Modelling Ecological Hazards and Causal Factors in the Yellow River Basin’s Key Tributaries: A Case Study of the Kuye River Basin and Its Future Outlook

Abstract

The Kuye River is the second largest tributary of the middle Yellow River. (1) Background: The Kuye River Basin, a typical erosion area of the Loess Plateau region, faces significant environmental challenges and intense human activities. Balancing environmental sustainability and economic development in this region is urgent. (2) Methods: This study analyses the phenomena, evolutionary processes, driving mechanisms, and future development trends. We assess ecological risks and drivers of land use change using data from 2000, 2005, 2010, 2015, and 2022. (3) Results: Farmland, grassland, and construction land are the main land use types, accounting for 85.63% of the total area. Construction land increased by 7.95 times over 22 years, mainly due to the conversion of woodland, grassland, and farmland. The landscape pattern increased in patches from 4713 in 2000 to 6522 in 2022. Patch density decreased from 0.0945 to 0.0771 between 2000 and 2015, then rose to 0.0788 in 2022. Post-2015, increased human intervention and urban development led to significant landscape fragmentation and higher ecological risk, expected to persist until 2030. Geographical detector analysis identified distance from roads, distance from cities, night light, and precipitation as key factors influencing landscape ecological risk. The interaction of anthropogenic disturbance with other factors showed a non-linear increase in risk, with combined factors having a greater impact than individual ones. (4) Conclusions: The Kuye River Basin’s landscape ecological risk is influenced by both natural conditions and human activities. To achieve sustainability, it is essential to protect critical areas, regulate development, and improve the adaptive management of ecological risks through innovative policies, integrated regulations, and technological solutions for ecosystem restoration. These findings provide empirical evidence to support decision-making and underscore the need for comprehensive strategies to mitigate ecological risks and promote sustainable development in the Kuye River Basin.

1. Introduction

Ecological risk means a combination of uncertain events or disasters in ecosystems [1], which can cause ecosystem function loss due to changes in the composition and structure of the ecosystems. Ecological risk can be caused by human activities or natural processes [2]. Ecological risk assessment dates to the 1980s, focusing on human health and the global environment [3,4]. It is a valuable method for measuring and assessing the adverse effects of human activities and natural disasters on ecosystems’ composition, organisation, and function [5]. Compared to the traditional ecological risk assessment, the landscape ecological risk assessment better addresses the potential negative outcomes associated with spatial factors [6,7,8]. As an emerging tool in natural resource management, landscape ecological risk assessment offers novel methods for statistically analysing the ecological impacts of land use change [9,10,11]. Additionally, the approach emphasises spatial visualisation, dividing an area into different risk zones using a grid system. This allows for more accurate spatial representations, and this measure has gradually emerged as a prominent subject of investigation in ecology and geography [12,13].

Landscape ecological risk assessment has been conducted in areas with human activity and essential risk control areas, including industrial mining zones and nature reserves [13,14,15]. Generally, there are two fundamental assessment approaches: the source–sink analysis method and the landscape pattern method. The source–sink analysis method is appropriate for assessment objectives with well-defined regional ecological stressors [16]. The landscape pattern method investigates the ecological risk effect caused by general disturbance and stress rather than a specific stress element [17]. Considering spatial variation, this approach broadens the scope of risk receptors to encompass the entire ecosystem. It identifies land use change as the primary driver of changes in ecological risk. In addition, it examines patterns of landscape ecological risk by analysing the spatial structure and dynamic changes in land use and cover [17,18,19]. Overall, the landscape ecological risk assessment is a multidisciplinary field that focuses on developing methodologies and indicators, utilising data and technology, resolving scale-related issues, and supporting practical applications and regulations. These studies have demonstrated the benefits of the landscape pattern method, including the comprehensive assessment and geographic representation of ecological hazards with limited resources.

Land use/land cover change is closely correlated with social, economic, and environmental issues and other elements [20]. Land use change primarily alters natural landscapes, mainly due to human activities, influencing humans’ ability to survive. The spatial and temporal change in land use results from the interaction between natural processes, human activities, and regional biological processes [19,21]. Land use changes can result in regional landscape ecological risks related to the land system without good feedback [22,23]. Additionally, landscape ecological risk can reflect the negative impacts of land use and land cover changes on the landscape function. When assessing the landscape ecological risks of land use change, it is important to consider the adverse impacts on risk receptors and the overall impact on landscape fragmentation and diversity [24]. Ji et al. [25] found that the ecological risk in Chaoyang County, a region in northeast China with significant forest cover, is primarily influenced by the conversion of farmland. Abdullah et al. [26] investigated the impact of land use change on landscape patterns in Malaysia. By exploring the driving mechanisms, they found that policies and human activities are the main reasons for landscape fragmentation. Austrheim et al. [27] proposed that appropriate land use patterns are related to the long-term sustainable development of agriculture and regional biodiversity by studying the relationship between land use patterns, biodiversity, and landscape patterns. The study focuses on the risks associated with construction activities and extensive deforestation, considering land use and landscape ecology. Consequently, landscape metrics that quantify regional landscape patterns are commonly used in landscape ecological risk assessment frameworks and related studies.

The landscape ecological risk analysis aims to protect environmental quality by identifying the primary drivers among the various factors that influence the ecological risk of a given region [28]. Thus, conducting a landscape risk assessment and analysing driving mechanisms can provide a comprehensive understanding of the factors contributing to spatial differentiation. This approach can also ensure the proper functioning of the ecosystems and the landscape’s ecological security. Moreover, the correlation between landscape ecological risk and environmental factors directly influences the identification and implementation of local ecological management strategies [25,29]. Understanding the effects of environmental factors, e.g., human activities, on watershed ecological conservation is essential in global change. The methods used to study the driving forces mainly include geographical detectors [30], partial least squares structural equation modelling [31], ordinary least squares [32], random forest regression [33], and geographically weighted regression [4]. The geographical detector method is highly effective in analysing the driving elements of the system of interest [34]. Currently, the geographical detector method is widely used to quantify the drivers of climate change, GHG emission reductions, vegetation changes, settlement distribution, and even the spread of the new crown epidemic (COVID-19) [35]. However, this method has yet to be applied to landscape ecological risk drivers. Zhou et al. [36] used the geographical detector method to detect the influence mechanisms and differences in natural and socio-economic factors with respect to PM2.5 change in Guangzhou. Based on the geographical detector, Chen et al. [37] analysed the spatial distribution and interaction intensity of the driving forces of soil heavy metals at the regional scale in Anshun City. Based on previous studies, the geographical detector can be effectively used in the fields of mining ecology, atmospheric environment, watershed, etc., and the scale of the study area is less strict, ranging from the township scale to large watersheds. Compared with other approaches, it detects and explains the interactions between these components. It is essential to explore the feasibility of the geographical detector model in comprehensively analysing the landscape changes caused by these potential factors, thereby elucidating the changes in landscape ecological risk at both global and local scales. Previous studies have conducted empirical analyses and investigated landscape ecological risks across regions, theories, and methodologies [38]. However, the precise causes and underlying mechanisms of risk emergence still need to be identified, necessitating further practical efforts to improve understanding and fill knowledge gaps, particularly regarding the watershed. Though landscape ecological risk assessments in watersheds have focused primarily on assessing current conditions, few have focused on future changes in landscape ecological risk. Predicting the potential environmental risks of landscapes can help formulate land use plans and promote environmentally sustainable development [39,40]. The landscape pattern index enables the simulation of future landscape ecological risk profiles using projected land use data [41].

Accurately predicting the spatial and temporal patterns of land use/land cover is essential for studying potential ecological risks in landscapes, e.g., watersheds. Various land use simulation models have been developed to predict future land use change patterns, including cellular automata (CA) [42], the conversion of land use and its effects modelling framework (CLUE) [43], and future land use simulation (FLUS) [44]. Applying multitype random patch seeds (CARS) can model numerous land use categories at a detailed resolution. The PLUS model has been empirically demonstrated to provide more accurate simulation results [40]. However, projecting land use and land cover change depends on robust economic factors and government policies characterised by high uncertainty and complexity. In the present study, land use change projections are implemented based on a novel model, i.e., the Patch-Generating Land Use Simulation (PLUS) model. The PLUS model introduces a novel land expansion analysis technique (LEAS), which enhances the existing rule-mining approaches compared to previous models. Lin and Wang [45] assessed landscape ecological threats and their causes from 2000 to 2020 in Guiyang, located in a sensitive rocky karst region, using remote sensing photos, landscape pattern indices, and the geographical detector model. In addition, the PLUS model was used to predict the landscape ecological risks associated with different urban development scenarios in 2030. As theoretical research in landscape ecology has progressed, scientists have recognised that landscape ecological risk is a complex issue spanning various scales and levels of analysis. Furthermore, the approach has prioritised using spatial visualisation techniques, namely creating a grid-like structure to delineate different risk zones. This enables precise and accurate spatial representations, which help this strategy effectively study future use changes [46,47,48]. Landscape ecological risk assessment critically supports decision-making in landscape conservation and ecosystem management. However, there needs to be more clarity in the assessment process. Different data sources, preprocessing methods, regional differences, and survey scales will affect the evaluation results to varying degrees. Therefore, it is crucial to carefully select the appropriate techniques and indicators adapted to the specific circumstances and consider regional characteristics when conducting landscape ecological risk assessments.

The Kuye River Basin is a representative river of the erosion-prone Loess Plateau region. The high precipitation in the area significantly increases the likelihood of significant flooding, making it one of the primary causes of flooding in the northern part of the Yellow River. It also significantly contributes to transporting coarse sediment (mainly silt) into the river, a crucial factor affecting the river’s flow. Besides, more than 95% of the watershed area in this region is affected by soil erosion. The average annual sediment transport volume is 111 million tons, with a maximum of 335 million tons and the highest recorded sediment concentration of 1700 kg/m³. At the same time, intensive human activities characterise the Kuye River Basin, as well as remarkable economic development and exceptional ecological conditions. The Shenfu-Dongsheng mega coal mine field, the seventh largest in the world, runs through the central part of the basin. The mining areas cover 28.51% of the Kuye River Basin [49]. In addition, rapid urbanisation has exacerbated the conflicts between farmland protection, economic development, and ecological conservation in the region. The resulting secondary environmental problems have seriously hindered the region’s sustainable development. Therefore, this paper focuses on assessing the ecological risk of the Kuye River Basin, which is the central area for constructing China’s cross-century “Jin, Shaanxi and Mongolia Energy and Heavy Chemical Industry Base” [49]. The analysis is based on the watershed’s land use structure and spatiotemporal development characteristics from 2000 to 2022. Then, Fragstats 4.2 and ArcGIS 10.8 (Environmental Systems Research Institute—Arc Geographic Information System) software were used to estimate the spatiotemporal changes of each landscape pattern index and landscape ecological risk based on the land use data. Then, the factors affecting landscape ecological risk were investigated using geoprobes. Geoprobes can detect the spatial differences and effects of individual variables and the cause–effect relationship between these variables. In addition, they can handle both quantitative and qualitative data to detect how these factors’ interaction affects the dependent variable. Finally, considering several scenarios, the PLUS model will forecast land use and assess landscape ecological concerns in the basin in 2030. Understanding these interrelationships provides valuable insights for policymakers. It lays the foundation for future strategies to balance urbanisation with sustainable agricultural and ecological practices, ensuring long-term regional stability and prosperity.

2. Materials and Methods

2.1. Overview of Study Area

The Kuye River Basin (109°28′~110°45′ E, 38°22′~39°50′ N) is located on the right side of the north mainstream of the Yellow River (Figure 1). The Kuye River Basin is a typical erosion area in the Loess Plateau, mainly caused by floods. The Kuye River is the second largest tributary in the middle part of the Yellow River, supplied by two major tributaries, the Wulanmulun River and the Quince Niuchuan River. It begins in Mixed Tree Village, Chaideng Township, Dongsheng District, Ordos City, Inner Mongolia Autonomous Region. Then, it flows toward the southeast, crosses 6 districts of Ordos and Yulin, and joins the Yellow River at Sha’an Village in Shenmu. The river spans 242 km in length and covers an area of 870,600 hectares. The river’s upper reach is in the sandy steppe region, while the middle and lower reaches are in the loess hills and gullies.

The flora across the entire basin is sparse, and well-formed erosive gullies are primarily arranged in a branching pattern. Gully density ranges from 5 to 9 km/km², with valleys over 100 m deep. The watershed is in the transitional zone between the Mao Wusu desert and the hills/gullies of the Loess Plateau. The terrain is higher in the northwest and lower in the southeast, with elevations ranging from 800 to 1300 m. The soil is sparse, consisting mainly of sandy terrain. The climate is arid and semi-arid, with an average annual precipitation of 415 mm. Precipitation is highly uneven in both spatial and temporal distribution. From June to September, it accounts for 75% to 81% of the annual total, with July and August alone contributing 50% to 60%. Winter sees minimal snowfall, while spring and fall have limited precipitation and early frost. Summer is marked by heavy rainfall. The Kuye River Basin also faces significant soil erosion due to hydraulic and wind erosion. The Kuye River Basin is characterised by a complex combination of environmental conditions, including intense soil erosion, frequent droughts, rainstorms, dust storms, and a significant amount of coarse sediment flowing into the Yellow River.

The Kuye River Basin is also home to the Shenfu Dongsheng mega coalfield, the eighth-largest coal field in the world. This coalfield stretches across the central part of the basin and covers 28.51% of the basin’s total area. As a result, the basin has become a focal point for developing the “Jin-Shaan-Mengguang Energy and Heavy Chemical Industry Base” in China for the coming century. However, rapid urbanisation, growing industrial and mining activities, and inadequate governance have led to significant unsustainable development. This has further exacerbated soil erosion, land fragmentation, and degradation.

2.2. Data Sources

To systematically assess landscape ecological risks, drivers, and future development scenarios, the present study applied remote sensing imagery, ecological and environmental data, and socio-economic data (

Table 1. Data sources.

Research Module	Data Details	Data Sources	Web Address	Data Processing Platform	Note
Land use structure	Cultivated land, woodland, grassland, construction land, waters, unutilized land	Geospatial Data Cloud	“http://www.gscloud.cn (accessed on 24 May 2024)”.	ArcGIS, ENVI	Produced the land use by using initial Landsat series images. Landsat TM 4–5 and Landsat 8 OLI_TRIS images were downloaded through the Geospatial Data Cloud, and image preprocessing was performed by ENVI 5.2 software with supervised classification to categorise land use types into farmland, woodland, grassland, construction land, waters, and unutilized land, with an image resolution of 30 m
Landscape Ecological Risk Drivers-Natural	Precipitation	National Meteorological Administration (NMA)	“http://www.cma.gov.cn/ (accessed on 24 May 2024)”.	ArcGIS	Precipitation data ware NMA public data. Data ware obtained by free download from the National Meteorological Bureau. Visualisation of precipitation data by interpolation with ArcGIS 10.8 software
	DEM (Digital Elevation Model)	Geospatial Data Cloud	“http://www.gscloud.cn (accessed on 24 May 2024)”.	ArcGIS	DEM data were Geospatial Data Cloud public data and were available for download upon registration. Considering that the data were downloaded based on latitude and longitude, they also needed to be cropped according to the watershed boundaries. This was performed using the cropping module under the data management tools of the ArcGIS 10.8 software
	Slope	Geospatial Data Cloud	“http://www.gscloud.cn (accessed on 24 May 2024)”.	ArcGIS	Based on the DEM data processed in ArcGIS 10.8 software, obtained by processing in the 3D analyst module of the ArcGIS Toolbox to obtain the slope data
	Air temperature	Scientific data platform on resources and environment	“https://www.resdc.cn/ (accessed on 24 May 2024)”.	ArcGIS	Air temperature were public data. The data were obtained from the China Meteorological Data Module of the Resource and Environmental Science Data Registration and Publishing System (RESDPS). Image projection, transformation, and resampling were carried out using ArcGIS 10.8 software to ensure an image resolution of 30 m
Landscape ecological risk drivers-socio-economic factors	Night light, GDP (Gross Domestic Product), population density	Scientific data platform on resources and environment	“https://www.resdc.cn/ (accessed on 24 May 2024)”.	ArcGIS	Night light, GDP, and population density were public data. The data were obtained from the Socio-Economic Data Module of the Resource and Environmental Science Data Registration and Publishing System (RESDPS), and the image projection, transformation, and resampling were carried out in ArcGIS 10.8 software to ensure an image resolution of 30 m
	Distance from road, distance from the city	Scientific data platform on resources and environment	“https://www.resdc.cn/ (accessed on 24 May 2024)”.	ArcGIS	Roadway data from Open Street Map, ArcGIS buffer and Euclidean distance module were used to derive the roadway distances; distance from the city was based on the Cave Creek Watershed administrative division, combined with ArcGIS 10.8 software analysis for derivation

).

2.3. Research Methodology

2.3.1. Research Framework

The research of this paper consisted of four parts (Figure 2). First, we acquired Landsat TM 4–5 and Landsat 8 OLI_TRIS remote sensing images of the Kuye River Basin. The images were processed using ENVI 5.2 software, and six land use categories in the Kuye River Basin were obtained through supervised classification. We analysed the land use structure, geographical and temporal evolution, and land use transfer status of the Kuye River Basin for 2000, 2005, 2010, 2015, and 2022 using ArcGIS 10.8 software. Secondly, Fragstats 4.2 software was used to calculate the number of patches, patch area, and patch density based on the derived land use data. Fishing nets were generated using ArcGIS 10.8 software, and a combination of algorithms was employed to produce various landscape pattern indices and ecological risk indices. We used ArcGIS 10.8 software to visually analyse the spatial and temporal differentiation characteristics and evolutionary elements of landscape ecological risk in the Kuye River Basin from 2000 to 2022. Thirdly, we used geographic detector analysis to investigate the driving factors of landscape ecological risk, e.g., precipitation, DEM, slope, air temperature, night light, GDP, population density, distance from different types of roads (high-speed, national, provincial, railway, and county), and distance from cities. Lastly, the land use and landscape ecological risk data were used to predict the land use structure and landscape ecological risk values of the Kuye River Basin in 2030 under different scenarios using the PLUS model.

2.3.2. Land Use Structure

(1)

Land Use Dynamics

Land use dynamics describe the quantitative shift of land use types over a certain period, mainly reflecting the regional differences in the extensity and intensity of land use change. There are two categories: comprehensive and individual land use dynamics [50]. The specific calculation method involves a formula where the dynamic degree of each land use type represents the rate of change of that particular land use type within a specified period in the region [51]. During the study period, the general state of land use transfer between land use categories was analysed using the dynamic degree of land use [52,53]. This can compare the comprehensive land use change between local and total areas and show the intensity of regional comprehensive land use change. The exact calculation method is described below.

$$LU{T}^{now-before}={\displaystyle \frac{LU{T}^{now}-LU{T}^{before}}{LU{T}^{before}\times Time}}\times 100\%$$

(1)

$$LU{T}^{synthesize}=\left[{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n\Delta LU{T}^{x-y}}}{2{\displaystyle {\sum }_{i=1}^{n}LU{T}^x}}}\right]\times {\displaystyle \frac{1}{Time}}\times 100\%$$

(2)

where LUT^now-before represents the area of a land use type at the beginning and end of the study period, while LUT prior denotes the dynamic degree of a single land use type over the study period. Time represents the study period; LUT^synthesize represents the overall dynamic degree of land use; LUT^x represents the number of land use types in the initial class x; and ∆LUT^x−y represents the area of land use type x converted to other land use types between the initial and final stages.

(2)

Land Use Transfer Matrix

The land use transfer matrix reflects the dynamic process of inter-transformation between land use type areas at the beginning and end of a period within a specified region [54]. In addition to the static data of a given location at a specific time, it includes more detailed information on the transfer of area for each category at the beginning of the study and the transfer of area for each category at the end of the study. The calculation method is shown in Formula (3).

$${\mathit{SQ}}_{\mathit{ij}}=\left[\begin{array}{ccccc}{\mathit{SQ}}_{11}& {\mathit{SQ}}_{12}& {\mathit{SQ}}_{13}& \dots & {\mathit{SQ}}_{1n}\\ {\mathit{SQ}}_{21}& {\mathit{SQ}}_{22}& {\mathit{SQ}}_{23}& \dots & {\mathit{SQ}}_{2n}\\ {\mathit{SQ}}_{31}& {\mathit{SQ}}_{32}& {\mathit{SQ}}_{33}& \dots & {\mathit{SQ}}_{3n}\\ \dots & \dots & \dots & \dots & \dots \\ {\mathit{SQ}}_{n1}& {\mathit{SQ}}_{n2}& {\mathit{SQ}}_{n3}& \dots & {\mathit{SQ}}_{\mathit{nn}}\end{array}\right]$$

(3)

where S is the area of land use, n is the number of land use types before and after the transfer, i, j = 1, 2, ..., n for the area of each type of land use, and SQ_ij is the area of the i land class both before and after the transfer.

2.3.3. Landscape Patterns and Landscape Ecological Risk

(1)

Creation of Risk Sample Plots

To calculate the two spatial indicators, landscape ecological risk (ERI) and human activity intensity (HAI), the study area and size of evaluation cells were comprehensively considered by using the Toolboxes–Data Management Tools–Sampling–Create Fishnet operation mode in the ArcGIS data platform. The study area was divided into 417 square grids as evaluation cells with a side length of 5 km, which were used to calculate the ecological risk index of the landscape in each independent cell. The ideal range for a landscape sample area is typically two to five times the average patch area [55], according to research on landscape ecology. As a result, we integrated the Kuye River Basin patch area from 2000 to 2022 and ultimately decided to use 5 km as the evaluation unit.

(2)

Landscape Ecological Risk

The landscape ecological risk index quantifies the ecological risk resulting from land use patterns in a region. The index calculation incorporates explanatory variables accounting for internal and external factors influencing landscape ecological risk. External factors pertain to the extent of ecosystem disturbance caused by external forces, as indicated by the degree of landscape disturbance. Internal factors, conversely, reflect the ecosystem’s susceptibility to pressure, as indicated by the degree of landscape fragility. We utilised Fragstats 4.2 software to develop a landscape ecological risk assessment model for the Kuye River Basin. This model was based on the study area’s specific conditions and previous research [56]. The model focused on the landscape disturbance degree index and vulnerability index. The calculations were performed as follows (

Table 2. Calculation formula and significance of landscape ecological risk indexes.

Name of Index	Formula	Explanation of the Meaning of the Formula	Meaning of Index
Landscape fragmentation index (Ci)	$C_i={\displaystyle \frac{n_i}{A_i}}$	Ci is the landscape fragmentation index; ni is the number of patches in landscape type i; Ai is the area of landscape type i	Complexity of spatial distribution of landscape types after encountering external disturbances
Landscape fractional dimension index (Fi)	$F_i={\displaystyle \frac{2\left[(n_i{\displaystyle \sum _{j=1}^n\ln p_{ij}^2)-{({\displaystyle {\sum }_{j=1}^n\ln p_{ij}})}^2}\right]}{\left[n_i{\displaystyle {\sum }_{j=1}^n(\ln p_{ij}\times \ln A_{ij})}\right]-\left[({\displaystyle {\sum }_{j=1}^n{p}_{ij})({\displaystyle {\sum }_{j=1}^n{A}_{ij})}}\right]}}$	ni is the number of patches in landscape type i; Aij is the area of landscape type i in the jth risk cell; Pij is the perimeter of landscape type i in the jth risk cell	Complexity of the shape of landscape types at a given scale
Landscape separation index (Ni)	$N_i={\displaystyle \frac{1}{2}}\sqrt{{\displaystyle \frac{n_i}{A_i}}}\times {\displaystyle \frac{A}{A_i}}$	Ni is the landscape separation index; ni is the number of patches in landscape type i; Ai is the area of landscape type i; A is the total area of all landscapes	Level of patch heterogeneity in a particular landscape
Landscape disturbance index (Ei)	$E_i=a{C}_i+b{N}_i+c{F}_i$	Ei is the landscape disturbance index; a, b, and c represent the weight of each landscape index, a + b + c = 1. In this paper, with reference to the results of many studies, such as by Tian et al. [57], and combined with the actual situation of the study area, the weight of a is set to 0.5, the weight of b is set to 0.3, and the weight of c is set to 0.2; Ci is the landscape fragmentation index; Ni is the Landscape separation index; Fi is the landscape fractional dimension index	Extent of anthropogenic disturbance of the landscape
Landscape vulnerability index (Qi)	The Landscape vulnerability index (LVI) was assigned to different landscape types with reference to the existing research results [58,59,60].		Sensitivity and vulnerability to and resistance to external disturbances
Landscape loss degree index (Ri)	$R_i=E_i\times Q_i$	Ri is the landscape loss degree index; Ei is the landscape disturbance index; Qi is the landscape vulnerability index	Ecological losses from external disturbance: the higher the degree of loss, the higher the degree of disturbance
Landscape ecological risk index (ERIi)	$ER{I}_i={\displaystyle \sum _{i=1}^n\frac{A_{ij}}{A_i}}\times R_i$	ERIi is the landscape ecological risk index; Aij is the area of landscape type i in the jth risk cell; Ai is the area of landscape type i; Ri is the landscape loss degree index	Landscape ecological risk profiles reflecting changes in ecological conditions

).

2.3.4. Geographical Detectors

The geographical detector is a geographical statistical method developed by Wang [61] to identify spatial variations and measure the influencing factors. Spatial heterogeneity reflects the expression of natural and socio-economic processes across different areas, indicating that the value of a characteristic varies among various types or locations.

(1)

Risk Detectors

Risk detection focuses on revealing which types of variables have significantly high or low values of liveability satisfaction, as measured by t-tests [62]. In this paper, we used this module to indicate risk areas.

$$t_{ij}={\displaystyle \frac{R_i-R_j}{\sqrt{{\sigma }_i^2}/n_i-{\sigma }_j^2/n_j}}$$

(4)

where t_ij is the t-test value,; R_i and R_j are the mean values of liveability satisfaction for attributes i and j, respectively; σ_i² and σ_j² are the variances of liveability satisfaction for attributes i and j, respectively; n_i and n_j are the sample sizes of the two attributes.

(2)

Power of Determinant

The power of determinant (PD), i.e., single-factor detection, measures the extent to which different perceived liveability factors and individual and family attribute factors explain liveability satisfaction [38], and is calculated as follows. In this paper, we used this module to quantitatively evaluate the explanatory power of different environmental factors.

$$P_{D,H}=1-{\displaystyle \frac{1}{n{\sigma }^2}}{\displaystyle \sum _{h=1}^L{n}_h}{\sigma }_h^2$$

(5)

where P_D,H is the explanatory power of influence factor D on liveability satisfaction H; n and σ² are the sample size and variance, respectively; n_h and σ²_h are the sample size and variance of the h (h = 1, 2, ..., L) stratum. P_D,H takes the value range of [0, 1], larger values indicate that the categorical factors have more substantial explanatory power on liveability satisfaction, a value of 0 indicates that the categorical factors are completely irrelevant to liveability satisfaction, and a value of 1 indicates that the categorical factors can fully explain the differences in the distribution of liveability satisfaction.

(3)

Interactive Detector

The interaction detector can identify the combined effect of two factors synergistically influencing landscape ecological risk [63]. We can determine the mode and interaction direction between various elements by comparing the q-value of two individual factors with the q-value of their interaction. The interactions can be classified into five categories: non-linear attenuation, single-factor non-linear attenuation, two-factor amplification, mutual independence, and non-linear amplification. This study used this module to determine the strength of environmental factor interactions (

Table 3. Type of interaction of arguments.

Relationship Description	Interaction
$q(X_1\cap X_2)<Min\left[q(X_1),q(X_2)\right]$	non-linear weakening
$Min\left[q(X_1),q(X_2)\right]<q(X_1\cap X_2)<Max\left[q(X_1),q(X_2)\right]$	single-factor non-linear attenuation
$q(X_1\cap X_2)>Max\left[q(X_1),q(X_2)\right]$	two-factor enhancement
$q(X_1\cap X_2)=q(X_1)+q(X_2)$	mutually independent
$q(X_1\cap X_2)>q(X_1)+q(X_2)$	non-linear enhancement

).

2.3.5. PLUS Model

The current study examined the regional patterns of the landscape ecological risk of different future scenarios by integrating the landscape ecological risk assessment model with the PLUS model (Figure 3). In the PLUS model, the land expansion analysis technique (LEAS) uses random forest classification (RFC) to determine the proportion of land expansion between two land use periods. It also extracts the development probability for each land use type by sampling from the increasing proportions. The PLUS model employs multi-type random patch seeds and a threshold reduction mechanism to generate patches in a spatiotemporal dynamic manner. It calculates the overall probability within the development probability constraint by simulating the automatic generation of patches using a cellular automata framework [64]. Therefore, this study utilised land use data from the Kuye River Basin for the years 2000, 2010, and 2020 as its foundation. We then simulated the land use conditions in 2020 and compared the results with the actual land use data from that year to validate the accuracy of the PLUS model’s simulations. After validation, we proceeded to simulate the land use structure in 2030. The land use data obtained from the simulation were utilised to compute the landscape ecological risk in 2030. The drivers (Table 1) for land use modelling were classified into socio-economic and natural factors.

(1)

Simulation Parameters and Neighbourhood Weight Settings for the CARS Module

The simulation parameters for the CARS module of the PLUS model were configured as follows: The new patch attenuation threshold was set to the default value of 0.5, the patch expansion coefficient was set to 0.3, and the seed percentage was set to 0.02. Markov chain projections captured the demand for each category during the accuracy validation phase in 2020.

The neighbourhood weight parameter measures the intensity of expansion for each land category, ranging from 0 to 1. A higher value indicates a greater capacity for expansion within the land category while a lower likelihood of transformation into the other land categories. It was used to indicate the ability of each type of land to expand, driven by both internal and external factors. Due to the complex relationship between the driving factors of land use and land use change, it is impossible to calculate each land type’s expansion intensity directly. However, the historical change pattern of each land use type in the Kuye River Basin can be used to indicate the expansion capacity of each land use type. Therefore, this study used the land use data from 2000, 2010, and 2020 to determine the degree of expansion for each land type. The results of the neighbourhood weight calculations are shown in Figure 3 and as follows.

$$X^{\ast }={\displaystyle \frac{X-X_{\min }}{X_{\max }-X_{\min }}}$$

(6)

where X* represents the standardised value of deviation; X is the area of change of each land category between different years of land use data; X_max is the maximum value of the change of the area of all land categories; and X_min is the minimum value of the change of the area of all land categories.

(2)

The Setting of Land Conversion Cost Matrix Parameters

The cost conversion matrix represents the conversion rules between land use classes. It indicates whether land classes can be converted to each other [65]. The matrix value is 1 if a land use class can be converted to another land class, and it is 0 if it cannot.

Table 4. Land conversion cost matrix setting.

Land Use Type	Farmland	Woodland	Grassland	Waters	Construction Land	Unutilized Land
Farmland	1	1	1	1	1	1
Woodland	1	1	1	0	1	1
Grassland	1	1	1	1	1	1
Waters	1	0	1	1	0	1
Construction land	1	1	1	0	1	1
Unutilized land	1	1	1	1	1	1

shows the transfer cost choices for the accuracy validation phase in the Kuye River Basin.

(3)

Simulation Accuracy Test

Land use modelling accuracy is usually tested using Kappa and FoM coefficients. A Kappa coefficient greater than 0.8 signifies optimal results [66]. The FoM coefficient is used to quantitatively assess simulation accuracy at the metacellular scale. A higher FoM value indicates greater accuracy in the simulation results, although its median value generally ranges between 0.01 and 0.25 [67]. The formula for the FoM coefficient is as follows.

$$FoM={\displaystyle \frac{B}{A+B+C+D}}$$

(7)

where A represents the area of error caused by the actual land use change in the Kuye River Basin, but it is predicted to remain unchanged; B represents the area of the area predicted accurately; C represents the area of the area of error caused by the wrong prediction; and D represents the area of the area of error caused by the change in prediction results even though there was no change in the actual change in land use.

(4)

Future Development Scenarios

The study established two future development scenarios to mimic the land use situation 2030, considering regional characteristics and development realities and land use data from the base period (Figure 4d).

Natural development scenario: This scenario implies that the geographical distribution of land use in the Kuye River Basin will follow a linear trajectory based on its historical development. This scenario examines land use changes and their driving factors, including natural and socio-economic factors, for the years 2000, 2010, and 2020. It does not consider the constraints imposed by policies, regulations, land use planning, urban development planning, and other factors. It forecasts the demand for each land use type in 2030 under a natural state using the Markov chain. This forecast is a parameter for land use demand in the PLUS model. This theoretical development scenario is the basis for additional scenarios.

Farmland protection scenario: Due to rapid urbanisation, farmland areas and the number of farmers have decreased. In the Kuye River Basin, only a tiny fraction of the population lives in rural areas. We implemented the Kuye River Basin Comprehensive Plan strategies to preserve agricultural land and convert forests back to farmland. In this scenario, we predicted the trajectory of land use change and the associated trends in landscape ecological development. The farmland protection scenario specifically targets preserving stable and premium farmland. Therefore, the farmland data from 2000, 2010, and 2020 were overlaid and identified as the long-term stable farmland within the watershed. Using the Agricultural Land Classification Regulations and previous studies [66,67,68], we identified farmland with a slope less than 6° as premium farmland. Stabilised and premium farmland were combined to form the limited conversion area.

3. Results

3.1. Spatiotemporal Distribution of Land Use

This essay highlights the strong correlation between land use change and social, economic, and environmental issues in the Kuye River Basin. The transfer and spatial dynamics of land use play a crucial role in altering both natural landscapes and human social contexts. The land use analysis showed that farmland, grassland, and construction land were the dominant land use types in the Kuye River Basin in 2000, 2005, 2010, 2015, and 2022 (Figure 4b). The three land use types accounted for approximately 85.63% of the total land area. The woodland and grassland areas together were the largest, totalling 602,576.5234 hm², 612,346.2168 hm², 620,415.5156 hm², 602,723.5285 hm², and 576,075.5507 hm² in 2000, 2005, 2010, 2015, and 2022, respectively, accounting for more than 65% of the total area. The western part of the northwestern and central areas exhibited a higher intensity level, resulting in a distinctive land use landscape pattern within the basin. Social and environmental variables have mainly driven it. The social impact stems from enforcing policies to convert farmland back to forests and to restrict grazing in the mountains since 2000. This has led to a significant increase in the basin’s forest and grassland areas. As for natural factors, there has been a noticeable precipitation increase in the Kuye River Basin. In addition, large areas of steep slopes and wasteland in the basin have been converted to woodland, improving the plant community hierarchy [69]. This suggests that woodland and grassland are the dominant landscape types in the watershed.

The farmland area constantly decreased by 174,236.2982 hm², 158,599.294 hm², 151,845.0701 hm², 149,024.3916 hm², and 146,599.1395 hm² in 2000, 2005, 2010, 2015, and 2022, respectively. The corresponding proportions of farmland in these years were 20.00% (2000), 18.20% (2005), 17.43% (2010), 17.10% (2022), and 16.83% (2000). The spatial distribution of farmland was relatively even in the watershed. Implementing the afforestation and farmland reforestation policy reduced farmland by 20.00% to 16.83% from 2000 to 2022.

The percentage of construction land, including urban land, rural settlements and other construction land areas, was smaller than grassland, although it rapidly increased. Over 22 years, the area of other construction land increased by 56,107.8799 hm². In 2000, it was 255.1545 hm² (0.03% of the total area). By 2005, it grew to 3499.402 hm² (0.40%); in 2010, it reached 11,670.263 hm² (3.15%). In 2015, the area expanded to 34,497.6585 hm² (3.96%); by 2022, it totalled 56,363.0344 hm² (6.47%). Over 22 years, the urban land area increased by 3449.8061 hm², and the rural settlement area increased by 575.6738 hm². In 2000, the urban land area was 1870.9991 hm² (0.21% of the total), and the rural settlement area was 6522.7291 hm² (0.75%). By 2005, the urban area grew to 3501.8804 hm² (0.75%), with the rural area unchanged. In 2010, the urban land area reached 5618.5674 hm² (0.64%), and the rural settlement area expanded to 10,156.0544 hm² (1.17%). In 2015, the urban land area increased to 6099.7272 hm² (0.70%), and the rural settlement area increased to 10,580.2773 hm² (1.21%). By 2022, the urban land area remained the same, while the rural settlement area decreased to 6522.7291 hm² (0.21%).

The primary cause of the extensive growth in other development land is the allocation of residential land for coal miners. The expansion of production and residential facilities, including transportation and support services, has substantially increased the extent of other development land. In addition, socio-economic growth and urbanisation have significantly expanded urban land and rural communities. The expanded built-up areas are mainly concentrated in the watershed’s central, northern, and northeastern parts. Overall, the extent of non-territorial areas showed a growing trend, mainly in the watershed’s north and northwestern regions. Furthermore, the importance and area of the watershed, a crucial resource for social production and the biological environment, steadily decreased. This included shrinking river channels, lakes, reservoirs, pits, ponds, and riverbank areas. In 2022, the areas were 6495.4066 hm² for river channels, 973.0268 hm² for lakes, 533.5782 hm² for reservoirs, and 13,735.2207 hm² for pits and ponds. Coal mining in the Kuye River Basin has lowered the groundwater level and significantly reduced soil water content, resulting in decreased runoff. This activity has also caused desiccation, further reducing basin runoff. Consequently, farmland and grassland now cover much of the land use landscape. Additionally, construction land has rapidly increased, while the area of water bodies has significantly decreased.

3.2. Land Use Dynamic Structure

Between 2000 and 2010, the Kuye River Basin experienced notable land use changes. Woodland, unutilized land, and construction land increased, while farmland, grassland, and water bodies decreased. Grassland, farmland, and woodland proportions were 89.15% in 2000, 88.48% in 2005, and 88.63% in 2010, indicating that these were the dominant land uses before 2010. Water bodies decreased consistently. Specifically, farmland decreased by 0.2879%, from 151,845.0701 hm² to 146,599.1395 hm². Forest area decreased by 0.372%, from 49,842.0514 hm² to 47,616.5412 hm². Grassland decreased by 0.6151%, from 570,573.4642 hm² to 528,459.0095 hm². Water bodies decreased by 0.4823%, from 23,072.5271 hm² to 21,737.2323 hm². The rapid industrial development, especially coal mining, has significantly contributed to the depletion of forest and grassland ecosystems. The basin is rich in coal, silica sand, and other minerals, predominantly found in hilly areas.

Part of the basin is in mountainous regions, and coal mining has significantly affected surface vegetation cover [70]. In addition, coal mining activities and extensive development have disrupted the soil structure by disturbing the soil surface, increasing the risk of soil erosion and drought in farmland. In addition, they have directly encroached on farmland.

On the other hand, the area of non-agricultural land increased from 48,588.7927 hm² in 2010 to 58,120.255 hm² in 2022, with a growth rate of 1.6347%. Construction land increased from 27,444.8848 hm² in 2010 to 68,782.2425 hm² in 2022, with a growth rate of 12.55%. The Kuye River Basin is a vital national hub for energy, heavy, and chemical industries, driving robust economic growth in the region [71]. Erdos and Yulin City are economically vibrant regions in China. The Kuye River upstream of Runlong Bay features an industrial park in Ordos City. The stretch from Runlong Bay to Shenmu is densely packed with industrial parks, highlighting advanced economic development and significant industrialisation.

The area of developed and other lands significantly increased, while other land types decreased to varying degrees. The developed land was primarily transferred from woodland and grassland, accounting for 5.17% and 28.41% of the increase, respectively. Escalating human activities, urbanisation, and rapid population growth have driven the expansion of developed land. Water resources, essential for ecosystems and economic development, decreased by 4046.3477 hm² from 2000 to 2022. This decline was due to agricultural irrigation, inefficient domestic water use, and water consumption for expanding forests and grasslands.

3.3. Landscape Ecological Risk Assessment Results

3.3.1. Landscape Pattern Index

Between 2000 and 2022, there was a consistent annual increase in the number of patches, while the density of patches showed a pattern of initial decrease followed by a rise (Figure 5). Specifically, patches increased from 4713 in 2000 to 6522 in 2022. However, the density of patches across all land types gradually declined from 0.0945 in 2000 to 0.0771 in 2015 before slightly rebounding to 0.0788 in 2022. This suggests that landscape fragmentation became increasingly severe after 2015 due to the expansion of human disturbance and urban development in the watershed, resulting in a more dispersed distribution of all landscape types. The changes in the landscape separation index mirrored the pattern of patch density, initially decreasing and then increasing, reaching a peak of 1.2252 in 2022. This means that landscape separation was most pronounced in 2022, indicating an increased fragmentation due to human activities. The landscape loss degree index showed a steady increase from 2000 to 2022, which means the natural attributes of landscape ecosystems experienced significant damage in the Kuye River Basin. Each year’s specific loss degree indices were 0.2391 (2000), 0.2588 (2005), 0.2750 (2010), 0.2827 (2015), and 0.2904 (2022). Overall, the landscape pattern index showed that human activities consistently affected the watershed, resulting in a more complicated and fragmented landscape.

Grassland, woodland, and farmland were the three main landscape types (Figure 5). However, the number of farmland patches was significantly higher than that of woodland and grassland, indicating that farmland was more dispersed, while woodland and grassland were more concentrated. At the same time, farmland, woodland, and grassland showed higher levels of dominance and loss than the other landscape types. The three land use types remained the primary landscape types in the Kuye River Basin, consistent with our previous analysis of land use structure (Section 3.2). The higher degree of loss also indicated more significant pressures. The expansion of urban structures and accelerated industrialisation have escalated the demand for land designated for development. To mitigate the environmental impact of industrial zones on densely populated areas, it is essential to construct urban industrial zones at a significant distance from residential areas. However, this approach leads to a higher degree of fragmentation. From 2010 to 2022, urbanisation and the expansion of industrial and mining activities, particularly coal mining, have gradually increased the area of construction land patches in the Kuye River Basin. In particular, the number of increased patches during 2010–2022 was significantly higher than that observed from 2000 to 2010. In addition, the construction land separation index has been consistently decreasing, suggesting a more concentrated distribution of construction land. In particular, the data from 2010 onwards indicated that human-induced disturbance became more severe, with a faster increase in the quantity and density of disturbed areas. The degree of clustering of construction land has increased. In contrast, the degree of dispersion has decreased, resulting in a shift from initial dispersion to current concentration in the construction land development scale.

The total area of water patches formed by Ulanmulun River, Xiaoniu River, and 19 other tributaries decreased by 4041.81 hm². Thus, the degree of water landscape loss was significant, leading to water vulnerability. This indicates that the water bodies were more vulnerable to environmental changes. In summary, the landscape pattern indices indicated a steady increase in built-up areas over time. Additionally, the landscape became more complex and fragmented. These data highlight the unique characteristics of different landscapes and the impact of human activities and environmental changes.

3.3.2. Spatiotemporal Dynamics of Landscape Ecological Risk

Landscape ecological risk is generally classified into five levels: low, lower, medium, higher, and high, indicating varying risks associated with different strategies and management measures [72]. According to the natural breakpoint methods of the classification of natural breaks by ArcGIS, the landscape ecological risk of the Kuye River Basin was automatically divided into five levels by ArcGIS and was characterised by spatial distribution and area (Figure 6). This classification was used to map the spatial distribution of landscape ecological risk in the study area from 2000 to 2022, highlighting changes in the areas at risk.

The areas of high ecological risk were primarily located in the northern parts of the basin. The reduced and low-ecological-risk areas were scattered across the basin. This high ecological risk area featured significant elevation changes and pronounced latitudinal and vertical zonation. It also served as a vast expansion zone for construction activities. The human-induced disturbances between 2000 and 2022 have resulted in a scattered, fragmented, and disjointed distribution of landscape types within this area. The fragmentation of landscape patches has disrupted ecosystem integrity and stability, leading to significant ecological risk. The Kuye River Basin features a low-ecological-risk area extending in all directions, dominated by stable and resilient woodland and grassland with high landscape integrity. Moderate ecological risk zones were scattered throughout the watershed.

3.4. Driving Factors of Landscape Ecological Risk

The intensity of human activities is closely linked to landscape patterns, directly and cumulatively influencing landscape changes. Topographic factors significantly shape land use patterns, while climate change, particularly air temperature and precipitation, greatly affects ecological processes and landscape functioning. Evaluating ecosystem responses to disturbances and assessing ecosystem sensitivity and instability can be achieved by studying vegetation changes. We investigated the influence of socio-economic and natural environmental factors shown in Table 1. These factors represent human activities, ecological conditions, terrain, and climate (Figure 7).

3.4.1. Power of Determinant

The geographical detector model was used to assess the influence of various indicators on landscape ecological risks. The factor detector quantitatively evaluated the impact of these factors on the overall landscape ecological risk pattern using Q values and examines their temporal variations. The results, including Q values and rankings, are presented in Figure 8, showing the contribution rate of each factor in 2000, 2005, 2010, 2015, and 2022.

Various factors influenced the spatial pattern of landscape ecological risk in the Kuye River Basin. The main determinants were socio-economic factors, including road distance and night light. From 2000 to 2005, road distance had the highest influence factor with a steady upward trend, with contribution values of 0.2538 and 0.2782. After 2010, night light surpassed road distance and became the primary influencing factor. The contribution reached its peak in 2010 and then decreased. Despite this decreasing trend, it remained the primary contributor to the area’s ecological risk, with values of 0.4538, 0.3420, and 0.293. The following determinants were the distance from the city and the precipitation. Therefore, precipitation was the primary natural factor determining the landscape ecological risk in the Kuye River Basin. Among the nine factors, precipitation consistently ranked in the top five contributors. Specifically, in 2000 and 2022, it held positions of 2 and 3, respectively. In addition, distance from the city was the third socio-economic factor influencing landscape ecological risk, followed by distance from the road and nighttime lighting. Except in 2000, the role of distance from cities was consistently ranked third. This clearly illustrates the significant impact of development growth on landscape ecological risk.

Air temperature, DEM, and slope showed a slightly weaker influence on landscape ecological risk, suggesting that socio-economic factors played a more significant role in determining landscape ecological risk in the Kuye River Basin. Topography, DEM, and climate type are important natural variables strongly influencing the regional landscape and shaping various ecosystems adapted to their environments. Different landscape types also affect the ecological risk of the associated environment. From 2000 to 2022, the average air temperature had a contribution value of 0.0721, a DEM of 0.0374, and a slope of 0.0210. The interaction of natural elements, such as the shape of the land and weather conditions, created multiple ecosystems in the watershed. In the Kuye River landscape, the distribution of ecological risks was characterised by low concentration.

3.4.2. Interactive Detection

Further interaction detection was conducted to assess whether different factors separately influenced the distribution of landscape ecological risk (Figure 8, interactive detection results for 2000, 2005, 2010, 2015, and 2022). An analysis of nine factors revealed non-linear interactions with no independent or weakened relationships. Instead, the interaction between any two factors significantly impacted the landscape ecological risk index more than individually. This indicates that multiple factors jointly influenced landscape ecological risk in the Kuye River Basin. The interaction between road distance and precipitation had the highest impact, as seen in the results for 2000–2022: 0.5618, 0.7038, 0.8486, 0.7057, and 0.6007. The significant influence of road distance and precipitation highlights the prominent effect of their interaction.

Furthermore, air temperature, DEM, and GDP showed low explanatory power individually. However, their explanatory power significantly increased when interacting with each other. Specifically, the interaction with road distance had the highest explanatory power of 0.9858, second only to the combined effect of road distance and precipitation. Road distance was the primary determinant of the geographic distribution of landscape ecological risk. The interaction between any two factors amplifies their impact on landscape ecological risk, with effects that cannot occur independently. The intersection between road distance and air temperature maintained values between 0.5618 and 0.8486, peaking in 2010 with a multi-year average of 0.6841. The road distance and GDP intersection ranged from 0.4676 to 0.7391, peaking in 2015 with a multi-year average of 0.588. The interaction between road distance and DEM ranged from 0.4944 to 0.6210, peaking in 2005 with a multi-year average of 0.4510.

3.4.3. Landscape Ecological Risk Zone Detection and Analysis

The landscape ecological risk zone detection is used to determine whether there is a significant difference between the different range intervals of each driving factor and the average landscape ecological risk. It can also identify the most appropriate type of landscape ecological area and range zoning. The experimental results of risk detection have been statistically tested at the 95% confidence level and can provide useful guidance; the results obtained are as follows (Figure 9).

Adequate precipitation is essential for vegetation growth and improves the landscape ecosystem of the Kuye River Basin. However, landscape ecological risk increases with higher precipitation. Among various factors, precipitation had the lowest ecological risk value of 1.5094 in sub-district 1 and the highest value of 1.6758 in sub-district 5. Air temperature showed the lowest risk in sub-district 2. A variable increasing trend was observed between DEM and ecological risk, with the highest risk in sub-district 5 and a lower correlation in sub-districts 1–2. Steep slopes are more prone to erosion and landslides; high precipitation can exacerbate this damage. Thus, landscape ecological risk increases significantly with slope gradient, with shallow (0–5°) and mild (6–15°) slopes showing gradual risk increases and steeper slopes showing significantly higher risks.

Regarding socio-economic factors, there was a positive correlation between the landscape ecological risk score and variables such as GDP, distance to road, distance to city, and population density. Rapid socio-economic growth was a primary factor contributing to increased landscape ecological risk in the Kuye River Basin. The rapid development increased the quantity and density of landscape patches, causing fragmentation and greater landscape loss. Consequently, landscape connectivity decreased, leading to higher ecological risk throughout the watershed.

3.5. Scenario Simulation Prediction of Future Development of Land Use and Landscape Ecological Risk

Using data from 2000, 2010, and 2020, we employed the PLUS model to predict landscape ecological risks in 2030 under two scenarios: farmland preservation and natural development. The spatial distribution of landscape ecological risk for 2030 was validated using available data, confirming the accuracy of the PLUS model with a kappa coefficient of 0.76 and an overall accuracy exceeding 85%. This high confidence level indicated that the PLUS model effectively simulated and predicted future land use and cover changes in the study area.

3.5.1. Land Use Prediction Results

Figure 10 displays the 2030 land use and land use class simulations for different scenarios in the study area using the PLUS model. The patterns of land use and land use classes in 2030 differed from those in 2020. Under natural development and farmland protection scenarios, construction land in the Kuye River Basin will increase to 3168.8605 hm² by 2030. As the basin’s total area remains constant, this expansion will reduce woodland, grassland, and water areas from 597,812.7830 hm² in 2022 to 572,250.30 hm² in 2030. However, the Kuye River Basin Comprehensive Plan, including policies on reforestation and returning land to cultivation, will expand farmland to 149,768 hm² by 2030.

The amount of vacant land will continue to increase in the densely populated, urbanised, and road-developed north, particularly in the northwestern and northeastern regions. The distribution of construction land will shift from being concentrated in the northern region in 2022 to gradually expanding southeast. The expansion of developed land in the Kuye River Basin aligns with provincial roads and subways, reflecting the influence of selected impact factors in the land use forecast. Accessible transportation facilitates potential development, while construction in the mountainous southern region with steep terrain has been unsuccessful. The increase in construction and unused land has reduced forest and grassland areas in the northeast, northwest, and southeast. Due to farmland protection scenarios, farmland will mainly expand in the central and southern areas by 2030, with no further expansion in the northern region due to its high population and urbanisation.

3.5.2. Landscape Ecological Risk Prediction Results

Using the modelling results of the land use projections, we calculated the expected landscape ecological risk of the Kuye River Basin in 2030 under two scenarios: natural development and farmland protection scenarios. The landscape ecological risk showed a consistent pattern in 2030 under both scenarios. Specifically, areas with low, higher, and high landscape ecological risk increased, while areas with lower and medium risk decreased, with varying magnitudes of change. Under the natural development scenario, lower- and medium-risk areas were 258,654 hm² and 114,274 hm², respectively, indicating more severe landscape ecological risk. In the farmland protection scenario, low and lower-risk areas were 58,017.4 hm² and 20,191.3 hm², respectively. Under the farmland protection scenario, the low-risk area was 233,632 hm², which was 162,203.2 hm² larger than the natural development scenario. Therefore, farmland protection measures resulted in a slower increase in landscape ecological risk (Figure 11).

The natural development scenario simulation results indicated that landscape ecological risk was primarily concentrated in higher- and high-risk areas, located mainly in the northern and southern regions, with some corridors in the central basin. Medium-risk areas were primarily in the central region. Overall, landscape ecological risk in the study area increased, ranging from medium to high, except for urban areas dominated by construction land. Conversely, landscape ecological risk generally decreased under the farmland protection scenario. The basin’s southern and northeastern parts exhibited low-risk levels, characterised by minimal fragmentation and sparse patches, indicating a large and connected landscape.

4. Discussion

The Kuye River Basin, a significant tributary of the Yellow River Basin, serves as a crucial component of China’s ecological security barrier in the northern region. Assessing and predicting the spatial and temporal distribution characteristics of land use structure and landscape ecological risks in watersheds and revealing their driving causes can safeguard the quality of the ecological environment, prevent and control ecological risks, and plan the key to further governance models. With population growth and the development of the coal industry, there has been a significant expansion of built-up land in the Kuye River Basin, leading to an increase in the landscape ecology of the basin, yet the regulation of ecological risks and ecological planning have lagged behind these developments. There is a need for innovative policies, planning guidance, integrated regulations, and technological innovations based on the characteristics of the natural and socio-economic development of the basin and the results of modelling of future development scenarios, which is the purpose of this study.

4.1. Analysis of Land Use and Landscape Ecological Risks

Between 2000 and 2022, woodland, grassland, farmland, and construction land were the dominant land use types in the Kuye River Basin. Woodland and grassland spanned significant portions of the northern, central, and southern areas. At the same time, farmland was concentrated in the south and central regions, serving as the primary food source for the residents of Shenmu City, Fugu District, and various districts of Ordos City. The introduction of the Kuye River Basin Comprehensive Plan, which emphasises ecological management, has improved environmental quality with increased vegetation coverage and enhanced land use structure. Programs like the “National Ecological Environment Protection Outline” [73], the “Yellow River Ecological Protection and Governance Battle Action Program” [74], and the “Inner Mongolia Yellow River Soil and Water Conservation Ecological Functions of Key Tributaries of the Kuye River Basin Ordos Project Area Governance Project” [75] have significantly expanded forest and grassland areas. Due to population growth and coal industry development, construction land has increased from 0.99% to 7.89% over 22 years. This includes housing for miners, transportation infrastructure, and other facilities, primarily in the basin’s central, northern, and northeastern parts. The rise in construction land reflects the impact of urbanisation and rapid population growth, driving further demand for development.

Due to population growth and the rapid expansion of the coal industry, the area and share of construction land have steadily increased [76]. This trend indicates that human activities, particularly urbanisation, have significantly impacted the expansion of construction land in the Kuye River Basin. Extensive coal extraction has further contributed to this growth, including infrastructure for miners, such as residential areas, transportation networks, and support services. As a result, human activities have escalated, landscape structure has become less rational, fragmentation has intensified, and ecological risk has increased. The study shows that landscape ecological risk and construction areas in the northwestern region have expanded from northwest to southeast, consistent with land use predictions. This expansion is most notable in central and southeastern areas, characterised by high human activity, population density, developed transportation, flatter terrain, and lower slopes. Urbanisation and hydropower projects have also contributed to the growth of residential and coal mining areas, leading to increased soil erosion and ecological risk. However, ecological risk regulation has lagged these developments. Therefore, this region is crucial for restoring and managing the Kuye River Basin. Future planning should prioritise regulating areas with intense human activities and high ecological risks, focusing on the northwestern half of the basin and gradually shifting to the southeast.

4.2. Analysing Natural and Social Economic Factors

Geographical detectors are advantageous for analysing study areas of any size, allowing for identifying and clearly visualising influencing factors with strong spatial representation [5]. This method outperforms alternatives in detecting and analysing interactions between components. Investigating the practicality of using geographical detectors to thoroughly analyse landscape changes caused by potential drivers is crucial for understanding the evolving ecological risks in watershed landscapes. Geographical detectors are used to determine the contribution rates of different influencing elements and to analyse the degree of influence that driving forces have on landscape ecological risk in various geographic areas. Each component has both positive and negative effects on landscape ecological risk.

The main elements influencing landscape ecological risk are distance from roads, cities, population density, and precipitation. In addition, the low-lying plain dominates the watershed and significantly influences landscape ecological risk. This is mainly due to its suitability for human settlement and intensive human activities. The expansion of urban areas and grazing activities in this region disrupts the natural ecological balance, resulting in increased landscape heterogeneity and fragmentation. This finding is consistent with our investigation of the risk and arrangement of landscape ecology. As elevation increases, human activities progressively decrease, resulting in the preservation of the biological integrity of the regional landscape and a lower level of landscape ecological risk. At higher elevations, the topography becomes very complex and human activities are mostly limited to the steep slope area. This is due to increased precipitation and favourable vegetation growth rather than a desire to minimise risk. As a result, the level of landscape ecological risk is lower in the central and southern parts of the basin, particularly in the southeastern fringe areas, than in the northern region. However, when the elevation reaches a significant height and the slope becomes excessively steep, the area’s geographical features become incredibly complex, decreasing air temperature, precipitation, and vegetation productivity. As a result, the terrain becomes more susceptible to natural disasters such as desertification, landslides, and debris flows, exacerbated by human activities at higher elevations.

It is important to note that natural elements such as DEM, slope, and air temperature have less of an impact on landscape ecology. However, when they interact with other elements simultaneously, the level of destruction increases significantly. In Section 3.4.2, we found that the effects of two factors on changes in landscape ecological risk are amplified by their interactions. Furthermore, we found that the impact of all components is interdependent and cannot occur in isolation. Therefore, our goal is to mitigate and minimise the detrimental effects on the landscape ecology of the watershed resulting from the interaction of the various factors.

4.3. Optimisation Suggestions for Future Regulatory Measures and Policy Formulation

We integrated land use, landscape patterns, and landscape ecological risk assessments in the Kuye River Basin from 2000 to 2022, using results from landscape drivers and future development scenario modelling. The following recommendations aim to guide future socio-economic development, resolve human–land conflicts, optimise land use allocation, and formulate policies and strategies for ecological and environmental management in the Kuye River Basin.

(1)

Actively promote the development of spatial master plans nationwide and expedite the implementation of “multi-planning integration [77]”. The government should abandon isolated planning concepts and accelerate the integration of various plans, including comprehensive land use, urban development, environmental protection, and specialised regional development plans. This approach will facilitate overall coordination and efficient integration, advance land spatial planning, improve spatial management in the basin, and optimise land resource distribution, organise land use patterns on a large scale to reduce landscape ecological hazards, and address the imbalance in land use structure caused by the rapid expansion of construction land. The Kuye River Basin, rich in mineral resources and a focal point for China’s “Jin, Shaanxi, and Mongolia Energy and Heavy Chemical Industry Base”, requires balanced planning to mitigate landscape ecological risks rather than allowing unplanned, organic development.

(2)

Strictly adhere to rigorous farmland protection measures to address land use disparity and urban-rural imbalance [78]. Optimise the arrangement of farmland and permanent basic farmland while maintaining food security requirements. Integrate the “conserving farmland and using advanced technology” strategy by implementing a comprehensive farmland protection system, safeguarding resources designated as farmland from 2000 to 2022 and high-quality agricultural land with slopes less than 6°. Strengthen control measures on farmland use, protect and develop high-quality farmland, prevent the conversion of farmland to non-agricultural uses, regulate non-food uses, and accurately delineate permanent basic farmland reserve zones. This approach aims to conserve valuable farmland, regulate construction land, reduce landscape fragmentation, and minimise ecological risk across the watershed.

5. Conclusions

Over the past 22 years, profound changes in land use have significantly altered the physical features and natural processes within the Kuye River Basin, posing potential risks to the region’s long-term development sustainability. This study confirms that natural conditions and anthropogenic activities have substantially influenced the spatiotemporal distribution of ecological risks across the basin. The delicate and fragile ecosystem in the upper Kuye River Basin exhibits a higher probability of ecological risk. Additionally, metropolitan areas in the northwest show clusters of heightened risk, underscoring the detrimental effects of human disturbances on ecosystems.

Projections based on the landscape ecological risk simulation under the natural development scenario indicate a continued expansion of areas with increasing ecological risks. Currently, nearly half of the study area requires urgent risk management intervention. To achieve sustainability, it is imperative to protect critical areas and regulate the rate of development. This can be achieved by improving adaptive management of ecological risks through innovative policies, planning guidance, integrated regulations, and technological innovations suitable for remote areas aimed at ecosystem restoration. The results of this study provide essential empirical evidence to support decision-making processes directly. These findings underscore the need for comprehensive strategies to mitigate ecological risks and promote sustainable development within the Kuye River Basin. Furthermore, the above study model and regional management strategies can be applied to other tributaries, particularly in regions that share the Kuye River Basin’s physical and socio-economic development characteristics.

Additionally, the potential harm and hazard that human activity and natural environmental conditions inflict on landscape ecosystems are referred to as landscape ecological risk. The effects of natural disasters and other small elements are also its driving forces. However, it was not included in this piece because it was hard to obtain statistics on it. Additionally, this is something that needs to be taken into account for future studies.