The Gunzburger Lab was awarded the best poster prize at the Accelerated Climate Modeling for Energy meeting. The lab’s research poster is entitled Exponential time differencing for large time stepping and localized approach for parallel implementation.
The research shows the promise in making climate simulations more accurate by increasing the spatial resolution at which they calculate meteorological phenomena. Climate model simulations which look millennia into the future are complex and costly, requiring large amounts of data and many cycles of intricate calculations. “We are trying to fix this problem by implementing mathematical methods that can untie this relation between the time step and simulation resolution, allowing us to take less snap-shots (data points) and get an answer from the simulation more quickly. In particular, we are implementing Exponential Time-Differencing methods to mitigate the CFL condition and create faster simulations,” said Chad Sockwell, SC doctoral candidate and Gunzburger Lab researcher.
In addition to decoupling the relation between time step and simulation resolution, the Gunzburger Lab is also studying a method that splits the simulation domain (spatial area of interest) into smaller domains, each with similar resolutions. “This allows us to deal with each small domain separately, each with its own CFL condition. Then resources can be appropriately assigned so the areas requiring more data points (snap-shots) can be simulated in time comparable to the areas that require less snap-shots. This is known as local time stepping.”