GSA Connects 2022 meeting in Denver, Colorado

Paper No. 141-5
Presentation Time: 9:10 AM

FORECASTING BUILDING DAMAGE USING POST-FIRE DEBRIS-FLOW RUNOUT MODELS


BARNHART, Katherine, Geologic Hazards Science Center, U.S. Geological Survey, Box 25046, MS 966, Denver Federal Center, Denver, CO 80225 and KEAN, Jason, U.S. Geological Survey, Geologic Hazards Science Center, P.O. Box 25046, MS 966, Denver Federal Center, Denver, CO 80225

Post-fire debris-flow runout poses a hazard to buildings located within the extent of inundated area. The impact force of a debris flow can generate building damage ranging from slight to complete. In addition to hazard assessments that identify areas likely to be inundated under hypothetical design scenarios, assessments that forecast building damage may support situational awareness for local decision makers. However, the ability of debris-flow runout models to reliably forecast building damage has not been extensively tested. We evaluated the ability of three debris-flow runout models (RAMMS, FLO2D, and D-Claw) to predict building damage in simulations of the 9 January 2018 Montecito, California, debris-flow event. We used simulations that used a wide range of debris-flow volumes and material properties. Observations of building damage after the event were combined with OpenStreetMap building footprints to construct a database of all potentially impacted buildings. The final building database contained 4002 unimpacted buildings, 127 buildings with 1%–9% damage, 126 buildings with 10%–25% damage, 114 with buildings 51%–75% damage, and 162 destroyed buildings. The use of both observed damage and OpenStreetMap-derived buildings allowed us to calculate the number of damaged buildings in locations with simulated, but not observed, inundation. We used existing fragility functions relating debris-flow depth to damage state for wood framed buildings that were based on observations from the Montecito event. For each simulation, we extracted the maximum flow depth and used it to classify a building into one of the above damage classes. We evaluated the ability of models to reach three increasingly difficult targets for building damage forecasts: the total number of damaged buildings, the number of damaged buildings in each category, and the damage state of individual buildings. At the estimated event volume, all models overpredict the total number of damaged buildings by a factor of 1.5–3x. Forecasts underpredict the number of undamaged buildings, overpredict the number of buildings with 10-75% damage, and are most accurate at predicting the number of destroyed homes. Threat score values are typically less than 0.20, indicating difficulty in predicting the correct damage state of individual buildings.