The Pawsey Supercomputing Centre


Full Waveform Inversion of 3D seismic data on HPC x86 clusters

3D seismic imaging is important to a wide variety of geophysical earth science problems, ranging from studies of the deep crust and mantle including plate tectonics and earthquakes; to the exploration, discovery, development and monitoring of valuable resources such as petroleum and minerals; to environmental challenges such as CO2 sequestration, groundwater studies and contaminant flow. In order to obtain accurate 3D images of the earth, a nonlinear inverse problem must be solved to estimate the 3D distribution of wave propagation velocity. Full waveform inversion (FWI) is an advanced approach to solving this 3D velocity inverse problem, however it requires significant HPC research to make it tractable since our 3D algorithms can require 10-100 GB of memory, seismic data sets can range in size from 10TB to 10PB, and the imaging/inversion algorithms require 10s to 1000s of Exaflops.

As an example, accurate 3D seismic imaging is extremely valuable for finding and producing petroleum reservoirs.  For example, $150 Billion is currently being invested to develop giant LNG gas fields offshore.

Comments are closed.