A Spatiotemporal Assessment of Cropland System Health in Xinjiang with an Improved VOR Framework

Abstract

Accurately identifying and comprehensively managing the health of cropland systems is crucial for maintaining national food security. In this study, a more suitable framework for evaluating the health status of cropland systems in arid areas was constructed, and a systematic diagnosis of the health status of a cropland system in Xinjiang was conducted by increasing cropland stress and extending the VOR model to the VOR-S framework. The principal driving factors and spatiotemporal heterogeneity of cropland system health were investigated by using geographic detectors and GTWR models. The results showed the following: (1) From 2001 to 2023, the health level of the cropland system in Xinjiang fluctuated and increased. The proportion of areas with higher health levels (health levels I and II) in the cropland system increased from 45.84% in 2001 to 50.80% in 2023. The overall environment of the cropland system thus improved. (2) From 2001 to 2023, in terms of stress on the cropland system in Xinjiang, the overall level of HAI (human activity intensity) exhibited an upward trend, while the overall SEI (soil erosion intensity) significantly decreased, and WEI (wind erosion intensity) remained relatively stable. (3) The explanatory power of driving factors for cropland system health is ranked by order of magnitude as follows: annual precipitation (0.641) > annual average temperature (0.630) > population density (0.619) > nighttime lighting (0.446) > slope (0.313) > altitude (0.267). In addition, the combination of climate and human activity factors plays a dominant role in the spatial differentiation of cropland system health. The research results can provide scientific reference for cropland protection policies in arid areas.

1. Introduction

As an important artificial ecosystem, croplands serve multiple purposes, such as providing production and living materials, maintaining soil and water, and conserving water sources, and are an important cornerstone for ensuring food security [1]. However, with the rapid development of the social economy and the continuous advancement of urbanization and industrialization, the intensity of the human utilization of cropland resources is increasing. Health problems in cropland systems, such as reductions in cropland resources, cropland fertility, and deterioration of the ecological environment, are becoming increasingly severe [2]. According to the Main Data Bulletin of the Third National Land Survey, the current cropland area in China is 1.28 × 108 hm², of which only 30% is high-quality cropland, and the reserve resources of cropland are relatively scarce [3]. It is estimated that the global population will exceed 9 billion by 2050 [4], and global food demand will also increase by nearly 70% [5]. The increasingly severe problems of cropland pollution, soil erosion, desertification, groundwater overexploitation, and declines in biodiversity pose severe challenges to food security [6,7,8,9,10]. To ensure the sustainable development of agriculture, it is necessary to analyze the overall health status of cropland systems, aiming to provide a scientific basis for achieving the sustainable utilization of cropland resources and optimizing cropland management plans [11].

The total cropland area in Xinjiang is 7.07 × 106 hm², ranking fifth in China; the per capita cropland area is 0.29 hm², which is significantly higher than the national average level (0.09 hm²) [12], thus making Xinjiang the province with the most potential for cropland development and increased grain production in China. It plays an important role in implementing the overall national security concept and maintaining national food security. However, Xinjiang is located in the arid northwest region of China, and its oasis-shaped cropland system reflects the typical characteristics of irrigated agriculture. Due to the fragile ecological background, it faces unfavorable climate conditions such as drought, frequent winds, and desertification, as well as prominent contradictions between urbanization expansion and cropland protection caused by economic development [13]. The utilization and protection of cropland have always faced significant pressure, posing great challenges to Xinjiang’s development as a strategic replacement area for grain as a national commodity. Therefore, it is crucial to explore the health issues of Xinjiang’s cropland system, to promote the rational utilization of cropland resources and ensure national food security.

In the 1940s, Aldo Leopold pioneered the theory of “land health”, emphasizing that land can maintain its basic functions under the influence of human development activities [11]. Inspired by this theory, scholars have gradually expanded their research fields to include cropland quality assessment, soil ecological health, ecosystem health, and cropland system health [14,15,16,17,18]. With the transformation of cropland protection from quantity to the “trinity” of quantity, quality, and ecology, people are becoming more aware of the close relationship between cropland health and sustainability. Some scholars believe that the health of cropland systems, as the cornerstone of national food security, can only be determined from the source by diagnosing the health of cropland, evaluating the reasons for unhealthy cropland, and implementing appropriate measures [19,20]. The precise identification and comprehensive management of the health status of cropland have become key to ensuring sustainable agricultural development. From a research scale perspective, scholars have conducted empirical studies on the health evaluation of cropland systems based on multiple scales such as the county, city, province, and watershed scales [17,18,19,21,22]. Regarding the research framework and methods, some researchers mainly use the vigor–organization–resilience (VOR) framework, the Pressure State Response (PSR) model, Soil Index Evaluation (SINDI), the Soil Health Index (SHI), etc., to evaluate the health status of cropland systems and optimize cropland management measures based on them [23,24,25,26]. Overall, research on the health of cropland systems has made some progress, but there are still some shortcomings. Firstly, the existing framework for evaluating the health of cropland systems still has room for improvement. Most studies focus on measuring the health level of the system itself [23], without considering the impact of human activities and natural pressures on the cropland system [27], or they emphasize the relationship between the environment and human society, mainly focusing on the pressure, state, and response of the system [24], without fully capturing the system’s own attributes [28]. There are relatively few studies that comprehensively consider both internal attributes and external pressures [16]. Secondly, most existing research has focused on the black soil areas in Northeast China, the Yellow River Basin, and the Yangtze River Basin, which have a long history of cultivation. There is relatively little research on cropland in the arid areas of Northwest China. As one of the few regions where cropland continues to expand, Xinjiang is an important strategic area for ensuring China’s food security. Understanding the health status of its cropland system can provide important references for governments at all levels and relevant departments to make decisions on cropland use and protection, implement prevention and control measures, and resolve risks.

The three-dimensional ecological functional framework of VOR (vigor–organization–resilience) is the core foundation for evaluating the health of cropland, but its drawback lies in its complete disconnection from external pressures. Therefore, this study incorporated the pressure dimension from the PSR model into the VOR model, preserving the depth of the VOR model’s understanding of ecological integrity. In addition, there are significant differences in the structure, function, and ecological issues faced by different types of regions in the system [29]. Using unified evaluation criteria to assess the health of cropland systems in different regions may lead to inaccurate evaluations [30]. Therefore, based on existing research, three typical stress indicators in arid regions were selected to quantify external stress (human activity intensity, soil erosion intensity, and wind erosion intensity), and a more comprehensive framework for assessing the health of cropland systems was established. On this basis, analyzing the spatiotemporal evolution characteristics of the health of Xinjiang’s cropland system from 2001 to 2023, and exploring the driving factors of the health evolution of Xinjiang’s cropland system, will help to comprehensively and accurately evaluate the overall and local health status of Xinjiang’s cropland, propose policy recommendations and strategies for sustainable development, and provide reference for the sustainable utilization and scientific management of other arid areas.

2. Materials and Methods

2.1. Study Area

Located in the arid northwestern interior of China, the Xinjiang Uygur Autonomous Region covers 1.665 × 10⁸ hm²—approximately one-sixth of the nation’s total land area [31]. Xinjiang is located in an arid/semi-arid region, with a temperate continental climate, strong evaporation, scarce precipitation, and frequent natural disasters such as sandstorms, salt alkali, floods, etc. The ecological environment is relatively fragile [32], but the region has long sunshine hours and sufficient light and heat, thus providing good natural conditions for the development of oasis agriculture. Affected by the landform of “three mountains and two basins”, the cropland is mainly distributed at the foot of the Tianshan Mountains, around the Tarim Basin and the Junggar Basin, and in the Ili River Valley. A total of 96% of the cropland is irrigated land, 3.15% is dry land, and 0.85% is paddy land [33]. An overview of the study area is shown in Figure 1.

2.2. Data Sources

This study utilized cropland data derived from Wuhan University’s CLCD (China Land Cover Dataset), extracting cropland portions at a 30-m resolution. DEM data was obtained from the Geographic Spatial Data Cloud (https://www.gscloud.cn, accessed on 13 October 2024), and it exhibits no significant temporal characteristics. Vegetation indices included annual mean EVI and NPP spatial distribution data (2001–2023) from MODIS products, which was preprocessed via Google Earth Engine with a 500 m spatial resolution. NDVI data originated from NASA EarthData (https://www.earthdata.nasa.gov, accessed on 5 November 2024). The remote sensing images for each target year underwent preprocessing steps such as cloud removal, stitching, cropping, and median synthesis. Administrative boundary data was acquired from the Standard Map Service of the National Bureau of Surveying, Mapping and Geoinformation (http://211.159.153.75, accessed on 10 October 2024) using the Chinese standard map [Approval No.: GS (2024)0650] without cartographic modifications. Climatic variables (precipitation, temperature) and nighttime light data were sourced from the National Earth System Science Data Center (http://www.geodata.cn, accessed on 12 November 2024). Soil texture data was provided by the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn, accessed on 12 November 2024). Population density data came from LandScan Global Population Database (https://landscan.ornl.gov, accessed on 15 November 2024). To ensure spatial consistency and comparability, the above data will be resampled to 500 m using bilinear interpolation (excluding the CLCD). Detailed parameters are listed in

Table 1. Data used in this study.

Data Type	Time	Spatial Resolution	Purpose of the Data	Data Sources
Land use data	2001–2023	30 m	Understand the distribution of cropland	Wuhan University’s CLCD data
Administrative division data	/	/	Determination of the extent of the study area	National Bureau of Surveying, Mapping and Geoinformation Standard Map Service Website
DEM	/	90 m	Measuring the topographic relief of cropland	https://www.gscloud.cn (accessed on 13 October 2024)
EVI	2001–2023	500 m	Quantify the vitality of cropland	MODIS Product Dataset
NPP	2001–2023	500 m	Quantify the vitality of cropland	MODIS Product Dataset
NDVI	2001–2023	500 m	Quantify the stress on the cropland system	https://www.earthdata.nasa.gov (accessed on 5 November 2024)
Precipitation data	2001–2023	1000 m	Quantitative driving factors	National Earth System Science Data Center
Temperature data	2001–2023	1000 m	Quantitative driving factors	National Earth System Science Data Center
Nighttime light data	2001–2023	500 m	Quantitative driving factors	National Earth System Science Data Center
Soil texture data	/	/	Quantify the stress on the cropland system	National Earth System Science Data Center
Population density data	2001–2023	1000 m	Quantitative driving factors	https://landscan.ornl.gov (accessed on 15 November 2024)

.

2.3. Research Methods

2.3.1. Improved VOR Model for Cropland System Health Assessment Framework

In recent years, the cropland in Xinjiang has shown an expansion trend, and most of the newly added cropland comes from projects such as wasteland development, saline alkali land remediation, and channel leveling. Such new croplands typically exhibit unstable quality [34]. Concurrently, urbanization and industrialization have intensified issues including non-grain production, non-agricultural land use, and cropland abandonment, thus posing significant threats to the health and sustainability of Xinjiang’s cropland system. In addition, there is significant spatial heterogeneity in the structural characteristics, functional capabilities, and ecological challenges of cropland in different geographical regions of Xinjiang [29], and it is difficult for traditional unified evaluation frameworks to accurately capture regional differences. Therefore, the cropland system health assessment framework needs to consider regional specific pressures and spatial heterogeneity, thus supplementing the shortcomings of existing models in characterizing the “pressure response” mechanism. In response to the limitations of the previous vigor–organization–resilience (VOR) model, which mainly focused on the internal state of the cropland system and did not fully integrate regional differentiated external pressures, in this study, the dimension of cropland system stress (CS) is added and the model is extended to the VOR-S framework (Figure 2). By incorporating the CS dimension, the model achieved a complete characterization of the “state (VOR)-stress (CS)” of the cropland system, which is more in line with the real scenario of cropland in Xinjiang.

Following established methodologies [16,35], equal weighting is assigned to vigor, organization, resilience, and stress components. The health of the cropland system is a comprehensive result of these four dimensions, and any weakness in any dimension will limit the overall health level. The multiplication form can amplify the short board effect, which is more in line with the reality of “short board constraints”. The cropland health index (CHI) is calculated as follows:

$$\mathrm{CHI} = \mathrm{CVI}\times \mathrm{COI}\times \mathrm{CRI}\times \mathrm{CSI}$$

(1)

In this formula, CVI, COI, CRI, and CSI represent the standardized comprehensive indices of cropland system vigor, organization, resilience, and stress, respectively. The evaluation results were classified into five levels using the natural breaks method.

Health evaluation indicators were categorized as positive or negative based on their directional relationship with health status: higher values of positive indicators correspond to better health, while negative indicators exhibit the inverse relationship. To eliminate dimensional differences among indicators, range normalization was applied to all data.

• Quantifying cropland system vigor

Cropland system vigor reflects metabolic capacity or primary productivity [30], which is typically measurable through aboveground vegetation growth status and biomass [36]. Remote sensing indices including the EVI and NPP are widely adopted in ecosystem research [37,38]. Compared with the NDVI, the EVI incorporates the blue band to better mitigate atmospheric scattering and soil background effects, thus overcoming the NDVI’s limitations in saturation-prone high-coverage areas and soil-sensitive low-coverage regions. Unlike GPP, which solely measures photosynthetic capacity, NPP more effectively characterizes ecosystem health by capturing plant growth, reproduction, and environmental responses. Therefore, the EVI and NPP were selected to quantify cropland system vigor. The annual average values of NPP and the EVI were normalized to [0, 100] and then added equally to represent the vigor of the cropland system. The larger the value, the higher the vigor.

• Quantifying cropland system organization

The stability of the organizational structure of the cropland system refers to the relatively orderly state of the cropland system in space. Based on existing research results [30,39], most studies employ 10 km × 10 km and 15 km × 15 km fishnet scales. However, this study area is much larger than those in previous research, and given Xinjiang’s vast area and fragmented cropland distribution, smaller fishnets may risk data redundancy while larger fishnets may cause information loss. This study utilized Fragstats 4.2.1 to divide the study area into 30 km × 30 km fishnets. The organization of the cropland system is mainly reflected in its regular shape, connectivity and the good connectivity of the fields. The key metrics included the following: the Perimeter–Area Ratio Mean (PARA_MN), which characterizes field shape regularity; the Euclidean Nearest Neighbor Distance Mean (ENN_MN) and Division Index (DIVISION), reflecting inter-patch connectivity; and the Aggregation Index (AI) and Cohesion Index (COHESION), measuring patch spatial aggregation [40]. Building upon prior research findings [18], an equal weighting of the five indicators from the above three levels was added to obtain the orderliness of the cropland system in terms of organization. The computational framework is defined as follows:

$$\mathrm{COI}={\displaystyle \frac{1}{3}}PARA\_MN+{\displaystyle \frac{1}{3}}\left(ENN\_MN+DIVISION\right)+{\displaystyle \frac{1}{3}}\left(AI+COHESION\right)$$

(2)

• Quantifying cropland system resilience

Cropland system resilience denotes the system’s capacity to maintain core functionality through self-regulation while resisting external natural variations and anthropogenic impacts [41]. This study posits that long-term stable or increasing productivity indicates robust system resilience. After calculating the coefficient of variation in NPP and the EVI, it is then used to characterize the long-term stability of cropland system productivity [42]. The specific calculation formula is as follows:

$$C_v={\displaystyle \frac{\sqrt{\frac{{\displaystyle \sum _{i=1}^n{(x_i-\overline{x})}^2}}{n}}}{\overline{x}}}$$

(3)

In this formula, C_v is the coefficient of variation of the variable; x is the actual value of the variable; $\overline{x}$ is the average value of the variable; and n is the number of variables.

To calculate the long-term trend in productivity changes, the mainstream Theil–Sen median Mann–Kendall calculation method is adopted. The specific formula is as follows:

$$Slope(\alpha )=Median({\displaystyle \frac{x_i-x_j}{t_i-t_j}})$$

(4)

In this formula, Slope(α) represents the trend in cropland productivity; x_i and x_j are the time series of cropland productivity, respectively; and t_i and t_j are sequential years of cropland productivity. When Slope(α) > 0, it indicates an increase in cropland productivity; when Slope(α) < 0, it indicates a decrease in cropland productivity.

Finally, following an established methodology [43], the coefficient of variation C_v, which represents the long-term stability of cropland productivity, and the trend value Slope(α), which represents the long-term trend in cropland productivity, were standardized and equally weighted to obtain the resilience of the cropland system.

• Quantifying cropland system pressure

Cropland system pressure denotes adverse impacts compromising system vigor, organization, and resilience, which ultimately drives productivity degradation. Based on existing research, pressure sources were selected from two categories: natural factors and human activities. Natural factors mainly include soil erosion intensity (SEI) and wind erosion intensity (WEI), while human activity intensity (HAI) includes nighttime light intensity (NLI) and population density (PD). The objective result of the entropy weight method is that the information contribution of the three indicators is almost the same, so equal weights are set for the three indicators. The calculation formula is as follows:

$$CSI={\displaystyle \frac{1}{3}}HAI+{\displaystyle \frac{1}{3}}SEI+{\displaystyle \frac{1}{3}}WEI$$

(5)

➀

Soil erosion intensity

Xinjiang’s arid environment subjects its croplands to a certain degree of soil erosion pressures. The Revised Universal Soil Loss Equation (RUSLE) was used in this study to quantify erosion severity [44]. The formula is as follows:

$$RUSLE=R\times K\times LS\times C\times P$$

(6)

In this formula, RUSLE is the annual average soil erosion index per unit area, measured in t/(hm²·a); R is the rainfall erosion factor, measured in MJ∙mm/(hm²·h·a); K is the soil erodibility factor, measured in t∙hm²∙h/(hm²∙MJ∙mm); LS is the terrain factor; C is the vegetation cover management factor; and P is the soil conservation factor.

$$R=1.735{\sum }_{i=1}^{12}\left({10}^{1.5lg\left({\displaystyle \frac{{P_i}^2}{P}}\right)-0.8188}\right)$$

(7)

In this equation, P_i represents the average monthly precipitation of the i-th month, and P represents the average annual precipitation of many years [45].

$$\begin{gathered} K_{EPIC}=\left(0.2+0.3\mathit{exp}\left(-0.0256\times {\displaystyle \frac{SAN}{1-\raisebox{1ex}{$SIL$}\!\left/ \!\raisebox{-1ex}{$100$}\right.}}\right)\right)\times {\left({\displaystyle \frac{SIL}{CLA+SIL}}\right)}^{0.3}\times \left(1-{\displaystyle \frac{0.25C}{C+exp\left(3.72-2.95C\right)}}\right) \\ \times \left(1-{\displaystyle \frac{0.7\left(1-SAN\right)}{\left(1-SAN\right)+exp\left(22.9\left(1-SAN\right)-5.51\right)}}\right) \end{gathered}$$

(8)

$$K=0.1317\times K_{EPIC}$$

(9)

In this formula, K_EPIC represents the K value expressed in US customary units, expressed as (ton·acre·hour)·(hundreds of acres·feet ton·inch)⁻¹, and K represents the converted K value in the International System of Units, expressed as (t·hm²·h)·(hm²·MJ·mm)⁻¹, using a conversion factor of 0.1317 [46]. SAN, SIL, CLA, and C correspond to the percentages of sand, silt, clay, and organic carbon content.

$$L={\left({\displaystyle \frac{\lambda }{22.13}}\right)}^{{\displaystyle \frac{\beta }{1+\beta }}}$$

(10)

$$\beta ={\displaystyle \frac{\raisebox{1ex}{$\mathit{sin}\theta $}\!\left/ \!\raisebox{-1ex}{$0.0896$}\right.}{3{\left(\mathit{sin}\theta \right)}^{0.8}+0.56}}$$

(11)

$$\begin{gathered} S=10.8\mathit{sin}\theta +0.03 \theta <5\% \\ S=16.8\mathit{sin}\theta -0.5 5\%<\theta <10\% \\ S=21.91\mathit{sin}\theta -0.96 \theta >10\% \end{gathered}$$

(12)

$$\lambda =l\times \mathit{cos}\theta$$

(13)

$$\mathit{sin}\theta =\mathit{sin}\left(slope\times 3.1415926/180\right)$$

(14)

Here, S is the slope steepness coefficient, and θ is the slope steepness (°). L is the slope length factor. λ is the slope length (m). The value 22.1 is the slope length of the standard. The parameter m varies with the change in slope θ.

$$FVC={\displaystyle \frac{NDVI-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}$$

(15)

$$\begin{gathered} C=1 FVC\le 0.01 \\ C=0.6508-0.34361\mathit{lg}\left(FVC\right) 0.01<FVC<0.783 \\ C=0 FVC\ge 0.783 \end{gathered}$$

(16)

The soil conservation factor (P) quantifies the impact of soil conservation measures on soil erosion. The selection of the P factor refers to previous research to determine the spatial distribution data of the P factor (

Table 2. Divide the P factor in study area.

Slope/(°)	1–8	8–16	16–25	25–30
p-Value	0.6	0.7	0.8	0.9

).

➁

Wind erosion intensity

The Xinjiang Uygur Autonomous Region in northwestern China ranks among the nation’s most severe wind erosion zones. This phenomenon exhibits pronounced spatial heterogeneity across the territory. To quantify cropland system wind erosion, this study employed the Revised Wind Erosion Equation (RWEQ) [47]:

Windproof and sand fixation amount

$$SR=S_p-S_L$$

(17)

Potential wind erosion amount

$$S_p={\displaystyle \frac{2Z}{S_p^2}}{Q}_{pMAX}{e}^{-(z/S_p{)}^2}$$

(18)

$$Q_{pMAX}=109.8[WF\times EF\times SCF\times K^{\prime }]$$

(19)

$$S_p=150.71(WF\times EF\times SCF\times K^{\prime }{)}^{-0.3711}$$

(20)

Actual wind erosion amount

$$S_L={\displaystyle \frac{2Z}{S^2}}{Q}_{MAX}{e}^{-({\displaystyle \frac{z}{s}}{)}^2}$$

(21)

$$S=150.71{(WF\times EF\times SCF\times K^{\prime }\times C)}^{-0.3711}$$

(22)

$$Q_{MAX}=109.8(WF\times EF\times SCF\times K^{\prime }\times C)$$

(23)

In these formulas, SR is the sand fixation amount, t/(km²∙a); S_p is the potential amount of wind erosion, t/(km²∙a); S_L is the actual amount of wind erosion, t/(km²∙a); Q_MAX is the maximum transfer amount, kg/m; Z is the maximum distance of wind erosion occurrence, m; WF is the climate factor, kg/m; K is the surface roughness factor; EF is the erodible factor of soil; SCF is the soil crust factor; and C is the vegetation coverage factor.

$$WF=wf\times {\displaystyle \frac{\rho }{g}}\times SW\times SD$$

(24)

$$WF=U_2\times {\left(U_2-U_1\right)}^2\times N$$

(25)

In those formulas, wf is the wind factor (m·s⁻³); ρ is the air density (kg·m⁻³); g is the acceleration due to gravity, with a value of 9.8 m·s⁻²; SW is the soil moisture factor; SD is the snow cover factor; U₂ is the wind speed at 2 m (m·s⁻¹); U₁ is the critical wind speed at 2 m (assuming 5 m·s⁻¹); and N is the number of days of the experiment.

$$EF=(29.09+0.31SAN+0.17SIL+0.33{\displaystyle \frac{SAN}{CLA}}-2.59OM-0.95{CaCO}_3)/100$$

(26)

In this formula, SAN is the soil sand content (%); SIL is the content of soil silt (%); CLA is the content of soil clay particles (%); OM is the soil organic matter content (%), calculated by multiplying the soil organic carbon content by 1.724 [48]; and CaCO₃ is the content of calcium carbonate (%).

$$SCF=1/(1+0.0066{CLA}^2+0.021{OM}^2)$$

(27)

$$K^{\prime }=\mathit{cos}\alpha$$

(28)

In this formula, α is the terrain slope.

$$C=e^{-0.0438FVC}$$

(29)

In this formula, C represents vegetation coverage (%).

➂

Human activity intensity

As a critical ecosystem component, cropland systems form an essential foundation for food security and exhibit heightened vulnerability to anthropogenic influences. Nighttime light intensity (NLI) data effectively characterizes regional economic development levels, providing significant reference value for analyzing spatial economic distribution patterns [49]. Concurrently, population density (PD) data quantifies demographic concentration. Therefore, NLI and PD were selected as representative proxies for human activity intensity, which is calculated as follows:

$$HAI={\displaystyle \frac{1}{2}}NLI+{\displaystyle \frac{1}{2}}PD$$

(30)

2.3.2. Spatial Autocorrelation Analysis

To examine the spatial heterogeneity and agglomeration effect in the health of Xinjiang’s cropland system, this study conducted spatial autocorrelation analysis using both global and local Moran’s I indices. The range of the Moran index (Moran I) is [−1, 1]—when Moran I > 0, this indicates that the spatial distribution exhibits clustering, and the intensity of spatial clustering increases with an increase in values. When Moran I = 0, the overall spatial distribution is discrete. When Moran I < 0, this indicates that there is no clustering in the spatial distribution [50]. Local Indicators of Spatial Association (LISA) were employed to test the correlation and spatial clustering level between adjacent regions and their neighboring regions [51].

2.3.3. Geographic Detector

The use of geographic detectors is a statistical method for detecting spatial heterogeneity and revealing its driving factors [52]. Compared to traditional correlation coefficient models, it can not only achieve quantitative data analysis and qualitative data processing but also analyze the interaction between various factors. When using a geographic detector to explore the impact of potential driving factors on the health of the cropland system, q represents the explanatory power of each factor on the target factor. The larger the q value obtained from factor detection, the stronger the explanatory power of the driving factors on the health of the cropland system [53]. The calculation formula is as follows:

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

(31)

In this formula, N is the total number of samples, N_h is the number of classified samples, σ² is the variance of all samples, ${\sigma }_h^2$ is the variance of classified samples, and L is the number of selected factors. The explanatory power q has a range of [0, 1], with higher values indicating stronger factor influence. A q value of 1 denotes the complete determination of the spatial distribution by the factor, while q = 0 indicates no association. To investigate the driving mechanisms of cropland health in Xinjiang’s arid regions, six indicators representing natural factors and human activities were selected based on prior research. Variance Inflation Factor (VIF) assessment confirmed no multicollinearity among these drivers [54] (

Table 3. Selection of driving factors and covariance test affecting changes in cropland system health.

Type	Driving Factors	VIF
Human factors	Population density	1.382
	Nighttime light	1.281
Natural factors	DEM	1.576
	Slope	1.531
	Annual precipitation	2.116
	Annual mean temperature	2.380

).

2.3.4. Spatiotemporal Geographic Weighted Model

The GTWR (Geographically and Temporally Weighted Regression) model extends traditional spatial regression by incorporating a temporal dimension and computing spatiotemporal weight matrices. This approach enables an investigation of the effects of independent variables on dependent variables across diverse spatiotemporal contexts [55]. In this study, grid-scale GTWR analysis was employed to examine spatial heterogeneity in driving factors affecting cultivated land system health. Modeling utilized the GTWR plugin developed by Huang [56] with the following formulation:

$$Y_i={\beta }_0(x_i,y_i,t_i)+{\displaystyle \sum _{k=1}^n{\beta }_k}(x_i,y_i,t_i)X_{ik}+{\epsilon }_i$$

(32)

In this equation, Y is the observed value; β₀(x_i, y_i, t_i) is the regression constant of unit i, which is the constant term in the GTWR model; β_k(x_i, y_i, t_i) is the kth regression coefficient of unit i; and X_ik is the value of the independent variable X_k at unit i. These are the values of quantitative indicators in the GWTR model indicator system. ε is the random error.

The Moran I test was used to evaluate the spatial autocorrelation of residuals in the GTWR model, and the values were not statistically significant (p > 0.05), thus indicating that there was no significant spatial autocorrelation and confirming the normativity of the model.

3. Results

3.1. Spatiotemporal Dynamics of Cropland System Health

From 2001 to 2023, Xinjiang’s cropland system exhibited a distinct health distribution pattern characterized by higher values in central regions and lower values in northern/southern areas. Higher health levels were clustered primarily in central Bozhou, the Ili River Valley, and southern Tacheng. These zones feature concentrated agricultural blocks, stable irrigation resources, sufficient light and heat conditions, and highly organic soils (predominantly irrigated silt, tidal soils, and irrigated desert soils) without significant health-limiting factors.

Temporally, the cropland health level demonstrated a fluctuating upward trajectory. Peak health values reached 81.75 (2001), 85.84 (2012), and 83.78 (2023), thus indicating an increase in fluctuations in health levels and overall improvement in the cropland system. Spatially, health gradients intensified—strengthening central while weakening in northern/southern peripheries (Figure 3). Altay’s central region continues to experience significant health constraints due to water scarcity, high salinity, severe desertification, limited irrigation efficiency, poor soil nutrients, and low productivity. Central Bozhou and Ili River Valley show expanding cultivated areas with increasing connectivity and consolidation. Abundant water resources and high-standard cropland construction policies are further driving the improvement in the health of the cropland system.

3.2. Spatiotemporal Differentiation of Cropland System Pressures

From 2001 to 2023, soil erosion intensity (SEI) decreased significantly, wind erosion intensity (WEI) remained relatively stable, and human activity intensity (HAI) exhibited an overall upward trend (Figure 4). In terms of SEI, severe soil erosion was concentrated in northern Tacheng, central Ili Kazakh Autonomous Prefecture, and central Bortala Mongolian Autonomous Prefecture. Erosion severity decreased overall during the study period, with particularly significant reductions in central Changji Hui Autonomous Prefecture. In terms of WEI, areas with high levels of wind erosion were mainly distributed in Altay, central Changji Hui Autonomous Prefecture, and central Turpan City, and the distribution range is relatively stable without significant changes. In terms of HAI, strong spatial correlation exists between human activities and cropland distribution. The temporal expansion patterns of cropland closely mirrored HAI expansion, thus indicating that the intensity of human activities has a positive effect on cropland expansion. However, as human activities exceed the intensity of sustainable development, the health level of the cropland system will show a downward trend.

Overall, the stress on the cropland system in northern Xinjiang is significantly higher than that in southern Xinjiang. Mainly due to the small population in southern Xinjiang, the degree of land development is relatively low, and cropland is concentrated in areas with flat terrain and that are less affected by sandstorms, resulting in the limited impact of soil erosion and wind erosion. The population distribution in northern Xinjiang is denser compared to southern Xinjiang, and there is a higher degree of land development and transformation with respect to human activities. Some areas that were originally unsuitable for cultivation are still being cultivated as cropland, and the quality of such cropland is less stable, thus putting greater stress on the cropland system.

3.3. Fishnet-Scale Patterns of Cropland System Health

Figure 5 illustrates spatiotemporal variations in cropland system health across the fishnet scale. In 2001, elevated health levels were clustered in Changji Hui Autonomous Prefecture, Bortala Mongolian Autonomous Prefecture, and Kashgar Prefecture, while Altay Prefecture exhibited lower values. Compared to 2001, health levels had declined moderately in Bortala and Kashgar but increased notably in Changji and northern Tacheng in 2012. By 2023, significant improvements emerged in Changji, Tacheng, Bortala, and Altay, while Kashgar did not show a significant increase, and there was no significant change compared to 2012.

Quantitatively, the proportion of high-health areas (levels I and II) shifted from 45.84% (2001) to 38.00% (2012) before rising to 50.80% (2023). Conversely, low-health areas (levels IV and V) increased from 28.72% (2001) to 32.88% (2012) then declined to 22.03% (2023). The health deterioration seen in 2001–2012 primarily stemmed from unreasonable land use patterns—disordered cropland expansion and overgrazing, which led to a decrease in the health level of the cropland system. The subsequent recovery (2012–2023) correlates directly with the implementation of China’s “strictest cropland protection system”, which is aimed at ensuring the effective protection of cropland resources. Multi-level governments were thus mobilized to take actions to protect cropland and promote the healthy development of the cropland system. Collectively, the health status of most fishnets in the cropland system exhibited an upward trend, with a minority maintaining stable conditions.

3.4. Spatial Autocorrelation Analysis of Cropland System Health

The global Moran I values for the health of the cropland system in the study area in 2001, 2012, and 2023 were 0.738, 0730, and 0.643, respectively (p < 0.01) (Figure 6). There is a strong positive correlation between the health of cropland systems and their spatial clustering at the scale of fishnets.

Figure 7 presents the LISA clustering results for local autocorrelation. The analysis revealed the following: Low-low clusters exhibited dispersed spatial distribution, which was characterized by limited cultivated acreage and severe wind and soil erosion. High–high clusters were concentrated primarily in Changji Hui Autonomous Prefecture, Bortala Mongolian Autonomous Prefecture, and Ili Kazakh Autonomous Prefecture, featuring extensive cropland areas and low levels of wind erosion and soil erosion. Temporally, low–low cluster occurrence decreased between 2001 and 2023, while high–high clusters demonstrated increasing spatial concentration.

3.5. Driving Factors of Cropland System Health Patterns

3.5.1. Factor-Health Correlations

As shown in Figure 8, cropland system health demonstrates significant positive correlations with population density, annual precipitation, and annual mean temperature. Precipitation and temperature exhibit the strongest correlations (r > 0.60). Conversely, the health of the cropland system shows a negative correlation with slope, and the correlation coefficient between them temporally decreases over time, indicating that the negative correlation is gradually weakening.

3.5.2. Driving Factor Impacts

Based on the exploration of spatial distribution factors affecting the health of cropland systems using geographic detectors, the explanatory power q values of single-factor effects were obtained (

Table 4. Explanatory power of driving factors on cropland system health.

Year	DEM	Slope	Population Density	Nighttime Light	Precipitation	Temperature
2001	0.265	0.331	0.630	0.452	0.638	0.640
2012	0.260	0.295	0.582	0.408	0.603	0.606
2023	0.275	0.314	0.644	0.478	0.682	0.645
Average	0.267	0.313	0.619	0.446	0.641	0.630

Note: Indicates statistical significance at 5%.

). The six driving factors selected in this study passed the significance test (p < 0.05). Overall, climate factors have the greatest explanatory power for the health of cropland systems among natural factors, followed by human activity factors. Among all driving factors, precipitation has the highest explanatory power for the health of the cropland system, while terrain has the lowest. From 2001 to 2023, the q-value of human activity factors gradually increased, thus indicating that the differences in human activity intensity between fishnets were gradually increasing, and that the impact on the heterogeneity of regional cropland system health was also strengthening.

There is only one interaction relationship between the driving factors, which is bilinear enhancement. The results showed that the strongest interaction factors explaining the health of the cropland system in 2001, 2012, and 2023 were slope and precipitation, precipitation and temperature, and precipitation and temperature, with q-values of 0.715, 0.666, and 0.738, respectively (Figure 9). Overall, this indicates that precipitation and temperature play a dominant role in the spatial differentiation of the health of the cropland system.

3.5.3. Spatiotemporal Heterogeneity of Driving Factors

To further investigate the spatiotemporal variation characteristics of the driving factors of the health of the cropland system at the scale of fishnets in Xinjiang, the GTWR model was used to conduct spatiotemporal weighted regression analysis on the health of the cropland system in the study area. The results showed that the R² in 2001 and adjusted R² were 0.7646 and 0.764, respectively; in 2012 were 0.726 and 0.725, respectively, and in 2023 they were 0.781 and 0.781, respectively, thus indicating that the independent variables jointly explained more than 70% of the variation in the dependent variable. The overall significance of the model is high (p < 0.05), indicating that the model has a good fitting effect. If the GTWR regression coefficient is positive, this indicates that the explanatory variable is positively correlated with the health of the cropland system; If the coefficient is negative, there is a negative correlation. In addition, the larger the absolute value of the regression coefficient, the stronger the correlation between the explanatory variable and the health of the cropland system.

The impact of driving factors on the health of cropland systems at the scale of fishnets in Xinjiang exhibits spatiotemporal heterogeneity (Figure 10, Figure 11 and Figure 12). The regression coefficients of slope in the study area are all negative, indicating a negative correlation between the health of the fishnet scale cropland system in Xinjiang and slope. Over time, the negative value area gradually decreases, indicating that the negative driving effect of slope on the health of the cropland system is gradually weakening. The regression coefficients of population density, annual precipitation, and annual mean temperature in the study area are all positive, indicating that human activities and climate factors are generally positively correlated with the health level of the cropland system.

From the spatiotemporal distribution of the GTWR coefficient of population density, it can be seen that the population density in the Aksu, Bazhou, and Tacheng regions has a greater explanatory power for the health of the cropland system. However, in 2023, the explanatory power showed a downward trend, indicating that the impact of human activities on the health of cropland systems in such areas has gradually weakened in recent years. The high-value areas of the annual precipitation GTWR coefficient are mainly distributed in the eastern part of Changji, Turpan, northern Bazhou and other areas. The health level of the cropland system in these areas is greatly affected by precipitation. From 2001 to 2023, there was no significant change in the distribution of high-value coefficient areas, thus indicating that the positive driving effect of annual precipitation on the health of cropland systems in this region remained stable; that is, water remains the main limiting factor for the health status of cropland systems in arid areas. The Altay and Tacheng regions in Xinjiang are areas with high annual mean temperature GTWR coefficients, and they maintained a stable trend from 2001 to 2023. One major factor limiting the health status of the cropland system in northern Xinjiang is temperature.

4. Discussion

4.1. Spatiotemporal Evolution of Cropland System Health in Xinjiang

From a temporal perspective, the health fluctuations in the cropland system have increased, and the overall environment of the cropland system has improved. This is consistent with the research findings of Li et al. [57], both showing a gradual improvement in the health of Xinjiang’s cropland system. Mainly due to the implementation of national cropland protection policies in the region in recent years, cropland protection work has been listed as one of the key tasks. Moreover, the health spatial heterogeneity of the cropland system in Xinjiang is strong, showing a spatial distribution pattern in which that of northern Xinjiang is higher than that of southern Xinjiang, which is consistent with the previous research conclusions of Jin et al. [13]. However, it is worth noting that there is still room for improvement in the overall level, as demonstrated by the research findings of Yoshizawa et al. [58], which show that most farmers still prioritize economic interests over ecological harmony development. Selecting three indicators, HAI, SEI, and WEI, to characterize the stress on the cropland system, it was found that the overall level of HAI showed an upward trend, while the overall level of SEI decreased significantly, and WEI remained relatively stable. With the deepening implementation of the Western Development Strategy, the level of urbanization continues to improve. The intensity of human activities has increased over time, leading to the escalation of regional cropland system risks [50], which is consistent with the conclusion of Zhang et al. [59]. The degree of soil erosion in Xinjiang is gradually decreasing, mainly due to the government’s measures in soil and water conservation, thus indicating that the local government has achieved significant results in ecological engineering for soil and water conservation. Xiong et al.’s research also confirms this conclusion [60]. In the key projects of soil erosion control, Xinjiang implements the principle of “prevention first, protection priority”; actively promotes the comprehensive management and prevention protection of soil erosion; and fully relies on the self-regulation and restoration capabilities of the ecosystem in the governance process.

4.2. Mechanisms Driving Cropland System Health

In terms of driving factors, the results of this study suggest that the combination of climate and human activity factors plays a dominant role in the spatial differentiation of cropland system health. The slope is negatively correlated with the health level of the regional cropland system, mainly because areas with higher slopes are not suitable for cropland activities and are not conducive to cropland expansion. Fei et al.’s research confirms this statement [61]. Overall, population density, annual precipitation, and average annual temperature are positively correlated with the health level of the cropland system. This is consistent with the research conclusion of Zhou et al. [18], who showed that the precipitation factor that has a negative driving effect on the health of the cropland system in the Yellow River Basin may have a positive driving effect in Xinjiang. The reason for this may be that, compared with the Yellow River Basin, Xinjiang is located in an arid area with relatively scarce annual precipitation, and water sources have become the main factor limiting the expansion of cropland. Therefore, it occupies a relatively important position in the cropland system in arid areas. Shen et al.’s research also confirms this conclusion [62]. Population density has a positive driving effect on the health of the cropland system. This research provides valuable support for policy-making regarding cropland protection and utilization in Xinjiang. Strategically measuring the health status of Xinjiang’s cropland system, combined with the spatial heterogeneity of driving factors, and proposing differentiated strategies for protecting Xinjiang’s cropland are key to achieving the sustainable development of cropland under the new situation of food security and rural revitalization.

4.3. Limitations and Prospects

The novelty of the improved VOR model lies in breaking through the “linear causality” logic of DPSIR, constructing a dynamic collaborative cycle, combining internal system attributes with external pressures, and more accurately depicting the dynamic interaction between “external pressures and cropland systems”. This “dynamic collaboration” perspective is the core novelty that distinguishes it from DPSIR and resilience indicators (which only focus on a single indicator of system recovery capability).

However, due to the current stage of research on cropland system health and the lack of a unified evaluation system, the results that can be used for reference are limited. Therefore, the driving factors selected in this study may not be perfect enough. Therefore, future research should expand the types of indicators and data sources, consider incorporating multiple factors such as policies and funding, and more accurately explore the spatiotemporal heterogeneity of driving factors for cropland system health. Due to incomplete data acquisition, six driving factors were selected for node years in 2001, 2012, and 2023 for research. Increasing the number of factors and shortening the time interval may yield more convincing results. In future research, further improvements will be made to address the above shortcomings.

5. Conclusions

Xinjiang was selected as the research area in this study. A more suitable framework for evaluating the health status of cropland system in arid areas was constructed, and a systematic diagnosis of the health status of cropland system in Xinjiang was conducted by increasing the cropland stress and extending the VOR model to the VOR-S framework. Geographic detectors and GTWR models were used to explore the impact of potential driving factors on cropland system health. The main conclusions are as follows: (1) From 2001 to 2023, the health level of Xinjiang’s cropland system fluctuated and increased. The proportion of areas with higher health levels (health levels I and II) in the cropland system increased from 45.84% in 2001 to 50.80% in 2023. The overall environment of the cropland system has thus improved. (2) From 2001 to 2023, in terms of stress on the cropland system in Xinjiang, the overall HAI level showed an upward trend, while the overall SEI level significantly decreased, and WEI remained relatively stable. (3) The explanatory power of driving factors for the health of cropland systems, in descending order, is as follows: annual precipitation (0.641) > annual average temperature (0.630) > population density (0.619) > nighttime light (0.446) > slope (0.313) > DEM (0.267). In addition, the combination of climate factors and human activity factors plays a dominant role in the spatial differentiation of cropland system health.