World of Science | Review ISSW 2018: Avalanche forecasting

This time: Avalanche Forecasting (Session 11)

This session is about forecasting the avalanche situation in operational mode. The contributions can be roughly divided into the following thematic subgroups:

• Data-based approaches to sharpening definitions.

• Local vs. regional avalanche forecasting and bridging this scale difference

• Modern technical tools for warning services

As usual, we will go through the abstracts in order and summarize them briefly.

Data-based approaches to sharpening spongy definitions, typical and atypical patterns

As skiers who use avalanche warnings and forecasts, we are used to the regional avalanche situation reports that are common in the Alps, which give us an overview of the situation in the region - be it a federal state or a mountain group. As winter sports enthusiasts, we of course know that snow conditions often vary greatly on a small scale and that it is therefore often not so easy to say something generally valid about a whole region, even if you just want to tell your buddy what the snow is like at the moment. But that's exactly what the warning services have to do every day. The well-known danger level scale, the avalanche problems and various other formalisms ensure the greatest possible consistency to ensure that everyone is largely talking about the same thing and understands the same thing. On the other hand, neither the subjectivity of human avalanche forecasters nor that of users can ever be completely eliminated. Although the danger levels of the five-part European scale are defined on the basis of the probability of triggering, avalanche size and distribution of danger spots, it is well known that words such as "possible" and "probable" leave room for interpretation.

An SLF team is analyzing at which danger level how many avalanches occur and how large they are in order to better quantify words such as "possible" and "probable". The frequency of spontaneous releases increases strongly with the danger level (non-linear). It is particularly interesting that the avalanche size hardly changes with the danger level in the Swiss data set. A higher danger level means more avalanches, not necessarily larger ones ( Quantifying the obvious: The avalanche danger level, Schweizer et al.). However, the situation appears to be different in Colorado: Here, an increase in avalanche size tends to be observed with the danger level. The increase in the number of avalanches observed is also more or less linear with the danger level. The American danger level scale differs slightly from the European scale, but it is not clear whether this is the reason for the differences (Patterns in avalanche events and regional scale avalanche forecasts in Colorado, USA, Logan and Greene).

If it snows a lot, avalanches will occur at some point. And in spring, the timing of wet snow avalanches is related to the diurnal variation in temperature. So far, so obvious. However, quantifying this correlation based on data and defining the temporal dependency of avalanches and weather events in more detail is not so easy. Another SLF study explains that information from increasingly available, automatic avalanche detection systems (radar, seismic) can help to identify corresponding patterns. This is mainly because they notice more outflows than human observers, who are dependent on good visibility. After a precipitation event, it can take up to several days for avalanches causally linked to the precipitation to occur. In the case of energy input in spring and wet snow avalanches, this usually only takes a few hours. The better the data basis, the better such patterns can be recognized and the better they can be integrated into avalanche forecasting (When do avalanches release: Investigating time scales in avalanche formation, van Herwijnen et al.).

Local vs. regional

The regional situation report with hazard levels and information that is as general as necessary and as specific as possible is therefore one thing (and you should consider yourself lucky if you get one). The individual slope assessment is the other. The local situation can differ greatly from that described in the regional LLB. This is why the LLB is not wrong, this is simply due to the difference in scale:

In Livigno, an attempt was made to compare the regional danger level of the LLB with individual slope assessments by mountain guides and the results will probably remind many of their own approach in the terrain: The guides noted down the danger level they would assign to each slope they skied. As a rule, the individual slope danger levels were 1 or 2, even with a higher level in the LLB. Only those slopes that were classified as comparatively harmless were skied. The discrepancy between local and regional is therefore primarily related to the terrain and group management on site and not to systematic differences in the danger level assessment (Regional versus local avalanche danger evaluation, Monti et al., no extended abstract).

Apart from the personal winter sports decision, "Do I go in there or not?", local avalanche warnings are important wherever a specific slope or avalanche line is concerned, not the entire region. This is practical, for example, for avalanche commissions or ski resorts that have to decide whether a certain road or slope should be closed. In contrast to the methodically at least theoretically uniform regional avalanche forecasts, there are considerable differences in the local variant. As a first step, the EAWS sent out a questionnaire to relevant institutions and organizations in Europe in order to roughly record these differences. Most, but not all, respondents produce their local assessments on the basis of the regional LLBs, even if the local hazard level may differ from the regional one. Snowpack models are rarely used in operational, local warning operations. There are major differences between the individual EAWS member countries as well as between individual institutions in terms of technical, financial and human resources. It will probably take several more questionnaires to create a Europe-wide overview of the "who, what, why and how exactly?" of local avalanche warning (Local avalanche warning in Europe, Jaedicke et al.).

Continue on the next page --->

The overall picture may still be somewhat opaque, but individual case studies show in detail how it can work: On December 19, 2015, an avalanche hit the village of Longyearbyen on Spitsbergen. 2 people were killed and 12 houses were destroyed. Another avalanche hit the village in 2017. Since the first accident, a local avalanche forecasting program and corresponding infrastructure have been developed at full speed, particularly with regard to guidelines for evacuations. Svalbard lies north of the Arctic Circle. In winter, it is therefore dark throughout and there is a reliance on automatic snow monitoring and weather stations. The automatic data was validated with snow profile images and serves as input for a snow cover model. As only one specific terrain chamber endangers the location in Longyearbyen, this newly established program is not a regional avalanche forecast as in the Alps, but rather a practical example of operational "single slope assessment" (Slope scale avalanche forecasting in the Arctic (Svalbard) Prokop et al.).

Another example of avalanche warning on a local scale is presented by the organizers of a ski touring race in Italy. It is a long way from planning the route and possible alternative routes to the race, which is divided into different phases. The route is selected a few weeks before the race based on the snowpack history and the terrain. The weather and snow conditions are then monitored and the route is adjusted if necessary. During the race, experts will also continuously assess whether and how the situation changes. It is also essential to ensure that the athletes carry functioning safety equipment and attend the safety briefing (Ski Alp Races: Avalanche Hazard evaluation and risk analysis, Raviglione et al.).

An avalanche warning concept for the ski resort in Tetnuldi was developed on a resort-wide scale in a French-Georgian cooperation. Due to a lack of infrastructure and financial resources, the resort is reliant on comparatively simple solutions. As interest in ski touring and freeriding is increasing in the region, it is nevertheless important for the ski resort to have a functioning safety and warning concept. In a special training program, the local ski patrol was trained to create a situation report for the ski area based on standardized criteria. The focus is on the type of avalanche problem and the spatial distribution of the problem, i.e. "what?" and "where?". It is based on conceptual guidelines established in Europe and North America, such as the Avaluator and the ATES classification, and now produces regular bulletins. The authors hope that similar programs will also be set up in other Georgian ski resorts (Local avalanche danger assessment in reduced means context: An example in Tetnuldi (Georgia-Caucasus), Escande et al.).

Conclusion

Leaving aside communicative issues (how do I present an avalanche forecast so that the user 1. notices and 2. understands it?), the challenge in predicting the development of the avalanche situation is to capture highly complex processes and then reduce them to the essentials. What the essentials are can depend on the current snow and weather situation, as well as the exact area of application and the spatial and temporal scale on which one is operating. New possibilities for data collection and documentation make things a lot easier for the warning services. Snowpack models are helpful on a case-by-case basis, but so far they are far from being able to replace human avalanche forecasters. The best possible recording of the current situation and extensive process knowledge in combination with communication concepts and definitions (hazard level scale, avalanche problems, etc.) that are as standardized as possible seem to be crucial.