Crossing the Scales of Subsurface Science via Parallel Computing

Mark White, Pacific Northwest National Laboratory

Through investigations into the migration of contaminants into drinking-water aquifers, the enhancement of fossil energy recovery, and the sequestration of carbon dioxide for climate change, subsurface science research impacts us locally, nationally, and globally. A key component of these investigations is the understanding of complex fluid flow and contaminant transport processes that occur in the heterogeneous environment beneath the earth’s surface. In a quest to bring modern computational technologies to bear on subsurface science problems, scientists have discovered a new tool for advancing their understanding — parallel computing. Inherent in subsurface science problems are the issues of process complexity, material variability, and property uncertainty. In subsurface science the key roles of numerical simulation in meeting the scientific challenges of these issues are hypothesis testing, high-resolution depictions, long-term predictions and engineering design. The promise of parallel computing for subsurface science was increased grid resolution and dimension of modeling domains. This promise is now being realized as computer codes for subsurface science that have been traditionally sequential implementations are being rewritten or converted into scalable algorithms for use on parallel computers. What scientists are discovering in applying these codes, however, is that parallel computing is having a much greater impact on subsurface science than anticipated: it’s changing how research is conducted, providing new insights to physical processes, bridging knowledge gaps, and dissolving scale boundaries. The manner in which parallel computing has altered subsurface science applications and research at the Pacific Northwest National Laboratory will be discussed by showing examples that cross the scales of subsurface science from pore-scale interactions through field-scale applications. The challenge for future scientists in solving the nation’s subsurface problems will be to incorporate the latest process science into robust, scalable computer codes that yield reliable predictions, allowing us to make the “right” environmental stewardship decisions for future generations.

Abstract Author(s): Mark White