TECHNOLOGY: Gigabites Give Power Boost To Waste Cleanup

Bill Siuru

It's not a bird: it's not a plane. It's Supercomputer! The latest technology to surface for cleaning up underground waste sites now involves sophisticated systems that can predict how far, how fast and in which direction liquid waste materials migrate in different subterranean environments.

However, three-dimensional modeling of underground waste migration is one the most complex problems ever tackled by computer scientists.

The supercomputers' potential to take on this problem recently was demonstrated by researchers David Alumbaugh and Greg Newman at the Sandia National Laboratories in Albuquerque, N.M. Using a 1,840-node Intel Paragon, one of the world's fastest-operating computers, they modeled the migration of a simulated underground waste plume. As an indication of the problem's difficulty, it took two years for the scientists to set up the problem, but only four hours, 17 minutes and 46 seconds for the supercomputer to obtain the answer.

A technique called "multiple instruction multiple data" (MIMD) was used to answer the oozy problem at lightning speed. MIMD is a form of massively parallel computing where large equations are broken into thousands of smaller ones and are solved simultaneously using hundreds of processors. In contrast, the typical desktop workstation solves problems one at a time. In this demonstration, the underground volume was divided into some 80,000 cells, each being 2-meter cubes.

While three-dimensional images of underground environments have been constructed in the past, these use a series of two-dimensional pictures stacked on top of one another. With the supercomputer's capabilities, enormous amounts of data with millions of unknown variables can be directly converted into a more precise three-dimensional model of the underground environment.

Experimental data was used to test the model's ac-curacy. For example, at the Univer-sity of California's (Berkeley) Richmond Field Station, five 60-meter wells were drilled on the corners and in the center a 45-meter square. To simulate a waste plume, 50,000 gallons of salt water was pumped through the center well into a gravel aquifer 30-meters below the surface.

The plume's migration distance and direction were tracked at different depths by measuring differences in the magnetic field's strength. These differences were created by the varying electrical conductivity between the salt water plume and the surrounding aquifer. The magnetic field itself was produced by a dipole source emitting a 18 kilohertz sinusoidal electromagnetic wave which was lowered into the center well. The computer took measurements before and after the salt water's injection.

Three pairs of three-dimensional color images resulted from the computer's simulation. One showed the aquifer's electrical conductivity before the salt water injection, another showed it after the injection and the third pair highlighted the differences in conductivity between the previous two.

The model revealed errors in the plume's direction found using traditional hydrology techniques to analyze experimental data. The demonstration proved that the supercomputer's three-dimensional modeling technique can help remediators clean up underground contamination, as well as assist hydrologists to explore aquifers, mining engineers to map mineral deposit boundaries and petroleum engineers to site wells for maxim oil extraction.