HR: 0800h
AN: U41A-0706    [Abstracts]
TI: Testing Models of Thermo-chemical Convection Against Models from Probabilistic Tomography
AU: * Deschamps, F
EM: deschamp@geo.uu.nl
AF: Utrecht University, Budapestlaan 4, PO box 80021, Utrecht, 3508 TA Netherlands
AU: Trampert, J
EM: jeannot@geo.uu.nl
AF: Utrecht University, Budapestlaan 4, PO box 80021, Utrecht, 3508 TA Netherlands
AU: Tackley, P J
EM: ptackley@ess.ucla.edu
AF: University of California, 595 Charles Young Drive East, Los Angeles, CA 90095 United States
AB: The main difficulty of a thermo-chemical interpretation of seismic tomography is the existence of a strong trade-off between temperature and composition. This trade-off can never be fully resolved, but we can identify all possible models compatible with data. In previous work, we have developed the technique of probabilistic tomography, which gives independent probability density functions for long wavelength models (spherical harmonic degree 2, 4 and 6) of bulk-sound and shear wave speed, density and boundary topography in the mantle. Using appropriate sensitivities (which take into account our ignorance on the thermodynamic reference state and the published range of mineral physics data), we have converted the results from probabilistic tomography into likelihoods of variations in temperature, perovskite and iron content throughout the lower mantle. Several robust features emerge which shed a new light on the nature of the lower mantle. Throughout the mantle temperature variations are much weaker than classically inferred from shear wave speed alone. Compositional variations are essential to explain the seismic data. In most places, the inferences are robust, i.e. the amplitudes of chemical and temperature variations are much larger than the uncertainties inferred from the width of the likelihoods. Below 2000 km, the correlation between relative shear wave speed variations and temperature is quite low, and we find that the much debated superplumes beneath the Pacific and Africa are due to an enrichment in perovskite and iron, rather than to high temperatures. These features are therefore denser than the surrounding mantle, and hence not buoyant. Our results clearly show that chemical variations are a key ingredient to model mantle dynamics, but they cannot yet discriminate between different possible models of thermo-chemical convection. We report first statistical comparisons between some chosen models of thermo-chemical convection, and likelihoods of seismic parameters observed by probabilistic tomography.
DE: 8180 Tomography
DE: 8120 Dynamics of lithosphere and mantle--general
DE: 8124 Earth's interior--composition and state (old 8105)
DE: 7207 Core and mantle
SC: Union [U]
MN: 2004 AGU Fall Meeting