Regional spatial patterns and influencing factors of Haze Pollution in the Pearl River Delta region

Based on the Pearl River Delta data from 2006 to 2016, this paper uses exploratory spatial data analysis technology to analyze the spatial correlation and spatial cluster characteristics of haze pollution, and uses the Spatial Durbin Model to analyze the influencing factors of haze pollution. The study finds that there is a certain spatial autocorrelation in the spatial distribution of haze pollution in the Pearl River Delta region. The direct, indirect and total effects of energy consumption on haze pollution are not significant, there is an inverted U-shaped relationship between economic development level and haze pollution, population size and industrial structure will increase haze pollution in the local and surrounding areas, technological innovation in the region has a significant impact on local haze pollution and has no significant impact on it in the surrounding areas, and the impact of FDI on haze pollution is not stable in the Pearl River Delta region.


Introduction
As one of the most dynamic economic zones in the Asia-Pacific region, the Pearl River Delta (PRD) region is an advanced manufacturing and modern service industry base with global influence [1]. It is also the country with the largest population concentration, the strongest innovation capability and the strongest comprehensive strength [2]. However, with the increasing of the regional economic development and economic strength, the PRD region is also facing serious environmental pollution problems, especially the PM2.5 (Fine Particulate Matter) and PM10 (Inhalable Particles) often appear since 2013, which seriously threatens social and economic development and people's health.
In response to air pollution, the Guangdong Provincial Government promulgated the regional air pollution guidance document "Guangdong Pearl River Delta Clean Air Action Plan" in 2010, and promulgated the air pollution joint control plan "Guangdong Province Pearl River Delta Air Pollution Prevention and Control Measures", and striving to make the air quality reach the livable environmental standard by the end of 2020 [3]. Under the guidance of this strategy, air pollution control in the Pearl River Delta region has achieved phased results. In 2015, the annual average concentration of PM2.5 in the region decreased from 120 micrograms per cubic meter at the highest peak to 34 micrograms per cubic meter, meets the national secondary standard and the World Health Organization level standard, and exceeds the national assessment target two years ahead of schedule [4]. Based on the above, by using exploratory spatial data analysis (ESDA), the paper aims to explore the spatial pattern of haze pollution and the interaction characteristics between neighboring regions and to examine the

exploratory spatial data analysis (ESDA)
The exploratory spatial data analysis method can be used to describe the spatial distribution characteristics and spatial correlation of each variable, which mainly includes Global Spatial Autocorrelation and Local Spatial Autocorrelation, and equations 1 and 2 show the calculation methods of the two, respectively.
(2) Where I and Ii are global and local spatial autocorrelation values, n is the number of spatial unit data, Xi and Xj represent the attribute values of spatial units i and j, respectively, and Wij is the spatial weight coefficient matrix, indicating the spatial unit neighboring relationship. This paper adopts geographic adjoining spatial weight matrix(W1) and geographic distance matrix(W2).

Model Building and Variable Selection
Based on the IPAT model, environmental pressure, population, wealth and technical level, energy consumption (EC), the total population at the end of the year (PO), the per capita GDP, industrial structure, technological innovation level(R&D) and foreign direct investment (FDI) are considered in the model. The panel data of this model is set as follows: Where, i represents provinces; t represents the period, lnPM10 refers haze pollution. LnPM10 means haze pollution. WlnPM10it is the spatial lag term of the explained variable (haze pollution), ∂ 3 means the spatial lag coefficients. wlnECit, wlnPOit, wlngdpit, wln2gdpit, wlnISit, wlnFDIit, wlnR&Dit are the spatial lag term of explanatory variables. , λ, are the area effect, time effect and random disturbance term respectively. obeys normal distribution. is the spatial weight matrix. lnEC, lnFDI, lnIS, lnPO and LnR&D means energy consumption, foreign direct investment, industrial structure, population size and technological innovation level respectively. Lngdp and ln 2 gdp represent the first and second terms of per capita GDP respectively. PM10 is used as a measure of haze pollution and PM2.5 is used as a confirmatory analysis, and the data comes from the report of "Guangdong-Hong Kong Pearl River Delta Regional Air Monitoring Network". Energy consumption is measured by total coal consumption in each region, and the actually amount of used foreign capital of each province is used to measure the level of investment introduction [1]. Industrial structure is measured by the proportion of the secondary industry in GDP, population size and technological innovation level are measured by the total population and research and experimental development funding respectively [1]. Gross Domestic Product (GDP) is indicated by actual GDP per capita. The data in this paper are mainly from the "Guangdong Province Statistical Yearbook".

Analysis of Factors Affecting Haze Pollution in the Pearl River Delta
It is found that the hausman statistical value is 12.7633 and the degree of freedom is 15, P The value is 0.6295, which indicates that the random effect model is more explanatory power, the random effects model is used to conduct empirical analysis, and the Spatial econometric test results are listed in Table  1. Note: ****，***, ** and * indicate the significance at 1%, 5%, 10% and 20%level, respectively.
First, the coefficients of W*dep are 0.4079 with the geographic adjoining spatial weight matrix (W1) and 0.3600 with the geographic spatial weight matrix(W2), both pass the significant test at 10%, which indicates that haze pollution has significant spatial correlation characteristics. Second, the impact of energy consumption on haze pollution is significant in the model with geographic adjoining spatial weight matrix, but the impact is not significant in the model with geographic distance weight matrix model, which indicates that the impact of energy consumption on haze pollution is unstable. Population size has a significant effect on haze pollution at the 1% significance level in the case of W1 and W2, and population size has increased the haze pollution from the coefficient symbol. Per capita GDP, FDI and technological innovation have an important effect on haze pollution at the 10% significance level in the case of W1 and W2, from the coefficients symbol it is found that Per capita GDP increased haze pollution, but FDI and technological innovation effectively inhibited haze pollution. Industrial structure has an effect on haze pollution at a 20% significance level. Table 2 showed the direct, indirect and total effects of Spatial Durbin Model. The direct, indirect and total effects of energy consumption on haze pollution are not significant in the case of W1 and W2. The direct, indirect and total effects of population size on haze pollution are significant at the 10% significance level in the case of W2, the total and indirect effects are not significant in the case of W1, and the direct effect is significant in the case of W1. The direct, indirect and total effects of per capita GDP and the square of per capita GDP on haze pollution are significant, and the level of economic development in the region has an important impact on haze pollution in the region and surrounding areas. From the coefficient symbol, per capita GDP increased haze pollution, but the square of per capita GDP inhibited haze pollution in the local and surrounding areas in case of W1 and W2, which means that there is an inverted U-shaped relationship between economic development level and haze pollution. The direct, indirect and total effects of industrial structure on haze pollution are significant at the 20% significance level in the case of W1 and W2, which means industrial structure will increase haze pollution in the surrounding areas. The total and direct effects of FDI on haze pollution are not significant in the case of W1and W2, which means the effect is not stable. The direct and total effects of technological innovation on haze pollution are significant, and the indirect effects are not significant in the case of W1 and W2, which means technological innovation in the region has a significant impact on local haze pollution and has no significant impact on it in the surrounding areas.

Robustness test
The robustness test was performed by using pm2.5 as surrogate variable, and table 3 showed the results. As can be seen from Table 3, the coefficients of W*dep are all significant at the level of 1%, the total effects of energy consumption, population size, industrial structure, FDI and technological innovation on haze pollution are significant in the case of W1 and W2,and the results of the empirical study are verified again.  112.9191 124.4409 Note: ****，***, ** and * indicate the significance at 1%, 5%, 10% and 20%level, respectively.

Conclusions
Based on the Pearl River Delta data from 2006 to 2016, this paper uses exploratory spatial data analysis technology to analyze the spatial cluster characteristics of haze pollution, and uses the Spatial Durbin Model to analyze the influencing factors of haze pollution. The study finds that the global Moran's I value are positive and ranges from 0.04666 to 0.4751, except the years of 2007, 2008 and 2012, and the coefficients of W*dep pass the significant test at 10%, which indicates that haze pollution has significant spatial correlation characteristics. The effects of energy consumption on haze pollution are not significant, and the effect of FDI on haze pollution is not stable. The direct, indirect and total effects of per capita GDP and the square of per capita GDP on haze pollution are significant, the level of economic development in the region has an important impact on haze pollution in the