New model harnesses the power of supercomputers for more accurate flood simulation
Researchers at ORNL used TRITON to simulate flooding in Houston, Texas, and surrounding areas caused by Hurricane Harvey in 2017. Light purples represent shallower water, and dark purples represent deeper water. Credit: Sudershan Gangrade/ORNL
A team of researchers from the Department of Energy’s Oak Ridge National Laboratory and Tennessee Technological University has created a 2D open-source flood inundation model designed for multi-architecture computing systems. The Two-dimensional Runoff Inundation Toolkit for Operational Needs, or TRITON, can use multiple graphics processing units, or GPUs, to model floods more quickly and accurately than existing tools.
Flood modeling is an important part of emergency preparedness and response. However, models must be fast and accurate—returning simulation results in minutes—to be a useful tool for decision making and planning. The higher the resolution of the model, the more computing power is required to run it, so organizations can use simpler models that sacrifice accuracy for speed. The computing power of the GPU allows computations with high-resolution models to run faster than simple models that use only the CPU.
As high-performance computing has developed into an indispensable tool for science, it has also become a requirement for modern flood models to harness the power of hybrid CPU + GPU architectures. TRITON, whose development was funded by the Air Force’s Numerical Weather Modeling Program, is specifically optimized for the multiarchitecture design of supercomputers such as the IBM AC922 Summit at the Oak Ridge Leadership Computing Facility.
“The unique thing about the TRITON isn’t just that it uses a GPU—it’s not the only flood model that the GPU has access to. But it is adapted to use multiple GPUs simultaneously, which makes it suitable for solving the flood problem at Summit,” said Shih-Chieh Kao, an ORNL group leader who led the project.
The team put the model through its paces at Summit to demonstrate its consistency, stability, and some of its unique capabilities, such as runoff hydrographs. This optional data allows TRITON to simulate pluvial flooding—that is, local flash flooding—in addition to river flooding. During river flooding, streams or rivers swell and inundate the floodplain. Using datasets from the Federal Emergency Management Agency’s 100-year flood zone as a benchmark, simulations using runoff hydrographs are more accurate than basic hydraulic models alone.
“To truly understand the impact of flooding, we need to understand inundation, which includes how deep a river is and explains various flood events: river flooding and flash flooding. Conventional flood models usually only address river flooding. TRITON can address both and provide more information about the impact of flooding,” said Kao. “If you have this inundation information, you can overlay it on assets and evaluate which ones are at risk and which are not.”
In another test case, the team simulated 2017 flooding in the Houston metropolitan area caused by Hurricane Harvey. The simulation lasted 10 days and was modeled on two different hardware configurations: one using multiple CPUs and the other using multiple GPUs. The results clearly demonstrate the advantages of flooded models designed to run on multi-GPU configurations. Even the smallest hardware configurations—one compute node with six GPUs—complete simulations faster than the most powerful 64 node multi-CPU configurations.
As an open-source toolkit, TRITON is available for free and can be used on a variety of computing platforms—from laptops and desktops to supercomputers. Research team members are constantly developing new features and working on algorithms to upgrade current capabilities to an operational level.
“TRITON will be the foundation on which we continue to build, and we call it a device for a reason. We keep building to make it more useful—that’s our vision. As computing power increases, and prices fall, eventually everyone should have more access. to use this ability to better simulate flooding,” said Kao.
Study explores uncertainty in flood risk estimates
M. Morales-Hernández et al, TRITON: An open source 2D multi-GPU hydrodynamic flood model, Environmental Modeling & Software (2021). DOI: 10.1016/j.envsoft.2021.105034
Provided by Oak Ridge National Laboratory
Quote: New model harnessing supercomputer power for more accurate flood simulation (2022, 28 July) retrieved 28 July 2022 from https://phys.org/news/2022-07-harnesses-supercomputing-power-accurate-simulations.html
This document is subject to copyright. Apart from any fair dealing for purposes of personal study or research, no part may be reproduced without written permission. Content is provided for informational purposes only.
#model #harnesses #power #supercomputers #accurate #flood #simulation
Comments
Post a Comment