Item: Stochastic Methods in Operational Avalanche Simulation - From Back Calculation to Prediction
-
-
Title: Stochastic Methods in Operational Avalanche Simulation - From Back Calculation to Prediction
Proceedings: International Snow Science Workshop 2016 Proceedings, Breckenridge, CO, USA
Authors:
- Valentin Hellweger [ University of Innsbruck, Innsbruck, Austria ]
- Jan-Thomas Fischer [ Austrian Research Centre for Forests - BFW, Innsbruck, Austria ]
- Andreas Kofler [ Austrian Research Centre for Forests - BFW, Innsbruck, Austria ]
- Andreas Huber [ Austrian Research Centre for Forests - BFW, Innsbruck, Austria ]
- Wolfgang Fellin [ University of Innsbruck, Innsbruck, Austria ]
- Michael Oberguggenberger [ University of Innsbruck, Innsbruck, Austria ]
Date: 2016-10-02
Abstract: Avalanche simulations are an integral part of hazard assessment. Determining the potential hazard requires a multidisciplinary approach, including different scientific fields such as geography, meteorology, physics, civil engineering and mathematics. The application of probabilistic methods allows one to develop a complete, comprehensive applicational concept for snow avalanche simulations, ranging from back calculation to prediction. In this context optimal parameter sets or runout distances are represented by probability distributions. Existing deterministic avalanche dynamics models contain several parameters (e.g. friction), some of them more conceptual than physical. Direct measurement of these parameters in the field is hardly possible. Hence, a parameter identification has to be undertaken, matching simulation results to field observations. This inverse problem can be solved by optimization or by a Bayesian approach (Markov chain Monte Carlo). An important task in snow avalanche simulation is to predict process intensities (runout, flow velocity and depth, ...). The identification process yields parameter distributions, that can be utilized for probabilistic reconstruction and prediction. Arising challenges include: the limited amount of observations, correlations appearing in model parameters or observed avalanche characteristics (e.g. velocity and runout) and the effective and accurate handling of ensemble simulations, always taking into account the related uncertainties.
Object ID: ISSW16_P4.50.pdf
Language of Article: English
Presenter(s):
Keywords: avalanche dynamics, parameter estimation, stochastic simulation, Bayes’ theorem, Metropolis-Hastings algorithm, posterior distribution, back calculation, prediction
Page Number(s): 1375-1381
Subjects: avalanche prediction avalanche simulation avalanche dynamics
-