Introduction

Wind erosion presents a profound challenge to global environmental stability, with studies indicating that roughly one-quarter of the world’s land surface is susceptible to this process1, particularly in arid and semi-arid regions2,3. Wind erosion is a critical issue in arid regions, where soil degradation exacerbates the vulnerability of ecosystems and human populations. In China, the situation is even more critical, as over half of the national territory experiences wind erosion4,5, with desert grasslands in Inner Mongolia facing heightened vulnerability. This is of particular concern as wind erosion not only degrades soil quality but also leads to the loss of valuable agricultural land, negatively affecting biodiversity, agricultural productivity, and the livelihood of millions of people in these regions. Consequently, understanding the key drivers of aeolian soil degradation in these areas is imperative, not only for the development of effective mitigation strategies but also for improving the long-term sustainability of these vulnerable regions.

A large body of academic work over several decades has provided insight into the fact that soil wind erosion is a complex process involving multiple aspects of meteorological conditions, soil properties, vegetation characteristics, and human activities6,7,8,9. Studies have identified meteorological factors as determinants of wind erosion to some extent10,11. Wind speed, as the main dust emission driver, directly influences dust emissions in dust source regions12,13, and soil erosion and dust emissions occur once wind speed reaches a certain level14, especially in arid regions with sparse vegetation15. Recent studies have shown that wind erosion and dust emissions are positively correlated with wind speed in eastern Asia16.

In addition to wind speed, meteor22ological elements such as precipitation and temperature also have important effects on wind erosion and dust emission processes. For example, dust emissions in eastern Asia show a significant negative correlation with precipitation17, and precipitation directly affects dust emissions in dust source regions13,18,19and may change the degree of vegetation cover, which in turn affects dust emissions20. Precipitation also affects surface humidity, which increases the difficulty of dust initiation and thus affects dust emissions21. Additionally, surface humidity increases after precipitation, promoting particle adhesion and reducing the probability of dust storms22. The highly negative correlation between dust events and precipitation in semi-arid regions23further supports the critical role of precipitation in regulating dust emissions and mitigating wind erosion. Temperature also plays a non-negligible role in the wind erosion process. Studies have shown that temperature is correlated with dust concentration12,24and can affect dust emission by influencing meteorological elements such as wind speed, precipitation, and soil moisture. In addition, temperature affects wind erosion by influencing soil moisture, surface evapotranspiration, and vegetation growth25, and even overall precipitation, temperature, and humidity variations can influence interannual variations in regional dust emissions26.

In addition to meteorological factors, vegetation has long been recognized as a key factor in protecting soils from wind erosion4,27,28,29,30 Vegetation acts as an anti-erosion effect by increasing surface roughness and absorbing the downward momentum of ambient air currents31,32, reducing the exposure of the soil surface, and decreasing the incidence of wind erosion and dust emissions33. The composition and canopy characteristics of vegetation also significantly affect sand and dust emissions. Vegetation cover tends to mitigate wind erosion more effectively at lower wind speeds, but this protective effect diminishes as wind speed increases, especially when vegetation is sparse4.

Soil erodibility factors play a crucial role in wind erosion and dust emission processes. For example, different fractions of soil organic matter, such as humic acids and saccharides, can promote soil agglomeration and thus enhance its erodibility34,35,36. In addition, soil organic matter can also affect the number and stability of water-stable aggregates, thereby influencing their resistance to wind erosion37,38. Therefore, soil organic matter can enhance the overall erosion resistance of soil through its own composition and interaction with other solid-phase materials. In addition, the content of different components in soil particles also affects their erosion resistance. Studies have shown that increasing the content of sand and powder particles leads to a decrease in the number of aggregates, which reduces the stability of the soil structure and makes the soil more susceptible to wind erosion39. In addition, it has been found that soil particles with particle size less than 0.84 mm40 or in the range of 0.075–0.4 mm29,41are more susceptible to erosive activities under wind action.

Numerous studies have shown that meteorological factors, vegetation conditions, and soil properties all have a significant impact on wind erosion and dust emissions. However, although the effects of these factors have been widely characterized, the vast majority of studies remain at the stage of qualitative analysis, and few studies have been able to quantitatively assess the relative importance of each factor in wind erosion events. At present, comprehensive assessments of the specific contributions of different factors to wind erosion and their importance are still insufficient2,42,43,44. One major gap in the current literature is the lack of research on the interactions between wind erosion influencing factors. Few studies have examined how these factors—such as precipitation, wind speed, temperature, and soil properties—interact to exacerbate soil degradation. Recent research has suggested that the combined effects of these climatic factors significantly increase the risk of wind erosion45,46. Yet, the complex dynamics of these interactions remain insufficiently explored. Studies have shown that reduced precipitation leads to soil moisture deficits and loose structure, while increased wind speed further enhances soil susceptibility to wind erosion, and increased temperature exacerbates the negative impacts on soil structure46,47. In addition, high organic carbon and moisture content in soils not only promote vegetation growth, but also significantly slow down the occurrence of wind erosion by enhancing the binding of soil particles. Especially in arid regions, soil desiccation exacerbates the wind erosion process48,49. However, the interactions between these factors have not been adequately clarified, leading to a significant gap in our understanding of wind erosion mechanisms and hindering the development of targeted and effective control strategies.

Therefore, the selection of suitable methods to quantify the effects of surface factors on processes such as wind erosion and dust emissions should be a current problem that needs to be urgently addressed. In recent years, new methods and techniques have been widely used and explored in the study of soil wind erosion problems. Geographic Detectors methods have emerged as a powerful tool in soil erosion research, capable of identifying the multifaceted contributions of different factors and spatially detecting correlations between them50,51. This method has been successfully applied in water erosion research to evaluate influential factors and pinpoint high-risk areas52,53,54. However, its application to wind erosion remains underexplored, highlighting an important gap in the literature.

The issue of desertification is becoming increasingly severe, with an area of 38 million square kilometers of land affected, and this number continues to rise, posing significant challenges to global ecological stability. In China, the area of wind erosion in the northern regions has reached 1.88 million square kilometers, with the desert grasslands of Inner Mongolia being a typical affected area. As a key ecological barrier in the north, the desert grasslands of Inner Mongolia play a critical role in maintaining ecological balance55,56. However, the ecosystem of Inner Mongolia’s desert grasslands is extremely fragile, and wind erosion is particularly pronounced55,57. Due to prolonged exposure to harsh climatic conditions and frequent wind and sand activities, the environment for vegetation growth is poor, and restoration faces numerous challenges58. These factors make the intensification of soil wind erosion a major driver of regional ecological degradation59,60.

Soil wind erosion has caused significant damage to the desert grassland ecosystem and has become a primary obstacle to ecological restoration61. Soil degradation not only weakens the foundation for vegetation growth but also leads to a decline in species diversity, further exacerbating the difficulty of vegetation recovery, creating a vicious cycle that ultimately affects the stability of the regional ecosystem. As global climate change intensifies and human activities increase, the frequency and intensity of wind erosion in desert grasslands have significantly risen. In recent years, China has invested substantial resources into ecological protection and restoration, implementing major projects like the “Three-North Shelterbelt Program,” which constructs green protective barriers through large-scale afforestation, and the Beijing-Tianjin Wind and Sand Source Control Project, which focuses on the governance of sand source areas in the Beijing-Tianjin region. These measures have somewhat slowed the degradation of desert grasslands and delayed the ecological deterioration process55. However, according to the latest reports, the desert grasslands still remain the most heavily eroded areas in the Beijing-Tianjin wind-sand source region, with many areas still experiencing severe erosion55. This suggests that, despite the existing efforts, the results of wind erosion control in desert grasslands remain inadequate, necessitating further enhancement and improvement.

Through in-depth analysis, we have found that the root cause of addressing wind erosion lies in the lack of a clear and comprehensive understanding of the key driving factors and their interrelationships62. Given the complexity and persistence of this ecological issue, relying solely on existing control measures is insufficient to address the underlying causes. Therefore, there is an urgent need to optimize and improve the control system based on a deeper understanding of the key factors of wind erosion and their interactions. For example, in-depth studies should be conducted on the roles of climatic factors63, soil factors64, vegetation factors29, and human activities65 in the wind erosion process and their interrelationships66,67. Through the development of scientific models and simulation analyses, we can accurately identify the primary driving factors of wind erosion in different regions, thus aiding in the formulation of targeted control strategies. Simultaneously, long-term monitoring and evaluation of existing control projects must be strengthened to identify problems in a timely manner and adjust control plans accordingly. By continuously optimizing the effectiveness and adaptability of control measures, we can achieve substantial improvements in the ecological environment of desert grasslands and promote their sustainable development. This will not only help protect this important ecological barrier but also contribute to maintaining ecological balance at both regional and global levels.

Under the severe challenges of wind erosion in the desert grassland region of Inner Mongolia, we carried out this work with the following objectives: firstly, to accurately identify the key factors affecting wind erosion by using Geographic Detectors, so as to provide new perspectives and methods for future wind erosion research. Secondly, the complex interactions between these factors can be explored in depth to provide theoretical support for the development of scientifically sound wind erosion models and to predict the wind erosion process more accurately. Finally, by accurately identifying the spatial distribution of key factors with high wind erosion risk, we can formulate effective management programs in a targeted manner to improve management efficiency, and contribute to the ecological environmental protection and sustainable development of desert grassland areas.

Materials and methods

Study area

This research focuses on a desert steppe region situated centrally within China’s Inner Mongolia Autonomous Region, spanning geographically from 41.3° to 43.8° N latitude and 107.4° to 114.8° E longitude and encompassing approximately 63,000 km2 (Fig. 1). The locale experiences a semi-arid desert-steppe climate. Thermal conditions average 2.5 °C annually, while Precipitation (PPT) levels approximate 217 mm per year, primarily concentrated between July and September. Gale-force winds are predominantly observed from March through May, averaging a wind velocity of 4.5 m s− 1 annually. Notably, winds reaching 8 m s− 1 at a 2-meter elevation occur on 23.9–51.4% of the annual timescale. The vegetative landscape is largely composed of Stipa klemenzii, Stipa breviflora, and Stipa plareosa species, characterized by low-growing herbaceous plants with a typical coverage of less than 40%. The predominant soil classifications include zonal light chestnut calcareous soils (21.4%), brown calcareous soils (20.3%), light brown calcareous soils (19.7%), and chestnut calcareous soils (13.1%), all primarily sandy in texture. Average sand and silt compositions are 45% and 50% respectively, contributing to a coarser surface texture due to the confluence of minimal rainfall and prevalent strong winds.

Fig. 1
figure 1

Spatial distribution distribution map of sampling points (Cartographic software: ArcGIS 10.5 https://www.esri.com/zh-cn).

Full size image

Data source and processing

Sampling point layout and sampling

In this study, a comprehensive sampling strategy was employed to ensure homogenous site distribution, representativeness of characteristics, and accessibility. Prior to the field survey, the 2019 and 1984 Landsat 15 m resolution images were utilized in Google Earth to compare land use changes, analyze topography and landforms, and pre-select potential 137Cs reference sites. The longitude and latitude of these sites were then imported to a GPS device to guide the field survey. Nine reference sites were initially selected, based on their topography and vegetation coverage, all of which were high-altitude, flat steppes with dense vegetation. After removing four sites with inadequate characteristics, five sites were finalized.

The selection of the 137Cs background value sample adhered to the criteria outlined by68. Prior to conducting the field survey for the background value samples, several potential sampling points were initially identified using the 2019 Magyar Technologies 3-meter resolution imagery available via Google Earth software. This allowed for the pre-selection of locations based on geographic features such as vegetation cover and elevation.

Subsequently, a field survey was conducted to verify the suitability of these pre-selected points. The field survey primarily focused on ensuring that the selected points met the criteria of having at least 70% vegetation cover. This threshold was chosen because vegetation cover exceeding 70% is generally regarded as sufficient to minimize soil disturbance and ensure a stable, undisturbed ecological environment. During the survey, any points with less than 70% vegetation cover were excluded, as these were considered to be too disturbed or inadequately covered by vegetation69.

Additionally, points with more than 70% vegetation cover but located in low-lying areas were also excluded. This was done to avoid areas that may be prone to waterlogging or water erosion, which could distort the background value due to hydrological effects. By applying these criteria, the points selected for the background value samples were ensured to have high vegetation cover and be situated at higher elevations, minimizing the risk of water-related disturbances.

In general, the points selected were unlikely to experience significant erosion or accretion, meeting the local criteria for background values and ensuring the reliability and accuracy of the sample data.

Sampling at each site included two main categories: 137Cs background value samples and soil erosion/sedimentation samples. A total of 30 background value samples and 552 erosion/sedimentation samples were collected. For the 137Cs background value samples, five sampling points were established, and soil samples were stratified into six layers (0–5 cm, 5–10 cm, 10–15 cm, 15–20 cm, 20–25 cm, 25–30 cm) using a root auger with an inner diameter of 8 cm and depth of 15 cm. The stratified sampling was designed to ascertain 137Cs activity levels and analyze soil particle composition and other physicochemical properties.

Additionally, the study focused on 137Cs activity analysis by collecting whole core and sectioned core samples at the reference sites. The whole core samples, collected at each site’s vertex and diagonal intersection points within a 10 m x 10 m grid, were used to measure the total 137Cs deposition. A total of 90 whole core samples were obtained, with 54 sectioned core samples taken to study the 137Cs distribution within soil profiles. After collecting the samples, the 137Cs activity was measured, and the erosion process was analyzed based on the 137Cs activity of the sectioned core samples. Four sites with clear erosion and deposition features were excluded, and five remaining 137Cs reference sites were used for further analysis.

This combined approach ensured that the data collected from the various sampling methods were comprehensive, with both soil physicochemical characteristics and vegetation data, which allowed for a thorough understanding of 137Cs distribution and its correlation with soil erosion and sedimentation processes.

To ensure the representativeness and accuracy of the soil samples, three sampling locations were randomly selected using the parabolic method, and the soil samples from each location were thoroughly mixed before being collected. Vegetation cover and height were surveyed using 1 m × 1 m sample plots, with five people estimating the cover and averaging the values. Vegetation height was measured at the four corners and center of the plot, and the average value was taken.

Determination of 137Cs activity and erosion modulus calculation

The collected 137Cs soil samples were firstly dried naturally in a cool and ventilated place, and then plant and animal residues, stones and other materials were removed, and after grinding, they were passed through a soil sieve with a pore size of 2 mm, and 400 g of the sieved soil samples were put into a cylindrical box of polyethylene material and waited for the determination of the 137Cs activity. The activity was determined using a high-resolution GMX50P4N high-purity germanium probe γ detector (measurement range 3Kev-10 Mev, detection efficiency 50%) produced by ORTEC, USA, with a test time of > 27,000 s.The full peak area method was used to calculate the value of the net peak area, and the mass activity of 137Cs was obtained based on the relative comparative measurement method. When selecting a model for estimating the erosion rate, there are numerous available models; however, the majority of them are designed for cropland scenarios, while there are relatively few models tailored for undisturbed soil environments. Given this limitation, we focused on two models: the profile distribution model70and the extended migration model71.

The extended migration model, while comprehensive, considers a variety of parameters that are difficult to measure or obtain in practical applications, limiting its widespread use. This complexity often requires extensive data input, making it less suitable for cases where data availability is constrained or where a simpler model is desired. In contrast, the profile distribution model simplifies the distribution of non-agricultural cropland profiles into an exponential form, requiring fewer parameters for estimation.

The simplicity of the profile distribution model does not compromise its accuracy. While it uses fewer parameters, the exponential form captures the main characteristics of 137Cs profile distribution effectively in undisturbed soils. Research has shown that this exponential model provides a close fit to experimental data for 137Cs distribution in a variety of soils, even though it lacks the complexity of other models. In fact, this simplification is advantageous because it reduces the risk of overfitting, which can occur with more complex models that rely on a larger number of parameters. The model’s simplicity allows for better generalization across different soil types and environmental conditions, ensuring its accuracy remains intact.

The profile distribution model has been widely used in studies of soil erosion and radionuclide migration due to its proven reliability and ease of use. Many scholars have used it in various soil types, demonstrating its ability to provide accurate estimations of erosion rates without requiring complex parameterization72,73,74,75.The model’s advantages—simplicity, fewer data requirements, and ease of implementation—make it a highly suitable tool for erosion rate estimation, particularly in areas with limited data on soil properties and erosion processes.Therefore, in this study, the profile distribution model70was chosen to calculate the soil erosion rate, and the model is shown in (Eq. 1)

$$\:\text{X}={\text{X}}_{0}\bullet\:{\text{e}}^{-{\uplambda\:}\bullet\:\text{h}\bullet\:(\text{T}-1963)}$$
(1)

where X is the current 137Cs area activity (Bq m− 2) at the sampling site; X0 is 137Cs background value (Bq m− 2); λ is 137Cs background value profile distribution morphology parameter (obtained from 137Cs content distribution in the profile); h (cm yr− 1) is the annual average soil wind erosion thickness since 1963; and T is the sampling year.

Acquisition of relevant indicators

To effectively assess the relative contributions of wind erosion factors, it is imperative to first delineate the pertinent indicators for each factor and acquire the corresponding data. Wind erosion is influenced by various elements including soil properties, vegetation, meteorological conditions, and topography. For soil parameters, Soil Organic Matter (SOM) is evaluated using the potassium dichromate oxidation method in concentrated sulfuric acid. Particle size distribution is analyzed using an R2000 laser particle size analyzer, while calcium carbonate content is determined through a gasometric method. Vegetation metrics, including Vegetation Height (VH) and cover, are derived from field surveys at the sampling points,

In order to ensure the accuracy of the measured index factors, soil samples were processed in accordance with the requirements of conventional soil physical and chemical experiments, and three parallel sample experiments were done for each soil sample index to ensure the accuracy of the index factors, such as soil organic matter, calcium carbonate content, and the percentage of clay, sand, and silt content, whereas Normalized Difference Vegetation Index (NDVI) is sourced from the Resource and Environmental Science and Data Center platform. Meteorological data, encompassing average annual temperature, average annual PPT, and potential evapotranspiration, are procured from Resource and Environmental Science and Data Center platform (https://www.resdc.cn/Default.aspx). Information on the frequency of windy days (categorized by wind speeds of 5 m s-1, 6 m s-1, 7 m s-1, and 8 m s-1) draws on wind speed data statistics from 48 meteorological stations within the study area, spanning from 1990 to 202076,77. Topographical data, specifically 12.5-m resolution Digital Elevation Model (DEM) data, are sourced from Advanced Land Observing Satellite (ALOS). Soil water content is ascertained using the standard drying method. The spectrum of values for each index within the study area is detailed in Table 1. This comprehensive approach ensures a robust dataset for a nuanced evaluation of the diverse factors influencing wind erosion in the region.

Table 1 Desert steppe wind erosion factor range.
Full size table

Geographic detector method and data processing

The geographic detector method reveals the potential determinants of the dependent variable by examining the inherent spatial heterogeneity of various factors. This technique requires the use of categorical (type) variables. Since the influencing factors in this study are numerical, they need to be converted into categorical variables. To do this, the natural breaks (Jenks) method is used to stratify the data, dividing continuous numerical data into several categories. Specifically, statistical analysis is performed on each influencing factor to determine appropriate stratification thresholds, ensuring that the sample size for each stratum is relatively uniform, while retaining the spatial heterogeneity of the data. The data, which include slope, elevation, average temperature, average precipitation (PPT), and potential evapotranspiration, are initially presented in raster format. These data can be directly converted into categorical variables through reclassification. For other data types, such as soil and vegetation, which are derived from point-based surveys, spatial interpolation is used to convert them into surface data, which are then reclassified.

The stratification thresholds for all influencing factors are determined using the natural breaks (Jenks) method, which minimizes internal variability within each class by identifying natural breaks in the data distribution and ensuring differences between classes. The reclassified data for each determinant is resampled into a 5 × 5 km resolution grid, ultimately aggregated into 3,117 raster datasets. These data serve as independent variables, while the soil erosion modulus serves as the dependent variable, undergoing the same treatment. Then, the geographic detector is used to explore the contribution of each independent variable.

This method principally assesses the explanatory power of independent variables over the dependent one, quantified by the q-value78. The q-value’s computational formula is delineated in Eqs. 2, 3, and 4. Results from interaction detection are instrumental in deciphering the interplay between pairs of factors, ascertaining whether their combined influence on the dependent variable is null, synergistic, or antagonistic. Moreover, risk detection facilitates the exploration of how disparate independent variables influence the dependent variable across various strata. The intensity of this influence is conveyed through the mean risk value among the strata, mirroring the spatial heterogeneity of the impact that different strata of the variables have on the dependent variable.

$$q = 1 - \frac{{\sum\nolimits_{{h = 1}}^{L} {N_{h} \sigma _{h} ^{2} } }}{{N_{\sigma } ^{2} }} = 1 - \frac{{SSW}}{{SST}}$$
(2)
$$SSW = \sum\nolimits_{{h = 1}}^{L} {N_{h} } \sigma _{h} ^{2}$$
(3)
$$SST = N_{\sigma } ^{2}$$
(4)

where h = 1,2…; L is the stratification of the independent variable X; Nh and N are the number of cells within layer h and in the region; σh2 and σ2 are the variance of Y values in layer h and the region, respectively; SSW is the sum of variance within the layer, and SST is the total variance in the region; and q represents the explanatory power of the independent variable on the dependent variable, ranging from 0 to 1. A higher value of q indicates a stronger explanatory power of the independent variable on the dependent variable. Analyzing the size of q can aid in identifying the dominant factor of soil erosion.

Results

Intensity and Spatial distribution of soil erosion modulus in the study area

Figure 2 shows the distribution of activity profiles for the five 137Cs background values of the study. The accuracy of the 137Cs background values is directly related to the accuracy of the wind erosion rates. To ensure the reasonableness of the 137Cs background values in this study, the data of this study were compared with the 137Cs background values of the research results in the neighboring areas. In order to ensure the reasonableness of the 137Cs background values in this study, the data of this study were compared with the Cs background values of research results in neighboring areas. It was found that the 137Cs background values in this study ranged from 1486.34 to 1772.04 Bq∙m− 2, with a mean value of 1680 Bq∙m− 2。.

The average soil erosion modulus of the desert steppe in Inner Mongolia is 1,680 tons per square kilometer per year, which is at a moderately high level. In terms of regional comparison, its degree of erosion is between that of the western desert, the typical grassland in Northeast China, and the typical grassland in China, and soil erosion is relatively significant. In the western desert area of the Mongolian Plateau, more than 72% of the land is affected by soil erosion of different degrees79. The proportion of moderate, severe, and very severe erosion reaches 49%, and the wind erosion severity in the Gobi desert area in eastern Mongolia is 5,000 tons per square kilometer per year80. In comparison, the degree of erosion in the desert steppe in Inner Mongolia is lower. The soil erosion modulus of the Hulunbuir Grassland in the typical grassland of Northeast China is 171 tons per square kilometer per year81, which is in the range of mild to moderate erosion, and the degree of erosion is significantly lower than that of the desert steppe in Inner Mongolia. The soil erosion modulus of the Xilingol Grassland in the typical grassland of China is 355 tons per square kilometer per year82, which is in the moderate erosion range and also lower than that of the desert steppe in Inner Mongolia. Nationwide, the soil erosion modulus of the desert steppe in Inner Mongolia is at a moderately high level. In different regional comparisons, its degree of erosion is lower than that of the western desert, but higher than that of the typical grassland in Northeast China and the typical grassland in China.

The 137Cs activity for soil erosion/accumulation samples ranged from 32.14 Bq m− 2 to 2575.8 Bq m− 2, with an average of 454.8 Bq m− 2. The calculated average soil erosion modulus for the Inner Mongolia desert grassland stood at 1592.5 t km− 2a− 1, accompanied by a standard error of 100.77 t km− 2a− 1, and a 95% confidence interval between 1383.3 t km− 2a− 1 and 1770.3 t km− 2a− 1. The spatial distribution of the soil erosion modulus displayed a distinct pattern, with heightened levels in the northern and southern regions, contrasted by lower levels in the central segment of the desert grassland (Fig. 3).

Fig. 2
figure 2

Distribution map of background activity profile for 137Cs (Cartographic software: OriginLab origin2022 https://www.originlab.com).

Full size image
Fig. 3
figure 3

Spatial distribution of soil erosion (Cartographic software: ArcGIS 10.5 https://www.esri.com/zh-cn).

Full size image

Soil erosion factor detection results

The detection of factors influencing soil erosion revealed variability in the explanatory power of each, as depicted by the q-values in Fig. 4. SOM emerged as the most significant, boasting an explanatory power of 0.6. VH and average annual PPT followed, each with an explanatory power exceeding 0.5, while factors such as evapotranspiration and temperature exhibited explanatory powers ranging from 0.4 to 0.5. In aggregate, factors with an explanatory power above 0.4 constituted 20% of all considered elements, encompassing soil, vegetation, meteorological, and additional variables, thereby underscoring the multifaceted nature of soil wind erosion processes.

An analysis comparing the spatial distribution of factors with an explanatory power above 0.4 (Fig. 4) against the soil erosion modulus’ spatial distribution suggested a pronounced alignment. Specifically, these influential factors mirrored the soil erosion modulus’ spatial distribution (Fig. 5), with a discernible south-to-north gradient in the data.

Fig. 4
figure 4

Explanatory power of factors on the spatial distribution of erosion (Cartographic software: OriginLab origin2022 https://www.originlab.com).

Full size image
Fig. 5
figure 5

Spatial distribution of factors with q ≥ 0.4 (Cartographic software: ArcGIS 10.5 https://www.esri.com/zh-cn).

Full size image

Interaction results of soil erosion factors

The results of the interaction analysis indicate that the combined effects of different factors significantly enhanced the explanatory power of soil wind erosion, surpassing the contribution of individual factors (Fig. 6). Specifically, the explanatory power range of the SOM group after interaction was between 0.62 and 0.79, 80% of the interactions in the VH group exceeded 0.6, and 75% of the interactions in the PPT group also surpassed this threshold. Overall, 65.7% of the interactions had an explanatory power greater than 0.6, indicating that the combined effects of the factors are more significant. The study further highlights that wind erosion is a process driven by the synergistic interaction of multiple factors, rather than being dominated by a single factor.

From a broader perspective, the results of these interactions reveal the complexity of the soil wind erosion process. Most factor interactions exhibited nonlinear enhancement, while a few showed linear enhancement, suggesting that wind erosion is not solely driven by a single factor but rather results from the combined effect of multiple factors. Nonlinear interactions indicate that the combined effects of different factors far exceed the impact of individual factors, especially under the joint influence of climate and soil factors. For example, the interaction between soil organic matter (SOM) and factors like precipitation and temperature suggests that the combination of these factors may intensify the risk of wind erosion under specific conditions. In contrast, linear enhancement interactions reflect a more direct and stable relationship between certain factors, such as the interaction between wind speed and slope. These combinations can amplify the likelihood of wind erosion within a certain range, and their enhancement effects are relatively predictable.

This finding emphasizes the multidimensional nature of soil wind erosion, underscoring the need to consider the interactions between factors when developing effective prevention and control strategies, rather than relying solely on the analysis of individual factors. Successful mitigation and prediction strategies require a comprehensive evaluation of these interactions to more accurately reveal the dynamic complexity of soil wind erosion. Therefore, effective environmental management strategies should adopt an integrated approach, considering the combined effects of all influencing factors, in order to address the multidimensional changes and complex interactions within the wind erosion process.

Fig. 6
figure 6

q value of each indicator factor after interaction (Cartographic software: OriginLab origin2022 https://www.originlab.com).

Full size image

Spatial distribution of erosion risk areas

Table 2 Erosion risk values for different groups of factors.
Full size table

The intra-stratum variance after stratification of each factor is the risk mean, which represents the level of risk that the intra-stratum variable has on the presence of the dependent variable, for a more detailed explanation, please refer to the literature78.The geographic detector analysis revealed that wind erosion risk levels fluctuate in response to different values of specific factors. As Table 2 illustrates, there is variability in erosion risk when the factors are assessed at different graded values. Notably, erosion risk reaches a peak value of 6.5 and descends to a minimum of 1.20 across the various factors. These findings facilitated the stratification of risk values into three distinct categories: high risk (risk value ≥ 4.5), medium risk (2.5 ≥ risk value > 4.5), and low risk (risk value ≤ 2.5). This categorization process allowed for the segmentation of each factor’s internal space into corresponding erosion risk classes. The detailed outcomes of this segmentation are presented in Table 3.

Table 3 Erosion risk classification based on various factors.
Full size table
Fig. 7
figure 7

Distribution of corresponding erosion risk levels for each factor (Cartographic software: ArcGIS 10.5 https://www.esri.com/zh-cn).

Full size image

Analyzing Table 4 in conjunction with Fig. 7 reveals that areas with lower levels of organic matter content, vegetation cover, and average annual rainfall exhibit a relatively high risk of wind erosion. Conversely, areas with higher values of temperature and evapotranspiration also present a heightened erosion risk. Predominantly, the spatial distribution of wind erosion risk follows a pattern of higher intensity in the north and lower in the south. To further elucidate the interactions among these factors, the spatial erosion risks of several factors with a q-value greater than 0.4 were overlaid. In these superimposed areas, the more severe risk assessment between two factors determined the assigned risk value. If a high-risk area coincided with a medium-risk zone, the combined area was designated as high risk. As depicted in Fig. 8, the composite distribution of erosion risk following factor interaction mirrors the single-factor trend of higher risk in the north and lower in the south. However, the area designated as high-risk for erosion is more expansive in the composite analysis than when factors are considered individually. For instance, the high-risk zones for factors like average annual PPT and temperature encompass those for SOM and VH, suggesting that this composite overlay provides a more precise spatial representation of wind erosion risk levels.

Fig. 8
figure 8

High explanatory power factor (q > 0.4) interaction results (Cartographic software: ArcGIS 10.5 https://www.esri.com/zh-cn).

Full size image

Discussion

Key determinants in the wind erosion dynamics

This investigation identifies SOM, VH, PPT, temperature, evapotranspiration, and their cumulative interactions as pivotal elements in the wind erosion process. It was found that the 137Cs background values in this study ranged from 1486.34 to 1772.04 Bq∙m− 2, with a mean value of 1680 Bq∙m− 2 .This is very close to the range of Liu ‘s data, which ranged from 1159 to 1944 Bq∙m− 2 with an average value of 1588.17 Bq∙m− 2 (Table 4). Meanwhile, 137Cs is dominated by wet deposition, so the general cesium deposition should follow the same pattern of change as rainfall73,83. The pattern of change of our 137Cs background values is indeed consistent with the change of precipitation in the region, showing a trend of gradual increase from west to east. This is also consistent with the results of Liu’s study72. Taken in consideration, the 137Cs background values in this study are relatively reliable.

Table 4 Results of existing studies on 137Cs background values in the neighboring areas of the study area.
Full size table

These findings align with the consensus in extant literature, underscoring the significance of these variables2,84,85,86,87,88,89. Previous research has dissected these factors from a mechanistic stance. For instance, components of SOM—such as polysaccharides90,91,92, soil humus37, and organic-inorganic complexes arising during the decomposition of apoplastic materials, along with biological crusts at the soil’s surface—alter soil aggregate content91,93,94.This transformation modulates soil erosion intensity95,96, reaffirming the critical role of SOM. VH, another key determinant, influences wind erosion by altering surface roughness97, thereby affecting airflow dynamics and wind speed profiles7,98,99.Moreover, climatic parameters such as air temperature, PPT, and evapotranspiration impact wind-driven soil erosion both by modifying air current structures and wind velocities100 and indirectly via their influence on vegetation characteristics101. These factors also mediate the soil moisture content, thereby altering inter-particle adhesion67,102,103. Consequently, their interplay can directly or indirectly affect soil moisture, structure, and activity, culminating in observable impacts on wind erosion. As the pattern found in this study, the influencing factors all significantly enhanced the explanatory power of the wind erosion modulus after interaction. Therefore. These phenomena make us have to face up to the fact that there are complex interrelationships and non-independence among the various factors in the wind erosion process, which will add challenges to our deeper understanding and analysis of the wind erosion process. For a long time, the models developed for wind erosion (e.g., WEQ, RWEQ, WEPS, etc.) seem to ignore this point, which may imply that there are certain limitations in these models for wind erosion prediction, and that the contributions of each element should not be simply calculated independently, but should be appropriately adjusted, and that we need to consider the models that can be applied under the condition of multifactorial coupling to improve the accuracy of model prediction.

As shown in this study, human factors such as grazing and land use significantly exacerbate the intensity of wind erosion. In areas with high grazing intensity, plant cover decreases, soil is exposed, and wind erosion becomes significantly more severe, especially during the non-growing season when vegetation is sparse and land exposure is more pronounced104. This phenomenon suggests that grazing not only directly reduces soil protection by damaging plant cover but also increases exposed areas, making the soil more susceptible to wind erosion105,106. Furthermore, high-intensity grazing leads to the loss of important elements such as carbon and nitrogen in the soil, further intensifying the wind erosion process107. The decline in soil fertility, particularly in heavily grazed areas, increases the soil’s vulnerability to wind erosion104.

However, the wind erosion process is not only significantly influenced by human factors, but natural factors also play a key role. Soil organic matter (SOM), vegetation height (VH), precipitation (PPT), temperature, and evapotranspiration all interact to influence the soil’s resistance to wind erosion. For example, SOM enhances the soil’s resistance to wind erosion by improving soil structure and increasing surface roughness91,93. VH, on the other hand, influences wind erosion by affecting airflow patterns and surface roughness, indirectly regulating wind speed and thus affecting the intensity of wind erosion98. Additionally, climatic factors such as precipitation and evapotranspiration not only directly affect soil moisture but also influence vegetation, thereby indirectly impacting the wind erosion process101,103. These factors, in conjunction with human activities, work together to drive the occurrence and development of wind erosion.

Wind erosion is a complex process influenced by multiple factors. In this study, several key factors were chosen to quantify the impact of soil wind erosion in desert steppe areas, based primarily on their physical mechanisms, ecosystem characteristics, and support from existing literature. The selected factors include Soil Organic Matter (SOM), Vegetation Height (VH), Precipitation (PPT), temperature, and evapotranspiration, all of which play an important role in the wind erosion process and have demonstrated strong explanatory power and substantial spatial variability in the Q-value analysis of the geographic detector method.Geographic Detector Method and Q-value Selection Mechanism The geographic detector method is a statistical tool used to reveal the spatial heterogeneity between geographical phenomena and driving factors. By calculating the Q-values of different factors, we can quantify each factor’s contribution to wind erosion. In this study, factors such as SOM, VH, PPT, temperature, and evapotranspiration showed strong performance in Q-value analysis, indicating their significant explanatory power in the wind erosion process and their substantial spatial variability. Therefore, these factors were selected as the focus of analysis, aligning with our scientific hypotheses and data support regarding the wind erosion process. Interactions of Factors and Multi-factor Complexity The wind erosion process itself involves complex interactions of multiple factors, and the effect of a single factor may not fully reflect its independent role in wind erosion. Through geographic detector analysis, we were able to identify significant associations between these factors and wind erosion intensity, especially when considering the interactions between factors. The explanatory power of the results was significantly enhanced when interactions were considered. Therefore, focusing on the interactions of these key factors enables us to more clearly identify the dominant factors in the wind erosion process and their mechanisms. Avoiding Overcomplicating the Model and Enhancing Model Effectiveness While wind erosion involves numerous factors, introducing too many factors into the model may complicate it, increasing the risk of data noise and overfitting. To address this, we chose factors with high Q-values, avoiding potential issues of model complexity. By reducing the number of factors, we could more precisely capture the main driving forces of wind erosion, thereby enhancing the model’s interpretability and predictive accuracy. The introduction of too many factors may cause cross-interference between them, making it difficult to identify the independent contribution of each factor. Therefore, focusing on the most representative and explanatory factors enhances the predictive capability and scientific accuracy of the wind erosion model. Comparison with Existing Literature The results of this study are consistent with existing literature, confirming the widespread impact of factors such as SOM, VH, and meteorological variables on wind erosion2,84,89. By focusing on these factors, we align with previous research and provide stronger data support. In conclusion, the selection of these factors as the focus of analysis is based on their significant role in the wind erosion process and their strong explanatory power in statistical analysis. This approach avoids model complexity, ensuring the accuracy and operability of the results.

In desert grasslands, the calcium carbonate (CaCO3) content in surface soil particles has not shown a significant impact on the erosion process. It is possible that a moderate amount of CaCO3 can increase the content of clay and aggregates, thereby improving the soil’s resistance to wind erosion. However, excessive CaCO3 increases the proportion of easily erodible parts, leading to poorer soil mechanical stability. As the CaCO3 content increases, the soil’s resistance to wind erosion decreases108,109 also suggested that areas with higher CaCO3 content are more susceptible to wind erosion compared to areas with lower content109. Additionally, in the Revised Wind Erosion Equation (RWEQ) by the U.S. Department of Agriculture, CaCO3 content is considered an important factor in assessing soil erodibility, and it is believed to have a linear negative correlation with wind erosion rates. This relationship has been widely applied by many researchers110.This relationship has been adopted by many scholars111,112,113,114,115.

Soil erosion, especially wind erosion, is influenced by multiple factors, including meteorological conditions, soil properties, vegetation, and human activities. The interaction of these factors determines the extent and distribution of soil erosion. In desert grassland areas, wind speed is usually the main factor contributing to soil erosion, as strong winds can carry away loose soil particles, leading to severe wind erosion.

Evaluation of wind erosion risk and strategic mitigation measures

This research undertook an evaluative approach to soil erosion risk, focusing on both individual factors and their aggregate impact. By discerning various erosion risk intensities across desert grassland zones through this dual-lens, the findings hold substantial implications for directing wind erosion mitigation strategies within these regions. Historically, wind erosion assessments predominantly relied on empirical models like WEQ116,117, and WEPS85. Despite their regional specificity118,119,120,121, these models often lack foundational grounding in theoretical and physical processes66,122and necessitate localized adaptations and confirmatory testing67. This contrasts with the more developed application of the USLE model in hydrological erosion studies123,124,125,126, highlighting a developmental discrepancy in wind erosion risk appraisal. Addressing this gap, this study innovates by pivoting the evaluative focus towards the determinants of wind erosion, thereby mitigating model simulation uncertainties. This factor-centric approach facilitates the delineation of erosion risk levels across different locales127,128,129,130,131,132, providing actionable insights for effective erosion management. The importance of soil organic matter (SOM) in this study, especially its significant impact on soil wind erosion in desert steppe areas, mainly stems from its profound influence on soil structure and stability133. The presence of SOM enhances the cohesion among soil particles, improving the soil’s resistance to wind erosion134. In desert steppe regions, due to the arid climate and low vegetation coverage, soil particles are loose and vulnerable to wind erosion. However, SOM promotes the formation of water - stable soil aggregates by improving the soil structure135. This not only helps reduce the soil’s sensitivity to wind erosion but also enhances its water - holding capacity136, further enabling it to resist wind erosion under arid conditions. The optimization effect of an increase in SOM on soil structure is particularly significant in arid and semi - arid regions. This is because the soil in these regions tends to become loose and susceptible to wind erosion due to long - term insufficient moisture137. Therefore, the content of SOM directly affects the wind - erosion resistance of desert steppe soil, which is the main reason for its high explanatory power in this study. In addition, the interaction between SOM and other factors such as precipitation, temperature, and wind speed further intensifies its impact on soil wind erosion137. Especially under arid climates and strong wind conditions, the role of SOM becomes even more prominent. For instance, the identification of SOM and VH as high-impact factors for wind erosion suggests that future mitigation efforts could strategically target these areas, enhancing the precision and efficacy of preventative measures.

The desert steppe in Inner Mongolia is an important ecological security barrier in northern China, with significant ecological functions. Grassland degradation can lead to aggravated soil erosion and a significant increase in the wind erosion modulus, thus disrupting the stability of the ecosystem. Research shows that, overall, the wind erosion modulus in Inner Mongolia showed a significant decreasing trend from 2000 to 2019, yet wind erosion still poses a threat to the ecological environment138. Moreover, grassland degradation can cause a reduction in soil organic matter and a deterioration of soil structure, further exacerbating soil and water loss. If the desert steppe in Inner Mongolia degrades into a desert, the wind erosion modulus will increase substantially. This will not only weaken the ecological barrier function but also have a negative impact on the regional climate and water resources. Studies indicate that grassland degradation can lead to a decrease in precipitation, an increase in surface runoff, and a decline in the groundwater level139. Therefore, protecting grassland vegetation is a key measure for maintaining ecological balance.

The desert steppe in Inner Mongolia is not only an important ecological barrier but also has certain social functions and economic value. The grassland provides abundant resources for animal husbandry and is an important source of livelihood for herdsmen. However, human activities such as overgrazing and land reclamation have led to the aggravation of grassland degradation and soil erosion problems140. Research shows that moderate grazing (for example, 0.15–0.3 sheep per hectare per month) contributes to soil nutrient accumulation and the sustainable development of the grassland ecosystem141. In addition, grassland degradation has an impact on the regional water supply and the sustainability of agricultural production. For instance, the expansion of farmland leads to an increase in the soil wind erosion modulus142.

Limitations and future directions

This study applies the geographic detector analysis method to reveal the spatial distribution patterns of soil wind erosion and its relationship with geographic factors, providing crucial spatial decision support for soil wind erosion management78. The geographic detector method has significant advantages in factor analysis, factor interaction, and risk assessment. The study found that key geographic factors such as topography, land use, and precipitation have a significant impact on soil wind erosion127,143, providing a scientific basis for future land management and ecological protection efforts. However, despite using a multi-source dataset, including field measurements, remote sensing, and meteorological data, there are still some limitations. Although the distribution of field sample points is relatively uniform, the sampling density is less than ideal, especially in high-risk areas, which may affect the representativeness of the results. Additionally, the remote sensing and meteorological data mainly come from public data-sharing platforms, leading to inconsistencies in time scale and spatial resolution, making the factor analysis process more complex.

To improve the accuracy and reliability of future research, it is necessary to increase the sampling density in high-risk soil erosion areas and conduct more detailed assessments. For example, in high-risk areas, ground measurement point density could be increased to ensure that more sample data are obtained from key regions, thus improving sample representativeness. In sampling design, special attention should be given to wind erosion hotspot areas, such as dune areas, bare soil regions, and areas with sparse vegetation, as these are areas that are more prone to wind erosion and impacted by climate change. Therefore, future research should consider using enhanced sensor networks and UAV sampling technologies to achieve higher-density sampling. Moreover, obtaining multi-source data with longer time coverage, higher precision, and better resolution will help provide a deeper analysis of the impacts of geographic factors. Future research should also further explore how key factors influence the underlying mechanisms of wind erosion, thereby improving the explanatory power of each factor. For example, while previous studies have mainly focused on vegetation cover, identifying its key role in mitigating wind and water erosion processes144,145,146,147, vegetation height (VH) is often overlooked. Particularly in desert steppe ecosystems, the role of VH in wind erosion has not been sufficiently explored, and its attention is lower than that of vegetation cover. Therefore, future research should prioritize considering the potential mitigating effects of VH on wind erosion dynamics to more accurately reflect its importance in these processes.

In terms of research technology, the continuous development of artificial intelligence and machine learning provides promising tools for predicting the rate and trends of soil wind erosion148. By building more complex and accurate prediction models, integrating multi-source data such as geographic information and meteorological data, and applying PFI technology to determine the importance of different factors, the accuracy and reliability of feature selection can be improved149. Additionally, combining SHAP technology helps determine the thresholds of specific features, deepening the understanding of wind erosion mechanisms150. Future research should combine emerging technologies such as deep learning and reinforcement learning, by constructing complex neural network models to capture the intricate nonlinear relationships in the wind erosion process, further improving prediction accuracy. The combination of geographic detector methods with interaction analysis and FI technologies not only considers the spatial correlation of data but also handles larger and more complex soil wind erosion datasets151. By integrating multiple methods, researchers can achieve more comprehensive and precise wind erosion predictions, reveal interactions between different factors, and uncover new mechanisms and influencing factors. This integrated approach can not only improve the understanding of wind erosion processes but also provide better theoretical support for the development of global soil wind erosion prevention strategies. Ultimately, based on these technological advancements, researchers will be able to build more accurate and reliable wind erosion prediction models, advancing soil wind erosion research toward broader and more influential results, and providing support for addressing global soil erosion issues.

Conclusion

This research employed the 137Cs tracer method to ascertain the soil erosion modulus within the desert steppe zone, accompanied by an analysis of the spatial distribution patterns of soil erosion. Concurrently, the study utilized the geographic detector method to explore the principal factors influencing the soil erosion modulus and the dynamics of their interplay. Key insights from this research include: The mean soil erosion modulus in Inner Mongolia’s desert grasslands was quantified at 1580.61 t km− 2a− 1. A spatial analysis revealed pronounced erosion in the northern, eastern, and western areas, while the southern and central regions experienced comparatively milder erosion. Critical factors governing soil wind erosion in this locale encompassed SOM, VH, annual average PPT, potential evapotranspiration, and annual mean air temperature. Interactions among these elements further amplified their collective impact on wind erosion. Given the heightened vulnerability to wind erosion in areas characterized by reduced SOM and lower VH, this study advocates for concentrated erosion mitigation efforts in these zones. Strategies should include the judicious management of grazing activities to foster carbon dynamics within the soil-vegetation nexus, enhance soil carbon sequestration, and bolster both SOM content and vegetation stature. These measures are integral to mitigating the wind erosion challenges confronting these regions.