Willie A. Deese College of Business and Economics

Multiscale Model for Hurricane Evacuation and Fuel Shortage


Recent hurricanes in the Southeast United States have led to mass evacuations. During the last year’s hurricane Irma, 23 counties in Florida issued mandatory evacuation orders, with the remaining 44 counties putting in place voluntary orders. Our analysis of hurricane Irma traffic data obtained from the Florida Department of Transportation (FDOT) indicates a net exodus of 550,000vehicles from the southern parts of Florida. It is estimated that approximately 6.8 million Floridians and tourists took to the roads in the days leading up to the storm. Such mass evacuations have also been observed during recent hurricane Florence affecting North and South Carolina as well as during this month’s Hurricane Michael.

Hurricane evacuees tend to make longer, intercity trips to stay with friends and family outside the impacted area and to completely move out of the storm path. The high volume mass evacuations, disruptions to the supply chain, long distances traveled, fuel hoarding from non‐ evacuees have led to localized fuel shortages lasting several days in all of the recent hurricanes including Irma, Florence and Michael. Fuel shortages further exacerbate the safe evacuation of residents in affected areas. While news reports have documented fuel shortages during past hurricanes, crowd‐sourced data from social media platform gas buddy has quantified the shortages during recent hurricanes. A predictive model of fuel shortages during hurricanes combined with on‐ground traffic and fuel data will be effective in providing policy suggestions for both effective evacuation and backup fuel storage locations. We propose a computational model at multiple scales to predict the fuel shortages that develop during a hurricane evacuation. Our preliminary analysis of such fuel shortages indicates that models developed for predicting epidemic dynamics can adequately represent fuel shortages. We will develop a multiscale framework with multiple components including:

  1. Compartmental stochastic epidemic spread model for fuel shortages at city and state level. 
  2. Traffic data from past hurricanes (e.g. Irma and Florence), combined with a Monte Carlo fuel consumption model to estimate fuel usage.
  3. Self‐excitation point process model to delineate and estimate baseline and contagious fuel
    consumption (e.g. fuel hoarding, the effect of nearby refueling stations emptying).
  4. Agent-based discrete event simulation model for detailed analysis at the scale of few intersections.
  5.  Policy analysis using a predictive model for future hurricanes based on the above

The project outcome will be a predictive capability for fuel shortages in the event of future hurricanes which can be used for effective analysis of available policy options and for guiding mass evacuations.

CATM Research Affiliates:
Sirish Namilae (ERAU: Lead)
Dahai Liu (ERAU)
Scott Parr (ERAU)