A probabilistic technique for exploring multi-scale spatial patterns in historical avalanche data by combining GIS and meteorological nearest neighbors with an example from the Jackson Hole Ski Area, Wyoming

McCollister, C., Birkeland, K., Hansen, K., Aspinall, R., & Comey, R. (2003). A probabilistic technique for exploring multi-scale spatial patterns in historical avalanche data by combining GIS and meteorological nearest neighbors with an example from the Jackson Hole Ski Area, Wyoming. Cold Regions Science and Technology, 37(3). https://avalanche.org/wp-content/uploads/2018/08/02_ISSW_McCollister.pdf

In this article, the researchers outline a method used for predicting avalanches at Jackson Hole Mountain Resort, Wyoming. Their method compares historical weather and avalanche data to generate avalanche probabilities for slopes within the resort. The authors call this the "modified meteorological nearest neighbors approach". Avalanche control personnel input weather data for a desired forecast day and the software references it against weather and avalanche data collected during the 1978-2002 ski seasons. This generates avalanche probability profiles for avalanche slide paths identified by resort operations and also derived from digital elevation models using GIS software.

The authors claim that current (2003) methods of discriminant analysis, cluster analysis, nearest neighbors, classification and regression trees are not adequate in avalanche prediction because they don't take into account the geographic component of slide paths, nor do they analyze data at an individual slide path scale. Results of this study indicated their method did improve on those previous techniques in generating usable avalanche probabilities and also in dealing with large data sets.

While I found the statistical analyses used in this research to be difficult to grasp, the general methodology of the "modified meteorological nearest neighbors approach" proved useful for my overall understanding of how GI systems work with large weather and terrain datasets. The meteorological nearest neighbors technique seems applicable to other areas of snow science work, and is something I plan to research further.