Texas Researchers Gathers for TACCSTER 2024

The Texas Advanced Computing Center (TACC) welcomed more than 100 attendees for the 7th annual Symposium for Texas Researchers, TACCSTER, September 23 - 25, held at the The University of Texas' J. J. Pickle Research Campus in Austin. TACCSTER brings together researchers from across Texas who use TACC's high-performance computing (HPC) resources, offering them a platform to discuss improvements, share insights, and highlight the wide-ranging research powered by their supercomputers.

Following a welcome from Dan Stanzione, executive director of TACC, featured talks from UT faculty Clint Dawson, Stella Offner and Tan Bui-Thanh, all from the Oden Institute for Computational Engineering and Sciences, got underway.

In his keynote address, How HPC Revolutionized Storm Surge Modeling, Dawson delved into his research on hurricane storm surge forecasting. He provided an overview of his group’s work, tracing its start from the early 1990s to present-day goals. “We’re always seeking algorithmic improvements, with the ultimate aim of achieving high-resolution or hybrid compound flood risk analysis,” said Dawson, department chair of aerospace engineering & engineering mechanics at Cockrell School of Engineering and lead of Computational Hydraulics Group at the Oden Institute.

Dawson also stressed TACC's crucial role in advancing his research. “When a hurricane is approaching, we need to run multiple high-resolution ensemble models. TACC has consistently provided the computational power necessary when resources are available.”

As Dawson gave his keynote, he stated that storm surge modeling was being run on TACC’s computers in real time as Hurricane Helene was quickly approaching Florida’s gulf coast. “We want a five-day hurricane simulation to finish in 15 minutes, because that gives us time to process the results and post them before the next forecast," said Dawson. The storm became a Category 4 hurricane that made landfall on September 26, breaking records for storm surge in Florida. Dawson and his team used TACC’s Frontera supercomputer to model Hurricane Helene using data released by the National Hurricane Center every six hours. 

Earlier this year, Dawson received the UT Austin President’s Research Impact Award for his research in data-driven storm-surge modeling, which has had a major impact on the response to natural disasters in Texas.

Currently, he is working on a global model—including 600,000 historical and synthetic hurricanes—to train artificial intelligence (AI) models and to understand the storm surging packs in different parts of the world.

“My group is heavily involved in machine learning to see what we can get out of these algorithms. Our goal is to develop a high resolution or hybrid compound flood risk analysis system,” Dawson concluded.

Stella Offner introduced herself as a “long-time TACC user,” who was using TACC systems even prior to coming to UT, and said it TACC's resources were a big reason she came to UT. Her presentation, Expanding the Frontera of Astrophysics: From Galaxies to Individual Stars, featured beautiful simulations of star system formation combined with dramatic music effects, drawing in the audience. Offner is assistant professor of astronomy and co-director of Center for Scientific Machine Learning at the Oden Institute.

Offner launched into her presentation, saying star formation is an extremely important process in astrophysics. It powers galaxy evolution, is responsible for the distribution of heavy elements in the universe, dictates planetary systems, and supports life on Earth. "It involves dynamic ranges and processes that take more than 10 million years and span 24 orders of magnitude in density and more than 11 orders of magnitude in spatial scale. These simulations are way beyond the conditions that we can achieve in labs—this is why we need high-performance computing.”

Offner is the newly minted director of the NSF-Simons AI Institute for Cosmic Origins funded by the National Science Foundation and the Simons Foundation to develop AI technologies for astronomical research and accelerate the pace of scientific discoveries. 

Bui’s talk highlighted the recent boom of digital twin technologies for real-time forecasting. Bui’s research group, the Probabilistic and High Order Inference, Computation, Estimation, and Simulation, focuses on using machine learning to estimate error at prediction steps, which is used to determine when the model requires retraining.

TACCSTER also featured poster presentations and lightning talks, showcasing the wide range of research conducted using TACC’s supercomputers. Several students and postdoctoral students from the Oden Institute were selected to present their work. 

Liting Huang and Ira Narang, members of Dawson’s group, presented their respective research on the impact of hurricanes and floods. Cole Nockolds, a CSEM student from Bui’s group, talked about his work on using linear parameterized latent space to simulate stiff systems. Joar Bagge, an O'Donnell Postdoc Fellow, discussed his research on GPU-accelerated simulations of deformable capsules flowing through pipes. Both Huang and Marissa Llamas, an undergraduate student who participated in a research program at TACC in Summer 2024, and who works with Dawson's group, received awards for Best Poster. For a full list of presentations, please visit TACCSTER. 

The event also included facility tours and a town hall Q&A where TACC leadership answered audience questions about high-performance computing. “This is an opportunity for TACC users and staff to interact and to learn about what we are doing right and what we need to improve on,” said Stanzione.