Seismic waveform tomography on continental scales based on the spectral-element and adjoint methods

数值地震学和计算机科学的最新进展使今天的地震学家能够以高精度常规地模拟弹性波在非均匀介质中的传播。这一发展建议超越一般的数据简化（例如提取旅行时间），并为高分辨率地震层析成像分析全波形。我们的研究项目主要是开发一种用于上地幔结构详细成像的全波形反演（FWI）技术，并将其应用于澳大利亚大陆。为此，我们建议：（1）建立澳大利亚大陆高质量波形数据库;（2）设计全波形错配，以便从地震图中提取尽可能多的可靠信息;（3）采用谱元法结合伴随法计算层析成像;（4）评估正确解释三维地球结构的重要性，特别是海洋-大陆边界，（5）比较FWI与经典射线层析成像的分辨率能力。FWI的主要创新之处在于，在具有横向变化的介质（如海洋-大陆过渡）中，正演模拟是准确的。因此，可以期望合成波形和所得到的断层图像更真实。此外，所利用的信息量明显高于基于射线或简正波理论的经典反演方法。利用这些丰富的信息将导致更好的上地幔结构的约束模型，对相关的地球动力学问题的影响。

Recent progress in numerical seismology and computer science enables today s seismologists to model elastic wave propagation through heterogeneous media routinely with high accuracy. This development suggests to go beyond the common data reduction (e.g. extracting travel times) and to analyse full waveforms for the purpose of high-resolution seismic tomography. Our research project is focused on the development of a full waveform inversion (FWI) technique for the detailed imaging of upper mantle structure and its application to the Australian continent. For this we propose (1) to compile a database of highquality waveform data for the Australian continent; (2) to design full waveform misfits in order to extract as much robust information as possible from the seismograms; (3) to employ a spectral-element method in conjunction with the adjoint method for the computation of tomographic images; (4) to assess the importance of correctly accounting for 3D Earth structure in general and for ocean-continent boundaries in particular, and (5) to compare the resolution capabilities of a FWI with those of classical ray tomographies. Key innovations of the FWI are that the forward modelling is accurate in media with lateral variations such as ocean-continent transitions. The synthetic waveforms and the resulting tomographic images can therefore be expected to be more realistic. Moreover, the amount of exploited information is significantly higher than in classical inversion methods based on ray or normal mode theory. Exploiting this wealth of information will lead to better constrained models of upper mantle structure with impact on the related geodynamic issues.