International Snow Science Workshop Proceedings - Montana State University Library

Item: Two Expert Systems to Forecast the Avalanche Hazard for a Given Region

• • Title: Two Expert Systems to Forecast the Avalanche Hazard for a Given Region

    Proceedings: Proceedings of the 1994 International Snow Science Workshop, Snowbird, Utah, USA

    Authors:

    • Jurg Schweizer  [ Swiss Federal Institute for Snow and Avalanche Research, CH-7260 Weissfluhjoch/Davos, Switzerland ]
    • Paul M.B. Fohn  [ Swiss Federal Institute for Snow and Avalanche Research, CH-7260 Weissfluhjoch/Davos, Switzerland ]

    Date: 1994

    Abstract: A commercially available software, CYBERTEK-COGENSYSâ„¢ Judgment Processor for inductive decision-making, was ;used to develop two different expert systems as supporting tools to forecast the avalanche hazard for a given region. Using weather, snow and snow cover data as input parameters the systems evaluate the degree of hazard, the aspect and altitude of the most dangerous slopes. So the output result is exactly what the avalanche forecaster needs. The avalanche hazard we use in the data base, is the verified hazard or the so-called verification, i.e. the day-to-day critical -a posteriori assessment of the avalanche hazard. The new models were developed, tested and rated in the Davos region (Swiss Alps) for several years from beginning of December to end ofApril. The first modelmaypartially be compared to statistically based systems. However the differences are: more input information about the snow cover including snow profiles and Rutschblock tests, the concise output result and the knowledge base that includes the verified degree of avalanche hazard. The performance is about 60%, i.e. the predicted degree of avalanche hazard coincides on 6 out of 10 days with the later on verified degree of hazard. The second model is more process oriented and includes partially implicit rules; it may be compared to a deterministic system. The system tries to model the decision making process of a pragmatic expert. It has a performance of 70 to 75%. In both models the snow cover data proved to be most decisive. Since some of the input parameters are not conventional measurements, the models do not run fully automatically, but the interactive use is highly instructive.

    Language of Article: English

    Presenters: Unknown

    Keywords: hazard, snow cover, decision making, snow profiles,

    Page Number(s): 295-309

    Subjects: avalanche hazard decision making avalanche forecasting

  • Get Full Article

    Digital Abstract Not Available

    Persistent Link

    Bookmark and Share

    Get Embed Code

Back to Home page