Bridge scour is the leading cause of bridge damage nationwide. Successfully mitigating bridge scour problems depends on our ability to reliably estimate scour potential, design safe and economical foundation elements that account for scour potential, identify vulnerabilities related to extreme events, and recognize changes to the environmental setting that increase risk at existing bridges. This study leverages available information, gathered from several statewide resources, and adds watershed metrics to create a comprehensive, georeferenced dataset to identify parameters that correlate to bridges damaged in an extreme flood event. Understanding the underlying relationships between existing bridge condition, fluvial stresses, and geomorphological changes is key to identifying vulnerabilities in both existing and future bridge infrastructure. In creating this comprehensive database of bridge inspection records and associated damage characterization, features were identified that correlate to and discriminate between levels of bridge damage. Stream geomorphic assessment features were spatially joined to every bridge, marking the first time that geomorphic assessments have been broadly used for estimating bridge vulnerability. Stream power assessments and watershed delineations for every bridge and stream reach were generated to supplement the comprehensive database. Individual features were tested for their significance to discriminate bridge damage, and then used to create empirical fragility curves and probabilistic predictions maps to aid in future bridge vulnerability detection. Damage to over 300 Vermont bridges from a single extreme flood event, the August 28, 2011 Tropical Storm Irene, was used as the basis for this study. Damage to historic bridges was also summarized and tabulated. In some areas of Vermont, the storm rainfall recurrence interval exceeded 500 years, causing widespread flooding and damaging over 300 bridges. With a dataset of over 330 features for more than 2,000 observations to bridges that were damaged as well as not damaged in the storm, an advanced evolutionary algorithm performed multivariate feature selection to overcome the shortfalls of traditional logistic regression analysis. The analysis identified distinct combinations of variables that correlate to the observed bridge damage under extreme food events.