Dr. Stephanie Isabelle Mayer

In the rapidly evolving mountain cryosphere, snow avalanches threaten livelihoods, settlements and infrastructure. In this Review, we analyse past and projected impacts of climate change on avalanche activity and the associated risks. The limited availability of comprehensive datasets, the potential confounding factors and the limitations of statistical approaches can make it difficult to identify trends in avalanche activity. However, available data indicate a general decrease in avalanche number, size, seasonality and active paths at low elevations, and an increase in the proportion of wet avalanches relative to dry avalanches. Increased snowfall at high elevations can lead to peaks in avalanche activity and an increase in the number of wet and slush-like avalanches. Activity patterns gradually shift from low to high elevations under continued warming. These changes affect avalanche risk; however, risk is also influenced by factors such as land use and the growth or decline of human settlements. The impact of these factors varies across diverse mountain environments, making it challenging to predict how risk will evolve under a changing climate. Therefore, future research should aim to couple an improved systemic understanding of the impacts of these factors with slope-scale projections of avalanche hazards and risks to support sustainable mountain development and adaptation strategies.

In a recent review of the available scientific literature, impacts of climate change on avalanche activity and the associated risks were analyzed. The work done includes an evaluation of existing data and approaches, a synthesis of past and future changes in avalanche hazards and risks in various mountain environments worldwide with focus on their relation to climate change, elaboration of a conceptual model of trends in avalanche activity and related risks and the formulation of research priorities to address critical knowledge gaps. This paper synthetizes the main outcomes of this review.

Anthropogenic climate change is rapidly altering high mountain environments, including changing the frequency, dynamic behavior, location, and magnitude of alpine mass movements. Here, we review three decades of scientific literature (1995 to early 2024) to assess to what degree observational records from the European Alps – as the region with the most comprehensive records – reveal these changes. We do this for the processes that are most common in this region, namely rockfall, rock avalanches, debris flows, ice avalanches, and snow avalanches. The systematic literature search and review yielded 335 publications, of which we omitted publications that did not focus primarily on observational records. The remaining 103 publications used observations from over 100 sites and 30 inventories to investigate the connection between climate change and mass movements. About one third of the relevant studies found a measurable impact of climate change on the investigated alpine mass movement processes (with the exception of large rock avalanches). The clearest climate-controlled trends are (i) increased rockfall frequency in high-alpine areas due to higher temperatures, (ii) fewer and smaller snow avalanches due to scarcer snow conditions at low and mid elevations, and (iii) a shift towards avalanches with more wet snow and fewer powder clouds. While there is (iv) despite a clear increase in debris-flow triggering precipitation, debris-flow activity has not been found to uniformly increase, though there is some evidence for increasing activity above treeline and at locations without historical precedence. The trends for (v) ice avalanches are spatially very variable with no clear direction. Ice temperatures are measurably increasing, but – despite a theoretical expectation – this has not impacted ice avalanche activity to date. The reviewed literature also reveals that quantifying the impact of climate change on these mass movements remains difficult in part due to the complexities of the natural system, but also because of limitations in the available datasets, confounding effects, and existing statistical processing techniques. Better assessments could be achieved if we would more broadly support the compilation and maintenance of large standardized data catalogs, bring together various dispersed datasets (in said catalogs), including from social and citizen science projects, invest in long-term natural observatories, and develop suitable processing techniques. Better observations will additionally support the development and performance of process-based models. If we can advance natural hazard research on these fronts, more quantitative predictions of future change are well within our reach.

The cryosphere in mountain regions is rapidly transforming due to climate change, yet the impact of these changes on snow avalanche activity remains uncertain. Here, we use a snow cover model driven by downscaled climate projections to evaluate future alterations in dry- and wet-snow avalanche occurrences throughout the 21st century in the Swiss Alps. We assess avalanche activity by employing machine learning models trained with observed records of avalanches. Our findings indicate an overall decline in the occurrence of dry-snow avalanches during the months of December to May that is partially compensated for by an increase in wet-snow avalanche activity. Depending on elevation and the emission scenario considered, we anticipate a net reduction in the total avalanche activity ranging from under 10 % to as much as 60 % by the end of the century compared to 45–75 avalanche days per year at the beginning of the century. Projections further reveal a shift of wet-snow avalanche activity to earlier winter months. Analysis of changes in prominent snow grain types offers a coherent explanation of projected changes beyond a mere decrease in snow depth and snow cover duration. Overall, our study quantifies for the first time the significant influence of climate change on snow avalanche activity in the Swiss Alps and may serve as a benchmark for further mountain regions with similar avalanche climates.

The snow and avalanche climate types maritime, transitional (or intermountain), and continental are well-established in North America, but rarely used in Europe and particularly not known for the Swiss Alps. While the Swiss climatology is well-established for snow precipitation, there is no classification of snow and avalanche climate types. Recent efforts have addressed the lack of consistent avalanche climatology for the European Alps by considering the occurrence of avalanche problem types, expanding beyond the standard snow climate classification based solely on weather variables. Simulated snow stratigraphy obtained with a numerical snow cover model allows taking into account not only meteorological data but also snowpack properties for classification. Following the approach established in North America, we analyzed the data from 128 automated weather stations, in total 2230 winter seasons, to derive the snow climate types maritime, transitional and continental. The snow climate in the Swiss Alps is primarily transitional (45%), with large parts having a strong maritime influence (34%), and a few regions with a continental influence (21%). There are distinct regional differences. On the northern slopes of the Alps, the snow climate is predominantly maritime, less so on the southern slopes. In between, the climate class is mainly transitional, and in some regions of the Engadine and the Valais, the snow climate is to a good part continental. To expand the classification of the snow climates, the snow stratigraphy was analyzed with regard to the avalanche problem types and the frequency of four classes of grain types, notably persistent grain types and melt forms. While in the maritime snow climate we found many melt form layers and few persistent grains, the snowpack classified as continental included primarily persistent grains and rarely showed melt forms. Considering the simulated avalanche problem types suggested that the snowpack in regions with a transitional snow climate is more avalanche-prone than in the maritime and continental snow climates. These findings manifest the large variations seen in snow and avalanche climates across the Swiss Alps. Establishing a closer link between snow climates in the Swiss Alps and avalanche characteristics holds promise for avalanche forecasting, avalanche training and accident analysis.

In recent years, the integration of physical snowpack models coupled with machine-learning techniques has become more prevalent in public avalanche forecasting. When combined with spatial interpolation methods, these approaches enable fully data- and model-driven predictions of snowpack stability or avalanche danger at any given location. This prompts the question: Are such highly detailed spatial model predictions sufficiently accurate for operational use? To explore this, we assess the performance of interpolated, model-based predictions of snowpack stability and avalanche danger, comparing them to human-generated public avalanche forecasts during the 2023/2024 winter season in Switzerland. To do so, we compare human forecasts and model predictions for locations in avalanche terrain (considering coordinates, aspect, elevation) where skiers triggered avalanches (244 events) or which were skied but where no avalanche was triggered (non-events, 3177 data points from GPX tracks). While this data reflects human behavior to some extent, we consider the event ratio as a proxy for the probability of avalanche release due to human load. We observed that with increasing model-predicted danger level or decreasing model-predicted snowpack stability, the event ratio increased. Comparing model predictions with human-made forecasts showed that the predictive performance of two operationally used models was similar to the performance of human avalanche forecasts: both predicted a strong increase in the probability of human-triggered avalanches. In summary, our results indicate that models capture regional patterns of snowpack (in)stability or avalanche danger well, and that these model chains should therefore be systematically integrated in the forecasting process.

Predicting natural avalanche activity and verifying avalanche forecasts have long been major challenges. However, the advent of new technologies, such as radar-based avalanche detection systems and prediction models, are beginning to fill these critical gaps. In this study, we describe the integration of these technologies into our operational workflow, assess their perceived value by forecasters, and provide a quantitative comparison between model predictions and radar detections in southern Valais. A recently developed natural avalanche model (NAM), which predicts natural avalanche activity, was used operationally by the Swiss Avalanche Warning Service in the 2023/24 winter season. The model was displayed next to avalanche observations reported by the observer network and avalanches detected by automated radar systems. This visualization was crucial in helping forecasters develop a better appreciation of the new model. The approach also provided immediate feedback when obvious deviations from predicted and actual conditions became apparent. In this work, after one winter of operational use, we quantitatively evaluated model predictions by comparing these with avalanche activity data from two radar avalanche detection systems in Southern Valais. The analysis confirmed that the natural avalanche model was related to regional avalanche activity, but also highlighted areas for improvement. In particular, after reaching a peak in predicted probability, the model often maintained a high probability while measured avalanche activity had ceased. Our findings suggest that incorporating a trend of predicted probability could improve the model’s accuracy. For further model development, we recommend integrating detection data from multiple sites and over multiple winters. In addition, we recommend the development of automated methods to classify detected avalanches as dry or wet and to link these events to explosives logs. This would greatly enhance the real-time utility of detection data for operational avalanche forecasting services and support ongoing analysis and model refinement.

Assessing snow instability is fundamental in avalanche forecasting. In the past, snow instability was seen as a balance between weak layer shear strength and the overlying load (slab and perhaps a skier). Nowadays, we recognize that this was an oversimplification. Snow instability is best understood in terms of the complex interplay between snow stratigraphy and external loading, influencing the fundamental processes of failure initiation and crack propagation. In short, an avalanche is triggered when a locally initiated failure propagates through a buried weak layer across the slope. While the basic concepts of avalanche release are relatively well understood, there is still no unified physics-based model that can be used for  ractical applications. Hence, machine learning models were developed to assess snow instability. Most of our knowledge on snow stability as well as the models we use for stability assessments come from alpine snow covers, i.e. mid-latitude mountainous regions with seasonal snow covers, such as in the European Alps or in North America. This raises pertinent questions regarding the transferability of these approaches to Arctic regions. Is snow simply snow, or does Arctic snow possess its own secrets?
Arctic snowpacks present unique challenges due to distinct boundary conditions, notably the absence of solar radiation during winter months and the frozen ground. These widely different boundary conditions affect internal snowpack processes with mostly unknown repercussions on snow stability. We therefore use the snow cover model SNOWPACK to explore the transferability of a recently established machine learning model to assess snow stability from alpine to Arctic contexts. First, we assess the viability of our model approach in Arctic regions by comparing stability predictions with local avalanche observations. Subsequently, we undertake a comparative analysis of modelled stability patterns between the Alps and the Arctic, scrutinizing differences and seeking to unravel the essence of Arctic snow stability.

Predicting the timing and size of natural snow avalanches is crucial for local and regional decision makers but remains one of the major challenges in avalanche forecasting. So far, forecasts are generally made by human experts interpreting a variety of data and drawing on their knowledge and experience. Using avalanche data from the Swiss Alps and one-dimensional physics-based snowpack simulations for virtual slopes, we developed a model predicting the probability of dry-snow avalanches occurring in the region surrounding automated weather stations based on the output of a recently developed instability model. This new avalanche day predictor was compared with benchmark models related to the amount of new snow. Evaluation on an independent data set demonstrated the importance of snow stratigraphy for natural avalanche release, as the avalanche day predictor outperformed the benchmark model based on the 3 d sum of new snow height (F1 scores: 0.71 and 0.65, respectively). The averaged predictions of both models resulted in the best performance (F1 score: 0.75). In a second step, we derived functions describing the probability for certain avalanche size classes. Using the 24 h new snow height as proxy of avalanche failure depth yielded the best estimator of typical (median) observed avalanche size, while the depth of the deepest weak layer, detected using the instability model, provided the better indicator regarding the largest observed avalanche size. Validation of the avalanche size estimator on an independent data set of avalanche observations confirmed these findings. Furthermore, comparing the predictions of the avalanche day predictors and avalanche size estimators with a 21-year data set of re-analysed regional avalanche danger levels showed increasing probabilities for natural avalanches and increasing avalanche size with increasing danger level. We conclude that these models may be valuable tools to support forecasting the occurrence of natural dry-snow avalanches.

Climate change strongly affects the seasonal snow cover in mountainous areas worldwide. Warming-induced changes of snowfall and snow cover are expected to influence the frequency and type of snow avalanches, but research on the effect of climate change on avalanches is scarce. To investigate the effect of climate change on avalanche activity in the region of Davos, Switzerland, we simulated snow stratigraphy until the end of the 21^st century using downscaled climate projections for two automatic weather stations at 2147 m a.s.l. and 2540 m a.s.l. Projections were based on eight different climate model chains and three different emission scenarios and snow stratigraphy was simulated with the snow cover model SNOWPACK. We interpreted simulated snow stratigraphy in terms of wet- and dry-snow instability using two recently developed random forest models. Results suggest that, depending on the emission scenario, dry-snow instability expressed as avalanche days will on average decrease by 20 to 60% until the end of the century compared to the reference period (1991-2020). Wet-snow instability, on the other hand, will slightly increase, by up to 20% for the higher-elevation site, and decrease by up to 50% for the lower-elevation site. Detailed analysis of the monthly changes of avalanche activity revealed a shift of wet-snow instability to earlier winter months. Overall, our results contribute to a better understanding of the influence of climate warming on avalanche activity which is crucial to revisit long-standing avalanche risk mitigation strategies.

Recent developments in snow stability modeling demonstrated the great potential of numerical snow cover modeling for avalanche forecasting. The recently developed metrics provided promising results, possibly even superseding traditional stability indices. To further validate these results, we compared the temporal evolution of various stability metrics to a unique data set of measurements including snow stratigraphy, snow microstructure and stability at the Steintälli site (Davos, Switzerland). From January to March 2016, we measured the shear strength of a prominent weak layer of depth hoar crystals with the shear frame and the critical crack length in propagation saw tests. Concurrently, we characterized snow microstructure with the snow micro-penetrometer (SMP) and ran the numerical snow cover model SNOWPACK with the data from the automatic weather station at the site. Our results show that initially, strength and toughness were low, but increased with time; they were correlated. Also, load increased with time, and strength and toughness were correlated with load. While in the field we monitored the most prominent weak layer in the snowpack, the modelled stability metrics indicated that different weak layers, closer to the snow surface, were more prone to fail later in the season. The three stability metrics we compared showed similar temporal trends and performance with regard to avalanche activity in the region of Davos and forecast danger levels. Our detailed site-specific validation indicates that numerical stability metrics provide valuable data for avalanche forecasting though different indices show different behavior at times calling for specific interpretation.

One of the main challenges in avalanche forecasting is the complexity of the snowpack and its interactions with the environment. Traditional methods rely on expert knowledge to analyze snow and weather data, field observations and weather predictions, which is time-consuming and partly subjective. In contrast, data-driven models can analyze large amounts of data faster and may identify patterns that are difficult for humans to detect. Such models are based on statistical or machine learning algorithms that learn from past data to make predictions about new situations. Data-driven models are increasingly used in avalanche forecasting, as they can provide more objective and timely predictions, assisting forecasters in decision-making. Here, we present three recently developed models used in operational avalanche forecasting in Switzerland. The data-driven models use machine-learning algorithms with meteorological and simulated snow stratigraphy data as input to predict (1) the avalanche danger level, (2) snowpack instability and natural avalanche probability, and (3) wet-snow avalanche probability. The models were trained on historical data and typically have an accuracy of about 75\%. During the last three winter seasons, we tested these models in operational avalanche forecasting for the Swiss Alps at SLF. Models 1 and 2 were consulted daily, while model 3 only in potential wet-snow avalanche situations. Preliminary results suggest that the models performed equally well in nowcast mode, when driven with measured data, as in forecast mode, when driven with data from numerical weather prediction models. Overall, the positive feedback we received from the forecasters shows that data-driven models can successfully be integrated into operational forecasting systems.

Modeled snow stratigraphy and instability data are a promising source of information for avalanche forecasting. While instability indices describing the mechanical processes of dry-snow avalanche release have been implemented into snow cover models, there exists no readily applicable method that combines these metrics to predict snow instability. We therefore trained a random forest (RF) classification model to assess snow instability from snow stratigraphy simulated with SNOWPACK. To do so, we manually compared 742 snow profiles observed in the Swiss Alps with their simulated counterparts and selected the simulated weak layer corresponding to the observed rutschblock failure layer. We then used the observed stability test result and an estimate of the local avalanche danger to construct a binary target variable (stable vs. unstable) and considered 34 features describing the simulated weak layer and the overlying slab as potential explanatory variables. The final RF classifier aggregates six of these features into the output probability P_unstable, corresponding to the mean vote of an ensemble of 400 classification trees. Although the subset of training data only consisted of 146 profiles labeled as either unstable or stable, the model classified profiles from an independent validation data set (N=121) with high reliability (accuracy 88 %, precision 96 %, recall 85 %) using manually predefined weak layers. Model performance was even higher (accuracy 93 %, precision 96 %, recall 92 %), when the weakest layers of the profiles were identified with the maximum of P_unstable. Finally, we compared model predictions to observed avalanche activity in the region of Davos for five winter seasons. Of the 252 avalanche days (345 non-avalanche days), 69 % (75 %) were classified correctly. Overall, the results of our RF classification are very encouraging, suggesting it could be of great value for operational avalanche forecasting.

Snow cover models support avalanche forecasting. However, despite much development to derive snow instability from snow cover model output, snow cover models presently do not provide information on avalanche problem types – an essential element to describe avalanche danger. We present an approach to detect, track and assess weak layers in snow cover model output data and assess the related avalanche problem types. We classify different avalanche situations into new snow, wind slab, persistent weak layer and wet-snow problems for either natural or artificial avalanche release. We demonstrate the applicability of this approach with both SNOWPACK and Crocus model output for several winter seasons at Weissfluhjoch, Switzerland, and two sites in Western Canada. At Weissfluhjoch, simulation results from both models coincided reasonably well with avalanche activity recorded with a seismic avalanche detection system. Over a 16-year period both models produced similar frequencies of avalanche problems. At the two Canadian sites frequencies of simulated avalanche problem types agreed fairly well with the observed frequencies. The results provide confidence that avalanche problem types can be simulated in different snow climates and that changes due to climate change can be studied. Moreover, the approach to assess avalanche problem types holds potential to support operational avalanche forecasting.

Forecasting avalanche danger at a regional scale is a largely data-driven yet also experience-based decision-making process by human experts. In the case of public avalanche forecasts, this assessment process terminates in an expert judgment concerning summarizing avalanche conditions by using one of five danger levels. This strong simplification of the continuous, multi-dimensional nature of avalanche hazard allows for efficient communication but inevitably leads to a loss of information when summarizing the severity of avalanche hazard. Intending to overcome the discrepancy between determining the final target output in higher resolution while maintaining the well-established standard of assessing and communicating avalanche hazard using the avalanche danger scale, avalanche forecasters at the national avalanche warning service in Switzerland used an approach that combines absolute and relative judgments. First, forecasters make an absolute judgment using the five-level danger scale. In a second step, a relative judgment is made by specifying a sub-level describing the avalanche conditions relative to the chosen danger level. This approach takes into account the human ability to reliably estimate only a certain number of classes. Here, we analyze these (yet unpublished) sub-levels, comparing them with data representing the three contributing factors of avalanche hazard: snowpack stability, the frequency distribution of snowpack stability, and avalanche size. We analyze both data used in operational avalanche forecasting and data independent of the forecast, going back 5 years. Using a sequential analysis, we first establish which data are suitable and in which part of the danger scale they belong by comparing their distributions at consecutive danger levels. In a second step, integrating these findings, we compare the frequency of locations with poor snowpack stability and the number and size of avalanches with the forecast sub-level. Overall, we find good agreement: a higher sub-level is generally related to more locations with poor snowpack stability and more avalanches of larger size. These results suggest that on average avalanche forecasters can make avalanche danger assessments with higher resolution than the five-level danger scale. Our findings are specific to the current forecast set-up in Switzerland. However, we believe that avalanche warning services making a hazard assessment using a similar temporal and spatial scale as currently used in Switzerland should also be able to refine their assessments if (1) relevant data are sufficiently available in time and space and (2) a similar approach combining absolute and relative judgment is used. The sub-levels show a rank-order correlation with data related to the three contributing factors of avalanche hazard. Hence, they increase the predictive value of the forecast, opening the discussion on how this information could be provided to forecast users.

In mountainous, cold temperate and polar sites, the presence of snow cover can affect relative seismic velocity changes (dV/V) derived from ambient noise correlation, but this relation is relatively poorly documented and ambiguous. In this study, we analyzed raw seismic recordings from a snowy flat field site located above Davos (Switzerland), during one entire winter season (from December 2018 to June 2019). We identified three snowfall events with a substantial response of dV/V measurements (drops of several percent between 15 and 25 Hz), suggesting a detectable change in elastic properties of the medium due to the additional fresh snow. To better interpret the measurements, we used a physical model to compute frequency-dependent changes in the Rayleigh wave velocity computed before and after the events. Elastic parameters of the ground subsurface were obtained from a seismic refraction survey, whereas snow cover properties were obtained from the snow cover model SNOWPACK. The decrease in dV/V due to a snowfall was well reproduced, with the same order of magnitude as observed values, confirming the importance of the effect of fresh and dry snow on seismic measurements. We also observed a decrease in dV/V with snowmelt periods, but we were not able to reproduce those changes with our model. Overall, our results highlight the effect of the snow cover on seismic measurements, but more work is needed to accurately model this response, in particular for the presence of liquid water in the snowpack.

Avalanche occurrences are unambiguous indicators of unstable snow conditions. Information on past and current avalanche activity is therefore crucial for avalanche forecasting. To continuously assess avalanche activity, automatic detection systems are required. In recent years, technological and signal processing advances have led to the development of operational infrasound avalanche detection systems. We evaluated the detection performance of four operationally running infrasound detection systems installed at three different sites in the Swiss Alps during two entire winter seasons. To this end, we collected a comprehensive data set of avalanche activity using a network of automatic cameras and supplementary field observations by local observers. The events automatically identified by the systems were then compared to the data set of visually observed avalanches. Only 3% of the 839 observed avalanches were associated with automatic detections and 21% of the automatic detections were confirmed by field observations. However, the majority of observed avalanches were small and most automatic detections occurred during periods of poor visibility. Furthermore, the probability of detection (POD) increased with avalanche size, decreased with distance, and varied with avalanche type. Dry-snow avalanches were generally better detected than wet-snow and mixed-type avalanches. Large avalanches (on the order of 300 m wide and 1000 m long) within a distance of 3 km from the array were typically well detected (POD ≈ 90%). The false alarm ratio was estimated to 13-30%.

Avalanche occurrences are unambiguous indicators of unstable snow conditions. Information on past and current avalanche activity is therefore crucial for avalanche forecasting. To continuously assess avalanche activity, automatic detection systems are required. In recent years, technological and signal processing advances have led to the development of operational infrasound avalanche detection systems. We evaluated the detection performance of four operationally running infrasound detection systems installed at three different sites throughout the Swiss Alps during two entire winter seasons. To this end, we collected a comprehensive data set of avalanche activity using a network of automatic cameras and supplementary field observations by local observers. The events automatically identified by the systems were then compared to the data set of visually observed avalanches. Only 3% of the observed avalanches were associated with automatic detections and 22% of the automatic detections were confirmed by field observations. However, the majority of observed avalanches were small and most automatic detections occurred during periods of poor visibility. Furthermore, the probability of detection (POD) increased with avalanche size and decreased with distance. Large avalanches (on the order of 100 m wide and 1000 m long) within a distance of 3-4 km from the array were typically well detected (POD ~ 80%). The false alarm ratio was estimated to 10- 30%.