Resource Exploration
Turning reflected sound into a map of what lies a kilometer underground, one iteration at a time.
Image: USGS Hawaiian Volcano Observatory
From Reflected Sound to Subsurface Certainty
Full Waveform Inversion (FWI) treats subsurface imaging as an optimization problem: start from a rough velocity model, simulate the wavefield, compare it to recorded seismic traces, and iteratively update the model until the misfit is minimized. Each iteration is itself a full 3D wave-propagation simulation — meaning a single FWI survey can require thousands of simulations across a dataset of tens of petabytes. Malgukke provides the GPU-dense compute and NVMe-oF storage fabric that make this iteration loop tractable for mineral, geothermal, and energy exploration.
HPC Solution Architecture: Survey to Subsurface Model
A Full Waveform Inversion pipeline, comparable to workflows run on TACC's Frontera system:
Running on the TOP500: Frontera
Frontera — Texas Advanced Computing Center (TACC), UT Austin
Frontera debuted at No. 5 on the June 2019 TOP500 list with 23.5 petaflops, making it the fastest academic supercomputer in the world at the time. It remains TACC's leadership-class system, delivering large-scale allocations to research spanning "cosmology to hurricanes, earthquakes" and subsurface geoscience. The Texas Consortium for Computational Seismology, based at the same university's Bureau of Economic Geology, develops FWI and reverse-time-migration methods that scale directly onto machines of this class.
Yes — this is a real, currently operating NSF system that has held a TOP500 top-5 position and continues to serve the seismic-imaging research community, even as newer systems have since surpassed its original ranking.
Voices from the Field
Researchers advancing computational seismic imaging for exploration:
Sergey Fomel
Wallace E. Pratt Professor of Geophysics, UT Austin / Director, Texas Consortium for Computational Seismology
On modern seismic interpretation and what still separates human geophysicists from machines.
Listen to podcastAndreas Fichtner
Professor of Seismology and Wave Physics, ETH Zurich
On extending full-waveform inversion to fiber-optic distributed acoustic sensing for subsurface characterization.
Read paper