Geographically Weighted Regression-Based Predictions of Water–Soil–Energy Nexus Solutions in Île-de-France

Round 1

Reviewer 1 Report

I found this manuscript very interesting and suitable for publication in this journal.  I have some suggestions, which hopefully would improve the manuscript.

·        I suggest reducing the number of paragraphs in the introduction section.  There are 29 paragraphs in the Introduction, which would distract readers.  There are similar issues in other sections.

 

·        Could you include brief discussions on the uncertainty quantification of the prediction of the EC pattern for the year 2054?

Author Response

Q1. I suggest reducing the number of paragraphs in the introduction section. There are 29 paragraphs in the Introduction, which would distract readers.  There are similar issues in other sections.

A1. Noted and amended accordingly in the whole manuscript.     

 

2. Could you include brief discussions on the uncertainty quantification of the prediction of the EC pattern for the year 2054?

A2. Noted and amended accordingly. The discussion is now presented in the text in page 2 as: “

The EC prediction for the year 2054 was based on (i) OLS regression for the year 2018 with a high determination factor of 84% and a weak spatial autocorrelation of standard residuals (p-value less than 5% and Moran’s Index = 0.228); and (ii) GWR for the year 2018 with a high determination factor of 85% and a weak spatial autocorrelation of standard residuals (p-value less than 5% and Moran’s Index = 0.172).

Author Response File: Author Response.docx

Reviewer 2 Report

It is an original manuscript with high scientific value. This manuscript presents an interesting and beneficial study. Additionally, the theme of the paper is within the scope of the journal. The paper is technically correct, well organized. The illustrations are adequate and the results look secure. I recommend it for publication.

Author Response

The authors would like to thank the Reviewer.

Reviewer 3 Report

The research conducts a meaningful work to explore possible explanatory factors of water-soil-energy solutions and do related predictions for some French regions. Authors employed a series of statistical methods to do analyses of different models. However, readers may better understand this work if the main logic could be presented more clear.

Some technical suggestions/questions:

1) Authors give two reasons (lines 191-194) for selecting 2054 as the year that is predicted. Two questions are:

               i) If using the same methods to do predictions for other years, are the predicted results still credible?

               ii) CAMCM seems an important method in the process, because it is this method that indicates 2054 as the “proper” year. If this is the point, why do not include CAMCM in figure 3 which illustrates the structure of methodologies.

2) The authors used the data of 2006 and 2012 to predict the year 2018 in which the true data were collected. The validation shows high accuracy for 2018. Does this imply the validation of the prediction for 2054 by using years 2006 and 2018?

3) Table 2 shows the GWR models effectively handle the spatial effects since most of the Moran’s I are near 0. Readers may be interested in those explanatory variables that have main impacts to the dependent variable (although different regions may have different main explanatory variables). So it is better if these results will be presented more noticeable.

And some details:

1) lines 266-269, and lines 294-297: the subscripts are not properly typeset.

2) lines 274, 435, 580: “spatial correlation” should be “spatial autocorrelation”, because it qualifies the correlation between the residuals which is one “variable” that distributed in different locations.  

3) lines 277-278: spatial autocorrelation tests if observations of a georeferenced variable are autocorrelated, and the test statistic (e.g., the Moran’s I) itself is asymptotic normal distributed, it does not test if the variable is normal or not. So the (1) may be deleted.

4) line 284: explanatory parameter-> explanatory variable. Parameter often means the coefficients of the explanatory variables or the variance/covariance of the error term.

5) line 304: “It is hypothesized…”. The “hypotheses” usually used in a statistical hypothesis testing context, so it may be better to use “assume” (or its synonym) here.

6) figure 4: what is the difference between the dashed frame and the solid frame? The (e) seems better to put in a full frame and be centered (so as to the title of this figure).

7) line 378: authors may need to double check the order number of each figure thereafter.

8) line 398: it may not be proper to say that the GWR and ANOVA are statistical econometric model although they are widely used in many fields including econometrics. Calling them “Statistical model” may be more safe.

9) line 446: R2, superscript

10) line 526: at the end of the line, “…” or “.”?

Author Response

Q1. Authors give two reasons (lines 191-194) for selecting 2054 as the year that is predicted. Two questions are:

   i) If using the same methods to do predictions for other years, are the predicted results still credible?

   ii) CAMCM seems an important method in the process, because it is this method that indicates 2054 as the “proper” year. If this is the point, why do not include CAMCM in figure 3 which illustrates the structure of methodologies.

 

A1.

   i) The following is added to the text:

“The time span of the prediction is 36 years, which is commonly used in the scientific literature for LULC predictions, mainly because 36 is a multiple of 12 (period extended be-tween the years 2006-2018), and because it constitutes a relatively long-term horizon”.

So, the prediction results are considered credible for a period of/less than 36 years. 

   ii) The authors would like to clarify that the CAMCM method is indirectly included in the first phase of the structure of the used methodology: “Identifying possible explanatory variables” which is based on data collection techniques comprising several methods as the CAMCM.

The authors think that adding explicitly “the data collection techniques” would steer the attention away from the main target of the “structure of the used methodology” intending to define a regression relationship structure among all dependent and exogenous variables.  

 

 

Q2. The authors used the data of 2006 and 2012 to predict the year 2018 in which the true data were collected. The validation shows high accuracy for 2018. Does this imply the validation of the prediction for 2054 by using years 2006 and 2018?

A2. The authors conducted two simulations to predict the LULC for the year 2054 using the input data of the years 2006-2012 and 2006-2018. Both predictions generate almost the same result. Thus, based on the development and LULC changes trend in the study area, the prediction using the CAMCM is considered correct.

However, knowing that the methodology could be used by other research, the authors used the data of 2006-2018 to indicate that in case of changing development trends over the years the biggest LULC transition period could represent better the development and LULC interactions trend.    

 

Q3. Table 2 shows the GWR models effectively handle the spatial effects since most of the Moran’s I are near 0. Readers may be interested in those explanatory variables that have main impacts to the dependent variable (although different regions may have different main explanatory variables). So it is better if these results will be presented more noticeable.

A3. The impacts of the explanatory variables are broadly presented and explained in section 3.4 “Effect size (Variance Analysis)”.

 

Q4. Details:

1. lines 266-269, and lines 294-297: the subscripts are not properly typeset.
2. lines 274, 435, 580: “spatial correlation” should be “spatial autocorrelation”, because it qualifies the correlation between the residuals which is one “variable” that distributed in different locations.
3. lines 277-278: spatial autocorrelation tests if observations of a georeferenced variable are autocorrelated, and the test statistic (e.g., the Moran’s I) itself is asymptotic normal distributed, it does not test if the variable is normal or not. So the (1) may be deleted.
4. line 284: explanatory parameter-> explanatory variable. Parameter often means the coefficients of the explanatory variables or the variance/covariance of the error term.
5. line 304: “It is hypothesized…”. The “hypotheses” usually used in a statistical hypothesis testing context, so it may be better to use “assume” (or its synonym) here.
6. figure 4: what is the difference between the dashed frame and the solid frame? The (e) seems better to put in a full frame and be centered (so as to the title of this figure).
7. line 378: authors may need to double check the order number of each figure thereafter.
8. line 398: it may not be proper to say that the GWR and ANOVA are statistical econometric model although they are widely used in many fields including econometrics. Calling them “Statistical model” may be more safe.
9. line 446: R2, superscript
10. line 526: at the end of the line, “…” or “.”?

 

A4. Noted and updated accordingly.

• For part (3), the authors would kindly clarify that the introduced hypotheses are used as an essential part of the research plan.
• For part (6), figure 4 is updated: please find the attached word file.
• For part (7), the authors would like to kindly ask the reviewer for more precise clarification. Thus, the authors can update accordingly.
• For part (10), the following is added at the end of line 526: “or other unidentified factors”.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

for art (7), the authors would like to kindly ask the reviewer for more precise clarification. Thus, the authors can update accordingly.

There are two figures numbered figure 5.

Author Response

The authors would like to thank reviewer 3.

The figure numbers are updated.