Towards A Next Generation High Resolution Global Elastic 3D Model of the Whole Mantle: Constraining Vp, Vs, and Density Structure

Knowledge of the 3D distribution of elastic parameters and density in the earth's mantle provides important insights into the evolution and dynamics of the planet. Variations in seismic velocity can be influenced by various factors, such as compositional heterogeneity or lateral variations in temperature. However, distinguishing the thermal and chemical effects is particularly challenging. Over the decades, seismic tomography has provided important constraints on mapping both isotropic and anisotropic shear velocity variations in the mantle. The dataset used in the construction of our latest whole mantle radially anisotropic model SEMUCB_WM1 (French and Romanowicz, 2014, 2015) includes long period (>60 s) first and second orbit fundamental and overtone surface wavetrains and long-period (>30 s) body waveforms from 273 globally distributed events, sensitive primarily to Vs and (Vsh/Vsv)2. However, in the construction of this model, density () and Vp are related to Vs through constant "standard" scaling factors (Rp=dlnVs/dlnVp and R=dlnVs/dln, where 1/Rp=0.5 and 1/R=0.3), because there is not enough sensitivity in the type of data considered to allow independent inversion for Vp and density structure. The variation in Rp and R are key parameters in thermal and compositional evolution of the earth when combined with insights from mineral physics. In this study, we aim to combine various data types previously used in separate studies to further constrain Vs, Vp, and models inasmuch as possible independently. We have extended our global collection of waveform data by including shorter period body waveforms (>18 s) and normal mode splitting functions, which are sensitive to Vp. We first focus on updating our (Vs, x) model to also include long wavelength heterogeneity in Vp, using SEMUCB_WM1 as our starting model. Forward modeling of waveforms is performed using the Spectral Element Method, and sensitivity kernels are computed using NACT (Non-linear Asymptotic Coupling Theory, Li and Romanowicz, 1995). In this preliminary study, splitting functions are modeled in the framework of first order perturbation theory. After finalizing the first step, the structure will be included in the inversion. By constructing Vs, Vp, and models independently, we are able to investigate the heterogeneity of Rp and R.