High-resolution Waveform Tomography of Mantle Transition Zone and Slab Structure beneath Northeast China

Abstract:

Northeast China has undergone widespread extension and magmatism since Late Cretaceous. There are many Cenozoic volcanoes in this region and a few of them are still active today, such as Changbaishan and Wudalianchi. Previous tomography models show stagnant slabs within the transition zone beneath NE China, and suggest deep slab control on the regional tectonics and volcanism. Proposed mechanisms for the magmatism include: 1) a mantle plume, 2) hot upwelling above the stagnant slab by deep dehydration and 3) upwelling induced by deep slab segmentation and detachment. To date, NE China seismic images still contain enough uncertainty to allow for multiple models. Using the dense seismic data coverage in NE China and adjacent regions our goal is to make high-resolution image of the transition zone and slab structure to test the origins of intraplate volcanism.

Recently Chen et al. (2015) developed a 3D model for P and S velocity structure beneath East Asia using adjoint tomography using the SPECFEM3D synthetic technique and cross-correlation time shifts as the objective function. We use their model as a starting model and further improve the resolution by fitting waveforms to a shorter period (from ~12s to ~5s) using the correlation coefficient as the objective function. The new objective function is closely related to the L2 waveform misfit but is insensitive to a constant amplitude ratio between the synthetic and data within each time window used. This feature is desirable because the absolute amplitude can be hard to model as it can be affected by many factors difficult to incorporate in simulations, such as site effects, source magnitude and mechanism error or even poor calibration of instruments. During inversion we focus specifically on the transition zone and the structure of slabs with the goal of fitting triplicated and multipath body waves. We have performed a waveform inversion experiment using data from a single deep earthquake. Excellent fits of the triplicated waveforms are achieved after 5 iterations, which show the feasibility of our waveform inversion procedure. Results from an inversion of a large data set will be presented.