Sentinel-2 Enables Nationwide Monitoring of Single Area Payment Scheme and Greening Agricultural Subsidies in Hungary
Round 1
Reviewer 1 Report
Dear Authors,
I read the manuscript with interest. In my opinion, it raises an interesting problem and its detailed solution. The undoubted advantage of the article is the practical application of Sentinel-2 photos to the Single Area Payment Scheme and Greening agricultural subsidies. It is worth publishing after minor revision. The article has already been reviewed once and thoroughly revised. My comments are as follows. The YCI index, as newly introduced, should be described in more detail. In Table 1, in the Average area column, there are values within a given range, such as 5.8 ± 8.4. Does this mean Average area can be negative? I would suggest rather to write 0-14.2.
Yours faithfully,
Reviewer
Author Response
Dear Editors and Reviewers,
many thanks first of all for your interest and positive comments, also for the rapid review. Please find attached a resubmission of the manuscript with your comments addressed, also below our responses to your suggestions.
Reviewer 1
-The YCI index, as newly introduced, should be described in more detail. --> we added some discussion to the description of this index as requested
- In Table 1, in the Average area column, there are values within a given range, such as 5.8 ± 8.4. Does this mean Average area can be negative? I would suggest rather to write 0-14.2. --> we modified Table 1 introducing the standard deviations in a separate column to avoid confusion, to avoid misleading the reader.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The manuscript is well written, structured and detailed, showing that it corresponde to a soundly carried out piece of work. The manuscript could be published considering the minor comments and suggestions that follow
Abstract
Line 11-12
· Sentinel-2 satellite 11 imagery are a promising data source
o Sentinel-2 satellite 11 imagery is a promising data source
Line 21
· (Overall Accuracy 88%)
o (overall accuracy 88%)
Line 22
· Does visual interpretation mean visiting the fields? Please explain
Line 22-23
· What is the main limitation of the size of fields? Too big? too small? Please do indicate it
Line 25-26
· What are the expected further improvements from integration with Sentinel-1? Please explain
Keywords
Line 28
· “national scale” could either be included in “Agriculture monitoring at national scale” or in “Common Agricultural Policy at national scale”
Introduction
The introduction is complete, clarifying and very well written.
Line 46-47
· the first sentence of the paragraph has just been mentioned in the previous paragraph (Line 41-42)
Line 74
· Please indicate the meaning of the acronym “convolutional neural network (CNN)”
Line 138-139
· by the Puglia region of Italy;
o by the Puglia region of Italy,
Line 157
· Please indicate the meaning of the acronym “Integrated Administration and Control System (IACS)”
Materials and Methods
Section 2 in Materials and Methods is very well organised and structured, perfectly giving an account of the study area, tasks to be performed with the remote sensing data, including a flowchart of the data analysis workflow. Reference farmers’ data and satellite data are also explained, as well as data preprocessing methods and the Random Forest classification method. The study is shown to be very robust by considering the extended number of vegetation classes considered. The evaluation of the classification accuracy is also explained in detail. And the methods for detection of agricultural practices and event dates are very well explained. Similarly, the authors provide a set of rules developed in the form of decision tree to be applied to the previously defined monitoring tasks.
Line 287
· Refer to and explain Figure 2 in the text, indicating the meaning of the different acronyms used
Line 359-360
· Equation 1 gives a significant index YCI. Please explain why that specific band combination achieves the meaning of the index. Adding complementary spectral diagram will also help
Results and Discussion
Results were also very well analysed and interpreted and the large number of monitoring operations, more that 2 million, were associated to the different monitoring tasks. The authors evaluate the distribution of these operations in terms of eligible, uncertain and ineligible. The discussion and conclusion sections are also well written, synthesising and giving detailed explanation of the work done and the results obtained
Author Response
Dear Editors and Reviewers,
many thanks first of all for your interest and positive comments, also for the rapid review. Please find attached a resubmission of the manuscript with your comments addressed, also below our responses to your suggestions.
Reviewer 2
Abstract
Line 11-12?: "Sentinel-2 satellite 11 imagery are a promising data source" - suggestion: "Sentinel-2 satellite 11 imagery is a promising data source" --> corrected according to request
Line 21:("Overall Accuracy 88%)" - suggestion: "(overall accuracy 88%)" --> corrected as requested
Line 22: Does visual interpretation mean visiting the fields? Please explain --> brief explanation ("on-screen visual interpretation") added as requested
Line 22-23: What is the main limitation of the size of fields? Too big? too small? Please do indicate it --> "The main limitation was the size of fields which were frequently small compared to the spatial resolution of the images" explanatory half-sentence added as requested
Line 25-26: What are the expected further improvements from integration with Sentinel-1? Please explain --> since this is the abstract, room for explanation is rather limited, but we extended this sentence slightly ("Based on these results we find that operational satellite-based monitoring is feasible for Hungary, and expect further improvements of classification accuracy from integration with Sentinel-1 due to additional temporal resolution.). We are also considering to remove the second half of the sentence that discusses the expected future improvements.
Keywords
Line 28: “national scale” could either be included in “Agriculture monitoring at national scale” or in “Common Agricultural Policy at national scale” --> we see the point of the reveiwer, since the scale of a study is rarely included in the keywords. Nevertheless, we propose to keep the keyword “national scale” in the current form as keywords are mainly intended as search terms. We believe our results are relevant both for sub-national scale agriculture monitoring and for national scale studies of other applications. If this is the case, "national scale" would be justified as a keyword on its own. A quick search of the keywords "national scale" in the archives of MDPI Remote Sensing provided 10 resulting papers.
Introduction
Line 46-47: the first sentence of the paragraph has just been mentioned in the previous paragraph (Line 41-42) --> this was corrected as requested by shortening lines 41-42 to avoid the repetition
Line 74: Please indicate the meaning of the acronym “convolutional neural network (CNN)” --> resolved as requested, also some NDVI in the introduction
line 69 was resolved.
Line 138-139: "by the Puglia region of Italy;" suggested correction "by the Puglia region of Italy," --> corrected as requested, with slightly different formulation to emphasize the exact time of each introduction: "After an early introduction of the monitoring checks in 2018 by the Puglia region of Italy, in 2019 Denmark, Malta, Flanders (Belgium) and various provinces of Spain and Italy communicated their intention to the EC to abandon traditional checks for Sentinel-based remote sensing "
Line 157: Please indicate the meaning of the acronym “Integrated Administration and Control System (IACS)” --> corrected as requested
Materials and Methods
Line 287: Refer to and explain Figure 2 in the text, indicating the meaning of the different acronyms used --> acronyms not used before in the text were now resolved in the figure caption, references to the figure added to each relevant paragraph.
Line 359-360: Equation 1 gives a significant index YCI. Please explain why that specific band combination achieves the meaning of the index. Adding complementary spectral diagram will also help --> more details on this index were also requested by reviewer 1. We added a brief description and a citation of a similar but not identical index, but in order to save space and not to hinder readability, we suggest to refrain from adding a separate figure.
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
The authors choose a very interesting and current theme. Undoubtedly the use of satellite data is a promising data source for agricultural verification and monitoring. For obvious reasons, they've made a pilot study based on Hungarian satellite data. From my point of view, the paper is well-written, the figures have good quality, and are well described in the text. The methodology used is coherent and produced good results. In the end, they obtained a crop classification with an overall accuracy of 88%. I agree with them that the main limitation found was the size of fields: more than 4% of the parcels had to be excluded. For sure they can focus their research next steps on this issue.
2.10: added comments
Undoubtedly the use of satellite data is a promising data source for agricultural verification and monitoring, although many companies believe that the use of aerial robots can reduce costs, increase image quality, and extract better and latest data. I believe that the use of satellite data can provide larger area data that can be faster processed (when compared with the time necessary to acquire drones' data from the same area). In case someone frequently gets good quality satellite images of an area, this can be a feasible solution not only for classifying crops, but also to detect the presence of diseases and plagues. Taking this into account, the use of satellite images is for sure not innovative, but the authors presented a good study case applied to their home country (for obvious reasons, they've made a pilot study based on satellite data of Hungarian territory). Although the authors focused their application on Hungary, the theme is interesting to the readers and to the field of knowledge. From my perspective, the methodology used is coherent, produced significant results, and can be applied to any country's data (this makes it interesting to the readers). In the end, they obtained a crop classification with an overall accuracy of 88%. I agree with them that their methodology still has limitations. For sure they can focus their research next steps on these limitations (maybe by combining satellite images with other image sources?). It would be nice if the authors could add more comments on this topic.
Reviewer 2 Report
(Remarks to the Author):
The authors developed a random forest classification method and applied it on Sentinel-2 for agriculture monitoring.
Overall, this research is very interesting and related to the aim of this journal. Moreover, the language and structure of this paper are good. I would like to recommend this paper for acceptance in the present form.
Here are my comments:
- Add a legend of the upper figure of figure 1
- It is better to add a flow chart of the workflow
- You can highlight all tasks # in the paper
2.10 added comments:
Please check my comments below. Thanks.
1. What is the main question addressed by the research? The authors developed a random forest classification method and applied it on Sentinel-2 for agriculture monitoring.
2. Do you consider the topic original or relevant in the field, and if
so, why?
Yes, this author applied a machine learning-based classifier (random forest) on remote sensing images. This paper proves that this method has the ability to distinguish the crops in the field.
3. What does it add to the subject area compared with other published
material? The method they developed is novel on agricultural remote sensing data.
4. What specific improvements could the authors consider regarding the
methodology? No, the method part is clean.
5. Are the conclusions consistent with the evidence and arguments
presented and do they address the main question posed? Yes, this paper can be published in this form.
6. Are the references appropriate? yes
7. Please include any additional comments on the tables and figures.
- Add a legend of the upper figure of figure 1
- It is better to add a flow chart of the workflow
- You can highlight all tasks # in the paper
Reviewer 3 Report
The authors have submitted a good manuscript about monitoring of subsidies using remote sensing. There is no innovation in this work, but it is well written, is interesting and is applicable. I only have some minor comments:
- Figure 1: please, add a legend to the map showing the agro-ecological zones of Hungary.
- L310: include the formulas of the indices.
- Figure 7: include a legend for the point colors.
Reviewer 4 Report
The manuscript presents the results of very extensive research on the application of Sentinel-2 imagery for monitoring agricultural subsidies in Hungary.
There are several suggestions and questions for the manuscript to be considered for publication.
- Introduction
The first paragraph of the Introduction (lines 28-35) should contain a reference.
In my opinion, more detailed information on the use of machine learning in agriculture should be included in the Introduction.
- Section 2. Study area
line 157: Was the study based only on data from 2020?
line 171: The reference for Figure 1b should be given in the Figure caption.
- Section 3. Data and Methods
lines 309-310: Why were these four spectral bands selected?
line 349: How many cases were in the initial set of training data?
line 446: Data augmentation should be described in more detail.
- Section 4. Results
When evaluating the results, the authors focused mainly on accuracy.
Why do the authors not compare other performance metrics, such as TP Rate, FP Rate, Precision, Recall, F-Measure, ROC Area, PRC Area, MCC?
Have the authors tried to use a different classifier besides Random Forest?
Added comments:
1. What is the main question addressed by the research?
The manuscript presents the results of very extensive research on the application of Sentinel-2 imagery for monitoring agricultural subsidies in Hungary. The developed procedures were intended to assess the feasibility of introducing operational monitoring. The authors checked what the strengths and weaknesses of satellite imagery for CAP monitoring in Hungary are and whether a Random Forest is useful for the crop classification and the detection of cultivation events.
2. Do you consider the topic original or relevant in the field, and if so, why?
The topic is original. A machine learning approach involving the application of a random forest-based classifier has been used as a monitoring tool to classify crops and detect the cultivation events based on NDVI. A lot of criteria, e.g., basic agricultural cultivation, minimum criteria for grasslands, crop diversification, cultivation of protected grasslands were checked.
3. What does it add to the subject area compared with other published material?
The approach involving the extensive use of Sentinel-2 in the remote sensing process taken by the authors is innovative. The procedures developed for data classification and decision making are a great contribution of the authors in the subject area.
4. What specific improvements could the authors consider regarding the methodology?
The methodology has been applied correctly. However, there are some specific comments:
- Section 2. Study area
line 157: Was the study based only on data from 2020? In my opinion, it would be good to repeat the experiments in another year.
- Section 3. Data and Methods
lines 309-310: Why were these four spectral bands selected?
line 349: How many cases were in the initial set of training data?
line 446: Data augmentation should be described in more detail.
5. Are the conclusions consistent with the evidence and arguments presented and do they address the main question posed?
The authors could consider the suggestions below to obtain more results and to strengthen the conclusions:
When evaluating the results, the authors focused mainly on accuracy.
Why do the authors not compare other performance metrics, such as TP Rate, FP Rate, Precision, Recall, F-Measure, ROC Area, PRC Area, MCC?
Have the authors tried to use a different classifier besides Random Forest?
6. Are the references appropriate?
The first paragraph of the Introduction (lines 28-35) should contain a reference.
In my opinion, more detailed information on the use of machine learning in agriculture should be included in the Introduction. Additional references should be cited.
7. Please include any additional comments on the tables and figures.
line 171: The reference for Figure 1b should be given in the Figure caption.
Reviewer 5 Report
In the section of introduction, while the authors clearly present the broad context and the highlight of this study, its purpose and its significance are not well defined. The authors should explain the method they used to process satellite imageries; they should explain why they choose that method compared with other algorithms/methods available. The main aim of the study is not well defined, in addition the authors should highlight the main conclusions
Line 15: Please correct “The satellite data were subject to Random Forest-based crop classification and the detection of cultivation events based on NDVI (Normalized Differential Vegetation Index) time series analysis.” to “The processing of the satellite data was conducted using Random Forest-based for crop classification and the detection of cultivation events was conducted based on NDVI (Normalized Differential Vegetation Index) time series analysis results.”
Lines 30-35, 83-85, 91- 99, 284: Please provide references
Lines 96-97: Please correct “High High Resolution (HHR) image data” to “Very High Resolution (VHR) image data”
Line 155: “Study area” Should be in the section of “Materials and methods “
Line 170: Please provide the legend for the Figure 1 map a.
Line 172: Please correct “Data and Methods” to “Materials and methods “
Line 217: Please correct “Classification of satellite imagery involves the class…” to “Classification of satellite imagery involves the classes definition…”
Line 238: Please provide the link of this online system. The authors should provide more descriptions of the reference data such as their precision, accuracy and exact date.
Line 261: Please correct “…measurements…” to “…images… “
Line 262: “…all available Sentinel-2 images…”, Please precise the period
Line 593: In the section of Results, please provide maps of classification results.
Lines 52-53: Please refer to “Felegari, S.; Sharifi, A.; Moravej, K.; Amin, M.; Golchin, A.; Muzirafuti, A.; Tariq, A.; Zhao, N. Integration of Sentinel 1 and Sentinel 2 Satellite Images for Crop Mapping. Appl. Sci. 2021, 11, 10104. https://doi.org/10.3390/app112110104.
And
Snevajs, H.; Charvat, K.; Onckelet, V.; Kvapil, J.; Zadrazil, F.; Kubickova, H.; Seidlova, J.; Batrlova, I. Crop Detection Using Time Series of Sentinel-2 and Sentinel-1 and Existing Land Parcel Information Systems. Remote Sens. 2022, 14, 1095. https://doi.org/10.3390/rs14051095 “ For crop detection and maping using Sentinel 2 and 1 satellite images.
