How is avalanche danger described in textual descriptions in avalanche forecasts in Switzerland? Consistency between forecasters and avalanche danger

Effective and efficient communication of expected avalanche conditions and danger to the public is of great importance, especially where the primary audience of forecasts are recreational, non-expert users. In Europe, avalanche danger is communicated using a pyramid, starting with ordinal levels of avalanche danger and progressing through avalanche-prone locations and avalanche problems to a danger description. In many forecast products, information relating to the trigger required to release an avalanche, the frequency or number of potential triggering spots, and the expected avalanche size is described exclusively in a textual danger description. These danger descriptions are, however, the least standardized part of avalanche forecasts. Taking the perspective of the avalanche forecaster and focusing particularly on terms describing these three characterizing elements of avalanche danger, we investigate first which meaning forecasters assign to the text characterizing these elements and second how these descriptions relate to the forecast danger level. We analyzed almost 6000 danger descriptions in avalanche forecasts published in Switzerland and written using a structured catalogue of phrases with a limited number of words. Words and phrases representing information describing these three elements were labeled and assigned to ordinal classes by Swiss avalanche forecasters. These classes were then related to avalanche danger. Forecasters were relatively consistent in assigning labels to words and phrases with Cohen's kappa values ranging from 0.64 to 0.87. Avalanche danger levels were also described consistently using words and phrases, with for example avalanche size classes increasing monotonically with avalanche danger. However, especially for danger level 2 (moderate), information about key elements of avalanche danger, for instance the frequency or number of potential triggering spots, was often missing in danger descriptions. In general, the analysis of the danger descriptions showed that extreme conditions are described in more detail than intermediate values, highlighting the difficulty of communicating conditions that are neither rare nor frequent or neither small nor large. Our results provide data-driven insights that could be used to refine the ways in which avalanche danger could be communicated. Furthermore, through the perspective of the semiotic triangle, relating a referent (the avalanche situation) through thought (the processing process) to symbols (the textual danger description), we provide an alternative starting point for future studies of avalanche forecast consistency and communication.

Data-driven automated predictions of the avalanche danger level for dry-snow conditions in Switzerland

Even today, the assessment of avalanche danger is by large a subjective, yet data-based decision-making process. Human experts analyze heterogeneous data volumes, diverse in scale, and conclude on the avalanche scenario based on their experience. Nowadays, modern machine learning methods and the rise in computing power in combination with physical snow cover modelling open up new possibilities for developing decision support tools for operational avalanche forecasting. Therefore, we developed a fully data-driven approach to predict the regional avalanche danger level, the key component in public avalanche forecasts, for dry-snow conditions in the Swiss Alps. Using a large data set of more than 20 years of meteorological data measured by a network of automated weather stations, which are located at the elevation of potential avalanche starting zones, and snow cover simulations driven with these input weather data, we trained two random forest (RF) classifiers. The first classifier (RF #1) was trained relying on the forecast danger levels published in the avalanche bulletin. Given the uncertainty related to a forecast danger level as a target variable, we trained a second classifier (RF #2), relying on a quality-controlled subset of danger level labels. We optimized the RF classifiers by selecting the best set of input features combining meteorological variables and features extracted from the simulated profiles. The accuracy of the danger level predictions ranged between 74 % and 76 % for RF #1, and between 72 % and 78 % for RF #2, with both models achieving better performance than previously developed methods. To assess the accuracy of the forecast, and thus the quality of our labels, we relied on nowcast assessments of avalanche danger by well-trained observers. The performance of both models was similar to the accuracy of the current experience-based Swiss avalanche forecasts (which is estimated to 76 %). The models performed consistently well throughout the Swiss Alps, thus in different climatic regions, albeit with some regional differences. A prototype model with the RF classifiers was already tested in a semi-operational setting by the Swiss avalanche warning service during the winter 2020-2021. The promising results suggest that the model may well have potential to become a valuable, supplementary decision support tool for avalanche forecasters when assessing avalanche hazard.

Avalanche danger forecast in little study territories using GIS

Methods for determining the zone of avalanche initiation using tools are proposed ArcGis 10.4.1. The process of creating a raster of the slope of heights and later on the territory itself for dividing them into existing gradations is described. Demonstrated the process of determining the area of avalanche foci and determining the boundaries of the avalanche collection. The areas of avalanche origination within the avalanche collection have been corrected. The method of creating a map of the exposure of slopes in order to identify the area of accumulation and drift of snow, its dynamics and characteristics of snow melting is shown. Based on this, it becomes possible to determine the maximum height of avalanche initiation and the range of the avalanche ejection.

Avalanche danger level characteristics from field observations of snow instability

Avalanche danger levels are described in qualitative terms that mostly are not amenable to measurements or observations. However, estimating and improving forecast consistency and accuracy require descriptors that can be observed or measured. Therefore, we aim to characterize the avalanche danger levels based on expert field observations of snow instability. We analyzed 589 field observations by experienced researchers and forecasters recorded mostly in the region of Davos (Switzerland) during 18 winter seasons (2001–2002 to 2018–2019). The data include a snow profile with a stability test (rutschblock, RB) and observations on snow surface quality, drifting snow, signs of instability and avalanche activity. In addition, observers provided their estimate of the local avalanche danger level. A snow stability class (very poor, poor, fair, good, very good) was assigned to each profile based on RB score, RB release type and snowpack characteristics. First, we describe some of the key snowpack characteristics of the data set. In most cases, the failure layer included persistent grain types even after a recent snowfall. We then related snow instability data to the local avalanche danger level. For the danger levels 1–Low to 4–High, we derived typical stability distributions. The proportions of profiles rated poor and very poor clearly increased with increasing danger level. For our data set, the proportions were 5 %, 13 %, 49 % and 63 % for the danger levels 1–Low to 4–High, respectively. Furthermore, we related the local avalanche danger level to the occurrence of signs of instability such as whumpfs, shooting cracks and recent avalanches. The absence of signs of instability was most closely related to 1–Low and the presence of them to 3–Considerable. Adding the snow stability class and the 3 d sum of new snow depth improved the discrimination between the lower three danger levels. Still, 2–Moderate was not well described. Nevertheless, we propose some typical situations that approximately characterize each of the danger levels. Obviously, there is no single easily observable set of parameters that would allow us to fully characterize the avalanche danger levels. One reason for this shortcoming is the fact that the snow instability data we analyzed usually lack information on spatial frequency, which is needed to reliably assess the danger level.

How is avalanche danger described in public avalanche forecasts? Analyzing textual descriptions of avalanche forecasts in Switzerland

Efficient communication in public avalanche forecasts is of great importance to clearly inform and warn the public about expected avalanche conditions. In Europe, avalanche danger is communicated using a pyramid, starting with ordinal categories of avalanche danger, and progressing through avalanche-prone locations and avalanche problems to a danger description. In many forecast products, information relating to the trigger required to release an avalanche, the frequency or number of potential triggering locations, and the expected avalanche size, are described exclusively in the danger description. These danger descriptions are, however, the least standardized part of avalanche forecasts. Taking the perspective of the avalanche forecaster, and focusing particularly on terms describing these three characterizing elements of avalanche danger, we investigate firstly which text symbols are used to describe these elements, and secondly how these descriptions relate to the forecast danger level. We do so through the perspective of the semiotic triangle, relating a referent (the avalanche situation) through thought (the processing process) to symbols (the textual danger description). We analyzed almost 6000 danger descriptions in avalanche forecasts published in Switzerland and written using a structured catalog of phrases with a limited number of words. Text symbols representing information describing these three elements were labeled and assigned to ordinal classes by Swiss avalanche forecasters. These classes were then related to avalanche danger. Forecasters were relatively consistent in assigning labels to words and phrases with Cohen's Kappa values ranging from 0.67 to 0.87. Nonetheless, even experts were not in complete agreement about the labeling of terms and were less likely to agree on terms not used in official definitions. Avalanche danger levels were categorized relatively consistently using words and phrases, with for example avalanche size classes increasingly monotonically with avalanche danger. However, especially for danger level 2-Moderate, information about key elements was often missing in danger descriptions. In general, the analysis of the danger descriptions showed that extreme conditions are more frequently described in detail than intermediate values, highlighting the difficulty of communicating conditions that are neither rare nor frequent, or neither small nor large. Our results provide data-driven insights that could be used to refine the ways in which avalanche danger could and should be communicated, especially to recreationalists, and provide a starting point for future studies of how users interpret avalanche forecasts.

A random forest model to assess snow instability from simulated snow stratigraphy

<p>Numerical snow cover models enable simulating present or future snow stratigraphy based on meteorological input data from automatic weather stations, numerical weather prediction or climate models. To assess avalanche danger for short-term forecasts or with respect to long-term trends induced by a warming climate, modeled snow stratigraphy has to be interpreted in terms of mechanical instability. Several instability metrics describing the mechanical processes of avalanche release have been implemented into the detailed snow cover model SNOWPACK. However, there exists no readily available method that combines these metrics to predict snow instability.</p><p>To overcome this issue, we compared a comprehensive dataset of almost 600 manual snow profiles with SNOWPACK simulations. The manual profiles were observed in the region of Davos over 17 different winter seasons and include a Rutschblock stability test as well as a local assessment of avalanche danger. To simulate snow stratigraphy at the locations of the manual profiles, we interpolated meteorological input data from a network of automatic weather stations. For each simulated profile, we manually determined the layer corresponding to the weakest layer indicated by the Rutschblock test in the corresponding observed snow profile. We then used the subgroups of the most unstable and the most stable profiles to train a random forest (RF) classification model on the observed stability described by a binary target variable (unstable vs. stable).</p><p>As potential explanatory variables, we considered all implemented stability indices calculated for the manually picked weak layers in the simulated profiles as well as further weak layer and slab properties (e.g. weak layer grain size or slab density).  After selecting the six most decisive features and tuning the hyper-parameters of the RF, the model was able to distinguish between unstable and stable profiles with a five-fold cross-validated accuracy of 88%.</p><p>Our RF model provides the probability of instability (POI) for any simulated snow layer given the features of this layer and the overlying slab. Applying the RF model to each layer of a complete snow profile thus enables the detection of the most unstable layers by considering the local maxima of the POI among all layers of the profile. To analyze the evolution of snow instability over a complete winter season, the RF model can provide the daily maximal POI values for a time series of snow profiles. By comparing this series of POI values with observed avalanche activity, the RF model can be validated.</p><p>The resulting statistical model is an important step towards exploiting numerical snow cover models for snow instability assessment.</p>