Particle-based simulation of snow drifting in an Eulerian wind field




Abstract:

In this work we simulate snow drift formation, as it is affected by a non-trivial wind
field. This could be to predict the drift formation in residential areas around houses and other
obstacles, or it could be to optimize the placing of snow drift fences in the landscape. The
granularity of the snow flow is abstracted away by viewing it as a non-Newtonian fluid flow.
This is done using a rheological material model, which is discretized using the Smoothed
Particle Hydrodynamics (SPH) method.

Simultaneously an Eulerian fluid flow field, representing the wind, is embedded in the
domain, and simulated using the Finite Volume Method.
To properly handle the interaction between the two methods, at their common interfaces,
we have developed a method, which lets the snow influence the wind field, and which lets the
wind field exert forces upon the snow.

There exist regions where snowfall, and the winds are commonly very strong.  At such
regions it is very beneficial to be able to quickly determine if a new building may end up
changing the wind patterns, in such as way that extra snow will be deposited in undesirable
locations. Simulation may additionally assist in the optimal placement of snow fences.
We have no knowledge of snow drifting being simulated using SPH previously.  While
models of flowing avalanches have often dealt with SPH, it appears to be a new method for
drifting snow.

We have collected or derived the the relevant material characteristics of snow, and im-
plemented a model with a more solid physical foundation than what is commonly seen in
computer graphics, dealing with the visualization of wind driven snow scenes. At the same
time it is a more general model than what is generally seen in snow engineering.
Our implementation of the snow simulator has been implemented in a massively parallel
fashion, and programmed in CUDA, in order for it to run on a GPU.


The paper can be viewed here as a pdf.

Č
Ċ
ď
Thomas Grønneløv,
Nov 11, 2011, 3:18 PM