the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Apr 2024
The Early Identification of Flash Flood Disasters: Mechanism, Model and Uncertainty
Abstract. Flash flood disasters are one of the major natural disasters in the world, and rapid and accurate early identification of flash flood disasters is the key to preventing and controlling them. In recent years, computer and spatial information technology development has promoted the advancement of early identification technology for flash floods. However, previous research has mainly focused on the impact of "water" and neglected the impact of "sediment" deposition on the rise of water levels. To gain a more comprehensive understanding of flash floods and improve the accuracy of early identification, this article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The research results have shown that the number of publications in this field has been increasing yearly and will continue to increase, but the cooperation between authors is not close. The coupling effect between sediment replenishment and floods cannot be ignored. Taking into account multiple uncertainties can greatly improve recognition accuracy. This study can provide a panoramic literature perspective and practical research experience for relevant researchers and decision-makers and support further improving flash flood prevention and control capabilities.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(2309 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on hess-2024-69', Anonymous Referee #1, 06 Jun 2024
This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The study has verified that considering multiple uncertainties can greatly improve recognition accuracy.
The review is very well written and in line with the scope of the Journal, and there is no need for significant changes before publication. However, before formal publication, I have some suggestions for consideration by the authors:
- Abstract: The research purpose should be further clarified and made explicit.
- Figure 3 shows a recent decrease in the number of relevant literatures, despite the conclusion drawn in the article that the literature exhibits exponential growth. Please explain the specific reasons for this apparent contradiction.
- The article mentions the uncertainty of the model (4.3.3), but it does not provide specific solutions or methods to reduce the uncertainty of the model.
- The conclusion section can include a summary of the limitations of this study and prospects for future research.
- Please further improve the titles of Figure 3 and Figure 5 in the article.
- Please use three-line tables for the tables in the article.
I am confident that addressing the above points could help in further improving an already very good manuscript.
Citation: https://doi.org/10.5194/hess-2024-69-RC1 -
AC1: 'Reply on RC1', Heng Lu, 18 Jun 2024
RESPONSES TO REVIEWER #1's COMMENTS:
General Comments:
This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The study has verified that considering multiple uncertainties can greatly improve recognition accuracy.
The review is very well written and in line with the scope of the Journal, and there is no need for significant changes before publication.
RESPONSE:
Thank you for taking the time and energy to help us improve the manuscript. We appreciate the positive evaluation of our manuscript and comments in the text. These comments were all valuable and very helpful for revising and improving our manuscript, as well as provided important guiding significance to our research. Below, we addressed the editions point-by-point accordingly and highlighted them in blue color in the letter and the revised manuscript.
Specific Comments:
Comment 1:
Abstract: The research purpose should be further clarified and made explicit.
RESPONSE:
Thank you for this point. We have rewritten this section as follows:
Flash flood disasters are one of the major natural disasters in the world, and rapid and accurate early identification of flash flood disasters is the key to preventing and controlling them. In recent years, computer and spatial information technology development has promoted the advancement of early identification technology for flash floods. However, previous research has mainly focused on the impact of "water" and neglected the impact of "sediment" deposition on the rise of water levels. The purpose of this study is to enable researchers to become more familiar with the research progress and trends of previous studies on disaster mechanisms, identification models, and uncertainties in the field of early identification of mountain floods, better explain the phenomenon of water sediment coupling leading to mountain flood disasters, and use this as the basis for researchers to think and explore new research methods. This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The research results have shown that the number of publications in this field has been increasing yearly and will continue to increase, but the cooperation between authors is not close. The coupling effect between sediment replenishment and floods cannot be ignored. Taking into account multiple uncertainties can greatly improve recognition accuracy. This study can provide a panoramic literature perspective and practical research experience for relevant researchers and decision-makers and support further improving flash flood prevention and control capabilities.
Comment 2:
Figure 3 shows a recent decrease in the number of relevant literatures,despite the conclusion drawn in the article that the literature exhibits exponential growth. Please explain the specific reasons for this apparent contradiction.
RESPONSE:
Thank you for this point. We conducted multiple repeated searches between July and August 2023 to obtain preliminary reference information. As the deadline for literature search is August 18, 2023, this article has determined the literature sample library based on this. Therefore, in Figure 3, the number of publications in 2023 did not decrease, but rather the total duration of literature publications in 2023 was less than one year.
Comment 3:
The article mentions the uncertainty of the model (4.3.3), but it does not provide specific solutions or methods to reduce the uncertainty of the model.
RESPONSE:
Thank you for your comments. We have added the descriptions in Section 4.3.3. as follows:
In addition to the uncertain influencing factors and multi-source spatio-temporal data, inconsistent input parameters, inaccurate initialization, and uncertain model structures will all lead to the uncertainty of early identification models (Jafarzadegan et al., 2021). In view of the strong uncertainty, further research and reliability evaluation of the models are needed to reduce the uncertainty of the model results (Abbaszadeh et al., 2021). The methods to solve the uncertainty problem of early identification models for flash floods include uncertainty estimation, Bayesian inference, visualization, etc. For example, using specific methods of uncertainty estimation can improve the training process of the model, introduce randomness and other methods to estimate or quantify the uncertainty of the model, thereby judging the reliability of the model results and improving the predictive ability of the model. By using Bayesian inference method, prior knowledge and mountain flood monitoring data can be introduced to judge the parameters and results of early identification models, obtaining posterior distribution and uncertainty information. The visualization method uses graphics and charts for data analysis, helping researchers quickly discover the relationships in the model data, and even identify uncertain sources, thereby providing effective decision-making basis. It is particularly important to carefully select the identification models in the calculation process. Under equal identification accuracy, different models will have other distribution characteristics. Therefore, it is still difficult to conclude which identification model is more conducive to modeling. Previous studies have shown that compared with a single model, based on balancing the model identification accuracy and computational burden as much as possible, various coupling models (Ahmadisharaf et al., 2018), integrated models (Tehrany et al., 2014) and hybrid models (Moftakhari et al., 2019) have more advantages in model fitting and prediction performance.
Comment 4:
The conclusion section can include a summary of the limitations of this study and prospects for future research.
RESPONSE:
Thank you for this point. We have added the last paragraph of the conclusion section as follows:
In conclusion, the research on the field is still in its infancy, and there are still a lot of gaps that urgently need to be studied. This paper enriches the document sample database, summarizes the research progress in this field, reveals the flash flood and sediment coupled disaster-causing mechanism, establishes early identification methods based on data-driven and water-sand dynamics, and analyzes the uncertainty sources of the model evaluation results and the suggestions for improving identification accuracy, so as to provide some reference for the early identification of flash flood disasters in the future. Due to the increasing number of literature publications, there is a need for timely updating of the literature database in this article. Future research can use similar literature review methods to reproduce research processes such as retrieval, screening, and visual analysis based on the data sample library in this article. On the other hand, with the rapid development of artificial intelligence technology, AI tools represented by generative AI ChatGPT will provide unlimited opportunities for early identification of flash floods.
Comment 5:
Please further improve the titles of Figure 3 and Figure 5 in the article.
RESPONSE:
Thank you for this suggestion. We changed "Figure 3. Number of publications: y is the number of publications, x is the year, the orange bar represents the number of publications, the light blue dashed line represents the trend of fitting the number of publications, and y = 2E-130e0.1503x is the formula corresponding to the trend of fitting the number of publications, R2 is the coefficient of determination." to "Figure 3. Publications by year: y is the number of publications, x is the year, the orange bar represents the number of publications, the light blue dashed line represents the trend of fitting the number of publications, and y = 2E-130e0.1503x is the formula corresponding to the trend of fitting the number of publications, R2 is the coefficient of determination. Statistics on the number of publications in different types of papers.". We changed "Fig. 5. Research topic clustering results." to "Figure 5. The clusters of the keyword co-occurrence (TOP11), as discussed in section 3.3."
Comment 6:
Please use three-line tables for the tables in the article.
RESPONSE:
Thank you for this point. We have used three-line tables for the tables in the article.
We appreciate all your insightful comments. We worked hard to be responsive to them. Thank you for taking the time and energy to help us improve the manuscript.
Citation: https://doi.org/10.5194/hess-2024-69-AC1
-
RC2: 'Comment on hess-2024-69 : A strange proposal of an article with an unclear focus and too many weaknesses', Anonymous Referee #2, 28 Jun 2024
One might expect from its title that the article would be concerned with flood impact forecasting. It's not. In fact, the authors take as their starting point the highly debatable observation that certain sediment transport phenomena are not sufficiently considered in past works on flash floods. In section 4, they propose a method for mapping the risks associated with flash floods, taking these phenomena into account. The authors propose a heterogeneous combination of a bibliometric analysis of the world literature on flash floods (sections 2 and 3) and a presentation of their personal work on mapping the risks associated with flash floods (section 4), mixing a synthesis of work already published and summarized with a proposal for improvement, unfortunately insufficiently described and evaluated to be of any real benefit to the reader. The link between these two parts does not seem clear to me, and each of them raises serious methodological problems.
Concerning the bibliometric analysis, the main conclusion of the authors is that sediment transport phenomena have generally been underestimated in the past. Two comments on this: first, the authors have one specific phenomenon in mind—locally massive sediment deposits filling the river bed and completely modifying the river cross-section. That is quite a narrow view of the geomorphic processes that may affect river beds and valleys during flash floods and increase local damage. A large diversity of phenomena may occur: mudflows and debris flow in upstream reaches, intense bedload, scour and erosion affecting the river banks and civil works, river bed incision, deposition of boulders, river divagation in its floodplain or on alluvial fans, appearance of secondary channels... Why such an emphasis on a specific phenomenon? Second, the extraction of the bibliographic references is puzzling. Crucial choices are not justified by the authors. They claim to have extracted 1,481 articles dealing with flash floods published between 1981 and 2023 from the Scopus and Elsevier databases. Their extraction does not include some of the “foundational papers” on the topic of flash floods according to Scopus AI. I could extract 7,294 references from Scopus with the keyword "flash flood." Why such a difference? Why have they chosen the additional keyword “identify*” for the selection (line 138)? What is the idea behind this choice that leads to the elimination of 4/5 of the existing literature references? Moreover, if the authors are interested in massive sediment transport and other geomorphic processes, they should have considered that this topic involves several research communities that do not always use the same terms to describe the natural phenomena. The geomorphological or mountain research communities do not always use the term “flash flood” in their publications, a term popularized in the hydrologic community in 1974 through the title of a red book of the IAHS. Scopus contains, for instance, 16,449 references dealing with debris flows, out of which only 346 share the keyword "flash flood." There are 603,316 references dealing with mountain floods, including the 7,294 explicitly mentioning the term "flash flood." Clearly, the bibliometric part of the proposed manuscript is not well thought out and is of little value.
As for section 4, the introductory literature review is far too limited. Only a few references are cited, which are focused on specific case studies, that probably correspond to the test area of the proposed method that is, by the way, not described in the manuscript. The problem and proposed risk mapping approach is presented as if it were of general nature, when it appears to be an ad-hoc proposal adapted to a specific case study. The proposed methodology, the results of which are summarized in figure 9, builds upon an already published work by Yuan et al. (2022). The presentation of the methodology and its justification and evaluation need substantial improvements. Section 4.2.1 presents this initial method in too much detail, including several figures. The reference to the initial publication should be given, but the manuscript should be focused on the new proposal. The criteria suggested on page 19 to identify potential places in the river network where massive deposition of sediments brought by landslides to the river may occur should be justified. What is the meaning of each proposed equation on page 20? Do they relate to previously published works that should be cited? Provide a clear definition and units for the variables in the equations. Finally, the part devoted to the evaluation should be much more detailed: the observed disaster database should be described—what type of data, what is considered a disaster, is it exhaustive for the considered event and valley? Has the observed damage database been split into a calibration and a validation dataset, and do the presented results really correspond to tests on an independent dataset not used for the training or calibration of the method? The evaluation criteria and methods should be precisely described. What do the authors mean by “disaster coverage rate” on line 431? What does “the method increased by 52.3%” mean? Increased what? The proportion of damage places detected by the method? Since the method produces various risk levels, which risk level has been considered to compute these figures? Some classical criteria for the evaluation of detection methods such as the ROC curve and AUC exist and should be implemented here. In conclusion, the proposed risk rating method may be interesting, but it should not be presented as a general method but rather as a case-specific method at this stage, since it seems to have been calibrated and tested for one single case study up to now. Moreover, the approach and its evaluation have to be presented in a much more comprehensive way. An annotated manuscript is attached to this comment.
Interactive discussion
Status: closed
-
RC1: 'Comment on hess-2024-69', Anonymous Referee #1, 06 Jun 2024
This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The study has verified that considering multiple uncertainties can greatly improve recognition accuracy.
The review is very well written and in line with the scope of the Journal, and there is no need for significant changes before publication. However, before formal publication, I have some suggestions for consideration by the authors:
- Abstract: The research purpose should be further clarified and made explicit.
- Figure 3 shows a recent decrease in the number of relevant literatures, despite the conclusion drawn in the article that the literature exhibits exponential growth. Please explain the specific reasons for this apparent contradiction.
- The article mentions the uncertainty of the model (4.3.3), but it does not provide specific solutions or methods to reduce the uncertainty of the model.
- The conclusion section can include a summary of the limitations of this study and prospects for future research.
- Please further improve the titles of Figure 3 and Figure 5 in the article.
- Please use three-line tables for the tables in the article.
I am confident that addressing the above points could help in further improving an already very good manuscript.
Citation: https://doi.org/10.5194/hess-2024-69-RC1 -
AC1: 'Reply on RC1', Heng Lu, 18 Jun 2024
RESPONSES TO REVIEWER #1's COMMENTS:
General Comments:
This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The study has verified that considering multiple uncertainties can greatly improve recognition accuracy.
The review is very well written and in line with the scope of the Journal, and there is no need for significant changes before publication.
RESPONSE:
Thank you for taking the time and energy to help us improve the manuscript. We appreciate the positive evaluation of our manuscript and comments in the text. These comments were all valuable and very helpful for revising and improving our manuscript, as well as provided important guiding significance to our research. Below, we addressed the editions point-by-point accordingly and highlighted them in blue color in the letter and the revised manuscript.
Specific Comments:
Comment 1:
Abstract: The research purpose should be further clarified and made explicit.
RESPONSE:
Thank you for this point. We have rewritten this section as follows:
Flash flood disasters are one of the major natural disasters in the world, and rapid and accurate early identification of flash flood disasters is the key to preventing and controlling them. In recent years, computer and spatial information technology development has promoted the advancement of early identification technology for flash floods. However, previous research has mainly focused on the impact of "water" and neglected the impact of "sediment" deposition on the rise of water levels. The purpose of this study is to enable researchers to become more familiar with the research progress and trends of previous studies on disaster mechanisms, identification models, and uncertainties in the field of early identification of mountain floods, better explain the phenomenon of water sediment coupling leading to mountain flood disasters, and use this as the basis for researchers to think and explore new research methods. This article first uses bibliometric methods to review the spatiotemporal distribution, internal relationships, and research hotspots of literature in this field over the past 42 years. Then, the research practice of considering the impact of sediment on the early identification of flash floods was introduced from three aspects: mechanism, model, and uncertainty. Finally, the existing problems in current research were discussed, and future research directions were proposed. The research results have shown that the number of publications in this field has been increasing yearly and will continue to increase, but the cooperation between authors is not close. The coupling effect between sediment replenishment and floods cannot be ignored. Taking into account multiple uncertainties can greatly improve recognition accuracy. This study can provide a panoramic literature perspective and practical research experience for relevant researchers and decision-makers and support further improving flash flood prevention and control capabilities.
Comment 2:
Figure 3 shows a recent decrease in the number of relevant literatures,despite the conclusion drawn in the article that the literature exhibits exponential growth. Please explain the specific reasons for this apparent contradiction.
RESPONSE:
Thank you for this point. We conducted multiple repeated searches between July and August 2023 to obtain preliminary reference information. As the deadline for literature search is August 18, 2023, this article has determined the literature sample library based on this. Therefore, in Figure 3, the number of publications in 2023 did not decrease, but rather the total duration of literature publications in 2023 was less than one year.
Comment 3:
The article mentions the uncertainty of the model (4.3.3), but it does not provide specific solutions or methods to reduce the uncertainty of the model.
RESPONSE:
Thank you for your comments. We have added the descriptions in Section 4.3.3. as follows:
In addition to the uncertain influencing factors and multi-source spatio-temporal data, inconsistent input parameters, inaccurate initialization, and uncertain model structures will all lead to the uncertainty of early identification models (Jafarzadegan et al., 2021). In view of the strong uncertainty, further research and reliability evaluation of the models are needed to reduce the uncertainty of the model results (Abbaszadeh et al., 2021). The methods to solve the uncertainty problem of early identification models for flash floods include uncertainty estimation, Bayesian inference, visualization, etc. For example, using specific methods of uncertainty estimation can improve the training process of the model, introduce randomness and other methods to estimate or quantify the uncertainty of the model, thereby judging the reliability of the model results and improving the predictive ability of the model. By using Bayesian inference method, prior knowledge and mountain flood monitoring data can be introduced to judge the parameters and results of early identification models, obtaining posterior distribution and uncertainty information. The visualization method uses graphics and charts for data analysis, helping researchers quickly discover the relationships in the model data, and even identify uncertain sources, thereby providing effective decision-making basis. It is particularly important to carefully select the identification models in the calculation process. Under equal identification accuracy, different models will have other distribution characteristics. Therefore, it is still difficult to conclude which identification model is more conducive to modeling. Previous studies have shown that compared with a single model, based on balancing the model identification accuracy and computational burden as much as possible, various coupling models (Ahmadisharaf et al., 2018), integrated models (Tehrany et al., 2014) and hybrid models (Moftakhari et al., 2019) have more advantages in model fitting and prediction performance.
Comment 4:
The conclusion section can include a summary of the limitations of this study and prospects for future research.
RESPONSE:
Thank you for this point. We have added the last paragraph of the conclusion section as follows:
In conclusion, the research on the field is still in its infancy, and there are still a lot of gaps that urgently need to be studied. This paper enriches the document sample database, summarizes the research progress in this field, reveals the flash flood and sediment coupled disaster-causing mechanism, establishes early identification methods based on data-driven and water-sand dynamics, and analyzes the uncertainty sources of the model evaluation results and the suggestions for improving identification accuracy, so as to provide some reference for the early identification of flash flood disasters in the future. Due to the increasing number of literature publications, there is a need for timely updating of the literature database in this article. Future research can use similar literature review methods to reproduce research processes such as retrieval, screening, and visual analysis based on the data sample library in this article. On the other hand, with the rapid development of artificial intelligence technology, AI tools represented by generative AI ChatGPT will provide unlimited opportunities for early identification of flash floods.
Comment 5:
Please further improve the titles of Figure 3 and Figure 5 in the article.
RESPONSE:
Thank you for this suggestion. We changed "Figure 3. Number of publications: y is the number of publications, x is the year, the orange bar represents the number of publications, the light blue dashed line represents the trend of fitting the number of publications, and y = 2E-130e0.1503x is the formula corresponding to the trend of fitting the number of publications, R2 is the coefficient of determination." to "Figure 3. Publications by year: y is the number of publications, x is the year, the orange bar represents the number of publications, the light blue dashed line represents the trend of fitting the number of publications, and y = 2E-130e0.1503x is the formula corresponding to the trend of fitting the number of publications, R2 is the coefficient of determination. Statistics on the number of publications in different types of papers.". We changed "Fig. 5. Research topic clustering results." to "Figure 5. The clusters of the keyword co-occurrence (TOP11), as discussed in section 3.3."
Comment 6:
Please use three-line tables for the tables in the article.
RESPONSE:
Thank you for this point. We have used three-line tables for the tables in the article.
We appreciate all your insightful comments. We worked hard to be responsive to them. Thank you for taking the time and energy to help us improve the manuscript.
Citation: https://doi.org/10.5194/hess-2024-69-AC1
-
RC2: 'Comment on hess-2024-69 : A strange proposal of an article with an unclear focus and too many weaknesses', Anonymous Referee #2, 28 Jun 2024
One might expect from its title that the article would be concerned with flood impact forecasting. It's not. In fact, the authors take as their starting point the highly debatable observation that certain sediment transport phenomena are not sufficiently considered in past works on flash floods. In section 4, they propose a method for mapping the risks associated with flash floods, taking these phenomena into account. The authors propose a heterogeneous combination of a bibliometric analysis of the world literature on flash floods (sections 2 and 3) and a presentation of their personal work on mapping the risks associated with flash floods (section 4), mixing a synthesis of work already published and summarized with a proposal for improvement, unfortunately insufficiently described and evaluated to be of any real benefit to the reader. The link between these two parts does not seem clear to me, and each of them raises serious methodological problems.
Concerning the bibliometric analysis, the main conclusion of the authors is that sediment transport phenomena have generally been underestimated in the past. Two comments on this: first, the authors have one specific phenomenon in mind—locally massive sediment deposits filling the river bed and completely modifying the river cross-section. That is quite a narrow view of the geomorphic processes that may affect river beds and valleys during flash floods and increase local damage. A large diversity of phenomena may occur: mudflows and debris flow in upstream reaches, intense bedload, scour and erosion affecting the river banks and civil works, river bed incision, deposition of boulders, river divagation in its floodplain or on alluvial fans, appearance of secondary channels... Why such an emphasis on a specific phenomenon? Second, the extraction of the bibliographic references is puzzling. Crucial choices are not justified by the authors. They claim to have extracted 1,481 articles dealing with flash floods published between 1981 and 2023 from the Scopus and Elsevier databases. Their extraction does not include some of the “foundational papers” on the topic of flash floods according to Scopus AI. I could extract 7,294 references from Scopus with the keyword "flash flood." Why such a difference? Why have they chosen the additional keyword “identify*” for the selection (line 138)? What is the idea behind this choice that leads to the elimination of 4/5 of the existing literature references? Moreover, if the authors are interested in massive sediment transport and other geomorphic processes, they should have considered that this topic involves several research communities that do not always use the same terms to describe the natural phenomena. The geomorphological or mountain research communities do not always use the term “flash flood” in their publications, a term popularized in the hydrologic community in 1974 through the title of a red book of the IAHS. Scopus contains, for instance, 16,449 references dealing with debris flows, out of which only 346 share the keyword "flash flood." There are 603,316 references dealing with mountain floods, including the 7,294 explicitly mentioning the term "flash flood." Clearly, the bibliometric part of the proposed manuscript is not well thought out and is of little value.
As for section 4, the introductory literature review is far too limited. Only a few references are cited, which are focused on specific case studies, that probably correspond to the test area of the proposed method that is, by the way, not described in the manuscript. The problem and proposed risk mapping approach is presented as if it were of general nature, when it appears to be an ad-hoc proposal adapted to a specific case study. The proposed methodology, the results of which are summarized in figure 9, builds upon an already published work by Yuan et al. (2022). The presentation of the methodology and its justification and evaluation need substantial improvements. Section 4.2.1 presents this initial method in too much detail, including several figures. The reference to the initial publication should be given, but the manuscript should be focused on the new proposal. The criteria suggested on page 19 to identify potential places in the river network where massive deposition of sediments brought by landslides to the river may occur should be justified. What is the meaning of each proposed equation on page 20? Do they relate to previously published works that should be cited? Provide a clear definition and units for the variables in the equations. Finally, the part devoted to the evaluation should be much more detailed: the observed disaster database should be described—what type of data, what is considered a disaster, is it exhaustive for the considered event and valley? Has the observed damage database been split into a calibration and a validation dataset, and do the presented results really correspond to tests on an independent dataset not used for the training or calibration of the method? The evaluation criteria and methods should be precisely described. What do the authors mean by “disaster coverage rate” on line 431? What does “the method increased by 52.3%” mean? Increased what? The proportion of damage places detected by the method? Since the method produces various risk levels, which risk level has been considered to compute these figures? Some classical criteria for the evaluation of detection methods such as the ROC curve and AUC exist and should be implemented here. In conclusion, the proposed risk rating method may be interesting, but it should not be presented as a general method but rather as a case-specific method at this stage, since it seems to have been calibrated and tested for one single case study up to now. Moreover, the approach and its evaluation have to be presented in a much more comprehensive way. An annotated manuscript is attached to this comment.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 436 | 105 | 27 | 568 | 17 | 17 |
- HTML: 436
- PDF: 105
- XML: 27
- Total: 568
- BibTeX: 17
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1