Disaster Relief Operations Planning

Effectively responding to disasters such as earthquakes, floods, tsunamis, avalanches, hurricanes, volcanic eruptions, forest fires, terrorist attacks etc. is of crucial importance for life and well-being of humans and deserves for this reason utmost efforts. Nevertheless, surprisingly enough, the logistics of disaster management seems to be a somewhat underdeveloped part of management science, compared to other branches of logistics. This holds also for the more specific area of model-based, computer-aided decision support for disaster relief operations planning. The reasons for this only moderate level of development certainly include the considerable methodological challenges one has to face when trying to apply quantitative models to the field of disaster relief and to implement them on computers: In disaster management applications, data are often uncertain, states are quickly changing, fast decisions based merely on partial information are required, and not only technological, but also social, psychological and cultural factors play a role. The task of disaster response and relief is usually mainly carried out by non-governmental organizations, supported sometimes by military units and/or police forces. In the process models of the aid organizations, several phases of a disaster management mission are distinguished. Whereas the first of these phases has inevitably to be covered by local staff and organizations and by the affected population itself, in the second phase, broader support is possible and sometimes necessary: in particular, large-scale disasters as well as disasters in poor countries need international help. In this context, the International Red Cross with its national member societies often plays a major role. In our proposed project, we do not focus on that part of aid and emergency response that has to be carried out immediately after the onset of a disaster, when local aid organizations as ambulances or fire fighters are in charge. Instead, we consider the problem situation faced at the time of the arrival of one or several international aid organizations. The main challenge at this time is to build up an auxiliary supply structure as a compensation for the structure destroyed by the disaster. People have to be accommodated with fresh water, food, medicine, clothes, sometimes also tents and other material. In view of damaged roads and reduced vehicle fleet, it is important to assure in a possibly unconventional way that all these materials reach the consumer. This can only be done by establishing an efficient distribution network. For the distribution process, intermediate stores (regional depots) have to be set up, vehicle tours and delivery stops on them have to be selected, and a schedule for the vehicles has to be designed. It is very important that the people in need have access to the basic care materials on a regular basis. For psychological reasons, every needy person should be confident about a regular schedule with fixed day and time when the assigned vehicle delivering the required goods arrives. Considering the literature on model-based decision making in transportation, especially three problem formulations turn out as particularly relevant for the application context: the Warehouse Location Routing Problem, the Covering Tour Problem, the Periodic Routing Problem. An important feature of disaster recovery planning, distinguishing it from some other application fields of logistic planning, is that the planning problem is stochastic by nature: Decisions have to be made under uncertainty. Therefore, besides a deterministic consideration of the problem, we devote attention also to the stochastic aspect by appropriate extensions of the used optimization models. Moreover, contrary to applications where all objectives can be reduced to a single benefit or cost measure, the diverse and incommensurable criteria that play a role in disaster recovery planning often disallow a quantitative treatment where the objectives are simply combined in a weighted average or another type of aggregation. Some of the relevant criteria are monetary, others express values as human life, health, security, or equity. Although trading them off against each other is finally inevitable, it seems preferable to apply a multi-criteria decision analysis which does not force the decision maker to decide in advance on weights, but offers instead a set of reasonable possible solutions that can be discussed on a human decision level after the application of the computer-supported analysis tool.

The project "Katastropheneinsatzplanung" focuses on disaster operations planning in developing countries. Three levels of disaster operations planning are investigated: strategic, tactical and operational. The different problem situations are caused by natural disasters, such as earthquakes, floods or tsunamis, which (partly) destroy the infrastructure in the affected regions. On a strategic level, a model for multi-objective decision analysis with respect to the location of public facilities, such as schools, is developed. For the location of the public facilities, tsunami hazards are taken into account. On a tactical level, a warehouse location routing problem and a multi- objective covering tour problem for the distribution of disaster relief supplies are considered. In a first step, critical items are transported with a homogeneous vehicle fleet. In a second step, a more realistic problem setting is investigated, where road networks are damaged and disaster relief supplies have to be carried by heterogeneous modes of transport, each of them being suitable for a specific category of road. On an operational level, five approaches for measuring the risk of the planned covering tours to become impassable are developed. Risk is considered as an additional objective and the solution method is extended by an enrichment phase in order to generate efficient compromise solutions. The multi-objective disaster operations problems are modelled and solved with commercial solvers. Only for small instances exact solutions can be determined. Therefore metaheuristic search techniques have to be developed. The heuristic solution approach are based on genetic algorithms and variable neighbourhood search for each level of disaster operations planning, taking the specifics of the different problem formulations into account. With the described solution methods decision makers could be supported in planning efficient disaster operations.