Exploration of the utilization of a new land degradation index in Lake Ebinur Basin in China

Abstract

Land degradation significantly impacts regional economic development and food security, particularly in arid river basins where soil and water conservation is crucial. Understanding the extent and causes of land degradation is pivotal for effectively prevention and management. This study employs the soil adjusted vegetation index (SAVI), the temperature vegetation dryness index (TVDI), and the salinization detection index (SDI), combined with the analytic hierarchy process and the entropy weight method, to construct a comprehensive land degradation index (LDI). Sen’s slope trend analysis and the Mann-Kendall significance test were used to analyze land degradation trends in the Ebinur Lake watershed from 2002 to 2022. Additionally, the optimal parameters-based geographical detector was used to examine the underlying mechanisms of land degradation. The results indicate the following: (1) From 2002 to 2012, the degree of land degradation in the Ebinur Lake watershed worsened, particularly in the eastern and southeastern parts, as well as in the southern region of Toli County. From 2012 to 2022, land degradation significantly improved, with a notable reduction in degraded land area. (2) Over the period of 2002-2022, $93.08\%$ of the land in the research region exhibited a declining LDI trend, $3.95\%$ showed no change, and only $2.96\%$ showed an increasing LDI trend. (3) Moderate, severe, and very severe degradation mainly occurred on grassland and unused land, while light degradation and non-degradation primarily occurred on forest land and cultivated land. (4) Unreasonable land use and overgrazing were identified as the primary factors influencing land degradation, with elevation being a secondary factor. The interaction between land use and other factors was found to be most significant, followed by the synergistic effects of grazing quantity with elevation, annual average temperature, gross domestic product, soil moisture, and elevation with annual average precipitation, and temperature. The results of this study offer an empirical basis and taking decisions assistance for land degradation control in the Ebinur Lake Basin, as well as examples and references for assessing land degradation in other places.

Introduction

The sustainable utilization of land resources is crucial for achieving environmental protection, economic development, and social progress^1,2. With the advancement of urbanization, land degradation has become a significant environmental problem threatening global sustainable development, affecting approximately one-third of the global land area^3,4. Land degradation involves natural processes such as soil erosion, salinization, and desertification, as well as the decline of ecosystem services due to human activities. This poses serious challenges to food security, biodiversity conservation, and ecological balance, particularly impacting the socio-economic development of developing countries^5,6. Since the United Nations introduced the concept of “Land Degradation Neutrality (LDN)” in 2015, countries have incorporated LDN into their sustainable development goals^7,8,9,10. By 2019, China had successfully achieved LDN ahead of schedule and plans to continue strengthening land restoration and ecological conservation efforts in the Ebinur Lake Basin according to the “14th Five-Year Plan” goals set by the Bortala Mongolian Autonomous Prefecture Government. Timely assessment and monitoring of land degradation in the Lake Ebinur region are crucial for effective restoration efforts.

Traditional land degradation monitoring studies have primarily focused on single processes such as desertification, vegetation degradation, and soil erosion. While these studies have advanced the understanding of the structural mechanisms of land degradation, they have limitations when applied to broader research on land degradation^11,12. In recent years, scholars have widely applied remote sensing technology to retrieve land degradation-related elements and construct comprehensive evaluation systems. By selecting representative key variables in the study area, researchers can more accurately explain the spatiotemporal changes in land degradation^13,14. Lu et al.¹⁵ employed a fuzzy logic modeling approach to create a comprehensive land degradation index and evaluate the land degradation status of the Loess Plateau. Similarly, Shao et al.¹⁶ explored the land degradation situation in arid and semi-arid areas by establishing an optimal land degradation index model. Additionally, some scholars have introduced the Analytic Hierarchy Process (AHP) to study the complex concept of land degradation. For example, Yadav et al.¹⁷ and Yang et al.¹⁸ applied the AHP to study the spatiotemporal variation characteristics of regional land degradation and proposed governance policies suitable for regional development. The innovation of this study lies in the integration of both subjective and objective methods to assess land degradation. Through the process of constructing land degradation indicators, it was found that relying on a single method has limitations. The hierarchical analysis method, for instance, carries certain subjectivity. Principal component analysis (PCA) and the entropy weight method (EWM), as systematic methods for dealing with uncertainty in complex problems, offer strong objectivity. However, compared to PCA, the EWM has better flexibility and adaptability, allowing it to more effectively highlight the importance of indicators in the overall evaluation. This flexibility has led to its widespread use by many scholars in various studies^19,20. In the process of selecting evaluation indicators, three evaluation indicators were selected. Among them, soil conditioned vegetation index (SAVI) effectively weakens the impact of soil surroundings on the vegetation index by introducing soil adjustment factors. This makes it particularly suitable for areas with low vegetation coverage, providing more accurate vegetation information²¹. Temperature vegetation dryness index (TVDI) has significant application value in drought monitoring and assessment, offering scientific backing for drought forecast and management²². Soil degradation index (SDI) exhibits higher sensitivity and dynamic monitoring capabilities than other salt indexes, reducing the impact of vegetation²³. The purpose of land degradation governance is to effectively minimize or control the process of land degradation by addressing its causes. Understanding the underlying causes of regional land degradation is critical for effective governance. Various methods are currently used in the study of influencing factors and driving mechanisms, such as correlation analysis²⁴, hotspot analysis²⁵, and geographically weighted regression²⁶. Although these methods have their own advantages, they have certain limitations compared to the Optimal Parameters-based Geographical Detector (OPGD) when dealing with complex relationships, spatial heterogeneity and comprehensive analysis. OPGD offers more advantages in handling multi-factor interactions and robustness. As a tool for exploring the relationship between geographical phenomena and environmental factors, OPGD has advantages over traditional methods in detecting and explaining nonlinear relationships, identifying spatial data heterogeneity, and revealing differences from multiple perspectives across various regions or conditions. This capability is particularly important in analyzing complex geographical phenomena^27,28.

The Ebinur Lake Basin is a region characterized by the integration of mountains, waters, forests, farmlands, lakes, grasslands, and sands, forming a diverse ecological community. This region faces severe land degradation issues, such as wetland shrinkage, salinization, and desertification^29,30. The hazard posed by the exposed lakebed salt dust, resulting from lake water shrinkage, is significant to downstream areas³¹. In addition, recent climate warming has led to increased the surface temperatures and widespread desertification in the basin³². The degradation of ecological and environmental quality due to these phenomena has received widespread attention from scholars^33,34. However, research on the spatiotemporal changes inland degradation in this basin remains limited. As national governance efforts progress, there is an urgent need to develop specific implementation plans on the changes and driving mechanisms of land degradation in the Ebinur Lake Basin. This study adopts the method of combining AHP with EWM, selects three indicators suitable for arid area research: SAVI, TVDI and SDI, and constructs a land degradation evaluation system in the Ebinur Lake Basin to evaluate the degree of land degradation more comprehensively and accurately. By analyzing the spatiotemporal characteristics and trends of land degradation in the basin, and using OPGD to explore the driving mechanism of land degradation, scientific suggestions are provided for the improvement of the basin’s ecological environment, enriching the existing research content of land degradation, and having important theoretical and practical significance for the rational use of land resources and the promotion of equitable development.

Materials and methods

Study area

The Ebinur Lake Basin $\left( 43^{\circ } 37^{\prime }-45^{\circ } 53^{\prime } {\textrm{N}}, 79^{\circ } 52^{\prime }-85^{\circ } 13^{\prime } {\textrm{E}}\right)$ is located in the northwest of the Xinjiang Uygur Autonomous Region, China, in the western section of the Tianshan Mountains, bordering Kazakhstan to the west (Fig. 1). The basin spans from east to west, encompassing Kuitun, Karamay’s Dushanzi District, Usu City, the southern part of Toli County, Jinghe County, Alashankou City, Shuanghe City, Bole City, and Wenquan County, covering an area of approximately $5 \times 10^{4}\, \mathrm {~km}^{2}$. The basin experiences long sunshine hours, scarce precipitation, and high evaporation rates, with an average annual evaporation of 1500 to $2000 \mathrm {~mm}$, indicating a typical temperate continental climate. Since the 1970s, the largest river flowing into Ebinur Lake, the Kuitun River, has completely dried up. The main tributaries remaining are the Bortala River and Jinghe River. The land use in the study area is divided into six categories: cropland, forest land, grassland, water, unused land, and building land. In 2022, grassland and unused land were the dominant types, covering $26,554.88 \mathrm {~km}^{2}$ and $14,472.17 \mathrm {~km}^{2}$ respectively, while water body and construction land were relatively small, with area of $983.35\, \mathrm {~km}^{2}$ and $344.35\, \mathrm {~km}^{2}$, respectively. With the rapid growth of the economic belt on the northern slope of the Tianshan Mountains, as well as the construction of the New Euro-Asia Continental Bridge, the land degradation pattern in the Ebinur Lake basin has been constantly shifting, impacting the basin’s ecosystem stability.

Figure 1

Ebinur Lake Basin (the map vector data used in this article are all from the China Standard Map Service: http://211.159.153.75/, and the map approval number is GS (2023) 2767).

Data sources

This study selected Landsat 7 and Landsat 8 data with less than $10\%$ cloud cover from June 1st to September 30th to calculate the SAVI, TVDI, and SDI. The Corine Land Cover Database (CLCD) land use data was used, and the study area was characterized into six land use types: cropland, forest land, grassland, building land, unused land, and water, through reclassification and manual correction. Combined with field surveys, sampling points are selected to verify the accuracy of CLCD land use data. The overall accuracy of the five periods ranged between 83 and $86\%$. The raster calculator was used to process precipitation, temperature, and soil moisture data on an annual basis. All data were projected to WGS_ 1984_UTM_Zone_44N and resampled at a spatial resolution of 30 m. Table 1 contains a list of the data sources.

Table 1 Data Source.

Research methods

This study employed the AHP and EWM to construct a new land degradation index, which was used to analyze the spatiotemporal changes in land degradation in the Ebinur Lake Basin. The trend analysis technique was utilized to explore the development trends of land degradation. Furthermore, the characteristics of land degradation in the Ebinur Lake Basin were investigated based on land use data. By exploring the influencing factors of land degradation, we aim to provide a scientific reference for regional land degradation management. The study structure is shown in Fig. 2.

Figure 2

Study structure diagram.

LDI construction based on AHP and entropy weight method

Land degradation results from both natural and human-induced factors. Considering the existing condition of land resources in the study area and prior research outcomes, three evaluation indices were selected: SAVI²¹, TVDI²² and SDI²³. These indices indicate land degradation from three aspects: vegetation, temperature, and salinization. They complement each other and address the limitations of a single indicator. The formulas for calculating these indices are as follows:

$$\begin{aligned} {\textit{SAVI}}&=\frac{( {\textit{NIR}}- {\textit{R}})(\textit{1}+ {\textit{L}})}{( {\textit{NIR}}+ {\textit{R}}+ {\textit{L}})} \end{aligned}$$

(1)

$$\begin{aligned} {\textit{TVDI}}&=\frac{ {\textit{TS}}- {\textit{T}} {\textit{S}}_{\min }}{ {\textit{TS}}_{\max }- {\textit{TS}}_{\min }} \end{aligned}$$

(2)

$$\begin{aligned} {\textit{SDI}}&=\sqrt{( {\textit{NDVI}}-\textit{1})^{2}+ {\textit{SI}}^{2}} \end{aligned}$$

(3)

In the formula: NIR is the reflection value of the near-infrared band; R is the reflection value of the red light band; L is the soil adjustment coefficient. According to the overall situation of this study area, the value of is 0.5; TS is the surface temperature; ${ {\textit{TS}}_{\max }}$ and ${ {\textit{TS}}_{\min }}$ are the highest and lowest values of the surface temperature corresponding to the value, respectively; SI is the salt index of the area.

The AHP is a subjective weighting method, which is highly subjective. The EWM calculates weights based on actual data, offering an objective basis for the calculation of indicator weights. Although there can be deviation between these calculated weights and the actual importance of the indicators, combining both methods can better reflect the objectivity and subjectivity of indicator weights. Therefore, here we combine the both methods to calculate the comprehensive weight. The following is the precise calculation formula:

$$\begin{aligned} W&=\alpha \omega _i+(1-\alpha )\mu _i, (0\le \alpha \le 1) \end{aligned}$$

(4)

$$\begin{aligned} minW&=\sum _{i=1}^n[(W_i-\omega _i)^2+(W_i-\mu _i)^2] \end{aligned}$$

(5)

In the formula: W is the comprehensive weight determined by AHP and EWM; $\omega _{i}$ is the weight calculated by analytic hierarchy process; and $\mu _{i}$ is the weight calculated by EWM. Without loss of generality, the two equations were combined and the value of a was obtained at 0.5. The final weights for SAVI, TVDI, and SDI are determined to be 0.547, 0.240 and 0.213 respectively. Consequently, the following formula can be used to determine the Land Degradation Index:

$$\begin{aligned} L D I=0.547 \times S A V I^{\prime }+0.240 \times T V D I^{\prime }+0.213 \times S D I^{\prime } \end{aligned}$$

(6)

The variables SAVI, TVDI, and SDI were standardized. A higher numerical value indicates a more severe degree of land degradation for that pixel, while a lower value indicates a lighter degree of land degradation. Following existing research findings^35,36,37, the LDI was categorized into five levels: Level I, Level II, Level III, Level IV, and Level V. These levels correspond to Non-degraded $(L D I<0.2)$, Slightly Degraded $(0.2 \le L D I<0.4)$, Moderately Degraded $(0.4 \le L D I<0.55)$, Heavily Degraded $(0.55 \le L D I<0.6)$, and Extremely Degraded ( $L D I \ge 0.6$ ), respectively.

Land degradation trend analysis based on Theil-Sen and Mann-Kendall methods

To analyze the trend of Land Degradation Index (LDI) changes in the Ebinur Lake Basin from 2002 to 2022, Theil-Sen slope estimation and the Mann-Kendall test were employed^38,39,40. The primary formulas for these analyses are:

$$\begin{aligned} \beta&={\text {mean}}\left( \frac{x_{j}-x_{i}}{j-i}\right) , \forall j>i \end{aligned}$$

(7)

$$\begin{aligned} Z&=\left\{ \begin{array}{cl} \frac{S-1}{\sqrt{{\text {Var}}(S)}}, &{} (S>0) \\ 0, &{} (S=0) \\ \frac{S+1}{\sqrt{{\text {Var}}(S)}}, &{} (S<0) \end{array}\right. \end{aligned}$$

(8)

In these formulas, $\beta$ represents the Theil-Sen slope estimation, indicating the trend of LDI changes, while $\textit{Z}$ represents the standardized test statistic from Sen and MK methods, and $\text {\textit{S}}$ represents the test statistic. This study categorizes the trend characteristics into nine types: extremely significant decrease, significant decrease, slightly significant decrease, not significant decrease, no change, not significant increase, slightly significant increase, significant increase, and extremely significant increase, represented by numbers from -4 to 4. A test result of $|Z| \ge 2.580, \beta >0$ indicates an extremely significant increase in the trend; while $\beta <0$ indicates an extremely significant decrease in the trend. Whereas, a test result of $1.960 \le$ $|Z|<2.580, \beta >0$ is considered to be a significant increase; $\beta <0$ indicates significant decrease in the trend. For $1.645 \le |Z|<1.960, \beta >0$, the trend is considered to be slightly significantly increased; while $\beta <0$ is considered to be slightly significantly decreased. A test result of $|Z|<1.645, \beta >0$ is considered to be not significantly increased; $\beta <0$ indicates no significant decreased.

Analysis of land degradation driving mechanisms based on OPGD

Geographical detectors are widely used to reveal spatial heterogeneity and spatial correlation in geographic phenomena across various fields^41,42. To address issues arising from the discretization of dependent variables and variations in spatial statistical units, Song et al.⁴³ proposed OPGD. The main advantages of OPGD over traditional geographic detectors include automated parameter optimization, stronger spatial heterogeneity detection capabilities, strong adaptability, and data robustness. These advantages make OPGD more effective in processing complex spatial data, identifying spatial distribution patterns, and determining their driving factors, providing more accurate and reliable spatial analysis results. OPGD includes factor detectors, interaction detectors, risk area detectors, and ecological detectors. In this study, the number of breakpoints in the OPGD model was set between 6 to 10. Through parameter optimization, the relationship between land degradation and various driving factors was explored, revealing deeper insights into the driving mechanisms behind land degradation in the Ebinur Lake Basin.

The factor detector is used to explore the spatial differentiation explanatory power of independent variables $\textrm{X} 1, \textrm{X} 2, \textrm{X} 3 \ldots \textrm{Xn}$ on the dependent variable $\textrm{Y}$, and the explanatory power is represented by q. Here is the specific calculation equation:

$$\begin{aligned} q=1-\frac{\sum _{h=1}^{L} N_{h} \sigma _{h}^{2}}{N \sigma ^{2}} \end{aligned}$$

(9)

In the formula, h represents the layer number of independent variable X; $N_{h}$ and N represent the layer number and the sample size within the study area, respectively; $\sigma _{h}^{2}$ and $\sigma ^{2}$ represent the layer number and the variance of dependent variable $\textrm{Y}$ within the study area, respectively; q ranges from 0 to 1 , where a q value closer to 1 indicates stronger explanatory power.

The interaction detector was used to explore the explanatory power of the combined effects of two independent variables $X_{a}$ and $X_{b}$ on the dependent variable Y. The explanatory power q was calculated to determine the interaction types of $X_{a}$ and $X_{b}$, with the criteria outlined in Table 2.

Table 2 Interaction type judgment table.

Land degradation future pattern prediction based on grey model

Grey model is analysis predictive and analytical method used when establishing precise mathematical models is challenging, particularly in situations with insufficient or incomplete data^44,45. In this study, the LDI data from 2002 to 2022 were processed using cumulative, average, and other treatments, converting the data into a form suitable for grey differential equations. Based on Deng’s grey system theory⁴⁶, a grey model was constructed to predict the LDI from 2023 to 2027. The main calculation formula of the grey model is shown in Equation 8. The model’s accuracy was verified employing the mean relative error method, resulting in a calculated mean relative error of $8.45\%$. This high level of accuracy indicates that the model is credible and suitable for predicting future LDI trends.

$$\begin{aligned} \hat{x}^{(1)}(t+1)=\left( x^{(0)}(1)-\frac{b}{a}\right) e^{-a t}+\frac{b}{a} \end{aligned}$$

(10)

In the formula: $\textrm{a}$ is the progression index; $\textrm{b}$ is the gray action amount.

Results

Spatiotemporal characteristics of land degradation

Vegetation coverage is represented by SAVI, where higher values denote greater vegetation and lower values denser vegetation. Figure 3 shows the spatiotemporal distribution of SAVI in the Ebinur Lake Basin from 2002 to 2022. Generally, high SAVI values are concentrated in flat agricultural areas and the region surrounding Lake Sayram, while low values are mainly distributed around Ebinur Lake and in the eastern desert and Gobi areas, including some areas on the periphery of oasis farmland. Over time, SAVI exhibited a trend of “first decrease, then increase”. From 2002 to 2012, there was a significant decrease in vegetation in the eastern part of Ebinur Lake, where vegetation was sparse and the ecosystem condition was poor. However, from 2012 to 2022, there was a noticeable improvement in these sparsely vegetated areas, with SAVI generally increasing across the basin, indicating an increase in vegetation richness and improvement in ecosystem conditions.

Figure 3

Spatiotemporal changes of SAVI in the Ebinur Lake Basin from 2002 to 2022.

The Soil Degradation Index (SDI) indicates the severity of land salinization, with higher values reflecting more severe salinization. Figure 4 illustrates the spatiotemporal variation pattern of SDI from 2002 to 2022. During the study period, regions with higher salinization levels were closely distributed near areas of lower SAVI values, mostly found in eastern and southeast regions of Ebinur Lake, exhibiting a significant large-scale continuous distribution. The areas surrounding Ebinur Lake, where salt fields are located, consistently exhibited high SDI values throughout the study period. Over time, there was a general trend of reduction in regions with higher salinization levels. From 2002 to 2012, the average SDI in the basin increased from 0.74 to 0.76. However, Between 2012 and 2022, the average SDI decreased from 0.76 to 0.71 , indicating an improvement in salinization.

Figure 4

Spatiotemporal changes of SDI in the Ebinur Lake Basin from 2002 to 2022.

The TVDI represents the degree of aridity in the region. Figure 5 illustrates the spatiotemporal variation characteristics of the TVDI during the study period. High TVDI values are concentrated in the desert and Gobi areas in the eastern part of Ebinur Lake, appearing as patchy aggregated distributions, while low-value areas are mainly distributed at the foot of high-altitude mountains. There is a significant difference in TVDI between the eastern and western parts of Ebinur Lake, with TVDI mostly maintained at a moderate level in the western region, while it is notably higher in the eastern region. Over the last two decades, TVDI has been steadily rising. The average TVDI value was 0.62 in 2002, but by 2022, the average value increased to 0.74, indicating a more severe loss of soil moisture..

Figure 5

Spatiotemporal changes in TVDI in the Ebinur Lake Basin from 2002 to 2022.

The three evaluation factors are incorporated into the LDI calculation formula, and the results are shown in Fig. 6. Between 2002 and 2022, the land degradation pattern in the Ebinur Lake Basin shows an initial upward trend followed by a subsequent decline. Severe degradation and extremely severe degradation mainly occur around Ebinur Lake and in the southern Gobi areas of Toli County, while moderate degradation is mainly found in the peripheral areas of oases with lower terrain. Mild degradation mainly occurs in the southwest of Toli County, while non-degraded areas are mainly located in oases. From 2002 to 2012, the average LDI value increased from 0.41 to 0.44 . By 2022, the average LDI value decreased to 0.38. All things considered, the LDI shows a clear pattern of spatial distribution, with greater levels in the east and lower levels in the west.

Figure 6

Spatiotemporal changes in LDI in the Ebinur Lake Basin from 2002 to 2022 ($\text {I}$ represents no degeneration; $\text {II}$ represents mild degeneration; $\text {III}$ represents moderate degeneration; $\text {IV}$ represents severe degeneration; $\text {V}$ represents extreme degeneration).

The combination of Sen’s slope and the MK trend test can visually illustrates the LDI trend in the Ebinur Lake Basin over the past 20 years, as shown in Fig. 7. From 2002 to 2022, the LDI trend in the Ebinur Lake Basin is mainly characterized by an extremely significant decrease and a significant decrease, accounting for $34.46\%$ and $30.35\%$ of the total area, respectively. This is followed by a slightly significant decrease and not significant decrease, accounting for $9.51\%$ and $18.76\%$, respectively. The unchanged area accounts for $3.95\%$ of the total area, while the area with no significant increase is $2.53\%$. The areas with a slightly significant increase, significant increase, and extremely significant increase are relatively small, accounting for $0. 13\%$, $0. 19\%$, and $0. 11\%$, respectively. Overall, it shows that $93.08\%$ of the research area’s LDI is decreasing, whereas areas with an increasing tendency account for only $0.43\%$. In terms of spatial distribution, areas with extremely significant decreases and significant decreases are mainly distributed in Jinghe County in the south of Ebinur Lake and the eastern part of Ebinur Lake. The northeastern and western parts of Ebinur Lake mainly show a slightly significant decrease and not a significant decrease. The boundaries of the research area, as well as regions around agricultural and construction land, exhibit a trend of increasing degradation.

Figure 7

The trend of LDI from 2002 to 2022 ($-4$ represents a very significant decrease; $-3$ represents a significant decrease; $-2$ represents a slightly significant decrease; $-1$ represents a not significant decrease; 0 represents no change; 1 represents a not significant increase; 2 represents a slightly significant increase; 3 represents a significant increase; and 4 represents a very significant increase).

The transfer of land degradation areas at different levels, based on the LDI classification, is shown in Fig. 8. From 2002 to 2022, the regions of moderate, severe, and extremely severe degradation initially increased before decreasing. From 2002 to 2012, areas of slight and moderate degradation decreased, while severe and extremely severe degradation areas increased by $5144.24 \mathrm {~km}^{2}$ and $3031.68 \mathrm {~km}^{2}$ respectively. From 2012 to 2022, non-degraded and slightly degraded areas significantly increased by $2181.89 \mathrm {~km}^{2}$ and $5770.58 \mathrm {~km}^{2}$ respectively. Compared to 2012, in 2022, In contrast, severe and extremely severe degradation areas decreased by $8235.44 \mathrm {~km}^{2}$ and $3211.88 \mathrm {~km}^{2}$ respectively. From 2002 to 2012, some non-degraded and slightly degraded land transitioned to moderate degradation, while moderately degraded land shifted to severe and extremely severe degradation. However, from 2012 to 2022, this trend reversed, with severely and extremely severely degraded land moving to moderate degradation, and moderately degraded land shifting to slight degradation and non-degradation.

Figure 8

Transfer situation of land degradation areas of different degrees from 2002 to 2022 ($\text {I}$ represents no degeneration; $\text {II}$ represents mild degeneration; $\text {III}$ represents moderate degeneration; $\text {IV}$ represents severe degeneration; $\text {V}$ represents extreme degeneration).

Land degradation analysis based on land use

Unreasonable or inappropriate land use practices can lead to land degradation. To gain a greater insight into the link between land use and land degradation, we calculated the areas of cropland, forest land, grassland, and unused land in the Ebinur Lake Basin from 2002 to 2022 (Fig. 9). During this period, the areas of cropland and forest land expanded annually. Until 2012, cropland expansion gradually decreased and stabilized, while government initiatives such as returning cropland to forest led to a continued increase in forest land. Over 20 years, cropland expanded by a total of $2618.19 \mathrm {~km}^{2}$, and forest land expanded by $327.71 \mathrm {~km}^{2}$ . From 2002 to 2022, grassland area decreased by $3370.78 \mathrm {~km}^{2}$, and unused land increased by $710.06 \mathrm {~km}^{2}$ due to the shrinking of Ebinur Lake’s surface area and drying surface runoff.

Figure 9

The land use change in the Ebinur Lake Basin from 2002 to 2022.

To visually demonstrate the differences in land degradation levels among different land use types, we extracted degradation levels for cropland, forest land, grassland, and unused land and calculated the areas of degradation at different levels for each land use type. The radar charts displaying these degradation areas under each land use type are displayed in Figure 10. During the study period, cropland and forest land primarily remained at non-degraded and slightly degraded levels. On average, non-degraded and slightly degraded cropland accounted for $72\%$ and $21\%$ of the total cropland area each year, respectively. Similarly, Non-degraded and slightly degraded forest land accounted for $54\%$ and $42\%$ of the total forest land area each year, respectively. Grassland mainly experienced slight and moderate degradation, with $30\%$ slightly degraded and $51\%$ moderately degraded annually. The degradation levels of unused land mainly remained at moderately degraded and severely degraded levels. Unused land primarily showed moderate and severe degradation, with $29\%$ moderately degraded and $51\%$ severely degraded annually. Given that grassland and undeveloped land make up the majority of all types of land use in the study area, their degradation levels directly impact the overall land degradation situation in the Ebinur Lake Basin.

Figure 10

Change map of degraded areas by land use type in Ebinur Lake Basin from 2002 to 2022($\text {I}$ represents no degeneration; $\text {II}$ represents mild degeneration; $\text {III}$ represents moderate degeneration; $\text {IV}$ represents severe degeneration; $\text {V}$ represents extreme degeneration).

Driving mechanisms of land degradation

Land degradation is affected by many factors. According to the United Nations Convention to Combat Desertification and the characteristics of the study area, we selected eight driving factors from both natural and human aspects (Table 3). Using LDI as the dependent variable representing land degradation in the Lake Ebinur Basin and eight driving factors as independent variables, we employed a geographic detector with optimal parameters to explore the explanatory power of single factors and the interaction of multiple factors.

Table 3 Drivers of Land Degradation.

Analyzing the explanatory power of the eight drivers of land degradation reveals their significance. Based on the results of single-factor analysis (Fig. 11), it is evident that land use and grazing intensity have a relatively high impact on land degradation during the study period, with mean explanatory power (q) values of 0.632 and 0.333 , respectively. Elevation also significantly impacts land degradation, with explanatory power exceeding 0.1 for all five periods, and a mean q value of 0.179. The other factors have relatively low explanatory power. From 2002 to 2022, there is a decreasing trend in the influence of elevation, land use, grazing intensity, and GDP on land degradation, with q values decreasing by 0.031, 0.05, 0.103, and 0.176, respectively, indicating a decline in their impact on land degradation.

Figure 11

Results of single factor detection of land degradation in Ebinur Lake Basin from 2002 to 2022.

In order to further clarify the synergistic effects of different factors on land degradation, we tested the interaction between the LDI and eight factors across five periods in the Lake Ebinur basin (Fig. 12). The results show that the interaction effects include both dual-factor enhancement and non-linear enhancement. Throughout all periods, the interaction between land use and other factors demonstrates dual-factor enhancement, with q values surpassing 0.59. Dual-factor enhancement is also observed between grazing intensity and elevation, annual average temperature, GDP, soil moisture, land use, annual precipitation, and population in some years. Additionally, elevation exhibits dual-factor enhancement with annual precipitation and annual temperature in all years. In terms of temporal changes, the strongest factor combinations for interaction in 2002, 2007, and 2012 are consistently land use and grazing intensity, with q values of 0.693, 0.697, and 0.686, respectively. In contrast, in 2017 and 2022, the strongest factor combinations for interaction are land use with annual temperature and land use with annual precipitation, with q values of 0.686 and 0.638, respectively.

Figure 12

Results of interactive detection of land degradation factors in Ebinur Lake Basin from 2002 to 2022.

Discussion

Assessment of land degradation indicators in dryland area watersheds

Different from other studies on land degradation indices^15,16, this study selected three evaluation indicators that are closely aligned with the actual conditions of the Ebinur Lake Basin. From 2002-2022, SAVI, which is less affected by soil brightness, proved suitable for sparsely vegetated arid areas. SDI is widely used to explore land degradation changes in saline-alkali areas. TVDI is an important indicator for measuring regional soil moisture and drought conditions. These three indicators address vegetation, salinization, and drought aspects, respectively. Using the AHP and EWM, we integrated SAVI, SDI, and TVDI indices to establish a new LDI. This LDI was then used to analyze land degradation characteristics in this typical arid region. The analysis results revealed that from 2002 to 2022, land degradation in the Ebinur Lake Basin initially worsened but then improved. After 2012, degraded areas around Sayram Lake, eastern Wensu County, and Ebinur Lake significantly decreased due to policies implemented by government departments, such as establishing the Ebinur Lake Wetland Nature Reserve, Sayram Lake and Wensu County Wetland Parks, and Ganjiahu Suaeda Forest Nature Reserve. Ecological restoration work has also been actively carried out in the Ebinur Lake Basin. However, severe and extremely severe degradation is concentrated in the eastern and southeastern parts of Ebinur Lake and the periphery of the oasis. The large-scale drying up of Ebinur Lake in recent decades has exposed high-salinity land, leading to the formation of extensive saline-alkali soil. Strong winds from the Alashankou Pass exacerbate this process by spreading salt particles and causing frequent sandstorms, furthering land desertification. Thus, scientific management of salinization and wind erosion control are critical issues for future management.

In analyzing trends and land degradation areas across different land use types, nearly $90\%$ of the land in the study area showed a decreasing LDI trend over the past 20 years. The most significant and severe declines occurred primarily in grassland and arable land areas, suggesting that vegetation restoration, returning farmland to forests, and soil and water conservation efforts play crucial roles in improving land degradation and soil erosion. However, $50\%$ of the land in the study area is unused, and changes in unused land directly affect land degradation. Future management measures should include protective measures for unused land, such as planting protective forests, especially in the eastern part of Ebinur Lake, to prevent wind erosion and secondary salinization of unused land. Comparing our findings with related studies on constructing LDI, we observed that land degradation is more severe in the downstream and surrounding areas of water bodies, while cultivated land and lush woodland and grassland are less likely to degrade. This is consistent with previous studies, indicating that LDI can better reflect the land degradation status of the study area. The research results based on LDI also support and supplement previous studies^47,48. The comprehensive system constructed based on both subjective and objective methods makes LDI more credible and scientific compared to previous studies. Future research should aim to integrate additional processes such as soil pollution and soil acidification and combine multi-source data to construct a more comprehensive evaluation system, addressing one of the key issues in assessing land degradation.

Integrated analysis of main impact factors of land degradation in the Ebinur Lake Basin

The dominant factors affecting land degradation in the Ebinur Lake Basin evolve with changes in the natural environment and human activities. However, in previous studies, scholars have often focused on analyzing land degradation factors for specific periods, neglecting the changing patterns of these dominant factors over time. Our single-factor detection results indicate that human activities are the leading factors influencing land degradation. Overgrazing and irrational land use further exacerbate the severity of land degradation. However, the explanatory power (q value) of land use increased annually, but since 2012, it has shown a decreasing trend. Similarly, the influence of grazing intensity has continuously declined throughout the study period. This indicates that while human activities remain the main factor influencing land degradation, their overall impact is gradually decreasing. In contrast, factors such as average annual temperature and soil moisture, although showing minimal changes in q values, have exhibited slight fluctuations and increases. The changes in both natural and human factors suggest that restoring desert vegetation through return grazing land to grassland and restocking in the Bortala Autonomous Prefecture have significantly improved the overall land degradation pattern.

The interactions among different factors exert a stronger influence than the individual impacts of natural and human factors, which aligns with the findings of researchers on land degradation⁴⁹. Although the interactions between land use, grazing intensity, and other factors have been relatively high over the past 20 years, they have shown a decreasing trend annually. However, the interactions between natural factors have shown an increasing trend. Since 2012, efforts have been made to establish nature reserves, and protective forests, and initiate the construction of the Bortala River Ecological and Cultural Corridor in the basin to improve land degradation conditions, reducing the impact of human activities on land degradation. Additionally, reports on climate research in Xinjiang⁵⁰ indicate that high summer temperatures and low precipitation have contributed to land degradation. The results of multi-factor detection support the findings of single-factor detection, highlighting the complex interplay between various factors in driving land degradation in the Ebinur Lake Basin.

Future trends in land degradation in the Ebinur Lake Basin and monitoring of priority areas

The factors influencing land degradation are complex, and their systemic relationships exhibit a certain degree of fuzziness. The process of change is gradual, and distant degradation levels cannot be directly converted^51,52. Grey models are particularly suitable for predicting data with fuzzy structures and uncertainties. Using continuous LDI data from the past 21 years, grey models were used to directly predict LDI changes. The model was trained to improve prediction accuracy, and the predicted results were visualized using Kriging interpolation. The figure shows the spatiotemporal change predictions of LDI from 2023 to 2027 (Fig. 13). According to the predictions, there will be significant improvement inland degradation around Lake Sayram, with mild degradation gradually expanding towards the center of the basin, while areas of moderate degradation are gradually shrinking. Most of the land along the Bolotola River shows mild degradation, indicating favorable vegetation growth and soil conditions. This also suggests that the restoration and management efforts in the region have been effective.

Figure 13

LDI prediction map of Lake Ebinur Basin from 2023 to 2027($\text {I}$ represents no degeneration; $\text {II}$ represents mild degeneration; $\text {III}$ represents moderate degeneration; $\text {IV}$ represents severe degeneration; $\text {V}$ represents extreme degeneration).

With the implementation of ecological restoration efforts, the Sayram Lake Basin has achieved zero growth in land degradation and has shown significant improvements, as confirmed by our predictive results. However, there has been persistently severe land degradation in the Sayram Lake Basin’s southeast and south. Field investigations and communication with forestry and grassland departments revealed that there several salt fields, such as the Jinghe Salt Field and the Wusu Cement Plant Salt Field, scattered throughout the southeastern part of Sayram Lake. Efforts are currently underway to dismantle these salt fields, and the next step involves rehabilitating and restoring the land they occupied. Additionally, the southern part of Tori County and the Ganzia Lake Desert region are also significant areas of severe degradation. In the next phase of restoration, suitable vegetation, such as poplar and tamarisk, will be planted in these areas to help prevent wind erosion and stabilize the sand.

Conclusions

Using the Ebinur Lake Basin as a case study, this research analyzed the shifts in land degradation severity and the underlying mechanisms from 2002 to 2022. The primary findings are summarized as follows:

• 1. From 2002 to 2012, land degradation in the Ebinur Lake Basin intensified, primarily concentrated in the eastern and southeastern areas of Ebinur Lake and the southern part of Toli County. However, from 2012 to 2022, there was a noticeable improvement in land degradation, with a decrease in the extent of degraded land.

• 2. Throughout the study duration, $93.08\%$ of the surveyed area witnessed a decline in LDI, $3.95\%$ experienced no change, and merely $2.96\%$ exhibited an increase.

• 3. In the study area, moderate, severe, and extremely severe degradation mainly occurred in grassland and unused land. In contrast, while mild degradation and non-degradation mainly occurred in forestland and cultivated land.

• 4. Improper land use and overgrazing are the primary factors influencing land degradation, with elevation differences being a secondary influencing factor. The most significant interaction was between land use and other factors, followed by the interaction between grazing intensity and elevation, average annual temperature, GDP, and soil moisture, as well as between elevation and average annual precipitation and temperature.