constraints on global mantle flow and sks ... - geodynamics · lithosphere net rotation from...

Constraints on Global Mantle Flow andLithosphere Net Rotation from Seismic Anisotorpy

Clinton P. ConradDepartment of Geology and Geophysics

SOEST, University of HawaiiHonolulu HI, 96816clintc@hawaii.edu

Mark D. BehnDepartment of Geology and GeophysicsWoods Hole Oceanographic Institution

Woods Hole MA, 02543mbehn@whoi.edu

Introduction: Asthenospheric AnisotropyViscous shear in the asthenosphere accomodates relativemotion between the Earth’s surface plates and underlyingmantle, generating lattice-preferred orientation (LPO). Thus,observations of anisotropy can be used to constrain shear flow inthe asthenosphere, which is produced by relative motionbetween the mantle and the tectonic plates. Anisotropyobservations may also be influenced by lithospheric anisotropy,as well as the finite strain history of asthenospheric flow.

Development ofAsthenosphericAnisotropyWhen exposed to simpleshear, the fast axis ofolivine (A-type fabric)orients 45º from themaximum shear direction.Simple shear rotates thisfabric into the infinite strainaxis (ISA, the orientationafter infinite deformation)at a rate ΩISA (top).

The Grain-Orientation Lag (Π)We measure the Grain-Orientation Lag Parameter, Π, todetermine where the infinite strain axis (ISA) approximates thelattice preferred orientation (LPO). We find that Π < 0.5 for mostof the asthenosphere because simple shear orients the LPO inthe direction of the ISA faster that the ISA itself rotates with theflow. By contrast, the slowly-deforming lithosphere can bedistinguished from the asthenosphere by its large values of Π.

SKS Splitting: A Constraint on Global Mantle Flow and Net RotationBy comparing the ISA direction with a global dataset of SKS splitting observations, weevaluate our global flow model’s ability to predict observed anisotropy. Becausecontinental anisotropy may be influenced by a lithospheric component, we use oceanicobservations only when calculating misfit. By varying α and β, we can determinerelative importance of plate-driven, density-driven, and net rotation flows for an optimalfit to observations. We find 0.3 < β < 0.8 and α < 0.6 provide the best fit for anasthenospheric viscosity 10 times smaller than the upper mantle viscosity. More netrotation is permissible if the asthenospheric viscosity is higher, because less of theshear produced by net rotation is accommodated in the asthenosphere.

Conclusions1. The Infinite Strain Axis (ISA) is a good approximation for

the Lattice Preferred Orientation (LPO) of olivine crystalsthroughout most of the asthenosphere:

ISO~LPO because Π<0.5This simplifies the anisotropy predictions because strainintegration along flow lines is not necessary.

2. Using seismic anisotropy observations, we find that thecombination of plate-driven and density-driven flowsconstrain upper mantle viscosity to ~0.5 x 1021 Pa s,consistent with other estimates.

3. A net lithosphere rotation of 2-3 cm/yr (60% of HS3) ispermitted by the anisotropy observations, consistnet withBecker’s [2008] constraint. Larger net rotation is possiblefor greater asthenospheric visosity, which reduces shear.

4. For oceans, anisotropy is dominated by asthenosphericshear flow, and the lithospheric contribution is small. Forcontinents, lithospheric anisotropy is more importantbecause continental lithosphere is thicker, older, and moredeformed than its oceanic counterpart.

References are available upon request. Also see:Conrad, C.P., M.D. Behn, and P.G. Silver, Global mantle flowand the development of seismic anisotropy: Differencesbetween the oceanic and continental upper mantle, J. Geophys.Res., 112, B07317, doi:10.1029/2006JB004608, 2007.

Oceanic vs.ContinentalAnisotropyWe compare a globaldataset of observed SKSsplitting observations (toppanel) with the predictedISA axis determined fromthe global flow models.Using an approximate“best fit” choice of α=0.5and β=0.4 (and anasthenospheric viscosityof ηasth / ηum = 0.1), wefind the distribution ofmisfits shown on the left.It is clear that whileoceanic stations are wellfit, the continentalstations are, on average,not well fit. We suggestthat continents are poorlyfit because theiranisotropic fabric isdominated by a fossillithospheric componentthat depends on a longgeologic history ofdeformation, and cannotbe predicted by mantleflow models.

However, olivine crystals may also rotate with the flow (at arate Ωflow), if the flow deviates from simple shear (bottom).Kaminski and Ribe [2002] define the ratio of these tworotation rates as:

Π = Ωflow / ΩISAThus, if Π < 1, the infinite strain axis is a good approximationfor the LPO. We measure Π for viscous mantle flow todetermine where the ISA may be used to estimate LPO.

a) Density-Driven Flow:We assign mantle density heterogeneity inferred from seismictomography (S20RTSb, Ritemsa et al., 2004) using aconversion factor of 0.15 g cm-3 km-1 s.

b) Plate-Driven Flow (NNR):We impose plate motions (NUVEL-1A, DeMets et al., 1994) inthe no-net-rotation (NNR) reference frame as velocityboundary conditions.

c) Net Rotation (HS3)We impose a net rotation of the lithosphere consistent with theHS3 model of Gripp & Gordon [2002], which features a ~5cm/yr westward net rotation of the lithosphere.

Viscosity Structure:The lower lower mantle and asthenosphere have viscosities50, and 0.1 times the upper mantle viscosity. The viscositytransition from lithosphere to asthenosphere is gradual anddetermined by the lithosphere thickness, which variesspatially (e.g., Conrad & Lithgow-Beretlloni, 2006).

Models of Global Asthenospheric FlowWe use the finite element code CitComS to solve for global mantle flow, usinga linear combination of factors that produce relative motion between the platesand the underlying mantle:

ηasth / ηum = 0.1

ηasth / ηum = 1ηasth / ηum = 0.3

ηasth / ηum = 0.03

Surface Wave Tomography: Constraint on Mantle Flow ModelsWe compare the surface wave tomography model of Debayle et al. [2005] (at 200 kmdepth) to our mantle flow model by varying α and β as we did for SKS splitting above.In oceanic regions where the anisotropy magnitude is more than 0.95% (25% of themaximum value, after Becker et al., [2003]), we find a similar pattern to what wefound for the SKS splitting measurements (above). For surface wave anisortopy, wefind best fits using 0.5 < β < 2 and α < 0.6. Again, an asthenospheric viscosity 10times smaller than the upper mantle viscosity provides the best fit, and higherasthenospheric viscosity permits larger amounts of net rotation (larger α).

ηasth / ηum = 0.1

ηasth / ηum = 1ηasth / ηum = 0.3

ηasth / ηum = 0.03

ηasth / ηum = 1

Viscosity Scale Factor (log10 β)Viscosity Scale Factor (log10 β)

Ref

eren

ce F

ram

e S

cale

Fac

tor (α

)R

efer

ence

Fra

me

Sca

le F

acto

r (α

)

Average Angular Misfit (Degrees)

Viscosity Scale Factor (log10 β)Viscosity Scale Factor (log10 β)

Ref

eren

ce F

ram

e S

cale

Fac

tor (α

)R

efer

ence

Fra

me

Sca

le F

acto

r (α

)

Average Angular Misfit (Degrees)

Ocean Basins (N0=21)

All Continents (N0=889)

North America (N0=356)

South America (N0=89)

Europe & Asia (N0=822)

Africa (N0=822)

N/N

0N

/N0

N/N

0N

/N0

N/N

0N

/N0

Misfit (Degrees)