Partial differential equations for seismic imaging.

The goal of the project is to implement specific numerical algorithms for rendering high resolution seismic images, using a parallel processing framework.

Seismic imaging is the standard technology used for creating accurate images of the earth’s subsurface, which is applied to the commercial exploration of oil and gas resources, monitoring carbon dioxide sequestering sites, and monitoring environmental resources such as groundwater. It is also closely related to medical imaging through ultrasound and microwave imaging. The imaging technology is based on accurate numerical modeling of the partial differential equations that describe the physics of seismic wave propagation in the earth, including the acoustic wave equation, the elastic wave equation, and the viscoelastic wave equation. These equations can only be solved numerically, and there is an extensive literature on a variety of numerical methods that give effective, accurate solutions that are good enough for commerically useful imaging.

To make this code useful to a wider audience, both academic and commercial, it is important to implement the research-level code in a state-of-the-art software framework that takes advantage of all the advance hardware resources present in the available computing environment. In particular, we will make use of both multicore central processors (CPUs) as well as available graphics processing cards (GPUs).

In tandem, the students will implement our existing wavefield propagator code into the parallel framework. The phase space methods we have naturally partition into frequency bands that can be processed independently, so the route to parallelism is clear. Individual shot records can again be processed independently, so there is an alternative parallel processing route to evaluate.

Rituraj Shukla
Faculty Supervisor: 
Dr. Michael Lamoureux