Location-allocation models for relief distribution and victim evacuation after a sudden-onset natural disaster
2015-03-04T10:14:49Z (GMT) by
Quick response to natural disasters is vital to reduce loss of and negative impact to human life. The response is more crucial in the presence of sudden-onset, difficult-to-predict natural disasters, especially in the early period of those events. On-site actions are part of such response, some of which are determination of temporary shelters and/ or temporary medical facility locations, the evacuation process of victims and relief distribution to victims. These activities of last-mile disaster logistics are important as they are directly associated with sufferers, the main focus of any alleviation of losses caused by any disaster. This research deals with the last-mile site positioning of relief supplies and medical facilities in response to a sudden-onset, difficult-to-predict disaster event, both dynamically and in a more coordinative way during a particular planning time horizon. Four mathematical models which reflect the situation in Padang Pariaman District after the West Sumatera earthquake were built and tested. The models are all concerned with making decisions in a rolling time horizon manner, but differ in coordinating the operations and in utilization of information about future resource availability. Model I is a basic model representing the current practice with relief distribution and victim evacuation performed separately and decisions made only considering the resources available at the time. Model II considers coordination between the two operations and conducts them with the same means of transport. Model III takes into account future information keeping the two operations separate. Model IV combines the features of Models II and III. The four models are approached both directly and by using various heuristics. The research shows that conducting relief distribution and victim evacuation activities by using shared vehicles and/or by taking into account future information on resource availability improves the current practice . This is clearly demonstrated by the experimental results on small problems. For large problems, experiments show that it is not practical to directly solve the models, especially the last three, and that the solution quality is poor when the solution process is limited to a reasonable time. Experiments also show that the heuristics help improve the solution quality and that the performances of the heuristics are different for different models. When each model is solved using its own best heuristic, the conclusions from results of large problems get very close to those from small problems. Finally, deviation of future information on resource availability is considered in the study, but is shown not to affect the performance of model III and model IV in carrying out relief distribution and victim evacuation. This indicates that it is always worthwhile to take into account the future information, even if the information is not perfect, as long as it is reasonably reliable.