Donna Calhoun (Boise State University)

Computational models used for tracking tsunamis and hurricanes, modeling blood flow in the human heart, or optimizing the shape of aircraft wings place huge, if not prohibitive, demands on computing resources. However, in many scenarios, the physical processes to be modeled are very localized. When a tsunami wave propagates across an ocean, for example, large regions of the ocean are either completely undisturbed, or quickly return to an undisturbed state once the propagating tsunami wave passes. An optimal strategy for taking advantage of these localized features is to dynamically apply computational resources only where they are needed, e.g. at the propagating tsunami wave front. One such strategy, known as "adaptive mesh refinement" (AMR) is widely recognized as an effective algorithmic and software strategy for modeling large classes of problems in scientific and engineering applications.

The ForestClaw library (D. Calhoun, C. Burstedde) is a general purpose software library that implements an AMR strategy for solving a variety of applications in environmental and engineering domains. In this talk, I will provide an overview of our implementation of AMR in ForestClaw, describe briefly the underlying solvers especially well-suited to AMR, and demonstrate some of the applications that use ForestClaw, including a DARPA (Defense Advanced Research Projects Agency) funded project in which we couple distinct ForestClaw codes to solve a complex multi-physics problems coupling the lower earth atmosphere with the ionosphere.