Avalanche forecasting — an expert system approach

Abstract

Avalanche forecasting for a given region is still a difficult task involving great responsibility. Any tools assisting the expert in the decision-making process are welcome. However, an efficient and successful tool should meet the needs of the forecaster. With this in mind, two models, were developed using a commercially available software: CYBERTEK-COGENSYSTM, a judgment processor for inductive decision-making–a principally data-based expert system. Using weather, snow and snow-cover data as input parameters, the models evaluate for a region the degree of avalanche hazard, the aspect and altitude of the most dangerous slopes. The output result is based on the snow-cover stability. The new models were developed and have been tested in the Davos region (Swiss Alps) for several years. To rate the models, their output is compared to the a posteriori verified hazard. the first model is purely data-based. Compared to other statistical models, the differences are: more input information about the snow cover from snow profiles and Rutschblock tests, the specific method to search for similar situations, the concise output result and the knowledge base that includes the verified degree of avalanche hazard. The performance is about 60%. The second, more-refined model, is both data- and rule-based. It tries to model the decision-making process of a pragmatic expert and has a performance of about 70%, which is comparable to the accuracy of the public warning.

Spatio-temporal avalanche forecasting with Support Vector Machines

“…In operational forecasting a mixture of all three approaches is often used, with process chains involving physical models and statistical or expert systems being relatively common (Durand et al, 1999;Purves et al, 2003;Bolognesi, 1993;Schweizer and Föhn, 1996).…”

“…Numerical methods used to assist in avalanche forecasting range from complex systems involving physical models (Durand et al, 1999;Lehning et al, 1999) to various heuristic and statistical decision-support methods (Schweizer and Föhn, 1996;Keiler et al, 2006;Bolognesi, 1993). Statistical methods of avalanche forecasting link several available input variables (meteorological factors, snowpack properties and stability factors as identified by LaChapelle (1980) to the target outputs.…”

“…However, the choice of the target is not evident. Although the verified avalanche hazard might be seen as an ideal choice (Schweizer and Föhn, 1996), forecasted avalanche hazard is much more commonly available (Brabec et al, 2001). Many approaches have used observations of avalanche activity as a target and assumed that it is related to avalanche hazard (Buser, 1989;Heierli et al, 2004).…”

Application of Neural Networks for Avalanche Forecasting

“…In early nineties, attempts were made to use artificial intelligence for avalanche forecasting, Schweizer 4 , et al developed a hybrid system, ALUDES, which clubs neural network and rules extracted from the database. Schweizer 5 , et al experimented with commercially available judgement processor COGENSIS TM MEPRA, a French expert system, analyses snowcover data for avalanche risk forecasting 1 . In Snow and Avalanche Study Establishment, Manali, attempts have also been made to use artificial intelligence techniques for avalanche forecasting [6][7][8][9] .…”

Modelling the system behaviour of wet snow avalanches using an expert system approach for risk management on high alpine traffic roads

“…Bolognesi (1993) and Bolognesi et al (1994) presented an expert system for the avalanche forecasting on a local scale (AVALOG), Schweizer and Föhn (1996), Schweizer et al (1994) and Brabec (2001) showed an approach for the avalanche forecasting on a regional scale (DAVOS, MODUL, NXD-REG). An expert system for the determination of avalanche release zones on the basis of digital terrain models was pointed out by Buisson and Charlier (1989).…”