A behavioural simulation study of allocating evacuees to public emergency shelters

Abstract

The opening of public shelters to provide evacuees with access to safety and welfare is a typical response to large-scale emergencies. In some countries, officials allocate people at risk to shelters based on the location of their residence or workplace and this advice is communicated to them in advance of an evacuation so they get to shelters quickly whilst avoiding overcrowding and associated secondary risks in the shelter. When an evacuation is called, some evacuees follow this advice and attend their allocated public shelter but others disregard the advice and select an alternative public shelter (e.g. the most convenient or nearest one) or go to a private shelter (e.g. a hotel or house of friends or family). Using data from Pittsburgh (Pennsylvania, USA), supplemented with data from interviews with 142 emergency officials from 10 countries, this paper explores how evacuee behaviour matters when planning public shelters. Our novel contribution is a model-based integer programming simulation analysis of public shelters from the combined effect of three evacuee behaviours: comply with official advice and attend their allocated public shelter; attend the nearest public shelter; attend a public shelter of personal preference. Sensitivity analyses identifies the importance of these behaviours on the time to evacuate the city, finding that if evacuees disregard the official allocation then widespread over-occupation of public shelters can result which significantly increases the time to evacuate. This paper discusses OR modelling issues and the implications of these results for disaster preparedness, evacuee behaviour and emergency response.