Complementarity between Textural and Radiometric Indices From Airborne and Spaceborne Multi VHSR Data: Disentangling the Complexity of Heterogeneous Landscape Matrix
Round 1
Reviewer 1 Report
I have completed my review of the manuscript by Lang et al. (“Complementary textural and radiometric indices from airborne and spaceborne multi VHSR data: Characterizing horizontal and vertical vegetation structure”). The manuscript applies a previously developed mapping methodology to aerial and satellite based data sources to examine the contributions of grain size and radiometric information on the characterization of landscape structure. The manuscript highlights the importance of fine-scale (~0.5 m) imagery is necessary to properly represent landscape heterogeneity in satellite-based mapping. This information is especially important for the development of mapping and monitoring frameworks in diverse landscapes dominated by differing plant functional types. The manuscript did a great job of laying out the justification, explaining methodologies, and discussing the key results. Furthermore, the manuscript provides detailed examples of how certain combinations of texture scale and radiometric information can lead to differing representations of vegetation structure. Discussion and conclusions seem well supported by the results, but the results could be presented a bit more explicitly to make everything clear (see below).
Despite the value of the paper, a few modifications could improve the impact even more.
(1) Grammatical errors were relatively common in this manuscript. I noted the first few I saw (see specific comments), but then stopped as they were just too common. The manuscript requires a careful copy-editing to ensure high quality.
(2) Classification accuracy for the four strata was rather impressive, though some additional information in the methods regarding the differentiation between LL and HL would be useful. For example, are these differentiated by a height threshold? Growth form of woody plants? Expert opinion?
(3) In the results section, the manuscript would benefit by a brief introductory sentence for each paragraph highlighting what component of the objectives/questions is being addressed. Sometimes this was done, but other times I found myself having to flip back to the introduction or methods sections to remind myself what a specific set of results were trying to show. I simple transition sentence at the beginning of each of these paragraphs would help. This would also help the reader assess whether the discussion and conclusions are well supported.
Even with these comments (and some additional ones below), I see a lot of value in this manuscript as an extension of the authors’ previous work and hope that these can be leveraged not just in France, but elsewhere.
#Specific Comments#
Line 14: Replace “shows” with “show”
Lines 37: Replace “over extent” with “over an extent”.
Line 52: Replace “us” with “as”, add a comma after “metrics”, and delete a space after “Nonetheless”
NOTE for Line 52: From this point on, I have stopped noted grammatical/spelling errors. There were relatively common in the first few pages of the manuscript. I suggest that the authors carefully copy-edit the manuscript.
Lines 51-61: The content in this paragraph is great and highlights the need for people to carefully consider what they are doing with their data.
Line 81: It might be good to put “shrubs” and “trees” in parentheses after “low” and “high”, respectively, just to make it abundantly clear what you are referring to. I think that “shrub” and tree” are more commonly used in terms of plant functional types. Or maybe it is a height cutoff?
Line 84: One thought regarding interpretability that the authors do not mention: inputs should had minimal correlation. The authors do not state this, but this is a key assumption encompassed within interpretability.
Line 89: This sentence lacks some specificity, such that I don’t know what is being described. Probably the authors are referring back to the previous work [44], but it would be good to be explicit.
Lines 20-106: The introduction is very well organized and provides a wonderful setup for the manuscript.
Line 164: By “Appendix”, do the authors mean “Appendix 1”?
Lines 238-240: This sentence is pretty awkward. Consider reorganizing.
Figure 2. Panel d is not referenced in the text and I am not sure exactly what I am supposed to be getting from it. In contrast, panels a-c are quite descriptive. I think that Figure 2d should be referenced in section 3.3.4, but that is just a guess.
Line 264: I think that the authors mean section 3.2, not 4.2.
Lines 315-316: It would be useful to remind the reader that PLAND refers to the frequency of the strata and LSI refers to the configuration of patches. I realize that the reader can go back and look, but a quick reminder is helpful.
Figure 5. I found the caption confusing. I assume that “(a – c)” and “(b – d)” should read “(a, c)” and “(b, d)”, respectively.
Lines 331-344: I found it difficult to link the text and the figure here. Additional text highlighting why (in an operational sense) the panels are different would help me interpret what is going on here. In other words, is this an effect of the angle arising out of the optimization? Or does this have to do with the PC1-PC2 relationship for each strata? This was not clearly stated for me, or at least I was not able to put all the pieces together in my own mind.
Figure 7-8. There is a p-value here, but no statement of a test. Also, given the large number of pixels being used, I do not think that a p-value is particularly meaningful. Another quick note: with this many points, it is generally better to present a heat map rather than a scatter plot. The scatter plot emphasizes the extreme observations.
Figure 12. The panel labels are incorrect.
Line 382: Starting off this, and later paragraphs, with the objective related to a specific comparison would be helpful. For example, the ANDVI and PNDVI comparison is mostly about the scale of imagery. This is done for some results (e.g., Influence of pansharpenning), but not all.
Lines 644-645: Were the training and validation datasets constructed at random or were there geographic holdouts. Due to spatial autocorrelation, the latter is often a more robust test of model or classification performance (generally described under the concept of transferability).
Lines 645-647: The same sentence is repeated.
Author Response
Please
find our answers to your comments in the pdf file attached.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript “Complementary textural and radiometric indices from airborne and spaceborne multi VHSR data : Characterizing horizontal and vertical vegetation structure” (remotesensing-442726) is extensively dealing with a comparison of aerial and Pleiades imagery for the detection of vegetation type and structure in a Mediterranean nature conservation area.
While the manuscript has undoubtly a high technical standard and is written well, it has two crucial disadvantages. While one is rather structural and solvable, I see no real solution for the second one. I really liked the manuscript and leave it with major revisions, but the task is not easy.
Major comments
The manuscript is hard to follow due to the several technical steps, resulting in a variety of data-products and vegetation classes with different abbreviations introduced at different stages in the text. Perhaps it would be good to have a table with all abbreviations in the beginning. Additionally, I found it hard to relate the steps (chapters) in the method section to those in the results section. Perhaps this could be a bit more aligned.
The aerial image and the Pleiades image, which are frequently compared, are from a complete different phenological period and even a different year. This is very obvious, e.g. in Figure 1b as well as in the scatter plot Figure 4. I guess that there is simply more vital biomass around in September that End of June. Instead, the authors explain that with “different radiometric units” (L329). The effects of phenology and radiometry/texture are not really separable, therefore the results cannot be interpreted in a pure technical/resolution way. The authors discuss this shortly in the conclusions (L584-589), but are probably not aware of the consequences this might have on the results. If there would be biophysical measurements (LAI, SPAD-chlorophyll, FVC, etc.) which could prove a similar state of the vegetation.
Minor comments
Title: I would skip the second part of the Title (Characterizing horizontal and vertical vegetation structure) as this might lead to other expectations the article is delivering. With this title, the reader might expect information on light and shade leafs of trees, etc.
There should be a mapping of the NATURA 2000 habitat types (4digit codes) available for this area. Please state the habitat types in the text (e.g. within the study area section). Perhaps you could even use these areas and compare your different data layers for these parts.
L4: “was tested” instead of “was to tested”
L 21: “ecosystem services” instead of “services”
L39-44: I am not sure that these points are generally accepted. For i: Especially in times of big data, often the other argument is used (if we have a lot of information, the algorithm will supply relevant information). For iii: It might be tempting to use no real classification methods any more, but then we need other validation measures (e.g. measurements of related biophysical variables).
L93-97: Very long sentence. Please make it shorter, especially because this is the research question, that should be easily understandable.
L103: “we indicate to” instead of “we will tried”
L104: you are not really using “different satellite images”. It is just a single Pleiades scene and aerial imagery (which are not satellite based)
Figure 1: Please add a legend for the NDVI (perhaps it might even be better to colour code it)
L175: Please give a reference for this statement
L196: Why not exactly 100 m instead of 108?
Chapter 3.3.3: This chapter is hard to follow. Please try to simplify and use less abbreviations.
Author Response
Please
find our answers to your comments in the pdf file attached.
Author Response File: Author Response.pdf
Reviewer 3 Report
The article deals with testing a methodological framework to VHSR satellite imagery. This framework was previously developed for airborne images. The method is based on textural and radiometric analysis using different indexes. In order to reach the main objective of the study, used available images from airborne and Pléiades satellite images. They focus on the Mediterranean area of south of France.
Broad comments
The topic of this article is relevant for monitoring biodiversity. In this way, it is also relevant for Remote Sensing Journal. The article is very dense written with lots of details but maybe not enough releveant detail. In my opinion, it is not clear enough whether the methodology could be applied to satellite images. I would not recommend this study for publication in Remote Sensing Journal, because it needs to be shorten with a more equilibrium between the results and the discussion. The latter is too short and not enough discussed with the perspective of the objective.
Specific comments
There are few typological problems so the specific comments concern mainly problem of comprehension.
L52 : “Nonetheless,….analysis”. What do you mean ? It could be useful to precise the limitaions of landscape metrics and how they are related to classification process and conceptual flaws.
L72 : “…careful parametrization…”. “Careful” is not a proper term in a scientific publication in my opinion.
L97 : “…it is important …” important for what ?
L163 : “…. Were assumed to be stable….” How far is this true ? It seems to be an important hypothesis
L174 : LSI ratio is not clearly explained.
L176 : a minimum of precision if required for the computation of the two landscape metrics with FRAGSTATS
L182 : “the window size is mainly constrained by ….”. It is not clear how it is constrainted.
L184-188 : this paragraph should be rewritten because it is not clear enough.
L193 : “…sufficient…” it is not rigorous
L196 : why 108 m for the window and not for instance 105 or 110 m ?
L201 :”….this is not the case for our study” Could you give some proves ?
L203 : Why in your study only two components were used ?
L260 : why the absolute values are expected to differ ?
L265 : replace “evidence” by “prove”.
L268-L277 : the paragraph has too many details and it is difficult to extract the main information
L291 : your hypothesis is not discussed in the following parts.
L313 : replace “BD” by “BS”.
L313 : in table 2, could you explain the reference data status ?
L329 : could you introduce the ranges of NDVI and especially why the NDVI are below zero ?
L645-647 : repetition.
Author Response
Please find our answers to your comments in the pdf file attached.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
I really appreciate your work and your thorough review, but I still see no way to solve the issue of the different year and phenological season, apart from acquiring scenes from the same year and date.
I did a short study on the phenology of your study area, to proof my concerns. Please have a look at the attached file.
I still do not reject the manuscript, but leave it to major revisions, because I am curious about your response. Perhaps you’ll find a way to solve it. I am very interested!
Comments for author File: Comments.pdf
Author Response
Please find the answer to the reviewer#2 in the pdf attached
Author Response File: Author Response.pdf
