Po Chen - US grants

0930040
Chen

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5). Granted funds will support acquisition of a cluster of NVIDIA Tesla S1070 1.44 GHz GPU computing nodes to augment an existing Linux CPU cluster at the University of Wyoming. Graphic Processing Units (GPUs) are designed to provide high throughput on parallel computations using many-core chips for which the primary market has been the graphic intensive gaming industry. NVIDIA has recently released the CUDA (Compute Unified Device Architecture) framework GPU programming interface making it possible to develop parallel codes to run on GPUs at much higher price/performance than is possible on CPUs. The PIs will explore parallelism and scalability of scientific computing on heterogeneous clusters composed of hardware with very different processing capability, speed and throughput and develop a set of open-source, GPU-accelerated modeling and inversion codes for geoscience research and to teach General-Purpose computation on GPUs (GPGPU) to graduate students at Geology & Geophysics Department, Computer Science Department and Mathematics Department in University of Wyoming. The PIs will develop GPU codes for seismic wave propagation simulations and fullwave seismic tomography and fluid dynamics simulations for oil reservoir modeling, subsurface fluid flow and solute transport and mantle convection.

This award facilitates scientific research using the large, new, computational resource named Blue Waters being developed by IBM and scheduled to be deployed at the University of Illinois in 2011. It provides travel funds to support technical coordination between the principal investigators, the Blue Waters project team and the vendor technical team.

The project involves porting to the Blue Waters system and refining two codes for the propagation of seismic energy and their use in constrained optimization problems to determine the geological structure of the Earth from seismic data. The award will enable the project team to evaluate whether the use of these codes on Blue Waters is likely to lead to breakthrough science in the following areas: the deep structure of the Earth, including the topography of the core-mantle boundary and the structure of the inner Core; a better understanding of the geological formations in regions with high seismic risks, and predictions of ground motion that will allow engineers to explore performance-based design of built structures; and more accurate pictures of geology in mineral-rich areas to reduce the risk and cost associated with drilling in complex geological environments.

The project is led by two new investigators and the insights gained into developing software for the next generation of large, parallel computers will be incorporated into new computational science teaching at the University of Wyoming. If the project is successful, it will provide a capability for better characterizing mineral resources in mineral-rich areas of the U.S. such as Wyoming as well as improving scientific understanding of the deep Earth.

The PI will conduct full-3D waveform tomography (F3DT) for the seismic velocity and attenuation structure on the San Andreas Fault zone around the Parkfield area in California. In the F3DT algorithm, both the reference structural model and the derived model perturbations are all 3D in space and the full-wave sensitivity (Fréchet) kernels are calculated from the full physics of 3D wave propagation. Joint inversions will be carried out for both seismic velocity and attenuation structures of the San Andreas Fault Zone using waveform recordings from natural and/or man-made earthquakes. By integrating high-quality waveform data into the F3DT algorithm, they expect to substantially enhance both the resolution and the accuracy of the 3D seismic velocity and attenuation structure in the Parkfield area.

Variations in physical properties of a fault zone may influence the generation, propagation and arrest of large earthquakes. Results from previous studies suggest that inelastic processes such as migration of fluids in fractures or microcracks and the associated changes in pore pressure and chemical effects such as stress corrosion and pressure solution may lead to mechanical failure and nucleation of earthquakes. Since attenuation structure is sensitive to fracturing, crack accumulation and partial saturation, the proposed high-resolution full-3D waveform tomography for fault-zone attenuation structure will provide important additional constraints to improve our understanding of fault-zone processes.

This project is funded jointly by the EPSCoR, Earthscope and Geophysics programs.

The principal investigators of this proposal are a collaboration of seismologists, computer scientists, and structural geologists who are developing computational platforms that can combine seismological and geological information into three dimensional (3D) representations of Earth structure. They derive the seismological information from earthquakes, controlled-source experiments, and observations of the ambient seismic field generated by ocean waves and other near-surface disturbances. They derive the geological information from an even wider variety of sources: field mapping, electromagnetic remote sensing, geotechnical measurements, drilling and well-logging, fault trenching, and laboratory measurements of material properties.

The systematic integration of such diverse datasets into unified structural representations (USRs) is a key problem of geoinformatics. The goal of this project is to develop for community use the complex cyberinfrastructure needed to manage the ?USR lifecycle?, including the effective application of high performance computing (HPC) to invert joint datasets in the iterative improvement of USRs. The lifecycle is initiated by the integration of several components into a starting model; the starting model is refined by full-3D tomographic inversion of seismic waveform data; the refined model is validated for its proposed applications and then disseminated to user communities. This new model becomes the main component of the starting model for the next lifecycle.