WATER RESOURCES RESEARCH, VOL. 30, NO. 4, PAGES 1153-i171, APRIL 1994 A physically based model for thetopographic control on shallowlandsliding David R. Montgomery Department of Geological Sciences and Quaternary Research Center, University of Washington, Seattle William E. Dietrich Department of Geology and Geophysics, University of California, Berkeley Abstract. A model for the topographic influence on shallowlandslide initiation is developed by coupling digitalterrain data with near-surface through flow and slope stability models.The hydrologic model TOPOG (O'Loughlin, 1986)predictsthe degree of soilsaturation in response to a steadystaterainfall for topographic elements defined by the intersection of contours and flow tube boundaries. The slopestability component uses this relative soil saturation to analyzethe stability of eachtopographic element for the case of cohesionless soils of spatially constant thicknessand saturated conductivity. The steady state rainfall predicted to cause instability in each topographic element provides a measure of the relative potentialfor shallowlandsliding.The spatial distribution of critical rainfall values is compared with landslidelocationsmapped from aerial photographs and in the field for three studybasins where high-resolution digital elevation data are available: TennesseeValley in Marin County, California; Mettman Ridge in the Oregon Coast Range; and Split Creek on the Olympic Peninsula, Washington. Model predictions in each of theseareas are consistent with spatial patterns of observed landslidescars,although hydrologic complexities not accounted for in the model (e.g., spatialvariabilityof soilproperties and bedrock flow) control specific sites and timing of debris flow initiation within areasof similar topographic control. Introduction The spatial and temporal distribution of shallow landslid- ing are important controls on landscape evolutionand a major component of both natural and management-related disturbance regimes in mountain drainage basins [e.g., Hack and Goodlet,1960; Dietrich and Dunne, 1978; Tsukamoto et al., 1982; Okunishi and Iida, 1983; Dietrich et al., 1986; Benda and Dunne, 1987; Crozier et al., 1990].The sudden failure andhigh speed of shallow landslides thatmobilize as debris flowsmake them particularly destructive to down- stream resources, property, andlives [e.g., SmithandHart, 1982; Ellen and Wieczorek, 1988; Brabband Harrod, 1989]. Debris flows mayalsoscour steep channels to bedrock and accelerate sediment deliveryto downstream, lower-gradient channels. Increasing pressure to use upland landscapes and concurrently to minimize downstreamimpacts necessitates development of objective methods for assessing the distri- bution of potential debrisflow source areas and run out paths. Debris flowstypicallyoccurduring intense storms or periods ofextended rainfall [e.g.,Caine, 1980], reflecting the effect of elevated soil moisture onsoil strength. Topography influences shallow landslide initiation through both concen- tration of subsurface flow andthe effect of gradient on slope stability. Other factorsthat also influence the spatialand temporal distributionof shallow landsliding include soil Copyright 1994 bythe American Geophysical Union. Paper number 93WR02979. 0043-1397/94/93WR-02979505.00 thickness, conductivity, and strength properties; rainfall intensity and duration; subsurfaceflow orientation; bedrock fracture flow; and root strength. While these factors are important controls, their spatial distribution are difficult to determine. On the other hand, most studies report that shallowlandslides only become important above a threshold hillslopegradient and that these landslides most commonly originatein areas of topographic convergence[e.g., Camp- bell, 1975; Reneau and Dietrich, 1987a; Ellen et al., 1988]. In this paper, we present a quantitative model for assess- ing the topographic influence on shallow landsliding.Models for the generation of soil saturation and slope instability are combined with digital terrain data to predict the steadystate rainfall necessary for slope failure throughout a catchment. Our primary assumption here is that while local properties surely affect the timing, size, and behavior of a shallow landslide, the dominant control on where they occur is the local surface topography, as it in turn defines local slopeand shallow subsurface flow convergence. The relative simplic- ity of the model is attractive for the typical case where little is known aboutthe spatial variability of the other important factorsthat affect slope stability. The coupled model delin- eates areas of the landscape with similar topographic control on shallow landslide initiation. Previous Work There are many approaches to assessing landslide haz- ards. The most widely used techniques include (1) field inspection usinga check list to identify sitessusceptible to 1!53