A Hybrid Model for Streamflow Forecasting in the Basin of Euphrates

Round 1

Reviewer 1 Report

This is a much improved version of the earlier submission.  I will recommend a slight modification for lines 33 and 34.  There is a missing word: "every".

This should read: Therefore, the existence and quality of water, which is necessary in every aspect of human life, is very crucial [1].  

Author Response

Therefore, the existence and quality of water, which is necessary in every aspect of human life, is very crucial [1]. 

Necassary recommendation was added.

Author Response File: Author Response.docx

Reviewer 2 Report

See attached file

Comments for author File: Comments.pdf

Author Response

"Please see the attachment."

Author Response File: Author Response.docx

Reviewer 3 Report

Reviewer’s Report on the manuscript entitled:

A Hybrid Model for Streamflow Forecasting in the Basin of Euphrates

The authors proposed a hybrid approach for river flow forecasting. Although the topic and results sound interesting, the presentation and structure of the manuscript must be improved.

Line 15. After Genetic Algorithm (GA), please say for streamflow forecasting.

Please restructure the manuscript as follow:

1. Introduction
2. Materials and Methods

     2.1 Study region

     2.2 Datasets and Pre-processing

     2.3 Methods

           2.3.1 Long Short-Term Memory (LSTM) Network  

           2.3.2 Genetic Algorithm

           2.3.3 The proposed hybrid method

3. Results and Discussion
4. Conclusions

Then update lines 150-155 as follows: The rest of the paper is organized as follows. In Section 2, the study regions ….

Also, please remove lines 157-159.

Line 41. Please also add the following recent articles for streamflow forecasting using the wavelet analysis and other AI techniques:

https://doi.org/10.1016/j.ejrh.2021.100847

https://doi.org/10.1007/s11269-019-02226-7

The first article above also describes how to process the streamflow time series with gaps and missing values without any interpolations that can be briefly discussed in the manuscript.

Line 176. “In Figure 1, …” please use comma.

Lines 177-180. When you refer to the equations, please use the following format: Equation (x). For example, line 178 should be Equations (1)-(4), line 179, Equation (5), etc.

Lines 183-188. The format of all the equations must be corrected. The equations are not readable. Please follow the MDPI guideline to fix the mathematical equations.

Figures 3 and 4. The text and labels in the figures must be clear and not stretched. Also, please ensure that all the figures have a resolution of at least 300 dpi.

Line 304. Where is Figure 5? The figures are not properly referred to in the text.

Figure 7. Please also show a panel below this panel that shows the residual series (observed minus predicted) for LSTM and GA-LSTM. This way the readers can visualize the differences easier.

Line 367. Please remove the dot before [40]. Please check for similar issues elsewhere.

Table 2 and 3. Typically, for indicating decimals, the use of dot is preferred nor comma. For example, 1.2682 not 1,2682. Please correct throughout the entire manuscript as you already did in lines 327-330.

Line 411. In Conclusions, please include the limitations of this study (how many hydrometric stations you used? One, two?) and future direction here.

Please define all the acronyms the first time the appear and be consistent with their style. Always, use the capital letter for the first words, for example, Singular Spectrum Analysis (SSA). Please also add an acronym table at the end of the manuscript listing all acronyms used.

Please check the references for the authors’ names, volume, page numbers, etc. and ensure that they follow the MDPI guidelines.

Finally, please very carefully proofread the manuscript.

Thank you for your nice contribution

Regards,

Author Response

"Please see the attachment."

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Thank you for the addition of the linear regression model. I think this paper is sufficiently improved for publication.

Author Response

''Please see the attachment''

Author Response File: Author Response.docx

Reviewer 3 Report

I would like to thank the authors for addressing my comments. The manuscript looks much better now and I recommend acceptance.

Please carefully proofread the article if accepted by the editor and kindly address the following editorial issues:

Lines 89, 94, etc. Please give the article number instead of the year. For example, "Xu et al. [21] ...". Please check and correct throughout the entire manuscript.

Line 463. It should read as "A long computational ..."

Line 494. It should be Least-Squares Wavelet Software

Please improve the resolution (at least 300 dpi) of ALL the figures, the numbers and texts should be clearly visible not blur/fade.

Thank you for your contribution

Regards,

Author Response

''Please see the attachment''

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This is an interesting topic which is of high interest to readers, however, the presentation and writing style can be improved.  I have made some hand-written editing (in red color) on the manuscript which I have scanned and attached to this review.  These are my suggestions on some ways to improve the quality of your presentation. 

Overall this is a very important and relevant topic, and I commend the authors for their work.   

Comments for author File: Comments.pdf

Reviewer 2 Report

GENERAL COMMENTS

Title: A Hybrid Model for Streamflow Forecasting in the Basin of Euphrates

The manuscript presented a hybrid model using LSTM and GA, and it was applied to forecast streamflow at the Euphrates River in Turkey. The major issue of this paper is that the idea of combining GA and LSTM or its application to streamflow forecast is not new. It suggests that authors might have not sufficiently investigated the literature. The comparison made in the results section was not proper. There are many inconsistent statements throughout the paper (e.g. how does GA improve the LSTM model), and the quality of presentations (e.g. figures) is low.

Major comments:

1. Introduction

Line 46-67, the relationship between ANN, AI, RNN, LSTM, and GA methods and their advantages and limitations are not presented clearly. Also, in the end, it concludes that it is necessary to propose a universal and automated artificial intelligence model. The reasoning is not strong.

Line 83-89, the importance of GA and why it is necessary for LSTM is not discussed intensively.

2. Related works

Following my comments above, there are a list of works without linking with each other or the focus of this work. More importantly, why GA is needed for LSTM is not mentioned at all.

3. Optimization and Hybrid-AI Model

From my point of view, Section 1 and 2 could be in one section and this section could be Section 2. Also, where are Equations (1) and (2)?

The major issue of this paper is that the idea of combining GA and LSTM or its application to streamflow forecast is not new. The literature is not sufficiently investigated (for example, see papers below). Coincidently, in Section 3. Optimization and Hybrid-AI models of Ibrahim, Huang, Ahmed, Koo, and El-Shafie (2021), they have already listed GA as one of the popular hybrid-AI models and some other works have been presented.

Ibrahim, K. S. M. H., Huang, Y. F., Ahmed, A. N., Koo, C. H., & El-Shafie, A. (2021). A review of the hybrid artificial intelligence and optimization modelling of hydrological streamflow forecasting. Alexandria Engineering Journal, 61(1), 279-303. doi:https://doi.org/10.1016/j.aej.2021.04.100

Moeeni, H., Bonakdari, H., Fatemi, S. E., & Zaji, A. H. (2017). Assessment of stochastic models and a hybrid artificial neural network-genetic algorithm method in forecasting monthly reservoir inflow. INAE Letters, 2(1), 13-23.

Chang, L. C., Chang, F. J., Wang, K. W., & Dai, S. Y. (2010). Constrained genetic algorithms for optimizing multi-use reservoir operation. Journal of Hydrology, 390(1-2), 66-74.

In a word, authors need to highlight your contribution and what is new in your hybrid-AI model.

4. Research Data and Experiment

Just wondered if Figures 4 and 5 can be combined into a single figure since it is not necessary to have two figures showing the study area and stations. Also, the resolution of both figures is low.  Although a number of FMS stations are listed in Table 1, only one of them was used in the study. The reasoning is briefly mentioned (due to the number of observations available), but according to the Observation (year), E21A31 has the shortest duration of records. Maybe authors should include the percentage of data missing for each station in the table as well.

In addition, only four stations listed in Table 1 were presented in Figure 5.

5. Prediction results for designed model

I am not sure what the authors mean by “Test data of ten-day streamflow chosen randomly selected from long-term annually”. Why not just show the whole test data (i.e. last four years). It is even more confusing in Table 3. Are they just 10 days of streamflow from the whole dataset? First, there is no MAEs for each day, but error. Second, the comparison should be made using testing data for verifying model performance. The ten randomly selected samples cannot show that the proposed method outperforms the other one.

Figure 6, the unit of streamflow is missing. Also, it is better to keep the x and y-axis the same between subplots for easy comparison.

Table 2, there is no need to show both MSE and RMSE. The unit for MSE and MAPE (%) is wrong in the table. Also, STD.DEV should include the standard deviation of observed streamflow, otherwise, it cannot show which model is better. In a word, the comparison of STD.DEV. should be made against observations.

Most importantly, how GA improve the LSTM model in choosing the optimal initial weights or determining the window size or number of units was not explicitly shown and discussed here, and it just included a range of applications in other works.

6. Conclusion

It is concluded that GA improves the parameterization of the LSTM model, but how it differs from the LSTM only model is not shown in the paper.

Also, the small standard deviation means the values are close to each other, but for streamflow data, they are highly skewed with lots of small values and several large values (e.g., flood events) so its standard deviation should be higher theoretically. Nevertheless, the standard deviation should compare against observation not look at their absolute value.

Minor comments:

Line 44-45: How does it link to human activity plans and energy policies? Any reference?

Line 53: ANN method; ?

Line 53:  such as modelling non-linearity?

Line 109: What does PCA stand for?

Line 118-120: What does TSA stand for?

Ibrahim, K. S. M. H., Huang, Y. F., Ahmed, A. N., Koo, C. H., & El-Shafie, A. (2021). A review of the hybrid artificial intelligence and optimization modelling of hydrological streamflow forecasting. Alexandria Engineering Journal, 61(1), 279-303. doi:https://doi.org/10.1016/j.aej.2021.04.100