5.1. Spatiotemporal characteristics of tertiary carbon emissions
Due to inconsistency in the caliber of statistical data of each city, yearbook data of Guangdong Province and night light DN value of the Province were used for fitting, as shown in Fig. 2a.
It can be seen from the figure that DN value of night light hads a good fitting effect with carbon emission, and the fitting effect reached 0.993. The carbon emission formula used was,
Accuracy test: The DN value of night light in the Province during 1997-1999 was calculated and compared with carbon emissions of the Province in the corresponding years based on CEADs of China carbon accounting data. The error from this comparison was less than 10.78%. The night light DN values of Dongzhou, Shenzhen, Zhaoqing, and Dongguan Cities in 2000, 2007, 2014, and 2020 were collected. The fitting formula was then used to estimate carbon emission. Compared with yearbook data of the same caliber, the error of this estimation was less than 7.97%, which met the accuracy requirements at provincial and municipal levels, and hence the model was reliable.
As shown in Fig. 2b, carbon emissions of Pearl River Delta urban agglomeration showed a fluctuating upward trend during 2000-2020. The total carbon emission decreased briefly in 2005, 2008, 2012, and 2016, and the decrease in 2008 might be related to the global economic crisis. The industrial transformation and industrial structure of the agglomeration led to reduction of carbon emissions in 2012. The implementation of the concept and policy of green and low-carbon development and green economy led to temporary reduction in carbon emissions in 2016. Similarly, the COVID-19 outbreak in 2019 had a profound impact on cities, with companies and, industries shutting down and social gatherings restricted , leading to reduced carbon emissions.
As shown in Fig.2c, carbon emissions of most cities in Pearl River Delta had been continuously increasing, with some cities, including Foshan, Dongguan, and Zhongshan Cities, showing negative growth. The growth trend of the Delta cities from 2014 to 2020 tended to be slow; and this was in line with our emphasis on environmental protection, a series of emission reduction measures, steady economic development, saturating population density, and so on. Among the Delta cities, total amount of carbon emission in Guangzhou was the largest, (which hads been firmly having this largest emission), echoing the needs of its economic development. The carbon emission of Zhuhai and Zhaoqing Cities was relatively low, with the geographical location of Zhuhai City, not suitable for the development of secondary industries. The low level of economic development of Zhaoqing City was one of the main reasons for its carbon emission to be at the lower side .
Since both Zhongshan and Dongguan Cities had achieved full urbanization, the vector map did not divide them into counties. In order to be consistent with their county-level city units, the DN value of night light and carbon emission of these cities were divided by 4. As shown in Fig. 2d, carbon emissions of county-level cities in Pearl River Delta increased continuously from 2000 to 2020. The counties with the highest carbon emissions were mainly located in Shunde, Nanhai, Longgang, Baiyun, Huadu, and Panyu Districts of Guangzhou, Foshan, and Huizhou. The cities with low carbon emission were mainly concentrated in Fengkai, Huaiji, Guangning, and Deqing Counties of Zhaoqing City, Yuexiu District of Guangzhou City, and Yantian District of Shenzhen City, among which Yuexiu District had low carbon emission due to its small area and less number of enterprises and industries. From 2000 to 2014, increase of carbon emissions of county-level cities was mainly concentrated in the central and eastern parts of the Delta cities, namely Guangzhou and Huizhou, and the Huicheng District, Boluo County, Huidong County, Nanhai District, Shunde District, Baiyun District, and other counties with high carbon emissions. From 2000 to 2007, Futian, Liwan, and Luohu Districts had a small reduction in carbon emissions, while Boluo County, Sanshui District, Huiyang District, and Nanhai District had the largest increment. From 2007 to 2014, carbon emissions in Guangning County and Yuexiu District decreased continuously, with the largest increment in Xinhui, Zengcheng, Huiyang, and Huadu Districts. Futian, and Yuexiu Districts had reduced carbon emissions from 2014 to 2020, while Huidong County, Boluo County, Taishan City, and Huiyang District had the largest increase in carbon emissions.
In general, carbon emissions of Pearl River Delta, municipal and county-level cities showed a fluctuating increase in the 20 years. At the macro level, there was little difference in the distribution of carbon emissions at municipal and county scales. Therefore, in the formulation and implementation of carbon emission policies, municipal and county scales can be used as main and auxiliary scales, respectively, so as to achieve precise emission reduction from macro and micro levels.
5.2. Spatial simulation of carbon emissions at grid cell scale
In previous studies, carbon emissions were only numerical values or spatialized by province, city, or county, and the spatial resolution of carbon emissions was too low. According to 2.4, carbon emissions were assigned to grid cells to realize grid scale spatial simulation of carbon emissions with a large resolution of 1000*1000 meters. Fig. 3, directly displays spatial distribution characteristics and differences of carbon emissions in Pearl River Delta urban agglomeration. From 2000 to 2020, carbon emissions of cities in the Delta had been increasing, and spatial distribution of carbon emissions tended to the center and southeast regions from the surrounding ones. From 2000 to 2007, overall carbon emissions were low, showing a star-dot discrete distribution trend. From 2007 to 2020, carbon emissions kept increasing, and the spatial distribution gradually changed from star-dot and scattered to network and concentrated distribution. Small carbon emission points were absorbed by siphon effect of the city center and disappeared, and carbon emissions gradually concentrated. The center and southeast of Pearl River Delta urban agglomeration were carbon emission concentration areas, while the northwest, southwest, and southeast regions were low carbon emission concentration areas. Further research showed that spatial distribution of carbon emissions corresponded to distribution of enterprises, economic development, and population size., These factors were mainly concentrated in Guangzhou, Dongguan, Zhongshan, Huizhou, and Shenzhen in the central and eastern parts of the Delta, and indicated that urban carbon emissions were strongly correlated with them.
5.3. Spatial autocorrelation analysis of carbon emissions
5.3.1. Moran's I exponential spatial autocorrelation
ArcGIS was used to calculate the global Moran's I index of carbon emissions of 50 county-level cities in Pearl River Delta from 2000 to 2020, and the results are shown in Table 2.
Table 2 Moran's I index of carbon emissions of county-level cities in the Pearl River Delta
scale
|
variable
|
2000
|
2003
|
2006
|
2009
|
2012
|
2015
|
2018
|
2020
|
municipal
|
Moran′s I
|
0.059
|
0.082
|
0.112
|
0.119
|
0.127
|
0.155
|
0.145
|
0.139
|
Z
|
1.019
|
1.204
|
1.392
|
1.431
|
1.512
|
1.115
|
0.995
|
0.924
|
P
|
0.1640
|
0.122
|
0.094
|
0.089
|
0.078
|
0.137
|
0.163
|
0.178
|
county level
|
Moran′s I
|
0.222
|
0.248
|
0.270
|
0.273
|
0.271
|
0.287
|
0.314
|
0.323
|
Z
|
2.554
|
3.076
|
3.310
|
3.376
|
3.340
|
3.468
|
3.704
|
3.502
|
P
|
0.014
|
0.005
|
0.0033
|
0.002
|
0.001
|
0.002
|
0.001
|
0.003
|
As can be seen from Table 2, Moran's I index of carbon emissions of cities in the Delta increased from 0.059 in 2000 to 0.155 in 2015, indicating that carbon emissions of neighboring cities in the Delta had increased due to spatial aggregation effect. The industries and enterprises in the Delta cities were in a period of rapid development from 2000 to 2015, with booming industries, continuous gathering of people in the cities, and increasing resource consumption. This development made the spatial aggregation effect of carbon emissions in the Delta to continuously deepen. However, from 2015 to 2020, Moran's I index continued to decrease, which was related to the promotion of balanced regional development, low-carbon development, and upgrading of industrial institutions in various cities. In addition, development speed of more developed cities slowed down, and gap between cities gradually narrowed, thus reducing carbon emission gap among cities.
The Moran's I index of carbon emissions of county cities in Pearl River Delta were all greater than 2.5, and the P values were all less than 0.04, passing Z-value and P-value tests. Passing of these tests indicated that under a confidence interval of 96%, adjacent county cities in the Delta had a significant spatial aggregation effect of carbon emissions, and Moran's I index increased over time. The spatial aggregation effect tended to increase gradually.
By calculating the global Moran's I index for county-level cities in Pearl River Delta, we could only study and judge the spatial clustering relationship of overall spatial carbon emissions, while local Moran's I index could well interpret spatial clustering of carbon emissions in the Delta cities at municipal and county-levels. Therefore, using spatial autocorrelation analysis function of OpenGeoDa and constructing spatial weight coefficient matrix, local Moran's I index was calculated for carbon emissions of municipal and county-level cities in the Delta in 2000, 2007, 2014, and 2020. In addition, 999 random substitutions were carried out. The LISA cluster graph of Moran's I index was obtained subsequently. and the graph is shown in Fig. 4. The four phenomena in the graph represent high-value aggregation, high-value containing low-value anomaly, low-value aggregation, and low-value containing high-value anomaly, respectively. The carbon emission indicated that Guangzhou in 2000 was a high-high-cluster city, while Huizhou was a low-high-cluster city. In 2007, carbon emission of Huizhou increased gradually, and its cluster type became insignificant. In 2014, Zhaoqing City became a low-high cluster city, consistent with its economic development and population density. In 2020, Jiangmen transformed into a high-low cluster city. On the whole, high carbon emission concentration area was distributed in Guangzhou and Jiangmen Cities, while low carbon emission area was mainly concentrated in Zhaoqing City.
Fig 5. shows that in 2000, high-high-value carbon emission cluster of county-level cities was mainly concentrated in Foshan, Guangzhou, and Dongguan Cities, while low-low-value cluster was mainly concentrated in Fengkai, Huaiji, and Deqing Counties and Gaoyao District, consistent with ow level of economic development of Zhaoqing City and ow level of development of county-level cities. Pengjiang, Chancheng, and Liwan Districts showed low values including abnormally high values. In 2020, high-high carbon emission cluster of county-level cities shifted from the central region to the southeast of Huizhou, mainly to Huadu District, all districts of Huizhou City, and Dongguan City. The carbon emission of all counties and regions of Guangdong and Foshan Cities gradually became flat, with balanced development of all regions and reduced carbon emission differences, corresponding to balanced urbanization development of Guangdong and Zhongshan Cities. The low-low cluster included Huaiji County and Futian District; low values included high values distributed in Gaoming District, Longmen County, Pengjiang District, and Chancheng District; but there was no significant clustering feature in other areas. Although Huaji County was the only low-low cluster in Zhaoqing City, carbon emission of this City had not caught up with that of other municipal cities in the 20 years, reflecting that there was still a certain gap between this City and other cities in population and economic development.
5.3.2. Local outliers and spatio-temporal hotspots at the county level
The spatio-temporal cube was the basis for local outlier and spatio-temporal hotspot analyses. The local outlier analysis can find the location of the cluster and outlier that have statistically significant differences with their neighborhood in both space and time in the research area. Spatio-temporal hotspot analysis is to add the element of time to detect the location of cold and hot spots in spatio-temporal dimension. First, we extracted carbon emission data of each county city center in Pearl River Delta, and created a spatio-temporal cube through the aggregation point of county city center. The local outlier and spatio-temporal hotspot analyses of carbon emission of each county city in the Delta in the 20 years were conducted by taking one year as the neighborhood step and distance from city center as the step.
As shown in Fig. 6a, through the analysis of local outliers, it was found that carbon emissions of county-level cities in Pearl River Delta were not significant all the time. In particular, the Nansha District, Zhongshan City, Taishan City, Xiangzhou District, Guangming District, Bao 'an District, Futian District, Luohu District, Yantian District, Pingshan District, Longgang District, and Huidong County, accounted for 26% of the total emissions in these county-level cities. In the 20 years, there was no abnormal value in spac-ime between the carbon emission of these regions and that of their surrounding regions, and their emissions were always in the stable range. In addition, Huadu District, Sanshui District, Baiyun District, Nanhai District, Yuexiu District, Tianhe District, Huangpu District, Panyu District, Shunde District, Chancheng District, Gaoxing District, Shunde District, Conghua District, Zengcheng District, Dongguan City, Boluo County, Huicheng District, and Huiyang District were all high-high clustering in the 20 years, accounting for 36% of the total emissions . Most high-carbon emission county-level cities were mainly concentrated in Guangzhou in the new region of the Delta and Huizhou in the eastern region, consistent with the developed economy and dense population of Guangzhou and Huizhou. There wasis no clustering of only high-low outliers for carbon emissions. The only low-high anomaly clusters included Dinghu District, Pengjiang District, Jianghai District, Liwan District, Haizhu District, and Longmen County. With the development of cities, the increase or growth rate of carbon emissions of these cities was less than that of the surrounding cities. In other words, developing cities were surrounded by high carbon emissions of surrounding cities. Only Huaiji County, Fengkai County, Guangning County, Sihui City, Deqing County, Duanzhou District, Enping City, Kaiping City, Doumen District, and Jinwan District had low-low clustering. Also, most county-level cities were distributed in Zhaoqing City, which corresponded to the Moran's I index analysis and economic development of county-level cities. Compared with other county-level cities in the Delta, urban carbon emission of these 10 county-level cities had been in low value region in the 20 years, and carbon emission of the neighboring region was also in low value region. Among various types, there was one new association area, indicating that there had been various types of abnormal clustering in the new association area in the 20 years. The analysis of local outliers could directly reflect spatio temporal clustering anomalies of carbon emissions in county cities of the Delta in the 20 years, and also indirectly reflect the economic, population, and industrial changes of these 50 cities to a certain extent.
As shown in Fig. 6b, the spatio temporal hotspot analysis effectively showed complex spatial hotspots and development evolution rules of carbon emissions of county cities in the Pearl River Delta in the 20 years. Five patterns were identified for the emissions , among which the high-power area was a new hotspot, of carbon emissions from county cities in the Delta in 2020. Before this year, there was no statistically significant hot spot for carbon emissions in the high-power area, which indirectly reflected the rapid development of the high-power area in recent years and increase of its energy demand. The continuous hot spots were mainly distributed in the central and eastern parts of the Delta, including Dinghu District, Sanshui District, Gaoming District, Nanhai District, Chancheng District, Shunde District, Pengjiang District, Panyu District, Liwan District, Yuexiu District, Tianhe District, Huadu District, Huangpu District, Tianhe District, Haizhu District, Zengcheng District, Conghua District, Boluo County, Huicheng District, Huiyang District, Pingshan District, Yantian District, Yantian District, Luohu District, and Futian District. The results showed that these regions had been the places with high carbon emissions for the 20 years, consistent with local outliers and Moran's I index analyses. There was one hot spot in Dongguan, indicating that Dongguan had been a hot spot in the 20 years, clustering intensity of carbon emissions had increased every year, and the increase of carbon emissions was statistically significant. Huaiji, Fengkai, and Deqing Counties had permanent cold spots, which were located in the northwest of Zhaoqing City., The presence of these cold spots indicated that 90% of the carbon emissions of these three cities were cold spots in the 20 years of development, and there was no obvious trend to show that clustering intensity of carbon emissions increased with time, consistent with low-low clustering outliers of local outliers . Guangning County was one of the gradually decreasing cold spots, indicating that this county was a significant cold spot area for 90% of the time. But, the clustering intensity of carbon emissions every year showed a decreasing trend on the whole, with this intensity related to economic development and population growth, and the trend is statistically significant during the study period. The Jianghai region was the only oscillating hot spot, with the carbon emission in this region decreased first and then increased gradually.
On the whole, in the 20 years, the carbon emission of county cities in Pearl River Delta showed hot spots and high-value clusters in Guangzhou, Shenzhen, Zhongshan, and Huizhou, the central and eastern parts of the Delta showed hot spot expansion and high-value clustering, and the northwest region of the Delta showed low value accumulation and cold-point expansion. Indirectly, itthese observations indicated that overall gap of carbon emissions in county-level cities was constantly expanding, and clustering trend of high and low levels formed a fragmentation trend. In city-level urban units, carbon emissions were relatively clustered, and the gap was constantly narrowing.
5.3.3. Spatial inequality of multi-scale carbon emissions
As shown in Table 2 and 3, PD and GDP were used as indicators respectively to establish Theil indices TP and TG of overall carbon emission intensity of municipal cities. The Theil index and contribution of per capita carbon emission between municipal cities and within cities were represented by TbP, TWP, WbP, and WWP, and Theil index and contribution of intra-city and inter-city carbon emission intensity were expressed by TbG, TWG, WbG, and WWG, respectively. WjP and WjG were used to represent contribution rates of each city in intra-city differences of per capita carbon emission and carbon emission intensity, respectively.
In Table 3, the per -capita carbon emission index between cities had been above 0.2 for the 20 years, (gradually dropped to 0.361 from 2000 to 2020). The per -capita carbon emission in Pearl River Delta urban agglomeration had always been different in space and region , but the difference showed a trend of fluctuation and decline. The Thiel index of within each city was 10 times that of among city , and the difference did not change with time. This observation indicated that per capita carbon emission and carbon emission intensity differences of cities in the Delta were mainly caused by urban differences, and the per capita carbon emission differences were increasing. The contribution within cities was greater than that between cities, and less than that between cities from 2014 to 2020. The development of cities at municipal level was unbalanced, and there were obvious differences in per capita carbon emissions. However, emission differences within cities had been decreasing in the 20 years, while emission differences between cities had gradually increased, which indirectly reflected the development imbalance and per capita differences between cities at the municipal level. The difference of Foshan, Shenzhen, Dongguan, Jiangmen, and Huizhou Cities to the Delta was large, which was consistent with population distribution among the nine municipal cities. The average internal differences of Guangzhou, Jiangmen, and Zhaoqing were 0.410, 0.571, and 0.558, respectively, indicating that the per capita carbon emissions of these three cities had the greatest impact on the Delta. In terms of contribution rate, intra-city contribution was obviously greater than inter-city contribution, but this gap was constantly narrowing.
Table 3 Thiel index of city level cities based on per capita carbon emissions
time
|
TP
|
TbP
|
wbp
|
Guangzhou
|
Shenzhen
|
Zhuhai
|
Foshan
|
Huizhou
|
Dongguan
|
Zhongshan
|
Jiangmen
|
Zhaoqing
|
2000
|
0.5600
0.2632
0.4775
0.3613
|
0.0580
|
0.0770
|
0.0334
|
0.1945
|
0.0415
|
0.0482
|
0.0168
|
0.1086
|
0.0046
|
0.4042
|
2007
|
0.0104
|
0.0252
|
0.0238
|
0.0783
|
0.0755
|
0.0538
|
0.0203
|
0.0213
|
0.0057
|
0.3957
|
2014
|
0.0457
|
0.0311
|
0.0611
|
0.0879
|
0.1292
|
0.0489
|
0.0202
|
0.0638
|
0.0236
|
0.38419
|
2020
|
0.0321
|
0.0567
|
0.0332
|
0.0229
|
0.1676
|
0.0481
|
0.0221
|
0.1041
|
0.0422
|
0.58424
|
|
TwP
|
wwp
|
Guangzhou
|
Shenzhen
|
Zhuhai
|
Foshan
|
Huizhou
|
Dongguan
|
Zhongshan
|
Jiangmen
|
Zhaoqing
|
2000
|
0.57467
|
0.26381
|
0.02180
|
0.03357
|
0.02329
|
0
|
0
|
1.1282
|
0.21925
|
0.59578
|
2007
|
0.35516
|
0.22060
|
0.06072
|
0.07413
|
0.04582
|
0
|
0
|
0.0278
|
0.24472
|
0.60427
|
2014
|
0.36205
|
0.13940
|
0.11462
|
0.10003
|
0.07219
|
0
|
0
|
1.0848
|
0.28592
|
0.61581
|
2020
|
0.34821
|
0.14067
|
0.12168
|
0.12866
|
0.01874
|
0
|
0
|
0.0462
|
0.21100
|
0.41575
|
As per Table 4, the Thiel index of carbon emission intensity of municipal cities increased from 0.379 to 0.639, indicating that difference of carbon emission intensity of municipal cities was continuously strengthened. Compared with regional difference of per capita carbon emission, regional difference of carbon emission intensity was continuously enhanced, indicating that carbon emission of cities in Pearl River Delta urban agglomeration had a high matching degree with GDP. The differences within cities were significantly greater than the differences between cities. The differences within Guangzhou, Shenzhen, Foshan, and Zhuhai were increasing, while those among other cities were narrowing, indicating that carbon emission intensity within these four cities was increasingly unbalanced. The contribution between cities was much greater than that within cities, and luctuation of contribution was small, indicating that the impact of economic development between cities on carbon emissions was large and stable.
time
|
TG
|
TbG
|
wbG
|
Guangzhou
|
Shenzhen
|
Zhuhai
|
Foshan
|
Huizhou
|
Dongguan
|
Zhongshan
|
Jiangmen
|
Zhaoqing
|
2000
|
0.379141
|
0.0807
|
0.1557
|
0.0108
|
0.1105
|
0.0969
|
0.0279
|
0.0113
|
0.0651
|
0.0639
|
0.6145
|
2007
|
0.495376
|
0.0097
|
0.1202
|
0.0220
|
0.0538
|
0.1644
|
0.0429
|
0.0162
|
0.1048
|
0.1044
|
0.5646
|
2014
|
0.547416
|
0.0050
|
0.1069
|
0.0358
|
0.0364
|
0.1426
|
0.0418
|
0.0189
|
0.1349
|
0.0802
|
0.4881
|
2020
|
0.639172
|
0.0426
|
0.1126
|
0.0254
|
0.0345
|
0.2352
|
0.0390
|
0.0113
|
0.1877
|
0.1158
|
0.6149
|
|
|
TwG
|
wwG
|
|
Guangzhou
|
Shenzhen
|
Zhuhai
|
Foshan
|
Huizhou
|
Dongguan
|
Zhongshan
|
Jiangmen
|
Zhaoqing
|
2000
|
0.294702
|
0.280533
|
0.018596
|
0.030969
|
0.104569
|
0
|
0
|
0.038319
|
0.100565
|
0.3855
|
2007
|
0.526373
|
0.159987
|
0.114176
|
0.085459
|
0.250606
|
0
|
0
|
0.042627
|
0.086164
|
0.4354
|
2014
|
0.51541
|
0.284263
|
0.21535
|
0.090546
|
0.504137
|
0
|
0
|
0.052004
|
0.111576
|
0.5119
|
2020
|
0.555594
|
0.257206
|
0.289633
|
0.100261
|
0.206935
|
0
|
0
|
0.078446
|
0.12567
|
0.3851
|
As shown in Fig. 8, among the contribution rates of cities at municipal level, there were great differences in the impact of the nine cities on the carbon emission in Pearl River Delta. In terms of contribution rates of carbon emission and carbon emission intensity per capita, the rates and intensity of Guangzhou, Foshan, Jiangmen, Shenzhen, and Huizhou were much higher than those of Zhuhai, Dongguan, Zhongshan, and Zhaoqing in the 20 years. Notably, Guangzhou was the capital of Guangdong Province; Shenzhen was a special economic zone, a national economic center city, and an international city; and Foshan was the third largest city in the Province. Located in the hinterland of the Delta, Foshan was in a superior geographical position. Factors like this led to the huge population, booming economy, and leading demand for energy of the above five cities. This might be the main reason for the huge differences in the per-capita carbon emission and energy-intensity carbon emission contribution of these cities. The contribution of carbon emission intensity in Shenzhen and Foshan Cities had been decreasing over time, largely due to full urbanization and balanced development. The contribution of per capita carbon emission and carbon emission intensity in Huizhou and Zhaoqing Cities had been increasing continuously, indicating that population and economy of these two cities had developed greatly in the 20 years. But, there were large spatial differences in the development of cities, and the development of spatial regions was unbalanced.
In general, the Thiel index of per capita carbon emissions had been decreasing in the 20 years, while the Thiel index of carbon emissions intensity had been increasing. Specifically, regional differences of per capita carbon emissions were greater than regional differences of carbon emissions before 2007. After 2007, regional differences of per capita carbon emissions had been decreasing, while regional differences of carbon emissions intensity had been increasing. This indicated that Pearl River Delta urban agglomeration was still facing the challenge of regional differences in the reduction of per capita carbon emissions. In terms of carbon emission contribution, contribution of per capita carbon emission had always been higher than that of carbon emission intensity. The per capita carbon emission had a greater impact on total carbon emission of cities. In addition, per capita carbon emission and Theil index of carbon emission intensity of all municipal cities decreased from 2014 to 2020. This might be related to industrial restructuring, green urban development, revision of the Environmental Protection Law of the People's Republic of China, and the 12th Five-Year Plan to strengthen energy conservation and emission reduction to achieve low-carbon development.
5.4. Analysis of influencing factors of carbon emission
According to 4.7, to adapt the independent variable of geographic detector as type variable, the impact factors were classified by natural discontinuous point classification method in ArcGIS., Accordingly, four factors including economic development level (GDP), population density (PD), number of enterprises above designated size (IL), and gross industrial output value (GV) were selected for the analysis based on [9, 35]. The influencing factors of carbon emissions in Pearl River Delta were analyzed from the perspective of spatial differentiation using geographic detectors. The influence value q of a single factor on spatial heterogeneity of carbon emissions is shown in Fig 8.
From 2000 to 2020, carbon emissions of Pearl River Delta urban agglomeration were mainly affected by level of economic development, number of enterprises above designated size, and total industrial output value. On the contrary, population density had the least effect on the emissions. The q value of gross industrial output value wasis from 0.28 at the beginning to 0.89 in 2020, and its explanatory power kept increasing, confirming that the gross industrial output was one of the leading factors. The explanatory power of GDP on spatial heterogeneity of carbon emissions had been relatively stable, with the corresponding q value between 0.7 and 0.9. Economic development exerted a leading and stable effect on carbon emissions in Pearl River Delta urban agglomeration. The q value of population density increased the fastest and was also in a stable range from 2006 to 2020, which was related to population distribution of county-level cities in the Delta. The population of the Delta was mainly concentrated in economically developed and narrow central and eastern areas and coastal areas. Even though the northwestern area was vast, the population was attracted by the "siphon effect" of the central and eastern areas. This attraction led to a large population density gap in county-level cities. The number of enterprises above the scale had occupied more than 50% in the early period of the study duration, which was related to China's vigorous development of real economy and promotion of our economic development in 2000, but could only occupy about 25% in the late period. The adjustment of industrial institutions, improvement of energy utilization level, and innovation of science and technology had greatly improved the efficiency of energy use during the study period. Also, the concept of synergistic development between man and nature, implementation of carbon emission reduction measures, and transformation and development of enterprises had weakened the explanatory power of the number of enterprises on carbon emissions. The influence of industrial output value increased from less than 20% to about 35% during the study period of 2000-2020 . During 2000-2003, industrial output value of Pearl River Delta was not high, while 2004-2010 was a period of rapid industrial development, which also resulted in a sharp increase in industrial energy demand and increased energy carbon emissions. In addition, energy demand of industrial production accounted for a large part of urban energy, which could explain why impact of industrial output on urban carbon emissions had accounted for about 35% in the 20 years.
The interactive exploration and analysis of the four types of impact factors were carried out by geographic detector, and the results are shown in Fig. 9. The interaction of any two factors was double-factor enhancement or nonlinear enhancement, that is, the interaction of any two factors had a greater impact on carbon emissions of Pearl River Delta urban agglomeration than that of a single factor [45], and the greater the value, the greater the impact. The interaction factor between population density and GDP had been among the highest values from 2000 to 2020, indicating that growth of population and GDP would greatly increase the demand for urban energy, that is, the impact on urban carbon emissions was among the highest. The value of population density as a single factor was not high, but in the interaction, q value of population density increaseds significantly and was in the high value area. Among influences of single factors, factors with low q value greatly increased after interaction, which indicated that interaction of population density, GDP, number of enterprises, and total industrial output had a spatial superposition effect on carbon emissions in Pearl River Delta urban agglomeration. In 2000, 2007, 2014 and 2020, the largest q value was GDP∩IL, IL∩GV\GDP, GDP∩IL\PD, and GDP∩IL, respectively. Therefore, the interaction between number of enterprises above the scale and GDP was the main driving force to promote carbon emission of Pearl River Delta urban agglomeration. The single factor q value of the total industrial output value was in the maximum line, while the q value of its interaction with various influencing factors was always at the low end. The development of emerging and high-precision technologies might require little energy consumption and bring great industrial output value, and the industrial development would be more energy-efficient and environmentally friendly, and was not particularly closely related to the number of large-scale enterprises and population density. The explanation of the interaction effect of total industrial output was not strong. As the population, economy, and science and technology center of Guangdong Province and an important region in greater Pearl River Delta region, Pearl River Delta urban agglomeration will require more energy for future development. While maintaining development, it can make policy adjustments to industrial structure, industrial distribution, and population distribution, achieve the goal of green development and low-carbon development, and strive to achieve carbon neutrality as soon as possible.