During the 20th century, the yearly average temperature over France increased by 1°C. One can say that higher temperature leading to less snow will cause fewer avalanches. Unfortunately, the avalanche activity response to climate changes is not that straightforward. On the contrary, it is very ambiguous because although the initial hypothesis might be correct in some areas, the opposite can be easily proved in others. This univocal behavior, along with the lack of long data series explains the limited knowledge about avalanche activity and the inability to detect behavioral trends using standard methodologies.

Although previous studies could highlight a form of correlation of the rather short snow avalanche series and the weather variables, no conclusion is drawn, and the need for longer data records and on-site calibration using the available historical chronicles is unavoidable. As a solution to this problem and many others, a Bayesian probabilistic calibration method has been proposed to reconstruct past avalanche activity and predict high return period avalanche. This approach was successful in reconstructing past high return period events and in deriving the relationship between runoff distance of an avalanche and its return period. However, this validation was made based on the assumption of stationarity and for a short time.

A rewarding yet tough approach is to use secondary, non-conventional data to enlarge the spatiotemporal coverage of snow and avalanche series. We can reach a better understanding of avalanche activity by combining tree-ring reconstitutions, historical archives, and instrumental data. However, in this case, a contextualization step is unavoidable due to the intrinsic nature of such data, and its direct link to several socio-environmental changes.

The objective of the Ph.D. is to study and investigate how the combined changes in environmental and social components can affect snow avalanche risk. This objective will be achieved by filling the gaps between the fully empirical approaches and the physically based numerical simulation (stated above, tested for stationarity). Within this context, a statistical numerical framework that explicitly addresses non-stationarity will be developed. This framework will be used to derive frequency-magnitude relationship by reconstructing past avalanche events (last 200 years or older). Finally, by combining the frequency-magnitude relationship with elements at risk and damage susceptibility functions, it will be possible to evaluate changes in the risk. The work represents an investigation of the social, environmental and climate drivers of avalanche risk which is very rare in the snow avalanche field. The project is a collaboration between IRSTEA, CEN (Center d’étude de la neige) and LARHRA. It is co-financed by IRSTEA (50%) and Trajectories (50%).