Hybrid Integration of Bagging and Decision Tree Algorithms for Landslide Susceptibility Mapping

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript's authors propose a hybrid ensemble-based methodological approach to evaluate different factors to produce a landslide susceptibility map for Yanchuan County within the hilly and gully region of China's Loess Plateau using 16 different geological and environmental indicators. The appropriate ones were selected for inclusion in 4 tested models based on consistent assessment and statistical correlation analysis. The study shows that the highest degree of confidence in the modeling is achieved by the Bagging-enhanced FT (BFT) model. The proposed holistic approach and conducted analysis contribute to increasing knowledge in hazard analysis and risk assessment in landslide areas in different locations around the world.

The reviewer has comments and questions for the authors, which are as follows:

· The study mainly used factors related to surface geomorphological features rather than geological ones, including tectonic features. No factors related to human activity were included in the study. Do they influence the activation of the landslide localities in the specific area of China's Loess Plateau that is studied?

· It is recommended to specify which geospatial transformation functions in the ArcGIS environment are used to combine, categorize, and classify and are used to prepare the different layers. How do the authors process heterogeneous and multi-scale data in different formats and with different precisions, particularly the so-called discontinuous and continuous data, to display them in a grid with a spatial resolution of 30x30m?

· The color scales used when drawing individual layers in a GIS environment do not allow to discern the difference in the min-max boundaries of the depicted factors; more specifically, this applies to figures 3d, e, f, g, h, i, and p. The accuracy with which the geographical coordinates are written on the figures is sufficient to be up to a degree minutes, namely Latitude: ° ', Longitude: ° '.

·           From the results presented in Table 1 regarding the predictive ability and importance of factors related to landslides, it is strange why geological factors such as lithology and soil type have a low Average Merit. Is there any reasonable explanation for such inferences based on the performed analysis?

 

· Of the four tree-based models tested, namely REPTree, BREPTree, FT, and BFT, it was found that for the study area of Yanchuan County, the BFT model provided the most accurate prediction of landslide susceptibility. What explanation can be given for the fact that none of the models exceeded 90% of the F-score in Table 6. do the values of the individual quantities given in this table have units?

Comments on the Quality of English Language

 ·   In some places in the text, some expressions could be redacted, such as in Line 174-175.

·   The explanation of acronyms of the terms used is given when they first appear in the manuscript text.

Author Response

Response letter to Reviewer 1

Dear reviewer, the following paragraphs include a point–to–point response to your comments and suggestions.

The manuscript's authors propose a hybrid ensemble-based methodological approach to evaluate different factors to produce a landslide susceptibility map for Yanchuan County within the hilly and gully region of China's Loess Plateau using 16 different geological and environmental indicators. The appropriate ones were selected for inclusion in 4 tested models based on consistent assessment and statistical correlation analysis. The study shows that the highest degree of confidence in the modeling is achieved by the Bagging-enhanced FT (BFT) model. The proposed holistic approach and conducted analysis contribute to increasing knowledge in hazard analysis and risk assessment in landslide areas in different locations around the world.

The reviewer has comments and questions for the authors, which are as follows:

 

Comment 1:

The study mainly used factors related to surface geomorphological features rather than geological ones, including tectonic features. No factors related to human activity were included in the study. Do they influence the activation of the landslide localities in the specific area of China's Loess Plateau that is studied?

 

Response to comment 1:

Dear reviewer, thank you for your comment. In response to your comment, we have added the following in the discussion section:

 

While our study has concentrated on surface geomorphological features as primary indicators for landslide susceptibility in China's Loess Plateau, we recognize that human activities can significantly influence the activation of landslides. It is well-documented that alterations in land use and land cover, such as deforestation, construction, and agricultural practices, can destabilize slopes and precipitate landslide events. In our analysis, land use and land cover data and distance from road network were utilized, however, more detailed aspects of human activity, including urban development, infrastructure expansion, and intensive agriculture, were not explicitly included. We acknowledge that this is an important dimension of landslide susceptibility, and that the exclusion of detailed human activity-related factors could be seen as a limitation of the current study. Future research should aim to integrate these factors to provide a more holistic understanding of the conditions leading to landslide occurrences. Incorporating detailed data on human activities will enhance the accuracy of predictive models and contribute to more effective risk management and mitigation strategies in regions similar to Yanchuan County.

Comment 2:

It is recommended to specify which geospatial transformation functions in the ArcGIS environment are used to combine, categorize, and classify and are used to prepare the different layers. How do the authors process heterogeneous and multi-scale data in different formats and with different precisions, particularly the so-called discontinuous and continuous data, to display them in a grid with a spatial resolution of 30x30m?

 

Response to comment 2:

Dear reviewer, thank you for your comment and suggestion. We acknowledge the lack of detail information concerning the preparation of the landslide related variables. Our response is as follows:

 

In more detail, from the DEM file (resolution 30x30m) obtained by the International Scientific & Technical Data Mirror Site of the Computer Network Information Center, Chinese Academy of Sciences, available at http://www.gscloud.cn., slope, aspect, plan and profile curvature, slope length, TPI, TRI and CI were constructed using the ArcGIS suite and the Spatial Analyst Tools toolbox. For these continuous raster files, the Slope length, TPI, TRI, and CI were reclassified into five classes using the natural break classification method having the same resolution (30x30) as the DEM file. For slope, aspect, elevation, plan and profile curvature, distance from river and distance from roads as well as rainfall, we applied a manual classification scheme, and the files were reclassified in the same resolution (30x30) as the DEM file. Slope was classified into seven classes, aspect, and elevation into nine classes, plan and profile curvature into 3 classes, distance from river and distance from roads into five classes and rainfall into six classes. Soil type, lithology and land use were digitized from the available maps and transformed into a raster format of 30x30 resolution to match with the rest of the files. The NDVI and land use maps were generated by processing Landsat 8 satellite images using ENVI 5.1 software. The Landsat 8 images, which had a spatial resolution of 30 × 30 meters, was also sourced from the same website as the ASTER Global DEM data, accessed on 22.08.2023. The NDVI was classified into five classes according to the natural break classification scheme, whereas the land use into six classes. Finally, the rainfall map was created using comprehensive rainfall data that covers the study area and reclassified into six classes. These data were collected and compiled in collaboration with the Meteorological Bureau of the region. Each of the raster format files were weighted according to the Certainty Factor values, as discussed in 3.3.2 (Fig.3).

 

Comment 3:

The color scales used when drawing individual layers in a GIS environment do not allow to discern the difference in the min-max boundaries of the depicted factors; more specifically, this applies to figures 3d, e, f, g, h, i, and p. The accuracy with which the geographical coordinates are written on the figures is sufficient to be up to a degree minutes, namely Latitude: ° ', Longitude: ° '

 

Response to comment 3:

Dear reviewer, we have modified the figures 3a, 3c, 3d, 3e, 3f, 3g, 3h, 3i, 3j and 3k. all figures illustrate the classes of each variable according to the applied classification scheme.

Concerning the geographical coordinates, we use a degree minute seconds projection thus it could not be modified by the ArcGIS suite.

 

Comment 4:

From the results presented in Table 1 regarding the predictive ability and importance of factors related to landslides, it is strange why geological factors such as lithology and soil type have a low Average Merit. Is there any reasonable explanation for such inferences based on the performed analysis?

 

Response to comment 4:

Dear reviewer, thank you for the comment. In response to your comment, we have introduced in the discussion section the following:

 

Concerning the low AM values of Lithology and Soil type covers, the study area, is characterized by a high degree of geological uniformity, the diversity and influence of lithology and soil types on landslide susceptibility might be less pronounced than they would be in areas with more geological diversity. In regions where geological features are relatively homogenous, other factors, such as topography or hydrological conditions, may play a more pivotal role in triggering landslides.

 

Comment 5:

Of the four tree-based models tested, namely REPTree, BREPTree, FT, and BFT, it was found that for the study area of Yanchuan County, the BFT model provided the most accurate prediction of landslide susceptibility. What explanation can be given for the fact that none of the models exceeded 90% of the F-score in Table 6. do the values of the individual quantities given in this table have units?

 

Response to comment 5:

The fact that none of the models exceeded 90% of the F-score and also other metrics can be attributed to various factors inherent to modeling complex natural phenomena like landslides. These include potential limitations in the data (such as resolution, coverage, or completeness), and the inherent challenges in accurately modeling environmental risks. This underlines the difficulty of achieving very high accuracy in predicting natural events and the continuous need for improving modeling techniques and data quality. This has been emphasized in the revised manuscript as follows (in the discussion section):

 

The observation that none of the models exceeded a 90% F-score, along with similar trends in other metrics, can be attributed to several factors intrinsic to the modeling of complex natural phenomena such as landslides. These factors encompass potential data-related limitations, including the resolution, coverage, or completeness of the dataset, as well as the inherent challenges posed by accurately modeling environmental risks. This situation underscores the inherent difficulty in attaining extremely high accuracy levels in the prediction of natural events. It also highlights the ongoing necessity to enhance both the methodologies employed in modeling and the quality of the data used for such analyses.

 

Concerning the introduction of units in Table 6, most of the values in the table are ratios or percentages and are thus unitless. This includes metrics like PPR (Positive Predictive Rate), NPR (Negative Predictive Rate), Sensitivity, Specificity, Accuracy, F-score, and TSS (True Skill Statistic), Also, the MCC (Matthews Correlation Coefficient) is also a unitless measure ranging from -1 to +1 and the RMSE (Root Mean Square Error) is typically expressed in the same units as the predicted variable; however, in classification problems like this, it's often unitless as it represents the square root of the average squared differences between the predicted and actual values.

TP, TN, FP and FN are actual numbers corresponding to the samples that are identified as positive, negative, incorrectly identified as positive, incorrectly identified as negative.

 

Comment 6:

In some places in the text, some expressions could be redacted, such as in Line 174-175.

 

Response to comment 6:

Dear reviewers, thank you for your comment. We have revised Lines 174-175 as follows:

A detailed description of our approach is provided below, complemented by Figure 2, which illustrates the flowchart of our hybrid-ensemble model approach.

 

Comment 7:

The explanation of acronyms of the terms used is given when they first appear in the manuscript text.

 

Response to comment 7:

Thank you for your comment. In response, we have thoroughly reviewed the document and made the appropriate revisions to ensure that all acronyms are clearly defined upon their initial introduction.

 

Dear reviewer, we appreciate your valuable comments and suggestions that enhance the quality of the paper.

On behalf of the co-authors,

best regards.

Paraskevas Tsangaratos.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Abstract should be rewritten. There are many key results that deserve more attention e.g. the covariates that are sensitive and important to hte susceptibliyt map. The justfiication of hte hybrid approach should also be highlighted 

The journal fit and context is also an issue.. It's clearly a method paper for landslide mapping. I was expecting to contextualize in the discussion any water management related implications of the findings but I found very minimal of such. This alone have made me decide at least a major revision

Line by line comments:

"due to various geological and environmental factors." -- not needed

"a Bagging algorithm that had as base learners the single based models, Reduced Error Pruning Decision Tree (REPTree) and Functional Tree (FT) " -- this comes up out of nowhere.. better to put more context about its relevance

"as evidenced by its high Area under 32 the curve area (AUC) value (0.895), compared to the other models" If it's a robust approach, i would expect more accuracy indicator

INtroduction

Line 54 "among experts and scholars" --choose one

Line 61 "ML is used not because statistics is progressing, stat is long"

"Despite the high performance of above methods" -- you didn't mention any "performance" in your "previous paragraph"

I suggest to have a strong justification of the proposed hybrid approach including any literature support supporting such. You don't introduce your proposed approach at the end of the introduction

Have a clear and concise definition / introduction of Bagging and Decision Trees as these are the core concepts of your approach

Define susceptibility also . I was expecting that MaxEnt as a susceptibility-based method is included in your literature review

Materials and methods

Diagram is almost unreadable.. 

Line 195.. which satellite imagery? What date?

How does your proposed hybrid approach compares similar bagging-based decision trees like random forest, quantile random forest etc?

Any consideration of uncertainties i.e. an uncertianty layer of your susceptability map?

Line 442. Why Jenks? Why not equal breaks or quantiles?

I suggest to have a table describing the four Models REPT, FT, BREPT, BFT.. 

Fig 5. Points doesn't look random to me? Could you also include a figure of your samples earlier like in the methods?

Improve readabiliyt and interpretabiliyt of Fig 6

Comments on the Quality of English Language

Some inconsistency in writing style e.g. Intro appears to be different than the rest. Would suggest proofreading of expert

Author Response

Response letter to Reviewer 2

Dear reviewer, the following paragraphs include a point–to–point response to your comments and suggestions.

Comment 1:

Abstract should be rewritten. There are many key results that deserve more attention e.g. the covariates that are sensitive and important to the susceptibility map. The justification of the hybrid approach should also be highlighted. 

The journal fit and context is also an issue. It's clearly a method paper for landslide mapping. I was expecting to contextualize in the discussion any water management related implications of the findings, but I found very minimal of such. This alone have made me decide at least a major revision.

Response to comment 1:

Dear reviewer, thank you for your suggestion. In response to that we have provided a more informative abstract as follows:

 

Landslides represent a significant natural hazard globally, threatening both human settlements and the natural environment. The primary objective of the study was to develop a landslide susceptibility modeling approach that enhances prediction accuracy and informs land-use planning decisions. To improve prediction accuracy and effectively capture the complexity of landslide susceptibility patterns, the study utilized a hybrid ensemble-based methodology. This approach harnessed the power of ensemble models, employing a Bagging algorithm with base learners, including the Reduced Error Pruning Decision Tree (REPTree) and Functional Tree (FT) models. Ensemble models are particularly valuable because they combine the strengths of multiple models, enhancing the overall performance and robustness of the landslide susceptibility prediction. The study focused on Yanchuan County, situated within the hilly and gully region of China's Loess Plateau, known for its susceptibility to landslides, using sixteen critical landslide conditioning factors, encompassing topographic, environmental, and geospatial variables, namely: elevation, slope, aspect, proximity to rivers and roads, rainfall, Normalized Difference Vegetation Index, soil composition, land use, and more. Model performances were evaluated and verified using a range of metrics, including Receiver operating characteristic (ROC) curves, trade-off statistical metrics, and Chi-Square analysis. The results demonstrated the superiority of the integrated models, particularly the Bagging FT (BFT) model, in accurately predicting landslide susceptibility, as evidenced by its high Area under the curve area (AUC) value (0.895), compared to the other models. The model excelled in both positive predictive rate (0.847) and negative predictive rate (0.886), indicating its efficacy in identifying landslide and non-landslide areas and also in F-score metric with a value of 0.869. The study con-tributes to the field of landslide risk assessment, offering a significant investigation tool for managing and mitigating landslide hazards in Yanchuan County and similar regions worldwide.

 

We understand your concern about the lack of direct water management-related implications in our study, as our primary focus has been on enhancing landslide susceptibility prediction through a novel hybrid ensemble-based methodology. The manuscript has been submitted to the special issue Risk Analysis in Landslides and Groundwater-Related Hazards. Our study's emphasis has been on the development and validation of predictive models using specific geological and environmental variables, including factors that indirectly relate to water, such as rainfall and proximity to rivers. In response to your comment, we have included the following in the revised manuscript:

Slope, TRI and Distance to Rivers have been identified as having the highest predictive capabilities. These factors are well-known as significant factors in landslide susceptibility modeling due to their direct physical and hydrological connections with land-slide occurrence. They are all physical characteristics of the terrain that are directly related to hydrological patterns and landslide susceptibility. Steeper slopes generally pose a higher risk of landslides, rugged terrain can increase landslide potential, and proximity to rivers can affect soil moisture levels and slope stability [7, 22, 79]. The above factors along with CI, aspect and rainfall are significant factors that overall are related with water related issues. Specifically, in most cases the presence of rivers with high volume water flow are associated with areas of higher landslide susceptibility [101, 113]. These interconnections highlight how accurately predicting landslide-prone areas is vital for managing and mitigating potential disruptions to water systems. Landslide occurrences can lead to significant consequences such as blockage of rivers and streams, contamination of water supplies from soil and debris, and alterations in hydrological patterns, affecting water availability. By addressing these factors in landslide susceptibility models, we can provide crucial insights for effective water resource management, particularly in mitigating the impacts of landslides on water flow and quality.

Comment 2:

"due to various geological and environmental factors." -- not needed

Response to comment 2:

Thank you for your suggestion. Based on your comment we have omitted the "due to various geological and environmental factors”.

Comment 3:

"a Bagging algorithm that had as base learners the single based models, Reduced Error Pruning Decision Tree (REPTree) and Functional Tree (FT) " -- this comes up out of nowhere.. better to put more context about its relevance.

Response to comment 3:

Dear reviewer, based on your comment we have revised parts of the abstract as follows:

 

Landslides represent a significant natural hazard globally, threatening both human settlements and the natural environment. The primary objective of the study was to develop a robust landslide susceptibility modeling approach that enhances prediction accuracy and informs land-use planning decisions. To improve prediction accuracy and effectively capture the complexity of landslide susceptibility patterns, the study utilized a hybrid ensemble-based methodology. This approach harnessed the power of ensemble models, employing a Bagging algorithm with base learners, including the Reduced Error Pruning Decision Tree (REPTree) and Functional Tree (FT) models. Ensemble models are particularly valuable because they combine the strengths of multiple models, enhancing the overall performance and robustness of the landslide susceptibility prediction. The study focused on Yanchuan County, situated within the hilly and gully region of China's Loess Plateau, known for its susceptibility to landslides, using sixteen critical landslide conditioning factors, encompassing topographic, environmental, and geospatial variables were selected, namely: elevation, slope, aspect, proximity to rivers and roads, rainfall, Normalized Difference Vegetation Index, soil composition, land use, and more.

 

Comment 4:

"as evidenced by its high Area under 32 the curve area (AUC) value (0.895), compared to the other models" If it's a robust approach, i would expect more accuracy indicator

Response to comment 4:

Dear author, we acknowledge your concern in respect with the accuracy of the models, however based on similar landslide susceptibility studies the 0.895 AUC value is considered that describes high accuracy models. This was why we used the term “robust”. Nevertheless, we have omitted the term robust.   

Comment 5:

Line 54 "among experts and scholars" --choose one

Response to comment 5:

Dear reviewer, based on your comment we have omitted “scholars”.

Comment 6:

Line 61 "ML is used not because statistics is progressing, stat is long"

Response to comment 6:

Dear reviewer, based on your comment we have revised as follows:

Machine learning techniques for predicting landslide susceptibility have been steadily re-placing traditional statistical methods, a trend that has gained significant attention from experts and scholars alike.

Comment 7:

"Despite the high performance of above methods" -- you didn't mention any "performance" in your "previous paragraph"

Response to comment 7:

Based on your comment we have revised the manuscript as follows:

Despite the successful implementation high performance of the above methods,

Comment 8:

I suggest to have a strong justification of the proposed hybrid approach including any literature support supporting such. You don't introduce your proposed approach at the end of the introduction

Response to comment 8:

Dear reviewer, although we have included a description of our approach at the end of the introduction, we have modified it as follows:

In this context, our research aimed to develop a robust and accurate landslide susceptibility model for Yanchuan County, situated in the hilly and gully region of the Loess Plateau in northern Shaanxi, China. This area is prone to landslides, which pose substantial threats to human life and the environment. To address this challenge, we introduced a novel hybrid ensemble-based methodology that combines machine learning models, including REPTree and FT, with the Bagging algorithm. What sets our study apart is our comprehensive approach to model landslide susceptibility and to evaluate the proposed models. While previous studies have examined the performance of FT, REPT and Bagging models individually, our research combines these models in a novel ensemble approach, leading to more accurate landslide susceptibility predictions. Concerning the evaluation approach, it includes the use of various metrics such as Receiver Operating Characteristic (ROC) curves, trade-off statistical metrics (Positive Predictive Rate (PPR), Negative Predictive Rate (NPR), sensitivity, specificity, etc.), and Chi-Square analysis for an overall evaluation. This comprehensive evaluation approach not only enhances prediction accuracy but also provides valuable insights into landslide susceptibility, supporting more in-formed land-use and planning decisions. By offering a more comprehensive assessment of model performance and verification using a range of metrics, our research provides a powerful tool for assessing landslide risks not only in the study area but also in other regions with similar geological and environ-mental characteristics.

Comment 9:

Have a clear and concise definition / introduction of Bagging and Decision Trees as these are the core concepts of your approach.

Response to comment 9:

Dear reviewer, based on your comment we have modified the manuscript as follows:

DT are predictive models that segment data into subsets based on feature value differences, constructing a tree for decision-making. The result is a tree with decision nodes and leaf nodes, where each node represents a feature in the instance to be classified, and each branch represents a decision rule, ultimately leading to a leaf node that indicates the outcome. The REPTree model refines this by pruning less accurate branches to enhance prediction. The model uses a regression tree that prunes using reduced-error pruning and is capable of learning from both nominal and numerical data. FT enhances the decision-making process by constructing a DT that classifies samples using the gain ratio as the split criterion. This allows for the integration of logistic regression functions at the leaves or nodes of the tree, combining the interpretability of DT with the predictive power of logistic regression. Bagging, short for Bootstrap Aggregating, is an ensemble machine learning technique that improves the stability and accuracy of machine learning algorithms. It works by creating multiple versions of a predictor model (such as DT) and then combining them to form an aggregated model with better performance. Bagging involves training each model on a random distribution of the dataset and then combining the models by averaging the output (for regression) or voting (for classification). Our modeling approach involved considering the strengths and weaknesses of each model to select the most suitable one for landslide susceptibility assessment. Further explanation concerning models will follow in the proceeding paragraphs.

 

Comment 10:

Define susceptibility also . I was expecting that MaxEnt as a susceptibility-based method is included in your literature review.

Response to comment 10:

Dear reviewer, thank you for your comment. Based on your comment we have introduced the following paragraph.

Landslide susceptibility refers to the assessment of the likelihood or probability of landslides occurring in a particular area based on geological, seismo-tectonic and environmental factors [5,6,7].

As for the MaxEnt method, although being a promising modelling approach, it is a presence-only method, it addresses presence-only data, thus having a different approach in comparison to the ones used in our study. However, it could be a challenge in a future work to compare the performance of presence-only methods with presence-absence models in landslide susceptibility assessments.

Comment 11:

Materials and methods

Diagram is almost unreadable.. 

Response to comment 11:

Dear reviewer, based on your comment we have provided an image of better resolution.

Comment 12:

Line 195.. which satellite imagery? What date?

Response to comment 12:

Dear author, we have included in the revised manuscript information concerning the satellite imagery, as follows:

Additionally, satellite imagery analysis of Landsat 8 images, obtained from the International Scientific & Technical Data Mirror Site of the Computer Network Information Center, Chinese Academy of Sciences, available at http://www.gscloud.cn (accessed on 22.08.2023), was employed to pinpoint regions characterized by stable vegetation cover, flat terrain, or solid bedrock—criteria that, in theory, denote areas less susceptible to landslides and thus suitable for classification as non-landslide areas.

Comment 13:

How does your proposed hybrid approach compares similar bagging-based decision trees like random forest, quantile random forest etc?

Response to comment 13:

Dear author, we haven’t done any comparison of with similar bagging-based decision trees like random forest, quantile random forest. However, in a future work we could procced in a through comparison of other bagging-based decision tree models. It has been discussed in the discussion section as follows:

Comparing the concept of FBT models with RF, which are known for their robustness and ability to handle diverse datasets, BFT extends the bagging concept by incorporating FT, which can capture complex nonlinear relationships within the data.

 

In conclusion, while other bagging- based tree models, like RF, appear as promising methods and are widely applicable, BFT may outperform them when there are complex, nonlinear relationships to uncover. However, as in any classification problem addressed through machine learning techniques, it is advisable to employ multiple methods with either similar or preferably diverse learning mechanisms. This approach allows for a comprehensive examination of the problem and facilitates a relative comparison between the performance metrics of each model to identify the optimal model.

 

Comment 14:

Any consideration of uncertainties i.e. an uncertainty layer of your susceptibility map?

Response to comment 14:

Dear author, we acknowledge that incorporating an uncertainty layer into the process of landslide susceptibility assessment would provide a more complete picture of the spatial distribution of landslide risks and model confidence, significantly enhancing the utility of the models for practical applications. In the present study we have provided (Tables 4 and 5) AUC values, standard errors, and confidence intervals, which showed that the BREPTree and BFT model which had minimal standard errors and significant p-values provide show higher confidence in their predictions. We have introduced in the discussion section the following concerning future work:

 

Furthermore, future work, may include a comprehensive project aimed at integrating an uncertainty analysis into the landslide susceptibility modeling, which will focus on quantifying the confidence levels associated with the susceptibility predictions by implementing advanced statistical techniques such as bootstrapping and Bayesian inference. This will allow us to identify zones where model predictions are less certain, thereby guiding targeted data collection to refine model inputs and improve predictive accuracy.

 

Comment 15:

Line 442. Why Jenks? Why not equal breaks or quantiles?

Response to comment 15:

Dear reviewer, in the case of our study, the natural breaks method was particularly useful for continuous data to identify meaningful classes based on natural groupings inherent in the data.

 

Comment 16:

I suggest to have a table describing the four Models REPT, FT, BREPT, BFT.. 

Response to comment 16:

Dear reviewer in response to your suggestion we have included in the revised manuscript in a table concerning the four models. Please see the revised manuscript for details.

Comment 17:

Fig 5. Points doesn't look random to me? Could you also include a figure of your samples earlier like in the methods?

Response to comment 17:

Dear author in response to your comment we have included in the revised manuscript a figure concerning the spatial distribution of landslide and non-landslide points that are also separated into training and validation subsets. Please, see in the manuscript figure 3. The randomness refers to the random procedure employed to create the training and validation in the region of interest using the Subset Features, in the Geostatistical Analyst toolbox of the ArcGIS suite.

Comment 18:

Improve readability and interpretability of Fig 6

Response to comment 18:

Based on your comment we have provided an image of better resolution and with different colors.

 

Comment 19:

Some inconsistency in writing style e.g. Intro appears to be different than the rest. Would suggest proofreading of expert.

Response to comment 19:

Dear reviewer, thank you for your suggestion. We have carefully reviewed the manuscript and revised it in parts that were necessary.

 

Dear reviewer, we appreciate your valuable comments and suggestions that enhance the quality of the paper.

 

On behalf of the co-authors,

best regards.

Paraskevas Tsangaratos.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Could you still revise FIg 1. Still unreadable text and check the map inset

Make equation 1 bigger  or at least reformat the notation so more readable

What happned to Table 1????

Emphasize more the uncertainties in spatial predictions 

Can you tell more about the possibility of a spatio-temporal model of your method? e.g. multi-date landslide susceptbility and change analysis

Author Response

Dear reviewer, the following paragraphs include a point–to–point response to your comments and suggestions.

Comment 1:

Could you still revise Fig 1. Still unreadable text and check the map inset.

Response to comment 1:

Dear reviewer, thank you for your suggestion. In response to that we have provided a batter figure concerning the study area

Comment 2:

Make equation 1 bigger or at least reformat the notation so more readable.

Response to comment 2:

Thank you for your suggestion. Based on your comment we have rewritten the equation. 

Comment 3:

What happened to Table 1????

Response to comment 3:

Dear reviewer, thank you for the comment. A last-minute change has caused a modification. In the revised manuscript, we have modified the table according to the settings of the journal.   

Comment 4:

Emphasize more the uncertainties in spatial predictions. 

Response to comment 4:

Dear author, we based on your response we have provided the following in the section of Discussion:

Furthermore, while our models incorporate multiple variables to predict landslide susceptibility accurately, the spatial predictions are subject to uncertainties due to potential variations in data quality and the dynamic nature of the environmental factors involved. Acknowledging these uncertainties is vital for interpreting the models’ outputs with caution, especially when applied to land-use planning and risk management decisions. Future research should focus on improving data quality, exploring more sophisticated modeling techniques, and incorporating uncertainty analysis methods, such as Monte Carlo simulations, to quantify and reduce these uncertainties, thereby enhancing the reliability of spatial predictions.

Comment 5:

Can you tell more about the possibility of a spatio-temporal model of your method? e.g. multi-date landslide susceptibility and change analysis.

 Response to comment 5:

Dear reviewer, based on your comment we have provided in the revised manuscript the following paragraph placed in the Discussion section:

The potential to extend our methodology into a spatio-temporal model offers an ex-citing avenue for advancing landslide susceptibility assessment. By incorporating multi-date landslide inventories and analyzing changes over time, a spatio-temporal model could capture the dynamics of landslide susceptibility, reflecting how it evolves in response to changes in environmental conditions and human activities. This approach would enable the identification of trends and patterns in landslide occurrences, providing insights into areas that are becoming more susceptible over time. Such a model would be particularly useful for monitoring the effectiveness of mitigation measures, planning future land use, and adapting to climate change impacts. Incorporating temporal data re-quires robust methodologies to handle the added complexity, including time-series analysis and dynamic modeling techniques, which could significantly enhance our under-standing of landslide risks and inform more proactive management strategies.

Dear reviewer, thank you very much for your time and valuable comments and suggestions that enhance certainly enhanced the quality of the paper.

On behalf of the co-authors,

best regards.

Paraskevas Tsangaratos.

Author Response File: Author Response.pdf