Review Reports

Round 1

Reviewer 1 Report

Review Summary:

I enjoyed reading this paper entitled "Mapping winter crops using phenology algorithm, time-series Sentinel-2, and Landsat 7/8 images, and Google Earth Engine” with a focus on extracting phenological characteristics of winter crops from composite remotely sensed and reference data in Henan Province, China. Authors have put extensive efforts to retrieve Landsat and Sentinel-2 imageries in the GEE platform, utilized survey questionnaires, agrometeorological data, and field surveys in order to generate landcover types and cropland maps of 2019 and then generated some phenological characteristics to extract out only the winter crops and their spatial distribution. I can see the study’s implication in other study areas and is in the benefit for the agricultural industry.  I have no doubt that the authors’ exploitation of the power of GEE in processing data in this project has made this massive task possible in a manageable way, which otherwise would have been a massive data management issue. I also like the fact the authors’ have a section on uncertainty outlining the limitations which would serve as a guide for the future researchers. I think if the authors use the review comments provided below and specifically describe some of their methods and technical terms, the articles will come out in a very strong shape.

I have outlined my comments below in two sections: Major Comments and Specific Comments

Major comments:

1. The authors mentioned they used ordinary least squares regression coefficients to convert and unify ETM+ and MSI data from Landsat and Sentinel-2 but have not considered providing details of what those coefficients are like and how the images were combined. It was also not clear which band designations from ETM+ or OLI and Sentinel-2 belong to the labeled bands used to calculate the indices.
2. The authors used the daily average temperature curve in figure 4, however have not described its relevance in the project.
3. Authors should also describe the effect of the Savitzky-Golay filter in improving their results.
4. I have mentioned in the specific sections below, and I would like to emphasize that the authors stated they developed new model based on LSWI and EVI. I suggest the authors consider rationalizing their use of the new algorithm development.
5. In this paper, SOS, SDP, EOS etc. are referred as phenological indicators and those are the important basis to extract the winter crop. First, I would like the authors to present the statistics of those indicators from their data. Second, I would clearly differentiate the indicators and the DOY of those indicators separately on their paper. It is confusing to read the SOS and EOS threshold of 0.1 and 018, respectively, in line 338 and used DOYs elsewhere (equation 9 and figure 8) to represent the same indicators. I am not against using these terms, but I would like the authors to conform with the consistency in the field. Probably provide more references on the usage of these terms.
6. Conclusion: Authors dedicated much of the conclusion section to the power of GEE, I suggest that they focus their conclusion to what they actually achieve in extracting out the planting area of winter crops and what they actually learned throughout this project.

 Specific Comments:

Line 25: The acronym EOS does not make a short form for the spelled-out words (end of date)

Line 41: use of the word “grain” do not properly represent the crop types mentioned in line 39

Line 61-63: These two parts of the sentence “low and medium resolution” ….. “at medium spatial resolutions” are confusing. Clarify it.

Line 67-68: What does “Chinese household responsibility cropland” mean? clarify

Line 81: The term VI is out of context and does not offer any meaning

Line 81-82: the sentence is not complete.

Line 84: VIs- spell it out for the first time you use it

Line 87-92: the description of this second method uses a lot of undefined words and is not easy to follow.

Line 110: Why the letter A is capitalized?

117-118: Your overall point so far was remote sensing is beneficial in mapping winter crops, and now you are saying due to the limitation of remote sensing technology. These are not consistent.

Figure 1: Spell out the acronyms such as SR, LSWI, mNDWI, etc which are not spelled out before on your paper.

Figure 2: What does the background color on figure (a) add for your purpose of showing the location of the study site?

Line 161: Correctly spell out OLI (Operational Land Imager)

Line 166: What is TOA data and what is SR data? First clarify that.

Line 173: The reference citation 40 does not seem like a very relevant reference for the Fmask algorithm. It would be appropriate to find and cite the original author instead of an not-very -closely related paper.  

Line 173-176: Find the original paper of Fmask and cite that.

Line 186: Acronym OLI is used without first showing its equivalent spelled out form

Figure 3: The figure 3 and preceding paragraphs do not mention which specific images (dates and site) are shown in a and b. Due to the mismatch in orientation of the two images shown in a and b, readers do not get clear sense of whether the preprocessing required the application of  geometric transformation between Sentinel and Landsat or not.

Line 222: Rather than saying high spatial-resolution, specify the spatial resolution of these images

Line 233: what are those problems? Do those problems severely affect the results of accuracy estimation?

Line 240: assist or assistant?

Line 242 and formula 3: be consistent in writing the acronym MNDWI

Line 286-290: The authors stated that they developed a new model to classify land cover as forest and impervious but have not elaborated the basis of using those cutoff values of LSWI, EVI, and NDVI. That is critical to mention.

Line 326-327: Incomplete sentence - to determine what?

Line 329-330: We adopts? determines? (English grammar)

Line 351-352: sentence structure (missing comma?)

Figure 6: Change the wording of figures d) and e) to make them nouns. Clarify in which figure the satellite from the Maxar Technology is used

Figure 7: Specify which landcover types do a,b,c, and d represent on the main figure.

Lines 418-425: Rewrite the sentence possibly breaking it into more than one sentence. There is a lot of confusion about what it really means.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

The paper has interesting ideas about the use of Google Earth Engine for analysing Sentinel and Landsat data of winter crops in China. The results of the analysis are encouraging.

The English in the paper does need to be improved, mainly for the flow of the text. But there are some stylistic points that need attention:

1. Page 1 line 40 refers to “less contradictory” but it is not clear what that means.
2. On pages 2-5 there are several places where an acronym is used without being first explained, for example VI, NDVI, TOA and SR. The text needs to be checked so that every acronym is explained at first use.
3. Page 8 line 259. The equation seems to include #(5) but I think this is not in the equation itself. This needs checking.
4. Page 11 line 358. The line starting 260<SOS is very confusing and its meaning is not clear. It should be improved.
5. Section 2.3.5 line 367. The references to crops should be references to fields.

Overall the English needs improvement beyond the points listed here.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The article “Mapping winter crops using phenology algorithm, time-series Sentinel-2 and Landsat-7/8 images, and Google Earth Engine” proposes the use of a set of aerial images to monitor the phenological evolution of the vegetation cover. The theme presented is interesting and is replicable to other areas of the globe. In general it is well organized, but needs some improvements. 

Although the Introduction correctly presents the problem and adequately justifies the interest of this study, the authors must adopt a more synthetic language in this topic. In good English and in order to facilitate reading, the sentences should not be too long, as in lines 63-68; 94-99; 102-109. 

Taking into account the temporal evolution of the vegetation cover, I suggest that the following article be consulted: https://doi.org/10.3390/environments8050044 

In the topic Materials and Methods the ideas are structured and objective. However, it is necessary to improve the quality of Figure 1, in order to allow an adequate reading.
Several processed maps are presented here, while the new data is expected to be presented in the Results topic.

The Results are well presented and do not raise any relevant doubts. 

The Discussion should focus on analyzing and comparing the data obtained with other similar works. Ordinary statements should be eliminated eg. 487-489, as they do not add new information. 

The Conclusion is based on the results and focuses on the main ideas that should be highlighted.

However, authors must change the context of line 571. Instead of stating that there will be a trend towards the use of Google Earth Engine in the future, they should say that it is a tool with great potential for use in the future.

The Bibliography used in this article is adequate and current. 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

The paper is about the use of combined Landsat and Sentinel multispectral imagery and Google Earth Engine (GEE) platform to map winter crops.
The topic is interesting but the paper has some issue that should be solved before its publication.

GENERAL COMMENTS:
-The introduction section is unbalanced. The frst part (until line 115) should be synthesised. What's missing, however, is an adequate introduction to the GEE platform and some examples of its usage. Please consider the following papers to improve the introduction section:
    doi: 10.1016/j.rse.2017.06.031 - General paper about GEE,
    doi: 10.3390/rs10101509 - GEE Review,
    doi: 10.1016/j.isprsjprs.2020.04.001 - GEE Review,
    doi: 10.3390/rs13040586 - Time-series, pixel-based approach classification and seasonal aspects,
    doi: 10.3390/rs12223776 - Object-based approach classification,
    doi: 10.1016/j.isprsjprs.2017.07.011 - Lansat7/8 and Sentinel integration;
-The authors used both Lansat 7 and 8 data. Resolution and wavelength of ETM+ and OLI data are quite the same in VIS-NIR-SWIR regions and the investigated time is just one year (covered by both satellites). why did the authors choose to use data from both Lansdsat 7 and 8?
-An important limit is that only one year was chosen as reference to perform all the analyses. One of the advatages of GEE platform is that it allows long time-series to be processed easily. I suggest to expand the considerd time-series to at least 3 years in order to improve and test the proposed methods;
-It is not clear the role of the GEE platform. Was it used just to collect images or was it used also for the analysis using its script-based language? What has been done inside GEE and what with thirt-party software and applications?
-The proposed methodology should be better explained to any potential readers;
-The integration between two different multispectral platform (Landsat and Sentinel) should be emphasised in Materials and Methods section;
- Conclusions section should be improved.

SPECIFIC COMMENTS:
-line 25: EOS should be end of season and not end of date;
-line 172: Did the authors use Fmask algorithm inside GEE or with third-party software?
-lines 239-243: Please put references for each adopted index;
-lines 245-247: it would be more appropriate to refer to reflectance values for each band;
-line 259: Please formatting equation 5 correctly;
-line 269: Why did the authors calculate the mean temperature? It seems that this data is not taken into account for other analyses;
-line 270: Please move the results section;
-lines 319-323: The authors wrote that peak and valley were defined only comparing a value with previous one and next one. In my opinion this method it is not accurate (NDVI values could have a floating behaviour for reason not strictly linked with crops growth) and this methodology could detect false peaks or valleys. I suggest to order values in ascending and descending mode to detect maximum and minimum values to detect peak and valley;
-line 352: please change the point with a comma;
-equation 9: Please clarify this line. Does it represent a formula? If no, please remove the numeration and format it as text. It seems to represent a list of thresolds and not a specific formula. Moreover, it is not clear what numbers represent (days? indices values? other?);
-lines 368-369: Why did the authors collect a different number of ground reference data for different LC? It is not explained into the text and should be clarified to any potential readers;
-lines 373-374: The authors present in this paper a pixel-based approach and not an object-based one. Why did the authors use polygons (objects) data as reference for the accuracy evaluation process? They should use pixels as reference. In this way they could also avoid the process of conversion from polygons to raster avoiding possible errors;

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 4 Report

Dear Authors,

thank you for the effort to consider the comments.

All previous comments have been adequately addressed