PHYSICS-BASED, PROBABILISTIC, AND PRACTICAL: LANDSLIDE SUSCEPTIBILITY MODELING WITH PISA-M

Physics-based landslide modeling is difficult and data intensive. Generating high quality and practical map results is often not feasible outside of small and thoroughly characterized study areas. Using the physics-based program Probabilistic Infinite Slope Analysis (PISA-m), users can perform expedient assessment of landslide hazard over large study areas where comprehensive geotechnical data may be lacking but other data inputs are robust. PISA-m uses an infinite slope equation and spatial layers of a digital elevation model (DEM), forest cover, and soil characteristics to calculate a probability of factor of safety (FS) being less than or equal to 1. A wide variety of spatial inputs are possible to use in PISA-m, for example layers such as mapped Unified Soil Classification System (USCS) extents, shale beds approximations, or bedrock geology can be used for soil characterization. The diverse inputs for this program demonstrate the potential wide use applications of PISA-m to create practical, first-order approximation landslide susceptibility models. This investigation considers two study areas of different landslide magnitudes and implements four different soil unit inputs; shale beds with alluvium and colluvium, 1:24,000 scale bedrock units, USCS with borehole derived geotechnical values, and USCS with general geotechnical values. We compared model results to landslides that post-date lidar DEM to evaluate the accuracy of PISA-m as regional landslide susceptibility model. PISA-m results modeled a high probability (0.75-1.0) of FS≤1 for a given landslide point location approximately 39% of the time. Approximating a more precise landslide location with a buffer increases intersection with high probabilities of FS≤1 to 96%. The intersection of landslides with indicative FS probabilities from the models demonstrate the practical use cases of PISA-m for regional landslide susceptibility assessment. While our PISA-m results are promising, these outcomes are dependent on the high-resolution input data and expert knowledge of ground characteristics to bolster the lack of precise and thorough geotechnical descriptions.