Reliable methods for imaging the subsurface are crucial in solving geotechnical problems. Imaging reduces the economic and environmental risk of drilling for natural resources (oil and gas, and geothermal energy) and is essential for site characterization and verification of environmental remediation practices. Geophysical imaging provides information from the subsurface, where borehole sampling is often restricted and where direct information cannot be obtained otherwise. Imaging produces a map of the geophysical attributes (electrical conductivity, density, and seismic velocity and/or acoustic impedance). These attributes are derived from geophysical measurements made on or above the Earth’s surface, on or above the ocean seafloor, and in isolated boreholes.
Both deterministic and stochastic imaging techniques have been successfully applied to a wide variety of geotechnical problems. We have expertise in Monte-Carlo Markov-Chain stochastic imaging methods and large-scale 3D imaging practices using gradient and Newton-based optimization schemes. Our imaging algorithms also exploit large parallel computing systems—utilizing tens of thousands of processors (necessary for realistic 3D imaging)—and effectively deal with imaging and data volumes of industrial size.
Subsurface imaging research is now focused on combining multiple types of geophysical data sets to better quantify the subsurface and reduce ambiguity. The degree to which joint images of geophysical attributes can be used successfully to infer rock properties (fracture orientation, fracture density, temperature and fluid saturations) from geophysical attributes is also an active area of research. A key component of such imaging is forward modeling of the subsurface. The focus of this research area is to continue the development of efficient 3D numerical codes for modeling geophysical attributes, mainly seismic and electromagnetic rock properties through seismic wave propagation and electromagnetic wave propagation and diffusion, respectively.
This strong Geophysics imaging capability is necessarily complemented by a robust computational capability. All of the applied sciences have specific computational challenges. The challenge for geophysical computation is to develop accurate and efficient computer codes capable of modeling the seismic and electromagnetic response in complex geologic media, including formation anisotropy or multiscale heterogeneities (in the form of fractures, faults, and/or patchy saturation and rough topography). The Geophysics Department has experience in a variety of computational approaches, including boundary integral equation, global matrix, finite difference, spectral element, discrete element, and asymptotic ray methods. In order to model realistic earth models, we have developed simulation codes for a variety of high-performance computing systems, including large parallel distributed- and shared-memory computing servers and many-core architectures (GPUs).
Our codes also serve as the computational engines for the next generation of modeling-based deterministic and stochastic inversion algorithms. This research is performed using the supercomputers at the National Energy Research Scientific—Computing Center (NERSC) at Berkeley Lab, and the high-performance computing servers maintained by Berkeley Lab’s Information Technology Division on behalf of the Geophysics Department.