Dryland Winter Wheat Production and Its Relationship to Fine-Scale Soil Carbon Heterogeneity—A Case Study in the US Central High Plains

Round 1

Reviewer 1 Report

Int.

I think there was a paragraph missing in the introduction to talk about the central High Plains. Characteristics of the location, production profile, of the producers.

Line 42 - typically not.

Soil chemical and physical properties are important for yield variations. It's a valid statement.

 

MM

Line 71-73 - What is the reference used for this classification of soils????

Results

Fundamental a Graph of soil water balance in the period of wheat cultivation. It will increase the understanding of the soil moisture behavior and the water stress suffered by the crop.

Discussion

Line 281 - "It should be noted that in this dryland environment, soil moisture dynamics are expected to play an important role in driving spatial and temporal variability in grain yield."

 

I completely agree, but how can I do this analysis without a graph of soil water balance in the wheat growing season?

Comments for author File: Comments.pdf

Author Response

1. I think there was a paragraph missing in the introduction to talk about the central High Plains. Characteristics of the location, production profile, of the producers.

Answer: We appreciate the reviewers' comments; we have rewritten the first paragraph in the context of the US Central High Plains framework (Lines 31-33)

2. Line 42 - typically not. Soil chemical and physical properties are important for yield variations. It's a valid statement.

 Answer: We are not sure what referring the reviewer. Certainly, soil and physical properties are relevant for yield variations. However, we were trying to address how the relative contribution of soil to yield can vary over the different spatial scales.

MM

3. Line 71-73 - What is the reference used for this classification of soils????

Answer: Thanks for pointing out the missing reference. Soil Series Descriptions and Series Classification, USDA were added in line 77

Results

4. Fundamental a Graph of soil water balance in the period of wheat cultivation. It will increase the understanding of the soil moisture behavior and the water stress suffered by the crop.

Answer: We understand that it can be discouraging not to find moisture data in this study. Certainly, we understand that while we emphasize soil nutrient and carbon data, we don't mention anything about water balance. Our decision to exclude moisture from this study does not undermine its importance for understanding yield variability. Collecting and analyzing moisture data over time can be complex and resource-intensive due to the number of sample points we used to study and the dynamic nature of this property. Instead, the main focus of our research was an examination of the role of soil carbon in soil heterogeneity because it is typically a more stable property that changes much more slowly than others, buffering rapid changes in moisture. While we recognize moisture's relevance, we believe its integration requires a separate investigation due to its complexity and the need for extensive data analysis. The only data we have that comes close to this is yearly precipitation, which has been included in the supplementary materials.

 

Discussion

5. Line 281 - "It should be noted that in this dryland environment, soil moisture dynamics are expected to play an important role in driving spatial and temporal variability in grain yield."

Answer: Please note the response to comment #4

6. I completely agree, but how can I do this analysis without a graph of soil water balance in the wheat growing season?

Answer : Please note the response to comment # 4

Reviewer 2 Report

This study investigates the relationship between soil attributes and wheat yield across eleven fields in the central Great Plains. Utilizing a range of statistical and machine learning methods, including Random Forests. The study finds that Total Carbon (TC), Phosphorus (P), and sand content are significant predictors of yield variability, although their impact is not uniform across all fields.

I think the paper is potentially quite interesting and relevant to the journal, however I have a number of concerns and queries regarding some parts of the work.  Also, much of the presentation of the article needs to be improved.

1) While the introduction does a good job of setting up the importance of the study, it could be more explicit about what gap in the literature the study aims to fill.

2) The objectives are listed at the end of the introduction, but it would be helpful to be clear about how each objective contributes to the overall aim of the study.

3) The paper mentions soil sampling at 0–15 and 15–30 cm depths. Why these specific depths were chosen.

4) The 30 m grid spacing is mentioned, but there's no justification provided for why this particular spacing was chosen.

5) There is mention of climate, but are there any control variables that were accounted for in the study?

6) The description of the study site is quite detailed, which is good. I think it will be helpful to explain why the 11 fields were chosen for the study.

7) The paper mentions that no definitive conclusions about climate effects should be derived from this dataset. This is an important limitation and should perhaps be highlighted more prominently.  I think this is probably my biggest concern with the work overall.  Clearly climate is surely a highly significant attribute.

8) The paper notes that fields were on a winter wheat-fallow rotation and could not be sampled in consecutive years. This is a significant limitation that could affect the interpretation of results, especially if soil properties change significantly year-to-year.

9) Various chemicals used for weed control are mentioned, but no discussion on how these might affect soil properties or wheat yield. Could this be a confounding factor?

10) Why was a 30 m x 30 m grid was chosen?

11) Why was Empirical Bayesian Kriging (EBK) chosen over others methods - I can't see reference to reasoning in the article.

12) The paper mentions testing normality but doesn't specify which tests were used or how this was achieved?

13) The paper mentions a significance level of p < 0.05 for the Kruskal-Wallis tests and Spearman's correlation but doesn't specify if this is a one-tailed or two-tailed test.

14) How did you ensure that the RF model didn't overfit the data?

15) The use of units need to be consistent throughout the document. eg. l170 has -1 as subscript, but l173 does not.  Also, l271 refers to "Figure 8" but l272 refers to "Fig. 8".  Against, please be consistent. These aren't the only examples, please check consistency throughout the entire document.

16) l180 - "significant differences in medians among fields" is mentioned, but it's not clear to me what test was used here?  Also, Bonferroni post hoc test was used, which is known to be conservative and produces the widest confidence intervals - why Bonferroni in this work?

17) The yields for fields in 2018 and 2019 are discussed, but it's unclear to me if there were differences between these two years, and if so were they statistically significant and if this is even relevant to the study?  The authors state earlier in the document that "no definitive conclusions about climate effects should be derived" (l91) and that "climate is often more significant than soil" (l44), but then pairwise compare fields between 2018 and 2019 without any real discussion - surely annual differences will play a major role in differences? 

18) The most important predictor variables selected are discussed but not why some expected important variables (eg. pH, clay, and EC) were not found to be significant predictors.

19) The importance of soil variables was not uniform across depths is mentioned. This is an interesting observation that should be discussed more.

20) The effect of slight elevation changes on yield is indicated, but it isn't fully explored why this might be the case. More discussion on the mechanisms behind this observation would be useful.

21) The discussion section mentions that soil moisture dynamics, which are likely crucial in a dryland environment, were not measured. This is a significant limitation, given that soil moisture could be a confounding variable affecting both soil properties and crop yield.

22) While the paper does address multicollinearity before conducting PCA, it's not clear how it was handled in other statistical analyses like Random Forest, which can also be affected by multicollinearity, although less much less so.

23) Figure 1 - I can't read the coordinates on the axes, the site labels and the distance marker has 0.12 and 0.23 so close together it's not completely obvious they are separate. The text size on Figure 2, although still small, are at least readible at 100%.

24) Figure 2 - There is an issue with precision on the axes. 40.16N appears twice, as does 40.15N. An extra decimal place might help here. Same problem on horizontal axis too.

25) Figure 3 - Outliers and Extreme values are listed as seperate legend entries, but with the same symbol.  Are these in fact the same thing and only a single legend entry needed?  I don't think significance level needs to be part of the graphic, the inclusion in the figure caption is enough.

26) Figure 4 - See Figure 2 regarding axis coordinates. The caption mentions the color blue, but I don't see any blue in this plot?

27) Figure 5 - I don't think significance level needs to be part of the graphic, the inclusion in the figure caption is enough.

28) Figure 7 - What are the lines on this plot?  Only a relationship is mentioned.  Was a linear model used?  A correlation (spearman or pearson or other?) is mentioned in the text, but then I see error bars.  Is there a statistically significant relationship here?  It's really difficult to see what's going on actually - I wonder if there is a better way to present this information?  The y-axis label uses t/ha rather than t ha-1 like almost the entire document.  Please be consistent.

29) Figure 8 - Does variable importance in this case refer how much the accuracy decreases when the variable is excluded?  Can be more specific what measure is used?

30) l58. What does "not significantly affected" mean exactly?

31) References should follow the MDPI style guide. eg. Journal names should be abbreviated.  There are quite a few simple mistakes too, eg. [32] "3rd editio."

32) The supplementary material doesn't seem to have been included with the submission, so I can't check these.

 

Largely well written with very few minor mistakes.

Author Response

We have carefully considered reviewer # 2 comments and put our best effort into addressing every one of them. We thank the reviewer for your time and effort in reviewing our manuscript. The feedback has been invaluable in improving the content and presentation of this paper.

Answers:

1. While the introduction does a good job of setting up the importance of the study, it could be more explicit about what gap in the literature the study aims to fill.

Answer: Thank you very much for pointing this out. We have adjusted the text to be clearer in Lines 55-57

 

2. The objectives are listed at the end of the introduction, but it would be helpful to be clear about how each objective contributes to the overall aim of the study.

Answer:  Modified. We appreciate the comment. We have adjusted the text to be clearer in lines 60-68

 

3. The paper mentions soil sampling at 0–15 and 15–30 cm depths. Why these specific depths were chosen.

Answer. These depths were established because surface soils are more responsive to changes in land management. The 0-15 cm soil depth represents a critical part of the soil profile due to is often more responsive to short-term environmental changes, such as land management practices, nutrient availability, and climate fluctuations. The majority of plant roots and associated nutrient uptake occur within this depth range, which can directly impact plant health and productivity. Therefore, layer 0–15 cm serves as a valuable indicator of soil dynamics. On the other hand, the 0-30 cm depth extends beyond the root zone into the subsoil, offering insights into soil properties that can influence plant growth indirectly. The 0-30 cm depth provides a broader perspective, encompassing deeper soil layers that may respond differently to environmental fluctuations.

 

4. The 30 m grid spacing is mentioned, but there's no justification provided for why this particular spacing was chosen.

Answer: It should be emphasized that the samples were obtained directly from the ground and were not interpolated (i.e., Kriging) predicted values. In this way, we were able to reduce the cost, time, and effort of data collection and analysis, instead of spacing the grid 15x15m. 

.

5. There is mention of climate, but are there any control variables that were accounted for in the study?

Answer: The only data we have that comes close to this is yearly precipitation, which has been included in the supplementary materials. We understand that it can be discouraging not to find moisture data in this study. Certainly, we understand that while emphasizing soil nutrient and carbon data, we didn't mention any about water balance. Our decision to exclude moisture from this study was not to dismiss its importance in understanding yield variability. Instead, it was for resource reasons collecting and analyzing moisture data over time can be complex and resource-intensive due to the amount of points we considered for this study and the dynamic nature of this property. Instead, our research sought to provide a focused examination of soil carbon role in soil heterogeneity because it is typically a more stable property and their changes occur much more slowly, buffering rapid changes in moisture.

 

6. The description of the study site is quite detailed, which is good. I think it will be helpful to explain why the 11 fields were chosen for the study.

Answer: It had been already explained in lines 97-99. We wanted to be able to obtain a greater sample size by including different fields in consecutive years, all planted with hard red winter wheat.

 

7. The paper mentions that no definitive conclusions about climate effects should be derived from this dataset. This is an important limitation and should perhaps be highlighted more prominently.  I think this is probably my biggest concern with the work overall.  Clearly, climate is surely a highly significant attribute.

Answer: Please note the response to comment # 5

 

8. The paper notes that fields were on a winter wheat-fallow rotation and could not be sampled in consecutive years. This is a significant limitation that could affect the interpretation of results, especially if soil properties change significantly year-to-year.

Answer: Our research sought to examine the role of soil carbon (C) in soil heterogeneity. The variations in C typically happen very slowly (over decades), and mostly is not detectable yearly. Therefore, it shouldn't affect the result interpretation.

 

9. Various chemicals used for weed control are mentioned, but no discussion on how these might affect soil properties or wheat yield. Could this be a confounding factor?

Answer: Yes, this is a valid question. Weed chemicals, including herbicides and pesticides, can indeed influence soil carbon variability. This can have direct effects on soil microbial communities, altering the rate of organic matter decomposition. Over time, the cumulative use of herbicides and pesticides can lead to gradual changes in soil carbon levels. However, the short-term impacts may be minimal. Therefore, this is not a factor that can affect our results.

 

10. Why was a 30 m x 30 m grid was chosen?

Answer: Please note the response to comment # 4

 

11. Why was Empirical Bayesian Kriging (EBK) chosen over others methods - I can't see reference to reasoning in the article.

Answer EBK is a powerful tool that often outperforms traditional kriging methods in cross-validation exercises, ensuring that the model's predictions align well with observed data. EBK can effectively smooth noisy yield data, reducing the impact of outliers or localized irregularities in the dataset. The performance of the spatial interpolation to predict wheat grain yield per field is now further explained in Table 1 in the supplementary information.

 

12. The paper mentions testing normality but doesn't specify which tests were used or how this was achieved?

Answer: Certainly, we didn't mention that the Shapiro-Wilk test was used to analyze normality. It was added to line 154

 

13. The paper mentions a significance level of p < 0.05 for the Kruskal-Wallis tests and Spearman's correlation but doesn't specify if this is a one-tailed or two-tailed test.

Answer: It was added to line 156

 

14. How did you ensure that the RF model didn't overfit the data?

Answer. We used a K-fold cross-validation to get a more robust estimate of the variable importance (line 170)

 

15. The use of units needs to be consistent throughout the document. eg. l170 has -1 as subscript, but l173 does not.  Also, l271 refers to "Figure 8" but l272 refers to "Fig. 8".  Against, please be consistent. These aren't the only examples, please check consistency throughout the entire document.

Answer. Thanks for the recommendation. All of them were corrected.

 

16. l180 - "significant differences in medians among fields" is mentioned, but it's not clear to me what test was used here?  Also, Bonferroni post hoc test was used, which is known to be conservative and produces the widest confidence intervals - why Bonferroni in this work?

Answer. Certainly, we didn't mention that an analysis of variance non-parametric Kruskal-Wallis was performed to evaluate statistically significant differences among groups. The Bonferroni is widely accepted in multiple comparisons for non-parametric tests. We also tried other Post-hoc test, i.e. Fisher's LSD; however, we observed false positive errors

 

17. The yields for fields in 2018 and 2019 are discussed, but it's unclear to me if there were differences between these two years, and if so were they statistically significant and if this is even relevant to the study?  The authors state earlier in the document that "no definitive conclusions about climate effects should be derived" (l91) and that "climate is often more significant than soil" (l44), but then pairwise compare fields between 2018 and 2019 without any real discussion - surely annual differences will play a major role in differences? 

Answer.  I would like to apologize for the oversight in not sharing the supplemental material along with the climate data as originally intended. We recognize the importance of supplementary information to understand the dataset and our research comprehensively. We have attached the supplementary material to this submission to rectify this situation.

18. The most important predictor variables selected are discussed but not why some expected important variables (eg. pH, clay, and EC) were not found to be significant predictors.

Answer. The main focus of our research was examining the role of soil carbon (C) in yield variability because it is typically a more stable property that changes much more slowly than others, buffering rapid changes in moisture. By narrowing our focus to soil C, we can go deeper into its role in yield variability and its interaction with other factors.

19. The importance of soil variables was not uniform across depths is mentioned. This is an interesting observation that should be discussed more.

Answer: We appreciate the suggestion, but we cannot consider going deeper into explaining the relationship of the soil variable at different depths. Broadening the discussion may divert attention from our primary objectives.

 

20. The effect of slight elevation changes on yield is indicated, but it isn't fully explored why this might be the case. More discussion on the mechanisms behind this observation would be useful.

Answer. I would like to express my gratitude for your taking the time to thoroughly review our work and for providing valuable feedback. I appreciate your guidance in this regard. However, we have carefully considered narrowing the focus of the discussion to ensure that it aligns more closely with the scope of our research. Broadening the discussion may divert attention from our primary objectives and research goals. To address your feedback, I have revised the discussion section to ensure that it remains closely aligned with our research objectives and does not stray into unrelated areas.

 

21. The discussion section mentions that soil moisture dynamics, which are likely crucial in a dryland environment, were not measured. This is a significant limitation, given that soil moisture could be a confounding variable affecting both soil properties and crop yield.

Answer: Please note the response to comment # 5

 

22. While the paper does address multicollinearity before conducting PCA, it's not clear how it was handled in other statistical analyses like Random Forest, which can also be affected by multicollinearity, although less much less so.

Answer. PCA was applied to identify dominant relationship capture similarities between fields as an exploratory analysis. The cluster separation makes visualizing the relation between different fields and soil variables easier. However, PCA does not always deal well with multicollinear data. In contrast, random forest can also be applied when predictor variables are highly correlated, allowing us to deal with model complex interactions among predictor variables.

 

23. Figure 1 - I can't read the coordinates on the axes; the site labels and the distance marker has 0.12 and 0.23 so close together it's not completely obvious they are separate. The text size on Figure 2, although still small, are at least readible at 100%.

Answer. Thank you for pointing out. We have included these edits in Figure 1

 

24. Figure 2 - There is an issue with precision on the axes. 40.16N appears twice, as does 40.15N. An extra decimal place might help here. Same problem on horizontal axis too.

Answer. Modified. We have corrected the axes in Figure 2

 

25. Figure 3 - Outliers and Extreme values are listed as seperate legend entries, but with the same symbol.  Are these in fact the same thing and only a single legend entry needed?  I don't think significance level needs to be part of the graphic, the inclusion in the figure caption is enough.

Answer. Thank you for pointing this out. We have deleted the significance level.

 

26. Figure 4 - See Figure 2 regarding axis coordinates. The caption mentions the color blue, but I don't see any blue in this plot?

Answer. We apologize for this error, and we have corrected the caption in Figure 4

 

27. Figure 5 - I don't think significance level needs to be part of the graphic, the inclusion in the figure caption is enough.

Answer. Modified. Thank you for pointing this out. We have deleted the significance-level text below the figure.

 

28. Figure 7 - What are the lines on this plot?  Only a relationship is mentioned.  Was a linear model used?  A correlation (spearman or pearson or other?) is mentioned in the text, but then I see error bars.  Is there a statistically significant relationship here?  It's really difficult to see what's going on actually - I wonder if there is a better way to present this information?  The y-axis label uses t/ha rather than t ha-1 like almost the entire document.  Please be consistent.

Answer. We apologize for this error. We took care of being consistent throughout the manuscript.

 

Figure 8 - Does variable importance in this case refer how much the accuracy decreases when the variable is excluded?  Can be more specific what measure is used?

Answer. Thank you very much for pointing this out. Indeed, we weren't clear of explaining this. We already added this info in the material and methods in line 171-176

 

29. What does "not significantly affected" mean exactly?

Answer. We have adjusted the sentence to be clearer (Line 62-63)

 

30. References should follow the MDPI style guide. eg. Journal names should be abbreviated.  There are quite a few simple mistakes too, eg. [32] "3rd editio."

Answer. Modified. They were now accordingly referenced. 

31. The supplementary materiaokl doesn't seem to have been included with the submission, so I can't check these.

Answer. We apologize. Now, it is included in the submission.

Reviewer 3 Report

accept after these:

1.      Change the color of sampling points in Figure 1.

2.      Show the scale in meters in Figure 1

 

3.      Please add a flowchart to clearly show the steps of the work.

Author Response

 

1. Change the color of sampling points in Figure 1.

Answer. Thank you for pointing this out. We have modified the color (grid points) in Figure 1

2. Show the scale in meters in Figure 1

Answer. Modified 

3. Please add a flowchart to clearly show the steps of the work.

Answer. Thanks for the suggestion. However, we believe the material and method sections clearly communicate our perspective without the need for an additional workflow Figure. However, we have reviewed and clarified various statements in the material and methods section.

 

Round 2

Reviewer 1 Report

The authors improved the quality of the article. They inserted a citation in the soil classification and improved the introduction.

However, the Figure:

"Figure S1. Monthly precipitation (cumulative water) and mean temperature for the 2017-18 and

18 2018-19 winter wheat growth seasons (April to July) in Akron, Colorado, US."

In my opinion, it should be part of the body of the article. I suggest inserting it into the article and discussing it.

Figure S1 is too important to be “supplemental.”

Comments for author File: Comments.pdf

Author Response

We thank reviewer #1 for understanding our work limitations and your expertise in reviewing our manuscript. Your feedback has been invaluable in improving the manuscript's quality and presentation.

"Figure S1. Monthly precipitation (cumulative water) and mean temperature for the 2017-18 and 2018-19 winter wheat growth seasons (April to July) in Akron, Colorado, US."

In my opinion, it should be part of the body of the article. I suggest inserting it into the article and discussing it.

Figure S1 is too important to be "supplemental."

Answer. Added. The Supplementary Figure S1 was included as part of the body of the article.

Reviewer 2 Report

I thank the authors for their updated manuscript.  Although most of the questions and concerns that were raised have been addressed, I feel there are still some outstanding issues that the authors must consider before the paper can be published.

Figure 2 - It isn't clear which axis tick the coordinates correspond to.  There are more ticks than coordinates!

Figure 4 - Same issue regarding axis ticks/labels as Figure 2.

4) It's still not clear why 30 m grid spacing was used.  In the response there is mention of 15 m grid spacing, but I'm not sure why this is relevant?  Was there prior experimentation that determined a 30 m grid spacing?  Or is this just an arbitrary number just randomly selected?  How are you sure you have sampled sufficiently to capture the variation in the underlying processes?  Please justify your experimental design.

5) I think this explanation should be explicit in the text.

25) I don't feel that the concern regarding Outliers and Extreme values has been address.  These have the same symbol.  If an outlier and an extreme value are different, then they should have different symbols.  If they are the same thing, then only one legend entry is needed.

28) There are still concerns regarding Figure 7.  The point in the 1st review about being able to see what is being presented in the graph is important.  I can barely see all of the linear model lines.  Even though colours are used, I find it extremely difficult to determine which points correspond to which line, particularly when there is point overlap.

I think the presentation of these results must be improved.  Is it necessary that they are on the same axes?  Would 11 small individual plots help here?  I don't know, but please generate a figure that is readable/interpretable.  What linear model is being used  mx+c? Least squares fit? It looks like a straight line, but the confidence limits seem to be non-linear?  There is mention of correlations in the document body(l.262) - Only spearman's correlation is mentioned in the rest of the document, but these mentioned correlations would be Pearson since derived from Linear Regression Model?

There are many observations outside of the error bars, which makes me wonder if these data fit would even be statistically significant if you were to calculate the p-values.  What is presented here isn't clear at all.

l.226, l.237 Spearman -> Spearman's

 

Author Response

We thank reviewer #2 for understanding our work limitations and your expertise in reviewing our manuscript. Your feedback has been invaluable in improving the manuscript's quality and presentation.

1. Figure 2 - It isn't clear which axis tick the coordinates correspond to.  There are more ticks than coordinates!
2. Figure 4 - Same issue regarding axis ticks/labels as Figure 2.

Answer #1 and #2 . Modified. We appreciate the feedback and suggestions to enhance the clarity and presentation of the figures. We have carefully addressed the revision recommended in Figures 2 and 4. However, locating the coordinates in the thicks, as you requested, is challenging. Since the degrees, minutes, and seconds were converted to decimal degrees, the GIS software located the coordinates in this way. We improved the figures as much as possible; however, we may not be able to make further adjustments. We believe the updated Figures now reflect the data and findings in the manuscript.

 

3. 4) It's still not clear why 30 m grid spacing was used.  In the response there is mention of 15 m grid spacing, but I'm not sure why this is relevant?  Was there prior experimentation that determined a 30 m grid spacing?  Or is this just an arbitrary number just randomly selected?  How are you sure you have sampled sufficiently to capture the variation in the underlying processes?  Please justify your experimental design.
4. 5) I think this explanation should be explicit in the text.

 

Answer #3 and #4. Added. The fields used in this study were converted to new rotations in 2017 to support the study of dryland precision management practices. This study is essential in order to allow us to establish preliminary precision management zones. We didn't have a priori knowledge about the spatial soil heterogeneity in the study area. Therefore,  this study adopted a 30 m x 30 m grid sampling as a baseline or initial survey. The 30 m grid spacing was acquired due to the extreme flat relief characteristics across the study area as well as the plot's sizes considered in this project. In a flat terrain scenario, we do not expect soil variability to change drastically over short distances to consider higher sampling density. On the other hand,  since our experimental plots are relatively small, larger grid intervals might not be appropriate for this study. We anticipated that the preliminary relationships between soil properties and crop yield established from 2 years of data would be instructive and allow preliminary zone delineation, but that successive years of data would clarify, and perhaps alter, the preliminary relationships. Certainly, this point is not defined in the text. We have clarity on this in line 120-124.

 

5. 25) I don't feel that the concern regarding Outliers and Extreme values has been address.  These have the same symbol.  If an outlier and an extreme value are different, then they should have different symbols.  If they are the same thing, then only one legend entry is needed.

Answer #5. Modified. Thank you for your thoughtful review and for bringing to our attention this error in the plot. The graph does not include the extreme values, which was an error in the legend. This was corrected in the updated Figure (now Figure 4)

 

6. 28) There are still concerns regarding Figure 7.  The point in the 1st review about being able to see what is being presented in the graph is important.  I can barely see all of the linear model lines.  Even though colours are used, I find it extremely difficult to determine which points correspond to which line, particularly when there is point overlap.
7. I think the presentation of these results must be improved.  Is it necessary that they are on the same axes?  Would 11 small individual plots help here?  I don't know, but please generate a figure that is readable/interpretable.  What linear model is being used  mx+c? Least squares fit? It looks like a straight line, but the confidence limits seem to be non-linear?  There is mention of correlations in the document body(l.262) - Only spearman's correlation is mentioned in the rest of the document, but these mentioned correlations would be Pearson since derived from Linear Regression Model?

There are many observations outside of the error bars, which makes me wonder if these data fit would even be statistically significant if you were to calculate the p-values.  What is presented here isn't clear at all.

Answer #6 and #7. Modified. Figure 7 (now Figure 8 in the updated manuscript version) was redone and added the linear equations.

8. 226, l.237 Spearman -> Spearman's

Answer #8. Modified. Now, Spearman's term is used consistently throughout the manuscript.