Retrieving Surface Soil Moisture over Wheat-Covered Areas Using Data from Sentinel-1 and Sentinel-2

Round 1

Reviewer 1 Report

Summary:

The authors use Sentinel-1 and -2 data to estimate surface soil moisture (SSM) of wheat fields in Hebi, Henan province, China.  In situ soil samples were collected at 28 locations on three different dates and analyzed for soil moisture using gravimetric methods, along with vegetation water content and density at the same sites.  The authors modified the water cloud model (WCM) to account for vegetation fraction and used multiple vegetation indices, including red-edge Sentinel-2 observations, to model vegetation water content (VWC), which allowed the authors to isolate the soil backscatter contribution from Sentinel-1 backscatter observations.  Then, the authors applied a support vector machine (SVR) machine learning method to relate soil backscattering to in situ soil moisture measurements, splitting their samples into 80% training/20% validation.  The authors found that the VV polarization was most effective at estimating SSM, and that the overall SSM remote sensing approach provided good agreement with in situ samples.

General Comments:

Overall, I think the manuscript is written in a fairly clear and straightforward manner.  The analysis has a logical flow, and I don’t think any major components are missing.  Generally, the figures effectively communicate information about the study.  The authors do a good job of literature review and comparing their results to other studies.  The analysis is somewhat novel in its handling of vegetation fraction in the WCM and the use of red-edge vegetation indices, but it is otherwise largely similar to some other studies that estimate SSM from active microwave observations.  However, the study is also somewhat narrow in scope (only one crop in one agricultural area), and should be viewed as a proof-of-concept.  I think the analysis is mostly sound, but I have a few reservations about the results and I would like to know the authors’ responses to these questions. 

Specific Comments:

There are no line numbers in the document, so it makes it difficult to indicate certain passages, but I will do my best.

Introduction, 1st & 2nd paragraph: The authors note that it is difficult to accurately estimate SSM over large scales, but make no mention of passive microwave methods.  Despite the coarse spatial resolution, these methods have proven quite effective.

Page 4, 1st full paragraph:  3 dates within only a 2-month period is little data to go on.  This study would be much stronger with periodic data over an entire growing season. 

Page 4, last full paragraph:  Do the authors have any concerns that Sentinel-2 data were not available until a week after their initial Sentinel-1 date?  Couldn’t wheat growth in the intervening period have some affect on their results?

Page 5, 1st paragraph:  What does “almost simultaneous” mean?  The authors should be more specific here.  Same day?

Equation 11: Typo “red-eage2”

Page 8, last paragraph:  I think this is the only location where the authors mention what is input to the SVR, and it is not especially clearly written.  The authors should be very explicit about which explanatory variables were included, and that soil moisture was the dependent variable.

Section 4.1, 1st paragraph:  16 validation samples is very little to go on.  I’m sure that the authors put in a lot of effort to collect soil samples at 28 sites on three different dates, but I worry that with so few validation samples in such a small region, there is a high risk of overfitting the SVM. 

Page 10:  Did the authors perform do any tests for the statistical significance of the variables in the multiple regressions for VWC (e.g. nested F-tests)?  With only 16 samples, it is very possible that one or more of these indices were not statistically significant.

Section 4.2, 1st paragraph:  What do the authors mean by “average variation”?

Page 13, multiple paragraphs:  Add the units (%) to the SSM RMSE values.

Figure 7:  The top and bottom portions of the map in panel (a) show SSM near zero.  This seems unrealistically low.  Is the colorbar incorrect?  Then, in panel (c) these are the wettest areas, at >0.4.  This also seems unrealistic (that these areas could flip so drastically, while the central region of the map changes little).  Do the authors have an explanation for this?

Additionally, Figure 1 shows that there were sampling sites in the bottom red area in panel (a), which appears to have SSM < 0.1, but no points in Figure 6 had values that low.  Is there an error somewhere?

Both of these things make me doubt the accuracy of the SSM predictions.  I don’t know if this would be from a lack of adequate training data for the SVM, or overfitting the SVM, or poor soil backscatter input (either from Sentinel-1 itself or from some step of the processing or WCM), but these seemingly contradictory observations give me pause.

Figure 7:  I only see panels (a), (c), and (e).  The frequency plots seem to be missing.

Figure 7:  Additionally, it is a bit confusing that SSM here is presented as a fraction, but in Figure 6 and Table 3 it is in units of percent.

Data Availability Statement:  The authors have listed “Not applicable”.  However, the authors have collected a lot of in situ data (e.g. SSM, VWC).  The availability (or lack thereof) of these data should be communicated here.

Author Response

Thank you for your time in handling this manuscript. We are thankful to the reviewers and the academic editor for pointing out some important modifications needed in the report. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made corresponding revisions which we hope meet with approval. Revised portions are marked in red in the new manuscript. An item-by-item response to the Reviewer’s comments is provided in the following pages.

Author Response File: Author Response.docx

Reviewer 2 Report

Congratulations on producing such a well organised and well written paper. You have done an excellent job of adding to the science of surface soil moisture measurements from remotely sensed data.

There is one misplaced period (should be a comma) in the last sentence of the discussion, and the introduction (and acknowledgements) contain the error of linking the Sentinel satellites exclusively to ESA. ESA are the space segment partners, but the satellites are actually part of the European Union's Copernicus Earth Observation Programme. Crucially it is Copernicus that assures the full free and open access to the data (also if you have not checked it out, there is an SSM product as part of the Copernicus Global Land Service). For the sake of correctness you should amend the exclusive reference to ESA, and you should certainly delete the text referring to ESA's global monitoring for environment and security in the introduction. GMES was again an EU programme, of which ESA is a key partner, and GMES became Copernicus a number of years ago.

Author Response

Dear  Reviewer:

Thank you for your time in handling this manuscript. We are thankful to the reviewers and the academic editor for pointing out some important modifications needed in the report. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made corresponding revisions which we hope meet with approval. Revised portions are marked in red in the new manuscript. An item-by-item response to the Reviewer’s comments is provided in the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript synthesized entinel-1 synthetic aperture radar and Sentinel-2 optical data to retrieve SSM in a wheat-covered area. The paper quality overall looks good. The methodology is well described and is scientifically sound. However, there are quite a few minor grammar issues, and the manuscript does not have line numbers, making it difficult to make detailed comments.  

Below are a few minor catches:

Abstract: the research gap and contributions from this paper are not mentioned.

In the introduction part Paragraph 2: Sentinel-2 should be mentioned.

There is also a gap of description for research of Wheat-covered area.

Figure 7: what is left and right in the caption?

Data Availability Statement: Not applicable is not accepted. The authors should at least try their best efforts to make the data publicly available.

Author Response

Dear  Reviewer:

Thank you for your time in handling this manuscript. We are thankful to the reviewers and the academic editor for pointing out some important modifications needed in the report. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made corresponding revisions which we hope meet with approval. Revised portions are marked in red in the new manuscript. An item-by-item response to the Reviewer’s comments is provided in the attachment.

Author Response File: Author Response.docx