It is well established in the risk communication literature that having an in-depth understanding of the characteristics and needs of the target audience is critical for the design of risk communication products and services that effectively address user needs. However, the avalanche safety community conventionally characterizes backcountry users predominantly by activity type and level of avalanche safety training. While helpful, this characterization provides only a very narrow perspective and is of limited use for the design and evaluation of avalanche safety information products. This is because it excludes integral contextual information on people’s backcountry experience preferences, personal circumstances, and avalanche safety needs. This poses a challenge as avalanche safety professionals navigate communicating with an ever growing backcountry recreation community. 

We propose a more holistics and informative approach to understanding and characterizing backcountry recreationists. In addition to activity and avalanche safety training, our characterization includes detailed information on recreationists’ terrain preferences, motivations and reasons for engaging in winter backcountry activities, experience level, trip planning practices, preferences for when and where they recreate, and basic socio-demographic information. Collectively, this information provides a much richer picture of recreationists’ avalanche safety needs; spotlights whether existing practices are adequate for specific desired experiences; and identifies where, when, and how to best communicate with certain cohorts of recreationists.

To illustrate our approach, we present data from the Euregio and Swiss avalanche forecast research panels (see related submission by Haegeli et al.), which was collected over the last three winters (n = 1975). After presenting an overview of the responses to individual questions, we discuss the results of various latent class analyses that identify common patterns in participants’ responses. These results create a foundation for evidence-based and context-specific backcountry user personas, a concept often utilized by marketers and developers to describe the characteristics of typical product users. For example, our detailed characterization allows us to distinguish between committed but untrained backcountry skiers interested in skiing aggressive terrain and reaching challenging summits from more casual backcountry skiers who only go into the backcountry during their holidays to relax and spend time in nature. Obviously, these two backcountry skier personas have very different avalanche safety needs and will respond to products differently. 

Hence, personas can be an effective tool for providing important context, identifying and describing at-risk cohorts, and designing more targeted avalanche safety information products, course curricula, and outreach campaigns. The concept of personas can also be used for targeted recruitment of research participants, meaningful characterizations of study participants, and provide an approachable way of presenting results.

Overall, our research aims to promote a more holistic understanding of the growing and increasingly diverse recreation community that can more effectively inform the development and evaluation of avalanche safety information products and initiatives.

The avalanche industry has long engaged with questions about public safety and the effectiveness of established risk communication and education services. Traditionally, avalanche experts have focused on accident data and fatality statistics (e.g., Greene et al., 2006; Niemann, Paul, & Rahman, 2022; Peitzsch et al., 2018; Techel, Zweifel, & Winkler, 2014). While fatality reports provide insightful case studies, accident data alone are fundamentally limited. First, the calculation of meaningful accident rates suffers from the lack of reliable estimates for the number of backcountry users, which are difficult and cost prohibitive. (Campbell & Haegeli, 2022; Langford, Haegeli, & Rupf, 2020). Second, reliable avalanche accident information is typically only available for fatal events. Thus, the resulting picture of public avalanche risk management promotes a knowledge deficit model, which spotlights failures and attributes them to a deficiency in individuals rather than limitations in public safety services (Simis, Madden, Cacciatore, &Yeo, 2016). To better target how well public avalanche safety services reduce the risk for the recreating public, we need to engage a different approach that guides us toward asking more constructive questions and targets improvements in services.

What is missing from the avalanche industry’s tool kit is a foundational, applied framework for systematically planning and evaluating public avalanche safety services. In this study, we aim to address this industry gap using a literature review. This approach allows us to look at the specifics of the avalanche context through various pertinent theoretical lenses and re-visit and define the fundamentals, including: 1) explicitly defining the public avalanche safety problem and the primary objective of services, 2) comprehensively detailing the problem landscape, and 3) identifying evaluative approaches that are best suited to the avalanche industry’s primary objective within the problem landscape.

Our framework is grounded in several established bodies of theoretical and practical work. To outline the purpose of public avalanche safety services and define clear objectives, we draw from the disaster risk reduction and risk governance literature, such as the Sendai Framework (United Nations Office for Disaster Risk Reduction, 2015). We look to Social-Ecological Systems theory (Berkes & Folke, 1998) to understand the nature of the problem, detail the problem landscape in our human-environment system, and organize its complexity from macro to micro scales across the multiple dimensions. To outline the components of avalanche risk management within the human-environment system, we refer to ISO 31:000 (ISO, 2018) and existing avalanche research. For guidance on evaluative approaches, we borrow concepts from public health research and community-engaged research practices to outline a process for conducting needs assessments and co-defining evaluative criteria that can be empirically tested.

As Carl Jung (1968) states, “to ask the right question is already half the solution to a problem.” Asking questions that target the effectiveness of services remains at the forefront of where the avalanche industry can make improvements. Our framework aims to provide a practical guide for informing more targeted and constructive questions to track and evaluate improvements.

Avalanche airbags can increase survival in an avalanche by preventing a critical burial or decreasing burial depth. Research has evaluated the effectiveness of this safety device for mortality reduction and their impact on risk attitudes and behaviors, but information is sparse on quantifying airbag ownership among the winter backcountry recreationists. The most comprehensive perspective to-date has been provided by Procter et al. (2014), who surveyed avalanche safety equipment use among recreationists in Northern Italy in 2011. Their analysis showed that at that time less than 4% of recreationists used avalanche airbags. However, there have been substantial developments in this space over the last 12 years, and the goal of our work is to provide an up-to-date perspective on the prevalence of airbag use and the characteristics of airbag users.  

We are using information from the Euregio and Swiss avalanche forecast research panels, two databases of avalanche forecast users interested in regularly participating in avalanche safety research. The signup survey for these research panels include detailed questions on recreational and professional activities in avalanche terrain, personal motivations and backgrounds, and avalanche safety practices including the use of safety equipment.

Focusing on research panel members with complete signup information (n=2154), we find that approximately one in three backcountry recreationists own an airbag (34.3%, n=739). Given that our sample is likely biased toward recreationists more engaged in avalanche safety, this result  asserts that avalanche airbags have become a common risk management tool, but not a safety standard among recreational backcountry users. 

Highest proportion of ownership is among 25-44 year old males. Education and experience levels did not show a significant relationship with ownership. However, when analyzing the data using conditional inference trees, we found compelling connections. Out-of-bound skiers and riders that are highly motivated by powder own airbags more commonly than ones that are less motivated by it. Alternately, skiers that have low motivation to pursue risk also own airbags at a higher than average proportion. Hence, airbag ownership can be associated with both risk-seeking and risk-avoidance strategies. 

Understanding the characteristics of avalanche airbag owners can improve the future design and guide appropriate promotion of this technology. Additionally, avalanche airbag information is valuable to avalanche educators who can pass relevant avalanche safety knowledge to backcountry users – who can then decide for themselves if owning an airbag is a right choice for their mountain adventures.

The influence of human perception, judgement, and behaviour on avalanche safety is well known and has been described in both the academic and applied literature for the long time. Early examples include Ed LaChapelle’s (1980) description of the psychological context of conventional avalanche forecasting, Jill Fredston and Doug Fesler’s (1994) listing of human factors contributing to accidents, and Munter’s (1992) chapter on “Dreizehn fatale Irrtümer des gesunden Menschenverstandes” (Thirteen fatal errors of common sense). One of the most influential works on human factors in the avalanche safety community has been the introduction of the concept of heuristics traps by Ian McCammon (2002), which illustrates how the subconscious use of well-known mental shortcuts identified by cognitive psychologists can lead to unsafe decisions and accidents in the backcountry. More than 20 years later, McCammon’s FACETS mnemonic remains the main reference for how our community conceptualizes human factors and the tool of choice for introducing the topic to recreationists (e.g., Tremper, 2018; Floyer & Robine, 2018; SLF, 2018).

While McCammon’s contribution to our community’s awareness of human factors is uncontested, there is much more to the human dimension of avalanche safety than just highlighting flaws in human decision-making. And even though the number of social science studies in our field has grown considerably in recent years, it is our opinion that the current perspective is rather limited, and many key questions remain unanswered or even unidentified. As a result, the contribution of social science research to risk management practices, curricula and product development is far from reaching its potential. Complementing the systematic overview of the existing research on human factors in avalanche terrain presented by Hetland et al. (this ISSW), our work provides a more critical review of this body of research by examining its particular perspectives, scientific rigor, and practical potential. We discuss common limitations of the existing body of work and reflect on the factors that seem to promote or hinder the adoption of social science research insights in our community. Combining our reflections with the experiences in neighboring interdisciplinary domains where the social sciences are more well established (e.g., risk communication, public health), we introduce a set of guiding principles and best practices to strengthen the quality, insight and relevance of social science research in our field.

We hope that our presentation will broaden our community’s understanding of the value of the social sciences, provide practitioners and researchers with tangible tools for recognizing high-quality research, and ultimately promote a more effective use of social science research in our community that can lead to more evidence-based decisions for the design and evaluation of avalanche safety information products and course curricula.

Having an in-depth understanding how backcountry recreationists comprehend and use avalanche safety products is critical for making informed decisions about how to improve these services or develop new ones. While there is a growing body of social science research examining the characteristics and avalanche risk management practices of backcountry users, the impact of this research on product development decisions has been limited so far. Possible reasons for this include study results that cannot be implemented directly, the fact that the design and analysis of insightful studies take considerable time that typically is not accounted for in development cycles, and the time consuming and challenging recruitment of participants beyond power-users.

Our work introduces the infrastructure we have developed to address these challenges and facilitate a more integrated and responsive contribution of social science research to avalanche safety information product design and evaluation. Our system consists of i) a research panel, ii) a standardized set of signup questions, and iii) a template for targeted research surveys. The research panel is a custom-built database of community members interested in regularly participating in avalanche risk communication research. A signup website allows for the continuous recruitment of participants through various channels to ensure the research panel is sufficiently large and meaningfully represents the increasingly diverse community of avalanche forecast users. The standardized signup questions facilitate a comprehensive characterization of panel members. This can be used for identifying target audiences for specific research studies or providing important context for the analysis of research surveys. The research survey template streamlines community engagement on specific topics by allowing members of the research panel to complete abbreviated survey versions focused on the key research questions, while other members of the backcountry communities complete the survey with a small set of crucial background questions. The full set of signup questions is integrated at the end of the research survey template to facilitate easy recruitment to the research panel at the same time.

The objective of our system is to allow avalanche warning services to deploy shorter and more targeted research surveys more frequently, but still have the scientific rigor necessary to provide informative insight. If implemented correctly, the system should facilitate a more continuous and structured two-way communication between avalanche warning services and their users, as well as more efficient collaborations between avalanche warning services and social scientists.

In addition to describing the components of our system in detail, we will elaborate on the lessons learned during our two-year collaboration. We hope that our experiences and reflections will make the integration of social science research easier for other warning services and help move our community towards a more evidence-based, community-engaged, and user-centered approach to avalanche safety information product design and evaluation.

To help recreationists make informed decisions about when and where to travel in the backcountry, avalanche warning services in many countries publish daily avalanche forecasts describing the nature of current and expected conditions. Several recent studies have started to examine forecast users’ ability to understand the provided hazard information, but they have so far not evaluated how users combine the information with additional avalanche knowledge to assess the severity of the conditions on individual slopes, which is a critical skill for the effective application of the forecast information during trip planning.

To provide insight into this knowledge gap, we conducted an online experiment with members of the Euregio (Tyrol, South Tyrol and Trentino) and Swiss avalanche forecast research panels where participants were presented with a series of hypothetical avalanche forecasts and asked to rank four slopes according to their avalanche hazard. Each slope was characterized by a different combination of aspect and elevation, both shown on a simplified map, and slope steepness, which was described using the standard qualitative terms defined by the European Avalanche Warning Services (EAWS). While the Graphical Reduction Methods (GRM) provided a reference for defining the correct answers, the more interesting aspect of our experiment is that it was designed in a way that the response patterns would reveal participants’ personal assessment approaches. To further complement the insights from this experiment, our survey included several debrief questions and a set of questions examining participants’ understanding of the qualitative steepness terms.

Our analysis of 4290 assessments from 2145 participants revealed that only 16% of the sample provided the “correct” GRM solution. Fifty-five percent used a sequential approach where they first split the slopes according to the provided danger rating and then ranked them according to steepness. While this approach is systematic, it does not always result in a meaningful assessment. We also found that many participants ignored the aspect information of the avalanche conditions (up to 40% of cases). The analysis of our questions onthe qualitative steepness terms showed that approximately half of our participants believe that ‘extreme terrain’ starts at slope inclines that are steeper than the 40 degree threshold defined by EAWS. This means that they potentially underestimate the severity of the terrain described in forecasts. Participants older than 45 years who spend more than 20 days in the backcountry each winter had a significantly higher proportion of individuals with more errors in the interpretation of the steepness terms but still high confidence in their answers.

The results of this study contribute to a growing body of research that aims to provide a user perspective on the effectiveness of avalanche risk communication products by examining how backcountry recreationists understand and apply avalanche forecast information. We conclude with a discussion of the practical implications of our study for the design of avalanche forecast products and messages.

Avalanche forecasters typically work in teams, relying on the continuity of their shared understanding of avalanche conditions throughout a season. Shift changes can disrupt this continuity as incoming forecasters need time to get up to speed with current conditions. While the conceptual model of avalanche hazard offers a structured way to describe existing conditions, it lacks important contextual information about assessment decisions, the evidence used, and critical considerations for forthcoming days. Without this context, forecasters can be left searching for clarifying information that could otherwise have been communicated. Handoff notes can provide this context but can be cumbersome to write because they require additional effort outside of the forecasters’ core hazard assessment process.

Collaborative Visual Analytics offers practical methods to address handoff challenges by streamlining the process of collecting and organizing contextual information within existing analytic workflows, easing the burden of note-taking. While these methods have already demonstrated their value in domains like healthcare and investigative intelligence analysis, their effectiveness critically depends on them being appropriately integrated into a specific context of application. In collaboration with Avalanche Canada, we developed and evaluated early design prototypes in practice to understand how they might perform operationally and to derive design guidelines for successful implementation.

We identified common themes in how forecasters discuss avalanche conditions and used these as a template for a lightweight screen capture tool, assisting forecasters in gathering and organizing contextual information as they conduct hazard assessments. Additionally, we explored how the commonly used representative snowprofile diagram or the structure of hazard assessments themselves could be employed as an organizing framework for communicating contextual information in a digital handoff support tool.

We discuss the relative strengths and weaknesses of alternative approaches and offer general design guidelines for tools that could improve handoff efficiency in a variety of avalanche forecasting contexts. Overall, forecasters thought our tools were easy to integrate into their daily workflows and helpful in planning and coordinating their work. This research highlights how dedicated digital handoff support tools could improve the efficiency and efficacy of team collaboration in avalanche forecasting.

Avalanche forecasting is primarily a human judgment process where a wide range of observations is synthesized into an overall picture of the nature and severity of avalanche hazard. While the conceptual model of avalanche hazard (CMAH; Statham et al., 2018) introduced a common language and qualitative framework for assessing avalanche hazard, operational experience has shown that there are considerable differences in how the CMAH and the concept of avalanche problems are applied by avalanche forecasters (Statham et al., 2018, Techel et al., 2018, Horton et al., 2020b, Hordowick, 2022). Since these inconsistencies can lead to serious miscommunications among forecasters themselves and with the recreational backcountry community, there is a need for improving the consistency and quality of the operational use of the CMAH.

Snowpack simulations offer the opportunity to extend the qualitative framework of the CMAH with quantitative links between weather, snowpack, and hazard, and thereby provide avalanche forecasters with an independent, reproducible perspective on the intricate details of avalanche problems. Existing research in this area include Reuter et al. (2021), who established a prescriptive approach to modeling avalanche problem types at individual grid points purely based on physical simulations of snow instability. More recently, Mayer et al. (in review) developed statistical models for predicting the probability and size of dry-snow avalanches in the proximity of weather sites used for snow stratigraphy simulations based on verified data sets of natural avalanche activity and stability tests related to human triggered avalanches. Both of these studies clearly demonstrate the potential of snowpack models for providing avalanche problem information.

The present study expands on these ideas by demonstrating how avalanche problem information can be extracted from spatially distributed snowpack simulations while preserving knowledge about individual layers across space and time. We tailor the output of the simulations to the needs of the North American practitioner community by mirroring the framework of the CMAH. Simulated hazard charts summarize the numerical information in ways that allow forecasters to quickly grasp the contribution of different regional-scale layers to the overall hazard at a given day. We then examine the agreement between simulations and human assessments for persistent and storm slab avalanche problem situations to simultaneously explore the capabilities of the model chain and gain further insight into the strengths and weaknesses of the human assessments. To do this, we identify avalanche problem periods with good and poor agreement and examine the effects of various context variables.

This paper contributes to a growing body of research that aims to enhance the operational value of snowpack simulations and provides insight in how snowpack simulations can help address some of the operational challenges of applying avalanche problems.

Snowpack models can provide detailed and continuous insight about the evolution of the snow stratigraphy in ways that are not possible with direct observations. However, the volume of data generated by the simulations can easily become overwhelming, and since the simulated snow profiles are characterized by a rather complex, multidimensional data format, it is challenging to analyze the rich information manually. The available information is therefore commonly reduced to bulk properties and summary statistics of the entire snow column or individual grid points. This is only of limited value for operational avalanche forecasting where knowledge about thin, critical avalanche layers plays an important role. In our opinion, the lack of efficient ways to access and mine large numbers of snow profiles is one of the key reasons for the limited operational use of spatially distributed snowpack simulations and ensemble systems.

We discuss recently developed tools for numerically processing snow profiles to make large volumes of snowpack model output more accessible for practitioners in relevant ways. This includes algorithms that compare and assess generic snow profiles by matching corresponding layers and aligning them before effectively synthesizing many profiles into a meaningful overall perspective (Fig. 1a, c). Our approach enables the compiling of informative summary statistics and distributions of snowpack layers (Fig. 1 b, d, e), as well as the dynamic clustering of profiles into groups with distinct conditions. Our algorithms are based on customized versions of Dynamic Time Warping (DTW) and DTW Barycenter Averaging (DBA), well established methods in the data sciences.

While some of these customized methods have already been published in academic journals and are publicly available in R software packages, this paper will focus on the tangible application of these tools for avalanche forecasters, which includes a discussion of our vision for using these new opportunities for supporting operational grouping of forecast micro-regions and meaningfully mining ensemble snowpack simulations.

Snow avalanches are the primary mountain hazard for mechanized skiing operations. Helicopter and snowcat ski guides are tasked with finding safe terrain to provide guests with enjoyable skiing in a fast-paced and highly dynamic and complex decision environment.  Based on years of experience, ski guides have established systematic decision-making practices that streamline the process and limit the potential negative influences that can occur due to time pressure and emotional investment. Their decision-making process can be broken into three levels: (1) deciding which runs are open for the day, (2) selecting the runs they ski based on the options that are open, (3) choosing how to manage the slope scale avalanche exposure within the selected runs. These three levels of decision-making consider different factors and take place on different spatial scales. While guiding teams pass on this expertise through mentorship, the current lack of a quantitative description of the process prevents the development of decision aids. To address this knowledge gap, we collaborated with guides at Canadian Mountain Holidays (CMH) Galena Lodge to catalog and analyze their multi-stage decision-making process using GPS tracking, automated avalanche hazard indication mapping, and Bayesian Networks.

In order to gain a deeper understanding of guide’s real world decision-making we include numerous datasets in our decision-making models to try and capture both the avalanche hazard factors as well as the logistical or operational factors that drive daily decisions. We leveraged operational data from CMH Galena that captures historic records of opening and closing runs, daily ski run usage, avalanche hazard conditions, and field observations to build decision-making models for each level of the process. Bayesian Networks are well suited to modeling complex decision-making scenarios because input from field experts can be incorporated in the design of the model structure and observed data from past decisions can be used to determine the prior distributions of the decision-making network. Building these models based on observed data allows us to quantitatively describe the guiding practices of CMH guides over the past eight seasons and document some of their institutional risk management knowledge.

Our results include 14,757 GPS tracks collected over a period of eight seasons (2015/16 – 2022/23). The GPS tracks provide precise spatial information about where the guides are traveling under a variety of avalanche hazard conditions. To describe the severity of avalanche terrain numerically we use high resolution avalanche hazard indication mapping methods to estimate forest cover, potential avalanche release areas, and simulate avalanche runouts. To understand the guide’s perspective of the terrain we also collected survey data, which provides critical context for the operational perspective of the decision-making process. Our results quantify some of the decision rules employed by professional guides under different conditions. These insights provide a baseline for development of future decision-making tools that can offer independent perspectives on operational terrain choices based on historic patterns or as a training tool for newer guides.

Over the past 20 years, avalanche safety operations in North America adopted common standards for assessing avalanche hazard, including the North American Public Avalanche Danger Scale, avalanche problems, and the conceptual model of avalanche hazard. While these systems aim to provide a consistent, structured way to assess and communicate avalanche hazard, practical experience and several studies have shown that there is considerable variability in how the standards are understood and applied. To examine the use and effectiveness of these standards more systematically, the Simon Fraser University Avalanche Research program conducted several projects focused on how avalanche forecasters apply danger ratings and avalanche problems. This paper synthesizes the key results of this research, with a focus on the practical implications for avalanche forecasters. We begin with a brief review of the individual projects, which span from statistical modelling of large datasets to qualitative interviews with forecasters. We then synthesize common themes found throughout the research, which include inconsistencies between forecasters, limitations of the published standards, and complications arising from the contextual nature of hazard assessments. Recommendations from these studies include improving published standards, developing and implementing evidence-based decision aids, increasing the level of training and communication amongst forecasters, and further research into how danger ratings and problems are used to manage risk in different contexts. We conclude the paper with thoughts on how the recommendations could be implemented to increase the quality and consistency of avalanche problem and danger ratings in the future.

Physical snowpack modelling has become an operational forecasting tool used by Avalanche Canada over the last decade. This paper reflects on the process of adopting snowpack models into the forecasting workflow, highlighting both successful implementation and challenges faced. Additionally, we present an outlook on future developments based on feedback from forecasters who have used the tools. The adoption process involved close collaboration between the SFU avalanche research program and other avalanche forecasting operations. Collaborative efforts focused on developing computer infrastructure to run SNOWPACK at a regional scale, designing effective ways to present model output, and delivering ongoing training. Over time, gradual exposure to this new source of information increased trust, especially after specific cases where the model offered additional insights into snowpack conditions that traditional data sources could not provide. However, limitations in understanding model uncertainty and the lack of meaningful verification data currently limit the weight placed on model predictions, leading to skepticism. To address this issue, future efforts should integrate the models with traditional data sources and establish workflows to regularly monitor model output and facilitate real-time validation. Despite these challenges, physical snowpack models have the potential to improve the accuracy and reliability of avalanche forecasting. The insights gained from this process may be applicable to the adoption of other types of new technologies into forecasting programs.