Spatiotemporal Evolution of County-Level Land Use Structure in the Context of Urban Shrinkage: Evidence from Northeast China

Abstract

Shrinking cities are a class of cities that show different trajectories in the urbanization process. Although many studies have examined shrinking cities from multi-dimensional perspectives, the spatiotemporal evolution of land use structure is still poorly understood. This study constructed an analysis framework for spatiotemporal evolution characteristics of land use structure over two 10-year periods based on 334 county-level administrative units in the context of northeast China, a region undergoing substantial population loss and urban shrinkage. This study analyses quantitatively measured population loss and land use conversion and investigated the rationality of the expansion mode of newly added construction land in research units. The results demonstrated that, first, the total population in northeast China continued to decline, but the total construction area continued to grow, and the various types of construction land in most shrinking units did not decline with the loss of population. Second, 67.09% of new construction land came from cropland, and compared with 2000–2010, the growth of new construction land in 2010–2020 slowed down. Third, during the study period, more than half of the expansion area for newly added construction land came from sprawling expansion, and areas in a state of shrinking were no exception.

1. Introduction

Land resources are the spatial carrier of urban social and economic development, and urbanization is one of the main methods of land use [1]. From 1978 to 2020, with the rapid advancement of industrialization and urbanization, China experienced a process of urban expansion [2], with its urban population increasing from 172 million to 902 million. In 2020, China’s resident urban population accounted for 63.88% of the total population. During this process, the land use structure underwent tremendous changes. The area of urban construction land in China surged from 6720 km² in 1981 to 58,307.7 km² in 2019, showing significant exponential growth and a rapid expansion trend. As a result, a series of problems have arisen, such as spatial sprawl, inefficient land use, increasingly tense relations between people and land, and escalating environmental pollution [3]. These changes have directly affected sustainable development and the quality of the living environment in the city. Under the circumstance of limited urban land stock, optimizing the allocation of urban land resources and improving utilization efficiency are important ways to promote the rational use of urban land and achieve sustainable urban development [4]. In the current process of urbanization and urban expansion, the structure of land use in urban areas is a growing concern for urban planners [5]. Based on different theoretical perspectives such as economics, sociology, and geography, many scholars have conducted studies on the concept [6], spatiotemporal evolution [7,8,9], driving factors [10,11,12,13,14], projection [15,16,17,18], and optimization schemes [19,20,21,22,23]. In sum, existing research has focused more on the expansion of urban land and the imbalance of urban land structure in the process of urbanization, while few studies have focused on the issue of land use structure under the shrinking scenario.

As a logical consequence, the expansion of rapid urbanization has been seen as the only succession path for China’s urban development (the traditional urban planning paradigm is also based on the “growth scenario simulation” to determine urban development planning) [24]. However, according to research by Long et al. [25,26], Wu et al. [27,28], Sun et al. [24], and Li et al. [29,30], there are various degrees of urban shrinkage, both in regions with rapid economic development such as the Pearl River Delta, Yangtze River Delta, and Beijing–Tianjin–Hebei and in relatively underdeveloped regions such as central, western, southwestern, and northeastern China. The concept of urban shrinkage was formally proposed in 1988 by German scholars, Huermann et al. [31], in an empirical study on population loss in the Drucker region of Germany. Since then, this concept has been widely accepted and used. At present, urban shrinkage is a hot issue in international research. Scholars have carried out extensive multi-dimensional and multi-scale research on aspects of urban shrinkage such as its conceptualization and methods for identification [32,33]; spatial changes and model refinement [34,35,36]; and causes, development effects, and planning responses [33,34,35,37,38,39]. In China, most empirical studies on shrinking cities take the municipality as the research unit [24,26]. However, analysis based on the population census has shown that variation at the county level is more pronounced, and using large administrative units would obscure the internal heterogeneity inside the area. More precisely, the population loss of counties under the jurisdiction of the city is relatively severe, while the population of districts, especially the central urban areas of sub-provincial capital cities, generally shows a state of continuous growth. However, existing research has ignored the distinction between the two types of administrative areas. Therefore, grounded in the context of China’s population shrinkage, there is a research gap in the exploration of shrinkage at the county level, including both districts and counties.

To address these research gaps, this study constructs an analysis framework for the evolution of the spatiotemporal patterns of land use structure under the shrinking scenario. Taking northeast China as an example, the county-level units are first divided into two categories: districts and counties; then the two types of analytical units are assessed in terms of population change to determine areas undergoing shrinkage. Spatial analysis and land use transfer matrices were subsequently used to measure the evolution characteristics of these spatiotemporal patterns of land use structure at the county and district levels. Then, based on the common edge measure method, the rationality of the expansion mode of newly added construction land under the shrinking scenario was investigated. This research contributes to the literature on urban shrinkage and related changes in land use structure; it also provides a basis for rational allocation of land resources under the shrinking scenario at the county level and provides a reference for the revitalization of northeast China.

2. Data and Methods

2.1. Study Area

Northeast China is one of China’s old industrial bases, and it is also a relatively independent and complete economic zone that was first developed after the founding of the People’s Republic of China (PRC). Most cities in the region are resource-based, which provided strong support for construction and development in the early days of the PRC. However, in recent years, due to the lack of locational advantages and over-exploitation of resources, urban shrinkage has become prominent in the region, including an economic downturn and low growth rate, continuous population loss, disorderly expansion of construction land, and inefficient use of land in general. This study, therefore, takes northeast China as the research area, as it has representativeness and practical significance. Figure 1 shows the study location and land use changes between 2000 and 2020.

The sub-provincial cities considered in this study indicate a level in the administrative hierarchy of cities in China. There are four sub-provincial cities in the study area: Harbin, Changchun, Shenyang, and Dalian. The county-level units considered in this study are the county-level administrative areas in China—that is, the administrative regions with the same administrative status as a county, including municipal districts, county-level cities, counties, autonomous counties, banners, and autonomous banners, all directly under the jurisdiction of the city. Counties are areas with a long history of development and a large proportion of the agricultural and rural population. County-level cities are areas with rapid development of secondary and tertiary industries, but they are geographically far away from the center of the prefecture-level city. Autonomous counties are places of ethnic autonomy, with a large number of ethnic minority populations. Banners and autonomous banners are applicable to Inner Mongolia and are equivalent to counties. Compared with other county-level administrative areas, municipal districts are characterized by a high degree of urbanization, developed secondary and tertiary industries, and a geographical location within the city center of a prefecture-level city. Based on the characteristics of all kinds of county-level administrative districts, this study divides them into districts and counties. Districts include only municipal districts, and counties refer to all county-level administrative regions except municipal districts.

This study selected 334 county-level units under the jurisdiction of northeast China as the research area, including 123 county-level units under the jurisdiction of Heilongjiang Province (56 districts and 67 counties); 60 county-level units in Jilin Province (21 districts and 39 counties); 100 county-level units in Liaoning Province (56 districts and 44 counties); and 51 county-level units in the “Three Cities and Two Leagues” of Inner Mongolia Autonomous Region (5 districts and 46 counties).

2.2. Data

The data of this study were collected from two sources: First, land use data of the study area in 2000, 2010, and 2020 with a resolution of 30 m × 30 m. The data were developed by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) (http://www.resdc.cn (13 October 2021)) based on Landsat–TM/ETM (2000) and Landsat 8 (2010, 2020) remote sensing image data. The classification system of land use data has three levels. Eight categories of data were used in this paper: cropland, forest, grassland, water, and unused land in the primary classification; and urban construction land, rural construction land, and other construction land in the secondary classification. The human–computer interactive interpretation method was used in extracting land use information. The field survey materials and field records were randomly chosen at a 10% ratio to the number of counties to assess the accuracy of the database [40]. The accuracy of the six classes of land use was above 94.3%, and the overall accuracy of the 25 subclasses was above 91.2%, which meets the requirement of user mapping accuracy [40,41,42]. Therefore, many studies used these data with different resolutions when analyzing land use at the regional- [43,44], city- [45], and county [46,47]-level scales.

Second, population data came from the “Population Censuses of China” (years 2000, 2010, and 2020). In the study period covered in this paper, the Chinese government fine-tuned the administrative divisions several times. Therefore in order to ensure the consistency and comparability of the data, this study used the county-level administrative divisions representing the latest county-level administrative divisions promulgated by the Chinese government in 2020, and 334 research units were finally obtained. Overall, 2.09% of county-level administrative divisions in northeast China have been fine-tuned, such as adding or deleting sub-districts. In order to ensure the comparability of data, this paper estimated populations for areas that underwent boundary changes using the interpolation method [48]. Since there were fewer administrative districts for fine-tuning, it had little effect on the results of this study.

2.3. Methods

2.3.1. Resident Population Change Rate

At present, academic circles generally agree that cities with continuous population loss are shrinking cities; however, the time and threshold for population loss are not uniform [34,49]. For example, the Shrinking Cities International Research Network has defined a shrinking city as a city with a population of more than 10,000 and a population loss for two consecutive years. This study determined the type of each research unit by calculating the population change of each research unit in two time periods [36]. The rate of change for the resident population is defined as follows:

$$R_1=\frac{P_{2010}-P_{2000}}{P_{2000}}\times 100\%$$

(1)

$$R_2=\frac{P_{2020}-P_{2010}}{P_{2010}}\times 100\%$$

(2)

where $P_{2000}$ is the resident population of a research unit in 2000; $P_{2010}$ is the resident population of a research unit in 2010; $P_{2020}$ is the resident population of a research unit in 2020; $R_1$ is the rate of change of the resident population of a research unit from 2000 to 2010; and $R_2$ is the change rate of the resident population of a research unit from 2010 to 2020. When $R_1\ge 0 \mathrm{and} R_2\ge 0$, the resident population has continued to increase, leading to the continuous growth type (G–G type). When $R_1<0 \mathrm{and} R_2\ge 0$, the resident population first decreased and then increased, thus belonging to the fluctuating growth type (R–G type). When $R_1\ge 0 \mathrm{and} R_2<0$, the resident population first increased and then decreased, thus belonging to the fluctuating shrinkage type (G–R type). When $R_1<0 \mathrm{and} R_2<0$, the resident population has continued to decrease, leading to the continuous shrinkage type (R–R type).

2.3.2. Land Use Transfer Matrix

To show the specific direction of land use dynamic changes, it is necessary to quantitatively study the number and ratios of the transformations of various land use types. The transfer matrix presents the transfer area based on the land use change type. The structure of land use types at the beginning and end of the research period can thus be examined. It can also show the specific transfer changes of different land types during the study period, thus clarifying the transfer direction and source composition of different land use types during the period. A two-dimensional transfer matrix was therefore used, and the mathematical expression is:

$$S_{ij}=\left[\begin{array}{cc}\begin{array}{cc}{S}_{11}& S_{12}\\ S_{21}& S_{22}\end{array}& \begin{array}{cc}\cdots & S_{1n}\\ \cdots & S_{2n}\end{array}\\ \begin{array}{cc}\cdots & \cdots \\ S_{n1}& S_{n2}\end{array}& \begin{array}{cc}\cdots & \cdots \\ \cdots & S_{nn}\end{array}\end{array}\right]$$

(3)

where S is the area; n is the number of land use types; and i and j refer to the land use types at the beginning and end of the research period, respectively. Each column element in the matrix represents the information about the area prior to transfer for land type j at the end of the research period, and each row element represents the information post transfer for land type i at the beginning of the research period.

2.3.3. Common Edge Measure Method

The growth of new construction land is to meet the needs of urban development, and different expansion modes of new construction land may be closely related to the development modes of different types of cities [50]. For example, for shrinking cities transferring vacant land and optimizing the land use structure is more important. In contrast, for expanding cities, the sprawling expansion model may be dominant and other types of land along the urban fringe may be converted into construction land to meet the demand of the population growth [51]. Therefore in order to explore the rationality of the expansion mode for newly added construction land in each research unit, this study used the common edge measurement method. The expansion mode was determined according to the ratio of the length of the common side of the newly added construction land plot and the original construction land plot to the total side length of the newly added construction land plot (Figure 2a). The specific formula is as follows [52]:

$$L=\frac{L_c}{L_a}$$

(4)

where $L_c$ is the common side length of the new construction land plot and the existing construction land plot; $L_a$ is the total side length of the new construction land plot; and L is the ratio of the two. When L ≥ 0.5, this indicates type L₁, or filled expansion; when 0 < L < 0.5, this indicates L₂, or sprawling expansion; and when L = 0, this indicates L₃, or enclave expansion [51] (Figure 2b).

3. Results

3.1. The Changing Characteristics of Resident Population in Northeast China

The research units were divided into two categories according to their attributes: districts and counties (Figure 3). From 2000 to 2010, the problem of population loss in northeast China was not apparent, and the overall resident population was still growing, with an increase of 4.59 million people or 3.77% of the total resident population in the region in 2010. However, the resident population of most research units was shrinking, accounting for 50.60% of the total research units, with a total decrease of 5.09 million people. County research units accounted for 71.60% of the research units whose resident population had decreased, and the resident population that decreased accounted for 64.17% of the total population decline. Just under half (49.40%) of the research units saw the resident population grow, an increase of 9.68 million people, of which 82.90% were district research units and 41.14% were district research units in sub-provincial cities. From 2000 to 2010, the total resident population in northeast China still increased. However owing to the level of economic development, the resident population in northeast China gradually gathered in district areas, especially those of the sub-provincial city.

The population change in the region in the second period was in sharp contrast to that of the first period. There was a total loss of 11.95 million resident population in 10 years, accounting for 10.86% of the total resident population in the region in 2020. The resident population of most research units fell, accounting for 78.44% of the total research units and a total decrease of 18.85 million people. Again, county research units were more likely to see declines, accounting for 70.23% of the research units with falling resident populations, and the reduced resident population accounted for 77.90% of the total population shrinkage. Furthermore, 21.56% of the research units saw resident population growth, an increase of 6.90 million people, of which 93.60% were in district research units and 54.47% were district research units in sub-provincial cities. From 2010 to 2020, population outflows from the region became prominent. At the same time the gathering of residents in district areas, especially in district areas of sub-provincial cities, also became increasingly obvious.

To characterize the population change patterns all research units were divided into four categories based on the population increase/decrease in the two periods (G–G, R–G, G–R, and R–R; Figure 4). Among the 138 district research units, 34.78% were G–G, of which 8.33% were located in the Inner Mongolia Autonomous Region, 39.58% in Liaoning Province, 20.83 in Jilin Province, and 31.25% in Heilongjiang Province. R–G research units accounted for 8.70% of all units, of which 58.33% were in Liaoning Province, 16.67% in Jilin Province, and 25.00% in Heilongjiang Province. Nearly a third (30.43%) were G–R research units, of which 2.38% were located in the Inner Mongolia Autonomous Region, 47.62% in Liaoning Province, 11.90% in Jilin Province, and 38.10% in Heilongjiang Province. R–R research units accounted for 26.09% of all units, of which 27.78% were located in Liaoning Province, 11.11% in Jilin Province, and 61.11% in Heilongjiang Province. The district study units were thus dominated by the G–G and G–R units, with resident population growth dominant from 2000–2010 and a shift to resident population decline from 2010–2020.

Among the 196 county research units, only a small proportion (3.57%) were G–G research units, of which 85.71% were located in the Inner Mongolia Autonomous Region and 14.29% in Jilin Province. A small fraction (2.55%) was R–G research units, of which 60.00% were located in the Inner Mongolia Autonomous Region, 20.00% in Liaoning Province, and in Heilongjiang Province separately. G–R research units accounted for 34.69% of all units, of which 14.71% were located in the Inner Mongolia Autonomous Region, 16.18% in Liaoning Province, 19.12% in Jilin Province, and 50.00% in Heilongjiang Province. A majority (59.18%) were R-R research units, of which 23.28% were in Inner Mongolia Autonomous Region, 27.56% in Liaoning Province, 21.55% in Jilin Province, and 27.59% in Heilongjiang Province. In contrast to the district-level units, the county study units were dominated by the R–R and G–R types, with the population loss becoming more severe in 2010–2020 compared to 2000–2010.

3.2. Changes in Land Use Structure in Shrinking Units

For the convenience of description, the following abbreviations are used to describe land use percentages: IL is the percentage of research units in an increasing state of construction land; UL is the percentage of units with a constant state of construction land; DL is the percentage of units in a decreasing state of construction land. The following abbreviations are used to describe land use structures: UCL is urban construction land, RCL is rural construction land, OCL is other construction land, TCL is total construction land, and CL is construction land.

CL showed persistent growth in most counties over the 20 years studied. The total amount of CL for about 90% of research units was in 2020 than in 2000, and more than 50% of the total CL of research units in the two research periods were in a state of continuous growth. Among the research units in a state of continuous growth, only about 20% were G–G type, while more than 40% were R–R type. Among the R–R type research units whose CL continued to grow, about 25% of the newly added CL was primarily UCL and about 48% was RCL. Most of the research units experiencing population shrinkage have not seen a decline in CL with the loss of population.

To study the difference in the proportion of CL in different types of research units, we calculated the change in the four types of CL in each research unit during the study period, and the statistics for the three values of IL, UL, and DL in the same type of research unit (Figure 5). The IL of research units under population shrinkage did not decrease with the continuous loss of resident population, and some values were even higher than in units that were in a state of population growth during the same period. For example, from 2010 to 2020, the IL values for UCL in G-R county and R-R county units in a state of population shrinkage were higher than that for a G–G county. Most of the research units in the shrinking state thus experienced a continuous expansion in CL, resulting in inefficient use of land and a serious waste of land resources.

It can be seen from

Table 1. Growth rate of new construction land in various research units, 2000–2010 and 2010–2020.

Type	District						County
Period	2000–2010			2010–2020			2000–2010			2010–2020
	OCL	NCL	G	OCL	NCL	G	OCL	NCL	G	OCL	NCL	G
G–G	2709.73	1399.41	51.64%	3501.93	1285.25	36.70%	286.03	281.46	98.40%	495.66	444.17	89.61%
R–G	693.76	320.57	46.21%	878.16	230.43	26.24%	304.24	225.56	74.14%	520	141.78	27.27%
G–R	1943.4	709.71	36.52%	2129.78	507.25	23.82%	8306.69	3302.52	39.76%	8620.75	2656.77	30.82%
R–R	1101.36	543.3	49.33%	1335.38	331.97	24.86%	13,064.65	6999.31	53.57%	14,996.48	4229.55	28.20%
Total	6448.26	2972.99	46.11%	7845.25	2354.9	30.02%	21,961.61	10,808.84	49.22%	24,632.89	7472.26	30.33%

OCL is original construction land (km²); NCL is new construction land (km²); G is growth rate, G = NCL/OCL (%).

that the growth rate of new construction land in 2010–2020 was slower than that in 2000–2010 for both districts and counties, which is in line with the continuous regulation of the disorderly spread of urban construction land, as well as with the optimization and adjustment of urban internal land use structure driven by national macro policies in recent years. In addition, compared with county research units, in the same period, the development of non-agricultural industry in the district research units did not depend on the new construction land, and owing to the impact of land scarcity, the growth rate of the new construction land in the county research units was higher than that in the district research units [53]. However, whether in district research units or county research units, the growth rate of newly added construction land was still higher in G–G research units during the research period. Although the growth rate of new construction land in the other three research units was slightly lower, it still reached about 30%.

The spatial distributions of CL change in northeast China were similar in the two study periods. The expansion of CL was concentrated in the southern part of the region with Liaoning Province at the core of a large aggregation distribution. However, in the central and northern regions, a small, decentralized pattern formed in some municipal districts of sub-provincial cities (Figure 6). From 2010 to 2020, the scope of CL expansion declined, showing a trend of concentration in cities with better economic development, and the gap between the growth rate of CL in small cities (especially peripheral) cities and in big cities widened further. The changing characteristics of CL in both the shrinking and growth research units were basically similar. The expansion of CL did not decrease with decreases in the permanent resident population, and the growth rate of CL in the shrinking research units was even slightly higher than in the growing units in some periods.

To explore the regularity of CL development and the spatial reconstruction process of land factors at different stages in northeast China, this study used a land use transfer matrix to analyze the conversion relationship between six land use types in shrinking and growing research units. As the total area of the different types of research units was not consistent, it is not significant to simply compare the differences in land use transfer areas in the different periods; rather, the ratio of land use transfer to the total area of the different types of research units in different periods is compared. To highlight the key points of analysis, we only extracted data related to CL when drawing the land use transfer matrix (Figure 7).

The main source of new construction land in the district research unit was the conversion of cropland to UCL and RCL; the transfer of construction land was mostly the conversion of RCL to cropland and UCL (Figure 7a). The proportion of G–G research unit cropland converted to UCL was higher, and the proportion of cropland converted to RCL in the other three research units was higher. The main source of newly added construction land in the county research unit was the conversion of cropland and grassland to RCL; the transfer of construction land is mostly the conversion of RCL to cropland and UCL (Figure 7b). Due to the G–G and R–G research units having a small amount of data, these two research units differed from other research units in the main source of newly added construction land, mainly from the conversion of grassland to OCL and RCL. However, the main source of new construction land for both district and county research units was cropland.

As shown in Figure 7, owing to the continuous development of cities, a large number of land resources in the suburbs were expropriated for both district and county research units. Many areas of cropland, especially those close to county-level administrative centers and main roads, were also within the scope of urban expansion, resulting in the reduction of a large number of cultivated land resources [54]. The expansion of rural CL is excessive, and non-agricultural construction such as housing and workshops, as well as infrastructure such as traffic roads, now occupy a large area formerly given over to agricultural use. From 2000 to 2020, 67.09% of the newly added CL in the study area came from cultivated land, occupying 9366.01 km² and accounting for 25.29% of total CL in northeast China in 2020. Once croplands have been converted to CL, the process of reversion is particularly difficult. As the largest commodity grain base in China, the grain production capacity of the northeast region is tied to national food security. Farmland should thus not be understood simply as one type of land use to be adapted for urban development—as current needs seem to dictate—but rather as a living part of the landscape that requires continued care and stewardship. Well-managed farmland in cities contributes to residents’ nourishment, as well as shaping the visual aesthetics of neighborhoods, providing recreational and educational opportunities for interaction with nature, and (as in the case of Kyoto) is vital in preserving culinary traditions and, thus, local cultural identity [55]. Therefore, northeast China should continuously increase the management and control of cropland resources by clarifying the priority for cropland use and focusing on combating the abuse and destruction of cropland so as to strictly adhere to the red line of cropland.

3.3. Expansion of Construction Land in Shrinking Units

Shrinking cities do not mean giving up new construction land, but this expansion should focus on the filled expansion mode. The shrinking city should rely on its idle land, adjust the structure and spatial layout of the urban land, improve the construction of public infrastructure within its borders, improve its livable level, and realize benign development by employing the concept of “smart shrinkage” (from “Key Tasks for New Urbanization Construction and Urban-Rural Integration Development in 2020 in China”). Therefore, it is particularly important to sort out the expansion mode of newly added construction land in each research unit under the shrinking scenario to discuss whether the expansion pattern of new construction land in shrinking cities tends to be rational.

On the whole, the expansion of urban, rural, and other CL in northeast China showed a spatial disparity in the high in the south and low in the north (Figure 8), which corresponds to the development reality in which the economic center moved continuously southward [24]. The common edge measurement method was used to analyze the expansion patterns for new CL in different types of research units in the study area, and the expansion patterns were divided into three categories: filled expansion, which is concentrated in the original CL or extends outward in one direction outside the original CL in the study area; sprawling expansion, which mainly spreads to the periphery of the original CL; and enclave expansion, in which development areas typically appear far away from the original CL with a relatively scattered spatial distribution.

The proportion of the expansion area and the number of patches of new CL under different expansion modes in 2000–2010 and 2010–2020 are summarized in

Table 2. The proportion of expansion areas and quantity of patches of new construction land in different expansion modes, 2000–2010 and 2010–2020.

Research Period	Sum		Filled Expansion		Sprawling Expansion		Enclave Expansion
	Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches
2000–2010	14,969.79	420,340	5.91%	51.16%	55.20%	38.88%	38.89%	9.97%
2010–2020	10,902.62	778,488	13.04%	73.20%	58.82%	25.54%	28.14%	1.24%

. In the two study periods, filled expansion accounted for a large proportion of the number of new CL patches, but only accounted for a small proportion of the area. The number of sprawling expansion patches accounted for a small proportion, but a large proportion of the area. During the whole study period, more than half of the CL expansion area appeared as sprawling expansion, and more than half of the number of patches appeared as filled expansion. Compared to 2000–2010, the expansion area for new CL in 2010–2020 decreased, while the area of filled expansion increased. This may reflect the effects of the national macro policy of continuous control of the disorderly sprawl of UCL, as well as the advocacy of optimization and adjustment of the urban internal land structure in recent years [56].

As shown in

Table 3. The proportion of expansion areas and number of patches of new construction land in different expansion modes in each research unit, 2000–2020.

Research Unit Type		Sum		Filled Expansion		Sprawling Expansion		Enclave Expansion
		Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches	Expansion Area (km2)	Number of Patches
District	G–G	2240.04	35,239	5.59%	67.04%	66.34%	26.72%	28.07%	6.24%
	R–G	531.70	7571	5.97%	44.23%	53.66%	44.97%	40.37%	10.79%
	G–R	793.45	28,325	7.38%	73.57%	56.47%	21.44%	36.15%	4.99%
	R–R	561.87	14,809	5.39%	77.26%	59.46%	16.94%	35.15%	5.79%
County	G–G	569.21	6801	1.92%	76.28%	20.15%	18.14%	77.92%	5.57%
	R–G	212.25	6181	5.18%	61.43%	37.28%	30.27%	57.54%	7.30%
	G–R	2692.66	193,872	9.19%	76.85%	49.99%	19.27%	40.82%	3.88%
	R–R	6248.48	320,812	7.60%	67.36%	47.50%	26.19%	44.91%	6.46%

, in both the district- and county-level research units, the number of filled expansion patches of new CL is relatively high. The sprawling expansion of district units accounted for a high proportion of the expansion area of CL, but the sprawling and enclave expansion both accounted for a high proportion of the CL expansion area in county units. In general, the proportion of the expansion area of CL for filled expansion was relatively small for all study units. In addition, the proportion of the expansion area and the number of patches of new CL for all three expansion modes did not notably differ between units that were shrinking and those that were growing. There is thus a long way to go to achieve the “slim and strong” and healthy development of shrinking cities (i.e., “smart shrinking”).

4. Discussion

Urban shrinkage is fairly new in China and has only taken place in recent years. The country has thus had little experience in the governance and control of shrinking cities. Since the mid-to-late 20th century, cities in Europe, America, and Asia have experienced urban shrinkage. Although governments have implemented a series of urban revival measures, urban shrinkage has still led to the bankruptcy of urban property in individual cities, such as Detroit [57]. Some shrinking cities, such as Youngstown, have actively adopted urban planning strategies to develop compact and intensive land use as a “smart shrinkage” approach to build a healthier and sustainable urban environment, so as to make the city revitalized [34]. It is thus particularly important to sort out the spatiotemporal changes of urban CL under urban shrinkage and explore the expansion modes for newly added CL in shrinking cities. However, at present, few scholars in China have paid attention to the changes in urban land use structure under urban shrinkage. Based on population census and land use data, this study constructed an analytical framework for the evolution of spatiotemporal patterns of land use structures under urban shrinking scenarios. The evolution characteristics of the temporal and spatial patterns of land use structure in different types of research units are also discussed, and the expansion model of newly added CL under urban shrinkage was also examined. This enriches the quantitative research methods and empirical research results on urban shrinkage and related changes in land use structures.

4.1. Causes of the Increasingly Severe Resident Population Loss in Northeast China

Compared with 2000–2010, there are two main reasons why the loss of the resident population in northeast China has become more serious in 2010–2020. First, the birth rate in the three northeastern provinces continued to decrease, while the death rate simultaneously rebounded. As a consequence, the natural growth rate continued to decrease, resulting in negative growth from 2010 to 2020 (Figure 9) and an increasingly serious loss of the resident population in the region. Second, in recent years, the economic development of northeast China has continued to be sluggish, and the economic growth rate has also lagged behind the national average. Economic factors in general are the main driving force for population mobility, thus causing serious population outflow problems in northeast China [58]. Since 2010, the proportion of inter-provincial migration in the northeast region has declined, and the destination of the migrant population has shifted from neighboring areas to major domestic urban agglomerations [59]. Due to the sluggish economic development, the general population—especially the talents who have received higher education or mastered professional skills—may choose to migrate, and the loss of such talents would in turn negatively affect regional economic development, thus forming a vicious circle [60].

4.2. Paradox of Population Loss and Construction Land Expansion

Evidence from the shrinking cities in the United Kingdom, United States, and other countries suggests that, with the decline of the local population, the driving force for new CL is weakened, and problems such as abandoned urban land and idle infrastructure appear [33]. China’s shrinking cities show the paradox of coexistence of population loss and expansion of construction land [33]. Northeast China is no exception. Against the background of the massive loss of resident population, construction land in about 90% of the research units is still growing. The growth rates for construction land among some areas experience a shrinking population that is higher than for areas experiencing population growth during the same period. Moreover, according to the statistics of the “China Urban Construction Statistical Yearbook” (years 2000 and 2020), whether it is a city with a continuously growing or shrinking resident population, UCL increased. The newly added UCL is mostly residential or industrial land. For example, from 2000 to 2020, the resident population of Shenyang increased by 1.95 million, and the UCL increased by 355.28 km², of which 42.65% was newly increased residential land and 26.31% was newly increased industrial land. This translated into a nearly threefold increase in residential land per capita from 9.59 to 25.20 m²/person. From 2000 to 2020, the resident population of Siping decreased by 0.35 million, and the UCL increased by 26.11 km², of which 42.36% was residential land and 26.73% was industrial land. Residential land per capita thus more than doubled from 6.28 to 15.14 m²/person.

Although compared with growth cities, shrinking cities have less new construction land, regional development is strongly dependent on new construction land [53], and residential and industrial are still used as the main component of new construction land. If new construction land in shrinking cities is still dominated by residential land, an oversupply of housing in shrinking cities can occur, resulting in large fluctuations in housing prices and an increase in housing vacancy rates. The increase in the vacancy rate in urban neighborhoods could lead to a further decline in maintenance services in such areas and an increased crime rate in the surrounding area [61]. Large fluctuations in housing prices exacerbate conflicts between developers and residents and even cause developers with less capital to go bankrupt, resulting in a large number of unfinished buildings in both counties and cities [34]. These results are not only a problem for real estate itself, but also for the whole local government in the vicious circle of urban shrinkage.

Due to the difficulty of obtaining socioeconomic data in districts. This study can only analyze the correlation between the resident population, construction land area, and GDP changes in the county units from 2000 to 2020 based on the data of the “China County Statistical Yearbook” (years 2001, 2011 and 2020) (

Table 4. Correlation analysis between resident population, construction land area, and GDP changes from 2000 to 2020.

Index	County			G–G County			R–G County			G–R County			R–R County
	P	L	G	P	L	G	P	L	G	P	L	G	P	L	G
P	1			1			1			1			1
L	0.338 **	1		0.062	1		0.990 *	1		0.549 **	1		0.190 *	1
G	−0.087	0.68	1	0.965 **	0.209	1	−0.759	−0.697	1	−0.276 *	−0.170	1	−0.199 *	0.212 *	1

P is the change in resident population from 2000 to 2020; L is the change in construction69 land area from 2000 to 2020; G is the change in GDP from 2000 to 2020. Note: * and ** mean significant correlation at p < 0.05 and extremely significant correlation at p < 0.01, respectively.

). It can be seen that in spite of the population declining, economic development continued. However, the economic development may not be sustainable in the future due to projected further loss of population. Furthermore, the economic development could be a result of shrinking units of land sales and over-expansion of construction land. This may be due to urbanization and land development have been closely coupled in China’s socioeconomic transition in the post-reform era [62]. Urban growth has created a demand for new construction land for further urban sprawl. Driven by market forces, massive quantities of cropland have been transformed into new construction land [63]. In addition, the significant relationship between land use change and economic growth is rooted in China’s land ownership and land use rights systems. Owned by the government, land can be used as a powerful macro-economic intervention tool [64]. Local governments accumulate land leasing fees by leasing converted land to developers, grabbing a high volume of profit deriving from the vast gap between the compensation for land-loss peasants and land premiums paid by developers [62]. The strong economic incentive simultaneously spurs a sharp eagerness for land accumulation [65], in part resulting from a more substantial financial burden to local economies [63], especially after the fascial system rearrangement in 1994.

4.3. Adjusting the Expansion Model for New Construction Land in Shrinking Cities

This study found that the proportion of the area of expansion and the number of patches of new construction land in the three expansion modes of resident population shrinking units were no different from those in growing units. More than half of the newly added construction land in northeast China during the study period came from sprawling expansion, but this expansion mode is mainly based on the all-around expansion of the original construction land, which is a typical urban planning and development paradigm under the urban growth scenario. If the new construction land in shrinking cities is still dominated by sprawling expansion, large quantities of cropland around the city will be converted to UCL, and the speed of land urbanization will be much higher than that of the population urbanization. Over time, this may lead to abandoned UCL and idle infrastructures, such as have formed in the shrinking cities of Europe and the United States, resulting in the waste of precious land resources due to blind development and construction.

Urban shrinkage is not, however, the same as urban decline: the coin has two sides. While shrinking cities experience population loss, a large amount of potential flexible land is formed, which creates the possibility for the transformation of land use structure, the improvement of industrial land layout, and the construction of urban green open space [66]. Shrinking cities thus do not mean giving up new construction land, but this expansion should focus on the filled expansion mode. Policymakers and urban planners in shrinking cities should rationally think about the role of cities in growth areas, moving away from traditionally convergent growth models, and exploring new functions and spatial growth points according to local characteristics. Shrinking cities rely on improving urban quality to cultivate sustainable development capabilities and unique regional competitiveness, and turn the region’s state of “when one is rising, the other is falling” into “reciprocal symbiosis” [67].

5. Conclusions

Urban shrinkage is a development path that is spreading widely across the world, so it is particularly important to monitor the temporal and spatial changes of construction land under the shrinking scenario and examine the modes of expansion for new CL in shrinking cities. This study considered northeast China, one of the areas with the most severe urban shrinkage in China, as the research area. Based on the resident population and land use data for the past 20 years, the temporal and spatial evolution characteristics of the land use structure of county-level units under the shrinkage scenario were explored. This research contributes to the literature on urban shrinkage and related changes in land use structure; it also provides evidence for the rational allocation of land resources under the shrinking scenario at the county level and can be used as a reference for the revitalization of northeast China.

Four main conclusions can be drawn. First, the loss of resident population in the county units was significantly more serious than that in the district units. There was an increased concentration of the resident population in district areas, especially the district areas of sub-provincial cities between the two periods. During the study period, the district research units were mainly G–G and G–R units, while the county research units were mainly R–R and G–R units. Second, against the background of massive loss of the resident population, about 90% of the research units still exhibited growth in construction land: construction land changes are not in line with changes in population. The growth rate of construction land in some shrinking areas was in fact higher than that in expanding areas. Third, the newly added construction land occupies a large quantity of former cropland in the study area during this period: 67.09% of the newly added construction land came from cropland. Compared with 2000–2010, the rate at which new construction land increased from 2010 to 2020 slowed down, which may reflect the effects of the national macro policy to constantly regulate the disorderly spread of UCL while advocating the optimization and adjustment of urban internal land use structure in recent years. Fourth, during the study period, more than half of the expansion area of construction land in northeast China came from sprawling expansion, while more than half of the number of patches came from filled expansion. The proportion of the expansion area and the number of patches for new construction land in the three expansion modes in shrinking units did not differ from those in growing units; as mentioned above, there is thus a long way to go to achieve healthy development of shrinking cities (i.e., “smart shrinking”).

This study has the following limitations and future research areas. First, because the research area is quite large, the CL in the land use data used in this study can only be classified into two levels, and the analysis of the urban internal land structure could thus be improved. Future research could thus be carried out in the following area: the typical cities (typical of shrinking and growth cities) could be screened to further refine their construction land types, including residential land, public facility land, and industrial land. Second, city shrinkage can be divided into Perforated Shrinkage [68], Doughnut Shrinkage [69], Peripheral Shrinkage [36], etc. Different types of shrinkage cities may have different characteristics in terms of population and land use structure changes, but this is not considered here. Future research could thus be carried out in the following area: the types of shrinking cities could be identified based on their population characteristics and the associated differences in land use structures could be analyzed. Third, there are two most representative identification systems for shrinking cities currently are population change theory and multi-indicator change theory [24]. However, as far as the core issue is concerned, most scholars believe that population reduction is the best criterion for determining whether a city is shrinking [39]. Although this study uses population change theory to identify the research unit types, many scholars still emphasize the multi-indicator change identification method in many empirical studies [24]. Future research could thus be carried out in the following area: identify the research unit types based on multi-indicator change, and then discuss the temporal and spatial pattern evolution characteristics of land use structure under the classification results. Fourth, since the government fine-tuned the administrative divisions during the study period, the resident population data for some years in county-level administrative divisions are estimated by interpolation. However, the overall impact on the results is minor due to the small number of counties undergoing such adjustments.