A sequential statistical decision model for hurricane disaster relief planning
by
Selda Taskin
Coauthors: Emmett J. Lodree, Jr.

This research introduces stochastic inventory control models that are relevant to planning for potential disaster relief activities associated with hurricane events. In particular, this research is motivated by manufacturing and retail firms whose demand for products such as flashlights, batteries, and gas-powered generators is influenced by the characteristics of the hurricane season. We develop a stochastic dynamic programming recursion with Bayesian updates whose solution yields the optimal inventory level for hurricane supplies as well as the optimal timing of the inventory decision. Our model is unique in that it leverages evolving forecast updates into the decision model. The proposed model accounts for the trade-off between forecast accuracy and logistics cost efficiency as a function of time. An example problem involving real hurricane wind-speed data is presented to illustrate the methodology.

Keywords: Sequential statistical decision processes, hurricane logistics planning, optimal stopping problem, Bayesian updating, stochastic dynamic programming

Date received: April 13, 2007