Estimation of Multi-Frequency, Multi-Incidence and Multi-Polarization Backscattering Coefficients over Bare Agricultural Soil Using Statistical Algorithms

Round 1

Reviewer 1 Report

Overall Assessment

The authors used multiple linear regressions and random forest to simulate the backscattering coefficients of bare agricultural soil and evaluate their performance. But your abstract did not highlight the significance of this paper. Meanwhile, the measurement data was very old (in 2010). Therefore, I suggest that you reorganize the abstract, keywords and Introduction section to highlight the innovation of the study. I strongly recommend resubmitting the paper after major revision. The specific comments are as follows:

Introduction

Point 1: Line 32 to 35. The transition between the two sentences is so sudden. What is the logic connection between them?

Point 2: I found that the entire introduction section of the paper focused on summarizing and reviewing the different modeling approaches, but the method used this paper is not very novel. So, I strongly suggest the authors reorganize and rewrite the Introduction section. I think the paper should clarify the importance and necessity of estimation of backscattering coefficients over bare agricultural soil, and then summarize the scientific research progress from the aspect of remote sensing sensors and inversion methods. Finally, the authors should highlight the innovation and aims of this study.

Methods

Point 3: Line 275. Why can analysis of variance analyze the importance of variables? Please cite high quality literature to support this approach and briefly explain its principle in this paper.

Point 4: Line 300. Why is your training and testing set split at 50%?

Point 5: I suggest you add the calculation formula of your accuracy evaluation index in Section 3.3.

Results and Discussions

Point 6: Where is the estimating equation for multiple linear regression modeling? I did not see it in the results section at all. Besides, whether your multivariate linear equation simulating the backscattering coefficient has been tested for significance. Finally, if the equation is significant, are the variables significant? Even if the importance of your variable is high in the results, it does not matter if the variable is not significant.

Point 7: Make sure the bias and RMSE calculations in each of your models are correct.

Point 8: Line 345. Make sure the results are in the same range in the Figure 8

Conclusions

Point 9: Line 483. You did not conduct correlation analysis between multiple variables in the entire paper, how do you know there is non-linear relations between variables rather than multicollinearity. These conclusions are not obtained from the experiments in this paper.

Point 10: Please keep the conclusion as short as possible, I suggest the authors rewrite the conclusion section.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

Worth publishing with some revisions

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I have read you revised manuscript, but I did not receive your response to my comments of first review. After carefully check the revised paper, I think several comments in first review did not get satisfactory revision. Meanwhile, there are still several problems for the revised paper in review v2. So, I suggest resubmitting the manuscript after a major revision. The specific comments are as follows:

Comments in the first review:

Point 1: Please cite high quality literature to support this approach that variance methods can analyze the importance of variables

Point 2: Why is your training and testing set split at 50%?

Point 3: Whether your multivariate linear equation simulating the backscattering coefficient has been tested for significance. Finally, if the equation is significant, are the variables significant? Even if the importance of your variable is high in the results (Section 4.2), it does not important if the variable is not significant.

Point 4: You did not conduct correlation analysis between multiple variables in the entire paper, how do you know there is non-linear relations between variables rather than multicollinearity. These conclusions are not obtained from the experiments in this paper.

New problems of paper V2:

Point 5: According to your formula that you given in Section 3.3, the final bias value could be positive or negative. However, the bias of each of your scatter plots was positive. Is this a coincidence?

Point 6: According to the formula you gave for bias, I can see from your scatter plot that your accuracy is definitely far from bias=0.00db and bias0.02db (for example, Figure 8 (a)). Please recalculate the evaluation metrics Bias and RMSE of the model in your full text and make sure the values are correct.

Author Response

Please see the attachment

Author Response File: Author Response.pdf