As the 2019 hurricane season escalates, three University of Texas at Austin researchers are addressing the problem of efficiently and safely evacuating hospitals before a hurricane strikes. Dr. Erhan Kutanoglu, of the Operations Research and Industrial Engineering (ORIE) program in the Cockrell School of Engineering, leads the research team. The group received a total of $330,000 from Planet Texas 2050 and the NSF CoPe (Coastlines and People) EAGER program to pursue an interdisciplinary effort. Their approach combines hydrological science and operations research techniques to more effectively integrate weather predictions, flooding models, and stochastic optimization to aid decision makers in the days before a hurricane makes landfall.

Dr. Kutanoglu’s expertise is in optimization and logistics -- that is, how to move goods, or people, quickly and safely through a network. A Co-Principal Investigator (Co-PI) on the project, Dr. John Hasenbein (ORIE), lends his expertise on stochastic network models. These models incorporate the random elements of a decision-making problem – for example, determining how the uncertain path of a hurricane affects preparations several days before landfall. The research team is rounded out by Co-PI Dr. Zong-Liang Yang of UT Austin’s Jackson School of Geosciences. Dr. Yang’s specialty is in hydrological models that predict river runoff and flooding, given soil conditions, surface features, and forecasts of precipitation.

Potentially preventable incidents from past hurricane events created a strong motivation to form the research team. During Hurricane Katrina, 35 nursing home residents died at St. Rita’s Nursing Home, in St. Bernard Parish. During the same event, thousands of patients were trapped for five days at Memorial Medical Center in New Orleans. Sheri Fink won a Pulitzer Prize for her reporting on this incident for the New York Times. During Hurricane Irma, 12 nursing home residents died when the air-conditioning failed at the Rehabilitation Center at Hollywood Hills in Florida. These events give rise to the question: Can flooding predictions and evacuation plans being improved by combining hydrological science and operations research models? The investigators’ two-year plan is to tackle this question and answer it in the positive.

The team has been laying the groundwork for the proposal for months, but the work started in earnest this September, when the two grants were awarded, allowing the team to support graduate students Kyoung Yoon Kim (ORIE) and Wen-Ying Yu (Jackson School) as they focus dissertation work on this topic. The proposed work involves combining several existing, computationally intense models into a single comprehensive decision-making tool. Ideally, hurricane forecasts produced by the National Oceanic and Atmospheric Administration several days before a hurricane’s landfall are fed into two different models: the Probabilistic Tropical Storm Surge model (P-Surge) and the Noah-MP land surface hydrological model linked with the Routing Application for Parallel Computation of Discharge (RAPID).

The P-surge model predicts coastal flooding due to wind and ocean surge at locations directly along the coast. The Noah-MP–RAPID model focuses on river flooding further inland, which is caused primarily by large amounts of precipitation and runoff from rivers and streams. In theory, these models can be combined with floodplain elevation data and data on hospitals and nursing homes using GIS technics, to make correlated predictions on the flooding probabilities, and levels, at each care facility. These “science-based” models can now feed the “engineering” models, which is where the ORIE expertise is brought in. Given a set of flooding scenarios, the ORIE team runs a model known as a stochastic integer program to determine the best place for regional authorities to set up staging areas for ambulances and emergency personnel. The model also determines the optimal routes for evacuation of patients from threatened facilities to safer locations. For now, running the full combination of models requires the powerful resources of the Texas Advanced Computing Center, although the PI’s hope to provide a slimmed down version of the model for implementation.

The team is also collaborating with organizations such as the SouthEast Texas Regional Advisory Council, which coordinates hurricane evacuations in the Houston area, to insure that the models and assumptions are reflective of real-life conditions. The plan is to first do some “hindcasting” on past hurricane events to see what could have been done in a more efficient way, had the models been in place at the time of Harvey, for example.

The initial goal is to have workable software in place for the 2020 hurricane season. The group also have their sights set on broader related work: predicting flooding effects on the electric grid and incorporation long-term climate change trends to inform regional planning between hurricane seasons. Hopefully, this work will lead to greatly reduced deaths, and fewer perilous patient situations, as society battles the inevitable arrival of future hurricanes.