Towards Inverting Seismic Waveform Data for Temperature and Composition in the Earth's Upper Mantle

Abstract

B: Unraveling the physical state of the upper mantle, including the transition zone, is one of the key factors for understanding the Earth's mantle dynamics. Knowledge of mantle temperature and composition is mainly based on the interpretation of seismological observations based on insights from mineral physics. Despite the progress made to image the 3-D seismic structure of the upper mantle, its interpretation in terms of physical parameters is still challenging and it requires a truly interdisciplinary approach. Due to the better knowledge of the elastic and anelastic properties of mantle minerals at high temperatures and pressures, such an approach is now becoming feasible. We propose a new waveform inversion procedure, based on a formalism previously developed at Berkeley for global elastic and anelastic tomography, and using our existing collection of long-period fundamental and higher mode surface waveforms. Here, we incorporate mineral physics data at an early stage of the process to directly map lateral variations in temperature and composition, using recent estimates of the temperature and composition derivatives of seismic velocities (∂lnV/∂lnT,C). Anelasticity introduces a non-linear dependence of the seismic velocities with temperature throughout the upper mantle, and phase-transitions confer a non-linear character to the compositional derivatives as well, therefore the kernels should be re-computed after each iteration of the inversion. We discuss ways to address the non-linearities, as well as uncertainties in the partial derivatives. In addition to constraining the lateral variations in temperature or composition, the models can have implications on the average structure of the upper mantle. The most-common accepted physical 1-D structure had problems to satisfactorily fit seismic travel time data, requiring a slower TZ to improve the fit. However, these data do not have sufficient coverage (and resolution) in the TZ. A complementary outcome of our models will be to shed light on whether the seismic data require a modification of the physical structure in the transition zone and if the three-dimensional heterogeneity introduces a significant shift of the average physical structure away from adiabatic pyrolite.