CSEDI: Collaborative Res.: Toward Mapping 3D Variations in Temperature and Chemical Composition in the Upper Mantle through a Mineral Physics-based Inversion of Seismic Waveforms

Yale University, New Haven CT

Investigators

Abstract

Romanowicz Karato This is a collaborative study between seismologists at UC Berkeley and mineral physicists at Yale University, to explore the feasibility of inverting seismic long period waveform data directly for 3D variations in effective temperature and major element chemistry in the upper mantle. Effective temperature will include the effects of temperature and water content, which cannot be easily separated from seismic data alone. Water content is likely to have large effects on rheology and therefore has significant influence on the dynamics of the Earth's mantle. They will use the mineral physics framework to delineate contributions from these various factors and provide estimates of relevant partial derivatives. Effects of major element chemistry can be distinguished from that of water, because water has a large influence on anelasticity whereas major element chemistry does not. On the other hand anelasticity is more strongly dependent on temperature than seismic velocity. At the heart of this study is the conjecture, supported by theoretical considerations, that the one step inversion of seismic waveforms for physical parameters such as composition C and (effective) temperature T*, should yield more stable results than the standard approach of first inverting for velocity and Q structure and then interpreting the obtained maps in terms of physical parameters. The latter is currently inhibited by the poor resolution in 3D upper mantle Q, due to contamination of amplitudes by largely unmodelled, but significant, effects of focusing and scattering. In the one step inversion the effects of C and T* are coupled in the inversion matrix elements, whereas in the standard approach, only the amplitudes depend on anelastic terms, while both phase and amplitudes depend on the elastic ones, leading to strong biases in the distribution of Q inferred. The coupled character of the equations in the one step approach, combined with the strong sensitivity of amplitudes to T* (through anelastic terms) should offset the approximate character of the available mineral physics partial derivatives and the errors due to inadequate theoretical treatment of focusing.