Retrieval of Surface Soil Moisture over Wheat Fields during Growing Season Using C-Band Polarimetric SAR Data

Round 1

Reviewer 1 Report

The manuscript reports on the use of empirical models for the retrieval of soil moisture in wheat fields. The study deals with the data set acquired during the Soil Moisture Active Passive Validation Experiment (SMAPVEX12) campaign, which was carried out in western Canada in 2012. The subject has been investigated in many past papers, though the authors provide interesting insights.

On the downside, the physical interpretation of the results is missing. For instance, the reason why the phase difference ???−?? showed the strongest sensitivity to soil moisture should be addressed. In the past, a lot of work on the information content of ???−?? has been carried out (e.g., Sarabandi, K. "Derivation of phase statistics from the Mueller matrix." Radio Science 27.5, 1992). Moreover, numerous past papers have investigated various soil moisture retrieval algorithms using the SMAPVEX12 data set. I’d recommend adding a session reviewing the main results obtained on the SMAPVEX12 data set and comparing them with the present study.

Below, I report a few specific comments that may help to prepare the revised version.

Specific comments

Lines 267-272. The threshold for the correlation (r) less than 0.9 to discriminate between dependent/independent variables seems too high. Could you please further elaborate on the choice of such a threshold?

Lines 323-324. Why the observed correlation between the phase difference ???−?? and soil moisture is negative?

Table 6 & Figure 8. Overall, the developed regression models show similar performance. This may be due to a limitation of the data set. Could you please elaborate on this aspect?

Author Response

Please see the attached file for our responses to the reviewer's comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors employ a straightforward multiple linear regression approach to construct four distinct models (3_Sigma, Sigma_Hv, MixPol, and SPol) by leveraging the variability present in the input features. The response variable for calibration or training of these models is derived from the SMAPVEX12 dataset. However, upon reviewing the manuscript, it is evident that there are substantial revisions required prior to considering it suitable for publication.

 

1.     The methodology section provides a comprehensive description of the approach, but it would be beneficial to further elucidate the motivation behind the selection of channel ratio as input features for training the multiple linear regression models. Additionally, it's notable that certain crucial features like NDVI are not considered in the models. Could you elaborate on the reasoning behind these choices? Clarifying why channel ratio was prioritized over other potential features, and discussing the implications of excluding features like NDVI, would provide readers with a deeper understanding of the modeling decisions made in the study. This insight could enhance the transparency of the methodology and the interpretation of the results. Cite credible sources.

2.     The study presents a valuable investigation into the development of models using multiple linear regression; however, a crucial aspect that requires attention is the absence of comparisons with other statistical models or machine learning algorithms. To ensure the robustness and effectiveness of the proposed models, it is advisable to extend the analysis by comparing the results obtained from the four sets of input features with those of standalone machine learning models. This comparison, commonly referred to as baseline models, would provide an essential benchmark to assess the performance of the developed models. Including results from various machine learning algorithms, such as decision trees, random forests, support vector machines, or neural networks, would offer insights into whether the multiple linear regression models outperform or are comparable to these more complex methodologies. The baseline models can help identify the strengths and limitations of the proposed approach, enhancing the study's overall contribution to the field. Therefore, I recommend that the authors consider this important aspect and incorporate a comparison of their results with those obtained from standalone machine learning algorithms. This addition will strengthen the paper's methodology and bolster the confidence in the findings presented.

3.     Elaborate on the observed link between model performance and growth stages of wheat. How might this observation be explained or further explored?

4.     Could you provide more insight into the potential limitations of the current study? Are there any sources of bias or error that might affect the validity of the conclusions?

5.     The discussion section of the manuscript is currently lacking in depth and structure, which hinders the comprehensive understanding of the potential of the multiple linear regression (MLR) model. To address this, I strongly suggest reorganizing the discussion into subsections to provide a clear and focused presentation of the findings. Each subsection can address specific aspects of the study's outcomes.

6.     Moreover, it is essential for the authors to go beyond simple observations and provide rigorous analysis to illustrate the potential of the MLR model. By delving deeper into the interpretation of the model's coefficients, the authors can shed light on the relationships between input features and soil moisture estimation. This could involve discussing the significance and implications of specific coefficients and how they align with theoretical expectations.

7.     Furthermore, given the high spatial variability of soil moisture, it is crucial to perform a spatial distribution analysis. This analysis would involve varying the calibration/training datasets and assessing the model's performance across different spatial settings. This approach will provide insights into the model's robustness and its ability to capture the dynamic spatial patterns of soil moisture. The discussion can elaborate on the implications of these variations for practical applications and highlight potential challenges or opportunities in using the MLR model in different regions or landscapes.

8.     The discussion section of the manuscript would benefit from a more comprehensive approach to contextualizing the results within the broader landscape of recent studies related to soil moisture estimation using C-band Synthetic Aperture Radar (SAR) data. To achieve this, I recommend introducing a subsection titled "Comparison with Recent Studies" within the discussion. In this subsection, the authors should include a table that summarizes and compares the findings of their study with those of other recent studies that utilize either statistical or machine/deep learning approaches for soil moisture estimation. Focus on studies that leverage polarimetric, non-polarimetric, or mixed features from C-band SAR data for similar purposes. The table can include key metrics such as root mean square error (RMSE), correlation coefficient (r), and explained variance (or bias), allowing readers to quickly grasp the relative performance of the MLR models developed in this study in comparison to recent work. To ensure the discussion is up-to-date and reflective of current trends, consider including studies published from 2022 onwards. A few examples of such recent studies are provided below for reference:

-Deep learning and data fusion to estimate surface soil moisture from multi-sensor satellite images. Scientific Reports, 13(1), 2251. (2023)

- A low-cost approach for soil moisture prediction using multi-sensor data and machine learning algorithm. Science of The Total Environment, 833, p.155066. (2022)

- Retrieval of farmland surface soil moisture based on feature optimization and machine learning. Remote Sensing, 14(20), p.5102. (2022)

Other comments:

9.     Line: Correct the unit “versus 0.098 m m3 /m3”

10.     Optimize the keywords.

Author Response

Please see the attached file for our responses to the reviewer's comments.

Author Response File: Author Response.pdf

Reviewer 3 Report

1. Some English Sentences are not clear in the abstract, methodology, Introduction parts.

2. Rewrite the sentences clearly mentioned in point no.1.

2. Methodology should have more clarity.

3. Check the operator in Line no. 233.

4. Explain the novelty in the used methodology.

 

 

Improve in few sentences is required. Read the whole manuscript thoroughly.

Author Response

Please see the attached file for our responses to the reviewer's comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have satisfactorily responded to all my comments and made the necessary changes to the manuscript. I congratulate the team for their effort. I recommend accepting the manuscript in its current form.