Next Article in Journal
Use of Amino Acids and Organic Waste Extracts to Improve the Quality of Liquid Nitrogen–Calcium–Magnesium Fertilizers
Previous Article in Journal
Geomechanical Characterization of Unwelded Volcanic Bimrock Materials for Sustainable Slopes: Application to Road Instability Problems in the Western Cordillera of Ecuador
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 17
Issue 15
10.3390/su17157082
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Upscaling Soil Salinization in Keriya Oasis Using Bayesian Belief Networks

by
Hong Chen
1,
Jumeniyaz Seydehmet
1,2,* and
Xiangyu Li
1
1
College of Geographical Science and Tourism, Xinjiang Normal University, Urumqi, 830017, China
2
Xinjiang Arid Area Lake Environment and Resources Laboratory, Key Laboratory of Xinjiang Uygur Autonomous Region, Urumqi 830054, China
*
Author to whom correspondence should be addressed.
Sustainability 2025, 17(15), 7082; https://doi.org/10.3390/su17157082
Submission received: 18 May 2025 / Revised: 31 July 2025 / Accepted: 1 August 2025 / Published: 5 August 2025
Browse Figures
Versions Notes

Abstract

Soil salinization in oasis areas of arid regions is recognized as a dynamic and multifaceted environmental threat influenced by both natural processes and human activities. In this study, 13 spatiotemporal predictors derived from field surveys and remote sensing are utilized to construct a spatial probabilistic model of salinization. A Bayesian Belief Network is integrated with spline interpolation in ArcGIS to map the likelihood of salinization, while Partial Least Squares Structural Equation Modeling (PLS-SEM) is applied to analyze the interactions among multiple drivers. The test results of this model indicate that its average sensitivity exceeds 80%, confirming its robustness. Salinization risk is categorized into degradation (35–79% probability), stability (0–58%), and improvement (0–48%) classes. Notably, 58.27% of the 1836.28 km2 Keriya Oasis is found to have a 50–79% chance of degradation, whereas only 1.41% (25.91 km2) exceeds a 50% probability of remaining stable, and improvement probabilities are never observed to surpass 50%. Slope gradient and soil organic matter are identified by PLS-SEM as the strongest positive drivers of degradation, while higher population density and coarser soil textures are found to counteract this process. Spatially explicit probability maps are generated to provide critical spatiotemporal insights for sustainable oasis management, revealing the complex controls and limited recovery potential of soil salinization.

1. Introduction

Land degradation is a global environmental problem that has an important impact on production, life, and the ecological environment [1,2]. Currently, saline land is widely distributed in more than 100 countries around the world, and it is estimated that about 7% of the Earth’s land is at risk of salinity, while 20% of irrigated areas are threatened by secondary salinity of soil [3,4]. Soil salinization is more widely distributed in arid and semi-arid regions [5]. China’s saline-alkali land area is about 3.6 × 107 hm2, mainly distributed in six provinces in northwest China [6]. In addition, drought, low rainfall, and high evaporation greatly hinder the sustainable development of agriculture in these areas [7,8]. Therefore, the prevention and control of salinization by prediction has become one of the important ways to solve global environmental and food security problems [9,10].
Soil salinization is a prominent contributor to land degradation. According to the cause, it can be divided into primary and secondary salinization. It occurs primarily through the upward migration of soluble salts from deeper soil layers or groundwater to the surface via capillary action, culminating in salt accumulation in the topsoil as water evaporates [11,12]. This phenomenon significantly diminishes agricultural productivity and environmental quality by reducing crop yields, soil fertility, vegetation coverage, and biodiversity [13,14]. Anthropogenic activities and geomorphological conditions play crucial roles in driving soil salinization in oasis regions [14,15], which exhibits notable dynamics and uncertainty across both horizontal and vertical spatial scales [16,17,18]. With the intensification of global warming and the continued expansion of irrigated land, soil salinization has become an increasingly complex ecological challenge, particularly affecting the sustainable development of arid oases [19]. Research has identified multiple sources of secondary salinization. These include the following: (1) parent geological material and inherent soil characteristics, (2) shallow, salt-rich groundwater, and (3) irrigation with saline water [20]. In studying spatial and temporal variations of salinity, 11 key influencing factors have been highlighted—most notably groundwater depth, surface water evaporation, irrigation practices, groundwater mineralization, and rainfall [21]. Notably, shallow groundwater depth, groundwater salinity, and agricultural practices such as over-irrigation and insufficient drainage are primary drivers of salinity expansion. The spatial pattern of soil salinity typically increases from oasis cores outward [22,23,24]. In summary, soil salinization is the result of the joint action of human activities and natural factors.
Several spatial modeling approaches have been employed to investigate salinization patterns. In the Manas River Irrigation Area, for instance, researchers found higher salt concentrations in both upstream and downstream regions and a decrease in variability with increasing soil depth [21]. In Egypt’s Siwa Oasis, a deep learning-enhanced U-Net algorithm revealed widespread salinity increases in all directions [25]. Given that salinization evolves at macro-environmental scales, a comprehensive and dynamic modeling approach is essential for supporting sustainable land management strategies [17]. Moreover, the increasing complexity of natural systems contributes to greater uncertainty in salinization risk distribution [26]. However, geostatistical techniques like probability kriging and ordinary cokriging have been applied to produce salinity probability maps [27,28]. But they often fall short in addressing the multidimensional interactions of contributing factors. Bayesian networks (BNs), when integrated with geospatial tools such as ArcGIS, have emerged as effective tools for modeling and predicting complex environmental systems [29,30,31]. The strength of BNs modeling lies in its ability to incorporate diverse quantitative and qualitative datasets, enabling the analysis of probabilistic causal relationships across scales [32]. BNs have been successfully applied to enhance the understanding and modeling of diverse environmental issue [33,34,35]. And BNs are becoming more and more mature as a tool for predicting soil salinization, for instance, scholars employed a combined approach of BNs and random forest regression to predict the electrical conductivity (EC) of saturated soil extracts in the coastal region of South Sindh Province, Pakistan (9707 km2) during the period from 2014–2015 to 2020–2021 [36]. BNs have been used by scholars to quantify the uncertainty in salinity associated with reducing the EC in peri-urban circulating water [37]. And a study using BNs to assess the use of water cycling to control soil salinity has also come to fruition [38]. Some studies have used BN models to predict the overall regional salinization status of the Keriya Oasis [39]. Despite the potential of such approaches, few studies have applied BNs in combination with field-based monitoring and remote sensing data to analyze soil salinization in oasis environments. This presents both a methodological gap and an opportunity to enhance our understanding of salinization dynamics in highly vulnerable regions [40]. And most of the existing studies focus on the region as a whole, but ignore the internal differences of oases.
The Keriya Oasis is located at the southern edge of the Taklimakan Desert in Xinjiang. It has an extremely arid temperate continental climate, with scarce precipitation and intense evaporation. The annual average precipitation is less than 50 mm [41,42]. Moreover, shallow groundwater drives salt accumulation; improper irrigation methods aggravate salt redistribution; and extreme arid climate amplifies ecological risks and makes it a salinization disaster area [43]. Therefore, the main objectives of the current study were to achieve the following: (1) Develop a GIS-integrated BNs model to analyze the spatial patterns and trends of soil salinization in the Keriya Oasis. (2) Estimate the likelihood of soil salinization trends—categorized as Improvement (Reduction in Salinity), stability, or degradation—and examine their relationships with eco-environmental driving factors. (3) Explore small-scale regional differences from the inside of the oasis, which provides some theoretical basis for soil salinization control the Keriya Oasis.

2. Materials and Methods

2.1. Study Area

The Keriya Oasis (81°09′–82°00′ E, 36°44′–37°12′ N) is located in the southern region of the Xinjiang Uygur Autonomous Region (XUAR), China (Figure 1) [39]. This oasis is primarily distributed across the alluvial plain of the Keriya River, covering an area of approximately 2000 km2. The river serves as its primary water source, forming the fundamental hydrological basis of the oasis [44,45]. The region is characterized by an arid continental climate, with an average annual precipitation of only 44.7 mm, a mean annual temperature of 11.6 °C, and a potential evaporation rate reaching 2498 mm. The soil structure is loose, the soil texture is sand, and the fertility is relatively low, while shallow groundwater is highly mineralized. The topography is relatively flat, further influencing hydrological and soil conditions [46]. The permanent population of the Keriya Oasis is approximately 283,000, concentrated in the southern part of the oasis, and it has a history of irrigation agriculture of about 3500 years [44]. The region faces substantial environmental pressures due to high population density and increasing anthropogenic activities. Rapid population growth and economic expansion have significantly heightened water demand, leading to unsustainable irrigation practices and insufficient drainage infrastructure [42]. These factors have exacerbated soil salinization, contributing to its expansion toward the periphery of the oasis and posing critical challenges for sustainable land use and ecological stability.

2.2. Data Sources

Modeling data for 13 indicators were extracted from three types of sources. The first includes remote sensing data sources such as the Earth Resources Data Cloud, National Tibetan Plateau Data Center, Land Scan dataset provided by the Oak Ridge National Laboratory under the U.S. Department of Energy, European Space Agency, OpenStreetMap, and Geospatial Data Cloud. The second source consists of published papers. The third source is from five field monitoring campaigns. Detailed information is provided here (Table 1). The precision of different data in the article has been resampled to 1 × 1 km, and the data of precipitation, evapotranspiration, average temperature, and soil moisture content are all averaged annually, but the NDVI data uses the maximum value for each month.
In this study, we extracted 6192 data points from Remote Sensing data sources and 288 data points from published papers using the extraction tool of ArcGIS 10.8, covering the years 2000 to 2023 (Figure 2) according to the 36 soil sampling points presented in Figure 1. There is a high regional consistency between the published data and the measured data. We then integrated this with 148 field monitoring data points spanning the years 2012 to 2021, and utilized the BNs software package NeticaTM (Version 5.24, http://www.norsys.com/.Netica (accessed on 15 May 2024)) to model the EC trend of each sampling point. And the spatial analyst tool ArcGIS 10.8 was used for spatial interpolation analysis. The likelihood of soil salinization trends and their relationships with eco-environmental driving factors was analyzed by Partial Least Squares Structural Equation Modeling (PLS-SEM) in R software (R version 4.2.0).

2.3. Field Soil Sampling and Laboratory Analysis

In this study, we conducted five field sampling activities using the Global Positioning System (GPS) to investigate the characteristics of the ecological transition zone in arid areas. The sampling design followed the principle of spatial stratified random sampling, stratifying into the core area of the oasis, the oasis–desert ecotone, and the desert area. Eight to ten random coordinate points were generated for each stratum to set up sample plots [49,50]. The soil profile was sampled using the three-point method. Specific sampling details are as follows: in July 2012, 24 soil profile samples were collected; In May 2013, 24 soil profile samples were collected. In May 2014, 36 soil profile samples were collected. In May 2015, 36 soil profile samples were collected. In July 2021, 26 soil profile samples were collected. There are a total of 148 samples. To ensure the consistency of the data as much as possible, no more diversity points were added. In this study, the surface soil was taken as the research object, and only 148 samples from the surface layer (0–20 cm) with the same spatial location in all sampling periods were selected. A series of laboratory procedures was carried out on the collected soil samples. Soil surface samples were sorted, air dried, ground, crushed, and screened with a 200-mesh sieve. 20 g of the screened sample was mixed with distilled water (conductivity < 20 mS/cm) in a 1:5 ratio, sealed, shaken over 200 times, and left to stand for 6 h. The supernatant was then filtered, and the filtrate was stored in a new container. Finally, the conductivity of the filtrate was measured using a conductivity meter (Leici DZS-706F-A) to obtain the actual conductivity data (Table 2).

2.4. Modeling Approach

The methodological framework of this research work comprises four parts (Figure 3). The first involves the collection and processing of remote sensing-based meteorological, socioeconomic, LUCC, and biophysical data, along with openly published paper-based data and some field soil sample data. The second part focuses on model development through the interconnected use of DPSIR, BNs, and GIS to examine the salinization trend at sampling points. The third part entails an interpolation analysis of the spatial distribution probability of salinization. The fourth part investigates the factors influencing the salinization trend using the PLS-MES model.

2.4.1. Indicator Selection

The study employs a research model that integrates the Driving–Pressure–State (DPS) components of the Driving–Pressure–State–Impact–Response (DPSIR) sustainability framework. For example, temperature and precipitation are the main factors that directly affect surface evaporation; while human activities, through altering vegetation coverage, soil moisture content, and surface runoff, etc., indirectly influence surface evaporation, and all these ultimately act on the process of soil salinization [51,52]. To populate each Driver, Pressure, and State variable, we analyzed previously published research findings on soil salinization, which provided indicator sources for the probability model [24,53,54,55]. Based on representativeness and data availability, relevant spatiotemporal indicators were identified (Figure 2), and three states for each indicator were established using equal interval classification methods to determine threshold values (Table 3).

2.4.2. Bayesian Network

A BN elucidates the dependencies among variables and enables predictions based on prior probabilities [56]. Also referred to as a belief network model, a BN links each node to a conditional probability distribution (CPD), which quantifies the influence of parent nodes on child nodes [30]. The parent node refers to the node that has a direct influence on other variables in the network structure, and the child node responds to the changes of the parent node; a complete set of parent-child nodes is called an evidence chain [57,58]. In practical inference, BNs typically depend on the CPD of variables and the BN theorem for calculations (Figure 4). The fundamental expression of the BN theorem is as follows:
P ( α │ β ) = ( P ( α │ β ) P ( α ) ) / P ( β )  
where P ( α ) and P ( β ) denote the probabilities of independently observing α and β , without accounting for their mutual influence; P ( α │ β ) signifies the conditional probability of α given β ; and P ( α │ β ) / P ( β ) represents the Bayesian factor or likelihood ratio.

2.4.3. Model Validation

BN node hypothesis testing was employed for model performance evaluation. This method assesses whether significant synchronous relationships exist between variables by introducing prior distributions and observed data to compute posterior distributions. It is particularly well-suited for large datasets where clear parent-child node relationships have been established based on existing knowledge [59,60]. We conducted BN hypothesis testing at 36 sampling points. Each test was performed node-by-node using observational data from 2000 to 2023, comparing prior and posterior probability distributions 300 times for each point (totaling 10,800 times) to analyze variable synchronization characteristics. Its equation is as follows:
B F 10 = ( P ( D │ H 1 ) ) / ( P ( D │ H 0 ) )
where, B F 10 represents the BN Factor, H 0 denotes the null hypothesis, and H 1 denotes the alternative hypothesis. P ( H │ D ) is the marginal likelihood of observing the data D under hypothesis H, which integrates over all possible parameter values.

2.4.4. Tension Spline Function

Spline interpolation is a deterministic interpolation method that can construct the model’s predicted results as a smooth surface passing through all the known points. There are two types of spline interpolation: regularization and tension. In this study, the tension spline interpolation method was adopted. Based on correlation and error analysis, it was found that the performance of the tension spline method was superior to that of the regularized spline method [61,62]. The tension spline function approximates a complex function or dataset using a series of polynomial segments. These segments are connected at specific nodes, forming a continuous curve [63]. The mathematical expression is as follows:
z x , y = ∑ i = 1 n λ i R r i , ϕ + T x , y
where z ( x , y ) represents the interpolated value at spatial location ( x , y ) ; λ i is the weight coefficient; R r i , ϕ is the radial basis function; r i is the Euclidean distance between the ith known point and ( x , y ) ; ϕ is a parameter that controls the shape of the function; and T x , y is the trend function, which describes the global trend of the data.

2.4.5. PLS-SEM Model

Partial least squares structural equation modeling (PLS-SEM) is one of the most widely used methods of multivariate data analysis [64]. The PLS-SEM model analyzes the causal relationships between multiple variables with the covariance matrix as the core, including measurement models and structural models as two basic models [65]. The measurement model describes the relationships between the latent variables and their corresponding observed variables with the following equations:
Υ = Λ y η + ε
X = Λ x ξ + δ
Equations (4) and (5) are the endogenous and exogenous variables, respectively. Λ x is the factor loading matrix of the exogenous measurable variable on the exogenous latent variable ξ ; Λ y is the factor loading matrix of the endogenous measurable variable on the endogenous latent variable η ; δ and ε are the measurement model error terms.
The structural model reflects the causal relationship between endogenous latent variables and exogenous latent variables with the following expressions:
η = B η + Γ ξ + ζ
where η is the endogenous latent variable, ξ is the exogenous latent variable, and B denotes the effect of the exogenous latent variable on the endogenous latent variable. Γ is the path coefficient, which measures the effect of the exogenous latent variable on the endogenous latent variable, and ζ is the structural model residual term [66].

3. Results

3.1. Model Sensitivity Analysis

The model sensitivity analysis was performed for fifteen chains of nodes using BNs node hypothesis testing (Equation (2)). The chains a, b, c, f, g, h, j, l, and n indicated relatively higher accuracy with an error rate of <20%; the chains i, k, m, and o showed accuracy with an error rate of <25%; and the chains d and e had an error rate between 25% and 35%. The highest sensitivity was found in chain a, while the lowest sensitivity was found in chain e, which requires additional investigation (Table 4).

3.2. Spatio-Temporal Changes of Soil Salinization

3.2.1. Temporal Changes of Soil Salinization

The changes in EC by 36 points in the Yutian oasis in 2012 were compared with the measured data in 2021. Among them, the EC of Y01, Y13, Y18, and Y21 sampling points decreased by 0.02, 0.26, 0.38, and 0.38, respectively, in the two-year data comparison. The EC at sampling points Y08, Y11, Y20, and Y35 increased by 14.68, 21.44, 8.83, and 8.44, respectively, which requires additional investigation (Table 5).

3.2.2. Spatial Characteristics of Salinization Probability Trend

Based on probabilistic theory (Equation (1)), we produced BNs to study the impact of twelve factors on salinization trends at 36 sampling points in the Keriya Oasis (Figure 1 and Figure 4). The software generated a total of 1049 conditional probabilities among 15 node chains for these points (Table 4). The salinization probability for each sampling point was calculated across three types of trends: degradation, stable state, and improvement. Subsequently, using the tension spline function of ArcGIS 10.8, we produced spatial distribution maps of salinization probability in the Keriya Oasis.
Results indicate that (Figure 5) the salinization probability in the Keriya Oasis exhibits clear spatial characteristics, with larger degradation areas located along the southern margin of the oasis in both the eastern and western parts, where intense anthropogenic activities occur in flat terrain with poor drainage. Stable state areas are predominantly found in the northwest and some northern regions of the oasis, characterized by sparsely populated natural vegetation and a dominant ecotone ecosystem. The improvement areas are located near degradation zones to the north, likely due to specific rational desalinization practices. The likelihood of degradation varies from 35% to 79%, stability from 0% to 58%, and improvement from 0% to 48%. In terms of degradation, stable state, and improvement, the minimum and maximum probability order is 79% > 58% > 48%, and 31% > 0% = 0%, respectively. This indicates that the probability of salinization degradation is higher than that of the stable state, which in turn is higher than the probability of improvement. Notably, the maximum probability of improvement in the oasis is less than 50%, indicating almost no chance of salinization enhancement under the current situation.

3.3. Area Analysis of Salinization Probability Trend

If the probability of the event P (Y = 1) is ≥50%, it is predicted to occur. This study divided the salinization probability into four groups based on the median value of 50%. The spatial distribution of salinization probability indicates that, as shown in (Figure 6), the predominant state of salinization trend is expected degradation and stable state events, with improvement being unlikely.
Among these, the degradation probability of 60–80% accounts for 15.47% of the total area, covering 284.02 km2; 50–60% accounts for 42.80% of the total area, covering 785.93 km2; and 40–50% accounts for 39.24% of the total area, covering 720.64 km2. The stable state probability of 50–60% accounts for 1.41% of the total area, covering 25.91 km2, while 40–50% accounts for 9.28% of the total area, covering 170.38 km2. The improvement probability of 40–50% accounts for 3.43% of the total area, covering 62.95 km2.

3.4. Analysis of Spatial Superposition with Different Probabilities

The geospatial overlay analysis identified distinct zones with varying probabilities of salinization states, allowing for the development of predictive models. The integrated probability distribution maps of three salinization trends delineate zones that can be interpreted in terms of salinization state probabilities (Figure 7). Figure 7 illustrates spatial domains where the probability of salinization exacerbation ranges from 50% to 77.8%. At the 77.8% probability threshold for exacerbation, the combined probability of the remaining two states converges to 22.2%, indicating a near-dominance of degradation dynamics in these zones.

4. Discussion

4.1. Mechanism of Drivers Coupling the Salinization Spatiotemporal Probability

To analyze the effects of multiple factors on the trend of salinization probability through various mechanisms, we studied the complex correlations between the dynamic probability of salinization and its key spatiotemporal drivers (13 spatial variables, ranging from 2000 to 2023) using PLS-SEM. This approach is valuable for compensating for the inadequate analysis of factor interactions in studies of complex salinization problems [67].
The arid oasis is a landscape unit with higher vegetation coverage and population density, particularly in the core area of the oasis, which experiences high population density and nearly all agricultural activities. However, in this part of the region, in order to ensure a suitable living environment, a large number of human and material resources are invested to ensure that the soil salinization is under control. Relevant research results show a clear correlation between vegetation density and population [68]. People tend to settle in areas with higher vegetation coverage, suitable topography and texture for drainage, and convenient hydrological conditions for cultivating crops. As the distance from farmland increases, population density decreases [69]. Agricultural activities within the oasis influence water allocation strategies, with farmland receiving more water than the ecotone, which increases soil moisture content in the farmland [70,71,72]. Subsequently, strong evaporation occurs in farmland compared to the ecotone, but salinization is less pronounced in the farmland. Frequent irrigation helps leach salts deeper into the soil layers, while the ecotone, due to sparse precipitation, has almost no water input to leach salts from the surface into deeper layers.
The Taklimakan adjacent to Keriya Oasis is dominated by desert soil with coarse particles, large pores, and rapid water infiltration. Following irrigation or precipitation, salts tend to migrate downward with the water, leading to low soil moisture content and, to some extent, reducing the risk of salinization [73]. In contrast, clay soils, composed of fine particles with a dense pore structure and low permeability, cause water to accumulate at the surface, and subsequent evaporation intensifies the accumulation of salts in the topsoil [74]. Therefore, frequent irrigation elevates the groundwater table in the ecotone, especially in flat, low terrain and low-lying areas where slopes descend sharply. A common landscape feature is a strongly salinized barren area surrounded by high grass at higher topographical positions [75].
Long-term irrigation with low-quality water sources leads to increased soil pH, electrical conductivity, and total inorganic carbon content, while reducing microbial activity, total organic carbon, and labile carbon pools, thereby resulting in the degradation of soil quality [76]. It is evident that a higher salinization hazard is closely related to a shallow groundwater table, flat terraces, poor drainage conditions, and salt-tolerant vegetation in marginal ecotone areas [24,77].
Furthermore, previous studies indicate that an increase in cultivation duration raises soil organic matter content, while other research suggests that salt-tolerant plants in ecotone areas also contribute to increased organic matter content in salinized soils. This highlights the role of organic matter in enhancing soil fertility and self-purification ability, thereby reducing salinization risk [78,79]. However, while organic matter content is higher in ecotone areas than in desert soils, the salinization risk is also higher in these areas [80,81].
The oasis marginal ecotone has a fragile ecosystem and lower environmental capacity than the inner cultivated oasis. However, the ecotone may exhibit some improvement in salinization due to natural drainage, higher topographical positions, and a beneficial surface water system [21,82]. It is clear that different management strategies are recommended for different landscape types based on the varying mechanisms of influencing factors. Figure 8 illustrates the impact mechanism of 10 significant spatiotemporal variables on soil degradation trends in the southern margin of the oasis, covering both the eastern and western parts.
There is a significant and direct negative correlation between salinization deterioration and the combined factors of increasing population density, elevated evaporation, and soil texture transitions, with population density exerting the strongest influence. Soil organic matter is directly and positively correlated with both salinization deterioration and population density, with population density again playing the dominant role.
Slope shows a direct positive correlation with both population density and salinization deterioration, while exhibiting a direct negative correlation with NDVI. Among these, the positive correlation with salinization deterioration is the most prominent. Distance from farmland is significantly and negatively correlated with both population density and soil moisture, while being positively correlated with NDVI. Consequently, it is indirectly and positively associated with salinization deterioration. Vegetation type is significantly and directly negatively correlated with NDVI, and subsequently, it is indirectly and positively associated with salinization deterioration (Figure 8).
The improvement areas are located near degradation zones to the north, likely due to specific rational desalinization practices, and are correlated with nine significant impact factors (Figure 9). Stable state areas are predominantly located in the northwest and some northern regions of the oasis, which are also correlated with nine significant impact factors (Figure 10). A significant negative correlation exists between population density and salinization improvement, while a significant positive correlation is observed between population density and salinization stability. Evaporation and soil texture show no significant relationship with salinization improvement, but they are significantly and positively correlated with the stable salinization state. Soil organic matter is directly and negatively correlated with the stable state; however, when mediated by population density, it exhibits a significant positive correlation with the stable state. Additionally, it is directly and positively correlated with salinization improvement. Distance from farmland is positively correlated with salinization improvement through the influence of population density, but negatively correlated with soil moisture. It is directly and negatively associated with the stable state, indirectly negatively correlated via population density, and positively correlated with the stable state through moisture conditions.
In summary, effective anthropogenic activities—such as the management of irrigation practices, organic matter levels, and vegetation coverage—are critical for mitigating soil salinization. By adopting optimized irrigation techniques and maintaining an appropriate balance of organic matter inputs, the long-term sustainability of cultivated lands can be ensured, reducing the adverse effects of salinization on agricultural productivity and soil health. In oases, each influencing factor exhibits a distinct spatial-geographic effect on soil salinization. Variations in landform, vegetation and population distribution, slope, and hydrological characteristics determine the intensity and feedback mechanisms of their influence on salinization. This study advocates that addressing soil salinization in oases requires analyzing its driving factors and mechanisms at a micro-scale. Specific mechanisms of salinization and their influencing factors should be examined within the context of individual landscapes and geomorphic units. By synthesizing the salinity characteristics and driving mechanisms across all landscape units within the oasis, an integrated model for comprehensive oasis-wide salinization management can be developed.

4.2. Regional Differences in Soil Salinization in the Keriya Oasis

The research results indicate that the deterioration of oases is concentrated in the peripheral areas, which are the transitional zones between oases and deserts [83]. The Keriya Oasis Interlaced Area is located downstream of the oasis or at the edge of the alluvial fan [84]. It is the final area for salt transport [85]. Due to the topography of the Keriya Oasis, the salt carried by surface runoff and groundwater accumulates in the interconnected areas at lower altitudes [86]. And due to the gradual flattening of the terrain and the absence of drainage channels, the salt cannot continue to migrate but can only remain and accumulate [87]. The irrigation water within the oasis also accumulates in the interlocking areas due to the terrain conditions [88]. In some areas within the oasis, a large amount of irrigation water infiltrates during the summer, resulting in a strong leaching effect, significantly reducing the salt content in the shallow soil [24]. The soil salinization remains in a stable state, and the well-developed drainage system also plays an important role [89]. The high vegetation coverage in the western part of the oasis reduces soil water evaporation and inhibits capillary rise [90]. The eastern part of the oasis is the core agricultural area [91]. The drip irrigation agriculture and the well-developed drainage facilities make it possible to improve the salinization problem [89].

4.3. Implications for Land Use Policy and Management

The spatial modeling of soil salinization presented in this study emphasizes the need for proactive and adaptive land use policies in arid regions like the Keriya Oasis. The integration of BNs with GIS offers a data-driven approach to identifying degradation-prone areas and prioritizing mitigation efforts [30,38]. The observed concentration of salinization in degraded and transitional land-use zones suggests the importance of regulating land conversion and promoting sustainable irrigation practices [54,92]. Policies should support the implementation of precision irrigation, salt-tolerant cropping systems, and managed aquifer recharge to mitigate the risk of secondary salinization [22,23,93,94]. Strategic land zoning, coupled with incentives for ecological restoration (e.g., vegetation cover, organic amendments), can enhance long-term soil resilience [95]. Moreover, participatory land-use planning involving local farmers and institutions has been shown to improve policy uptake and management outcomes [96,97]. This study’s spatial-probabilistic framework supports evidence-based policy development by enabling decision-makers to visualize salinization hotspots and evaluate future land-use scenarios under uncertainty.

5. Conclusions

The purpose of this study was to investigate the differences in soil salinization in internal regions of the Keriya Oasis. The Keriya Oasis is a small-scale, ecologically fragile arid area, and the existence of internal differences is the main reason for the difference in salinization trends. Using a BNs model combined with remote sensing, field-based EC data, and ArcGIS analysis, the spatial distribution and trends of salinization—improvement, stability, and degradation—were mapped and quantified. The model demonstrated strong predictive performance, with most variable chains showing errors below 25%, validating its robustness for complex eco-environmental analysis. Salinization is influenced by multiple interacting factors, including population density, soil texture, slope, and vegetation cover, as revealed through PLS-SEM analysis. Farmland tends to show lower salinization due to irrigation and organic matter input, whereas ecotone areas are more vulnerable due to poor drainage and limited water availability. Nonetheless, some ecotone zones exhibit improvement due to favorable topography and natural drainage. The findings emphasize that salinization processes and their drivers vary across landscape units, requiring tailored management strategies. Effective practices such as optimized irrigation and organic matter enhancement are critical to reducing salinity and sustaining agricultural productivity. This research supports the development of integrated, landscape-specific salinization management approaches and highlights the importance of high-resolution monitoring for future refinement.

Author Contributions

H.C.: Writing—original draft, Visualization, Methodology, Conceptualization. J.S.: Conceptualization, Supervision, Writing—review and editing, Visualization, Methodology. X.L.: Visualization, Methodology. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Natural Science Fund of China (Grant No. 32160319); Funding Project for Young Top-notch Talents in Xinjiang Normal University 2023 (Grant No. XJNUQB2023-10). Major Science and Technology Program of Xinjiang Uygur Autonomous Region (2024A03006-2).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

As the data was obtained through cooperation with other institutions, it is not possible to make this data public without obtaining their consent.

Acknowledgments

We would like to thank our other research partner from Xinjiang University for their support and help during field work and laboratory data processing.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Hertwich, E.G.; Pease, W.S.; Koshland, C.P. Evaluating the environmental impact of products and production processes: A comparison of six methods. Sci. Total Environ. 1997, 196, 13–29. [Google Scholar] [CrossRef]
  2. Gisladottir, G.; Stocking, M. Land degradation control and its global environmental benefits. Land Degrad. Dev. 2005, 16, 99–112. [Google Scholar] [CrossRef]
  3. Sakadevan, K.; Nguyen, M.-L. Extent, impact, and response to soil and water salinity in arid and semiarid regions. Adv. Agron. 2010, 109, 55–74. [Google Scholar] [CrossRef]
  4. Hopmans, J.W.; Qureshi, A.; Kisekka, I.; Munns, R.; Grattan, S.; Rengasamy, P.; Ben-Gal, A.; Assouline, S.; Javaux, M.; Minhas, S.P.; et al. Critical knowledge gaps and research priorities in global soil salinity. Adv. Agron. 2021, 169, 1–191. [Google Scholar] [CrossRef]
  5. El Hasini, A.; El Amrani, A.; Bouri, H. Soil Salinity under Climate Change: Challenges for Sustainable Agriculture and Food Security. Agr. Ecosyst. Environ. 2019, 287, 106786. [Google Scholar] [CrossRef]
  6. Yan, H.; Fan, Y.; Ning, Y.; Wei, J.; Yong, C. Analysis and Prospects of Saline-alkali Land in China from the Perspective of Utilization. Chin. J. Soil Sci. 2023, 54, 489–494. [Google Scholar] [CrossRef]
  7. Sivakumar, M.V. Climate extremes and impacts on agriculture. Agroclimatol. Link. Agric. Clim. 2020, 60, 621–647. [Google Scholar] [CrossRef]
  8. Hillel, D. An overview of soil and water management: The challenge of enhancing productivity and sustainability. Soil Manag. Build. A Stable Base Agric. 2011, 1–11. [Google Scholar] [CrossRef]
  9. Ondrasek, G.; Rengel, Z. Environmental salinization processes: Detection, implications & solutions. Sci. Total Environ. 2021, 754, 142432. [Google Scholar] [CrossRef] [PubMed]
  10. Yang, G.; Li, X.; Wang, S.; Liu, Y. Monitoring the Seasonal Dynamics of Soil Salinization in the Yellow River Delta of China Using Landsat Data. Nat. Hazards Earth Syst. Sci. 2019, 19, 1499–1510. [Google Scholar] [CrossRef]
  11. Pitman, M.G.; Läuchli, A. Global impact of salinity and agricultural ecosystems. In Salinity: Environmental Plants-Molecules; Lauchli, A., Luttge, U., Eds.; Springer: Dordrecht, The Netherlands, 2004; pp. 3–20. [Google Scholar] [CrossRef]
  12. Guo, H.; Ling, H.; Xu, H.; Guo, B. Study of suitable oasis scales based on water resource availability in an arid region of China: A case study of Hotan River Basin. Environ. Earth Sci. 2016, 75, 1–14. [Google Scholar] [CrossRef]
  13. Singh, A. Soil salinization and waterlogging: A threat to environment and agricultural sustainability. Ecol. Indic. 2015, 57, 128–130. [Google Scholar] [CrossRef]
  14. Singh, A. Soil salinity: A global threat to sustainable development. Soil Use Manag. 2022, 38, 39–67. [Google Scholar] [CrossRef]
  15. Wang, R.; Wan, S.; Kang, Y.; Dou, C. Assessment of secondary soil salinity prevention and economic benefit under different drip line placement and irrigation regime in northwest China. Agric. Water Manag. 2014, 131, 41–49. [Google Scholar] [CrossRef]
  16. Huang, J.; Subasinghe, R.; Malik, R.; Triantafilis, J. Salinity hazard and risk mapping of point source salinisation using proximally sensed electromagnetic instruments. Comput. Electron. Agric. 2015, 113, 213–224. [Google Scholar] [CrossRef]
  17. Besma, Z.; Christian, W.; Didier, M.; Pierre, M.J.; Mohamed, H. Soil salinization monitoring method evolution at various spatial and temporal scales in arid context: A review. Arab. J. Geosci. 2021, 14, 283. [Google Scholar] [CrossRef]
  18. Tarolli, P.; Luo, J.; Park, E.; Barcaccia, G.; Masin, R. Soil salinization in agriculture: Mitigation and adaptation strategies combining nature-based solutions and bioengineering. iScience 2024, 27, 108830. [Google Scholar] [CrossRef]
  19. Codling, E.A.; Dumbrell, A.J. Bridging the Gap between Theory and Data in Ecological Models. Ecol. Complex. 2013, 16, 1–8. [Google Scholar] [CrossRef]
  20. Nachshon, U. Cropland soil salinization and associated hydrology: Trends, processes and examples. Water 2018, 10, 1030. [Google Scholar] [CrossRef]
  21. Li, W.; Kang, S.; Du, T.; Ding, R.; Zou, M. Optimal groundwater depth and irrigation amount can mitigate secondary salinization in water-saving irrigated areas in arid regions. Agric. Water Manag. 2024, 302, 109007. [Google Scholar] [CrossRef]
  22. Cuevas, J.; Daliakopoulos, I.N.; del Moral, F.; Hueso, J.J.; Tsanis, I.K. A review of soil-improving cropping systems for soil salinization. Agronomy 2019, 9, 295. [Google Scholar] [CrossRef]
  23. Wang, D.; Zhao, C.; Zheng, J.; Zhu, J.; Gui, Z.; Yu, Z. Evolution of soil salinity and the critical ratio of drainage to irrigation (CRDI) in the Weigan Oasis in the Tarim Basin. Catena 2021, 201, 105210. [Google Scholar] [CrossRef]
  24. Lu, T.; Luo, P.; Wang, J.; Lu, Y.; Huo, A.; Liu, L. Soil salinity accumulation and groundwater degradation due to overexploitation over recent 40-year period in Yaoba Oasis, China. Soil Tillage Res. 2025, 248, 106398. [Google Scholar] [CrossRef]
  25. El-Rawy, M.; Sayed, S.Y.; AbdelRahman, M.A.; Makhloof, A.; Al-Arifi, N.; Abd-Ellah, M.K. Assessing and segmenting salt-affected soils using in-situ EC measurements, remote sensing, and a modified deep learning MU-NET convolutional neural network. Ecol. Inform. 2024, 81, 102652. [Google Scholar] [CrossRef]
  26. Zhang, Y.; Hou, K.; Qian, H.; Gao, Y.; Fang, Y.; Xiao, S.; Tang, S.; Zhang, Q.; Qu, W.; Ren, W. Characterization of soil salinization and its driving factors in a typical irrigation area of Northwest China. Sci. Total Environ. 2022, 837, 155808. [Google Scholar] [CrossRef]
  27. Shaddad, S.M.; Buttafuoco, G.; Castrignanò, A. Assessment and mapping of soil salinization risk in an Egyptian field using a probabilistic approach. Agronomy 2020, 10, 85. [Google Scholar] [CrossRef]
  28. Ashrafzadeh, A.; Roshandel, F.; Khaledian, M.; Vazifedoust, M.; Rezaei, M. Assessment of groundwater salinity risk using kriging methods: A case study in northern Iran. Agric. Water Manag. 2016, 178, 215–224. [Google Scholar] [CrossRef]
  29. Grafius, D.R.; Corstanje, R.; Warren, P.H.; Evans, K.L.; Norton, B.A.; Siriwardena, G.M.; Pescott, O.L.; Plummer, K.E.; Mears, M.; Zawadzka, J.; et al. Using GIS-linked Bayesian Belief Networks as a tool for modelling urban biodiversity. Landsc. Urban Plan. 2019, 189, 382–395. [Google Scholar] [CrossRef]
  30. Liu, Y.; Yuan, X.; Li, J.; Qian, K.; Yan, W.; Yang, X.; Ma, X. Trade-offs and synergistic relationships of ecosystem services under land use change in Xinjiang from 1990 to 2020: A Bayesian network analysis. Sci. Total Environ. 2023, 858, 160015. [Google Scholar] [CrossRef]
  31. Chen, B.; Duan, Q.; Zhao, W.; Wang, L.; Zhong, Y.; Zhuang, Y.; Chang, X.; Dong, C.; Du, W.; Luo, L. Oasis sustainability is related to water supply mode. Agric. Water Manag. 2023, 290, 108589. [Google Scholar] [CrossRef]
  32. Sun, L.; Erath, A. A Bayesian network approach for population synthesis. Transp. Res. Part C Emerg. Technol. 2015, 61, 49–62. [Google Scholar] [CrossRef]
  33. Kaikkonen, L.; Parviainen, T.; Rahikainen, M.; Uusitalo, L.; Lehikoinen, A. Bayesian networks in environmental risk assessment: A review. Integr. Environ. Assess. Manag. 2020, 17, 62–78. [Google Scholar] [CrossRef]
  34. Chang, J.; Bai, Y.; Xue, J.; Gong, L.; Zeng, F.; Sun, H.; Hu, Y.; Huang, H.; Ma, Y. Software. Dynamic Bayesian networks with application in environmental modeling and management: A review. Environ. Model. Softw. 2023, 170, 105835. [Google Scholar] [CrossRef]
  35. Barton, D.N.; Kuikka, S.; Varis, O.; Uusitalo, L.; Henriksen, H.J.; Borsuk, M.; de la Hera, A.; Farmani, R.; Johnson, S.; Linnell, J.D.; et al. Bayesian networks in environmental and resource management. Integr. Environ. Assess. Manag. 2012, 8, 418–429. [Google Scholar] [CrossRef] [PubMed]
  36. Gao, Z.; Peña-Arancibia, J.L.; Siyal, A.A. Three-dimensional soil salinity mapping with uncertainty using Bayesian Hierarchical Modelling, Random Forest Regression and remote sensing data. Agric. Water Manag. 2025, 309, 109318. [Google Scholar] [CrossRef]
  37. Rahman, M.M.; Hagare, D.; Maheshwari, B. Bayesian Belief Network analysis of soil salinity in a peri-urban agricultural field irrigated with recycled water. Agric. Water Manag. 2016, 176, 280–296. [Google Scholar] [CrossRef]
  38. Rahman, M.M.; Hagare, D.; Maheshwari, B. Framework to assess sources controlling soil salinity resulting from irrigation using recycled water: An application of Bayesian Belief Network. J. Clean. Prod. 2015, 105, 406–419. [Google Scholar] [CrossRef]
  39. Seydehmet, J.; Lv, G.H.; Nurmemet, I.; Aishan, T.; Abliz, A.; Sawut, M.; Abliz, A.; Eziz, M. Model Prediction of Secondary Soil Salinization in the Keriya Oasis, Northwest China. Sustainability. 2018, 10, 656. [Google Scholar] [CrossRef]
  40. Shi, H.Y. To reduce soil salinity: The role of irrigation and water management in global arid regions across development phases. arXiv 2022, arXiv:2204.02029. [Google Scholar]
  41. Wang, J.; Zhang, F.; Luo, G.; Guo, Y.; Zheng, J.; Wu, S.; Wang, D.; Liu, S.; Shi, Q. Factors influencing seasonal changes in inundation of the Daliyaboyi Oasis, lower Keriya River valley, central Tarim Basin, China. Remote Sens. 2022, 14, 5050. [Google Scholar] [CrossRef]
  42. Wang, J.; Zhang, F.; Luo, G.; Guo, Y.; Zheng, J.; Wu, S.; Keram, Q.; Liu, S.; Shi, Q. Influence of natural and anthropogenic controls on runoff in the Keriya River, central Tarim Basin, China. PLoS ONE 2022, 17, e0269132. [Google Scholar] [CrossRef]
  43. Xiang, Y.; Nurmemet, I.; Lv, X.; Yu, X.; Gu, A.; Aihaiti, A.; Li, S. Multi-Source Attention U-Net: A Novel Deep Learning Framework for the Land Use and Soil Salinization Classification of Keriya Oasis in China with RADARSAT-2 and Landsat-8 Data. Land 2025, 14, 649. [Google Scholar] [CrossRef]
  44. Yang, X. The oases along the Keriya River in the Taklamakan Desert, China, and their evolution since the end of the last glaciation. Environ. Geol. 2001, 41, 314–320. [Google Scholar] [CrossRef]
  45. Yang, X.; Zhu, Z.; Jaekel, D.; Owen, L.; Han, J. Late Quaternary palaeoenvironment change and landscape evolution along the Keriya River, Xinjiang, China: The relationship between high mountain glaciation and landscape evolution in foreland desert regions. Quat. Int. 2002, 97, 155–166. [Google Scholar] [CrossRef]
  46. Nurmemet, I.; Sagan, V.; Ding JianLi, D.J.; Halik, Ü.; Abliz, A.; Yakup, Z. A WFS-SVM model for soil salinity mapping in Keriya Oasis, Northwestern China using polarimetric decomposition and fully PolSAR data. Remote Sens. 2018, 10, 598. [Google Scholar] [CrossRef]
  47. Li, S.; Nurmemet, I.; Seydehmet, J.; Lv, X.; Aili, Y.; Yu, X. Spatiotemporal Dynamics and Driving Factors of Soil Salinization: A Case Study of the Yutian Oasis, Xinjiang, China. Land 2024, 13, 1941. [Google Scholar] [CrossRef]
  48. Mamat, Z.; Halik, U.; Muhtar, P.; Nurmamat, I.; Abliz, A. Temporal variation of significant soil hydrological parameters in the Yutian oasis in Northwest China from 2001 to 2010. Environ. Earth Sci. 2016, 75, 287. [Google Scholar] [CrossRef]
  49. Franklin, J.; Keeler-Wolf, T.; Thomas, K.A.; Shaari, D.A.; Stine, P.A.; Michaelsen, J.; Miller, J. Stratified sampling for field survey of environmental gradients in the Mojave Desert Ecoregion. In GIS and Remote Sensing Applications in Biogeography and Ecology; Springer: Boston, MA, USA, 2001; pp. 229–253. [Google Scholar] [CrossRef]
  50. Boschetti, L.; Stehman, S.V.; Roy, D.P. A stratified random sampling design in space and time for regional to global scale burned area product validation. Remote Sens. Environ. 2016, 186, 465–478. [Google Scholar] [CrossRef]
  51. Guo, B.; Liu, Y.; Fan, J.; Lu, M.; Zang, W.; Liu, C.; Wang, B.; Huang, X.; Lai, J.; Wu, H. The salinization process and its response to the combined processes of climate change—Human activity in the Yellow River Delta between 1984 and 2022. Catena 2023, 231, 107301. [Google Scholar] [CrossRef]
  52. Eswar, D.; Karuppusamy, R.; Chellamuthu, S. Drivers of soil salinity and their correlation with climate change. Curr. Opin. Environ. Sustain. 2021, 50, 310–318. [Google Scholar] [CrossRef]
  53. Kosmas, C.; Kairis, O.; Karavitis, C.; Acikalin, S.; Alcalá, M.; Alfama, P.; Atlhopheng, J.; Barrera, J.; Belgacem, A.; Solé-Benet, A.; et al. An exploratory analysis of land abandonment drivers in areas prone to desertification. Catena 2015, 128, 252–261. [Google Scholar] [CrossRef]
  54. He, J.; Ju, X.; Han, C.; Liu, L.; Zhang, H.; Adilai, S.; Li, C. Spatiotemporal Dynamics and Driving Mechanism of Ecological Environment Quality in Piedmont-Oasis-Desert Ecotone Based on Long-term Harmonized Remote Sensing Ecological Index-Take Korla-Tiemenguan Oasis in Xinjiang as an Example. Environ. Sustain. Indic. 2025, 26, 100611. [Google Scholar] [CrossRef]
  55. Pessoa, L.G.; Freire, M.B.d.S.; Green, C.H.; Miranda, M.F.; de A Filho, J.C.; Pessoa, W.R. Assessment of soil salinity status under different land-use conditions in the semiarid region of Northeastern Brazil. Ecol. Indic. 2022, 141, 109139. [Google Scholar] [CrossRef]
  56. Garvey, M.D.; Carnovale, S.; Yeniyurt, S. An analytical framework for supply network risk propagation: A Bayesian network approach. Eur. J. Oper. Res. 2015, 243, 618–627. [Google Scholar] [CrossRef]
  57. Xu, X.-K.; Wang, X.; Xiao, J. Inferring parent–child relationships by a node-remove centrality framework in online social networks. Phys. A Stat. Mech. Its Appl. 2018, 505, 222–232. [Google Scholar] [CrossRef]
  58. Cowell, R.G.; Verrall, R.J.; Yoon, Y.K. Modeling operational risk with Bayesian networks. J. Risk Insur. 2007, 74, 795–827. [Google Scholar] [CrossRef]
  59. Marcot, B.G. Metrics for evaluating performance and uncertainty of Bayesian network models. Ecol. Model. 2012, 230, 50–62. [Google Scholar] [CrossRef]
  60. Andrés-Hernández, L.; Baten, A.; Azman Halimi, R.; Walls, R.; King, G.J. Knowledge representation and data sharing to unlock crop variation for nutritional food security. Crop Sci. 2020, 60, 516–529. [Google Scholar] [CrossRef]
  61. Kamaruddin, S.A.; Zainolabdin, S.N.; Abd. Aziz, K.N.; Roslani, M.A.; Zohir, N.S.A.M.; Al-Bakri, N.Y.M. A comparative study of the accuracy of regularized and tension spline interpolation methods to map the surface water temperature of Pulau tuba, Langkawi, Kedah. In Proceedings of the Charting the Sustainable Future of ASEAN in Science and Technology: Proceedings from the 3rd International Conference on the Future of ASEAN (ICoFA) 2019-Volume 2; Springer: Singapore, 2020; pp. 285–295. [Google Scholar]
  62. Ajvazi, B.; Czimber, K. A comparative analysis of different DEM interpolation methods in GIS: Case study of Rahovec, Kosovo. Geod. Cartogr. 2019, 45, 43–48. [Google Scholar] [CrossRef]
  63. Perperoglou, A.; Sauerbrei, W.; Abrahamowicz, M.; Schmid, M. A review of spline function procedures in R. BMC Med. Res. Methodol. 2019, 19, 46. [Google Scholar] [CrossRef]
  64. Memon, M.A.; Ramayah, T.; Cheah, J.-H.; Ting, H.; Chuah, F.; Cham, T.H. PLS-SEM statistical programs: A review. J. Appl. Struct. Equ. Model. 2021, 5, 1–14. [Google Scholar] [CrossRef]
  65. Ma, X.; Zhang, S.; Guo, L.; Yang, J. The influence of traditional ethnic villages on forest structure based on PLS-SEM: A case study of Miao inhabited area. Forests 2023, 14, 2011. [Google Scholar] [CrossRef]
  66. Wang, C.; Ma, L.; Zhang, Y.; Chen, N.; Wang, W. Spatiotemporal dynamics of wetlands and their driving factors based on PLS-SEM: A case study in Wuhan. Sci. Total Environ. 2022, 806, 151310. [Google Scholar] [CrossRef] [PubMed]
  67. Song, J.; Aishan, T.; Ma, X. Coupled water-habitat-carbon nexus and driving mechanisms in the Tarim River Basin: A multi-scenario simulation perspective. Ecol. Indic. 2024, 167, 112649. [Google Scholar] [CrossRef]
  68. Du, H.; Li, M.; Xu, Y.; Zhou, C. An ensemble learning approach for land use/land cover classification of arid regions for climate simulation: A case study of Xinjiang, Northwest China. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 2413–2426. [Google Scholar] [CrossRef]
  69. Zhao, S.; Banerjee, S.; White, J.F.; Liu, J.-J.; Zhou, N.; Tian, C.-Y. High salt stress increases archaeal abundance and network connectivity in saline agricultural soils. Catena 2022, 217, 106520. [Google Scholar] [CrossRef]
  70. Xue, B.; Jiang, Y.; Wang, Q.; Ma, B.; Hou, Z.a.; Liang, X.; Cui, Y.; Li, F. Seasonal transpiration dynamics and water use strategy of a farmland shelterbelt in Gurbantunggut Desert oasis, northwestern China. Agric. Water Manag. 2024, 295, 108777. [Google Scholar] [CrossRef]
  71. Sahab, S.; Suhani, I.; Srivastava, V.; Chauhan, P.S.; Singh, R.P.; Prasad, V. Potential risk assessment of soil salinity to agroecosystem sustainability: Current status and management strategies. Sci. Total Environ. 2021, 764, 144164. [Google Scholar] [CrossRef]
  72. Yu, Y.; Yu, R.; Chen, X.; Yu, G.; Gan, M.; Disse, M. Agricultural water allocation strategies along the oasis of Tarim River in Northwest China. Agric. Water Manag. 2017, 187, 24–36. [Google Scholar] [CrossRef]
  73. Zhang, Z.; Xu, E.; Zhang, H. Complex network and redundancy analysis of spatial–temporal dynamic changes and driving forces behind changes in oases within the Tarim Basin in northwestern China. Catena 2021, 201, 105216. [Google Scholar] [CrossRef]
  74. Li, X.; Chang, S.X.; Salifu, K.F. Soil texture and layering effects on water and salt dynamics in the presence of a water table: A review. Environ. Rev. 2014, 22, 41–50. [Google Scholar] [CrossRef]
  75. Seydehmet, J.; Lv, G.-H.; Abliz, A. Landscape design as a tool to reduce soil salinization: The study case of Keriya oasis (NW China). Sustainability 2019, 11, 2578. [Google Scholar] [CrossRef]
  76. Madureira, H.; Pacheco, M.; Sousa, C.; Monteiro, A.; De-Sario, M. Sustainability. Evidences on adaptive mechanisms for cardiorespiratory diseases regarding extreme temperatures and air pollution: A comparative systematic review. Geogr. Sustain. 2021, 2, 182–194. [Google Scholar] [CrossRef]
  77. Li, W.; Gao, S.; Pei, D.; Wen, Y.; Mu, X.; Liu, M.; Wang, Z. Spatio-temporal evolution and simulation of soil salinization in typical oasis water-saving irrigation area based on long series data. Agric. Water Manag. 2025, 307, 109275. [Google Scholar] [CrossRef]
  78. Ding, J.; Jiang, X.; Ma, M.; Zhou, B.; Guan, D.; Zhao, B.; Zhou, J.; Cao, F.; Li, L.; Li, J. Effect of 35 years inorganic fertilizer and manure amendment on structure of bacterial and archaeal communities in black soil of northeast China. Appl. Soil Ecol. 2016, 105, 187–195. [Google Scholar] [CrossRef]
  79. Wang, X.; Ding, J.; Han, L.; Tan, J.; Ge, X.; Nan, Q. Biochar addition reduces salinity in salt-affected soils with no impact on soil pH: A meta-analysis. Geoderma 2024, 443, 116845. [Google Scholar] [CrossRef]
  80. Chen, X.; Luo, M.; Tan, J.; Zhang, C.; Liu, Y.; Huang, J.; Tan, Y.; Xiao, L.; Xu, Z. Salt-tolerant plant moderates the effect of salinity on soil organic carbon mineralization in a subtropical tidal wetland. Sci. Total Environ. 2022, 837, 155855. [Google Scholar] [CrossRef]
  81. Xu, E.; Zhang, H.; Xu, Y. Exploring land reclamation history: Soil organic carbon sequestration due to dramatic oasis agriculture expansion in arid region of Northwest China. Ecol. Indic. 2020, 108, 105746. [Google Scholar] [CrossRef]
  82. Gao, Y.; Tariq, A.; Zeng, F.; Sardans, J.; Peñuelas, J.; Zhang, Z.; Islam, W.; Xu, M. “Fertile islands” beneath three desert vegetation on soil phosphorus fractions, enzymatic activities, and microbial biomass in the desert-oasis transition zone. Catena 2022, 212, 106090. [Google Scholar] [CrossRef]
  83. Luo, G.; Feng, Y.; Zhang, B.; Cheng, W. Sustainable land-use patterns for arid lands: A case study in the northern slope areas of the Tianshan Mountains. J. Geogr. Sci. 2010, 20, 510–524. [Google Scholar] [CrossRef]
  84. Brookes, I.A. Geomorphology and Quaternary geology of the Dakhla Oasis region, Egypt. Quat. Sci. Rev. 1993, 12, 529–552. [Google Scholar] [CrossRef]
  85. De Boer, A.; Volkov, V. Logistics of water and salt transport through the plant: Structure and functioning of the xylem. Plant Cell Environ. 2003, 26, 87–101. [Google Scholar] [CrossRef]
  86. Rapiya, M.; Ramoelo, A. Groundwater–Vegetation Interactions in Rangeland Ecosystems: A Review. Water 2025, 17, 1174. [Google Scholar] [CrossRef]
  87. Chhabra, R. Nature and origin of salts, classification, area and distribution of salt-affected soils. In Salt-affected Soils and Marginal Waters: Global Perspectives and Sustainable Management; Springer International Publishing: Cham, Switzerland, 2021; pp. 1–47. [Google Scholar] [CrossRef]
  88. Unger, P.W.; Howell, T.A. Agricultural water conservation-A global perspective. J. Crop Prod. 2000, 2, 1–36. [Google Scholar] [CrossRef]
  89. Mohanavelu, A.; Naganna, S.R.; Al-Ansari, N. Irrigation induced salinity and sodicity hazards on soil and groundwater: An overview of its causes, impacts and mitigation strategies. Agriculture 2021, 11, 983. [Google Scholar] [CrossRef]
  90. Ali, A.M.; Salem, H.M. Salinity-induced desertification in oasis ecosystems: Challenges and future directions. Environ. Monit. Assess. 2024, 196, 696. [Google Scholar] [CrossRef]
  91. Liu, Y.; Xue, J.; Gui, D.; Lei, J.; Sun, H.; Lv, G.; Zhang, Z. Agricultural oasis expansion and its impact on oasis landscape patterns in the southern margin of Tarim basin, Northwest China. Sustainability 2018, 10, 1957. [Google Scholar] [CrossRef]
  92. Askri, B.; Khodmi, S.; Bouhlila, R. Impact of subsurface drainage system on waterlogged and saline soils in a Saharan palm grove. Catena 2022, 212, 106070. [Google Scholar] [CrossRef]
  93. Gebrehiwot, K.A. A review on waterlogging, salinization and drainage in Ethiopian irrigated agriculture. Sustain. Water Resour. Manag. 2018, 4, 55–62. [Google Scholar] [CrossRef]
  94. Wang, W.; Chen, Y.; Wang, W.; Zhu, C.; Chen, Y.; Liu, X.; Zhang, T. Water quality and interaction between groundwater and surface water impacted by agricultural activities in an oasis-desert region. J. Hydrol. 2023, 617, 128937. [Google Scholar] [CrossRef]
  95. Aishan, T.; Song, J.; Halik, Ü.; Betz, F.; Yusup, A. Predicting Land-Use Change Trends and Habitat Quality in the Tarim River Basin: A Perspective with Climate Change Scenarios and Multiple Scales. Land 2024, 13, 1146. [Google Scholar] [CrossRef]
  96. Thiam, H.I.; Owusu, V.; Villamor, G.B.; Schuler, J.; Hathie, I. Farmers’ intention to adapt to soil salinity expansion in Fimela, Sine-Saloum area in Senegal: A structural equation modelling approach. Land Use Policy 2024, 137, 106990. [Google Scholar] [CrossRef]
  97. Wang, J.; Zhou, Q.; Chen, R.; Wang, Z. Rationale saline-water irrigation also serves as enhancing soil aggregate stability, regulating carbon emissions, and improving water use efficiency in oasis cotton fields. Ind. Crops Prod. 2025, 223, 120144. [Google Scholar] [CrossRef]
Sustainability 17 07082 g001
Figure 1. (a) People’s Republic of China; (b) Xinjiang of China; (c) study area.
Figure 1. (a) People’s Republic of China; (b) Xinjiang of China; (c) study area.
Sustainability 17 07082 g001
Sustainability 17 07082 g002
Figure 2. Spatiotemporal and Spatial maps for modeling of Keriya Oasis salinization probability. Note: Dot is sampling points; LUCC is labeled by 0–6, 1 is grassland, 2 is shrubland, 3 is water body, 4 is bareland, 5 is building area, 6 is field. SOM, Soil texture, Drainage distance, Field distance, and Slope are spatial difference data, and they will not change in a short period of time.
Figure 2. Spatiotemporal and Spatial maps for modeling of Keriya Oasis salinization probability. Note: Dot is sampling points; LUCC is labeled by 0–6, 1 is grassland, 2 is shrubland, 3 is water body, 4 is bareland, 5 is building area, 6 is field. SOM, Soil texture, Drainage distance, Field distance, and Slope are spatial difference data, and they will not change in a short period of time.
Sustainability 17 07082 g002
Sustainability 17 07082 g003
Figure 3. Technical framework applied in this study (***, p < 0.01; **, p < 0.05; *, p < 0.1).
Figure 3. Technical framework applied in this study (***, p < 0.01; **, p < 0.05; *, p < 0.1).
Sustainability 17 07082 g003
Sustainability 17 07082 g004
Figure 4. Model structure of the BNs of the Keriya Oasis salinization probability.
Figure 4. Model structure of the BNs of the Keriya Oasis salinization probability.
Sustainability 17 07082 g004
Sustainability 17 07082 g005
Figure 5. Spatial distribution map of soil salinization probability trend in Keriya Oasis.
Figure 5. Spatial distribution map of soil salinization probability trend in Keriya Oasis.
Sustainability 17 07082 g005
Sustainability 17 07082 g006
Figure 6. Glacier area at different elevations in the PP section of the CPEC from 2000 to 2022.
Figure 6. Glacier area at different elevations in the PP section of the CPEC from 2000 to 2022.
Sustainability 17 07082 g006
Sustainability 17 07082 g007
Figure 7. Cumulative spatial impact layers map of salinization probability spatial differentiation.
Figure 7. Cumulative spatial impact layers map of salinization probability spatial differentiation.
Sustainability 17 07082 g007
Sustainability 17 07082 g008
Figure 8. Path of salinization degradation and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance, (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Figure 8. Path of salinization degradation and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance, (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Sustainability 17 07082 g008
Sustainability 17 07082 g009
Figure 9. Path of salinization improvement and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance, (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Figure 9. Path of salinization improvement and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance, (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Sustainability 17 07082 g009
Sustainability 17 07082 g010
Figure 10. Path of salinization stable state and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Figure 10. Path of salinization stable state and its drivers identified by PLS-SEM; blue and red arrows represent significant positive and negative correlations, respectively; the width of arrows indicate the relationship strength; and asterisks represent statistical significance (***, p < 0.01; **, p < 0.05; *, p < 0.1), no significant correlation is ignored.
Sustainability 17 07082 g010
Table 1. Data sources for modeling of Keriya Oasis salinization probability.
Table 1. Data sources for modeling of Keriya Oasis salinization probability.
Data TypesIndicatorTime ScaleResolutionData QuantitySource
MeteorologicalTemperature (°C)2000–20231 km24http://www.gis5g.com/data/dzsjj (accessed on 21 April 2024)
Evaporation (mm)2000–20231 km24http://www.gis5g.com/data/dzsjj (accessed on 21 April 2024)
Precipitation(mm)2000–20231 km24https://data.tpdc.ac.cn/home (accessed on 1 April 2024)
SocioeconomicDensity of population (person/km2)2000–20231 km24https://landscan.ornl.gov/ (accessed on 15 April 2024)
LUCCVegetation type trend2000–2023300 m24European Space Agency
BiophysicalSoil moisture content (g/kg)2000–2022100 m23https://data.tpdc.ac.cn/home (accessed on 1 April 2024)
Natural drainage distance (km)20201 km1https://www.openstreetmap.org/ (accessed on 20 April 2024)
Slope (°)202012.5 m1https://www.gscloud.cn/search (accessed on 23 April 2024)
Field distance (km)2020300 m1European Space Agency
NDVI (Normalized Difference Vegetation Index)2000–202330 m24http://www.gis5g.com/data/dzsjj (accessed on 21 April 2024)
Soil texture2019250 m1http://www.gis5g.com/data/dzsjj (accessed on 21 April 2024)
Soil organic matter (SOM) (mg/kg)20191 km1https://data.tpdc.ac.cn/home (accessed on 1 April 2024)
salinizationEC (mS/cm)2000, 2005, 2010;
2012, 2013, 2014,
2015, 2021, 2001,
2006, 2011, 2016,
2021		-13[47,48], This study field survey experimental data
Table 2. Descriptive statistics for soil EC by different years as measured over 0–20 cm soil surface profile.
Table 2. Descriptive statistics for soil EC by different years as measured over 0–20 cm soil surface profile.
IndicatorTimeMaximumMinimumMeanMediumSDSkewnessNumber
EC (mS/cm)20126.360.090.800.551.254.1824
2013134.30.1618.633.7135.742.4724
201431.508.173.888.901.1636
201526.408.655.248.860.6736
202135.70.176.483.239.102.0926
Table 3. Indicators category and its threshold information.
Table 3. Indicators category and its threshold information.
CategoryIndicatorIndicator States and Determination of Threshold
DriveTemperature Decrease :   P _ y e a r s < P _ ( y e a r s − 1 ) ;   increase :   P _ y e a r s > P _ ( y e a r s − 1 )
Soil organic matterLow: <0.83 g/kg; medium: 0.83–1.66 g/kg; high: >1.66 g/kg
PrecipitationLow: >42.3 mm; medium: 42.3–68.1 mm; high: >68.1 mm
Density of population decrease :   P _ y e a r s < P _ ( y e a r s − 1 ) stable   state :   P _ y e a r s ≈ P _ ( y e a r s − 1 ) ;   increase :   P _ y e a r s > P _ ( y e a r s − 1 )    
Slopesmall: <4.73°; middle: 4.73–6.20°; big: >6.20°
Natural drainage distancenearly: <5.76 km; middle: 5.76–11.29 km; far: >11.29 km
Soil textureclay: <0.2 mm; sand: 0.05–2 mm; silt: >2 mm
PressureEvaporationlow: >1255 mm; medium: 1255–1352 mm; high: >1352 mm
Soil moisture contentlow: <0.080 m3; medium: 0.080–0.13 m3; high: >0.13 m3
Vegetation type trendharmed: bare land; unchanged: cultivated land; improvement: grassland
Field distancenearly: <0.95 km; middle: 0.95–1.18 km; Far: >1.18 km
NDVIlow: <30%; medium low: 30–40%; medium: 40–60%; high: >60%
StateEC degradation :   P _ y e a r s < P _ ( y e a r s − 1 ) ;   Stable   state :   P _ y e a r s ≈ P _ ( y e a r s − 1 ) ;   improvement :   P _ y e a r s > P _ ( y e a r s − 1 )
Table 4. Results of the model validation exercise.
Table 4. Results of the model validation exercise.
No.Chain of BN NodesAccuracy (%)Number of Tests
aSoil texture—NDVI90720
bSoil organic matter—NDVI84
cSoil organic matter—LUCC87
dPrecipitation—LUCC73
ePrecipitation—Soil moisture content66
fDensity of population—LUCC83
gDensity of population—Field distance81
hField distance—Soil moisture content83
iSoil moisture content—Evaporation75
jNDVI—Evaporation80
kLUCC—Evaporation78
lTemperature—Evaporation87
mEvaporation—EC77
nSlope—EC83
oNatural drainage channels—EC76
Average80.2Total 10,800
Table 5. Comparison Results of Sampling Points’ EC between 2012 and 2021.
Table 5. Comparison Results of Sampling Points’ EC between 2012 and 2021.
Sample IDY01Y02Y03Y04Y05Y06Y07Y08Y09Y10Y11Y12
EC change from 2012 to 2021 (mS/cm)−0.020.350.1514.680.165.211.1916.651.340.0521.442.28
Sample IDY13Y14Y15Y16Y17Y18Y19Y20Y21Y22Y23Y24
EC change from 2012 to 2021 (mS/cm)−0.262.685.481.910.33−0.386.248.83−0.383.522.474.84
Sample IDY25Y26Y27Y28Y29Y30Y31Y32Y33Y34Y35Y36
EC change from 2012 to 2021 (mS/cm)5.284.89−0.325.582.542.471.546.041.020.968.444.61
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Chen, H.; Seydehmet, J.; Li, X. Upscaling Soil Salinization in Keriya Oasis Using Bayesian Belief Networks. Sustainability 2025, 17, 7082. https://doi.org/10.3390/su17157082

AMA Style

Chen H, Seydehmet J, Li X. Upscaling Soil Salinization in Keriya Oasis Using Bayesian Belief Networks. Sustainability. 2025; 17(15):7082. https://doi.org/10.3390/su17157082

Chicago/Turabian Style

Chen, Hong, Jumeniyaz Seydehmet, and Xiangyu Li. 2025. "Upscaling Soil Salinization in Keriya Oasis Using Bayesian Belief Networks" Sustainability 17, no. 15: 7082. https://doi.org/10.3390/su17157082

APA Style

Chen, H., Seydehmet, J., & Li, X. (2025). Upscaling Soil Salinization in Keriya Oasis Using Bayesian Belief Networks. Sustainability, 17(15), 7082. https://doi.org/10.3390/su17157082

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop