A Case Study: Groundwater Level Forecasting of the Gyorae Area in Actual Practice on Jeju Island Using Deep-Learning Technique
Round 1
Reviewer 1 Report
This paper presents deep-learning techinque to forecast groundwater level in an island. This paper is well orgainzed and the study is also sigifnicant. However, some factors are not well introduced. My comments are as follows.
(1) My biggest concern is the reason why you choose mean wind spped, sun hours, evaporation, minimum temperature, and daily precipitation as predictors. The rehcarge and discharge items of groundwater should be explained in detail and thus the relationship between each factor and groundwater level changes should be anlayzed.
(2) Keywords are suggested to revise.
(3) Hydrogeology conditions in the study area should be added.
(4) A list of data used in the study is suggested to add.
(5) There are many observation wells, and only JH Gyorae 1 well is discussed. The paper aimed at the actual practice. How about the effectiveness of other wells?
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
With interest I have read this manuscript on "A Study of the Groundwater Level Forecasting in Actual Practice Using Deep-Learning Technique of the Gyorae Area in Jeju Island".
The authors might benefit their manuscript’s presentation if they would avoid being overly specific with information that would require citations (eg “ about 81% of the available water resources in Jeju Island ”) In the abstract. It is advisable to share generally applicable observations that are not location specific in the first sentences of the abstract, to engage and sustain the interest of potential readers.
Unclear if this is a research article or a case study. If it is submitted like a research article, then the results presented by this manuscript, which read more like a practical case study, should be significantly expanded, sharing recommendations and abstraction/generalisations of the major findings, to help guide the decision making approach of practitioners in similar situations.
I would suggest a careful proofreading from an experienced English writer, to help with any instances of poor grammar, such as the use (or lack) of “the”. For example, when referring to a specific item, one needs use “the”: eg line 72 “With selected hydro-meteor … ” becomes “With the selected hydro-meteor”. Also carefully check for typos eg line 160 replace “serious” with “series”.
Please use the full expanded nomenclature for any abbreviated words used: eg for “ANFIS” replace with “adaptive neuro-fuzzy inference systems (ANFIS)” on the first instance it is used. Likewise for all abbreviations used in the main text.
At the end of the introduction the authors attempt to introduce the objectives of the presented research, however they do not substantiate the targeted novelty nor the need for “pragmatic” GWL forecasting. Can the authors elaborate on both of these aspects further? Doing so in the introduction will help keep the reader engaged with this research and interested in the findings.
It is interesting to see the use of Granger’s causality test to justify the use of certain time series as input for the proposed model. Can the authors compare their results with other in the literature to help broaden the discussion and understand how costly (in terms of predictive ability) would be to not have all the presented input offered herein.
May I suggest restructuring the manuscript as follows: section 3.1 would become more of a “Application and results” section, followed by 3.2 and 3.3 which will now form an independent Discussion section.
Text font in the Figures need be checked and revised for legibility. For example the legends in Fig.1, 2, 8 and axis labels Fig.4 and 5.
Line 458: remove “Please add”
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
In this paper, a prediction study of GWL was conducted using deep learning techniques, and a good result was obtained. However, there are still some problems in the study that need to be explained.
1. The abstract is more descriptive of the basic design of the study and lacks the summary of the results.
2. Lines 67-72, this part of the description on the selection of variables is recommended to be put into the method.
3. This paper presents an empirical analysis of Jeju Island using LSTM and measured data, and its applicability to other study areas is unknown.
4. LSTM enables long-term prediction. After training and validation of historical data in this paper, there is a lack of predictive applications for the future. What is the practical guidance of this study for future GWL changes in the region?
5. Please further elaborate on the innovation and value of the paper, or the innovative approach used in this paper.
6. Separate numbering is recommended for each formula.
7. Lines 154-160, please add references.
8. You conducted the cross-wavelet analysis between the collected hydro-meteorological data and groundwater level data. How the hydro-meteorological indices affect the groundwater level? Is there any interaction mechanism between the hydro-meteorological indices the groundwater level? Please clarify.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors repsoned my comments well. However, I have one more comment.
(1)in Lines 68-69, you mentioned the reason why you carried out the study is that GWL forecasting based on deep learning is raely applied. You should check whether it is true. To my knowledge, there is a lot of studies on the application of deep learning in groundwater hydrology.
Author Response
→ That sentence was edited to more clearly meaning.
→ Author agree to the comments, there are numerous studies about GWL forecasting based on deep learning having been conducted. The line 68-69 means that there are many studies but it is still hard to apply in actual practice due to its uncertainty. That sentence was edited
Author Response File: Author Response.docx
