Signal Processing Options for High Resolution SAR Tomography of Natural Scenarios

Round 1

Reviewer 1 Report

The authors present a detailed discussion on the Synthetic Aperture Radar Tomography (TomoSAR), a method to provide 3D imaging of illuminated targets by processing SAR data from multiple trajectories. They provide in depth discussion of the 1D, 2D and 3D approaches of TomoSAR image processing including the benefits and costs of each approach. They also provide a new 3D processing approach with substantially low computational burden than the established approach based on fast defocusing and refocusing, utilizing parallelized processing on GPUs. They show that this proposed approach can provide substantial speed-ups without sacrificing on the image quality. 

The paper is well-written with detailed mathematical explanation of the approaches. The results are clearly presented and show the proposed approach providing substantial benefits. I would only like to ask the authors to expand on the simulation section of the paper which feels like it is slightly weak. Firstly, while the true 3D approach is very computationally heavy, it would be better if for a small simulation, the other approaches and the proposed approach are compared in terms of speed and computational complexity. Secondly, in a similar fashion, for the AlpTomoSAR data, the results should be expanded to including the 1D, 2D approaches instead of only including the "Global approach" to show the effectiveness of the given approach. I think that these changes would make the paper more robust and give the readers a chance to truly understand the power of the algorithm. 

Author Response

Dear Reviewers, 

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Signal Processing Options for High-Resolution SAR Tomography of Natural Scenarios”. 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 comments carefully and reviewed the paper accordingly.  Revised text is marked in the revised paper. The main corrections in the paper and the response to the reviewer’s comments are detailed as follows.

Answers to comments from reviewers 1:

1. I would only like to ask the authors to expand on the simulation section of the paper which feels like it is slightly weak. Firstly, while the true 3D approach is very computationally heavy, it would be better if, for a small simulation, the other approaches and the proposed approach are compared in terms of speed and computational complexity. Secondly, in a similar fashion, for the AlpTomoSAR data, the results should be expanded to including the 1D, 2D approaches instead of only including the "Global approach" to show the effectiveness of the given approach. I think that these changes would make the paper more robust and give the readers a chance to truly understand the power of the algorithm.

Answer: We agree that the including of the 1D, 2D approaches do help to show the efficiency improvement of the proposed approach. Since the focus of this paper is not 1D, 2D approaches, we believe it is sufficient to present the comparison of speed once. Therefore, we recorded the CPU processing time required by 1D and 2D approaches in the processing of AlpTomoSAR data and placed the analysis at Line 448.

Reviewer 2 Report

Signal Processing Options for High Resolution SAR Tomography of Natural Scenarios

REVIEW 1

I am interested in remote sensing, have used and use it currently in mineral exploration and geological projects. More recently, I have been studying SAR and its applications, especially from satellite platforms (Sentinel 1). In the past I have also contacted with seismic tomography. My comments on this paper are, naturally, more on the method’s application and on its formal aspects. I won’t comment of the theoretical underpinnings, as other reviewers and the authors are better qualified.

That said, my main comment concerns the method’s applications. I would to have seen a (even if short) discussion of the practical applications of SAR Tomography – what is the added value of this method? What useful additional information can we obtain from it?

The submitted paper is very well written and structured. The comments I have are few and succinct:

• The abstract is too long. Please consider reducing it.
• Section 3.4 is superfluous. Please consider placing the (one line) text without a header.
• Please consider rearranging the text in section 3 so that Table 1, which refers to that section, is not placed in section 4.
• Figure 12 – red dashed lines used are hardly visible. I suggest a larger figure and using another (more contrasting) color.
• Figure 13 - I suggest a larger figure, as in Figure 12.
• Line 604 is not applicable and must have been part of a general template – “Sample Availability: Samples of the compounds ...... are available from the authors.” . Please remove.

Congratulations for the work done.  

Comments for author File: Comments.pdf

Author Response

Answers to comments from reviewers 2:

1.That said, my main comment concerns the method’s applications. I would to have seen a(even if short) discussion of the practical applications of SAR Tomography – what is the added value of this method? What useful additional information can we obtain from it?

Answers:  concerning applications, the added value of this work lies in the description of an exact 3D processing algorithm that can be implemented using standard computing machines and without the need for raw data. We deem this an important achievement, as it will allow researchers to carry out accurate tomographic analyses of extended natural targets, like glaciers and dense forests, using normally-available computing machines and focused SAR data, which are generally the only data available to final users. Using the AlpTomoSAR data-set, we demonstrated that conventional 1D or 2D processing approaches fail in retrieving internal ice features, which would lead to totally wrong conclusions about the internal structure of the illuminated scene. We deem these concepts were made sufficiently clear in the submitted paper, and therefore made no further modifications in this direction.

Concerning the analysis of AlpTomoSAR data, we decided not to insist with many details since results have already been discussed in depth in the paper:

Tebaldini S , Nagler T , Rott H , et al. Imaging the Internal Structure of an Alpine Glacier via L-Band Airborne SAR Tomography[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(12):7197-7209.

It is important to note that the above reference provided a detailed analysis of glacier features obtained by applying a close relative of the 3D algorithm discussed in this paper, but no description of the algorithm was provided since the focus of that paper was on the physics of ice scattering. Therefore, the contribution of this study focused less on the application contents and more on the mathematical derivation and technical details.

2. red dashed lines used are hardly visible. I suggest a larger figure and using another (more contrasting) color.

Answers: We have tried a lot of color. We did not find a more contrasting color in the jet colormap. But we used thicker red dashed lines and a larger figure to make the DTM lines clearer.

The paper has been revised according to all other suggestions.