Sources of Power: How People Make Decisions

International snow science workshop

Snow avalanche hazard prediction using machine learning methods

This conceptual model of avalanche hazard identifies the key components of avalanche hazard and structures them into a systematic, consistent workflow for hazard and risk assessments. The method is applicable to all types of avalanche forecasting operations, and the underlying principles can be applied at any scale in space or time. The concept of an avalanche problem is introduced, describing how different types of avalanche problems directly influence the assessment and management of the risk. Four sequential questions are shown to structure the assessment of avalanche hazard, namely: (1) What type of avalanche problem(s) exists? (2) Where are these problems located in the terrain? (3) How likely is it that an avalanche will occur? and (4) How big will the avalanche be? Our objective was to develop an underpinning for qualitative hazard and risk assessments and address this knowledge gap in the avalanche forecasting literature. We used judgmental decomposition to elicit the avalanche forecasting process from forecasters and then described it within a risk-based framework that is consistent with other natural hazards disciplines.

/pdf/a-conceptual-model-of-avalanche-hazard-4demv5axy2.pdf

A Conceptual Model of Avalanche Hazard

This study sought to produce an accurate multi-hazard risk map for a mountainous region of Iran. The study area is in southwestern Iran. The region has experienced numerous extreme natural events in recent decades. This study models the probabilities of snow avalanches, landslides, wildfires, land subsidence, and floods using machine learning models that include support vector machine (SVM), boosted regression tree (BRT), and generalized linear model (GLM). Climatic, topographic, geological, social, and morphological factors were the main input variables used. The data were obtained from several sources. The accuracies of GLM, SVM, and functional discriminant analysis (FDA) models indicate that SVM is the most accurate for predicting landslides, land subsidence, and flood hazards in the study area. GLM is the best algorithm for wildfire mapping, and FDA is the most accurate model for predicting snow avalanche risk. The values of AUC (area under curve) for all five hazards using the best models are greater than 0.8, demonstrating that the model's predictive abilities are acceptable. A machine learning approach can prove to be very useful tool for hazard management and disaster mitigation, particularly for multi-hazard modeling. The predictive maps produce valuable baselines for risk management in the study area, providing evidence to manage future human interaction with hazards.

/pdf/a-machine-learning-framework-for-multi-hazards-modeling-and-36mqqqvsw1.pdf

A machine learning framework for multi-hazards modeling and mapping in a mountainous area.

/pdf/a-conceptual-model-of-avalanche-hazard-3p96g8xof3.pdf

A conceptual model of avalanche hazard

An exceptionally high number of avalanche fatalities during the winter of 2003 forced the Canadian avalanche community to question the effectiveness of existing public avalanche safety programs in Canada. In response to the recommendations of several avalanche safety reviews, the Canadian Avalanche Association launched the ADFAR (Avalanche Decision Framework for Amateur Recreationists) Project for the development of a practical, science-based decision framework for amateur recreationists when planning for, or traveling in avalanche terrain. The goal of the project was to reduce recreational avalanche fatalities by improving risk communication and risk awareness among the fast growing number of winter backcountry enthusiasts in Canada. The Avaluator is a new rule-based decision support tool for amateur recreationists, including backcountry skiers and snowboarders, snowmobile riders and out-of-bounds skiers and snowboarders. A key part of the Avaluator is a pocket card that assists with planning backcountry trips and facilitates field decisions. The paper provides an overview of the ADFAR project, describes the usage of the Avaluator and discusses the underlying design principles.

/pdf/the-avaluator-a-canadian-rule-based-avalanche-decision-3zg1kvm4i1.pdf

The Avaluator – A Canadian Rule-Based Avalanche Decision Support Tool for Amateur Recreationists

The Avalanche Danger Scale is an ordinal, five-level warning system that is a cornerstone of public avalanche information. The system was developed in Europe in 1993, and introduced to North America in 1994. Although both Canada and the United States adopted the system, different descriptors of the danger levels were developed in each country. Fifteen years of practical use revealed numerous deficiencies in this danger scale, most notably a lack of clarity during low probability/high consequence avalanche conditions. In 2005, a group of Canadian and American avalanche forecasters and researchers began to revise the system, with the goal of improving clarity and developing a single standard for North America. Initial explorations to define the problem resulted in more questions and uncovered an almost complete absence of formal underpinnings for the danger scale. The magnitude of the project subsequently changed, and in 2007 the project objectives were clarified as: 1) definitions of avalanche hazard, danger and risk; 2) methodology for assessing avalanche danger; and 3) revisions to the danger scale as a public communication tool. This paper concentrates on the third and final objective, and describes the methods and results of producing the North American Public Avalanche Danger Scale. Emphasis is placed on best practice in warning system design and the principles of risk communication, which helped reshape the avalanche danger scale into a more effective communication tool. The revised danger scale will be implemented across Canada and the United States for the 2010/11 season.

/pdf/the-north-american-public-avalanche-danger-scale-4m8m5hq92d.pdf

The north american public avalanche danger scale

Guides of commercial backcountry skiing operations manage the physical risk from avalanches by assessing the local avalanche hazard and carefully choosing terrain and travel procedures to keep the residual risk at an acceptable level. A daily evaluation of predefined ski runs by the guiding team about whether the particular terrain is open or closed for guest skiing under the expected avalanche conditions represents the foundation for all subsequent guiding decisions during the daily skiing program. This study uses daily hazard assessments and run list records from a heli-skiing operation to quantitatively explore patterns in large-scale terrain preferences under different avalanche hazard conditions. Our data set spans four winters (2012/13 to 2015/16) with a total of 339 days and 93,647 documented run list ratings. The patterns in the daily terrain preferences revealed by our analysis provide interesting insights on how professional mountain guides deal with different avalanche situations. Our visualization of the 2013/14 winter season shows that both daily run list ratings for different terrain categories and run usage were closely linked to the avalanche hazard rating and avalanche character. Different ski line categories on a run were progressively opened or closed with decreasing and increasing hazard level, respectively. Using the entire data set of four winters, we show that runs with more options are skied earlier in the season and more frequently skied for the rest of the season. Contrasting run use on days with and without persistent avalanche problems, we see that both repeated use of terrain that was skied previously and the number of run passes per group on the same ski run are higher on days with persistent avalanche problems and groups tend to use the same runs more frequently.

/pdf/exploring-large-scale-terrain-preferences-for-managing-198cnmsby1.pdf

Clair Israelson

Papers

A Conceptual Model of Avalanche Hazard

A conceptual model of avalanche hazard

The Avaluator – A Canadian Rule-Based Avalanche Decision Support Tool for Amateur Recreationists

The north american public avalanche danger scale

Exploring Large-Scale Terrain Preferences for Managing Avalanche Risk in a Commercial Backcountry Skiing Operation