Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones

Dang, Kinh Bac; Nguyen, Hieu; Do, Thanh Dat; Pham, Thi Phuong Nga; Giang, Tuan Linh; Nguyen, Thi Dieu Linh; Ngo, Huu Hao; Forino, Giuseppe

doi:10.5194/egusphere-2026-160

Preprints

https://doi.org/10.5194/egusphere-2026-160

Preprints

27 Jan 2026

| 27 Jan 2026

Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones

Kinh Bac Dang, Hieu Nguyen, Thanh Dat Do, Thi Phuong Nga Pham, Tuan Linh Giang, Thi Dieu Linh Nguyen, Huu Hao Ngo, and Giuseppe Forino

Abstract. Current forecasting models for landslides and debris flows mostly look at environmental or socio-economic factors on their own. They rarely combine both into a single probabilistic framework that might give warning in complicated and uncertain situations. This constraint is especially clear in Vietnam, where intense subtropical rain, steep and extensively dissected mountainous terrain, and quick changes in land use and infrastructure are the main causes of landslides and debris flows. This research introduces a novel approach using a Bayesian Belief Network (BBN) to enhance landslide-risk prediction through the integrated analysis of environmental and socioeconomic data. The developed BBN model incorporates inputs from diverse sources, including Geographic Information Systems (GIS), remote sensing, and field survey observations. Structural Equation Modeling was employed to align the BBN with established relationships between landslides and influencing factors. The analysis explored different scenarios by combining rainfall intensity with land-use patterns and assessing the protective role of embankments. Results indicate that precipitation exceeding 130 mm over a period longer than three days markedly increases the likelihood of landslides and debris flows, particularly in agricultural regions. Gabion embankments were found to be highly effective in mitigating risks to both human safety and built environments.

Received: 12 Jan 2026 – Discussion started: 27 Jan 2026

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 paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Kinh Bac Dang, Hieu Nguyen, Thanh Dat Do, Thi Phuong Nga Pham, Tuan Linh Giang, Thi Dieu Linh Nguyen, Huu Hao Ngo, and Giuseppe Forino

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-160', Anonymous Referee #1, 17 Feb 2026
Dear Authors,
I have carefully read your manuscript "Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones". Unfortunately, I have identified several critical issues and shortcomings throughout the manuscript that, in my opinion, make it unsuitable for publication in its current form.
A major concern relates to discrepancies and confusion in the terminology. Based on widely accepted classification schemes, debris flows are a type of landslide. However, the manuscript differentiates two types of phenomena: landslides and debris flows. What do the Authors refer to when using the term landslides? Besides debris flows, what types of landslides does this manuscript address? I think this choice/description should be included in the Introduction or in another section.
A further significant issue is related to the term risk, which appears to be both overused and, in several instances, misapplied. According to Varnes et al. (1984), landslide risk is the expected number of lives lost, persons injured, damage to properties and disruption of economic activity due to a particular phenomenon (in this case, landslides) for a given area and time period. That said, I have found that Authors often use risk as a synonym for the likelihood of landslides. Notably, this probability represents hazard, which is only one component of risk (as correctly illustrated in Figure 4). I strongly suggest the Authors to carefully revise the terminology as they are presenting a research paper.
Another shortcoming lies in how the manuscript describes the novelty and the results. In all fairness, I consider the purpose of boiling down the risk assessment into a single statistical model quite interesting. However, by reading the paper, I had the impression that this objective has not been fully achieved. Specifically, this lack is largely due to terminological inconsistencies and insufficient clarity in both the methodological description and the presentation of results. For instance, if the study aims to produce a landslide risk analysis, why is the well-established risk equation never mentioned in the text? This omission is particularly "strange" as Figure 4 clearly displays risk as the main objective. Furthermore, it remains unclear whether vulnerability and exposure were actually assessed. How were they assessed? The manuscript lacks a description of this part. Both vulnerability and exposure should be analysed differently depending on the type of element at risk.
The outcomes are also not sufficiently supported. As currently presented, the results may appear somewhat predictable. For example (line 373-374): " the higher the perceived rainfall and property risks, the higher the perceived landslide risk" or (line 387) "that weather conditions are an important factor in risk assessment and mitigation". These statements describe relationships that are largely expected. Expanding the supporting analysis would make their contribution clearer.
Ultimately, there is significant confusion regarding the form and scale of the results. Given the landslide context, I would have expected a cartographic translation of the outcomes displaying the spatial distribution of hazard/risk across the study area. CInstead, the results are roughly presented as tables or diagrams. This form limits their interpretability.
For the above reasons, I regret to say that I suggest rejection in its current form. Below, I provide further comments that I hope will help Authors improve the overall quality of their work. Good luck!
Line 31-32. I would include a definition of landslides instead of describing triggering/predisposing conditions, which are quite heterogeneous. For instance, rotational/translational landslides may occur even on gentle slopes. Moreover, as mentioned above, pay attention to landslides vs debris flows: debris flows are a type of landslide!

Line 33-38. I think it is good to report data on losses/damage due to landslides. However, I would follow a more logical flow: world data, Asia data, Vietnam data. There is an increasing availability of data and studies on landslide damage worldwide.

Line 40. How many landslides? What type of landslides? Do economic losses include damage due to floods?

Line 45-118. This part is an example of the confusion in the scientific terminology, as well as the core of the manuscript. What is this work addressing? The Authors leverages the term Risk in the title and throughout the paper; however, the state of the art presented in this paragraph is just describing the framework of landslide susceptibility/hazard. To which, moreover, I would suggest substantial changes and improvements. That said, the literature overview should be refined.

Figure 1, Figure 2. I think these Figures are not suitable within the Introduction section of a research paper. They would have been appropriate for a review.

Section 2.2. This part is the core of the work, but it should be improved. I did not find any definition of warning, for instance, or a clear description of how a Bayesian Belief Network model works.

Line 180. This definition may belong to resilience, not to vulnerability. Moreover, how is vulnerability estimated?

Figure 5. This Figure is not mentioned in the text. What do the Authors mean by thresholds? What does it mean by "landslide hazard higher than 50%"? How are these correlation plots obtained?

Line 255. Is the landslide risk (potential economic/social losses) the dependent variable of the model? This part further supports my idea of general confusion within the work.

Line 263 - 283. The procedures of the landslide risk assessment should be re-written, they are not clear. Moreover, I've never found the risk equation.

What is the spatial resolution of this study? What type of mapping unit is adopted by the Authors?

Section 3.1. I think the relationships reported in this section are expected. For instance, I think anyone would expect slope, or curvature, to be correlated with SPI, since they are products derived from DEM. Moreover, slope and curvature may be a part of the SPI computation. The same consideration applies to the correlation between DEM and TRI (the Authors forgot to define the acronym). Even in this case, anyone would expect a strong relationship between DEM and roughness.

Figure 7. The reading of this Figure is challenging due to its complexity. Moreover, this Figure gives me the opportunity to raise a few points that are not well presented in the methods. Are the predisposing/triggering factors classified? If yes, what kind of classification scheme/method did the Authors use? In the case of data-binning, I think that the entire work is characterized by a degree of subjectivity, which is not addressed.

Overall, the results and their discussion need to be improved.

General comment. The references and literature overview on risk, susceptibility, hazard, vulnerability, and exposure should be improved.
Citation: https://doi.org/10.5194/egusphere-2026-160-RC1
- CC2:
  'Reply on RC1', Hieu Nguyen, 21 Apr 2026
  
  Letter to editors and reviewers:
  Dear editors and reviewers,
  Thank you very much for your useful comments, we are now pleased to resubmit the revised version of Egusphere-2026-160 title: “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones”. Based on the comment of reviewer, the authors changed the title to: “Integrating multidimensional factors through Bayesian Belief Networks for landslide risk reduction in subtropical zones”.
  We would like to thank the reviewers for careful and thorough reading of this manuscript and for your suggestions, which helped us to improve the manuscript. We have carefully considered all the suggested changes and revised the manuscript accordingly.
  Please refer to "Detailed response to reviewers' comments" below for studying the changes.
  Yours Sincerely,
  On behalf of all authors who read and agreed on the revised manuscript.
  Nguyen Hieu
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-CC2
  - CC4: 'Reply on CC2', Hieu Nguyen, 21 Apr 2026
    
    Dear editors and reviewers,
    Because the authors do not know how to upload the revised manuscript, we uploaded it here in the "reply" option.
    Thank you so much for your understanding.
    Best regards,
    Nguyen Hieu
    
    Citation: https://doi.org/10.5194/egusphere-2026-160-CC4
  - AC1: 'Reply on CC2', Kinh Bac Dang, 14 May 2026
    
    Dear editors and reviewers,
    After the authors revised the manuscript based on comments from the second reviewer, we would like to update the new version.
    Thank you so much.
    Best regards,
    Dang Kinh Bac
    
    Citation: https://doi.org/10.5194/egusphere-2026-160-AC1
- AC4: 'Reply on RC1', Kinh Bac Dang, 14 May 2026
  
  Dear editors and reviewers,
  Thank you very much for your useful comments, we are now pleased to resubmit the revised version of Egusphere-2026-160 title: “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones”. Based on the comment of reviewer, the authors changed the title to: “Integrating multidimensional factors through Bayesian Belief Networks for landslide risk reduction in subtropical zones”.
  We would like to thank the reviewers for careful and thorough reading of this manuscript and for your suggestions, which helped us to improve the manuscript. We have carefully considered all the suggested changes and revised the manuscript accordingly.
  Please refer to "Detailed response to reviewers' comments" below for studying the changes.
  Yours Sincerely,
  On behalf of all authors who read and agreed on the revised manuscript.
  Dang Kinh Bac
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-AC4
CC1:
'Comment on egusphere-2026-160', Muhammad Yasir, 19 Apr 2026

I have carefully read your manuscript. This paper proposes a suitable and meaningful approach to landslide risk assessment in subtropical regions. The integration of multiple data sources in the BBN model, supported by SEM, demonstrates a sound scientific basis and potential for practical application. Results such as rainfall thresholds and the role of reinforcement structures are valuable references for risk management.
Overall, the study meets the requirements and can be considered acceptable after revisions to clarify the novelty of the method, the role of the components in the model, and enhance the interpretation of the results.
- The concept of “multidimensional factors” needs to be more clearly defined from the outset, including the scope of variable groups and selection criteria, to avoid the perception of the concept being purely qualitative.
- This study should consider grouping landslides and mudflows together, collectively referred to as landslides. Because mudflows are essentially a type of landslide, while mudslides need to be considered a separate category. - The choice of the BBN model is reasonable, but the presentation doesn't clearly explain why this method is preferred over other popular machine learning models like Random Forest or Neural Networks. The author only needs to add clarification on the advantages of BBN, such as its ability to handle uncertainty, integrate expert knowledge, or build scenarios, thereby helping readers better understand the basis for choosing the method. The combination of BBN and SEM is a strength, but the specific role of SEM in the network structure construction process (e.g., determining causal relationships or parameter adjustment) needs to be clarified to avoid misunderstandings about the level of integration between the two methods.
- Although the paper emphasizes the integration of socio-economic factors, the specific role of this group of factors is not clearly demonstrated in the results section. The author could add a brief statement to clarify the extent of the influence of these factors in the model, thereby highlighting the significance of the multidimensional approach the study aims for.
- The scenario analysis is a strong point; however, the logic behind constructing the scenarios (combining rainfall, land use, and protective structures) is not clearly explained. Adding a short paragraph explaining the reasons for choosing these combinations would help readers understand the representativeness and practical significance of the constructed scenarios.
- The role of reinforcement structures such as gabion embankments is presented in a very positive light. To achieve better academic balance, the author could add a brief statement about the application conditions, scope of effectiveness, or potential limitations (e.g., cost, suitable terrain), thereby making the conclusions more objective.
- One of the advantages of BBN is its ability to handle uncertainty, but this aspect is not clearly emphasized in the methods section. Adding a sentence explaining how the model reflects and propagates uncertainty would further highlight the advantages of the methodology used in the study.
- Field data is mentioned as an important input source; however, its specific role in the model (calibration, validation, or structural construction) is not clearly presented. The author could clarify this point with a brief statement, thereby increasing the reliability and transparency of the study.
- The conclusion currently mainly summarizes the results but does not truly emphasize a key scientific message. The author could add a highly generalized summary sentence clarifying the core contribution of the study to the field of landslide and mudflow risk assessment, thereby increasing the academic weight of the paper.

Citation: https://doi.org/10.5194/egusphere-2026-160-CC1
- CC3: 'Reply on CC1', Hieu Nguyen, 21 Apr 2026
  
  Letter to editors and reviewers:
  Dear editors and reviewers,
  Thank you very much for your useful comments, we are now pleased to resubmit the revised version of Egusphere-2026-160 title: “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones”. Based on the comment of the first reviewer, the authors changed the title to: “Integrating multidimensional factors through Bayesian Belief Networks for landslide risk reduction in subtropical zones”.
  We would like to thank the reviewers for careful and thorough reading of this manuscript and for your suggestions, which helped us to improve the manuscript. We have carefully considered all the suggested changes and revised the manuscript accordingly.
  Please refer to "Detailed response to reviewers' comments" below for studying the changes.
  Yours Sincerely,
  On behalf of all authors who read and agreed on the revised manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-CC3
- CC5:
  'Reply on CC1', Hieu Nguyen, 21 Apr 2026
  
  Dear editors and reviewers,
  Because the authors do not know how to upload the revised manuscript, we uploaded it here in the "reply" option.
  Thank you so much for your understanding.
  Best regards,
  Nguyen Hieu
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-CC5
  - AC3: 'Reply on CC5', Kinh Bac Dang, 14 May 2026
    
    Dear editors and reviewers,
    After the authors revised the manuscript based on comments from the second reviewer, we would like to update the new version.
    Thank you so much.
    Best regards,
    Dang Kinh Bac
    
    Citation: https://doi.org/10.5194/egusphere-2026-160-AC3
- AC5: 'Reply on CC1', Kinh Bac Dang, 14 May 2026
  
  Dear editors and reviewers,
  Thank you very much for your useful comments, we are now pleased to resubmit the revised version of Egusphere-2026-160 title: “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones”. Based on the comment of the first reviewer, the authors changed the title to: “Integrating multidimensional factors through Bayesian Belief Networks for landslide risk reduction in subtropical zones”.
  We would like to thank the reviewers for careful and thorough reading of this manuscript and for your suggestions, which helped us to improve the manuscript. We have carefully considered all the suggested changes and revised the manuscript accordingly.
  Please refer to "Detailed response to reviewers' comments" below for studying the changes.
  Yours Sincerely,
  On behalf of all authors who read and agreed on the revised manuscript.
  Dang Kinh Bac
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-AC5
RC2:
'Comment on egusphere-2026-160', Anonymous Referee #2, 12 May 2026
这篇题为“基于贝叶斯信念网络的多维因素整合方法在亚热带地区滑坡和泥石流风险降低中的应用”的论文，运用贝叶斯信念网络进行了滑坡灾害概率预测，并在越南取得了可靠的结果。然而，缺乏关键信息是该论文的主要缺陷。因此，在被本期特刊“滑坡清单质量对滑坡易发性和灾害地图可靠性的影响”接收发表之前，该论文需要进行大幅修改。
主要评论：
在引言部分，作者应该给出更有条理的解释，说明为什么使用 BBN 网络，以及它的优势是什么。

目前的稿件需要由英语母语人士或专业的语言编辑服务进行润色。现有版本段落过长，导致读者难以理解作者想要传达的重要信息。例如，第 2.2 节“用于预警滑坡和泥石流的贝叶斯网络开发”中存在过多技术描述，虽然部分过程已可视化（图 4 显示了步骤 2），但如果能提供一张展示整体框架的图表，将有助于读者更好地理解。

在2.1节介绍案例研究时，作者应提供越南滑坡和泥石流的总体或详细列表。建议删除讨论部分的第145段，因为它与数据无关。相关的讨论内容应放在引言中。

在讨论结果时，用可视化图表代替表格来展示结果，可以增强读者的理解。
Citation: https://doi.org/10.5194/egusphere-2026-160-RC2
- RC3:
  'Reply on RC2', Anonymous Referee #2, 12 May 2026
  Here is the english version:
  In this manuscript titled “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones” performed landslide hazards probability forecasting based by a Bayesian Belief Networks, showed a reliable result on Vietnam. However, lack of cis the major flaw of the manuscript. Therefore, major revision has be done before this manuscript could be accepted for publication in this special issue: The influence of landslide inventory quality on susceptibility and hazard map reliability.
  Major Comments:
  In the introduction section, the author should present a more structured explanation and provide why the BBN network is used, and what its advantages are.
  
  The current manuscript needs to be polished by a native English speaker or professional language editing service. The existing version has very long paragraphs, making it difficult for readers to follow the important information the author wants to convey. For example, section 2.2 BBN development for warning landslides and debris flows. There are too many technical descriptions, and although some processes are visualized (Fig 4 shows step2), a diagram showing the overall framework would help readers understand it better.
  
  When introducing their case study in section 2.1, the author should provide a general or detailed list of landslides and debris flows in Vietnam. It is recommended to delete paragraph 145 of the discussion, as it is irrelevant to the data. The relevant discussion should be included in the introduction.
  
  When discussing results, adding visualizations to the results instead of table can enhance the reader's understanding.
  
  Citation: https://doi.org/10.5194/egusphere-2026-160-RC3
  - AC2: 'Reply on RC3', Kinh Bac Dang, 14 May 2026
    
    Dear editors and reviewers,
    Thank you very much for your useful comments, we are now pleased to resubmit the revised version of Egusphere-2026-160 title: “Integrating multidimensional factors through Bayesian Belief Networks for landslide and debris-flow risk reduction in subtropical zones”. Based on the comment of first reviewer, the authors changed the title to: “Integrating multidimensional factors through Bayesian Belief Networks for landslide risk reduction in subtropical zones”.
    We would like to thank the reviewers for careful and thorough reading of this manuscript and for your suggestions, which helped us to improve the manuscript. We have carefully considered all the suggested changes and revised the manuscript accordingly.
    Please refer to "Detailed response to reviewers' comments" below for studying the changes.
    Yours Sincerely,
    On behalf of all authors who read and agreed on the revised manuscript.
    Dang Kinh Bac
    
    Citation: https://doi.org/10.5194/egusphere-2026-160-AC2

Kinh Bac Dang, Hieu Nguyen, Thanh Dat Do, Thi Phuong Nga Pham, Tuan Linh Giang, Thi Dieu Linh Nguyen, Huu Hao Ngo, and Giuseppe Forino

Viewed

Total article views: 1,944 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
1,105	736	103	1,944	65	179

HTML: 1,105
PDF: 736
XML: 103
Total: 1,944
BibTeX: 65
EndNote: 179

Views and downloads (calculated since 27 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	505	115	35	655
Feb 2026	247	251	31	529
Mar 2026	246	298	20	564
Apr 2026	86	56	16	158
May 2026	21	16	1	38

Cumulative views and downloads (calculated since 27 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	505	115	35	655
Feb 2026	247	251	31	529
Mar 2026	246	298	20	564
Apr 2026	86	56	16	158
May 2026	21	16	1	38

Viewed (geographical distribution)

Total article views: 1,930 (including HTML, PDF, and XML) Thereof 1,930 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 23 May 2026

Short summary

This study aims to better predict landslides and debris flows in mountainous regions where heavy rain and human activities increase risk. We combined environmental conditions, land use, and local infrastructure in a probabilistic network model using maps, satellite images, and field information. Results show that prolonged heavy rainfall greatly raises risk, especially in farming areas, while protective embankments can strongly reduce threats to people and property.


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