Snow Avalanche Prediction in Grasdalen, Norway : Application of wind drift factors and classification trees

Abstract

Snow avalanches are a common hazard experienced in many mountainous environments around the world during the winter months. It is therefore often of great importance to be able to predict these events to reduce the risk they pose to the population and infrastructure. Work done in this thesis has involved detailed analysis of various weather parameters within Stryn in western Norway in order to establish common triggering factors for avalanches in Grasdalen. Two large sets of data; a gridded extrapolated data set and an observed set of data from the study area have been compared. From these data sets, several wind drift factors have been derived as these parameters are considered important avalanche triggering elements by several authors. These combined factors have not previously been analysed for the observed Fonnbu data set. Statistical procedures include cumulative probability plots which have provided threshold values, a Kruskal-Wallis test, and additionally, a number of classification trees. The latter were used to highlight the most important weather parameters used to classify data in terms of dry avalanche days or non avalanche days which has not been undertaken with the Fonnbu data set previously. Results of these indicate the primary splitting factor to be various sums of precipitation over the preceding days, particularly the two day sum and four day sum. Following this, the maximum temperature measured on the preceding day is considered important for classification tree splits between dry avalanche and non avalanche days. In terms of the combined wind drift factors, these appear in 32 % of the classification trees within the top three splits, the most predominant being calculated using maximum wind speeds on the day of the avalanche. Although this thesis is not in total agreement with the results of previous work which outlines the importance of the wind drift factor, rather, it highlights the complicated relationship between preceding weather conditions and avalanche occurrence indicating the vast array of factors to be considered for avalanche prediction.