CAREER: Transcending Barriers between Natural Hazard Researchers, Educators, and Practitioners - An Integrative Approach to Multi-Hazard Probabilistic Assessment

This Faculty Early Career Development (CAREER) grant will create a strong quantitative and qualitative foundation that enables researchers, students, and other science/engineering professionals from traditionally separate disciplines to work together to develop better strategies to improve the resilience of infrastructure. An essential step toward improving infrastructure resilience involves assessing risks from numerous natural hazards and then using the resulting risk-insights to plan for and respond to events in real-time.

Probabilistic hazard assessments serve as the starting point for risk assessments. Typically, these assessments focus on specific hazards such as earthquakes or hurricanes. However, major disaster events often result from combinations of multiple hazards.

Probabilistic hazard assessments that do not capture “multi-hazard events” can ultimately lead to suboptimal risk mitigation and event response. Research and education gaps currently prevent fully integrated multi-hazard assessments. These gaps exist because the approaches used for various natural hazards were developed in relative isolation.

This has led to meaningful differences in the tools professionals use to assess hazards and how hazard assessment is taught, researched, and addressed in regulation.

This project involves research and education activities centered around three aims that provide a strong foundation for creating dynamically updatable, multi-hazard risk assessment models for spatially-distributed infrastructure. Aim #1 will focus on developing Bayesian-network-based formulations to assess multiple, spatially-distributed hazards. This will provide a unified quantitative and graphical structure for “linking” hazard-specific models based on their probabilistic dependencies.

Aim #2 will focus on computation enabling strategies needed to implement Bayesian networks for probabilistic assessments of multiple, spatially-distributed hazards (modeled as random fields) at practical scales. Physically-informed surrogate models will be developed using statistical and machine-learning methods to reduce the computational (processing) demands associated with generating conditional probability tables required by the Bayesian networks.

Surrogate model development research will be coupled with a project-based university course on the subject. Finally, Aim #3 will focus on creating a dynamically updateable, multi-hazard Bayesian network case study to demonstrate, validate, communicate, and seek feedback on project outcomes.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.