Next Article in Journal
Litter Selfie: A Citizen Science Guide for Photorecording Macroplastic Deposition along Mountain Rivers Using a Smartphone
Previous Article in Journal
Utilizing Entropy-Based Method for Rainfall Network Design in Huaihe River Basin, China
 
 
Article
Peer-Review Record

Groundwater Level Prediction with Deep Learning Methods

Water 2023, 15(17), 3118; https://doi.org/10.3390/w15173118
by Hsin-Yu Chen 1, Zoran Vojinovic 2, Weicheng Lo 1,* and Jhe-Wei Lee 1
Reviewer 1: Anonymous
Reviewer 2:
Water 2023, 15(17), 3118; https://doi.org/10.3390/w15173118
Submission received: 22 July 2023 / Revised: 20 August 2023 / Accepted: 28 August 2023 / Published: 30 August 2023
(This article belongs to the Section Hydrogeology)

Round 1

Reviewer 1 Report

Dear authors, I had the pleasure of revising your manuscript entitled "Groundwater Level Prediction with Deep Learning Methods". Although I believe the AI can represent an advanced method for missing values imputation and for data prediction in hydrology to support water management, I noticed some parts of the paper that must be improved before possible publication. I highlighted the significant concerns in the PDF attached.

Best.

Comments for author File: Comments.pdf

No significant issues about the English Language are observed. Please see the attached PDF for minor suggestions.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, authors have used various deep learning networks into prediction of  groundwater level. Application of deep learning is quite simple and the paper has no sufficient novelty for publication. Additionally, evaluation of groundwater/AI models have been frequently done. Overall, the paper is not suitable for publication in this state:

(1) Definition of observed datasets should be more expanded.

 

 (2) Why did authors apply the optimized structures of the deep learning-NN model than other robust AI models such as ANFIS, M5, GEP, MARS, SVM, ensemble learning, RF, RT, and EPR?? This issue should be clarified in this paper.

 

 (3) Authors are strongly recommended to enhance literature review because there are a variety of investigations on the prediction of groundwater level by disparate AI models.

 

 (4) Performance of evaluations by adding scatter index, BIAS and DR to the present measures: Ocean Engineering 111, 128-135, 2016 and following research works can be helpful for improving results

 

 https://link.springer.com/article/10.1007/s11269-019-02463-w

 

 (5) Authors need to compare their results with literature (at least 20 papers published in this journal and other reputable journals) in terms of quality and quantity

 

 (6) Novelty of this work is very low. In this way, application of hybrid model of deep learning-NN should be justified for groundwater prediction.

 

 (7) What is unit of RMSE in whole of paper? insert units through the paper

 

 (8) Why did not authors use Wavelet function?? This issue should be clarified.

 

 (9) Setting parameters/performance platform of hybrid deep learning-nn model should be detailed.

 

 (10) Literature review is very poor and authors should add more related works.

 

 (11) The best performance should be added to the conclusions and abstract sections.

 

 (12) Authors are emphatically recommended to use SVR, RF, RT, M5, EPR, MARS, GEP, and Ensemble learning Technique (in both training and testing stages) for comparisons and therefore their performances should be added to all sections of paper.

 

 (13) After comparisons, Violin diagram and Tailor diagram should be added to the results and discussions.

 

 (14) Authors are emphatically recommended to compare the present results with related works (at least 20 papers published in the reputable journals) in terms of accuracy level and structural complexity of the used AI models.

 

 (15) After performing Comment#13, authors should design a Stacked Machine Learning model for more robust comparisons.

 

 (16) Authors should perform sensitivity analysis to define the least the most effective parameters (finding optimum number of lag time for groundwater level prediction) on the output of the AI models in this study.

 

 (17) What are criteria for selection of input parameters? possibility of PCA application can be investigated in this paper.

 

 (18) Uncertainty, reliability, and resiliency of the proposed RBF networks should be investigated.

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Accept as is

Back to TopTop