Anisotropic seismic tomography: Potentials and pitfalls Mark Panning University of Florida CIDER Research Talk 7/5/2010

Cartoon land motivation: tomography of scientists

What is seismic anisotropy? ?

Origins of mantle anisotropy Single crystal has anisotropic elastic properties But large regions of the Earth appear nearly isotropic to seismic waves!

Origins of mantle anisotropy A random mix of orientations makes seismic waves see an isotropic average

Origins of mantle anisotropy Deformation can lead to preferential orientation (LPO) and seismic anisotropy

Complications Anisotropy depends on deformation mechanism –Varies by stress state and grain size –Varies by volatile content Depends on integrated strain history Requires many model parameters to describe

Fabric development from Karato et al, 2008

Not all gloom and doom Natural samples (e.g. Montagner and Anderson, 1989) and numerical modeling (e.g. Becker et al, 2006) suggest hexagonal symmetry is dominant

Why we like hexagonal symmetry Reduces number of elastic coefficients from 21 to 5 (2 isotropic properties, 3 anisotropic ones) plus 2 orientation angles With scaling, we can reduce the number of parameters even further (scale Vp to Vs, and the various anisotropic parameters to each other)

Why we like finite strain ellipses from Becker et al, 2003

“Vectorial tomography” Arbitrarily oriented hexagonal medium Can be linearized – with assumptions to reduce number of parameters Also can invert directly for anisotropic strength and orientation angles symmetry axis

Nonlinearity Sensitivity to strength and orientation of anisotropy depends on the starting model

Potential? Chevrot and Monteiller, 2009 synthetic tests with non-linear inversion of body wave splitting data

Matching models from Gaboret et al, 2003

Matching models from Becker, 2008

12% 7% 4% Upper mantle anisotropy

Correlation with ridges

Inconsistency of radial anisotropy models From Becker et al., 2008 Correlation of VS models above 350 km Correlation of ξ models above 350 km

Poor crustal corrections - source of some inconsistency? Inversions of synthetic data using Crust2.0 but no mantle anisotropy show anisotropy From Bozdağ and Trampert, 2008 From Lekic et al, 2010

The crust and anisotropic models All seismic data is influenced by crustal structure Varying crustal models has similar effect on data fit as mantle radial anisotropy (Ferreira et al, 2010) Corrections based on linear perturbations from 1D crustal models are inadequate for long-period data

Testing the impact of crustal corrections SAW642AN (as well as S362WMANI) incorporated non-linear crustal corrections based on regionalized mode calculations Other methods of non-linear crustal corrections exist We can compare models using different corrections and look at stability of model parameters.

V S model SAW642ANSAW642ANb

Changing ξ model

What remains General pattern of radial anisotropy beneath oceanic and continental lithosphere remains. Ridge signature also remains.

Troublesome details – D” structure SAW642AN New corrections – more regularization Same dataset with linear corrections and longer wavelengths New corrections – less regularization

Takeaway message Anisotropic modeling has great potential for constraining flow patterns (and therefore mantle rheology, etc.) Inverse approach and crustal correction matter and can strongly affect anisotropic models In order to resolve anisotropic structure (and other secondary effects like attenuation), we need to figure out the crust!