The EAWS matrix, a decision support tool to determine the regional avalanche danger level (Part B): operational testing and use

Abstract

To support public safety and risk management in snow-covered mountains, regional avalanche forecasts must deliver reliable information on avalanche conditions, including regional danger levels representing the avalanche danger across warning regions. To promote greater transparency and consistency in avalanche danger level assessment across European avalanche warning services, a revised version of the EAWS Matrix was developed based on expert elicitation. The Matrix, a structured decision support tool that combines the Matrix input factors snowpack stability, the frequency of snowpack stability, and avalanche size, is used to determine the regional danger level. To support the development of the Matrix described in detail in the companion paper, we analyzed its operational use over the first three winters following implementation by 26 European avalanche warning services. Our aim was to identify inconsistencies in Matrix application and to provide empirically based guidance for further refinement. In operational use, forecasters predominantly assigned a consistent single danger level to most Matrix input factor combinations. However, two factor combinations (poor-some-size 2 and very poor-some-size3) were commonly assigned to one of two adjacent danger levels, indicating that these combinations function as transition zones between danger levels. Analyses based on finer-grained assessments of the input factors, that is, using sub-classes of the predefined coarse factor categories, revealed systematic tendencies within these classes. While application of the Matrix was relatively consistent for avalanche problems relating to dry-snow conditions, pronounced inconsistencies emerged in the classification of snowpack stability for wet-snow and gliding snow avalanche problems. These findings underscore the need for community-wide discussion and harmonization in Matrix application, particularly with respect to stability assessment practices. Assessing input factors at a finer scale shows potential for preserving important nuances in expert judgment and may enable more targeted guidance on when to assign the higher or lower of two danger levels indicated by the Matrix. However, because neither the danger level nor its input factors can be measured independently, a formal validation of Matrix logic and operational application is not possible. Despite some inconsistencies, our results suggest that European forecasters generally align with the Matrix logic, supporting its operational utility.