High-Performance Computational Methods for Coupled Fields and GAFD Turbulence

Grand Challenge Applications Group (GCAG)

University Colorado at Boulder (UCB) and Denver (UCD), National Center for Atmospheric Research (NCAR), University of Minnesota (UM)

GAFD Turbulence Overview

We are studying highly turbulent flows, typically constrained by rotation and stratification, which play crucial roles in determining the large-scale dynamics within oceans, atmospheres and stars. Remarkably coherent structures and circulations are a part of such turbulence. Understanding how they arise, and their effects, is a major challenge shared by all branches of geophysical and astrophysical fluid dynamics (GAFD). High-performance computing offers the opportunity to investigate such an important class of turbulent flows at a fundamental level.

The GAFD Turbulence project is motivated by four general grand challenge applications: geostrophic turbulence, oceanic convection, deep convection in planetary atmospheres, and stellar compressible convection, all influenced by effects of rotation and stratification. With such studies should come breakthroughs in assessing the effects of such turbulence variously on ocean circulations, on fast zonal jets in the giant planets, and on differential rotation and dynamo action in stars like our sun. The emphasis is to exploit new massively parallel architectures to increase the spatial resolution in three-dimensional simulations employing variously pseudo-spectral, finite-difference, multi-grid and piecewise-parabolic-method (PPM) approaches in studying the intense turbulence encountered in planetary and stellar settings, including the large-scale coherent structures and mean flows that can coexist with such turbulence.

The scale of these simulations requires corresponding progress in the computational sciences, both in order to develop and optimize software for massively-parallel computers and to capture and visualize the resulting massive data sets. Computer science challenges paced by these problems include performance modelling and analysis, performance evaluation tools, program transformation and automatic data distribution, heterogeneous computation, very large scientific databases and high-performance visualization. Because the computational science developments are occurring in the context of realistic applications, they are widely applicable to other large-scale simulations on parallel computers.

This GAFD Turbulence study involves an interdisciplinary team of researchers combining strengths in the necessary physical and computational disciplines from the University of Colorado, the National Center for Atmospheric Research and the University of Minnesota. We are also joined by a team of engineering scientists working on Coupled Fields. The joint team is utilizing a wide range of high performance computational resources, including a range of shared and distributed memory parallel machines, some in the largest available configurations. Results of the research are being shared with the community through a series of yearly workshops and by providing access to the multi-terabyte 4--D turbulence data sets.

Current Projects

The following topics are currently being studied in detail, using a variety of different codes and various computing platforms:

This page prepared by Nic Brummell and Juri Toomre, Laboratory for Computational Dynamics, University of Colorado.