Home Page of Thomas Zwinger

start about me science
ice sheets

Avalanche Research


During my master thesis and thereafter within my doctoral thesis I have been studying the physics of avalanche dynamics. This scientific field basically can be described as geophysical Fluid Dynamics.
Dry snow avalanches in particular have been the subject of my research. Dry snow avalanches consist of relatively small extremely cold ice particles. The missing cohesion between the particles has two effects:

  1. reduced bed friction and consequently longer run-out distances
  2. tendency to build up turbulent particle-air suspension, better known as powder snow avalanches

This makes dry snow avalanches a very dangerous natural hazard.

A Coupled Physical Model for Dry Snow Avalanches

Physical avalanche models can act as an supporting tool in risk assessment of such natural hazards. In between my doctoral thesis I was involved in the development of the first Austrian physical model for dry snow avalanches SAMOS® (Snow Avalanche MOdeling and Simulation). SAMOS® is a joint project of AVL-List and Institute for Fluid Mechanics and Heat Transfer at Vienna University of Technology in cooperation with the Institute for Avalanche- and Torrent Research.

Layer structure of a dry snow avalanche

Naturally, a physical model distinguishes between the different flow regimes of a dry snow avalanche. The dense flow avalanche (DFA) at the bottom often is superimpose by a powder snow avalanche (PSA). Every dry snow avalanche starts as a DFA. The particles are re-suspended during the event and - if gaining sufficient density - then form the PSA. This re-suspension process takes place in a, compared to the PSA, shallow layer, the so called re-suspension layer (RES).

The aims in the project have been the following:

  • Applicability on arbitrarily shaped terrain
  • Determination of a set of fixed model parameters (essential for a prognostic avalanche model)
  • Effective code in order to be run on PCs

top of page

Dense Flow Avalanche

A dense flow avalanche (DFA) is characterized by a high particle volume fraction. Consequently, the interstitial air has no impact on the dynamics of this layer, which can be described applying models related to dense granular flows. In SAMOS® a modified version of the well established Savage-Hutter model for quasi static shallow granular flow has been adopted. The governing equations have been formulated in their global form. The numerical solution procedure utilizes the Finite Volume Method on an Lagrangian grid (i.e., the grid moves with the avalanche mass). The time integration step is explicit.

top of page

Powder Snow Avalanche

The governing equation for the powder snow avalanche (PSA) have been obtained by the Reynolds-averaged equations for the gas-particle mixture. Additionally, to account for the varying density caused by the particle load, a balance for the particle volume fraction has to be solved. The closure relation for the turbulent exchange properties, i.e., the components of the Reynolds stress for the mixture and the turbulent particle volume flux occurring in the balance equation for the volume fraction, are obtained applying the k-epsilon turbulence model. Additional terms in order to account for buoyancy effects have been introduced in this model.
The numerical solution procedure of the PSA is based on the CFD solver package FIRE® of AVL-List. It solves the Reynolds averaged Navier-Stokes equations on a fixed, structured hexahedral grid applying a time-implicit Finite Volume scheme. Density-pressure coupling is obtained utilizing the SIMPLE method.

top of page

Re-suspension Process

The re-suspension process has been modeled utilizing an analogy between turbulent momentum transport (i.e., Reynolds-shear stress) and turbulent mass transport (i.e., turbulent particle volume flux) close to the free surface of the DFA or the snow cover. As a result, the re-suspension flux of particles from the free surface of the DFA into the PSA could be expressed in terms of the wall friction velocity. A heuristic turbulent wall Schmidt number accounts for the finite Stokes number of the particles.

top of page


Initially, SAMOS® has been validated by comparing simulation results with data of real avalanche events. Within the range of the inaccuracy of the field data, the model was able to reproduce the scenarios of all events with a fixed set of model parameters. The pictures and animations have been produced by P. Sampl using SAMOS® postprocessing tools.

Galtür, Austria 1999

Click to magnify (300 Kb)

Deposition heights of the DFA

Click to magnify (300 Kb)

Pressure impact of PSA 2.4 meters above ground

Ischgl, Austria 1984

Click to play MPEG (5 Mb)

Animation of the coupled simulation of the Madlein avalanche in Ischgl. The filled iso-contours represent the current flow depth of the DFA. The density of the white particles above the DFA is proportional to the local volume fraction in the PSA. It is nicely to be seen that the PSA originates from the DFA after a certain travel distance. The whole event lasted about 150 seconds.

Another animation (2.4 Mb) with a projected map has been made by me using the PovRay raytracer (the encoding might not work well with all versions of the Windows MediaPlayer).

top of page

Material found on this page is for private and academic use only!

© 2003 TTZ