A Novel Method for Predicting Landslide-Induced Displacement of Building Monitoring Points Based on Time Convolution and Gaussian Process
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn this research, a a predictive framework was introduced for displacement estimation, and this is suitable for the journal. Questions are listed as follow,
1) qualities of some figures should be improved, such as Fig.5 and Fig.9
2) The amount of training data has a huge influence on the accurency in forecasting results of slope displacements, and the authors should have a deeper discussion on this.
3) It seems that the results shown in Fig.13 is better than that in Fig.14. However, the error and value in Table 3 show different. Please have a check.
4) More cases should be employed to verify the new proposed method.
Readable
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper presents a new methodology for landslide prediction using a new hybrid predictive framework that synergistically integrates Gaussian Process Regression (GPR) with Temporal Convolutional Neural Networks (TCN), called the GTCN model.
The paper follows a logical structure, presenting a logical sequence of introduction, Deep Learning Methods, Case Study (methodology), results/discussion and conclusion. However, it would be interesting to correct some aspects in the abstract. I suggest that the results be presented in the abstract in a descriptive way and not in a numbered way, as it is not common to do so in this way. An excessive number of keywords is also observed, which could be reduced.
The paper presents the necessary and appropriate tables and figures, but it would be interesting to add a map of the province showing the location of the elements (reservoir, river and district). This map could also contain the geology of the region. It would also be interesting to present the data demonstrated in the conclusion, which describes comparisons between the techniques applied, in graphs or tables, throughout the results section.
The introduction is well written and presents the problem, the scientific gap and the innovative proposal of the study. The text demonstrates up-to-date knowledge when addressing the advancement of deep learning algorithms and their applications in landslide prediction. It also presents 35 references, which indicates a good bibliographic review on the subject. However, it would be interesting to include quantitative data on loss of lives or values/damages caused by landslides. In addition, we have a relatively large section, between lines 66 - 75 in which we have no references.
The Deep Learning Methods section and the methodology are well organized and provide a good description of the techniques adopted in the study, which are in line with the most advanced practices in data science applied to geotechnical engineering. The explanation of the equations is pertinent and all this context contributes to the reliability, comparability and reproducibility of the results.
However, there are some points that could be improved. Firstly, the study area could include a better description of the geology, since this constitutes an extremely relevant factor for landslides. The inclusion of rainfall and climate classification data would help to better contextualize the characteristics of the region. It would also be interesting to present a map of the province, including the reservoir, the river and the district, to better situate the reader in the context.
When presenting the criteria that led to the choice of JCD11 as a monitoring point, no reference is observed, but it is necessary to describe and reference which other studies adopted these criteria to choose the main point.
A similar situation occurs when choosing the data collection period (December 2011 to December 2013). It would be useful to explain the reason for using this period, presenting references that confirm that this time is sufficient. Another important issue to be described is what would be a time index. It should also be noted that in part of the methodology and in the presentation of the equations we do not have references, this should be verified. All these issues are described in the attached PDF.
The "Results" chapter is well organized and divided into sections that facilitate understanding. The finding that all models capture the growth trend behavior well and the criticisms of GGRU for its poor performance with scarce data are pertinent. The section on Cumulative Displacement Forecasting reports that the analysis of imputation methods has a decisive impact on the final forecast, which is an interesting finding.
However, several aspects of the presentation and discussion of the results could be improved. In Section 4.1, lines 316 to 319, it is reported that, when the Time Index assumes values ​​close to 105, both models present relatively lower predictive accuracy. It would be useful to say whether this has already been observed in other studies, that is, whether these data converge or diverge from the existing literature on the subject.
In Section 4.2, line 339, it is stated that "the trend change, characterized by positional changes influenced by geological conditions, typically shows a gradual increase over time", but it is not stated who stated this. It is important to provide references that confirm the statements.
A similar problem is observed in lines 349-350, where it is stated that "In addition, the distribution of forecast errors is relatively uniform, with no significant anomalies observed", but it is not discussed whether this agrees or differs from other studies.
The results need a more detailed quantitative discussion, describing why certain models are more sensitive and comparing them with other studies.
The lack of a comparative analysis with other studies is also a relevant gap in the paper. Several sentences are presented, but it is not described whether that finding has already been observed in other studies, or whether these data converge or diverge from the existing literature on the subject.
There are practically no references cited throughout the results/discussion. The last reference (44 and 45) is cited in line 277, still in the methodology section, and from there we have the entire section 4 without references.
Regarding the conclusion, I suggest that this section be rewritten, as we have parts that should be in the results and other sentences such as "With an increase in the volume of data, the stability of predictions is influenced by historical data, which may lead to increased prediction errors" that could be in the introduction, since this statement is an issue already defined in the literature and not a finding of this research.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsIn this paper, the landslide-induced displacement of a building was assessed using a novel hybrid predictive framework known as the GTCN model, which integrates Gaussian Process Regression (GPR) with Temporal Convolutional Neural Networks (TCN). The GTCN model was validated using time-series displacement monitoring data from a slope prone to instability near the Zhongliang Reservoir in Wuxi County, Chongqing.
The rationale for this investigation is well presented through a review of current challenges and relevant literature. The methodologies are clearly described, the figures and the tables included in the manuscript are appropriate and necessary. The referencing is relevant and supports the study’s context. In conclusion, the study contributes to a more comprehensive understanding of prediction of landslide-induced displacement of engineering structures, thus offering valuable insights for the geotechnical hazard mitigation and resilient infrastructure management. Therefore, minor revisions are recommended to improve clarity and ensure alignment between the text and visual data.
Comments to the authors given below:
- 10: Please clearly define the term discontinuity site to enhance understanding of its role in the analysis.
- The author only considered rainfall and reservoir water level as influencing factors for the displacement. How do geological factors influence the displacement trends? This aspect should be discussed to provide a more comprehensive analysis.
- In line 379, the manuscript states: "It provides reasonable predictions of landslide displacement, characterized by a reduction in unexpected peaks or abrupt changes." What factors contribute to the reduction in these unexpected peaks or abrupt changes? Please clarify and support with evidence.
- The authors should correlate the training and testing displacement data with rainfall, reservoir water level, and other potential influencing factors. Furthermore, they should discuss the displacement mechanisms in relation to these variables, as defined in Equation (8).
- Accurate prediction of landslide-induced displacement is essential for assessing the stability of rock slopes, particularly in complex geological environments. The trends and tendencies of surface displacement/continuous rock slope deformation are influenced by a variety of factors, including geological structure, weather conditions, and human activities. I recommend that the authors explore these additional factors in greater detail. Please refer to the following papers for further reference
Amagu, C.A., Zhang, C., Sainoki, A. et al.Analysis of Excavation-Induced Effect of a Rock Slope Using 2-Dimensional Back Analysis Method: A Case Study for Clay-Bearing Interbedded Rock Slope. Geotech Geol Eng 42, 6315–6337 (2024). https://doi.org/10.1007/s10706-024-02893-3
- Were there any specific challenges or limitations encountered during the analysis?
- The authors should suggest practical methods to prevent continuous slope displacement within the studied landslide area.
- Limitations and Future Research: The article should make the potential for future research more specific
Comments on the Quality of English Language
Although the language of the manuscript is generally understandable, some improvements seem necessary.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe current version could be accepted.
