Snow–Pack Structure: Stability Analyzed by Pattern–Recognition Techniques

Abstract

AbstractInternal snow–cover structure depicted by conventional snow profiles is commonly held to be important for avalanche forecasting but it is largely omitted from numerical forecasting schemes owing to quantification difficulties. A subjective test given to practising forecasters illustrated the complex roles of measured profile parameters in describing instability. A clustering technique of pattern recognition, after introducing a snow–profile training set, has demonstrated an objective ability to sort correctly 86% of the layer boundaries associated with slab–avalanche bed surfaces or lubricating layers. Further improvement in stability recognition comes from a newly devised shear test driven by a wedge which quickly locates failure planes within the snow cover. The cluster analysis is then able to sort with similar accuracy between stable and unstable (avalanching) failure planes so located.