Next Article in Journal
Ownership Patterns and Landscape Diversity: Conservation Implications in Maryland
Previous Article in Journal
From Stated Importance to Revealed Preferences: Assessing Residential Property Features
Previous Article in Special Issue
A BiLSTM-Based Hybrid Ensemble Approach for Forecasting Suspended Sediment Concentrations: Application to the Upper Yellow River
 
 
Article
Peer-Review Record

The Spatiotemporal Evolution, Driving Mechanisms, and Future Climate Scenario-Based Projection of Soil Erosion in the Southwest China

Land 2025, 14(7), 1341; https://doi.org/10.3390/land14071341
by Yangfei Huang 1,2,*, Chenjian Zhong 1, Yuan Wang 1 and Wenbin Hua 1
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Land 2025, 14(7), 1341; https://doi.org/10.3390/land14071341
Submission received: 6 May 2025 / Revised: 20 June 2025 / Accepted: 22 June 2025 / Published: 24 June 2025
(This article belongs to the Special Issue Artificial Intelligence for Soil Erosion Prediction and Modeling)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors address an important topic by analyzing soil erosion using the Revised Universal Soil Loss Equation (RUSLE) model, which is one of the most widely applied models globally. They assess erosion for both past and future periods under different climate scenarios. A notable contribution of this study is the use of the SHAP model to examine the interaction among factors and their influence on soil erosion processes.

Following a detailed review of the manuscript, several ambiguities and methodological issues were identified. Please find below my specific comments and suggestions:

  1. Since the RUSLE model expresses soil loss in t·ha⁻¹·yr⁻¹, this unit should also be used consistently throughout the manuscript. Additionally, it is recommended that the units for R and K factors be presented by the original methodology.
  2. In Figure 1, improve the visibility of the legend for elevation and precipitation.
  3. In Table 1, it is advisable to include references alongside the listed data sources. This would allow readers to consult original publications or websites for further information.
  4. In the Results section, it remains unclear how soil erosion was calculated for the years 2000, 2005, 2010, 2015, 2020, and 2023, given that only the input data for 2023 are presented. Considering that dynamic factors such as rainfall and land use/land cover (LULC) significantly influence soil erosion, and static factors such as K and LS generally remain unchanged (with some exceptions), it is unclear why these were not shown for each time slice. Although the P factor can be static or dynamic depending on conservation practices, it would be beneficial to present R and C factors for each year, both spatially (as maps) and in tabular form.
  5. Regarding Figure 4, if erosion was modeled at 5-year intervals (2000, 2005, 2010, etc.), why does the graph include data for years such as 2002, 2004, 2006, 2008, and so on?
  6. What is the explanation for the observed decrease in soil erosion between 2000 and 2023?
  7. Lines 435–437 and 443–444 mention efficient control and management of erosion, but this is not supported by evidence since the relevant factors are not shown for each year.
  8. The application and structure of the XGBoost model are not sufficiently explained in the Methods and Results sections. A general overview is provided, but further clarification is required to ensure reproducibility and understanding among readers.
  9. In Figure 5, specifically subfigure (f), the purpose of the dots on the map should be explained, either in the figure caption or within the main text.
  10. In section 3.3.1 Influence of climate factors on soil erosion, it is unclear whether the discussion refers to Figure 8 or Figure 9. This ambiguity also applies to sections 3.3.2 and 3.3.3.
  11. In section 3.3.3, when referring to a “generally positive influence on soil erosion,” it should be clarified whether this refers to an increase in erosion rates and sediment yield or to a beneficial reduction in erosion intensity.
  12. The SHAP model appears to be applied only for the year 2023. Why was it not used to analyze factor influence for earlier years (2000–2020)?
  13. Lines 513–516 state that erosion potential is higher in areas with more than 75% built-up land. Clarification is needed here: does "built-up land" refer to impervious surfaces such as buildings, glass, or concrete? If so, such areas do not undergo typical soil detachment due to raindrop impact, and the main concern would be increased surface runoff rather than soil erosion per se.
  14. Section 3.4 Soil erosion projections contains numerous inconsistencies. Firstly, the R factor values used for future climate scenarios should be clearly described, including whether there was an increase or decrease, along with minimum, maximum, and mean values. A spatial representation of these changes should be included either in the main manuscript or as supplementary material.
  15. Another critical point is the treatment of LULC in future erosion modeling. Did the authors use static LULC from 2023 across all future periods? If so, the manuscript should state this assumption clearly. If LULC were simulated for future scenarios, the modeling approach should be fully described. The same applies to the C factor. If NDVI was used to calculate the C factor in the past and present, how was it estimated for the future?
  16. Line 583 is the first instance where the Modified Universal Soil Loss Equation (MUSLE) is mentioned. However, throughout the manuscript, the RUSLE model is referenced. Since MUSLE and RUSLE are methodologically distinct, this needs to be corrected for consistency and accuracy.
  17. In section 4.2, Discussion, the figures, tables, and graphs should not be included. The discussion should interpret the results and compare them with findings from previous studies. Figure 11 and its accompanying description should be moved to the Results section, and only the interpretive discussion should remain here.

Author Response

The authors address an important topic by analyzing soil erosion using the Revised Universal Soil Loss Equation (RUSLE) model, which is one of the most widely applied models globally. They assess erosion for both past and future periods under different climate scenarios. A notable contribution of this study is the use of the SHAP model to examine the interaction among factors and their influence on soil erosion processes.

Following a detailed review of the manuscript, several ambiguities and methodological issues were identified. Please find below my specific comments and suggestions:

  1. Since the RUSLE model expresses soil loss in t·ha⁻¹·yr⁻¹, this unit should also be used consistently throughout the manuscript. Additionally, it is recommended that the units for R and K factors be presented by the original methodology.

According to the reviewer’s suggestion, the unit of soil loss throughout the manuscript has been standardized to t·ha⁻¹·yr⁻¹. Additionally, the units of the R factor and K factor have been consistently revised to MJ·mm·(ha·h·yr)⁻¹ and t·ha·h·(ha·MJ·mm)⁻¹, respectively.

 

  1. In Figure 1, improve the visibility of the legend for elevation and precipitation.

According to the reviewer’s suggestion, the legends for elevation and precipitation in Figure 1 have been redesigned to enhance visual clarity by improving color contrast and font readability. Additionally, to strengthen the spatial representation and completeness of the map, latitude and longitude coordinates have been added to the figure, further optimizing its overall presentation.

 

  1. In Table 1, it is advisable to include references alongside the listed data sources. This would allow readers to consult original publications or websites for further information.

According to the reviewer’s suggestion, we have added the corresponding reference numbers next to each data source in Table 1. Additionally, the original sources or website links have been listed in the reference section at the end of the manuscript to facilitate further consultation by readers.

 

  1. In the Results section, it remains unclear how soil erosion was calculated for the years 2000, 2005, 2010, 2015, 2020, and 2023, given that only the input data for 2023 are presented. Considering that dynamic factors such as rainfall and land use/land cover (LULC) significantly influence soil erosion, and static factors such as K and LS generally remain unchanged (with some exceptions), it is unclear why these were not shown for each time slice. Although the P factor can be static or dynamic depending on conservation practices, it would be beneficial to present R and C factors for each year, both spatially (as maps) and in tabular form.

According to the reviewer’s suggestion, we have added a detailed explanation in the results section regarding the calculation method for soil erosion data across different years. Specifically, for the years 2000, 2005, 2010, 2015, and 2020, we used a five-year moving average of the surrounding years to represent the values (e.g., the soil erosion value for 2000 is the average of 1998–2002). For 2023, the soil erosion data were calculated based on actual data from that year. This methodological approach has been clearly stated in the main text.

 

Furthermore, this study primarily focuses on the spatiotemporal evolution of soil erosion under climate change, with a particular emphasis on the role of rainfall erosivity (R factor). Therefore, we included spatial distribution maps of the R factor across different years to highlight the driving effect of rainfall changes. Due to space limitations, we did not present the multi-year spatial distribution of the C factor. However, future studies will consider incorporating dynamic vegetation factors to enrich the analysis.

 

  1. Regarding Figure 4, if erosion was modeled at 5-year intervals (2000, 2005, 2010, etc.), why does the graph include data for years such as 2002, 2004, 2006, 2008, and so on?

This study actually calculated soil erosion data continuously from 1998 to 2023. To analyze the spatiotemporal evolution characteristics of soil erosion under the background of climate change, six key years (2000, 2005, 2010, 2015, 2020, and 2023) were selected for spatial distribution mapping. Figure 4 aims to reflect the interannual variation trend of soil erosion modulus during 2000–2023; therefore, soil erosion data at two-year intervals were included to improve the resolution of the time series analysis and enhance trend identification.

 

  1. What is the explanation for the observed decrease in soil erosion between 2000 and 2023?

Based on this suggestion, we have supplemented the main text in Section 3.2.2 with an explanation of the decreasing trend of soil erosion from 2000 to 2023. This explanation elaborates on factors such as the implementation of ecological engineering, optimization of land use management, and improvement of vegetation cover.

 

  1. Lines 435–437 and 443–444 mention efficient control and management of erosion, but this is not supported by evidence since the relevant factors are not shown for each year.

Regarding the “effective control and management of soil erosion” mentioned in the manuscript, we have provided a detailed discussion in Section 4. This section incorporates typical regional governance policies (such as the Grain for Green Project) and previous research findings, supported by relevant citations. The analysis explores possible reasons for the decline in soil erosion intensity from the perspectives of policy responses, land use changes, and the implementation effects of ecological engineering.

Concerning the issue of “displaying related factors for each year,” the manuscript systematically analyzes the spatial variation of the R factor across multiple years in the Results section (Section 3) to illustrate the impact of rainfall changes on soil erosion. Although some factors (e.g., the C factor) are not displayed annually in the figures, their processing methods are thoroughly described in both the Methods and Results sections. Overall, the evidence sufficiently supports the conclusion that soil erosion has been effectively controlled.

 

  1. The application and structure of the XGBoost model are not sufficiently explained in the Methods and Results sections. A general overview is provided, but further clarification is required to ensure reproducibility and understanding among readers.

In response to the comment, we have added relevant details regarding the XGBoost model setup and validation in Section 3.3.1. Specifically, the XGBoost algorithm in this study is configured with the following parameters: the objective function is set to “reg:squarederror” for regression tasks. The learning rate (eta) is 0.05, and the maximum tree depth (max_depth) is 7, controlling model complexity. The gamma parameter is set to 0.4 to regulate the minimum loss reduction required for a split. Both subsample and colsample_bytree parameters are set to 0.8 to reduce overfitting by randomly sampling 80% of the data and features for each tree. Additionally, L2 regularization (lambda = 1.2) and L1 regularization (alpha = 0.3) are applied to control complexity and prevent overfitting. The model is trained using cross-validation (xgb.cv), with early stopping triggered after 30 rounds without improvement. The final model is trained based on the best iteration identified by cross-validation. Finally, SHAP (SHapley Additive exPlanations) analysis is used to assess and quantify the contribution of each feature to the model’s predictions.

 

  1. In Figure 5, specifically subfigure (f), the purpose of the dots on the map should be explained, either in the figure caption or within the main text.

The black dotted symbols in Figure 5f represent areas of statistical significance (q < 0.05), highlighting spatial locations where changes in soil erosion are significantly correlated with driving factors. We have added this explanation to the caption of Figure 5 to improve the clarity and readability of the figure.

 

  1. In section 3.3.1 Influence of climate factors on soil erosion, it is unclear whether the discussion refers to Figure 8 or Figure 9. This ambiguity also applies to sections 3.3.2 and 3.3.3.

In response to the comment, we have carefully reviewed the relevant discussions in Section 3.3. We have now explicitly indicated the corresponding Figure 8 panels for each discussion in Sections 3.3.2 (climatic factors), 3.3.3 (topographic factors), and 3.3.4 (human activity factors) to avoid ambiguity and enhance consistency between the text and figures.

 

  1. In section 3.3.3, when referring to a “generally positive influence on soil erosion,” it should be clarified whether this refers to an increase in erosion rates and sediment yield or to a beneficial reduction in erosion intensity.

The phrase “overall positive effect on soil erosion” in the manuscript refers to a beneficial impact in terms of reducing soil erosion. Specifically, under certain conditions, human activities—such as ecological engineering projects and optimized land management—contribute to mitigating the soil erosion process. We have revised the related statements in Section 3.3.3 to clearly clarify this meaning.

 

  1. The SHAP model appears to be applied only for the year 2023. Why was it not used to analyze factor influence for earlier years (2000–2020)?

Regarding the time range of the SHAP model application, in fact, the modeling process used data from five years—2000, 2005, 2010, 2015, and 2020—for an overall analysis rather than building separate models for each individual year. This approach was chosen because analyzing multiple years’ data collectively better captures the long-term trends and persistent effects of soil erosion.

 

  1. Lines 513–516 state that erosion potential is higher in areas with more than 75% built-up land. Clarification is needed here: does "built-up land" refer to impervious surfaces such as buildings, glass, or concrete? If so, such areas do not undergo typical soil detachment due to raindrop impact, and the main concern would be increased surface runoff rather than soil erosion per se.

The term "built-up areas" in the manuscript indeed refers to impervious surfaces, including buildings, roads, and concrete pavements. We agree with the reviewer’s point that such areas no longer experience traditional raindrop splash erosion. However, due to their significant alteration of surface hydrological processes, these areas increase surface runoff intensity and concentration, which may indirectly exacerbate erosion risks in adjacent non-paved regions. Therefore, the phrase “higher erosion potential” in the manuscript primarily emphasizes the amplified impact on the erosion risk of neighboring areas. We have clarified and revised the relevant statements accordingly.

 

  1. Section 3.4 Soil erosion projectionscontains numerous inconsistencies. Firstly, the R factor values used for future climate scenarios should be clearly described, including whether there was an increase or decrease, along with minimum, maximum, and mean values. A spatial representation of these changes should be included either in the main manuscript or as supplementary material.

We appreciate the reviewer’s valuable comment. In response, we have added Section 3.4.1 to explicitly analyze the rainfall erosivity (R factor) under different future climate scenarios. This section now includes a discussion of the overall variation trends (increase or decrease) in R values, along with the average values corresponding to each scenario and year. Considering that the minimum and maximum values of R are often influenced by localized extreme precipitation events, we focused on the analysis of average R values. Additionally, spatial distribution maps of the R factor under all scenarios have been incorporated as Figure 10 to visually support our findings.

 

  1. Another critical point is the treatment of LULC in future erosion modeling. Did the authors use static LULC from 2023 across all future periods? If so, the manuscript should state this assumption clearly. If LULC were simulated for future scenarios, the modeling approach should be fully described. The same applies to the C factor. If NDVI was used to calculate the C factor in the past and present, how was it estimated for the future?

Under future scenarios, the trends of land use and land cover (LULC) changes are relatively uncertain. Therefore, we chose 2023 as the reference baseline for predictive modeling. Regarding the treatment of the C factor, we used the annual average method because estimating the C factor typically requires consideration of long-term environmental change trends. Using the annual average can, to some extent, reflect the inter-annual variability.

 

  1. Line 583 is the first instance where the Modified Universal Soil Loss Equation (MUSLE) is mentioned. However, throughout the manuscript, the RUSLE model is referenced. Since MUSLE and RUSLE are methodologically distinct, this needs to be corrected for consistency and accuracy.

Regarding the RUSLE model, since the MUSLE and RUSLE methods differ, it is necessary to unify the terminology to ensure accuracy. We confirm that the model used throughout this study for soil erosion estimation and analysis is the Revised Universal Soil Loss Equation (RUSLE). The expression “Modified Universal Soil Loss Equation (MUSLE)” in line 583 has been corrected to “RUSLE,” and a thorough review has been conducted to ensure consistent use of terminology throughout the manuscript.

 

  1. In section 4.2, Discussion, the figures, tables, and graphs should not be included. The discussion should interpret the results and compare them with findings from previous studies. Figure 11 and its accompanying description should be moved to the Results section, and only the interpretive discussion should remain here.

In accordance with the suggestion, we have revised the manuscript by moving Figure 11 and its descriptive content to the Results section. Only the explanatory discussion and analysis related to Figure 11 remain in the Discussion section, ensuring that the Discussion focuses on interpreting and elaborating the significance of the results.

 

Reviewer 2 Report

Comments and Suggestions for Authors

This study analyzes the characteristics and trends of soil erosion in Southwest China based on the RUSLE model and GIS methods. The findings are significant for accurately understanding regional soil erosion characteristics and risks. Specific revision suggestions are as follows:

  1. In the Introduction section, regarding the quantitative evaluation of soil erosion, the authors point out many shortcomings in current research. For example, key factors such as rainfall erosivity (R) and cover management (C) are often treated as static or annual values, failing to capture intra-annual variability (lines 62-63). Soil erosion assessments often neglect the interactions between natural and human systems, such as feedback loops between vegetation degradation, soil hardening, and runoff generation (lines 81-82), models have deficiencies in addressing subsurface hydrology issues within karst systems (lines 99-101). The influence of subsurface hydrology in karst ecosystems does not seem to be addressed in this study. I am confused about which specific shortcomings of existing research this study aims to address, the innovation points of this study are not explicitly stated in the Introduction. Based on this, I believe the authors should reorganize and revise the Introduction section.
  2. In Figure 1, the fonts in the two small maps on the left are too small to read clearly.
  3. Line 357-364: the rationale for comparing erosion conditions across different prefecture-level cities is unclear. While environmental protection strategies may differ between local governments, the subsequent Discussion section scarcely addresses the results of this comparison.
  4. Line 584-584: "higher in the northwest and lower in the southeast." This phrase is duplicated on line 606.
  5. Line 624: "in-creased"? (Suspected typo, likely "increased").
  6. Line 664: The results show that soil erosion risk in the study area has clearly decreased over the past 20 years, largely attributed to the implementation of ecological conservation practices. However, the authors' analysis predicts a future increase in soil erosion risk. The paper does not provide an in-depth explanation for this predicted outcome.
  7. In the Discussion, the authors propose many strategies for reducing soil erosion risk. Given that numerous ecological construction measures (such as the Grain for Green Program) have already been implemented, which specific measures are more critical for this region, considering its characteristics?

Author Response

This study analyzes the characteristics and trends of soil erosion in Southwest China based on the RUSLE model and GIS methods. The findings are significant for accurately understanding regional soil erosion characteristics and risks. Specific revision suggestions are as follows:

  1. In the Introduction section, regarding the quantitative evaluation of soil erosion, the authors point out many shortcomings in current research. For example, key factors such as rainfall erosivity (R) and cover management (C) are often treated as static or annual values, failing to capture intra-annual variability (lines 62-63). Soil erosion assessments often neglect the interactions between natural and human systems, such as feedback loops between vegetation degradation, soil hardening, and runoff generation (lines 81-82), models have deficiencies in addressing subsurface hydrology issues within karst systems (lines 99-101). The influence of subsurface hydrology in karst ecosystems does not seem to be addressed in this study. I am confused about which specific shortcomings of existing research this study aims to address, the innovation points of this study are not explicitly stated in the Introduction. Based on this, I believe the authors should reorganize and revise the Introduction section.

This paper clearly identifies the specific limitations of the current study and emphasizes the issues it aims to address, including the incorporation of the spatiotemporal dynamics of R and C factors, the use of machine learning techniques to reveal the coupled effects of natural and human drivers, and the prediction of future soil erosion trends under multiple climate scenarios. Meanwhile, considering the scope and data limitations of this study, the influence of groundwater hydrological processes in karst areas has not been included in the analysis, which has been acknowledged and explained in the manuscript.

 

  1. In Figure 1, the fonts in the two small maps on the left are too small to read clearly.

In response to the suggestion, we have enlarged the fonts in the two smaller maps on the left side of Figure 1 to ensure clarity and improve the readability of the legends.

 

  1. Line 357-364: the rationale for comparing erosion conditions across different prefecture-level cities is unclear. While environmental protection strategies may differ between local governments, the subsequent Discussion section scarcely addresses the results of this comparison.

Regarding the basis and explanation for comparing soil erosion conditions across different prefecture-level cities, the manuscript provides a detailed discussion in the Discussion section. This includes an analysis of driving factors and future climate scenario projections to elucidate the causes of regional differences in erosion intensity and propose corresponding response strategies. Specifically, the analysis highlights trends of intensified erosion in high-altitude and increased-precipitation areas, alongside recommendations for protective measures in varying terrain conditions such as steep slopes. Additionally, considering the indirect impacts of population growth, economic development, and land use changes on erosion, the study proposes integrated management approaches including optimizing land use structure, restricting construction expansion in high-risk zones, and promoting ecological agriculture.

 

  1. Line 584-584: "higher in the northwest and lower in the southeast." This phrase is duplicated on line 606.

Regarding the repetition of the phrase “higher in the northwest and lower in the southeast” in lines 584 and 606, the former appears in the Results section to describe specific research findings, while the latter is located in the Conclusion section to summarize the core conclusions of the paper. Although the expressions are similar, they serve different purposes and are considered a reasonable and necessary repetition to help readers clearly understand the research results and conclusions in their respective sections.

 

  1. Line 624: "in-creased"? (Suspected typo, likely "increased").

The spelling error “in-creased” in line 624 has indeed been identified and corrected to “increased.”

 

  1. Line 664: The results show that soil erosion risk in the study area has clearly decreased over the past 20 years, largely attributed to the implementation of ecological conservation practices. However, the authors' analysis predicts a future increase in soil erosion risk. The paper does not provide an in-depth explanation for this predicted outcome.

Regarding the difference between the decrease in soil erosion risk over the past 20 years and the projected increase in future risk, this study analyzed the positive effects of ecological protection measures implemented during the historical period, which have indeed significantly reduced soil erosion risk. The future projections, however, are based on scenario simulations under the background of climate change and do not account for potential ecological protection policies that may be implemented going forward. Therefore, the predicted increase in future soil erosion risk primarily reflects the potential impact of climate factors and does not negate the effectiveness of current protection measures.

 

  1. In the Discussion, the authors propose many strategies for reducing soil erosion risk. Given that numerous ecological construction measures (such as the Grain for Green Program) have already been implemented, which specific measures are more critical for this region, considering its characteristics?

In view of the extensive ecological engineering projects already implemented in the study area (such as returning farmland to forest), we have further supplemented the Discussion section by emphasizing the importance of site-specific and classified measures based on the region’s topography, climate, and land use characteristics. Specifically, in high-altitude and ecologically fragile zones with significant increases in precipitation, it is necessary to strengthen soil and water conservation infrastructure, including terraces, check dams, and vegetative slope protection. In steeper slope areas, priority should be given to terrace construction and slope stabilization projects to reduce slope erosion. These targeted protection measures, combined with existing ecological engineering efforts, will more effectively reduce soil erosion risks and enhance the sustainability of the regional ecological environment. Relevant content has been accordingly improved in the manuscript.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors of this paper have revised it according to the comments and suggestions provided by the reviewer, responding to each point individually. They have enhanced the paper by adding relevant references that address this topic and have improved the Abstract, Introduction, and Discussion sections. Additionally, they have included some graphical representations, which, in my opinion, strengthen the significance of the paper. I believe that the paper can be accepted for publication in its current form.

Author Response

Dear reviewer
Thank you very much for your feedback. We will continue to work hard to conduct research.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript is acceptable.

Author Response

Dear reviewer
Thank you very much for your feedback. We will continue to work hard to conduct research.

Back to TopTop