Adjoint-state traveltime tomography in Central Java enhanced by a machine-learning-assisted catalog


引用 0|浏览3
<p>As one of the most populous regions, Central Java in Indonesia is prone to high seismic and volcanic hazards, mostly due to the progressive northward subduction of the Australian plate beneath the Sunda plate. Detailed velocity structure of the crust and uppermost mantle beneath Central Java is critical for an improved understanding of the subduction processes and the associated seismicity and volcanism. Despite several independent isotropic velocity models proposed for the region, crustal-scale anisotropic structure, which reflects past and ongoing deformation, has rarely been investigated. The reasons behind this include that 1) conventional ray tracing may fail in strongly anisotropic crust, especially in a heterogeneous forearc setting; and 2) reliable anisotropy tomography requires sufficient data coverage, while the existing seismic networks in Central Java are restricted.</p> <p>In this study, we target on the crustal-scale P-wave azimuthally anisotropic structure beneath Central Java. The acquired seismic data were recorded by more than 200 seismic stations from multiple projects with different execution periods (from 6 months to > 2 years). To make full use of the open access data, machine learning phase picking and subsequent event association and location were applied to build a local earthquake catalog. The machine-learning-based workflow detects more than 1500 events, roughly double the amount by previous manual picking. Notably, the preliminary catalog includes a large number of earthquakes that were located in the offshore areas but were recorded by land stations to the north, resulting in huge back azimuthal gaps and potential bias in earthquake relocation. The current study attempts to involve depth phases, such as sPn and sPg, for more accurate earthquake locations and thus more reliable tomographic images. With the expanded and refined seismicity catalog, a ray-free adjoint-state traveltime tomography package called TomoATT will be used for the 3-D velocity heterogeneity and azimuthal anisotropy beneath Central Java.</p>
AI 理解论文