High-Performance Computing to Evaluate Hierarchical Heterogeneity Paradigms in Sedimentary Aquifer Systems

点击翻译按钮获取中文摘要

The scaling of processes governing ground-water flow and mass transport in the near subsurface is being linked to the hierarchical sedimentary architecture found within geo-reservoirs. The PIs have developed a code that generates three-dimensional models that represent the heterogeneity of hydrogeologic attributes arising from hierarchical sedimentary architecture in braid-belt deposits. The sedimentary architecture represented ranges in scale from centimeters to kilometers. This architecture does not represent all deposits, aquifers, or reservoirs, but it does represent important ones, including the Fortymile Wash alluvium at Yucca Mountain, glaciofuvial aquifers in the northern United States, the Ringold Formation at the Hanford site, and the Ivishak Formation, an Alaskan north-slope petroleum reservoir. The model is not meant to represent any one site exactly, but to represent important aspects of heterogeneity common to many sites. The full model represents a domain 2.5 km by 2.5 km by 10 m thick with voxels of 1 cubic centimeter on a 62.5 trillion-voxel lattice. While parts of the domain have been created on desktop workstations, we are currently collaborating with scientists at the Pacific Northwest National Laboratory (PNNL) to generate the full 62.5 trillion node domain using the high performance supercomputer located at the Environmental Molecular Sciences Laboratory (EMSL) at PNNL. The full model domain will be freely distributed to the research community as a resource for testing ideas related to the upscaling problem, the inverse problem, and for other computational research requiring high-resolution grids. In our work, we are generating simulations to test the hypothesis that the percolation of highly permeable open-framework gravels must be understood not only from their proportions and geometry, but also from the proportions, geometry, and spatial distribution of the strata at multiple scales (unit bars, compound bars, channel fills, etc.). Furthermore, we are generating numerical transport simulations to test the hypothesis that solutions to upscaling problems may require accounting for the presence of percolation, as well as the hierarchy of unit types defined at different scales.If successful, the full model domain and numerical tracer tests will be freely distributed to the research community. Thus, the research will promote the testing of emerging ideas in computational research on ground water flow and transport. The hypotheses we will test using these models are relevant to heterogeneity and upscaling issues in environmental cleanup and to enhanced petroleum recovery. The project also benefits human resources and education in science and engineering through the training of a post-doctoral scholar in the areas of high-performance computing, hydrogeology, and sedimentology, and through mentoring from university and national laboratory scientists.