Earthquake Detection Using Stacked Normalized Recurrent Neural Network (SNRNN)

Round 1

Reviewer 1 Report

1.       I recommend authors to merge, add more scenarios, etc. Discuss logically and methodically when comparing to real-world events, not just stating that the RMSE in detecting, which shows that SNRNN method is effective and complete.

2.       Section 3 and 4 lacks sufficient consistency; it is advised that it be reasonably and methodically revised. Take into account the clear flow of facts and the sentences' connections.

3.       The conclusion almost reads like the outcome of a short report and that is perfect.  Make the title as Conclusions and Future work.

4.       Table [1] can ordered with paper references as ascended otherwise group it based on the models it looks clumsy not ordered in any form.

5.       Can elaborate normalization and batch normalization appropriately.

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript and provide your valuable feedback. We have carefully considered all of your comments and suggestions, and have made the necessary revisions to address each one.

In the attached document, you will find a detailed response to each of your comments, along with the specific changes we have made to the manuscript. We hope that these revisions address your concerns and improve the quality of the manuscript.

Thank you for your time and consideration.

Sincerely,

Muhammad Atif Bilal

Author Response File: Author Response.docx

Reviewer 2 Report

On the one hand, the research topic is of great interest. On the other hand, the article is written in a completely incomprehensible language, making it impossible to adequately understand the content of the research and the obtained results. While the introduction, literature review, and method description are somewhat comprehensible and coherent, the seismic part of the article suggests that the authors are not experts in seismology. What do the authors mean by "Figure 3 (c) shows the distribution of events according to their earthquake types. Body wave magnitude (MB) types of earthquakes are most likely to occur in Turkey."? What do they mean by earthquake types? Do they take a certain magnitude type as an earthquake type? How can MB magnitude be the most likely to occur? What do they mean by "Figure 4 presents a risk map of the earthquake catalog in the Turkey region"? The map is constructed in peak accelerations, what does it have to do with the catalog? Only RMSE values are provided in the results. What about seismic results? It is entirely unclear what the authors are detecting when they talk about earthquake detection. The method description mentioned the processing of three-component seismograms, but then there was no further discussion on this matter. Therefore, it is unclear what the authors have obtained. From the text of the article, it appears that the authors are predicting earthquake types, including magnitude type (according to a precise understanding of the article's text). AND THE MOST IMPORTANT QUESTION FOR THE REVIEWER IS: WHAT DO THE AUTHORS MEAN WHEN THEY TALK ABOUT PREDICTION, PREDICTION OF MAGNITUDE, DEPTH, AND TYPE?

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript and provide your valuable feedback. We have carefully considered all of your comments and suggestions, and have made the necessary revisions to address each one.

In the attached document, you will find a detailed response to each of your comments, along with the specific changes we have made to the manuscript. We hope that these revisions address your concerns and improve the quality of the manuscript.

Thank you for your time and consideration.

Sincerely,

Muhammad Atif Bilal

Author Response File: Author Response.docx

Reviewer 3 Report

The paper presents an ensemble learning model based on a stacked normalized recurrent neural network (SNRNN) for earthquake detection. The model can detect the depth, type, and magnitude of an earthquake event. Here are some additional questions/points to be addressed in the paper:

 1. The statement “GRU is an RNN model that uses less memory and is faster when data has longer sequences” is not fully true. A GRU cell has more parameters than an RNN cell.

2. As layer normalization is a central point, the authors should detail the method and compare it with batch normalization.

3. The authors should suggest why the use of the ensemble method (i.e., three layers of SimpleRNN, GRU, and LSTM in a stack) outperforms the performance of independent layers.

4. Why “In order to address the gradient descent problem, GRUs employ the update gate and reset gate.”? What is the relationship between gradient descent and update and reset gates?

5. The authors should justify the use of the softmax function after the GRU layer.

6. In Figure 3, the authors should include the unit of Magnitude.

7. The authors should detail the database, especially, the waveform (e.g., sampling, number of points, etc.).

The article should be handled carefully from the beginning (including revision of English style/grammar and typos).

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript and provide your valuable feedback. We have carefully considered all of your comments and suggestions, and have made the necessary revisions to address each one.

In the attached document, you will find a detailed response to each of your comments, along with the specific changes we have made to the manuscript. We hope that these revisions address your concerns and improve the quality of the manuscript.

Thank you for your time and consideration.

Sincerely,

Muhammad Atif Bilal

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have approached the task of responding to my comments and making corrections to the article in a “lazy” manner. In one or two places, they have made terminological corrections to the text, and in 20-30 other places, they have left it unchanged, for example, Figure 3c. As a reader-reviewer, I am still unclear about what is being fed into neural networks and what is being produced as output. Purely mathematically, everything is probably understandable. Neural networks were taken, “magicked” with the sequence of their use, and obtained a result from the “black box”. But what has been achieved in the field of seismology? Is the article about mathematics or seismology? What type of earthquake is being predicted??? What is the article about?

The authors claim a high level of their article, but it is their right...

In my opinion, the article is insanely confusing and contains “terminological mush”. I consider myself a person of science and cannot give a positive review of this article. The article needs to be either completely rewritten or rejected for further consideration.

I believe it is appropriate to involve one or two more seismologists in the review process. This will allow the journal's editorial board to be even more objective towards this article.

Author Response

We want to express our deepest gratitude to the reviewers for their invaluable feedback. We appreciate your time and effort in reviewing our work and providing constructive comments. We have carefully considered your suggestions and made revisions to the manuscript. As a result, we believe that our research design is appropriate, our methods are adequately described, and our conclusion accurately reflects the results. We have provided detailed responses to each question below. 

Author Response File: Author Response.docx

Reviewer 3 Report

Now, after the round of corrections, the paper is suitable for Applied Sciences.

Now, after the round of corrections, the paper is suitable for Applied Sciences.

Author Response

We want to express our deepest gratitude to the reviewers for their invaluable feedback. We appreciate your time and effort in reviewing our work and providing constructive comments. We have carefully considered your suggestions and made revisions to the manuscript. As a result, we believe that our research design is appropriate, our methods are adequately described, and our conclusion accurately reflects the results. We have provided detailed responses to each question below. 

Author Response File: Author Response.docx