Revealing the Driving Mechanism of Urban Development Based on the 3D Building Morphological Changes in the Beijing–Tianjin–Hebei Region

Highlights

What are the main findings?

• Use concentric circle analysis to explore the spatiotemporal variations in building height and extract new building heights from the building height data.
• Use panel analysis to examine the drivers of changes in building height.

What are the implications of the main findings?

• The mechanisms underlying changes in the 3D morphology of comparable cities in the Beijing–Tianjin–Hebei region exhibit similarities.
• Various social factors influence building heights in new development areas in different ways and to varying degrees.

Abstract

Since the late 1970s, during the period of reform and opening-up, China has emerged as one of the fastest-urbanizing countries in the world. The high speed of urbanization and three-dimensional (3D) morphological changes in urban buildings may lead to significant urban challenges, including air and light pollution, the urban heat island effect, and strained human-land relationships. It is therefore crucial to elucidate the pattern of urban 3D morphological change and to investigate the factors that influence these changes. This study proposed a novel method of filtering new buildings using the ALOS and GBH building height data, combined with statistical data and panel data, focusing on spatial-temporal variations and their driving factors of 3D building morphological changes between 2002 and 2020 for 11 prefecture-level cities and 2 municipalities in the Beijing–Tianjin–Hebei (BTH) region. Our findings indicate that the 3D morphological change mechanisms of the same types of cities in the BTH region are similar. The total population negatively impacts the height of newly constructed building areas, and the optimization of the industrial structure positively influences the height of newly developed land. Based on our findings, we propose actionable recommendations to foster sustainable urban development in the BTH region.

1. Introduction

As a widespread global phenomenon, urban land expansion poses far-reaching influences on sustainability and resource management [1]. China has undergone an extraordinary urbanization boom since 1978 [2]. Existing literature confirms that the expansion of urban land in China outpaces that of all other countries globally [3]. China gradually established its socialist market economic system in the 1990s, which greatly boosted industrialization and urbanization [4]. China’s built-up area expanded from 22,439.3 km² in 2000 to 62,420.5 km² in 2021, almost tripling within 21 years and demonstrating the rapid progress of urbanization. Its urbanization rate is projected to hit 70% by 2035 [5]. Such prospects point to ample room for urban development, highlighting the importance of elaborate planning for long-term sustainability. The rapid expansion of urban construction land and the resulting decline of ecological space may generate adverse impacts on the environment and society [6], and further weaken the region’s capacity for sustainable development [7]. Against this background, the dramatic expansion of urban land in China has drawn considerable research attention [8]. At the same time, a more comprehensive understanding of urban sprawl and its underlying drivers is essential to evaluate the effects of potential future development on the environment, economy, and society [9]. Urban sprawl has been extensively modeled through panel data analysis and multiple linear regression, providing insights into its dynamics [10,11].

Studies have shown that local urban expansion is significantly influenced by population growth and economic development [12], both of which are key drivers of impervious surface formation in prefecture-level cities [13]. Furthermore, research indicates that market forces, policy factors, geographical conditions, and underlying urban vitality play a crucial role in influencing urban sprawl in China [2,14,15,16]. Among these factors, the land market is a particularly important aspect of marketization. The growth of the land market can enhance the efficiency of land use for urban construction and optimize the distribution of land resources [17,18,19]. In contrast, other studies indicate that land sales and financing play a positive role in promoting the growth of land for urban development [20,21]. At the same time, some researchers have found that the land market’s contribution to urban construction land growth varies significantly across different regions and urban contexts [22,23,24], and that the extent of this contribution also influences the role of government in regulating land use and urban development. Therefore, it is necessary to examine the mechanisms through which land markets influence urban expansion at the regional and urban levels, so that regions and cities can adopt appropriate macro-regulatory measures to guide the direction and pace of urban expansion and achieve coordinated, sustainable urban development.

However, most current studies focus primarily on urban boundary expansion in two-dimensional space, whereas urban expansion actually encompasses both horizontal and vertical expansion [25,26,27]. Compared to the flat two-dimensional form, the three-dimensional urban form is better able to express the main characteristics of a city. Horizontal expansion is closely linked to vertical expansion, as a land unit may construct high-rise buildings on its original site if it seeks to increase its scale and no nearby land is available for development [28]. A three-dimensional perspective is essential because the rise of high-rise buildings makes it impossible to accurately capture urban expansion’s spatial and temporal characteristics from a two-dimensional view. The three-dimensional expansion of cities is primarily reflected in the vertical growth of buildings [8].

In recent years, many empirical studies have noticed this and focused more on the three-dimensional expansion of cities. Some scholars have explored the three-dimensional spatial expansion patterns and mechanisms of area in Yangzhou City, Jiangsu Province, Guangdong–Hong Kong–Macao Greater Bay Area, and the main city of Wuhan based on geographic information system (GIS) tools and high-resolution remote sensing imagery, land-use simulation models, and random forest algorithms [29,30,31]. Meanwhile, some scholars have studied the spatial and temporal characteristics of the three-dimensional expansion of Chinese prefecture-level cities and used geo-probes to investigate their influencing factors, and found that governmental and economic factors have the greatest influence on the three-dimensional expansion of Chinese cities, while social factors are in the middle, and natural factors have the weakest explanatory power [32].

Although abundant existing studies have probed into urban sprawl from three-dimensional morphological perspectives using various open-access building height datasets, most existing works directly employ height data for statistical analysis rather than developing specialized algorithms to dynamically extract newly increased built-up land. Against this research gap, this study assembles a decadal two-period dataset integrating publicly available 2010 building height data and our research group’s self-developed open-access 2020 building height dataset, and centrally constructs a refined extraction approach targeting incremental construction land as the core technical contribution of this paper. With incremental land information acquired via the proposed extraction method, this study further supplements land market indicators into panel regression and explores the driving mechanisms of three-dimensional urban sprawl across the BTH region. Accordingly, the specific objectives of the current study are as follows:

(1)

Using the BTH region as the study area, a panel linear regression model was used for quantitative analysis;

(2)

Building height in China of Advanced Land Observing Satellite-1 World 3D (AW3D) in 2010 and Global maps of building heights (GBH) in 2020 are selected to describe building heights in 2002–2010 and 2011–2017, respectively. Within ten years, we made reasonable assumptions about the mechanism of urban expansion and urban renewal;

(3)

Excluding the impact of rural and suburban areas, which have large data errors, only the built-up part of the city is analyzed.

In this paper, we start from the perspective of three-dimensional urban morphology, explore the change mechanism of spatial expansion and its influencing factors, and analyze the influence and system of land market development on the role of vertical urban expansion. This research holds significant implications for studies on urban sprawl and land market development. Additionally, we seek to propose a sound regulatory framework for urban expansion and effective policy recommendations, which play a crucial role in promoting rational urban expansion and fostering coordinated regional sustainable development.

2. Materials and Methods

2.1. Study Area

The Beijing–Tianjin–Hebei Metropolitan Economic Circle, also known as the Beijing–Tianjin–Hebei (BTH) Urban Agglomeration, is situated between 113°27′ and 119°50′ east and between 36°03′ and 42°40′ north. The BTH region encompass 3 provinces and 13 cities with a covering area of 216,000 km², a population of 109 million, and a GDP of 10 trillion yuan. The urbanization process in the BTH region has progressed steadily characterized by integrated and coordinated development. The construction area has consistently expanded, with urban land use exhibiting an overall trend of growth, and an increase in impervious surfaces [33]. The cooperative development of the BTH region is a major national strategy, as it constitutes a significant metropolitan area. Consequently, the region’s development status serves as a barometer for the nation’s overall growth, making research at the city level crucial. Such research not only provides valuable insights into the region’s development but also lays the foundation for future studies covering the entire country. Therefore, we selected the BTH region as our study area (Figure 1).

2.2. Data

This study extensively utilizes both statistical and remote sensing data. Statistical data primarily supports panel analysis, while remote sensing data is mainly employed for economic ring analysis and the extraction of newly constructed building land after processing.

2.2.1. Remote Sensing Data

The building height data mainly used in this study include Chinese building height with a spatial resolution of 30 m (AW3D), estimated by ALOS [34], and global building height data with a spatial resolution of 150 m (GBH), derived from spaceborne LiDAR [35]. These datasets represent the building height of the BTH region in 2010 and 2020 respectively (Figure 1b,c). Compared with other contemporary data, AW3D has less no mask data and broken pixels, which is easier to analyze and process; whereas GBH is trained based on GEDI-based samples, which reflects the trend more accurately.

To prevent excessive systematic differences in building heights between the two sources and inversion methods, we first resampled the AW3D data to a 150 m resolution. Using Google Earth Pro, we randomly selected 55 building grids that remained unchanged between 2010 and 2020, and evaluated the systematic error by calculating the mean and standard deviation of the two datasets. The results showed an average of 0.4 m and a standard deviation of 7.96 m. Overall, the global systematic deviation between the two datasets was extremely small, and the overall height reference was largely consistent, but the standard deviation was relatively large. After excluding high-rise buildings (over 27 m) and recalculating the data, the average was found to be 2.54 m, with a standard deviation of 3.33 m. This indicates that the relatively large standard deviation is primarily attributable to high-rise buildings, whereas the systematic error for low- and mid-rise buildings is less than 3 m (approximately the height of one story), resulting in significantly reduced variability and good data consistency. Overall, the error in the building height data for both groups is acceptable.

Additionally, we need to use some auxiliary data in the analysis process (

Table 1. Remote sensing data definition, spatial resolution, period and source.

Remote Sensing Data	Spatial Resolution	Period	Source
Global maps of building heights (GBH)	150 m	2020	Ma et al. (2023) [35]
Advanced Land Observing Satellite-1 World 3D (AW3D)	30 m	2010	Huang et al. (2022) [34]
A global dataset of annual urban extents (built-up area)	1 km	1992–2020	Zhao et al. (2022) [37]
Global Artificial Impervious Area (GAIA)	30 m	1985–2018	Gong et al. (2020) [36]
China Building Rooftop Area (CBRA)	2.5 m	2016–2021	Liu et al. (2023) [38]

). Global Artificial Impervious Area (GAIA) with a spatial resolution of 30 m [36] is required to remove the influence of non-building factors such as vegetation, water bodies, and roads. Compared to other impervious surface data, GAIA has better mapping performance for semi-arid climate zones. At the same time, GAIA can well capture the growth of urban artificial impervious areas with different growth rates, which is more applicable to our study area and research objectives. In acquiring the city area, we use the built-up area data with a spatial resolution of 1 km [37] to extract the city center and the main urban area that characterizes the city. Compared with other publicly available built-up area data, it has longer product time series and data continuity. In the process of extracting new building sites, we need to determine which parts of the city are buildings, so we use building boundary data of China Building Rooftop Area (CBRA) with a spatial resolution of 2.5 m [38], where we choose the building boundaries of 2020 because the accuracy of the data in this year is relatively higher.

2.2.2. Statistical Data

Marketization is considered a significant factor in urban expansion [2], while population growth and economic development are the primary drivers of urbanization in China [39]. In this study, the total population (in ten thousand persons), actual per capita urban GDP (yuan), and industrial structure are used as statistical indicators to measure economic development. Conversely, the total area of primary land market transactions (in hectares), the level of land marketization, and the average land transaction price (in ten thousand yuan per hectare) are employed to assess land market development (

Table 2. Name, definition and source of statistical data for 2005, 2008, 2011, 2014, 2017.

Statistical Data	Name	Definition	Source
Total population	POP	Total population of the municipal district (10,000 persons)	China City Statistical Yearbook
Actual per capita GDP	RPGDP	Gross regional product per capita of the municipal district (yuan)	China City Statistical Yearbook
Industrial structure	INDS	The proportion of tertiary and secondary industries in the gross regional product	China City Statistical Yearbook
Total area of primary land market transactions	PLMA	Total area transferred by agreement, tender, auction, listing, allocation, leasing and other means of supply (Ha)	China Land and Resources Statistical Yearbook
Land marketization level	CLM	Proportion of land sold in the primary land market through “auction” to the total number of compensated land sales in the primary land market	China Land and Resources Statistical Yearbook
Average land transaction price	CLP	Ratio of total urban land concessions in compensation to total land (10,000 yuan/Ha)	China Land and Resources Statistical Yearbook

). The data regarding total urban population, per capita urban GDP, and industrial structure upgrading are sourced from the China City Statistical Yearbook, while the data on primary market transactions, land marketization level, and average land transaction price are obtained from the China Land and Resources Statistical Yearbook.

(1)

Population growth. Population growth is recognized as a direct driver of urban built-up area expansion in developing countries [40]. In this study, total population is employed as the indicator for measuring population growth.

(2)

Economic growth. Economic development significantly influences the expansion of urban built-up areas [12,41]. While GDP per capita serves as the primary indicator of a city’s economic development [42], the labor force also transitioned from the primary sector to the secondary and tertiary sectors during the study period [43]. Therefore, this study utilizes actual per capita urban GDP and industrial structure as measures of economic development.

(3)

Land marketization. Theoretically, the expansion of land for urban development is affected by the size of the land market, the degree of marketization, and the price of land transactions [22]. This study employs the total area of primary land market transactions is gauge land market size, the level of land marketization to assess the degree of marketization, and the average land transaction price to evaluate land transaction costs.

2.3. Methodology

Three primary research approaches were employed in this study: concentric ring analysis, newly added building land extraction, and panel analysis. Figure 2 illustrates the research framework. The details are shown below.

2.3.1. Concentric Rings

A technique widely used to analyze urban form and expansion involves dividing a city into several concentrated zones. Researchers frequently discuss changes in spatial scale and density characteristics beyond the city center [44]. According to [45], the concentric ring method uses the city center as a focal point and extends outward by a specific distance to delineate a series of buffer zones. The buffer rings should encompass the principal areas of the city to maintain equidistant spacing [46]. The city center is typically defined in a meaningful context, usually centered on the birthplace of the city or the city’s central business district (CBD) [47]. However, since not all cities possess a strictly defined CBD in the sense of the word, in the concentric rings analysis, we first determined the urban centers of each city by calculating the center of gravity of the main urban areas within the built-up zones as of 1992, and then combining this with the original city centers and CBDs of the old urban areas. Concentric ring buffers are then constructed at intervals of every 1 km radius distance from the city center and extending to either the outermost perimeter of the main city or a distance of 20 km from the city center, and the maximum distance determined based on previous research [48]. Then, we calculated the mean building heights within each buffer ring for the two years 2010 and 2020, intercepting the built-up area main urban district portions of the AW3D and GBH data for each city. Finally, we generated line graphs to illustrate changes in the three-dimensional morphology of the buildings in large, medium, and small cities in the BTH region over the previous ten years. The classification of cities into large, medium, and small categories is based on population. However, due to the small size of the BTH region, there are no small cities in the region according to China’s classification standards. Instead, according to the 2020 urban population, a large city is defined as that with more than 10 million (>10 million) inhabitants, a medium city as that with 5–10 million (5–10 million) inhabitants, and a small city as having between 1 and 5 million (≤5 million) inhabitants.

2.3.2. Newly Added Building Land Extraction

When using a one or two-year interval, the amount of newly added building land is minimal, resulting in weaker explanatory power. At the same time, since part of the data acquisition is only updated to 2017, we use 2017 as the end point of the experimental period; therefore, if four or more years are used as time interval, it will lead to a small number of periods used for the panel data, which may affect the accuracy of the model. Consequently, a three-year interval was selected to extract the newly added building land from the AW3D data for the periods of 2002–2005, 2005–2008, and the GBH data for the periods of 2008–2011, 2011–2014, and 2014–2017. Additionally, we masked both datasets with built-up area data and municipal district data of each period to eliminate the influence of rural and suburban areas, which are not accounted for in the statistical indicators and may introduce significant errors.

The specific extraction method is as follows: we take the intersection of the building boundary data with the newly added impervious surface data for each period to obtain the new building data. Based on the results of multiple tests, we selected pixels in each image where the ratio of new construction area to building boundary area is greater than or equal to 16%, and the ratio of new construction area to total image area is greater than or equal to 2%. This approach more accurately reflects changes in building height within the image. The reason for choosing 2% is that we determine threshold values based on the physical size of pixels. In this study, the area of a single pixel is either 22,500 square meters or 900 square meters. If the area proportion threshold is set to 1%, the corresponding minimum identifiable area is only 9 m², making it highly likely that scattered noise and minor land features will be misidentified as new buildings, resulting in significant false positives; If set to 3%, the minimum identifiable area reaches 675 m², which would filter out a large number of small new building patches, resulting in severe false negatives. After weighing the pros and cons of these two extreme scenarios, 2% was ultimately selected as the threshold for this study, striking a balance between suppressing false positives and reducing false negatives. To determine the threshold for the ratio of new buildings to the building boundary, we first conducted gradient tests, examining three gradients: 10%, 20%, and 50% (Figure 3). Visual interpretation results indicate that the 20% extraction performance is clearly superior to that of 10% and 50%. We further narrowed the testing range around 20% by adding three adjacent thresholds: 15%, 16%, and 17%. Comparison revealed that while all three thresholds performed quite well, the 15% threshold was more likely to incorrectly classify unchanged building areas as new additions; 17% is more likely to omit patches of building edge expansion, whereas 16% effectively avoids false positives while maximizing the retention of genuine new building information (Figure 3). Additionally, by counting the number of extracted pixels at each gradient, it was observed that the total number of pixels fluctuated very little within the 15–17% range (

Table 3. The number of pixels corresponding to each threshold for the proportion of new floor area and building boundaries across different time periods.

	Ratio	10%	15%	16%	17%	20%	50%
Period
2002–2005		147,509	139,551	138,065	136,653	132,352	98,000
2005–2008		140,916	134,341	133,091	131,794	128,294	99,416
2008–2011		25,928	21,508	20,808	20,127	18,308	8489
2011–2014		23,520	19,673	19,046	18,479	16,920	8308
2014–2017		17,787	14,172	13,576	13,041	11,718	4962
Total		355,660	329,245	324,586	320,094	307,592	219,175

), demonstrating the stability of this threshold range. Consequently, 16% was selected as the optimal threshold.

To prevent the same pixel from characterizing the newly added building land in different periods, we designate the period with the largest area ratio for each pixel as the final period, thereby identifying the newly added building land. Finally, we calculate the average height of newly constructed buildings for each city during each period. This approach is based on the assumptions that (1) the BTH region has experienced rapid urbanization over the past two decades, with most cities expanding year-on-year, and (2) new buildings in the BTH region were relatively new in the period 2002–2017, with a low probability of renewal in the short term.

2.3.3. Panel Multivariate Regression Analysis

Panel analysis has been employed in numerous studies investigating the factors influencing urban sprawl [12,49,50,51]. To examine the impact of the land market on the building height of newly developed land in the BTH region, this study utilized panel multivariate regression analysis. The mean building height of newly developed land in each city during each period serves as the dependent variable, while the independent variables include total population, actual per capita GDP, industrial structure, total area of primary land market transactions, land marketization level, and average land transaction price in each city at the end of each period. The following panel multivariate regression model is constructed:

$$y_{it}={\alpha }_0+\alpha X_{it}+{\epsilon }_{it}$$

(1)

$${Height}_{it}={\alpha }_0+{\alpha }_1{POP}_{it}+{\alpha }_2{RPGDP}_{it}+{\alpha }_3{INDS}_{it}+{\alpha }_4{PLMA}_{it}+{\alpha }_5{CLM}_{it}+{\alpha }_6{CLP}_{it}+u_i+{\epsilon }_{it}$$

(2)

where y is the dependent variable, i denotes each city, and t denotes each period. α₀ denotes the intercept, α is the k × 1 matrix of coefficients, X is the 1 × k vector of independent variables, and ε denotes the random error.

The process of constructing the panel multivariate regression model is as follows:

To test the multicollinearity problem, we first calculate the variance inflation factor (VIF) for each variable. Next, we compute the dependent variable’s Moran’s I for each period to assess its spatial correlation and determine whether selecting a spatial panel model is required. Finally, according to the results of the Hausman test (

Table 4. Results of Hausman test.

Chi-Sq Statistic	p-Value	Null Hypothesis	Regression Model Effects
0.26	0.9679	Accept	Random effects

), we choose a random effect.

3. Results

3.1. Three-Dimensional Changes in Urban Form

Figure 1 shows the BTH region’s urban structure in three dimensions in 2010 and 2020, respectively (Figure 1c). In general, whether in 2010 or 2020, the highest building heights in the BTH region were primarily concentrated in city centers. Only the center of Beijing had a large concentration of high building height values in 2010, with other clusters of tall buildings were scattered across the city centers of other cities. Outside of these areas, building heights were generally lower, with most falling within the range of 2–6 m. In 2020, a wide range of high building values can be seen in downtown Tianjin and downtown Shijiazhuang, while the range of high building values in other cities is also generally larger than in 2010. Meanwhile, on other territories covered by buildings, the low values observed in 2010 have been replaced by a preponderance of building median values between 9 and 15 m, with a smaller number of low values found virtually exclusively at the periphery of cities. The three-dimensional morphology of the areas closer to the Beijing city center has remained almost unchanged over the past ten years (Figure 1b), while the areas outside Beijing’s fourth ring road have primarily changed, as evidenced by the emergence of taller buildings in these areas and the filling in of previously empty spaces. The central urban area of Tianjin features multi-layered morphological expansion. Not only did the average building height increase remarkably, but mid-rise and high-rise buildings also expanded their spatial scope. By 2020, low-rise to high-rise buildings had replaced a large number of vacant plots recorded in 2010.

Figure 4 shows the average building heights in the BTH region’s cities between 2010 and 2020. As shown in Figure 4a, the regional average building height was marginally greater in 2020 than in 2010, with the mean difference falling between 2 and 5 m. Around the core areas of large cities (Figure 4b,c), the average building heights of the two periods were close to each other. Although 2020 generally saw higher values, some distance nodes presented higher building heights in 2010. In large cities, zones with obvious height gaps were primarily located away from urban centers. Medium-sized cities (Figure 4d–f) had smaller buffer rings than large cities. Their average building heights mostly increased by about 3 m from 2010 to 2020. At certain distance nodes, there are instances where the building heights of the two phases are close to each other. Most small cities (Figure 4g–n) witnessed buffer ring expansion, with Cangzhou (Figure 4i) expanding its urban boundary by 6 km. Overall, small cities had the narrowest buffer rings, with the minimum coverage of 6 km. Their average building heights in 2020 were consistently higher than in 2010, and the inter-period gaps were greater than those in larger and medium cities. The maximum height difference in Chengde (Figure 4h) was 13 m, and the maximum growth of average building height reached 8 m. Unlike large and medium cities, large height differences in small cities mainly occurred near urban centers.

A number of representative cities exhibit unique spatial patterns. In Beijing (Figure 4b), the two average building height curves intersect. In the 0–9 km buffer ring, 2010 had larger mean values, with their difference forming a U-shaped trend. In the 9–20 km zone, building heights were higher in 2020, with the disparity expanding gradually and eventually stabilizing at 3–4 m. Shijiazhuang (Figure 4d) featured a relatively high average building height in 2010. Its building height rose steadily from 2010 to 2020. While the two curves are visually similar, the average height increased by nearly 7 m in the 6 km buffer ring. Finally, when looking at Tangshan (Figure 4l), it shows no intersection between the two curves, with a persistent gap. Starting from a low level in 2010, Tangshan witnessed a substantial rise in building height over the decade. The growth was concentrated within 6 km of the city center, where the average increase reached 9 m. Eventually, its 2020 average building height exceeded the overall average of the BTH region.

According to the Unified Standard for Design of Civil Buildings (GB 50352-2019) [52], 27 m is the threshold height distinguishing high-rise residential buildings from mid-rise residential buildings, and low-rise buildings generally refer to structures with 1 to 3 stories. However, some studies use 10 m as the boundary between low-rise and mid-rise buildings [53]. Considering that the typical height of a single story is 3 to 4 m, and given that the Guidelines for Skyline Planning in Key Urban Areas of Tianjin (2022) stipulate that the height of historic and cultural districts must not exceed 12 m, we have decided to adopt 12 m as the boundary between low-rise and mid-rise buildings.

Figure 5 illustrates the decadal changes in the proportional composition of low-rise, mid-rise, and high-rise buildings across different city tiers. At the regional scale of the BTH region, the proportion of high-rise buildings remained relatively stable, rising slightly from 2.71% to 2.73%. The primary structural transition occurred from low-rise to mid-rise buildings. In large cities, the share of low-rise buildings decreased from 77.77% to 66.6%, while the proportion of mid-rise buildings increased from 16.83% to 28.3%, and high-rise buildings declined moderately from 5.4% to 5.1%. This indicates that large cities have undergone vertical transformation dominated by low-to mid-rise replacement, accompanied by a slight reduction in high-rise building proportion. For medium-sized cities, the high-rise building ratio remained nearly unchanged, with only a small fraction of low-rise buildings converted to mid-rise structures. By contrast, small cities experienced a decline in low-rise building coverage and a simultaneous growth in mid-rise and high-rise proportions, reflecting a dual transition from low-rise to mid-rise and high-rise buildings. In addition, in 2010, small cities featured the highest proportion of low-rise buildings, whereas large cities dominated in mid-rise and high-rise building shares across the region.

3.2. Building Heights on Newly Added Building Land

The average height of buildings on newly added building land in each city across the five periods is shown (Figure 6a). For both large and small cities, the average height of newly added buildings in the later three periods was generally higher than that in the first two periods. In contrast, medium-sized cities exhibited relatively stable average building heights across all five periods. For most cities, building height variations were limited during 2002–2005 and 2005–2008, while a notable growth point emerged in 2008–2011. The three consecutive periods of 2008–2011, 2011–2014, and 2014–2017 shared a consistent changing trend, which generally increased first and then decreased.

Figure 7 shows the changes in the percentage of low, medium, and high structures in each type of city’s newly constructed region throughout five periods. At the regional scale of the BTH region, the proportion of low-rise buildings in new construction follows a U-shaped trend over the five periods, decreasing first and then rebounding. By comparison, the share of mid-rise buildings fluctuates repeatedly in the sequence of decline, growth, decline and growth, yet its period-to-period variation remains modest. Compared to the tendency for low-rise structures, the proportion of high-rise buildings exhibits an inverted U-shaped pattern that first increases and then decreases. The single factor that sets the large cities apart from the rest of the BTH region is the patterns regarding low-rise buildings, which, in contrast to the U-shaped pattern observed throughout the region, have been declining in the large cities. For medium-sized cities, the share of mid-rise buildings fluctuates in a wave pattern: decreasing, rising and then falling again. Their high-rise building proportion follows an inverted U-shaped trend, consistent with that of large cities. Meanwhile, the low-rise building ratio in medium-sized cities also fluctuates, with a trend of rise, fall and subsequent rise. As for small cities, they share a similar low-rise building trend with medium-sized cities, whereas their mid-rise and high-rise building proportions show distinct patterns. The share of high-rise buildings in small cities fluctuates by falling first, then rising and declining again. In contrast, mid-rise buildings follow an inverted U-shaped trend, which rises at first and then drops.

When examining building height composition, the proportion of low-rise buildings rose most markedly in large cities during the first two periods, with small cities recording similar growth over the same timeframe. Both large and small cities experienced the largest increase in the share of mid-rise buildings in the third period. The proportion of high-rise buildings grew the most in small cities in the fourth period, and large cities ranked second in terms of growth magnitude during this period.

Figure 8 and Figure 9 present the results of construction land extraction, along with historical remote sensing images covering parts of Beijing, Tianjin and Shijiazhuang. The results reveal that the area of newly developed construction land gradually declined over the study period, and its spatial distribution became increasingly dispersed. Based on the same datasets (AW3D2010 and GBH2020), comparisons of remote sensing images of the same city across different periods prove that our extraction results can effectively capture the growth of construction land in each phase. This is especially evident in areas with obvious building clusters, which are marked with data frames in the figures. Despite rigorous quality control, a small number of extraction errors, omissions and inaccurate judgments on the development time of new construction land may still exist.

At the same time, we conducted stratified random sampling of the results for new construction sites, yielding a total of 200 samples. We then compared these against historical imagery in Google Earth Pro (Version 7.3.7), ultimately obtaining 178 valid samples. Among these, 6 samples were incorrectly identified as having undergone changes when they had not, and 17 were misclassified; the overall accuracy rate was 87.08%. We then generated a confusion matrix (Figure 10). The results showed that the producer accuracy for each building height change period was above 87%, indicating stable overall recognition performance. Misclassifications were primarily concentrated between adjacent periods, with no severe cross-period confusion, suggesting that the model effectively distinguishes between different construction phases. Misclassifications occurred only at the boundaries of adjacent periods due to slight feature similarity. At the same time, we used Google Earth Pro to determine whether a building confirmed to have been newly constructed between 2002 and 2005 or between 2005 and 2008 had been demolished by 2010; or, when a building was indeed newly constructed during the periods 2008–2011, 2011–2014, or 2014–2017, whether it was demolished in 2020, to verify the second hypothesis listed in Section 2.3.2. The results are shown in

Table 5. Number of buildings constructed during each of the five periods that were demolished in 2010 or 2020.

Period	Number of New Buildings	Number of Buildings Demolished
2002–2005	28	1
2005–2008	30	1
2008–2011	37	2
2011–2014	39	2
2014–2017	38	1
Total	172	7

. The results show that the overall removal rate was only 4.06%, confirming the validity of our hypothesis.

3.3. Panel Multivariate Regression Analysis Results

A great deal of work went into figuring out the model’s ultimate form (

Table 6. Results of multicollinearity test and spatial autocorrelation test.

Multicollinearity Test		Spatial Autocorrelation Test
Variable	VIF	Period	Moran’ I	p-Value
POP	4.03	1	−0.208	0.287
RPGDP	2.03	2	−0.023	0.366
INDS	2.46	3	−0.205	0.287
PLMA	1.87	4	−0.693	0.001
CLM	1.58	5	−0.420	0.069
CLP	2.60	-	-	-

Note: Period 1: 2002–2005, Period 2: 2005–2008, Period 3: 2008–2011, Period 4: 2011–2014, Period 5: 2014–2017.

). The variance inflation factor (VIF) calculation findings indicate that there is essentially no multicollinearity issue because all of the independent variables’ VIF values are less than 5. Then, we calculated the Moran’s I of the dependent variable for each period. The results show that only the value of Moran’s I for the period 2011–2014 is −0.693 and is significant at 99% level, which represents a very strong negative spatial correlation. The results of the segmented global Moran’s I test indicate that significant spatial negative correlation was observed in only a single time period during the study period, while no obvious spatial dependence was observed among cross-sectional units in the remaining periods. This suggests that spatial correlation is merely a temporary, localized phenomenon and does not result in stable spatial interactions over the entire cycle; thus, the assumption of sample independence in traditional OLS holds true for the vast majority of observation periods. Therefore, the OLS regression results effectively reflect the overall impact of the explanatory variables on building height. With regard to the specific period during which spatial negative correlation occurred, this paper will provide a detailed analysis of the specific effects observed across regions during that phase in the discussion section, taking into account factors such as policy restrictions and regional development. In light of this, we arrived at the regression model equation shown in Formula (2).

Table 7. Panel estimation result.

Variable	Coefficient
POP	−0.00252 **
RPGDP	0.00002
INDS	1.25271 **
PLMA	0.00022 *
CLM	5.72519 ***
CLP	0.00009 *
Constant	8.66559 ***
R2	0.5380

Note: * indicates significance at the 90% level. ** indicates significance at the 95% level. *** indicates significance at the 99% level.

displays the model regression results. Except for the actual per capita urban GDP index, which is not significant, the coefficients of the total area of primary land market transactions and average land transaction price are 0.00022 and 0.00009 at a 90% confidence level, which suggests that both factors have a weak association with the height of new building sites. And by a confidence level of 95%, the coefficients of total population and industrial structure are −0.00252 and 1.25271. Statistically, total population correlates negatively with the height of newly built sites: each additional 10,000 residents are associated with a 0.00252 m decline in new building height. By contrast, the industrial structure index shows a positive correlation with new building height, with a 1 percentage-point rise in industrial structure linked to a 0.0125271 m rise in new building height. At the same time, we find that with the significance of 99% level, the coefficient of land marketization level is 5.72519, revealing a strong positive statistical association between land marketization and new building height; a 1% growth in land marketization level is accompanied by a 0.0572519 m increase in the height of newly constructed sites. Finally, The R² of the whole model is 0.5380, which indicates that the regression results are reliable. Overall, population growth lacks a statistically robust correlation with new building height, whereas economic factors represented by industrial structure are positively correlated with vertical construction. All three land market indicators of core research interest display statistically significant positive correlations with new building height, among which land marketization level presents the strongest positive statistical association with the dependent variable.

4. Discussion

4.1. Influential Factors of Three-Dimensional Vertical Morphology in BTH

The Moran’s I test for spatial autocorrelation shows that only Period 4 (2011–2014) exhibits a significant negative Moran’s I (−0.693, p < 0.01), indicating a “high-low” alternating spatial pattern of newly built building heights, which is consistent with our finding (Figure 6) that new buildings in small cities were generally higher than those in large and medium-sized cities during this period. This negative correlation can likely be explained from two perspectives: building height regulation and urban development. First, strict height limits and historic preservation policies in major cities like Beijing have restricted the height of new buildings, whereas smaller cities face fewer such restrictions. Moreover, as these smaller cities were undergoing rapid urban development during this period, they prioritized high-rise development to create urban landmarks. Second, large cities enforce stricter floor area ratio (FAR) controls than small cities, while Hebei is vigorously promoting new-style urbanization, allowing for increased FAR in new development zones to encourage high-rise construction. At the same time, the concept of the “Beijing Metropolitan Economic Circle” has been gaining increasing attention during this period. Expectations regarding housing prices have risen in the small cities and county towns surrounding Beijing, and developers have tended to boost property values by constructing high-rise buildings. Small cities have primarily focused on urban development centered around high-rise construction projects. Consequently, new buildings in small and medium-sized cities tend to be taller, while those in large cities are not as tall as those in smaller cities, resulting in a “high-low” contrast characterized by negative spatial autocorrelation. Although this significant spatial effect is limited to a single phase and is not a universal characteristic across the entire cycle, and given the context of height restriction policies and urban development as revealed by the heterogeneity analysis, this spatial anomaly appears to be a temporary phenomenon. However, it is important to note that this negative spatial autocorrelation may still have a potential impact on the overall panel regression results. Specifically, if this spatial pattern of alternating high and low building heights in new construction is driven primarily by factors such as land marketization and industrial structure, ignoring this spatial characteristic could lead to biased assessments of the statistical significance of key variables. In future research, we should strive to address this issue.

For new urban construction sites, a complicated multifactorial ground cover type, there is still a deficiency in change detection research [54], and the extraction of building heights for newly constructed sites is even less studied. Our results show that there is a negative correlation between Total population and urban sprawl, whereas, in previous studies, the population factor plays an important positive driving role [49,55,56]). Population growth urgently requires cities to increase their capacity to accommodate residents, and areas with higher building heights can provide a greater number of dwellings to meet housing demand [57]. When the rate of land expansion does not match the rate of population increase, urban land expansion requires a shift from low density to compactness [58], one manifestation of which is an increase in building heights [59]. The size of the urban population should be the underlying cause of urban expansion [60]. However, we find that in each period, the total population of medium-sized and small cities increases slowly or remains unchanged or even shows a decreasing posture, while the built-up area is mostly in a state of continuous increase, indicating that the rate of land expansion in small and medium-sized cities has not been slower than the rate of population increases. At the same time, in large cities, there may be a marginal effect: when the population density exceeds a certain threshold, its effect on building heights is attenuated [57]. On the other hand, based on the results of the concentric ring analysis shown in Figure 4 and previous studies, at the beginning of the study, the study area has already experienced the emergence of many new cities and new districts in the form of enclave expansion [58]. The expansion rate of the main urban areas of large cities is significantly faster than that of other cities [61], and the earlier start of urbanization, the bigger the original building land area [62]. During the period under study, large cities expanded at a slower rate, mainly in suburban areas, while small and medium-sized cities showed medium- to high-speed expansion, with expansion areas close to urban centers [62]. At this time, a large number of people from small and medium-sized cities in Hebei province flowed into Beijing and Tianjin, and there were signs of population flow from small cities to medium-sized cities in Hebei province [63]. At the same time, the inflow of people from large cities was mainly concentrated near urban centers [64]. While the population of large cities increases, the height of new buildings in the suburbs is not high; the population of small and medium-sized cities remains unchanged or decreases, but the height of new buildings in the city center is higher, which leads to a negative correlation between population and the height of new buildings.

Analyzed in terms of economy, industrial structure is significantly and positively correlated with the height of new buildings in the city. This is largely consistent with the findings of previous studies on the relationship between industrial structure and urban sprawl, both at the two- and three-dimensional levels [32,65,66,67]. In the study interval, the value added by the tertiary industry in most cities has increased faster than that of the secondary industry, which means that the BTH region has gradually become more industrialized and the tertiary industry has flourished in recent years. This continuous optimization and upgrading of the industrial structure intrinsically affect the change of building land [68], which is manifested in the shift in consumption from subsistence consumption to a stage of developmental and enjoyment-based consumption [69], which leads to more and more high-rise buildings such as offices, financial centers, and high-rise housing, and a higher mean value of the height of new buildings. Office buildings and financial centers, which primarily serve the tertiary sector, are typically located near a city’s CBD. As shown by the results of our concentric ring analysis (Figure 4), areas closer to the city center generally have higher average building heights. At the same time, the variation in building heights in city centers of small and medium-sized cities over the past decade has been significant. This may be because, compared to large cities, the tertiary sector in small and medium-sized cities had a lower baseline in 2010 but experienced rapid growth over the decade, resulting in a larger height disparity. This also confirms the positive role of industrial structure in the development of new building land. The actual per capita GDP, on the other hand, is weak and insignificant in this study, which is inconsistent with most of the studies [55,70]. Conventionally, regional gross domestic product (GDP) is treated as a core proxy for regional economic development. Economically advanced regions tend to invest in high-rise construction, implying an expected positive correlation between per capita GDP and newly built building height. Nevertheless, Gao et al. [22] documented that the linkage between GDP and construction land scale remains relatively weak even if statistically significant, demonstrating that such correlation is inherently susceptible to research scope and time frame as an objective feature of the variables themselves. On this substantive ground, one plausible reason for the insignificant coefficient in our linear regression is that the real nexus between GDP and new building height may follow a nonlinear pattern across the BTH region; the adoption of a linear empirical framework therefore fails to capture such nonlinear association and yields an insignificant coefficient. Future studies could adopt nonlinear modeling to further examine the GDP-building height nexus.

As a factor that is the focus of this study, the development of the land market exhibits a strong positive association with the three-dimensional vertical form of the city. This is generally consistent with the results of previous studies [32,71]. Land marketization has driven the functional differentiation of urban land use, and the business and service functions of urban centers have been intensified, resulting in the formation of central business districts (CBDs) lined with employment and commercial centers [72], which has led to a rise in the height of new buildings. This result, however, is not quite the same as most previous studies describing the drivers of urban sprawl from a two-dimensional perspective [7,19,22,23,24]. These studies generally agree that the development of land marketization will promote the intensive use of urban land and inhibit urban sprawl, but this is a conclusion based on the horizontal sprawl of the city, and as mentioned earlier, the horizontal and vertical sprawl of the city are related, but it does not mean that they can be discussed in the same way. As the level of land marketization increases, land users will consciously improve the intensive use of land, and land grantors will optimize the efficiency of land resource allocation under the influence of competition and price mechanisms [22], which will make land resources limited and challenging. To overcome this challenge, cities have shifted their goals and directions from rapid development to efficient development [73], and this has led to strategies and paths that have transferred from incremental expansion to stock optimization, resulting in a development pattern that makes full use of the vertical space of the city [74]. This is manifested in the fact that the development of land marketization has had an inhibiting effect on the horizontal expansion of cities, while it is positively correlated with vertical expansion.

It is worth noting that the average building height of newly developed land varied across the five periods, with the height of newly developed buildings reaching a significant peak in most city categories between 2008 and 2011 (Figure 6). This may be attributed to the fact that in 2008, in response to the financial crisis, China implemented a four-trillion- yuan infrastructure stimulus plan. The same year, the State Council’s executive meeting issued policies such as the “Several Opinions on Promoting the Healthy Development of the Real Estate Market,” which accelerated the construction of affordable housing and encouraged reasonable housing consumption. This series of measures drove up housing prices, stimulated large-scale urban construction projects, and led to the simultaneous launch of high-rise residential development projects nationwide, resulting in a significant increase in the number of high-rise buildings. In the land market, policies governing land auctions, tenders, and public listings were fully implemented in November 2007. The primary land market expanded rapidly, and fierce competition among real estate developers drove up land costs, prompting developers to favor the construction of high-rise buildings to reduce the cost per unit of land. This trend also led to an increase in the floor-to-ceiling heights of new buildings. Consequently, this resulted in a sharp rise in the height of newly developed buildings during the third phase in most cities.

4.2. Influential Factors of Three-Dimensional Morphology of BTH Cities

Most of the cities in the BTH region have shown a tendency to expand outwards with the city center as a circle during the decade. This is because cities need to continuously absorb the positive externalities of spatial agglomeration for stable development and long-term stability [75,76], and they need to carry out orderly lateral expansion in a compact form to efficiently promote the development of urbanization [77], whereas a circle is the most compact form with relatively minimal internal spatial barriers and a more concentrated and regular distribution [78]. A few cities including Zhangjiakou and Chengde are exceptions. Their urban cores lie on either side of river valleys, leading to strip-shaped expansion along river banks. The average building height across the BTH region increased, which is consistent with findings from previous studies [32], and it is worth mentioning that the proportion of low-rise buildings decreases, and the proportion of medium- and high-rise buildings increases. This is because with the acceleration of economic development and urbanization, the urban population increases, and the construction of medium- and high-rise buildings becomes a necessary path for urban development [59].

At the level of horizontal expansion, large cities with large original built-up areas have slow rates of horizontal expansion, which is consistent with previous studies [62]. This is due to their large land size, a stable urban spatial structure, and a spatial expansion pattern dominated by infill [58]. The increase in such a spatial expansion pattern is mainly in low and medium-rise buildings, resulting in a lower proportion of high-rise buildings and little change in average height. However, also as large cities, Tianjin and Beijing have different development trends. As the political, economic, and cultural center of China, Beijing has the unique advantage of having established a relatively up-to-date industrial structure and a high level of economic development, while Tianjin has a large gap in urban development compared to it [79]. Therefore, the urban form of Tianjin at the beginning of the study is similar to that of a medium-sized city. In the past decade, thanks to the development and opening up strategy of Binhai New Area, Tianjin has entered an unprecedented period of urban expansion [80]. The intensity and speed of urban expansion have gone through a process of ‘medium-high speed growth–high-speed growth–low-speed growth’ [81], and the pattern of expansion is mainly fringe-type and infill-type, which has resulted in a much greater increase in building height than that of Beijing. It is worth noting that, according to the concentric ring analysis (Figure 4), building heights across Tangshan increased significantly between 2010 and 2020. This may be attributed to a combination of factors, including urban planning and the city’s historical built environment. Tangshan faces unique development conditions: the city relies on low-rise, single-story housing built after the earthquake, and due to constraints on the scale of developable land, it is difficult to expand outward in a disorderly manner. As a result, development has shifted toward vertical, intensive development. Coupled with the implementation of new urban planning, which lacks the height restrictions typical of historic old cities, this has led to the emergence of large-scale high-rise residential and commercial buildings. Consequently, the city’s average building height growth rate is significantly higher than that of most cities in the BTH region.

Medium-sized cities primarily adopt edge expansion combined with infill expansion. Their urban cores are beginning to take shape, presenting the characteristics of a strong core and a weak periphery [58]. Edge expansion is dominated by low-rise and mid-rise buildings, whereas infill development within urban cores mainly consists of mid-rise and high-rise buildings. Accordingly, the overall proportion of high-rise buildings remains nearly stable, while that of low-rise and mid-rise buildings increases. In contrast, small cities expand mainly in the form of enclave and edge growth. Their urban cores are not fully developed and exert limited attraction to surrounding areas, so new construction is largely concentrated within core zones [58]. For this reason, small cities witness a substantial growth in mid-rise and high-rise buildings, and register the most remarkable rise in average building height over the decade.

The spatial evolution of cities can be described as a two-stage process of diffusion and agglomeration, with the evolution of cities beginning with the expansion of the urban seed or core [82], which has a certain regularity, although the manifestations show slight differences due to the different contexts of individuals [83]. Large cities always change from small cities. From this, we can speculate that in the next cycle, today’s small cities may begin to expand along their edges and accompanying infill of their urban cores; while the relative suburbs of today’s medium-sized cities have developed almost to the point of spill-over effects, with the main mode of expansion changing to infill. For large cities, on the other hand, there may be a trend towards decentralization and re-centralization, as in other global cities [84], or they may undergo a process of polycentricity [85].

4.3. Innovations

Compared to existing studies on three-dimensional urban expansion, the core advancement of this work lies in the development of a dedicated extraction framework for newly developed land, supplemented by improved two-temporal datasets and expanded analytical perspectives to support this central objective. Previous investigations into three-dimensional urban expansion have largely relied on single-point-in-time building height data, which were directly fed into regression analyses without the development of targeted tools to distinguish incremental built-up areas from the overall urban built environment. This study first combined the publicly available 2010 AW3D building height product with our group’s proprietary 2020 GBH dataset to construct a 10-year, two-phase dataset, thereby establishing a reliable data foundation for identifying long-term dynamics in new building land use. Unlike traditional studies that rely solely on building height data for simple quantification calculations, the proposed framework fully leverages spatiotemporal datasets to precisely identify new construction sites, thereby establishing a mature and practical technical approach for the long-term quantitative monitoring of three-dimensional urban expansion. Building on the incremental land datasets generated by the aforementioned extraction methods, this study further incorporates land market variables into the assessment of driving factors and conducts a cross-city comparative analysis of the BTH region using panel regression. This provides a valuable supplement to the existing research on three-dimensional urban expansion mechanisms in terms of both the factor dimension and regional empirical evidence.

4.4. Shortcomings and Prospects

This study has certain limitations and future research directions. One data limitation of this study stems from inconsistent spatial statistical coverage across variables. The dependent variable of building height is extracted via spatial masking based on built-up areas and physical urban construction boundaries. Total population, real per capita GDP and industrial structure adopt the statistical scope of urban districts; this design is intended to eliminate disturbances from suburban peripheries and better align their spatial coverage with that of the built-up-area-based building height dataset. However, constrained by official data availability, three land market indicators—total area of primary land market transactions, land marketization level and average land transaction price—can only be collected for the full administrative territory of each prefecture-level city. This mismatch in spatial scope for land-related variables may lead to potential measurement bias and impair the robustness of regression results. In addition, missing data on secondary land market transactions prevents a comprehensive depiction of local land market development. As for the threshold selection method, our study adopts fixed empirical thresholds for new building extraction, and the two core thresholds are determined based on pixel physical characteristics and gradient tests in the experimental area. In this paper, only simple verification is carried out for the intervals adjacent to the optimal thresholds, while systematic sensitivity analysis of full-dimensional and multi-combination thresholds is not conducted. Although these thresholds perform well within the study area, their generalizability may still be limited by the local building morphology and building density. Future research will complete the sensitivity analysis of global threshold combinations and attempt to construct an adaptive threshold method to improve the stability and universality of the model in different scenarios. This study uses building heights from 2010 and 2020 to characterize the heights of buildings constructed between 2002 and 2017. This approach may be subject to the effects of urban renewal, although our tests indicate that the demolition rate for newly constructed buildings within a short period is only 4.06%, if such buildings are demolished shortly after construction, the original number of new buildings will be significantly overestimated. Furthermore, the rule for pixel allocation based on the maximum proportion rule makes it possible to miss smaller-scale expansions, which affects our subsequent analysis. Future research should attempt to dynamically and comprehensively identify urban renewal projects to reduce errors caused by these factors.

5. Conclusions

Using two phased building height datasets (AW3D for 2010 and our group’s self-developed and publicly available GBH dataset for 2020) and long-term urban panel data from 2002 to 2017, this study’s core contribution lies in developing a targeted optimized extraction method for newly increased urban building sites oriented toward long-term three-dimensional urban expansion monitoring. Relying on the incremental construction information derived from this newly proposed extraction approach, we further adopt concentric circle analysis and panel multiple regression modeling to explore dynamic evolution and driving factors of three-dimensional urban morphology across the BTH region. Our customized workflow can effectively pinpoint newly added urban built-up space. Quantitative validation shows that the proposed extraction method achieves an accuracy rate of 87.08%, which strongly confirms the reliability and technical advantages of this core innovative framework.

The results show that in the BTH region, cities of the same type have similar patterns of three-dimensional morphological changes, while they are unique in some details; the three-dimensional morphological changes in the cities are not the same in different periods. From a statistical perspective, demographic indicators are weakly negatively associated with newly built building height, whereas industrial structure optimization and land market development show significant positive statistical correlations with the vertical height of newly constructed buildings. On the basis of these observed statistical correlations, this study puts forward tentative targeted reference suggestions. Relevant departments may take the quantified correlation features as auxiliary reference when drafting differentiated urban three-dimensional development management policies for the BTH region, so as to facilitate sustainable urban construction and mitigate potential drawbacks stemming from irrational vertical expansion. For cities facing mismatches between human and land resources, population outflow, and rapid three-dimensional urban expansion, it may be advisable to formulate talent attraction and incentive policies to retain the resident population, mitigate the risks of underutilized urban space and underutilization of construction land, thereby optimizing the spatial alignment of human and land resources and supporting steady urban development.