Parcel-Level Mapping of Horticultural Crop Orchards in Complex Mountain Areas Using VHR and Time-Series Images

Round 1

Reviewer 1 Report

This article can be accepted now

Author Response

Dear reviewer, thank you very much for your recognition of this article. About “English language and style” in open review，We received a request that “English language and style are fine/minor spell check required”. For this suggestion, we used the editing services listed at https://www.mdpi.com/authors/english. All the revisions in the manuscript have been marked up with “Track Changes”.

Author Response File: Author Response.docx

Reviewer 2 Report

This work is characterized by complying with the elements of a scientific and technological research, appropriate to be published in this journal.

On the other hand, it is highlighted that the authors have carried out rigorous research in relation to data management and mathematical and computational methods that allow demonstrating the proposed hypothesis, achieving reliable results for decision-making in the study area.

 

Author Response

Dear reviewer, thank you very much for your recognition of this article. About “English language and style” in open review，We received a request that “English language and style are fine/minor spell check required”. For this suggestion, we used the editing services listed at https://www.mdpi.com/authors/english. All the revisions in the manuscript have been marked up with “Track Changes”.

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This study focused on the parcel-level mapping of horticultural crop orchards using time-series images and VHR images. This study has some sense but it is a mess of the methods and approaches whose reliability and reasoning is very questionable. Moreover, this is one of the studies which tries to do multiple sensitive segments at the same time but fails to do so. I suggest that the authors refocus this content on a more specific study objective and to analyze and reflect their research much more thoroughly according to the recent scientific articles.

The Introduction is unnecessarily long and lacks relevant information at the same time. While it is not important to mention unsuitability of UAV images, as well as MODIS and Landsat, the explanation about the relevance of the Sentinel-2 images in this study is unclear. You focused on the VHR images in the title and the vast majority of the Introduction but the exact interaction and complementarity of VHR and Sentinel-2 is unclear.

The Materials and methods are not novel and well-structured in my opinion. The topic of the research is clear, but the combination of the VHR and Sentinel-2 images is not reasonably explained in my opinion, nor the combination of these data regarding their spatial resolutions, which can put the reliability of such results into question. Only the second segment of the method (lines 223-225) is a very sensitive and difficult procedure, which is here performed using a somewhat primitive approach (vegetation indices from multitemporal Sentinel-2 data). Moreover, combining the results obtained based on the 0.55 m and 10 m spatial resolution is somewhat explained in the figure 6 but the robustness and reliability of this approach are very questionable. What about the orthorectification from the image suppliers? How does this affect this approach? The study area subsection should be shortened and much of its content would be better represented with figure(s) in my opinion. Also, this subsection completely lacks references, which questions reliability of these information.

The primary scientific contribution of this study in my opinion deals with the parcel extraction. The methodology regarding the crop classification is not adequately explained and represented. For example, lines 369-394 are unnecessary besides having “flashy” formulas in the paper. Also, no references were used in this entire part, implying that the authors developed and proposed this method (?!). Meanwhile, the Results section dominantly focuses on the results of classification and spectral information based on the Sentinel-2 imagery! You collected samples from apple orchards, cherry orchards and greenhouses (Figure 1) and how did exactly and reliably differentiate these (spectrally very specific) classes from other land cover types or other orchards?

The Discussion section is abysmal as it analyzes these results in a sandbox, which zero (!?) references to previous similar studies.

This manuscript has twelve authors but the Author Contributions (lines 649-653) are not written according to specifications and the justification for many authors here is questionable. “Helped to prepare sample data” might not be enough for authorship, but it is up to the editor to evaluate that.

The English language throughout the paper is very poor and must be significantly improved.

Specific comments:

Lines 18-19: Unnecessary repetition of the same information.

Lines 46-47: Display these values in other measurement units improve readability.

Lines 186: GPS or GNSS? This is very important. What was the accuracy of such georeferencing?

Figure 1. This figure must be thoroughly improved. The international readers are dominantly not familiar with the left part of the figure. Instead, the relative position of the study area in a single sub-image within China would be much more informative. The middle subfigure should be increased and the color scale of DEM or samples must be changed to ensure clear visibility of all items. The two subfigures on the right are too small to give adequate information and are not explained. It might be the best to remove them.

Table 1. This table is poorly formatted. I suggest creating a figure representing a timeline instead.

Figure 2. “indexs” is wrong.

Lines 665-790: References are not formatted according to the specification and the lack of doi number makes it harder for the reviewer to check used references.

Reviewer 2 Report

1) In the abstract, it is mentioned that "DL models are used to extract the parcels". What models do you use? No explanation is given on them. The features seem to be convolutional features but what about the extraction process?

2) Several block diagrams are shown but the individual results of them are not clearly given

3) Is Table 3, a gold standard data?

4) What is the reason for some type providing good classification results and other less quality results in the confusion matrix?

5) Your data seems to be a colour image. Do you incorporate 3D DL models for your implementation?

6) A comparative analysis with other existing works must be given to validate the superior nature of your proposed method