Supporting Pro-Poor Reforms of Agricultural Systems in Eastern DRC (Africa) with Remotely Sensed Data: A Possible Contribution of Spatial Entropy to Interpret Land Management Practices

Round 1

Reviewer 1 Report

This is a well-researched and presented manuscript which makes a valuable contribution by detailing the construction of a remote-sensing classification system for distinguishing business-oriented agricultural (BOA) land use from small-farmer-oriented agricultural (SOA) land use in Eastern Congo, a region  where a lack of records and physical access make remote analysis important. The construction of the classification system is rigorous and well-presented and the construction of the HNVI index of entropy/hetergeneity makes good use of a selection of spectral signatures. Cloud cover and some difficulties in distinguishing small fields in agricultural and pasture use complicates the analysis of satellite images, but methods are presented to deal with the complications in the final analysis. The map of the BOA vs. SOA classification for the study region provides a nice conclusion of the exercise with a policy-relevant interpretation. 

Two relatively minor points to raise:

1) the paper notes that the final classification results are similar to other survey results from an unpublished Caritas survey, but I wondered if information about control points could also be leveraged to give a measure of predicted vs. observed. 

2) the title of the manuscript might be reconsidered as the "analysis of entropy" seems to emphasize the technical building-block of the method but the purpose of the article seems more like answering the question, "can a remotely-sensed land cover and entropy classification approach support pro-poor land tenure reforms in the DRC?"

 

Author Response

We thank the reviewer for the revision and especially for pointing out possible improvements in the description of how we validated the results and how we titled the whole paper. Please find the attached point-by-point answer to the proposed revisions.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper used an object-based classification approach on Sentinel-2 imagery to classify land cover., Furthermore, to classify production systems in the “agriculture” and “pasture” land use classes, binary classification based on an entropy value threshold was performed, and applied to in the eastern Democratic Republic of Congo. This study is of significance for supporting adapted intervention strategies in development cooperation and pro-poor agrarian land tenure reforms in conflict ridden landscapes. However, there were still some problems should be discussed further.

• The detailed process during the LU/LC CLASSIFICATION PHASES in Figure 5 should be supplemented. How to realize the process from segmentation to LULC classes aggregation and accuracy computation?
• What’s the relationship between “object-based classification” in section 2.4 and “agent-based visual correction” in Figure 5? Please clarify the process of “agent-based visual correction” used in this study.
• In Table 5, 10 wetlands and burned areas (especially wetlands) are classified as the urbanized and bare soil. Please explain and demonstrate the reasonability.
• For Eq. (1), please add the reference. In addition, please further supplement the physical meanings of ρNIR and ρRED.
• For Eq. (2), i(j) from 0 to N-1, why not from 1 to N? Please explain. In addition, please demonstrate the applicability of entropy theory in distinguishing the agricultural production system in this study.
• Please further explain the meaning of “the patch level of the SEG layer” in lines 321-322.
• How to determine the threshold values of HNDVI for agriculture LULC and pasture LULC in lines 339 and 340? Please supplement the detailed calculation process and scientific evidence, and meanwhile verify the reliability of these threshold values.
• In Table 6, the names of LULC classes are not consistent with those in the previous contents, such as Table 5.
• Please supplement the determination process and evidence of the threshold value (0.7) for both agriculture and pasture in lines 383 and 384, and demonstrate the reliability of the value.
• The reasonability and reliability for the results of agricultural production system identification in this study should be further analyzed and demonstrated quantitatively.

Author Response

We thank the reviewer for the in-depth analysis of the manuscript, revealing several inconsistencies in formula wording, referencing, and some points worth clarification. We also thank the reviewer for guiding us to a more comprehensible text throughout the workflow description. Some of the suggested revisions were accepted only partially due to the impossibility to rethink and re-implement the experience, but more efforts have been put to clarify and argue the authors ‘choices. We hope these will meet the satisfaction of the reviewer. Please find the attached point-by-point answer to the suggested revisions.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors use an object-based image analysis approach to generate a binary classification of pastoral and agriculture classes using an identified entropy criteria on Sentinel 2A imagery.

The manuscript is well written however, my primary concern with the paper is the lack of spatial or temporal analysis to see if the optimization method is replicable in different (but environmentally similar) geographic areas, as well as replicable in the presence of phenological changes. 

Is this method unique to this particular case study?

There should be some replication, even on the same area/region to see if the method’s results are consistent and not unique to this single dataset.

 

There were also several missing/incomplete references throughout the text

Author Response

We thank the reviewer for highlighting very interesting research perspectives that we were pleased to add to the conclusion section. Moreover, we thank the reviewer for asking for an appropriate check of all references that allowed correcting missing ones. Please find the attached point-by-point answer to the suggested revisions.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thanks for the authors for their revisions. However, I think that the authors have partially revised the manuscript according to the comments. The contents in lines 345-355 (now) and Figure 9 cannot explain and solve all the problems. I consider that the authors should carefully supplement relevant contents in the manuscript and respond to following comments before accepting for publication.

• Supplement the principle of entropy theory and equation (2) in reflecting homogeneous or heterogeneous vegetation surface before equation (2).
• Supplement the calculation results of the other threshold values (from 0.3 to 1.2 for agriculture LULC and from 0.3 to 0.9 for pasture LULC) and the relevant analysis and discussions in the section of results and discussion, and then demonstrate the reliability of the threshold value 0.7.
• Supplement the quantitative analysis and demonstration of the reasonability and reliability for the results of agricultural production system identification in this study in the section of results and discussion (for examples, comparing with the fact, results obtained by other methods, or other similar studies, and so on). Not lines 345-355 and it is not enough to see Figure 9 by ourselives.

In addition, there are lots of mistakes in writing of the revised manuscript (such as HNDVI or H_{NDVI（subscript）}? m\ in line 348, and so on), please check and modify carefully.

Author Response

We thank the reviewers for giving us the possibility to improve our manuscript. Please see the attached file for detailed answers.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors did not truly address the replication issue brought up in the previous round, however seeing as they have modified the title and purpose of their paper to be more of a case study application approach than a method paper, replication may not be critical to the paper as the method is no longer central focus of the manuscript.  

Author Response

We thank the reviewer for highlighting that the purpose of the paper is now made more clear.

Author Response File: Author Response.docx