Size independence of the strength of snow.

The mechanical reliability of 600 randomly taken snow samples follows Weibull distributions: If sigma(max) is the maximum stress present in a specimen of given density, the fraction of specimens that fail at stresses below sigma(max) is P=1-exp[-(sigma(max)/sigma(0))(m)]. The scale parameter sigma(0) evaluated by the maximum likelihood method increases nearly quadratically with the density rho of snow, but, unlike predicted by the weakest link model, is independent of size and shape of the specimen: there is no size dependence of the strength of snow. The Weibull parameter m is independent of density, size, and shape of the snow sample, m=1.5+/-0.5. This implies, on the one hand, that the results of laboratory scale tests can be used for avalanche prediction, but on the other hand, that these predictions remain contaminated with large statistical errors. Snow is a fragile, weak, and unreliable material.