Impervious surface (IS) is a key indicator to measure the urbanization process and ecological environment. Many studies have observed an urbanization process based on IS at the city scale. Understanding the changes in the IS over a period at a regional level offers an alternative and effective approach to characterize and quantify the spatial process of urban agglomeration. This study focuses on the urban agglomeration of the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) by utilizing the sensor-based Landsat data during 1987-2017 and investigates the spatiotemporal distribution of IS expansion at both regional and city scales. The modified linear spectral mixture analysis (MLSMA) method is used to extract the IS of the GBA. Then, the IS mapping accuracies were assessed after comparison with high-resolution historical data. The spatiotemporal and directional changes of IS surfaces for GBA are analyzed by using Gravity Center (GC) and Standard Deviational Ellipse (SDE). Finally, Shannon’s Diversity Index (SHDI) is used to analyze the overall characteristics of landscape level, and the Patch Density (PD) and Landscape Shape Index (LSI) are used to describe the characteristics of different classes of the IS. The results show that the IS of the whole region experienced rapid and massive expansion during the past 30 years and exhibited a distinct characteristic along the Pearl River Delta (PRD) and the coastline. Furthermore, the IS area increased rapidly in the PRD, while it is relatively stable in Hong Kong and Macao. We believe that the findings of this study can help policy makers to better understand and maintain the sustainable development of the GBA.
In 1950, 30% of the world population resided in the urban areas due to global urbanization which has progressed with unprecedented speed and thus resulted in the increased urban population to 54% in 2014 [
Since the 1980s, a substantial amount of research has focused on understanding the urban agglomeration spatial structure to cover issues such as the spatial expansion mode [
Furthermore, understanding the long-term land expansion and spatial structure change in various geographical scales and locations is better to understand the historical development and the future trend. Therefore, it is necessary to investigate the spatiotemporal IS change of urban agglomeration in various scales. The IS greatly effects the regional climate [
GBA is one of the major bay areas in the world, and it has been experiencing an intensive urbanization process since the reform and opening-up policy in 1978 [ Comparing the extension differences between PRD and two special administration regions by calculating the IS growth speed and rate Quantifying the spatiotemporal structure evolution of three metropolitan areas in GBA by using the GC and SDE analysis
Finally, the similarities and differences of spatial and temporal change at the landscape pattern of core cities in GBA were compared by using Shannon’s Diversity Index (SHDI), Patch Density (PD), and Landscape Shape Index (LSI).
The organization of the rest of the paper is as follows: Section
To strengthen the cooperation between the Chinese mainland and Hong Kong-Macao and establish a new development pattern, GBA was proposed in the government work report of the 12th National People’s Congress [
Study area.
The remote sensing images utilized in the current study are acquired by Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 OLI sensors over the period 1987 to 2017 with the 30 m spatial resolution. The administrative boundary of the GBA is located within eight scenes (path/row: 121/44, 121/45, 122/43, 122/44, 122/45, 123/43, 123/44, and 123/45). The datasets used in the study are the surface reflectance dataset provided by USGS [
In the current study, we used historical high-resolution images acquired from Google Earth [
In the current study, a modified linear spectral mixture analysis (MLSMA) method proposed by Xu et al. [
Mapping the dynamics of the impervious surface and measuring orientation and direction of IS in the study area is important to understand the spatial and temporal change characteristics of urban development during the past 30 years. The change of the IS area is mapped using the MLSMA in three steps, as shown in Figure Calculate the modified normalized difference water index (MNDWI) and then mask the water body of Landsat images by the result of MNDWI. The threshold values of the MNDWI for 56 Landsat images of different times were experimentally determined in this study Extract the fractions of high and low albedo, vegetation, and soil by the conventional LSMM method Extract the IS, vegetation, and soil fractions with the unmixing results of the conventional LSMM, Normalized Difference Built-up Index (NDBI), and Normalized Difference Vegetation Index (NDVI). The threshold values of NDBI for 56 Landsat images of different times were experimentally determined in this study. A previous study has shown that pixels with a NDVI value less than 0.2 are classified as bare soil [
The flowchart of mapping the IS using Landsat data by MLSMA.
The IS mapping accuracies were assessed after comparison with high-resolution historical data acquired from Google Earth, Institute of Surveying and Mapping, Department of Natural Resources of Guangdong Province and Guangzhou Jiantong Surveying, Mapping, and Geoinfo Co., Ltd. The available historical images from Google Earth are limited and cannot cover the whole study area before 2007. Thus, the data of 2012-2017 are used to assess the accuracy of the contemporary IS classification result. The three steps involved to assess the accuracy are as follows: Firstly, we carried out the space registration of reference and classified. Secondly, we collected 500 IS testing samples and 500 pervious surface testing samples in each city and the whole region randomly by stratified random sampling [
Confusion matrix model.
The higher the value of sensitivity and specificity, the smaller the probability of misclassification and omission of IS, although the precision of IS during a few periods cannot be assessed due to the absence of high-resolution images. Thus, it is assumed that they are similar to the result of 2012-2017.
Detecting the spatiotemporal and directional changes of the IS can provide important information for optimizing the regional planning and management. In the current study, the spatiotemporal and directional changes of IS on three metropolitan areas Guang-Fo-Zhao, Shen-Guan-Hui, and Zhu-Zhong-Jiang, and three development poles Guangzhou-Foshan, Hong Kong-Shenzhen, and Macao-Zhuhai in the bay area are analyzed by using GC and SDE. The GC is employed to identify the weighted center of geographic elements [
The gravity center is calculated as follows:
The azimuth “
The standard deviations of the ellipse “
By using the equidistant division [
In the current study, the landscape pattern analysis is adopted to indicate the IS features and the difference in the biophysical composition of the same IS category to better understand the difference of spatiotemporal characteristics among the core cities. To analyze the overall characteristics of landscape level, SHDI is used, and the PD and LSI are used to describe the characteristics of different classes of the IS [
Overview of landscape indices used in the current study.
Scale | Name | Range | Implication |
---|---|---|---|
Class level | PD | Representing the complexity of spatial structure, the higher PD is, the more fragmental the class is [ | |
LSI | In the measurement of patch aggregation, the larger LSI is, the more discrete the patches are [ | ||
Landscape level | SHDI | Representing the richness and complexity of the land surface, the higher the value, the higher the complexity of the whole landscape [ |
Based on the sensor-based Landsat images, the distribution of the IS surface is extracted for 1987, 1992, 1997, 2002, 2007, 2012, and 2017. To calculate the overall accuracy, sensitivity, and specificity, high-resolution images from Google Earth are selected to assess the classified accuracy and the reliability of the IS mapping for 2012-2017. All accuracy metrics of the classification were greater than 85% in 2012 and 2017 (Table
A summary of the assessment of the classification’s accuracy.
Date | Overall accuracy | Sensitivity | Specificity |
---|---|---|---|
2012 | 88.64% | 87.71% | 88.02% |
2017 | 88.27% | 88.67% | 87.53% |
Average | 88.45% | 88.19% | 87.78% |
Figure
IS area distribution from 1987 to 2017.
Figure
The IS area (a), growth rate (b), and speed (c) of the IS of the bay area at different periods.
The IS area of GBA increased significantly (see Figures
The IS area (a), growth rate (b), and speed (c) of the IS of the cities in the bay area at different periods.
IS expansion for the cities in the GBA from 1987 to 2017.
IS expansion of the cities in GBA is mapped in Figure
In Hong Kong and Macao, the two special administration regions, as a geographical environment restricts the urban expansion, IS changing was dominated by internal adjustment. The development and expansion of IS mainly occurred from 1987 to 1997 and then developed steadily, and the structure of these two cities becomes stable. Moreover, the growth rate and speed of IS in these two cities are lower than those in the other nine cities in mainland China.
In Shenzhen and Zhuhai, the IS area was primarily distributed at the intersection with Hong Kong and Macao in the early period, then expanded to multicore. From 1987 to 1997, the IS increased with a high growth rate and speed, which in turn generated the core of the city. However, the growth rate and speed of IS began to decline after 2002, and the structure of Shenzhen tends to stabilize. Moreover, the IS increased at the intersection with Zhuhai during 1987-2002. From 2007 to 2017, the IS experienced internal growth and almost covered the whole city. As a special economic zone, the IS in Zhuhai mainly appeared at the intersection with Macao. The IS expanded significantly with a high growth rate and speed from 1987 to 2002.
Foshan and Guangzhou showed an obvious development trend of the core region driving the edge region. In Foshan, the growth rate and speed of the IS showed “M”-type development, while after 2002 it increased significantly as compared to the period of 1987-2002. Moreover, the IS was mainly distributed along the border between Guangzhou and Foshan in the early period, which then expanded to the whole city. From 1987 to 2002, most of the IS can be observed in Guangzhou and extended near the deviation of the PRD, where both the growth rate and speed of IS decreased, while after 2007, the IS expanded to cover the whole city, except the mountain and forest region. Furthermore, the growth rate and speed of the IS continued to increase, then decreased a little in 2017.
It can be observed that in Dongguan, IS experienced expansion based on multicore in the early period due to the policy of “Rural Economy.” The growth rate and speed of the IS experienced a “down-up-down” trend, while after 2002, the growth speed became faster and the urban construction was mainly around the core region, with the region closed to Shenzhen expanding rapidly. The urban expansion of Huizhou, Jiangmen, Zhaoqing, and Zhongshan presented a trend of the existing core regions driving the construction and expansion of new town. In Huizhou, the growth rate of the IS experienced a “down-up-down” trend, and the growth speed showed “W”-type development. The IS increasingly spread from the middle and southwest region to the western part of the city during 1987-2002. After 2002, northwest expansion can be observed, and the growth speed increased significantly. However, the growth rate and speed of IS in Jiangmen fluctuated with a high margin of growth speed and the IS clusters were dispersed during 1987-1997. Furthermore, the IS concentrated in the northeastern region and then expanded to the whole city. In Zhaoqing, the growth rate and speed increased firstly and then decreased, and the growth speed after 2002 was significantly faster. However, the increase in IS concentrated in the northwestern and southeastern regions of the city, while the growth rate of the IS in Zhongshan tends to decrease gradually but growth speed experienced an “up-down” trend.
Based on the PRD integrated development zone, Hong Kong and Macao were included in the urban agglomeration of GBA. However, due to the different administrative nature, the IS change differs, as shown in Table
The growth rate and speed of IS in the Pearl River Delta and two special administration regions (Hong Kong and Macao) from 1987 to 2017.
Period | Pearl River Delta | Hong Kong and Macao | ||
---|---|---|---|---|
Growth rate (%) | Speed (km2/a) | Growth rate (%) | Speed (km2/a) | |
1987-1992 | 14.64 | 255.17 | 6.61 | 6.34 |
1992-1997 | 4.97 | 150.16 | 2.03 | 2.59 |
1997-2002 | 6.29 | 237.23 | 3.07 | 4.32 |
2002-2007 | 7.86 | 389.60 | 4.07 | 6.61 |
2007-2012 | 8.93 | 616.56 | 2.87 | 5.60 |
2012-2017 | 4.27 | 426.73 | 3.68 | 8.21 |
The nine cities in the PRD have completed the transformation from the traditional agricultural economic regions to the important manufacturing centers. The IS of Hong Kong and Macao increased with an annual average speed of 3.72 km2/a, and they expanded mainly on the basis of the old city due to the limitation of administrative size; the growth speed of IS from 2002-2017 is slightly higher than that from 1987-2002. Moreover, The IS experienced rapid growth with a speed of 8.21 km2/a during 2012-2017. In the beginning, the urbanization processes are totally different between the PRD and special administration regions due to the different political systems. However, after implementation of “Closer Economic Partnership Arrangement (CEPA)” [
Coordinated development of urban agglomeration has become the major impetus of global economic. Development and construction of GBA have become the national development strategy of China. We have found that there is a huge difference of urban evolution between the PRD and special administration regions. Constructing and developing GBA based on the differentiation of political institution and economic foundation have a significant impact on China’s economic development.
In mainland China, around the GBA, there are three metropolitan areas: (1) Guang-Fo-Zhao: Guangzhou, Foshan, and Zhaoqing; (2) Shen-Guan-Hui: Shenzhen, Dongguan, and Huizhou; and (3) Zhu-Zhong-Jiang: Zhuhai, Zhongshan, and Jiangmen. To reveal the spatiotemporal characteristics of IS, the increasing area and the growth speed from 1987 to 2017 are calculated and presented in Figure
IS expansion of three metropolitan areas in GBA from 1987 to 2017.
To further explore the spatiotemporal change of IS for these three metropolitan areas, the SDE and GC of IS during 1987-2017 are calculated and are shown in Figure In Guang-Fo-Zhao, from 1987 to 2002, the azimuth is distributed at 110°. However, from 2002 to 2007, the azimuth shifted to 114.04° with an increased speed of 0.85°/a, which indicates that the IS increased with a directivity trend towards northwest-southeast orientation. Moreover, the azimuth maintained around 112° after 2012. Furthermore, the flattering ratio first increased and then decreased, indicating that the IS area is distributed mainly along the long axis. There was a weak polarization phenomenon towards the short axis after 2012. The areas of ellipse deviation increased volatility over time; this indicated that the imperious surface continued to expand outward. It can also be observed that the IS GC moved from the border between Guangzhou and Foshan towards the border between Foshan and Zhaoqing in the past 30 years, which is mainly close to the geometric center. Thus, the GC moved 31.73 km northwest. This trend also indicates that the IS of Guang-Fo-Zhao increased mainly in the downtown of Guangzhou then extended towards the border between Guangzhou and Foshan in the 1990s. Finally, after the implementation of the Guang-Fo city plan in 2002 [ In Shen-Guan-Hui from 1987 to 2002, the azimuth shifted and thus decreased from 73.16° to 70.74°. There is a significant counter-clockwise shift, where the azimuth decreased to 55.74° in 2007. This indicates that the orientation of IS distribution changed from east-west to northeast-southwest. However, the azimuth is maintained at 72° from 2012 to 2017, which shows that the spatial direction of IS expansion is towards east-west again, but this trend is gradually weakening. Moreover, the flattering ratio shows an M-type decreasing trend from 1.21 in 1987 to 1.15 in 2017, which reflects an uncertain principle expansion direction of the IS. Furthermore, the IS GC laid between Huizhou and Dongguan border and moved 3.16 km towards the northeast. In order to overcome the limitation of available space in Shenzhen, the Shen-Guan-Hui metropolitan area was established so that Dongguan and Huizhou can provide developed space to Shenzhen [ In Zhu-Zhong-Jiang, the azimuth experienced an increasing-decreasing trend, and the azimuth shifted to 76.93° at an increasing speed of 0.64°/a from 1987 to 2007, which indicated that the IS expanded towards northeast-southwest orientation. However, this spatial structure started to decrease with 73.55° azimuth in 2017. Moreover, the flattering ratio presents a decreasing-increasing-decreasing trend from 1.52 to 1.82. Furthermore, the GC for Zhu-Zhong-Jiang is concentrated in Jiangmen, which surrounded the geometric center in U-type distribution, and the GC moved to the northeast with the movement of 13.21 km. There is no core city in Zhu-Zhong-Jiang and is only considered a geographical combination. The orientation of IS distribution moved from the northeast-southwest with the development of Macao and Zhuhai. Also, IS expansion is affected by the geomorphology and environment of the coastal region.
SDEs and GCs of IS in the metropolitan areas.
Generally, the spatial and temporal patterns of metropolitan areas in GBA are different. The IS distribution of Guang-Fo-Zhao and Shen-Guan-Hui has a high degree of aggregation, but IS distribution of Zhu-Zhong-Jiang is relatively dispersed. Zhu-Zhong-Jiang has natural and ecological advantage among these three metropolitan areas. The economic scale and industrial innovation strength of Guang-Fo-Zhao and Shen-Guan-Hui are in the domestic leading position. Multicore and group developing weaken the boundaries of cities in GBA, and the boundaries between downtowns and rural regions in GBA have a fusion trend. In the future, these three metropolitans should take industrial development and environmental protection into account, to solve the uneven resource allocation and urban development, and ultimately achieve the integrative development of GBA.
In order to improve the integral strength and global influence of GBA, three development poles (namely, Guangzhou-Foshan, Hong Kong-Shenzhen, and Macao-Zhuhai) were created for the GBA [
IS expansion for the three development poles in GBA from 1987 to 2017.
The SDE and GC of IS on three development poles during 1987-2017 are calculated and shown in Figure During 1987-2017, the azimuth decreased from 52.76° to 51.17°, which indicates that the IS of Guangzhou-Foshan is distributed towards the northeast-southwest. The flattering ratio increased during 1987-2017, which indicated that the IS expanded towards the long axis. The area of SDE increased overtime, revealing that the IS experienced rapid development. The GC of the IS was mainly situated towards the border of Guangzhou and Foshan, and the GC moved 5.29 km towards the southeast. However, near the downtown of Guangzhou and Foshan, the increasing IS is distributed. The spatial structure of Guangzhou-Foshan urban integration became stable after 2012; then, the IS started to extend towards the coastline In Hong Kong-Macao, the azimuth shifted from 125.07° to 85.69° with a decreasing angle of 39°, which revealed that the orientation of the distribution of the IS during 1987-1992 shifted from northwest-southeast to east-west. However, the azimuth increased to 115.69° in 1997, which revealed that the IS shifted towards northwest-southeast. Moreover, the azimuth experienced an increasing trend indicating that the northwest-southeast spatial structure was enhanced during 2002-2017. The flattering ratio decreased from 1.27 to 1.10 overtime; in other words, the short axis length was increasing closer to the long axis length, and the area of SDE increased over time, which reflected the increasing diffusion degree of IS. The GC of Hong Kong-Macao is mainly distributed in Shenzhen towards the north of the geometric center of the development pole and moved towards the southeast with a distance of 3.30 km. In general, the increased IS was mainly distributed along the border of Hong Kong and Shenzhen; the IS of Shenzhen experienced a multicenter and leap forward development because of the flat terrain. However, Hong Kong, as a multi-island city, experienced an expansion by leaps and bounds without the obvious directivity, thus making it hard to understand the increase in the directivity The azimuth of Macao-Zhuhai decreased from 104.94° to 95.26°, which exhibits the weak spatial structure of IS towards the northwest-southeast. The flattering ratio decreased from 1.84 to 1.56 during 1987-2007. Moreover, it increased to 1.68 in 2012, and then it decreased slightly in 2017, which shows the change in the SDE area with obvious directivity of the distribution of the IS. The GC of Macao-Zhuhai is located in Zhuhai, while the geometric center is located west of Macao-Zhuhai. However, the GC moved 4.42 km towards the southwest. Macao-Zhuhai and Hong Kong-Shenzhen have a similar trend of expansion, while in the beginning, the increasing IS towards the border of Macao and Zhuhai can be observed. Furthermore, the IS extended towards the whole region of Zhuhai
SDEs and gravity centers of IS at the development poles scale from 1987-2017.
Currently, the development of GBA has formed a pattern that development poles promote the development of whole region. The cities in development poles of GBA are well integrated that prompted the regional cooperative construction and industrial economic development. Although integrative development of GBA is taking shape due to the positive influence of development poles, the connections of development poles and other cities in GBA are still weak because of their great disparity of development. Therefore, development poles should continue to lead industry and play important roles on the development of the whole region.
Hong Kong, Macao, Guangzhou, and Shenzhen are considered to be the “core engines” [ The PD and LSI of IS in Guangzhou decreased with the increase of IS density, which implies that the higher the density of the IS, the stronger is the aggregation. It also indicates a simple and more stable structure of the IS. The PD of very low-density IS decreased, and others did not change a lot. The LSI of very low-density IS decreased then increased while others increased. The SHDI of IS in Guangzhou experienced an M-type trend, which indicated that the IS in Guangzhou experienced the replacement process of the landscape. Finally, the natural surface had a high dominant position, but it changed to high-density IS From 1987 to 2017, the PD and LSI of low-density IS in Hong Kong experienced a “decreasing-increasing” trend, but the very low-density IS had an opposite trend, and the trend of other densities is not steady, indicating that the degree of fragmentation and aggregation of different densities IS are not stable. Moreover, the PD of Hong Kong was lower than the PD of Guangzhou and Shenzhen, but LSI values were similar to those of Shenzhen, which indicates that the IS of Hong Kong had a lower degree of fragmentation. The SHDI of Hong Kong showed a similar changing trend with that of Guangzhou before 2002. However, the SHDI decreased again during 2007-2017, which might resulted in the replacement of lower density IS with higher IS In Macao, the PD and LSI of lower density IS are higher than those of higher density IS, the PD of very low-density IS had a similar variation trend with Guangzhou, and the PD of other densities IS varied, indicating that the degree of fragmentation of IS declined gradually. The LSI of very low and low density IS increased first, then decreased and finally reached a stable state, which indicates that the degree of aggregation increased first then decreased; the LSI of other densities’ IS had a fluctuating trend with low values. Moreover, the value of PD and LSI is lower than those of Guangzhou and Shenzhen, which indicates that the IS had a low degree of fragmentation and a high degree of aggregation. Due to the “up-down-up” trend, the SHDI of Macao started a steady development; it shows that in 1987, the different densities’ IS had a uniform pattern. Finally, it can be observed that the higher IS had a high dominant position in 1992 with a homogenized pattern of IS Shenzhen had a similar PD variation trend with Guangzhou, and PD of very low-density IS decreased from 1987 to 2017; others did not change a lot; it means that the degree of fragmentation of IS declined gradually. LSI of very low-density IS experienced the process of “decreasing-increasing”; others went through an “increasing-decreasing-increasing” process while the change range was smaller than that of Guangzhou, indicating that the degree of aggregation tends to be stable. The change characteristic of SHDI in Shenzhen was similar to that in Guangzhou. However, it tended to be stable after 2002, which indicated that the IS pattern is in a stable state
The landscape pattern index of IS in the core cities of GBA from 1987-2017.
In summary, both Guangzhou and Shenzhen experienced a replacement process of the main landscape. Therefore, the dominant position of the spatial structure changed from lower IS to higher IS. The landscape of Hong Kong and Macao had a relatively low degree of fragmentation, and a high degree of aggregation compared with those of Guangzhou and Shenzhen and the spatial structure of the IS is more stable. In the future, Guangzhou, Shenzhen, Hong Kong, and Macao should pay more attention to rational urban development and ecological environment protection and also focus on optimizing the industrial structure to further enhance their effects for other regions in GBA. With the positive influence of core cities, GBA will become a real world-class bay area.
Information on IS area change in the context of rapid urbanization is key to understanding the process of urban agglomeration and deal with these serious challenges in terms of the environment, climate, and natural resources. However, previous studies have mostly focused on spatial structure changing in the city scale. In this study, we focused on GBA and investigated how the IS expanded during 1987, 1992, 1997, 2002, 2007, 2012, and 2017 at regional and city scales by using sensor-based Landsat images (i.e., TM, ETM+, and OLI). Furthermore, we discussed the spatiotemporal change characteristics of GBA growth at different spatial scales and indicated the evolution process of IS. Moreover, we attempted to understand the spatial structure evolution of urban agglomeration. We found that GBA has experienced a spatial structure evolution as follows: mega city, metropolitan area, and an urban agglomeration with rapid urbanization and industrialization. In addition, GBA has experienced a spatial reconstructing during its growth process, which is different compared with other urban agglomeration.
The results indicate that the IS of GBA experienced a rising trend during the past 30 years and exhibited a distinct characteristic along the PRD and the coastline. The IS area of GBA increased 749.68% during 1987 to 2017, the IS distributed along the Pearl River and coastline, and the ratio of the impervious surface decreased with the increasing distance from the axis. Furthermore, the IS of different cities in GBA changed. However, small cities developed by leapfrog development, while developed cities tended to expand on the edge of exiting IS. The findings of the current study reveal the evolution process in urban agglomeration. It not only provides the reference for GBA planning management but also supports a reference for a better understanding of other urban agglomerations. The current study only focused on multiscale analysis to observe the spatiotemporal patterns of the IS of urban agglomeration and preliminarily discussed the causes. However, the driving factors of sustainable development and ecological influence of urban agglomeration are more complicated with different political institutions, functions, and degrees of development.
These change progressions and orientations in IS patterns during the study period show the urbanization evolution from the pattern to the process in GBA, which thus provides useful information for urban sustainable planning and environmental protection. Further studies may be conducted on revealing them quantitatively.
The topographic map of the Institute of Surveying and Mapping, Department of Natural Resources of Guangdong Province and the Digital Orthophoto Map (DOM) data of Guangzhou Jiantong Surveying, Mapping, and Geoinfo Co., Ltd., used to support the findings of this study were supplied by Xiaoding Liu and Mingqiang Chen under license and so cannot be made freely available. Requests for access to these data should be made to Fan Liu (
The authors declare that there are no conflicts of interest regarding the publication of this paper.
The study was supported by the National Natural Science Foundation of China (41871292), the Science and Technology Program of Guangdong Province, China (2018B020207002), and the Science and Technology Program of Guangzhou, China (201803030034 and 201802030008).