Development of Empirically-Based Fragilities of Residential Damage in the 2011 Joplin, Missouri, Tornado

Abstract

Performance-based engineering (PBE) is a methodology that requires specification on a range of performances or target reliabilities for structures of different importance. Information on these ‘performance levels’ require a probabilistic assessment of the potential factors that may influence a design, including information on the hazard, load, resistance, loss estimates, expert opinion and public perception. This paper describes one such probabilistic assessment in the development of empirically-based fragility functions for tornadoes using damage assessment data and a tornado wind field model for the 22 May 2011 Joplin, MO tornado. The damage assessment data was collected during field surveys of more than 1,240 structures in the aftermath of the tornado, using provisions of the Enhanced Fujita (EF) Scale to assess the damage. The wind field model was developed from the tree-fall patterns noted in the damage path of the tornado. Fragility functions were developed for the Degrees of Damage (DOD) associated with One- and Two-Family Residences in the EF Scale. The empiricallyderived fragility functions were progressive in nature, with median wind speeds varying from 33.6 m/s for initiation of visible damage to 85.2 m/s for complete destruction. These functions were compared to existing fragility functions for straightline winds to evaluate potential differences in failure mechanisms for structures exposed to tornadoes. Wind speeds associated with the median failure probability were used to estimate load factors, defined as the square of the ratio of the straightline wind speed to the tornado wind speed. Structures tended to fail at lower wind speeds in tornadoes than in straightline winds, with load factors between 1.32 and 1.51. A fragility assessment in the context of PBE naturally requires attribution and quantification of all uncertainties. Uncertainties in the both the damage and wind speed estimation in the development of fragilities are quantified and assessed using Monte Carlo methods. Preliminary results show variance in fragility parameters is higher for higher damage states but all damage states have relatively low coefficients of variation.