Validation of an Augmented Parcel Approach for Hurricane Regional Loss Assessments

• DOI: 10.1061/nhrefo.nheng-1649
• Publication Date: 2023-04-18T16:27:57Z
• Authors: Tracy Kijewski-Co...; Barbaros Cetiner; Kuanshi Zhong; Chaofeng Wang; Adam Zsarnoczay; Yunhui Guo; Meredith Lochhead; Frank McKenna
• Keywords: Ground truth; Natural hazard; Resilience

While simulation environments for the study of community resilience are rapidly advancing, they remain constrained by completeness inventory data. This paper presents an augmented parcel approach leveraging various sources open data, machine learning modules, and time-evolving rulesets to support Hazus-compatible risk assessments on a wide class buildings under hurricane wind flood hazards. These techniques implemented within open-source regional loss assessment workflow Natural Hazards Engineering Research Infrastructure (NHERI) SimCenter. Illustrative examples demonstrate building generation in both data-rich data-scarce environments. The study's validations computer vision–based modules underscore importance training "in wild" images labeled with explicit knowledge region representative architectural nuances such as carports. Validations further reveal challenges simplifying complex contemporary roof geometries simplified shapes adopted Hazus criticality accurate year built given approach's reliance code-based rulesets. Published field observations collected Lake Charles, Louisiana, following landfall Hurricane Laura, that use SimCenter's yields damage states consistent ground-truth minor moderate states. Simulations extreme (characterized fewer observations) bias toward undamaged structures plateau at even severely damaged collapsed buildings. trend persists when considering uncertainty hazard intensity, well low rates shutter compliance. Root causes inconsistencies revealed this validation exercise will require processing street-level panoramic generate more samples post-2007 construction.