An exploration of the nearest-shelter assumption in shelter location models

Abstract

Many existing shelter location models assume displaced persons will go to the nearest shelter. This paper examines the nearest-shelter assumption for actual response shelter configurations in three North Carolina cities following Hurricane Florence and two model-prescribed configurations for the same event. The focus is on shelter placement decisions for evacuees whose accommodation type is public shelter and whose destination choice is local. Case study results indicate that a displaced person's second-nearest shelter often requires less than one mile of incremental travel beyond the distance to the first-nearest shelter. Further, nearest-shelter behavior results in demand imbalance across open shelters. Finally, nearest-shelter predictions of the proportions of shelter seekers choosing each open shelter do not align with historical client data. Taken together, these observations do not support a one-size-fits-all assumption of nearest-shelter behavior. Empirical research to understand factors influencing shelter destination choice is needed, along with new shelter location planning models capable of considering these new and more comprehensive descriptors of shelter-seeking behavior.