Constraints on Earth's Radial Velocity Structure from the Global Correlation Wavefield

Several Earth's radial velocity models have been developed as references for a wide range of studies on Earth's interior over the last few decades. These models provide a good representation of the overall features of the seismic wavefield in a broad frequency band. However, due to different datasets and approaches used in the studies, some discrepancies exist near the internal sharp discontinuities (e.g., the core-mantle boundary and the inner-core boundary) among these models. To place tighter constraints on the Earth's radial structure, here we rely on the global earthquake-coda correlation wavefield, which has shown a promising potential to study Earth's deep structure. Firstly, we construct the observed global correlogram (a 2D representation of the correlation wavefield) by stacking cross-correlations of the late-coda seismic wavefield (3-9 hour after the event origin time) from the ten selected large earthquakes recorded by seismic stations globally. The correlograms derived from these earthquakes all display a multitude of prominent correlation features that have been shown to exist due to many cross-terms of reverberating body-waves, a principle fundamentally different from the reconstruction of surface waves in the ambient-noise correlograms. Then we select the pronounced correlation features as observations and fit both the travel-times and waveforms of these features by computing synthetic correlograms through a series of models, including PREM, ak135, ek137, etc. Cross-correlation coefficients for individual correlation features are computed as misfit measurement to search for the best-fitting model. Our preliminary results show, inter alia, that the ak135 model gives a better fit to the correlation features that are sensitive to core structures than the PREM.