Evaluation of Cultivated Land Fragmentation and Analysis of Driving Factors in the Major Grain-Producing Areas of the Middle and Lower Yangtze River Basin

Abstract

Cultivated land fragmentation has become a critical constraint on regional agricultural sustainable development. Revealing its spatial patterns and driving mechanisms is of great significance for optimizing the utilization and management of cultivated land resources and enhancing regional agricultural productivity. This study focuses on the main grain-producing areas in the middle and lower reaches of the Yangtze River Basin. It constructs a Cultivated Land Fragmentation Index (CLFI) using an integrated method that combines landscape index analysis with an entropy-weighted approach, based on 2023 land-use data. The optimal analytical grain size and extent were determined before employing geographic detectors to identify dominant factors influencing cultivated land fragmentation. The key findings include the following: (1) The appropriate spatial resolution for fragmentation analysis was identified as 330 m, with an optimal analysis extent of 8910 m. (2) CLFI values ranged from 0.001 to 0.973, exhibiting significant spatial heterogeneity. The central plains and northeastern regions demonstrated low fragmentation levels and better contiguous cultivated land distribution, while the western and peripheral areas showed higher fragmentation. A provincial-scale comparison revealed that Jiangxi Province had the highest fragmentation level (0.255), whereas Jiangsu Province had the lowest (0.146). The topographic gradient analysis indicated a decreasing trend from the Guizhou Plateau (0.503) to the North China Plain (0.125), with plateaus and basins showing significantly higher fragmentation than hilly and plain regions. (3) Dominant controlling factors varied among provinces: In provinces with greater topographic relief (Anhui, Hubei, Hunan, Jiangxi), natural factors like elevation, slope gradient, and NDVI primarily controlled fragmentation patterns; in contrast, socioeconomic factors such as nighttime light intensity dominated in Jiangsu Province, characterized by flat terrain and high urbanization. Multi-factor interactions generally enhanced explanatory power regarding spatial patterns, confirming that cultivated land fragmentation is a result of comprehensive multi-factor interactions. This study reveals the spatial distribution characteristics of cultivated land fragmentation at the pixel scale in the study region, providing theoretical foundations and decision-making references for the efficient utilization of cultivated land resources and rural land system reforms.

1. Introduction

Cultivated land resources constitute the most fundamental material basis for agricultural production, playing a critical role in food supply, ecological cycles, and resource–economic activities. They exert profound impacts on ensuring national food security, ecological safety, and sustainable human development [1]. Economic development has accelerated urban spatial expansion and intensified factor flows between urban and rural areas, leading to the severe fragmentation of China’s cultivated land resources. This phenomenon manifests specifically as reduced total area and increased number of plots [2]. Such fragmentation not only hinders the advancement of large-scale and intensive agricultural operations but also further escalates technical difficulties in farming practices and increases cultivation costs [3,4].

In response to cultivated land fragmentation, China has implemented a stringent cultivated land protection system to ensure that land-use purposes remain unchanged and that comprehensive agricultural production capacity remains intact. This approach aims to maintain stability in land utilization while enhancing agricultural productivity. A thorough analysis of cultivated land fragmentation has significant implications for improving agricultural production efficiency, promoting the optimal allocation of arable land resources, and safeguarding national food security.

In recent years, research on cultivated land fragmentation has primarily focused on fragmentation assessment [5,6] and on analyses of influencing factors [7,8]. Regarding the evaluation of cultivated land fragmentation, studies can be categorized into two perspectives: tenure fragmentation and landscape fragmentation [9,10]. The former adopts a micro-level perspective from the producers’ viewpoint, emphasizing that fragmentation represents an infinite subdivision process of property rights. Research on cultivated land fragmentation from a tenure perspective typically employs methods that combine field surveys with statistical analyses [11]. Investigators collect sample data through household interviews, including information on farmers’ land-use patterns and property ownership, and then use statistical analysis to identify the phenomena and causes of cultivated land fragmentation [12]. This approach emphasizes social tenure attributes of cultivated land and facilitates farmer decision making; however, its reliance on survey and interview data, which are often difficult to acquire, constrains the feasibility of large-scale fragmentation studies. Landscape fragmentation focuses on spatial morphology, patch distribution, and landscape pattern characteristics of cultivated land. With advancements in GIS technology and landscape ecology, current studies increasingly analyze landscape fragmentation using land-use data and landscape metrics [13,14,15]. Given limitations in accessing large-scale tenure data, this study primarily concentrates on cultivated land landscape fragmentation.

Existing studies have demonstrated that landscape indices can effectively characterize cultivated land fragmentation and have been widely applied in related research [16,17,18,19,20]. However, landscape indices exhibit significant scale dependency, and neglecting their scale effects may lead to differentiated impacts on research outcomes [21,22]. In relevant studies, scale is primarily described in terms of grain size [23] and extent [24], yet it lacks universality [25]. For instance, scholars have determined optimal grain sizes and extents for specific analyses—including land-use conflict analysis, cultivated land fragmentation assessment, and landscape pattern interpretation—across diverse regions such as Poyang Lake Basin [26], Shaanxi Province [27], Bosten Lake watershed [28], and Southwest China [29]. These identified scale ranges vary from tens of meters to several kilometers, thereby highlighting the pronounced regional specificity inherent in determining optimal scales for landscape analysis.

Currently, the primary methods for determining optimal granularity include the granularity extrapolation method [30,31] and the area loss evaluation method [32,33]. Although the former is widely applied, its decision-making process has been controversial due to reliance on researchers’ subjective expertise. While the latter can quantify area loss during resampling processes, it tends to overlook a comprehensive representation of landscape characteristics. Subsequent studies often integrate both approaches to balance their respective advantages and disadvantages when determining appropriate granularity levels [34,35,36,37]. Based on clearly defined optimal granularity, most researchers further delineate the optimal extent of study areas by fitting semivariograms [38,39,40].

Research on influencing factors of cultivated land fragmentation has transitioned from early qualitative analysis to quantitative approaches. Commonly employed quantitative methods primarily focus on geographically weighted regression models, principal component analysis, and multiple linear regression [41]. Cultivated land fragmentation exhibits significant spatial heterogeneity and nonlinear characteristics, rendering traditional statistical and spatial regression methods based on linear relationship assumptions inapplicable. In contrast, the geographical detector method, which offers advantages for addressing multivariate nonlinear problems, has been widely adopted by scholars in subsequent studies [42].

Based on the existing research, scholars have rarely considered both analytical scale and amplitude when selecting landscape indices to characterize cultivated land fragmentation. Additionally, most studies focus on specific provinces or municipalities, with limited investigation into cultivated land fragmentation in key regions. The middle and lower reaches of the Yangtze River Basin, as one of China’s three major grain-producing regions, hold an irreplaceable strategic position in the national food security framework and bear significant responsibility for ensuring it [43]. In recent years, urbanization has led to frequent instances of “occupying high-quality farmland while compensating with low-quality land,” resulting in severe degradation of cultivated land quality [44]. Cultivated lands now exhibit varying degrees of fragmentation, which seriously hinders the advancement of agricultural production scaling and mechanization levels, reduces grain production efficiency, and increases farming costs [45]. Therefore, systematically researching cultivated land fragmentation in this critical grain-producing region demonstrates an irreplaceable practical necessity for ensuring food security and promoting appropriately scaled agricultural operations. Analyzing the spatial patterns and driving mechanisms of cultivated land landscape fragmentation in this region can provide scientific references for comprehensive land consolidation and sustainable utilization. Evaluating cultivated land fragmentation holds substantial research significance for improving land quality and overcoming agricultural development barriers in this vital grain-producing area, while also contributing to achieving rural revitalization goals characterized by thriving industries and affluent livelihoods.

This study, based on land-use data and employing a combination of Geographic Information System (GIS) technology and landscape index methods, investigates the characteristics of cultivated land fragmentation in the main grain-producing areas of the middle and lower reaches of the Yangtze River. The research aims to address the following scientific areas: (1) scientifically determining the optimal granularity and extent for cultivated land fragmentation analysis in the study area; (2) systematically exploring the spatial distribution patterns of cultivated land fragmentation in the main grain-producing areas of the middle and lower reaches of the Yangtze River based on the optimal granularity and reasonable extent; (3) using the geographical detector method to explore the determinant factors influencing the spatial distribution of cultivated land fragmentation in the study area.

2. Materials and Methods

2.1. Study Area and Data Sources and Preprocessing

2.1.1. Study Area

The major grain-producing region in the middle and lower reaches of the Yangtze River includes 71 prefecture-level cities across five provinces: Jiangxi, Hunan, Hubei, Anhui, and Jiangsu. This region covers a total area of approximately 806,800 km², bordered by the sea to the east, extending westward to the Three Gorges, northward to the Huaiyang Mountains, and southward to the Jiangnan Hills. It is located between 24° N and 35° N latitude, and 108° E to 122° E longitude (Figure 1). The region experiences a subtropical monsoon climate characterized by distinct seasons, abundant solar radiation and thermal resources, prolonged growing seasons, extended frost-free periods, annual precipitation exceeding 700 mm, and an average annual temperature above 3 °C.

The region is rich in cultivated land, accounting for approximately 7% of China’s total arable land. The elevation ranges from 0 to 3061 m above sea level, with most areas being relatively flat. The overall topography slopes from southwest to northeast, with higher elevations in the southwest and lower elevations in the northeast. The Yangtze River serves as a critical waterway, flowing eastward across the entire region. It not only provides abundant irrigation water for agricultural production but also plays a crucial role in navigation and aquaculture. The major grain-producing areas in the middle and lower reaches of the Yangtze River feature diverse topography, encompassing plains, hills, basins, plateaus, and mountains. The unique geographical environment provides natural advantages for agricultural production, establishing this region as a core production area for grains, oil crops, and cotton in China. In 2023, the grain output reached 159.9181 million tons, with an agricultural gross output value of approximately CNY 1941.62 billion, playing a crucial role in ensuring national food security.

2.1.2. Data Sources and Processing

The multi-source data employed in this study encompass the administrative boundaries of the major grain-producing areas in the middle and lower reaches of the Yangtze River, land-use data, monthly precipitation data, monthly temperature data, Digital Elevation Model (DEM) data, slope data, soil type data, population density data, nighttime light data, NDVI data, river and road network data, and statistical data including grain yield and total agricultural output value. The administrative boundary data were obtained from the National Geographic Information Public Service Platform (https://www.tianditu.gov.cn/). Land-use data were derived from the CLCD land cover dataset, publicly released by Wuhan University, with a spatial resolution of 30 m [46] (https://doi.org/10.5281/zenodo.12779975). Meteorological data, including temperature and precipitation, were sourced from the National Earth System Science Data Center (https://www.geodata.cn/).

The DEM and soil type data were obtained from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/), and the slope was derived from surface analysis. Population density data originated from the Oak Ridge National Laboratory (ORNL) of the U.S. Department of Energy (https://landscan.ornl.gov/). Nighttime light data were sourced from the publicly available dataset by Chen Zuoqi et al., with a spatial resolution of 500 m [47]. NDVI data were acquired from NASA’s regularly updated MOD13A3 dataset (https://search.earthdata.nasa.gov/). River and road data were extracted from Open Street Map (https://www.openstreetmap.org/). Food production and total agricultural output value were collected from the website of the National Bureau of Statistics (https://www.stats.gov.cn). All datasets were projected into the WGS_1984 coordinate system using the ArcGIS 10.7 platform for technical preprocessing.

2.2. Methods

2.2.1. Evaluation Method for Habitat Fragmentation Degree

Landscape indices can quantitatively reflect the overall structure and spatial distribution of landscapes and have been widely applied in studies of cultivated land fragmentation. However, single landscape indices are insufficient to comprehensively characterize the fragmentation status of cultivated land within the study area. To holistically capture the landscape features of cultivated land fragmentation while minimizing redundancy in pattern information, this study draws on previous research [27,48,49,50,51]. It integrates the distribution characteristics of cultivated land in the study area by considering three key aspects: area, shape, and the degree of aggregation of cultivated land patches. Based on these considerations, ten landscape indices at the class level (

Table 1. Definition of the cultivated land fragmentation index.

Landscape Index	Description of Indicators	Index Direction
Class area (CA)	The measurement of cultivated land area size indicates that larger values correspond to lower degrees of cultivated land fragmentation.	-
Mean perimeter–area ratio (PARA_MN)	Characterize the distribution patterns of patch shapes to reveal the impacts of human activities on landscape patterns. The weaker the patch aggregation, the more complex the shape tends to be, and the higher the fragmentation level.	+
Patch density (PD)	Characterizes the number of cultivated land patches within a certain range; higher values indicate greater fragmentation of cultivated land.	+
Landscape division index (DIVISION)	The degree of separation between cultivated land patches indicates that higher values correspond to greater fragmentation of cultivated land.	+
Patch cohesion index (COHESION)	Represents the degree of physical connectivity between cultivated land patches; higher values indicate better connectivity of cultivated land.	-
Aggregation index (AI)	The description of cultivated land landscape aggregation degree indicates that higher values represent greater aggregation.	-
Mean patch Area (AREA_MN)	The arithmetic mean of cultivated land patch areas, where smaller values indicate higher fragmentation levels.	-
Effective mesh size (MESH)	The ratio of the sum of squared patch areas to the total landscape area indicates that a higher value corresponds to a lower degree of cultivated land fragmentation.	-
Largest patch index (LPI)	This indicator represents the proportion of landscape area occupied by the largest cultivated land patch; a higher value indicates lower fragmentation of cultivated land.	-
Landscape shape index (LSI)	This indicator describes the complexity of cultivated land patch shapes. Higher values indicate more irregular shapes and greater fragmentation of cultivated land patches.	+

) were selected for analysis. The Fragstats 4.3 × 64 software was used to calculate class-level landscape indices for cultivated land, which served as a foundation for computing the comprehensive fragmentation index for cultivated land within the study area.

• Selection of Spatial Scale for Cultivated Land Fragmentation

Spatial scale is primarily characterized by grain and extent, where grain is typically defined as the spatial resolution of data, corresponding to pixel size. Excessively fine grain significantly increases data volume by computing ecologically insignificant patches and introducing redundant information, whereas overly coarse grain may result in loss of critical details. Selecting an appropriate grain size is essential for accurately analyzing cultivated land fragmentation patterns within the study area.

To determine the optimal grain size for cultivated land fragmentation analysis in the study area, a response curve of landscape indices to varying grain sizes was generated from the 2023 land-use data using an upward-scaling approach. This method involves “coarse-graining” the land-use data by increasing the raster cell size and reducing the resolution, thereby selecting an appropriate analytical grain size. Within the suitable grain size range, the trend of the landscape index response curve remains relatively stable, with little fluctuation [52]. The resampling tool in ArcGIS 10.7 was employed with an initial resolution of 30 m and a step size of 30 m, progressively increasing to a maximum resolution of 900 m, resulting in 30 raster datasets representing different grain sizes. Landscape metrics at each grain level were calculated using the standard methods in Fragstats 4.3 × 64 software. Since landscape metrics typically vary with changing grain sizes, a coefficient of variation (CV), calculated as the ratio of the standard deviation to the mean, was used to identify metrics sensitive to grain-size changes, as reported in previous studies. The selected sensitive landscape metrics were then used to construct response curves for determining appropriate analysis grain sizes, with selection criteria detailed in

Table 2. Criteria for determining the sensitivity of landscape indices to grain size changes.

Coefficient of variation	<0.01	0.01–0.1	0.1–0.5	0.5–1
Sensitivity	Insensitive	Low sensitivity	Moderately sensitive	Highly sensitive

[53].

During resampling, information loss is inevitable. To minimize accuracy loss as much as possible, an area-loss evaluation model was employed to assess area loss within the appropriate scale range. The total landscape area loss in the study area at different granularity levels served as the basis for determining a suitable granularity partition [54]. The optimal granularity for cultivated land fragmentation analysis was selected based on minimal area loss, with its calculation formula expressed as follows:

$$L=(A-B)/A\times 100\%$$

(1)

In Equation (1): $A$ represents the area of cultivated land at a certain granularity after resampling, $B$ denotes the area of cultivated land under the original granularity, and $L$ signifies the relative value of area loss before and after the granularity conversion.

Grain size can be conceptualized as the spatial extent of the study area. When grain size is too small, landscape indices tend to focus on local features and fail to capture overall characteristics; conversely, excessively large grain sizes significantly increase computational demands by smoothing local information, thereby obscuring landscape details.

To determine the appropriate scale, landscape indices were calculated for different window sizes using the moving-window method in Fragstats 4.3 × 64, based on the optimal grain size. The initial window size needed to be larger than the grain size to avoid local noise dominance. After multiple trials, an initial window size which was five times the optimal grain size was selected, with subsequent increments at odd multiples of the optimal grain size: this generated 15 raster maps of landscape indices under varying window sizes. Subsequently, a fishnet tool was created in ArcGIS 10.7 according to the study area’s dimensions, with grid points spaced at multiples of the grain size. Landscape index values were extracted from raster data to these points using the multi-value extraction tool, resulting in 2982 valid sampling points after removing null values. These sample data were imported into GS+ 9 software to fit spherical semivariogram models, recording the nugget-to-sill ratio $(C/(C0 + C\left)\right)$ for each landscape index. Variation curves of $C/(C0 + C)$ ratios across different window sizes were plotted to identify the optimal analytical scale. When most landscape indices reached stable ratios between the nugget and sill parameters, indicating stabilized spatial structures, the corresponding scale was determined to be the suitable analytical extent for landscape pattern analysis.

The semivariogram primarily characterizes spatial distribution patterns of ecological elements by quantifying the relationship between the degree of spatial attribute variation and the distance separating two points. This method is commonly employed to identify characteristic scales of landscape spatial structures. By observing periodic characteristics of parameter variations across scales, the optimal scale for landscape index analysis can be determined [55]. The calculation formula is as follows:

$$Y\left(h\right)=\left({\displaystyle \frac{1}{2N\left(h\right)}}\right)+{\sum }_{i=1}^{N\left(h\right)}{[Z\left(x_i\right)-Z\left(x_i+h\right)]}^2$$

(2)

In Equation (2), $Y\left(h\right)$ represents the semivariogram value, while $Z\left(x_i\right)$ and $Z\left(x_i+h\right)$ denote the landscape indices at sample points $x_i$ and $x_i+h,\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{l}\mathrm{y}$. The semivariogram primarily consists of several key parameters: nugget effect ($C_0)$, still ($C_0+C$), and range. Here we focus on observing the ratio between $C$ and ($C_0+C$) to determine the analytical scale of the landscape index.

2.

Calculation of Cultivated Land Fragmentation Index

To comprehensively characterize cultivated land fragmentation in the study area, this research selected 10 indicators related to cultivated land area, shape, and degree of aggregation to construct the Cultivated Land Fragmentation Index (CLFI). The entropy weight method was employed to determine indicator weights, which has the advantage of preserving the objectivity of the original data and reducing the inherent human errors in subjective weighting approaches. Its core principle is to assign higher weights to grid cells with greater dispersion by precisely measuring indicator dispersion levels. The specific steps are as follows:

(1) Standardization of the evaluation index.

Normalize using the raster calculator by applying the standardization formula to calculate $Z_{ij}^{\prime }$:

$$\mathrm{Positive} \mathrm{indicator}: Z_{ij}^{\prime }=\left(0.999{\displaystyle \frac{Z_{ij}-min{Z}_{ij}}{{maxZ}_{ij}-{minZ}_{ij}}}\right)+0.001$$

(3)

$$\mathrm{Negative} \mathrm{indicator}: Z_{ij}^{\prime }=\left(0.999{\displaystyle \frac{{maxZ}_{ij}-Z_{ij}}{{maxZ}_{ij}-{minZ}_{ij}}}\right)+0.001$$

(4)

In Equations (3) and (4), $Z_{ij}^{\prime }$ represents the standardized indicators, $Z_{ij}$ denotes the original indicators, $i$ corresponds to the evaluation objects, the number of grid cells, $(i=\mathrm{1,2},\cdots ,n)$, $j$ indicates the number of evaluation indicators ($j=\mathrm{1,2},\cdots ,m)$. ${maxZ}_{ij}$ is the maximum value calculated from the original data using a zonal statistics tool, while ${minZ}_{ij}$ represents the minimum value derived through zonal statistics. The translation by 0.001 is applied to prevent division by zero in subsequent calculations.

(2) Calculate the proportion of the characteristic value.

Using the raster calculator tool, calculate the proportion of feature values.

$$P_{ij}={\displaystyle \frac{{Z_{ij}}^{\prime }}{\sum _{i=1}^n{Z}_{ij}^{\prime }}}$$

(5)

In Equation (5), $P_{ij}$ represents the proportion of the eigenvalue for the $i$ raster cell under the $j$ indicator.

(3) Calculate the information entropy of the j-th indicator.

$$e_j=-K{\sum }_{i=1}^n{P}_{ij}{lnP}_{ij}$$

(6)

In Equation (6), $e_j$ represents the information entropy, $0{\le e}_j\le 1$, and $K$ denotes a positive constant. Here, we set $K=-1/lnn$.

(4) Calculate the index weight.

Calculate the weight of each indicator using formulas within the raster calculator:

$$W_j={\displaystyle \frac{\left(1-e_j\right)}{{\sum }_{j=1}^m\left(1-e_j\right)}}$$

(7)

In Equation (7), $W_j$ denotes the weight of the j-th indicator, where $j=(\mathrm{1,2},\cdots ,m)$.

The fragmentation evaluation indicators were calculated using the aforementioned computational steps (Appendix A).

The final comprehensive index calculation formula for cultivated land fragmentation is as follows:

$$CLFI={\sum }_{j=1}^m{W}_j{Z}_{ij}^{\prime }$$

(8)

In Equation (8), $CLFI$ represents the comprehensive evaluation index of the assessment indicators, $w_j$ denotes the weight of the j-th indicator, and $Z_{ij}^{\prime }$ signifies the standardized indicator.

3.

Fragmentation Difference Analysis

To clarify the spatial differentiation characteristics of cultivated land fragmentation across different regions, sample points were extracted from raster layers generated by the comprehensive index of cultivated land fragmentation at both provincial and terrain-type levels. Based on these sampling data, boxplots were created in Origin 2024 to visually illustrate the distribution patterns. One-way analysis of variance (ANOVA) followed by Duncan’s multiple range test was conducted using IBM SPSS 27.0 (p < 0.05) to test intergroup differences in fragmentation values among provinces within the study area and across different topographic units.

2.2.2. Analysis Method of Fragmentation Influencing Factors

This study selected 11 factors (

Table 3. Influencing factors of cultivated land fragmentation.

Influencing Factor Category	Impact Factor	Abbreviations
	Elevation	DEM
	Vegetation cover	NDVI
	Slope	SLOP
Natural factors	Precipitation	PRE
	Temperature	TEM
	Soil type	SOIT
	Distance from the river	DTR
Socioeconomic factors	Distance from the railway	DTRL
	Distance from the highway	DTH
	Population density	POP
	Night lights	NL

) from natural and socioeconomic aspects based on previous research. It employed the Geodetector model to examine the factors influencing cultivated land fragmentation in the major grain-producing provinces of the middle and lower reaches of the Yangtze River [27,42]. This model is primarily used to analyze spatial heterogeneity and identify driving factors. However, traditional geographic detectors require manual categorization of variables before analysis, which introduces strong subjectivity [56]. To address this limitation, this study implemented optimal parameter processing using the Optimal Parameter Geodetector (OPGD) model (GD package) developed by Song et al. [57] in R. Three discretization methods (geometric interval method, quantile classification method, and equal interval method) were applied to transform continuous variables into categorical data, with classification numbers ranging from 4 to 10 categories.

Compared to conventional geographical detectors, the optimal parameter geographical detector automatically determines the optimal discretization method and the number of categories for each variable. Based on these optimal parameters, it employs factor detection and interaction detection to reveal the influencing factors of cultivated land fragmentation in the main grain-producing areas of the middle and lower reaches of the Yangtze River. In the geographical detector framework, the q statistic quantifies the explanatory power of each influencing factor with respect to the dependent variable.

$$q=1-{\displaystyle \frac{\sum _{h=1}^L{N}_h{{\sigma }^2}_h}{N{\sigma }^2}}$$

(9)

In Equation (9), the value range of $q$ is 0–1; values closer to 1 indicate a stronger influence, while those closer to 0 represent a weaker influence. $L$ denotes the number of layers for dependent variables and influencing factors, $N$ represents the unit count, while ${{\sigma }^2}_h$ and ${\sigma }^2$ correspond to the variance of dependent variable values at layer h and the global variance across all regions, respectively.

To further analyze the mechanisms through which various influencing factors affect cultivated land fragmentation in the study area, and to identify the impacts of these factors under their combined effects on cultivated land fragmentation across provinces in the major grain-producing regions of the middle and lower reaches of the Yangtze River, this study employs the interactive detection method of geographical detectors (

Table 4. Types of interaction detection.

Type	Description
Nonlinear weaken	$q(X1\cap X2)<Min\left\{q\left(X1\right), q\left(X2\right)\right\}$
Uni-weaken	$Min\left\{q\right(X1), q(X2\left)\right\}<q(X1\cap X2)<Max\left\{q\right(X1), q(X2\left)\right\}$
Bi-enhance	$q(X1\cap X2)>Max\left\{q\right(X1), q(X2\left)\right\}$
Independent	$q(X1\cap X2)=q\left(X1\right)+q\left(X2\right)$
Nonlinear enhance	$q(X1\cap X2)>q\left(X1\right)+q\left(X2\right)$

). This approach analyzes whether the interactions between paired factors strengthen or weaken their influence on cultivated land fragmentation within the study area.

3. Results

3.1. Evaluation of the Degree of Cultivated Land Fragmentation

Based on the magnitude of coefficients of variation (

Table 5. Coefficient of variation for each landscape index.

Landscape Index	Coefficient of Variation
Class area (CA)	0.001
Mean perimeter–area ratio (PARA_MN)	1.223
Patch density (PD)	1.700
Landscape division index (DIVISION)	0.035
Patch cohesion index (COHESION)	0.001
Aggregation index (AI)	0.063
Mean patch area (AREA_MN)	0.788
Effective mesh size (MESH)	0.426
Largest patch index (LPI)	0.248
Landscape shape index (LSI)	0.687

), the following landscape metrics were selected: mean perimeter area ratio (PARA_MN), patch density index (PD), mean patch area (AREA_MN), largest patch index (LPI), landscape shape index (LSI), and effective mesh size index (MESH) (all with coefficients of variation > 0.2). These metrics demonstrate high sensitivity to changes in spatial grain size and effectively capture critical characteristics of cultivated land landscape pattern dynamics across varying spatial resolutions. Using these six selected landscape metrics, a gradient in grain size ranging from 30 to 900 m was established, and corresponding grain-size response curves were generated (Figure 2). This methodological framework provides a scientific basis for determining optimal analytical grain sizes in subsequent investigations.

As shown in Figure 2, the landscape index values exhibit regular variation with changes in spatial grain size and show distinct trends. Overall, the grain size effect curves of landscape indices can be summarized into two categories: The first category shows response curves that linear functions can fit with an overall increasing trend (Figure 2a–c). These curves exhibit relatively gentle changes between 150 m and 510 m, during which landscape indices do not fluctuate dramatically with grain-size variation, indicating relatively stable land-use boundaries and landscape structures within this grain-size range. The second category displays approximately power-law-decreasing trends (Figure 2d–f), characterized by a rapid decline between 30 m and 330 m, followed by gradual stabilization after 330 m, suggesting relative stability of landscape structures beyond this threshold. Considering the study area scope and area-loss analysis (Figure 2g), cultivated land area loss remains minimal at 330 m, with high data fidelity. Additionally, all landscape indices show stable intervals without significant fluctuations near 330 m. Therefore, 330 m is selected as the optimal analytical grain size for assessing farmland fragmentation in the main grain-producing areas of the middle and lower Yangtze River region, effectively balancing the reduction in cultivated land area loss while avoiding computational redundancy.

A smaller ratio of $C$ to ($C_0+C$) indicates that spatial heterogeneity caused by random factors dominates, while a larger ratio suggests that structural factors play a primary role in determining the degree of spatial heterogeneity. As shown in Figure 3, with increasing moving window size, the ratio of nugget-to-sill for each index increases. When this ratio reaches a certain threshold, the increase in the $C$/($C_0+C$) ratio decelerates. According to Figure 3, when the window size exceeds 8910 m, the nugget-to-sill ratios of most landscape indices stabilize, indicating that the spatial structure has reached a steady state. Considering that excessively large window sizes may lead to significant information loss, a moving window size of 8910 m was selected as the analytical scale for investigating cultivated land fragmentation within the study area.

Based on the optimal granularity and magnitude analyses, the Cultivated Land Fragmentation Index (CLFI) for the study area was calculated using Equations (3)–(9), and a spatial distribution map of cultivated land fragmentation was generated (Figure 4). Lower CLFI values indicate lower fragmentation levels within the region, with cultivated land more concentrated and contiguous; conversely, higher CLFI values indicate a dispersed distribution of cultivated land and increased fragmentation intensity. According to the cultivated land fragmentation map, significant spatial heterogeneity exists in the grain-producing areas of the middle and lower reaches of the Yangtze River Basin. The fragmentation index ranges from 0.001 to 0.973 across the study area, with most cultivated lands exhibiting low fragmentation values, particularly in the central plain and northeastern regions. This indicates that these areas maintain a relatively concentrated land distribution pattern, characterized by contiguous cultivated land patches and low fragmentation. Conversely, high fragmentation indices are predominantly observed in western and peripheral zones, suggesting that these regions likely face more severe land division issues, with cultivated lands distributed in scattered, fragmented patterns.

3.2. Regional Differences in Cultivated Land Fragmentation Characteristics

3.2.1. Provincial Differences in Cultivated Land Fragmentation

From the perspective of average cultivated land fragmentation across provinces in the study area (Figure 5), Jiangxi Province exhibits the highest degree of fragmentation, while Jiangsu Province demonstrates the lowest level. Specifically, the descending order is Jiangxi Province (0.255) > Hunan Province (0.249) > Hubei Province (0.204) > Anhui Province (0.166) > Jiangsu Province (0.146). In terms of box length, Hubei Province has the longest boxplot, indicating significant internal variation in cultivated land fragmentation values with distinct high-fragmentation and low-fragmentation zones. Conversely, Jiangsu Province shows the shortest boxplot length, suggesting a more concentrated distribution of fragmentation levels and minimal regional disparities.

The results of one-way ANOVA indicate that, except for the absence of significant difference in cultivated land fragmentation between Hunan and Jiangxi Provinces, all other provinces exhibit significant variations in their degrees of cultivated land fragmentation.

Among these cases, the western regions of Jiangxi Province demonstrate high fragmentation due to substantial topographic relief in surrounding areas, which prevents the contiguous distribution of cultivated land. Low fragmentation values within the province are primarily concentrated around Nanchang City and the central regions. In the mountainous areas of western Hunan Province, cultivated land is divided into numerous small plots, making it difficult to form large-scale, contiguous cultivated land and leading to a high fragmentation index. The northeastern and central regions of the province have relatively low fragmentation values, concentrated across the province.

In Hubei Province, high fragmentation values of cultivated land are primarily distributed in the western regions, characterized by significant topographic relief. In contrast, low fragmentation values occur mainly in the central and southern areas, where cultivated land exhibits better contiguity and clustering. For Anhui Province, low fragmentation indices are predominantly found in the northern and central plains with gentle topography, whereas high fragmentation characterizes the western and southern regions. This phenomenon can be attributed to the dispersed agricultural production patterns in mountainous zones, where small cultivated land patches are more susceptible to natural environmental influences, ultimately leading to increased land fragmentation. The Cultivated Land Fragmentation Index in Jiangsu Province is predominantly characterized by low fragmentation, with good contiguity and clustering of arable land. High fragmentation values occur only in urban construction areas and proximal zones of rivers. Overall, the degree of cultivated land fragmentation across the study region’s provinces shows significant regional heterogeneity.

3.2.2. Differences in Cultivated Land Fragmentation Across Different Topographic Regions

From the perspective of mean values (horizontal lines in the boxplot), the fragmentation index shows a spatial gradient. The highest degree of cultivated land fragmentation occurs on the Guizhou Plateau, followed by the Hanzhong Basin, which also shows relatively high fragmentation. The Sichuan Basin shows moderate fragmentation, while the Nanling Mountains and Jiangnan Hills exhibit relatively low fragmentation. The North China Plain displays the lowest fragmentation level. Specifically, the descending order of mean fragmentation indices is as follows: Guizhou Plateau (0.503) > Hanzhong Basin (0.398) > Sichuan Basin (0.337) > Nanling Mountains (0.306) > Jiangnan Hills (0.232) > Huainan and Middle-Lower Yangtze Plain (0.163) > Central Shandong Uplands (0.148) > North China Plain (0.125).

The results of one-way ANOVA further confirmed the significance of differences among various geomorphological regions, indicating that there are significant variations in cultivated land fragmentation levels across different terrain zones (Figure 6). The analysis of fragmentation disparities revealed that, except for the Huainan region and the middle-lower Yangtze Plain where no significant difference was observed compared to the central Shandong hilly region, all other geomorphological zones exhibited statistically significant differences in fragmentation degrees. This demonstrates distinct differentiation in cultivated land fragmentation under varying topographic conditions. Specifically, plateau and basin areas showed significantly higher fragmentation levels than hilly and plain regions due to pronounced terrain undulations and scattered distribution of cultivated land. In contrast, plain areas displayed the lowest fragmentation levels owing to flat topography and contiguous cultivation patterns. These findings highlight the critical role of physical geography in shaping regional spatial patterns of cultivated land.

3.3. Analysis of Influencing Factors on Cultivated Land Fragmentation

3.3.1. Factors Influencing Cultivated Land Fragmentation in the Study Area

The factor detector in the geodetector was employed to evaluate the impact of various factors on cultivated land fragmentation across the entire study area. All selected factors passed the significance test (p < 0.01), indicating that these factors exert significant influences on the spatial distribution of cultivated land fragmentation within the study region. As illustrated in Figure 7, the explanatory power of each factor follows the descending order: slope > elevation > NDVI > population density > soil type > annual precipitation > annual temperature > nighttime light intensity > distance to railways > distance to highways > distance to rivers. Notably, slope and elevation exhibit the highest q-values, demonstrating their predominant explanatory power regarding the spatial pattern of cultivated land fragmentation. This suggests that topographic conditions primarily determine the spatial distribution pattern of cultivated land fragmentation in the study area. The undulation of terrain and variations in altitude fundamentally shape the contiguity and degree of fragmentation of cultivated land, serving as foundational drivers of spatial heterogeneity in regional cultivated land fragmentation.

3.3.2. Factors Influencing Cultivated Land Fragmentation in Different Provinces

The results of the geodetector factor analysis are presented in Figure 8. All factors passed the significance test (p < 0.01), indicating that the selected factors exert significant influences on the spatial distribution of cultivated land fragmentation within the study area. The explanatory power of influencing factors for cultivated land fragmentation shows notable provincial differences. Specifically, the spatial distribution of cultivated land fragmentation in Anhui and Hubei Provinces is primarily influenced by elevation and slope, suggesting that topographic relief determines the contiguity of cultivated land patches. In Hunan and Jiangxi Provinces, NDVI emerges as the dominant influencing factor, reflecting the correlation between vegetation coverage and cultivated land distribution. The interlacing of densely vegetated areas (e.g., forests) and cultivated lands intensifies the fragmentation of cultivated land. Notably, in economically developed Jiangsu Province, nighttime light data exhibit stronger explanatory power than other factors, indicating that urbanization intensity and human economic activities are the primary drivers of cultivated land fragmentation in this region.

The interaction detection results are presented in Figure 9, where pairwise interactions between factors across provinces exhibit two distinct characteristics: bifactor enhancement and nonlinear enhancement. This indicates that the explanatory power of any two interacting factors for the spatial distribution of cultivated land fragmentation across different provinces is significantly greater than that of individual factors. As shown in the interaction detection results for Anhui Province (Figure 9a), the strongest interactive effect occurs between elevation and nighttime light, with a maximum influence value of 0.428. The interaction detection results for Hubei and Hunan Provinces (Figure 9b,c) show that the highest q-values of interactions between NDVI and slope are 0.580 and 0.515, respectively. In Jiangsu Province, all pairwise factor interactions (Figure 9d) demonstrate that the effects of nighttime light interacting with other factors exceed 0.340, indicating these combined factors with nighttime light exert a significant influence on the spatial distribution of cultivated land fragmentation in Jiangsu. The analysis of pairwise interactions in Jiangxi Province (Figure 9e) reveals that elevation and NDVI exhibit the highest q-value among all interactive factors, reaching 0.537. The interaction-detection results across provinces collectively indicate that cultivated land fragmentation results from multiple interacting factors rather than single-factor effects, with notable provincial differences in the patterns of these interactions.

4. Discussion

To determine the appropriate analytical grain size for the study area, this research constructed a grain-size range of 30–900 m and plotted grain-size response curves. Compared to previous studies, this extended range more comprehensively captures the complete response process of landscape indices across varying grain sizes, thereby avoiding the omission of optimal scales caused by overly narrow grain-size ranges. By identifying grain-size intervals with relatively stable landscape indices, integrating an area-loss evaluation model, and considering the actual spatial extent of the study region, 330 m was ultimately determined to be the optimal analytical grain size for characterizing farmland fragmentation in the main grain-producing areas of the middle and lower Yangtze River Basin. Based on this optimal grain size determination, the best analytical extent was further identified through semivariogram fitting. This approach demonstrates greater scientific rigor compared to Li Yunlu et al.’s method of deriving optimal extents by curve-fitting mean landscape index values. The advantage lies in the semivariogram’s capacity to simultaneously account for both spatial randomness and the autocorrelation of regionalized variables, enabling clear revelation of the spatial variation characteristics of landscape indices at different extents. Consequently, 8910 m was determined as the optimal analytical extent for the study region.

The spatial scales identified in this study differ significantly from those reported in other regions. For instance, Zhao Yu et al. determined the optimal grain size for analysis to be 90 m with an analytical extent of 3000 m in Shaanxi Province [27]. Li Yunlu et al. determined the optimal grain size to be 150 m and the optimal extent to be 600 m through their investigation in Guizhou Province [48]. These discrepancies primarily stem from heterogeneity across study areas in physical geography, ecosystem types, intensity of human activity disturbances, regional area dimensions, and the spatial distribution patterns of cultivated land. Both Shaanxi and Guizhou Provinces are characterized by mountainous and plateau terrain and relatively fragmented regional areas, where human disturbances show a localized distribution pattern. Consequently, smaller analytical grain sizes and extents are appropriate for accurately capturing details of localized farmland fragmentation.

In contrast, the main grain-producing regions in the middle and lower reaches of the Yangtze River cover extensive territories spanning multiple provinces, where cultivated land distribution exhibits both large-scale contiguity and localized fragmentation. Therefore, larger analytical grain sizes and extents become necessary to maintain regional integrity and data reliability simultaneously. This further confirms that the scale effects of landscape indices lack universal applicability across different geographical regions.

The results of this study indicate that the spatial distribution of cultivated land fragmentation exhibits a decreasing trend from west to east, suggesting that topographic conditions exert significant constraints on cultivated land contiguity. Areas with higher elevation and greater terrain undulation (such as western Jiangxi Province, western Hunan Province, and western Hubei Province) generally exhibit higher fragmentation values, which are closely associated with complex mountainous topography and the scarcity of arable land. In contrast, Jiangsu Province, characterized by flat terrain and well-developed agricultural infrastructure, exhibits a concentrated, contiguous distribution of cultivated land with lower fragmentation. This finding aligns with the conclusion of Chen et al. [51] that cultivated land landscape fragmentation in plain regions is generally lower than in plateau areas. Topographic conditions represent the core natural driving factor for cultivated land fragmentation: flat terrain provides inherent advantages for forming large-scale contiguous cultivated land, while complex terrain impedes its formation. These geographical characteristics constitute the common natural foundation for fragmentation patterns across different regions. Plain areas, due to their flat, open topography, have conditions favorable to large-scale agricultural production and land reclamation, thereby facilitating the formation of extensive, contiguous cultivated land parcels.

This study further reveals that the fragmentation index of cultivated land in the research area exhibits significant spatial differentiation across different provinces and under varying topographic conditions. Specifically, the degree of fragmentation in plateau and basin regions is significantly higher than that in hilly and plain areas, which aligns with Wang Xue et al.’s observation regarding notable regional disparities in Chinese farmland fragmentation [14]. The underlying cause lies in the regionalized characteristics of farmland fragmentation, driven by differential natural conditions and socioeconomic development across regions. Plateau and basin regions, characterized by complex terrain and scattered farmland distribution, combined with relatively lower socioeconomic development levels and underdeveloped agricultural infrastructure, face challenges in implementing large-scale cultivation, thereby maintaining higher fragmentation indices. In contrast, hilly and plain regions benefit from comparatively favorable topographical conditions, higher socioeconomic development levels, advanced agricultural scaling and intensification practices, and better farmland contiguity, collectively contributing to lower levels of fragmentation.

The research findings of this study further confirm that there are significant differences in the explanatory power of natural and socioeconomic factors across different regions. In provinces with pronounced topographic relief, such as Anhui, Hubei, Hunan, and Jiangxi, elevation and slope consistently rank among the top influencing factors, with q-values remaining high throughout the analysis. This phenomenon can be attributed to the dominant role of natural terrain conditions in driving land fragmentation in these areas: regions with higher elevations and steeper slopes pose greater challenges for selecting suitable agricultural zones. In contrast, Jiangsu Province, characterized by flat terrain with minimal topographic variation, exhibits a nocturnal light q-value of 0.340, the highest among all factors, indicating that socioeconomic development and human intervention are core regulatory forces driving land fragmentation in this region. The entirety of Jiangsu consists primarily of plains, where natural terrain exerts minimal fragmentation on cultivated land, as evidenced by an elevation q-value of 0.112. Here, fragmentation primarily originates from anthropogenic activities, with higher urbanization levels driving the degree of land demand for construction, causing the localized fragmentation of cultivated land. These results demonstrate that the formation mechanisms of cultivated land fragmentation exhibit distinct regional characteristics and complexity. This phenomenon further supports the view that cultivated land fragmentation results from the combined effects of physical geographic factors and human activities [8]. This conclusion aligns with the perspective proposed by Zeng Jinwei et al., who argued that cultivated land resources maintain spatial stability. At the same time, natural factors such as slope and elevation provide the foundational framework for the spatial distribution of cultivated land fragmentation, while human activities drive its expansion and evolution [7].

The analysis of inter-factor interactions reveals that the combined effects of multiple factors exhibit significantly higher explanatory power for cultivated land fragmentation than individual factors. This enhanced explanatory capacity stems from the fact that farmland fragmentation is not independently driven by a single factor, but rather results from the synergistic interaction between natural baseline conditions and human activity intensity. Topographic factors (slope, elevation) establish the fundamental spatial distribution patterns of farmland fragmentation, constituting foundational constraints and prerequisite conditions for its spatial differentiation. Meanwhile, anthropogenic factors, as reflected in nighttime light data, further drive the formation and evolution of fragmentation patterns beyond these topographic constraints. Consequently, the explanatory power of multi-factor interactions proves significantly superior to that of individual factors, indicating that these influences are interconnected rather than independent. Although changes in any single factor may alter the degree of farmland fragmentation, this complexity underscores the intricate nature of influencing mechanisms, a finding consistent with Liang Jialiang et al.’s research on influencing factors in the Huai River Basin of China [49].

5. Conclusions and Policy Recommendations

5.1. Conclusions

This study, based on 2023 land-use data, constructs a Cultivated Land Fragmentation Index (CLFI) by integrating the landscape index method with the entropy weight method, after determining the optimal granularity and amplitude. It systematically reveals the status of cultivated land fragmentation and its spatial differentiation characteristics in the main grain-producing areas of the middle and lower reaches of the Yangtze River. Using the optimal parameter geodetector, this research analyzes factors contributing to cultivated land fragmentation and arrives at the following conclusions:

1. Different landscape indices exhibit distinct variation characteristics and patterns in response to changes in grain size and extent, primarily manifesting as linear increasing trends and power-law-decreasing trends. The optimal grain size for cultivated land fragmentation analysis in the study area is 330 m, with an optimal extent of 8910 m.

2. The Cultivated Land Fragmentation Index in the study area ranges from 0.001 to 0.973, exhibiting significant spatial heterogeneity overall. Most regions demonstrate low fragmentation levels, particularly in the central plain and northeastern areas where cultivated land is concentrated and contiguous. High fragmentation values are primarily concentrated in western and peripheral regions, with a scattered distribution pattern. At the provincial scale, Jiangxi Province shows the highest fragmentation level (0.255), while Jiangsu Province records the lowest (0.146). Apart from Hunan and Jiangxi Provinces, inter-provincial differences are statistically significant. At the topographic zone level, except for the Huainan region, the middle and lower Yangtze Plain, and central Shandong hills, all other topographic zones exhibit significant variations in fragmentation degrees. Plateaus and basins, characterized by complex terrain and substantial elevation changes, exhibit significantly higher fragmentation levels than hilly and plain regions. A gradient in fragmentation intensity decreases from the Guizhou Plateau (0.503) to the North China Plain (0.125).

3. The optimal parameter geodetector results indicate that the primary controlling factors of cultivated land fragmentation vary across different provinces. In provinces with significant topographic relief (Anhui, Hubei, Hunan, Jiangxi), the spatial distribution of cultivated land fragmentation is predominantly influenced by natural factors, including elevation, slope, and NDVI. In contrast, in provinces characterized by flat terrain and high urbanization rates (such as Jiangsu), nocturnal light intensity emerges as the dominant socioeconomic factor driving cultivated land fragmentation. Furthermore, the combined effects of multiple factors enhance the explanatory power of the spatial patterns of cultivated land fragmentation, demonstrating that this phenomenon results from synergistic interactions among these factors.

5.2. Policy Recommendations

The degree of cultivated land fragmentation varies significantly across provinces in the study area, and the dominant influencing factors exhibit distinct spatial patterns. Differentiated cultivated land protection policies should be implemented to promote the contiguous consolidation and spatial optimization of fragmented cultivated lands in Anhui and Hubei Provinces, where fragmentation is primarily influenced by topographic factors such as elevation and slope. The degree of terrain undulation determines the contiguity of cultivated land. These regions should implement comprehensive slope-cultivated land rehabilitation and systematically convert steep-slope cultivated land by returning it to forest or grassland programs, thereby mitigating the disruptive effects of terrain undulation on cultivated land contiguity. Simultaneously, a one-size-fits-all approach characterized by “large-scale contiguous rehabilitation” should be abandoned in favor of small-scale contiguous plot strategies that minimize fragmentation. Engineering measures, including merging small plots into larger ones, transforming slopes into terraced fields, optimizing field ridges, and other techniques, should be implemented to consolidate scattered cultivated lands and enhance their concentration and contiguity. In addition, it is necessary to strictly regulate the development of cultivated land in topographically sensitive areas by delineating a red line for cultivated land protection under topographic constraints. This approach prohibits unauthorized reclamation of cultivated land in regions with high elevation and steep slopes, thereby preventing fragmentation of newly added cultivated land.

In Hunan and Jiangxi Provinces, where NDVI strongly influences cultivated land fragmentation, prevention strategies should integrate vegetation conservation with arable land-use. This requires a dual approach combining landscape pattern optimization with ecological improvement of cultivated lands to mitigate fragmentation caused by the interdigitated distribution of vegetation and cultivated lands. On the one hand, it is essential to establish a coordinated symbiosis between cultivated lands and vegetation by delineating appropriate ecological buffer zones in areas with concentrated, contiguous farmlands and standardizing boundary management between forests and cultivated lands to prevent disorderly intermingling. On the other hand, the ecological transformation of cultivated lands should be advanced by implementing ecological revetments and constructing cultivated land shelterbelts around agricultural areas, which not only maintain vegetation’s ecological functions but also reduce the fragmentation effects of forest vegetation on the contiguity of cultivated lands.

For Jiangsu Province, an economically developed region significantly impacted by human activities, urbanization expansion should be strictly controlled, and disturbances from anthropogenic economic activities should be minimized to prevent cultivated land fragmentation. The region must rigorously implement the policy of balancing cultivated land conversion with compensation measures, strengthen law enforcement to enforce ecological red lines and permanently designated essential cultivated land areas, and prevent construction land from encroaching on contiguous cultivated land. Strict control over the increase in urban construction land is required. On one hand, it is necessary to reinforce the binding constraints of territorial spatial planning, strictly adhere to the “Three Zones and Three Lines” regulatory framework, delineate clear boundaries for urban development and cultivated land protection, prohibit disorderly urban sprawl from occupying high-quality agricultural land, and safeguard against the fragmentation of cultivated land into scattered plots during the urbanization process.

5.3. Limitations and Future Directions

This study aimed to enhance the accuracy of cultivated land fragmentation research by determining the optimal grain size and extent for fragmentation analysis at the pixel scale within the main grain-producing areas of the middle and lower reaches of the Yangtze River. The fragmentation results were visualized using a moving window method. However, this research has certain limitations.

Due to practical constraints in acquiring large-scale data on land ownership attributes, the selection of indicators for characterizing cultivated land fragmentation focused exclusively on landscape metrics, thereby omitting fragmentation caused by land tenure patterns, potentially leading to an underestimation of actual fragmentation levels in the study area. Additionally, this study only examined the degree and spatial distribution of cultivated land fragmentation in 2023, whereas fragmentation is inherently a long-term dynamic process. Future research should incorporate multi-temporal data to explore its temporal evolution.

Regarding driving factors, cultivated land fragmentation results from complex interactions between natural and anthropogenic forces. Subsequent studies should integrate policy regulations and farmers’ cultivation preferences into the analytical framework. Moreover, the intricate interactions among different driving factors pose significant challenges for the comprehensive investigation of fragmentation. Although this study quantitatively assessed intensity differences in interactive effects among various influencing factors, methodological limitations prevented elucidation of the underlying driving mechanisms behind these interactions, necessitating further optimization and refinement in future research.