Assessing Interannual Urbanization of China’s Six Megacities Since 2000

: As a large and populous developing country, China has entered the rapid urbanization stage since 2000. Until 2018, China has accounted for nearly 1 / 5 of global megacities. Understanding their urbanization processes is of great signiﬁcance. Given the deﬁciencies of existing research, this study explored the interannual urbanization process of China’s six megacities during 2000–2018 from four aspects, namely, the basic characteristics of urban land expansion, expansion types, cotemporary evolution of urban land–population–economy, and urbanization e ﬀ ects on the local environment. Results indicated that (1) urban lands in China’s six megacities increased by 153.27%, with distinct di ﬀ erences across megacities; (2) all of six megacities experienced the expansion processes from high-speed to low-speed, but they varied greatly in detail; (3) the speeds of urban land expansion in China’s megacities outpaced the population growth but lagged behind in GDP increase; and (4) urbanization has triggered an environmental crisis, which is represented by the decline in vegetation coverage and the increase in land surface temperature in newly expanded urban lands. This study enriched the content of urbanization, supplemented the existing materials of megacities, and provided a scientiﬁc reference for designing rational urban planning. of vegetation coverage and land surface temperature in China’s megacities during 2000–2018.


Introduction
Global urbanization has progressed considerably with the proportion of urban population increasing from 33% in 1950 to 55% in 2018 and urban lands expanding at twice the rate of population growth [1]. Urbanization has both advantages and disadvantages. On the one hand, it provides increased urban housing, convenient transportation, sophisticated education, and excellent medical treatment and social services. On the other hand, the irrational urbanization may result in negative effects on socioeconomic and eco-environment development, such as increasing housing prices, traffic congestion, and emissions of domestic garbage, and vehicle exhaust. All these issues pose both opportunities and challenges to the further development of urban areas in the 21st century [2]. Researches of the urbanization processes could provide basic materials as references when designing rational urban planning and building the inclusive, safe, resilient, and sustainable human settlements, which is also the goals of the 2030 Development Agenda [3].
Urbanization, characterized by the demographic migration from rural to urban areas and the conversion of territorial resources from other types to urban lands [4][5][6], has remarkably improved cities in both number and scale, thereby facilitating the formation of megacities. According to the United Nations [7], only two megacities existed in the world in 1950. This number increased to 29 in 2014 and will exceed 40 by 2030. Megacities are defined as cities with over 10,000,000 inhabitants [8].

Materials and Methods
The main workflow of our study includes five steps. Firstly, all datasets were downloaded and preprocessed. Secondly, urban lands of China's six megacities were delineated (Section 3.2). Thirdly, urban expansion types were characterized (Section 3.3). Fourthly, the cotemporary evolution of urban land-population-GDP was analyzed (Section 3.4). Fifthly, the effect of urban expansion on environment was monitored (Section 3.5).

Data Acquisition and Preprocessing
In this study, 135 scenes of multisource remotely sensed images with 30-80 m spatial resolutions, less than 10% cloud cover, and vigorous vegetation growth [31] were applied to delineate the urban lands of China's six megacities during 2000-2018 (Appendix A Table A1). With a spatial resolution of nearly 1000 m, Terra MOD13A3 and MOD11A2 8-day composite products were selected to obtain the VC and LST in summer from 2000 to 2018 (Appendix A Table A2). Urban populations and GDPs of China's six megacities during 2000-2016 were collected from the China City Statistical Yearbook (Table

Materials and Methods
The main workflow of our study includes five steps. Firstly, all datasets were downloaded and preprocessed. Secondly, urban lands of China's six megacities were delineated (Section 3.2). Thirdly, urban expansion types were characterized (Section 3.3). Fourthly, the cotemporary evolution of urban land-population-GDP was analyzed (Section 3.4). Fifthly, the effect of urban expansion on environment was monitored (Section 3.5).

Data Acquisition and Preprocessing
In this study, 135 scenes of multisource remotely sensed images with 30-80 m spatial resolutions, less than 10% cloud cover, and vigorous vegetation growth [31] were applied to delineate the urban lands of China's six megacities during 2000-2018 (Appendix A Table A1). With a spatial resolution of nearly 1000 m, Terra MOD13A3 and MOD11A2 8-day composite products were selected to obtain the VC and LST in summer from 2000 to 2018 (Appendix A Table A2). Urban populations and GDPs of China's six megacities during 2000-2016 were collected from the China City Statistical Yearbook (Table 2). During data preprocessing, all multisource remotely sensed images and MODIS products were resampled to 30 or 1000 m pixel sizes by using Albers equal-area conic coordinate system, respectively.

Urban Land Extracting
Visual interpretation method was used to extract urban lands in China's six megacities, and a subsequent procedure that consists of four steps is deployed ( Figure 2). Step 1: Apart from the data preprocessing, a clear definition of urban lands is necessary. Our research focused on the central built-up areas, which were defined as urban lands where buildings have been developed contiguously with available municipal utilities and public facilities [32,33].
Step 2: The band composition was executed based on standard false-color synthesis and the images were enhanced using linear contrast stretching and histogram equalization. By employing this step, differences (i.e., color, and hue) between various land use/cover types seemed more obvious. Step 1: Apart from the data preprocessing, a clear definition of urban lands is necessary. Our research focused on the central built-up areas, which were defined as urban lands where buildings have been developed contiguously with available municipal utilities and public facilities [32,33].
Step 2: The band composition was executed based on standard false-color synthesis and the images were enhanced using linear contrast stretching and histogram equalization. By employing this step, differences (i.e., color, and hue) between various land use/cover types seemed more obvious. Besides, the accuracy of geometric correction in terms of the relative position error or the same feature point does not exceed 2 pixels.
Step 3: In accordance with various interpretation symbols as shown in Figure 2, urban lands were separated from other land use/cover types. When interpreting the remotely sensed images of next year, the original urban lands were applied as the basic layers and the newly developed urban lands during this year were delineated. Step 4: Quality control was executed by adopting field validation and repeated interpretation. Among which, field verification mainly employed in 2000, 2005, 2008, 2010, and 2015, by taking photos and recording the situation of local land use/cover on tables. The repeated interpretation was employed annually by referring to interpretation symbols, Google Earth platform and topographic maps. If the urban lands showed low accuracies (less than 90%), they should be re-interpreted.
This procedure was performed on the Modular GIS Environment (MGE) platform, which was developed by the Intergraph Company of America and had strong image processing functions. The visual interpretation was accomplished by professional interpreters with rich experience. More detailed information of the procedure had been elaborated by Zhang et al. [34]. Noticeably, given the difficulty in obtaining high-quality remotely sensed images of China's six megacities in some years, the interpolation data on urban land areas in these years were applied as a supplement by executing the method of Liu et al. [9].

Measurement for Characterizing Urban Land Expansion Types
As the result of the comprehensive development of factors, such as economy, society, culture, and national policies, the urban land expansion in different cities generally exhibited distinct differences in morphology. Area and perimeter dynamics are two of the basic manifestations of urban land expansion. In this study, the combination of growth rates of urban land area (GRA) (Formula (1)) and urban land perimeter (GRP) (Formula (2)) were applied to characterize the urban expansion types of China's megacities [35].
where A t1 and A t2 are the urban land areas in t1 and t2, respectively; P t1 and P t2 are the urban land perimeters in t1 and t2, respectively. When characterizing urban land expansion types, the following steps were used according to the method proposed by Shi et al. [35]: (1) Calculating the GRA and GRP, (2) standardizing the GRA and GRP by employing Z-score normalization in the SPSS 2.0 software, and (3) dividing the normalized GRA and GRP into four levels, including Levels 1 (minimum, −1), 2 (−1, 0), 3 (0, 1), and 4 (1, maximum). Figure 3 shows that each block is combined by using two-digit codes. The first and second numbers indicated the GRA and GRP levels, respectively. High values of GRA and GRP indicated high expanding speeds and decreased compact morphologies. Then, all newly expanded urban lands were categorized into four expanding types. In this study, Types A, B, C, and D represented the loose expansion at high speed, compact expansion at high speed, loose expansion at low speed, and compact expansion at low speed, respectively. (1, maximum). Figure 3 shows that each block is combined by using two-digit codes. The first and second numbers indicated the GRA and GRP levels, respectively. High values of GRA and GRP indicated high expanding speeds and decreased compact morphologies. Then, all newly expanded urban lands were categorized into four expanding types. In this study, Types A, B, C, and D represented the loose expansion at high speed, compact expansion at high speed, loose expansion at low speed, and compact expansion at low speed, respectively.

Growth Rates of Urban Population (GRPOP) and Gross Domestic Product (GRGDP)
Urbanization is a complex process involving multiple developments in physical, demographic, and socioeconomic dimensions [2]. Therefore, apart from the scale and morphology of urban land, urban population and GDP were also regarded as representative indicators of the urbanization process. Researches about the cotemporary evolution of these indicators could help further understand the urbanization process of China's megacities. Given the accessibility of statistical data and the definition of urban lands, year-end household-registered populations, and GPDs in city districts from 2000 to 2016 were applied as urban populations and GDPs of China's six megacities. The growth rate of urban population (GRPOP) and the growth rate of urban GDP (GRGDP) were calculated by using Formulas (3) and (4), respectively.
where O t1 and O t2 are the urban populations in t1 and t2, respectively; D t1 and D t2 are the urban GDPs in t1 and t2, respectively.

Vegetation Coverage (VC) and Land Surface Temperature (LST)
As a crucial factor in achieving urban sustainability, the dynamics of urban environment have attracted considerable attention from researchers in the remote sensing community during the past several years [36]. The previous researches have shown that VC and LST were two important indicators of environmental conditions [36,37]. To explore the urbanization effects on the local environment of China's megacities, the dynamics of VC and LST in both pre-grown urban lands in 2000 ("R1" hereinafter) and newly expanded urban lands during 2000-2018 ("R2" hereinafter) were calculated by using Formula (5).
where ∆ i indicates the changed VC or LST of the ith pixel from 2000 to 2018; I i_2000 and I i_2018 is the VC or LST of the ith pixel in 2000 and 2018, respectively. The VC in 2000 and 2018 is calculated using Formula (6).
where V i indicates the VC of the ith pixel; N i is the NDVI value of the ith pixel in MOD13A3 products; N s and N l are the minimum and maximum NDVI values of MOD13A3 products, respectively. The LST in 2000 and 2018 is calculated using Formula (7) [38].
where T i indicates the LST of the ith pixel; and D i is the DN value of the ith pixel in MOD11A2 products.

Basic Characteristics of Urban Land Expansion
The magnitude of urban lands varied greatly in China's six megacities (Figure 4a). In 2000, Beijing had the largest urban land area of 830.78 km 2 , followed by Shanghai (598.78 km 2 ) and Guangzhou (491.75 km 2 ). Urban land in Shenzhen (461.82 km 2 ) and Tianjin (266.23 km 2 ) ranked third and second from the end. Chongqing had the smallest urban land area of 161.43 km 2 , which was less than 1/5 of that in Beijing. In 2003, urban land areas in Shenzhen surpassed Guangzhou and ranked third.  Spatially, the newly expanded urban lands of Beijing mainly distributed in the north, east, and south directions (Figure 1). Tianjin had witnessed urban land expansion at all directions. For Shanghai, urban land expansion mainly emerged in the west, south, and east directions. Urban lands in Chongqing and Guangzhou mainly expanded along the south-north directions. For Shenzhen, the urban lands expanded minimally in the south direction, however, the north direction had undergone dramatic expansion.  During 2000-2018, the distributions and contribution rates of four urban land expansion types were uneven in the six megacities ( Figure 6). Type A was the dominated expansion type of Beijing with the proportion of 43.63%, followed by Types B (30.94%), D (17.12%), and C (8.31%). The four expansion types distributed evenly in the north, east, and south directions. For Tianjin, Types A (33.18%) and B (31.27%) were its main expansion types, supported by Types C (10.46%) and D (25.09%). Although Tianjin had witnessed urban land expansion at all directions, the distributions of four expansion types were distinct. Among which, the southeast direction of Tianjin mainly expanded by Type B. Nearly 2/3 of the newly developed urban lands in Shanghai were from Type A and B, meanwhile, the other 1/3 were from Types C and D. Type A dominated the west direction, while Type D mainly took place in the east direction. Types A, C, and D contributed 32.03%, 22.97%, and 45.00% to the urban land expansion of Chongqing, respectively. Similar to Chongqing, Guangzhou also expanded to the south by Type C and D, and to the north by Type A. Totally, 75.04%, 18.97%, and 6.00% of the newly developed urban lands adopted Types A, D, and C, respectively. For Shenzhen, the north direction had undergone dramatic expansion and mainly adopted Type A. Overall, Type A was the main urban land expansion type of China's megacities, Type C contributed relatively little to urban land expansion, whereas, Type B had no contribution to urban land expansion in Guangzhou and Chongqing. During 2000-2018, the distributions and contribution rates of four urban land expansion types were uneven in the six megacities ( Figure 6). Type A was the dominated expansion type of Beijing with the proportion of 43.63%, followed by Types B (30.94%), D (17.12%), and C (8.31%). The four expansion types distributed evenly in the north, east, and south directions. For Tianjin, Types A (33.18%) and B (31.27%) were its main expansion types, supported by Types C (10.46%) and D (25.09%). Although Tianjin had witnessed urban land expansion at all directions, the distributions of four expansion types were distinct. Among which, the southeast direction of Tianjin mainly expanded by Type B. Nearly 2/3 of the newly developed urban lands in Shanghai were from Type A and B, meanwhile, the other 1/3 were from Types C and D. Type A dominated the west direction, while Type D mainly took place in the east direction. Types A, C, and D contributed 32.03%, 22.97%, and 45.00% to the urban land expansion of Chongqing, respectively. Similar to Chongqing, Guangzhou also expanded to the south by Type C and D, and to the north by Type A. Totally, 75.04%, 18.97%, and 6.00% of the newly developed urban lands adopted Types A, D, and C, respectively. For Shenzhen, the north direction had undergone dramatic expansion and mainly adopted Type A. Overall, Type A was the main urban land expansion type of China's megacities, Type C contributed relatively little to urban land expansion, whereas, Type B had no contribution to urban land expansion in Guangzhou and Chongqing.         (Figure 8). The urban population grew faster than urban land areas in Shenzhen and Chongqing with average GRPOPs of 7.02% and 7.30% and GRPOP dispersions of 0.42 and 0.06, respectively. The average GRPOPs in the four other megacities ranged from 1.57% (Shanghai) to 2.84% (Tianjin) and clearly lagged behind their corresponding GRAs (3.35-7.77%). Beijing, Chongqing, Guangzhou, Shanghai, Shenzhen, and Tianjin showed high average GRGDPs of 16.70%, 20.96%, 14.84%, 12.91%, 16.92%, and 17.45%, respectively, and their GRGDP dispersions ranged from 0.08 (Guangzhou) to 0.19 (Beijing). By contrast, all these six megacities presented higher GRGDPs than GRAs. Overall, the speed of urban land expansion in China's megacities outpaced their corresponding population growth but lagged behind their GDP increase from 2000 to 2016.

Urbanization Effects on Local Environment
From 2000 to 2018, VC dynamics in R1 and R2 varied greatly in China's six megacities ( Figure  9). The VCs of six megacities in R1 showed increasing trends and their average value grew by 3.32% during the past 18 years. VC increases in Beijing (8.09%) and Tianjin (4.27%) were larger than the average. VC increase in Shanghai was 3.02% and ranked third. However, VC increases in Shenzhen, Guangzhou and Chongqing had not reached 3.00%. On the contrast, the VCs of six megacities in R2 mainly exhibited the descending trends with their average value reducing by 5.07%. The most obvious decline of VC values emerged in Chongqing (15.40%), followed by Shanghai (6.82%) and Guangzhou (5.40%). The VC of Tianjin reduced by 3.55%, which was lower than the average. Beijing showed the minimal VC decrease with 0.90%, whereas, VC in Shenzhen had not even decreased.
Moreover, the LSTs of the six megacities presented increasing trends, gentle in R1 and dramatic in R2. The averaged LST increased by 1.65 °C in R1 and 2.12 °C in R2, respectively. For R1, the highest LST increases emerged in Chongqing (3.92 °C), while the lowest occurred in Tianjin (0.30 °C). LST increases in Guangzhou and Shanghai ranked second and third, which grew by 2.17 °C and 2.02 °C, respectively. LST increases in Beijing and Shenzhen were lower than 1.00 °C, far below the average.

Urbanization Effects on Local Environment
From 2000 to 2018, VC dynamics in R1 and R2 varied greatly in China's six megacities (Figure 9). The VCs of six megacities in R1 showed increasing trends and their average value grew by 3.32% during the past 18 years. VC increases in Beijing (8.09%) and Tianjin (4.27%) were larger than the average. VC increase in Shanghai was 3.02% and ranked third. However, VC increases in Shenzhen, Guangzhou and Chongqing had not reached 3.00%. On the contrast, the VCs of six megacities in R2 mainly exhibited the descending trends with their average value reducing by 5.07%. The most obvious decline of VC values emerged in Chongqing (15.40%), followed by Shanghai (6.82%) and Guangzhou (5.40%). The VC of Tianjin reduced by 3.55%, which was lower than the average. Beijing showed the minimal VC decrease with 0.90%, whereas, VC in Shenzhen had not even decreased.
Moreover, the LSTs of the six megacities presented increasing trends, gentle in R1 and dramatic in R2. The averaged LST increased by 1.65 • C in R1 and 2.12 • C in R2, respectively. For R1, the highest LST increases emerged in Chongqing (3.92 • C), while the lowest occurred in Tianjin (0.30 • C). LST increases in Guangzhou and Shanghai ranked second and third, which grew by 2.17 • C and 2.02 • C, respectively. LST increases in Beijing and Shenzhen were lower than 1.00 • C, far below the average. For R2, the LST increases in Chongqing, Shanghai, and Guangzhou have surpassed the average, with 4.38 • C, 2.70 • C, and 2.22 • C, respectively. Meantime, the LST increases in Beijing (1.26 • C), Shenzhen (1.35 • C), and Tianjin (0.84 • C) was lower than the average. LST increases emerged in Chongqing (3.92 °C), while the lowest occurred in Tianjin (0.30 °C). LST increases in Guangzhou and Shanghai ranked second and third, which grew by 2.17 °C and 2.02 °C, respectively. LST increases in Beijing and Shenzhen were lower than 1.00 °C, far below the average. For R2, the LST increases in Chongqing, Shanghai, and Guangzhou have surpassed the average, with 4.38 °C, 2.70 °C, and 2.22 °C, respectively. Meantime, the LST increases in Beijing (1.26 °C), Shenzhen (1.35 °C), and Tianjin (0.84 °C) was lower than the average.

Discussions
The 2030 Development Agenda has devoted a specific goal to cities, which aims to "make cities and human settlements inclusive, safe, resilient and sustainable" [3,39]. Understanding the urbanization processes might help to achieve the goal. As the important city forms carrying dense population and social activities, megacities in China was selected as the study areas in this work. Urban lands in China's six megacities were delineated from multi-source remotely sensed images using visual interpretation method. To ensure the accuracy of monitoring results more than 90%, this procedure was executed based on strict criteria and accomplished by professional interpreters with rich experience. Section 4.1 elaborated the differences of urban lands in China's six megacities, in terms of magnitudes and expansion directions. Apart from the various historical, socioeconomical and political backgrounds of six megacities illustrated in Section 2, natural terrain (Table 3) might be the other vital factor that influenced the urban land expansion. The interpretation results provided the data base for subsequent research. However, the spatial resolutions of multi-source remotely sensed images used in this work are 30-80 m, therefore, the urban land products were only applied as the referenced extents when analyzing urbanization effects on environment. In the future, more remotely sensed images with various spatial resolutions should be used to obtain urban land products to meet the needs of multi-scales. Table 3. Basic information of China's six megacities.

Megacity Terrain Reference
Beijing Covers a wide topographic gradient from 83 the mountainous areas in the north and west to the plain areas in the central, south, and east [5] Tianjin Relatively flat [40] Shanghai Relatively flat [26] Chongqing The famous "mountainous city", with many SN-trending mountains and a complex elevation ranging from 75 m to 2800 m [41] Guangzhou Higher in the northeast and lower in the southwest [19] Shenzhen Higher in the southeast and lower in the northwest [42] Dynamics of physical features (i.e., scale, and morphology) of urban lands are vital indicators to understand urbanization processes. The previous studies (i.e., [9,43]) mostly characterized urban land sprawl types via edge-expansion, infilling, and outlying patterns. Shi et al. [35] provided a simple approach to identify expansion types by synergistically considering areas and perimeters, and divided China's 340 cities into four types based on the newly developed urban land during 1987-2010. In this work, the interannual newly grown urban land of each megacity was applied as a basic unit to characterize expansion types, and the monitored epoch was updated to 2018. Therefore, more detailed information could be acquired. Figure 5 showed that Type A was the main expansion type in six megacities before 2004, while Type D dominated the urban land expansion after 2012. Although urban land expansion in six megacities exhibited the similar tendency from high GRA/GRP to low GRA/GRP, they underwent a diverse sprawl process. For instance, expansion types presented distinct differences during 2004-2012, those changed frequently in coastal megacities (i.e., Tianjin, and Shanghai) but gently in four other megacities. Accordingly, four types contributed distinctly to urban land expansion in the six megacities. By employing this method, the expansion types of more cities in China or other countries could be characterized in the future.
Urbanization is a complex process involving many aspects. Apart from the uncontrollable urban sprawl, the remarkable economic growth, continuous population explosion, and negative effects on environment are also characteristics considered in the urbanization process. Results of this work revealed two problems existing in China's six megacities at the onset of the 21st century. From a statistical perspective, urban lands, populations, and GDPs in China's six megacities grew at different speeds, and their specific manifestation was that the speed of urban land expansion outpaced the population growth but lagged behind in GDP increase. This finding also supported the standpoints of Fei et al. [2]. From an environmental perspective, the rise in LST has become a problem that cannot be ignored. Besides, VC in R2 presented obvious decline because proportional vegetation had been encroached by the newly expanded urban lands. However, VC in R1 showed a slight increase. According to Qian et al. [44], this phenomenon could be ascribed to the great efforts in increasing urban greenspace by the local government. In this study, researches about the VC and LST were just a preliminary attempt. However, it provided referenced materials and new ideas for further studies. More exploring about urbanization effects on the environment should be executed in the future.
Specially, urban sustainability is defined as an adaptive capacity that can balance social wellbeing, economic development, and environmental protection [45]. Li et al. [46] stated that the sustainable development of megacities has four major challenges, including land subsidence, environment, traffic, and energy aspects. According to the results of this work, there was still a distance to achieve sustainability for China's six megacities. Natural conditions are stable factors that cannot be changed frequently during a short period, but social factors can be rationally regulated and controlled. In the future, helpful measures and special urban planning should be applied to the six megacities to mitigate the negative effects of urbanization on the local environment. Overall, obtaining sustainability in China's cities is an expected achievable goal that requires the joint efforts of the government and ordinary people. In addition, this work characterized China's six megacities from the limited aspects, providing some materials as references when designing rational urban planning. However, these aspects could not thorough respect urbanization of China's six megacities. More key indicators (i.e., population density, air temperature [47], food production [48], cropland losses [14], etc.) of cities and megacities should be investigated in the future.

Conclusions
and Chongqing (370.33 km 2 ), respectively. (2) The contributions of urban land expansion types varied greatly in China's six megacities. Two expansion types-loose expansion at high speed and compact expansion at low speed-dominated the urban land expansion in the early and later years, respectively. (3) Urban land-population-GDP showed an uneven evolution. GDP increased the fastest (1140.57 billion RMB), followed by urban land expansion (4036.33 km 2 ), whereas population growth (31.80 million persons) was the slowest. (4) Urbanization resulted in distinct environmental effects. Vegetation coverage in newly expanded urban lands decreased significantly, whereas those in pre-grown urban lands increased slightly. Land surface temperatures in newly expanded urban lands exhibited a higher increase than those in pre-grown urban lands. This study enriched the content of urbanization, supplemented the existing materials of megacities, and provided a scientific reference in designing rational urban planning.