GPS tracks to determine risk
The new study from Switzerland uses GPS tracks from the database of Skitourenguru.ch for this purpose. The GPS data is combined with the accidents, danger levels and avalanche problems of the bulletins of the respective days in order to quantify the risk as a ratio of accidents to ascents depending on the danger level and the terrain. The following questions, among others, are to be answered:
How does the risk change from danger level to danger level?
How does the risk in the altitudes and exposures mentioned in the avalanche bulletin differ from the risk in the other parts of the terrain?
Is the risk dependent on the prevailing avalanche problem?
Are the reduction factors of the rule-based methods correct, or do they change when the number of ascents is taken into account?
In order to get the best out of the data and to avoid comparing apples with oranges, the data first had to be prepared and filtered:
Data for the winters 2005/06-2018/19 were analyzed, before that there are no GPS tracks.
The GPS tracks come from ski or snowboard tours in open terrain. Therefore, only accidents from tours were used, and accidents from off-piste terrain were excluded.
In addition, wet and sliding snow avalanches were excluded, so the study refers to dry avalanches, or the avalanche problems of fresh snow, drifting snow, old snow and "no pronounced avalanche problem".
For most of the 784 accidental avalanches (at least one person recorded), only the starting point, i.e. the highest point, is known. In order to better take the terrain into account, an approximate avalanche path was calculated in each case and an average value for exposure and steepness was determined from this.
Over 7000 recorded tours are included in the evaluation. After excluding dense forest and very flat terrain, this results in a good 2 million individual GPS points in potential avalanche terrain.
For all points, i.e. accidents and GPS points, the danger level valid at the time, the avalanche problem and the particularly critical exposures and altitude ranges were determined on the basis of the archived bulletins, as well as the terrain characteristics (steepness, exposure). A certain area around the individual points was also taken into account (same method as Skitourenguru.ch).
The authors of the study counter the argument that the danger level cannot apply to one point or individual slope with a kind of law of large numbers: For very many points (or individual slopes), there must be a correlation between the regional level and the local hazard despite spatial variability, otherwise the hazard level as a concept would be useless.
For the calculation of the risk, each GPS point was counted as a "non-accident". Thus, the risk in this evaluation corresponds to the number of avalanches divided by the number of GPS points, each for certain conditions (e.g.: level 3, >30°, north sector, in the critical altitude range of the bulletin).
As not all winter sports enthusiasts record their tours, the GPS tracks only reflect a fraction of the actual ascents. The risk is therefore overestimated. Assuming that the same percentage of tours always found their way to the ski touring guru as GPS tracks, regardless of the conditions, a relative risk can still be determined for different conditions. For example: How does the risk at "level 3, >30°, north-facing slope, in the critical altitude range of the bulletin" differ from the risk at "level 3, >30°, southwest-facing slope, in the critical altitude range of the bulletin"?
Results
Over 90% of the accidents considered in the study happened at danger level 2 or 3. The accident location was almost always within the core zone of the bulletin, i.e. in the altitudes and exposures assessed as particularly critical.
The analysis of the GPS data shows that different altitudes and exposures are traveled to with different frequencies. The assumption implicit in Munter's reduction method that the same amount of skiing takes place everywhere is therefore not correct. North-facing slopes were skied 1.7 times more often than south or south-west-facing slopes. At level 2, 71% of the touring activity took place within the core zone of the bulletin. For level 3, even 86%. However, as higher altitudes and/or more exposures fall within the particularly critical core zone at level 3, this does not necessarily mean that less attention is paid to the bulletin at level 3.
The risk increases sharply with increasing danger level: at level 2, the risk is over 5 times as high as at level 1, at level 3 around three times as high as at level 2.
According to study author Kurt Winkler, the avalanche problem has only correlated with the risk in the last two winters (2019/20 and 2020/21), i.e. since the avalanche warning has been rule-based and thus more uniform in its assessment of avalanche problems. The latest data not yet included in the study shows that the old snow problem is rightly feared: the risk is 1½ times higher than for the other avalanche problems at the same danger level. In comparison to the danger level, however, this influence is significantly smaller.
Dependence on altitude and exposure
In relation to the number of GPS points ("non-accidents"), there are more accidents with increasing altitude, so the risk increases with altitude. No further increase was found above 2700m. Below the altitudes indicated as critical, the risk is over 5 times lower than at the critical altitudes.
Accidents occurred 3.6 times more often on north-facing slopes than on south-facing slopes. However, as north-facing slopes are also skied more often, the corresponding accident risk (accidents / skis used) is "only" 2.1 times higher - this clearly shows the influence that taking the number of skis used can have on the risk calculation.
While avoiding the northern sector (NW-N-NE) or the northern half (W-N-E) are important reduction factors in the reduction method according to Munter, the authors of the study conclude that the risk reduction achieved in this way is lower than assumed. The reduction factors "no northern sector/half" should therefore not be exhausted. Better results are achieved by avoiding the exposures specified in the bulletin.
According to Munter, the "hazard potential" within the core zone of the bulletin is 4 times higher than outside. This ratio corresponds to a difference of about two hazard levels. In the assessment, the one-level rule is common, i.e. the assumption of a reduction by one level outside the core zone (e.g. in the graphical reduction method or the SLF interpretation aid for the bulletin). The results of the study also tend to correspond to a reduction of one level.
The difference in risk between slopes within the core zone mentioned in the bulletin (i.e. slopes within the specified altitude as well as exposures) is clear. However, an even better differentiation is possible if altitude and exposure are considered separately: Even a little below the critical altitude zone, the risk is significantly lower than in the core zone, even if one remains in the critical exposure. The authors of the study therefore propose a method that incorporates altitude and exposure more clearly separately in rule-based decision aids.
