Hurricane evacuation

Background

Natural or manmade disasters present multiple risks to a community, with evacuation being one of the most effective means of reducing mortality and morbidity associated with natural disasters. Evacuation has particular importance for events with forewarning, such as the tropical storms and hurricanes which threaten the southern and eastern coastlines of the United States. In response to the devastating hurricane seasons in 2004 and 2005, Florida developed one of the most comprehensive evacuation plans in the country. However, this did not prevent the traffic management problems derived from hurricane evacuations. During Irma, about 6 million residents in Florida were evacuated from coastal areas, creating the largest mass evacuation in U.S. history and significant traffic jams and fuel shortages. 

Recent negative evacuation experiences, however, have also presented an opportunity for improving evacuations through the availability of new data sources. Most notably, the proliferation of cell phones and connected navigation devices have created data sources that provide the traffic data necessary to improve evacuation modeling. The newly available presence of these data has led to the emergence of a field that is dedicated to using mobile location data to model human mobility This provides a research opportunity to use mobile location data to improve hurricane evacuation modeling.

Research Objectives

  • Review the social factors influencing evacuation and determine how well they are captured in prominent composite indices measuring Disaster Resilience, Social Vulnerability, and Social Capital currently used in disaster planning efforts.
  • Analyze the influence of social demographic variables on travel behavior during a hurricane evacuation using mobile location data.
  • Propose a protocol demonstrating the implications of different data manipulation and belief updating methods on computational results when using Evidence Theory.
  • Demonstrate the application of Evidence Theory to incorporate highly uncertain sensor data in pavement condition assessment.

Contributions​

Research began with a literature review of evacuation research. Results revealed the need to include more human-centric indicators into evacuation modeling. This motivated a study of the ability to analyze social influences in evacuation data. This analysis used mobile location data and observed traffic counts to analyze evacuation behavior during Hurricane Michael in 2018.

Results demonstrate that social influences in evacuation can be observed using mobile location data. The results show that at certain traffic sites, certain social groups of different income and race exhibit different travel behavior during evacuation as compared to normal conditions. The variability of results across the traffic sites analyzed also highlights the uncertainty associated with determining social influences.

The need to understand the uncertainty of social influences and combine diverse data sets, such as traffic counts and evacuee behavioral surveys, motivated research into alternative methods of uncertainty analysis. Probability Theory-based methods may not be well suited to address the uncertainty present in social data.

One method well suited for the combination of uncertain data is Evidence Theory. A review of Evidence Theory computation and combination methods led to the development of a protocol for Evidence Theory applications. The protocol addresses assignment of belief mass, computational implication of combination methods, and the commonalities of different methods. The protocol facilitates sensitivity analysis of Evidence Theory output. The protocol, therefore, enables a secondary analysis of the results of Evidence Theory applications, highlighting uncertainty among possible outcomes.

The protocol was then applied to predict pavement condition, demonstrating the concepts addressed by the protocol. The demonstration application also allowed a comparison of Evidence Theory to Probability Theory-based methods, such as Markov Decision Process (MDP). The comparison demonstrates the effectiveness of Evidence Theory and the value of sensitivity analysis.

Overall, this research contributes to the field of transportation system decision making by identifying needs and applicability of data in social analyses. This research applies Evidence Theory to combine uncertain data. The awareness of data needs and data combination methods support decision-making and communication with uncertain data.

Funding

US Department of Education Graduate Assistance in Areas of National Need (GAANN) Fellowship

Publications

  • Seites-Rundlett, W.; Corotis, R.; Torres-Machi, C. (2022) Development of a Protocol for Engineering Applications of Evidence Theory. Submitted to: ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 8(3): 04022036, DOI: 10.1061/AJRUA6.0001241.
  • Seites-Rundlett, W.; Bashar, M.; Torres-Machi, C.; Corotis, R. (2022) Combined Evidence Model to Enhance Pavement Condition Prediction from Highly Uncertain Sensor Data. Reliability Engineering & System Safety, 217, 108031, DOI: 10.1016/j.ress.2021.108031
  • Seites-Rundlett, W.; Garcia-Bande E.; Alvarez-Mingo, A.; Torres-Machi, C.; Corotis, R. (2020) Social Indicators to Inform Community Evacuation Modeling and Planning. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 6(3):03120001, DOI:10.1061/AJRUA6.0001069.