Automatic Detection of Regional Snow Avalanches with Scattering and Interference of C-band SAR Data

Round 1

Reviewer 1 Report

The paper investigates the potential of Sentinel-1 SLC images to detect avalanche debris. The idea behind this work, i.e. to combine backscatter and coherence observations to better detect avalanche debris, is very relevant. Backscatters are sensitive to several properties of the snowpack, such as its water content and roughness and most of avalanche detection algorithms are based on change detection methods that identify pixels of significant backscatter increase between 2 dates while the surrounding area remains unchanged. Coherence is a correlation measure of complex SAR images that determines the degree of similarity of the phase of two images and depends on the target backscattering that may change between two satellite passes.

So combining backscatters, interferometric coherence, ... should be very useful. The problem is that all the elements of understanding of the method, the analysis of the results and their evaluation are not sufficiently clear in the paper. I think the paper needs an in-depth revision in order to better target the objectives of the paper, the methodology being followed, to distinguish the calculation steps, the analysis of the detection results and their evaluation with independent data.

Some other important elements necessary for a full understanding of the paper are listed bellow:

• Detailed Workflows of SAR processing is needed.

• A lot of calculations and statistics are provided, but often without any guiding thread, which makes it difficult to fully understand some parts of the article.

• The significance of the results should be increased with more Sentinel-1 scenes.

• The authors argue that the combination of measurements (backscatters, coherence, entropy, …) allows them to improve the detection of debris. Did the authors compare the detection results with a simpler method based only on backscatters? Which of the parameters used in the detection is most informative?

• What is the distribution of detected debris by orientation? This would probably explain the differences in performance depending on the direction of the orbit.

• I did not fully understand how debris detections were evaluated with independent data (in situ or optical satellites). This is a key point to address to be able to conclude whether the method, subject of the article, is performing well or not.

Author Response

    Thank you very much for your constructive and scientific suggestions, which will benefit me for life. We have carefully studied your suggestion and realized that our research does have shortcomings, and your suggestion can make it more perfect. The suggestions you put forward have been revised one by one（Please see the attachment）, and I hope to make you satisfied.

Author Response File: Author Response.docx

Reviewer 2 Report

The contents and results of the paper are interesting and important. I think that the paper is nearly publishable as it is. One comment is as follows; Full spellings of some acronyms are expected such as SLC, FPR, and TPR for general readers. 

Author Response

Point 1.Full spellings of some acronyms are expected such as SLC, FPR, and TPR for general readers.

Response1: Correct full name writing is necessary to avoid misunderstanding of the reader due to confusion with terms of the same abbreviation. Thanks to your careful discovery of this problem.

The author traversed all abbreviations in the text and ensured that the correct full name was supplemented where they first appeared. The following is a summary of all modifications:

• Line 89: SVM (Support Vector Machine)
• Line 119: SLC(Single Look Complex)
• Line157: PCA (Principal Component Analysis)
• Line 165: IW (Interferometric Wide swath)
• Line 177: POD(Precise Orbit Determination)
• Line 179: SAR(Synthetic Aperture Radar)
• Line 409: FPR(False Positive Rate) 
• Line 410: TPR(Ture Positive Rate)
• Line 516: GRD(Ground Range Detected)

Among them, we move the full name position of SVM, SLC and PCA because the position changing of the first occurrence in the text; IW, FPR, TPR and GRD are abbreviations inadvertently not given with the full name explanation, which have been supplemented in this modification; POD and SAR are abbreviations introduced when modifying other problems, and the full name is added at the position of their first occurrence. The acronyms in the full text have been carefully revised and the corresponding full names have been given.

Reviewer 3 Report

The study aim is to present the research on “Automatic Detection of Regional Snow Avalanches with Scatter and Interference of C-band SAR Data” The manuscript is presented clearly and nicely. I would like to suggest minor revisions.  

1. Are there any past research based on the avalanche in the study area. If so, better to incorporate why needs additional research. It will help you to show your originality.
2. Figure 1 needs to be modified by adding China as a part of the figure. It will help the reader to understand the location of the study area,

Author Response

 Point 1.Are there any past research based on the avalanche in the study area. If so, better to incorporate why needs additional research. It will help you to show your originality.

Response1: As you suggest, only by clarifying the necessity of conducting this study in the study area can its originality be stand out.

Line 98-116 describes the work carried out in the study area in the past, the achievements obtained, the shortcomings as well as the necessity and urgency of carrying out this study.

First, researchers have carried out a large number of systematic avalanche basic information collection efforts in the study area, and manual field surveys are the only method used. However, subjective field manual operations are inevitably limited by harsh natural conditions. The obtained data often fail to meet the needs in quality and quantity, and the avalanche information in most regions is still unknown. This method has no longer satisfied the decision-making needs of disasters in the dual background of increasing disasters as well as rapid development (Line 98-104). In addition, the research carried out in the study area mainly focused on the 19th century, and no further technical mining and application were carried out except for conventional manual observation operations after entering the 20th century. Therefore, it is needed to propose methods that can accurately obtain avalanches within the region in order to provide effective decision support for local decision makers and improve public perception of the real situation of avalanches in the study area.

Second, as climatic conditions changes and surface transformation result from human activities, old avalanche thematic maps are no longer of use and update is needed. (Line 105-111).

Finally, from the theoretical analysis, most of the studies have realized the automatic detection of avalanche by detecting the change of a single index. However, the scattering and interference characteristics of avalanches have not been excavated and utilized; from the analysis of focused snow types, the main direction is oceanic snow, while the study of continental dry and cold snow is still very lacking. The study area, on the other hand, is precisely typical of this type of snow (line 111 – 116). Therefore, it is of great significance and originality to carry out automatic detection of continental dry cold snow in the study area in combination with avalanche scattering and interference characteristics.

Point 2.Figure 1 needs to be modified by adding China as a part of the figure. It will help the reader to understand the location of the study area.

Response2: Clear elaboration and graphical representation of the country and regional location of the study area can strengthen readers' understanding of the study area. Thanks to the reviewers for such important suggestions.

The original study area profile maps do not clearly indicate their geographical location. The revised version, according to the reviewers' comments, can intuitively present that the study area is located within the territory of Kazak Autonomous Prefecture of Ili, Xinjiang Uygur Autonomous Region, China. In addition, in order to make the study area profile more artistic, some detailed modifications have been made such as reformatting, enlarging fonts, and using mountain shadows to highlight terrain characteristics.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.