|
|
|
| News |
| Advanced Scientific Computing Research |
| A New DOE ASCR SciDAC Initiative: ISICLES |
| The DOE Office of Advanced Scientific Computing Research (ASCR) recently funded six projects in response to the national and international need for better inclusion of dynamic ice sheet modeling in Earth System and Climate models. Collectively grouped as ISICLES (Ice Sheet Initiative for CLimate ExtremeS), this ASCR SciDAC Initiative is intended to accelerate computational science research for state-of-the-science algorithms and codes for ice sheet models that are scalable at extreme scales of computing. |
 |
| Stock Photo |
| Figure 4. The Ice Sheet Initiative for Climate ExtremeS (ISICLES) is a set of ASCR/SciDAC projects intended to accelerate computational science research for state-of-the-science algorithms and codes for ice sheet models that are scalable at extreme scales of computing. |
|
| A kickoff workshop for the ice sheet projects was held September 16–17, 2009, in Annapolis, Maryland. The goal was to coordinate the projects into a collective initiative to maximize their success and impact. The workshop opened with “View from the Office of the Under-Secretary for Science, Department of Energy” presented by Dr. David Dean, Senior Advisor, Office of the Under-Secretary for Science, DOE. This was followed by “Perspectives from the Office of Advanced Scientific Computing Research (OASCR), DOE” presented by Dr. Walt Polansky, Division Director for Research, OASCR. |
| Organized by the DOE ASCR Program Manager, Dr. Lali Chatterjee, the workshop included presentations from Dr. Robert Bindschadler, Chief Scientist, NASA Goddard Space Flight Center, and Dr. Tony Payne, University of Bristol, United Kingdom. These addressed the critical importance of computational science advances on ice sheet dynamics and their inclusion in climate models. Dr. David Keyes, University of Columbia, talked about key mathematical techniques and software issues, and Dr. Jim Hack, Director, National Center for Computational Sciences at Oak Ridge National Laboratory (ORNL), discussed the regional impacts of ice sheet melting. Dr. W. Lipscomb, Los Alamos National Laboratory (LANL) presented on the existing models, and each project highlighted its objectives. DOE Program Managers – Dr. Chatterjee (ASCR) and Dr. A. Bamzai (BER) – led the discussions on future projections, milestones, and expectations. |
This ISICLES initiative also represents an expansion of ASCR’s existing Computational Partnerships with the Office of Biological and Environmental Research (OBER), for DOE SC Climate Modeling efforts as computationally strong and robust ice sheet and climate codes are naturally, and critically, dependent on the correct science input. The projects are led by four national labs (Argonne, Lawrence Berkeley, Oak Ridge, and Pacific Northwest) and two universities (Columbia and the University of Texas); LANL is collaborating with five projects, and Sandia National Laboratories with two. Planned investigations include:
- Working within SEA-CISM, a project led by ORNL (Kate Evans, principal investigator) will implement the Trilinos framework around a hierarchical blocking structure for increased parallelism, using implicit solver capabilities and physics-based preconditioning to maximize scalability
- B-ISICLES, an effort led at Berkeley Lab (Esmond Ng, principal investigator) will leverage expertise in block-structured adaptive mesh refinement to develop a parallel and scalable Community Ice Sheet Model (CISM) in the Chombo solver framework
- SISIPHUS (Scalable Ice Sheet Solvers and Infrastructure for Petascale) is a project led at Argonne (Tim Tautges, principal investigator) will develop techniques for solving the full three-dimensional Stokes problem using hp-adaptive finite element methods, with the PETSc solvers as a base and careful consideration of the interfaces to allow forward and adjoint solutions
- A group based at the University of Texas (Omar Ghattas, principal investigator) is developing a full-Stokes flow model with adaptive grid refinement, high-order discretizations, forward and inverse problem solving capability, and scalable uncertainty quantification techniques
- Using extended finite elements to capture discontinuities with “self-correcting” and algebraic multigrid methods, a project led by Columbia University (Haim Waisman, principal investigator) will develop scalable models of the complex fracture of ice within larger ice sheet models
- A project led by Pacific Northwest Laboratory (Alexander Tartakovsky, principal investigator) will develop a three-dimensional Lagrangian particle ice sheet model based on the fundamental conservation equations, using algorithms based on smooth particle hydrodynamics to achieve scalability
|
Since inclusion of “fully dynamic ice sheet models and ocean/ice shelf interactions,” prioritized by the “Joint ASCAC-BERAC Report” in March 2008, are still in their early stages of development, ISICLES has the potential to make a notable difference. Incorporation of advanced techniques of high-performance and extreme scale computing in the early stages of code creation will better ensure their effective use, rather than introduction of these techniques into established codes. ISICLES is not only expected to have a strong impact on DOE’s climate models and codes, but also planned to extend to other national and international models.
|
