A Bayesian probabilistic framework for avalanche modelling
based on observations
, Adrienne Grêt-Regamey
Department of Civil and Environmental Engineering, University of California, Berkeley, CA 94720-1710, USA
WSL Swiss Federal Institute for Snow and Avalanche Research, Division, 7260 Davos, Switzerland and ETH Zürich,
Institute for Spatial and Landscape Planning, 8093 Zürich, Switzerland
Received 30 May 2006; accepted 22 August 2006
Applied avalanche models are based on parameters which cannot be measured directly. As a consequence, these models are
associated with large uncertainties, which must be addressed in risk assessment. To this end, we present an integral probabilistic
framework for the modelling of avalanche hazards. The framework is based on a deterministic dynamic avalanche model, which is
combined with an explicit representation of the different parameter uncertainties. The probability distribution of these uncertainties
is then determined from observations of avalanches in the area under investigation through Bayesian inference. This framework
facilitates the consistent combination of physical and empirical avalanche models with the available observations and expert
knowledge. The resulting probabilistic spatial model can serve as a basis for hazard maping and spatial risk assessment. In this
paper, the new model is applied to a case study in a test area located in the Swiss Alps.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Avalanches; Natural hazards; Bayesian analysis; Risk assessment; Uncertainty modelling; Land-use planning
Mountainous areas are subject to various gravitational
hazards, including avalanches, debris flows, land slides
and rock-falls. It is increasingly recognized by decision
makers that the effective and rational management of such
natural hazards requires a risk-based strategy which
explicitly addresses the involved uncertainties together
with the consequences of such events, e.g. PLANAT
(2004). An important part of such a strategy is construct-
ing a probabilistic hazard model, which should be based
on all available information, including physical models
and observations of past events, following the principles
of Bayesian decision theory, for more information refer to
DeGroot (1970) or Benjamin and Cornell (1970).
The purpose of this paper is to provide such a proba-
bilistic hazard model for avalanche hazards. The
probabilistic model presented in this paper can be con-
sidered as a framework to accommodate any existing
one- or two-dimensional phenomenological avalanche
model. To demonstrate the implementation of this
method, we apply a state-of-the-art two-dimensional
dynamic simulation model, the AVAL-2D, used in
Switzerland for avalanche prediction and hazard zoning
(Gruber, 1999). In Grêt-Regamey and Straub (2006),we
demonstrate how the model may be included in a risk
analysis framework using Bayesian Networks.
Cold Regions Science and Technology 46 (2006) 192 – 203
E-mail addresses: firstname.lastname@example.org (D. Straub),
email@example.com (A. Grêt-Regamey).
0165-232X/$ - see front matter © 2006 Elsevier B.V. All rights reserved.