Estimating the forecasting success of artificially triggering of avalanches with the combination of cluster and discriminant analysis.

Birmensdorf

2009

S. 366-370 + Poster

Artikel aus einer ZeitschriftBandaufführung

10534S

159582

Meteorological conditions play an important role for the mechanical stability of the snowpack. Statistical algorithms can be used to determine the likelihood of the occurrence of avalanches which was shown in several previous studies. The common use of two statistical methods offers the possibility to analyze the data set in regard to specific weather conditions. In the first step the k-mean cluster analysis selects days with similar weather conditions; so that each day is assigned to a predefined group. Significant weather conditions are used for the definition of the initial conditions of each cluster. Each group represents a typical weather situation. In the second step the discriminant analysis is used to separate between avalanche and non-avalanche days for each group. Consequently the coefficients of the discriminant functions can differ and the best fit of the discriminant function depending on the predominant weather situation is applied. Results of the clustering algorithm and the hit rate of the subsequent discriminant analysis are shown.