Mapping Paddy Rice Fields by Combining Multi-Temporal Vegetation Index and Synthetic Aperture Radar Remote Sensing Data Using Google Earth Engine Machine Learning Platform

Round 1

Reviewer 1 Report

Since many researchers are paying attention to the combination of optical remote sensing with SAR, the reviewer thinks this manuscript would interest many researchers. However, the analysis does not meet the hypothesis presented by the author (L100-110). Further analysis or improvement of presentation seem necessary.

1. The authors should show how the detailed class in table 2 improved the classification. The authors just show the difference among types and did not show the difference among the detailed classification. 
2. If the three vegetation indices actually differentiate paddy rice from non paddy fields, it is quite interesting (L252-253). However, the Figure 3 just shows the difference between paddy and non-paddy is quite limited. If consider the variation of the vegetation indices, the differentiation would be difficult.
3. To respond the hypothesis that the authors indicated, the detailed analysis for why the procedure miss the classification shown in Figure 7 must be necessary.
4. The reviewer thinks how did adding optical image improves the classification of paddies by SAR. Because many studies suggested that SAR is the best way to detect paddies.
5. Figure 9 seems have no meaning in the manuscript because if the classification is improved, the water body extraction must be improved. 
6. Most figures and tables needs improvement. The font size is too small, legends are necessary for Figure 6, the symbols are not distinguishable in Figure 7, and tables need which is observed and estimated.  

Author Response

Please see the attachment。

Author Response File: Author Response.pdf

Reviewer 2 Report

The submitted research work aims to investigates the results of combining three vegetation indices (NVDI, EVI, ELWIS) and SAR data using Sentinel satellite on the Google Earth Engine. I found this paper very interesting where Several technical aspects were nicely implemented and explained sufficiently. Undoubtedly, authors invested huge amount of time and have made a great effort to produce this high-quality of research which is clearly structured and the language used is largely appropriate. Nevertheless, I found few gaps that need correction and improvements. As final decision, I see that this manuscript in its form and level deserves to be accepted for publication in MDPI-RS BUT after addressing below MINOR COMMENTS.

DETAILED COMMENTS PER SECTION:

• The title is adequate for the content of the paper.
• ABSTRACT:
• The abstract gives a good overview about the undertaken work. Authors mentioned clearly the aim and objectives of their work and included all the required info in the abstract. The list of keywords can be improved by adding words like “Precision farming, SAR data”.
• INTRODUCTION:
• The introduction is well written and very organised.
• Material and Materials
• The figure 1 of the study area must be improved especially the legends.
• Under section >2.2.1. Sentinel data. Please improve the description and add clearly the pre-processing steps and the all the corrections implemented to the EO data.
• Improve the quality of the figure 2.
• Improve the key of the classification map on figure 4.
• RESULTS:
• The produced figures have been discussed sufficiently.
• DISCUSSION
• The discussion section of this paper is well written and the authors explained/supported sufficiently their findings.
• Conclusions: Authors made a very good conclusion and very interesting recommendations.
• As final general comment, please make sure to define ALL the acronyms form their first appearance in the paper. Also, all the references MUST BE CHECKED and formatted as required by MDPI-RS, also make sure that all the references have DOI number unless it is not available.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

 

General Comments: Missing satellite imageries or cloud contaminated imageries are big issues in applications of satellite imageries for processing and mapping areas of interests. Issue become more challenging when variables of interest are dynamic in space and time. Method presented in this study could help to overcome missing/cloud-contaminated imageries in space/time domain of interest. However, on the downside of manuscript: this study site is selected in only one region and hence application of this methodology in a complex vegetation areas could be challenging to obtain high classification accuracy.

Regardless, methodology and concept are advancements to science of image processing and classification. I would recommend acceptance of this manuscript for publication after minor revision.

 

Line 30:  However, accurately mapping paddy rice XX a long-term   (please insert “ is”)

 

Line 104:  “….in one area (Dong and Xiao 2016).”  (Add here following sentences and a reference)

Advance function of Geostatistics can be used to interpolate space/time missing data (Jat et al., 2018). However, high spatiotemporal discontinuities limit application of geostatistical methods as these methods rely on spatial and temporal autocorrelations. Additionally, “the intra-class variability ……

 

 

Figure 1: need high resolution figure.

 

Tables 1 and 2: Tables look scattered in their current form.

Line 195: The selected set of sample data was randomly split into 30/70 percent separately for training and classification accuracy assessment, respectively. (This is unusual. Usually training data should contain 70% and test/classification should consist remaining 30%).  Please make sure that  correct data fractions are used for each process.

 

Citation: Jat, P., Serre, M.L. A novel geostatistical approach combining Euclidean and gradual-flow covariance models to estimate fecal coliform along the Haw and Deep rivers in North Carolina. Stoch Environ Res Risk Assess 32, 2537–2549 (2018). https://doi.org/10.1007/s00477-018-1512-6

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The revised manuscript unfortunately does not meet the recommendation of the reviewer. The authors almost just showed the classification results by machine learning with SAR and MSI. Information is quite limited for the reader to improve their analysis. For the purpose, the reviewer recommend following analysis.

1. The reviewer is interested in how the analysis was improved by the detailed classification of crops in Table 2. Please compare the results between the analysis based on the detailed classification and on the non-detailed classification (paddy rice, non-rice fields, forest, built-up area and water body).
2. Please include items of the detailed classification in Table 5. The reader will understand what is the problems in the classification.

Followings are also needed to be revised. 

1. evi, ndvi and lswi of figure legends must be EVI, NDVI and LSWI.
2. The differences among maps were hardly distinguishable in Figure 4.
3. Table 5 needs labels of "estimated classification" for the first row, and "observed classification" for the first column. 
4. L410 - L427 is hardly understandable. Is the discussion base on Figure 7? No light green in the new figure. Any regional distribution of classification error is hardly found in the figure.
5. JRC Monthly Water History obviously does not include paddy fields as water body. The comparison is meaningless. If the authors would like to criticize it, it is better to describe elsewhere. The water body extraction seems not included in the purpose of this study. 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf