Off-system bridges are essential for rural and underserved communities, providing crucial connectivity to larger transportation networks. Disasters such as tornadoes can severely disrupt these bridges and isolate communities, underscoring the need for robust maintenance and prioritization strategies. This study introduces a digital shadow framework to assess tornado impacts on bridge substructures, integrating data from multiple sources and applying a gravity-based spatial interaction model to estimate economic losses. A generative adversarial network (GAN) with focal loss simulates pre- and posttornado scenarios, generating synthetic postdisaster data that enhance the data set with realistic tornado-induced damage scenarios and shifts in bridge maintenance priorities. The framework showed significant improvements in F1 scores, especially for Condition 5, a minority class with initial F1 scores of 0.39 in both scenarios. Posttuning, the Condition 5 F1 score improved to 0.82, demonstrating the GAN’s effectiveness in representing this vulnerable class. Confusion matrices indicated an increase in bridges classified as Condition 5 posttornado, from 47 to 63, suggesting that the tornado caused considerable damage to previously better-conditioned bridges that now were in this critical category. The study also found that bridges in Conditions 5 and 6 were the most affected before and after the tornado, aligning with digital shadow expectations. The framework identified shifts in maintenance priorities, particularly for bridges in densely populated and tornado-impacted areas, and highlighted disparities in infrastructure prioritization, with minority communities potentially receiving fewer resources than predominantly White areas. To address these disparities, this study recommends an equitable allocation of resources for bridge maintenance, ensuring that minority and underserved communities receive adequate attention. This digital shadow framework provides a valuable tool for proactive infrastructure management, offering insights to support data-driven decisions by federal and local agencies in preparation for extreme weather events.
Bayat et al. (Fri,) studied this question.