LANDSLIDE FREQUENCY AND LOSS ESTIMATES IN EVALUATING LANDSLIDE HAZARD ASSESSMENTS

Many areas experience heavy losses from landslides, including deaths, injuries, and property damage. Losses may result from a landslide’s sudden onset and rapid movement or from insidious destruction of property through slow, persistent landslide movement. Cost of historical landslide losses for several U.S. metropolitan areas and Puerto Rico exceed traditionally accepted levels of risk by factors of roughly 10 – 250 at comparable probabilities. Great potential exists to reduce losses in such areas if high-hazard locations can be distinguished sufficiently to manage risk. Landslide susceptibility maps are often subjected to statistical tests to validate them and evaluate their performance. The tests evaluate the predictive skill of a map but may not indicate how well it delineates areas of high and low landslide probability (hazard). The choice of how many susceptibility (or hazard) categories and the boundaries between them may be arbitrary. Combining landslide density with estimates of direct losses and landslide frequency provides a check on hazard categories and boundaries for any susceptibility map. For example, in Seattle, Washington, 93% of historical landslides initiated on landslide landforms, which account for 15% of the land area (Schulz, 2007, https://doi.org/10.1016/j.enggeo.2006.09.019). Landslides occurred at average annual rates of 0.2 – 1.1 y^-1km^-2 in this area and 0.004 y^-1km^-2 elsewhere. In addition, historical loss costs in the highly susceptible landslide landforms exceeded acceptable risk whereas losses elsewhere were acceptable (well below the risk-acceptance threshold) assuming uniform cost per landslide. Thus, the landform-defined susceptibility units successfully define high-hazard areas. The situation is more challenging at a study area in Puerto Rico where susceptible land holding 86% of landslides covers 35% of the total land area and susceptible land holding 97% of landslides covers 51%. Hazard delineation comparable (similar separation between high and low densities and annual frequencies) to that achievable in Seattle with 15% of the land requires 63% of the land (containing 99% of the landslides) at the Puerto Rico site. This approach can be applied to susceptibility maps of areas for which sufficient landslide density, frequency, and loss data are available.