Remote Quantification of Soil Organic Carbon: Role of Topography in the Intra-Field Distribution

Round 1

Reviewer 1 Report

Please see the following comments/suggestions:

Line 54-72, why informing about the comparison between sentinel data and hyperspectral data when no such comparison was done in this study? Why was the comparison important in the context of this study?

Line 63, there are two references presented, but going forward only the results of one are mentioned, why mention the other study when no results are provided?

Castaldi et al. [18] found that the prediction accuracy obtained by Sentinel-2 data was similar to that retrieved by APEX airborne hyperspectral data. Can you please provide some more references to studies with similar results?

Line 83-91, Can you please provide concise objectives? And why the site area/location was selected?

Line 99-101, Can you please provide small descriptions of these soil types?

Can you please provide the GIS coordinates of the site locations?

Please provide the name of the GPS device.

What procedure was adopted for collecting and transporting etc. of soil samples?  Please provide in the description.

What was the type of sampling? How were the samples selected? Please provide in the description.

Why only a 10 cm soil layer was selected? Please provide in the description.

Line 158, which software was used for the resampling of the spatial resolutions?

Line 179, reference (‘European space agency, 2010’) is not according to mdpi requirements.

Please also give the full forms of abbreviations for the indices (Table 2)?

Formula for GNDVI would include NIR instead of RE?

Formula for NDRE would include NIR instead of RE?

Which software was used for extracting the information from indices?

Line 254: Please describe why the radial basis function kernel for SVM was chosen.

Line 275: Please expand the description for RF and SVR i.e., separate headings.

How was the spectral information for each sampling point using Sentinel 2 image extracted?

Please bring the section “Assessment of the model performance” to the last of the “Materials and Methods” section.  

Please provide the equations for R2, RMSE, etc.

What are the hyperparameters used for RF and SVR i.e., their values?

Line 299: how predictive modeling was performed for SOC? Which software?

Why 5-fold cross-validation was used? Please provide in description.

What are the descriptive statistics of SOC for each site? Please provide in the results section.

Line 366: can you please provide the model details in the methodology section instead?

Can you please also explain why SVR performed better than RF models?

Line 608, Can you also please describe why B11 was also the better indicator for SOC?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

General Comments:

 The study aims to enhance the understanding of SOC inversion at the intra-field scale, a relevant topic for precision agriculture. The research sets clear objectives, particularly in examining the role of topographical factors, the innovative aspects, such as the use of LiDAR data and multitemporal Sentinel-2 data, could benefit from further clarification to fully grasp their contribution and novelty. The experiment's design, focusing on these technological applications, suggests a thoughtful approach.

 Specific Comments:

 Lines 78-80: The significance of employing both single and multitemporal data in the context of SOC formation variables needs more elaboration. A more detailed background could help justify the choice of methodology.

 Lines 120-121: The manuscript should offer a clearer explanation of how the percentage of crop residue cover is determined. The influence of a high crop residue cover on Sentinel-2's spectral bands and its implications for SOC prediction lacks depth in the current exposition.

 Line 122: "soil samples" instead of "said samples".

 Lines 207-211: The rationale behind using multitemporal images should be articulated more clearly, and the statement regarding the SOC content being below the detection threshold of spectral methods requires stronger justification.

 Lines 249-252: The discussion would benefit from identifying the most suitable set of topographical features for modeling SOC values, including the methodology for determining this set.

Providing visualizations of the DEM data and topographical features for the fields studied could enhance the comprehensiveness of the analysis.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Some minor comments:

1. Line 99-101, Can you please provide small descriptions of these soil types?

 - What is the Wickham 4 association?

 - A general reference to the Upton series i.e., why this classification etc., would be good  here.  https://www.landis.org.uk/soilsguide/series.cfm?serno=2004

2. What was the type of sampling? How were the samples selected? Please provide in the description. What procedure was adopted for collecting and transporting etc. of soil samples?  Please provide in the description.

  - 2.0 x 2.0, should be consistent throughout all descriptions.  

  - Third paragraph should go with the first paragraph.  

3. Line 299: how predictive modeling was performed for SOC? Which software?

 - Thanks, but which software i.e., programming language, for producing maps? Can you also please add? 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf