An Integrated Logistic Model for Predictable Disasters

In the aftermath of a disaster, the relief items are transported from temporary warehouses (Staging Areas, SAs) to the Points of Distribution (PODs). Reducing the response time to provide relief items to disaster victims and cost minimization are two important objectives of this study. We propose an integrated optimization model for simultaneously determining (1) locations of staging areas, (2) inventory assignments to these SAs, (3) selecting sizes and numbers of trucks, and (4) routing of trucks from SAs to PODs. We also introduce another variable, a value function, which forces the model to reduce the logistics response time. We study the interactions among these variables through extensive sensitivity analysis. The time horizon for supply of relief items to disaster areas is usually limited to six days after the disaster occurs. Therefore, we use the proposed optimization model in a rolling‐horizon manner, one day at a time. This reduces daily demand uncertainty. We analyze three disaster scenarios: (1) a low impact disaster, (2) a medium impact disaster, and (3) a high impact disaster. We conduct 720 experiments with different parameter values, and provide answers to the following questions that are useful for the logistic managers: (i) What are the right sizes (in terms of storage capacities) of SAs closer to the PODs? (ii) How should the budget be allocated in a disaster scenario? (iii) What mix of different types (in terms of sizes) of trucks should be selected in a given scenario? The most important managerial insights include: (i) operational budget beyond a limit does not improve the operational efficiency, (ii) when the budget is very low, it is essential to select smaller SAs close to the PODs in order to carry out operations in a feasible manner, (iii) when the impact of disaster is high, it is always beneficial to select larger SAs close to the PODs (as long as the budget is not very low), (iv) when the budget is high and the impact of disaster is not very high, the emergency management administrators need to select SAs prudently based on the tradeoff between the operational cost and the humanitarian value, and (v) the cost of operations is higher when all the trucks are of the same type compared to the case when there is a mix of different types of trucks. Also, we find that the optimal selection of SAs is not impacted by different combinations of the types of trucks. The focus of this study is on disasters that can be forecasted in advance and provide some lead time for preparations, for example, hurricanes. In order to understand the disaster management process of such disasters and develop our model, we (i) interviewed several emergency management administrators, and (ii) studied the disaster management processes available in documents released by various government agencies.