Dendrogeomorphic reconstruction of snow avalanche regime and triggering weather conditions: A classification tree model approach

While dendrogeomorphology has been recognized as a useful tool to identify past avalanche activity, there is only a handful of papers that focus on the assessment of weather or climatic triggers of tree-ring reconstructed avalanche events. This paper compares the potential of logistic regression and classification tree algorithms to highlight weather scenarios responsible for the occurrence of high-magnitude avalanche activity in the Presidential Range of the White Mountains, New Hampshire (USA). Our tree-ring procedure improves the modern GD-I-t threshold with the implementation of a second criteria based on the Moran index. 450 trees sampled in seven different avalanche paths allowed us to reconstruct 45 avalanches that occurred during 19 different years for the period 1936-2012. The results show that while statistically significant, the logistic regression models are less accurate than classification trees to assess avalanche activity based on annual and monthly weather variables. Moreover, even if snow related covariates are located at the root node of every classification tree model, the addition of temperature and wind predictors increases their robustness. This suggests that high-magnitude avalanches in the Presidential Range not only respond to snow, but also to atmospheric conditions responsible for the creation of weak layers within the snowpack.