Date of Graduation


Document Type


Degree Name

Master of Science in Industrial Engineering (MSIE)

Degree Level



Industrial Engineering


Ashlea Bennett Milburn

Committee Member

Sarah Nurre Pinkley

Second Committee Member

Shengfan Zhang


Disaster Response Shelter Location, Hurricane Florence, Optimization, Potential Spatial Ability


In the immediate response phase of a natural disaster, local governments and nonprofit agencies often establish shelters for affected populations. Decisions regarding at which locations to open shelters are made ad hoc based on available building inventory, and may result in high travel impedance to reach shelters and congestion. This thesis presents a shelter location optimization model based on the two-step floating catchment area (2SFCA) method. The 2SFCA method creates a shelter accessibility score for each areal unit (e.g., census block group) which represents the ability for persons in the unit to access shelter capacity with low travel impedance, relative to persons in other units competing for the same shelter capacity. A distance decay function within the 2SFCA method models the propensity of a person to visit a shelter based on the distance to the shelter. The optimization model recommends locations at which to open shelters so as to optimize some function of the 2SFCA accessibility scores. Three single-objective models and one bi-objective model are considered. Across all areal units, the alternative models: (i) maximize the sum of accessibility scores; (ii) minimize the disparity in accessibility scores; (iii) maximize the minimum accessibility score; and (iv) maximize the sum of all scores and minimize disparity. These models are demonstrated via a case study based on Hurricane Florence, which struck North Carolina in 2018. The optimization model outputs are compared with actual shelter openings during Hurricane Florence in four North Carolina cities, and also with outputs of classic p-Median and p-Center facility location models. Case study results demonstrate that, across the range of parameter values included in a sensitivity analysis, the bi-objective model achieves the best tradeoff between efficient and equitable shelter locations, while also achieving a higher minimum accessibility score than either of the two single objective models on their own.