Collaborative Research: ITR/NGS: Multiresolution High Fidelity Earthquake Modeling: Dynamic Rupture, Basin Response, Blind Deconvolution Seismic Inversion, and Ultrascale Computing

这个项目的目的是开发一个三维地震反演模型，然后使用它来获取关于大洛杉矶盆地的三维结构和最近南加州主要地震(如兰德斯和北岭地震)中断层破裂的性质的信息。在项目结束时，这些工具还将应用于关东盆地和1923年关东地震。该项目将利用最近和预计在南加州部署的地震网络的增长，以及超级计算系统能力的增加。反问题是非常大的。该项目的一部分涉及构建逆模型，以便它在大型并行系统上运行良好。大多数工作将在单个系统上完成，例如国家科学基金会和能源部计算中心提供的那些系统；然而，由于反向模型需要多次正向和伴随运行的迭代，研究人员计划通过在TeraGrid的分布式计算环境中部署反向模型来利用它的粗粒度。这项工作的长期好处可能是更好地了解大洛杉矶盆地的地质结构，更好地了解断层破裂，并提高预测强烈地面运动的能力。这里开发的这类复杂的地震反演工具将帮助研究人员从美国阵列将在未来几年开始上线的数据中获得最大利益。

This project is aimed at developing a three-dimensional seismic inverse model and then using it to obtain information about the both three-dimensional structure of the Greater Los Angeles Basin and the nature of fault rupture during recent major Southern California earthquakes such as the Landers and Northridge earthquakes. At the end of the project these tools will also be applied to the Kanto Basin and the 1923 Kanto Earthquake. The project will capitalize on both the recent and projected growth in seismic networks deployed in Southern California and the increase in the capabilities of supercomputing systems. The inverse problem is very large. Part of the project involves structuring the inverse model so that it will run well on large parallel systems. Most of this work will be done on single systems such as those available at National Science Foundation and Department of Energy computing centers; however, because the inverse model requires many iterations of forward and adjoint runs, it has a coarse granularity that the investigators plan to exploit by also deploying the inverse model on the distributed computing environment of the Teragrid. The long-term benefits of this work are likely to be a better understanding of the geological structure of the Greater Los Angeles Basin, a better understanding of fault rupture, and improvements in the ability to predict strong ground motion. Sophisticated seismic inversion tools of the sort developed here will help researchers gain maximum benefit from the data that will start to come on-line in the next few years from the United States Array.