Belt and Road Initiative and Urban Landscapes: Quantifying Land Use Changes and Development Strategies in Minsk, Vientiane, and Djibouti

Abstract

This paper presents a comparative study of land use in different regional types under the Belt and Road Initiative (BRI) through a case study of Minsk, Vientiane, and Djibouti. The BRI has created significant opportunities for economic cooperation between China and participating countries. As key instruments for advancing the BRI and fostering regional economic development, overseas industrial parks play a crucial role in attracting investment, facilitating workforce training, and promoting local infrastructure development. With the growing number of overseas industrial parks, the need for their efficient and scientifically guided development has attracted increasing scholarly attention. However, a pressing issue remains of how to formulate context-specific and scientifically grounded development plans for these parks that account for diverse regional natural conditions and economic foundations. This study examines the China-Belarus Industrial Park in Minsk, the Saysettha Development Zone in Vientiane, and the Djibouti International Free Trade Zone, using remote sensing data and the patch-generating land use simulation (PLUS) model to analyze land use changes and driving forces in these three cities from 2000 to 2020, thereby revealing regional land-use patterns and their developmental differences. The findings indicate that population density, transportation infrastructure, and topographic features are the primary driving forces behind land use changes in the areas surrounding the industrial parks, significantly influencing the expansion of construction land and the utilization patterns of adjacent land types. Moreover, variations in natural conditions and economic strategies across regions result in divergent land use changes, giving rise to distinct urbanization characteristics. The coordinated development of ports, industrial parks, and urban areas plays a vital role in advancing regional economic integration. Through quantitative analysis, this study provides theoretical support and practical guidance for the planning and development of overseas industrial parks along the BRI. It advocates for tailored development strategies that align with specific regional contexts, in order to avoid the shortcomings of one-size-fits-all policies under varying natural and economic conditions.

1. Introduction

Since the global financial crisis in 2008, the overall pace of global economic growth has slowed down. China’s development has also experienced a deceleration, primarily due to structural factors [1,2]. Perkins (2015) pointed out that China’s manufacturing value-added as a share of GDP had reached its peak. However, during the transition from manufacturing to services, the service sector has been unable to sustain the same high growth rate as manufacturing once did [3]. Cai and Lu (2013) argued that China’s economic growth over the past three decades was largely driven by its demographic dividend, while the ongoing demographic transition has become a major factor slowing the country’s economic growth [4]. Multiple studies have suggested that under the broader context of global economic downturn, Chinese enterprises should adopt a “going global in groups” strategy [5]. Through the BRI, China aims to strengthen economic ties with countries along the route, promoting regional development, mutual benefit, and shared prosperity [6].

Since its launch in 2013, the BRI has aimed to enhance regional economic connectivity and promote multilateral cooperation across economic, political, and cultural domains [7]. Against the backdrop of deepening globalization, the demand for international cooperation has increased significantly, with growing interdependence among countries. At the same time, after decades of rapid growth, China’s economy is undergoing a transitional phase, during which cooperation with developing countries has intensified to facilitate a new phase of globalization.

In response to these challenges, China has invested in multiple overseas industrial parks along the BRI corridor and participated in the construction of local railways, highways, and ports, thereby promoting local economic development and achieving mutual benefit [8,9].

Currently, overseas industrial parks exhibit a range of characteristics, including a coexistence of government-led and market-driven development models [10], a combination of localization and international cooperation [11] industrial diversity and technology transfer [12], as well as a growing emphasis on green development and sustainability [13]. Since the inception of the BRI, Chinese enterprises have actively engaged in the development of industrial parks, resulting in a steady increase in the number of national-level parks, expanded spatial distribution, and a more diversified and extended industrial chain in overseas parks.

However, the growing number of overseas parks has introduced challenges stemming from complex regional environments during their development. Due to the absence of a globally coordinated planning perspective and the complexity of both geographic and geopolitical conditions, overseas industrial parks face various challenges, including financial constraints, underdeveloped agglomeration economies, discrepancies between policy implementation and practical conditions, insufficient planning capacity, and uneven resource allocation [14,15]. These issues have resulted in the uneven distribution of parks, disconnecting between industrial and urban functions, and a lack of unified planning standards, which collectively hinder the high-quality advance of the BRI. At present, the development of overseas parks urgently requires more precise quantitative analysis to establish scientifically sound development models that can provide effective guidance for their construction and operation.

Industrial parks serve as growth poles that drive urban development and transformation [16]. During rapid urbanization and industrialization, there is a dynamic interplay between the development of industrial parks and surrounding land-use conditions [17]. The development of industrial parks not only drives broader land-use changes, but evolving land use patterns also have significant long-term implications for the future development of such parks. A study from Argentina examined the environmental and social impacts of converting rural land into urban and industrial uses. Barenboim and Zamler (2017) found that changes in land use and infrastructure can significantly alter the growth trajectory of industrial areas. Taking the Rosario Business Park as an example, land use transformation spurred economic activity within the park and led to infrastructure improvements in surrounding areas. However, these developments also caused adverse environmental and social effects, such as increased noise pollution and infrastructure stress, contributing to continuous dynamic changes in nearby land use [18].

Early studies (e.g. Fischer & Sun, 2001) had already demonstrated a direct relationship between economic development and land use change [19]. Scholars have employed various methods to simulate land use change and explore its underlying driving forces [20,21,22]. Alcamo et al. (2011) used the Land SHIFT model and existing datasets to analyze the driving forces of land use change across Africa and demonstrated the model’s ability to formulate consistent land-use scenarios at the continental scale [23]. Chen et al. (2018) applied the CA-Markov model to accurately predict land-use patterns in Zhanle County, China, from 2025 to 2036, proving that the integration of GIS and remote sensing technologies is an effective forecasting approach. With continued advances in land use research, methodological integration has emerged as a key trend. Wang et al. (2021) combined the logistic regression–cellular automata–Markov chain (LR-CA-Markov) model with the FLUS model to simulate land use change in arid regions [24]. A growing body of research has confirmed that continuous improvements in land use modeling have significantly enhanced the ability to manage regional development and provide scientifically grounded and reliable guidance.

However, due to the limited availability, low accuracy, and difficulty of accessing fundamental data in some overseas regions, academic research on China’s overseas industrial parks under the BRI has largely focused on qualitative analyses, such as park development, institutional environments, and industrial clustering [25]. For instance, topics such as site selection, policy mobility, policy transfer [26], and comparisons between domestic and overseas parks have been examined, but there remains a general lack of quantitative analysis of land use change and limited investigation into the driving mechanisms behind such changes. Moreover, given that the BRI spans Asia, Europe, and Africa, the geographical conditions of cities hosting the parks vary widely, resulting in distinct land-use development trajectories across regions. Consequently, a unified development model cannot serve as a scientifically valid basis for guiding the operation of all parks. The lack of quantitative research has left overseas park development models in an exploratory phase, and their adaptability remains largely unverified due to insufficient studies on land use dynamics.

This study investigates the driving forces of land use change to conduct an in-depth analysis of regions where China’s overseas parks are located along the BRI, with the aim of identifying key drivers and summarizing regional advantages. A critical question arises: can these findings provide precise and scientifically grounded recommendations for industrial park development in regions where open-source data are scarce and land use studies are insufficient? This question undoubtedly presents a compelling subject for further research.

The core objective of this study is to select three representative regions across different continents covered by the BRI and, using remote sensing data and the PLUS model, reveal the characteristics of land use change in the areas where overseas parks are located between 2000 and 2020. This study further analyzes the contribution of various driving forces and calculates the land-use transfer matrix over the 20-year period. By comparatively analyzing the land use patterns and driving forces in overseas parks across different geographical regions, this study aims to identify region-specific strengths and challenges, thereby offering valuable insights for the future planning and development of industrial parks.

The theoretical framework of this study is structured around three core components: land use change theory, regional development theory, and spatial planning theory. Land use change theory emphasizes the dynamic and heterogeneous nature of land use, asserting that transformations are driven by an interplay of natural conditions, socioeconomic development, and institutional policies. This study applies this theoretical perspective to analyze the key driving forces behind land-use type changes across different geographic regions, including population growth, economic activity, infrastructure development, and environmental conditions. Regional development theory—particularly the growth pole theory and theories of regional cooperation—provides a crucial basis for exploring how industrial parks drive regional economic development. Overseas industrial parks function not only as economic growth poles but also as platforms for international economic cooperation and strategies for integrated regional development. From this theoretical perspective, this study analyzes the impacts of BRI-linked overseas parks on regional economies and land resource allocation. Spatial planning theory offers guidance for the spatial configuration and land use planning of industrial parks, emphasizing the role of rational planning and policy tools in enhancing land use efficiency and sustainability. This study applies spatial planning theory to optimize park layout, exploring how land use strategies can facilitate coordinated development between industrial parks, cities, and ports. A comprehensive theoretical framework is thus constructed to gain a deeper understanding of land use changes and their underlying drivers across BRI regions, providing both theoretical and practical support for the planning and development of overseas parks.

This study continues and deepens the spatial analytical tradition within geography by innovatively applying GIS technology and the PLUS model to quantitatively simulate and predict regional land-use dynamics, thereby overcoming the limitations of traditional qualitative and descriptive geographic methods. It explicitly challenges conventional descriptive and geopolitical narrative frameworks by employing empirical data and quantitative analysis to better elucidate the complex economic–geographical phenomena arising from the development of overseas industrial parks under the BRI.

This research contributes both significant theoretical and practical implications. Theoretically, most existing studies on China’s overseas parks along the BRI are qualitative in nature, lacking quantitative and spatial analytical approaches. By integrating remote sensing technology and the PLUS model, this study systematically reveals land-use characteristics in various geographic regions and provides an in-depth analysis of the associated driving mechanisms. Spatiotemporal simulations are employed to quantify the impacts of natural, economic, demographic, and transportation factors on land use change. These are then used to compare land-use patterns and driving forces across industrial parks in Africa, Asia, and Europe, offering a scientific basis for future regional development strategies. Practically, this study offers quantitative data support for the planning, construction, and management of overseas industrial parks, helping to avoid resource waste and redundant competition while improving construction efficiency. At the same time, the research findings will contribute to optimizing the spatial layout and functional positioning of China’s overseas industrial parks, support the development of more scientifically grounded and context-appropriate models for different parks, promote the high-quality advancement of the BRI, and provide empirical evidence to enhance the standardization and normalization of park planning, thereby facilitating the high-quality development of overseas parks.

2. Research Area and Data

2.1. Study Area

To address the research gap in land use changes associated with overseas industrial parks, this study selects three representative case areas: the Saysettha Development Zone in Vientiane (Asia), the China-Belarus Industrial Park in Minsk (Europe), and the Djibouti International Free Trade Zone (Africa). Minsk, the capital and economic hub of Belarus, holds a strategic position within the Eurasian Economic Union and has significantly benefited from initiatives such as the China-Belarus Industrial Park, which aims to promote advanced manufacturing and high-tech industries. Vientiane, the capital of Laos, has maintained strong economic ties with China and serves as the core of the China-Laos Economic Corridor initiative, which seeks to accelerate infrastructure and industrial cooperation under the China–ASEAN framework. Djibouti, as a vital gateway connecting Africa, the Middle East, and Europe, holds significant geopolitical and economic importance. As a strategic port city at the core of China’s maritime Silk Road, it has experienced accelerated infrastructure development and economic integration.

As major economic initiatives, overseas industrial park developments directly influence urban land-use patterns, drive regional infrastructure investment, reshape land use intensity, and accelerate urbanization processes. For example, in Minsk, the high-tech industries within the China-Belarus Industrial Park have stimulated the development of surrounding commercial and residential land, increasing land use intensity. In Djibouti, the integration of the industrial park with port infrastructure has enhanced logistics and infrastructure capacity, significantly facilitating the conversion of unused land into construction land. In Vientiane, the construction of the industrial park has led to moderate urban expansion and increased land use intensity, with infrastructure improvements and industrial agglomeration promoting urbanization.

The selection of study areas is based on the following four key criteria: (1) strategic location and economic representativeness, (2) consistency and comparability of development models, (3) diversity and representativeness of socioeconomic contexts, (4) and data availability and research feasibility. These three regions represent critical strategic nodes along the BRI: Minsk lies at the core of the Eurasian Economic Union; Vientiane is a key location along the China–ASEAN Economic Corridor; and Djibouti serves as a major maritime trade hub connecting Africa, the Middle East, and Asia. All three regions have adopted the “Port–Park–City” (PPC) development model, offering methodological consistency and reliability for cross-regional comparative analysis. The three areas represent different stages of regional development and governance structures, encompassing the diverse economic and political contexts of Europe, Asia, and Africa, thereby enabling a broad comparative analytical perspective. Each region possesses relatively complete and reliable data, ensuring both methodological feasibility and analytical depth for this study.

Vientiane, the capital of Laos, is situated on the northern bank of the Mekong River’s middle reaches in a valley plain in Southeast Asia (Figure 1). The city has relatively flat terrain, an elevation of 174 m, a total area of 3920 km², and a population of approximately 989,000. The city experiences a tropical and subtropical climate, characterized by year-round heat and humidity. The Saysettha Comprehensive Development Zone serves as a key node in China’s effort to promote the southward channel of the BRI, significantly enhancing economic cooperation between China and ASEAN. Minsk, located in central Belarus on the banks of the Svisloch River, a tributary of the upper Dnieper River, is the capital of Belarus and the country’s political, economic, scientific, and cultural center. It is also the most economically developed city in Belarus.

Minsk has a temperate continental humid climate, covering an area of approximately 349 km², with an elevation of 280.4 m, a population of around 2.01 million, and an annual average temperature of 7.8 °C (Figure 2). The China-Belarus Industrial Park is located at the core of the Eurasian Economic Union, providing strategic convenience for Chinese enterprises to access the European market. This industrial park is the largest one established by China in Belarus, located near Minsk, focusing on high-tech and manufacturing industries and aiming to attract high-end sectors such as electronics, machinery manufacturing, and biomedicine.

Djibouti City is located along the shores of the Gulf of Tadjoura, covering an area of 630 km² (Figure 3). It is the capital of Djibouti and one of the largest seaports in East Africa. The city has a population of approximately 620,000 and experiences a tropical desert climate, with July and August temperatures frequently exceeding 45 °C. Djibouti City has significant topographical variation, with its lowest point being Lake Assal at −155 m, the lowest point on the African continent, and its highest point being Mount Mousa Ali at 2020 m. The Djibouti International Free Trade Zone is located on the eastern Red Sea coast, controlling the entrance and exit of the Red Sea shipping routes. It serves as a vital hub for China’s trade routes in Africa and the Middle East. Positioned near the entrance to one of the busiest Red Sea shipping lanes globally, it spans over 48 square kilometers and is one of the largest free trade zones in Africa.

Additionally, all three parks are operated by the China Merchants Group and adopt the innovative PPC model, which emphasizes the integration of ports, parks, and cities to form a synergistic new economic ecosystem. Given the long-term operation of these three parks, the PPC model has already accumulated measurable results, which facilitates the control of the variables in land use development during this study and enables in-depth analysis and conclusions.

2.2. Data Source

Detailed information on data sources is provided in

Table 1. Data information and sources.

Data Name	Data Source
GlobeLand30 Land Cover Data	China National High
geoBoundaries Global Political Administrative Boundaries	GeoBoundaries by William & Mary GeoLab
Global Bathymetry DEM	Earth System Science Data Center
Slope	Derived from DEM data
Aspect	Derived from DEM data
Annual Precipitation Data from Global Weather Stations	CHIRPS Pentad
Modis_LST Land Surface Temperature	MOD11A1 Version 6
LandScan Global Population Dynamics Data	East View Cartographic
Global GDP	Global Electricity Consumption GDP Dataset
Long-Term NDVI Data	GIMMS NDVI3g V1
Global Night-time Light Time-Series Data	National Centers for Environmental Information (NCEI)
Railway Network	Open Street Map
Road Network	Open Street Map
River and Lake Distribution	Open Street Map

. The remote sensing land cover data from GlobeLand30 were uniformly processed using ArcGIS 10.7, with reclassification based on the “Land Use and Land Cover Classification System Standard” developed by the Chinese Academy of Sciences. The data used in this study include land cover, precipitation, temperature, population, GDP, normalized difference vegetation index (NDVI), night-time light data, railway network, road network, and river distribution (See

Table 2. Impact factors and treatment methods.

Driving Factors	Data Processing Methods and Descriptions
Land cover type	Classified using remote sensing imagery to generate categorical rasterdata, serving as baseline inputs for land-use change modeling.
Precipitation and temperature	Collected from meteorological station data and interpolated spatially(e.g., Kriging interpolation) to create continuous raster datasets.
Population and GDP	Derived from statistical yearbooks or socio-economic datasets based onadministrative units, and converted into raster data via spatialinterpolation or population density allocation methods (e.g., spatialweighted distribution)
Normalized Difference Vegetation Index	Calculated from remote sensing imagery for each year and standardizedto reflect regional vegetation coverage and ecological quality.
Nighttime lighting intensity	Extracted from nighttime remote sensing imagery, standardized intoraster data, representing regional economic activity intensity andurbanization level.
Railway and road network	Calculated using GIS to determine the nearest distance or kernel densityvalues from each raster cell to railways and roads, quantifyingtransportation accessibility
River distribution	Using hydrological data, calculated the distance from each raster cell tothe nearest river, reflecting the hydrological influence on land-use changes.
Elevation	Extracted elevation values from Digital Elevation Models (DEM) foreach raster cell, indicating topographic constraints on land use.

). These datasets were collected for the cities hosting the parks at three time points: 2000, 2010, and 2020.

2.3. Data Processing

All the aforementioned data were uniformly processed as raster data in ArcGIS 10.7. In accordance with the “Land Use and Land Cover Classification System Standard” of the Chinese Academy of Sciences, each temporal dataset was reclassified in ArcGIS. The land use types across the three study areas were ultimately categorized into six major classes: cultivated land, forest land, grassland, water area, construction land, and unused land. The pixel distribution for each land use category during data processing is shown in

Table 3. Number of spatial pixels for six land use types.

Land Use Type	Year	Cultivated Land	Forest Land	Grassland	Water Area	Construction Land	Unused Land
Minsk	2000	22,533,309	19,985,478	201,859	773,931	981,878	22,533,309
	2010	22,204,372	20,091,775	223,191	798,153	1,158,964	22,204,372
	2020	22,198,379	19,868,088	230,454	810,288	1,369,246	22,198,379
Djibouti	2000	7701	681	152,893	2532	8667	690,546
	2010	8739	676	152,341	2488	13,065	685,711
	2020	11,136	578	148,175	2475	26,725	673,931
Vientiane	2000	2,119,484	11,335,907	489	447,676	19,183	2,119,484
	2010	2,223,676	11,214,741	503	450,494	33,325	2,223,676
	2020	2,288,712	11,099,279	535	469,389	64,824	2,288,712

.

In this study, due to favorable climatic conditions in Minsk and Vientiane, land types typically classified as unused land—such as deserts, Gobi areas, saline–alkali lands, swamps, bare land, or rocky gravel zones—were virtually absent within their administrative boundaries. As a result, the category of unused land was not included in the land use classifications for these two regions.

3. Research Methods

3.1. PLUS Model

The research framework is shown in Figure 4. In this study, the PLUS model was used to simulate and predict land use changes [27]. The PLUS model integrates the Land Expansion Analysis Strategy (LEAS) and an improved multi-type Random Seed Cellular Automata (CARS) model, overcoming the limitations of traditional Cellular Automata (CA) models in transition rule mining and dynamic change simulation. The model consists of a land quantity prediction module based on the Markov model and a spatial distribution simulation module [28]. The quantity prediction module forecasts the future land use demand based on historical land use data, while the spatial distribution module allocates space by analyzing the expansion probability of current land use, combined with planning scenarios and neighborhood effects.

3.2. Driving Forces of Land Use Change

The driving forces of land use change refer to various factors influencing the transformation of land use types, with changes being shaped by intrinsic physical and chemical properties as well as external natural and socio-economic factors [29]. Driving forces typically encompass natural, socio-economic, and policy-related factors. For instance, topography is a fundamental determinant of land use spatial patterns, with plains primarily used for cropland, low-lying areas for wetlands, and mountainous regions for grassland. Climate conditions, such as temperature and water resource distribution, are critical driving factors influencing land use pattern changes in agro-pastoral transition zones. Socio-economic conditions play a significant role in the short-term transitions of land use types, while policy preferences affect the relationships between cropland, forest, and grassland [30]. In this study, the driving force analysis of land-use change aims to identify which factors play key roles in land use changes across different regions. Natural factors such as climate conditions and topographic features, as well as socio-economic factors like population growth, economic development levels, and infrastructure construction, are included in the analysis. When selecting driving factors, a comprehensive consideration of topography, temperature, water resources, soil physical and chemical properties, and socio-economic factors is required [31]. Based on this, 11 driving factors were selected for impact analysis in this study.

3.3. Field Weights

Domain weights represent the potential suitability or competitive ability of different land-use types during spatial expansion. The values range from 0 to 1, with larger values indicating greater difficulty in converting that land type into other types, signifying stronger expansion capability [32,33]. Domain weights are determined based on the dimensionless value of △TA:

$$w_i={\displaystyle \frac{∆{TA}_i-∆{TA}_{min}}{{∆TA}_{max}-{∆TA}_{min}}}$$

(1)

In the formula, $∆{TA}_i$ represents the change in expansion area of each land type; $∆{\mathrm{T}\mathrm{A}}_{\mathrm{m}\mathrm{i}\mathrm{n}}$ is the minimum value of the expansion area change; and ${∆TA}_{max}$ is the maximum value of the expansion area change [34].

3.4. Transfer Matrix

The land use transition matrix, as a tool in land-use change studies, is typically used to describe the conversion relationships and dynamic changes between different land-use types. It is characterized by its intuitiveness, quantifiability, and flexibility. In the matrix, the rows and columns represent land use types at different times, with each matrix element indicating the area or proportion of one land use type converted to another over a specific period [35].

The specific rules are illustrated in

Table 4. Transfer matrix.

		Land Use Type
		A	B	C	D	E	F
Land Use type	A	1	1	1	1	1	1
	B	1	1	1	1	1	1
	C	1	1	1	1	1	1
	D	0	0	0	1	0	0
	E	0	0	0	0	1	0
	F	1	1	1	1	1	1

: when a land type can be converted to another, the corresponding matrix value is 1; if conversion is not allowed, the value is 0 [36].

To identify the intensity and direction of land use changes across different categories, land use transition matrices were constructed for two time periods: 2000–2010 and 2010–2020. A transition is defined as the change in land use of any given parcel from category i to category j. The data source is the GlobeLand30 land cover classification product with a spatial resolution of 30 m, which provides high spatial consistency and classification accuracy.

3.5. Accuracy Verification

The Kappa coefficient is a method used to evaluate the consistency of classification results. It is widely applied in research fields requiring accuracy assessment, such as geographic science and medical diagnostics [37]. Its calculation formula is as follows:

$$Kappa={\displaystyle \frac{P_a-P_b}{1-P_b}}$$

(2)

In the formula, Pa represents the observed agreement, which is the actual agreement rate of the classifier and indicates the proportion of correctly classified grids in the simulation [38]. Pb represents the expected agreement, which is the expected agreement rate under completely random classification and is used to estimate the proportion of correctly classified grids. The Kappa coefficient ranges from 0 to 1, where 1 indicates perfect agreement and 0 indicates no difference from random agreement [39]. The simulation accuracy classification follows the standards proposed by Pontius and Millones (2011) and Landis and Koch (1977) [40,41], as shown in

Table 5. Classification and evaluation standards for Kappa coefficients.

Kappa coefficient	<0.00	0.00~0.20	0.21~0.40	0.41~0.60	0.61~0.80	0.81~1.00
Level	Poor	Slight	Fair	Moderate	Substantial	Almost Perfect

.

Using land use data from 2000 and 2010 as the basis, the Markov model was used to simulate the land use situation in 2020. The Kappa coefficient was then applied to compare the simulated results with the actual 2020 data. Validation results for the three cities are shown in

Table 6. Confusion matrix for actual and predicted land use patterns in Minsk (2020).

Land Use Forecast for 2020
	Land Use Types	Cultivated Land	Forest Land	Grassland	Water Area	Construction Land	Total
Actual Land Use Situation In 2020	Cultivated Land	21,663,031	37,068	3034	18,751	159,667	21,881,551
	Forest Land	325,156	19,828,652	4214	3114	32,364	20,193,500
	Grassland	21,104	36	223,123	28	1	244,292
	Water Area	31,170	1747	77	788,321	617	821,932
	Construction Land	157,201	38	0	52	1,176,556	1,333,847
	Total	22,197,662	19,867,541	230448	810,266	1,369,205	44,475,122

Based on the analysis of the confusion matrix and the formula, the Kappa coefficient was calculated to be 0.967, with an overall accuracy of 0.982. Comparing this to the table, it is evident that Kappa > 0.8, indicating a good simulation performance.

,

Table 7. Confusion matrix for actual and predicted land use patterns in Djibouti (2020).

Land Use Forecast for 2020
	Land Use Types	Cultivated Land	Forest Land	Grassland	Water Area	Construction Land	Unused Land	Total
Actual Land Use Situation In 2020	Cultivated Land	473	0	23	0	2	20	518
	Forest Land	0	40	0	0	6	0	46
	Grassland	66	0	7336	1	239	58	7700
	Water Area	0	0	0	139	1	1	141
	Construction Land	2	0	7	2	738	98	847
	Unused Land	43	0	181	0	304	33,573	34,101
	Total	584	40	7547	142	1290	33,750	43,353

Based on the analysis of the confusion matrix and the formula, the Kappa coefficient was calculated to be 0.932, with an overall accuracy of 0.976. Comparing this to the table, it is evident that Kappa > 0.8, indicating a good simulation performance.

and

Table 8. Confusion matrix for actual and predicted land use patterns in Vientiane (2020).

Land Use Forecast for 2020
	Land Use Types	Cultivated Land	Forest Land	Grassland	Water Area	Construction Land	Total
Actual Land Use Situation In 2020	Cultivated Land	2,067,781	225,842	8	2825	24,166	2,320,622
	Forest Land	219,435	10,857,548	151	15,987	6409	11,099,530
	Grassland	0	149	368	0	0	517
	Water Area	1215	1679	2	450,562	13	453,471
	Construction Land	206	13,709	6	0	34,235	48,156
	Total	2,288,637	11,098,927	535	469,374	64,823	13,922,296

Based on the analysis of the confusion matrix and the formula, the Kappa coefficient was calculated to be 0.891, with an overall accuracy of 0.963. Comparing this to the table, it is evident that Kappa > 0.8, indicating a good simulation performance.

below:

4. Result

4.1. Comparative Analysis of Land Use Change Driving Forces Across Different Regions over 20 Years

This study analyzed the land use changes and their driving forces in Minsk, Vientiane, and Djibouti during the periods 2000–2010 and 2010–2020. The results are as follows:

As shown in Figure 5, the analysis of driving factors of land use change in Minsk reveals that population and road infrastructure are major driving forces influencing most land use types, with particularly high contributions to changes in cultivated land, forest land, and construction land.

Specifically, changes in grassland and water areas are primarily driven by natural environmental factors such as precipitation, elevation, and temperature, whereas changes in construction land are more strongly influenced by socioeconomic factors such as population, GDP, and transportation infrastructure. The contribution of various driving forces to land use change in Minsk exhibits dynamic variations over time. For instance, between 2010 and 2020, the influence of population on cultivated land increased significantly, while economic factors such as GDP showed a notably stronger impact on changes in construction land and water areas during the same period.

As shown in Figure 6, in Djibouti City, both population and distance to railways significantly influenced all six land use types, indicating that population density and infrastructure development played a critical role in land use changes.

The primary drivers of changes in cultivated land and unused land remained stable across both time periods (2000–2010 and 2010–2020). In contrast, changes in water areas were mainly influenced by precipitation, with the impact of this natural factor becoming notably stronger during 2010–2020. Moreover, the influence of driving factors exhibited temporal dynamics. For instance, construction land was primarily affected by railway proximity during 2000–2010, whereas in 2010–2020, the influence of population and distance to roads increased significantly compared to that of railways.

As shown in Figure 7, analysis of the driving factors in Vientiane indicates that topography and distance to roads are key drivers for most land use types, with particularly high contributions to changes in cultivated land, forest land, and construction land.

Considerable differences exist among the driving forces of various land-use types: grassland and water areas are primarily influenced by environmental factors such as river proximity and precipitation, while construction land is more strongly associated with socioeconomic factors like population and GDP. Temporal variation is also observed in the contribution of driving forces. For example, between 2000 and 2010, elevation and population had increasing impacts on cultivated land, whereas from 2010 to 2020, economic factors such as GDP showed stronger influence on both water areas and construction land.

The correlation analysis of driving factor contributions for different land-use types between 2000 and 2010 and 2010 and 2020 is summarized in

Table 9. Correlation analysis of land use driving force contribution.

Minsk		Djibouti		Vientiane
Land Use Types	Correlation Coefficient	Land Use Types	Correlation Coefficient	Land Use Types	Correlation Coefficient
cultivated land	0.932	cultivated land	0.956	cultivated land	0.931
forest	0.967	forest	0.942	forest	0.907
grassland	0.952	grassland	0.932	grassland	0.784
water area	0.789	water area	0.81	water area	0.792
construction land	0.981	construction land	0.856	construction land	0.991
unused land	NONE	unused land	0.951	unused land	NONE

.

The results indicate a generally high consistency in driving force contributions for the same land-use type across the two time periods. Among the three cities, the correlation of driving forces for construction land change over the 20-year period was the strongest, approaching 1. This suggests that the primary driving forces affecting construction land remained largely consistent over time in both Minsk and Vientiane. In contrast, the correlations for grassland and water area drivers were relatively lower, particularly in Vientiane.

4.2. Analysis of Land Use Transfer in Different Regions Under the Time Dimension

4.2.1. Land Use Change in Minsk

In Minsk, land use changes were relatively evenly distributed, while changes were more concentrated in the central urban area (Figure 8). The transition from forest land to cultivated land, and subsequently to construction land, demonstrated a spatially uniform distribution across Minsk.

Over the 20-year period, construction land in Minsk increased from 883.96 km² to 1232.32 km², representing a 39.53% growth. This expansion absorbed 300.0474 km² from cultivated land and 48.0114 km² from forest land. By 2020, construction land accounted for 3.08% of the total area. The transition from cultivated land to construction land reflects a clear trend of urbanization and reduction in farmland over the past two decades, driven by urban expansion and infrastructure development in Minsk (See

Table 10. Land use transition matrix in Minsk from 2000 to 2020 (km²).

Land Use Types in 2000	Land Use Types in 2020
	Grassland	Cultivated Land	Construction Land	Forest Land	Water Area	Total
Grassland	181.67	0.00	0.00	0.00	0.00	181.67
Cultivated Land	18.16	19,697.91	300.05	232.28	31.58	20,279.98
Construction Land	0.00	0.00	883.69	0.00	0.00	883.69
Forest Land	7.52	278.87	48.01	17,647.10	5.44	17,986.93
Water Area	0.05	1.77	0.57	1.90	692.25	696.54
Total	207.41	19,978.54	1232.32	17,881.28	729.26	40,028.81

).

Cultivated land, forest land, and water areas remained relatively stable during the 20-year period (Figure 9). Despite gradual conversion into cultivated and construction land, forest land maintained a high degree of stability, with 17,647.0983 km² remaining unchanged, suggesting the potential impact of ecological restoration efforts or forest policy interventions.

4.2.2. Land Use Change in Vientiane

In Vientiane, land use changes were primarily concentrated in the southern region, particularly around the edges of cultivated land (See Figure 10). As of 2020, construction land in Vientiane measured 58.34 km², accounting for only 0.4% of the city’s total area. The proportion of land converted to construction land over the past two decades was just 0.14%, indicating a relatively low level of urbanization and substantial potential for future development.

Cultivated land in Vientiane experienced notable expansion over the past 20 years. Some of this land was subsequently converted, primarily into forest land and construction land. Of the 1696.69 km² of cultivated land recorded in 2000, a large portion remained unchanged by 2020. However, 31.31 km² was converted to construction land, 175.3884 km² to forest land, and 4.14 km² to water area. Details shown in

Table 11. Land use transition matrix of Vientiane from 2000 to 2020 (Km²).

Land Use Types in 2000	Land Use Types in 2020
	Grassland	Cultivated Land	Construction Land	Forest Land	Water Area	Total
Grassland	0.44	0.00	0.00	0.00	0.00	0.44
Cultivated Land	0.01	1696.69	31.31	175.39	4.14	1907.54
Construction Land	0.00	0.00	17.26	0.00	0.00	17.26
Forest Land	0.03	361.74	9.76	9813.83	16.96	10,202.32
Water Area	0.00	1.41	0.01	0.13	401.35	402.91
total	0.48	2059.84	58.34	9989.35	422.45	12,530.47

and Figure 11.

4.2.3. Land Use Change in Djibouti

In Djibouti City, land use changes were concentrated in the northeastern coastal port zone (Figure 12). The port’s expansion played a significant role, and land use changes were clearly oriented around the expansion of existing construction land in the northeast. Over the past two decades, Djibouti City has undergone substantial urbanization, with construction land increasing from 7.80 km² in 2000 to 24.05 km² in 2020, marking a 208.33% growth.

Compared with the other two study areas, the arid climate of Djibouti has resulted in extensive areas of unused land, with cultivated and forest land accounting for only a small proportion. Grassland and unused land were the main sources of construction land expansion. Specifically, 6.35 km² of grassland and 9.75 km² of unused land were converted to construction land, making them the primary contributors to construction land growth. Further details are provided in

Table 12. Land use transition matrix of Djibouti City from 2000 to 2020 (Km²).

Land Use Types in 2000	Land Use Types in 2020
	Grassland	Cultivated Land	Construction Land	Forest Land	Water Area	Unused Land	Total
Grassland	128.12	1.70	6.35	0.01	0.01	1.41	137.60
Cultivated Land	0.02	6.86	0.04	0.00	0.00	0.01	6.93
Construction Land	0.00	0.00	7.79	0.00	0.01	0.00	7.80
Forest Land	0.02	0.01	0.09	0.46	0.00	0.03	0.61
Water Area	0.01	0.00	0.03	0.00	2.18	0.06	2.28
Unused Land	5.19	1.45	9.75	0.05	0.03	605.04	621.49
Total	133.36	10.02	24.05	0.52	2.23	606.54	776.72

and Figure 13.

5. Discussion and Conclusions

5.1. Comparative Investigation on Discrepancies and Similarities in Land Use Across Diverse Regions

5.1.1. Key Drivers of Land Use in Overseas Parks Regions: Population, Traffic, and Topography

According to the 2020 Global Ecological Environment Remote Sensing Monitoring Report published by China’s Ministry of Science and Technology [42], global urban land expanded by 117.49% from 2000 to 2020. Asia experienced the highest expansion (161.67%), followed by Africa (116.58%) and North America (111.21%). Against this global backdrop, Minsk’s 39.53% increase in built-up land appears moderate, likely shaped by planning controls and constrained industrial growth. In contrast, Djibouti’s 208% increase far exceeds the global average, indicating a port-driven expansion model, which, while growth-oriented, may pose challenges in terms of land use efficiency and environmental sustainability. By analyzing land use changes in Vientiane, Minsk, and Djibouti, this study reveals the land use patterns and driving forces of land use change in regions hosting Chinese overseas industrial parks under different geographic and economic conditions.

This study reveals that population density, transportation accessibility, and topographic conditions serve as the core driving forces of land use change. These factors fundamentally determine future land transformation trajectories within each region and play a crucial role in promoting high-quality development and efficient resource utilization. Under the combined influence of these three variables, land use development across different overseas industrial park regions exhibits a set of common patterns. This can be explained by the following reasons:

The rationale lies in the fact that population density reflects the concentration of labor and market demand, which is closely associated with the scale and speed of construction land expansion, serving as a fundamental basis for land development. The transportation infrastructure influences regional accessibility and connectivity. A well-developed transport network significantly enhances land value, facilitates industrial agglomeration, and improves the efficiency of park development. Topographic features act as critical natural constraints that affect the developability and cost of land, while also determining the feasibility and rationality of land use planning.

Therefore, the planning and development of overseas industrial parks must comprehensively incorporate these three categories of driving forces. By leveraging data analysis and spatial optimization, it is essential to formulate scientifically sound and context-specific land use strategies to achieve high-quality park construction and long-term sustainable development.

5.1.2. Natural Conditions, Development Strategies and Urbanization Variations in the Three Regions

A comparative analysis of land use from 2000 to 2020 in Vientiane, Minsk, and Djibouti reveals that differing natural conditions and development strategies have led to varied urbanization patterns. Vientiane and Minsk exhibited relatively stable driving force contributions over the 20-year period, whereas grassland and unused land in Djibouti showed greater temporal dynamics, resulting in differing land transition models across the three cities: Minsk and Vientiane generally followed a sequential transformation from forest land → grassland and cultivated land → construction land, while Djibouti displayed a transition pattern from unused land → grassland and cultivated land → construction land. Minsk and Vientiane generally followed a sequential transformation from forest land → grassland and cultivated land → construction land, while Djibouti displayed a transition pattern from unused land → grassland and cultivated land → construction land.

Although all three cities have undergone significant urbanization, the pace of expansion varied: Vientiane experienced relatively slow construction land growth, whereas Djibouti’s urban expansion was more rapid. In Djibouti, large areas of grassland and unused land were converted into construction land. However, the harsh climatic and geographic conditions impose substantial pressure on natural land resources, and the vast presence of unused land underscores potential challenges for future land development. By contrast, Vientiane must focus on how to rationally develop and utilize forest land to achieve sustainable land management.

5.1.3. Decoding the Influence of Varied Climate Types on the Patterns of Land-Use Change

Compared to the other two cities, Djibouti has a typical tropical desert climate, with an average annual temperature of 37 °C, peak temperatures exceeding 45 °C, and scarce annual precipitation of only around 150 mm. These extreme climatic conditions result in relatively slow transitions between different land-use types. Analysis of land-use change data from 2000 to 2020 suggests that Djibouti’s land development trajectory is unlikely to deviate significantly over the next decade.

Due to minimal land type conversions, Djibouti’s land-use confusion matrix values were noticeably lower than those of the other two cities. Zeros in the matrix indicate that certain land use types did not undergo any change. Effectively developing and utilizing unused land has thus emerged as a critical issue for Djibouti’s future development.

5.1.4. Comparative Analysis of Land Use Change Across Study Areas

Minsk primarily exhibited land use transitions from forest and cultivated land to construction land, driven mainly by population growth and economic factors. This reflects steady urban expansion under favorable geographic and infrastructural conditions. Vientiane experienced moderate urban expansion, largely influenced by infrastructure development and environmental constraints, with cultivated land being converted to construction land at a relatively slower pace. In contrast, Djibouti underwent rapid urbanization, mainly through the conversion of large areas of unused land and grassland into construction land. This was driven by population growth and strategic infrastructure development but constrained by harsh climatic conditions and limited natural resources.

A comparative analysis of Minsk, Vientiane, and Djibouti reveals distinct trends in land use change across the three cities. Djibouti exhibited the fastest rate of urban land expansion (39.5%), largely due to rapid infrastructure and port development under the BRI and substantial inflows of foreign direct investment. Minsk showed a stable and incremental land-use change pattern, influenced by balanced urban planning, advanced infrastructure networks, and industrial park-driven economic strategies. Vientiane’s urban expansion was more moderate, reflecting gradual infrastructure improvements and a slower economic development pace due to topographical constraints and limited population growth. Despite these differences, all three regions shared a common trend of significant construction land expansion, driven by intensified economic interactions under the BRI.

Our analysis shows that the driving forces behind land use change vary substantially across Minsk, Vientiane, and Djibouti. Population density had an especially prominent influence in Djibouti, where rapid population growth and highly concentrated urban land demand accelerated conversion to construction land. In contrast, Vientiane’s lower population density limited the short-term demand for land conversion, resulting in a slower rate of urban expansion. In terms of transportation infrastructure, Minsk benefits from a mature and well-developed transport network, which facilitates gradual and stable urban land expansion. Djibouti’s rapid development of infrastructure—particularly ports and railways—has significantly improved regional connectivity and accelerated changes in land use types. Infrastructure improvements in Vientiane have been more moderate, promoting incremental urban land expansion under geographic constraints. In addition, economic factors vary considerably: Minsk’s industrial park-centered development strategy has significantly promoted the intensive use of urban land, whereas Djibouti’s land use dynamics are primarily driven by port-based economic development and foreign investment. These differences underscore the necessity of formulating targeted regional policies and development strategies.

The findings on land-use change offer important insights when linked to broader socioeconomic and political contexts. For instance, the rapid expansion of construction land in Djibouti is closely associated with substantial foreign direct investment, which has increased economic dependency, potentially exacerbated socioeconomic inequality, and influenced local governance structures. In Minsk, the stable and incremental pattern of land use change suggests the presence of effective governance and balanced economic growth, reflecting deliberate spatial planning and industrial policy. In Vientiane, moderate land expansion constrained by topography highlights specific socioeconomic challenges, including limited job growth and restricted economic diversification. These cases illustrate how physical land-use dynamics are closely intertwined with economic transitions, social change, governance capacity, and policy decisions. Integrating socioeconomic and political dimensions greatly enriches our understanding of observed land use outcomes and highlights key directions for future research and policymaking.

5.2. Development Recommendations for the Three Overseas Parks Based on Driving Force Analysis

In Minsk, a city with a population of approximately two million, the road network is identified as a key factor influencing land use patterns. Future development of industrial parks should be closely aligned with the regional transportation system. Special attention must be paid to the construction of the transportation infrastructure within the parks to improve accessibility, enhance operational efficiency, and promote coordinated regional development.

Topography and distance to roads are the key factors influencing land use types in Vientiane. The future development of industrial parks in Vientiane should fully leverage the potential of areas surrounding existing urban development in the southern part of the city, while adopting a scientifically grounded approach to planning around the regional transportation network. Emphasis should be placed on improving the transportation infrastructure within industrial parks to enhance accessibility and foster integrated development with surrounding areas.

In Djibouti City, population and proximity to railways are particularly significant driving forces influencing land use change. Among them, population density and infrastructure development have played a critical role in shaping land transformation. The port demonstrates a strong spillover effect, serving as a key growth pole. Therefore, the planning and construction of overseas industrial parks should be anchored around existing construction land, with careful attention to integration and coordination with the port-centered development zone, to achieve efficient regional resource allocation and high-quality land use development.

Given the rapid growth of construction land in Djibouti, it is recommended that future industrial park expansion or site selection prioritize infrastructure carrying capacity planning, ecological risk mitigation, and high-quality resource management, to prevent the environmental and governance challenges associated with uncontrolled expansion. In Minsk, where land use change exhibits a stable and incremental growth pattern, future development strategies should focus on enhancing industrial clustering efficiency and strengthening the integration between industrial park development and local spatial planning policies, to maintain a trajectory of high-quality growth. In Vientiane, where land-use change is relatively moderate due to topographical constraints, it is particularly important that future development strategies for industrial parks address geographic and terrain limitations by adopting adaptive infrastructure planning and environmentally friendly development approaches.

The distinct land-use change patterns observed in Minsk, Vientiane, and Djibouti underscore the need for tailored planning strategies for each industrial park. Djibouti’s rapid urban expansion suggests an urgent need for sustainable land management practices in light of its environmental constraints. By contrast, the relative stability of land use in Minsk and Vientiane indicates that strategic efforts should focus on enhancing infrastructure and optimizing urban–rural integration.

5.3. Limitations

Due to significant differences in natural conditions, economic structures, and infrastructure development along the BRI corridor, development models for industrial parks must be tailored to local contexts. Therefore, site selection and development strategies for overseas parks should be grounded in region-specific analyses and supported by context-sensitive theoretical frameworks. However, due to limitations in data availability, this study could only utilize land use data from 2000, 2010, and 2020, making it difficult to capture year-by-year changes with high temporal accuracy. In the future, the integration of higher-resolution datasets could improve model validation and enable a more detailed assessment of long-term land use dynamics.

During the research process, significant deficiencies in data infrastructure across BRI regions were identified. Many regions lack diverse and reliable datasets necessary for academic research. For example, two-dimensional spatial data often lack the precision needed for quantitative analysis at the park scale. This issue directly limits the number and depth of case studies. Under such data scarcity, a key challenge for future research is how to leverage the existing datasets to enhance the scientific validity and practical feasibility of overseas park development.

In addition, overseas industrial parks are generally small in scale and may have limited short-term impact on host city economies. Under the current constraints of limited data precision in BRI regions, conducting precise quantitative analyses of industrial parks remains challenging, and comprehensive insights through direct quantitative observation are still difficult to achieve. However, as a major force in the global economy, China possesses the capacity to upgrade production technologies in low-income economies [43] Therefore, identifying development models that are appropriate for overseas parks holds critical practical relevance and urgency.

From the perspective of critical theory, the BRI and its associated overseas park developments have come under scrutiny, particularly regarding their potential implications for economic dependency, sovereignty risks, and spatial governance dynamics. Theories of economic colonialism and debt-trap diplomacy suggest that such overseas investments may increase financial obligations or foster dependence on external capital and expertise, thereby creating economic and political vulnerabilities for host countries. Spatial planning theory warns that large-scale overseas development may disrupt local planning priorities or compromise spatial governance autonomy. While this study primarily focuses on empirical land use analysis, acknowledging these critical theoretical perspectives provides important context, highlighting the broader geopolitical and economic layers that shape local land use outcomes. Future studies should systematically incorporate these critical lenses to enable a more comprehensive evaluation of international industrial park investments.

As more Chinese overseas parks enter the construction phase, future research should take advantage of newly available datasets to explore the interactions between parks and regional development at more granular spatial scales. Development strategies should also be tailored to local development needs to increase the embeddedness of local enterprises within the global network of Chinese overseas parks and promote diversified development paths that avoid the limitations of one-size-fits-all policies across varying environmental and economic contexts.