A Global System for Avalanche Risk Assessment

Abstract

Several decision-support methods exist to assist ski touring practitioners in their choice of the safest possible route to take. This paper proposes approaches to solve two different challenges presented by decision-support methods: 1) the description and assessment of the parameters used in the methods and 2) the combination of the parameters into a final result. Specifically, this paper focuses on recent avalanche observations. Indeed, this parameter is a particularly effective indicator of the current danger level and is considered in several decision-support methods but is not well formalized yet. The developed process, based on unsupervised statistical analysis and machine learning methods, evaluates both the weather trends and the criticality of different areas. It aims to positively impact and improve the assessment of this parameter in the existing methods. Further, this paper presents a global system based on fuzzy logic and developed to combine all parameters into a final result. We have constructed this system in collaboration with a domain expert and applied it to the CRISTAL approach, one of the existing decision-support methods, whose final result is the vigilance mode to adopt when practicing ski touring.