Just over 10 years ago, the United States experienced one of the most damaging hurricanes in U.S. history, Hurricane Katrina. According to the National Oceanic and Atmospheric Administration (NOAA), Katrina claimed 1,833 lives, cost over $100 billion in damages and, as stated in a U.S. Census Bureau report, displaced over 400,000 people living in and around New Orleans and the Mississippi Golf Coast. More recently, Hurricane Sandy tore along the east coast of the U.S. in 2012, causing an estimated 147 deaths and over $50 billion in damages, as mentioned in a 2013 NOAA Service Assessment.
Hurricanes and coastal floods have left life-altering wreckage in their wake long before Katrina and Sandy, and they will continue to pound into the coasts long after. These natural disasters cannot be thwarted, but the work of one research group at the University of Notre Dame is making a difference in being able to better prepare for these situations by modeling the physics of the coastal ocean to forecast coastal storms. The Computational Hydraulics Laboratory (CHL) devotes their time to developing high performance codes and coastal circulation models to advance hurricane hazard modeling to help protect people, infrastructures and plan for future construction. Developing next-generation codes that are more flexible and efficient is a continual effort of the CHL. Improving the accuracy of codes and minimizing computational costs helps in the effort to make these technologies more widely available.
The CHL is led by Joannes Westerink, the Joseph and Nona Ahearn Professor of Computational Engineering and Science and the Henry J. Massman Chairman of Civil & Environmental Engineering & Earth Sciences. Developing models of the coastal ocean requires many computational hours. Westerink and the CHL team recently celebrated the acquisition of their new computer cluster. Housed at Union Station Technology Center downtown South Bend and maintained by the Center for Research Computing, the Lenovo Intel Xeon E5-2680 cluster adds a total of 1,512 cores on the front end, increasing CHL cores to 3,000. The typical run time of a hurricane event simulating 20 days real time on 2,000 cores is approximately four hours, assuming five simulation days per hour. With the chip speed improvement of the newly acquired cores, the CHL is quadrupling their capacity.
“Exponentially expanding our computer horsepower allows us to delve deeper and deeper into the physics of the coastal ocean,” says Westerink.
The hurricane modeling process begins with the development of codes to simulate the equations that describe the physics of the coastal ocean environment. The code most widely used by government, universities, and for consulting, is ADCIRC+SWAN. This model has been used for the FEMA Great Lakes Flood Insurance Study, the East Coast Flood Insurance Study, the Gulf Coast Flood Insurance Study, and by the Nuclear Regulatory Commission. ADCIRC, a coastal ocean circulation code, is the leading modeling technology in evaluating coastal flood risk and was co-developed by the CHL with the University of North Carolina at Chapel Hill and the University of Texas at Austin.
Once the code is developed, a grid is created to figure out the geometry and hydrodynamic flow characteristics of the region being studied. Once the code and grid have been generated, they are combined with metadata in a high performance, massively parallel computing environment. Results are compared to multiple data sets from previous events to validate whether the process was captured correctly. Currently, the CHL is working on projects for NSF to advance code development, and is partnering with AON, FMGlobal, Office of Naval Research, NOAA, Arcadis, Baker and SURA on projects to model the coastal ocean and shoreline environments of the New York Harbor, Western Alaska, Puerto Rico and the western North Pacific. These models will help measure the risk and vulnerabilities of flooding and storm surges that could cause major damage to the ecosystems of these coasts.