Detection of Irrigated and Rainfed Crops in Temperate Areas Using Sentinel-1 and Sentinel-2 Time Series
Round 1
Reviewer 1 Report
Overall a sound paper, especially that it tackles the identification of irrigation areas in moderate climatic zones. Please consider the following points:
Rephrase first sentence:
The detection of irrigated areas by means of Remote Sensing is essential to improve agricultural water resource management.
line 52-56:
move this paragraph to the method section
line 53-54:
Please remove this statement or provide a reference with regards on a frequency analysis of different supervised classification
algorithms within the domain of Remote Sensing/Earth Observation.
line 54-55:
That statement is OK, however, usually the model selection is very dependent on the application target, calibration data and feature dataset.
line 63-64:
maybe rephrase: In order to make this discrimination, we relied on the phenological development of the vegetation cover as a explanatory variable.
Section 2:
Please provide a crop calendar figure.
line 79: It is highly doubtful, that sunflowers are not irrigated
line 99: remove double-dot
line 102-103: please rephrase:
The obtained images were temporally aggregated to regular 10-day interval composite datasets using...
line 103: please add more information, which statistical aggregated method did you use for the creation of the composite images (SUM,MAX,MEDIAN?)
Section 2.4:
The study area is 1500km², so we would have in total 23 spatial features for the rainfall data.
This is actually not sufficient, but the whole methodology would be applicable in a larger area.
line 122: rephrase: It includes 4 main steps
Section 3.1:
Please provide time series plots for each of the feature datasets, depicting the phenology of the crops (NDVI, VV, etc), precipitation data, stratified for irrigated vs. rainfed areas, etc.
3.1.3. Cumulative indices:
although a reasonable procedure, I am not sure, whether such a dimensionality reduction would decrease the information extraction,
as unique irrigation events (which would require a higher temporal granularity) may not be captured by such a generalized dataset.
Moreover, using cumulative indices for RS data is not suitable, as model transferability (to another year) may be difficult, therefore using monthly MAX or MEDIAN would be better. Could you please redo the analysis with the proposed features?
line 164-166:
the limitation makes sense. However you should have probably used stratified random sampling with the same number of samples per class.
3.2.: please include more information on the parameters of the RF algorithms (depth, max-features, number of trees, etc.)
Author Response
Thank you for taking the time to review the publication. All your comments have been taken into account in the new version of the document. The response to each of your comments is available below. You will also find a version of the paper with coloured corrections and additions. Corrections/additions are in blue and deletions are in red.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript "Detection of irrigated crops in temperate areas with Sentinel-1 and Sentinel-2 time series" proposes a new methodology for distinguishing irrigated and rainfed crops at plot scale in temperate areas using the joint use of optical, radar, and meteorological data. It has important theoretical and practical significance and maybe of international interest. The authors demonstrated a good knowledge of the problematic and of the related literature.
The “Discussion“ section is not well structured.
The authors have to discuss the results and how they can be interpreted in perspective of previous studies.
Author Response
Thank you for taking the time to review the publication. All your comments have been taken into account in the new version of the document. The response to each of your comments is available below. You will also find a version of the paper with coloured corrections and additions. Corrections/additions are in blue and deletions are in red.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors of the manuscript entitled “Detection of irrigated crops in temperate areas with Sentinel-1 and Sentinel-2 time series” introduce a novel approach for mapping irrigated crops in the temperate area of south-western France using optical, radar and meteorological time series. In particular, they employed the Random Forest classifier and evaluated five different scenarios in order to examine the resulting classification accuracy derived from either the individual or synergistic use of the aforementioned data.
This is a very well-written manuscript with good use of English. The introduction section includes all significant related information, sufficiently highlighting the importance of the topic and the novelty of this work. All parts of the methodology are described in a comprehensive manner and the results are properly discussed.
Nevertheless, since a number of typographical errors can be found, I strongly recommend a thorough proofreading of the entire manuscript. Some additional issues include the following:
- Section 2.2: Please include some general information about the MAJA processor (preferably 1-2 sentences).
- Section 2.2: Please provide the full form of the “TOC” abbreviation as well.
- Figure 2: I recommend attributing different colors to the dashed lines of the cumulative indices’ and respective datasets. According to the current presentation, it seems like all the indices were simultaneously derived from both datasets. Please make the appropriate adjustments, so that the dashed line’s color of each grey box (i.e. Sentinel 1 or Sentinel 2) matches the one of the corresponding yellow boxes (i.e. derived indices).
- L.266: As illustrated in the graphs of Figure 3, the Kappa coefficient of the first two scenarios ranged between 0.3 and 0.49. Please modify.
- The Kappa coefficient computed for all classification results is relatively low given that a Kappa value of at least 0.7 is considered as acceptable. Please comment these results in the discussion section and provide possible reasons.
Author Response
Thank you for taking the time to review the publication. All your comments have been taken into account in the new version of the document. The response to each of your comments is available below. You will also find a version of the paper with coloured corrections and additions. Corrections/additions are in blue and deletions are in red.
Author Response File: Author Response.pdf
Reviewer 4 Report
General comments
I have been invited to review this paper by Pageot et al. The paper aimed to detect and distinguish between different irrigated and rainfed crops in a temperate climate area in Southern France using a combination of Sentinel-1 and Sentinel-2 data along with meteorological data. The paper used a Random Forest classifier with good validation evaluation. The main objective of the paper is clear enough. Generally, the manuscript is organized with interesting findings. Having said that, in my opinion, the discussion section needs to be re-written in a better manner. Furthermore, the paper has an issue with referencing system, in which the numbers are not showing correctly according to their actual place within the text, therefore, will need to be redone. The paper is too short to be considered as an article with less than 13 pages (excluding the refs), according to the journal’s instructions, it should be at least 18 pages. While the study presents an interesting topic with interesting findings and I appreciate the time spent in processing and classifying the used images, I still have several concerns regarding this study. Please see my specific comments below:
Specific Comments
Title
Considering the remaining of the paper, I think the article’s title should be something like “Detection of irrigated and rainfed crops in temperate areas using Sentinel-1 and Sentinel-2 time series”
Abstract
Line 3: What the authors mean by S1 and S2? If it is Sentinel-1 and Sentinel-2, so, abbreviations should be written in full when first mentioned.
Introduction
Line 24: What is this weblink for? Also, it should be cited in the references list.
Line 31: The refs numbers should have been [6-38]. Anyway, I do not think citing 32 refs or so for one sentence is a good idea!!
Line 32: Again here, citing 16 refs for one sentence is a bit too much! In my opinion, citing 3-4 refs for one sentence (fact) is an overkill!
Lines 31-33: I agree with the authors that most of these studies were conducted in semi-arid regions. However, I strongly disagree when they say that there are very little studies conducted in temperate areas. There are several studies out there which discuss mapping irrigated agricultural areas using different remote sensing techniques in temperate/Mediterranean areas.
Line 37: Ref(s) needed.
Line 41: I think “can” is missing before “improve”.
Line 42: (the Sentinel-2 time series). Replace “the” with “a”.
Line 44: Replace “on” with “in”.
Study site and dataset
Line 71: What is the name of your study site? Within a village/town/city?
Line 71: (In the figure 1). Remove “the”.
Lines 72-73: Ref(s) needed.
Lines 78-82: Ref(s) needed.
Figure 1: It is probably okay, however, what is the dark red shape on the top right corner of the map? Is it France? Why not add it to the legend?
Lines 87-88: Why did not the authors sample the same plots with the same total number for 2017 and 2018? Why did they use a different number of samples for 2017 and 2018? How did the authors choose the location of each plot? Was it just random?
Line 88: (in the table 1). Remove “the”.
Line 92: Replace “on” with “in”.
Line 97: Sentinel-2 images: The sub-title should be “Sentinel-2”. What are the tiles numbers of the images obtained? I am aware that they have been mentioned in figure 1, however, it must be mentioned within the text as well. Furthermore, why not adding the weblink you used to download these images?
Line 106: Sentinel-1 images: The sub-title should be “Sentinel-1”. How many images were used?
Line 107: Should be “Sentinel-1”.
Methods
What about the applicability of this method in a much larger area or with denser and several types of crops? 71-78% of overall accuracy achieved in a 1,500 km2 area is not exceptional and this implies that applying this method on a larger scale area would result in a much less overall accuracy!
Line 130: (in the table 2). Remove “the”.
Figure 2: The caption needs to be changed to something like “The workflow of the methodology used in this study”.
Table 2: Within the caption, “with the equation” could be removed.
Lines 150-151: I think the authors are gambling with this assumption. What about the morphological/physiological variations between different crops? What about the variations in the water requirements needed by each crop?
Line 180: (in the table 3). Remove “the”.
Results
Line 216: What the authors mean by “global performance”? General or overall, maybe?
Line 223: Which scenario is the best overall in your opinion? The optical SAR rainfall scenario?
Figure 3: Within the caption, should be “correspond to”.
Line 233: “Except for”.
Discussion
The discussion section is probably okay, however, I think the authors should also discuss their findings with the literature’s.
Conclusions
Conclusions not conclusion.
References
Line 342: I do not think that ref 1 is correctly cited. Is this FAO perspective really published in Land Use Policy?
Author Response
Thank you for taking the time to review the publication. All your comments have been taken into account in the new version of the document. The response to each of your comments is available below. You will also find a version of the paper with coloured corrections and additions. Corrections/additions are in blue and deletions are in red.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
No further comments! Thank you for considering the changes
Reviewer 2 Report
The authors have successfully addressed reviewers' comments. Overall, the quality of the manuscript is significantly improved.
I don't have any comments.
Reviewer 4 Report
I would like to thank the authors for the efforts put to improve the manuscript. I can confirm that the current iteration of the paper hs improved over the original one. I was expecting the authors to include thanks to the anonymous reviewers who contributed with comments and suggestions to improve the paper as a matter of courtesy. Furthermore, I think it would be best if the conclusions section could be more concise.
