Examination of Single- and Hybrid-Based Metaheuristic Algorithms in ANN Reference Evapotranspiration Estimating

Round 1

Reviewer 1 Report

Overall comment: This is an interesting study. However, more details are needed to show how the authors conducted the research.

The authors used ANN model to predict reference ET0. ET0 that calculated from FAO-56 PM was input into the model, and ET0 was output from the model (which is hard to understand). The ANN model was integrated with PSOGWO to optimize the hyperparameters.

Comment 1: In the introduction section, the author took long paragraphs to introduce other research, while YOUR OWN IDEAS were less written. As a research paper, the article does not have an introductory text with the hypothesis, research objectives, and methodology overview.

Comment 2: Please provide a figure to illustrate the study area in Section 2.1.

Comment 3: The authors said that “Due to exceptional circumstances i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was loss”, so they used secondary data (Line 155-157). Please explain the difference between the so-called “lost data” and the secondary data.

Comment 4: The author used secondary data as the dataset, with many meteorological data (Line 155-162). It would be advisable to match them with the parameters in Equation 1 (Line 167-171) to make the readers know which data match which parameter.

Comment 5: The symbols in Equation 1 did not match the text (e.g. Tave and T in Line 168).

Comment 6: Please elaborate on how to preprocess the data using SSA and MI in Section 2.3.

Comment 7: In Section 2.4., the authors spent too much space on introducing the algorithms proposed by OTHERS (e.g. GWO, PSO, PSOGWO). However, I do not understand how the authors used those methods in their research and estimation of ET0. Moreover, Fig 1 only showed the algorithm flowchart of PSOGWO that was taken from OTHER RESEARCHERS. The key points of the authors’ original work were poorly written in the manuscript.

Comment 8: It is hard to understand why the author input ET0 from FAO-56PM into the model (Line 240) and acquired ET0 as output data (Line 242).

Comment 9: How did the authors integrate PSOGWO with ANN? Please explain this in the method section.

 

Comment 10: In the method section, please describe the software that you used to perform statistical analysis and generate graphs, and report the settings and workflow as you did for the other parts. Also, please provide the source code on GitHub to ensure scalability and allow evaluation of the code adopted.

Moderate editing of the English language is required.

Author Response

 

 

 

10^th September, 2023

 

 

 

Dear Dr. Lyra Xu,

 

Thank you for giving us the opportunity to submit a revised draft of our manuscript titled “Examination of Single- and Hybrid-based Metaheuristic Algo-rithms in ANN Reference Evapotranspiration Estimating” [sustainability-2546001] to Sustainability Journal. We appreciate the time and effort that you and the reviewers have dedicated to providing valuable feedback on our manuscript. We are grateful to the reviewers for their insightful comments, which have improved the paper. We have incorporated changes that reflect all the suggestions provided by the reviewers. All changes are highlighted in green in the revised manuscript. Please see below for point-by-point responses to the reviewers’ comments. If any responses are unclear or you wish additional change, please do not hesitate to let us know.

 

 

Thank you for your consideration.

 

 

 

 

 

Thank you for your consideration.

Sincerely,

Asst. Prof. Dr. Salah L. Zubaidi

Department of Civil Engineering, Wasit University, Wasit, 52001, Iraq.

Email: [email protected]

On behalf of the rest of co-authors.

Point-by-point response to the comments:

 

 

 

Reviewer #1 comment:

Overall comment: This is an interesting study. However, more details are needed to show how the authors conducted the research.

The authors used ANN model to predict reference ET0. ET0 that calculated from FAO-56 PM was input into the model, and ET0 was output from the model (which is hard to understand). The ANN model was integrated with PSOGWO to optimize the hyperparameters..

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction section, the author took long paragraphs to introduce other research, while YOUR OWN IDEAS were less written. As a research paper, the article does not have an introductory text with the hypothesis, research objectives, and methodology overview.

 Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- Please provide a figure to illustrate the study area in Section 2.1.

 Response: In the current version of the paper, Figure 1 illustrate the study area.

 

3- The authors said that “Due to exceptional circumstances i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was loss”, so they used secondary data (Line 155-157). Please explain the difference between the so-called “lost data” and the secondary data.

 Response: We revised the sentence to be “Due to exceptional circumstances, i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was missing. Secondary (i.e., satellite) data offered by the National Aeronautics and Space Administration” .

4- The author used secondary data as the dataset, with many meteorological data (Line 155-162). It would be advisable to match them with the parameters in Equation 1 (Line 167-171) to make the readers know which data match which parameter.

  Response: This has now been addressed in the current version of the manuscript.

 

 

5- The symbols in Equation 1 did not match the text (e.g. Tave and T in Line 168).

  Response: This has now been addressed in the current version of the manuscript.

 

 6- Please elaborate on how to preprocess the data using SSA and MI in Section 2.3.

  Response: This has now been addressed in the current version of the manuscript (Section 3.1).

 

7- In Section 2.4., the authors spent too much space on introducing the algorithms proposed by OTHERS (e.g. GWO, PSO, PSOGWO). However, I do not understand how the authors used those methods in their research and estimation of ET0. Moreover, Fig 1 only showed the algorithm flowchart of PSOGWO that was taken from OTHER RESEARCHERS. The key points of the authors’ original work were poorly written in the manuscript.

  Response: You have raised an important point. All these issues are amended as below.

• In Section 1, we added some paragraphs and revised the aim and ojectives
• Figure 2 presents a workflow diagram showing the steps involved in univariate ETo simulation.
• Section 4.2 ANN-based MHAs

 

 8- It is hard to understand why the author input ET0 from FAO-56PM into the model (Line 240) and acquired ET0 as output data (Line 242).

  Response: We work according to the methodology that was applied successfully by Ferreira et al., 2020, Nourani et al., 2020, and Sayyahi et al., 2021 by using previous data (Lags) to simulate future data (i.e., univariate technique). However, we have taken this advice and added Figure 2. Also, Figure 5 shows how to determine the best possible input scenario for the model using the MI technique. Moreover, Section 5 clarifis this issue.

• Ferreira, L.B.; da Cunha, F.F. Multi-step ahead forecasting of daily reference evapotranspiration using deep learning. Computers and Electronics in Agriculture 2020, 178, doi:10.1016/j.compag.2020.105728.
• Nourani, V.; Elkiran, G.; Abdullahi, J. Multi-step ahead modeling of reference evapotranspiration using a multi-model approach. Journal of Hydrology 2020, 581, 124434.
• Sayyahi, F.; Farzin, S.; Karami, H.; Cai, N. Forecasting Daily and Monthly Reference Evapotranspiration in the Aidoghmoush Basin Using Multilayer Perceptron Coupled with Water Wave Optimization. Complexity 2021, 2021, 6683759 doi:10.1155/2021/6683759.

 

 

9- How did the authors integrate PSOGWO with ANN? Please explain this in the method section.

  Response: This has now been addressed in the current version of the manuscript (Section 3.4).

 

10- In the method section, please describe the software that you used to perform statistical analysis and generate graphs, and report the settings and workflow as you did for the other parts. Also, please provide the source code on GitHub to ensure scalability and allow evaluation of the code adopted.  

  Response: This has now been addressed in the current version of the manuscript Lines 257-259, 408, and 428. Also, The code of PSOGWO-ANN algorithm was attached as supplementary materials.

 

 

 

Reviewer #2 comment:

 

In the present study, some metaheuristic algorithms are evaluated in the estimation of reference evapotranspiration using artificial neural networks.

In recent years, many works based on artificial intelligence have been presented in the estimation of reference evapotranspiration, both at monthly and daily scales. Also, combinations have been made with other algorithms to adjust their hyperparameters, among the most used are the so-called evolutionary algorithms

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction there is no mention of the existing problem to be solved.

Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the problem and novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- The meteorological input variables to the models are not mentioned.

Response: We used the meteorological variables to calculate the ETo time series (Section 2.1). Afterthat, Figure 5 shows the best possible input scenario (Lags of ETo) for the model (univariate technique). Also, Section 5 clarifies this issue.

3- It would be worthwhile to compare the proposed model with an AI model where the hyperparameters are not adjusted, or to use an empirical model such as Hargreaves-Samani, since it seems that the study complicates the estimation of the reference evapotranspiration.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. Also. Using a trial-and-error procedure to establish these hyperparameters can be risky be-cause of the high computational complexity and potential for mistakes [77]. Also, it is possible to overfit the data if a high number of neurons are added to the hidden layer using the trial-and-error procedure [78]. (lines 337-340). Moreover, three recent review papers in the different fields of hydrology showed that hybrid models outperformed single models. However, we have taken the advice of “Reviewer 3” and applied MPSO-ANN model. Thanks

 

4- In general, in the form in which the work is presented, it does not make a novel contribution to the study of evapotranspiration.

Response: We thank the reviewer for this suggestion, as we feel now that the novelties are much more clearly elaborated in the paper. In the current version of the paper, we have made a stronger effort to highlight the novelties of the study. The edits related to this point can be found in:

• Add several paragrphs and even revised the aim and objectives in Section 1 to highlights the novility.
• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 178-181 to clarify the main subjects of methodology.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.
• little amendments to the title and revised the manuscript in different sections accordingly.

 

 

Reviewer #3 comment:

The manuscript introduces a novel hybrid model that integrates the Slime Mould Algorithm (SMA) with the Artificial Neural Network (ANN) to predict Reference Evapotranspiration (ET0). The study also compares the performance of SMA-ANN with four other hybrid models (PSO-ANN, GWO-ANN, WOA-ANN, and MFO-ANN) in predicting ET0.

Overall, while the research direction is promising and aligns with the scope of this journal, there are areas in the manuscript that require improvement in terms of readability and addressing certain limitations.

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

Limitations:

 

1- The study is limited to testing on four sites in Iraq, which might restrict its applicability in other regions.

Response: This was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks.

 

2- The research only considers four hybrid models for comparison. There might be other potentially more effective hybrid models that have not been considered.

Response: This is a good point to increase the prediction range and decrease the uncertainty. As we mentioned in section 3.4, “Selecting an effective MHA is complex and presents additional difficulties, requiring different MHAs”. We applied an additional hybrid model (i.e., MPSO-ANN) and compared its performance with the other hybrid models (Sections 4.2 and 4.3).

 

Considering the above, I believe that the manuscript can be accepted for publication in the journal after revisions.

 

Specific Comments and Suggestions:

 

3- Language and Readability: The language of the manuscript needs polishing and further enhancement. The current readability is somewhat low, and it would benefit from a thorough proofreading.

Response: We have taken this advice and fully revised the manuscript to a much better standard of English and checked by a native speaker. We believe the paper now is much better.

 

4- Abstract: The abstract should state the research hypothesis or theoretical basis. Additionally, it would be beneficial to include the scientific significance of the study at the end of the abstract.

Response: Amendments have been made accordingly.

 

5- Materials and Methods: More details about the specific research area, including images and information, should be added. This section should be enriched with relevant content.

Response: Thank you very much, it makes the methodology clearer. The edits related to this point can be found in:

• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 219-223 to clarify the main subjects of methodology.
• Drew Figure 1 to illustrate the study area.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.

.

6- Geographical Coverage: I recommend broadening the geographical coverage of the study. Extend the research to more geographical locations to validate the model's applicability under different climatic and geographical conditions.

Response: We are looking to maximise forecast accuracy. So, this manuscript considers one of the procedures reported in Hajirahimi and Khashei (2022), which is “Hybridisation of parameter optimisation-based with preprocessing-based hybrid models (HOPH)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, this was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks

 

7- Model Comparison: Consider comparing with other advanced hybrid models or machine learning techniques, beyond the ones mentioned in the manuscript.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, we have taken this advice and applied MPSO-ANN model. Thanks.

 

8- Field Validation: Conduct field validations or case studies to validate the effectiveness of the model in real-world applications.

Response: You have raised an important point. We divided the data into three sets, including training, testing, and validation. During the training stage, the network of the ANN model (i.e., weight and bias) is located. While, in the testing stage, the network of the ANN model (i.e., weight and bias) are tested. In both stages, the ANN model sees the actual target and compares it with the simulated one to calculate the error for each epoch to reach the best network. Accordingly, the prediction is in the training and testing stage because the ANN see the target.

 But, in the validation stage, we examine the generalisation of the ANN model by simulating the target through unseen data before. So, we called the forecast for the validation stage. Please see Figure 2.

 

 

9- Model Training Strategies: Explore different model training strategies or techniques, such as transfer learning, reinforcement learning, etc., to further enhance the performance of the model.

Response: We work according to the methodology that was applied successfully by Zubaidi et al., 2018 and Alawsi et al., 2022 by using supervised learning to train the data. However, we have taken this advice as a future work. Section 5. Thanks.

• Zubaidi et al, Short-Term Urban Water Demand Prediction Considering Weather Factors. Water Resources Management 2018, 32, 4527-4542, doi:10.1007/s11269-018-2061-y.
• Alawsi, M.A.; Zubaidi, S.L.; Al-Bdairi, N.S.S.; Al-Ansari, N.; Hashim, K. Drought Forecasting: A Review and Assessment of the Hybrid Techniques and Data Pre-Processing. Hydrology 2022, 9, doi:10.3390/hydrology9070115.

 

 

 

 

 

Reviewer 2 Report

In the present study, some metaheuristic algorithms are evaluated in the estimation of reference evapotranspiration using artificial neural networks. 

 

In recent years, many works based on artificial intelligence have been presented in the estimation of reference evapotranspiration, both at monthly and daily scales. Also, combinations have been made with other algorithms to adjust their hyperparameters, among the most used are the so-called evolutionary algorithms. 

 

The following is observed:

1. In the introduction there is no mention of the existing problem to be solved. 

2. The meteorological input variables to the models are not mentioned.

3. It would be worthwhile to compare the proposed model with an AI model where the hyperparameters are not adjusted, or to use an empirical model such as Hargreaves-Samani, since it seems that the study complicates the estimation of the reference evapotranspiration. 

4. In general, in the form in which the work is presented, it does not make a novel contribution to the study of evapotranspiration. 

Author Response

 

 

 

10^th September, 2023

 

 

 

Dear Dr. Lyra Xu,

 

Thank you for giving us the opportunity to submit a revised draft of our manuscript titled “Examination of Single- and Hybrid-based Metaheuristic Algo-rithms in ANN Reference Evapotranspiration Estimating” [sustainability-2546001] to Sustainability Journal. We appreciate the time and effort that you and the reviewers have dedicated to providing valuable feedback on our manuscript. We are grateful to the reviewers for their insightful comments, which have improved the paper. We have incorporated changes that reflect all the suggestions provided by the reviewers. All changes are highlighted in green in the revised manuscript. Please see below for point-by-point responses to the reviewers’ comments. If any responses are unclear or you wish additional change, please do not hesitate to let us know.

 

 

Thank you for your consideration.

 

 

 

 

 

Thank you for your consideration.

Sincerely,

Asst. Prof. Dr. Salah L. Zubaidi

Department of Civil Engineering, Wasit University, Wasit, 52001, Iraq.

Email: [email protected]

On behalf of the rest of co-authors.

Point-by-point response to the comments:

 

 

 

Reviewer #1 comment:

Overall comment: This is an interesting study. However, more details are needed to show how the authors conducted the research.

The authors used ANN model to predict reference ET0. ET0 that calculated from FAO-56 PM was input into the model, and ET0 was output from the model (which is hard to understand). The ANN model was integrated with PSOGWO to optimize the hyperparameters..

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction section, the author took long paragraphs to introduce other research, while YOUR OWN IDEAS were less written. As a research paper, the article does not have an introductory text with the hypothesis, research objectives, and methodology overview.

 Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- Please provide a figure to illustrate the study area in Section 2.1.

 Response: In the current version of the paper, Figure 1 illustrate the study area.

 

3- The authors said that “Due to exceptional circumstances i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was loss”, so they used secondary data (Line 155-157). Please explain the difference between the so-called “lost data” and the secondary data.

 Response: We revised the sentence to be “Due to exceptional circumstances, i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was missing. Secondary (i.e., satellite) data offered by the National Aeronautics and Space Administration” .

4- The author used secondary data as the dataset, with many meteorological data (Line 155-162). It would be advisable to match them with the parameters in Equation 1 (Line 167-171) to make the readers know which data match which parameter.

  Response: This has now been addressed in the current version of the manuscript.

 

 

5- The symbols in Equation 1 did not match the text (e.g. Tave and T in Line 168).

  Response: This has now been addressed in the current version of the manuscript.

 

 6- Please elaborate on how to preprocess the data using SSA and MI in Section 2.3.

  Response: This has now been addressed in the current version of the manuscript (Section 3.1).

 

7- In Section 2.4., the authors spent too much space on introducing the algorithms proposed by OTHERS (e.g. GWO, PSO, PSOGWO). However, I do not understand how the authors used those methods in their research and estimation of ET0. Moreover, Fig 1 only showed the algorithm flowchart of PSOGWO that was taken from OTHER RESEARCHERS. The key points of the authors’ original work were poorly written in the manuscript.

  Response: You have raised an important point. All these issues are amended as below.

• In Section 1, we added some paragraphs and revised the aim and ojectives
• Figure 2 presents a workflow diagram showing the steps involved in univariate ETo simulation.
• Section 4.2 ANN-based MHAs

 

 8- It is hard to understand why the author input ET0 from FAO-56PM into the model (Line 240) and acquired ET0 as output data (Line 242).

  Response: We work according to the methodology that was applied successfully by Ferreira et al., 2020, Nourani et al., 2020, and Sayyahi et al., 2021 by using previous data (Lags) to simulate future data (i.e., univariate technique). However, we have taken this advice and added Figure 2. Also, Figure 5 shows how to determine the best possible input scenario for the model using the MI technique. Moreover, Section 5 clarifis this issue.

• Ferreira, L.B.; da Cunha, F.F. Multi-step ahead forecasting of daily reference evapotranspiration using deep learning. Computers and Electronics in Agriculture 2020, 178, doi:10.1016/j.compag.2020.105728.
• Nourani, V.; Elkiran, G.; Abdullahi, J. Multi-step ahead modeling of reference evapotranspiration using a multi-model approach. Journal of Hydrology 2020, 581, 124434.
• Sayyahi, F.; Farzin, S.; Karami, H.; Cai, N. Forecasting Daily and Monthly Reference Evapotranspiration in the Aidoghmoush Basin Using Multilayer Perceptron Coupled with Water Wave Optimization. Complexity 2021, 2021, 6683759 doi:10.1155/2021/6683759.

 

 

9- How did the authors integrate PSOGWO with ANN? Please explain this in the method section.

  Response: This has now been addressed in the current version of the manuscript (Section 3.4).

 

10- In the method section, please describe the software that you used to perform statistical analysis and generate graphs, and report the settings and workflow as you did for the other parts. Also, please provide the source code on GitHub to ensure scalability and allow evaluation of the code adopted.  

  Response: This has now been addressed in the current version of the manuscript Lines 257-259, 408, and 428. Also, The code of PSOGWO-ANN algorithm was attached as supplementary materials.

 

 

 

Reviewer #2 comment:

 

In the present study, some metaheuristic algorithms are evaluated in the estimation of reference evapotranspiration using artificial neural networks.

In recent years, many works based on artificial intelligence have been presented in the estimation of reference evapotranspiration, both at monthly and daily scales. Also, combinations have been made with other algorithms to adjust their hyperparameters, among the most used are the so-called evolutionary algorithms

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction there is no mention of the existing problem to be solved.

Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the problem and novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- The meteorological input variables to the models are not mentioned.

Response: We used the meteorological variables to calculate the ETo time series (Section 2.1). Afterthat, Figure 5 shows the best possible input scenario (Lags of ETo) for the model (univariate technique). Also, Section 5 clarifies this issue.

3- It would be worthwhile to compare the proposed model with an AI model where the hyperparameters are not adjusted, or to use an empirical model such as Hargreaves-Samani, since it seems that the study complicates the estimation of the reference evapotranspiration.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. Also. Using a trial-and-error procedure to establish these hyperparameters can be risky be-cause of the high computational complexity and potential for mistakes [77]. Also, it is possible to overfit the data if a high number of neurons are added to the hidden layer using the trial-and-error procedure [78]. (lines 337-340). Moreover, three recent review papers in the different fields of hydrology showed that hybrid models outperformed single models. However, we have taken the advice of “Reviewer 3” and applied MPSO-ANN model. Thanks

 

4- In general, in the form in which the work is presented, it does not make a novel contribution to the study of evapotranspiration.

Response: We thank the reviewer for this suggestion, as we feel now that the novelties are much more clearly elaborated in the paper. In the current version of the paper, we have made a stronger effort to highlight the novelties of the study. The edits related to this point can be found in:

• Add several paragrphs and even revised the aim and objectives in Section 1 to highlights the novility.
• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 178-181 to clarify the main subjects of methodology.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.
• little amendments to the title and revised the manuscript in different sections accordingly.

 

 

Reviewer #3 comment:

The manuscript introduces a novel hybrid model that integrates the Slime Mould Algorithm (SMA) with the Artificial Neural Network (ANN) to predict Reference Evapotranspiration (ET0). The study also compares the performance of SMA-ANN with four other hybrid models (PSO-ANN, GWO-ANN, WOA-ANN, and MFO-ANN) in predicting ET0.

Overall, while the research direction is promising and aligns with the scope of this journal, there are areas in the manuscript that require improvement in terms of readability and addressing certain limitations.

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

Limitations:

 

1- The study is limited to testing on four sites in Iraq, which might restrict its applicability in other regions.

Response: This was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks.

 

2- The research only considers four hybrid models for comparison. There might be other potentially more effective hybrid models that have not been considered.

Response: This is a good point to increase the prediction range and decrease the uncertainty. As we mentioned in section 3.4, “Selecting an effective MHA is complex and presents additional difficulties, requiring different MHAs”. We applied an additional hybrid model (i.e., MPSO-ANN) and compared its performance with the other hybrid models (Sections 4.2 and 4.3).

 

Considering the above, I believe that the manuscript can be accepted for publication in the journal after revisions.

 

Specific Comments and Suggestions:

 

3- Language and Readability: The language of the manuscript needs polishing and further enhancement. The current readability is somewhat low, and it would benefit from a thorough proofreading.

Response: We have taken this advice and fully revised the manuscript to a much better standard of English and checked by a native speaker. We believe the paper now is much better.

 

4- Abstract: The abstract should state the research hypothesis or theoretical basis. Additionally, it would be beneficial to include the scientific significance of the study at the end of the abstract.

Response: Amendments have been made accordingly.

 

5- Materials and Methods: More details about the specific research area, including images and information, should be added. This section should be enriched with relevant content.

Response: Thank you very much, it makes the methodology clearer. The edits related to this point can be found in:

• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 219-223 to clarify the main subjects of methodology.
• Drew Figure 1 to illustrate the study area.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.

.

6- Geographical Coverage: I recommend broadening the geographical coverage of the study. Extend the research to more geographical locations to validate the model's applicability under different climatic and geographical conditions.

Response: We are looking to maximise forecast accuracy. So, this manuscript considers one of the procedures reported in Hajirahimi and Khashei (2022), which is “Hybridisation of parameter optimisation-based with preprocessing-based hybrid models (HOPH)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, this was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks

 

7- Model Comparison: Consider comparing with other advanced hybrid models or machine learning techniques, beyond the ones mentioned in the manuscript.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, we have taken this advice and applied MPSO-ANN model. Thanks.

 

8- Field Validation: Conduct field validations or case studies to validate the effectiveness of the model in real-world applications.

Response: You have raised an important point. We divided the data into three sets, including training, testing, and validation. During the training stage, the network of the ANN model (i.e., weight and bias) is located. While, in the testing stage, the network of the ANN model (i.e., weight and bias) are tested. In both stages, the ANN model sees the actual target and compares it with the simulated one to calculate the error for each epoch to reach the best network. Accordingly, the prediction is in the training and testing stage because the ANN see the target.

 But, in the validation stage, we examine the generalisation of the ANN model by simulating the target through unseen data before. So, we called the forecast for the validation stage. Please see Figure 2.

 

 

9- Model Training Strategies: Explore different model training strategies or techniques, such as transfer learning, reinforcement learning, etc., to further enhance the performance of the model.

Response: We work according to the methodology that was applied successfully by Zubaidi et al., 2018 and Alawsi et al., 2022 by using supervised learning to train the data. However, we have taken this advice as a future work. Section 5. Thanks.

• Zubaidi et al, Short-Term Urban Water Demand Prediction Considering Weather Factors. Water Resources Management 2018, 32, 4527-4542, doi:10.1007/s11269-018-2061-y.
• Alawsi, M.A.; Zubaidi, S.L.; Al-Bdairi, N.S.S.; Al-Ansari, N.; Hashim, K. Drought Forecasting: A Review and Assessment of the Hybrid Techniques and Data Pre-Processing. Hydrology 2022, 9, doi:10.3390/hydrology9070115.

 

 

 

 

 

Reviewer 3 Report

The manuscript introduces a novel hybrid model that integrates the Slime Mould Algorithm (SMA) with the Artificial Neural Network (ANN) to predict Reference Evapotranspiration (ET0). The study also compares the performance of SMA-ANN with four other hybrid models (PSO-ANN, GWO-ANN, WOA-ANN, and MFO-ANN) in predicting ET0.

Overall, while the research direction is promising and aligns with the scope of this journal, there are areas in the manuscript that require improvement in terms of readability and addressing certain limitations.

Limitations:

1. The study is limited to testing on four sites in Iraq, which might restrict its applicability in other regions.
2. The research only considers four hybrid models for comparison. There might be other potentially more effective hybrid models that have not been considered.

Considering the above, I believe that the manuscript can be accepted for publication in the journal after revisions.

Specific Comments and Suggestions:

1. Language and Readability: The language of the manuscript needs polishing and further enhancement. The current readability is somewhat low, and it would benefit from a thorough proofreading.

2. Abstract: The abstract should state the research hypothesis or theoretical basis. Additionally, it would be beneficial to include the scientific significance of the study at the end of the abstract.

3. Materials and Methods: More details about the specific research area, including images and information, should be added. This section should be enriched with relevant content.

4. Geographical Coverage: I recommend broadening the geographical coverage of the study. Extend the research to more geographical locations to validate the model's applicability under different climatic and geographical conditions.

5. Model Comparison: Consider comparing with other advanced hybrid models or machine learning techniques, beyond the ones mentioned in the manuscript.

6. Field Validation: Conduct field validations or case studies to validate the effectiveness of the model in real-world applications.

7. Model Training Strategies: Explore different model training strategies or techniques, such as transfer learning, reinforcement learning, etc., to further enhance the performance of the model.

I hope these comments and suggestions will help improve the manuscript.

 

There are really some areas in the manuscript that require improvement in terms of readability and addressing certain limitations.

 

Author Response

 

 

 

10^th September, 2023

 

 

 

Dear Dr. Lyra Xu,

 

Thank you for giving us the opportunity to submit a revised draft of our manuscript titled “Examination of Single- and Hybrid-based Metaheuristic Algo-rithms in ANN Reference Evapotranspiration Estimating” [sustainability-2546001] to Sustainability Journal. We appreciate the time and effort that you and the reviewers have dedicated to providing valuable feedback on our manuscript. We are grateful to the reviewers for their insightful comments, which have improved the paper. We have incorporated changes that reflect all the suggestions provided by the reviewers. All changes are highlighted in green in the revised manuscript. Please see below for point-by-point responses to the reviewers’ comments. If any responses are unclear or you wish additional change, please do not hesitate to let us know.

 

 

Thank you for your consideration.

 

 

 

 

 

Thank you for your consideration.

Sincerely,

Asst. Prof. Dr. Salah L. Zubaidi

Department of Civil Engineering, Wasit University, Wasit, 52001, Iraq.

Email: [email protected]

On behalf of the rest of co-authors.

Point-by-point response to the comments:

 

 

 

Reviewer #1 comment:

Overall comment: This is an interesting study. However, more details are needed to show how the authors conducted the research.

The authors used ANN model to predict reference ET0. ET0 that calculated from FAO-56 PM was input into the model, and ET0 was output from the model (which is hard to understand). The ANN model was integrated with PSOGWO to optimize the hyperparameters..

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction section, the author took long paragraphs to introduce other research, while YOUR OWN IDEAS were less written. As a research paper, the article does not have an introductory text with the hypothesis, research objectives, and methodology overview.

 Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- Please provide a figure to illustrate the study area in Section 2.1.

 Response: In the current version of the paper, Figure 1 illustrate the study area.

 

3- The authors said that “Due to exceptional circumstances i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was loss”, so they used secondary data (Line 155-157). Please explain the difference between the so-called “lost data” and the secondary data.

 Response: We revised the sentence to be “Due to exceptional circumstances, i.e., wars and terrorism in Iraq, most of the data from 1990–2020 was missing. Secondary (i.e., satellite) data offered by the National Aeronautics and Space Administration” .

4- The author used secondary data as the dataset, with many meteorological data (Line 155-162). It would be advisable to match them with the parameters in Equation 1 (Line 167-171) to make the readers know which data match which parameter.

  Response: This has now been addressed in the current version of the manuscript.

 

 

5- The symbols in Equation 1 did not match the text (e.g. Tave and T in Line 168).

  Response: This has now been addressed in the current version of the manuscript.

 

 6- Please elaborate on how to preprocess the data using SSA and MI in Section 2.3.

  Response: This has now been addressed in the current version of the manuscript (Section 3.1).

 

7- In Section 2.4., the authors spent too much space on introducing the algorithms proposed by OTHERS (e.g. GWO, PSO, PSOGWO). However, I do not understand how the authors used those methods in their research and estimation of ET0. Moreover, Fig 1 only showed the algorithm flowchart of PSOGWO that was taken from OTHER RESEARCHERS. The key points of the authors’ original work were poorly written in the manuscript.

  Response: You have raised an important point. All these issues are amended as below.

• In Section 1, we added some paragraphs and revised the aim and ojectives
• Figure 2 presents a workflow diagram showing the steps involved in univariate ETo simulation.
• Section 4.2 ANN-based MHAs

 

 8- It is hard to understand why the author input ET0 from FAO-56PM into the model (Line 240) and acquired ET0 as output data (Line 242).

  Response: We work according to the methodology that was applied successfully by Ferreira et al., 2020, Nourani et al., 2020, and Sayyahi et al., 2021 by using previous data (Lags) to simulate future data (i.e., univariate technique). However, we have taken this advice and added Figure 2. Also, Figure 5 shows how to determine the best possible input scenario for the model using the MI technique. Moreover, Section 5 clarifis this issue.

• Ferreira, L.B.; da Cunha, F.F. Multi-step ahead forecasting of daily reference evapotranspiration using deep learning. Computers and Electronics in Agriculture 2020, 178, doi:10.1016/j.compag.2020.105728.
• Nourani, V.; Elkiran, G.; Abdullahi, J. Multi-step ahead modeling of reference evapotranspiration using a multi-model approach. Journal of Hydrology 2020, 581, 124434.
• Sayyahi, F.; Farzin, S.; Karami, H.; Cai, N. Forecasting Daily and Monthly Reference Evapotranspiration in the Aidoghmoush Basin Using Multilayer Perceptron Coupled with Water Wave Optimization. Complexity 2021, 2021, 6683759 doi:10.1155/2021/6683759.

 

 

9- How did the authors integrate PSOGWO with ANN? Please explain this in the method section.

  Response: This has now been addressed in the current version of the manuscript (Section 3.4).

 

10- In the method section, please describe the software that you used to perform statistical analysis and generate graphs, and report the settings and workflow as you did for the other parts. Also, please provide the source code on GitHub to ensure scalability and allow evaluation of the code adopted.  

  Response: This has now been addressed in the current version of the manuscript Lines 257-259, 408, and 428. Also, The code of PSOGWO-ANN algorithm was attached as supplementary materials.

 

 

 

Reviewer #2 comment:

 

In the present study, some metaheuristic algorithms are evaluated in the estimation of reference evapotranspiration using artificial neural networks.

In recent years, many works based on artificial intelligence have been presented in the estimation of reference evapotranspiration, both at monthly and daily scales. Also, combinations have been made with other algorithms to adjust their hyperparameters, among the most used are the so-called evolutionary algorithms

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

1- In the introduction there is no mention of the existing problem to be solved.

Response: In the current version of the paper we have made a stronger effort to amended the issues that related the introduction and even highlight the problem and novelty of the study. The edited related to this point can be found in section 1. We thank the reviewer for this suggestion, as we feel now that this points are much clearly elaborated in the paper.

 

2- The meteorological input variables to the models are not mentioned.

Response: We used the meteorological variables to calculate the ETo time series (Section 2.1). Afterthat, Figure 5 shows the best possible input scenario (Lags of ETo) for the model (univariate technique). Also, Section 5 clarifies this issue.

3- It would be worthwhile to compare the proposed model with an AI model where the hyperparameters are not adjusted, or to use an empirical model such as Hargreaves-Samani, since it seems that the study complicates the estimation of the reference evapotranspiration.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. Also. Using a trial-and-error procedure to establish these hyperparameters can be risky be-cause of the high computational complexity and potential for mistakes [77]. Also, it is possible to overfit the data if a high number of neurons are added to the hidden layer using the trial-and-error procedure [78]. (lines 337-340). Moreover, three recent review papers in the different fields of hydrology showed that hybrid models outperformed single models. However, we have taken the advice of “Reviewer 3” and applied MPSO-ANN model. Thanks

 

4- In general, in the form in which the work is presented, it does not make a novel contribution to the study of evapotranspiration.

Response: We thank the reviewer for this suggestion, as we feel now that the novelties are much more clearly elaborated in the paper. In the current version of the paper, we have made a stronger effort to highlight the novelties of the study. The edits related to this point can be found in:

• Add several paragrphs and even revised the aim and objectives in Section 1 to highlights the novility.
• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 178-181 to clarify the main subjects of methodology.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.
• little amendments to the title and revised the manuscript in different sections accordingly.

 

 

Reviewer #3 comment:

The manuscript introduces a novel hybrid model that integrates the Slime Mould Algorithm (SMA) with the Artificial Neural Network (ANN) to predict Reference Evapotranspiration (ET0). The study also compares the performance of SMA-ANN with four other hybrid models (PSO-ANN, GWO-ANN, WOA-ANN, and MFO-ANN) in predicting ET0.

Overall, while the research direction is promising and aligns with the scope of this journal, there are areas in the manuscript that require improvement in terms of readability and addressing certain limitations.

Response: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted).

 

Limitations:

 

1- The study is limited to testing on four sites in Iraq, which might restrict its applicability in other regions.

Response: This was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks.

 

2- The research only considers four hybrid models for comparison. There might be other potentially more effective hybrid models that have not been considered.

Response: This is a good point to increase the prediction range and decrease the uncertainty. As we mentioned in section 3.4, “Selecting an effective MHA is complex and presents additional difficulties, requiring different MHAs”. We applied an additional hybrid model (i.e., MPSO-ANN) and compared its performance with the other hybrid models (Sections 4.2 and 4.3).

 

Considering the above, I believe that the manuscript can be accepted for publication in the journal after revisions.

 

Specific Comments and Suggestions:

 

3- Language and Readability: The language of the manuscript needs polishing and further enhancement. The current readability is somewhat low, and it would benefit from a thorough proofreading.

Response: We have taken this advice and fully revised the manuscript to a much better standard of English and checked by a native speaker. We believe the paper now is much better.

 

4- Abstract: The abstract should state the research hypothesis or theoretical basis. Additionally, it would be beneficial to include the scientific significance of the study at the end of the abstract.

Response: Amendments have been made accordingly.

 

5- Materials and Methods: More details about the specific research area, including images and information, should be added. This section should be enriched with relevant content.

Response: Thank you very much, it makes the methodology clearer. The edits related to this point can be found in:

• We separated the materials in section 2 and methodology in section 3 for more clarifications and renumbered the subsections accordingly.
• We added lines 219-223 to clarify the main subjects of methodology.
• Drew Figure 1 to illustrate the study area.
• Drew Figure 2 to show a process flow depicting the actions needed to forecast ETo.

.

6- Geographical Coverage: I recommend broadening the geographical coverage of the study. Extend the research to more geographical locations to validate the model's applicability under different climatic and geographical conditions.

Response: We are looking to maximise forecast accuracy. So, this manuscript considers one of the procedures reported in Hajirahimi and Khashei (2022), which is “Hybridisation of parameter optimisation-based with preprocessing-based hybrid models (HOPH)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, this was highlighted as a future works in lines 519-521 “hybrid PSOGWO-ANN model is a promising technique for estimating monthly ETo in other agro-Iraqi provinces such as Kirkuk and Salahaddin and is therefore strongly recommended for this purpose”. Thanks

 

7- Model Comparison: Consider comparing with other advanced hybrid models or machine learning techniques, beyond the ones mentioned in the manuscript.

Response: We looking for a new framework that maximises forecast accuracy. So, we apply the ANN model because “In a review of ETo prediction models by Krishnashetty, et al. [4], it was found that the artificial neural network (ANN) technique outperformed other machine learning techniques when simulating reference evapotranspiration. (lines 333-335)”. Our foxing here is to examine the performance of single- and hybrid-based metaheuristic algorithms. However, we have taken this advice and applied MPSO-ANN model. Thanks.

 

8- Field Validation: Conduct field validations or case studies to validate the effectiveness of the model in real-world applications.

Response: You have raised an important point. We divided the data into three sets, including training, testing, and validation. During the training stage, the network of the ANN model (i.e., weight and bias) is located. While, in the testing stage, the network of the ANN model (i.e., weight and bias) are tested. In both stages, the ANN model sees the actual target and compares it with the simulated one to calculate the error for each epoch to reach the best network. Accordingly, the prediction is in the training and testing stage because the ANN see the target.

 But, in the validation stage, we examine the generalisation of the ANN model by simulating the target through unseen data before. So, we called the forecast for the validation stage. Please see Figure 2.

 

 

9- Model Training Strategies: Explore different model training strategies or techniques, such as transfer learning, reinforcement learning, etc., to further enhance the performance of the model.

Response: We work according to the methodology that was applied successfully by Zubaidi et al., 2018 and Alawsi et al., 2022 by using supervised learning to train the data. However, we have taken this advice as a future work. Section 5. Thanks.

• Zubaidi et al, Short-Term Urban Water Demand Prediction Considering Weather Factors. Water Resources Management 2018, 32, 4527-4542, doi:10.1007/s11269-018-2061-y.
• Alawsi, M.A.; Zubaidi, S.L.; Al-Bdairi, N.S.S.; Al-Ansari, N.; Hashim, K. Drought Forecasting: A Review and Assessment of the Hybrid Techniques and Data Pre-Processing. Hydrology 2022, 9, doi:10.3390/hydrology9070115.

 

 

 

 

 

Round 2

Reviewer 1 Report

The new manuscript has shown some improvements compared with the previous version. However, some of my comments from the previous review were not addressed.

The author spent long paragraphs introducing other researchers’ work, with detailed figures (e.g., Line 87, Line 94-95), which is unnecessary. In addition, the structure of the literature review is unclear, and it would be better to separate a new paragraph for it.

The author did not elaborate on THEIR OWN IDEAS. The research objectives were directly proposed, without any logical reasoning （Line 164）, which makes it hard for readers to understand how the authors thought out the methodology.

The author used secondary data as the dataset, which included many meteorological data. It would be much clearer to present a table for data specification.  I still argue that it is better to match secondary data with the parameters in Equation 1 to help readers know which data corresponds to which parameter.

In Line 208, it should be 2.2 instead of 2.1.

Author Response

 

 

21^st September, 2023

 

RE: sustainability-2546001 “Examination of Single- and Hybrid-based Metaheuristic Algorithms in ANN Reference Evapotranspiration Estimating”.

 

Dear Dr. Zachary Zhang,

 

The authors would like to express their sincere thanks to the Editor and the reviewers for the received feedback that has considerably improved the paper. We hope to have addressed all the comments appropriately.

 

Please, find attached to this letter a point-by-point response to the comments.

 

 

 

 

 

 

 

 

Thank you for your consideration.

Sincerely,

Asst. Prof. Dr. Salah L. Zubaidi

Department of Civil Engineering, Wasit University, Wasit, 52001, Iraq.

Email: [email protected]

On behalf of the rest of co-authors.

 

 

Point-by-point response to the comments:

 

Reviewer #1 comment:

 

The new manuscript has shown some improvements compared with the previous version. However, some of my comments from the previous review were not addressed.

Authors’ comment: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted). Also, we believe that we addressed all the comments raised by the reviewers, and we truly want to thank the reviewers because they opened the door for future research.

 

1. The author spent long paragraphs introducing other researchers’ work, with detailed figures (e.g., Line 87, Line 94-95), which is unnecessary. In addition, the structure of the literature review is unclear, and it would be better to separate a new paragraph for it.

Authors’ comment: All these issues are amended as below.

• Both “detailed figures (e.g., Line 87, Line 94-95), which is unnecessary” were deleted.
• We divided the introduction section into four sub-sections: Research background, Applied machine learning methods for ETo forecasting, Research significance and motivation, and Research objectives.

 

2. The author did not elaborate on THEIR OWN IDEAS. The research objectives were directly proposed, without any logical reasoning （Line 164）, which makes it hard for readers to understand how the authors thought out the methodology.

Authors’ comment: As mentioned above, we have divided the introduction into four sub-sections.

We systematically wrote the introduction section. We highlighted the problem globally and specifically in Iraq. After that, we presented the application of ML models in the field of univariate ETo prediction. Next, we reported the drawbacks of a single ML model and how MHAs can improve the ML models. Then, we mentioned different types of MHAs, and how they work (i.e., exploration and exploitation) for single- and hybrid-based types. After that, we highlighted the gap based on a recent systematic review paper, which stated far too little attention has been paid to hybrid-based MHAs (6%) compared with single-based MHAs. Finally, we present the paper's aim and the objectives which should be performed to achieve the aim.

 

3. The author used secondary data as the dataset, which included many meteorological data. It would be much clearer to present a table for data specification. I still argue that it is better to match secondary data with the parameters in Equation 1 to help readers know which data corresponds to which parameter.

Authors’ comment: All these issues are amended as below.

• In section 2, we added Table 1, which presents the statistical parameters of the meteorological time series.
• In the first revision, we amended some of the abbreviations of metrological factors to match Equation 1. However, reference evapotranspiration was calculated following the Penman-Monteith equation [1] using ETo Calculator FAO version 3.2 [2]. Available online: https://www.fao.org/land-water/databases-and-software/eto-calculator/en/ .The climate variables required for it include monthly minimum and maximum air temperature, solar radiation, relative humidity, and wind velocity.

 

1. Allen, R.G.; Pereira, L.S.; Raes, D.; Smith, M. Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56. Fao, Rome 1998, 300, D05109.
2. Raes, D. Reference manual-ETO calculator. Food and Agriculture Organization of the United Nations Land Water Division. Rome, I., Ed. 2012.

 

Accordingly, the abbreviations of metrological factors match Calculator FAO version 3.2.

 

 

 

4. In Line 208, it should be 2.2 instead of 2.1.

Authors’ comment: This has now been addressed in the current version of the manuscript. Thanks.

 

 

 

Reviewer #2 comments:

 

The version presented by the authors has improved considerably, and they also clarify some doubts. I am satisfied with the present version.

Authors’ comment: Many thanks for your positive opinion about the paper.

 

 

Reviewer 2 Report

The version presented by the authors has improved considerably, and they also clarify some doubts. 

I am satisfied with the present version. 

 

Author Response

 

 

21^st September, 2023

 

RE: sustainability-2546001 “Examination of Single- and Hybrid-based Metaheuristic Algorithms in ANN Reference Evapotranspiration Estimating”.

 

Dear Dr. Zachary Zhang,

 

The authors would like to express their sincere thanks to the Editor and the reviewers for the received feedback that has considerably improved the paper. We hope to have addressed all the comments appropriately.

 

Please, find attached to this letter a point-by-point response to the comments.

 

 

 

 

 

 

 

 

Thank you for your consideration.

Sincerely,

Asst. Prof. Dr. Salah L. Zubaidi

Department of Civil Engineering, Wasit University, Wasit, 52001, Iraq.

Email: [email protected]

On behalf of the rest of co-authors.

 

 

Point-by-point response to the comments:

 

Reviewer #1 comment:

 

The new manuscript has shown some improvements compared with the previous version. However, some of my comments from the previous review were not addressed.

Authors’ comment: Thank you very much for your feedback. We believe the paper now is much better. All these issues are amended as below (all amended items are highlighted). Also, we believe that we addressed all the comments raised by the reviewers, and we truly want to thank the reviewers because they opened the door for future research.

 

1. The author spent long paragraphs introducing other researchers’ work, with detailed figures (e.g., Line 87, Line 94-95), which is unnecessary. In addition, the structure of the literature review is unclear, and it would be better to separate a new paragraph for it.

Authors’ comment: All these issues are amended as below.

• Both “detailed figures (e.g., Line 87, Line 94-95), which is unnecessary” were deleted.
• We divided the introduction section into four sub-sections: Research background, Applied machine learning methods for ETo forecasting, Research significance and motivation, and Research objectives.

 

2. The author did not elaborate on THEIR OWN IDEAS. The research objectives were directly proposed, without any logical reasoning （Line 164）, which makes it hard for readers to understand how the authors thought out the methodology.

Authors’ comment: As mentioned above, we have divided the introduction into four sub-sections.

We systematically wrote the introduction section. We highlighted the problem globally and specifically in Iraq. After that, we presented the application of ML models in the field of univariate ETo prediction. Next, we reported the drawbacks of a single ML model and how MHAs can improve the ML models. Then, we mentioned different types of MHAs, and how they work (i.e., exploration and exploitation) for single- and hybrid-based types. After that, we highlighted the gap based on a recent systematic review paper, which stated far too little attention has been paid to hybrid-based MHAs (6%) compared with single-based MHAs. Finally, we present the paper's aim and the objectives which should be performed to achieve the aim.

 

3. The author used secondary data as the dataset, which included many meteorological data. It would be much clearer to present a table for data specification. I still argue that it is better to match secondary data with the parameters in Equation 1 to help readers know which data corresponds to which parameter.

Authors’ comment: All these issues are amended as below.

• In section 2, we added Table 1, which presents the statistical parameters of the meteorological time series.
• In the first revision, we amended some of the abbreviations of metrological factors to match Equation 1. However, reference evapotranspiration was calculated following the Penman-Monteith equation [1] using ETo Calculator FAO version 3.2 [2]. Available online: https://www.fao.org/land-water/databases-and-software/eto-calculator/en/ .The climate variables required for it include monthly minimum and maximum air temperature, solar radiation, relative humidity, and wind velocity.

 

1. Allen, R.G.; Pereira, L.S.; Raes, D.; Smith, M. Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56. Fao, Rome 1998, 300, D05109.
2. Raes, D. Reference manual-ETO calculator. Food and Agriculture Organization of the United Nations Land Water Division. Rome, I., Ed. 2012.

 

Accordingly, the abbreviations of metrological factors match Calculator FAO version 3.2.

 

 

 

4. In Line 208, it should be 2.2 instead of 2.1.

Authors’ comment: This has now been addressed in the current version of the manuscript. Thanks.

 

 

 

Reviewer #2 comments:

 

The version presented by the authors has improved considerably, and they also clarify some doubts. I am satisfied with the present version.

Authors’ comment: Many thanks for your positive opinion about the paper.