Monitoring Harvesting by Time Series of Sentinel-1 SAR Data

Round 1

Reviewer 1 Report

Summary

The article proposes an effective algorithm which can be used for detection of harvesting events by Sentinel-1 SAR satellites.

 

Broad comments

The study was performed on a high technical and scientific level.

The authors provide all the information about the methodology they used and data processing steps they took in a clear manner.

Overall, the article is written well, but some editing is needed.

 

Specific comments

Despite the overall high level of the study, there are the following comments.

 

Sentinel-1 acquires images during ascending and descending orbits. Authors used images acquired with the time interval of 12 days which means that all the images belong to only one type of the orbit. If so, why authors did not try to use images belong to another overpass? The most probable reason is the problem of incidence angle normalization, but authors should describe explicitly what orbit was used and why. SNAP is a constantly evolving software. Please specify the version. The Results section is very brief.

Author Response

Dear Reviewer!

Thank you for your attention to the article and valuable comments. The authors have carefully reviewed the comments and have revised the manuscript accordingly.

Please see the authors’ point-by-point response in attachment.

Kind regards,
O. Kavats, D. Khramov, K. Sergieieva, and V. Vasyliev

Author Response File: Author Response.pdf

Reviewer 2 Report

General remarks

All in all, an interesting study with given relevance. Needs improvement regarding its presentation and documentation to help the reader acknowledge its benefits over other approaches.

Language: I recommend language revisions by a native speaker to eliminate the errors regarding style and grammar, especially in the first part of the manuscript. Unfortunately, the current quality of the language does meet the current standards of “Remote Sensing”.  

Abstract: The abstract does not state a research hypothesis or present a clear aim of the study. Why is it important to know the harvesting date? This should be made clear here as well to underline the relevance of your study and already target a specific audience.

Abbreviations: To cite the "Instructions for Authors" of MDPI: "Abbreviations should be defined in parentheses the first time they appear in the abstract, main text, and figure or table captions". Please add them in the text accordingly. Having a list of abbreviations in the end is fine, but I think it's easier for the reader if you sort it alphabetically and not by occurrence in the manuscript.

Framework: The introduction is too vague and misses a clear definition of the problem (detailed comments below). Your literature review is too short. Please clearly outline what has been done in this field (SAR and agriculture) and to what extent you are offering something new (detailed comments below).

Structure: The different parts are not clearly separated yet. Please try to follow a clear structure: Problem and setting, aims of your study, data and methods, results, discussions. Currently, these switch too often which makes it hard for the reader to follow your approach. Sections 1.3-1.6 do already contain some parts of the method (as far a I understand it). 

Methods: The description of your workflow is not transparent enough. The reader should be able to reproduce it with own data. This does not necessarily require the detailed description of each step, but it should be clear how exactly you used the various data sources (optical, SAR, IMEG) came to the presented results (detailed comments below). As suggested below, a flow chart with includes all the input data required to implement your suggested workflow and how it is combined to achieve the results would be great in terms of understanding and transparency. Figure 5 is an attempt but should include the data sources and how they are combined at which stage. To give examples, please have a look at
Figure 2 of this paper: https://www.mdpi.com/2072-4292/11/17/2047 or
Figure 3 of this paper: https://www.mdpi.com/2072-4292/11/6/648.

Credibility: The validation of your approach needs more transparence. The method seems feasible and there is a pattern observable in Figure 7. But it would be good to demonstrate how you measure the accuracy. For example, how many of the areas were tested and how this leads to the presented MAE and the RMSE. Would it make sense to cross-check with time-series of PlanetScope, https://www.planet.com/explorer/) to increase the number of available reference images?

Balance: The share of results and discussion in the paper is quite small. Despite the suggestions above on things to include in the introduction and methods chapters, there is still room for more content, results and discussion in chapters 3. and 4 (detailed comments below). Also, some discussion about the accuracy assessment would be nice. Are the validation points sufficient in terms of numbers and also representative for the entire study area?

Outreach: It is not clear enough what are your main findings and for whom they are relevant. You did a good job deriving the harvesting dates, but the presentation could be more confident and clearer addressing potential readers and users of your paper.

 

 

Detailed remarks

1 Earth: lower case spelling

2 please spell out Sigma0 (no unexplained variables in the abstract)

6 registers: quite unfortunate terminology, please consider using a clearer verb here

7 and is related: This part does not refer to the subject of the sentence (this study). Please rephrase.

10 SLC IW: abbreviations in the abstract should be avoided. As IW is the most common imaging mode, I think simply "Sentinel-1 data" be sufficient here.

12 can be deemed satisfactory: Please be precise here and use selected numbers. People which are only reading the abstract should be able to determine the accuracy as well.

17 The overall context of the study (precision agriculture?) is not presented very clear in the introduction. Why do you mention GPS measurements, UAV, or MODIS data here when your study is about using SAR? If SAR is used as an alternative to these methods, please make it clearer. Why should radar be used instead of optical data? Please try to make clear who is going to use your result and for what purpose. Measuring agricultural parameters (spatially and temporally) is fine, but it is important to guide the reader through a story: Why is your study important, what exactly is the problem, and how are you going to address it?

18 machinery: you use this word three times in a row. If it is really relevant to your study, it should be made clearer what is meant. How does it relate to what you are going to present?

18 any required parameters: Please specify.

36 volume of vegetative biomass: This is the first crucial parameter which really relates to high resolution SAR. Please explain how backscatter intensity and volume of agricultural fields are related and name references.

39 A number of methods: If you say 'a number of' you have to give a bit more examples. Please remember that in this part of the paper you discuss the state of the art of the field of study, where there are research gaps and open questions and which of these gaps are addressed by your approach. What is new about it and which similar studies exist?

44 The problem of clouds is why your study is relevant. Make this clearer and at an earlier stage of the paper.

49 Earth's surface: lower case letter

52 Combined interaction of the: Consider removing "combined" and please rephrase the sentence to make clear that backs5catter intensity is influenced by surface characteristics (not vice versa).

56 In the majority of the published works...: The two references [5,6] do not represent studies working with backscatter intensity to describe agricultural phenomena very well. Please carefully select significant publications in this field. What about polarimetric approaches? Give a more extensive review on what is already done. This helps you to argue why having a look at coherence might be an interesting idea.

57: Backscatter coefficient σ0: Please add an explanation what Sigma0 really stands for and add a suitable reference. The one with the water cloud model [7] is not ideal in my opinion.

58 Here would be a good point to relate time-series of backscatter intensity over agricultural areas to the determination of harvesting dates. How is this solved in other studies (without coherence)?

73 the use of coherence is described: Give examples how it is used in the mentioned studies. What are their findings and where are the challenges? As this is your main point of analysis, these points should be shortly presented here.

76 These are the elaborations on coherence which I would have expected earlier. Once you clarified how it interacts with agricultural periods, you can refer to the studies, what they investigated and how.

82 AOI: This is not a suitable title for a chapter. How about "Study area"?

83 Tell a bit about the study area. How is it used in general, what is the predominant land-use and landcover, what is its meaning for Kazakhstan?

83 sample areas: Please explain how the spatial distribution of the sample areas was determined and what is done here. Was any field work conducted here, by whom, when, how? This paragraph leaves too many questions open.

84 harvesting dates: Please specify how you determined the date of harvesting. As Sentinel-2 and Landsat-8 do not deliver images at a daily base: What is the temporal accuracy that can be expected here?

86 satellite data: A table listing all data sources would help the reader to understand.

87 SLC IW: It would make more sense to me to first name the swath and then the slice number.

90 Harvesting: In 1.2 you say the harvesting date was determined by visual analysis. Here you mention NDVI calculations based on MODIS data. This is confusing the reader who is not familiar with your study design. Consider creating a flow chart of your analysis: What data you used, what is done with it. Figure 2 is a good start, but it should encompass the whole study: How optical data were used for training and when they were brought together with the SAR data.

96 precipitation: Again, too vaguely explained. What kind of data is it and how do you use it?

99 Figure 2: Was flat-earth-phase subtracted for calculation of coherence? Was topographic phase removed? The results in SNAP can be considerably different.

99 SNAP: Please include the version number in the reference and maybe an URL.

99 GPT: not listed in the abbreviations

99 bash scripts: It is welcomed by the journal to share the scripts of pre-processing as online supplementary.

104 marked as dubious: What does this mean? Generally, you switch too often between the data sets used in this study, which makes it hard to follow. Can this be aggregated so parts of the processing belonging to one dataset are written within one paragraph? It is still not clear what the IMERG data is exactly used for.

105 Peculiarities of the AOIs: Why don't you put all this in 1.2? I see no need in splitting up the location of the study area and the description.

109 Figure 3/4: The colors are unfortunate. It is hard to interpret the image without prior knowledge of the study area. Basic map elements are missing. At least one scale bar is required to get an idea about proportion: Additionally, north arrow and a coordinate frame would be good. Instead including the last two points, you can put a frame of the extent of Figure 3/4 into the map of Figure 1 and refer to it. This helps the reader putting the image in a context.

What is the yellow lines represent? This should be clear before the image comes up in the manuscript or at least be named in the figure caption.

110 buffer zone around each area: What does "area" here refer to? Each of the yellow points of Figure 1? Please clarify.

112: at least 30 points: Can this be specified? How many points were used in total in the end? Number of areas x number of points?

113: Figure 3 and 4: The IDs do not make sense to me, what do they refer to?

116: Here you name the use of the NDVI. As suggested earlier, it would be easier to follow if these things would be described more aggregated in the manuscript so that things which belong together are explained within one section. I think the general description of the methodology could be more linear and straightforward.

123: Is there a reason why you only used VH and, for example, not the ratio of VH/VV to enhance the cross-polarized information normalized against co-polarized?

128 Figure 5: I suggest putting the noun at the beginning: Selection of control points, detection of observation start points.

146 Would it make sense to draw a red line for the harvesting dates. It is included in the caption but would be easier for interpretation if it was somehow marked in the time-series as well.

146 Can you also show a representative time-series of the HV backscatter so it could be compared over time with the coherence time-series. I think this is something very interesting to the readers.

146 Soybean and sunflower: This is the only time these words occur in the manuscript - are these crops relevant in your study area in terms of occurrence, financial or ecological aspects? Why did you select it as an example? Generally, as mentioned earlier: A description on the land-use of your study area is important to know what is grown here, maybe a simple table with shares of different crops to give the reader an idea why your work is important here.

146 Please include a coherence image of a harvested and a non harvested field to demonstrate the decrease of coherence and to proof that what you are measuring at the points is not just coincidence. This also helps to assess the robustness of your approach and the reader might be interested to compare your coherence maps to his own.

149 Please also number this equation.

150 Can you also make a graph about your ∆Ci calculations. I think this is an interesting index, but it is not very evident how it can be used. Maybe you find a way to make it more visual. I think such graphs could as well be put into the Results section, because they were derived within your work and present new insights on the temporal development of backscatter and coherence over time for specific crops. To my opinion, these are the type of contents which are expected by a typical reader of "Remote Sensing" under such a title.

155-167: I like the concept of stepwise determination of harvesting from the time-series. Could you make a simple decision-tree which illustrates the checks and thresholds in a graphical form?

178 As suggested above, it would be nice to know how MAE and RMSE were calculated.

179 Figure 7: Please include a scale bar for Fyodorov district.

180 The discussion section is a bit short. Besides the thematical aspects I miss a critical discussion of the research design itself, the data sources used and how feasible it is to retrieve similar results for other parts of the world. Readers are interested if your approach would help them for their own research, so anything you can say about potential error sources and recommendations are welcome.

181-185 Would it be possible to increase the accuracy by the integration of repeat-pass data from different tracks? This could be added to the discussion, e.g. what is the shortest interval achievable in your study area if image pairs of different tracks would be used.

186-193 What mechanisms could be integrated to eliminate false alarms? Even if they are not implemented in your approach, it would be fair to discuss them shortly or to propose ideas for future studies.

201 An increase in frequency: Can you give an example (e.g. in days) what would be needed for higher accuracy? You can discuss this against the revisit rates of th upcoming new satellite missions (CSK second generation, Radarsat constellation mission, ICEYE, …).

214 I really like your conclusions because it straightly leads through your whole idea and findings within a short time and breaks down the complexity of the approach to a well understandable level. If you could be similarly precise and clear in the abstract and introduction, the reader would get a better image about your idea earlier.

Author Response

Dear Reviewer!

Thank you for your attention to the article and valuable comments. The authors have carefully reviewed the comments and have revised the manuscript accordingly.

Please see the authors’ point-by-point response in attachment.

Kind regards,
O. Kavats, D. Khramov, K. Sergieieva, and V. Vasyliev

Author Response File: Author Response.pdf

Reviewer 3 Report

The main point related to the discussion there is no link between the sigma not with erosion or/and crop management, in deep. The discussion came from the indirect data (satellite images), not from the field. For me the discussion became weak.

Author Response

Dear Reviewer!

Thank you for your attention to the article and valuable comments. The authors have carefully reviewed the comments and have revised the manuscript accordingly.

Please see the authors’ point-by-point response in attachment.

Kind regards,
O. Kavats, D. Khramov, K. Sergieieva, and V. Vasyliev

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I appreciate your efforts to implement the suggestions. I really think the quality of the paper has improved substantially. I still have a few smaller points/suggestions to be considered before publication. I again marked it as "major revisions", so the editors will grant you enough time to go through the points, especially as you mentioned in your response that you didn't have enough time to implement all comments.

Again, I believe that this manuscript is getting close to the standards of Remote Sensing, and if the points below are considered respectively, I will recommend it for publication.

 

In case I suggest rephrasing, I highlighted new or crucial words by _underscores_.

 

21 Shouldn't it be "are high costs"?

24 "are particularly cost advantageous" How about "are attractive to farmers with low budgets in particular"?

27 "after harvesting there appears a striped texture" How about "that after harvesting a striped texture appears"?

36/37 "According to the [3]": If you use a reference mid-sentence please write down the name of the author followed by the reference: "According to Sood et al. (2015)[3],..." to preserve a correct syntax of the sentence. This also applies to other points in the manuscript, e.g. [19].

43 "However, this approach is useless for establishing the fact of harvesting." I think I understand what you want to say, but maybe you can rephrase it to make it clearer.

46 "images is concerned with the problem that is inevitable for optical sensors - cloudiness." How about "...is inevitably linked to limitations of cloud cover"?

48 I would prefer "spaceborne" as it is more common.

51 Maybe it is worth to mention here that independency from cloud cover allows reliable image acquisitions at regular intervals which is decisive for agricultural applications.

73 "Valuable information about the vegetation state can be obtained from coherence [14]." I would put this sentence at the beginning of the next paragraph to introduce it as an additional factor.

87 "The present paper aims to develop results of the [19] research" Maybe you can shortly explain what points in the referenced approach you want to improve with this study.

88 "The idea of the method" Consider adding one word for clarification "The idea of the _proposed_ method".

Your introduction has significantly improved. The reader is now guided along your research topic, the hypotheses it is based on and what you want to achieve. Well revised. A few words on what is actually new in your study could be presented more confidently.

100 "were under wheat" Maybe you rephrase it to "cultivated with wheat". F

107 "According to the crop calendar" I think this sentence does not need to be an own paragraph, you could move it at the end of the previous paragraph. The reader now gets a better understanding of your study area and its use.

109 Figure 2 Is "field boundary" the correct term to describe it? If I understood correctly, the cultivated area could be larger but is interrupted by soil erosion and lea areas. If this describes it more correctly, consider renaming the green areas to "cultivated areas" or something similar. I like the scale bar which helps to put the image into a context.

111 This section (Harvesting dates) has improved, thank you. Maybe you could add "Harvesting dates in the area were _initially_ determined by..."

119-121 Please use whole sentences to describe the data.

130 I appreciate that you included Figure 5 with the workflow of the study, well done. It is worth mentioning this figure earlier, for example a tht end of section 2.3 "A structural diagram presenting all data sources and work flows of this approach will be presented in chapter 3 (Figure 5)" so that the reader knows that it will be clarified at a later point.

134 I would have preferred if the flat-earth phase was removed because often the coherence image becomes clearer. But it is your choice and I leave it up to you.

139 "marked as dubious" I now understand what you mean but maybe you can phrase it more objectively, e.g "were considered unusable" or simple "were replaced by..."

146 Consider "At least 30 such points were chosen for _each field_".

149 Please remove the coma after "then"

154 Thank you for adding the error measures and what they are based on. I would prefer to directly present the MAE and RMSE values (from line 180) at this stage. However, it would be nice to refer to it at later stage and shortly discuss what 6.5 days and 8 days mean in practice. Is this good enough?

160 "Vertical transmit and horizontal receive". Please stick to lower case spelling of polarizations.

167 Figure 5 As I said, very well done, this really makes your paper more attractive. Maybe you could mark the parts of the workflow which serve to train, calibrate and validate of the approach (e.g. control determination of harvest date) with a grey background to clearly separate them from the actual analysis of SAR backscatter and coherence. Alternatively, you give short labels to the different boxes, such as "control", "identification of...", "extraction of...", "validation" to guide the reader through the figure.

186 I am aware of the short revision time provided by MDPI in general. Please still try to include an image pair of radar coherence (maybe overlaid by the field boundaries) dropping during the harvesting process. As you have processed all these data, a two-image figure with a before and after situation should be possible. I will ask the editors to grant you enough time to include it.

188 Thank you for inserting the harvesting date. Please mention it in the figure caption. I like the new graphs showing dC over time. Interesting!

188 As you did to me, please also explain to the reader why you selected Soybean and sunflower here. Your argumentation is valid and worth to be included in the text.

200 please change to "(_equation_ 4)"

198-214 You nicely explain how the assumptions, equations and decisions which you define lead to the characterization of coherence and the determination of harvesting/no-harvesting. Would it be possible to express this in a graph as well (as it is not covered by Figure 5) at the end of this section? Examples:

http//www.geo.cornell.edu/geology/SIRC_Pat/zones_tree.gif

http//docsdrive.com/images/ansinet/itj/2009/fig3-2k8-5841.gif

https//journals.plos.org/plosone/article/figure?id=10.1371/journal.pone.0080989.g004

222 Here you call it "class other". It would be good to mention this definition in the methods description and how you identified these areas. 

227 "The harvesting completion dates were computed..." As this refers to the determination of the threshold ɛ used in the previous section, shouldn't it be part of the section "3. Proposed algorithm"?

231-278 I like the discussion section very much. You mention which points contributed to the results and how strong their impact on the quality could be.

 

Again, congratulations to the substantial improvements on this interesting article. It is worth including the points above to make it a publication of high quality.

Author Response

Dear Reviewer!

The authors would like to thank you for the careful and thorough reading of the manuscript and providing your valuable comments. We have carefully reviewed the comments and have revised the manuscript accordingly.

Please see the authors’ point-by-point response in attachment.

Sincerely,
O. Kavats, D. Khramov, K. Sergieieva, and V. Vasyliev

Author Response File: Author Response.docx

Reviewer 3 Report

The text was improved a lot in the beginning but in the Item Results/Discussion the text continue poorly discussed. There no related with the literature, which is vast. The contribution would be great if this relationship was done. 

The authors describe in the Introduction lines 75 to 93 about the coerence and sigma not, responses related to some variables but in the discussion does not any results from those literatures were brought (at least). But other works from the literature could have been inserted.

Author Response

Dear Reviewer!

The authors would like to thank you for the careful and thorough reading of the manuscript and providing your valuable comments. We have carefully reviewed the comments and have revised the manuscript accordingly.

Please see the authors’ point-by-point response in attachment.

Sincerely,
O. Kavats, D. Khramov, K. Sergieieva, and V. Vasyliev

Author Response File: Author Response.docx