Short-Term Streamflow Forecasting Using Hybrid Deep Learning Model Based on Grey Wolf Algorithm for Hydrological Time Series
Round 1
Reviewer 1 Report
- Fit Figure 1 and 2 into one figure.
- You can mentioned also in Introduction part machine learning methods like regression and model trees (i.e. Štravs and Brilly, 2007, Development of a low-flow forecasting model using the M5 machine learning method, Hydrological Sciences Journal...)
- Can you show results (measured vs predicted values) similar as graph on Figure 3? On Figure 3 show "Date" on x-axis.
- Can you use your model to predict flows for year 2021 and compare it? It would be interesting to show how your model behaves...
- Why didnt use as input air temperature and precipitation data? Better explain statement in row 566-568. (you can see paper Kompare. 1997. Prediction of rainfall runoff from catchment by intelligent data analysis with machine learning tools within the artificial intelligence tools…).
- Do you have some future plans to improve the model and use it for management purposes, flood alerts…?
Author Response
''Please see the attachment''
Author Response File: 
Author Response.docx
Reviewer 2 Report
1) Hydrological predicting plays a crucial role in water resources management, which, nevertheless, faces a huge challenge due to high uncertainties on climate change. In this manuscript, author employed a deep learning model integrating grey wolf algorithm and gated recurrent unit to forecast short-term streamflow. However, the manuscript still needs a big revise to touch a scientific paper standard such as review history study, modeling and description, making charts, drawing conclusion and language.
2) In introduction, author should further focus on research progress concerning with short-term streamflow forecasting including methodology and models. In addition, some citations are not canonical.
3) In materials and method, it is notable that the methods should be clearly described, especially integrated process of GWO and GRU. And text description needs to be consistent with corresponding figure. In addition, some parameters’ values to be determined by evidences, for example, why was the dataset divided into training and test datasets by 80% and 20%, and what principle is it based on? Training sample size plays an important role in training result of deep learning model.
4) In results, I don’t think that it makes much significance to compare the improved model (GRU-GWO) with RNN model and GRU model in hydrology science research perspective, because all of them are “black box” model. To prove its advantage, including predicting accuracy and universality to application, in hydrological forecasting research, this model should compare with some models based on physical processes. In addition, the results of simulation and forecasting of these models need to show in chart format.
5) The conclusion needs a big revise to answer the proposed scientific question, not review.
6) The language needs further polishing.
Special comments:
Line 12: I think that river flow management is not a key driver of sustainability.
Line 19: Why the GWO-GRU model is more effective than other models is not clear?
Line 21: GRU-GWO or GWO-GRU? Expression of name of model needs to be consistent in whole manuscript.
Line 59: Citation is incorrect, should be [8-10].
Line 62: Citation is incorrect.
Line 69: Citation is incorrect.
Line 74: Citation is incorrect.
Line 122: Expression of LSTM needs to be consistent in whole manuscript.
Line 209: Fig. 1 requires more map elements to enhance the it's readability, and suggest to merge Fig1 and Fig 2.
Line 211: 2.2 section needs to be condensed.
Line 247: In Table 1, the observation time of training data is inconsistent between the two hydrological observation stations.
Line 314: In figure 4, more legend information is required to improve readability of this figure.
Line 315-317: In the formulas 1,2 and 3, some key parameters need to be described detailly in text.
Line 400: Why was the dataset divided into training and test datasets by 80% and 20%, and what principle is it based on? Training sample size plays an important role in training result of deep learning model.
Line 413: The figure 6 is not cited in text.
Line 455: The figure 7 needs an improvement.
Author Response
''Please see the attachment''
Author Response File: Author Response.docx
Reviewer 3 Report
This is an interesting article; however some changes/improvements are necessary.
In the abstract you should add some numerical results as derived from the study.
In the introduction, when state the problem also mentions the application of Grey Wolf Algorithm in other application and fields.
- Negi, G., Kumar, A., Pant, S., & Ram, M. (2021). GWO: a review and applications. International Journal of System Assurance Engineering and Management, 12(1), 1-8.
Highlight the importance of drought on streamflow and the reported changes of seasonal drought conditions with statistical techniques.
- Myronidis, D., Ioannou, K., Fotakis, D., & Dörflinger, G. (2018). Streamflow and hydrological drought trend analysis and forecasting in Cyprus. Water resources management, 32(5), 1759-1776.
- Stefanidis, S., & Alexandridis, V. (2021). Precipitation and Potential Evapotranspiration Temporal Variability and Their Relationship in Two Forest Ecosystems in Greece. Hydrology, 8(4), 160.
Author Response
''Please see the attachment''
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Most comments have been revised, nevertheless, there are some suggestions as followed.
1) Some citation in text still weren’t revised, such as in line 78.
2) Quality of figure 1 needs to be improved including resolution of image, the spatial layout of main and subordinate images.
3) In Figure 6, it is better for comparing which method is better to put all predicting results from three methods including benchmark, regression and deep learning method into one chart. In addition, merging Figure 6 and Figure 6 Cont. is more suitable for scientific paper expression requirements.
Author Response
''Please see the attachment''
Author Response File: Author Response.docx
Reviewer 3 Report
None of my suggestion adressed by the authors,
Please make the necessary changes to the article as below.
In the abstract you should add some numerical results as derived from the study.
In the introduction, when state the problem also mentions the application of Grey Wolf Algorithm in other application and fields.
Negi, G., Kumar, A., Pant, S., & Ram, M. (2021). GWO: a review and applications. International Journal of System Assurance Engineering and Management, 12(1), 1-8.
Highlight the importance of drought on streamflow and the reported changes of seasonal drought conditions with statistical techniques.
Myronidis, D., Ioannou, K., Fotakis, D., & Dörflinger, G. (2018). Streamflow and hydrological drought trend analysis and forecasting in Cyprus. Water resources management, 32(5), 1759-1776.
Stefanidis, S., & Alexandridis, V. (2021). Precipitation and Potential Evapotranspiration Temporal Variability and Their Relationship in Two Forest Ecosystems in Greece. Hydrology, 8(4), 160.
Author Response
''Please see the attachment''
Author Response File: Author Response.docx
