Characteristics of Avalanche Occurrence
Avalanche forecasting tries to predict the probability of avalanche occurrence relative to a given area and time. As a mountain lover, I am curious about knowing which characteristics influence the most when referring to avalanche danger level.
The used data set, which is accessible here and previously assessed by Schweizer et al., 2020, gives the avalanche activity in the region of Davos (Switzerland) during the years 1998-2019 (21 years). It contains a total of 13'918 avalanches with their descriptive features such as size, aspect, trigger or snow type, among others. For more information about the data we are going to analyse you can download this pdf. Furthermore, a detailed avalanches glossary can be found in this website.
Within this blog I will try to find in the data the answer of these questions: Is there a relation between the different characteristics of an avalanche? How the Avalanche Danger Level changes during the year? How does the Avalache Activity Index looks like? How the avalanche danger relates to avalanche snow type, trigger type and size?
Let’s get started!
Avalanche features correlation
In order to have a clear overview of the data set, we can create the correlation map shpwing the relationship between pairs of features.
Since we are working with categorical values (e.g. trigger_type_HUMAN, trigger_type_NATURAL, trigger_type_EXPLOSIVE) the correlation coefficients between values from the same category are omitted.
Even though there are no major dependencies between features, there are some which are more correlated than others. For instance, we can state that the snow type WET is associated to the season SPRING whereas the snow type DRY is linked to WINTER, which makes a lot of sense.
Surprisingly to my expectations, the avalanche size has not shown much correlation with respect to the avalanche danger level. This may suggest that this feature might remain to a secondary importance when assessing the risk level.
Let’s have a deeper look into the data!
Insight I: Avalanche danger level
The European avalanche danger scale ranges from 1-Low to 5-Very High and it is based on the snowpack stability and the avalanche triggering probability. The common sense tells us that the higher the risk, the greater the avalanche activity.
Another way to represent this data is by the Avalanche Activity Index (AAI) which indicates the number of avalanches per day.
As shown in the plot, the AAI mean per ascending danger level is 1.73, 2.95, 8.85, 46.91 and 27.33, respectively. Hence, there is a tendance of higher AAI with higher risk.
The danger level 5-Very High is documented to be most likely incomplete, so it might be the reason why it does not follow the mentioned tendancy.
Last but not least, another interesting plot would be to show how many avalanches has been released within all days in a specific danger level. Unfortunately, the data set does not contain the danger information of those days in which any avalanche was released (AAI=0).
A good insight would be to know how the snow type and avalanche trigger type contribute to the avalanche danger level.
Let’s continue seeking information in the data.
Insight II: Snow and trigger types
For a better readability of the plots the columns whose danger level is 1-Low and 5-Very High will be ignored. As seen before, none of those cases contain many numbers of entries.
The snow condition is the liquid water content in the starting zone, and it has three categories: dry, wet or mixed.
The tendancy of wet-snow avalanche consistently increased with an increase of the danger level. Contrarily, we cannot claim the same regarding the dry-snow whose avalanches are more frequent at risk 3-Considerable.
Not surprisingly, the number of natural avalanches is linear to the avalanche danger level.
On the other hand, when the danger level is 4-High there were almost no avalanches triggered by humans. That demonstrates that people are aware of the risk and when danger level is beyond 3-Considerable they do not engage in winter sports outside open ski runs and trails.
Part III: Avalanche size
As seen in the heat map at the beginning of this post, the avalanche size does not have a significant effect in the avalanche danger. The avalanche size is based on its width and length and they are divided into 4 levels: small, medium, large and very large avalanches.
The plot above displays the recorded avalanches with respect to their size. In addition, the black dots represent those avalanches which were triggered when the danger level was 4-High. Interestingly, the size distribution is pretty much independent of the danger level.
Summary
Within this post we have seen there are some avalanche features that correlates more than others. On one hand, we found some tendencies between the snow type, trigger type or AAI with respect to the avalanche danger level. On the other hand, there are other features such as the avalanche size which shows no correlation to the avalanche risk.
Furthermore, a good insight was to prove that even when having a forecasted avalanche risk of 3-Considetable, there are a quite large number of releases, which puts the remark to not underestimate this risk level.
Regarding the data set, I would like to mention that it is collected from visual observations which might be often biased. Therefore, it does not provide a full picture of avalanche activity.
For a more exhaustive description of avalanche danger using the same data set you can have a look to this paper here.
Feel free to check the code from this project in this GitHub repository. Thanks for reading and take care out dare!