Attribution and Risk Assessment of Wind Erosion in the Aral Sea Regions Using Multi-Source Remote Sensing and RWEQ on GEE

Abstract

The rapid desiccation of the Aral Sea has transformed the region into one of the world’s most severe soil wind-erosion hotspots. Despite growing concern, long-term, high-resolution assessments and driver attribution remain insufficient. This study integrates the Revised Wind Erosion Equation (RWEQ) with multi-source remote sensing data on the Google Earth Engine (GEE) platform to simulate wind erosion dynamics from 1990 to 2020. The residual trend method was used to disentangle the contributions of climate change and human activities, while erosion risk was assessed using the Information Quantity model and Analytic Hierarchy Process (AHP). This study reveals five key findings: (1) wind erosion increased significantly after 2011, peaking in 2015 with an annual growth rate of 2.418 kg/m². (2) The Aral Sea Basin’s relative contribution to regional erosion declined sharply, indicating a shift in dominant erosion zones to peripheral deserts. (3) Climate change emerged as the primary driver, contributing 70.19% overall, and up to 92.13% in recent years, while human activities showed a peak influence (55.53%) in 2005. (4) Spatial attribution showed climate dominance in desert areas and localized human impact in exposed lakebeds. (5) High-risk erosion zones expanded rapidly into the Kyzylkum Desert after 2010, due to rising wind speeds and vegetation loss. This study provides a robust remote sensing–based framework for wind erosion monitoring and attribution, offering critical insights for erosion mitigation and ecological restoration in arid, climate-sensitive regions.

1. Introduction

With the continuous advancement of industrialization and urbanization, ecological and environmental security issues have become increasingly prominent, especially in arid regions. Central Asia, located in the inland arid zone of Eurasia, exerts a crucial influence on worldwide ecological stability and atmospheric regulation. However, it is also one of the most ecologically vulnerable regions globally, characterized by severe land degradation and frequent dust storms [1,2]. The Aral Sea is a closed inland lake in Central Asia maintaining the ecological balance and stability of the region [3]. Over the past six decades, the Aral Sea has experienced dramatic shrinkage, with its surface area reduced by 88% and water volume decreased by over 92% [4]. The lake’s drying has exposed vast areas of fine-grained sediments, contributing to one of the largest sources of dust emissions in the world [5,6]. The exposed lakebed of the Aral Sea has become a significant source of dust storms, with dust emission rates approximately ten times higher than those in surrounding areas. Recent studies indicate that the emission of salt-rich dust from the Aral Sea contributes significantly to air pollution and human health risks in downwind regions [7,8]. Dust storms originating from the Aral Sea Basin have been linked to significant reductions in crop yields and soil degradation, with long-term consequences for food security and the expansion of desertification [9]. Moreover, delicate particulate matter transported over long distances contributes to air pollution and respiratory diseases [10,11]. Therefore, understanding and assessing the space–time variability of aeolian soil erosion processes in the Aral Sea Basin and its surrounding regions is critical for developing effective land management and ecological restoration strategies.

Similar desiccation-induced wind erosion patterns have been documented globally. In the Mojave Desert, USA, field measurements of playa surfaces demonstrate that dry playas can emit dust at fluxes reaching 120 μg m⁻² s⁻¹ under erosive wind conditions, while wet or crusted playa surfaces exhibit significantly lower emissions below 15 μg m⁻² s⁻¹ [12]. The Aral Sea exhibits unique dynamics due to its hypersaline sediments—surface salt content reaches 20–40% [5], driving emission rates 10-fold higher than surrounding deserts [7]. Persistent northeasterlies further differentiate its erosion regime from these analogues [13].

Substantial research has been conducted on wind erosion mechanisms in the Aral Sea region. Micklin (2007) identified three key factors driving erosion after the lake dried up: exposed fine sediments, reduced soil moisture, and vegetation loss [3]. Later research by Indoitu et al. (2012) found the dust contains high salt levels (20–40%), harming downwind areas [5]. Studies show wind speeds as low as 5.5 m/s can lift dust here [14], much lower than other dry regions, because salt breaks up the soil. These findings help explain why the Aral Sea has become such a major dust source.

Despite extensive research on wind erosion, most existing models rely on coarse-resolution data, limiting their accuracy in large-scale assessments [15]. Traditional approaches, such as the Wind Erosion Prediction System (WEPS) and the Soil Erosion Estimation Model (SEEM), require extensive calibration, making them less practical for large-scale applications in arid regions [16]. In contrast, the Revised Wind Erosion Equation (RWEQ) effectively integrates climatic, soil, and vegetation factors, and has been widely used for quantifying soil erosion intensity [17,18]. Extensive research has demonstrated that integrating RWEQ with geospatial technologies (remote sensing, GIS, and GPS) significantly enhances the precision of erosion assessments across various regions and spatial scales, particularly in arid and semi-arid environments [19,20,21,22].

Additionally, traditional GIS-based approaches struggle with large-scale, multi-source data integration [23]. To address this, we employ Google Earth Engine (GEE), a cloud-based platform that efficiently processes remote sensing, meteorological, and land cover data at high spatial and temporal resolutions [24]. GEE has been successfully applied in land-cover change analysis, vegetation monitoring, and soil erosion modeling [24,25,26]. Integrating GEE with empirical wind erosion models, for instance, the widely-used Revised Wind Erosion Equation (RWEQ), enables the efficient analysis of large-scale wind processes [21,22,26], without altering the fundamental wind erosion modeling methodology.

The process of soil wind erosion involves complex interactions between natural drivers and anthropogenic influences. Quantifying the contributions of these factors to erosion remains a significant challenge. Traditional approaches often fail to distinguish the impacts of climate change and human activities accurately. In recent years, efforts to disentangle the effects of climate variability and human activities on land degradation have led to the development of more refined attribution methods. Among them, the residual trend (RESTREND) approach, first proposed by Evans and Geerken [26], utilizes satellite-derived NDVI and climate variables to isolate vegetation changes not explained by climatic variability. This method has since been refined and widely adopted for detecting anthropogenic degradation signals under natural fluctuations [27,28]. RESTREND has proven particularly effective in arid and semi-arid regions, including southern Africa and Central Asia, where it has been applied to assess the relative influence of land-use change, irrigation reduction, and climate factors on vegetation cover and dust activity [15,29,30].

Meanwhile, erosion risk assessment has evolved toward the integration of multi-criteria decision-making models and spatial data analysis. The Analytic Hierarchy Process (AHP) and information value models are widely used to assign weights to variables like slope, vegetation, soil, and rainfall, improving erosion susceptibility mapping [31,32]. Applied within GIS environments, they offer reliable assessments, especially in complex or data-limited areas. Case studies in Mediterranean and North African regions show these methods effectively identify high-risk zones by combining key environmental layers [33]. Additionally, multi-resolution remote sensing enhances temporal monitoring, supporting dynamic erosion and degradation analysis [34]. These tools establish a solid foundation for the integrated attribution and risk assessment approach used in this study.

Building on these advances, this study aims to comprehensively assess soil wind erosion in the Aral Sea Basin by integrating remote sensing data, meteorological reanalysis data, and ground observations. We employ the Revised Wind Erosion Equation (RWEQ) model to simulate the spatiotemporal dynamics of erosion, and use the residual trend method to separate the contributions of climate change and human activities. Additionally, we develop an erosion risk assessment framework using the Information Quantity model and Analytic Hierarchy Process (AHP) to identify high-risk areas and inform regional management strategies.

2. Materials and Methods

2.1. Study Area

The Aral Sea Basin, situated at the border between Kazakhstan and Uzbekistan, is a central feature of the region (Figure 1). Geographically, it lies between Kazakhstan and Uzbekistan, within the heart of the Kyzylkum Desert. The Kyzylkum Desert borders the basin to the east, the Syr Darya and Amu Darya deltas to the northeast and south, respectively, and the Betpak-Dala Desert to the north. The western coastline is defined by the Ustyurt Plateau, which has an average elevation of 150 m and is primarily composed of rocky badlands.

The formation of the Aral Sea dates back to the late Pliocene when tectonic subsidence created a depression that accumulated water [35]. Since the Pleistocene, the inflows of the Syr Darya and Amu Darya rivers have stabilized the Aral Sea’s water level. However, since the 1960s, the extensive diversion of water from these rivers for agricultural and industrial purposes, combined with persistent arid climatic conditions, has led to a dramatic reduction in the lake’s surface area [36]. These changes lead to increased salinity of the lake water and the accumulation of vast amounts of dry salt in the surrounding regions, severely altering regional hydrology, ecosystem balance, and atmospheric dust emissions [37]. Wind erosion in the Aral Sea Basin and its periphery has become a significant source of salt and dust storms, posing one of Central Asia’s most severe threats to human health and agricultural activities [38,39].

2.2. Data Collection and Sources

This study integrated a comprehensive suite of multi-source datasets, including remote sensing products, meteorological reanalysis data, field surveys, and in situ observations, to evaluate aeolian soil erosion dynamics in the Aral Sea Basin over the 1990–2020 period (

Table 1. Data collection and sources.

Data Type	Data Source	Data Content	Spatial Resolution	Time Range
Ground Observation	NOAA	Meteorological station data (wind speed, etc.)	-	1990–2023
Field Sampling	Kazakhstan Aral Sea Dry Lakebed	Soil sampling data (29 sites)	-	June 2018
Dust/Salt Monitoring	Aral Sea Dust Monitoring Stations	Dust/salt deposition data (2000–2005)	-	2000–2005
Remote Sensing	MODIS	Land surface temperature, vegetation cover, aerosol optical depth	1 km	1990–2023
Meteorological Data	ECMWF ERA5-Land	Wind speed, temperature, precipitation, snow depth	1 km	1990–2023
Water Body Data	Global Surface Water	Waterbody changes in the Aral Sea	1 km	1990–2023
Topographic Data	SRTM	Elevation data	30 m	-
Land Cover Data	ESA CCI	Global land-cover data	300 m	1990–2023
Soil Data	HWSD	Soil properties (texture, organic matter, etc.)	1 km	-

).

Meteorological data were primarily derived from the ERA5-Land reanalysis dataset, provided by ECMWF, which offers hourly estimates of wind speed, temperature, and precipitation at a resolution of 0.1° (~10 km). To enhance accuracy, ERA5 wind speed was bias-corrected using daily observations from NOAA ground stations in the study region. The bias correction employed a quantile mapping approach, achieving high consistency (R² > 0.87) between adjusted and observed values.

Remote sensing datasets formed the core of the spatial analysis. MODIS (Moderate Resolution Imaging Spectroradiometer) data were used to extract NDVI, AOD, and LST at a 1 km spatial resolution. Vegetation indices were derived using the MOD13A2 product (NASA/USGS, Washington, DC, USA) and further normalized using z-score transformation to minimize sensor-specific bias. Land cover data were obtained from ESA’s Climate Change Initiative (CCI) at 300 m resolution and aggregated to 1 km using majority resampling. Topography was captured via the Shuttle Radar Topography Mission (SRTM) at 30 m resolution and upscaled to 1 km using mean value resampling to match other inputs. To ensure temporal consistency and interannual comparability, all datasets were reprojected to the WGS 1984 geographic coordinate system and resampled to a uniform 1 km spatial resolution using bilinear interpolation for continuous variables and nearest-neighbor for categorical data. NDVI, AOD, and precipitation were temporally smoothed using a five-year moving average to reduce the impact of short-term anomalies and sensor noise.

Soil property information was extracted from the Harmonized World Soil Database (HWSD) at 1 km resolution, including soil texture, organic matter, and calcium carbonate content, which are key inputs for calculating erodibility factors in the RWEQ model. Soil parameters were verified against 29 field samples with strict tolerances (texture ± 5%, organic matter ± 0.5%). To assess baseline changes in hydrological and erosion conditions, the JRC Global Surface Water dataset was utilized to track spatiotemporal variations in water extent from 1984 to 2020.

Field and ground-based observations were utilized for calibration and validation purposes. Specifically, dust flux and PM₁₀ concentration data from Aral Sea Dust Monitoring Stations (2000–2005) were compared with simulated erosion outputs. Additionally, soil samples from 29 sites collected in 2018 across the former lakebed were analyzed for salinity, texture, and crusting characteristics to validate soil data from reanalysis datasets, with deviations controlled within ±10%.

Quality assurance for remote sensing data incorporated the following: (1) MODIS cloud masking using QA band confidence thresholds >90%; (2) temporal consistency checks removing values beyond ±3σ of 5-year moving averages; and (3) cross-validation between MODIS NDVI and ground observations (R² > 0.82, n = 29).

All preprocessed datasets were ingested into the Google Earth Engine (GEE) platform to implement the Revised Wind Erosion Equation (RWEQ). The cloud-based infrastructure of GEE enabled efficient storage, parallel computation, and temporal stacking of long-term raster time series. Overall, this rigorous preprocessing ensured that input data maintained high spatial fidelity and temporal integrity, enhancing the robustness of wind erosion modeling and facilitating accurate spatial attribution analysis.

2.3. Methodology

2.3.1. Technical Flowchart of This Study

Based on the objective of this research, the research framework is organized into four sequential procedures, as shown in the technical flowchart (Figure 2). First, the Revised Wind Erosion Equation (RWEQ) model was implemented on the Google Earth Engine (GEE) platform to simulate monthly Soil Wind Erosion Potential (SWEP) across the Aral Sea Basin from 1990 to 2020, using multi-source geospatial data including MODIS NDVI, ERA5 wind speed, HWSD soil data, and SRTM elevation. Model reliability was assessed using ground-based dust flux observations, satellite-derived AOD data, and comparison with similar arid regions. Second, the spatiotemporal patterns and long-term trends of SWEP were analyzed, and the residual trend method was applied to separate the impacts of climate change and human activities. Finally, a wind erosion risk index was constructed using the information value model and AHP, enabling the identification of high-risk zones and analysis of their driving factors.

2.3.2. The Revised Wind Erosion Equation (RWEQ) Model

The Revised Wind Erosion Equation (RWEQ) is an empirical model that integrates climate, soil, and vegetation factors to estimate wind erosion intensity and has been widely used for quantifying soil erosion intensity in arid and semi-arid environments [26,27].

In this research, the key input parameters for the RWEQ model were derived from remote sensing and meteorological datasets; the wind erosion rate per unit area was computed as

$$Q_R={\displaystyle \frac{2z}{S^2}}{Q}_{\mathrm{m}\mathrm{a}\mathrm{x}}{\mathrm{e}}^{-{\left({\displaystyle \frac{z}{S}}\right)}^2}$$

(1)

$$S=150.71{\left(WF\times EF\times SCF\times K^{\prime }\times C\right)}^{-0.3711}$$

(2)

$$Q_{\mathrm{m}\mathrm{a}\mathrm{x}}=109.8\left(WF\times EF\times SCF\times K^{\prime }\times C\right)$$

(3)

where Q_R is the soil erosion rate (t/(hm²·a)), Q_max is the maximum soil transfer rate (kg/m), and z is the distance downwind where maximum erosion occurs (m). The model assumes that maximum wind erosion occurs at the midpoint of the field (z ≈ S/2), and S is the plot length (m).

The climate factor (WF) is calculated as

$$WF={\displaystyle \frac{SW\times SD\times {\sum }_{i=1}^N{u}_2{(u_2-u_1)}^2\times \frac{\rho }{\mathrm{g}}\times N_d}{500}}$$

(4)

where SW represents soil moisture, SD is the snow cover factor, u₂ indicates measured wind speed at 2 m (m/s), u₁ is the threshold wind speed at 2 m (assumed 5 m/s), N is the number of wind-speed-observation periods (u₂ > u₁), N_d refers to the number of days, ρ is the air density (kg/m³), and g represents gravitational acceleration (m/s²).

The soil erodibility factor (EF) and soil crust factor (SCF) are given by

$$EF={\displaystyle \frac{29.09+0.31+SA+0.17\times SI+0.33\times \frac{SA}{CL}-2.59\times OM-0.95\times {\mathrm{C}\mathrm{a}\mathrm{C}\mathrm{O}}_3}{100}}$$

(5)

$$SCF={\displaystyle \frac{1}{1+0.0066\times C{L}^2+0.021\times O{M}^2}},$$

(6)

where SA, CL, and SI represent sand, clay, and silt content (%), respectively, OM represents soil organic matter content (%), and CaCO₃ denotes the calcium carbonate content (%).

The vegetation cover factor (C) is computed for five land cover types (forest, shrubland, grassland, cropland, and bare land,) using

$$C={\mathrm{e}}^{a_i\left(SC\right)}$$

(7)

where a_i is the vegetation-specific coefficient, and SC is the vegetation cover fraction derived from NDVI. The vegetation-specific coefficients a_i were adopted from Fryrear [17] for arid regions, with values of 0.1535 (forest), 0.0921 (shrubland), 0.1511 (grassland), 0.0438 (cropland), and 0.0768 (bare land), which are consistent with the land cover types in our study area.

While the empirical models provided practical estimates of erosion, it is worth noting the uncertainties associated with the model parameters. First, the fixed wind threshold (u₁) may underestimate erosion in crusted lakebed areas where thresholds could be higher. Second, vegetation coefficients (a_i) from arid regions may not fully capture seasonal variations in cover effectiveness. Future studies could employ Bayesian calibration or ensemble modeling to quantify these uncertainties better.

2.3.3. Application of the Google Earth Engine (GEE) Platform

The adoption of Google Earth Engine (GEE) in this study was predicated on its distinctive computational architecture that addresses three fundamental challenges in large-scale wind erosion modeling: first, its cloud-based parallel processing capability facilitates efficient analysis of three-decade (1990–2020) time-series data across the extensive Aral Sea region; second, the platform provides direct access to, and seamless integration of, multi-source geospatial datasets, including MODIS surface reflectance products, ERA5-Land meteorological reanalysis, and SRTM elevation data; third, its optimized spatial analysis functions enable pixel-level computation at 1 km resolution, while maintaining computational feasibility for regional-scale assessments. This integrated GEE framework demonstrated exceptional computational efficiency, completing the 30-year simulation analysis in just one day, while maintaining rigorous analytical standards. Such processing capacity is particularly crucial for environmental modeling studies.

The computational framework consisted of four main steps. Firstly, in data preprocessing, meteorological data (ERA5, NOAA) and land cover data (MODIS, ESA CCI) were resampled to 1 km resolution for spatial consistency, and wind speed data were corrected using bias-adjustment techniques to improve accuracy. Secondly, the RWEQ model equations were applied to calculate the spatio-temporal distribution of wind erosion rates from 1990 to 2020. Thirdly, simulated erosion rates were validated against dust flux observations and satellite-derived Aerosol Optical Depth (AOD) data for validation and calibration. The final risk classification utilized GEE’s optimized Jenks Natural Breaks algorithm, which automatically determined optimal thresholds by statistically analyzing the erosion modulus distribution to minimize within-class variance while maximizing between-class differences, thereby generating distinct and meaningful risk categories for regional assessment.

2.3.4. Residual Trend Method

The Residual Trend Method (RTM) is used to quantify the relative contributions of climate change and human activities to soil wind erosion by isolating climate-induced trends [40,41]. This method assumes that vegetation cover responds predictably to climatic factors, and deviations from this trend indicate the presence of anthropogenic influences.

First, the actual soil erosion modulus (QA) was derived from RWEQ model simulations, while the climate-driven erosion modulus (QW) was estimated using historical meteorological data, keeping 1990 land-use conditions fixed. The human-induced erosion component (QL) was then computed as

$$QL=QA-QW$$

(8)

The relative contributions of climate change (CW) and human activities (CL) were calculated as

$$CL=QL/QA$$

(9)

$$CW=1-CL$$

(10)

All input datasets were standardized to a common spatial resolution (1 km) and temporal resolution (annual scale) to ensure methodological robustness. Land-use data from 1990 was fixed to establish a stable baseline, while meteorological variables and vegetation indices (NDVIs) were updated annually to capture climate variability. A sensitivity test was conducted by adjusting NDVI and precipitation inputs within ±10% to assess method reliability, yielding stable results (R² = 0.91, p < 0.001).

The residual trend method (RTM) effectively partitions climate and human influences, providing a clear framework for understanding their respective roles. But, it assumes linear climate–vegetation relationships; non-linear responses (e.g., drought-induced tipping points) may introduce uncertainty in areas undergoing rapid ecological shifts.

2.3.5. Information Quantity Model and Analytic Hierarchy Process (AHP)

This study integrates the Information Quantity Model and AHP to assess soil wind erosion risk. The Information Quantity Model quantifies the relative contribution of multiple environmental factors to wind erosion, while AHP assigns weights to these factors based on their significance, optimizing risk evaluation accuracy [42,43]. It provides a powerful tool for understanding complex interactions and prioritizing risk factors. This model has been successfully applied across various fields, including geological hazard assessment and environmental risk management. However, its application in soil wind-erosion risk assessment remains underexplored.

Key influencing factors were selected, based on their relevance to wind erosion processes, including elevation, temperature, land cover type, wind speed, vegetation cover, precipitation, soil moisture, and soil type. The AHP weighting procedure followed four key steps: (1) hierarchical structure development with goal, criteria (inducing/controlling factors), and indicator levels (7 environmental factors); (2) pairwise comparison matrix construction using Satty’s 1–9 scale through expert surveys; and (3) weight calculation via eigenvector method with consistency verification; the final weights were computed through consistency ratio (CR) testing, ensuring a CR < 0.1 for reliability.

The soil wind-erosion risk index (I) was calculated as

$$I={\sum }_{i=1}^{n}I\left(X_i,A\right)={\sum }_{i=1}^n{w}_i\mathit{ln}{\displaystyle \frac{N_i/N}{M_i/M}}$$

(11)

where w_i is the weight of the ith influencing factor, and N_i is the total wind erosion amount corresponding to factor X_i. N is the total wind erosion amount in the study area, M_i is the area associated with factor X_i, and M is the total study area.

The Jenks Natural Breaks method, optimized using Goodness of Variance Fit (GVF), was applied to classify wind erosion risk. The resulting five risk levels were Low risk (I < −2.036), Moderate risk (−2.036 < I ≤ 1.758), Moderate–high risk (1.758 < I ≤ 4.315), High risk (4.315 < I ≤ 6.218), and Very high risk (I > 6.218).

This classification provides a quantitative basis for regional wind-erosion prevention strategies, enabling targeted mitigation efforts in high-risk areas, such as the Kyzylkum Desert and the dry lakebeds of the Aral Sea Basin.

2.3.6. Spatial Attribution Analysis

To differentiate the spatial impacts of climate change and human activities on wind erosion, we conducted a spatial attribution analysis using a decomposition approach based on three erosion moduli: the actual erosion modulus (C), the simulated erosion modulus under climate conditions (W), and the residual erosion modulus (L) representing human-induced effects [44,45].

The actual erosion modulus (C) represents the synergistic impacts of climatic variations and land use/cover modifications on soil erosion dynamics. Its temporal trend (CS) reflects their integrated influence, where a positive CS indicates increasing erosion and a negative CS suggests decreasing erosion. The simulated erosion modulus (W) estimates erosion under climate-only conditions, isolating the impact of climatic factors. Its trend (WS) determines whether climate change exacerbates (WS > 0) or mitigates (WS < 0) erosion. The residual erosion modulus (L) quantifies the influence of land-use/cover change, with its trend (LS) indicating whether these changes intensify (LS > 0) or reduce (LS < 0) erosion.

Based on the trends of these variables (CS, WS, and LS), erosion drivers were categorized into six scenarios (

Table 2. Scenarios of Climate and Land Use Change Involvement in Soil Wind Erosion Processes.

Scenario		Dominant Factor
WI	CS > 0, WS > 0, LS < 0	Climate
WD	CS < 0, WS < 0, LS > 0
LD	CS > 0, WS < 0, LS > 0	Land Use/Cover
LI	CS < 0, WS > 0, LS < 0
CI	CS > 0, WS > 0, LS > 0	Combined Climate and Land Use/Cover
CD	CS < 0, WS < 0, LS < 0

).

3. Results

3.1. Spatiotemporal Dynamics of Soil Wind Erosion

Figure 3 shows the interannual spatial distribution of the soil wind erosion modulus in the study area from 1990 to 2020. The Aral Sea Basin, particularly the eastern dried lakebed, is identified as the primary source of erosion, with increasing intensity over time. Since 2012, erosion within the lakebed has declined, while surrounding areas have seen intensified erosion, indicating an expansion of the erosion range and a shift in the primary erosion source, to the lakebed periphery.

Between 1990 and 2020, the soil wind erosion modulus in the study area exhibited significant spatiotemporal variability (Figure 4, Figure 5, Figure 6 and Figure 7), with a spatial variation rate ranging from −1.2 to 5.09 kg/m² and an overall increasing trend of 0.33 kg/m² per year. The highest increases were observed in the central dried Aral Sea Basin, the Kyzylkum Desert, and parts of the western plateau (annual growth rate > 1.8 kg/m²) [46]. In contrast, minimal changes (<0.1 kg/m² per year) occurred in the northern Aral Sea and the delta regions of the Syr Darya and Amu Darya rivers.

Notably, a declining trend in the soil wind erosion modulus was identified in the eastern region of the Aral Sea, with an average decrease of 0.301 kg/m² per year. It corresponds to the areas that had dried up in the Aral Sea Basin before 1990.

Although extreme erosion values (>200 kg/m²) were detected in localized hotspots, their limited spatial extent and rare occurrence suggest they represent short-term, particular erosion events, rather than a regional trend. These extreme values, likely resulting from brief but intense windstorms or transient surface disturbances, do not reflect the general erosion pattern across the study area.

Figure 4 presents the interannual variation in the soil wind erosion modulus in the study area from 1990 to 2020. The erosion modulus exhibited significant interannual fluctuations and an upward trend (red line in the figure), consistent with previous studies on intensified wind erosion in the Aral Sea region [47,48,49]. The maximum erosion modulus was recorded in 2015 at 19.57 kg/m², while the minimum value of 3.65 kg/m² occurred in 2005. Specifically, from 1990 to 2005, the erosion modulus decreased slowly, at a rate of −0.289 kg/(m²·a), resulting in a 55.9% reduction from 8.28 kg/m² in 1990 to 3.65 kg/m² in 2005. In contrast, the period from 2005 to 2020 was marked by a significant increase in erosion modulus at 0.776 kg/(m²·a). Notably, a substantial jump in erosion modulus occurred around 2010, peaking in 2015, with an average annual increase of 2.418 kg/m². This surge correlates with a 9.6% increase in wind speeds, a 17.8% decline in precipitation, and a 30.5% reduction in vegetation cover, which increased soil exposure and susceptibility to erosion [50].

To assess the statistical significance of temporal trends, we applied the non-parametric Mann–Kendall test to annual wind-erosion modulus time series, which confirmed a significant increasing trend after 2011 (p = 0.012, 95% CI [0.024, 0.156]). Pettitt’s test identified 2011 as a critical change point (p = 0.021), corroborating the visual surge in erosion rates observed in Figure 4. The pre-2011 mean modulus (6.69 ± 2.02 kg/m²) differed markedly from the post-2011 mean (16.07 ± 2.25 kg/m²; two-sample t-test, p = 0.008), with a large effect size (Cohen’s d = 4.38, 95% CI [2.77, 5.99]).

Figure 5 demonstrates that the sharp increase in the soil wind erosion modulus after 2011 was closely associated with significant enhancement of average wind speed. During 2011–2015, the mean wind speed in the study area increased by 31%, with wind speed exhibiting a peak correlation with erosion modulus (r = 0.655) from 2006 to 2015. Notably, other factors, such as precipitation, temperature, and land surface conditions, also affect wind erosion by altering soil moisture and vegetation. Future studies will investigate how these factors interact with wind speed, to gain a deeper understanding of regional erosion patterns.

3.2. Validation of the GEE-RWEQ Model

This study uses multiple validation methods to assess the reliability of soil wind erosion simulations in the Aral Sea region. First, horizontal dust flux data from the Aral Sea monitoring station (1990–2005) were used to evaluate simulation accuracy. The results show a strong correlation (r = 0.72, p < 0.05) between the simulated wind erosion modulus and the observed dust flux (Figure 6a), demonstrating that the model effectively captures the temporal variability of wind erosion [51].

AOD data served as a proxy for regional wind erosion intensity. The scatter plot of the annual mean AOD and soil wind erosion modulus (Figure 6b) reveals a highly significant correlation between AOD and the soil wind erosion modulus in terms of their annual variations (r = 0.85, p < 0.001), further confirming the reliability of the simulation results [52].

To further assess the reliability of the RWEQ simulation, this study conducted a comparative analysis with soil wind erosion studies from other arid and semi-arid regions.

Table 3. Comparison of Soil Wind Erosion Modulus in Similar Regions.

Reference	Region	Method	Period	Soil Wind Erosion Modulus [kg/(m2·y)]
				Cropland	Grassland	Forestland	Bare Land
This study	Aral Sea region	RWEQ	1990–2020	0.98	3.33	1.82	11.84
Li et al. [4]	Central Asia	RWEQ	1986–2005	0.47	1.56	0.34	4.51
Wang et al. [35]	Central Asia	RWEQ	2000–2019	0.52	0.88	0.20	10.36
Chi et al. [13]	Arid regions of China	RWEQ	2000–2010	1.77	0.67–2.81	1.60	5.76
Zhang et al. [45]	Inner Mongolia, China	RWEQ	1990–2015	1.13	2.421	0.30	10.20
Hu et al. [37]	Inner Mongolia, China	137CS	2003	7.99	1.81–4.27	NA	NA
Gong et al. [15]	Northern China	RWEQ	2000–2010	0.50	0.87–2.09	0.51	5.02
Lin et al. [42]	Hexi Corridor, China	RWEQ	1982–2015	2.14	4.01	0.95	8.52
Hagen [16]	Arid regions of the United States	WEPS	1989–1997	0–3.98	NA	NA	NA

compares the soil wind erosion modulus worldwide in different semi-arid and arid regions. The values obtained in this research are generally higher than those reported for Inner Mongolia, but comparable to those observed in Central Asia [53,54]. This discrepancy may stem from the distinct climatic conditions and soil textures of the Aral Sea Basin, which are characterized by a higher proportion of fine, salt-rich sediment deposits. Although variations exist in study periods and data sources, the observed consistency in wind erosion patterns across different land cover types supports the reliability of the simulation results [55].

3.3. Contribution of the Aral Sea Basin to Regional Erosion

The Aral Sea Basin’s contribution to regional erosion declined over time. From 1990 to 2011, the average annual erosion in the basin (1.968 × 10⁹ t) was 1.88 times higher than outside the basin (1.049 × 10⁹ t). However, after 2012, the basin’s erosion decreased, and its contribution dropped to 0.43 times that of the surrounding areas [56] (Figure 7).

As shown in Figure 7, the proportion of basin erosion volume in the total erosion volume gradually increased from 1990 to 2011, fluctuating from a 2000 low of 55.8% to a 2006 peak of 76.2%, with an average of 64.4%. This highlights that the Aral Sea Basin was the primary source of soil wind erosion in the study area during this period. However, after 2011, this proportion decreased annually, reaching a minimum of 20.6% in 2018, with an average of 29.2% from 2012 to 2020. This trend suggests that the contribution of the Aral Sea Basin to the total erosion volume has diminished over time, while the area outside the basin has gradually become the dominant source of wind erosion [57].

3.4. Contributions to Climate Change and Human Activities

Table 4. Changes in Soil Wind Erosion Volume Under the Joint Influence of Climate and Land Use/Cover.

Year	Baseline Value	Simulated Value (×109 t)	Change Amount (t)	Climate Change			LUCC
				Change Amount (×109 t)	Contribution Rate	Average Contribution Rate	Change Amount (t)	Contribution Rate	Average Contribution Rate
1995	1.574	1.722	147.72	91.73	62.10%	70.19%	55.99	37.90%	29.81%
2000		2.377	803.44	507.71	63.19%		295.73	36.81%
2005		1.708	134.05	59.61	44.47%		74.44	55.53%
2010		2.370	795.57	534.20	67.15%		261.37	32.85%
2015		9.156	7581.70	7033.11	92.76%		548.59	7.24%
2020		7.278	5703.82	5218.79	91.50%		485.03	8.50%

presents the contributions of climate change and human activities to soil erosion in the Aral Sea region. The combined effects of these two factors have led to an overall increase in soil erosion in the study area. Specifically, the dominance of climate change (70.19% overall) is spatially heterogeneous: high contributions (>80%) coincide with regions of increasing wind speeds, whereas human activities prevail (55.53% in 2005) in locales with intensive irrigation or lakebed exposure. This divergence underscores the need for region-specific mitigation strategies. Simultaneously, this suggests that climate change is a dominant driving factor of soil erosion in the region, which is consistent with findings from those who used a Bayesian model to assess the contributions of climatic factors [58].

Notably, the contribution of climate change has shown an increasing trend over time, while that of human activities has decreased, correspondingly. This trend was particularly pronounced in 2015 and 2020, when the average contribution rate of climate change reached 92.13%. However, an exception was observed in 2005, when human activities contributed 55.53% to soil erosion, surpassing the influence of climate change and becoming the dominant factor [59,60].

3.5. Spatial Attribution of Erosion Drivers

Figure 8 presents a detailed spatial attribution analysis of soil wind erosion under various scenarios influenced by climate change and human activities. The results reveal that the spatial contribution rates to enhanced wind erosion are in the order of Combined Impact (CI, 61.72%) > Wind-driven Impact (WI, 28.98%) > Land-use Impact (LI, 2.72%). This indicates that the combined effects of climate change and human activities play a dominant role in exacerbating wind erosion.

Specifically, regions where the synergistic effects of climate change and human activities lead to enhanced wind erosion are primarily located in the western plateau and eastern desert areas of the Aral Sea. These areas are characterized by sparse vegetation cover, low soil moisture, and high wind velocities, which collectively facilitate the detachment and transport of soil particles. In contrast, areas where enhanced wind erosion is attributed solely to climate change are sporadically distributed in the northern part of the Aral Sea, the Amu Darya, and the Syr Darya deltas. These regions are likely influenced by changes in precipitation patterns and increased evaporation rates, which reduce soil moisture and improve soil susceptibility to wind erosion.

Notably, regions where enhanced wind erosion is driven predominantly by human activities are concentrated in the eastern dried-up lakebed of the Aral Sea, particularly before the 1990s. This is attributed to historical land-use practices such as excessive water diversion for irrigation, which led to the desiccation of the lakebed and the exposure of fine, erodible sediments.

3.6. Soil Erosion Risk Assessment

In assessing soil wind erosion risk, we employed the Weighted Information Quantity Model to calculate the information quantity for each pixel within the study area. The Jenks Natural Breaks method, optimized with the Goodness of Variance Fit (GVF), was utilized to classify the wind erosion information quantity into five risk levels: low risk (I ≤ −2.036), moderate risk (−2.036 < I ≤ 1.758), moderate–high risk (1.758 < I ≤ 4.315), high risk (4.315 < I ≤ 6.218), and very high risk (I > 6.218). This approach yielded monthly-scale soil wind erosion risk maps at a 1 km resolution from 1990 to 2020.

Analysis reveals a consistent spatial distribution of wind erosion risk across the study region throughout the three-decade study period, although interannual variations were notable in some local regions (Figure 9). Areas with very high risk were primarily concentrated in the eastern dried-up lakebed and central islands of the Aral Sea, while high-risk regions were distributed across the dried lakebed and the eastern Kyzylkum Desert, showing an expanding trend. Moderate-to-high-risk areas were primarily located around the Aral Sea lakebed and in the east region, with moderate-risk areas in the western part of the lakebed. Low-risk areas were predominantly located in the Amu Darya and Syr Darya deltas, as well as the northwestern plateau.

From 1990 to 2010, soil wind erosion was predominantly characterized by low and moderate risk levels, accounting for 41.06% of the total area. High- and very high-risk regions were relatively small, occupying only 1.32% of the total area. In contrast, from 2015 to 2020, moderate and moderate–high risk levels became more prevalent, covering 32.29% of the total area. Low- and moderate-risk areas decreased by an average of 38.7% compared to 1990–2010, indicating a shift towards higher-risk categories. High- and very high-risk areas increased significantly, accounting for 4.76% of the total area, which is 2.61 times higher than in the 1990–2010 period.

Low- and moderate-risk areas showed a general decreasing trend, with average annual change rates of −11% and −9%, respectively (Figure 10). The area of low risk was highest in 1990 (44.26%) and lowest in 2000 (23.22%). The moderate-risk area peaked in 2005 (58.78%) and reached its minimum in 2015 (27.23%). In contrast, moderate–high, high, and very high-risk areas exhibited fluctuating increases, with average annual change rates of 21%, 34%, and 40%, respectively. The higher the risk level, the more significant the increase in area. The moderate–high, high, and very high risk reached their maximum in 2015, accounting for 32.14%, 13.16%, and 0.83% of the total area, respectively, while the minimum area was observed in 2005 (10.28%). The area of very high risk was smallest in 2005 (0.09%) and most extensive in 2015 (1.89%).

4. Discussion

This study comprehensively evaluates soil wind erosion in the Aral Sea Basin and its neighboring zones. Our results indicate that wind erosion intensity varied significantly across different regions during the study period, particularly after 2011, with a peak in 2015. This trend is consistent with previous studies on Central Asian desertification and wind erosion dynamics [41,42,43], which also identified an increase in wind-driven erosion following major hydrological and climatic shifts. Our findings align with those of researchers who reported a rise in global wind speeds since 2010, contrasting with the relatively stable conditions observed between 1970 and 2010.

Spatial differences in wind erosion trends are also remarkable. The central dried lakebed of the Aral Sea, the Kyzylkum Desert, and parts of the western plateau experienced significant increases in wind erosion intensity (1.8 kg/m²/year). In contrast, the eastern region of the Aral Sea exhibited a decrease in wind erosion intensity (−0.301 kg/m²/year), which coincided with areas that had dried up before 1990. This spatial variation is primarily attributed to differences in soil texture. The lakebed that dried up before 1990 consists of primitive desert soils with high clay and silt content, which enhances soil stability and vegetation establishment, thereby reducing erosion susceptibility.

In contrast, the lakebed that dried up after 1990 is predominantly composed of loose, highly erodible saline soils, which are more susceptible to wind erosion. These differences have also led to a decline in the Aral Sea Basin’s contribution to overall wind erosion. While the erosion volume within the basin was 1.88 times higher than that of surrounding areas between 1990 and 2011, this ratio dropped to 0.43 after 2012. This indicates that the dominant erosion zones are shifting from within the lake basin to the surrounding regions, a previously unreported finding.

The spatial shift in erosion hotspots from the Aral Sea dried lakebed to surrounding deserts mirrors trends observed in other post-desiccation landscapes, driven by fundamental changes in sediment availability and wind dynamics. As newly exposed lakebeds initially stabilize through crust formation, prolonged wind erosion leads to progressive depletion of fine, transportable sediments (<50 μm) in older exposed areas, reducing their erosion potential. Concurrently, increased wind speeds and lower entrainment thresholds activate previously stable sediments in surrounding desert margins. This dual process of source depletion and peripheral activation is evident in the Lop Nur region and Badain Jaran Desert [61], where central lakebed erosion declined as adjacent desert margins became more active following similar fine sediment loss and wind speed increases of 15–20%. The Sahara Desert shows similar patterns, with shifting erosion zones as sediment depletion in degraded areas activates new dust sources under changing wind conditions [62]. These global comparisons demonstrate that post-desiccation landscapes undergo predictable shifts in erosion governed by these interacting physical processes, reinforcing the broader significance of our findings.

The relative contributions of climate change and human activities to erosion have also been widely debated. Our findings suggest that climate change (70.19%) is the dominant driver, increasing to 92.13% between 2015 and 2020. This aligns with those who used Bayesian modeling to quantify the effects of temperature, precipitation, and wind dynamics on wind erosion in Central Asia [58]. However, some studies highlight a more substantial role for human activities. For example, it argued that land use changes, rather than climate change, were the primary drivers of wind erosion in Inner Mongolia, where overgrazing and cropland expansion significantly degraded surface stability [59]. Our study finds that human activities contributed up to 55.53% of total erosion in 2005, coinciding with rapid lake shrinkage and intensified land use. However, after 2010, climate change surpassed human activities as the primary driver of erosion, highlighting the phased nature of these interactions.

These findings suggest that while climate change is the primary driver of wind erosion in the Aral Sea Basin, the role of human activities varies across regions and periods. Future studies should explore how regional land-use policies and restoration strategies can mitigate wind erosion in post-desiccation landscapes.

The spatial attribution analysis revealed significant regional differences in the impacts of climate change and human activities on wind erosion. Specifically, climate change was the dominant driver of increased wind erosion in the western plateau and eastern desert areas of the Aral Sea, where intensified wind speeds and reduced precipitation—combined with human activities—were the primary causes (CI scenario: 61.72%). In contrast, human activities had a more pronounced impact on the dried lakebed in the eastern region, where changes in land use were closely linked to enhanced wind erosion (LI scenario: 2.72%). This aligns with those who highlighted the fact that human activities lead to vegetation degradation and soil destabilization [59].

The study further elucidated the spatial pattern and evolution of soil wind erosion risk in the Aral Sea region. Results indicated that over the past 30 years, the spatial pattern of wind erosion risk remained relatively stable, yet high-risk areas expanded. The extremely high-risk zones were primarily concentrated in the eastern dried lakebed of the Aral Sea and the Kyzylkum Desert, gradually extending outward from the lake basin. This trend suggests that as the Aral Sea continues to shrink and desertification intensifies, the risk of wind erosion spreads from within the lake basin to surrounding areas, posing a potential threat to regional ecosystems and land-use security [50]. Notably, between 2015 and 2020, the Kyzylkum Desert saw a significant increase in wind erosion risk due to enhanced wind speeds and reduced vegetation cover, highlighting this region as a priority for wind erosion control.

Despite the comprehensive nature of our analysis, several limitations should be acknowledged. Data uncertainties arise from the coarse resolution of MODIS and ERA5, which may not fully capture localized erosion dynamics. In contrast, AOD-based estimates can be influenced by non-dust aerosols, necessitating higher-resolution data and field validation. RWEQ model constraints include oversimplified representations of soil roughness, moisture, and wind thresholds, lacking feedback mechanisms between erosion and vegetation loss [16,17]. Integrating machine learning with physical models could enhance accuracy [2]. Attribution bias in the Residual Trend Method may misrepresent climate–human interactions due to CO₂ fertilization and delayed land-use effects, suggesting that Bayesian modeling could be a potential improvement. Future research should refine data integration, modeling approaches, and validation efforts, to achieve more robust erosion assessments [11,18].

5. Conclusions

This study employed the RWEQ model and the GEE remote sensing platform to analyze the spatiotemporal evolution of soil wind erosion in the Aral Sea Basin and its surrounding regions from 1990 to 2020. The relative contributions of climate change and human activities were systematically quantified, and an evaluation of erosion risk was conducted. The key findings are as follows:

• Wind erosion exhibited an overall increasing trend from 1990 to 2020, with a significant intensification after 2011, peaking in 2015 (an annual increase of 2.418 kg/m²). The contribution of the Aral Sea Basin to regional erosion declined, while surrounding arid areas, particularly the Kyzylkum Desert, emerged as the primary sources of erosion. These findings suggest that erosion control strategies should expand from lakebed restoration to broader ecosystem management in adjacent vulnerable regions.
• Climate change accounted for 70.19% of the total erosion contribution, increasing to 92.13% from 2015 to 2020, primarily driven by rising wind speeds, decreasing precipitation, and increasing temperatures. In contrast, human activities contributed up to 55.53% in 2005, which was associated with rapid lake shrinkage and changes in land use. As the lake recession rate slowed, the impact of human activities declined. Effective erosion control requires strategies that integrate the long-term influence of climate change with the short-term effects of human interventions.
• Spatial attribution analysis revealed significant regional differences in erosion drivers. Climate change was the primary driver in the western plateau and eastern desert regions (61.72%), whereas human activities had a more significant influence on the dried lakebed of the Aral Sea (2.72%). In oasis regions, erosion was jointly driven by climate and land-use changes. These findings underscore the need for region-specific management strategies that integrate ecological restoration with wind-erosion control measures.
• Over the past three decades, extremely high-risk erosion areas have expanded outward from the lakebed, primarily affecting the dried lakebed and the Kyzylkum Desert. After 2015, rising wind speeds and declining vegetation cover exacerbated erosion risks in peripheral areas. This trend underscores the need for a comprehensive erosion mitigation strategy that extends beyond lakebed restoration to broader landscape-scale interventions.

This study offers new insights into the spatiotemporal dynamics and driving mechanisms of wind erosion in the Aral Sea Region, providing scientific support for erosion prevention and ecological restoration in arid regions. Furthermore, the methodological framework proposed in this study can be extended to other desertification-prone areas for monitoring and risk assessment. Future research should integrate multi-source remote sensing data and high-resolution field observations to enhance erosion modeling accuracy and further investigate the long-term impacts of vegetation restoration and land-use changes on wind erosion, thereby supporting sustainable land management in arid ecosystems.