Exploring the Use of Pattern Classification Approaches for the Recognition of Landslide-Triggering Rainfalls
Round 1
Reviewer 1 Report
- Use a more representative dataset. The dataset used in this study is limited to 1090 rainfall time series that triggered landslides. This is a relatively small number of data points, and it is possible that the model is not able to generalize well to other landslide events. A more representative dataset would be one that includes a wider variety of landslide events, from different locations and with different triggering mechanisms.
- Consider other factors that can trigger landslides. The model in this study only considers rainfall as a trigger for landslides. However, there are other factors that can also contribute to landslides, such as soil type, slope angle, and vegetation cover. A more comprehensive model would need to consider all of these factors.
- Use a more sophisticated model. The model in this study uses a simple LSTM model. There are more sophisticated models available, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs). These models may be able to learn more complex patterns in the data and provide better predictions of landslide triggering.
- Evaluate the model on a larger dataset. The model in this study was only evaluated on a small dataset of 1090 rainfall time series. A more rigorous evaluation would involve testing the model on a larger dataset of landslide events. This would help to ensure that the model is not overfitting to the training data.
- Make the model more interpretable. The model in this study is a black box model, meaning that it is not possible to understand how it makes its predictions. This can make it difficult to trust the predictions of the model. A more interpretable model would be one that allows users to understand how the model is making its predictions.
- This study uses rainfall stations near slope disasters to analyze the rainfall that causes disasters. However, the spatial range of the data used is very wide, and the author states that the geology and rainfall in the study area are very heterogeneous in space. Therefore, is it appropriate to use the entire space of data to train the same model? If we first classify according to terrain slope, and then use slope hazard rainfall records with similar slopes for modeling, the results should be better and more reasonable.
- Section 3.2 is the model performance evaluation. It is recommended to add the equation for each evaluation indicator.
- This study applies LSTM to estimate disaster-causing rainfall. In order to have a clearer understanding of the performance of LSTM, it is recommended to compare traditional rainfall warning values ​​with the results of this study. That is to say, you should find a benchmark for comparison instead of just applying indicators for self-evaluation.
- Figure 1 should include a north arrow.
There are many typos and verb and grammatical errors in the article, showing the author's carelessness in writing. Please first use the built-in grammar and spelling check functions of the document software to conduct a comprehensive check.
Author Response
Dear reviewer, I would than you for your valuable comments.
Please find attached below the replies to your comments
Author Response File: Author Response.docx
Reviewer 2 Report
The study is very interesting. I am inclined to accept the manuscript after addressing the following comments. All the comments are suggestion for improvement.
“Model performances have been evaluated by several statistical indicators, which showed that the defined model was capable of identify the rainfall condition associated with landslide with a high degree of accuracy and a low rate of false positives” Please add the results of the indicators”.
Introduction is very short. It should cover the following: Brief introduction, Literature review, Problem statement, Objectives. Also recent works need to be included, for example: The authors can support the work by highlighting the advantages of LSTM models: A novel application of transformer neural network (TNN) for estimating pan evaporation rate AND Modelling monthly pan evaporation utilising Random Forest and deep learning algorithms
Test site better to be under Methodology.
In the Methodology the authors are encouraged to show the equations that used to evaluate the models. The authors can refer to https://doi.org/10.1038/s41598-023-41735-9
“Figure 3 Flow chart of the data processing chain. The whole procedure was repeated for each time interval.” Not clear.
“Figure 7 Schematic representing the temporal distribution of landslide probabilities over the vali- dation period.” Not clear
Author Response
Dear reviewer, I would like to thank you for your valuable comments.
Please find in attachment the replies to your comments.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
All of the revision suggestions I gave to the author have been fully addressed and appropriately incorporated into the text. I have no other comments.