ABR: Toward Seismic Tomography Based Upon Adjoint Methods

该项目的主要目标是利用地震数据，基于现代数值方法和成像技术，对地球结构和震源进行成像。我们将进一步开发和增强用于模拟三维地震波传播的软件，特别强调图形处理单元的计算，而不是传统的中央处理单元，可能会在模拟速度上提供一个数量级的增加。该提案的更广泛影响包括用于模拟地震波传播的开源软件包。基于这些包的模拟可用于研究地震危险，并有助于确定地球内部的层析成像和震源参数。在计划研究期间开发的所有软件将通过地球动力学计算基础设施（geodynamics.org）免费提供。地震波传播的模拟考虑了地壳和地幔的非均质性、地形和测深、地震各向异性、衰减、流固相互作用、自重力、自转和海洋。这个项目的主要智力价值围绕着利用正演建模工具的力量来提高地球内部和地震破裂过程的图像质量。该方法是基于所谓的伴随技术结合共轭梯度方法，我们称之为“伴随层析成像”的方法，最大限度地减少模拟和观测地震图之间的剩余。具体来说，在南加州的成功应用之后，我们将1)开发基于开源gpu的正演和伴随谱元求解器，2)对欧洲进行伴随层析成像，3)进一步开发基于伴随方法的噪声互相关层析成像，4)向整个地球的伴随层析成像迈进，5)扩展南加州和全球“ShakeMovie”网络基础设施。

The primary goal of this project is to use seismic data to image Earth structure and seismic sources based on modern numerical methods and imaging techniques. We will further develop and enhance software for the simulation of 3D seismic wave propagation, with a particular emphasis on computing on Graphics Processing Units, rather than traditional Central Processing Units, potentially providing an order of magnitude increase in simulation speed. The broader impacts of the proposal include open-source software packages for the simulation of seismic wave propagation. Simulations based on these packages may be used to investigate seismic hazard and aid in the determination of tomographic images of Earth's interior and earthquake source parameters. All software developed during the course of the proposed research period will be made freely available via the Computational Infrastructure for Geodynamics (geodynamics.org).The simulations of seismic wave propagation account for heterogeneity in Earth's crust and mantle, topography& bathymetry, seismic anisotropy, attenuation, fluid-solid interactions, self-gravitation, rotation, and the oceans. The main intellectual merit of this project revolves around harnessing the power of the forward modeling tools to enhance the quality of images of Earth's interior and the earthquake rupture process. The approach is to minimize remaining between simulated and observed seismograms based on so-called adjoint techniques in combination with conjugate gradient methods, an approach we refer to as "adjoint tomography". Specifically, following a successful application in southern California, we will 1) develop open-source GPU-based forward and adjoint spectral-element solvers, 2) perform adjoint tomography of Europe, 3) further develop noise cross-correlation tomography based on adjoint methods, 4) move towards adjoint tomography of the entire planet, and 5) extend the southern California and global "ShakeMovie" cyber infrastructure.