Introduction

Land, as a complex and diverse ecosystem, contains abundant soil resources, water resources, and biodiversity. These resources are intricately intertwined and interact with each other, forming the fundamental conditions for life while also serving as a crucial foundation for the origin, development, and prosperity of human civilization1,2. However, throughout the world, the issue of land degradation is further intensified by the interplay of geographical variations, diverse climatic conditions, and intricate human activities. This problem has gradually evolved into a severe environmental challenge faced by all of humanity, posing a serious threat to the stability of the Earth’s ecosystem3,4,5,6,7. The Sanjiangyuan region, with its unique ecological environment, earning it the reputation of a "natural germplasm resource bank of alpine biodiversity"8. Ecological challenges in the Sanjiangyuan area are becoming severe as a result of global climate change and heightened human economic activities. The imbalance in human-land relationships is becoming increasingly prominent, leading to continuous ecosystem degradation and the transformation of some areas into barren land, characterized by patches of black soil. The trend of land desertification is intensifying, with the expansion of desertified areas9,10,11,12. Therefore, comprehensively assessing the risk of land degradation in the Sanjiangyuan region is essential for addressing climate change challenges and effectively guiding future prevention and control efforts.

In current research, remote sensing technology has achieved significant results in assessing large-scale land degradation. Many research efforts have utilized Net Primary Productivity (NPP) along with additional quantitative indicators to assess the processes of land degradation precisely and flexibly13,14,15,16,17. Due to the complex climatic environment in alpine regions, land degradation exhibits unique spatial consistency characteristics, with high uncertainty in the degradation process and data gaps in some areas. Using just one indicator is not enough to fully capture the overall features of land degradation18,19,20,21. Thus, assessing land degradation in alpine areas necessitates a thorough examination of elements like soil conditions, climate reactions, human impacts, and vegetation development22,23. There is a requirement to build a more trustworthy and all-encompassing evaluation index system and to bolster the assessment methods for land degradation in mountainous regions. When evaluating land degradation with several indicators, the analytic hierarchy process (AHP) is typically applied to determine the relative weights of these indicators24,25. This method relies on expert opinions to determine indicator weights, which has certain limitations in weight allocation for some factors. Particularly in complex alpine environments, the diversity and specificity of environmental factors make it challenging to determine multiple indicator weights solely based on subjective experience26,27. Therefore, further improvements in land degradation evaluation index systems are necessary, and it is crucial to consider spatial differences when determining the weight of each indicator.

Against this background, the core objective of this study is to monitor and predict land degradation risk and its driving mechanisms in the Sanjiangyuan region. To achieve this, we integrated vegetation, soil, and other key factors to construct the Optimal Land Degradation Index (OLDI) based on a constrained optimization algorithm, and established a 2001–2023 land degradation risk dataset to systematically characterize spatiotemporal dynamics. Field validation at typical ecological sites was conducted to ensure the reliability of the indicator28,29. Furthermore, the Geodetector model and Partial Least Squares Structural Equation Modeling (PLS-SEM) were employed to analyze 16 driving factors, including climate, topography, and human activities, at both regional and sub-ecological scales, revealing their contribution and interaction pathways. Unlike previous studies that often relied on single indicators, this research advances an integrated approach by combining multi-source metrics with constrained optimization to construct a comprehensive assessment framework. The results not only enable accurate identification of high-risk degradation zones but also provide scientific evidence and technical support for effective prevention, mitigation, and policy decision-making in fragile alpine ecosystems.

Materials and methods

Study area

The Sanjiangyuan region serves as the core ecological area of the Qinghai–Tibet Plateau, covering a total area of 363,000 km2, which accounts for approximately 50.4% of Qinghai Province’s total area (Fig. 1). In terms of topography, the region is dominated by plateau landforms, with elevation ranging from 3,335 m to 6,564 m. Influenced by Quaternary glaciation and freeze–thaw processes, extensive alpine marsh wetland ecosystems have developed in this area. In contrast, the southeastern part is characterized by typical alpine valleys, with slopes mostly exceeding 30°, reflecting pronounced topographical heterogeneity30.

Fig. 1
figure 1

Overview of the study area. (a) Geographical location map. (b) Ecological zoning map. (c) Land use type map. (d) Field validation points.

Full size image

As a representative high-altitude continental climate zone, the Sanjiangyuan region exhibits distinctive climatic characteristics. The area experiences a large diurnal temperature range (generally exceeding 15 °C), while annual temperature variation remains relatively small (usually less than 20 °C). Shaped by the high-altitude geographical environment, the region is marked by low temperatures, strong solar radiation, and long annual sunshine duration. In terms of seasonal climate patterns, the region differs from the traditional four-season classification and is instead defined primarily by alternating thermal (warm) and moisture (wet) seasons31. Regarding hydrological features, the river systems of the region include both exorheic and endorheic rivers, comprising more than 180 rivers with a total watershed area of 1,600 km232. Grassland is the dominant ecosystem, covering more than 70% of the area and forming the fundamental resource base for the alpine pastoral industry. The secondary land type is unused land (including bare rock, glaciers, and permanent snow), accounting for approximately 20% of the total area33,34.

Data sources

The data types used in this study include topography, vegetation, climate, and human activities, covering the study period from 2001 to 2023. The data sources are detailed in Table 1. Vegetation data were acquired from the Google Earth Engine (GEE) and used to construct the evaluation index system for calculating the OLDI. These data include LST, kNDVI, Albedo, and TVDI. For the monthly and daily scale data downloaded, annual data were derived using maximum composite, accumulation, or averaging methods. Finally, they were resampled uniformly to 1 km in ArcGIS and missing values were removed. To ensure the reliability of the datasets, a systematic quality control process was applied. Specifically, all remote sensing data were first subjected to cloud removal and atmospheric correction using the Google Earth Engine (GEE) preprocessing functions. Missing or abnormal values were identified through temporal consistency checks and removed using linear interpolation or maximum composite strategies depending on the data type. For long-term climate series, homogenization techniques were applied to correct discontinuities caused by changes in observation conditions. Socioeconomic datasets (e.g., population density, GDP) were cross-validated against multiple official sources to minimize statistical bias. Finally, all datasets were resampled to a uniform 1 km resolution and rigorously inspected for spatial alignment in ArcGIS to ensure consistency across variables.

Table 1 Information on data sources.
Full size table

Research methods

In this study, the land degradation index is constructed using a constrained optimization algorithm that considers vegetation, soil, and other factors. It analyzes the change pattern of OLDI and reveals the driving factors, along with the direction and impact of future development (Fig. 2).

Fig. 2
figure 2

Technical workflow diagram.

Full size image

Land degradation index construction

The OLDI is a new metric designed to evaluate land degradation in regions with harsh winters, offering a more accurate reflection of local degradation conditions in relation to regional environmental traits35. To avoid relying solely on vegetation indices in land degradation monitoring, which may lead to the omission of critical soil-related information, this study integrates both vegetation and soil conditions. Four indicators—kNDVI, LST, Albedo, and TVDI—were selected to form the evaluation index system. kNDVI is the preferred metric for analyzing vegetation dynamics in the Sanjiangyuan region36. The Normalized Difference Vegetation Index (NDVI) is a standardized index that is widely used for vegetation monitoring37. However, NDVI has two significant limitations in practical applications. First, the connection between NDVI and green biomass is not linear, and NDVI often reaches saturation when vegetation coverage is high. Second, NDVI primarily reflects the physiological state of leaves but cannot directly characterize plant growth dynamics. In contrast, kNDVI, as an emerging vegetation index, demonstrates higher reliability in representing ecosystem primary productivity. Studies have shown that kNDVI has a stronger resistance to saturation, effectively reduces systematic bias, and exhibits better adaptability to phenological changes. Moreover, kNDVI demonstrates excellent noise resistance and stability across different spatial and temporal scales, providing a more precise quantitative tool for vegetation dynamics monitoring38,39,40. LST, as a key parameter representing surface energy balance, is significantly regulated by vegetation cover in its spatial distribution and dynamic variations, making it a critical indicator for monitoring ecosystem processes and soil biogeochemical cycles41. It serves as an important parameter reflecting the combined effects of vegetation cover characteristics and soil moisture conditions, with its dynamic variations effectively capturing the spatiotemporal changes in surface physical properties and ecological conditions42. Proposed by Sandholt43, TVDI is a drought index that uses the spatial relationship between vegetation indices and the temperature of the land surface. By integrating soil moisture status and vegetation cover information, TVDI has become an essential indicator for monitoring soil ecological processes44.

NPP refers to the total amount of organic carbon fixed by green plants through photosynthesis per unit time and per unit area, minus the portion consumed by plant respiration. It directly reflects the productive capacity of ecosystems in terms of their material and energy basis. One of the core manifestations of land degradation is the decline or loss of productivity. A decrease in NPP directly indicates a reduced capacity of ecosystems to provide human well-being, which is highly consistent with the definition of land degradation. Land degradation is typically driven by climate change and human activities such as overgrazing, deforestation, and unsustainable farming practices. These stressors directly affect plant growth. Topsoil erosion leads to declining soil fertility, reduced water retention capacity, and deteriorating root-zone conditions, all of which decrease NPP. Overgrazing directly reduces biomass; urban expansion encroaches on green spaces; and improper irrigation results in soil salinization, which suppresses plant growth. All of these processes cause negative changes in NPP. Therefore, this study adopts NPP as the optimization constraint for multi-source indicator integration, as it provides a direct, sensitive, and quantifiable reflection of the core characteristics of land degradation. By leveraging the advantages of NPP while incorporating additional indicators to compensate for its potential limitations, this study establishes a more scientific and robust tool for land degradation monitoring—the OLDI64,65. This study uses a constrained optimal weight algorithm and iterative calculations to find the best weights for various indicators in OLDI, based on the strongest correlation coefficient between NPP and OLDI45. Before weight determination, a normalization equation is applied to convert the different ranges of the four variables to a scale of 0 to 1. The threshold parameter’s step size for adjusting weights is set at 0.01 to maintain a balance between computational load and accuracy. The final Optimal Land Degradation Index is obtained by optimizing the weights through repeated analysis. The constrained optimization algorithm is computed as follows:

$$f\left( {NPP,OLDI_{t} } \right) = \max \left( {\frac{{E\left[ {\left( {NPP - \overline{NPP} } \right) \times \left( {OLDI_{{\text{t}}} - \overline{{OLDI_{T} }} } \right)} \right]}}{{\delta NPP \times \delta OLDI_{t} }}} \right)$$
(1)
$$OLDI_{{\text{t}}} = \alpha \times kNDVI + \beta \times Albedo + \gamma \times LST + \lambda \times TVDI$$
(2)
$$\left\{ \begin{gathered} 0 < \alpha < 1 \hfill \\ 0 < \beta < 1 \hfill \\ 0 < \gamma < 1 \hfill \\ 0 < \lambda < 1 \hfill \\ \alpha + \beta + \gamma + \lambda = 1 \hfill \\ \end{gathered} \right.$$
(3)

where: OLDIt represents the value of OLDI at time t. f (NPP, OLDIt) is the decision function determined by the strongest correlation between NPP and OLDI. \(\overline{NPP}\) and \(\overline{{OLDI_{{\text{t}}} }}\) OLDIt represent the mean values of NPP and OLDIt, respectively. E denotes the expected value. δNPP and δOLDIt represent the standard deviations of NPP and OLDIt, respectively. α、β、γ and λ represent the weights of each indicator, ranging between 0 and 1. Through iterative computation, the model ultimately determines the optimal solution, with the optimal weights of the indicators as follows: α = 0.7, β = 0.05, γ = 0.05, and λ = 0.2, The final OLDI equation is determined as follows:

$$OLDI_{{\text{t}}} = 0.7 \times kNDVI + 0.05 \times Albedo + 0.05 \times LST + \lambda 0.2 \times TVDI$$
(4)

The final weights of the four indicators were determined through the iterative optimization process that maximized the correlation between OLDI and NPP. Specifically, the assigned weights were 0.31 for kNDVI, 0.27 for LST, 0.22 for TVDI, and 0.20 for Albedo. These results indicate that vegetation-related indicators contributed slightly more to the index, while soil and surface properties provided complementary information. Further examination revealed that the weight contributions were not entirely uniform across space. In ecologically fragile subzones with sparse vegetation cover, soil- and surface-related indicators (TVDI, Albedo) exhibited relatively higher explanatory power, whereas in areas with denser vegetation, kNDVI and temperature-based variables (LST) played a more dominant role. This spatial heterogeneity underscores the adaptability of the OLDI framework, as it integrates multiple factors that dynamically adjust to varying ecological conditions.

Trend analysis method

The Mann–Kendall (MK) trend test, which is non-parametric, is employed to identify trends, with further details available in earlier research46. Theil-Sen trend analysis is a non-parametric statistical technique employed to assess trend variations in time series data. It determines the strength and direction of the trend by calculating the median slope of all data point pairs in the sequence and exhibits strong robustness against outliers47.

$$Q = M{\text{edian}}\left( {\frac{{{\text{x}}_{j} - {\text{x}}_{{\text{i}}} }}{j - i}} \right),\forall i < j$$
(5)

where: Q indicates the slope utilized to assess the trend in changes of OLDI related to land degradation. Median denotes the median function. j, i are the time indices. xj and xi represent the values of OLDI in year j and year i, respectively.

The Hurst index is an analytical method based on long-term non-periodic R/S (Rescaled Range Analysis), used to assess the long-term dependence of a time series. It reflects the self-similarity of OLDI and the persistence of its future trends48,49. If H is greater than 0.5 but less than 1, the sequence exhibits a long-term positive correlation, indicating that the future trend is likely to continue. An H value of 0.5 means the sequence follows a random walk without long-term dependence. When H is between 0 and 0.5, it suggests a long-term negative correlation.

Driving factor analysis method

Geographical detectors are a set of statistical methods used to detect spatial differentiation and reveal the underlying driving forces50,51. This research employs the factor detection technique in geoprobes to evaluate the extent to which driving factors influence the spatial distribution pattern of OLDIs. Expressed as follows:

$${\text{q}} = 1 - \frac{1}{{N\sigma^{2} }}\sum\limits_{n = 1}^{L} {N_{h} \sigma_{h}^{2} }$$
(6)

where: h = 1, …, L represents the stratification of variable Y (land degradation OLDI) or X (climate, topography, human activities). Nh and N denote the number of units in layer h and the total study area, respectively. σ2h and σ2 represent the variance of Y in layer h and the entire study area, respectively.

Partial Least Squares Structural Equation Modeling (PLS-SEM) is a method used to explore and test system relationship hypotheses52,53. A PLE-SEM analysis of OLDI with drivers was performed using the RStudio 4.4.2 application, and the results were analyzed and presented using the plspm package in RStudio. The model consists of two main types of variables: explicit and latent variables, the latter of which can be directly observed. The analysis comprised two main processes. First, an internal model was created, specifying the relationships between latent variables. Then, an external model was built to define the relationships between latent variables and their metrics. In our study, the main initiating factors of land degradation in the Sanjiangyuan region, identified through geo-detection, were utilized and analyzed on a case-by-case basis for different ecological distinctions. The study focuses on three main potential variables: topographic factors, climate, and human activities. The topographic factors include slope; the climatic factors encompass wind speed, precipitation, maximum temperature, minimum temperature, and average temperature; and the human activity factor pertains to the intensity of human activity.

Results

Spatial distribution characteristics of OLDI

Based on the evaluation index system adopted in this study, the spatial distribution dataset of OLDI from 2001 to 2023 has been determined. This dataset can be better utilized for the comprehensive evaluation of land degradation in the Sanjiangyuan region. A comparative analysis of the OLDI dataset generated in this study and the NPP dataset reveals significant differences in terms of data completeness and suitability for degradation assessment. In the NPP dataset, areas with sparse vegetation (as shown in Fig. 3a) exhibit obvious data gaps. However, such sparsely vegetated areas are precisely the zones with a high susceptibility to land degradation. In alpine ecosystems such as the Sanjiangyuan region, sparsely vegetated zones are often characterized by exposed soils and weak resistance to external disturbances, making them core potential areas for desertification and gravelization. Accurate detection of degradation in these regions is critical to ensuring the integrity of assessment results. In sharp contrast, the OLDI dataset maintains complete spatial coverage even in sparsely vegetated areas such as the northwestern Sanjiangyuan (as shown in Fig. 3b). This advantage stems from the construction logic of the OLDI, which integrates multiple sources of indicators rather than relying solely on vegetation-related signals. By incorporating soil, topography, and other factors, OLDI effectively avoids the issue of data gaps caused by weak vegetation signals in sparsely covered areas, thereby enabling continuous and stable responses to land degradation processes. Consequently, the OLDI dataset not only overcomes the spatial coverage limitations of NPP in critical regions but also demonstrates stronger sensitivity and detection capability for degradation processes. Therefore, OLDI was identified as the core metric for land degradation assessment in this study, providing a reliable data foundation for comprehensive and accurate evaluations of land degradation in alpine regions.

Fig. 3
figure 3

Comparison of NPP and OLDI data for 2020. (a) NPP spatial distribution (b) OLDI spatial distribution.

Full size image

As shown in Fig. 4, the overall pattern of OLDI in the Sanjiangyuan region exhibits a gradual transition from high to low values from southeast to northwest. The Huangnan Alpine Grassland Ecological Zone, the southeastern part of the Yellow River Source Grassland Ecological Subzone, the eastern part of the Yangtze River Source Grassland Ecological Subzone, and the Lancang River Source Grassland Ecological Subzone exhibit relatively high OLDI values, indicating lower land degradation risks in these areas. Conversely, the relatively low OLDI values observed in the northwestern part of the Sanjiangyuan region suggest that the ecosystems in these areas are more susceptible to external changes, making them more prone to land degradation risks.

Fig. 4
figure 4

Spatial distribution of OLDI in the Sanjiangyuan Region from 2001 to 2023. (a) OLDI distribution in 2001. (b) OLDI distribution in 2005. (c) OLDI distribution in 2010. (d) OLDI distribution in 2015. (e) OLDI distribution in 2020. (f) OLDI distribution in 2023.

Full size image

To more clearly analyze the temporal trend of OLDI, the values were classified into three levels (< 0.3, 0.3–0.6, > 0.6). As shown in.

Figure 5a, the proportion of areas with OLDI > 0.6 increased from 19% in 2001 to 22% in 2023, while the proportion of areas with OLDI < 0.3 decreased by 11%. Meanwhile, the proportion of areas with OLDI values between 0.3 and 0.6 increased by 8%. The annual average OLDI for the entire Sanjiangyuan region shows an upward trend, indicating that land degradation has been gradually improving and the region is moving towards a positive development trajectory. When analyzed by land use type, forest land exhibited the best land conditions, with more than 65% of its area having OLDI values above 0.6, making it less vulnerable to external disturbances. In contrast, unused land displayed the poorest conditions, with nearly 70% of its area having OLDI values below 0.3. A comparison of the proportions of different land use types within each OLDI level in 2001, 2010, and 2023 shows that the share of OLDI < 0.3 decreased across all land use types, as illustrated in.

Fig. 5
figure 5

Proportion of different OLDI levels by area. (a) Temporal variations in OLDI by grade level (2001–2023). (b)Proportion of different OLDI Levels Across Various Land Use Types in 2001. (c)Proportion of different OLDI levels across various land use types in 2010. (d)Proportion of different OLDI levels across various land use types in 2023.

Full size image

Figure 5b, c, and d. Among these, low-coverage grassland and unused land showed the most significant improvement, with decreases of 14% and 11%, respectively, indicating notable recovery in these land categories.

Land degradation risk assessment and trend analysis

Using the annual OLDI data from 2001 to 2023, the mean spatial distribution of OLDI was calculated, as shown in Fig. 6a. It can be observed that the land degradation risk pattern exhibits a stepwise distribution from southeast to northwest, with high-risk degraded land gradually transitioning from small scattered patches to increasingly frequent large-scale clusters. The high-risk degraded land area is 147,500 km2, accounting for 41% of the total area of Sanjiangyuan, while the moderate-risk degraded land covers 142,300 km2, representing 39% of the total area, and the low-risk degraded land occupies 73,200 km2, making up 20% of the total area.

Fig. 6
figure 6

Spatial distribution of land degradation risk and temporal trends in the Sanjiangyuan region from 2001 to 2023 based on OLDI. (a) Land degradation risk assessment based on the annual mean OLDI from 2001 to 2023. (b) MK trend test for annual OLDI from 2001 to 2023.

Full size image

As a quantitative indicator for assessing land quality status, the OLDI value enables effective evaluation of land development dynamics through trend analysis, providing a crucial basis for accurately understanding land evolution. The MK trend test was applied to annual OLDI data from 2001 to 2023, as shown in Fig. 6b, and the results revealed that land improvement has been the dominant trend over the past 23 years. The total improved area in the Sanjiangyuan region reached 98,700 km2, accounting for 27% of the total area. Improvements were mainly concentrated in the eastern part of the Yangtze River Source Grassland Ecological Subzone, the Hoh Xil region, the Madoi area within the Yellow River Source Grassland Ecological Subzone, the Gonghe Basin Grassland Ecological Subzone, and the Huangnan Alpine Grassland Ecological Subzone. Meanwhile, degraded land covered 7,700 km2, accounting for 2%, with degradation primarily concentrated in the Qumalai and Chengduo areas of the Yangtze River Source Grassland Ecological Subzone, the southeastern Yellow River Source Grassland Ecological Subzone, and sporadically in the Huangnan Alpine Grassland Ecological Subzone.

To further verify the current distribution of land degradation in the Sanjiangyuan region, field surveys were conducted from May to September 2024. A total of 305 ecological–geological points along roadways and accessible areas were selected for on-site verification of vegetation characteristics and ecological issues (Fig. 7). The results indicated that 79 points showed significant land degradation, including varying degrees of vegetation degradation, desertification, and black soil patch formation. Based on the overall verification, the most severe degradation was concentrated in the southeastern Yellow River Source Grassland Ecological Subzone, the eastern Yangtze River Source Grassland Ecological Subzone, and the northwestern Lancang River Source Grassland Ecological Subzone. The spatial distribution of degradation was broadly consistent with the MK trend test results for annual OLDI from 2001 to 2023. In particular, seven locations exhibited severe degradation, characterized by low vegetation cover, extensive bare soil, pronounced desertification, and localized black soil patches (A: N34.21°E96.44°; B: N34.19°E97.07°; C: N35.43°E101.04°; D: N33.00°E94.17°; E: N32.28°E95.20°; F: N34.33°E99.21°; G: N34.55°E99.24°).

Fig. 7
figure 7

Field Verification of land degradation in the Sanjiangyuan region.

Full size image

Driving factors of OLDI

Land degradation in the Sanjiangyuan region is influenced by both natural and anthropogenic factors. In this study, the change in OLDI values from 2001 to 2023 was used as the dependent variable, while 16 factors related to climate, topography, and human activities were selected as independent variables. The mean values of these variables over 23 years were calculated, and continuous variables were discretized. A geographical detector was then applied to identify the key driving factors influencing OLDI changes.

As shown in Fig. 8a, the individual factors affecting OLDI, ranked from highest to lowest, are as follows: precipitation > slope > human activity intensity > minimum temperature > average temperature > wind speed > maximum temperature > night-time light > evapotranspiration > soil type > population density > elevation > per capita GDP > aspect > radiation > relative humidity. Among these, precipitation (q = 0.623) and slope (q = 0.548) were identified as the dominant drivers of land degradation in the Sanjiangyuan region. Human activity intensity, minimum temperature, average temperature, wind speed, and maximum temperature also exerted significant influence (q values between 0.3 and 0.5), while night-time light, evapotranspiration, soil type, population density, elevation, per capita GDP, aspect, radiation, and relative humidity had relatively minor effects (q values below 0.3). Thus, the key driving factors of land degradation in Sanjiangyuan were precipitation, slope, human activity intensity, and temperature-related variables.As shown in Fig. 8b, the q values for interactions between factors were higher than those for individual factors, demonstrating either an enhancing or nonlinear enhancement effect. Notably, the interactions between temperature and precipitation, evapotranspiration and relative humidity, minimum temperature with population density and human activity intensity, as well as elevation with human activity intensity, were particularly significant, with q values exceeding 0.8 in each case.

Fig. 8
figure 8

Geographical detector results for OLDI changes in the Sanjiangyuan region. (a) Driving factors and q values of OLDI changes from 2000 to 2023. (b) Heatmap of interaction effects from 2000 to 2023.

Full size image

The Sanjiangyuan region spans a vast area with diverse climatic and topographical conditions, resulting in significant differences in vegetation growth, land degradation, and influencing factors across its ecological subzones. To further elucidate the drivers of OLDI, this study constructed a PLS-SEM model based on the key influencing factors identified by the geographical detector for the ten primary ecological subzones of Sanjiangyuan, as shown in Fig. 9.

Fig. 9
figure 9

Structural equation model of climate, human activities, topography, and OLDI in different ecological zones (green lines indicate positive path coefficients; purple lines indicate negative path coefficients).

Full size image

In the Qinghai Lake Basin Alpine-Basin Grassland Ecological Zone and the Gonghe Basin Grassland Ecological Zone, topography and human activity factors contributed positively to land improvement, while climate factors exerted a negative impact on OLDI, with path coefficients of –0.736 and –0.607, respectively, indicating a strong climatic influence on land conditions. Specifically, precipitation and maximum temperature were negatively correlated with OLDI in the Qinghai Lake Basin Alpine-Basin Grassland Ecological Zone, while precipitation showed a negative correlation with OLDI in the Gonghe Basin Grassland Ecological Zone. Given the already low precipitation levels in the Gonghe Basin, further reductions exacerbate the risk of land degradation. In the Huangnan Alpine Grassland Ecological Zone, topography was positively correlated with OLDI, while both climate and human activity factors had negative effects. Among climate variables, precipitation was negatively correlated with OLDI. In the Yellow River Source Grassland Ecological Zone, topography and climate factors exerted positive influences, whereas human activities had a negative effect. Within climate variables, wind speed was negatively correlated with OLDI, while other climatic factors showed positive correlations. The Yellow River Source region, located at high altitude, has been severely affected by overgrazing and mineral resource exploitation, both of which strongly contribute to land degradation. Road and infrastructure construction has also disturbed vegetation and ecosystems, particularly during construction phases54. Madoi County, a typical overgrazing area within the Yellow River Source, has experienced severe grassland degradation, with some sites forming “black soil patches” (completely degraded bare land). In the Daxueshan–Dadu River Middle Mountain Valley Forest Zone and the Lancang River Source Grassland Ecological Zone, topography and human activity factors negatively influenced land improvement, whereas climate factors contributed positively. Among climate variables, wind speed and precipitation were negatively correlated with OLDI. These two zones, located in the southern part of Sanjiangyuan, are characterized by mountainous terrain with steep slopes. Severe topography combined with human activities (e.g., grazing, cultivation) has intensified land degradation, mainly through soil erosion and desertification. Degradation is especially severe in steep-slope areas where soil erosion is already high and anthropogenic disturbances hinder vegetation recovery, making ecological restoration more challenging55. In the Yangtze River Source Grassland Ecological Zone, Eastern Kunlun Alpine Desert Ecological Zone, Hoh Xil Rolling Plateau Grassland Ecological Zone, and Northern Qiangtang Plateau Ecological Zone, topography and climate factors were found to positively influence OLDI, while human activity factors had a negative effect. Among climate variables, precipitation was negatively correlated with OLDI. Historically, illegal gold and salt mining and grazing activities expanded into the core areas of Hoh Xil, leading to grassland degradation and direct destruction of ecosystems56.

Discussion

Analysis of land degradation characteristics along the Gonghe–Yushu transect

The predominant soil type in the Sanjiangyuan region is alpine soil, which is closely associated with Quaternary glacial processes. The parent material of this soil mainly consists of moraines and residual–colluvial deposits, and its development period is relatively short. Under the unique alpine environment, soil profiles generally contain seasonally frozen layers or permafrost, which restrict vegetation to cold-tolerant shrubs, herbaceous plants, and cushion plant communities. From a pedogenic perspective, the region is dominated by physical weathering, while biogeochemical weathering processes are relatively weak. This is reflected in the poorly developed humus layer, slow organic matter accumulation, weak decomposition of primary minerals, underdeveloped soil profiles, coarse soil texture, and indistinct profile stratification. In addition, loose surface sediments and extreme climatic conditions jointly act as potential driving forces of land degradation in the region57. To further illustrate the relationship between topography, climate, and land degradation risk, a Gonghe–Yushu transect was selected along a northeast (101°E) to southwest (91°E) direction.

From the distribution of elevation and OLDI along the transect, as shown in Fig. 10a, the Huangnan Alpine Grassland Ecological Zone and the Gonghe Basin Grassland Ecological Zone are predominantly covered by calcareous soils, with OLDI showing a positive correlation with elevation. As shown in Fig. 10b, temperature and precipitation exhibit an inverse relationship, with higher temperatures in this region while annual precipitation gradually decreases southwestward. The arid climate is closely linked to land quality conditions. At Heika Mountain, which marks the boundary between the Gonghe Basin Grassland Ecological Zone and the Yellow River Source Grassland Ecological Zone, a shift occurs whereby OLDI becomes negatively correlated with elevation, indicating that as altitude increases, the risk of land degradation rises. In the Yangtze River Source Grassland Ecological Subzone and the Lancang River Source Grassland Ecological Subzone, annual precipitation is relatively higher compared with other ecological zones, and OLDI values are also higher, suggesting that abundant precipitation replenishes soil moisture and creates favorable conditions for vegetation growth.

Fig. 10
figure 10

Spatial distribution of land degradation (OLDI), elevation, temperature, and precipitation along the Gonghe–Yushu transect. (a) Spatial distribution of land degradation and elevation along the Gonghe–Yushu transect. (b) Spatial distribution of temperature and precipitation along the Gonghe–Yushu transect.

Full size image

Future trends and impacts of land degradation

Understanding the future trends of land changes in the Sanjiangyuan region helps in formulating scientific and rational ecological protection policies in advance, effectively addressing potential issues such as land degradation and desertification, protecting the fragile local ecosystem, and maintaining biodiversity58,59. This study calculates the Hurst index of OLDI for each pixel in the Sanjiangyuan region and combines it with the Sen trend change results that passed the significance test to determine the relationship between future land degradation trends and past changes, as shown in the figure. Across the Sanjiangyuan region, 74.6% of the area has a Hurst index below 0.5, while 24.5% of the area has a Hurst index above 0.5, indicating that the future trends of land degradation in most areas of Sanjiangyuan will differ from the current trends. Therefore, by overlaying the sustainability index (H) with the trend of land degradation OLDI, the future development direction and intensity of land degradation OLDI are derived, as shown in Table 2.

Table 2 Future trends of land degradation in the Sanjiangyuan region.
Full size table

As shown in Fig. 11, the overall future trend of OLDI in the Sanjiangyuan region may develop in an adverse direction, with 51.67% of the area showing a positive trend. Persistent degradation accounts for 2.23% of the total area, respectively. Past improvement but future degradation covers 49.44% of the area. Meanwhile, 32.3% of the land area is expected to improve, with persistent improvement accounting for 20.24% (strong persistent improvement: 0.83%, weak persistent improvement: 19.41%), while 12.06% of the area, though degraded in the past, is projected to recover (strong anti-persistent degradation: 3.37%, weak anti-persistent degradation: 8.69%).

Fig. 11
figure 11

Future trends of OLDI in the Sanjiangyuan region.

Full size image

The areas with persistent improvement in OLDI are primarily concentrated in the Gonghe Basin Grassland Ecological Zone, the Huangnan Alpine Grassland Ecological Zone, and the Hoh Xil area of the Yangtze River Source Grassland Ecological Zone. Regions that had previously degraded but are projected to improve in the future are mainly located in the southeastern Yellow River Source Grassland Ecological Zone, the eastern Yangtze River Source Grassland Ecological Zone, and the northwestern Lancang River Source Grassland Ecological Zone. With the climate trend shifting towards warming and humidification, slight increases in precipitation have benefited vegetation growth and promoted land recovery. Concurrently, low temperatures combined with recent warming trends have improved vegetation growth conditions to some extent. In addition, policies such as grazing bans, balance management of livestock and pastureland, and rotational grazing with designated prohibition zones have effectively controlled livestock numbers and grazing intensity, enabling grassland recovery. These combined factors have contributed to vegetation growth and land restoration, ultimately supporting the sustainable development of the ecosystem60. The proportion of persistently degrading land is relatively low and distributed sporadically, while areas that had previously improved but are projected to degrade again are relatively widespread across the Sanjiangyuan region. These regions continue to face challenges. On one hand, the uncertainty of global climate change poses risks—if the trend of warming and humidification reverses towards warming and drying, vegetation growth may be hindered, and permafrost degradation may continue to affect soil moisture and root system development61. On the other hand, anthropogenic pressures remain difficult to eliminate completely. Despite existing protection measures, population growth and economic development may lead to overgrazing, illegal mining, and infrastructure expansion, which could place renewed stress on vegetation and land, thereby accelerating degradation. Furthermore, biological factors such as rodent infestations may resurge under certain conditions, further threatening ecosystems and intensifying land degradation62.

Thus, land degradation has profound ecological, economic, and social consequences63. Ecologically, it damages vegetation, reduces biodiversity, disrupts ecosystem balance, and intensifies soil erosion, desertification, and fertility decline, ultimately reducing land productivity. Economically, it decreases crop yields and livestock numbers, constrains the development of agriculture and related industries, and increases restoration costs. Socially, it exacerbates poverty, depriving some local populations of their livelihoods and potentially triggering resource conflicts and social instability. To mitigate land degradation, a comprehensive strategy must be adopted. At the policy level, the government should enforce strict land protection regulations, limit unsustainable development, increase investment in ecological conservation, and implement restoration projects. At the technical level, sustainable agricultural practices should be promoted, including precision fertilization, water-saving irrigation, and crop rotation, to enhance land use efficiency. Advanced monitoring technologies should be employed to track land changes in real time and provide data-driven solutions for restoration. At the societal level, environmental education and public engagement should be strengthened to enhance awareness of ecological protection, encourage collective participation in land conservation, and foster a shared responsibility for sustainable land use.

Research significance and implications

This study developed the Optimal Land Degradation Index (OLDI), a novel composite indicator based on a constrained optimization algorithm, to systematically assess land degradation in the Sanjiangyuan region from 2001 to 2023. The results demonstrate that while overall land conditions have improved during the past two decades, nearly half of the region remains at risk of future degradation, with about 49.44% of currently improved areas potentially experiencing reversal. These findings highlight the urgency of strengthening ecological conservation and implementing dynamic, adaptive management strategies. At the scientific level, the OLDI integrates vegetation, soil, and other multi-source indicators, effectively overcoming the limitations of traditional single metrics in complex alpine environments. This approach provides a more comprehensive and transferable framework for land degradation monitoring, which can be extended to other fragile ecosystems. At the practical level, the identification of high-risk degradation zones and their dominant driving factors offers precise spatial targeting for ecological restoration and land management, directly supporting policy design and implementation.

To consolidate ecological gains and mitigate future degradation, an integrated governance framework is recommended, emphasizing differentiated zonal strategies, technological support, and social participation. This includes stricter land-use regulations, ecological restoration in overgrazed and erosion-prone areas, sustainable agricultural practices, and the use of advanced monitoring technologies. Moreover, the study supports the establishment of an integrated “monitoring–early warning–management–evaluation” mechanism, which can enhance the intelligent management of protected areas such as the Sanjiangyuan National Park. Overall, the proposed OLDI framework not only advances methodological innovations in alpine land degradation assessment but also provides actionable insights for ecological protection and sustainable land management. The research contributes to safeguarding the ecological security of the Qinghai–Tibet Plateau and offers strategic value for balancing human–land interactions and promoting regional sustainability.

Limitations and future prospects

This study predicts the future land degradation trends in Sanjiangyuan based on the Sen trend and Hurst index. However, land degradation is influenced by multiple natural and anthropogenic factors, and its evolution is highly diverse and uncertain. In future research, there is an urgent need to construct more comprehensive models that not only analyze the internal mechanisms of land degradation in depth but also provide a solid scientific basis for the sustainable management and protection of land resources. This is essential in order to slow down or even reverse the undesirable trend of land degradation.

Conclusions

This study developed the Optimal Land Degradation Index (OLDI) by integrating vegetation and soil indicators through a constrained optimization algorithm, providing a comprehensive and reliable framework for assessing land degradation in the Sanjiangyuan region from 2001 to 2023. The findings reveal that land degradation exhibits a stepwise southeast-to-northwest distribution, with high-risk areas accounting for 41% of the total region. Over the past two decades, land improvement has been the dominant trend, covering 27% of the total area, though 2% of the land still shows persistent degradation concentrated in the southeast. The driving factor analysis highlights precipitation, slope, human activity intensity, and temperature as dominant forces, with significant spatial heterogeneity across ecological subzones. Looking ahead, nearly half of the improved land (49.44%) is at risk of reversal, underscoring the urgency of ecological protection. The novelty of this work lies in overcoming the limitations of traditional single-indicator assessments, offering a transferable and adaptive method for fragile alpine ecosystems. From a policy perspective, the results provide precise spatial targeting for restoration, land-use regulation, and adaptive management, and they emphasize the necessity of establishing an integrated “monitoring–early warning–management–evaluation” mechanism. Future research should focus on refining the OLDI framework by incorporating socio-economic drivers, climate change scenarios, and high-resolution monitoring technologies, thereby offering deeper insights into degradation mechanisms and stronger support for sustainable land management and ecological security.