SHALSTAB is a physically-based digital terrain model for mapping the relative shallow slope stability potential across a landscape. Extensive testing of the model and application in practical contexts suggest that the model can be successfully used to delineate observed landslide scar locations and provides an objective procedure for delineating future potential sites of instability. It can be used as a parameter free model in which the only decision is how to rank the mapped pattern of relative stability into such categories as "high", "medium" and "low" for the practical purpose of prescribing some land management practice. This utility is accomplished by eliminating many processes or factors that do matter to slope instability but require too much local parameterization to be useful in a practical context for application over large areas. Hence, this model only routes water through the landscape at steady state, rather than dynamically modeling storm events. It is the underlying hypothesis here, seemingly well supported with observation, that the overriding influence of topography on the local evolution of a perched water table, permits the steady state model to emulate the effects of dynamic storm response. This is not to say that dynamic modeling of landscape response to storms is not a valuable enterprise. Furthermore, the model does not consider the effects of root strength on slope stability. The assumption here is that the model is capturing the condition of low root strength that would occur after cutting, disease and death, or a fire, and that shallow landslides are still most commonly associated with steep, high drainage area sites. Again, comparison of the model with mapped landslide scars supports this assumption in general, although many local cases will differ.

The most important practical decision in using the model (after selecting a topographic base) is how to categorize the relative slope stability rating into hazard ratings that invoke specific landuse prescriptions, including the possibility of no harvest zones in forests. In the end it is unrealistic in most settings to set the log (q/T) threshold such that all mapped landslide scars are part of the high zone. Compromise about perceived risk is necessary and there should be the expectation that the spatial pattern of risk will change with intensive field work or improved topographic information.

Useful extensions to the model already exist and more are anticipated. We have developed the code to include root cohesion by using a spatially constant soil depth. This permits using field data on root strength as influenced by forest practices to illustrate the possible controls on pattern of hillslope instability. We have called this model SHALSTAB.C and a description will soon be made available. A more advance model, reported by Dietrich et al. (1995), employs a process-based soil production and transport model to predict soil depth in order to use that outcome in a model of slope instability that includes root strength and vertically varying saturated hydraulic conductivity. This model we now call SHALSTAB.V. Both models add insight about causality, but require parameterization that the base model was designed to avoid. With high resolution topographic data, the addition of cohesion can greatly reduce the area considered to be in the highest slope stability category. We have also developed a debris flow runout algorithm for use in the contour-based version of SHALSTAB (Montgomery and Dietrich, 1994). This model is being converted to a grid-based model and will be called SHALSTAB.D.

In the next several years we also anticipate significant improvements in the quality of the base maps that drive digital terrain models. We have illustrated here how SHALSTAB works using one such data set derived from laser altimetry. It is technologically possible now to generate such maps for large areas. With such maps, landslide hazard delineation will be much improved and so too will many other things we wish to infer, model or route from digital terrain data.