1. Introduction
As a consequence of the processes of urbanisation and socioeconomic development, urban area has expanded significantly worldwide. In the last few decades, dramatic increases in built-up areas have been experienced globally [
1,
2], with a total built-up area of 797,076 km
2 in 2018, 1.5 times that in 1990 [
3]. As urbanisation continues, the demand for land is expected to increase [
4]. It is forecasted that urban land cover will increase by 1.2 million km
2 by 2030, nearly tripling the global urban land area circa 2000 [
5]. Such expansion is becoming increasingly unsustainable, with urban areas expanding much faster than the population [
6], especially in India, China, North America, and Europe [
7]. Analyses have suggested that, due to the decrease in urban population densities, an estimated 125,000 km
2 of land was converted to urban land-uses that could have otherwise remained in cultivation or as natural vegetation [
7]. Though occupying only a small proportion of the global land cover, the expansion of built-up area significantly affects Earth system processes and, thus, has tremendous ecological and environmental consequences, such as arable land loss [
7,
8,
9], biodiversity loss [
10,
11], local and regional climate change [
12,
13,
14], and higher air pollution concentrations [
15,
16]. The scarcity of land makes better governing a necessity, promoting sustainable planning [
17]; for which, studies focusing on understanding the spatiotemporal characteristics of built-up area expansion at different levels is essential.
In China, the expansion of built-up areas has its own characteristics, due to the joint effects of the government and the market. Since the reform and opening up in 1978, the pace of urbanisation in China has increased significantly. In 2015, China became the country with the most built-up area in the world [
3]. Due to differences in market attractiveness and policy emphasis, there has been a significant heterogeneity in the expansion of built-up areas over time and across regions. Temporally, the expansion speed increased in the 2000s [
18,
19,
20], while the change in the 2010s remains controversial, owing to the study areas and data sources [
21,
22]. Spatially, expansion has mainly been observed in coastal China, with Beijing–Tianjin–Hebei (BTH), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD) experiencing the largest expansion agglomeration [
19,
23,
24]. Cities with larger size tend to expand at higher speed [
25]. However, owing to the difficulty in obtaining urban expansion information over multiple years on a large spatial scale, studies with a national coverage and a long time period—in particular, covering the whole time span since China’s marketisation—is rare.
Statistic data and remote sensing data are the main data sources in studies of built-up area expansion, while the latter has fine spatial information and can help analysing the change in urban built-up areas objectively from multitemporal images. Landsat data is widely used, with impervious area identified based on surface reflectance. However, with a 30 m spatial resolution, it has difficulty in covering a large spatial scale. While night-time light (NTL) data is mainly at medium-low resolution and frequently used in analysing urban expansion at large scales. Of which, Defense Meteorological Satellite Program and Operational Linescan System (DMSP-OLS) covers the longest time span (from 1992 to the present), is of interest by researchers. With DMSP-OLS data, the built-up area expansion based on night light intensity can be summarised. The national rapid growth and great disparities among regions are consistent with studies using Landsat data and statistic data [
26,
27,
28,
29,
30]. Hot points are also prone to concentrate around BTH, YRD, PRD, and provincial capitals, which are the highpoints of urban growth as well [
28]. DMSP-OLS data can also help identify the general formation of the strategic urban pattern characterised by two horizontal axes and three vertical lines [
29]. Generally, night-time light data directly reflects socioeconomic development and human activities and can help in identifying the expansion dynamics at the national scale.
To characterise the pattern of built-up area, numerous landscape indices have been developed, such as landscape shape index (LSI) [
2], largest patch index [
25], and fractal dimension index [
24]. Researchers have used these indices to quantify the geometric and spatial properties of categorical map patterns but have seldom used them to obtain information about the dynamic change processes of landscape patterns. Indices for quantifying urban dynamics at two or more time points are essential. Researchers have used indices to capture information on temporal changes of built-up areas, including dynamic degree [
19], (annual) expansion areas [
20,
25,
31], and (annual) expansion rates [
20,
25,
32], which have most frequently been used to observe temporal changes in built-up areas, thus representing the speed of expansion. However, the spatiotemporal pattern of changes in speed has never been mapped, which can be interpreted as the change in driving forces of built-up area expansion [
20]. This paper develops a new index, acceleration, to measure the change in expansion speed, to represent the change in driving forces. Speed and acceleration, together, can give us insight to better understand spatiotemporal land-use dynamics in fast-growing regions. It is expected that speed and acceleration can be used to identify the characteristics of temporal changes of a certain landscape (in this study, we selected built-up area) using multitemporal remote sensing data.
In this paper, using the multitemporal built-up area data with Landsat imagery and night-time light (NTL) data as data sources, we mapped and quantified the dynamics of built-up area expansion in China over the past four decades using expansion speed and acceleration as indicators, as well as investigating the driving forces and environmental effects of built-up area expansion, based on the mapping results. We have four objectives: (1) to develop a data set of expansion speed and acceleration at fine resolution over 40 years; (2) to find new information from the mapping of the two indicators, including the spatiotemporal characteristics of built-up area expansion and its differences between regions and urban sizes, and expansion types from the perspective of temporal changing dynamic; (3) to examine the driving factors and their spatiotemporal changes; (4) to explore the relationships between different expansion indicators and environmental changes.
3. Results
3.1. Mapping Results of the Speed and Acceleration of Built-Up Area Expansion
The mapping results show that the areas of high expansion speed and acceleration have enlarged dramatically (
Figure 5). The coastal region and some large cities in central China are areas with rapid expansion speed. Large urban agglomerations, such as Beijing–Tianjin–Hebei (BTH) and the Yangtze River Delta (YRD), have experienced accelerating expansion in the last 40 years. Acceleration in the North China Plain and northeast region increased significantly between periods 3 and 4. Detailed information at the city level can also be provided in the mapping results. Taking Shanghai as an example, the rapidly expanding area has changed from the central region in the 1990s to the outlying new towns in the 2010s, with its acceleration having a toroidal structure.
3.2. Temporal Variation of Built-Up Area Expansion at the National Scale
According to the calculation based on the mapping results, the built-up area expanded significantly, from 19,434 km
2 in 1978 to 245,362 km
2 in 2017 (
Figure 6). The 40 years were divided into four periods, based on which the speed and acceleration of expansion were observed. The results show that the highest annual growth occurred between 2010 and 2017, with an expansion speed of 13,808 km
2/yr—3.38 times more than in the previous time period—followed by 2000–2010 (5391 km
2/yr), 1990–2000 (3495 km
2/yr), and 1978–1990 (3384 km
2/yr). The accelerations were 10.1, 187.6, and 937.4 km
2/yr
2, respectively, indicating a considerable increase in overall driving forces, especially after 2010.
In terms of spatial distribution, the values of Moran’s I were positive, and the spatial dependence was highly significant over the four decades, indicating the spatial clustering of expansion speed and acceleration (
Table 3). The Moran’s I values for expansion speed were 0.566, 0.667, 0.603, and 0.617 in the four respective periods, with the highest spatial dependence in the 1990s. There was a gradual increase in spatial clustering for expansion acceleration, with Moran’s I values of 0.293, 0.454, and 0.543, respectively, indicating an increase in spatial clustering of socioeconomic resources driving built-up area expansion.
3.3. Comparison among Regions and Urban Sizes
Comparison of urban expansion amongst various regions and urban sizes enables more insights into urban expansion (
Figure 7a,
Table 4). Between 1978 and 2017, built-up areas in all regions increased (
Figure 7b). The eastern region exhibited the highest proportion of built-up area, followed by the central region, the northeast region, and the western region, respectively. The average expansion speeds were 3098, 1375, 89, 786 m
2/yr per square kilometre of land in the eastern, central, western, and northeast regions, respectively. The expansion speed in the eastern region increased continually for 40 years, while the speed in central and northeast regions remained almost unchanged until 2010. In terms of acceleration, the eastern region had the highest acceleration in all periods. During the first two periods, only the eastern region had a positive acceleration, reflecting the development priority under limited resources. The acceleration of the eastern region from the 1990s to the 2000s was still much higher than the other regions (118 m
2/yr
2 vs. 33.1, 4.3, and 17.5 m
2/yr
2 per square kilometre of land). The last two periods saw significant increases in acceleration in all regions.
Figure 7c displays the characteristics of urban expansion for differently sized urban cities. The variation is quite large in all three indicators, with megacities ranking first, followed by large cities, medium cities, and small cities. The average expansion speeds were 4126, 953, 134, and 25 m
2/yr per square kilometre of land for mega, large, medium, and small cities, respectively. Acceleration prior to 2010 was mainly observed in megacities, indicating the concentration of development resources in major growth poles.
To further analyse the differences in urban expansion based on urban size within different regions, we summarise the regional and urban size differences in
Appendix A. The results show that different-sized cities in the eastern region have experienced sustained urban expansion after the reform and opening up, while expansion in other regions was mainly concentrated in megacities, with megacities and large cities accelerating after 2000.
3.4. Type of Built-Up Area Expansion
The number of thriving cities increased by approximately 50% since 2000 (
Figure 8). As a result of rational investment, the changes in speed and acceleration were consistent, with no emerging city being identified as particularly noteworthy. The spatial and temporal pattern changed over the 40 years. The thriving cities between 1978 and 1990 were mainly concentrated in coastal China and the North China Plain, with the highest speed and acceleration observed in the Pearl River Delta (PRD; see
Figure 9). In the 1990s, the number of thriving cities in these regions increased; the highest expansion speed occurring in the PRD, and the fastest acceleration occurring in the Yangtze River Delta (YRD) and Beijing–Tianjin–Hebei (BTH). In the 2000s, many cities in the central and northeast regions, and some of the major cities in the western regions, were also categorised as thriving. After 2010, acceleration failed to continue in some megacities in the PRD; thus, these cities became stable following 20 years of thriving development.
In addition, in mature urban agglomerations, the neighbouring cities, such as Zhongshan, Foshan, Wuxi, and Suzhou expanded concurrently. Cities in other agglomerations were distributed independently in the axes, indicating the stage of development in the growth pole.
3.5. Socioeconomic Drivers behind Built-Up Expansion
After mapping and comparing the spatiotemporal changes in driving forces indicated by acceleration, we further explored the specific drivers, in terms of the period (Models 1–4), region (Models 5–8), and urban size (Models 9–12). The results of F test are significant all the time, indicating the difference among cities, so that pooled model is rejected. The modified Hausman test indicated that the fixed effects model performed better, with the exception of Models 5 and 12. Thus, Models 6–11 were based on the fixed effects model, while Models 5 and 12 were based on the random effect model.
The results of the total samples with a panel data model showed that changes in fixed asset investment were positively correlated with built-up area expansion (Model 0 in
Table 5). However, the effects of different factors vary in different periods (Model 1–4 in
Table 6). In the early period of the reform and opening up, built-up area expansion was mainly related to population growth. In the 2000s, the increase in GDP and fixed asset investment were significantly related to built-up area expansion, where the proportion of secondary industry production strengthened the effect of GDP. After 2010, fixed asset investment had a significantly positive effect, while the proportion of secondary industry production diminished the effect of GDP.
The drivers varied, in terms of region and urban size. In the eastern region, population growth, fixed asset investment, and secondary industry production were positively related to built-up area expansion. The central region had a significant relationship with fixed asset investment. In the western region, GDP and fixed asset investment were significantly related to built-up area expansion, where the negative effect of GDP was mitigated with higher secondary industry production. In the northeast region, fixed asset investment was positively related, and the secondary industry production played a negative role in mediating the effect of GDP.
Cities of different sizes experienced different expansion paths. In small cities, population played a negative role. In medium cities, fixed asset investment was a positive driver. In megacities and large cities, GDP and fixed asset investment were positively related to built-up area expansion; whereas, in large cities, the proportion of secondary industry production diminished the effect of GDP.
The validation result indicates that the modelling is relatively stable with the panel data at national scale, with outliers in certain cities (
Figure 10). Despite the relatively small adjusted R-square, we can say that the model is reliable and applicable. The RMSE, MAE, and pseudo R-squared in each period are listed in
Table 7. They can serve as a comparison for future research.
3.6. The Environmental Effects of Built-Up Area Expansion
The built-up area continued to rise between 1990 and 2017. At the same time, wastewater discharge increased from 3.54 billion tons in 1990 to 7.35 billion tons in 2015, with an average annual increase of 3.0%. The average area of natural reserves showed a decreasing trend, especially after 2001, from 8.37 to 5.35 km2 in 2017, with an average annual decrease of 2.8%. Industrial waste gas emissions revealed a rising trend, with the greatest increase occurring between 2000 and 2010.
The change in wastewater discharge, industrial waste gas emissions, and average area of natural reserves was consistent with that of built-up area expansion (
Figure 11a). The Spearman’s correlation coefficient passed the significance test and the coefficients between built-up areas and industrial waste gas emission, wastewater discharge, and average area of natural reserves were 0.997, 0.997, and −0.989, respectively. In terms of speed, the growth speed of environmental factors was lower than that of built-up area after 2010 (
Figure 11b), showing an opposite acceleration direction. Taking industrial waste gas emissions as an example, different characteristics were shown in three periods. In the 1990s, its growth trend was similar to that of built-up areas. In the 2000s, there was a significant increase in its growth rate, becoming much higher than that of built-up areas, mainly due to industry-driven expansion [
20]. Along with the industrial structural transformation and the introduction of clean production policies, its growth rate slowed after 2011.
5. Conclusions
Remote sensing data with long-term series and large geographic coverage make it possible to observe urban expansion from a spatial and temporal perspective. In this study, we used the concept of acceleration to represent the change in driving forces and developed a new data set providing the speed and acceleration of expansion between 1978 and 2017 in China. Based on the mapping results, we investigated the dynamics, driving forces, and environmental effects of built-up area expansion.
We found that (1) there has been considerable growth in the built-up area in China over 40 years, with the eastern region and the megacities experiencing the highest growth speed. Thriving cities, with high expansion speed and acceleration, and which are mainly in the eastern region, were observed. (2) Driving forces tend to increase in rapidly expanding cities and, in analysing the relationship between expansion speed and acceleration, an inverted U-shape relationship was uncovered. (3) The overall driving forces increased dramatically over the four analysed decades. The main drivers changed from population to economic development and fixed asset investment. The driving forces and drivers varied amongst regional distribution and urban size. (4) The environmental factors changed consistently within built-up areas. The difference in speed reflects changes in driving effect, while the difference in acceleration reflects the roles of other drivers.
In this study, we developed a method for mapping the speed and acceleration of urban built-up areas. The combination of these two indicators was shown to be useful for identifying expansion dynamics and changes in driving forces, with potential for further application in the study of environmental effects. They are suitable for longitudinal data and we hope that they will be used to generate more interest in related subjects or on different scales.