Geographic knowledge discovery techniques for exploring historical weather and avalanche data

Abstract

Many ski areas, backcountry avalanche centers, highway departments, and helicopter ski operations record and archive daily weather and avalanche data. The objective of this thesis is to present probabilistic techniques that allow avalanche forecasters to better utilize weather and avalanche data by incorporating a Geographic Information System with a modified meteorological nearest neighbors approach. This nearest neighbor approach utilizes evolving concepts related to visualizing geographic information stored in large databases. The resulting interactive database tool, Geographic Weather and Avalanche Explorer, allows the investigation of the relationships between specific weather parameters and the spatial pattern of avalanche activity. In order to validate these new techniques, two case studies are presented using over 10,000 individual avalanche events from the past 23 years that occurred at the Jackson Hole Mountain Resort. The first case study explores the effect of new snowfall, wind speed, and wind direction on the spatial patterns of avalanche activity. Patterns exist at the slide path scale, and for groups of adjacent slide paths, but not for either the entire region as a whole or when slide paths are grouped by aspect. Since wind instrumentation is typically located to measure an approximation of the free air winds, specific topography around a given path, and not aspect, is more important when relating wind direction to avalanche activity. The second case study explores the spatial variability of hard slab and dry loose avalanches, and characterizes these avalanche types with respect to their geographic location and associated weather conditions. I analyzed these data with and without the incorporation of three weather parameters (wind speed, 24-hour maximum temperature, and new snow density). Slide paths near each other often had similar proportions of hard slabs and a higher proportion of hard slabs occurred on exposed ridges. The proportion of loose avalanches also was similar for adjacent slide paths, and these paths were typically sheltered from strong winds. When I incorporated the three weather parameters I found significant increases in the average proportion of hard slabs with increases in new snow density, but not for changes in the 24-hour maximum temperature or wind speed. When I analyzed the proportion of loose avalanches associated with the three weather parameters I found a more direct relationship than with hard slabs. Changes in both wind speed and density significantly changed the average proportion of loose avalanches, with low wind and low density resulting in higher proportions of loose avalanches. My results quantify what operational avalanche forecasters have long known: Geographic location and weather are both related to the proportion of hard slab and dry loose avalanches.