Research on Water Level Anomaly Data Alarm Based on CNN-BiLSTM-DA Model
Round 1
Reviewer 1 Report
Reviewer’s Comments
Minor revision is being suggested for the manuscript id: Water-2318938 titled, “Research on Water Level Anomaly Data Alarm Based on CNN-BiLSTM-DA Model”. The following are specific comments; the author must revise the manuscript and prepare a rebuttal to the comments for further review.
1. Title: Author must avoid unnecessary capitalization, both in title and text, kindly check the whole manuscript and update.
2. Abstract: Line no 12, DA stands for Residual difference?
3. Line 15: pre-diction should be prediction.
4. Line 78-79, 101, 102, 117,… : Author should avoid the abbreviations along with the meaning at multiple times instead the abbreviations can be specified at firs use and later need not to be specified. kindly check the whole manuscript and update.
5. Line 174: check the sentence and rewrite, “A of the difference value sequence;”.
6. Section 2.3.3: Author mentioned the Pearson’s Correlation for the evaluation of the developed model performance. But the Pearson’s Correlation is the measure of the linear association. Instead correlation coefficient may better describe developed model performance. Kindly explore and if significant, please add. Also, it is suggested to refer following articles:
a. https://doi.org/10.2166/wst.2019.257
b. https://doi.org/10.1016/j.jwpe.2020.101477
7. Line no 260-262: please redraft the sentence, “To avoid this adverse effect, this paper uses the difference between the predicted water level value and the corresponding monitoring value at the same time step as input data for residual analysis in the DA model.”
8. Section 4. Discussion, is this really discussion? Author drafted it as conclusions? kindly check it and update its section heading.
9. If this is discussion, then authors need to add conclusions section in the manuscript, which is missing. Also, while writing it author should write conclusions in max 150 word presenting the major findings of the presented research work.
I wish authors a great success.
Comments for author File: 
Comments.pdf
Author Response
Author's Reply to the Review Report (Reviewer 1)
- Title: Author must avoid unnecessary capitalization, both in title and text, kindly check the whole manuscript and update.
The unnecessary capitalization in the title and text has been revised based on the reviewer's comments, and the revised version can be found for details.
- Abstract: Line no 12, DA stands for Residual difference?
Due to lack of further checking after translation, this error occurred. DA stands for Difference Analysis, which has been revised in the text. Please refer to the revised version for details.
- Line 15: pre-diction should be prediction.
The text has been revised according to the reviewer's comments. Please refer to the revised version for details.
- Line 78-79, 101, 102, 117,… : Author should avoid the abbreviations along with the meaning at multiple times instead the abbreviations can be specified at firs use and later need not to be specified. kindly check the whole manuscript and update.
The unnecessary abbreviations and meanings in the text have been removed based on the reviewer's comments. Please refer to the revised version for details.
- Line 174: check the sentence and rewrite, “A of the difference value sequence;”.
This sentence has been rewritten based on the reviewer's comments:” To derive the difference sequence, the measured value sequence is subtracted from the predicted value sequence, followed by computation of the absolute value sequence, denoted as ‘A’, from the resulting difference value sequence.” Please refer to the revised version for details.
- Section 2.3.3: Author mentioned the Pearson’s Correlation for the evaluation of the developed model performance. But the Pearson’s Correlation is the measure of the linear association. Instead correlation coefficient may better describe developed model performance. Kindly explore and if significant, please add. Also, it is suggested to refer following articles:
- https://doi.org/10.2166/wst.2019.257
- https://doi.org/10.1016/j.jwpe.2020.101477
After careful reading of the two recommended references in the review comments, it was decided to refer to these two references and introduce the MAD and RMSE to evaluate the accuracy of the model prediction, instead of using Pearson coefficient . The specific modifications can be found in Section 2.3.3 of the revised version.
- Line no 260-262: please redraft the sentence, “To avoid this adverse effect, this paper uses the difference between the predicted water level value and the corresponding monitoring value at the same time step as input data for residual analysis in the DA model.”
The redrafted sentence is as follows,”To avoid unfavorable impacts of this phenomenon, this study utilized the difference be-tween the water level monitoring value of the present moment and the predicted water level value of the next moment as input data for difference analysis in the DA model.”
- Section 4. Discussion, is this really discussion? Author drafted it as conclusions? kindly check it and update its section heading.
This section is a discussion. Based on the reviewer's comments, additional conclusions have been added in Section 5, which can be found in the revised version's Section 5 conclusion for details.
- If this is discussion, then authors need to add conclusions section in the manuscript, which is missing. Also, while writing it author should write conclusions in max 150 word presenting the major findings of the presented research work.
As per the reviewer's comments, the conclusion section has been supplemented. Please refer to Section 5 of the revised version for details.
Author Response File: Author Response.pdf
Reviewer 2 Report
The paper proposde a CNN-BiLSTM-DA model for water level prediction. The accuracy have been verified in Yancheng city, China.
1. Please draw a figure of daily water level series in Section 2.1.
2. The water level time serise is enssentially a one-dimensioinal data. However, the input data of CNN architecture shown in Figure 1 is a two-dimensional data. Maybe it is not suitable for the water level prediction problem.
3. The detail of the proposed CNN-BiLSTM-DA model architecture is not described clearly. How to integrate the three block, CNN, BiLSTM and DA, into a united model?
4. The training set, test set and validation set in Section 2.1 are not reasonable.
5. line 15: pre-diction;
6. line 117: Long and Short-term Memory Neural Network LSTM.
Author Response
Author's Reply to the Review Report (Reviewer 2)
- Please draw a figure of daily water level series in Section 2.1.
The Figure of daily water level series has been added in Section 2.1 based on the reviewer's comments. Please refer to Section 2.1 of the revised version for details.
- The water level time serise is enssentially a one-dimensioinal data. However, the input data of CNN architecture shown in Figure 1 is a two-dimensional data. Maybe it is not suitable for the water level prediction problem.
The 2D CNN structure diagram has been replaced with a 1D CNN structure diagram. Please refer to Figure 2 in the revised version for details.
- The detail of the proposed CNN-BiLSTM-DA model architecture is not described clearly. How to integrate the three block, CNN, BiLSTM and DA, into a united model?
To provide better clarity on how to integrate CNN, BiLSTM, and DA into a unified model, we have included additional details in the manuscript. Please refer to the revised sections 2.2.4, 2.2.5, and 2.3.4 for further information.
- The training set, test set and validation set in Section 2.1 are not reasonable.
After reviewing some relevant literature, adjustments have been made to the setup of the training set, test set, and validation set, with particular focus on the test set and validation set. For more information, please refer to the revised Section 2.1 of the manuscript.
- line 15: pre-diction;
This section has been revised and the details can be found in line 15 of the revised version.
- line 117: Long and Short-term Memory Neural Network LSTM.
This section has been revised and the details can be found in line 137 of the revised version.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors have prepared an article related to Water Level Anomaly Data Alarm Based on CNN-BiLSTM-DA Model. Please address the following points.
1. In Section 2.1, the dataset description should be elaborated. Since the data is collected from 2015 onwards, explain the different parameters that are considered. Or is it only the water level recorded in the dataset? There can be several parameters that will promote the occurrence of flood event. Hence a detailed dataset description is needed.
2. Section 2.3.4 should be readdressed by elaborating on the architecture diagram of the proposed model. I did not see why CNN-BiLSTM-DA Model is proposed for water level prediction.
3. Section 2.2 is about the explanation of existing models, which is theoretical information freely available on the internet. This section does not link to further sections and to the proposed model. Basically, the justification of the proposed model is missing in the article.
Since the description of the dataset is missing, and the novelty and research contribution is weak, I have to reject the article.
Author Response
Author's Reply to the Review Report (Reviewer 3)
- In Section 2.1, the dataset description should be elaborated. Since the data is collected from 2015 onwards, explain the different parameters that are considered. Or is it only the water level recorded in the dataset? There can be several parameters that will promote the occurrence of flood event. Hence a detailed dataset description is needed.
The study area is located in a plain region where water level is a crucial indicator for flood warnings. Considering the following factors, this paper chooses water level as the warning indicator for flood disasters in plain areas: water level is a direct indicator that reflects flood risk and is easy for people to understand and accept, as well as for warning information to be conveyed; water level data collection is relatively simple and convenient; as a crucial warning indicator for flood disasters in plain areas, the use of water level can better reflect the actual situation; water level change is sensitive and can reflect environmental changes in a short period, providing better timeliness for flood warning compared to other indicators. The emergence of extreme rainfall will cause ab-normal changes in water level, and the realization of water level anomaly data warning plays an indispensable supportive role in achieving prediction and warning of flood dis-asters in plain areas.
Additional descriptions of the research area and data have been added in the manuscript. Please refer to the revised Section 2.1 for more details.
- Section 2.3.4 should be readdressed by elaborating on the architecture diagram of the proposed model. I did not see why CNN-BiLSTM-DA Model is proposed for water level prediction.
A detailed explanation of why the CNN-BiLSTM-DA model is used for water level prediction and flood warning research in plain areas has been added in Section 2.3.4. Please refer to the revised version in Section 2.3.4 for more details.
- Section 2.2 is about the explanation of existing models, which is theoretical information freely available on the internet. This section does not link to further sections and to the proposed model. Basically, the justification of the proposed model is missing in the article.
The rationale for the proposed model has been incorporated into the manuscript, and additional descriptions have been included in Section 2.2 to enhance the overall coherence. Please refer to the Section 1 and Section 2.2 of revised version for more details.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors have addressed all the comments in the revised work. The revised manuscript can be accepted for publication.
