Exciting news! We're transitioning to the Statewide California Earthquake Center. Our new website is under construction, but we'll continue using this website for SCEC business in the meantime. We're also archiving the Southern Center site to preserve its rich history. A new and improved platform is coming soon!

Poster #078, Seismology

Moment tensor inversion for deep earthquakes at the Tonga-Kermadec subduction zone using 3-D Green’s functions

Xueyan Li, & Hejun Zhu
Poster Image: 

Poster Presentation

2020 SCEC Annual Meeting, Poster #078, SCEC Contribution #10358 VIEW PDF
Seismic moment tensors from the Global CMT catalog have been widely used to infer source mechanisms and physical properties around local regions where earthquakes occur. However, the zero-trace constraint used in the source inversion algorithm may be not suitable for earthquakes occurred at depths greater than 300 km, considering that mechanisms for deep earthquakes are different from shallow events. Besides, the velocity model in use is quite simple, without considering complicated and rapidly varying velocity structure at subduction zones, which may introduce errors in the moment tensor solutions of deep earthquakes. In this study, we choose 232 deep earthquakes that occurred at the Tonga-...Kermadec subduction zone and perform moment tensor inversion using 3-D Green’s functions. The misfit function is defined as least-squares waveform differences for 20-60s body waves and 50-150s surface waves. Moment tensors from the Global CMT catalog are used as starting models, while Green’s functions are computed using SPECFEM3D_GLOBE with a recently constructed three-dimensional tomographic model AU32, including fine-scale velocity structures as well as seismic anisotropy beneath the Australasian region and the Tonga-Kermadec subduction zone. Our results show reductions in scalar moment for most of the selected deep earthquakes, and a large portion of selected events has non-double-couple components, which is consistent with the input catalog as well as seismic observations over decades. Meanwhile, about two-thirds of these moment tensors contain negative trace while the rest one-third present positive trace. We further perform control experiments using a subset of 70 events to illustrate that the usage of a 3-D velocity model is more essential to cause data misfit reduction compared with dropping the zero-trace regularization.