Landslide Susceptibility Mapping Using Remote Sensing Interpretation and a Blending-XGBoost-CNN Model

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study area should be presented on a map with topography and hydrography to provide a sense of its geographic configuration. In the Discussion section, the results should be compared with the work of more authors, particularly regarding the Interaction Detector function of the Geodetector. Moreover, it is important also considering the increase in risk factors when analyzing the variables together, because this notation, among other aspects presented in the paper, may be a key for understanding how to consider reproducing this work in other areas.

Author Response

Comments 1: The study area should be presented on a map with topography and hydrography to provide a sense of its geographic configuration.
Response 1: Figure 2 (Overview of the study area) has been added in the revised manuscript. This figure not only shows the location of the study area but also displays details such as its elevation and hydrographic features.

Comments 2: In the Discussion section, the results should be compared with the work of more authors, particularly regarding the Interaction Detector function of the Geodetector.
Response 2: In the Discussion section of the revised manuscript, the findings of Li et al. were discussed and compared with the results of this paper, as shown below: Li et al. [39] conducted a similar study in Yiyuan County, Zibo City, Shandong Province. The results showed that elevation and land use were the key determinants of landslide susceptibility; while SPI and PLC had almost no effect. These conclusions differed significantly from those obtained in this paper. The main reasons include: (1) The study area is located on the Loess Plateau, where vegetation coverage is generally low. Vegetation can intercept rainfall and reduce flow velocity. Therefore, slopes with higher NDVI are less likely to experience landslides, while the opposite is true for slopes with lower NDVI; (2) Loess has high porosity and well-developed vertical joints, allowing it to remain stable in a near-vertical state under natural conditions. However, when saturated with water, the structure of loess rapidly disintegrates, leading to significant changes in gradient.

Comments 3: Moreover, it is important also considering the increase in risk factors when analyzing the variables together, because this notation, among other aspects presented in the paper, may be a key for understanding how to consider reproducing this work in other areas.
Response 3: In the Discussion section of the revised manuscript, the interactions of hazard factors on landslide susceptibility were discussed as follows: The interactions between eleven pairs of hazard factors are BE, including (slope aspect, NDVI), (PLC, NDVI), (PLC, slope aspect), (distance from river, slope aspect), (distance from river, PLC), (distance from road, PLC), (lithology, TWI), (lithology, PLC), (lithology, distance from river), (lithology, distance from road), (land use, lithology). Among them, the interactions between lithology and the other five factors, as well as between PLC and the other five factors, are BE. Similarly, the interactions between slope aspect and the other three factors, and between distance from river and the other three factors, are also BE. This indicates that these four hazard factors have more complex triggering mechanisms for landslides. The interaction of hazard factors not only increases landslide risk but also lead to more complex disaster patterns. Therefore, when formulating landslide mitigation strategies, it is essential to fully consider the complex relationships among hazard factors. This approach enables more effective landslide prediction and prevention.

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents the landslide risk assessment based on remote sensing interpretation and the Blending-XGBoost-CNN model. The following comments are proposed:

• In the introduction:

o Add the reasons for choosing these models.

o Add the factors influencing landslides.

• Add a map showing the location of the study area in the Study Area section.
• Figures 4 and 5 are not clear.
• Figure 5 shows a multicolored scale, while the accompanying text describes only the green, red, and blue colors.
• Figures 6, 7, 8, and 9 are unnecessary.
• Describe the conditions for selecting the 6160 additional non-landslide grid cells.
• In Section 4.1 “Selection of hazard factors,” specify the basis on which these factors were selected.
• Present the STI, distance to faults, and profile curvature maps; after the correlation analysis, these can be removed if necessary.
• Move Table 3 to follow Figure 12.
• The section 4.4. Accuracy analysis, which describes the evaluation method, should be moved before the Results section.
• Add a more in-depth discussion—the current discussion mainly presents the study’s results. Discuss the factors controlling the highly susceptible zones.
• Present the interaction of factors involved in mapping the high-susceptibility zones.
• Add a discussion on the limitations of the method.

Author Response

Comments 1: Add the reasons for choosing these models.
Response 1: In the Discussion section of the revised manuscript, the reasons for choosing the XGBoost model, CNN model and Blending-XGBoost-CNN model were added as follows: Compared to the traditional Gradient Boosting Decision Tree (GBDT) algorithm, the XGBoost model utilizes a second-order Taylor expansion to simplify the calculation when solving the loss function and introduces regularization terms such as L1 and L2 into the objective function, resulting in higher computational efficiency and improved overfitting prevention capabilities [40]. CNN leverages the structural advantages of parameter sharing and local connectivity, enabling efficient processing of grid-like data without relying on additional features. Compared to traditional fully connected networks, CNN exhibits fewer parameters and stronger representational power when handling high-dimensional data. The Blending-XGBoost-CNN model integrates XGBoost model and CNN model through the Blending framework, which can enhance the overall performance [41]. The XGBoost model, CNN model and Blending-XGBoost-CNN model were constructed based on Python for LSM.

Comments 2: Add the factors influencing landslides.
Response 2: In the 4.1 (Selection of hazard factors) section of the revised manuscript, the factors influencing landslides were added as follows: The landslide disaster-pregnant environment involves topographic and geological factors (elevation, gradient, slope aspect, plane curvature (PLC), profile curvature (PRC), lithology, distance from fault), hydrological and vegetation factors (distance from river, topographic wetness index (TWI), sediment transport index (STI), stream power index (SPI), NDVI), and human activity factors (distance from road, land use) [33,34]. Among them, TWI quantifies the influence of terrain on hydrological processes, describing the degree of surface saturation runoff, which increases with contributing area accumulation. STI represents the comprehensive terrain variable for slope sediment transport, indicating the degree to which surface sands and other materials are transported by water flow. SPI is an important parameter that characterizes the surface water erosion capacity and is used to identify strong flow paths formed by water accumulation and locations prone to gully erosion [35,36].

Comments 3: Add a map showing the location of the study area in the Study Area section.

Response 3: Figure 2 (Overview of the study area) has been added in the revised manuscript. This figure not only shows the location of the study area but also displays details such as its elevation and hydrographic features.

Comments 4: Figures 4 and 5 are not clear.

Response 4: In the revised manuscript, Figures 4 and 5 have been renumbered as Figures 5 and 6. Both figures have been redrawn with significantly improved clarity.

Comments 5: Figure 5 shows a multicolored scale, while the accompanying text describes only the green, red, and blue colors.

Response 5: The figure has been redrawn, but there is indeed a color conflict between the main image and the legend. The reason for this has been explained in the revised manuscript, as follows: It should be specifically noted that Figure 5 employed a false color composite technique. The colors shown are not the true colors of the features but are instead generated by combining different bands (Red: Band_1, Green: Band_2, Blue: Band_3) to enhance the visualization of feature characteristics. When these three bands are synthesized in RGB mode, the differences in reflectance of various features across the bands are converted into combinations of red, green, and blue, resulting in mixed colors such as purple and green.

Comments 6: Figures 6, 7, 8, and 9 are unnecessary.

Response 6: Figures 6, 7, 8, and 9 have been removed from the revised manuscript.

Comments 7: Describe the conditions for selecting the 6160 additional non-landslide grid cells.

Response 7: The condition for selecting the 6160 additional non-landslide grid cells has been added in the revised manuscript as follows: The area of the 246 potential landslides is 0.427km², the volume is 2.161×10⁶m³. The largest potential landslide is the Zhangmaocai Landslide, with an area of 0.013km² and a volume of 8.332×10⁴m³. The DEM of the study area was resampled into 10m×10m grids using ArcGIS 10.2, and the number of potential landslide grids is 6,160. Since subsequent landslide susceptibility modeling requires both positive and negative samples, an additional 6,160 grids were randomly selected from non-landslide areas to serve as negative samples. The method for selecting non-landslide grids was as follows: set the elevation to no more than 1200m and the gradient to no more than 5°, then manually select non-landslide grids that are relatively distant from the landslide grids from the eligible grids.

Comments 8: In Section 4.1 “Selection of hazard factors,” specify the basis on which these factors were selected.

Response 8: In the revised manuscript, the theoretical basis for selecting the hazard factors has been added as follows: The landslide disaster-pregnant environment involves topographic and geological factors (elevation, gradient, slope aspect, plane curvature (PLC), profile curvature (PRC), lithology, distance from fault), hydrological and vegetation factors (distance from river, topographic wetness index (TWI), sediment transport index (STI), stream power index (SPI), NDVI), and human activity factors (distance from road, land use) [33,34]. Among them, TWI quantifies the influence of terrain on hydrological processes, describing the degree of surface saturation runoff, which increases with contributing area accumulation. STI represents the comprehensive terrain variable for slope sediment transport, indicating the degree to which surface sands and other materials are transported by water flow. SPI is an important parameter that characterizes the surface water erosion capacity and is used to identify strong flow paths formed by water accumulation and locations prone to gully erosion [35,36].

Comments 9: Move Table 3 to follow Figure 12.

Response 9: In the revised manuscript, Table 3 has been moved to its corresponding position.

Comments 10: The section 4.4. Accuracy analysis, which describes the evaluation method, should be moved before the Results section.

Response 10: In the revised manuscript, the sequence of Sections 4.3 and 4.4 has been extensively rearranged as requested.

Comments 11: Add a more in-depth discussion—the current discussion mainly presents the study’s results. Discuss the factors controlling the highly susceptible zones.

Response 11: In the revised manuscript, the factors controlling the highly susceptible zones have been added as follows: NDVI and gradient are key factors determining landslide occurrence; elevation, slope aspect, distance from river and land use also play significant roles in landslide occurrence; the contributions of TWI and lithology to landslide occurrence are relatively small; PLC and distance from road have q values close to 0, indicating their influences on the landslide occurrence are minimal.

Li et al. [39] conducted a similar study in Yiyuan County, Zibo City, Shandong Province. The results showed that elevation and land use were the key determinants of landslide susceptibility; while SPI and PLC had almost no effect. These conclusions differed significantly from those obtained in this paper. The main reasons include: (1) The study area is located on the Loess Plateau, where vegetation coverage is generally low. Vegetation can intercept rainfall and reduce flow velocity. Therefore, slopes with higher NDVI are less likely to experience landslides, while the opposite is true for slopes with lower NDVI; (2) Loess has high porosity and well-developed vertical joints, allowing it to remain stable in a near-vertical state under natural conditions. However, when saturated with water, the structure of loess rapidly disintegrates, leading to significant changes in gradient.

Comments 12: Present the interaction of factors involved in mapping the high-susceptibility zones.

Response 12: In the revised manuscript, the interaction of factors has been analyzed as follows: The interactions between eleven pairs of hazard factors are BE, including (slope aspect, NDVI), (PLC, NDVI), (PLC, slope aspect), (distance from river, slope aspect), (distance from river, PLC), (distance from road, PLC), (lithology, TWI), (lithology, PLC), (lithology, distance from river), (lithology, distance from road), (land use, lithology). Among them, the interactions between lithology and the other five factors, as well as between PLC and the other five factors, are BE. Similarly, the interactions between slope aspect and the other three factors, and between distance from river and the other three factors, are also BE. This indicates that these four hazard factors have more complex triggering mechanisms for landslides. The interaction of hazard factors not only increases landslide risk but also lead to more complex disaster patterns. Therefore, when formulating landslide mitigation strategies, it is essential to fully consider the complex relationships among hazard factors. This approach enables more effective landslide prediction and prevention.

Comments 13: Add a discussion on the limitations of the method.

Response 13: In the Conclusion section of the revised manuscript, the limitations of this study have been added as follows: (4) This study utilized remote sensing interpretation to obtain potential landslides and conducted LSM based on the XGBoost model, CNN model and Blending-XGBoost- CNN model, respectively. The results can provide a theoretical basis for landslide prevention-oriented land use planning. However, further improvements can be made in the following aspects: (1) With the advancement of artificial intelligence technology, advanced brain-like neural network models can offer more approaches for LSM, and there is still significant room for improving the accuracy of the results; (2) This study analyzed the interaction of hazard factors on landslide susceptibility and categorized them into NE and BE. However, this research was still in its early stages, and the specific mechanisms of influence remain unclear; (3) Although the method proposed in this paper demonstrated high applicability to the study area, the limited size of the area and its unique geological characteristics mean that the applicability of the method to other regions with distinct geological features remains unclear. The authors intend to conduct further in-depth researches on these topics in the future.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents a comprehensive study integrating SBAS-InSAR and object-oriented classification for identifying potential landslides, followed by susceptibility mapping using a novel Blending-XGBoost-CNN model. The topic is relevant and timely, aligning with the increasing application of remote sensing and hybrid machine learning in geohazard assessment. The structure is generally clear, and the methodology appears systematic. However, the paper lacks depth in scientific interpretation, validation rigor, and discussion of limitations. The novelty, while claimed, is only moderately supported by evidence.

The following comments aim to help the authors strengthen the technical quality and scientific robustness of their work:

• Abstract: Too long and partially redundant. It should be condensed, emphasizing the problem, proposed method, and main quantitative results (AUC, Precision, % area of each susceptibility class);
• Introduction: Some cited works are outdated or loosely related (e.g., references on UAV-based detection). Include more recent studies on hybrid ML–DL landslide models from 2023–2025 to contextualize your contribution.

Novelty: The paper claims novelty by integrating XGBoost and CNN through a Blending approach, yet similar hybrid frameworks have been explored in recent years (e.g., DBN-MLP, CNN-LSTM, RF-XGBoost). The authors should explicitly state how this hybridization advances prior models — for instance, in interpretability, convergence, or feature extraction capacity. The novelty could be better emphasized through a comparative performance analysis with other hybrid or ensemble models.

• Data: The SBAS-InSAR data processing section is extensive but lacks quantitative quality assessment. Indicators such as coherence values, standard deviation of residual phase, or deformation accuracy (mm/year) should be reported to validate the reliability of the deformation results. The paper states that 246 potential landslides were confirmed through field validation, but no information is provided on the validation protocol — number of ground truth points, GPS accuracy, or uncertainty in labeling;
• Models: The CNN architecture is described in general terms (one convolutional and one pooling layer), but critical hyperparameters (kernel size, stride, learning rate, optimizer, batch size, epochs) are missing. Similarly, XGBoost parameter tuning (number of trees, depth, learning rate, regularization parameters γ and λ) is not documented. Authors should include a summary table of model parameters and training conditions;
• Results: A deeper geotechnical interpretation — linking soil type, infiltration, and hydrological responses to landslide occurrence will be interesting;
• Conclusions: The study area (Jingchuan County) is relatively small (≈1,464 km²). The authors should discuss the model’s transferability to other regions with different geomorphologic and climatic conditions;
• Figures: Improve figure clarity and ensure that all visuals support the text effectively;
• The reference formatting does not comply with the Applied Sciences template. In the current version, in-text citations are not properly numbered according to the journal’s referencing style (e.g., [1], [2], [3], …). Instead, the manuscript presents author–year citations or incomplete references without numerical indicators, which is inconsistent with the MDPI referencing guidelines.

Author Response

Comment 1: Abstract: Too long and partially redundant. It should be condensed, emphasizing the problem, proposed method, and main quantitative results (AUC, Precision, % area of each susceptibility class);

Response 1: The Abstract was rewritten in the revised manuscript as follows: The accuracy of historical landslide data is a key factor affecting the precision of landslide susceptibility mapping. The degree of conformity between mathematical models and disaster-pregnant environment cannot be predetermined, and the optimal model needs to be determined through comparative studies. In this paper, SBAS-InSAR and object-oriented classification method were integrated to provide data for landslide susceptibility mapping: SBAS-InSAR was used to process Sentinel-1 images, while object-oriented classification method was applied to interpret Landsat 8 images. Eleven hazard factors were selected for landslide susceptibility modeling, and the best-performing model was determined. The influences of single and multiple hazard factors on landslide susceptibility were analyzed using Geodetector. The results showed that 246 potential landslides were identified, with the total area was 0.427km², the total volume was 2.161×10⁶m³. The Blending-XGBoost-CNN model achieved the highest AUC and Precision, outperforming the XGBoost model and CNN model. The extreme high susceptible areas, high susceptible areas, moderate susceptible areas, minor susceptible areas and extreme minor susceptible areas accounted for 6.24% (91.4km²), 15.07% (220.6km²), 29.15% (426.8km²), 30.58% (447.7km²) and 18.96% (277.8km²) of the total area, respectively. NDVI and gradient were key factors determining landslide occurrence. Elevation, slope aspect, distance from river and land use played significant roles in landslide occurrence. The contributions of TWI and lithology to landslide occurrence were relatively small, while those of plane curvature and distance from road were minimal. The interaction of hazard factors exhibited NE or BE relationships, not only increasing landslide risk but also potentially leading to more complex disaster patterns. This study can provide a theoretical basis for landslide prevention-oriented land use planning.

Comment 2: Introduction: Some cited works are outdated or loosely related (e.g., references on UAV-based detection). Include more recent studies on hybrid ML–DL landslide models from 2023–2025 to contextualize your contribution.

Response 2: The Introduction section has been substantially revised in the revised manuscript, with a large number of irrelevant or outdated references removed and several highly relevant references from the last two years added, as shown below:

Choi et al. [13] presented a case study approach to fully leverage variety of multi-source remote sensing technologies for analyzing the characteristics of a landslide. The chosen multi-source technologies encompassed digital photogrammetry using RGB and multi-spectral imageries, 3D point clouds acquired by light detection and ranging (LiDAR) mounted on UAV, and InSAR. Handwerger et al. [14] revisited the 2017 Mud Creek landslide in California using radar interferometry, pixel tracking, and elevation change measurements from satellite and airborne radar, LiDAR, and optical data. The results showed that pixel tracking of optical imagery captured the transition from slow motion to runaway acceleration starting-1 month before catastrophic failure-an acceleration undetected by satellite InSAR alone.

1. Longoni, L.; Scaioli, A.; Panzeri, L.; Arosio, D.; Corti, M.; Hojat, A.; Papini, M. A new Landslide Investigation and Simulation Archive through downscaled landslide experiments. Scientific Data 2025, 12, 1668.
2. Li, Y.C.; Chen, J.P.; Tan, C.; Li, Z.H.; Zhang, Y.S.; Yan, J.H. Deformation and potential failure analysis of a giant old deposit in the southeastern margin of the Qinghai-Tibet Plateau based on SBAS-InSAR and numerical simulation. Bulletin of Engineering Geology and the Environment 2023, 82, 58.
3. Zheng, X.S.; Lu, W.J.; Jiang, R.C.; Li, J.H.; Zhang, L.M. Analysis of landslide on Meizhou-Dapu expressway based on satellite remote sensing. Geoenvironmental Disasters 2025, 12, 25.
4. Choi, S.K.; Ranirez, R.A.; Lim, H.H.; Kwon, T.H. Multi-source remote sensing-based landslide investigation: the case of the August 7, 2020, Gokseong landslide in South Korea. Scientific Reports 2024, 14, 12048.
5. Handwerger, A.L.; Lacroix, P.; Bell, A.F.; Booth, A.M.; Huang, M.H.; Mudd, S.M.; Bürgmann, R.; Fielding, E.J. Multi-sensor remote sensing captures geometry and slow-to-fast sliding transition of the 2017 Mud Creek landslide. Scientific Reports 2025, 15, 29831.
6. Yang, X.W.; Chen, D.N.; Dong, Y.H.; Xue, Y.M.; Qin, K.X. Identification of potential landslide in Jianzha county based on InSAR and deep learning. Scientific Reports 2024, 14, 21346.
7. Mao, X.C.; Liu, P.; Deng, H.; Liu, Z.K.; Li, L.J.; Wang, Y.S.; Ai, Q.X.; Liu, J.X. A Novel Approach to Three-Dimensional Inference and Modeling of Magma Conduits with Exploration Data: A Case Study from the Jinchuan Ni-Cu Sulfide Deposit, NW China. Natural Resources Research 2023, 32, 901-928.
8. Sun, D.; Zhang, C.Y.; Zhang, S.A.; Xu, J.D.; Tao, Z.G.; Wu, M.L.; Li, D.X.; Zhang, G.H.; Liu, Y.P.; Wang, F.N.; He, M.C. Hydraulic Fracturing In-Situ Stress Measurements and Large Deformation Evaluation of 1000m-Deep Soft Rock Roadway in Jinchuan No. 2 Mine, Northwestern China. Rock Mechanics and Rock Engineering 2025, 58, 2781-2801.
9. Chen, W.; Wang, J.L.; Xie, X.S.; Hong, H.Y.; Trung, N.V.; Bui, D.T.; Wang, G.; Li, X.R. Spatial prediction of landslide susceptibility using integrated frequency ratio with entropy and support vector machines by different kernel functions. Environmental Earth Sciences 2016, 75, 1344.
10. Ijaz, Z.; Zhao, C.; Ijaz, N.; Rehman, Z.; Ijaz, A. Novel application of Google earth engine interpolation algorithm for the development of geotechnical soil maps: a case study of mega-district. Geocarto International 2022, 37, 18196-18216.
11. Wang, Y.; Zhou, C.; Cao, Y.; Meena, S.R.; Feng, Y.; Wang, Y. Utilizing deep learning approach to develop landslide susceptibility mapping considering landslide types. Bulletin of Engineering Geology and the Environment 2024, 83, 430.
12. Huang, F.M.; Yang, Y.; Jiang, B.C.; Chang, Z.L.; Zhou, C.B.; Jiang, S.H.; Huang, J.S.; Catani, F.; Yu, C.S. Effects of different division methods of landslide susceptibility levels on regional landslide susceptibility mapping. Bulletin of Engineering Geology and the Environment 2025, 84, 276.
13. Li, Z.B.; Yin, C.; Tan, Z.Y.; Liu, X.L.; Li, S.F.; Ma, X.B.; Zhang, X.X. Landslide Susceptibility Assessment Considering Time-Varying of Dynamic Factors. Natural Hazards Review 2024, 25, 05024004.
14. Ijaz, Z.; Zhao, C.; Ijaz, N.; Rehman, Z.; Ijaz, A. Development and optimization of geotechnical soil maps using various geostatistical and spatial interpolation techniques: a comprehensive study. Bulletin of Engineering Geology and the Environment 2023, 82, 215.
15. Wang, P.; Deng, H.W.; Li, Y.Y.; Pan, Z.; Peng, T. Enhancing landslide susceptibility modelling through predicted InSAR deformation rates. Environmental Earth Sciences, 2025, 84, 347.
16. Wang, P.; Deng, H.W. The impact of different sampling strategies on landslide susceptibility assessment: an explainable hybrid BO-XGBoost model. Earth Science Informatics 2025, 18, 440.
17. Yuan, H.P.; Ji, S.J.; Li, H.Z.; Zhu, C.Q.; Zou, Y.Y.; Ni, B.; Gu, Z.A. Classification forecasting research of rock burst intensity based on the BO-XGBoost-Cloud model. Earth Science Informatics 2025, 18, 95.
18. Chen, J.G.; Yin, C.; Sun, T.Q.; Li, J.X. Seismic Damage Risk Assessment of Reinforced Concrete Bridges Considering Structural Parameter Uncertainties. Coatings 2025, 15, 1242.
19. Liu, L.C.; Wang, P.Q.; Su, L.B.; Li, F. Landslide data sample augmentation and landslide susceptibility analysis in Nyingchi City based on the MCMC model. Scientific Reports 2025, 15, 25624.

Comment 3: Results: A deeper geotechnical interpretation — linking soil type, infiltration, and hydrological responses to landslide occurrence will be interesting;

Response 3: The following content has been added to the revised manuscript as requested:

Li et al. [39] conducted a similar study in Yiyuan County, Zibo City, Shandong Province. Their results indicated that elevation and land use were the key determinants of landslide susceptibility, while SPI and PLC demonstrated negligible effects. These conclusions differ significantly from the findings presented in this paper. The main reasons for these discrepancies include: (1) The study area is situated on the Loess Plateau, characterized by generally sparse vegetation coverage. Vegetation plays a crucial role in intercepting rainfall and reducing flow velocity. Consequently, slopes with higher NDVI values exhibit lower landslide susceptibility, whereas the opposite applies to slopes with lower NDVI values; (2) Loess possesses high porosity and well-developed vertical joints, enabling it to maintain stability in near-vertical slopes under natural conditions. However, when saturated with water, the loess structure rapidly disintegrates, resulting in significant gradient alterations.

Eleven pairs of hazard factors demonstrate bivariate enhancement (BE) in their interactions, including: (slope aspect, NDVI), (PLC, NDVI), (PLC, slope aspect), (distance from river, slope aspect), (distance from river, PLC), (distance from road, PLC), (lithology, TWI), (lithology, PLC), (lithology, distance from river), (lithology, distance from road), and (land use, lithology). Notably, the interactions between lithology and the other five factors, as well as between PLC and the other five factors, all exhibit BE characteristics. Similarly, the interactions between slope aspect and the other three factors, and between distance from river and the other three factors, also demonstrate BE patterns. This indicates that these four hazard factors possess more complex landslide-triggering mechanisms. The interactions among hazard factors not only elevate landslide risks but also lead to more intricate disaster patterns. Therefore, when developing landslide mitigation strategies, it is essential to comprehensively consider the complex interrelationships among hazard factors. This approach will enable more effective landslide prediction and prevention measures.

Comment 4: Conclusions: The study area (Jingchuan County) is relatively small (≈1,464 km²). The authors should discuss the model’s transferability to other regions with different geomorphologic and climatic conditions;

Response 4: In the Conclusion section of the revised manuscript, the limitations of this study have been added as follows: (4) This study utilized remote sensing interpretation to obtain potential landslides and conducted LSM based on the XGBoost model, CNN model and Blending-XGBoost- CNN model, respectively. The results can provide a theoretical basis for landslide prevention-oriented land use planning. However, further improvements can be made in the following aspects: (1) With the advancement of artificial intelligence technology, advanced brain-like neural network models can offer more approaches for LSM, and there is still significant room for improving the accuracy of the results; (2) This study analyzed the interaction of hazard factors on landslide susceptibility and categorized them into NE and BE. However, this research was still in its early stages, and the specific mechanisms of influence remain unclear; (3) Although the method proposed in this paper demonstrated high applicability to the study area, the limited size of the area and its unique geological characteristics mean that the applicability of the method to other regions with distinct geological features remains unclear. The authors intend to conduct further in-depth researches on these topics in the future.

Comment 5: Figures: Improve figure clarity and ensure that all visuals support the text effectively;

Response 5: In the revised manuscript, Figure 2 has been added to show the location, topography, and hydrographic network of the study area. Several unnecessary figures have been removed to make the paper more concise, and originally unclear graphics have been redrawn.

Comment 6: The reference formatting does not comply with the Applied Sciences template. In the current version, in-text citations are not properly numbered according to the journal’s referencing style (e.g., [1], [2], [3], …). Instead, the manuscript presents author–year citations or incomplete references without numerical indicators, which is inconsistent with the MDPI referencing guidelines.

Response 6: In the revised manuscript, the formatting of the references has been fully revised according to the requirements.

Reviewer 4 Report

Comments and Suggestions for Authors

The study presents an interesting integration of SBAS-InSAR and object-oriented classification for landslide identification combined with a Blending-XGBoost-CNN model. However, the novelty compared to existing hybrid deep learning or ensemble models (e.g., Stacking, DBN-MLP, or CNN-LSTM approaches) is not clearly highlighted.

The authors should explicitly discuss what technical advancement or methodological innovation distinguishes this work from recent studies in 2023–2024.The authors should condense the general software-based steps and focus more on how model hyperparameters were optimized and validated.

The authors should incorporate cross-validation, ROC comparison significance testing, or alternative evaluation metrics (e.g., F1-score, Kappa, RMSE).

The reference list is extensive but leans heavily on 2023–2025 Chinese case studies. To strengthen the paper’s global relevance, include more recent international works on hybrid landslide susceptibility modeling. Also, Study missing latest work on spatial modeling consider citing https://doi.org/10.1007/s10064-023-03244-x and https://doi.org/10.1080/10106049.2022.2138566 while discussing mapping.

The authors should include a more physically grounded explanation linking these findings with slope stability and geomorphic processes.

 

 

Author Response

Comment 1: The authors should incorporate cross-validation, ROC comparison significance testing, or alternative evaluation metrics (e.g., F1-score, Kappa, RMSE).

Response 1: Several new metrics have been added to Table 5 of the revised manuscript, as shown below:

Table 5. AUC-related statistical values.

Model	AUC	Standard error	Asymptotic significance	Approaching the 95% confidence interval
				Lower limit	Limit
XGBoost model	0.882	0.021	0.000	0.814	0.898
CNN model	0.900	0.017	0.000	0.862	0.941
Blending-XGBoost-CNN model	0.912	0.016	0.000	0.883	0.947

Comment 2: The reference list is extensive but leans heavily on 2023–2025 Chinese case studies. To strengthen the paper’s global relevance, include more recent international works on hybrid landslide susceptibility modeling. Also, Study missing latest work on spatial modeling consider citing https://doi.org/10.1007/s10064-023-03244-x and https://doi.org/10.1080/10106049.2022.2138566 while discussing mapping.

Response 2: As requested, extensive revisions have been made to the References section in the revised manuscript. Irrelevant and outdated literatures have been removed, and in addition to the two literatures suggested by the reviewers, numerous other relevant literatures have been added, as shown below.

1. Longoni, L.; Scaioli, A.; Panzeri, L.; Arosio, D.; Corti, M.; Hojat, A.; Papini, M. A new Landslide Investigation and Simulation Archive through downscaled landslide experiments. Scientific Data 2025, 12, 1668.
2. Li, Y.C.; Chen, J.P.; Tan, C.; Li, Z.H.; Zhang, Y.S.; Yan, J.H. Deformation and potential failure analysis of a giant old deposit in the southeastern margin of the Qinghai-Tibet Plateau based on SBAS-InSAR and numerical simulation. Bulletin of Engineering Geology and the Environment 2023, 82, 58.
3. Zheng, X.S.; Lu, W.J.; Jiang, R.C.; Li, J.H.; Zhang, L.M. Analysis of landslide on Meizhou-Dapu expressway based on satellite remote sensing. Geoenvironmental Disasters 2025, 12, 25.
4. Choi, S.K.; Ranirez, R.A.; Lim, H.H.; Kwon, T.H. Multi-source remote sensing-based landslide investigation: the case of the August 7, 2020, Gokseong landslide in South Korea. Scientific Reports 2024, 14, 12048.
5. Handwerger, A.L.; Lacroix, P.; Bell, A.F.; Booth, A.M.; Huang, M.H.; Mudd, S.M.; Bürgmann, R.; Fielding, E.J. Multi-sensor remote sensing captures geometry and slow-to-fast sliding transition of the 2017 Mud Creek landslide. Scientific Reports 2025, 15, 29831.
6. Yang, X.W.; Chen, D.N.; Dong, Y.H.; Xue, Y.M.; Qin, K.X. Identification of potential landslide in Jianzha county based on InSAR and deep learning. Scientific Reports 2024, 14, 21346.
7. Mao, X.C.; Liu, P.; Deng, H.; Liu, Z.K.; Li, L.J.; Wang, Y.S.; Ai, Q.X.; Liu, J.X. A Novel Approach to Three-Dimensional Inference and Modeling of Magma Conduits with Exploration Data: A Case Study from the Jinchuan Ni-Cu Sulfide Deposit, NW China. Natural Resources Research 2023, 32, 901-928.
8. Sun, D.; Zhang, C.Y.; Zhang, S.A.; Xu, J.D.; Tao, Z.G.; Wu, M.L.; Li, D.X.; Zhang, G.H.; Liu, Y.P.; Wang, F.N.; He, M.C. Hydraulic Fracturing In-Situ Stress Measurements and Large Deformation Evaluation of 1000m-Deep Soft Rock Roadway in Jinchuan No. 2 Mine, Northwestern China. Rock Mechanics and Rock Engineering 2025, 58, 2781-2801.
9. Chen, W.; Wang, J.L.; Xie, X.S.; Hong, H.Y.; Trung, N.V.; Bui, D.T.; Wang, G.; Li, X.R. Spatial prediction of landslide susceptibility using integrated frequency ratio with entropy and support vector machines by different kernel functions. Environmental Earth Sciences 2016, 75, 1344.
10. Ijaz, Z.; Zhao, C.; Ijaz, N.; Rehman, Z.; Ijaz, A. Novel application of Google earth engine interpolation algorithm for the development of geotechnical soil maps: a case study of mega-district. Geocarto International 2022, 37, 18196-18216.
11. Wang, Y.; Zhou, C.; Cao, Y.; Meena, S.R.; Feng, Y.; Wang, Y. Utilizing deep learning approach to develop landslide susceptibility mapping considering landslide types. Bulletin of Engineering Geology and the Environment 2024, 83, 430.
12. Huang, F.M.; Yang, Y.; Jiang, B.C.; Chang, Z.L.; Zhou, C.B.; Jiang, S.H.; Huang, J.S.; Catani, F.; Yu, C.S. Effects of different division methods of landslide susceptibility levels on regional landslide susceptibility mapping. Bulletin of Engineering Geology and the Environment 2025, 84, 276.
13. Li, Z.B.; Yin, C.; Tan, Z.Y.; Liu, X.L.; Li, S.F.; Ma, X.B.; Zhang, X.X. Landslide Susceptibility Assessment Considering Time-Varying of Dynamic Factors. Natural Hazards Review 2024, 25, 05024004.
14. Ijaz, Z.; Zhao, C.; Ijaz, N.; Rehman, Z.; Ijaz, A. Development and optimization of geotechnical soil maps using various geostatistical and spatial interpolation techniques: a comprehensive study. Bulletin of Engineering Geology and the Environment 2023, 82, 215.
15. Wang, P.; Deng, H.W.; Li, Y.Y.; Pan, Z.; Peng, T. Enhancing landslide susceptibility modelling through predicted InSAR deformation rates. Environmental Earth Sciences, 2025, 84, 347.
16. Wang, P.; Deng, H.W. The impact of different sampling strategies on landslide susceptibility assessment: an explainable hybrid BO-XGBoost model. Earth Science Informatics 2025, 18, 440.
17. Yuan, H.P.; Ji, S.J.; Li, H.Z.; Zhu, C.Q.; Zou, Y.Y.; Ni, B.; Gu, Z.A. Classification forecasting research of rock burst intensity based on the BO-XGBoost-Cloud model. Earth Science Informatics 2025, 18, 95.
18. Chen, J.G.; Yin, C.; Sun, T.Q.; Li, J.X. Seismic Damage Risk Assessment of Reinforced Concrete Bridges Considering Structural Parameter Uncertainties. Coatings 2025, 15, 1242.
19. Liu, L.C.; Wang, P.Q.; Su, L.B.; Li, F. Landslide data sample augmentation and landslide susceptibility analysis in Nyingchi City based on the MCMC model. Scientific Reports 2025, 15, 25624.

Comment 3: The authors should include a more physically grounded explanation linking these findings with slope stability and geomorphic processes.

Response 3: The following content has been added to the revised manuscript as requested:

Li et al. [39] conducted a similar study in Yiyuan County, Zibo City, Shandong Province. Their results indicated that elevation and land use were the key determinants of landslide susceptibility, while SPI and PLC demonstrated negligible effects. These conclusions differ significantly from the findings presented in this paper. The main reasons for these discrepancies include: (1) The study area is situated on the Loess Plateau, characterized by generally sparse vegetation coverage. Vegetation plays a crucial role in intercepting rainfall and reducing flow velocity. Consequently, slopes with higher NDVI values exhibit lower landslide susceptibility, whereas the opposite applies to slopes with lower NDVI values; (2) Loess possesses high porosity and well-developed vertical joints, enabling it to maintain stability in near-vertical slopes under natural conditions. However, when saturated with water, the loess structure rapidly disintegrates, resulting in significant gradient alterations.

Eleven pairs of hazard factors demonstrate bivariate enhancement (BE) in their interactions, including: (slope aspect, NDVI), (PLC, NDVI), (PLC, slope aspect), (distance from river, slope aspect), (distance from river, PLC), (distance from road, PLC), (lithology, TWI), (lithology, PLC), (lithology, distance from river), (lithology, distance from road), and (land use, lithology). Notably, the interactions between lithology and the other five factors, as well as between PLC and the other five factors, all exhibit BE characteristics. Similarly, the interactions between slope aspect and the other three factors, and between distance from river and the other three factors, also demonstrate BE patterns. This indicates that these four hazard factors possess more complex landslide-triggering mechanisms. The interactions among hazard factors not only elevate landslide risks but also lead to more intricate disaster patterns. Therefore, when developing landslide mitigation strategies, it is essential to comprehensively consider the complex interrelationships among hazard factors. This approach will enable more effective landslide prediction and prevention measures.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The suggested revisions have been implemented; only one minor comment remains to be addressed :

The text « The ROC curves for the XGBoost model, CNN model and Blending-XGBoost-CNN model are shown in Figure 12, and the AUC-related statistical values are shown in Table 5. The Precisions for the XGBoost model, CNN model and Blending-XGBoost-CNN model are 0.77, 0.79 and 0.83, respectively. The Blending-XGBoost-CNN model has the highest AUC and Precision, indicating that this model outperforms the XGBoost model and CNN model. Therefore, it can be promoted and applied in landslide monitoring and early warning systems », Figure 12. ROC curves and Table 5. AUC-related statistical values should be moved to follow Section 4.4. LSM results.

Author Response

Comment: Figure 12. ROC curves and Table 5. AUC-related statistical values should be moved to follow Section 4.4. LSM results.

Response: The manuscript has been revised according to the reviewers' comments, with Sections 4.3 and 4.4 merged to improve the overall structure of the paper.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have satisfactorily addressed all previous comments and substantially improved the manuscript. The revisions are clear and consistent with the suggestions. I am satisfied with the current version and have no further comments. The manuscript is now suitable for publication.

Author Response

Comment: The authors have satisfactorily addressed all previous comments and substantially improved the manuscript. The revisions are clear and consistent with the suggestions. I am satisfied with the current version and have no further comments. The manuscript is now suitable for publication.

Response: Thank you for your efforts in the revision process of this article, and also for recommending its acceptance.

Reviewer 4 Report

Comments and Suggestions for Authors

Accept in present form

Author Response

Comment: Accept in present form.

Response: Thank you for your efforts in the revision process of this article, and also for recommending its acceptance.