1. Introduction
Land use change (LUC) results from the interaction between human activities and natural biophysical processes and reflects human–land relationship changes. The Anthropocene epoch has caused changes in global land use, from primitive landscapes to urban and cultivated landscapes [
1,
2]. LUC extensively affects the ecological environment, biodiversity conservation, agricultural economy, climate, and sustainable development [
3,
4,
5,
6,
7]. LUC is part of the complex human–land system, and multiscale LUC analysis is of great significance for model building, change prediction, and land management [
8,
9,
10,
11]. Extensive studies on LUC have been conducted at the international, national, regional, and urban levels [
12,
13,
14,
15].
Due to rapid population growth, industrialization, and urbanization, the metropolitan areas have developed rapidly worldwide, leading to large-scale spatial transformation and drastic LUC [
16,
17]. The application of remote sensing image recognition technology and the establishment of evaluation indicators has enabled the quantitative analysis of the process, dynamics, and mode of spatial change within metropolitan areas [
18,
19,
20,
21]. Many studies have used regression analysis, geographically weighted regression, and other econometric methods to explore the driving factors behind LUC in metropolitan areas [
22,
23,
24], as well as the relationship between LUC and the changing process of ecosystem services and urban heat islands [
25,
26,
27]. Simultaneously, other studies have used cellular automata, artificial neural networks, random decision forests, and Markov chains to simulate and predict the complex process of LUC in metropolitan areas [
28,
29,
30].
Within metropolitan areas, government cooperation between cities and the spatial integration of the labor and housing markets are conducive to land use management [
31,
32]. However, cases from different fields worldwide have shown that socioeconomic contradictions and imbalances inevitably and universally exist within metropolitan areas [
33,
34,
35,
36,
37,
38,
39]. Internal imbalance results in competition and conflict in land policy and development and brings new land use management challenges to metropolitan areas [
40]. This imbalance phenomenon is concentrated in three areas: urban and rural regions, core and peripheral regions, and cross-boundary regions. Regarding the urban and rural regions, rapidly increasing urbanized areas that have significant resource concentrations attempt to seize water, agricultural land, and ecological resources from rural areas, leading to intense land use conflict [
41,
42,
43,
44]. Regarding the core and peripheral regions, the core areas continue to agglomerate resources and facilities, which exacerbates socioeconomic inequality and reduces social mobility [
45,
46,
47,
48,
49]. Alternatively, the relative decline of the core areas results in the division of competitive, autonomous local governments [
50,
51]. Regarding the cross-boundary regions, due to market regulations and the complex relationships between governments at various levels, cross-boundary cooperation is often quite difficult to achieve, especially in the context of land use management. Meanwhile, cross-boundary cooperation and management in some regions may also promote segregation as well as socioeconomic inequality [
52,
53]. However, compared with the urban/rural and core/peripheral regions, the literature has not paid enough attention to the cross-boundary regions within metropolitan areas.
Differences in economic conditions and management policies between administrative bodies result in complex dynamics in cross-boundary regions, which brings challenges in planning and management [
54]. The extant research on cross-boundary regions falls into two categories. The first category comprises cross-boundary planning and management practices, particularly in Western countries where such practices are more common [
55,
56]. For example, Europe has established cross-border cooperation organizations and a supra-regional institution—the European Territorial Cooperation Group [
57]. Meanwhile, in China, attention has been paid to the regions that cross provincial boundaries, such as the Beijing-Tianjin-Hebei Greater Beijing Region, the Guangdong-Hong Kong-Macao Greater Bay Area, and the Fujian-Taiwan boundary region [
58,
59]. The second category comprises the causes of variability in the development status, land use, and ecological landscapes within cross-boundary regions. Some cross-boundary regions have become hotspots for regional development because of the effect of factor flows and agglomeration, which have more complex driving factors [
60]. The rapid development of these cross-boundary regions has also been accompanied by problems such as a sharp increase in house prices and inadequate cross-boundary transportation provision. A common difficulty includes coordinating the conflicting interests of local governments, real estate developers, and other parties in land allocation to realize the optimal use of land in the border regions of urban areas in various countries [
61].
In summary, the literature has mainly focused on socioeconomic development at the national and sub-national scale and has paid little attention to land use and spatial changes at the regional and sub-regional scale. A large gap exists between the complex LUC and the underlying mechanisms in the cross-boundary regions of metropolitan areas in fast-urbanizing countries and the existing research. Therefore, this study aims to fill this gap by focusing on the LUC and its driving mechanisms in the Tongzhou-Wuqing-Langfang (TWL)—a typical cross-boundary region of the Beijing-Tianjin-Hebei Metropolitan Region in the Greater Beijing Region. As a rapidly urbanizing metropolitan area, the unbalanced development of the Greater Beijing Region and the contradiction in its human–land relationships appear to be prominent in the TWL region. The land management in the TWL region faces difficulties which are similar to those of other cross-boundary regions in fast-urbanizing metropolitan areas. First, the basic socioeconomic conditions vary between districts and counties, making it difficult for the local governments to implement land and planning policies in a coordinated manner. Second, as residents on both sides of the administrative boundary are mobile and may have access to each other’s information, the implementation of policies that have regional differences involving public interests will encounter significant resistance. Therefore, the case study of TWL may shed new light on understanding the current situation and the challenges of land management for metropolitan areas in fast-urbanizing countries.
2. Methods
2.1. Research Area
The Beijing-Tianjin-Hebei urban agglomeration is the “capital circle” of China. It is the largest and most dynamic region in northern China and has an important strategic position. Local governments initiated regional cooperation in the 1980s, while in the early 2000s the National Development and Reform Commission (NDRC) promoted regional coordination and communication and created regional development plans. In 2014, the coordinated development of Beijing-Tianjin-Hebei was promoted as a national strategy. The central government established the Beijing-Tianjin-Hebei Coordinated Development Leading Group and promoted transportation integration, industrial transfer, and joint air prevention and control in several vital sectors. However, the region still faces great problems, such as weak ecological and environmental protection, unbalanced development of the urban system, and a widening gap between regional and urban development, particularly due to the emergence of the poverty and pollution belts around Beijing. According to the Outline of China’s 14th Five-Year Plan and 2035 Vision Goals, China will accelerate the coordinated development of the Beijing-Tianjin-Hebei Metropolitan Region, build a high-quality sub-center in Beijing, and promote the integrated development of Beijing, Tianjin, and Hebei. This collaborative spatial development and sustainable urbanization of the Greater Beijing Region will have reference significance for the sustainable development of urban agglomerations in both China and other developing countries.
The TWL region is located at the intersection of Beijing, Tianjin, and Hebei and is a pilot demonstration area for the coordinated development of Beijing, Tianjin, and Hebei, as well as for regional policies (
Figure 1). As critical areas connecting the Greater Beijing Region, Beijing’s Tongzhou District, Tianjin’s Wuqing District, and Hebei’s Langfang City are closely interconnected spatially. They are important core areas for the coordinated development of Beijing-Tianjin-Hebei and are known as the TWL Golden Triangle. With a total land area of 8895 km
2, the TWL region is located at the lower end of the North China alluvial plain. It has a gentle terrain and a total population of 7.78 million. The expansion of urban construction land in recent years has resulted in the rapid development of the suburban and rural construction land areas and persistent LUC.
Although they share a geographical border, these three areas of the TWL region fall under different administrative jurisdictions. As districts of the centrally-administered municipalities, the Tongzhou and Wuqing District governments cannot independently formulate and implement land management policies and are subject to the unified deployment of the Beijing and Tianjin municipal governments. Moreover, due to the particularity of Beijing’s status, land use in Wuqing District and Langfang City often give way to Beijing’s planning needs.
The basic characteristics of Tongzhou, Wuqing, and Langfang (in 2019) are as follows. Tongzhou District is the eastern gate of Beijing. It borders Langfang’s three northern counties (Sanhe City, Dachang Hui Autonomous County, and Xianghe County) in the east and Tianjin’s Wuqing District and Langfang’s Guangyang District in the south. Tongzhou District covers a total area of 904 km2 and has a permanent population of 1.675 million. The per capita GDP of Tongzhou District is CNY 63,236, and the percentage contributions of the primary, secondary, and tertiary industries to the total GDP are 1.2%, 39.9%, and 59.0%, respectively. Wuqing District is located on the northwest edge of Tianjin, at the center of Beijing and Tianjin, and covers an area of 1575 km2. Its permanent population totals 1.183 million, and its per capita GDP is CNY 74,932. The composition of the GDP by the three industries is 3.9%, 33.1%, and 63.0%. Langfang lies between Beijing and Tianjin and has 10 counties (cities and districts) under its jurisdiction that border Beijing and Tianjin. Langfang’s three northern counties (Sanhe City, Dachang Hui Autonomous County, and Xianghe County) are enclaves, which are separated from the Langfang central urban area (Guangyang District and Anci District) and other counties and cities by Beijing and Tianjin. Langfang covers an area of 6415 km2 and has a permanent population of 4.921 million. Its per capita GDP is CNY 65,512, and the composition of the GDP by the three industries is 6.7%, 32.9%, and 60.4%.
2.2. Data Sources
The TWL region’s land use data were obtained by interpreting the Landsat series 1 remote sensing image product (Collection 1, Level 1). The research on the land use in the TWL region used a time series that spanned from 1990–2020, and the interpretation and sampling interval was approximately 10 years. To minimize the seasonal impact on vegetation, the image acquisition period was controlled from June–September, and the cloud cover was less than 10%. The data for each year were obtained from the latest available satellite remote sensing images that had the highest accuracy. This included Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 OLI satellite remote sensing data that had a spatial resolution of 30 m, Universal Transverse Mercator image projection, and a WGS84 datum level. We used the urban and rural construction land area data from the Beijing Land Use Survey (2008–2018) and the China Urban Construction Statistical Yearbook (2010–2020) to supplement the remote sensing interpretation results.
We obtained the socioeconomic data of Tongzhou, Wuqing, and Langfang (including the GDP, population, urbanization rate, and fixed asset investment) from the Beijing Statistical Yearbook (2010–2020), the Beijing Tongzhou District Statistical Yearbook (2010–2020), the Tianjin Statistical Yearbook (2010–2020), and the Hebei Economic Statistical Yearbook (2010–2020).
2.3. Research Methods
2.3.1. Research Framework
This research studies the LUC of the TWL region at the junction of Beijing-Tianjin-Hebei at the district and county level and explores the mechanism of administrative barriers on the land use differences. We adopt a human–computer interaction method to process the remote sensing image. By direct interpretation of Landsat remote sensing images and a corresponding analysis, this study can avoid the inaccuracy of the production and processing of the existing remote sensing land use data. We carry out the quantitative analysis of LUC using, for example, the land-use dynamic degree and kernel density analysis, as well as qualitative analysis of the land management policies and regulations, to reveal the temporal and spatial LUC within TWL and in each district/county. Furthermore, we adopt principal component analysis and stepwise multiple linear regression to analyze the driving factors of construction land change. Based on the analysis results, we provide beneficial suggestions for planning and controlling the boundary regions in metropolitan areas for fast-urbanizing countries (
Figure 2).
2.3.2. Pre-Treatment and Supervised Classification
First, we pre-processed the original remote sensing image data and then conducted a registration, correction, and image fusion to establish a sample selection and classification system. By referring to the main classification categories of several past LUC studies [
62,
63], we divided the land types into five categories: construction land, unused land, cultivated land, forest land, and water bodies.
Supervised classification is also known as training area classification, in which the training area is established according to the defined land pixels; the discriminant function is trained by using the statistical information features of the training area; and then, the trained function is used to identify the land types of other pixels. Considering both the large area and the complex land use types in the TWL region, we made a comparative attempt to obtain more accurate classification and interpretation results using the Mahalanobis distance, the K-nearest neighbor method, maximum likelihood, and other classifiers. We adopted the maximum likelihood method classifier, which proved to have the best effect in classical supervised classification, to conduct a preliminary classification of the land use types using the TWL region’s remote sensing images [
64]. We used ENVI Classic software to correct the results via a visual interpretation, and the remaining unrecognized land categories were reclassified. We then compared the final interpretation results with the latest national land survey results in China. The accuracy rate was over 85%; so, the results were considered reliable. The supervised classification allowed us to obtain the spatiotemporal evolution of LUC in the TWL region by comparing the spatial distribution land use maps in the four periods, while we could analyze the influence mechanism of the land management policies on LUC in the corresponding periods based on government documents and other materials.
2.3.3. Land-Use Dynamic Degree
The land-use dynamic degree (LUDD) is adopted to analyze the intensity of LUC within the TWL region as it directly reflects the intensity of a certain type of LUC within a certain time frame [
65]. The formula is expressed as follows:
where Pi represents the dynamic change degree of i-type land, |∆
Si →
j + ∆
Sj →
i| represents the absolute sum of conversions between this land use type and other land use types,
Si represents the total area of this land use type, and
t is the study period. The larger the Pi value, the greater the conversion intensity between the i-type land and other land types.
The comprehensive LUDD reflects the overall change in the land use quantity. The formula is expressed as follows:
where P represents the degree of dynamic change of all land use types in the study area, |
Si −
Si0| represents the difference between the final land area and the initial area for type
i,
S is the total land area in the study area, and
t is the study period. The larger the value of P, the more drastic the change in the whole land use type in the study area in a specific time frame.
2.3.4. Kernel Density Analysis
We adopt the kernel density analysis to analyze the urbanization process and the spatial expansion of built-up areas during the research period. As a non-parametric method, kernel density analysis can be easily implemented and can better reflect the distance attenuation effect in the spatial distribution of geographical phenomena [
66]. This study employed the kernel density analysis module in ArcGIS 10.5 to conduct a weighted kernel density analysis of the construction land in the TWL region in the four phases, taking the construction land patch area as the weight.
2.3.5. Analysis of Driving Factors
Principal component analysis (PCA) is adopted to analyze the driving factors of the LUC in the TWL region. PCA refers to the dimensionality reduction in complex socioeconomic multi-variables. It calculates feature vectors to remove information overlap among similar factors and extract variable combinations that significantly contribute to the LUC of the construction land in the TWL region.
The formula for the multiple stepwise linear regression model regarding the LUC of the construction land is as follows:
where Yi is the TWL region’s total construction land area in year
i;
xin is the observed value of the nth explanatory variable, that is, the driving factor in year
i; and
βi is the parameter to be estimated. Stepwise regression analysis requires that each time a new independent variable is introduced, the old independent variable should be individually tested to eliminate the independent variables with an insignificant partial regression square sum. To establish the optimal multiple linear regression equation of the driving factors of the LUC of the construction land in each region of the TWL, we introduced and removed the variables until no new variables could be introduced and no variables could be deleted. Based on the conclusions from the previous empirical studies [
67,
68], and by considering the availability of statistical data, we selected eight indicators from the Tongzhou, Wuqing, and Langfang statistical yearbooks (2010–2020) as the driving factors of the construction LUC: the output value of secondary industry (X1); the output value of tertiary industry (X2); the investment in fixed assets (X3); the total retail sales of social consumer goods (X4); the population of permanent residents (X5); the urbanization rate (X6); the per capita disposable income of the residents (X7); and the local general financial budget expenditure (X8).
4. Discussion
Although the terrain and locations of the TWL’s three administrative regions are similar, the LUC has shown great differences over the past 30 years. Tongzhou District and the three northern counties have had the fastest expansion of construction land over the past 20 years, showing a significant trend of mutual convergence and even mutual mergers. The development of Wuqing has been relatively slow, with no contiguous development area outside the central city. This has been mainly caused by administrative management and land policy differences. Tongzhou District and the three northern counties were fully guaranteed the construction land index supply from Beijing and Langfang City to construct Tongzhou New Town and North Langfang New Town. To a certain extent, the development of Wuqing has been restricted by the positioning of Tianjin’s main agricultural production areas and key ecological protection areas. The problems caused by the differences in the subordination and administrative ranks regarding the coordination of the borderline transboundary regions have also been commonly encountered in the borderline transboundary regions of metropolitan circles worldwide. The urban planning practices in many countries show that although constitutions and laws emphasize the need for local governments to collaborate in areas such as infrastructure, collaborative governance is often challenged by the long-standing reliance on the decision making of central governments; insufficient cooperation between local governments; and power struggles between different levels of government over decision making, funding, and responsibility [
74].
The leading factors of the construction land in the TWL region are the changes in the population, economic growth, urban–rural structure, and income level. The main driving factors are the changes in the resident population and the output values of the tertiary and secondary industries. On one hand, determining land demand via population remains the core of the land development policy in the TWL region, which is similar to other regions in China [
75]. Furthermore, population-structure prediction and land-demand prediction constitute the core of general land use planning. On the other hand, with economic growth as an assessment-oriented incentive, local governments are motivated to convert more land into commercial land to obtain higher unit outputs; therefore, they increase the amount of construction land [
76]. Moreover, fixed asset investments and urbanization rates are the main drivers of some regions. The growth of fixed asset investments improves the capital density of the urban unit land areas and facilitates the expansion of the urban built-up areas [
77]. The main driving factors of construction land expansion vary between different administrative regions as a result of the differences in land policies and the management of different administrative bodies.
To better promote local resource integration, the concept of cross-district management and control by means of spatial governance has gradually become the theoretical consensus and practical choice of all involved parties [
78]. In recent years, cross-administrative regional planning has been practiced in many countries worldwide to effectively alleviate the resource shortage problem in some cities and promote the transfer of administrative power and resources among cooperative cities. According to the relative strength of political and economic forces on both sides of an administrative boundary, cross-district management can be divided into four modes: strong–weak control, strong–strong control, weak–strong control, and weak–weak control [
79]. Due to the existence of Tongzhou District, which is the sub-center of Beijing, the cross-district management represented by the TWL region demonstrates a typical strong–weak control mode. In this mode, the cross-district spatial control of a weak region is led by an economically advantaged region and emphasizes a developed region’s cross-district control of an underdeveloped region in terms of the labor commuters and the land and real estate development. Specifically, developed regions pay more attention to the spatial transfer of production factors, such as industry, capital, and technology, while underdeveloped regions pay more attention to the outflow of the labor force. Meanwhile, the developed regions’ facilities and service capacities are strengthened, which also shows why the main driving factors of construction land expansion vary among different administrative regions.
Faced with the lack of unified policy-implementation subjects for land management in the TWL region, the Beijing-Tianjin-Hebei Coordinated Development Leading Group has made a series of beneficial attempts to reform cross-district control systems in recent years. Since 2016, the Hebei Provincial Government has made plans for transportation and environmental protection in Langfang’s three northern counties, as well as other adjacent counties to Beijing, to strengthen their coordination with Beijing’s sub-center planning. Moreover, a joint planning review mechanism has been established for the cross-boundary area under the participation of Beijing and Hebei. Under this system, the planning examination and approval power of the three northern counties is concentrated upward and is directly managed by the Hebei Provincial Government in order to effectively solve the problem of the lack of common interest in the coordination subjects caused by the unequal administrative levels of Beijing, Tianjin, and Langfang. Moreover, the “Framework Agreement on Promoting Tong-Wu-Lang Strategic Cooperation and Development (2017)” requires Tongzhou, Wuqing, and Langfang to cooperate in tourism, education, health, and other sectors. However, regarding the implementation of relevant planning in recent years, some practical difficulties remain in managing the coordinated development of the TWL region. Moreover, there is much consensus but not much action [
80]. The overall objectives and tasks of the coordination framework are clear but lack specific policy implementation plans, feasible measures, and detailed guidance on the division of the local governments’ responsibilities. Meanwhile, great institutional inertia remains in the single hierarchical territorial management system [
81]. Many local governments still make relatively independent decisions within their respective jurisdictions, while Tongzhou and Wuqing have made little progress in their coordinated governance attempts. Beijing and Tianjin, the superior governments to which Tongzhou and Wuqing belong, are both powerful governments that are strongly independent, and their awareness of the competition is usually greater than the awareness of cooperation. The concept of protecting local interests still exists; thus, it is difficult to achieve spontaneous coordination in regional development.
5. Conclusions
In the cross-boundary regions of metropolitan areas, LUC is a complex system that is affected by the market economy, national policy, population migration, and infrastructure construction, among other factors. This study explores the spatiotemporal LUC in the TWL region from 1990–2020 and analyzes the driving mechanism of the construction land from 2019–2019. The results show that: (1) The construction LUC in the TWL region shows “disordered expansion, rapid expansion, and incremental slowdown”, while each administrative region shows different and independent land use characteristics during different periods. From 2010, Tongzhou, Wuqing, and Langfang show a consistent trend of slowing the growth rate of construction land. (2) The driving factors of construction LUC in the TWL region from 2010–2019 are characterized by the interaction of “changes in population, economic growth, the urban–rural structure, and income level”, and the comprehensive score of these socioeconomic drivers has increased over the past 10 years. (3) According to the stepwise regression analysis results, the construction land growth in the TWL region is mainly driven by three factors: the permanent resident population, the output value of the tertiary industry, and the output value of the secondary industry. However, the dominant driving factors are different in the different administrative regions; that is, the construction land area in Tongzhou District is mainly affected by fixed asset investment; the construction land area in Wuqing District is mainly affected by the permanent population; and the construction land area in Langfang City is mainly affected by the urbanization rate. This indicates the impact of administrative boundaries on land use development and implies the necessity of strengthening coordinated development.
This study is an attempt to analyze the LUC in cross-boundary regions using remote sensing data in order to show the impact that different administrative subordinations have on land use development. However, some limitations remain. First, due to the limited resolutions of the available remote sensing image data, this study’s classification of land use types was relatively rough, and multiple land use types within the construction land could not be further distinguished. The follow-up research should use higher-quality remote sensing data and combine it with other data sources to provide a more detailed analysis. Second, this study revealed that the dominant driving forces of the construction land within different administrative regions were different and attempted to explore the associated policy factors. However, the follow-up research should show the correlations between them in a more systematic manner in order to suggest cross-administrative land management in metropolitan areas, which requires urgent attention.