Optimization of the issuance of evacuation orders under evolving hurricane conditions

This paper develops a bi-level programming model to optimize the issuance of evacuation orders with explicit consideration of (i) the highly uncertain evolution of the storm, and (ii) the complexity of the behavioral reaction to evolving storm conditions. A solution procedure based on progressive hedging is developed. A realistic case study for the eastern portion of the state of North Carolina is presented. Through the case study we demonstrate (1) the value of developing an evacuation order policy based on the evolution of the storm in contrast to a static policy; (2) the richness in the insights that can be provided by linking the behavioral models for evacuation decision-making with dynamic traffic assignment-based network flow models in a hurricane context; and (3) the computational promise of a progressive hedging-based solution procedure to solve large instances of the model.