A paper recently co-authored by Md Murad Hossain Khondaker, a civil and environmental engineering Ph.D. candidate at the University of Houston, may hold key insights that allow for better predictions of catastrophic flooding produced by hurricanes in the United States.
“Weather prediction models still struggle to accurately forecast hurricanes, even with recent advances in computational power,” said Khondaker. “Our research focused on the idea that current weather models overestimate how much hurricanes’ kinetic energy spreads out or diffuses.”
They hypothesized that over-estimations of diffusivity lead to weaker intensity forecasts, which creates a discrepancy between forecast and reality.
“Our research was motivated by the significant impact of hurricanes: the most destructive and costly natural disasters in U.S. history. We observed that current models often underestimate hurricane intensity, leading to critical gaps in disaster preparedness. Our goal was to improve existing forecasting methods to reduce the devastating effects of hurricanes by enhancing both intensity and rainfall predictions,” he added.
Khondaker and his supervisor, assistant professor of civil and environmental engineering Mostafa Momen, Ph.D., saw the paper published in the August 2024 issue of the Journal of Hydrometeorology. It was also highlighted by phys.org and the Pittsburgh Supercomputing Center (PSC), whose flagship supercomputer, Bridges-2, played a pivotal role in the project.
“We needed significant computational resources to run a coupled atmospheric and hydrological model. For example, simulating 17 days of Hurricane Irma at 8 km horizontal resolution took 22 hours using 128 processors. Although we have access to dedicated computing nodes at the University of Houston, such as Carya and Sabine, the scale and complexity of our research required additional resources. This led us to rely on the Bridges-2 system at PSC as well as the National Center for Atmospheric Research supercomputers.
“The computational power provided by Bridges-2 was essential for running our simulations, and its high-memory nodes were critical in handling our large datasets, enabling us to perform detailed analyses,” Khondaker explained.
Their time on Bridges-2 was allocated by the National Science Foundation’s (NSF) Advanced Cyberinfrastructure Coordination Ecosystem: Services and Support (ACCESS) program. ACCESS helps researchers and educators utilize the United States’ collective advanced computing systems and services at no cost.
“One of the most important discoveries from our research was the significant impact that adjusting turbulence parameters had on improving hurricane intensity forecasts, with enhancements of up to 40% compared to the default weather models. This improvement will be crucial in preparing for damages associated with high winds and storm surges,” he added. “An interesting finding was that in the more intense hurricanes, the total amount of rainfall does not increase; instead, precipitation becomes more localized and intense.”
This phenomenon was observed in 2017 with Hurricane Harvey, highlighting the vulnerability of urban areas with more impervious surfaces, such as concrete, to severe flooding as a result of concentrated rainfall. Khondaker and Momen’s adjusted model significantly improved forecast accuracy - hurricane intensity forecasts by up to 40%, and flood predictions by up to 34% - and the resulting improved rainfall predictions may be able to mitigate some of the danger and risk to life for individuals living in vulnerable areas of hurricane-prone regions.
“These improvements are crucial, as better forecasts can lead to more effective evacuation and reduce the overall damage caused by these extreme weather events,” he said.
“The future of our research will focus on two key objectives. First, we aim to deepen our understanding and improve the prediction of compound precipitation and storm surge floods, a major contributor to the damage caused by hurricanes along coastal areas. Second, we plan to develop a general turbulence diffusion adjustment for hurricane simulations by considering the rotational dynamics of these storms and incorporating more advanced physics-based models. These efforts are designed to enhance the accuracy of hurricane predictions and strengthen preparedness, ultimately reducing the risks and impacts on coastal communities,” Khondaker concluded.