Dynamic Changes and Prediction of Land Use Driven by Socioeconomic Activities in Bazhong City, Southwest China (2004–2024)

Abstract

Land use systems are closely coupled with socioeconomic activities. To explore the interactions between land use and socioeconomic activities in Bazhong City, clarify the characteristics, drivers, and future trends of land use change, and provide scientific support for optimizing regional land resource allocation, ecological conservation, and food security, this study analyzes land cover data from 2004 to 2024, identifies economic drivers via principal component analysis, and predicts future land use trends for 2025, 2030, and 2035 using the GM(1,1) model. The results indicate the following: (1) Cropland decreased by 1338.69 km², while forest increased by 1304.88 km², with the largest area of mutual conversion occurring between these two types. (2) The comprehensive index of land use exhibited a fluctuating decline. The quality and continuity of cropland decreased, while the expansion of forest increased ecosystem services. (3) Principal component analysis identified the Comprehensive Economic Development and Urbanization Factor (e.g., GDP, urbanization rate, etc.) as the long-term core driver, with the land use driving system evolving through three stages. (4) Projections indicate that forest will increase, while cropland will decrease by 263.83 km². While the cropland is projected to remain above the planned target by 2035, the persistent downward trend will nonetheless pose a threat to food security. This study provides insights for harmonizing land use planning with socioeconomic progress and ecological conservation with cropland protection and may also serve as a reference for related decision-making in similar regions.

1. Introduction

With the rapid advancement of urbanization and the increasing prosperity of the socio-economy, human living standards have significantly improved [1]. However, this development model, which heavily relies on massive resource consumption, has also profoundly impacted regional ecosystems. In this process, land, as the core spatial carrier supporting various human activities, has seen the rationality and sustainability of its utilization patterns become a key factor influencing the quality of development. Driven by population growth and over-exploitation, issues such as deforestation [2], loss of arable land [3], land pollution, and ecological degradation [4] are becoming increasingly severe. These challenges endanger the ecological security of the region and limit the potential for high-quality socio-economic development. Land not only underpins economic activities including agriculture, as well as industry, mining, and urban construction but also sustains cultural heritage and human health [5].

Socioeconomic activities and land use changes demonstrate a bidirectional coupling relationship [6]. The evolution of socioeconomic systems drives transformations in land use by reshaping land demand patterns—economic growth and shifts in consumption elevate the need for urban construction land and commercial forestry [7]; rising urbanization rates lead to rural-to-urban migration, indirectly causing cropland abandonment and the restoration of natural forests in rural regions. Industrial restructuring, such as an expanding service sector, reduces dependence on industrial land while supporting forest conservation through ecotourism development [8]. Conversely, the optimization or imbalance of land use structures constrains socioeconomic development quality. Expanding forest cover enhances ecosystem service values, providing a foundation for green industries. However, excessive reduction in cropland threatens regional food security and hinders agricultural modernization [9]. Uncontrolled expansion of construction land may result in inefficient resource use and intensify human-land conflicts [10]. Therefore, how to scientifically coordinate human-land relationships and optimize land use patterns has become a crucial and urgent task for achieving sustainable development goals.

Against this backdrop, gaining a deep understanding regarding the intrinsic patterns of land use change is particularly important. Land use change is typically manifested across multiple dimensions, including quantity, structure, intensity, and spatial form, illustrating the dynamic interplay relationship among human activities with the natural environment [11]. As the stages of development evolve, driven by natural conditions and socio-economic factors, the patterns of land use have become increasingly complex and diverse, forming intricate interrelationships. Consequently, systematically revealing the patterns of land use change, predicting future trends, and constructing rational land use allocations have become core tasks in this field of research [4,12,13,14].

To achieve these objectives, land use change, its prediction, and simulation have long been focal points in international academic research. Numerous scholars have conducted extensive empirical and methodological explorations from different regional perspectives. For instance, some studies have employed multi-temporal land use classification methods, based on Landsat data from 1984 to 2016, to systematically track the expansion processes of 18 Canadian cities and provide multi-scale assessments of land change from pixels to broader census areas [15]. Other research combined long-term aerial imagery with field surveys, applying semi-automatic classification and principal component analysis to identify land cover changes, and further processed and analyzed the data using GIS layers and principal component analysis [16]. In a study of the Guder watershed, researchers used remote sensing and GIS, as well as the Google Earth Engine platform, to forecast land use scenarios for the years 2039 and 2057 using Random Forest and Neural Network models [17]. Further studies, based on CA-Markov models and Landsat imagery, conducted in-depth analyses of land use configurations and anthropogenic drivers in rapidly changing areas: one study constructed a predictive model for land use/cover in downtown Mamuju to forecast its 2034 land use scenario [18]; another study, through Landsat-based analysis, identified cropland expansion, unauthorized settlements, deforestation, charcoal production, and livestock grazing as the primary drivers of local land use change [19].

Research within the Chinese context has likewise seen substantial advancements, employing a diverse suite of models to simulate, and forecast land use dynamics. Early efforts, such as one focusing on the Baihe River Basin, applied the model of CA-Markov to simulate its 2020 land use pattern based on historical data from 1996 to 2008 [20]. Other work employed remote sensing techniques to evaluate land use dynamics in China’s arid northwest from 1990 to 2020, quantifying distribution patterns and predicting future scenarios to inform regional strategies [21]. The methodological toolkit has since expanded significantly. For instance, a study developed a coupled ANN-CA-Markov model for multi-period land use simulation in Dongguan City, projecting future patterns for 2025, 2030, and 2035 to offer actionable planning suggestions [22]. Furthermore, the PLUS model was utilized to forecast Harbin’s urban land use for 2029, offering essential guidance for the city’s future planning [23]. Further analyses of three cities which are along the ancient Shu Road at Jianmen Pass, based on the MCE-CA-Markov model, not only delineates the trajectory of land use development between 2012 and 2022, but also forecasts the scenario for 2027 to optimize resource allocation [24]. Additionally, research employing the FLUS model and grey relational analysis predicted and interpreted land use change trends for 2030 in Poyang Lake wetland, while also using the grey relational model to analyze the underlying driving factors [25]. The GM(1,1) model is extensively utilized in land use prediction. It enables diversified forecasts for multiple regions based on measured data from various time series. Examples include the prediction of land use status in Luzhou District of Shanxi Province for 2020 [26], the forecast of land cover area in the black soil region of Northeast China for 2025 [27], the projection of land ecological security status in Shenzhen from 2020 to 2025 [28], and the simulation of land use/cover change (LUCC) in the Pearl River Delta region [29]. The results of these predictions not only elucidate the changing trends in regional land use quantity and the evolution of ecological security but also offer scientific references for the protection of regional land ecosystems, comprehensive management, and the formulation of related policies.

Although domestic and international land use research has established a relatively mature methodological framework and accumulated a wealth of regional case studies, systematic research on land use change remains relatively scarce for prefecture-level cities in Western China with unique geographical and ecological characteristics, such as Bazhong City. The interactive relationship between land use and socio-economic development in these areas has not yet been fully elucidated. The land use data in Bazhong City lacks granularity but shows stable changes in indicators. The GM(1,1) model, designed for small sample sizes and sparse information systems, does not rely on assumptions about data distribution. This model effectively captures gradual trends, provides simple modeling, and ensures precise short-term forecasting. Hence, it is the preferred approach for making forecasts in Bazhong City.

To gain a comprehensive understanding of land use evolution patterns and their linkages with socioeconomic development in Bazhong City, this study examines the period from 2004 to 2024. It quantifies the characteristics of land use changes during this time, identifies the key socioeconomic drivers behind these changes, and forecasts future trajectories of major land use types using the GM(1,1) grey model in conjunction with local spatial planning objectives. This provides a scientific basis for coordinating regional land resource allocation, enhancing ecological protection, and fostering sustainable urban socioeconomic development. The innovations of this study are primarily evident in the following aspects: First, the utilization of 30 m resolution annual land cover data addresses the gap in long-term, high-resolution land use research for cities situated in the mountainous regions of western China. Second, the implementation of principal component analysis elucidates the evolutionary patterns of the driving system, thereby providing empirical evidence for comprehending the human-environment relationship in these western mountainous cities. Third, the application of the GM(1,1) model to forecast future land use trends serves as a valuable reference for decision-making regarding the optimization of regional land resource allocation and the balancing of ecological protection with agricultural production.

2. Materials and Methods

2.1. Study Area

Bazhong City, a prefecture-level city in Southwest China under Sichuan Province’s jurisdiction, is situated in the northeastern region of the province, with geographical coordinates spanning from 106°20′ to 107°20′ E and 31°15′ to 32°15′ N (Figure 1). It located in the transitional zone between the hilly and mountainous regions surrounding the Sichuan Basin, with Shaanxi Province on the north, Guangyuan City on the west, Nanchong City on the south, and Dazhou City on the east. The city covers a total area of 12,265.34 square kilometers and administers five county-level divisions: Tongjiang County, Nanjiang County, Pingchang County, Bazhou District, and Enyang District [30]. Bazhong is positioned at the geographical intersection of Chongqing, Chengdu, and Xi’an, serving as a pivotal link between the Chengdu-Chongqing Economic Zone and the Guanzhong-Tianshui Economic Zone. It performs a pivotal linking function in encouraging regional collaborative advancement.

Bazhong experiences a northern subtropical monsoon climate, featuring distinct seasons, a temperate climate, and abundant precipitation. The annual mean temperature ranges between 16.9 °C and 17.4 °C, while the mean annual precipitation is between 1113.0 mm and 1425.2 mm. Influenced by monsoon circulation, precipitation is unevenly distributed throughout the year, concentrated mainly in the summer and autumn, with relatively less rainfall in spring and winter. Droughts are common in spring, whereas floods frequently occur in summer. Despite these seasonal climatic challenges, the total water resources in Bazhong amount to 6.742 billion cubic meters, sufficient to meet production and domestic water needs. The area also enjoys favorable sunlight conditions, with annual sunshine hours ranging from 1368 to 1502, providing advantageous light and heat conditions for crop growth. The humid monsoon climate, combined with ample light, heat, and water resources, lays a solid foundation for the growth of forest vegetation. Ecological surveys indicate that the region is rich in plant diversity, with over 3000 recorded species, including more than 300 woody plants. Notable among these are rare and protected species such as Ginkgo biloba, Pinus bashanensis, and Fagus bashanensis, forming a germplasm resource pool of significant conservation value [31].

As of 2024, Bazhong has a resident population of 2.613 million, with an urbanization rate of 48.91%. The population is primarily concentrated in Bazhou District and Pingchang County. In recent years, Bazhong has made rapid progress in urban construction, transportation infrastructure, water conservancy projects, healthcare, and education, achieving notable advancements in both material and cultural progress. However, constrained by its geographical location and complex terrain, the city’s economic development remains relatively nascent. Compared to other prefecture-level divisions in Sichuan Province, Bazhong’s comprehensive competitiveness is relatively weak. There is an urgent requirement to stimulate intrinsic growth catalysts, hasten industrial reorganization and enhancement, and encourage the joint progression of the primary, secondary, and tertiary domains to thoroughly augment the region’s comprehensive might.

2.2. Data

To investigate how land use evolves dynamically in Bazhong City under the influence of socio-economic activities and predict future trends, this study systematically integrates geospatial and statistical data from multiple sources. The fundamental spatial framework was established by integrating administrative boundary data from the National Geospatial Information Sharing Platform with a 30-metre spatial resolution Digital Elevation Model (DEM) obtained from the Geospatial Data Cloud Platform.

For the core analysis of land use dynamics, the study utilized the 30 m spatial resolution annual land cover dataset for China (1985–2024) [32]. The developed dataset attains a total classification accuracy of over 85%, with both user’s accuracy and producer’s accuracy for individual land-use categories exceeding 80%. It delineates the spatiotemporal variation patterns observed between 2004 and 2024 and offers a basis for subsequent predictive analyses. In analyzing the magnitude of land use change, single-unit dynamics, and transition matrices, the primary land cover data utilized spans the years 2004, 2014, and 2024. The year 2004 marks a pivotal juncture for the full implementation of China’s Grain for Green Project, 2014 represents a significant milestone with the introduction of the New Urbanization Strategy and the accelerated advancement of ecological civilization construction, and 2024, as the most recent temporal cutoff in this study, effectively reflects land use patterns across distinct policy phases. The most representative areas of cropland and forest show no sudden increases or decreases (Figure 2). These three years are separated by a 10-year interval to capture long-term land use trends while minimizing the impact of short-term fluctuations.

Furthermore, Due to the lack of continuous data from 2004 to 2024 for other indicators such as employment distribution, land prices, and policy implementation intensity, this study selected 11 types of sample data for analysis (

Table 1. Socioeconomic indicator system.

Indicator Code	Indicator Name	Unit
X1	Total Population	10,000 people
X2	GDP	100 million yuan
X3	Proportion of Secondary Industry	%
X4	Proportion of Tertiary Industry	%
X5	Total Retail Sales of Consumer Goods	100 million yuan
X6	Urbanization Rate	%
X7	Highway Mileage in Service	km
X8	Gross Agricultural Output Value	100 million yuan
X9	Per Capita Grain Possession	kg
X10	Per Capita Net Income of Farmers	yuan
X11	Investment in Real Estate Development	100 million yuan

Source: Bazhong Statistical Yearbook.

), key indicators—including Total Population, Gross Domestic Product (GDP), Proportion of Secondary Industry, Proportion of Tertiary Industry, Total Retail Sales of Consumer Goods, Urbanization Rate, Highway Mileage in Service, Gross Agricultural Output Value, Per Capita Grain Possession, Per Capita Net Income of Farmers, Investment in Real Estate Development—from 2004 to 2024, were compiled from the official Bazhong Statistical Yearbook. The systematic integration of these diverse datasets has constructed a robust, multi-dimensional foundation for quantitatively analyzing land use changes, elucidating their underlying drivers, and developing predictive scenarios for Bazhong City.

2.3. Methods

2.3.1. Land Use Area Change Indicator

The variation in the area of different land types is a fundamental characteristic of land use dynamics. Analyzing the area distribution among various land use categories elucidates the overarching trends and structural evolution inherent to land use systems. The single land use dynamic degree quantifies the rate of change in area for a particular land category over a defined period within a given region. The expressions are as follows:

$$R_L={\displaystyle \frac{U_b-U_a}{U_a}}\times 100\%$$

(1)

$$K={\displaystyle \frac{U_b-U_a}{U_a}}\times {\displaystyle \frac{1}{T}}\times 100\%$$

(2)

where R_L denotes the aggregate alteration in area for a specified land use category; U_a and U_b denote the area at the initial and final stages of the study, respectively; K denotes the dynamic degree of a specific land use type; and T denotes the length of the study period [33,34].

2.3.2. Comprehensive Land Use Degree Index

The comprehensive land use index can provide a more systematic interpretation of changes in the degree of land use [35]. The degree and rate of land use change serve as indicators of land utilization patterns within a region, reflecting alterations in various land use types over time. The variation (∆L_b−a) and rate (R) of land use degree are used in this study to explore land use changes and their overall level, facilitating a precise grasp of the current land use situation and its change patterns. The expressions are as follows:

$$L=100\times {\displaystyle \sum _{i=1}^n{A}_i\times C_i}$$

(3)

$$\Delta L_{b-a}=L_b-L_a=100\times \left[{\displaystyle \sum _{i=1}^n\left(A_i\times C_{ib}-A_i\times C_{ia}\right)}\right]$$

(4)

$$R=\left[{\displaystyle \sum _{i=1}^n\left(A_i\times C_{ib}-A_i\times C_{ia}\right)}\right]÷{\displaystyle \sum _{i=1}^n\left(A_i\times C_{ia}\right)}$$

(5)

where L denotes the comprehensive index for the degree of land use; L_a and L_b denote the comprehensive index of regional land use at times a and b; A_i denotes the grading index corresponding to the i-th level of land use degree; C_i denotes the area proportion of the i-th level of land use degree; n is the count of land use grades; C_ia and C_ib denote the area proportion of land classified under the i-th degree grade at times a and b [36,37,38,39,40].

The extent and spatial scope of land resource utilization is comprehensively evaluated by this index, with a range of continuous values from 100 to 400 being encompassed [41]. The classification system sets out a range of four land use levels, based on how much the balance of natural land areas has been affected by social factors. [42]. Each land cover type corresponds to a specific grade index. Bare land and snow are given an index of 1; forest, shrub, grassland, water body and wetland correspond to 2; cropland to 3; and impervious surface to 4.

2.3.3. Land Use Transfer Matrix

The land use transfer matrix uses a two-dimensional matrix to analyze the exchange between different land use types over time, clearly showing how and where each land use type is converted [43,44,45]. The expression is given below:

$$S_{ij}=\left(\begin{array}{ccc}{S}_{11}& \dots & S_{1n}\\ ⋮& \ddots & ⋮\\ S_{n1}& \dots & S_{nn}\end{array}\right)$$

(6)

where S denotes the land area; n represents total land use category count; i indicates the land use type of origin; j signifies the destination type following the transition (both i and j are integer sequences ranging from 1 to n); S_ij denotes the total area flux transferred from land use type i to j over the specified time interval [46,47].

2.3.4. Principal Component Analysis

To address information redundancy and multicollinearity stemming from multiple indicators when examining the socioeconomic determinants of land use, this study utilized Principal Component Analysis (PCA). This method reduces high-dimensional driving metrics into a few uncorrelated comprehensive principal components while preserving the core information of the original economic data. It avoids the one-sidedness of single metric analysis and streamlines the data structure for subsequent research on driving mechanisms [48,49,50].

In the present study, the year serves as the primary observational unit for principal component analysis (PCA), with each observation corresponding to a distinct year spanning from 2004 to 2024. The final dataset consists of 21 observations and incorporates 11 variables, which are employed to examine the driving effects of annual socioeconomic development on changes in land use. To enhance the understanding of the relationship between economic indicators and land use types,

Table 2. Correlation between socioeconomic indicators and land use type.

Indicator Code	Associated Land Use Type	Influence Mechanism	Expected Direction of Change
X1	Cropland, Forest	Population migration leads to the abandonment of cropland, with some of the unused plots being converted into forest cover.	Cropland reduction, forest expansion
X2	Cropland, Impervious Surface	Economic growth drives the expansion of urban areas, which encroaches on cropland resources for the construction of impervious surface.	Impervious surface expansion, cropland reduction
X3	Grassland, Impervious Surface	Industrial sector growth calls for industrial land development, extending impervious areas and occupying grassland	Impervious surface expansion, grassland reduction
X4	Forest, Impervious Surface	Service industry advancement drives infrastructure construction, while eco-tourism initiatives bolster forest protection	Impervious surface expansion, forest expansion
X5	Cropland, Impervious Surface	Thriving commercial activities fuel the expansion of commercial land, leading to the occupation of suburban cropland	Impervious surface expansion, cropland reduction
X6	Cropland, Impervious Surface	Urban expansion encroaches on peri-urban cropland, transforming it into constructed surfaces like buildings and transport networks	Cropland reduction, impervious surface expansion
X7	Cropland, Forest	Highway construction directly occupies cropland and forest resources along the transportation routes	Cropland reduction, slight forest reduction
X8	Cropland, Water Body	Agricultural development requires cropland preservation, while irrigation projects and aquaculture activities expand water areas	Cropland stability, slight water body expansion
X9	Cropland, Shrub	Grain security needs maintain stable cropland acreage, with partial shrubland reclaimed for agricultural use	Cropland stability, slight shrub reduction
X10	Forest, Impervious Surface	Farmers participate in the Grain for Green Program, and rising rural housing demand expands impervious surfaces in rural areas	Forest expansion, impervious surface expansion
X11	Cropland, Impervious Surface	Real estate projects generate large-scale impervious surfaces, occupying high-quality cropland in urban fringe zones	Impervious surface expansion, cropland reduction

Source: Statistical Yearbook and land cover data, analyzed and compiled by the authors.

presents the relevant influencing mechanisms and anticipated changes. To address the differences in measurement units between the indicators, the eleven socioeconomic variables were first standardized using z-scores to remove scale-related discrepancies. Next, the calculation of eigenvalues, factor contribution rates and cumulative contribution rates was performed. These results helped to identify the main driving factors and derive their corresponding scores.

2.3.5. Grey Prediction Model

The GM(1,1) model, a predictive model, suits situations characterized by sparse data and partial information [51,52]. The impact of randomness is diminished by this model through the aggregation of the initial data, the identification of patterns through parameter-solving differential equations, and the subsequent restoration of the data to generate predictions [53,54]. Importantly, this approach is straightforward and does not hinge on assumptions about data distribution [55]. The Grey Model GM(1,1) is defined as follows:

Let the original non-negative time series be

$$x^{\left(0\right)}=\left\{x^{\left(0\right)}\left(1\right),x^{\left(0\right)}\left(2\right),\dots ,x^{\left(0\right)}\left(n\right)\right\}$$

(7)

Perform a cumulative generation on $x^{\left(0\right)}$ (1-AGO) to obtain the new sequence $x^{\left(1\right)}$:

$$x^{\left(1\right)}\left(k\right)={\displaystyle \sum _{i=1}^k{x}^{\left(0\right)}}\left(i\right),k=1,2,\dots ,n$$

(8)

Define the sequence of adjacent mean values for $x^{\left(1\right)}$:

$$z^{\left(1\right)}\left(k\right)=0.5x^{\left(1\right)}\left(k\right)+0.5x^{\left(1\right)}\left(k-1\right),k=2,3,\dots ,n$$

(9)

The differential equation for GM(1,1) in grey is expressed as follows:

$$x^{\left(0\right)}\left(k\right)+a{z}^{\left(1\right)}\left(k\right)=b,k=2,3,\dots ,n$$

(10)

Express the gray differential equation in matrix form $Y=Bu$:

$$\left[\begin{array}{c}{x}^{\left(0\right)}\left(2\right)\\ x^{\left(0\right)}\left(3\right)\\ ⋮\\ x^{\left(0\right)}\left(n\right)\end{array}\right]=\left[\begin{array}{cc}-z^{\left(1\right)}\left(2\right)& 1\\ -z^{\left(1\right)}\left(3\right)& 1\\ ⋮& ⋮\\ -z^{\left(1\right)}\left(n\right)& 1\end{array}\right]\times \left[\begin{array}{c}a\\ b\end{array}\right]$$

(11)

Minimum-squares estimation of parameters $\widehat{a}$ and $\widehat{b}$:

$$\widehat{u}={\left(B^{T}B\right)}^{-1}{B}^{T}Y$$

(12)

Solving a whitening differential equation:

$${\displaystyle \frac{d{x}^{\left(1\right)}\left(t\right)}{dt}}+a{x}^{\left(1\right)}\left(t\right)=b$$

(13)

$${\widehat{x}}^{\left(1\right)}\left(t\right)=\left(x^{\left(0\right)}\left(1\right)-{\displaystyle \frac{\widehat{b}}{\widehat{a}}}\right)e^{-\widehat{a}\left(t-1\right)}{\displaystyle \frac{\widehat{b}}{\widehat{a}}}$$

(14)

Setting $t=k$ to discretize the time response function leads to the predicted sequence $x^{\left(1\right)}$:

$${\widehat{x}}^{\left(1\right)}\left(k\right)=\left(x^{\left(0\right)}\left(1\right)-{\displaystyle \frac{\widehat{b}}{\widehat{a}}}\right)e^{-\left(k-1\right)\widehat{a}}+{\displaystyle \frac{\widehat{b}}{\widehat{a}}}$$

(15)

Cumulative reduction yields the predicted value for the original sequence $x^{\left(0\right)}$:

$${\widehat{x}}^{\left(0\right)}\left(k\right)=\left(1-e^{\widehat{a}}\right)·\left(x^{\left(0\right)}\left(1\right)-{\displaystyle \frac{\widehat{b}}{\widehat{a}}}\right)·e^{-\left(k-1\right)\widehat{a}}$$

(16)

where ${\widehat{x}}^{\left(0\right)}\left(k\right)$ denotes predicted value; k is the time sequence number; $x^0\left(1\right)$ is the first data point in the original sequence; $a$ denotes the development coefficient; $b$ denotes the quantity of the grey effect; e is a natural constant approximating 2.71828.

Model accuracy is typically evaluated using either the mean relative error or the posterior error method. A mean relative error (MRE) of less than 20% is considered acceptable, while below 10% indicates high predictive accuracy [56]. The posterior variance test involves the posterior ratio C and the small error probability p. These two metrics are used to evaluate model reliability, with prediction accuracy categorized into four levels, as summarized in

Table 3. Grades for grey system prediction accuracy test.

Model Accuracy Grade	C	p
1 (Excellent)	C ≤ 0.35	p ≥ 0.95
2 (Good)	0.35 < C ≤ 0.5	0.8 ≤ p< 0.95
3 (Qualified)	0.5 < C ≤ 0.65	0.7 ≤ p < 0.8
4 (Unqualified)	C > 0.65	p < 0.7

Abbreviations: C, the posterior ratio; p, the small error probability. Source: Compiled based on the accuracy verification grading standards derived from grey system theory.

.

$$S_1=\sqrt{{\displaystyle \frac{1}{n}}{\displaystyle \sum _{k=1}^n{\left(x^{\left(0\right)}\left(k\right)-{\overline{x}}^{\left(0\right)}\right)}^2}}$$

(17)

$$S_2=\sqrt{{\displaystyle \frac{1}{n}}{\displaystyle \sum _{k-1}^n{\left(\epsilon \left(k\right)-\overline{\epsilon }\right)}^2}}$$

(18)

$$C={\displaystyle \frac{S_1}{S_2}}$$

(19)

$$p=p\left(\left|\epsilon \left(k\right)-\overline{\epsilon }\right|<0.6745S_1\right)$$

(20)

where S₁ denotes original sequence’s standard deviation; S₂ denotes residual sequence’s standard deviation; $\epsilon \left(k\right)$ is the residual of the k-th point; $\overline{\epsilon }$ is the mean of the residual sequence. The model can be diagnosed based on the values of C and p: a smaller C value and a larger p value indicate better predictive performance of the model [52].

2.4. Research Process

To systematically elucidate the technical approach and logical framework of this study, a technical flowchart is presented in Figure 3. This diagram comprehensively illustrates the entire process, from data foundation to final outputs. At the citywide scale of Bazhong City, multi-period land cover data were employed to extract areas and conduct dynamic analyses, including dynamic degree, composite index, and transition matrix, utilizing ArcGIS. SPSS 27.0.1 software was used to perform driving force analysis on socioeconomic indicators that influence land use. GM(1,1) models were constructed using area data through MATLAB R2016a and an MCDM online platform for predictive modeling.

3. Results

3.1. Dynamic Changes in Land Use

3.1.1. Quantity Change in Land Use

GIS technology was utilized to extract land use area data for Bazhong City across three time periods (2004, 2014, and 2024), and the proportions and alterations in each land use category were computed (

Table 4. Area proportion and changes in land use types in Bazhong City (km², %).

Type	2004	2014	2024	2004–2014	2014–2024	2004–2024
	Proportion	Proportion	Proportion	Area Change	Area Change	Area Change
Cropland	43.93	40.53	33.02	−417.65	−921.05	−1338.69
Forest	54.96	58.27	65.60	405.85	899.03	1304.88
Shrub	0.27	0.10	0.04	−21.43	−7.68	−29.11
Grassland	0.06	0.04	0.03	−2.69	−1.04	−3.74
Water Body	0.44	0.54	0.52	12.02	−2.84	9.18
Impervious Surface	0.33	0.53	0.80	23.89	33.58	57.47

Source: Calculated based on China’s 30 m interannual land cover dataset (2004, 2014, 2024).

). By applying Equations (1) and (2), the magnitude of change and the single land use dynamic degree in Bazhong City were deter-mined and are detailed in

Table 5. Magnitude of change (R_L, %) and single dynamic degree (K, %/a) of land use types in Bazhong City.

Type	2004–2014		2014–2024		2004–2024
	RL	K	RL	K	RL	K
Cropland	−7.75	−0.78	−18.53	−1.85	−24.84	−1.24
Forest	6.02	0.60	12.58	1.26	19.36	0.97
Shrub	−64.11	−6.41	−64.01	−6.40	−87.08	−4.35
Grassland	−37.91	−3.79	−23.70	−2.37	−52.62	−2.63
Water Body	22.22	2.22	−4.29	−0.43	16.97	0.85
Impervious Surface	58.40	5.84	51.83	5.18	140.50	7.03

Source: Calculated using Equations (1) and (2) based on the data in Table 4.

.

Analysis of Table 4 and Table 5 reveals that the proportional distribution of land types in Bazhong City remained fundamentally consistent between 2004 and 2024, with cropland and forest land consistently constituting the dominant components of the land use structure, collectively accounting for over 95% of the city’s total area.

From 2004 to 2024, the R_L value for cropland remained negative, indicating a continuous reduction in area. The dynamic degree K for cropland was also consistently negative, and its absolute value showed an increasing trend, suggesting an acceleration in the rate of cropland loss. The RL value for the forest showed a consistent positive trend from 2004 to 2024, indicating a continuous expansion of the forested area. The dynamic degree K also maintained positive values throughout this period, reaching its peak between 2014 and 2024 at K = 1.26%/a. Between 2004 and 2024, both the R_L value and the dynamic degree K for shrub exhibited a consistent negative trend, indicating a decline in shrub coverage. The R_L value for shrub notably plummeted to −87.08%, marking the most substantial reduction among all land cover types. From 2004 to 2024, grassland consistently showed negative R_L values, indicating a continuous decrease in area. The dynamic degree K remained negative during this period, with its absolute value demonstrating a declining trend, reflecting a progressive slowdown in the rate of grassland reduction. From 2004 to 2024, the R_L value for water body was 16.97% and the dynamic coefficient K was 0.85%/a, indicating an increase in total water area. From 2004 to 2014, the R_L value was 22.22% and K was 2.22%/a, while for 2014 to 2024, R_L = −4.29% and K = −0.43%/a, indicating water area first expanded and then contracted. Between 2004 and 2024, there was a continuous increase in both the R_L and K values for impervious surface, signifying consistent expansion.

In summary, from 2004 to 2024, the change rates for cropland, shrub, and grassland were negative, while those for forest, water body, and impervious surface were positive. Shrub experienced the largest decrease (R_L = −87.08%), and impervious surface showed the greatest increase (R_L = 140.50%). Considering the initial total area differences, cropland experienced the largest net loss in area, decreasing by 1338.69 km². Conversely, forest area showed the highest net increase, expanding by 1304.88 km².

3.1.2. Changes in Land Use Degree

Employing Equation (3), the comprehensive land use degree index for Bazhong City between 2004 and 2024 was ascertained. Figure 4 presents a schematic representation elucidating the fluctuations in the degree of land utilization. According to Equations (4) and (5), the amount and rate of change in land use degree in Bazhong City were obtained, as demonstrated in

Table 6. Amount and rate of change in land use degree for Bazhong City.

Year	Change Amount	Change Rate
2004–2009	−4.22	−1.73%
2009–2014	1.21	0.50%
2014–2019	−3.97	−1.64%
2019–2024	−2.99	−1.26%
2004–2024	−9.98	−4.08%

Source: Calculated based on Equations (3)–(5).

.

As shown in Figure 4, Bazhong City’s overall land use degree exhibited a downward trend between 2004 and 2024. The comprehensive index peaked at 244.60 in 2004 and decreased to its lowest point of 234.14 in 2022. The rate of change varied over different periods. Between 2004 and 2008, the index decreased from 244.60 to 239.19 at a relatively rapid pace. From 2008 to 2018, the index generally exhibited an upward trajectory with modest growth. Between 2018 and 2022, the index experienced a sharp decline from 239.62 to 234.14. From 2022 to 2024, the index showed a rising trend. Table 6 shows that land use degree change was positive solely during 2009–2014. The smallest change occurred between 2004 and 2009, with a change value of −4.22 and a change rate of −1.73%. From 2004 to 2024, the overall change in land use degree was −9.98, with a change rate of −4.08%.

3.1.3. Changes in Land Use Types

Statistical analysis was conducted on the land use status, and the spatial distribution maps for 2004, 2014, and 2024 were plotted (Figure 5). The GIS-derived transfer matrix (

Table 7. Land use transition matrix for Bazhong City (2004–2024) (km²).

	2024
	Type	Cropland	Forest	Shrub	Grassland	Water Body	Impervious Surface	Total
2004	Cropland	3594.33	1719.98	0.70	2.00	10.22	61.05	5388.28
	Forest	443.87	6289.85	2.34	0.42	0.58	4.45	6741.50
	Shrub	2.29	29.56	1.23	0.35	0.01	0.00	33.43
	Grassland	1.19	5.10	0.05	0.60	0.07	0.09	7.10
	Water Body	6.69	1.87	0.00	0.00	44.09	1.47	54.12
	Impervious Surface	1.23	0.02	0.00	0.00	8.34	31.32	40.91
	Total	4049.59	8046.38	4.32	3.36	63.31	98.38	12,265.34

Source: Land use type transition data was generated using ArcGIS spatial analysis.

) clearly demonstrates land use transition in Bazhong City between 2004 and 2024, and land use transfer Sankey diagram among these years was generated (Figure 6).

As shown in Table 7 and Figure 6, the largest area of conversion throughout the study periods (2004–2014, 2014–2024, and 2004–2024) occurred between cropland and forest, while conversions among other types were relatively minor. The following details the land type conversions from 2004 to 2024:

Between 2004 and 2024, an area of 1719.98 km² of cropland in Bazhong City was converted to forest. Additionally, 61.05 km² of cropland was transformed into impervious surface. Conversions from cropland to shrub of 0.70 km², grassland of 2.00 km², and water body of 10.22 km² were relatively limited. The primary transformation of forest was to cropland, with 443.87 km² being converted. The main transformation process for shrub was to forest, covering 29.56 km². Variations in grassland, water body and impervious surface were slight. These land types constitute a small proportion of the overall land use structure and exhibited limited spatial transfer.

3.2. Driving Forces Analysis

Common factors are usually identified by extracting eigenvalues. Factors with eigenvalues greater than 1 are retained, while those below this threshold are excluded [57]. The eigenvalues and variance contributions of the components are detailed in

Table 8. Eigenvalues and variance contributions of the principal components.

Component	Initial Eigenvalues			Sum of Squares Loadings Extracted
	Total	Percentage of Variance	Cumulative	Total	Percentage of Variance	Cumulative
1	8.26	75.12	75.12	8.26	75.12	75.12
2	1.90	17.26	92.38	1.90	17.26	92.38
3	0.39	3.56	95.94
4	0.26	2.33	98.28
5	0.12	1.09	99.37
6	0.04	0.41	99.77
7	0.01	0.11	99.88
8	0.01	0.06	99.94
9	0	0.04	99.98
10	0	0.02	99.99
11	0	0.01	100

Source: Derived from Principal Component Analysis conducted using SPSS 27.0.1, based on standardized socioeconomic indicators.

.

As indicated in Table 8, two principal components with eigenvalues exceeding 1 were retained. The eigenvalue of Principal Component 1 (PC1) was 8.26, explaining 75.12% of the total variance, while that of Principal Component 2 (PC2) was 1.90, explaining 17.26%. Together, the cumulative contribution rate reached 92.38%; therefore, PC1 and PC2 were selected as composite factors for further analysis [58].

Based on the eigenvectors of the eigenvalues, the loadings of each factor on PC1 and PC2 were calculated. The factor score coefficients were derived through computational processing and are presented in

Table 9. Factor loadings and score coefficients of the principal components.

	Factor Loading		Score Coefficient
	1	2	1	2
X1	−0.741	0.574	−0.090	0.303
X2	0.986	0.155	0.119	0.082
X3	−0.074	0.959	−0.009	0.505
X4	0.957	0.103	0.116	0.054
X5	0.992	0.030	0.120	0.016
X6	0.974	0.197	0.118	0.104
X7	0.942	0.125	0.114	0.066
X8	0.939	−0.240	0.114	−0.127
X9	0.833	−0.358	0.101	−0.188
X10	0.995	0.028	0.120	0.015
X11	0.659	0.610	0.080	0.322

Source: Derived from Principal Component Analysis conducted using SPSS 27.0.1, based on standardized socioeconomic indicators.

.

As shown in Table 9, PC1 exhibits strong positive correlations (loadings > 0.9) with variables X2 (GDP), X4 (Tertiary Industry), X5 (Retail Sales), X6 (Urbanization Rate), X7 (Highway Mileage), X8 (Agricultural Output), and X10 (Farmer Income). These variables reflect the overall scale and structural modernization level of regional development; hence, PC1 is named the Comprehensive Economic Development and Urbanization Factor. PC2 exhibits a significant positive correlation with variable X3 (Proportion of Secondary Industry), revealing the degree of emphasis on industry within the economic configuration. Therefore, PC2 is named the Industrialization Emphasis Factor.

The driving force scores corresponding to PC1 and PC2, denoted as F1 and F2, respectively, were calculated. Subsequently, the comprehensive driving force (F) was derived from the weighted combination of F1 and F2 [59], as presented in Figure 7.

Figure 7 indicates an overall gradually increasing trend in the scores of PC1, which turned positive in 2016. The PC2 score demonstrated an upward movement, peaking at 1.22 in 2017 before decreasing. From 2004 to 2010, the comprehensive score ranged from −1.27 to −0.66. This period was characterized by synchronous low growth in both PC1 and PC2. This indicates that there was generally a weak driving force on land use, as both factors were in their nascent stages and had not formed an effective synergistic system. Between 2010 and 2018, the composite score increased from −0.66 to 0.78, achieving a breakthrough by transitioning from negative to positive values, with a stable growth trend. During this period, PC2 provided the main impetus for land use drivers, confirming the continuously strengthening impact of industrial development on land use. From 2018 to 2024, the composite score increased from 0.78 to 1.39, with growth accelerating further over this period. During this period, PC1 emerged as the dominant factor, with its steadily increasing score directly propelling the rapid rise in the composite score. This suggests that the impact of PC1 on land use intensified during this period, with its driving force becoming more pronounced and stable.

3.3. Prediction of Land Use Change Trends

3.3.1. Validation of GM(1,1) Model

In order to gauge the reliability of the GM(1,1) model, this study validated the predicted values against observed data for different land use types in Bazhong City between 2021 and 2024.

As shown in

Table 10. Fitting results of the GM(1,1) model for various land use types (km², %).

Year		Cropland	Forest	Shrub	Grassland	Water Body	Impervious Surface
2021	Observed	4048.49	8049.49	4.63	5.53	65.91	91.29
	Predicted	4062.31	8035.67	4.50	5.20	66.24	91.42
2022	Observed	3994.79	8100.81	4.42	4.61	64.42	96.30
	Predicted	4008.47	8087.13	4.29	4.28	64.75	96.42
2023	Observed	4013.64	8082.17	4.17	3.83	63.91	97.61
	Predicted	4027.30	8068.51	4.04	3.50	64.24	97.74
2024	Observed	4049.59	8046.38	4.32	3.36	63.31	98.38
	Predicted	4063.25	8032.72	4.19	3.03	63.64	98.51
	MRE	0.34%	0.17%	2.97%	7.89%	0.51%	0.13%
	C	0.31	0.29	0.35	0.35	0.33	0.35
	p	1.00	1.00	0.97	0.95	0.98	1.00

Abbreviations: MRE, mean relative error; C, the posterior ratio; p, the small error probability. Source: Based on historical area data, predictions were made using the GM(1,1) model and compared with actual observed values.

, the posterior ratio C ranges from 0.29 to 0.35, while the probability of small error p ranges from 0.95 to 1.00. The MRE between the actual measurements and the simulated results of the model remains low across various land use types. Based on the accuracy ratings in Table 3, the posterior ratio C and the small error probability p achieve the highest grade (Excellent for most types, Good for Shrub and Grassland), guaranteeing the stability and reliability of the model’s predictive outcomes. The MRE values were given below: cropland at 0.34%, forest at 0.17%, shrub at 2.97%, grassland at 7.89%, water body at 0.51%, and impervious surface at 0.13%. These minor discrepancies suggest that the prediction accurately captured changes in land use areas from 2021 to 2024. The model achieved an excellent accuracy grade overall and possesses high reliability.

3.3.2. Prediction of Land Use Scenarios

The validated model was employed to forecast land use changes in Bazhong City for the years 2025, 2030, and 2035, as detailed in

Table 11. Projected Area of Land Use Types in Bazhong City (km²).

Year	Cropland	Forest	Shrub	Grassland	Water Body	Impervious Surface
2025	4035.12	8058.21	4.18	3.12	62.89	101.82
2030	3931.73	8149.52	3.13	2.05	60.04	118.87
2035	3785.76	8285.38	2.33	1.34	56.94	133.59

Source: Forecasts for future years (2025, 2030, 2035) based on the GM(1,1) model.

. The magnitude of land area change was calculated for the periods 2004–2024, 2025–2030, 2030–2035, and 2024–2035 (

Table 12. Magnitude of Change in Land Use Type Areas across Different Periods (km²).

Type	2004–2024	2025–2030	2030–2035	2024–2035
Cropland	−1338.69	−103.39	−145.97	−263.83
Forest	1304.88	91.31	135.86	239.00
Shrub	−29.11	−1.05	−0.80	−1.99
Grassland	−3.74	−1.07	−0.71	−2.02
Water Body	9.19	−2.85	−3.10	−6.37
Impervious Surface	57.47	17.05	14.72	35.21

Source: Calculated based on the forecast data and historical data in Table 11.

).

As shown in Table 11 and Table 12, from 2004 to 2024, the areas of forest, water body, and impervious surface increased, whereas the areas of cropland, shrub, and grassland decreased. Cropland experienced the most substantial net loss of 1338.69 km², whereas forest showed the largest net expansion of 1304.88 km².

Based on the prediction results, during 2025–2030 and 2030–2035, the areas of forest and impervious surface will continue to increase, while the areas of cropland, shrub, grassland, and water body will decrease. Throughout the 2024–2035 period, a decrease of 263.83 km² is projected for the cropland area, standing out as the greatest reduction among all types of land use. In contrast, the area of forest is forecast to increase by 239.00 km², marking the most significant increment within the aforementioned period.

4. Discussion

4.1. Domestic and International Comparative Analysis

This study reveals a distinct “three increases and three decreases” pattern in land use changes in Bazhong City from 2004 to 2024, characterized by sustained expansion of forest, water body, and impervious surface, alongside consistent reduction in cropland, shrub, and grassland. This trend aligns with findings previously reported in the same study area [31].

The land use change pattern in Bazhong City, characterized by a reduction in cropland and an expansion of forest, exhibits similarities with ecologically vulnerable regions in western China, including Guizhou and Gansu. This pattern across these areas is primarily driven by both ecological conservation policies, such as the Grain for Green Program, and ongoing urbanization. During the transformation process, forests in these cities fall primarily into two categories: ecological forests, which focus on environmental conservation, and economic forests, which focus on producing commercial goods. Bazhong City’s forest expansion is characterized by a high proportion of economic forests; Gansu Province’s expansion is primarily driven by natural factors, with Longnan City placing greater emphasis on ecological forest restoration [12,47]; Guizhou Province’s expansion of construction land is more pronounced [60].

In comparison to more developed regions such as Zhengzhou, Hubei (centered on Wuhan), Guangzhou and Jinan, the main cause of cropland reduction is the expansion of construction land due to urbanization and industrialization [14,46,61,62]. In contrast, the main cause of cropland reduction in Bazhong is the Grain for Green Program, reflecting differences in developmental stages and policy orientations.

Notable differences exist in comparison to foreign regions. For instance, coastal erosion and rising sea levels have degraded 76.04% of cropland in the coastal areas of Bangladesh [63]. In the Meshginshahr region of Iran, agricultural expansion and urbanization are the main drivers of land use change, with the conversion of pasture and forest land into cropland being the main development strategy [64]. These two regions are therefore representative of “passive losses” and “agricultural development dominance”, respectively, contrasting with the land use patterns observed in Bazhong City.

4.2. Summary and Implications

Between 2004 and 2024, forest and cropland in Bazhong City experienced the most significant changes in area. Cropland was primarily converted into forest and impervious surface, while conversions between other land types were relatively minor.

The observed land transformation was driven by multiple factors: From 2004 to 2024, Bazhong City’s population decreased by nearly 400,000, particularly among young rural laborers. This demographic shift resulted in significant natural restoration of remote sloping cropland into forested areas. Consequently, this process led to a reduction in the total area of cropland and an increase in forest coverage. Accelerated urbanization and transportation infra-structure development directly converted peri-urban cropland, while the effective implementation of the Grain for Green Program—resulting in approximately 1400 km² of converted land—facilitated the transition of sloped and low-yield cropland into forest. These processes collectively led to progressive diminution of agricultural land and simultaneous expansion of forest area. The reduction in cropland far exceeds the increase in impervious surface, resulting in a fluctuating downward trend in regional land use degree and an overall low level of land development intensity. The modest decline in water body following 2014 primarily resulted from reduced precipitation in certain years attributed to climate change, the encroachment of water body due to urbanization and infrastructure development, and the intensified exploitation of water resources. From 2004 to 2024, forest coverage increased consistently, reaching approximately 65% by the end of 2024. This expansion contributed to the enhancement of ecosystem service values, particularly improving water conservation and soil retention capacity [65,66].

The concurrent processes of urbanization and forest expansion have revealed conflicting land use dynamics. Prime cropland surrounding urban areas has been encroached upon, while portions of cropland have been converted to forest. Given that Bazhong’s topography is predominantly mountainous, the availability of contiguous cropland is limited. Much of the remaining cropland consists of fragmented, sloped plots with low productivity, further constraining cropland reserve resources, a situation similar to the quality degradation of cropland noted in others research [67,68].

In this study, two principal components (PC1 and PC2) were extracted via principal component analysis (PCA). These components can explicitly reveal the driving mechanism of socioeconomic development on the evolution of land use patterns in Bazhong City between 2004 and 2024.

From 2004 to 2010, during the accumulation phase, Bazhong City focused on developing transportation infrastructure and sustaining traditional agriculture. Industrial development centered on small-scale operations and primary processing, while the urbanization rate remained below 30%. During this period, the expansion of land for construction proceeded slowly, maintaining the stability of the overall land use structure.

The period from 2010 to 2018 was a time of robust growth. A pivotal turning point came in 2011, when the Bazhong Economic Development Zone was expanded and relocated, followed by construction. This strategic move spurred a dramatic surge in industrial output value, which jumped from under 100 million yuan to 2.6 billion yuan. The implementation of industrial projects further fueled the restructuring of regional land use configurations, with both industrial and urban construction land witnessing considerable expansion during this period.

The period between 2018 and 2024 was characterized by a surge in growth, marked by a significant shift towards specialization and environmentally—friendly practices in the industrial sector. The completion of the Ba-Shan Expressway, the opening of Enyang Airport and the introduction of high-speed rail services improved transport links. The “One City, Three Districts” urban spatial pattern was formed because of these developments, which led to a significant increase in the urbanization rate to 48%. During this period, the development of urban construction land will shift towards refinement, while the growth of industrial land will decelerate. It is important to note that problems such as unoccupied rural properties and deserted cropland resulting from population migration will further decrease the amount of cropland, indicating the need for urgent action to protect cropland. Subsequent policy implementation will be required to achieve the dual objective of safeguarding the quantity and enhancing the quality of cropland.

According to the forecast results, the area of forested and impervious surface is expected to increase, while the area of cropland, grassland, shrub, and water body is projected to decrease. The amount of cropland maintained is fundamental to sustaining agricultural out-put, ensuring food security, and maintaining social stability. Based on the “Bazhong Territorial Space Master Plan (2021–2035),” the city’s cropland must remain at or above 2520.07 km² by 2035. Although the projected cropland area for 2035 (3785.76 km²) exceeds the conservation target (2520.07 km²), Bazhong City continues to suffer a net loss of cropland. This trend will reduce the buffering role of regional land reserves, compromise cropland quality and spatial contiguity, and ultimately weaken the region’s grain production capacity, thereby posing a threat to food security. Future strategies should include strict enforcement of the permanent prime cropland protection system to safeguard the minimum quantity requirement. Spatial layout optimization through land consolidation and scientifically delineated urban development boundaries can help revitalize existing land resources. Additionally, implementing high-standard cropland construction and promoting eco-agricultural technologies are crucial to enhancing cropland quality and compensating for quantitative losses.

An increased forested land area helps to reduce soil erosion, enhance water conservation and protect biodiversity. Efforts to consolidate achievements in the Grain for Green Program and strengthen the protection of ecological forests should continue. The reduction in cropland and the outflow of rural populations reinforce each other, which could accelerate the depopulation of rural areas. To reverse this trend, policy measures must improve integrated urban-rural development strategies to encourage population return, facilitate land transfers, and promote large-scale operations, thereby enhancing land use efficiency.

4.3. Limitations and Future Directions

Building upon previous methodologies, this study has yielded certain findings regarding land use in Bazhong City. However, due to the broad scope of the research, limitations in data availability, and insufficient flexibility in methodological application, the following improvements could be made in future studies on land use dynamics and prediction:

Regarding data selection, a 30 m spatial resolution is insufficient for accurately representing the boundaries and internal variability of small ground features. To enhance classification accuracy, future efforts should focus on integrating diverse data sources and advancing sub-pixel unmixing and deep learning methodologies [69]. The principal component analysis covered 21 years of data and 11 variables. The stability of the analytical results was somewhat compromised by the relatively small sample size and limited number of variables. Key socioeconomic variables such as employment distribution, land prices and policy intensity were excluded during the selection process due to data unavailability, which could result in deviations in the interpretation of the driving mechanisms. This study concentrates on macro trends, considering cropland, forest, and other areas as uniform entities. Neglecting the functional variations within these areas and the consequent requirement for tailored policies could lead to imprecise policy development. Additionally, the absence of consistent quantitative data on the intensity of policy implementation implies that policy outcomes are deduced from contextual correlations rather than empirical testing based on models, leading to constraints in assessing policy impact. Hence, pertinent policy implementations should incorporate functional classification at the parcel level alongside field validation, improve metrics, and adjust them dynamically to improve policy precision and efficacy. Future efforts will focus on creating a comprehensive, long-term, cross-departmental database that integrates multi-source patio-temporal data [70]. This will enable a more systematic and stable examination of the complex interactions between driving factors. The main constraint in using data from three particular time points—2004, 2014, and 2024—to judge area alteration and build the land use transfer matrix is the incapability to capture the evolving processes of land use transformation over the two decades. Future endeavors will focus on establishing a sequence of datasets with increased temporal granularity, spanning consecutive years or quarters, to accurately depict the ongoing trajectory of land use alterations.

Regarding methodology, the index of land use degree is simple to calculate, and the results are easy to understand. They are also comparable across different regions and can be used in many situations. This index can quickly quantify the intensity of regional land development and utilization. However, it may obscure significant spatial heterogeneity. Future research will incorporate landscape indices such as the Shannon diversity index and the fragmentation index, as well as slope-based stratified analysis and topographic constraint analysis [71,72,73]. The analysis in this study primarily focuses on correlations rather than causality when examining the socioeconomic drivers and land use change. While principal component analysis revealed associations among key factors, including GDP, urbanization rate and land use change (Table 2), the research did not validate causal relationships directly through regression analysis or causal inference models. The correlations identified in this study should not be equated with unidirectional causality. It is important to avoid overinterpreting the driving effects of socioeconomic factors on land use change. During principal component analysis, influential factors are often regarded as variables assumed to be independent of land use changes, neglecting the effects of time lags and spatial arrangements. Future research will integrate methods such as spatio-temporal lag models and spatial Durbin models to establish a comprehensive framework for understanding the driving forces of land use, taking into account both spatial and temporal dynamics [74,75]. The GM(1,1) grey prediction model is known for its low data requirements, lack of specific statistical distribution assumptions and simple calculation process. It mitigates the randomness of the initial data through cumulative generation. It exhibits notable accuracy and stability when forecasting short-term and minor trends in land use changes. The forecast period from 2025 to 2035 aligns with the core applicability range of the GM(1,1) model, with the forecasting error falling within an acceptable range, surpassing long-term predictions exceeding a decade. The GM(1,1) model demonstrates high predictive accuracy for medium-to-short-term land use change trends between 2025 and 2035. However, it is still limited by the inherent uncertainty of forecasting. Based on the cumulative generation principle of time series data, the model mitigates the impact of random influences, yet it fails to account for potential future shocks, such as policy shifts or extreme climate events. Consequently, the deterministic trajectories it generates do not incorporate uncertainty boundaries, such as confidence intervals. The model assumes continuity of historical trends from 2004 to 2024 and does not consider multiple potential development scenarios, such as enhanced ecological conservation or accelerated urbanization. Therefore, applying the findings of this study to long-term planning beyond 2035 requires caution, given that uncertainty in long-term projections increases significantly. However, this model performs poorly in terms of fitting abrupt land-use transitions driven by policy changes, and it fails to incorporate spatial characteristics such as location and neighborhood effects. This makes it difficult to reflect the spatial heterogeneity of regional land-use changes. Future efforts will focus on incorporating additional predictive models (e.g., CA-Markov, FLUS, PLUS) to enhance the accuracy and temporal scope of land use projections [76,77,78,79,80].

5. Conclusions

Based on 30 m spatial resolution land cover data and utilizing multiple methods including Principal Component Analysis, this study has quantified the status and change trends in land use in Bazhong City, identified the primary economic drivers of land use change, and employed the grey GM(1,1) model to predict future land use changes in the city. The primary conclusions are as follows:

(1)

Between 2004 and 2024, Bazhong City’s land use followed a clear “three increases and three decreases” pattern. Specifically, forest, water body and impervious surface expanded, while cropland, shrub and grassland shrank. The transformation of cropland into forest was the main driver of this shift in land use, demonstrating a clear trade-off dynamic where one increased as the other decreased. The conversion of cropland primarily involves its transformation into forest or impervious surface, while forest mainly originates from converted cropland. This trend is similar to the common characteristics seen in ecologically fragile regions of western China. These regions differ from developed areas, where cropland reduction patterns are dominated by construction land expansion, and from some foreign regions, where land use conversion is driven by agriculture and is passive in nature.

(2)

During this period, the degree of land use in the study area showed a variable downward trend, suggesting a change in development priorities towards ecological restoration and the efficient use of construction land. The decline in cropland far exceeds the increase in impervious surface. The quality and spatial continuity of cropland has been compromised by this reduction, resulting in a diminished buffering capacity of land reserves. Conversely, the sustained growth of forest has effectively improved regional ecosystem services, such as water conservation and soil retention. This highlights the positive outcomes of ecological conservation efforts.

(3)

A clear three-stage pattern is exhibited by the evolution of drive systems: From 2004 to 2010, during the accumulation phase, development momentum was generally weak, and land use patterns remained stable. The period from 2010 to 2014 marked the beginning of a growth phase, where industrialization emerged as the primary driver, fueling a continuous expansion of industrial land use. Entering the accelerated growth stage from 2018 to 2024, the combined forces of economic development and urbanization came into play. Enhanced transportation infrastructure and optimized urban spatial configurations jointly drove a profound transformation in the structure of land use.

(4)

Projections from the GM(1,1) model indicate that, from 2025 to 2035, forest and impervious surface in Bazhong City will increase, while cropland, shrub, grassland and water body will decrease. The difference between cropland and forest will become increasingly pronounced. While the protection targets outlined in national spatial planning may still be satisfied by the projected cropland area in 2035, regional grain production capacity will be diminished by the ongoing decline in cropland and a potential threat to food security will be posed. Furthermore, forest expansion will reinforce the gains achieved in ecological protection, creating a conflict between ecological preservation and food security.

(5)

The interaction between ecological conservation policies and the urbanization process has fundamentally shaped land use transitions in mountainous areas. Future regional development must adhere to the bottom line of cropland protection. This objective can be realized through the rigorous implementation of a permanent basic cropland protection system, the advancement of land consolidation, and the development of high-standard cropland, in conjunction with the optimization of the spatial arrangement of territories. These measures should ensure that gains in ecological protection are maintained while safeguarding food security. It is imperative to enhance policies for integrated urban-rural development, direct land transfers and large-scale operations, mitigate the detrimental cycle of rural population outflow and fallow farmland, and attain sustainable land resource utilization and harmonized ecological and economic development. These findings could inform land use planning and sustainable development in similar ecologically vulnerable regions.