The Avalanche Terrain Exposure Scale (ATES) v2 is an update to a 20-year-old system introduced in Canada as a risk management tool for backcountry travel. This new version of ATES offers improvements that combine developments over the past two decades into a single model.

After reading the manuscript along with the original ATES model article (Statham et al., 2006) and subsequent developments (e.g., Campbell and Gould, 2013), I’ve gained an appreciation for the deep refinement of ATES. However, I’m unclear about why, with version 2.1, there is a need for a peer-reviewed publication, when there was not for the original ATES nor its interim conference papers. Likewise, I have concerns about the re-use of previously published material. As outlined in the review criteria, publication in NHESS requires new work that is scientifically significant, like the original ATES. ATES v2.0 has some new features, but appears as an incremental improvements on two decades of work. A figure or table summarizing the evolution of ATES since its inception would be helpful. The requirement for a substantial new advance is why there are often not peer-reviewed manuscripts for new versions of a model. There are examples (e.g., Sturm et al., 1995; Sturm and Liston, 2021; Dozier, 2022; Dozier et al., 1981), but the authors must convince the reviewers that the changes are substantial enough to change the understanding of the topic. Conversely, given that the previous ATES work has been published mostly in conference proceedings, there are issues with preservation. For example, the “permalinks” for ISSW Proceedings are not an authoritative archival method, whereas the Digital Object Identifier that is produced for this discussion article is. The authors should consider stating their motivation for a peer-reviewed article directly in the manuscript.

The article contains reprinted material, sometimes verbatim. As cited, Tables 4-6 are from Campbell and Gould (2013) and Figure 12 appears verbatim in Sykes et al. (2024, Figure 10) as well as in parts of Toft et al. (2024, Figure 4). In fact, most of the figures and tables (Figure 9, Figure 11, Table 8, Table 10, Table 11) are from previous publications. At the least, these reproductions beg the question: what’s new here?

Given that ATES is likely to be applied to remote areas, the AutoATES v2 open-source code is a significant new contribution. It is discussed in Section 5.2.1 and should have been included in a “Code availability” section (https://www.natural-hazards-and-earth-system-sciences.net/submission.html#manuscriptcomposition). Yet there is already a publication for AutoATES v2.0 (Toft et al., 2024). Thus, this section too contains previously published information.

Not present, although maybe present in previous ATES literature which could have been cited, is a discussion of the limitations of the static hazard model. If there is no snow, there is no snow avalanche hazard, regardless of terrain. Given that the majority of backcountry use (i.e., hiking) comes in the summertime, that is a significant limitation. Further, as someone who has spent considerable effort mapping snow from satellites, I find the omission of the changing snowcover to be a major model shortcoming. A static map is certainly easier to produce, but the authors could at least discuss how a dynamically modeled or remotely-sensed snowcover could be incorporated into the ATES model.

In short, I commend the authors efforts with ATES v2, however this manuscript requires considerable revision to justify how it is a new contribution. Currently, it appears to be a summary of previously published work rather than an original research article. I’m stopping short of rejecting the manuscript and suggesting it be resubmitted as a review article, but the authors should consider that route.

NB 6/5/24

Minor:

Abstract

What’s new here? Consider adding l 530-536 and maybe l 37 – 40 to the Abstract

Introduction: Element-at-risk sounds like agency jargon. Define.

2 Background

I’m still unclear why impact-based hazard mapping can’t be used for backcountry recreation. Impact-based models can generate runouts without historical observations. Perhaps it’s because impact-based models rely on low probability events, i.e., 10-300 year return periods?

L 71 Correct ATES Zoning model to ATES Zoning Model

Section 3.1 This class 0 issue highlights the shortcomings of a model that doesn’t use current conditions. No snow means no avalanche hazard, regardless of terrain. This limitation should be discussed.

Table 3 : 1:1 years? Please provide units. 10:1

Section 4

The inputs to these tables are likely not available in many parts of the world, e.g., route options seems to require pre-mapped routes or trails. This could use some further explanation.

Table 4-6 are reprints of Campbell and Gould (2013). What’s new here?

214 –Unsupported slope needs clarification? With respect to what axis? Even a serac is supported in the slope normal direction.

214 – “possible source”. It’s not a possible source. It is a source of tensile stress.

Table 10 – representing an area with a point is an oxymoron

5.1 Spatial scale

General comment, not the authors fault. The ISSW proceedings are not resolving on my browser without manually typing in. These proceedings need DOIs, not “persistent links” which don’t resolve well and do not have changeable underlying links.  What happens when we are no longer using PHP, or arc.lib.montana.edu is no longer the base URL? Retrievability, especially for older works, is a major problem with grey lit citations. See my comments on preservation.

Works cited

Campbell, C. and Gould, B.: A proposed practical model for zoning with the Avalanche Terrain Exposure Scale, Proceedings of the 2013 International Snow Science Workshop, Grenoble, France, 385-391,

Dozier, J.: Revisiting topographic horizons in the era of big data and parallel computing, IEEE Geoscience and Remote Sensing Letters, 19, 8024605, 10.1109/LGRS.2021.3125278, 2022.

Dozier, J., Bruno, J., and Downey, P.: A faster solution to the horizon problem, Computers and Geosciences, 7, 145-151, 10.1016/0098-3004(81)90026-1, 1981.

Statham, G., McMahon, B., and Tomm, I. l. a. A., 2006.: The Avalanche Terrain Exposure Scale, Proceedings of the 2006 International Snow Science Workshop, Telluride, USA, 491-497,

Sturm, M. and Liston, G. E.: Revisiting the Global Seasonal Snow Classification: An Updated Dataset for Earth System Applications, Journal of Hydrometeorology, 22, 2917-2938, 10.1175/jhm-d-21-0070.1, 2021.

Sturm, M., Holmgren, J., and Liston, G. E.: A seasonal snow cover classification system for local to global applications, Journal of Climate, 8, 1261-1283, 10.1175/1520-0442(1995)008<1261:ASSCCS>2.0.CO;2, 1995.

Sykes, J., Toft, H., Haegeli, P., and Statham, G.: Automated Avalanche Terrain Exposure Scale (ATES) mapping – local validation and optimization in western Canada, Nat. Hazards Earth Syst. Sci., 24, 947-971, 10.5194/nhess-24-947-2024, 2024.

Toft, H. B., Sykes, J., Schauer, A., Hendrikx, J., and Hetland, A.: AutoATES v2.0: Automated Avalanche Terrain Exposure Scale mapping, Nat. Hazards Earth Syst. Sci., 24, 1779-1793, 10.5194/nhess-24-1779-2024, 2024.

 General comments

The paper outlines the changes to the ATES terrain assessment based on the experience of the last 20 years. The main improvements are as follows:

1. a) additional classes 0 and 4,
2. b) exclusion of glaciers and
3. c) two types of definition (communication model and technical model).

The paper explains the ATES scale and how the classes are defined in a simple and comprehensible manner. The general outcome is not new.

The communication model provides a straightforward description of the scale, which helps less experienced people to recognise avalanche terrain. However, experience is required to understand the definition. E.g. Challenging: " With careful route finding, options exist to reduce or eliminate exposure".

The parameters "exposure" and "frequency magnitude" are subjective. I question whether "frequency magnitude" is a useful way to classify ATES. If the terrain characteristics are optimal for avalanches at a specific starting zone, then avalanches can occur if the snowpack is unstable. The question of frequency is more relevant when assessing objects with observed records. A remote starting zone with no records from past avalanches is not necessarily less complex than a starting zone were records are available.

The authors say that the ATES rating is subjective and that some criteria require experience and local knowledge to be properly assessed. This leads to different assessments both for manually and automatic mapping. 

As some parameters of the technical model describe frequencies, this is only suitable for classifying an area or a route as a whole. It is not possible to assess individual cruxes or objects using some of the assessment criteria.  E.g. "many very large starting zones", "some open slopes > 35°", "options exist to avoid avalanche path".

The possibilities of high-resolution terrain models and avalanche dynamics models to describe the runout, such as used in the Swiss avalanche terrain maps, are not mentioned. Furthermore, the different approach to classifying avalanche terrain is not mentioned, for example, "classified avalanche terrain, CAT" (e.g. in Introduction).

Specific comments

1 Introduction or 2 Background

• Mention other automatic approaches, such as “Classified avalanche terrain, CAT” (Harvey et al., 2018), incorporating high-resolution digital terrain models, avalanche data and numerical avalanche dynamic model

3 Principal changes to the Avalanche Terrain Exposure Scale

L80: has ATES Zoning Model been introduced before?

4 Avalanche Terrain Expore Scale v.2

Tables 1 and 2 : The colours are not ideal:

• Why is complex black and extreme red. Black is rather used for extrem in hazard rating
• Problem with red/green colour blindness.

L160-167: A bit short and confusing. Maybe pull up as matrix with ATES classes and 8 parameters defining 40 criterias. How are criterias ideally combines to rate ATES? I would like more guidance than in this respect.

Table 3:

• Is “Frequency magnitude” necessary for backcountry ATES rating?
• Convoluted terrain leads to higher ATES rating than planar terrain (terrain shape). This contradicts with route options. In convoluted terrain there often are more options for less exposed route than in planar terrain. Furthermore planar terrain often leads to widespread crack propagation in unstable snowpacks and therefore to larger avalanches.
• Difficulty of defining "Exposure" and "Frequency magnitude". Also low frequence can be extrem terrain….

L220: Convoluted terrain: What do you want to say in this section? 

L224: What are propagation spots? Do you mean trigger spots?

L263: What about remote terrain without records? A slope with no records is not inherently less prone to avalanches.

The paper “The Avalanche Terrain Exposure Scale (ATES) v.2” by Statham and Campbell offers a valuable overview of the developments and applications of the ATES, focusing on the latest advancements. It presents ATES v.2 and the changes from ATES v.1, including applications to routes and areas. The paper effectively introduces ATES and its purpose, emphasizing the classification and communication of avalanche terrain exposure. The main updates from v.1 to v.2 are highlighted (5 classes instead of 3, removal of glaciation as a classification criterion) and the motivation behind the updates seems well argumented. By providing background information on the technical model and the communication model(s) the manuscript is of interest to a wide target-audience ranging from avalanche professionals and educators to individuals working or recreating in potential avalanche terrain.

However, the main intention of the paper is partly unclear, which may cause potential readers to get lost between the general ATES review and the focus on developments specific to ATES v.2. While the historical evolution and widespread application of ATES are emphasized, it would be beneficial to also briefly discuss the similarities and differences to other (automated) avalanche terrain classification schemes (e.g. Harvey et al., 2018; Schmudlach and Köhler, 2016) and their reception by the target audience. To gain better insight into the ATES methodology, the systematic combination of terrain parameters to produce a rating is discussed in detail, but it would also be helpful to provide more guidance on how these parameters could be or are assessed.

The paper devotes considerable attention to the communication model and potential (risk management) applications, but the description of the updated technical model (thresholds, description) sometimes lacks consistency. One general point needing clarification in a revised version is the terminology. Specifically, the term “exposure” seems to be used inconsistently or with different meanings. On one hand, terrain exposure (potential vs. actual) towards avalanches is a crucial part of the terrain classification, while on the other hand, temporal exposure (of the element at risk) is used in terms of ATES as a risk management tool. Additionally, the interplay between avalanche size and frequency could be discussed more in depth, including the limitations of a static representation of a dynamic problem (how the maps are connected to different avalanche hazard/problems/size scenarios).

Overall, the paper is timely, fits the target audience, and fills a gap in peer-reviewed literature on avalanche terrain classification schemes. Many recent scientific publications have been based on ideas developed by the authors over the past 20 years and as such a peer-reviewed reference to ATES is certainly of interest to the community. However, figures, captions, and referencing leave some room for improvement. Please refer to the specific line-by-line comments for more information.

Specific line by line comments

p.1 Abstract: The abstract is (nearly) identical to the one in the corresponding extended abstract in the ISSW23 proceedings. Please compare with the abstract in the corresponding ISSW proceedings and revise accordingly. Two main questions appear in the abstract but also throughout the paper (see comments above):

• l. 7-8: Are ATES ratings only independent of daily hazard conditions or also independent of the (temporal) exposure of the element at risk?
• l. 13-14: Is ATES actually risk management tool or is ATES a tool for risk management (comparable to what slope maps are for classical risk reduction methods)?

p.2 l.19-23: Briefly explain the difference between hazard and risk, noting that hazard, vulnerability, and exposure are all key factors. Describe how these concepts are related to ATES and how exposure is defined within this context.

p.2 l.30: Providing a more in-depth review of different applications of ATES, including manual and automated approaches, and applications in different regions would be interesting. Also an overview of different (spatial) application scales (location, size, etc.) could be of interest to the readers.

p.2 l.40: Correct “Larson” to “Larsen.” (generally check correct spelling and formatting of references in the manuscript)

p.2 l.41: Insert a comma in “(Klassen, 2012)” and specify if this refers to ATES v.2 or ATES in general.

p.2 l.48: Include both recreationists and professionals in the discussion.

p.2 l.58-60: Expand on why quantitative methods become impractical in this context (due to the highly mobile element at risk). Clarify the impact and exposure-based approaches within ATES (see comments above on exposure).

p.3 l.64: Clarify what is meant by “river ratings”.

p.3 l.70: Reformulate the sentence for clarity.

p.3 l.76: Question the certainty that avalanches do not occur in class 0 terrain; suggest it may be very unlikely instead (compare table 2 “small sluffs”).

p.8 l.173: Discuss how the ATES scenario (potential maximum avalanche size, expected average avalanche size, etc.) is implicitly considered when developing a spatial and temporal exposure rating. So any individual trying to come up with a spatial and temporal exposure rating for a certain location (for both ATESlinear and ATESspatial) must implicitly have some sort of “Avalanche Scenario” in mind (i.e. at least potential avalanche sizes, which are to some extent related to expected avalanche frequencies).

p.10 l. 191 ff. @Forest effects on avalanche formation:

• The first argument you introduce in the discussion of forest effects is the mechanical anchoring of the snowcover in dense forest stands. However, this effect might be secondary to modification of snowpack structure by influencing wind re-distribution and micro-climatic conditions in forest stands. You also state that ATES uses a simplified model of the involved processes; maybe you can expand on which forest effects are included in your assessment and which are neglected (e.g. do you mainly consider forest effects on avalanche formation, or do you also consider potential forest effects on avalanche runout behavior?).
• Also a discussion of the extent of the spatial evaluation area might be of interest here, since average tree spacing might be substantially different when assessed over different spatial scales/extents (e.g. 10m pixel in GIS or a slope in a manual delineation).
• How would you e.g. classify the slope depicted in Fig. 3 according to tables 5 and/or 6? Can you comment on how to assess average tree spacing in the scope of pratical applications?

p.11 l.204: Provide more information on the reasons for combining ATES v.1 with the zoning model (such as automatic classification with more objective thresholds).

p.11 l.211: Define “high spots” and “steep, unsupported slopes” for non-expert readers and explain their significance in relation to avalanche exposure (overhead hazard).

p.12 l.220 ff.: Clarify the meaning of slope shape classes, as the definitions of classes between “flat” and “cliffy” are less clear. Discuss the slope shape factor’s relevance to avalanche triggering and potential avalanche sizes and overhead hazards. All in all slope-shape is probably the least justified of the 8 ATES factors. The area reference remains rather unclear (while parts of a slope can be convex, the whole slope might be convoluted or intricate or cliffy?). The discussion of terrain shape is mainly linked to likelihood of avalanche triggering by limited additional load (e.g. skier-loading) and does not really consider factors such as potential avalanche sizes and overhead hazard?

p.13 @terrainTraps: Consider adding the Harvey et al. (2018) CAT, ATHM reference.

p.13 l.239, 245: Elaborate on what is meant by “harmless.”

p.14 @Avalanche Frequency: Suggest a more nuanced formulation regarding the stability of avalanche frequency at a location, considering potential shifts due to climate change and the significance of the observation period.

p 15. l. 274-276, Tab 8: Is e.g. typical impact pressure not rather a measure of intensity?

p.16, section 4.2.7: Refine the definition of remote triggering. Use clear definitions for runout zones, distinguishing them from the track and relating them to international classification schemes (e.g., zone of origin, transit, deposition, see De Quervain, R et al.: Avalanche atlas , Unesco, Paris, 1981 ).

p.16 l.302: Specify whether “The ATES Technical Model” refers to the old or the new model.

p.17 @route options: Clarify the meaning of “route options” when applied to ATES_spatial as opposed to ATES_linear.

p. 17 l. 320-321: Please check for consistency “Class 1” -> Class 1 Terrain, by checking for uniform usage throughout the text.

p.18: Consider providing specific color codes (RGB, hex) for clarity, as color descriptions alone can be vague.

p.19 l.345: Refer to the corresponding figure (Figure 11?) when discussing North American colors.

p.22 @actual vs. potential exposure: Clarify the difference between actual and potential exposure in the context of ATES, and how avalanche size scenarios play a role in identifying exposed segments. Additionally, (temporal) exposure in terms of avalanche risk may have a different meaning than (spatial) exposure to avalanche terrain (see comment above)?

p.22 @spatial scale: Discuss the importance of spatial scale and how well the eight ATES evaluation criteria can be assessed at different spatial scales (i.e. forest densitiy on a synoptic scale might be difficult, …).

p.23 l.445: Provide a brief explanation of autoATES and related automated criteria and algorithms if it has not been referenced previously.

p. 26 l.501: also refer to Larsen et al 2020 who developed autoATES v1

p.26 l.501: Discuss the limitation of automated algorithms for autoATES due to the ‘subjective’ selection of parameterization.

p.26 l.504-505: Potential & true (terrain or spatial) exposure: Please comment on the differences between the types of exposure and double check on the wording throughout the paper. Following your thought “because there is no route” further implies that without an element at risk there is no (temporal) exposure (see comment on “Is ATES a risk management tool or tool for risk management”?).

p.26 l.506: Highlight the importance of being aware of resolution differences in maps and corresponding uncertainties.

p.27: Mention additional benefits of ATES compared to classical slope-angle-based terrain-choice strategies, particularly emphasizing the role of overhead hazard and terrain exposure to avalanches (one of the highlights of ATES, which appears to be underrated throughout the paper).

Figures

General: Review all figures and their captions. Highlight key features in the images and provide appropriate scales for the maps.

Fig. 1: Describe the blue trails (ATES class 2?) and explain why the white trails are class 0 (due to flat and forested terrain?). Include data sources for trails and maps. Add an overview map for context, orientation (north arrow), and scale (scale bar).

Fig. 2, 4: Indicate if all displayed terrain is the same class and highlight accordingly.

Fig. 3: Provide more context, such as slope angle, and show how an ATES v2.0 map would look in this area. Highlight different forest densities and slope angles.

Fig. 5: Highlight the zone of deposition and discuss terrain rating.

Fig. 6, 7, 8: Highlight different zones and terrain ratings, particularly areas that might propagate into nearby starting zones.

Fig. 10: Include overview maps, orientation, and scale. Ensure 10b is not arbitrarily cut off and is easily interpretable without local knowledge; consider redesigning the figure. 

References

Harvey, S., Schmudlach, G., Bühler, Y., Dürr, L., Stoffel, A., and Christen, M.: Avalanche Terrain Maps for Backcountry Skiing in Switzerland, in: Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018, pp. 1625 – 1631, Innsbruck, Austria, 2018.

Schmudlach, G. and Köhler, J.: Method for an automatized Avalanche Terrain Classification, in: Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 2016, pp. 729 – 736, Breckenridge, Colorado, 2016.

De Quervain, R et al.: Avalanche atlas , Unesco, Paris, 1981