Snow Layering and Spatial Heterogeneity

Snow packs are made up of snow layers, each differing in physical and microstructural properties from those above and below. The sequence and characteristics of the layers affect the electromagnetic, thermal, physical and mechanical properties of the pack. Layer boundaries are also important in determining the strength of the pack and the transport of air, water and heat through it, though relatively little attention has been focused on the nature of the boundaries themselves. In general, layers are used (some times tacitly) as the basis for spatial extrapolation of properties, with the assumption that layers are laterally homogenous. On ice sheets and large glaciers, this assumption may be valid, but in seasonal snow covers the layers vary laterally at multiple scales (10^-1 to 10^-3 m) in ways strongly dependent upon deposition conditions and substrate micro- and meso-topography. Some "slow" processes, like snow settlement and kinetic growth due to temperature gradients, produce spatial variations that are predictable where the layer geometry is gently varying. Other "faster" processes, like the development of dunes during wind transport or melt water percolation, produce high lateral gradients in properties over short (10^-1 m) distances that are currently beyond our ability to either measure (except with intense effort) or predict. A number of important snow processes (i.e., wind pumping, avalanche release, melt water routing) seem to be focused in or dominated by these sharp gradient zones and narrow facies boundaries. Future progress in modeling snow covers is going to require a better understanding of spatial variability of snow layers, as well as an improved identification of those problems requiring prediction of average vs. outlier properties. New approaches in modeling will also be needed. Fortunately, there are currently a number of new tools (micro-penetrometers, snow radars, mini-data loggers, digital imaging systems) that if we can figure out how to use in concert, should be able to generate the extensive data sets we will need in order to advance.