Abstract
<jats:p>Community resilience assessment under a given hazard scenario is commonly performed by representing the building inventory through archetypes and assigning fragility functions for multiple limit states to each archetype. However, because these archetypes and fragility functions may be developed from design provisions, mechanics-based analyses, expert judgment, or historical event data, their transferability to other risk contexts remains uncertain, and the errors introduced by such generalization are not yet well understood. This paper presents a framework for quantifying modeling error and evaluating the transferability limitations of archetype-based fragility models when applied beyond their original risk contexts. The framework integrates exposure data, hazard characterization, fragility-based damage analysis, post-event reconnaissance observations, and risk quantification to compare predicted and observed damage states. The framework is demonstrated across five tornado-affected communities from three tornado events, exposure conditions, and community contexts. Model performance is assessed at both community and building scales by characterizing discrepancies between field observations and model predictions. Statistical associations within the examined testbed conditions are evaluated to assess how effectively the models reproduce observed damage patterns. The resulting community testbeds provide quantitative evidence regarding modeling error, uncertainty, and transferability limitations under the examined event conditions, while supporting greater consistency between post-event reconnaissance and community model development. The findings demonstrate that current fragility frameworks, in their existing form and without systematic multi-event calibration, are not yet ready for generalized operational prediction and deployment across heterogeneous tornado environments.</jats:p>