between the proportion of the urban population and ecological environment pressures in the Beijing–Tianjin–Hebei region also conformed to the EKC theory, and it could effectively alleviate ecological environment pressure until it reached 80%, which was the turning point in EKC for most cities. By 2019, the proportion of the urban population in BJ and TJ exceeded 80%, while others were within the scope of 40–60%, below the threshold, indicating that we still have a long way to go in the urbanization process.

SI is the value-added of Secondary Industry and is used to represent the industrial structure. There was a significant positive correlation between SI and PM2.5 concentrations in 2015, 2017, and 2019. According to the statistical results of the output of the secondary industry, as shown in Figure S4, it had been decreasing or first increasing and then decreasing in AY, BJ, BD, LC, JNN, LF, PY, SJZ, TJ, and TA during 2015–2019, while it increased in other cities. These cities were often accompanied by severe PM2.5 pollution, which indicated that these cities may have already carried out the elimination of backward production capacity or the transfer of secondary industry to alleviate local PM2.5 pollution. The national development strategy has significantly increased the proportion of tertiary industries in the Beijing–Tianjin–Hebei region through the relocation and replacement of traditional secondary industries, which is consistent with our results. The results of Hao and Liu [28] are similar to ours, and they believe that PM2.5 concentrations in Chinese cities are also strongly influenced by secondary industry. In 2019, the average ratio of secondary industry to GDP in the study area was 41.97 percent. In addition, energy-intensive industries characterized by high emissions have a large-scale base, and the effect of industrial transformation and upgrading is not obvious in the short term. Therefore, to effectively reduce the level of urban PM2.5, it is necessary to accelerate the transformation of economic structures and reduce the dependence on secondary industries, especially heavy industries.

RD, road length per unit area, is often used to represent the impact of traffic factors on PM2.5. During the study period, there was a significant positive relationship between PM2.5 concentration and RD. According to the statistical results, as shown in Figure S4, the road length of most cities in the study area kept increasing in 2015–2019, except for BJ and TJ. A dense urban road traffic network promotes the increase in vehicle ownership, and pollutants from vehicle exhaust, such as NOx, are important sources of PM2.5 [29,30]. In addition, the increase in roads also enhances road dust, which is also a source of PM2.5 [31]. In this regard, traffic will continue to have an impact on continuing PM2.5 levels. There are also related studies [24] that use other indicators to characterize the influence of traffic factors and obtain similar results. Ding et al. [32] used per capita vehicle ownership to characterize traffic impacts, which determined that this factor had a driving effect on PM2.5 pollution and that it fluctuated during the study period.

In this study, BA and GR did not pass the significance test and were not statistically significant, so the results were not credible. BA is the ratio of the built-up area to the area of the municipal district. Due to the jurisdiction of the county, BA cannot completely represent the overall situation of cities in the research region. The GR of all cities was about 40% with slight distinctions. This may be why the results were not statistically significant. In addition, some studies used related indicators to explore the influence on PM2.5. For example, Wang, Yao, Xu, Sun, and Li [25] found an inverted U-shaped relationship between

built-up area and PM2.5 levels but lacked in-depth discussions. Qin et al. [33] simulated the impact of urban greening on atmospheric particulate matter, and the results showed that reasonable tree cover could reduce PM by 30%. In addition, there are still many deficiencies in this study. First, in addition to socio-economic factors, PM2.5 is also affected by topography, meteorology, pollution emissions, and other factors, which are not involved in this study. Secondly, the social and economic data used in this study are from various statistical yearbooks and bulletins, which may have certain deviations and bring certain uncertainties. In future studies, more factors should be considered to ensure the accuracy of the results.

## **4. Conclusions**

This study used PDFs to analyze the temporal variation trends and spatial distribution differences of PM2.5 concentrations in the Beijing–Tianjin–Hebei region and its surrounding provinces from 2015 to 2019. Then, the spatial distribution characteristics of PM2.5 concentrations were analyzed using Moran's I and Getis-Ord-Gi\*. Finally, SLM was adopted to quantify the driving effect of socioeconomic factors on PM2.5 levels. The main results were as follows:

(1) From 2015 to 2019, PM2.5 in the study area showed an overall downward trend. The Beijing–Tianjin–Hebei region and Henan Province decreased for the period of 2015 to 2019; Shanxi and Shandong Provinces expressed a variation trend of an inverted U-shape and U-shape, respectively. In a word, air quality in the study area had been improving from 2015 to 2019.

(2) From the perspective of spatial distributions, PM2.5 concentrations in the study area indicated an obvious positive spatial correlation with "high–high" and "low–low" agglomeration characteristics. The high-value area of PM2.5 was mainly concentrated in the junction of Henan, Shandong, and Hebei Provinces, which had a characteristic of moving to the southwest. The low values were mainly distributed in the northern part of Shanxi and Hebei Provinces, and the eastern part of Shandong Province.

(3) Socio-economic factor analysis showed that POP, UP, SI, and RD had a positive effect on PM2.5 concentration, while GDP had a negative driving effect. In addition, PM2.5 was also affected by PM2.5 pollution levels in surrounding areas.

Although PM2.5 levels in the study area decreased, PM2.5 pollution was still a serious problem until 2019. The significance of this study is to highlight the spatio-temporal heterogeneity of PM2.5 concentration distributions and the driving role of socioeconomic factors on PM2.5 pollution in the Beijing–Tianjin–Hebei region and its surrounding areas. Identifying the differences in PM2.5 concentration caused by socioeconomic development is helpful to better understand the interaction between urbanization and ecological environmental problems.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/atmos12101324/s1, Table S1: Names and abbreviations of cities in the study region, Figure S1: the percentage of exceeding standard days in each city from 2015 to 2019, Figure S2: PM2.5 concentration in each city and province from 2015 to 2019, Figure S3: Decreasing rate of PM2.5 concentration in 2019 compared with 2015, Figure S4: Statistics of social and economic factors in each city from 2015 to 2019.

**Author Contributions:** Data curation, C.F.; formal analysis, K.X.; investigation, J.W.; methodology, R.L.; project administration, J.W.; software, R.L.; visualization, K.X.; writing—original draft, R.L.; writing—review and editing, C.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ecology and Environment Department of Jilin Province. The project numbers are 2018-19 and 2019-08.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The datasets supporting the conclusions of this article are included within the article and its additional file.

**Conflicts of Interest:** The authors declare no conflict of interest.
