The Avalanche Review printed this article in spring 2024. Minor revisions here.

Confidence? Are you kidding? Hell yeah I’m confident! I understand the snow and I can ski that slope. If it avalanches, I’ll ski out. What’s the problem?

Backcountry skiers have confidence. Lots of it. How else can we venture into big dangerous mountains where avalanches pummel down? This (over)confidence has been engrained in our psyche over millions of years for survival. The problem is, it doesn’t always help us avoid avalanches when backcountry skiing. To better avoid avalanches, it helps to shift our thinking to the antonym of confidence: uncertainty. By acknowledging that we don’t fully understand the avalanche problems, we can identify gaps in our knowledge, work toward reducing those gaps, and add margins for safety. For this article, uncertainty is defined as the lack of information, knowledge or understanding about avalanche problems.

“One thing we know about risk management is that it tends to get worse when our uncertainty increases.”

— Ian McCammon

Currently, a scale of avalanche problem uncertainty doesn’t exist. A simple uncertainty scale would allow us to better incorporate uncertainty into our thinking and discussion, both at the recreational and professional level. This article describes the sources of uncertainty, proposes an avalanche uncertainty scale, and how to use it.

CAA InfoEx Confidence Scale

The closest thing to an uncertainty scale are the Confidence Ratings (below) for the Hazard Assessment charts from the Canadian Avalanche Association InfoEx. Their definition of confidence is “an expression of the degree of certainty about a prediction of expected conditions in the future.” One option for an uncertainty scale is to invert the Confidence Ratings, so high confidence would become low uncertainty, and low confidence would become high uncertainty. This would be a good start, but more specifics would help.

High Confidence

The forecast is based on high-quality information and the nature of the issue makes it possible to render a solid judgement. A 'high confidence' rating does not imply fact or complete certainty however, and such judgements still carry the risk of being wrong.

Moderate Confidence

The information used to produce the forecast is credibly sourced and plausible, but it is not of adequate quality or sufficiently corroborated to warrant a higher level of confidence.

Low Confidence

The credibility or plausibility of the information used to produce the forecast is questionable, or the information is too fragmented or poorly corroborated to make solid judgements, or there are significant concerns regarding problems with the sources.

Sources of Uncertainty

Uncertainty can be from natural sources or knowledge sources (CAA 2016, Jamieson and others 2015). Natural sources (aleatory) of uncertainty include weather and snowpack variability over terrain. Knowledge sources (epistemic) of uncertainty come from limited field data or limited understanding about the topic. To best suit both recreational and professional users, a subset of sources are used in this scale. These sources include:

1. Accuracy of field data.

Collecting different data reduces uncertainty, while more of the same data does not reduce uncertainty. Credibly-sourced field data will also reduce uncertainty better than poor field data. For example, the quality of data at Teton Pass can be as high as from the Alaska Range, but the accuracy at Teton Pass will be better because of the larger sample size.

2. Problem uncertainty.

Each of the nine avalanche problems—or five in the European Avalanche Warning System model (EAWS 2017)—have different inherent levels of uncertainty. For example, dry loose problems have low uncertainty (higher predictive snow behavior), while deep slab problems have high uncertainty (low predictive snow behavior) (Wagner and Hardesty 2014). Uncertainty can also arise from any attribute of the current problem including type, location, likelihood (sensitivity and distribution), size, and danger (Statham and others, 2017). For example, weak layer distribution can have low uncertainty, as when a uniform layer drapes the terrain. Or distribution can have high uncertainty, as when a surface hoar layer was partially blown down before burial.

Another way of looking at problem uncertainty is through the summary statement of avalanche danger. The middle of the danger scale, at moderate and considerable, tends to have higher uncertainty…it might avalanche. Contrast that with low danger, where uncertainty is usually low and it probably won’t avalanche. Uncertainty is also low at high or extreme danger where avalanches are likely or even certain…it probably will avalanche.

3. Effect of the next weather system.

Weather forecast further into the future, or for larger areas has greater uncertainty. For example, weather forecasters and/or models may be unsure if the approaching system will result in 0.5 or 2 inches of snow water equivalent, or how the next storm layer will bond to the old snow surface. Climate change adds further uncertainty to avalanche forecasts as historical records may not indicate current trends.

4. Skier knowledge, experience, and understanding.

A backcountry skier’s level of uncertainty about an avalanche problem may vary based on their level of knowledge, experience, and their understanding of avalanches. For example, a new backcountry skier may not be able to recognize an avalanche problem, while a more experienced skier could recognize an avalanche problem, but dismiss the uncertainty due to a high risk tolerance.

Avalanche Problem Uncertainty Scale

This scale is for the uncertainty of avalanches, including type, location, likelihood or size. Since avalanche uncertainty can not be calculated (Atkins 2023), users must use their own judgment to determine if a rating is appropriate for a scenario. The sources of uncertainty can be one or more of the sources listed above, in addition to the complexity of the interaction of the factors involved. This uncertainty is not the uncertainty, but it is your uncertainty, or the team’s uncertainty.

This uncertainty scale can be used while trip planning, on guide meeting forms, in avalanche courses, and in avalanche-avoidance language. It is a sense-making aid to help skiers choose appropriate terrain and routes for the the conditions and team. It became an important addition to guide meetings and avalanche courses at our Alaska Guide Collective. Uncertainty discussions usually involve a reason for the uncertainty rating. For example, “I’d say we have high uncertainty above treeline because we haven’t been there in a week.” While this scale could be more specific, with checkboxes for each category, I have kept it simple for ease of use. This scale is not a refined or accepted scale, but rather the starting point to further discussion.

Low Uncertainty

The accuracy of field data is sufficient for confident decisions; the avalanche problem(s) have a predictable behavior; the effect of the forecast weather on avalanche conditions is well understood; or the team has sufficient knowledge and experience with the problem. Travel advice: Low uncertainty does not imply certainty and this judgment still carries the risk of being wrong. Apply normal caution and margins for safety.

Moderate Uncertainty

The accuracy of field data is limited or of moderate quality; the avalanche problem(s) have a varied behavior; the effect of the forecast weather on avalanche conditions is uncertain; or the team has some knowledge and experience with the problem. Travel advice: Collect different field data which may reduce uncertainty. Use extra caution and wide margins for safety.

High Uncertainty

The accuracy of field data is limited or of poor quality; the avalanche problem(s) have an unpredictable behavior; the effect of the forecast weather on avalanche conditions is poorly understood; or the team has little knowledge and experience with the problem. Travel advice: Plan route options that account for the low confidence. Maintain wide margins for safety including turning around. Collect different field data which may reduce uncertainty.

References

Canadian Avalanche Association InfoEx.
Bruce Jamieson, Pascal Haegeli and Grant Statham. Uncertainty in Snow Avalanche Risk Assessments, GeoQuebec. 2015.
Dale Atkins. Uncertainty Versus Risk, The Avalanche Review 42(1) p28-32. 2023.
European Avalanche Warning System, Typical avalanche problems, approved by General Assembly of EAWS. 2017.
Statham and others. A Conceptual Model of Avalanche Hazard. Natural Hazards. 2017.
Wendy Wagner and Drew Hardesty. Travel Advice for Avalanche Problems, Proceedings of the ISSW. 2014.
Technical Aspects of Snow and Avalanche Risk Management: Resources and Guidelines for Avalanche Practitioners in Canada. Canadian Avalanche Association. 2016.

Thank You for Helping With this Article

Aaron Diamond, Andrew Schauer, Bruce Tremper, Dale Atkins, Elliot Gaddy, Henry Munter, and Nick D’Alessio.