The goal of this project is to improve understanding of and decision support for evacuation and mass case sheltering in hurricanes. The task of moving tens or even hundreds of thousands of people from a wide geographic area in only a few days or hours under uncertain, dangerous conditions, getting them to safe locations, and keeping them safe until they can return home is an extraordinarily complicated process, and as Hurricane Katrina made abundantly clear, the stakes are high. Despite a lot of progress, recent events and unchecked population growth in hurricane-prone regions assure us that many challenges remain. The traditional, conservative approach of evacuating everyone thought to be at risk is no longer feasible in many areas in which there are simply too many people and too little transportation capacity. We propose a fundamentally new approach. In the past, math modeling in this application has been limited to estimating the time required to clear a region, assuming many characteristics of the problem are uncontrollable input (e.g., shelter locations). Instead, we will develop sophisticated optimization models with an expanded decision frame that focuses on higher-level objectives, such as minimizing life loss, cost, and inequity, and considers the full range of strategic and operational evacuation and sheltering strategies in meeting those objectives, including for example, vertical evacuation and strategically locating shelters. These models will be developed through a tight interaction between sociologists and engineers to ensure they are firmly grounded in the reality of people?s behavior. For the first time, the models will be based on individual hurricane scenarios instead of conservative aggregations of many events, and they will be dynamic, accounting for the fact that officials can update their decisions as an event unfolds and information about the situation changes. The project has 5 main steps: (1) determine a set of hurricane scenarios for use in evacuation and shelter models such that they appropriately represent the full range of possible events, but are few enough to allow detailed analysis with each; (2) conduct focus groups of key decisionmakers and stakeholders to identify and characterize appropriate decision objectives, constraints, assumptions, and possible evacuation and shelter management strategies; (3) using the focus group input, develop two mathematical optimization models?one long-term strategic and one short-term operational?for evacuation and sheltering decisions; (4) conduct surveys of affected citizens to ensure that the optimization model assumptions and results make sense; and (5) demonstrate the models through case studies in North Carolina and Florida. Any evacuation and sheltering planning effort is only as effective as its weakest link, so it requires a broad range of expertise from marine science, transportation engineering, risk modeling, optimization, and behavioral research collaborating closely. We have assembled this expertise on the project team. This project will help begin to transform the way hurricane evacuation and sheltering are conducted in the U.S., addressing many of the known limitations of the current approach. The new understanding and optimization models developed in this project will help local and state emergency managers better plan for hurricane evacuation and sheltering, thus reducing the deaths, injuries, and unnecessary expense associated with poorly planned or executed response in future hurricanes. By collaborating throughout the project with state and local emergency management departments and the American Red Cross, the key agencies in charge of hurricane evacuation and sheltering, we will ensure that study results are disseminated to practitioners as quickly and effectively as possible. Three graduate students will participate in all aspects of the research, each with at least two of the co-PIs on their committees to ensure tight integration. By providing a substantive example of truly interdisciplinary disaster research, the project will help facilitate the transformation of the well-known Disaster Research Center, historically based in sociology, into an interdisciplinary center. It will also help to launch the new interdisciplinary graduate program in Disaster Science and Management at the University of Delaware.
SHELTER/EVACUATION MODELING. We developed two computer optimization models—one for long-term planning for mass care shelters, the other for short-term operational evacuation decisionmaking; and a traffic model that is part of both. We demonstrated all models for real case study applications in North Carolina. Together the results of this project provide a fundamentally new framework to think about hurricane evacuation and mass care sheltering. The shelter model determines which shelters should be maintained in the long-term, and which should be opened in a specific hurricane. Currently a long list of churches and public buildings are considered possible shelters, but none were specifically designed for that purpose. With a smaller number of carefully selected shelters, effort can focus on ensuring they are properly prepared to provide sufficient safe, comfortable shelter for any possible hurricane. We developed an integrated evacuation/shelter model that selects those shelters so they are robust given the range of possible hurricane events and to minimize travel times by directing shelter traffic away from the traffic going to non-shelter destinations. The evacuation model determines, as a hurricane approaches shore, who should stay, and who should leave and when, so as to minimize both risk and travel time. Novel features include: (1) refocusing the decision on the true objectives of minimizing both risk and travel time; (2) allowing direct comparison of more alternatives, including for the first time, sheltering-in-place (i.e., not evacuating); (3) using a hurricane-scenario based analysis that represents the critically important uncertainty in hurricane track, forward speed, and intensity; and (4) including our new traffic model. TRAFFIC MODELING. The evacuation model requires knowing where people are over time during the evacuation so their locations relative to flood waters and high winds can be determined. To accomplish that, we developed a new dynamic traffic model that for the first time is efficient enough to be applied to real regional, multi-day hurricane evacuations and does not use the unrealistic assumption that users cooperate to minimize total system travel times. We validated the model with data from Hurricane Katrina and found it to be about 1000 times faster than an alternative model with no loss in accuracy. HAZARD MODELING. Instead of using conservative aggregations of many hurricanes as is standard practice, the new models use carefully selected sets of individual hurricane scenarios that together represent the full range of possible storms while being few enough in number that the models can be run. We developed the new optimization-based probabilistic scenario (OPS) method to determine that set of hurricane scenarios, considering wind and storm surge together. When applied to North Carolina, the analysis produced a set of 100 hurricane scenarios to input into the shelter planning model. For each scenario, it produces a likelihood of occurrence and wind speeds and flood depths at every location in the study region, as they change during the event. For the short-term evacuation model, since it is to be applied after a hurricane has formed, the scenarios must all be consistent with the track up to the time of application. We developed a set of events consistent with Hurricane Isabel for the evacuation model case study. The traffic and hazard models can be useful for many other applications as well. BEHAVIORAL ANALYSIS. Focus groups were conducted with key community stakeholders, including government and non-governmental decisionmakers. Quick response field work was conducted following the March 2011 Tohoku earthquake in Japan. Telephone surveys were conducted with North Carolina households both before and after Hurricane Irene threatened the state in August 2011. Quantitative data analysis is underway, however, findings point to differential decisionmaking for distant future events compared with near future events; disagreements over the feasibility of contraflow solutions, fuel availability, and the suitability of shelters to accommodate people with disabilities or health considerations. Framing of essential personnel is important to effective evacuation and sheltering planning. Some strategies, which constitute segments of disaster planning, indirectly rely on the presence and participation of among non-essential personnel, whose potential role to an organization or community in a disaster actually serves a critical node in the response’s success. Data also point to fail points with contracts, such as Memoranda of Understanding and Mutual Aid Agreements. Even when planning takes place, reliance on identical resources (e.g., fuel sources, vendors) as other organizational entities can render plans as fantasy documents, despite good intentions. In Japan, we observed the challenges of planning for events seen as outside the probability of occurrence, and how evacuation and sheltering strategies for more routine disasters did not always achieve safety for residents or responders. One post-doc, nine graduate students, and many undergraduates were funded through this project, enhancing their research skills and advancing development of the next generation of scholars and practitioners. Results have already been presented in six peer-reviewed journal articles, a book chapter, and numerous conferences.
