University of Texas at Austin

Research

Improving the Efficiency of Wave and Surge Models via Adaptive Mesh Resolution


PIs: Clint Dawson and Casey Dietrich, North Carolina State University
Sponsor: Department of Homeland Security Coastal Resilience Center of Excellence
University of North Carolina, Chapel Hill


Coastal communities rely on predictions of flooding caused by storms, but these predictions can take hours on even the fastest supercomputers. In our project, we are speeding up the models to predict hurricane waves and coastal flooding, and we are improving the forecast guidance for use by decision makers. In Year 10, we will support the widespread adoption of these technologies and transfer them to end users.

Our CRC project has enabled flooding predictions to be faster.
There are obvious benefits to speeding up the predictions of coastal flooding due to hurricanes. The largest benefit occurs during storms – if the predictions can be completed faster, then they can be shared faster with emergency managers and other decision-makers, who can then have more time to include them in their real-time response. There are also benefits between storms – long-term adaptation studies can combine flooding predictions from hundreds of possible future storms, and if these predictions can be completed faster, then they can lead to adaptations that are faster (by finishing early) and/or more accurate (by considering more future storms). In our ongoing CRC project, we speeded up the surge and wave models used for predictions of coastal flooding. We developed technologies to improve the efficiency of:
– ADCIRC (ADvanced CIRCulation) is used widely to predict storm surge and coastal flooding. By using a dynamic load balancing, a coarse-grain mesh adaptivity, and/or subgrid corrections, the model speed can be decreased by 50 percent or more.
– SWAN (Simulating WAves Nearshore) is used widely to predict hurricane waves. By quantifying the use of previous-generation physics and an alternative model formulation; the model speed can be decreased by 30 percent or more.
– Kalpana is used widely to visualize the flooding predictions. By downscaling the flood maps to the small scales of critical infrastructure (e.g. homes, roads), we can tailor guidance to the needs of end users, without having to run ADCIRC at those scales.


Future research on this project includes:

  1. Support for widespread adoption of subgrid corrections and downscaling.

  2. Expand the probabilistic predictions of downscaled flood maps.

  3. Improve mass-conservation in ADCIRC with DG/CG formulation.
  4. Improve efficiency of DG/CG formulation.

Back