Land Use Evolution and Multi-Scenario Simulation of Shrinking Border Counties Based on the PLUS Model: A Case Study of Changbai County

Abstract

The sharp decline in the population along the northeastern border poses a significant threat to the security of the region, the prosperity of border areas, and the stability of the social economy in our country. Effective management of human and land resources is crucial for the high-quality development of border areas. Taking Changbai County on the northeastern border as an example, based on multi-source data such as land use, the natural environment, climate conditions, transportation location, and social economy from 2000 to 2020, the land use transfer matrix, spatial kernel density, and PLUS model were used to analyze the spatio-temporal evolution characteristics of land use and explore simulation scenarios and optimization strategies under different planning concepts. This study reveals the following: (1) During the study period, the construction land continued to increase, but the growth rate slowed down, mainly transferred from cultivated land and forest land, and the spatial structure evolved from a single center to a double center, with the core always concentrated along the border. (2) The distance to the port (transportation location), night light (social economy), slope (natural environment), and average annual temperature (climate conditions) are the main driving factors for the change in construction land, and the PLUS model can effectively simulate the land use trend under population contraction. (3) In the reduction scenario, the construction land decreased by 1.67 km², the scale of Changbai Town slightly reduced, and the contraction around Malugou Town and Badagou Town was more significant. The study shows that the reduction scenario is more conducive to the population aggregation and industrial carrying capacity improvement of shrinking county towns, which is in line with the high-quality development needs of border areas in our country.

1. Introduction

Urban shrinkage is a prevalent phenomenon in the global urbanization process, characterized by population decline and accompanied by negative externalities such as inefficient land use, vacant housing, and economic downturns. This trend poses significant challenges to the high-quality development of cities in the 21st century [1]. Influenced by globalization, urbanization, and deindustrialization, urban shrinkage is not confined to specific countries or regions. Developed nations, including the United States, Germany, the United Kingdom, and France, as well as developing countries such as Brazil, Russia, India, and China, are all facing shrinkage pressure, and the number of global shrinking cities is increasing year by year [2,3,4]. As a rapidly urbanizing country, China has seen a rapid increase in the number of shrinking cities [5,6,7,8], and with the population center gradually concentrating in the eastern coastal areas [9], shrinking cities are mainly distributed in the Northeast (32.93%), East (28.05%) and West (21.95%) regions [10]. Among them, in the Northeast region, where the shrinkage phenomenon is widespread and particularly significant [11,12,13,14], due to the location disadvantage and continuous population outflow in the border areas, the shrinkage trend shows a continuous distribution feature [15,16,17,18]. In addition, due to policy development contradictions [19], the contradiction between population and land development in the border areas has become increasingly prominent. Against this background, Changbai County, as a typical shrinking border city, due to its unique population structure, shows a more significant double shrinkage feature of population and economy [20].

2. Literature Review

2.1. Measurement and Identification of Shrinking Cities

Research on shrinking cities begins with defining the concept and quantitatively identifying it. At the international level, since the term “shrinking cities” was introduced by German scholars in 1988 [21], early studies primarily focused on a single indicator of population loss [22]. Over time, the research expanded to encompass a multidimensional indicator system that includes population, economic, and social factors [23,24]. With advancements in spatial information technology, the focus has shifted to spatial dimensions such as built-up area, land use, and facility utilization [25,26]. Techniques such as geostatistics [27], satellite image analysis [28], and cellular automata simulation [29] have facilitated a transition from “result identification” to “spatial prediction” [30]. Although domestic research started later [31], it has been based on national conditions and characteristics. Early studies mainly focused on indicators such as permanent resident population and urban population [32,33] and later integrated economic and social indicators such as industry, employment, and finance [34]. In recent years, it has further introduced multi-source data such as vacant houses [35], controlled land use [36], built-up area, and nighttime vitality [26,37,38,39] and improved accuracy through Markov models [40] and spatial econometric models [41], forming a Chinese characteristic identification method and technical path for shrinking cities. The research direction has gradually shifted to the prediction of land and spatial resources under population shrinkage.

2.2. Causes and Mechanisms of Urban Shrinkage

From a global perspective, the driving mechanisms of urban shrinkage exhibit significant regional differences. In developed countries, such as in the Rust Belt in the United States and eastern Germany, the primary drivers are external factors, including globalization, aging populations, and deindustrialization [42,43,44]. In contrast, developing countries, such as the former industrial bases in China and northeastern Brazil, face more internal constraints, such as imbalanced urbanization, industrial transformation, and environmental challenges [45]. In terms of research on these mechanisms, scholars both domestically and internationally have focused on the non-linear characteristics of urban shrinkage. International studies have increasingly adopted a non-linear dynamic perspective [46,47], exploring the feedback mechanisms between the causes and effects of shrinkage through heuristic models [48]. Domestic research has further emphasized the notion of cities as complex adaptive systems, revealing that external environments and internal factors exacerbate shrinkage through non-linear cumulative effects [49,50,51]. Researchers have employed hierarchical linear models [11] and random forest models [52] to analyze the intricate correlations among multiple factors, thereby deepening the understanding of the causes of urban shrinkage.

2.3. Strategies and Practices for Urban Shrinkage

International practices have established two distinct models. Cities like Detroit, categorized as “resistance type, represent “expansion-oriented” planning driven by government and enterprise initiatives, aiming to counteract economic contraction through industrial transformation [53]. In contrast, cities like Youngstown, classified as “adaptation type, represent “adaptation-oriented” planning, utilizing green infrastructure, land banks, and other strategies to facilitate resource reallocation [54,55,56]. Domestic research aligned with the objective of high-quality development has proposed top–down governance strategies such as brownfield redevelopment [57] and adaptive housing reform [58]. Additionally, it has suggested bottom–up renewal models, including urban self-construction, urban agriculture, and cultural and creative research and development [59], emphasizing a developmental approach characterized by “reducing quantity while enhancing quality”.

2.4. Existing Limitations and Purposes

Although both domestic and international research has established an analytical framework for urban shrinkage, three significant deficiencies remain. First, in the realm of measurement and identification, the perspective has shifted from a singular focus on population to a multi-dimensional, comprehensive approach. While the predictive identification technology for land use and spatial resources has been initially explored, the precise simulation of the dynamic evolution of land and spatial resources requires further development. Second, regarding the mechanisms of causation, border counties represent a unique category of cities. In addition to the general internal factors [33], these areas are also influenced by external factors such as the effects of their border location, natural environmental constraints, and national security considerations. This has exacerbated the distinct characteristics of the nonlinear and cumulative mechanisms that contribute to the shrinkage of border counties [60,61,62]. Although existing models, such as cellular automata, hierarchical linear models, and random forest models, can address some nonlinear relationships, they still exhibit limitations in stochastic simulation and in capturing complex nonlinear interactions. Third, in the implementation of response strategies, while the concept of reduction planning has been introduced in China, current land management and planning norms predominantly follow an expansion-oriented approach and lack specific strategic designs tailored for shrinking scenarios. Furthermore, comparative studies examining land use outcomes under different planning concepts are insufficient.

In response to the aforementioned issues, this study focuses on the typical case of Changbai County, located in the northeastern border region. Utilizing land use, natural environment, climate, transportation infrastructure, and socio-economic data from 2000, 2010, and 2020, the PLUS model is employed to analyze the spatio-temporal evolution characteristics of land use. Additionally, a land use change simulation model tailored to the context of population decline is developed. Various land use scenarios guided by different planning concepts are presented to provide scientific support for the high-quality development of border areas in China.

Research gaps include (1) inadequate and imprecise simulation of the dynamic evolution of land and spatial resources, (2) the fact that the nonlinear and cyclical cumulative characteristics of the formation mechanisms of shrinking border counties have not been thoroughly elucidated, and (3) a lack of adaptability and comparative analysis regarding planning strategies for border cities in the context of contraction.

The following research questions are addressed in this article: (1) What spatial distribution characteristics does the expansion of construction land in Changbai County exhibit in the absence of cross-border interactions? (2) Can the PLUS model effectively capture the nonlinear impact mechanisms of multiple driving factors on shrinking border cities in the context of population decline, and what are the key factors that significantly influence changes in construction land? (3) In comparison to traditional incremental planning, does the decreasing development scenario enhance population concentration and the industrial carrying capacity of border counties?

3. Materials and Methods

3.1. Research Area

Changbai County is situated in the northeastern border region, south of the Democratic People’s Republic of Korea, with the Communist Party of Korea located across the river. The length of the border line measures 260.6 km, encompassing an area of 2505.96 square kilometers (Figure 1). The county is governed by seven towns and one township and features a state-class port (Changbai Port), which serves as a crucial gateway for economic cooperation in northeast Asia. It is a significant node in the One Belt, One Road, playing an essential role in the northern strategy, particularly in the development and opening of the Yalu River economic zone [63]. Changbai County is dedicated to transforming and developing its tertiary industry, leveraging the Border Economic Cooperation Zone to promote mutual trade, specialty ethnic culture, and tourism. The county aims to play a more active role in northeast Asian economic cooperation through five key industries: tourism, border trade, diatomaceous earth processing, food and medicine, and health and modern services. According to national census data, the Jilin Provincial Statistical Yearbook, and the Changbai County Statistical Bulletin, the population was 58,300 in 2020, with a per capita GDP of only CNY 43,900, accounting for 6.66% of Baishan City’s GDP. Local financial income decreased by 13.21%, while tax revenue fell by 7.4%. Additionally, the resident population declined by 14,300 people from 2010 to 2020, reflecting an average annual decrease of 2.17%. This trend is characteristic of a border economic cooperation zone, marking it as a typical example of a shrinking border city.

3.2. Data Sources and Pre-Processing

3.2.1. Land Use Data

The study utilized land use data from 2000, 2010, and 2020 (source: Resources and Environmental Science Data Registration and Publishing System [64], with a resolution of 30 × 30 m). This selection was based on two primary considerations: first, the period from 2000 to 2020 represented a critical phase of rapid urbanization and land contraction in China, and the three datasets effectively capture the long-term evolutionary trajectory of land use “expansion–contraction” in Changbai County. Second, the 30 m resolution is well-suited to the research scale (county level), allowing for the avoidance of data redundancy while accurately reflecting the spatial details of key land types, such as construction land and cultivated land. Using ArcGIS 10.8, the data were reclassified into six categories: cultivated land, forest land, grassland, water bodies, construction land, and unused land.

3.2.2. Natural Geographic Substrate Dataset

This study uses the 2020 elevation data provided by the Geospatial Data Cloud and slope data for that year extracted using the slope tool in ArcGIS 10.8. Additionally, the Chinese soil dataset from the National Glacial Permafrost Desert Science Data Center was employed [65], which includes data types such as soil type and soil texture. The accuracy of the soil data was adjusted to align with the aforementioned land use data using the resampling tool in ArcGIS 10.8. Furthermore, the 2020 river data provided by the National Geographic Information Resources Inventory Service System was incorporated, and the Euclidean distance tool in ArcGIS 10.8 was utilized to calculate the distance of various pixels from the river. The resampling tool in ArcGIS 10.8 was again used to ensure that the accuracy of the soil data was consistent with the accuracy of the previously mentioned land use data. The reasons for selecting the aforementioned data are outlined as follows: Elevation and slope directly impact the developability of construction land; soil type determines the quality of agricultural land and indirectly influences the expansion direction of construction land; and distance from rivers indicates ecological sensitivity. The natural geographical substrate dataset from 2020 was utilized to facilitate synchronous analysis of the land use data.

3.2.3. Climatic Condition Dataset

The study utilized monthly precipitation, evapotranspiration, and temperature data provided by the National Tibetan Plateau Science Data Center in 2020 [66,67,68]. It employed ArcGIS 10.8 to create NetCDF raster layers, generate raster layers, and perform image statistics to extract the average values of the monthly data from various images. These values were then summed to obtain the annual data. Additionally, the resampling tool in ArcGIS 10.8 was used to adjust the precision of the precipitation, evapotranspiration, and temperature data, ensuring consistency with the accuracy of the aforementioned land use data. The reasons for selecting the aforementioned data are outlined as follows: Climate conditions directly influence regional ecological carrying capacity and the willingness of populations to migrate. The climatic condition dataset from 2020 was utilized to facilitate synchronous analysis of the land use data.

3.2.4. Transportation Location Dataset

The study utilized 2020 government data on residences, city centers, ports, and road networks, as provided by the National Geographic Information Resource Catalog Service System. It extracted the actual distances of various pixels to government residency, Baishan City Center, and Changbai Port using the Baidu Maps API for path planning, along with the Inverse Distance Weighting tool in ArcGIS 10.8. Additionally, the study employed the Fishing Nets tool, Intersection tool, Raster Turning Point, and Euclidean Distance tool in ArcGIS 10.8 to determine the road network density for different pixels. The reasons for selecting the aforementioned data are outlined as follows: The distance between the government residency and the city center reflects the administrative and city centers’ influence, the distance to the Changbai port directly correlates with the cross-border connectivity of the border regions, and the density of the road network indicates the efficiency of transportation. The transportation location dataset from 2020 is utilized to facilitate synchronous analysis with the land use data.

3.2.5. Socio-Economic Dataset

The study utilized the 2020 spatial distribution kilometer grid data of China’s GDP [69], which was provided by the Resource and Environmental Science Data Registration and Publishing System (RESDPS). Additionally, it incorporated annual nighttime lighting data [70] for China and 2020 population density data from WorldPop. The reasons for selecting the aforementioned data are outlined as follows: GDP reflects the intensity of regional economic activities, and nighttime lights serve as a direct representation of the economic output of developed land, while population density is directly correlated with the extent of population contraction. The resampling tool in ArcGIS 10.8 was employed to standardize the image units (see

Table 1. Data sources and descriptions.

Datasets	Data Name	Data Type
Land Use Dataset	Land Use Data of 2000, 2010, and 2020	raster data/30 m
Natural Geographic Substrate Dataset	Slope	raster data/30 m
	Soil	raster data/1 km
	Distance from the River	raster data/30 m
Climate Condition Dataset	Average Annual Precipitation	raster data/1 km
	Average Annual Evapotranspiration
	Average Annual Temperature
Transportation Location Dataset	Distance from the Government Residency	raster data/30 m
	Distance from the Baishan City Center
	Distance from the Changbai Port
	Road Network Density
Socio-economic Dataset	Population density data	raster data/1 km
	Spatial Distribution Kilometer Grid Data of China’s GDP	raster data/1 km
	Nighttime Lighting	raster data/1 km

). The socio-economic dataset from 2020 is utilized to facilitate synchronous analysis with the land use data.

3.3. Methods

First, this study utilized land use data from Changbai County for 2000, 2010, and 2020 to reveal the spatial and temporal evolution characteristics of land use in the region. This was achieved through the application of the land use transfer matrix and spatial kernel density analysis methods. Secondly, using data extracted on land use expansion between 2010 and 2020 and incorporating key driving factors such as the natural environment, climatic conditions, transportation infrastructure, and socio-economic factors, the development probabilities for each type of land use were calculated. The contribution of these expansion factors was assessed using the Land Use Change Analysis Tool (LEAS) within the PLUS model. Again, using the 2010 land use data and the calculated development probabilities, the land use in 2020 was simulated with the help of the CA modeling tool (CARS) for multiple classes of random plate seeds in the PLUS model. In order to verify the accuracy of the simulation results, the simulation accuracy validation tool in the model was used, and the overall classification accuracy, user accuracy, producer accuracy, and Kappa coefficient were calculated through the confusion matrix, confirming the reliability of the model. Finally, the Markov chain tool and CARS in the PLUS model were combined to predict the land use demand of Changbai County in 2030 under the guidance of different planning concepts, and the corresponding land use scenarios were simulated. These prediction results provide a scientific basis for land use planning in Changbai County and a reference for future land management strategies. The following data and the technical route of the research methodology are shown in Figure 2.

3.3.1. Land Use Transfer Matrix

The land use transition matrix is a fundamental method for visualizing the temporal evolution characteristics in the study of spatio-temporal land use dynamics. This method involves comparing land use data from two distinct periods to construct a two-dimensional matrix that categorizes “source type–target type” and systematically reveals the conversion relationships among different land types, including the area of conversion, direction, and proportion. The primary advantage of this approach is its capacity to reflect both the outcomes of land use changes and the processes underlying those changes. The analysis is performed using the Raster Calculator tool in ArcGIS 10.8, and the calculation formula is expressed as follows:

$$S=({S_{ij})}_{n\times n}=\left[\begin{array}{cccc}{S}_{11}& S_{12}& S_{\dots }& S_{1n}\\ S_{21}& S_{22}& S_{\dots }& S_{2n}\\ \dots & \dots & \dots & \dots \\ S_{n1}& S_{n2}& S_{\dots }& S_{nn}\end{array}\right]$$

(1)

In Equation (1), $S$ represents the area of each land use type;$n$ denotes the number of land use types; and $i,j$ refer to the land use types at the beginning and end of the study period, respectively. Additionally, $S_{ij}$ indicates the area transferred from land use type $i$ at the beginning of the study to land use type $j$ at the end.

3.3.2. Spatial Kernel Density Analysis

Spatial kernel density is employed to visualize the spatial evolution characteristics in the study of spatio-temporal land use features. This analysis method is based on the spatial distribution density of point features. By calculating the density of point features surrounding each raster cell, the results can illustrate the degree of aggregation of various land use types within the spatial distribution. The primary advantage of this method is its ability to overcome the limitations of traditional static descriptions, allowing for a visual representation of the spatial heterogeneity characteristics of different land use types through continuous density surfaces. The analysis is performed using the kernel density analysis tool in ArcGIS 10.8. The calculation formula is expressed as follows:

$$Density={\displaystyle \frac{1}{{\left(radius\right)}^2}}\sum _{i=1}^n[{\displaystyle \frac{3}{\pi }}\times {pop}_i{(1-{\left({\displaystyle \frac{{dist}_i}{radius}}\right)}^2)}^2]$$

(2)

In Equation (2), $Density$ represents the kernel density value for a specific land use type. $i=1\dots$, n denote the input points, which include only those points that fall within a specified radius of the $\left(x,y\right)$ coordinates. The ${pop}_i$ variable indicates the number of land use pixels of that type at point $i$, while ${dist}_i$ refers to the distance between point $i$ and the $\left(x,y\right)$ position.

3.3.3. The PLUS Model

• Land Expansion Analysis Strategy (LEAS)

The Land Expansion Analysis Strategy (LEAS) is employed to analyze the various components of land use change between two distinct periods and to sample from the areas of increase. The random forest algorithm is utilized to identify the factors contributing to each type of land use expansion and to determine the driving forces behind them individually. This approach yields the development probability of each land use type and assesses the contribution of driving factors to the expansion of each land use category during the specified period. Given that transformation rules possess temporal attributes, they effectively describe the characteristics of land use changes within a defined timeframe [71]. The strength of this method lies in its ability to transcend the assumptions of traditional regression models, regarding linear relationships, enabling it to capture the nonlinear interactions among driving factors.

The detailed calculation steps are outlined as follows: First, the Extract Land Expansion (ELE) module of the model is utilized to identify expansion areas by comparing land use data from 2010 and 2020. Next, the LEAS module spatially correlates the expansion data with 13 driving factors categorized into four groups: natural geographical matrix, climate conditions, transportation locations, and socio-economic factors. The parameters for the random forest regression are configured to execute the model, quantifying the contributions of all land use types and each driving factor. Finally, the driving factors are classified and summarized by their attributes, and the contribution ratios of major categories (such as the natural geographic substrate) and minor categories (such as terrain slope) to all land use types and to the specific type of construction land are statistically calculated. Through secondary integration, Table 6 and Figures 5 and 6 are generated to provide a comprehensive analysis of the varying intensity of different driving factors on various types of land use, particularly construction land. The calculation formula is expressed as follows:

$$P_{i,k}^d\left(x\right)={\displaystyle \frac{{\sum }_{n=1}^{M}I(h_n\left(x\right)=d)}{M}}$$

(3)

In Equation (3), $x$ is a vector comprising multiple drivers, while $d$ can take on a value of either 1 or 0. A value of 1 signifies a transition from other land use types to land use type $k$, whereas a value of 0 indicates the occurrence of other shifts. The $h_n\left(x\right)$ function represents the predicted type ($n$) of decision tree $x$, and $M$ denotes the total number of decision trees.

2.

CA based on multiple random patch seeds (CARS)

Cellular automata based on multiple stochastic seeds (CARS) are utilized to simulate future land use change scenarios. When the domain effect of a specific type of land use is zero, changes are generated with respect to the development probability surface of each land use type. A cellular model employing multi-type random patch seeds and threshold reduction rules is implemented, enabling new patches to evolve under rule-based constraints. In accordance with anticipated future land use demands, a transfer cost matrix, adjacency coefficients, and other relevant parameters are established. The strength of this method lies in its flexibility to incorporate various development scenario constraints, facilitating multi-factor interactive simulations of future scenarios. The calculation formula is expressed as follows:

$$({P_{ij})}_{n\times n}=\left[\begin{array}{cccc}{P}_{11}& P_{12}& P_{\dots }& P_{1n}\\ P_{21}& P_{22}& P_{\dots }& P_{2n}\\ \dots & \dots & \dots & \dots \\ P_{n1}& P_{n2}& P_{\dots }& P_{nn}\end{array}\right]$$

(4)

In Equation (4), $n$ represents the number of land use types, and $P_{ij}$ denotes the probability of transitioning from land use type $i$ to land use type $j$, where 0 ≤ $P_{ij}$ ≤ 1. Additionally, the sum of the elements in each row equals 1.

3.

Model Validity Test

• The confusion matrix

The confusion matrix is a widely used tool in statistics for evaluating model performance, presenting the computational effectiveness of the model in the form of an error matrix. This method’s advantage lies in its ability to visually display the classification accuracy of the model across different land use types by comparing the classification matrices of predicted values with true values. The confusion matrix is structured as a $k$ rows and $k$ columns, where $n$ represents the total number of evaluation samples. Each predicted sample is assigned to one of the $k$ classes in the mapping (typically represented by the rows) and independently to one of the same $k$ classes in the true dataset (usually represented by the columns). The $n\_ij$ term denotes the number of samples classified as belonging to class $i$ ($i$ = 1, 2, …, $k$) in the mapping and class $j$ ($j$= 1, 2, …, $k$) in the reference dataset [72] (See

Table 2. The confusion matrix.

	Predicted Value					Row Totals
	Type	1	2	…	k
True Value	1	$n_{11}$	$n_{12}$	…	$n_{1k}$	$n_{1+}$
	2	$n_{21}$	$n_{22}$	…	$n_{21}$	$n_{2+}$
	…	…	…	…	…	…
	k	$n_{k1}$	$n_{k2}$	…	$n_{kk}$	$n_{k+}$
Total Columns		$n_{+1}$	$n_{+2}$	…	$n_{+k}$	$n$

).

Based on the confusion matrix, a series of evaluation metrics can be statistically evaluated for the classification extraction results, and the basic evaluation metrics are outlined as follows:

①

Overall Accuracy (OA) is calculated by the following formula:

$$OA=\sum _{k=1}^n{\displaystyle \frac{n_{kk}}{n}}$$

(5)

In Equation (5), the overall precision is the sum of the major diagonal (i.e., correctly classified sample units) divided by the total number of sample units in the entire error matrix, and an overall precision greater than 85% indicates that the classification results are valid [73].

②

User’s accuracy and producer’s accuracy are calculated as follows:

$$n_{i+}=\sum _{j=1}^k{n}_{ij}$$

(6)

$$n_{+j}=\sum _{j=1}^k{n}_{ij}$$

(7)

$${user’saccuracy}_i={\displaystyle \frac{n_{ii}}{n_{i+}}}$$

(8)

$${producer’saccuracy}_j={\displaystyle \frac{n_{jj}}{n_{+j}}}$$

(9)

In Equations (6) and (7), $n_{i+}$ represents the number of samples classified as class i in remote sensing classification, with the plus sign indicating the total values in the row. Equation (8) defines user accuracy as the proportion of correctly classified samples within a category to the total number of images classified as belonging to that category. This metric reflects the accuracy of the simulation results for each type of land use [74]. When user accuracy exceeds 70% [75], the simulation results are considered reliable. Equation (9) addresses producer accuracy, which indicates the proportion of correct classifications of a category to the total number of image elements classified as belonging to that category in the reference data. This measure reflects the accuracy of the land use data source [76,77]. Producer accuracy exceeding 59% [78] suggests that the accuracy of the land use source data reaches 90%, thereby meeting the accuracy requirements for land use data [79]. Additionally, $n_{+j}$ denotes the number of samples classified as belonging to class j in the reference dataset, with the plus sign indicating the total values in the column.

• Kappa coefficient

The objectivity of overall accuracy, user accuracy, and producer accuracy indicators depends on the sampling methods and samples used. After analyzing these indicators, a more objective measure is still needed to evaluate classification quality.

The Kappa coefficient is a commonly used statistical measure that assesses the consistency between two variables. The advantage of this method lies in its ability to quantify the overall accuracy of the model simulation, and it is frequently employed to test the consistency level between predicted and actual results. The Kappa coefficient exceeding 0.80 indicates a high level of consistency between the simulated image and the real image, suggesting that the simulation is effective [80]. The calculation formula is expressed as follows:

$$Kappa={\displaystyle \frac{n\sum _{i=1}^k{n}_{ii}-\sum _{i=1}^k\left(n_i+n_{+i}\right)}{n^2-\sum _{i=1}^k\left(n_{i+}{n}_{+i}\right)}}$$

(10)

In Equation (10), $n$ is the total number of columns in the confusion matrix (i.e., the total number of categories); $n_{ii}$ is the number of samples in row $i$ and column $i$ in the confusion matrix, i.e., the number of correctly categorized samples; $n_{i+}$ and $n_{+i}$ are the total numbers of samples in the row $i$ and column $i$, respectively; and n is the total number of samples used for the accuracy assessment.

3.3.4. Multi-Scenario Design

Four land use scenarios were established based on distinct planning concepts: natural development, incremental development, stock development, and decremental development. These four scenarios correspond to the varied development logic of the “natural trend–expansion–intensification–contraction response”, from inertia continuation to active intervention and addressing diverse demands. This framework provides systematic support for the comprehensive assessment of land use changes and their impacts under different development models. The specific details are outlined as follows (

Table 3. Land use transfer matrix.

Land Use Type		Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land
Natural development scenario	Cultivated Land	1	1	1	1	1	0
	Forest	1	1	1	1	1	1
	Grass	1	1	1	1	1	0
	Watershed	1	1	1	1	1	0
	Construction Land	1	1	1	1	1	0
	Unutilized Land	0	1	1	0	0	1
Incremental Development Scenario	Cultivated Land	1	1	1	1	1	0
	Forest	1	1	1	1	1	1
	Grass	1	1	1	1	1	0
	Watershed	1	1	1	1	1	0
	Construction Land	1	0	0	0	1	0
	Unutilized Land	0	1	1	0	1	1
Stock Development Scenario	Cultivated Land	1	1	1	1	0	0
	Forest	1	1	1	1	0	0
	Grass	1	1	1	1	0	0
	Watershed	1	1	1	1	0	0
	Construction Land	0	0	0	0	1	0
	Unutilized Land	0	1	1	0	0	1
Reduced Development Scenario	Cultivated Land	1	1	1	1	0	0
	Forest	1	1	0	0	0	0
	Grass	1	1	1	1	0	0
	Watershed	1	1	1	1	0	0
	Construction Land	1	1	1	1	1	1
	Unutilized Land	0	1	1	0	0	1

): (1) Natural development scenario: According to the transfer probability matrix of land use from 2010 to 2020, the current trend of land use change will continue through 2030. (2) Incremental development scenario: The incremental development scenario is a plan that uses new construction land supply as the main means to promote urban development, mainly by expanding the scale of land use [81]. The economic development scenario restricts the transfer of construction land to land use types other than arable land and increases the probability of conversion of arable land, forest land, grassland, water area, and unutilized land to construction land by 40%, 30%, 10%, 10%, and 30%, respectively [82]. (3) Stock development scenario: The stock development scenario is a plan to realize urban development mainly through the revitalization, optimization, tapping, and upgrading of the stock of land under the conditions of keeping the total scale of construction land unchanged and not expanding the urban space [83]. Based on the scenario of protecting cultivated land while promoting economic development, this policy restricts the conversion of other land types into construction land. It establishes a 50% probability for the conversion of grassland to cultivated land and forest land, as well as a 50% probability for the conversion of unutilized land to forest land. Additionally, there is a 30% probability of the conversion of forest land and water areas to cultivated land [71]. (4) Reduced development scenario: The reduced development scenario is an important planning option to cope with local economic decline and population scale reduction, and regional resource integration is carried out on the premise of construction land scale reduction [84,85]. Under the ecological protection scenario, while restricting the transfer of other land to construction land and increasing the probability of transferring construction land to other land, the probability of transferring construction land to forest land, grassland, waters, and unutilized land is set to increase by 40%, and restricting the transfer of forest land to types of land other than cultivated land reduces the probability of transferring forest land to cultivated land by 20%, and the probability of transferring grassland and waters to forest land increases by 30% [86].

4. Results

4.1. Characteristics of the Spatial and Temporal Evolution of Land Use

4.1.1. Time Evolution Characteristics

Combined with the land use transfer matrix from 2000 to 2020 (

Table 4. Area and percentage of land use in Changbai County, 2000–2020.

Land Use Type	2000		2010		2020
	Area	Share	Area	Share	Area	Share
Cultivated Land	77.61	3.13	183.75	7.40	194.74	7.85
Forest	2251.64	90.74	2255.61	90.90	2238.43	90.20
Grass	127.96	5.15	16.66	0.67	16.06	0.65
Watershed	0.39	0.02	9.94	0.40	13.10	0.53
Construction Land	5.45	0.22	15.05	0.61	15.35	0.62
Unutilized Land	18.45	0.74	0.47	0.02	3.82	0.15

Area is measured in square kilometers (km²), and share is expressed as a percentage (%).

and

Table 5. Long Branch County land use transfer matrix, 2000–2020.

2000	2010							Turnover
	Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land	Total
Cultivated Land	57.02	10.35	0.03	3.68	6.53	0.00	77.61	20.59
Forest	108.98	2121.74	12.72	4.81	3.37	0.00	2251.62	129.88
Grass	16.57	105.02	3.29	1.14	1.46	0.47	127.95	124.66
Watershed	0.01	0.13	0.00	0.26	0.00	0.00	0.39	0.14
Construction Land	1.17	0.54	0.00	0.05	3.69	0.00	5.45	1.76
Unutilized Land	0.00	17.84	0.61	0.00	0.00	0.00	18.45	18.45
Total	183.75	2255.61	16.66	9.94	15.05	0.47	—	—
Inflow	126.73	133.88	13.37	9.68	11.36	0.47	—	—
2010	2020							Turnover
	Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land	Total
Cultivated Land	165.79	15.79	0.1	1.22	0.85	0.00	183.75	17.96
Forest	27.59	2220.62	0.63	2.82	0.59	3.37	2255.61	35
Grass	0.12	1.17	15.30	0.05	0.01	0.00	16.66	1.35
Watershed	0.43	0.58	0.01	8.90	0.03	0.00	9.94	1.04
Construction Land	0.81	0.27	0.01	0.08	13.88	0.00	15.05	1.17
Unutilized Land	0.00	0.01	0.01	0.00	0.00	0.45	0.47	0.02
Total	194.74	2238.43	16.06	13.07	15.35	3.82	—	—
Inflow	28.95	17.81	0.76	4.17	1.48	3.37	—	—

Area is measured in square kilometers (km²).

), the areas of arable land, watershed, and construction land have increased by 117.13 km², 12.70 km², and 9.90 km², respectively. (1) Notably, the increase in arable land is 11.83 times greater than that of construction land. Conversely, the areas of grass, unutilized land, and forest land have decreased by 111.90 km², 14.63 km², and 13.21 km², respectively, with the total decrease in grass being 8.47 times that of forest land. Cultivated land is primarily converted from forest land and grass, while grass is mainly converted to forest land and cultivated land, with only a small portion of grass being transferred to construction land. (2) Combining the changes in land use dynamics in the two phases of 2000–2010 and 2010–2020, it is found (Figure 3) that cultivated land, watershed, and construction land all show an inertial development trend of continuous increase and that the development rate slows down in the second phase, with the slowdown of construction land being the most obvious. Unutilized land shows a V-shaped trend of first decreasing, then increasing, and the growth rate in the second stage is smaller than the decrease rate. Forest land shows a trend of increasing first, then decreasing, and although grass shows an inertial shrinkage trend of continuous decrease, the shrinkage rate slows down obviously in the second stage. (3) Under the background of population contraction and development, the area of construction land in Changbai County from 2000 to 2020 is mainly transferred from cultivated land and forest land, accounting for 55.15% and 32.40%, respectively. the growth rate of construction land from 2000 to 2010 is 86.79 times that of 2010 to 2020, and the types of land transferred in the two phases are both cultivated land (57.47%) and forest land (30.84%). The area transferred from cultivated land (57.47%) and forest (30.84%) in the two phases is 7.38 km² and 3.96 km², respectively, and the area transferred from other types of land is 1.50 km² in total.

4.1.2. Spatial Evolution Characteristics

Combining the kernel density distribution of land use from 2000 to 2020 (Figure 4), the following observations can be made: (1) The spatial agglomeration of construction land, cultivated land, forest, and watershed areas has increased. Construction land and cultivated land have developed two primary centers in the southeast and the west, while water areas have formed a relatively decentralized polycentric structure. Additionally, the center of forest land has gradually shifted towards the central–northern region. (2) The spatial aggregation of both grass and unutilized land has diminished, with grass transitioning from a polycentric to a monocentric distribution in the north and unused land migrating from the north–central region to the eastern part of the country. (3) The primary center of construction land remains consistently located in Changbai Town along the border, exhibiting spatial characteristics of synergistic development with Malugou Town. The secondary center is situated in the inland Badaogou Town and extends towards the central city of Baishan.

4.2. Construction and Testing of the PLUS Model

4.2.1. Analysis of Land Use Drivers

Combining the contribution levels of various drivers to land use change in 2020 (

Table 6. The contribution levels of various drivers to types of land use change.

Primary Driving Factors	Secondary Driving Factors	Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land
Natural Geographic Substrate Dataset	Slope	0.05	0.04	0.09	0.02	0.06	0.00
	Soil	0.13	0.18	0.02	0.01	0.01	0.00
	Distance from the River	0.07	0.06	0.05	0.08	0.03	0.00
Climatic Condition	Average Annual Precipitation	0.05	0.10	0.07	0.40	0.01	0.09
	Average Annual Temperature	0.02	0.03	0.06	0.05	0.03	0.00
	Average Annual Evapotranspiration	0.11	0.14	0.26	0.28	0.03	0.01
Transportation Location	Distance from the Government Residency	0.22	0.14	0.05	0.00	0.26	0.14
	Road Network Density	0.06	0.04	0.01	0.03	0.35	0.00
	Distance from Changbai Port	0.05	0.10	0.10	0.01	0.35	0.01
	Distance from Baishan City Center	0.07	0.06	0.03	0.01	0.04	0.01
Socio-economic	GDP	0.05	0.03	0.14	0.04	0.01	0.30
	Population	0.07	0.04	0.02	0.04	0.14	0.45
	Nighttime Lighting	0.07	0.04	0.10	0.03	0.17	0.00

, Figure 5 and Figure 6), the following findings emerge: (1) Among the primary drivers, climatic conditions (28.98%), socio-economic factors (28.64%), and transportation location (27.60%) significantly impact land use change, while the natural geographic substrate (14.78%) has a comparatively smaller effect. (2) Among the secondary driving factors, distance to the port, population density, soil conditions, and average annual evapotranspiration exhibit notable effects on land use change. (3) Changes in construction land use are primarily influenced by transportation location (52.21%) and socio-economic factors (30.86%). Additionally, among the secondary driving factors, distance to the port, nighttime lighting, slope, and average annual temperature have significant effects on changes in construction land use.

4.2.2. PLUS Model Validity Test

Based on the actual land use data from 2010, the land use changes for 2020 were simulated with a sampling rate of 10% [87]. A comparison of the simulation results with the land use confusion matrix of the actual situation is presented in

Table 7. Long Branch County land use confusion matrix, 2020.

	Predicted Value							Row Totals	Producer’s Accuracy
	Land Use Type	Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land
True Value	Cultivated Land	18,754	2560	40	133	86	0	21,573	86.93%
	Forest	2630	244,574	89	316	119	341	248,069	98.59%
	Grass	8	114	1642	5	0	0	1769	92.82%
	Watershed	43	382	0	918	6	0	1349	68.05%
	Construction Land	94	87	0	12	1477	0	1670	88.44%
	Unutilized Land	0	387	1	0	0	57	445	12.81%
	Total Columns	21,529	248,104	1772	1384	1688	398	274,875	—
	User’s accuracy	87.11%	98.58%	92.66%	66.33%	87.50%	14.32%	—	—
Overall Accuracy = 97.29%

[77,88]. The confusion matrix test results, shown in Table 6, indicate an overall accuracy of 97.29%. The producer accuracy varied significantly, ranging from 98.58% for forest to 14.32% for unutilized land. The highest user accuracy was observed for forest, at 98.59%, followed by grassland, at 92.82%, and cropland, at 86.93%. Watershed had a user accuracy of 68.05%, while unutilized land exhibited the lowest accuracy, at 12.81%.

The results of the Kappa coefficient test (

Table 8. Land use simulation accuracy verification for Long Branch County, 2020.

Kappa Coefficient	Producer’s Accuracy
	Cultivated Land	Forest	Grass	Watershed	Construction Land	Unutilized Land
0.85	0.87	0.99	0.93	0.68	0.88	0.13

) indicate that the Kappa coefficient for different land use types is 0.85, with a simulation accuracy of 88.44% for construction land. These results demonstrate that the simulation is successful and that the model effectively reflects the trends in land use change in the context of population contraction.

4.3. Comparison of Land Use Scenario Simulations

First, according to the Land Space Plan of Baishan City (2021–2035) and the Land Space Plan of Changbai Korean Autonomous County (2021–2035), Badaogou Town, Changbai Town, and Malugou Town are identified as key areas for import and export processing, as well as new material research and production. These towns represent the core of comprehensive services along the Yalu River development axis of Baishan City. Secondly, Changbai Town, Badaogou Town, and Malugou Town are the principal towns in Changbai County, with a combined population of 53,600, accounting for 73.12% of the total population of the county. Finally, based on the multi-scenario simulation results of the PLUS model, significant changes in construction land within Changbai County are primarily concentrated in Badaogou Town in the west (Zone A) and in Changbai Town (Zone B) and Malugou Town (Zone C) in the southeast. Therefore, these areas were selected for a comparative analysis of the different scenario simulation results.

Combining the simulation results of Changbai County’s land use spatial distribution in 2030 under various planning concepts (Figure 7, Figure 8 and Figure 9), the following findings emerge: (1) Natural development scenario: The areas of unutilized land, watersheds, arable land, and construction land have increased by 83.83%, 21.63%, 5.06%, and 2.19%, respectively, while the areas of grasslands and forests have decreased by 2.95%, and 0.70%, respectively. (2) Incremental development scenario: The expansion of construction land is primarily concentrated around government sites, border crossings, and the city center, which features a dense road network. In comparison to the natural development scenario, the growth of built-up land is predominantly observed in Zones A, B, and C, with Zone A experiencing the most significant increase. This trend can be attributed to two main factors: first, the northern parts of Zones B and C are largely forested and situated at higher elevations, which restricts human activities and development. Second, Zones B and C, being closer to the China–North Korea border, lack interaction with neighboring international cities and are farther from the city center of Baishan City, resulting in limited benefits from urban development dynamics. (3) Stock development scenario: The area of built-up land has remained stable, while the areas of unutilized land and arable land have increased by 82.77% and 9.39%, respectively. Concurrently, the areas of grassland, water, and forest land have decreased by 6.84%, 3.73%, and 0.89%, respectively. Compared to the natural development scenario, the overall growth trend of arable land is more pronounced, and its spatial distribution is more decentralized, while the decline in the scale of forest land and grassland is slightly more pronounced. (4) Reduced development scenario: The scale of construction land has decreased by 10.89%. In comparison to the natural development scenario, the construction land in the central area of Changbai Township has experienced a slight reduction, while the construction land in the central areas of Zone C and Zone A has decreased more significantly, by 0.40 km² and 0.35 km², respectively. This change may be attributed to two reasons: first, as the core area of Changbai County’s integrated town development, the construction land area in Zone B has seen a lesser reduction; second, guided by the spatial development strategy of the central urban area of the east-connected, west-expanded, north-introduced, and south-connected areas, Changbai County concentrates on the development of Changbai Town with resources, which leads to the most significant reductions in the construction land areas of Zone A and Zone C.

5. Discussion

5.1. Unique Spatial Distribution Characteristics of Border-Shrinking County Towns

This study elucidates the spatial evolution mechanism of the “dual-center reconstruction” of construction land in Changbai County using the PLUS model, thereby challenging the traditional contraction theory’s reliance on inland conditions. The research indicates that the construction land in Changbai County has exhibited a heterogeneous expansion pattern characterized by a “transformation from a single center to dual centers”. Spatial reconfiguration logic significantly differs from the traditional contraction patterns of perforation and circle-pie observed in shrinking cities [42].

Most studies on shrinking cities have concentrated on resource-depleted inland areas, where spatial contraction typically manifests in two predominant patterns: “punctured” or “ring-shaped” [42]. In contrast, the built-up area of Changbai County has exhibited a heterogeneous expansion pattern, characterized by a transition from a single center to a dual center—specifically, the secondary growth poles formed by the border port (Changbai Port) and the county’s sub-center (Badaogou Town). This phenomenon is primarily the result of proactive administrative intervention, which diverges from the traditional urban shrinkage processes in cities primarily influenced by market mechanisms [89]. The establishment of this “policy-driven growth pole” not only reflects the spatial governance strategies aimed at achieving the dual objectives of “security and development” in border regions but also challenges the traditional contraction theory’s simplistic framework of “shrinkage = decline” [90]. Given the stringent constraints of national security, the spatial governance of shrinking counties must move beyond a “passive response to shrinkage and develop a dual-track mechanism of “bottom-line control + flexible reserve”. On one hand, it is essential to strictly adhere to the ecological red line of 260.6 km along the border to ensure national security; on the other hand, it is crucial to foster new growth opportunities through the port economic corridor, thereby transforming “shrinkage pressure” into “development momentum”. These insights not only offer practical pathways for the spatial governance of Changbai County but also enhance the understanding of the unique experiences of border-shrinking areas in implementing the “Border Enrichment and People’s Prosperity Action Plan”, deepening the theoretical comprehension of the dialectical relationship between “shrinkage and development”.

5.2. The Advantages and Adaptability of Proactive Contract Planning

The multi-scenario simulation of the PLUS model validates the nonlinear characteristics of “shrinkage and efficiency enhancement” in border areas, thereby expanding the applicability of traditional shrinkage theory. The governance of global shrinking cities is undergoing a strategic shift from “resisting shrinkage” to “smart shrinkage” [91]. This transition has resulted in varied practical approaches in notable international cases: Leipzig, Germany, has implemented a “scale adjustment” strategy, utilizing planning measures to resize the city in response to its declining urban population and to integrate the surplus real estate market. In contrast, Youngstown, United States, has adopted an “active adaptation to shrinkage” strategy, focusing on enhancing the city’s image and improving residents’ quality of life as central elements of its urban development framework while aligning with the regional economic structure. Meanwhile, Liverpool, UK, has pursued a “city renewal” strategy to tackle the challenges of shrinkage, employing a plan that combines an “industrial park + urban development company + urban development fund” to revitalize the city’s land and housing market [50]. Domestic explorations in shrinking areas have further enriched the localized connotations of “smart shrinkage”. As a typical resource-exhausted city, Hegang in Heilongjiang Province has maintained the stability of its population and urban development by transforming abandoned land into “land banks” through the construction of urban green infrastructure and the revitalization of neglected sites [92]. Muleng, also in Heilongjiang Province, has leveraged its border port advantages to adjust its urban development strategy to a “defensive” shrinkage planning approach, achieving high-quality urban and rural development through the strategy of “cultivating internal driving forces and building livable urban and rural areas” [59]. Genhe in Inner Mongolia, adhering to the principle of “ecological protection first”, has implemented a model that combines the integration of industry and city, guiding the population to concentrate in the central urban area and focusing efforts on its development [93].

Through multi-scenario simulations of the PLUS model, this study further verified that implementing “active reduction and contraction” can significantly enhance spatial efficiency. When the construction land area of the Changbai-Malu Gou comprehensive service core area was reduced by 1.67 square kilometers, the implementation of “active reduction and contraction” improved spatial efficiency. As the construction land area decreased by 1.67 square kilometers, not only did the population density of the Changbai-Malu Gou comprehensive service core area increase, but the intensive development of newly introduced import and export processing and R&D industries was also promoted in Badao Gou Town. This approach facilitated a virtuous cycle of “spatial reduction–industrial intensification–population concentration”, demonstrating the applicability of the “smart contraction” theory in border areas. This empirical result not only aligns closely with the core logic of “smart contraction” in international cases—activating existing resources and reconstructing economic resilience through spatial optimization—but also addresses the unique challenges of border areas, such as balancing security constraints with development needs. It supplements the Chinese solution by exploring a path that transitions from “contraction pressure” to “development momentum”, taking into account national security, ecological protection, and economic efficiency. This provides a differentiated practical reference for the global governance of shrinking cities.

5.3. The Differences in Land Use Strategies Between Population Outflow Areas and Population Growth Areas in Northeast China: A Future Perspective

Due to population outflow and ecological pressures, land use in border cities primarily manifests as agricultural expansion (the conversion of forested areas to farmland) [94] and ecological degradation (the transformation of forest land into construction land). The driving mechanisms have shifted from being predominantly influenced by socio-economic factors in the early stages to being primarily driven by natural factors in later stages, with these driving factors exhibiting a non-linear enhancement pattern. In contrast, growth regions that depend on resource endowments and policy support are characterized by the continuous agglomeration of construction land and the efficient expansion of industrial land. Relevant studies utilizing the Geodetector model have found that in these regions, the driving mechanisms are consistently dominated by socio-economic factors, with the driving factors displaying a linear superposition effect [95]. The government expands the scale of construction land by facilitating the conversion of other types of land into construction land. Additionally, studies employing the CLUE-S and Markov-CA models have demonstrated [96] that in areas experiencing population outflow (such as Jiamusi City and Jixi City), there is a trend of forest land being converted to farmland; conversely, in areas with population growth (such as Changchun City and Dalian City), there is a significant loss of forest land, primarily transferred to construction land. This disparity fundamentally reflects the dual differentiation model of “ecological constraint-induced contraction” and “resource-driven growth” in Northeast China [97]. Therefore, in the future, Northeast China must adopt differentiated policy approaches to achieve human–land coordination and sustainable development.

5.4. Planning and Development Strategies for Declining Border Counties

Focusing on the specific goal of enhancing population aggregation and industrial carrying capacity, in accordance with the requirements of the “Three-Year Action Plan for the Development of the Technical Standards System for National Spatial Planning (2021–2023)”, the following planning and development strategy for contraction-type border counties is proposed: (1) At the general planning level, the proposed approach is based on the “Action Plan for Enriching Border Areas and Their Inhabitants”, combined with the real needs of stabilizing and fixing the border for development, to enhance the population density of the areas along the border. The first goal is to build a center-node strip system of strip towns along the border. Guided by the principle of “ecological protection and settlement consolidation”, the economic vitality and regional competitiveness of the areas along the border will be enhanced through urban renovation, ecological compensation, and relocation of villages and settlements. The second aim is to establish a land use control policy that combines compact and reduced development. With “compact utilization and moderate reduction” as the guideline, through the “tight + shrink” land use structure optimization orientation, on the one hand, the center and node towns establish the land use optimization direction, focusing on compact development, combining the population and land use status quo, and adopting the “small group” approach to optimize land use. On the other hand, the town development boundaries of other shrinking towns are adjusted for shrinkage, and moderately shrinking planning schemes are formulated to place boundaries according to the local conditions so as to comprehensively improve the level of intensive land use in border areas under the background of population shrinkage. The final goal is to establish a strategy to improve the attractiveness of cities and towns by perfecting functions and classifying and clustering, under the guidance of “precise optimization and zoning guidance”, by upgrading the public environment and infrastructure of the central towns and cultivating the science and education industries, as well as medical care and material processing, in the nodal towns to provide new types of jobs. the attractiveness of different towns will be enhanced so as to improve the functions of the towns in the border areas and support the attraction of the border areas to the agglomeration of the population. In response to the dual challenges of population decline and border stability, a multi-center network and contraction policy can be implemented to guide spatial optimization. The vitality of the border zone is enhanced through a three-dimensional approach of “ecological protection, functional reorganization, and intensive development”, supporting the sustainable development of both the population and the economy. (2) At the level of detailed planning, we propose an approach based on the “Five-Year Action Plan for the Comprehensive Implementation of the New Urbanization Strategy Focused on People”, combined with the needs of domestic urban system construction, to enhance the development of internal circulation of industries, with advantages in the border areas and inland. The first goal is to accelerate the transportation link between the center and node towns and the inland neighboring towns. Taking “comprehensive network and grid layout” as a guideline full use should be made of the existing road transportation network by connecting major arterial roads and establishing a multi-channel transportation network, in addition to reserving land for roads and transportation facilities in accordance with the demand for development of road transportation so as to establish a number of fast distribution channels for foreigners and strengthen the ability of interconnection and mutual communication of resource elements between the border areas and inland cities and towns. This can facilitate interconnection and intercommunication capacity between border areas and inland towns. Secondly, the center and node towns and the inland neighboring towns have strengthened the inner-circulation links between the deep processing of characteristic agricultural products and the border culture and tourism industry. With “regional agglomeration and key development” as the guideline, through regional synergy and tourism zoning policy guidance, among other means, can give full play to the advantages of the production of regional characteristic agricultural products in border counties and the cultural advantages of ethnic minority villages and towns, in addition to developing differentiated construction land indexes in accordance with major projects and general projects. Differentiated construction land indexes have been formulated to extend and improve the characteristic industrial chain and supply chain along the border and to enhance the synergistic development of industries and internal economic circulation between border areas and inland cities and towns. Finally, we suggest strengthening lifeline facilities between the center and node towns and the inland neighboring towns. Taking “interconnection and key layout” as a guideline, based on the belt-shaped town system along the border, we recommend the establishment of an interconnection and dual-use infrastructure guarantee system through the establishment of comprehensive pipeline corridors for lifelines and the construction of a national key emergency reserve so as to strengthen the sustainable and renewable capacity of advantageous industries in border areas. In response to the integration of industrial chains and regional connectivity, it is essential to focus on the upgrading of transportation networks and linking characteristic industries. These efforts should aim to transform border counties from “peripheral areas” into “open hubs”. (3) At the level of special planning, based on the “Notice on Several Measures for Supporting the Construction of a New Development Pattern and Promoting High-Quality Development of Border (Cross-Border) Economic Cooperation Zones”, combined with the demand for the construction of the national FTZ, we recommend enhanced development of external circulation between the border areas and foreign markets in the energy resource industry. First of all, the role of the FTZ window should be strengthened. Taking “comprehensive improvement of soft and hard environments” as a guideline, we recommend the promotion of the construction of soft and hard environments through the construction of comprehensive international logistics hubs, the construction of public service facilities, the creation of an internationalized business environment, the establishment of a higher level of opening platforms, etc., in addition to supporting the expansion of FTZs with a better development trend to lay a solid foundation for the development of FTZs in border areas. Secondly, we recommend a forward-looking FTZ layout and advanced manufacturing export trade. Taking “international trade and cross-border platform development” as a guideline, through the development of a port economy; the cultivation of warehousing, transit, processing and distribution; and the construction of cross-border characteristics of the industrial chain and other means, the aim is not only to develop externally oriented industrial clusters, open up of the bases of each focus, and realize the optimal allocation of resources but also to further promote the common industrial chain and complement each other and to realize the interests of all parties involved in the cross-border trade of the FTZ. FTZ cross-border trade can maximize the interests of all parties and play a strong and driving role in local economic development. The final aim is to strengthen the role of FTZ county coordination and linkage. Taking “industrial linkage and external radiation” as a guideline, establishing the FTZ in the central town and coordinating the distribution of the functions of the FTZ between the central town and the node towns enhances the spillover effect of cross-border industrial development within the border areas and amplifies the overall radiation-driven effect of the FTZ (Figure 10). In response to the dual demands of constructing free trade zones and fostering innovation in cross-border cooperation mechanisms, an approach that emphasizes cross-border collaboration and external circulation empowerment is adopted. This strategy aims to transform border counties into pivotal nodes within the “dual circulation” concept.

5.5. Innovation and Limitation

Compared with traditional studies that mostly focus on urban expansion or shrinking cities in general, the innovation of this paper lies in four aspects: First, it is the first to apply the PLUS model to a shrinking border county, and through the simulation of four scenarios—natural development, incremental development, stock development and decremental development—it was found that under the decremental scenario, the scale of construction land decreases by 1.67 km², while the population concentration and industrial carrying capacity of Changbai Town increase simultaneously, breaking through the traditional incremental planning paradigm. Second, it integrates multi-source data (climate, terrain, socio-economic, etc.) to construct a driving model, deepening the under-standing of the driving mechanism of land use in border areas. Third, the proposed decremental development planning strategy involves the construction of a three-dimensional strategic system comprising a “strip-shaped urban system”, internal circulation of advantageous industries, and the establishment of national free trade zones in response to the development paradox faced by shrinking border cities. This research not only provides a new theoretical paradigm and practical path for land use planning in shrinking border counties but also has important practical significance for maintaining border security, promoting the notion of “prosperous border and enriching the people” and high-quality territorial space governance. Finally, in contrast to traditional GIS-based representations of the evolution of complex land phenomena, which have limitations [98,99], the PLUS model utilizes cellular automata (CA) and multi-type random seed mechanisms to optimize the quantitative structure and spatial pattern simulation of land use. Moreover, the PLUS model not only flexibly adjusts constraint conditions and adapts to long-term land use simulations, but it also effectively accommodates data deficiencies or fluctuations. This capability allows it to generate optimization schemes that align more closely with specific objectives.

This study also has the following shortcomings. Firstly, to make the simulation results comparable to actual land use, the classification of land use data was adjusted; however, this adjustment may limit the reliability of the research findings. Secondly, although the study employed a confusion matrix and Kappa coefficient to comprehensively assess the model’s effectiveness, it did not address the issue of spatial distribution of classification errors. Future research could utilize spatial autocorrelation statistics to analyze the spatial distribution of these errors or incorporate field verification data for a more thorough accuracy evaluation [100]. Finally, the driving factors considered by the model may not be comprehensive enough. Future research needs to delve deeper into driving factors, in combination with policy planning, optimization of model parameters, and improvements in simulation accuracy.

6. Conclusions

In this study, land use scenarios were predicted under the following four conditions utilizing the PLUS model: natural development, incremental development, decremental development, and stock development,. The empirical conclusions are presented as follows:

(1)

Incremental development exacerbates the conflict between people and land. The dual centers of Changbai Town and Malugou Town have not achieved effective synergy, resulting in a significant spatial mismatch between the expansion of construction land and higher-level planning.

(2)

Stock development presents a paradox. The fragmentation rate of cultivated land is high, leading to a conflict between the preservation of ecological spaces and agricultural areas. This situation reveals the dilemma of “protecting grain versus protecting forests” during the transformation process. The reduction of forest and grassland areas coexists with the successes of the policy aimed at returning farmland to forest, underscoring the stark contradiction between development and protection.

(3)

Reduced development emphasizes its benefits. Implementing decremental planning can enhance the efficient use of construction land, promote population concentration, and increase industrial carrying capacity, thereby validating the feasibility of “shrinkage and quality improvement.

Based on the empirical conclusions presented above and considering the unique characteristics of the shrinking border counties, this paper proposes four-dimensional recommendations for future planning. These recommendations align with the core pathways for sustainable urban development suggested by UN-Habitat and include the following four-dimensional suggestions for future planning. First, we suggest the implementation of a land space governance model characterized by “reduction planning + flexible control”. We recommend the delineation of three key zones and lines: the ecological protection red line, permanent basic farmland, and urban development boundaries. Concurrently, construction should be cleared while exploring the potential of existing land. Through high-density and mixed-use land development and urban planning, we recommend the creation of flexible and expandable urban spaces to enhance urban resilience. Second, a regional coordination strategy that combines “cross-border linkage and industrial restructuring” should be adopted at border ports to establish distinctive industrial corridors and implement differentiated industrial development based on the characteristics of economic shrinkage, such as cross-border logistics and cultural tourism integration. This approach aims to create a cycle of “prospering the border through industry” and facilitate the transformation of the urban economy towards high-value-added sectors and technological innovation, thereby reducing reliance on a single industry. Third, we recommend the initiation of ecological restoration projects that prioritize ecological health and resilience construction, in addition to the designation of biological corridor buffer zones in ecologically sensitive areas, the promotion of the use of cold-resistant vegetation for slope protection, and the implementation of smart monitoring technologies. Natural ecosystems should also be integrated into urban planning, utilizing forest and wetland ecosystems to mitigate climate risks and foster sustainable urban environmental development. Finally, we propose further innovation in the coordination mechanism by developing a digital twin early warning platform. This platform will dynamically monitor the inter-relationships between population flow, land use, and industrial changes, guiding the rational distribution of populations and industries while curbing the unregulated expansion of cities. In conclusion, this study has integrated the essential pathways for the sustainable future development of cities as proposed by UN-Habitat. It has established a comprehensive policy framework that encompasses four key components: spatial governance, industrial synergy, ecological resilience, and mechanism innovation. This framework provides a replicable, high-quality development plan for similar border regions.

However, this study did not account for the compounded effects of sudden public events, such as the COVID-19 pandemic, on border shrinkage. Additionally, the PLUS model has limitations in data accuracy when simulating cross-border element flows. Future research could focus on several areas, including enhancing the classification and accuracy assessment of land use data, implementing policy interventions, and establishing a long-term monitoring and evaluation mechanism. These efforts aim to develop planning strategies that integrate resistance and adaptation, as well as conduct empirical research in accordance with the planning practices of Changbai County. This approach seeks to explore a spatial planning strategy and optimization pathway for border counties facing shrinkage, with the hope of providing a model for the transformation and development of similar towns.