Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds&nbsp;

Ebrahim, Kyrillos M. P.; Gomaa, Sherif M. M. H.; Zayed, Tarek; Alfalah, Ghasan

doi:https://doi.org/10.5194/egusphere-2024-4160

Preprints

https://doi.org/10.5194/egusphere-2024-4160

Preprints

24 Jan 2025

| 24 Jan 2025

Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds

Kyrillos M. P. Ebrahim, Sherif M. M. H. Gomaa, Tarek Zayed, and Ghasan Alfalah

Abstract. Landslides rank among the most devastating hazards, leading to loss of life and destruction of infrastructure, with rainfall being a primary triggering factor. Global climate change has increased landslide occurrence; accordingly, accurate landslide prediction is crucial to reduce damage and losses. Since landslides account for 17 % of all natural hazard fatalities, several studies have been done across the globe to predict these events better. Despite the considerable number of review articles, a comprehensive comparison between empirically, physically, deterministically, and phenomenologically based prediction models is still missing. Moreover, they lack adopting mixed methodology. Accordingly, a mixed review that comprised scientometric, systematic, and bibliometric analysis was employed. This study (Part I of a two-part review) examines two approaches for analyzing local-scale landslides: empirical-statistical methods and physically based causative threshold models. Deterministic and phenomenologically based prediction models are discussed in part ii and have been published (Ebrahim et al., 2024a). This study explores the practicality and constraints associated with the aforementioned methodologies. As a result, critical insights into rainfall-induced landslides are examined. Macroscopically, antecedent rainfall surpasses the intensity-duration thresholds. Physically based causative thresholds can be utilized when geotechnical or hydrological data are limited. Microscopely, hybrid artificial intelligence models provide higher prediction accuracies. Finally, research suggestions are highlighted, as modeling artificial intelligence models with extensive datasets to achieve high prediction accuracy is still needed for further development.

Received: 27 Dec 2024 – Discussion started: 24 Jan 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Kyrillos M. P. Ebrahim, Sherif M. M. H. Gomaa, Tarek Zayed, and Ghasan Alfalah

Status: closed

CC1: 'Comment on egusphere-2024-4160', Fayrouz Elwassief, 03 Feb 2025

Great work, and very informative clear figures.

Citation: https://doi.org/10.5194/egusphere-2024-4160-CC1
RC1:
'Comment on egusphere-2024-4160', Anonymous Referee #1, 16 Feb 2025

Dear Authors, after careful reading of your manuscript "Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds", which falls in my area of expertise, I am sorry to recommend rejection. The main reasons are below:
Incompleteness. This is a review article, and reading the reference list I was surprised to see the absence of many works that I consider very important (for innovation, originality or relevance of the results). Also, I saw the absence of many important authors working on this field, or some of them who have published dozens of papers are present only with one or two works. My feeling was confirmed while reading the text: the sample of the articles analyzed is not representative of the whole literature body and I think you missed a lot. Therefore, this review is no complete and brings biased results which may miss the most relevant findings.
Method. I think one of the possible reasons of the incompleteness is some flaw in the search methodology. You searched only "landslide prediction", therefore you missed all works that use "forecast" (more or less it has the same meaning), "model-modeling-modelling" (somebody does the modeling, somebody does the modeling to predict), "early warning" (prediction is usually included in papers about landslide early warning), and others linked terms. Also, limiting the search only to the field of "engineering" is a heavy limitation. A landslide is a geomorphologic process. I would have added geology-related disciplines as well: I'm sure you have missed a lot.
Irrelevance. In your text you provide some details about the studies you have identified, without proper background. Nothing you write is wrong, but everything misses a wider context and it is not relevant information that can be used by somebody else in an effective way. Just an example: you write down equations 1, 2 and 3, but you fail to explain what is the purpose of the different thresholds, which landslide types are encompassed, which area they refer to, which procedure (e.g., rainfall data, threshold model...) has been used to define them. In this way, it is not clear what the reader can do with the limited information you provide. This is just an example of the general scheme used in the manuscript: you summarize every work in three lines but there is not a clear scientific or technical context and the information provided is of little practical use.

Citation: https://doi.org/10.5194/egusphere-2024-4160-RC1
- AC1:
  'Reply on RC1', Kyrillos Ebrahim, 25 Mar 2025
  Thank you for taking the time to provide detailed feedback on our manuscript. We greatly value your insights and will carefully consider them. However, we would like to take this opportunity to justify our work.
  Please find our responses below:
  We acknowledge that citations serve to support the authors' claims rather than act as a direct measure of a study’s quality. In our review, we aimed to be as comprehensive as possible by referring to all available existing reviews to identify gaps in the literature (please refer to Table 1).
  
  Our study includes peer-reviewed manuscripts and provides a detailed analysis through a scientometric approach. We believe that contributions to scientific knowledge are not limited to a specific group of authors, and every peer-reviewed work brings valuable insights to the field.
  
  To ensure an unbiased and systematic review, we adopted the well-established PRISMA methodology, incorporating forward and backward snowballing techniques. This method was carefully designed to minimize selection bias. We assure you that our selection process was rigorous and aimed at maintaining objectivity.
  
  Our literature search followed the PRISMA guidelines, incorporating multiple strategies, including keyword mapping and scientometric analysis, to ensure comprehensive coverage. While "landslide prediction" was a primary search term, our approach also considered related terms, as reflected in the keyword maps (please see Figure 9). Additionally, our review includes studies from geology-related disciplines that contribute to the understanding of landslide prediction (through the snowballing technique which is further visualized and illustrated in the Scientometric analysis).
  
  The quality of the selected studies is supported by multiple indicators, including journal impact, citation count, and geographical distribution. For example, Figure 7 illustrates that our review includes highly reputable journals such as Landslides and Engineering Geology. Furthermore, Figure 8 highlights the contributions from key research hubs such as China and Italy, while Table 5 lists the most cited papers, further demonstrating the robustness of our selection.
  
  We have provided comparative tables, illustrative charts, and a structured discussion to enhance clarity (refer to Figures 14, 15, 17, 18, 20, 24, 25, 26, and Table 6). While it is not feasible to provide a detailed breakdown of every paper, we believe the included visualizations and summaries offer a clear and useful overview of the topic.
  
  The equations presented in our manuscript, including Equations 1, 2, and 3, are part of a broader discussion on intensity-duration (ID) thresholds. Figure 14 provides additional context to help readers understand the applicability of these thresholds.
  
  Our review aims to provide researchers with a clear understanding of the strengths and limitations of various prediction approaches. By summarizing key insights and highlighting research gaps, we hope to assist future studies in selecting appropriate methodologies (please refer to Figure 29 and the Discussion section).
  
  Thank you for your time and consideration.
  Best regards,
  
  Citation: https://doi.org/10.5194/egusphere-2024-4160-AC1
RC2:
'Comment on egusphere-2024-4160', Anonymous Referee #2, 23 Feb 2025

Review of NHESS submission by Ebrahim et al. entitled “Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds
Abstract: If the first part of this review was published in Bulletin of Engineering Geology and the Environment, why would you not submit this paper to that journal? (especially given your emphasis on comparing landslide models)
Overall comments: I found this review paper to be very unorganized and inconsistent in terms of addressing important aspects of rainfall-initiated landslide models. These issues are articulated in my summary that follows. Admittedly, I am not a big proponent of the plethora of papers being published these days that rely almost solely on automated literature searches. Many such papers simply present literature search metrics without deep insights into the topic under consideration. I would categorize this paper into this list. The authors made few attempts to succinctly synthesize the breadth of findings into meaningful insights associated with landslide initiation and modelling. Ignoring pre-2000 papers in my opinion, was a major oversight. Many of the newer papers that employ so-called advanced algorithms (e.g., machine learning, AI, evolving statistical methods) often overlook the fundamental processes inherent in landslide initiation. This biases creeps into this review paper as well. Even the authors’ writing employed seemingly new catch phrases and sometimes awkward English expressions: e.g., lines 43, 134-135, 606, 930. Overall, I do not see this review paper contributing to an advanced appreciation of landslide modelling.
General and specific comments:
L 40 Missing in this definition is that gravity is the driving process.
L 60 Rainfall-triggered landslides can also be quite deep; maybe you are referring to shallow-rapid landslides that respond to individual storm events. Even in Fig. 2b where you show a deep-seated rotational landslide, rainfall will contribute to activation, however the timing may be a bit lagged from the rain event. What is the point of this figure?
L 77-80 I do not see the need to focus on these structural mitigation measures in a review paper on rainfall-initiated landslide models.
L 85-91 Here you mention spatial and temporal prediction of landslides: however, some models are not spatially explicit, while others do not consider temporal aspects.
L 102-109 Repeated sentences with very strange wording; it seems that the first review paper (this one) may have been rejected elsewhere.
L 110-131 Overly wordy, repeating items previously discussed; reads more like a graduate thesis than a journal paper.
Table 1 presents a narrow overview of review articles and books that cover landslide prediction techniques.
Figures 3 and 4 are not needed, and contain a lot of unhelpful jargon and some misspellings. Furthermore, the details presented in L150-220 about how to perform a literature search seem irrelevant. I realize such semi-automated searches are now quite popular, but they do little to ensure that relevant literature is included without incorporating deep process understanding. The section that follows (Scientometric Analysis), is first of all a strange terminology and appears to depend on the outcomes from the initial identification process and screening, which I have already expressed concerns with. The graph in Figure 6 shows the extreme to which authors are now relying on publication metrics – this adds no value to a systematic scientific review. Likewise, just because your automated search results targeted the journals shown in Figure 7, does not mean the most important papers on landslide modeling reside there – I don’t see Water Resources Research, Earth Surface Processes and Landforms, Geomorphology, or JGR Earth Surface for example. This plus your automated findings in Table 2 indicate to me that you have not applied any historical context or deep personal insights into this field of study. Furthermore, just because certain countries are more prolific based on the search you conducted, does not mean that these represent the best modelling findings – as such, I see no value in Figure 8 or Table 3.
L 277-289 Restricting your literature from 2000 to 2024 excludes some of the key landslide modeling studies in the 1990’s and even earlier.
Figure 9 Enough with the cloud bubble mapping, please. What does this really tell us without deeper insights into landslide modelling.
L 307-320 & Figure 10; For a review paper, this is highly site-specific information that contributes little to the overall knowledge of rainfall-initiated landslide modelling.
L 331-425 I-D Thresholds were succinctly reviewed in a book published in 2006 (Sidle & Ochiai, Landslides: Process, Prediction, and Land Use, American Geophysical Union Water Resources Monograph 18) in which the importance of antecedent soil moisture (2-day API) was demonstrated. While this current review by Ebrahim et al. includes updated references, it ignores some of the more fundamental work and reviews including the work of Mike Crozier in New Zealand. As such, I see very little in terms of new insights presented here except for the sub-section on I-D accuracy – unfortunately this was poorly presented (including Figure 14) making it difficult to ascertain how the results were achieved.
L 427-442 Important references to landslide causation factors were ignored; many of these papers were published pre-2000.
L 448 Why not say ‘pore water pressure’; this is the most common cause of rainfall-initiated landslides.
Figure 16 is not needed.
L 468-477 I do not think these measurement methods are needed here. Also, there is an ‘overload’ of references from Three Gorges – please remember, a good ‘review paper’ needs to synthesize information. What I see here is just a ‘download’ of compiled references with no insights. I do not see the relevance of these references and the material presented in L 468-477. The paragraph does not flow well after this point into the material presented in L 478-484.
Model selection section (L 494-542): Here I was surprised to see no reference to the spatial scale of application of such regression models. How can these be incorporated into spatially distributed models? Finally, on L 542-548 there is some mention of including these algorithms in landslide susceptibility maps and ‘regional areas’, but the connection between modelling type and spatial scale application is vague.
L 552-557: This is mostly just another ‘overload’ of references with little insight. Same issue in L 567-574.
The reference to soil creep appears rather suddenly without proper introduction (L 559, 580, 588, etc.). Soil creep is the plastic deformation of the soil mantle (or rock mantle) and one of the most important factors affecting creep rate is cumulative soil moisture. The topic of soil creep (if you wish to include this in this landslide model paper) needs to be more systematically introduced. Here you go into excessive detail with statistical and automated methods to assess creep without a more process-focused discussion about what causes creep and how it is manifested. In the entire subsection on “Training and testing data split ratio” it is never mentioned if this refers to creep (I assume it does) and just presents another download of geographically focused references) and Figure 23, which contributes very little.
Paragraph beginning on L654: Now we go back to regression models it seems. I do not think everything presented prior to this was related only to regression models. Again, this underlines the need for better organization.
L 665-710 This section seems to focus on soil creep. However, the header “Displacement models…” has a much wider meaning in landslide science, incorporating landslide movement and debris flow runout.
L 723-725 The time lag between rainfall inputs and slope failure is a major issue that most all models do not address well. Lag time depends on soil depth, but also, soil properties, particularly the presence of interconnected preferential flow networks. In addition to being within the soil, these can induce failure due to preferential flow in fractured bedrock via exfiltration. These issues should have been addressed.
L 740-814 This section discusses the importance of antecedent soil moisture on landslide initiation and how this has been incorporated into models. However, the formative work on this topic has largely been ignored. As early as 1980 Crozier and Eyles discussed this and a paper by Crozier in 1999 showed a clear distinction between days with and without landslides in a plot of daily rainfall vs. soil water. Furthermore Sidle & Ochiai (2006) showed the effects of antecedent 2-day rainfall in I-D plots related to landslide thresholds. While I am not negating the importance of some of the newer work cited, a review should cover the formative research. Additionally, I don’t think enough credit was given to the fairly recent work on I-D thresholds by Bogaard & Greco (2018).
L 816-822 An important point here is that empirical models such as those based on I-D thresholds are typically not used in spatially-distributed models.
In the Discussion or elsewhere in the paper, there is no mention about the modelling research related to vegetation management effects on landslides. This is incredibly important in many parts of the world and is a conspicuous omission.
L 891-892 What about the earlier landslide modeling research that addressed these issues?; some of these studies were more comprehensive than the recent studies you present that only focus on statistical manipulations and machine learning.
L 894-896 This is an artifact of your time-limited automated search. If you examine the landslide literature, you will see that the first two physically based landslide models (SHALSTAB and dSLAM) were cited much more than Merghadi et al.’s paper. And both of these other studies were research papers, not reviews.
L 900-902 As noted earlier, antecedent thresholds were incorporated into I-D thresholds in the text on “Landslides: Processes, Prediction and Land Use”
Table 7 Interestingly, you note neglecting the role of vegetation cover on infiltration, but this is one of the few parts of the paper where vegetation influences is even mentioned. What about the role of root reinforcement?

Citation: https://doi.org/10.5194/egusphere-2024-4160-RC2
- AC2: 'Reply on RC2', Kyrillos Ebrahim, 25 Mar 2025
  
  Thank you for taking the time to provide detailed feedback on our manuscript. We greatly value your insights and will carefully consider them. However, we would like to take this opportunity to justify our work.
  Please find our responses below:
  We acknowledge your concerns regarding the automated literature search and the exclusion of pre-2000 studies. While our intent was to focus on recent advancements. Our literature search followed the PRISMA guidelines, incorporating multiple strategies, including keyword mapping and scientometric analysis, to ensure comprehensive coverage. The quality of the selected studies is supported by multiple indicators, including journal impact, citation count, and geographical distribution. For example, Figure 7 illustrates that our review includes highly reputable journals such as Landslides and Engineering Geology. Furthermore, Figure 8 highlights the contributions from key research hubs such as China and Italy, while Table 5 lists the most cited papers, further demonstrating the robustness of our selection
  Your comments on the organization and clarity of certain sections are well noted. We will work on improving the coherence of the manuscript, eliminating redundant content, and ensuring a clearer distinction between different modeling approaches. However, we have provided comparative tables, illustrative charts, and we believe the included visualizations and summaries also offer a clear and useful overview of the topic (refer to Figures 14, 15, 17, 18, 20, 24, 25, 26, and Table 6).
  Thank you for your time and consideration.
  Best regards,
  
  Citation: https://doi.org/10.5194/egusphere-2024-4160-AC2

Status: closed

CC1: 'Comment on egusphere-2024-4160', Fayrouz Elwassief, 03 Feb 2025

Great work, and very informative clear figures.

Citation: https://doi.org/10.5194/egusphere-2024-4160-CC1
RC1:
'Comment on egusphere-2024-4160', Anonymous Referee #1, 16 Feb 2025

Dear Authors, after careful reading of your manuscript "Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds", which falls in my area of expertise, I am sorry to recommend rejection. The main reasons are below:
Incompleteness. This is a review article, and reading the reference list I was surprised to see the absence of many works that I consider very important (for innovation, originality or relevance of the results). Also, I saw the absence of many important authors working on this field, or some of them who have published dozens of papers are present only with one or two works. My feeling was confirmed while reading the text: the sample of the articles analyzed is not representative of the whole literature body and I think you missed a lot. Therefore, this review is no complete and brings biased results which may miss the most relevant findings.
Method. I think one of the possible reasons of the incompleteness is some flaw in the search methodology. You searched only "landslide prediction", therefore you missed all works that use "forecast" (more or less it has the same meaning), "model-modeling-modelling" (somebody does the modeling, somebody does the modeling to predict), "early warning" (prediction is usually included in papers about landslide early warning), and others linked terms. Also, limiting the search only to the field of "engineering" is a heavy limitation. A landslide is a geomorphologic process. I would have added geology-related disciplines as well: I'm sure you have missed a lot.
Irrelevance. In your text you provide some details about the studies you have identified, without proper background. Nothing you write is wrong, but everything misses a wider context and it is not relevant information that can be used by somebody else in an effective way. Just an example: you write down equations 1, 2 and 3, but you fail to explain what is the purpose of the different thresholds, which landslide types are encompassed, which area they refer to, which procedure (e.g., rainfall data, threshold model...) has been used to define them. In this way, it is not clear what the reader can do with the limited information you provide. This is just an example of the general scheme used in the manuscript: you summarize every work in three lines but there is not a clear scientific or technical context and the information provided is of little practical use.

Citation: https://doi.org/10.5194/egusphere-2024-4160-RC1
- AC1:
  'Reply on RC1', Kyrillos Ebrahim, 25 Mar 2025
  Thank you for taking the time to provide detailed feedback on our manuscript. We greatly value your insights and will carefully consider them. However, we would like to take this opportunity to justify our work.
  Please find our responses below:
  We acknowledge that citations serve to support the authors' claims rather than act as a direct measure of a study’s quality. In our review, we aimed to be as comprehensive as possible by referring to all available existing reviews to identify gaps in the literature (please refer to Table 1).
  
  Our study includes peer-reviewed manuscripts and provides a detailed analysis through a scientometric approach. We believe that contributions to scientific knowledge are not limited to a specific group of authors, and every peer-reviewed work brings valuable insights to the field.
  
  To ensure an unbiased and systematic review, we adopted the well-established PRISMA methodology, incorporating forward and backward snowballing techniques. This method was carefully designed to minimize selection bias. We assure you that our selection process was rigorous and aimed at maintaining objectivity.
  
  Our literature search followed the PRISMA guidelines, incorporating multiple strategies, including keyword mapping and scientometric analysis, to ensure comprehensive coverage. While "landslide prediction" was a primary search term, our approach also considered related terms, as reflected in the keyword maps (please see Figure 9). Additionally, our review includes studies from geology-related disciplines that contribute to the understanding of landslide prediction (through the snowballing technique which is further visualized and illustrated in the Scientometric analysis).
  
  The quality of the selected studies is supported by multiple indicators, including journal impact, citation count, and geographical distribution. For example, Figure 7 illustrates that our review includes highly reputable journals such as Landslides and Engineering Geology. Furthermore, Figure 8 highlights the contributions from key research hubs such as China and Italy, while Table 5 lists the most cited papers, further demonstrating the robustness of our selection.
  
  We have provided comparative tables, illustrative charts, and a structured discussion to enhance clarity (refer to Figures 14, 15, 17, 18, 20, 24, 25, 26, and Table 6). While it is not feasible to provide a detailed breakdown of every paper, we believe the included visualizations and summaries offer a clear and useful overview of the topic.
  
  The equations presented in our manuscript, including Equations 1, 2, and 3, are part of a broader discussion on intensity-duration (ID) thresholds. Figure 14 provides additional context to help readers understand the applicability of these thresholds.
  
  Our review aims to provide researchers with a clear understanding of the strengths and limitations of various prediction approaches. By summarizing key insights and highlighting research gaps, we hope to assist future studies in selecting appropriate methodologies (please refer to Figure 29 and the Discussion section).
  
  Thank you for your time and consideration.
  Best regards,
  
  Citation: https://doi.org/10.5194/egusphere-2024-4160-AC1
RC2:
'Comment on egusphere-2024-4160', Anonymous Referee #2, 23 Feb 2025

Review of NHESS submission by Ebrahim et al. entitled “Review Article: Rainfall‑Induced Landslide Prediction Models, Part I: Empirical-Statistical and Physically Based Causative Thresholds
Abstract: If the first part of this review was published in Bulletin of Engineering Geology and the Environment, why would you not submit this paper to that journal? (especially given your emphasis on comparing landslide models)
Overall comments: I found this review paper to be very unorganized and inconsistent in terms of addressing important aspects of rainfall-initiated landslide models. These issues are articulated in my summary that follows. Admittedly, I am not a big proponent of the plethora of papers being published these days that rely almost solely on automated literature searches. Many such papers simply present literature search metrics without deep insights into the topic under consideration. I would categorize this paper into this list. The authors made few attempts to succinctly synthesize the breadth of findings into meaningful insights associated with landslide initiation and modelling. Ignoring pre-2000 papers in my opinion, was a major oversight. Many of the newer papers that employ so-called advanced algorithms (e.g., machine learning, AI, evolving statistical methods) often overlook the fundamental processes inherent in landslide initiation. This biases creeps into this review paper as well. Even the authors’ writing employed seemingly new catch phrases and sometimes awkward English expressions: e.g., lines 43, 134-135, 606, 930. Overall, I do not see this review paper contributing to an advanced appreciation of landslide modelling.
General and specific comments:
L 40 Missing in this definition is that gravity is the driving process.
L 60 Rainfall-triggered landslides can also be quite deep; maybe you are referring to shallow-rapid landslides that respond to individual storm events. Even in Fig. 2b where you show a deep-seated rotational landslide, rainfall will contribute to activation, however the timing may be a bit lagged from the rain event. What is the point of this figure?
L 77-80 I do not see the need to focus on these structural mitigation measures in a review paper on rainfall-initiated landslide models.
L 85-91 Here you mention spatial and temporal prediction of landslides: however, some models are not spatially explicit, while others do not consider temporal aspects.
L 102-109 Repeated sentences with very strange wording; it seems that the first review paper (this one) may have been rejected elsewhere.
L 110-131 Overly wordy, repeating items previously discussed; reads more like a graduate thesis than a journal paper.
Table 1 presents a narrow overview of review articles and books that cover landslide prediction techniques.
Figures 3 and 4 are not needed, and contain a lot of unhelpful jargon and some misspellings. Furthermore, the details presented in L150-220 about how to perform a literature search seem irrelevant. I realize such semi-automated searches are now quite popular, but they do little to ensure that relevant literature is included without incorporating deep process understanding. The section that follows (Scientometric Analysis), is first of all a strange terminology and appears to depend on the outcomes from the initial identification process and screening, which I have already expressed concerns with. The graph in Figure 6 shows the extreme to which authors are now relying on publication metrics – this adds no value to a systematic scientific review. Likewise, just because your automated search results targeted the journals shown in Figure 7, does not mean the most important papers on landslide modeling reside there – I don’t see Water Resources Research, Earth Surface Processes and Landforms, Geomorphology, or JGR Earth Surface for example. This plus your automated findings in Table 2 indicate to me that you have not applied any historical context or deep personal insights into this field of study. Furthermore, just because certain countries are more prolific based on the search you conducted, does not mean that these represent the best modelling findings – as such, I see no value in Figure 8 or Table 3.
L 277-289 Restricting your literature from 2000 to 2024 excludes some of the key landslide modeling studies in the 1990’s and even earlier.
Figure 9 Enough with the cloud bubble mapping, please. What does this really tell us without deeper insights into landslide modelling.
L 307-320 & Figure 10; For a review paper, this is highly site-specific information that contributes little to the overall knowledge of rainfall-initiated landslide modelling.
L 331-425 I-D Thresholds were succinctly reviewed in a book published in 2006 (Sidle & Ochiai, Landslides: Process, Prediction, and Land Use, American Geophysical Union Water Resources Monograph 18) in which the importance of antecedent soil moisture (2-day API) was demonstrated. While this current review by Ebrahim et al. includes updated references, it ignores some of the more fundamental work and reviews including the work of Mike Crozier in New Zealand. As such, I see very little in terms of new insights presented here except for the sub-section on I-D accuracy – unfortunately this was poorly presented (including Figure 14) making it difficult to ascertain how the results were achieved.
L 427-442 Important references to landslide causation factors were ignored; many of these papers were published pre-2000.
L 448 Why not say ‘pore water pressure’; this is the most common cause of rainfall-initiated landslides.
Figure 16 is not needed.
L 468-477 I do not think these measurement methods are needed here. Also, there is an ‘overload’ of references from Three Gorges – please remember, a good ‘review paper’ needs to synthesize information. What I see here is just a ‘download’ of compiled references with no insights. I do not see the relevance of these references and the material presented in L 468-477. The paragraph does not flow well after this point into the material presented in L 478-484.
Model selection section (L 494-542): Here I was surprised to see no reference to the spatial scale of application of such regression models. How can these be incorporated into spatially distributed models? Finally, on L 542-548 there is some mention of including these algorithms in landslide susceptibility maps and ‘regional areas’, but the connection between modelling type and spatial scale application is vague.
L 552-557: This is mostly just another ‘overload’ of references with little insight. Same issue in L 567-574.
The reference to soil creep appears rather suddenly without proper introduction (L 559, 580, 588, etc.). Soil creep is the plastic deformation of the soil mantle (or rock mantle) and one of the most important factors affecting creep rate is cumulative soil moisture. The topic of soil creep (if you wish to include this in this landslide model paper) needs to be more systematically introduced. Here you go into excessive detail with statistical and automated methods to assess creep without a more process-focused discussion about what causes creep and how it is manifested. In the entire subsection on “Training and testing data split ratio” it is never mentioned if this refers to creep (I assume it does) and just presents another download of geographically focused references) and Figure 23, which contributes very little.
Paragraph beginning on L654: Now we go back to regression models it seems. I do not think everything presented prior to this was related only to regression models. Again, this underlines the need for better organization.
L 665-710 This section seems to focus on soil creep. However, the header “Displacement models…” has a much wider meaning in landslide science, incorporating landslide movement and debris flow runout.
L 723-725 The time lag between rainfall inputs and slope failure is a major issue that most all models do not address well. Lag time depends on soil depth, but also, soil properties, particularly the presence of interconnected preferential flow networks. In addition to being within the soil, these can induce failure due to preferential flow in fractured bedrock via exfiltration. These issues should have been addressed.
L 740-814 This section discusses the importance of antecedent soil moisture on landslide initiation and how this has been incorporated into models. However, the formative work on this topic has largely been ignored. As early as 1980 Crozier and Eyles discussed this and a paper by Crozier in 1999 showed a clear distinction between days with and without landslides in a plot of daily rainfall vs. soil water. Furthermore Sidle & Ochiai (2006) showed the effects of antecedent 2-day rainfall in I-D plots related to landslide thresholds. While I am not negating the importance of some of the newer work cited, a review should cover the formative research. Additionally, I don’t think enough credit was given to the fairly recent work on I-D thresholds by Bogaard & Greco (2018).
L 816-822 An important point here is that empirical models such as those based on I-D thresholds are typically not used in spatially-distributed models.
In the Discussion or elsewhere in the paper, there is no mention about the modelling research related to vegetation management effects on landslides. This is incredibly important in many parts of the world and is a conspicuous omission.
L 891-892 What about the earlier landslide modeling research that addressed these issues?; some of these studies were more comprehensive than the recent studies you present that only focus on statistical manipulations and machine learning.
L 894-896 This is an artifact of your time-limited automated search. If you examine the landslide literature, you will see that the first two physically based landslide models (SHALSTAB and dSLAM) were cited much more than Merghadi et al.’s paper. And both of these other studies were research papers, not reviews.
L 900-902 As noted earlier, antecedent thresholds were incorporated into I-D thresholds in the text on “Landslides: Processes, Prediction and Land Use”
Table 7 Interestingly, you note neglecting the role of vegetation cover on infiltration, but this is one of the few parts of the paper where vegetation influences is even mentioned. What about the role of root reinforcement?

Citation: https://doi.org/10.5194/egusphere-2024-4160-RC2
- AC2: 'Reply on RC2', Kyrillos Ebrahim, 25 Mar 2025
  
  Thank you for taking the time to provide detailed feedback on our manuscript. We greatly value your insights and will carefully consider them. However, we would like to take this opportunity to justify our work.
  Please find our responses below:
  We acknowledge your concerns regarding the automated literature search and the exclusion of pre-2000 studies. While our intent was to focus on recent advancements. Our literature search followed the PRISMA guidelines, incorporating multiple strategies, including keyword mapping and scientometric analysis, to ensure comprehensive coverage. The quality of the selected studies is supported by multiple indicators, including journal impact, citation count, and geographical distribution. For example, Figure 7 illustrates that our review includes highly reputable journals such as Landslides and Engineering Geology. Furthermore, Figure 8 highlights the contributions from key research hubs such as China and Italy, while Table 5 lists the most cited papers, further demonstrating the robustness of our selection
  Your comments on the organization and clarity of certain sections are well noted. We will work on improving the coherence of the manuscript, eliminating redundant content, and ensuring a clearer distinction between different modeling approaches. However, we have provided comparative tables, illustrative charts, and we believe the included visualizations and summaries also offer a clear and useful overview of the topic (refer to Figures 14, 15, 17, 18, 20, 24, 25, 26, and Table 6).
  Thank you for your time and consideration.
  Best regards,
  
  Citation: https://doi.org/10.5194/egusphere-2024-4160-AC2

Kyrillos M. P. Ebrahim, Sherif M. M. H. Gomaa, Tarek Zayed, and Ghasan Alfalah

Viewed

Total article views: 443 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
272	151	20	443	13	14

HTML: 272
PDF: 151
XML: 20
Total: 443
BibTeX: 13
EndNote: 14

Views and downloads (calculated since 24 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	64	24	5	93
Feb 2025	68	28	6	102
Mar 2025	56	30	2	88
Apr 2025	40	25	0	65
May 2025	25	23	3	51
Jun 2025	19	21	4	44

Cumulative views and downloads (calculated since 24 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	64	24	5	93
Feb 2025	68	28	6	102
Mar 2025	56	30	2	88
Apr 2025	40	25	0	65
May 2025	25	23	3	51
Jun 2025	19	21	4	44

Viewed (geographical distribution)

Total article views: 439 (including HTML, PDF, and XML) Thereof 439 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 23 Jun 2025

Short summary

Rainfall induced landslides pose increasing risks to lives and infrastructure due to climate change. This study presents a framework for predicting landslides, combining statistical and physically based models. A flowchart demonstrates how large-scale models transition to localized. By analyzing thresholds and model performance, the study highlights the role of artificial intelligence and big data in improving predictions, helping to reduce global landslide risks.


Total:	0
HTML:	0
PDF:	0
XML:	0