Rapid sensitivity analysis for reducing uncertainty in landslide hazard assessments

One of the challenges in assessing temporal and spatial aspects of landslide hazard using process-based models is estimating model input parameters, especially in areas where limited measurements of soil and rock properties are available. In an effort to simplify and streamline parameter estimation, development of a simple, rapid approach to sensitivity analysis relies on field measurements of landslide characteristics, especially slope and depth. This method is demonstrated for a case study in Puerto Rico where widespread destruction resulted from tens of thousands of debris flows induced by Hurricanes Irma and María in Puerto Rico in 2017. The approach can be applied to estimation of shear strength as well as hydrologic parameters that control infiltration and flow of water in the subsurface and ultimately the timing of landslides resulting from heavy rainfall. Results narrow the possible range of cohesion and friction parameters as well as hydraulic conductivity and other soil water parameters by counting the fraction of field observations that can be explained by each combination of parameters. For cases studied in Puerto Rico, the method identified combinations of cohesion and friction values that explain more than 80–90% of observed landslide source areas.