Groundwater Level Trend Analysis and Prediction in the Upper Crocodile Sub-Basin, South Africa

Round 1

Reviewer 1 Report (New Reviewer)

This is an interesting study. The following comments may be considered by the authors:

1- Literature review is not critical. The authors may explain the results of reported studies revealing their limitations. 2- The superiority of present study in comparison with the existing ones needs to be clarified. 3- Explain the reliability of data, management of outliers and the extension of results for other regions of the World. 4- Interpret the results based on theoretical considerations. Why is R2 =0.9 for GB and R2 = 0.125 for SVR model? 5- The higher correlation coefficient does not necessarily show a good model, that is why more physical sound of the results is required in the text. 6- The assumptions of this study and its limitations should be provided.  

7- Calibration and validation of GB model is required. Please discuss about the results of calibration and calibration in the text.

8-  How can  we extend the results of GB model  in this study to other regions?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

I sincerely appreciate the dedication and effort invested in preparing the manuscript. The research presented in this work utilizes machine learning and statistical analysis to address a highly significant global concern pertaining to groundwater resources. However, I have several comments and suggestions to improve the clarity and overall quality of the manuscript:

·       The manuscript contains numerous typos that should be thoroughly addressed.

·        It is essential to number the references in the order of their appearance in the text and listed individually at the end of the manuscript.

·        The maps and figures would benefit from additional effort to enhance clarity. Specifically, the Long and Lat axis should be included in all maps of Figure 1, and the font size of the text should be increased for improved readability. Additionally, the same figure repeated four times in the manuscript seems redundant (Figure 1).

·        There are multiple errors in the Cross-reference of Tables and Figures, making it difficult for readers to follow the content seamlessly.

·        If equations are presented in the manuscript, it is recommended to use either the Microsoft Equation Editor or the Math Type add-on for better presentation and consistency.

·        Acronyms, abbreviations, and initialisms should be defined the first time they appear in the abstract, main text, or first figure/table. After their initial definition, they should be added in parentheses next to the written-out form.

·        The introduction should be expanded to highlight the significance of the proposed model for predicting future groundwater levels (GWL) based on the Gradient Boosting (GB) Algorithm and inputs of rainfall and antecedent GWL. Comparisons with key publications in the field would help clarify the novelty and contributions of this study.

·        In the Materials and Methods section, it is unclear whether the method of determining rainfall through cross-correlation with GWLs is novel or previously used by other researchers. Further clarification on this matter is needed.

·        Regarding the results (Lines 452-454), the statement about the GB and SVR models performing better in calibration than validation due to the dataset size difference raises a question. It would be valuable to explore how changing the sizes of the calibration and validation datasets affects the predictive model's results.

·        The conclusions and recommendations section should elaborate on how climate change is linked to the research findings and its implications on groundwater resources (Lines 508-510). Expanding on this aspect would enhance the manuscript's relevance and significance.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

This study predicted groundwater levels using Gradient Boosting regression modelling after analyzing its historical trend from 2010 to 2020 in the Upper Crocodile Sub-Basin, South Africa. Corresponding conclusions will benefit the development of groundwater utilization strategies. I recommend this MS to be published after a minor revision.

But I’m not sure if the temporal trend of groundwater levels observed in the studied region and the GB model are applicative in other regions. Is there something special here?

There are some minor editing errors, e.g., Fig. 1 appears three times; What’s the mean of “Error! Reference source not found” in the manuscript?

none

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report (New Reviewer)

The authors improved the manuscript and I recommend the acceptance of this interesting research. 

Author Response

The authors would like to once again thank the reviewer for the comments provided in improving the manuscript. 

Reviewer 2 Report (New Reviewer)

 

The revised manuscript is significantly improved compared to the first version. I suggest considering for publication.

Please check the comment: The conclusions and recommendations section should elaborate on how climate change is linked to the research findings and its implications on groundwater resources. Expanding on this aspect would enhance the manuscript's relevance and significance.

Author Response

The authors would like to once again thank the reviewer for the valuable comments provided in improving the manuscript.

The conclusions and recommendations section should elaborate on how climate change is linked to the research findings and its implications on groundwater resources. Expanding on this aspect would enhance the manuscript's relevance and significance.

• Please refer to line 509 to line 518 of the conclusion.  Rainfall, a climatic parameter is used in explaining the behaviour of groundwater levels. 

 

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

Round 1

Reviewer 1 Report

The authors did not give sufficient attention to the aquifer (hydrogeology) and geology of the area. Their focus is on the numbers (groundwater level and rainfall). The result revealed also that groundwater levels in some boreholes do not respond to rainfall (at least 25%) and another 25% gave negative results which means with the presence of rainfall, there is a decline in gw level. Therefore, the authors missed the contribution of regional groundwater circulation since dolomitic aquifers have high transmissivity owing to karst occurrence.  In the introduction section, the authors must clearly present if they are dealing with the whole upper crocodile river basin or A21D (upper crocodile river sub-catchment.

There are lots of published works in International journals that have not been referred and can help the authors to improve the manuscript. 

Detailed comments are included in the annotated PDF file.

Comments for author File: Comments.pdf

Good

Reviewer 2 Report

This paper applied the Mann Kendall test to analyze the temporal trend of the monthly ground water level and the Gradient Boosting (GB) regression to predict the monthly ground water level in the upper Crocodile Basin of South Africa. The analysis of the temporal trend in the monthly ground water level and the prediction of the monthly ground water level are apparently two individual components. It is lack of linkage between these two components. This raises a question why the authors include these two into this paper. According to the results, it appears that the models do not perform well at a number of stations, questioning the feasibility and/or reliability of the models developed. The modeling results are presented, but the discussion is very limited and did not provide insights to the groundwater in the study region.

Other comments:

11) L28-59 are not directly related to the topic of this paper, and thus can be removed.

22) L140-142: It is not clear how the results can be used to build adaptive and mitigation capacity to climate-related disasters.  

33)  L167-172: The dams and main rivers should be shown in Fig. 1.

44) Table 1: Nine stations are listed in Table 1, while there are only 8 stations in Fig. 1. Check the stations used in this study.

55)  L219-221: Although the tool “Panda” was used to fill the missing data points, the method/algorithm used should be provided.

66) L222-245: The Mann-Kendall test is well known and commonly used. This section can be removed.

77) L246-273: Cross correlation analysis is well known and commonly used. This section is not necessary.

88) L274-284: Autocorrelation analysis is well known and commonly used. This section is also not necessary.

99) L285: GB regression is the core method used in this paper. More details are desired, for example, the algorithm of the GB regression, the hyperparameters, etc.

110) L294-295: It is not adequate to justify the selection of the input variables just according to the previous studies. The authors need to justify the selection for their study site. The selection of input variables always plays a critical role when developing data-driven models.

111) L298-300:  When developing data-driven models, the data division for training and validation is also one of important steps. Justification for the data division is desired.

112) Figs. 2-3: Not all plots in these figures are necessary.

113) Table 2: It is not necessary to report the intermediate results from the trend analysis.

114) L421: Report the values of the hyperparameters for the models.

115) Table 5: The model failed to model GWL at three stations. This raises a question on the use of the models for predicting monthly groundwater level.

116) Model performance in both training and validation needs to be reported.

Quality of English language is fine.