Designing Valid Humanitarian Logistics Scenario Sets: Application to Recurrent Peruvian Floods and Earthquakes

Introduction

A variety of approaches, ranging from analytical models and theories to case studies, have been considered to manage risks during disaster operations. In the field of Humanitarian Logistics (HL), mathematical programming is the most frequently used research methodology (Galindo and Batta, 2013). While the use of optimization tools and algorithms has been shown to have a great potential to improve disaster management practices, they are rarely used in the field (Laguna Salvadó et al., 2015; Laguna Salvadó et al., 2016). Hence, the lack of an easy-to-use and established approach to risk assessment means that in practice, decision-makers often refer to their experience and intuition, which can lead to a range of biases and loss of performance (Comes, 2016). As demonstrated by (Charles et al., 2016), this statement is mainly due to research works that frequently use fictitious scenarios and data compensating for the lack of information. This approach fails to validate whether decision support systems can be successfully applied in the actual context of disaster relief (Charles et al., 2016). Real cases, or at least realistic ones, with accurate data are necessary to enable practitioners to be confident with the results of scholar and to start to use them concretely in the field. This chapter tackles this issue by suggesting an innovative methodology able to generate valid and realistic scenario sets on future disaster trends as suggested by (Galindo and Batta, 2013; Pedraza-Martinez and Van Wassenhove, 2013). We therefore develop a series of requirements that are designed to support researchers in producing valid and plausible scenarios for their quantitative decision support systems that are tailored to fit the needs and standards of field-based decision-makers. Basically, such systems should be able to support HL decisions such as location-allocation, routing or inventory management for instance.

When referring to disasters, most of us will intuitively refer to mega-disasters such as Indonesia’s tsunami in 2004, Haiti’s earthquake in 2010, Japan’s earthquake / tsunami in 2011 or Nepal’s earthquake in 2015. Although all those cases have had dramatic consequences, they are far from typical for disaster response. Ferris et al. (2013) define the notion of “recurrent disaster” as “the repeated occurrence of a unique natural hazard in the same geographical region”. Since 2000, each year, more than 400 disasters have been recorded in the disaster database EM-DAT (http://www.emdat.be). More than 90% of those disasters recur in the same regions: cyclones in the Caribbean, earthquakes in the Pacific Ring of Fire or floods in South-Eastern Asia. In this chapter, we focus on recurrent disasters, which constitute the great majority of disasters.

To conduct empirically grounded work that enables HL practitioners to analyse the implications of their HL decisions (such as planning, routing, allocating…), we suggest using a scenario-based approach. Scenario based reasoning has been advocated for its flexibility and appeal to the user, particularly in complex situations (Comes et al., 2015). Scenarios are understood as a means for exploring eventualities before they occur. They support users to think through a variety of different situations, and as such are well-positioned for HL decisions support. We here propose an approach that avoids some of the most common pitfalls of a too narrow or biased set of scenarios, which reflects opinions of a small number of experts, or is subject to groupthink (Wright et al., 2009; Comes et al., 2012). Our approach guarantees that each individual scenario is sufficiently plausible (i.e. a good assessment of truth (Bosch, 2010)) and relevant for feeding HL decision support systems.