Comparison of Three Imputation Methods for Groundwater Level Timeseries
, Giulia Passadore 2, Silvia Bertoldo 1, Andrea Sottani 1 and Andrea Rinaldo 2,3Round 1
Reviewer 1 Report
Comments for author File: 
Comments.pdf
Author Response
Response to the reviewer’s comments in the attached pdf file
Author Response File: Author Response.pdf
Reviewer 2 Report
Hydraulic head data are probably the most important directly knowledge we can get in term of subsurface water status and movement, hence are crucial in hydrogeological studies. The present study looked at one of the very common problems when dealing with groundwater head data, namely missing data in time series. It compared 3 different methods for missing data imputation, which are spline interpolation, autoregressive linear model, and patched kriging. There was also a fourth approach which used the average value the above three. The whole idea is very simple and the methods used are already existing, so I would recommend that the authors polish the paper more and dig into the implications of the results.
First, to evaluate the performance of different data imputation methods, the authors chose several criteria such as Euclidean distances, Fourier analysis, autocorrelation, microscale. Among them, only Euclidean distances represent accuracy but the other only show similarity. We need some common criteria such as RMSE or R2 in order to better evaluate the results.
Second, as far as I understand, there are interpolation methods better suitable for spatial data and some methods for time series data. Take kriging for instance, it is a method for spatial data. However, the current study mixed them up, and discussion on this issue is considerably lacking. When filling in the groundwater data gap, whether you can borrow information from the neighborhood in space is very important. If you just have one standalone borehole, will you still get the same result?
Third, when dealing with groundwater data, it is not enough to just consider data alone, otherwise groundwater data would be the same as any other data. I would strongly recommend the authors take geological/hydrogeological condition of the research area also into account.
Fourth, according to my understanding, the study used only 3 methods instead of 4. Averaging should not be considered as an independent method; it is only a post-processing step. The paper did not showcase whether the fourth approach (averaging) actually make any difference.
Fifth, the paper did not discuss in detail about different types of missing data, e.g. a point of data, a block of data etc. It is strongly recommended that the sensitivity of data on different missing data forms are discussed.
Figures:
Please add a flow chart for the whole procedure including data preprocessing, interpolation method used, comparison method used etc. Also, there is lacking detailed descriptions on figures in general.
Legend is missing in Figure 1.
Figure 2-5, please use dots instead of lines for observed data.
Figure 6-8, Please add observed data in panel (a), and all the legends are missing from panel (b), (c) and (d).
Author Response
Response to the reviewer’s comments in the attached pdf file
Author Response File: Author Response.pdf
Reviewer 3 Report
The submitted paper by Meggiorin et al. titled “Comparison of four imputation methods for groundwater level time series” provides a work about four different imputation methods for groundwater level records analysis. The manuscript in general is well written. Nevertheless, I would like to mention that the introduction is in a higher level from the following chapters. Thus, I suggest a thorough syntax and academic English check. In the end, no real conclusions are highlighted.
My recommendation is major revisions.
General comments
-Always referred in the third person.
-Avoid using keywords that are already in the title.
-Abstract needs some improvements. The lines 8-12 needs reformulation. Please highlight the results.
-Introduction: Great work. To the point analysis and important general information are mentioned.
L.116-117: Please be more specific. Mention step by step the methodology.
L.123-126: Avoid add sentences that are used in technical reports. Please change this paragraph.
-Data about the study area are missing (ex., type of aquifers).
-Conclusion: Provide the take home message.
- A graphical abstract could attract the reader’s interest.
-The chapter of methodology has no sense. The authors have to gain the reader’s interest.
Suggested literature
Modeling groundwater and surface water interaction: An overview of current status and future challenges (2022). Science of The Total Environment. 846, 157355.
Formulation of Shannon- Entropy model averaging to groundwater level prediction using Artificial Intelligence models (2021). International Journal of Environmental Science and Technology.
Multivariate statistical analysis for the assessment of groundwater quality under different hydrogeological regimes (2017). Environmental Earth Science. 76: 349.
Author Response
Response to the reviewer’s comments in the attached pdf file
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have done a nice job responding to previous questions and comments. I appreciate the plots for sensor 92 to help clarify the tables that follow. I also like the maps despite not showing a strong geographic pattern, Figure 12b may suggest that confined versus unconfined conditions are important for the imputation method.
Reviewer 3 Report
The manuscript has been improved according to the reviewers' comments and in my opinion, is ready for publication.
