Next Article in Journal
The Multiscale Assessment of Infrastructure Vulnerability to River Floods in Andean Areas: A Case Study of the Chibunga River in the Parish of San Luis, Ecuador
Next Article in Special Issue
Fracture Process in Conceptual Numerical Geological Rock Mass System Model and Its Implications for Landslide Monitoring and Early Warning
Previous Article in Journal
Tracing the Evolutionary and Migration Pathways of Economically Important Turkish Vicia L. Species: A Molecular and Biogeographic Perspective on Sustainable Agro-Biodiversity
Previous Article in Special Issue
Hydro-Meteorological Landslide Inventory for Sustainable Urban Management in a Coastal Region of Brazil
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 17
Issue 17
10.3390/su17177913
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

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

Index-Driven Soil Loss Mapping Across Environmental Scenarios: Insights from a Remote Sensing Approach

by
Nehir Uyar
Department of Architecture and Urban Planning, Zonguldak Vocational School, Zonguldak Bülent Ecevit University, 67600 Zonguldak, Turkey
Sustainability 2025, 17(17), 7913; https://doi.org/10.3390/su17177913
Submission received: 3 August 2025 / Revised: 26 August 2025 / Accepted: 29 August 2025 / Published: 3 September 2025
(This article belongs to the Special Issue Landslide Hazards and Soil Erosion)
Browse Figures
Versions Notes

Abstract

Soil erosion is a critical environmental issue that leads to land degradation, reduced agricultural productivity, and ecological imbalance. This study aims to assess soil loss under various land surface conditions by developing 11 distinct scenarios using the RUSLE (Revised Universal Soil Loss Equation) model integrated within the Google Earth Engine (GEE) platform. Remote sensing-derived indices including NDVI, EVI, NDWI, SAVI, and BSI were incorporated to represent vegetation cover, moisture, and bare/built-up surfaces. The K, LS, P, and R factors were held constant, allowing the C factor to vary based on each index, simulating real-world landscape differences. Soil loss maps were generated for each scenario, and spatial variability was analyzed using bubble charts, bar graphs, and C-map visualizations. The results show that vegetation-based indices such as NDVI and EVI lead to significantly lower soil loss estimations, while indices associated with built-up or bare surfaces like BSI predict higher erosion risks. These findings highlight the strong relationship between land cover characteristics and erosion intensity. This study demonstrates the utility of integrating satellite-based indices into erosion modeling and provides a scenario-based framework for supporting land management and soil conservation practices. The proposed approach can aid policymakers and land managers in prioritizing conservation efforts and mitigating erosion risk. Moreover, maintaining and enhancing vegetative cover is emphasized as a key strategy for promoting sustainable land use and long-term ecological resilience.

1. Introduction

Soil erosion by water remains one of the most critical land degradation processes affecting ecosystems globally. While physically based models are often employed to simulate erosion events over short timescales and small catchments [1,2], broader spatial assessments typically rely on semi-empirical models calibrated at plot or field scales. These models are crucial for estimating long-term erosion risk across regional to continental extents [3,4,5]. Erosion modeling is inherently complex, as it requires integrating diverse drivers, including climate variability, land use dynamics, topographic gradients, and anthropogenic pressures [6]. Furthermore, the physical and chemical properties of topsoil layers—such as texture, organic matter content, and pH—play a pivotal role in determining soil susceptibility to erosion [7]. The degradation of topsoil not only reduces agricultural productivity but also impairs vital ecosystem services such as water retention, nutrient cycling, and biodiversity support [8,9].
Globally, soil erosion has emerged as one of the most pressing environmental challenges, contributing to severe land degradation, loss of soil fertility, and increased sedimentation in water bodies. The Mahi River Basin in western India, characterized by seasonal monsoon patterns, variable topography, and expanding anthropogenic activities, is especially vulnerable to erosion. With climate variability and land use change intensifying, there is an urgent need for high-resolution, spatially explicit models to predict soil loss and guide sustainable land management. The Revised Universal Soil Loss Equation (RUSLE) is widely used for such purposes, yet its performance can be significantly enhanced when integrated with remote sensing indices that represent vegetation cover, soil moisture, and surface reflectance conditions.
From a global perspective, the most significant contributor to soil degradation is water-induced erosion, which is intensifying under the dual pressures of climate change and unsustainable land management [10,11,12,13]. This phenomenon is particularly alarming in semi-arid and mountainous regions, where soil formation rates are often outpaced by erosion rates, leading to irreversible losses of soil capital. In this context, integrative and spatially explicit models are required to more accurately assess erosion risk and to inform sustainable land use planning strategies. Recent advancements in remote sensing and geospatial technologies have facilitated the inclusion of vegetation indices (e.g., NDVI, NDWI, NDBI), albedo, and terrain-derived parameters into erosion models such as RUSLE, thereby enhancing both their spatial resolution and predictive performance [6,14,15]. These tools now provide powerful opportunities to monitor erosion processes and design targeted land management interventions across vulnerable landscapes.
Recent studies have demonstrated the utility of spectral indices such as the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Bare Soil Index (BSI) in estimating soil loss patterns. However, their effectiveness often varies depending on sensor type, temporal resolution, and landscape context. Moreover, few studies have systematically compared multiple indices derived from different satellite platforms within a scenario-based modeling framework. To address this research gap, the present study develops 11 scenarios using Sentinel-1, Sentinel-2, and Landsat 8 datasets, each integrating a distinct spectral index with RUSLE parameters through the Google Earth Engine (GEE) platform. By leveraging GEE’s cloud-based processing capacity, this study aims to evaluate the influence of different indices on soil loss prediction and to identify the most reliable configurations for erosion assessment in the Mahi Basin.

2. Materials and Methods

2.1. Study Area

This study was conducted in the Mahi River Basin in western India, covering parts of Gujarat, Madhya Pradesh, and Rajasthan. The basin has an area of about 35,000 km2, with elevations ranging from lowlands to hilly areas in the upper reaches. The climate is mostly semi-arid, with rainfall concentrated during the monsoon season (June–September).
The landscape includes forests in the upper basin, agricultural lands and scrublands in the middle, and urban areas in the lower basin. These natural and human-influenced conditions make the basin suitable for studying soil erosion. The basin boundary was obtained from the HydroSHEDS dataset, and analyses were carried out using a consistent coordinate system. A map of the study area is shown in Figure 1. The study area corresponds to the Mahi River Basin, located in western India between approximately 22° N–25° N latitude and 73° E–75° E longitude. The basin boundary was extracted from the HydroSHEDS database and visualized in Figure 1 to illustrate its geographic location in the global context. Only the outer boundary of the basin is displayed to highlight its spatial extent.

2.2. Workflow of the Study

This study employed the Revised Universal Soil Loss Equation (RUSLE) integrated with remote sensing and GIS techniques within the Google Earth Engine (GEE) cloud platform to estimate soil erosion under eleven distinct input scenarios. Each scenario was designed by combining static biophysical parameters with different vegetation or surface indices derived from various satellite sensors (Sentinel-2, Landsat 8, Sentinel-1), enabling a comparative evaluation of input sensitivity on soil loss estimation. Figure 2 illustrates the workflow of the study. The input datasets used for RUSLE factor estimation were derived from multiple open-source platforms, combining rainfall, soil, topography, vegetation, and land use information. Each factor was selected to ensure both spatial consistency and methodological reliability across the study area (Table 1).
In the first phase, the study area, the Mahi River Basin, was delineated using the HydroSHEDS basin boundary dataset. The temporal window was defined as 1 January 2018 to 1 January 2019. All inputs were harmonized spatially and temporally for this period.
The second phase involved the calculation of RUSLE factor:
  • R factor (rainfall erosivity) was computed using CHIRPS pentadal precipitation data, aggregated annually and adjusted through an empirical equation.
  • K factor (soil erodibility) was derived from the OpenLandMap soil texture classification using a lookup table-based logic reflecting USDA soil taxonomy.
  • LS factor (slope length and steepness) was calculated using the SRTM DEM and the modified Desmet and Govers approach, which considers percent slope and a fixed slope length.
  • C factor (cover management) was the primary variable across the scenarios. In each of the 11 scenarios, a different index was applied depending on the satellite data source. Vegetation indices such as NDVI, EVI, SAVI, NDWI, and BSI were derived from Landsat 8 and Sentinel-2. Sentinel-1 SAR data supported the Radar Vegetation Index (RVI). Each index was then converted into a normalized C factor through logarithmic or exponential scaling techniques.
  • P factor (support practices) was computed using Dynamic World LULC data, cross-referenced with slope classes to assign conservation values, particularly for croplands. Non-agricultural and natural covers were assigned constant values based on standard RUSLE assumptions.
In the third phase, the final soil loss for each scenario was estimated by integrating all RUSLE factors through pixel-wise multiplication. Each of the 11 scenarios (S1–S11) followed the same procedure with only the C factor varying.
In the final stage, the output maps were categorized into five soil loss severity classes. Area statistics were computed using pixel counting, and mean soil loss values were extracted for the basin and sub-basins using zonal statistics. Additionally, pie charts were generated to visualize the areal distribution of erosion severity for comparative analysis.
This modular and replicable workflow enabled a comprehensive assessment of how the choice of vegetation index and satellite platform affects RUSLE-based soil erosion modeling, offering valuable insights for remote sensing-based conservation planning.

2.3. Indices Calculation

The C factor, which represents cover management, was derived from remote sensing indices as follows:
  • NDVI, EVI, NDWI, SAVI, BSI from Sentinel-2 and Landsat 8;
  • RVI from Sentinel-1.
The C factor, which reflects the effect of vegetation and land cover on erosion, was calculated using multiple remote sensing indices derived from Sentinel-1, Sentinel-2, and Landsat 8. These indices provide complementary information on vegetation cover, soil moisture, and bare soil conditions, allowing for a more robust representation of surface dynamics in erosion modeling. The equations and references for the indices applied in this study are presented in Table 2.

2.3.1. Normalized Difference Vegetation Index (NDVI)

The Normalized Difference Vegetation Index (NDVI) has become one of the most established and widely applied remote sensing indices in fields such as forestry, ecology, and agriculture. Its primary strength lies in its ability to provide a simple yet effective means of assessing vegetation health by exploiting the contrast between red and near-infrared reflectance—an approach that reflects the physiological state of vegetation at the canopy level [22]. Since its initial development [23,24], NDVI has been extensively utilized to monitor vegetative biomass and detect areas under stress due to environmental or anthropogenic factors. Alongside NDVI, the Enhanced Vegetation Index (EVI) has gained prominence in recent years, particularly for its improved sensitivity in regions with dense canopy cover or where atmospheric distortions might affect reflectance values. Both indices are now central tools in vegetation monitoring and health assessment across a variety of biomes and temporal scales [25,26]. Their widespread use underscores the growing importance of satellite-based vegetation indices in understanding ecosystem dynamics under changing environmental conditions.

2.3.2. Enhanced Vegetation Index (EVI)

The Enhanced Vegetation Index (EVI) was developed as an improvement over NDVI, particularly to address limitations related to soil background, atmospheric conditions, and canopy saturation in areas of dense vegetation [27]. By incorporating additional correction factors, EVI enhances the accuracy of vegetation monitoring in regions where NDVI tends to lose sensitivity, especially in high biomass zones or under variable atmospheric influences [25,26]. The index produces values ranging from 0 to 1, where values approaching 1 indicate healthy and dense vegetation, while those closer to 0 suggest degraded or sparse vegetative cover. EVI’s robustness against external noise makes it a valuable tool for assessing subtle variations in plant health and productivity across diverse ecosystems. While both NDVI and EVI are extensively used in vegetation studies, EVI is particularly advantageous in complex environments where background signals may otherwise obscure the true vegetation signal.

2.3.3. Soil-Adjusted Vegetation Index (SAVI)

The Soil-Adjusted Vegetation Index (SAVI) was introduced to enhance vegetation monitoring in areas where vegetation is sparse and soil reflectance significantly influences spectral readings. Particularly useful in arid and semi-arid regions, SAVI incorporates a soil brightness correction factor to reduce the background noise caused by exposed soil surfaces [19]. By utilizing reflectance values from the red and near-infrared (NIR) bands—similar to NDVI—SAVI introduces an adjustment parameter (commonly L = 0.5) to account for variations in soil brightness. This makes it especially effective in low-vegetation environments, where traditional indices like NDVI may misrepresent actual vegetation cover due to high soil exposure. Subsequent studies have highlighted SAVI’s improved reliability under such conditions, emphasizing its role in minimizing red-NIR spectral distortions that stem from soil influence [28]. Overall, SAVI provides a refined approach for evaluating vegetation health and distribution in challenging landscapes.

2.3.4. Normalized Difference Water Index (NDWI)

The Normalized Difference Water Index (NDWI), originally developed by [14], is a widely recognized remote sensing index designed to detect changes in the water content of vegetation. It operates by analyzing the difference in reflectance between the near-infrared (NIR) and shortwave infrared (SWIR) bands, which are particularly sensitive to variations in leaf water content [14,15]. This index has demonstrated its utility at multiple spatial scales—from single-leaf measurements that reveal physiological responses under stress conditions [29,30], to broader satellite-based applications that monitor ecosystem-scale moisture dynamics [31]. Because of its strong sensitivity to internal leaf water status, NDWI is frequently employed in drought monitoring, vegetation health assessment, and agricultural water stress analysis. Its capability to reflect subtle shifts in canopy moisture makes it a vital tool in climate-resilient vegetation monitoring strategies.

2.3.5. Bare Soil Index (BSI)

The Bare Soil Index (BSI) is a remote sensing-based spectral index that integrates reflectance values from the blue, red, near-infrared (NIR), and shortwave infrared (SWIR) bands to effectively capture surface soil characteristics. While the SWIR and red bands are particularly sensitive to soil mineral content and surface brightness, the blue and NIR bands help to suppress the influence of vegetation, thus enhancing the contrast between bare soil and vegetated areas [32,33]. This combination allows BSI to serve as a reliable indicator for distinguishing bare soil from other land cover types, especially in heterogeneous landscapes. It also demonstrated the utility of BSI in land classification tasks, highlighting its effectiveness in isolating non-vegetated areas [32]. Building upon this foundation, [34] employed BSI in the construction of the Barest-Pixel Composite (BPC), a method that aims to identify and extract the most exposed soil pixels over time. Such applications underscore the relevance of BSI in land surface monitoring, particularly in studies focusing on land degradation, desertification, and agricultural field dynamics.

2.3.6. Ratio Vegetation Index (RVI)

The growing capabilities of remote sensing technologies have significantly advanced our ability to monitor vegetation structure and function. Among the numerous vegetation indices developed, the Normalized Difference Vegetation Index (NDVI) and the Ratio Vegetation Index (RVI) stand out as two of the most commonly utilized metrics for assessing vegetation density and growth dynamics. NDVI has gained widespread popularity due to its effectiveness in capturing vegetation patterns across broad spatial and temporal scales, offering a consistent measure of plant health and cover [35]. However, its performance may decline in regions with very dense vegetation, where spectral saturation reduces its sensitivity. In such cases, the Ratio Vegetation Index (RVI), calculated as the simple ratio of NIR to red reflectance, provides a valuable alternative. RVI has shown improved sensitivity in high-biomass environments, making it more reliable for monitoring vigorous plant growth and estimating biomass where NDVI may plateau [36]. As such, RVI complements NDVI by extending vegetation monitoring capacity in ecologically productive or heavily forested areas.

2.4. Calculation of RUSLE Factors

The Revised Universal Soil Loss Equation (RUSLE) is one of the most widely adopted empirical models used to estimate long-term average annual soil erosion by water. It evaluates soil loss by integrating five key factors: rainfall erosivity (R), soil erodibility (K), slope length and steepness (LS), crop management (C), and conservation practices (P) [37]. Despite its origin in the agricultural landscapes of the United States, RUSLE’s conceptual simplicity, empirical formulation, and adaptability have enabled its widespread application across diverse geographical and climatic settings [38,39]. The model relies heavily on extensive experimental data, primarily derived from standardized plots measuring 22 m in length and 1.8 m in width, with a 9% slope under bare soil conditions, where tillage is conducted in the upslope–downslope direction [40,41]. This controlled experimental basis lends robustness to the RUSLE factors, but also highlights the importance of local calibration when applied in regions with differing topographic, climatic, and land use characteristics. As such, while RUSLE remains a foundational tool in soil erosion research, its effective implementation requires thoughtful adaptation to site-specific environmental conditions. The RUSLE model was implemented by combining rainfall erosivity (R), soil erodibility (K), topographic factor (LS), cover and management factor (C), and support practice factor (P). Each factor was derived using established equations and datasets, ensuring consistency with previous applications of the model. For instance, the C factor was calculated from vegetation indices such as NDVI using an exponential function relating vegetation density to soil protection. The equations and references used for factor derivation are summarized in Table 3.
In the RUSLE, soil loss estimation is conceptualized through a combination of interacting environmental and land management variables. Specifically, A denotes the predicted average annual soil loss per unit area (expressed in tons per hectare per year), serving as the primary output of the model. The R factor reflects the erosive power of rainfall, capturing the intensity and frequency of precipitation events that contribute to surface runoff (measured in MJ mm ha−1 h−1 year−1). Meanwhile, K represents the soil erodibility coefficient, which quantifies the susceptibility of soil particles to detachment and transport under rainfall impact and runoff conditions, taking into account the physical and chemical properties of the topsoil.
Additionally, LS combines slope length and steepness into a single topographic factor, reflecting the influence of terrain characteristics on erosion potential. The C factor accounts for vegetation cover and land management practices, providing insight into how different land uses and crop rotations reduce or exacerbate erosion risk. Finally, the P factor evaluates the effectiveness of erosion control measures such as contour farming, terracing, or strip cropping. Altogether, these parameters allow RUSLE to simulate soil loss across diverse landscapes by integrating biophysical processes with anthropogenic land use decisions, making it a valuable tool for both research and applied land management.

2.4.1. Rainfall–Runoff Erosivity Factor (R)

The rainfall–runoff erosivity factor (R) captures the influence of rainfall characteristics on soil erosion, particularly the energy exerted by raindrop impact and the subsequent force of surface runoff. As one of the key components of the RUSLE model, the R factor reflects the climatic aggressiveness that drives the detachment and transportation of soil particles [43]. High-intensity and prolonged rainfall events are especially effective in initiating splash and sheet erosion, where the kinetic energy of falling raindrops dislodges soil particles from unprotected surfaces. These detached particles are then easily carried downslope by surface runoff, contributing to sediment yield and land degradation [44]. The standard method for calculating the R factor involves combining the kinetic energy of individual rainfall events with the maximum 30 min rainfall intensity, providing a measure of erosive potential [35]. Understanding and accurately estimating the R factor are crucial for predicting soil loss, particularly in regions prone to intense or irregular precipitation patterns under changing climatic conditions. R is the rainfall erosivity in MJ mm ha−1 h−1 year−1, Pm is the monthly precipitation (mm), and Pa is the yearly precipitation (mm).

2.4.2. Soil Erodibility Factor (K)

The soil erodibility factor (K) represents the intrinsic susceptibility of soil to erosion processes, particularly under the impact of rainfall and surface runoff. This factor is fundamentally linked to the physical and chemical properties of the soil, including its texture (sand, silt, and clay content), organic matter percentage, structure, and permeability [45,46]. Soils with high silt and low organic matter content tend to be more prone to erosion due to their poor structural stability and limited cohesion. Consequently, the K factor quantifies this vulnerability, with values typically ranging between 0 (resistant) and 1 (highly erodible), where higher values indicate greater erodibility [47]. In practical applications, such as in the RUSLE framework, K is often calculated using empirical equations that incorporate soil texture and organic carbon data. For instance, in the case of the Rupnagar district, soil data were derived from the FAO’s digital soil map, and the K factor was estimated using the equation developed by [48], as part of the EPIC (Erosion Productivity Impact Calculator) model. The equation utilizes parameters such as sand (Sd), silt (Si), clay (Ci), and organic carbon (C) content to compute soil erodibility (Ke), expressed as a function of potential soil loss per unit rainfall energy. Such calculations are vital for spatially explicit erosion risk mapping, particularly in heterogeneous landscapes with variable soil properties.

2.4.3. Slope Length and Steepness Factor (LS)

Topography plays a pivotal role in controlling soil erosion, as slope characteristics directly influence the detachment, transport, and deposition of sediment. In the RUSLE framework, the combined effects of slope length and steepness are captured by the LS factor, which quantifies the topographic contribution to rill and inter-rill erosion [49]. The LS factor increases with longer slope lengths and steeper gradients, both of which enhance the velocity and erosive power of surface runoff. Traditionally, soil loss estimates were standardized based on experimental plots with a 22 m slope length and a 9% gradient [50,51], serving as a baseline from which site-specific topographic adjustments are derived. For spatial modeling applications, particularly in GIS environments, LS is often calculated using flow accumulation and slope raster data. In this study, the LS factor was computed using the equation proposed by [51] where Facc represents the flow accumulation matrix, M refers to the cell size (12.5 × 12.5 m), and S denotes slope in percent. This approach enables detailed topographic characterization across heterogeneous terrains and supports the generation of high-resolution soil erosion risk maps.

2.4.4. Cover Management Factor (C)

Land cover and agricultural practices exert a fundamental influence on soil erosion processes. In this context, the cover management factor (C) serves as a critical variable in determining the extent of soil loss. The C factor represents the protective capability of existing land use and vegetation cover, measuring the effectiveness of conservation practices relative to conventional tillage methods [52]. As a dimensionless coefficient in the Revised Universal Soil Loss Equation (RUSLE), it integrates the interaction between vegetative canopy, ground cover, and soil disturbance frequency, offering a proxy for erosion susceptibility under different land management scenarios.
The C value typically ranges between 0 and 1, where values near 0 correspond to dense vegetation and well-managed conservation systems, while values approaching 1 indicate bare soil surfaces that are highly vulnerable to erosion [40]. For instance, forested regions often exhibit C values as low as 0.001 due to their high interception and root reinforcement, whereas barren lands or fallow agricultural fields may reach values close to 1.0, reflecting minimal surface protection [44]. Thus, beyond merely describing land cover, the C factor plays a strategic role in evaluating the sustainability of agricultural practices and guiding erosion control efforts. Its dynamic nature is closely linked to vegetation density and the intensity of land disturbance, making it an essential component for both predictive modeling and landscape management.

2.4.5. Conservation Support Practice Factor (P)

The conservation support practice factor (P) represents a critical parameter in adjusting soil loss estimates based on differences in slope direction and agricultural techniques [53]. Incorporated into the RUSLE model, this factor captures the effectiveness of mechanical soil conservation practices such as contour farming, terracing, and strip cropping in reducing erosion risk. The P value ranges from 0 to 1, where values closer to 0 indicate the implementation of effective conservation measures, while values near 1 suggest the absence of any erosion control strategies [40]. Research has shown that appropriate soil conservation practices, particularly on sloped terrains, can significantly minimize soil loss by reducing surface runoff [1,52].

2.5. Google Earth Engine (GEE)

Google Earth Engine was used for preprocessing, index calculation, and soil loss map generation. SRTM data were used to compute the LS factor, while FAO soil maps provided K-factor estimates. The outputs include soil loss maps (tons/ha/year), bar graphs, and bubble charts for statistical visualization, and C-map matrices showing inter-index relationships. All preprocessing, coding, and visualization steps were conducted in GEE using JavaScript API.

3. Results

3.1. Soil Loss Patterns

Soil loss values exhibited considerable variation across all scenarios, with higher values observed in areas characterized by steep slopes, sparse vegetation, or bare soil. Scenarios incorporating vegetation indices such as NDVI (S1, S6) and EVI (S2, S7) consistently resulted in the lowest mean soil loss values, highlighting the protective role of vegetation in reducing erosion. In contrast, scenarios utilizing the Bare Soil Index (BSI; S3, S8) produced the highest soil loss estimates, particularly in degraded or urbanizing regions where soil exposure was extensive. NDWI (S4, S9) and SAVI (S5, S10) demonstrated intermediate patterns, reflecting the role of soil moisture and sparse vegetation in mitigating erosion. The radar-based RVI (S11) provided stable estimates across vegetated areas, although less sensitivity was observed in bare land zones.
Figure 3 presents soil loss maps and corresponding charts for all scenarios. These maps reveal distinct spatial differences: vegetation-dense zones were consistently characterized by low soil loss, while open fields, urban fringes, and steep agricultural slopes showed high erosion risk.
The integration of remote sensing indices such as NDVI, EVI, BSI, NDWI, SAVI, and RVI is essential for improving the accuracy of soil erosion modeling. These indices provide valuable insights into vegetation density, surface moisture, bare soil exposure, and land cover conditions. For example, vegetation-based indices like NDVI and SAVI help estimate the protective capacity of vegetation cover, while BSI aids in identifying areas more vulnerable to erosion due to surface exposure. NDWI reflects soil moisture content and is particularly useful for assessing runoff potential in saturated areas. RVI, derived from radar datasets such as Sentinel-1, offers a complementary perspective by capturing vegetation structure even under cloudy conditions.
Scenario-Based Soil Loss Simulations
  • Scenario 1 (S1, Sentinel-2 NDVI): Areas with moderate to dense vegetation cover exhibited substantially lower soil loss, while exposed areas showed elevated erosion values.
  • Scenario 2 (S2, Sentinel-2 EVI): EVI captured vegetation density more effectively than NDVI, leading to broader distributions of low soil loss across forested regions.
  • Scenario 3 (S3, Sentinel-2 BSI): Bare soil areas, particularly on steep slopes, showed sharply increased soil loss, underlining the strong influence of surface exposure.
  • Scenario 4 (S4, Sentinel-2 NDWI): Moisture-rich soils exhibited reduced erosion, while dry zones corresponded to higher soil loss.
  • Scenario 5 (S5, Sentinel-2 SAVI): SAVI effectively represented sparse vegetation conditions, providing balanced soil loss predictions in semi-arid areas.
  • Scenario 6 (S6, Landsat-8 NDVI): Similar patterns to Sentinel-2 NDVI were observed, though finer spatial variability was less visible due to lower resolution.
  • Scenario 7 (S7, Landsat-8 EVI): Demonstrated low soil loss in vegetated areas, with moderate variability across open landscapes.
  • Scenario 8 (S8, Landsat-8 BSI): Produced the highest soil loss estimates, particularly in urbanizing and degraded zones.
  • Scenario 9 (S9, Landsat-8 NDWI): Soil loss was notably lower near water bodies and moist zones, highlighting the protective role of soil moisture.
  • Scenario 10 (S10, Landsat-8 SAVI): Results were spatially homogeneous, capturing moderate vegetation as an effective buffer against erosion.
  • Scenario 11 (S11, Sentinel-1 RVI): Radar-based assessment provided consistent results across cloudy and forested areas, although less effective in bare surfaces compared to optical indices.
An integrated analysis of the scenarios highlights the dominant influence of vegetation in reducing soil erosion. Vegetation-based indices (NDVI, EVI, SAVI) consistently indicated lower soil loss, while bare soil indices (BSIs) revealed the highest erosion risks. NDWI emphasized the protective role of soil moisture, whereas RVI demonstrated the advantage of radar data in persistent cloud conditions. Overall, the predictive capacity of erosion models varied considerably depending on the selected remote sensing index, underscoring the importance of index choice in enhancing spatial accuracy and reliability of soil erosion predictions.

3.2. Index Maps

Remote sensing indices NDVI, EVI, SAVI, BSI, NDWI, and RVI were integrated to improve modeling accuracy and spatial detail. Vegetation-related indices (NDVI, EVI, SAVI) effectively highlighted areas with protective plant cover, whereas BSI identified bare or urbanized zones prone to erosion. NDWI captured soil moisture dynamics that affect runoff and detachment, and RVI provided vegetation information independent of atmospheric conditions, ensuring reliable assessments even under cloud cover (Figure 4).
Scenario-based analysis demonstrated the performance of each index:
  • Scenario 1—NDVI (Sentinel-2): High NDVI areas, representing dense vegetation, exhibited minimal soil loss, whereas low NDVI regions on slopes or agricultural lands showed increased erosion.
  • Scenario 2—EVI (Sentinel-2): EVI improved differentiation in densely vegetated areas, showing low soil loss in forests and moderate loss in agricultural zones.
  • Scenario 3—BSI (Sentinel-2): High BSI values highlighted bare soil and urban areas, corresponding to elevated erosion risk.
  • Scenario 4—NDWI (Sentinel-2): Dry areas with low NDWI values experienced substantial soil loss, while moisture-rich zones showed limited erosion.
  • Scenario 5—SAVI (Sentinel-2): SAVI captured sparse vegetation effectively, revealing higher soil loss in semi-arid and agricultural regions.
  • Scenario 6 and 7—NDVI and EVI (Landsat 8): Despite coarser resolution, patterns were consistent with Sentinel-2 results, showing low soil loss in vegetated areas.
  • Scenario 8—BSI (Landsat 8): High BSI values again corresponded to erosion-prone bare and urban areas.
  • Scenario 9—NDWI (Landsat 8): Dry soils exhibited higher erosion, confirming the role of moisture in reducing soil loss.
  • Scenario 10—SAVI (Landsat 8): Sparse vegetation in agricultural and semi-natural areas resulted in higher soil loss.
  • Scenario 11—RVI (Sentinel-1): Radar-based RVI allowed consistent evaluation of vegetated areas even under cloud cover, though fine details in bare land were less clear than optical indices.
Overall, vegetation indices (NDVI, EVI, SAVI) consistently corresponded to lower soil loss, while bare soil (BSI) and moisture (NDWI) indices effectively highlighted high-risk zones. RVI provided a reliable alternative in areas where optical imagery may be limited. These findings emphasize the importance of selecting appropriate indices according to landscape characteristics and demonstrate that integrating multiple remote sensing indices enhances the accuracy and robustness of erosion modeling. Importantly, this approach supports sustainable land management by linking erosion risk to vegetation, soil, and moisture dynamics.

3.3. Index Contribution to C Factor

The C factor derived from remote sensing indices effectively captured vegetation density, surface exposure, and soil moisture conditions, which are central determinants of erosion vulnerability. Indices such as NDVI and SAVI exhibited strong inverse relationships with soil loss, confirming their robustness in representing vegetation cover as a protective element. In contrast, BSI consistently emphasized bare and built-up areas, aligning with elevated erosion risk. The radar-based RVI provided complementary information, particularly for vegetation structure under cloudy conditions, although it showed weaker sensitivity to bare surfaces compared to optical indices.
Figure 5 presents the C-factor maps for the 11 scenarios. These maps highlight distinct spatial differences among indices, reflecting their capacity to capture landscape features relevant to soil erosion modeling.
  • Scenario 1 (NDVI—Sentinel-2): NDVI effectively represented vegetation gradients, producing low C values in forested zones and high values in exposed agricultural and urban landscapes.
  • Scenario 2 (EVI—Sentinel-2): EVI enhanced vegetation contrast compared to NDVI, reducing atmospheric interference and producing clearer delineations of low-erosion regions in densely vegetated areas.
  • Scenario 3 (BSI—Sentinel-2): BSI emphasized bare soils and built-up zones, resulting in consistently high C values over sparsely vegetated regions and highlighting areas of increased erosion vulnerability.
  • Scenario 4 (NDWI—Sentinel-2): NDWI reflected moisture-related surface variation, indicating reduced erosion risk in saturated or water-adjacent areas and higher risk in dry exposed soils.
  • Scenario 5 (SAVI—Sentinel-2): SAVI provided balanced results in sparsely vegetated zones, such as semi-arid fields, by reducing soil background effects while maintaining vegetation sensitivity.
  • Scenario 6 (NDVI—Landsat 8): While spatial resolution was coarser than Sentinel-2, Landsat-8 NDVI maps preserved general vegetation–erosion patterns, though with reduced local detail.
  • Scenario 7 (EVI—Landsat 8): EVI continued to distinguish vegetation gradients effectively, even at lower resolution, capturing forested regions as zones of low erosion susceptibility.
  • Scenario 8 (BSI—Landsat 8): Similarly to Sentinel-2, BSI with Landsat-8 strongly emphasized bare and urban surfaces, clearly identifying erosion-prone landscapes in agricultural frontiers and settlement peripheries.
  • Scenario 9 (NDWI—Landsat 8): Moisture-rich areas were consistently associated with lower C values, whereas arid landscapes displayed higher values, reinforcing the moisture–erosion relationship.
  • Scenario 10 (SAVI—Landsat 8): SAVI reduced atmospheric and soil background influences, yielding stable erosion estimates particularly in agricultural and semi-arid environments.
  • Scenario 11 (RVI—Sentinel-1): Radar-derived RVI maps demonstrated the advantage of cloud-independent monitoring, effectively representing vegetation structure in forests and croplands, though less responsive to bare surfaces compared to optical indices.
C-map analyses were instrumental in visualizing the role of different indices across spatial domains. Vegetation-related indices (NDVI, EVI, SAVI) consistently aligned with reduced soil erosion susceptibility, while indices emphasizing surface exposure (BSI) corresponded with elevated risk. NDWI added an important hydrological perspective by reflecting moisture availability, and RVI underscored the potential of radar data in supplementing optical observations.
Bar charts and bubble plots further supported these findings by illustrating how vegetation and moisture indices clustered toward low-erosion scenarios, whereas bare-soil and urban indices were more frequently associated with higher erosion estimates. These comparisons highlight the importance of carefully selecting indices for erosion modeling, as each provides unique insights into land surface conditions.
Overall, the scenario-based evaluation demonstrated that integrating multiple indices enhances both the reliability and interpretability of soil erosion modeling. Vegetation and moisture indices emerged as the most effective in reducing uncertainty, while structural indices such as BSI and RVI added critical perspectives for identifying erosion hotspots under varying landscape and climatic conditions.
The modeling experiments conducted under different scenarios demonstrated that both Random Forests (RFs) and Gradient Boosted Trees (GBTs) provided high predictive accuracy for soil loss estimation (results are presented in Table 4). In particular, the RF model generally exhibited more stable performance in terms of R2 values, often exceeding 0.85 (e.g., Scenarios 1 and 11). By contrast, although the GBT model produced competitive results in some cases (e.g., Scenarios 5 and 11), it was generally characterized by higher error values.
When error metrics (RMSE, MSE, and MAE) are examined, RF consistently yielded lower error values compared to GBT, making it the more reliable predictor. For instance, in Scenario 11, the RF model achieved an R2 of 0.857 with an MAE of 0.196, indicating strong predictive capability, while in the same scenario the GBT model performed slightly less effectively (R2 = 0.848, MAE = 0.196).
A noteworthy finding across the scenarios is that those incorporating EVI (Enhanced Vegetation Index) in Scenarios S2 and S7, and NDWI (Normalized Difference Water Index) in Scenarios S4 and S9, yielded negative or statistically insignificant results. This indicates that these indices provided limited contribution to soil loss prediction, and in some cases, even reduced the overall predictive power of the models. Specifically, NDWI, being highly sensitive to water content and soil moisture, appears inadequate in directly reflecting soil erosion processes. Similarly, EVI was found to be less discriminative than NDVI in capturing vegetation-related controls on erosion in the study area.
Overall, the RF model consistently outperformed GBT, delivering more robust and reliable results across scenarios, particularly in Scenarios 1 and 11. Conversely, the inclusion of certain indices (EVI and NDWI), either individually or in combination, negatively affected model performance. These findings highlight the critical importance of variable selection in erosion modeling and suggest that indices must be evaluated not only for their theoretical relevance but also for their empirical effectiveness under site-specific conditions.

4. Discussion

This study presents a spatial assessment of soil erosion risk by integrating the Revised Universal Soil Loss Equation (RUSLE) with remote sensing and GIS-based parameters. The model’s primary factors—rainfall erosivity (R), soil erodibility (K), slope length and steepness (LS), land cover (C), and conservation practices (P)—were derived from multiple datasets, including CHIRPS, SRTM, MODIS, and OpenLandMap. The resulting erosion risk maps revealed pronounced spatial heterogeneity, with hotspots predominantly occurring in steep-sloped, sparsely vegetated, and poorly managed agricultural lands.
Among the RUSLE factors, the LS factor was the most influential in controlling erosion intensity, especially in areas with complex topography. High NDVI and SAVI values, indicative of dense vegetation, corresponded to low C-factor values and reduced soil loss, highlighting vegetation’s critical role in stabilizing soil and mitigating runoff. Similarly, areas implementing effective conservation practices—such as terracing, contour plowing, or strip cropping—exhibited lower erosion estimates, demonstrating the model’s sensitivity to human interventions.
Rainfall erosivity (R) also contributed significantly to soil loss patterns. While the CHIRPS dataset provided comprehensive spatial coverage, its temporal resolution may have underrepresented short-term, high-intensity rainfall events, which could slightly affect erosion estimates.
Integrating multi-source satellite imagery enhanced understanding of spatial erosion patterns. Nevertheless, inherent limitations in dataset resolution and potential classification uncertainties in NDVI and LULC layers underscore the importance of field validation to improve reliability in future analyses.
Importantly, the findings highlight the relevance of sustainable land management practices. By identifying erosion-prone areas and promoting vegetation conservation alongside effective soil control measures, this study supports strategies aimed at maintaining long-term ecosystem health and agricultural productivity.
Furthermore, the scenario-based modeling with RF and GBT revealed that variable selection significantly influences predictive performance. While vegetation indices such as NDVI consistently enhanced the accuracy of soil loss estimation, the use of EVI and NDWI in certain scenarios produced negative or non-significant results, suggesting that not all spectral indices are equally effective in representing erosion dynamics. These outcomes reinforce the need for critical evaluation of input variables, as theoretically relevant indices may not always translate into improved model performance under specific environmental and climatic conditions. The superior and more stable performance of the RF model across scenarios underscores its suitability for soil erosion prediction in heterogeneous landscapes, where complex interactions between topography, vegetation, and land management practices prevail.

5. Conclusions

The RUSLE-based modeling framework combined with remote sensing and GIS techniques proved effective for spatially explicit soil erosion assessment. The key findings are summarized as follows:
  • High-risk erosion zones were mainly located in steep-sloped and deforested areas, where the combined effect of high LS and C factors increased erosion potential.
  • Vegetative cover, as indicated by NDVI- and SAVI-derived C factors, played a crucial role in reducing soil loss, with densely vegetated areas consistently showing lower erosion.
  • Anthropogenic interventions captured by the P factor, including terracing and contour plowing, effectively reduced erosion, emphasizing the importance of sustainable land management strategies.
  • Integration of CHIRPS, MODIS, SRTM, and soil datasets provided detailed insight into spatial variability of erosion risk, although some uncertainty remains due to resolution limitations and the lack of field validation.
  • The model outputs offer practical guidance for policymakers, land managers, and conservation planners to identify erosion-prone areas and implement sustainable interventions that balance agricultural productivity with ecosystem preservation.
Overall, this RUSLE-based framework provides a cost-effective, scalable, and sustainability-oriented tool for promoting soil conservation, supporting long-term land use planning, and enhancing ecological resilience in regions facing land degradation challenges.
In addition to the RUSLE-based framework, the scenario analyses conducted with machine learning models (RF and GBT) further emphasized the importance of variable selection in soil erosion modeling. The RF model consistently outperformed GBT in terms of stability and predictive accuracy, achieving R2 values above 0.85 in several scenarios (e.g., Scenario 1 and 11). Conversely, scenarios incorporating EVI and NDWI produced negative or insignificant results, underscoring that not all spectral indices contribute positively to erosion prediction. These findings highlight the necessity of carefully selecting input variables that truly capture the driving factors of erosion. Taken together, the integration of RUSLE with advanced machine learning approaches strengthens the capacity for reliable soil erosion risk assessment and provides valuable insights for evidence-based land management and conservation planning.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data are available upon request.

Conflicts of Interest

The author declare no conflicts of interest.

References

  1. Morgan, R.P.C.; Hann, M.J. Design of diverter berms for soil erosion control and biorestoration along pipeline rights-of-way. Soil Use Manag. 2005, 21, 306–311. [Google Scholar] [CrossRef]
  2. Mitasova, H.; Barton, M.; Ullah, I.; Hofierka, J.; Harmon, R.S. GIS-based soil erosion modeling. In Remote Sensing and GIScience in Geomorphology; Elsevier Inc.: Amsterdam, The Netherlands, 2013; pp. 228–258. [Google Scholar]
  3. Panagos, P.; Borrelli, P.; Meusburger, K. A new European slope length and steepness factor (LS-Factor) for modeling soil erosion by water. Geosciences 2015, 5, 117–126. [Google Scholar] [CrossRef]
  4. Borrelli, P.; Robinson, D.A.; Fleischer, L.R.; Lugato, E.; Ballabio, C.; Alewell, C.; Panagos, P. An assessment of the global impact of 21st century land use change on soil erosion. Nat. Commun. 2017, 8, 2013. [Google Scholar] [CrossRef]
  5. Benavidez, R.; Jackson, B.; Maxwell, D.; Norton, K. A review of the (revised) universal soil loss equation (RUSLE): With a view to increasing its global applicability and improving soil loss estimates. Hydrol. Earth Syst. Sci. 2018, 22, 6059–6086. [Google Scholar] [CrossRef]
  6. Panagos, P.; Ballabio, C.; Himics, M.; Scarpa, S.; Matthews, F.; Bogonos, M.; Borrelli, P. Projections of soil loss by water erosion in Europe by 2050. Environ. Sci. Policy 2021, 124, 380–392. [Google Scholar] [CrossRef]
  7. Delgado, D.; Sadaoui, M.; Ludwig, W.; Mendez, W. Depth of the pedological profile as a conditioning factor of soil erodibility (RUSLE K-Factor) in Ecuadorian basins. Environ. Earth Sci. 2023, 82, 286. [Google Scholar] [CrossRef]
  8. Ding, B.; Zhang, J.; Zheng, P.; Li, Z.; Wang, Y.; Jia, G.; Yu, X. Water security assessment for effective water resource management based on multi-temporal blue and green water footprints. J. Hydrol. 2024, 632, 130761. [Google Scholar] [CrossRef]
  9. Bach, E.M.; Ramirez, K.S.; Fraser, T.D.; Wall, D.H. Soil biodiversity integrates solutions for a sustainable future. Sustainability 2020, 12, 2662. [Google Scholar] [CrossRef]
  10. Hurni, H. Current international actions for furthering the sustainability use of soils. In Symposium Papers; University of Bern: Bern, Switzerland, 2002; Volume 63, pp. 1–8. [Google Scholar]
  11. Piccarreta, M.; Capolongo, D.; Boenzi, F.; Bentivenga, M. Implications of decadal changes in precipitation and land use policy to soil erosion in Basilicata, Italy. Catena 2006, 65, 138–151. [Google Scholar] [CrossRef]
  12. Brady, N.C.; Weil, R.R. The Nature and Properties of Soils, 13th ed.; Prentice Hall: Upper Saddle River, NJ, USA, 2008; pp. 662–710. [Google Scholar]
  13. Rodrigo, C.J.; Brings, C.; Lassu, T.; Iserloh, T.; Senciales, J.M.; Martínez, M.J.F.; Ruiz, S.J.D.; Seeger, M.; Ries, J.B. Rainfall and human activity impacts on soil losses and rill erosion in vineyards (Ruwer Valley, Germany). Solid Earth 2015, 6, 823–837. [Google Scholar] [CrossRef]
  14. Gao, B.C. NDWI—A normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens. Environ. 1996, 58, 257–266. [Google Scholar] [CrossRef]
  15. Otgonbayar, M.; Clement, A.; Jonathan, C.; Amarsaikhan, D. Mapping pasture biomass in Mongolia using partial least squares, random forest regression and Landsat 8 imagery. Int. J. Remote Sens. 2019, 40, 3204–3226. [Google Scholar] [CrossRef]
  16. Carlson, T.N.; Ripley, D.A. On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sens. Environ. 1997, 62, 241–252. [Google Scholar] [CrossRef]
  17. Bannari, A.; Morin, D.; Bonn, F.; Huete, A. A review of vegetation indices. Remote Sens. Rev. 1995, 13, 95–120. [Google Scholar] [CrossRef]
  18. Gao, B.C. Normalized difference water index for remote sensing of vegetation liquid water from space. In Imaging Spectrometry, Proceedings of the SPIE, Orlando, FL, USA, 5 June 1995; SPIE: Bellingham, WA, USA, 1995; Volume 2480, pp. 225–236. [Google Scholar] [CrossRef]
  19. Huete, A.R. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ. 1988, 25, 295–309. [Google Scholar] [CrossRef]
  20. Nguyen, C.T.; Chidthaisong, A.; Diem, P.K.; Huo, L.Z. A Modified Bare Soil Index to Identify Bare Land Features during Agricultural Fallow-Period in Southeast Asia Using Landsat 8. Land 2021, 10, 231. [Google Scholar] [CrossRef]
  21. Charbonneau, F.; Trudel, M.; Fernandes, R. Use of Dual Polarization and Multi-Incidence SAR for Soil Permeability Mapping. In Proceedings of the 2005 Advanced Synthetic Aperture Radar (ASAR) Workshop, St-Hubert, QC, Canada, 15–17 November 2005. [Google Scholar]
  22. Jeevalakshmi, D.; Reddy, S.N.; Manikiam, B. Land cover classification based on NDVI using LANDSAT 8 time series: A case study Tirupati region. In Proceedings of the 2016 International Conference on Communication and Signal Processing (IC-CSP), Melmaruvathur, India, 6–8 April 2016; pp. 1332–1335. [Google Scholar]
  23. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring vegetation systems in the Great Plains with ERTS. In Proceedings of the 3rd Earth Resources Technology Satellite Symposium, Washington, DC, USA, 10–14 December 1973; NASA: Greenbelt, MD, USA, 1973; Volume 1, pp. 309–317. [Google Scholar]
  24. Tucker, C.J. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 1979, 8, 127–150. [Google Scholar] [CrossRef]
  25. Li, Z.; Xin, X.; Tang, H.; Yang, F.; Chen, B.; Zhang, B. Estimating grassland LAI using the random forests approach and Landsat imagery in the meadow steppe of Hulunber, China. J. Integr. Agric. 2017, 16, 286–297. [Google Scholar] [CrossRef]
  26. Ding, L.; Li, Z.; Wang, X.; Yan, X.; Shen, B.; Chen, B.; Xin, X. Estimating grassland carbon stocks in Hulunber, China using Landsat-8 OLI imagery and regression kriging. Sensors 2019, 19, 5374. [Google Scholar] [CrossRef]
  27. Liu, H.Q.; Huete, A. A feedback-based modification of the NDVI to minimize canopy background and atmospheric noise. IEEE Trans. Geosci. Remote Sens. 1995, 33, 457–465. [Google Scholar] [CrossRef]
  28. Jin, Y.; Yang, X.; Qiu, J.; Li, J.; Gao, T.; Wu, Q.; Zhao, F.; Ma, H.; Yu, H.; Xu, B. Remote sensing-based biomass estimation and its spatio-temporal variations in temperate grassland, northern China. Remote Sens. 2014, 6, 1496–1513. [Google Scholar] [CrossRef]
  29. Fourty, T.H.; Baret, F. On spectral estimates of fresh leaf biochemistry. Int. J. Remote Sens. 1998, 19, 1283–1297. [Google Scholar] [CrossRef]
  30. Ceccato, P.; Flasse, S.; Tarantola, S.; Jacquemoud, S.; Grégoire, J.M. Designing a spectral index to estimate vegetation water content from remote sensing data: Part 1: Theoretical approach. Remote Sens. Environ. 2002, 82, 188–197. [Google Scholar] [CrossRef]
  31. Serrano, J.; Shahidian, S.; da Silva, J.M. Evaluation of normalized difference water index as a tool for monitoring pasture seasonal and inter-annual variability in a Mediterranean agro-silvo-pastoral system. Water 2019, 11, 62. [Google Scholar] [CrossRef]
  32. Rikimaru, A.; Roy, P.S.; Miyatake, S. Tropical forest cover density mapping. Trop. Ecol. 2002, 43, 39–47. [Google Scholar]
  33. Mzid, N.; Pignatti, S.; Huang, W.; Casa, R. An analysis of bare soil occurrence in arable croplands for remote sensing topsoil applications. Remote Sens. 2021, 13, 474. [Google Scholar] [CrossRef]
  34. Diek, S.; Fornallaz, F.; Schaepman, M.; de Jong, M. Barest pixel composite for agricultural areas using Landsat time series. Remote Sens. 2017, 9, 1245. [Google Scholar] [CrossRef]
  35. Jin, K.; Wang, F.; Han, J. Contribution of climatic change and human activities to vegetation NDVI change over China during 1982–2015. Acta Geogr. Sin. 2020, 75, 961–974. [Google Scholar]
  36. Gao, G.; Wang, S. Compare analysis of vegetation cover change in Jianyang city based on RVI and NDVI. In Proceedings of the 2nd International Conference on Remote Sensing, Environment and Transportation Engineering, Nanjing, China, 1–4 June 2012. [Google Scholar]
  37. Renard, K.G.; McCool, D.K.; Cooley, K.R.; Foster, G.R.; Istok, J.D.; Mutchler, C.K. Rainfall–runoff erosivity factor (R). In Predicting Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation (RUSLE); US Department of Agriculture, Agricultural Research Service: Washington, DC, USA, 1997. [Google Scholar]
  38. Morgan, R.P.C. Soil Degradation and Soil Erosion in the Loamy belt of Northern Europe; A.A. Balkema: Rotterdam, The Netherlands, 1986; Volume 2, pp. 165–172. [Google Scholar]
  39. Lal, R. Soil and Water Conservation Society. In Soil Erosion Research Methods; St. Lucie Press: Ankeny, IA, USA, 1994. [Google Scholar]
  40. Wischmeier, W.H.; Smith, D.D. Predicting Rainfall Erosion Losses: A Guide to Conservation Planning; USDA Handbook No. 537; U.S. Department of Agriculture: Washington, DC, USA, 1978.
  41. Foster, G.R.; Renard, K.G.; Yoder, D.C.; McCool, D.K.; Weesies, G.A. RUSLE User’s Guide; The Soil and Water Conservation Society: Ankeny, IA, USA, 1996. [Google Scholar]
  42. Williams, J.R. EPIC: The erosion-productivity impact calculator. In Proceedings of the 1989 Summer Computer Simulation Conference; Clema, J.K., Ed.; The Society: Austin, TX, USA, 1989; pp. 676–681. [Google Scholar]
  43. Morgan, R.P.C. Soil Erosion and Conservation; John Wiley & Sons: Hoboken, NJ, USA, 2009. [Google Scholar]
  44. Arekhi, S. Evaluating long-term annual sediment yield estimating the potential of GIS interfaced MUSLE model on two micro watersheds, Pakistan. Pak. J. Biol. Sci. 2008, 11, 270–274. [Google Scholar] [CrossRef]
  45. Millward, A.A.; Mersey, J.E. Adapting RUSLE to model soil erosion potential in a mountainous tropical watershed. Catena 1999, 38, 109–129. [Google Scholar] [CrossRef]
  46. Shabani, F.; Kumar, L.; Esmaeili, A. Improvement to the prediction of the USLE K factor. Geomorphology 2014, 204, 229–234. [Google Scholar] [CrossRef]
  47. Mhaske, S.N.; Pathak, K.; Dash, S.S.; Nayak, D.B. Assessment and management of soil erosion in the hilltop mining dominated catchment using GIS integrated RUSLE model. J. Environ. Manag. 2021, 294, 112987. [Google Scholar] [CrossRef]
  48. Williams, J.R.; Singh, V.P. The EPIC model. In Computer Models of Watershed Hydrology; Singh, V.P., Ed.; Water Resources Publications: Highlands Ranch, CO, USA, 1995; pp. 909–1000. [Google Scholar]
  49. Getnet, T.; Mulu, A. Assessment of soil erosion rate and hotspot areas using RUSLE and multi-criteria evaluation technique at Jedeb watershed, Upper Blue Nile, Amhara region, Ethiopia. Environ. Chall. 2021, 4, 100174. [Google Scholar] [CrossRef]
  50. Kaltenrieder, J.; Hurni, P.D.H.; Herweg, K.D. Adaptation and Validation of the Universal Soil Loss Equation (USLE) for the Ethiopian-Eritrean Highlands. Master’s Thesis, University of Bern, Bern, Switzerland, 2007. [Google Scholar]
  51. Rejani, R.; Rao, K.V.; Osman, M.; Rao, C.S.; Reddy, K.S.; Chary, G.R.; Pushpanjali; Samuel, J. Spatial and temporal estimation of soil loss for the sustainable management of a wet semi-arid watershed cluster. Environ. Monit. Assess. 2016, 188, 143. [Google Scholar] [CrossRef]
  52. Ganasri, B.P.; Ramesh, H. Assessment of soil erosion using the RUSLE model using remote sensing and GIS—A case study of the Nethravathi Basin. Geosci. Front. 2016, 7, 953–961. [Google Scholar] [CrossRef]
  53. Renard, K.G. Predicting Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation (RUSLE); US Department of Agriculture, Agricultural Research Service: Washington, DC, USA, 1997.
Sustainability 17 07913 g001
Figure 1. Study area.
Figure 1. Study area.
Sustainability 17 07913 g001
Sustainability 17 07913 g002
Figure 2. Workflow of the study.
Figure 2. Workflow of the study.
Sustainability 17 07913 g002
Sustainability 17 07913 g003aSustainability 17 07913 g003bSustainability 17 07913 g003c
Figure 3. Soil loss maps and charts under different scenarios: (a) Scenario 1; (b) Scenario 2; (c) Scenario 3; (d) Scenario 4; (e) Scenario 5; (f) Scenario 6; (g) Scenario 7; (h) Scenario 8; (i) Scenario 9; (j) Scenario 10; (k) Scenario 11.
Figure 3. Soil loss maps and charts under different scenarios: (a) Scenario 1; (b) Scenario 2; (c) Scenario 3; (d) Scenario 4; (e) Scenario 5; (f) Scenario 6; (g) Scenario 7; (h) Scenario 8; (i) Scenario 9; (j) Scenario 10; (k) Scenario 11.
Sustainability 17 07913 g003aSustainability 17 07913 g003bSustainability 17 07913 g003c
Sustainability 17 07913 g004aSustainability 17 07913 g004b
Figure 4. Indices maps: (a) Sentine 2 Ndvi; (b) Sentinel 2 Evi; (c) Sentinel 2 Bsi; (d) Sentinel 2 Ndwi; (e) Sentinel 2 Savi; (f) Landsat 8 Ndvi; (g) Landsat 8 Evi; (h) Landsat 8 Bsi; (i) Landsat 8 Ndwi; (j) Landsat 8 Savi; (k) Sentinel 1 Rvi.
Figure 4. Indices maps: (a) Sentine 2 Ndvi; (b) Sentinel 2 Evi; (c) Sentinel 2 Bsi; (d) Sentinel 2 Ndwi; (e) Sentinel 2 Savi; (f) Landsat 8 Ndvi; (g) Landsat 8 Evi; (h) Landsat 8 Bsi; (i) Landsat 8 Ndwi; (j) Landsat 8 Savi; (k) Sentinel 1 Rvi.
Sustainability 17 07913 g004aSustainability 17 07913 g004b
Sustainability 17 07913 g005aSustainability 17 07913 g005b
Figure 5. The maps for Scenarios C: (a) Scenario 1 C map; (b) Scenario 2 C map; (c) Scenario 3 C map; (d) Scenario 4 C map; (e) Scenario 5 C map; (f) Scenario 6 C map; (g) Scenario 7 C map; (h) Scenario 8 C map; (i) Scenario 9 C map; (j) Scenario 10 C map; (k) Scenario 11 C map.
Figure 5. The maps for Scenarios C: (a) Scenario 1 C map; (b) Scenario 2 C map; (c) Scenario 3 C map; (d) Scenario 4 C map; (e) Scenario 5 C map; (f) Scenario 6 C map; (g) Scenario 7 C map; (h) Scenario 8 C map; (i) Scenario 9 C map; (j) Scenario 10 C map; (k) Scenario 11 C map.
Sustainability 17 07913 g005aSustainability 17 07913 g005b
Table 1. RUSLE factors and dataset.
Table 1. RUSLE factors and dataset.
RUSLE FactorDescription/MethodData SourceSpatial
Resolution
R (Rainfall erosivity)Annual rainfall aggregated from pentadal precipitation, converted using erosivity equationCHIRPS v2.0 (Climate Hazards Group)~5 km
K (Soil erodibility)Derived from soil texture class and organic matter content, USDA lookup tableOpenLandMap 250 m
LS (Topographic
factor)		Slope length and steepness computed using Desmet & Govers’ methodSRTM DEM v330 m
C (Cover
management)		Derived from vegetation/surface indices (NDVI, EVI, SAVI, NDWI, BSI, RVI)Sentinel-2 MSI, Landsat 8 OLI, Sentinel-1 SAR10–30–10 m
P (Support practices)Conservation factor estimated by integrating slope with land use/land coverDynamic World LULC 10 m
Table 2. Index calculations.
Table 2. Index calculations.
IndicesEquationReference
NDVI(NIR − RED)/(NIR + RED)[16]
EVI2.5 NIR − RED/(NIR + 6RED − 7.5BLUE) + 1[17]
NDWI(NIR − SWIR)/(NIR + SWIR)[18]
SAVI(NIR − RED)/(NIR + RED+ 0.5) × (1 + 0.5)[19]
BSI((SWIR2 + RED) − (NIR + BLUE))/((SWIR2 + RED) + (NIR + BLUE))[20]
RVI4 VH/(VV + VH)[21]
Table 3. RUSLE factors and their equations with references.
Table 3. RUSLE factors and their equations with references.
FactorsEquationReference
Rainfall Runoff Erosivity Factor (R) 12   (1.5*log(Pm2/Pα) − 0.08188)
R = ∑1.73*10
 I = 1 		[40]
Soil Erodibility Factor (K)K = {0.2 + 0.3*exp[(−0.0256*SAN*(1.0 − SIL/100))]}*(SIL/CLA + SIL)0.3
*{1 − (0.25*C)/(C + exp(3.72 − 2.95*C))}*(1 − (0.7*Sn/Sn + exp(22.9*Sn − 5.51)))*0.1317 		[42]
Slope Length and Steepness Factor (LS)LS = (Flowaccumulation*(Cellsize/22.13)0.4*{(sin(slope)*0.01748)})1.4[41]
Cover Management Factor (C)NDVI = NIR − R/NIR + R/c = exp(−αNDVI/β − NDVI)[41]
RUSLE ModelA = R × K × LS × C × P[40,41]
Table 4. Summary of scenarios and corresponding model results.
Table 4. Summary of scenarios and corresponding model results.
Scenario NoModelR2RMSEMSEMAE
1RF0.857220.407080.165710.15429
3RF0.795630.503510.253520.23944
5RF0.854390.478710.229170.20139
6RF0.768580.668730.447200.32298
8RF0.749290.503610.253620.22464
10RF0.817660.503150.253160.21519
11RF0.857050.443200.196430.19643
1GBT0.783370.501430.251430.19429
3GBT0.767240.537340.288730.27465
5GBT0.854390.478710.229170.21528
6GBT0.713930.743500.552800.37888
8GBT0.777940.473960.224640.21014
10GBT0.776630.556890.310130.25949
11GBT0.848390.456440.208330.19643
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Uyar, N. Index-Driven Soil Loss Mapping Across Environmental Scenarios: Insights from a Remote Sensing Approach. Sustainability 2025, 17, 7913. https://doi.org/10.3390/su17177913

AMA Style

Uyar N. Index-Driven Soil Loss Mapping Across Environmental Scenarios: Insights from a Remote Sensing Approach. Sustainability. 2025; 17(17):7913. https://doi.org/10.3390/su17177913

Chicago/Turabian Style

Uyar, Nehir. 2025. "Index-Driven Soil Loss Mapping Across Environmental Scenarios: Insights from a Remote Sensing Approach" Sustainability 17, no. 17: 7913. https://doi.org/10.3390/su17177913

APA Style

Uyar, N. (2025). Index-Driven Soil Loss Mapping Across Environmental Scenarios: Insights from a Remote Sensing Approach. Sustainability, 17(17), 7913. https://doi.org/10.3390/su17177913

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

Article Metrics

Back to TopTop