October 2010

Program aims to get ice right

Ice sheets are shrinking in Greenland and Antarctica, noted the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) in 2007.

Melting land ice could have a huge impact, raising sea levels enough to flood coastal cities, the report said. Even a modest amount of melting could change the oceans enough to affect climate.

But the panel, says Lawrence Berkeley National Laboratory’s (LBNL) Esmond Ng, also “acknowledged that they did not have the right computing capability or the modeling effort to understand ice sheet dynamics.”

The Department of Energy Office of Advanced Scientific Computing Research (ASCR) responded in 2009 with ISICLES – Ice Sheet Initiative for CLimate ExtremeS.

“The problem of getting an accurate, high-resolution modeling of land ice has not really been looked at in a grand-scale way,” says Ng, who heads one of the projects from LBNL’s Computational Research Division.

“That’s the motivation for ISICLES. The funding supports applied mathematicians and computer scientists so they can work with climate scientists to further advance modeling capability.”

Many of the six initiatives are designed to capitalize on software tools and algorithms developed in other ASCR-backed programs, such as Applied Mathematics and Scientific Discovery through Advanced Computation (SciDAC).

“Some projects,” says Katherine Evans, who heads an ISICLES effort based at Oak Ridge National Laboratory (ORNL), “are designed to start from scratch and develop something completely new. Others are taking an existing capability and extending it in several different ways.”

ASCR, she says, “wanted to leverage the money it was spending on ISICLES to have some quick rewards, some longer-term rewards, some more high-risk and some more low-risk projects – diversifying their portfolio.”

Some of the projects are designed to provide improved data for the next IPCC report, due in 2013-14, whereas others will take longer to bear fruit.

The ISICLES projects:

  • LBNL’s Ng heads Berkeley-ISICLES (BISICLES), an effort to develop high-performance adaptive algorithms for ice sheet modeling. BISICLES will employ and extend existing algorithmic tools, especially adaptive mesh refinement (AMR), for high-resolution ice sheet modeling at relatively low computational cost. AMR casts fine meshes of data points in areas of interest and coarser meshes in other areas. Computers calculate changes in physical properties at each point.
  • Pacific Northwest National Laboratory applied mathematician Alexandre Tartakovsky directs a study on Lagrangian smoothed particle hydrodynamics for ice sheet dynamics. He and colleagues will develop an ice sheet model based on computationally tracking the movement of representative particles rather than calculating properties in a data mesh.
  • Haim Waisman, a Columbia University engineering professor, and his team will model the fracture of ice sheets on parallel computers. Parallel algorithms can efficiently track and predict fractures and iceberg formation. The project also will model how lakes atop glaciers quickly drain through ice cracks.
  • SEACISM (Scalable, Efficient and Accurate Community Ice Sheet Model), headed by ORNL’s Evans, will build on an extended version of Glimmer, an ice sheet model developed at the University of Bristol in the United Kingdom. The researchers will apply algorithmic tools to improve the model’s ability to run efficiently on ever-larger parallel computers.
  • SISIPHUS (Scalable Ice Sheet Solvers and Infrastructure for Petascale, High-resolution, Unstructured Simulation), led by Timothy Tautges of Argonne National Laboratory, will develop algorithms capable of efficiently solving the three-dimensional fluid dynamics equations at the heart of most ice sheet models. The project will tap solvers and services developed in other DOE programs.
  • Omar Ghattas of the University of Texas will lead research in uncertainty quantification for large-scale ice sheet modeling and simulation. This project will devise techniques to assimilate “noisy” data from observations into fluid dynamics models and to infer the uncertain parameters that drive ice sheet behavior.

If all goes well, Evans says, the six efforts will improve the overall picture of human activity’s impact on Earth’s climate. Some systems, like oceans, are so large that effects can take hundreds of years to appear. But “ice sheets are small in comparison and they seem quite sensitive to changes in Earth’s temperature. That’s why we think it’s important to get it right.”