M. Casadei, W.E. Dietrich & N. Miller
University of California at Berkeley, Department of Earth and Planetary Sciences, Berkeley, CA, USA
Keywords: Shallow Landslides, Slope Stability, Scar Width, Root Strength ABSTRACT: Most distributed slope-stability models for shallow landslide prediction neglect the forces acting on the sidewalls of landslide scars.Back-calculations and field observations, however, show that lateral root strength is a primary control on size and location of shallow landslides in soil mantled hillslopes. Here we report a theory that estimates landslide width, assuming that root strength acts primarily through a perimeter boundary. The model predicts that landslide width increases with increasing root strength and decreasingslope, as larger masses of soil are needed to overcome resisting forces. Perhaps surprisingly, the drier the soil, the larger the landslide mass (and width), whereas the water table rise reduces the size needed for failure. The comparison of the model results with field data suggests that landslide size is controlled by the local patchiness of soil thickness, root strength and topographically-drivenrelative saturation.
Digital terrain models are becoming widely used to map the relative potential for shallow landsliding across a landscape by both process-based models and statistical analyses (e.g. review in Dietrich et al., 2001, Gritzner et al., 2001). Process-based models have been based on the infinite slope form of the Mohr-Coulomb failure law in which landslidedimensions are ignored and relative stability is performed for each individual grid cell. While the influence of spatially variable soil depth and vegetation, and temporarily variable vegetation and precipitation have been included (e.g. Hsu, 1994; Terlien et al., 1995; Wu and Sidle, 1995; Duan, 1996; Iida, 1999; Casadei et al., in press), these models nonetheless treat each grid cell independentlyand, therefore, depend strongly on the quality of the topographic data and the relative size of landslides compared to grid cell dimensions (e.g. Dietrich et al., 2001; Gritzner et al., 2001). Figure 1 shows the results of digital terrain based slope stability model that uses high-resolution topographic data (obtained through airborne laser swath mapping), a predicted soil depth field, a uniformfield of root strength and a steady state hydrology. A calibrated threshold rainfall to transmissivity ratio is used. Mapped landslide scars generally occur in areas classified as unstable (potentially unstable areas require larger rainfall, stable areas will not fail when saturated), but there are extensive areas shown as unstable without scars and the mapped landslide scars are frequently adifferent size than the local unstable areas. In maps with coarser grid cells, landslides are commonly smaller than the local unstable areas (e.g. Casadei et al., in press). An improvement in grid-based models would be to predict which cluster of cells may fail together because of scale controls associated with strength effects on the boundaries and the spatially variable material and hydrologicproprieties. Prediction of landslide
size would improve debris flow hazard modeling, possibly reduce the general tendency to over predict areas of instability, and be useful in landslide-driven landscape evolution models.
Figure 1. Delineation of landslide potential for a site in Coos Bay, Oregon, US. calculated with a version of the SHALSTAB model (Dietrich et al., 1995) which uses a calculatedsoil depth field The area calculated as unstable and potentially unstable is significantly larger than the actual landslides observed during the last 15 years. Potentially unstable sites are areas that could fail if greater precipitation caused sufficient pore pressure rise. This problem (overprediction of instability) is common to all current landslide hazard models.
The issue of shallow...