Laureano F. Escudero (Universidad Rey Juan Carlos), M. Araceli Garín (Universidad del País Vasco), Juan F. Monge (Universidad Miguel Hernández) and Aitziber Unzueta (Universidad del País Vasco).

Abstract. A preparedness resource allocation model and an algorithmic approach are presented for a three-stage stochastic problem for managing natural disaster mitigation. That preparedness consists of warehouse location and capacity assignment and the procurement of commodities on the one hand and refurbishing the rescue network infrastructure on the other. Two types of uncertainty are considered: exogenous uncertainty which is due to the lack of full knowledge about the probability and intensity of the disaster for each focal point in a given network; and endogenous uncertainty which is based on the decision-maker’s investment to obtain greater accuracy in regard to the occurrence of the disaster and to reinforcing the network infrastructure. A stochastic mixed 0-1 bilinear optimization model is presented. Additionally, a time-consistent stochastic dominance-based risk-averse measure for a set of profiles in a multifunction setting is introduced. Both types of elements imply large-sized problems, so some kind of decomposition algorithmic should be used. Based on the special features of the three-stage problem subject of this work, we introduce the Cluster Dual Descent Algorithm for obtaining feasible solutions based on duality theory. Computational results are reported for a well-known real-life case by comparing the performance of the models based on the alternatives given by the risk-neutral and risk-averse versions jointly with exogenous and endogenous uncertainty.