Capability and Robustness of Novel Hybridized Artificial Intelligence Technique for Sediment Yield Modeling in Godavari River, India
, Devendra Joshi 1, Vinod Kumar 1, Hitesh Mohapatra 1, Celestine Iwendi 2,* and Thippa Reddy Gadekallu 3,*
Round 1
Reviewer 1 Report
Dear Authors,
The paper submitted for review titled: Capability and Robustness of Novel Hybridized Artificial Intelligence Technique for Sediment Yield Modeling in Godavari River, India presents an interesting approach to the topic presented.
The paper consists of 5 chapters, which are structured into a logical whole. The strong point of the paper is the described subject matter, in which the authors have adopted modified neural networks GA-ANN .
Despite this, the article has several shortcomings that should be taken into account in order to improve its attractiveness to the reader:
- The abstract should be formulated in such a way that it reflects the brief content of the article. It should not begin with bullet points.
- There are places in the paper where there are too many spaces between words.
- Spaces should be inserted between tinings placed in square brackets - it reads better.
- Mathematical equations should be centred.
- The paragraph below the variables describing the formula should be preceded by a space.
- The descriptions in Figure 3 are not very legible. You may want to change the font colour to black.
- There should be a space between the axis descriptions and the units.
- Please consider whether it is worth reducing the range of the horizontal axis to make the graph in Figure 5 more readable.
- Figure 6 is distorted.
- For what purpose are negative values shown in the graphs shown in Figures 7 and 8?
Conclusion:
In order to proceed with the further publication process, the reviewer would like to encourage the authors to follow the comments/suggestions. A thorough revision of the paper is required.
Author Response
Comment 1: The scientific paper is valuable and I honestly admire your research done.
Response: Thank you for your comments.
Comment 2: Although, I want to give you an advice regarding the sources of information used. A big majority of them are a little bit old and others very old. So, I advise you to use as references more titles from the ones published in this particular journal in the last 1-3 years and also to use especially references from the ones published in Q1 and Q2 , eg SCOPUS or WoS/CA indexed journals in the last 1-3 years.
Response: Thank you for your comments. We have made change in the revised manuscript. The first author of this paper is already published the paper in 2021 which is SCI Q1 Journal paper. It is already cited in reference index of 49 which is given below.
- Yadav, A.; Chatterjee, S; Equeenuddin, SM.; Suspended sediment yield modeling in Mahanadi River, India by multi-objective optimization hybridizing artificial intelligence algorithms. Int J Sediment Res 2021; 36(1):76-91. . https://doi.org/10.1016/j.ijsrc.2020.03.018.
We updated the references from the published in this “Water” journal in the last 1-3 years.
The following references are added in the revised manuscript.
- Hsieh, T-C.; Ding, Y.; Yeh, K-C.; Jhong R-K. Numerical Investigation of Sediment Flushing and Morphological Changes in Tamsui River Estuary through Monsoons and Typhoons. Water 2022; 14(11):1802. https://doi.org/10.3390/w14111802.
- Lee, F.Z.; Lai, J.S.; Sumi, T. Reservoir Sediment Management and Downstream River Impacts for Sustainable Water Resources—Case Study of Shihmen Reservoir. Water 2022; 14(3), p.479.
- Hsieh, T.C.; Ding, Y.; Yeh, K.C.; Jhong, R.K. Investigation of morphological changes in the tamsui river estuary using an integrated coastal and estuarine processes model. Water 2020., 12(4), p.1084.
Comment 3: The abstract should be formulated in such a way that it reflects the brief content of the article. It should not begin with bullet points.
Response: Thank you for your comments. It is corrected in revised manuscript. We removed the bullet points of the abstract.
Comment 4: There are places in the paper where there are too many spaces between words.
Response: Thank you for your comment. Now, we removed the spaces between the words. We have made change in the revised manuscript.
Comment 5: Spaces should be inserted between tinings placed in square brackets - it reads better.
Response: Thank you for your valuable comments and suggestions. We have made change in the revised manuscript.
Comment 6: Mathematical equations should be centred.
Response: Thank you for your comments. We have made change and updated in the revised manuscript.
Comment 7: The paragraph below the variables describing the formula should be preceded by a space.
Response: Thank you for your comments. We have made change in the revised manuscript.
Comment 8: The descriptions in Figure 3 are not very legible. You may want to change the font colour to black.
Response: Thank you for your comments. We have made change in the revised manuscript.
Comment 9: There should be a space between the axis descriptions and the units.
Response: Thank you for your comments. We have made change in the revised manuscript.
Comment 10: Please consider whether it is worth reducing the range of the horizontal axis to make the graph in Figure 5 more readable.
Response: Thank you for your comments. We have modified the figure. We have made change in the revised manuscript.
Comment 11: Figure 6 is distorted.
Response: Thank you for your comments. Now, we modified the Figure 6. We have made change and updated in the revised manuscript.
Comment 12: For what purpose are negative values shown in the graphs shown in Figures 7 and 8?
Response: Thank you for your comments. We have made change in the revised manuscript. The negative value of sediment is not possible. Sediment can never be negative in reality. Negative value of sediment is found corresponding to lower observed sediment values which indicates that model is not capable to predict the sediment at lower sediment value or close to zero sediment value.
Reviewer 2 Report
Dear authors,
The scientific paper is valuable and I honestly admire your research done.
Although, I want to give you an advice regarding the sources of information used. A big majority of them are a little bit old and others very old. So, I advise you to use as references more titles from the ones published in this particular journal in the last 1-3 years and also to use especially references from the ones published in Q1 and Q2 , eg SCOPUS or WoS/CA indexed journals in the last 1-3 years.
Please, continue your research in the near future.
That is all for the moment.
EG
Author Response
Comment 1: The scientific paper is valuable and I honestly admire your research done.
Response: Thank you for your comments.
Comment 2: Although, I want to give you an advice regarding the sources of information used. A big majority of them are a little bit old and others very old. So, I advise you to use as references more titles from the ones published in this particular journal in the last 1-3 years and also to use especially references from the ones published in Q1 and Q2 , eg SCOPUS or WoS/CA indexed journals in the last 1-3 years.
Response: Thank you for your comments. We have made change in the revised manuscript. The first author of this paper is already published the paper in 2021 which is SCI Q1 Journal paper. It is already cited in reference index of 49 which is given below.
- Yadav, A.; Chatterjee, S; Equeenuddin, SM.; Suspended sediment yield modeling in Mahanadi River, India by multi-objective optimization hybridizing artificial intelligence algorithms. Int J Sediment Res 2021; 36(1):76-91. . https://doi.org/10.1016/j.ijsrc.2020.03.018.
We updated the references from the published in this “Water” journal in the last 1-3 years.
The following references are added in the revised manuscript.
- Hsieh, T-C.; Ding, Y.; Yeh, K-C.; Jhong R-K. Numerical Investigation of Sediment Flushing and Morphological Changes in Tamsui River Estuary through Monsoons and Typhoons. Water 2022; 14(11):1802. https://doi.org/10.3390/w14111802.
- Lee, F.Z.; Lai, J.S.; Sumi, T. Reservoir Sediment Management and Downstream River Impacts for Sustainable Water Resources—Case Study of Shihmen Reservoir. Water 2022; 14(3), p.479.
- Hsieh, T.C.; Ding, Y.; Yeh, K.C.; Jhong, R.K. Investigation of morphological changes in the tamsui river estuary using an integrated coastal and estuarine processes model. Water 2020., 12(4), p.1084.
Reviewer 3 Report
See attached!
I recommend MODERATE revision. Since no such option is available, I pick MAJOR revision but, again, mean by it MODERATE revision.
Comments for author File: 
Comments.doc
Author Response
Comment 1: Figure 1 is of OK, but its caption, which should stand (be understood) on its own, uses two unestablished acronyms and forces the reader to seek their definition. SPELL THEM OUT! •
Response: Thank you for pointing out that. We have corrected it. We have made change in the revised manuscript.
Comment 2: Figures 2: poor quality raster; too small font size making it practically illegible.
Response: Thank you for valuable comments. We have made change and updated in the revised manuscript.
Comment 3: Figures 3: poor quality raster ; too small font size making it practically illegible; no need to crowd the horizontal axis—display every second (or third) number of days.
Response: Thank you for valuable comments. We have made change and updated the modified figure in the revised manuscript.
Comment 4: Figure 4: add axis labels; increase the font size; don’t crowd the vertical axis—display every second number.
Response: Thank you for comments. Figure 4 is changed. Now, we have made change in the revised manuscript.
Comment 5: Figure 5: increase the font size; What is the meaning of “Best: 0.0305102 Mean: ...”? Increase the data point size; re-scale the axis (log-linear?) so that the gravity center of the figure is not located close to the origin.
Response: Thank you for valuable comments. Now, figure is changed as per your comments. It is updated in the revised manuscript. This Matlab auto generated figure in hybrid GA-ANN model during training period. It represents Generation-wise fitness function profile during GA-based neural network learning. The “best” represents the best fitness function and “mean” represents the mean fitness function value corresponding the best fitness value among all generations of the hybrid GA-ANN model.
Comment 6: Figure 6, 7, 8, and 9: practically illegible for the analogous reasons.
Response: Thank you for comments. Now, we changed these 4 figures and increased the DPI of all these figures to make it clear visible. Now, it is updated in revised manuscript.
Particular Comments
Comment 1: Line 329: Equation (4): the subscript is likely incorrect: S(t) ==> S(t)
Response: Thank you for pointing out that. Now equation 4 is corrected and updated in the revised manuscript.
Comment 2: Line 635: RMSE (root mean square error?) UNDEFINED symbol; DEFINE, IT AS THERE ARE DIFFERENT DEFINITIONS in the literature (e.g., dividing by N or N-1)
Response: Thank you for comments. RMSE represents the root mean square error. Now, it is updated in revised manuscript. The standard formula of all error statistics which are used in the manuscript are given below. The following formulas were added in the revised manuscript:
(8)
where, , , and are measured, measured mean, predicted and predicted mean values respectively. The N value represents the number of samples. The E and represent the error and mean error values.
Comment 3: Line 635: MAE UNDEFINED symbol
Response: Thank you for comments. MAE represents the mean absolute error. Now, it is updated in revised manuscript. The following formulas were added in the revised manuscript:
where, and are measured and predicted values, respectively. The N value represents the number of samples.
Comment 4: Line 384: The daily mean of WD ((Q)), WHY TWO SETS OF PARENTHESIS?
Response: Thank you for comments. Now, Q is deleted and only one set WD is corrected in revised manuscript.
Comment 5: VAR UNDEFINED symbol
Response: Thank you for comments. VAR represents the error variance. Now, it is updated in revised manuscript. The following formulas were added in the revised manuscript:
The N value represents the number of samples. The E and represent the error and mean error values.
Comment 6: The captions should stand on their own. Make sure that they can be understood without reading the paper, as the Abstract, Conclusions, and the captions should stand on their own! So, no undefined acronyms in these sections.
Response: Thank you for comments. All are defined acronyms in abstract section.
Comment 7: The novelty of the paper should be flashed out and clearly communicated in the abstract, discussion, and conclusions.
Response: Thank you for comments. It is mentioned in the abstract, discussion and conclusions sections.
Comment 8: The paper would benefit greatly from a comparison of the presented artificial intelligence methodology to one more closely based on a physical model, perhaps even one based on the simple power equation, (5). I urge the authors to consider this expansion, which with 7 authors shouldn’t be a problem.
Response: Thank you for comments. In this paper, two regression-based models’ multiple linear regression (MLR) and sediment rating curve (SRC) methods are used for comparison. Tradition Artificial intelligence-based model like ANN is also used for comparison with hybrid GA-ANN model. Equation 5 of traditional SRC method represents the power equation models for sediment estimation which are used by various researcher successfully for comparison in the field of hydrology for sediment estimation ([17-19],[49-51]) which references are given in the manuscript. In this paper, there are 6 authors.
Reviewer 4 Report
In the manuscript “Capability and Robustness of Novel Hybridized Artificial Intelligence Technique for Sediment Yield Modeling in Godavari 3 River, India” authors are dealing with important subject because development of models of high accuracy in hydrology has many practical applications.
However, I have some comments:
1. In line 25 Authors stated that “model was developed using daily water discharge, water level and SSY data to estimate the daily SSY”. What kind of SSY as input parameter was used? It should be explained in section 3 or removed. This aspect can be seen also in line 194. Authors declare that they were changing the number of input parameters (1, 2 or 3).
2. Why Authors decided to use list of input parameters as changing/adjusted factor of the model? It must be explained in the text. Usually, the list of inputs is based on our knowledge about the modelled phenomena. If there are many of them, we can remove some based on eg. correlation analysis. But it shouldn’t be done randomly.
3. In Table 2 and 3 Authors presented high correlations between WD and WL. Then, based on ANOVA, Authors stated that WD(Q) and WL are important inputs features to predict the SSY. According to my knowledge, the decision about including/excluding input model parameters should be made based on the correlation and one of highly correlated parameters should be excluded.
4. I think that the information about input model parameters should be declared for every model when it’s presented for the first time, not only in section 4.5. It would make text clearer for readers.
5. In section 4.3 the MLR equations should be presented, it would make text clearer for readers.
6. It’s not clear how R2 (determination coefficient) was calculated in Table 8. It’s much higher than in tables 4-7. It should be revised or explained.
7. There are some linguistic mistakes such as in line 412.
8. I suppose that lines 483-489 should be removed.
Author Response
Comment 1: In line 25 Authors stated that “model was developed using daily water discharge, water level and SSY data to estimate the daily SSY”. What kind of SSY as input parameter was used? It should be explained in section 3 or removed. This aspect can be seen also in line 194. Authors declare that they were changing the number of input parameters (1, 2 or 3).
Response: Thank you for pointing out that. Now, it is corrected in revised manuscript. In this manuscript water discharge and water level data are used as inputs. Suspended sediment yield(SSY) is output data. In his paper, both inputs (water discharge and water level) and actual output (SSY) are used for developing all models. Past SSY data is used as output data not as input data. The ANN and GA-ANN models are supervised learning models so both past inputs and outputs data are used for developing the models by analysing the inputs and output data. In line no. 194, it is also corrected, and 1 and 2 inputs parameters are selected.
Comment 2: Why Authors decided to use list of input parameters as changing/adjusted factor of the model? It must be explained in the text. Usually, the list of inputs is based on our knowledge about the modelled phenomena. If there are many of them, we can remove some based on eg. correlation analysis. But it shouldn’t be done randomly.
Response: Thank you for bringing this issue. we have gone through many numbers of literature which suggest that, and water level (WL) and water discharge (WD) plays a major role on suspended sediment yield (SSY). We have also performed the correlation tests (Table 2 and Table 3). We also found significant linear and nonlinear correlation between water discharge and water level with sediment yield. We saw that the water discharge and water level have relatively strong correlation with sediment yield (both linear and non-linear). Therefore, we have hypothesised that the water discharge and water level might have major effect on sediment yield, and thus we include water discharge and water level as input for the model. It should be noted that we have selected that best model by changing the input variables to select the best possible combination of parameters which impact on sediment yield using global optimisation genetic algorithm (GA).
Comment 3: In Table 2 and 3 Authors presented high correlations between WD and WL. Then, based on ANOVA, Authors stated that WD(Q) and WL are important inputs features to predict the SSY. According to my knowledge, the decision about including/excluding input model parameters should be made based on the correlation and one of highly correlated parameters should be excluded.
Response: Thank you for your comments. The Pearson correlation coefficient (PCC)(r) and Spearman rank correlation (SRC) are presented in Tables 2 and 3, respectively. The values of PCC between WD and SSY are found to be higher than the PCC value of WL and SSY. It is concluded that SSY is highly linearly correlated to the WD. In contrast rank correlation coefficient are between WD and SSY is higher as compare to SRC value of WL and SSY. Thus, it may be inferred that the influence of WL on the SSY is higher compared to the other WL parameters. It is seen in the ANOVA box plots of Figure 4 that all the input parameters have different distributions due to differences in the central lines in the boxes of each input. Therefore, it was decided to consider all these three input parameters initially for predicting the SSY. Finally, the optimum input parameters from these WD and WL is selected using global search optimisation algorithm i.e. genetic algorithm (GA) instead of trial and error approaches and gride search algorithm which generated local optimum and computational expensive and time consuming process. In total, three ANN and GA-ANN models, namely, WD, WL and WD + WL were developed by selecting different combinations of two input variables: water discharge (WD) and Water level (WL) and the best model was selected from this input space configuration. Best inputs (WD+WL) are selected using GA algorithm among all three possible combinations.
Comment 4: I think that the information about input model parameters should be declared for every model when it’s presented for the first time, not only in section 4.5. It would make text clearer for readers.
Response: Thank you for your comments. These following sentences were added in the introduction section of the revised manuscript. In this study, WD and WL data are used as inputs data for all models for prediction of SSY at Polavaram station of the GRB.
Comment 5: In section 4.3 the MLR equations should be presented, it would make text clearer for readers.
Response: Thank you for your comments. The MLR equation is given in equation 4, The detail description about the MLR method is given in section 2.3 of methodology.
Comment 6: It’s not clear how R2 (determination coefficient) was calculated in Table 8. It’s much higher than in tables 4-7. It should be revised or explained.
Response: Thank you for your comments and valuable suggestions. In Table 8, By mistaken, R2 is written in place of correlation coefficient(r). Now, it is corrected in Table 8. Now, the r value in Table 8 is same as r values of GA-ANN, ANN, MLR and SRC models from Table 4 to Table 7.
Comment 7: There are some linguistic mistakes such as in line 412.
Response: Thank you for pointing out that. We have made change in the revised manuscript.
Comment 8: I suppose that lines 483-489 should be removed.
Response: Thank you for pointing out that. Now, we removed the lines 483-489 from the revised manuscript.
Round 2
Reviewer 3 Report
The authors have IGNORED a significant part of my review concerning an inordinate number of acronyms, some used merely a few times.
As a reviewer, my job is to find errors and ways to help authors improve their paper for the READERS and the JOURNAL. Your job is to respond to EVERY issue raised by the reviewers. Please DO SO!
I thus recommend one more time a MODERATE revision. Since no such option is available, I pick MINOR revision but mean MODERATE one.
Author Response
Comment 1: The language is generally passable, but it is often excessively wordy, and it would benefit from fine-tuning. For example:
- Lines 695-697: This research only used the information from only Polavaram station and further research may be needed to strengthen these findings by using more information taken from different gauge stations. AWKWARD/REDUNDANT; 29 words
- COMPARE to: Information only from Polavaram station was used in this research; its findings should be strengthened with data from other gauge stations (21 words).
Response: Thank you for your valuable comments and suggestions. Now, it is modified and updated in revised manuscript.
Comment 2: However, the paper suffers from a FLOOD of unnecessary and UNESTABLISHED acronyms. Acronyms usually involve the first letters of well-established phrases. Here are the acronyms (most of them) that the authors use:
- Suspended sediment yield (SSY) used 50+ times
- artificial neural network (ANN) used 50+ times
- artificial neural network ... (GA-ANN) used 50+ times
- Multiple linear regression (MLR) used 42 times
- sediment rating curve (SRC) used 42 times
- Water discharge (WD) used 35 times
- RMSE UNDEFINED symbol used 31 times
- water level (WL) used 29 times
- genetic algorithm (GA) used 28 times
- Godavari River Basin (GRB) used 24 times
- artificial intelligence (AI) used only 9+ times
- Levenberg Marquardt (LM) used only 9 times
- Multi-Layer Perceptron (MLP) used only 7 times
- multi-layer perceptron neural network (MLPNN) used only 3 times
- least square regression (LSR) used only 3 times
- ANOVA (Analysis of Variance) used only 3 times
- Co active neuro fuzzy inference system (CANFIS) used only 2 times
- multiple non-linear regressions (MNLR) used only 2 times
- National Water Development Agency (NWDA) used only 1 times
With the exception of, perhaps, RMSE and ANOVA, none of your acronyms are well and widely established or broadly known, so, effectively you introduce 19 NEW WORDS. What is the benefit of using them? Clearly, you don't need acronyms that are used less than 20-25 times? And many of yours are used just a few times. Since this Journal has a very broad audience, assume that not one of your acronyms, even RMSE and ANOVA, will be known to the majority of the readers. The focus should be on the convenience for the READERS rather than the authors!
And, as opposed to most other journals, this one is generous and does NOT impose word-count limits, so there is ABSOLUTELY NO REASON to obfuscate your paper with unestablished acronyms. SPELL them ALL out & make your paper READER-FRIENDLY! And then benefit from a higher citation count.
Response: Thank you for valuable comments and suggestions. Now, we have made change and updated in the revised manuscript.
