High Performance Computing for Geoscience Problems

这项研究的重点是开发计算技术和高性能的工具，涉及地震动力学和火山地震学中的波传播的社会相关问题。与这些地球物理过程相关的灾害越来越多地通过对底层物理过程的详细建模来了解。这是计算上的挑战，由于耦合多物理场性质的问题，存在许多时间和空间尺度，以及大的模拟量相比，小规模的过程，需要解决。地球中的非均匀材料性质和结构复杂性（如地形和断层网络）对数值方法施加了额外的限制。将开发复杂几何结构中地震波传播的计算框架。这会延长私家侦探的时间吗的新开发的，可证明稳定的高阶有限差分方法和多块网格耦合结构和非结构网格。这些工具将用于地震动力学和火山地震学中的问题，但可以扩展到涉及波传播的其他问题（例如，地震成像、断裂力学和增强的地热能生产）。将开发一个自适应网格加密（AMR）代码，用于使用不变的实验室测量摩擦定律参数模拟大震级地震。这种研究对于理解改变摩擦参数的影响是必要的，这在情景地震模拟中是常规的。AMR是必需的，因为模拟体积是数十万立方千米，在mm尺度上，需要在传播破裂尖端附近的百分之几的区域中的分辨率，即，断层从锁定到滑动的转变。此外，P.I.将协助指导学生，参与课程开发，并讲授计算科学的主题。

This research focuses on developing computational techniques and high-performance tools for societally relevant problems involving wave propagation in earthquake dynamics and volcano seismology. Increasingly, the hazard associated with these geophysical processes is being informed by detailed modeling of the underlying physical processes. This is computationally challenging due to the coupled multiphysics nature of the problem, existence of many temporal and spatial scales, and the large simulation volumes compared to the small-scale processes that need to be resolved. Additional constraints are placed on the numerical methods by the heterogeneous material properties in the earth and structural complexities such as topography and fault networks. A computational framework for seismic wave propagation in complex geometries will be developed. This will extend the P.I.?s newly developed, provably stable high-order finite difference method and multi-block grids using coupled structured and unstructured grids. These tools will be used for problems in earthquake dynamics and volcano seismology, but can be extended to other problems involving wave propagation (e.g., seismic imaging, fracture mechanics, and enhanced geothermal energy production). An adaptive mesh refinement (AMR) code will be developed for the simulation of large magnitude earthquakes using unaltered laboratory measured friction law parameters. Such studies are necessary to understand the effect of altering friction parameters as is routinely done in scenario earthquake simulations. AMR is required as the simulation volume is hundreds of thousands of cubic kilometers, with resolution, at the mm scale, needed in a few percent of the domain near the propagating rupture tip, i.e., transition from locked to sliding on the fault. Additionally, the P.I. will assist in advising of students, participate in course development, and lecture on topics in computational science.