Near-Real Time Automatic Snow Avalanche Activity Monitoring System Using Sentinel-1 SAR Data in Norway

Round 1

Reviewer 1 Report

The authors report on a number of developments in the field of avalanche research and the background and application of a complete automated near-realtime monitoring system.

As much as I enjoyed reading this manuscript and learning about the authors' work I also felt the need to learn more about a number of aspects as this monitoring in all its design and sophistication remains closed-source. The relevance of this work is undisputed and it remains to find out how a broader readership can benefit from these developments. In order to achieve that, a manuscript of such kind needs to be detailing conceptual features as well implementation aspects that allow to actually take out approaches.

I believe that this manuscript succeeds in doing so for most of the parts reported on. I would request/suggest a few additions and moving of paragraphs to make things a bit clearer.

The introduction is a bit underwhelming. As the relevance of this work is relatively obvious, existing approaches could be emphasized a bit more by providing a more comprehensive review. Some aspects appear in section 5.2 and although it is useful to re-discuss them there, it would be extremely helpful to put them upfront to better understand the state of the art. Also 5.1 provides important detail on previous developments that would be important to know earlier. I believe it would be helpful if readers can see more image examples comparing manual and automatic detection, in particular at slightly higher resolution (cf. figure 10). The Introduction could benefit from some national background information on the actual process of monitoring and in how far monitoring actually helps to assess risks for the future. Provide some context. On line 54 when citing some research it would help if the authors could briefly summarize the main results here. The same applies to Line 84 (reference 12) which could require some detail In particular when citing in-house developments that form the foundation of this work it would help to provide more background. SAR processing is done with in-house software, change detection is done using in-house tools and results are published in conference proceedings only. This feels a bit like a black box. As these contributions are the foundation of all your developments, this demands for more details. You report on the national DEM (2.1.2) but you do not mention its details (how obtained, accuracy, limitations,...). A full description would be certainly appreciated. The conclusions would benefit from some words detailing how this monitoring scheme can actually help to assess and eventually prevent avalanches in the future. How are these products used? Some of the abbreviations require to be introduced before first use (VV, VH, and all abbreviations in the abstract). Please check again. I noticed a few minor English mistakes. "Data" is plural, please check again. Also, satellites do not "fly" on their orbit. Line 254: check figure caption, it probably does not belong here.

Author Response

Dear reviewer 1,

Thank you very much for your interest in our study and the provided comments to improve its readability for a broader audience.

In the following we will reply to your requests/suggestions point-by-point:

 

Introduction:

Our goal was to keep the introduction concise, however, we understand that a broader audience requires more background information. We will move some of the referenced work from us and others in the field into the introduction to improve understanding of the state-of-the-art. We will also give a more thorough introduction about our processing system being used in Norway for public avalanche warning and explain in detail how avalanche warning benefits from satellite-borne avalanche detection. As requested, we will explain the importance of avalanche detections also in the conclusion in greater detail and give examples of how the dataset is used and how building a dataset of avalanche detections over several winter can help reduce inherent uncertainty in avalanche warning and forecasting.

 

In-house SAR processing software:

The entire processing system is dependent on our in-house SAR processing software “gdar”. We explain what this software does and that for example the geocoding routine is very similar to SNAP. “gdar” is used in many ESA proof-of-concept studies that we are leading. All our previous publications on SAR remote sensing of avalanches are also based on processing SAR data in “gdar”. We explain the most essential geocoding steps done with “gdar”, however, we can give more technical details on the scene pairing and change detection to improve taking out approaches. In general, “gdar” could be replaced by SNAP and it is likely that the same results would be achieved.

 

Examples of manual vs automatic detection:

We have an additional figure that graphically explains the age-tracking process. It shows some more examples of manual vs automatic detection of avalanches.

 

National DEM:

We will give more details on the DEM!

 

Abbreviations:

We will carefully go through the manuscript and introduce all abbreviations.

 

On behalf of all co-authors,

 

Markus Eckerstorfer

Reviewer 2 Report

 

Authors presented manuscript on the snow avalanche activity monitoring using Sentinel SAR data in Norway. The system is pre-operationally used by the Norwegian Avalanche Warning Services.

The structure of the manuscript is not clear, please reorder the structure.

 

General remarks:

Please define what avalanche is. Please clarify the relation between identification, observation, forecasting and monitoring of the avalanche. Please add a review of available methods to observe avalanche phenomena. Please provide geological setting of the test site as understand prone to avalanche and test site location. Please explain avalanche prone areas. Please provide any quantitative data on avalanche events in Norway, Europe, etc…as well as quantitative/qualitative classification of the avalanche. Sections on products produced using monitoring system are split (e.g. Line 286 – avalanche activity map, Line 368 – avalanche activity map). Please be clear what products you prepared using monitoring system. Section “Results” with “field validation” is not correct? Please reorder it.

Detail remarks:

Line 66 – I see three output products, please clarify.Line 77 - I do not see “black” Lines 90-92 – What data was used for DTM production, please provide. Please add cited websites to references. Line 107 and Table 1 – What forecasting season is please clarify. (vide general remark: forecasting, monitoring, or observation season – please be precise). Table 1 - please provide dates, periods, units – not clear. Line 286 – Is winter season is forecasting season? Please clarify. Line 330 - -What True Skill Score is? Figure 6 – Please add proper caption. Line368 – Are the same time series as Figure 6? So, please move Figure 6 to section 4.4 Line 435 – please explain what soft snow is. Figure 9 – provide references Line 537 – What small avalanche is, please clarify. Is Section 5.4 an introduction? Please reorder the manuscript.

Author Response

Dear reviewer 1,

Thank you very much for your interest in our study and the provided comments to improve its readability for a broader audience.

In the following we will reply to your requests/suggestions point-by-point:

 

Avalanche definition:

We will include a definition in the beginning of the introduction

 

Identification, observation, forecasting and monitoring:

We understand that all these terms are confusing. We will make sure that we are reducing the use of all these terms and explain them when necessary.

 

Avalanche observation methods:

For a complete review of remote sensing of avalanches we refer already to Eckerstorfer et al., 2016. However, we can give a concise overview over field-based monitoring vs. satellite-based monitoring and also explain other remote sensing techniques that are currently used in the introduction.

 

Geological setting:

We will include a concise introduction to our study site and the most avalanche prone areas. By doing so, we will also explain the term “avalanche prone areas”.

 

Quantitative data on avalanches in Norway and Europa:

We do not understand what the point is of providing quantitative data on avalanches. The point of this study is to quantify avalanche activity in Norway using Sentinel-1. An analysis of past avalanche activity is beyond the scope of this study.

 

Line 286 vs 368:

In the first line mentioned here, we explain our methods and what the products are as an end result of the processing system. In 368 we present these end results.

 

Field validation as a result section:

This comment is not entirely clear to us. We see the field validation that we carried out as a result, that’s why we put it into the result section. We can add an explanation of the method used in the method section if that helps clarifying the structure of the manuscript.

 

Line 66:

This is a mistake, you are correct, there are only three output products

 

Line 77:

This is a bad description. The areas without coverage appear as dark, shaded areas.

 

Lines 90-92:

This is a comment we have gotten from reviewer 1 too. We will provide more technical details on the DTM.

 

Cited websites to references:

We will do that!

 

Table 1 on forecasting season:

We will carefully go through the manuscript to make sure that we concisely use only one term for forecasting season, which is basically the winter season defined from 1 December to 31 May.

We will provide more dates and unites to the table to clarify all the numbers.

 

Line 330:

We will provide a concise description of how a True Skill Score is calculated.

 

Figure 6 caption:

We will update the caption to explain all variables shown in the graphs. They show the same time series as we mention in line 368. We can merge sections 4.3 and 4.4, as we agree that it makes no sense to separate them.

 

Soft snow avalanches:

We will explain what we mean. It basically describes the density of the snow in the avalanche debris. We infer it from the surface roughness we saw in the field.

 

Line 537:

We will provide the reference to Figure 9. A small avalanche is capable of burying a person, as we write. We can additionally give information on typical area covered by a small avalanche.

 

Section 5.4:

We don’t see this section as an introduction but moreover an upscaling of our results from Norway to worldwide application. First we showed what is possible in terms of using Sentinel-1 for avalanche detection and then we show worldwide Sentinel-1 coverage and discuss what is needed to apply our processing system to anywhere on Earth.

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.