A new 3-D mantle density model from recent normal-mode measurements

<p>Constraints on the 3-D density structure of Earth’s mantle provide important insights into the nature of seismically observed features, such as the Large Low Shear Velocity Provinces (LLSVPs) in the lower mantle under Africa and the Pacific. The only seismic data directly sensitive to density variations throughout the entire mantle are normal modes: whole Earth oscillations that are induced by large earthquakes (M<sub>w</sub> > 7.5). However, their sensitivity to density is weak compared to the sensitivity to velocity and different studies have presented conflicting density models of the lower mantle. For example, Ishii & Tromp (1999) and Trampert et al. (2004) have found that the LLSVPs have a larger density than the surrounding mantle, while Koelemeijer et al. (2017) used additional Stoneley-mode observations, which are particularly sensitive to the core-mantle boundary region, to show that the LLSVPs have a lower density. Recently, Lau et al. (2017) have used tidal tomography to show that Earth's body tides prefer dense LLSVPs.</p><p>A large number of new normal-mode splitting function measurements has become available since the last density models of the entire mantle were published. Here, we show the models from our inversion of these recent data and compare our results to previous studies. We find areas of high as well as low density at the base of the LLSVPs and we find that inside the LLSVPs density varies on a smaller scale than velocity, indicating the presence of compositionally distinct material. In fact, we find low correlations between the density and velocity structure throughout the entire mantle, revealing that compositional variations are required at all depths inside the mantle.</p>

Viscoelasticity of the lower mantle from forward modeling of normal modes and solid Earth tides

<p>Grain scale diffusive processes are involved in the rheology at convective timescales, but also at the transient timescales of seismic wave propagation, solid Earth tides and post-glacial rebound. Seismic and geodetic data can therefore potentially provide constraints on lower mantle properties such as grain size that are unconstrained otherwise. Current models of the transient viscosity of the lower mantle infer an absorption band of finite width in frequency. Seismic models predict a low frequency end to the absorption band at timescales corresponding to the longest normal modes of about an hour. By contrast, geodetic models infer the onset of an absorption band at these frequencies to cover anelastic deformation at timescales up to 18.6 years. A difficulty in extracting frequency dependence from mode and tide data is its convolution with depth dependence.</p><p>To circumvent this problem we select a distinct set of seismic normal modes and solid Earth body tides that have similar depth sensitivity in the lower mantle. These processes collectively span a period range from 7 minutes to 18.6 years. This allows the examination of frequency dependent energy dissipation over the lower mantle across 6 orders of magnitude. To forward model the transient creep response of the lower mantle we use a laboratory-based model of intrinsic dissipation that we adapt to the lower mantle mineralogy. This extended Burgers model represents an empirical fit to data principally from olivine, but also MgO and other compounds. The underlying microphysical model envisions a sequence of processes that begin with a broad plateau in dissipation at the highest frequencies after the onset of anelastic behavior, followed by a broad absorption band spanning many decades in frequency. The absorption band transitions seamlessly into viscous behavior. Since dissipation both for the absorption band and for (Newtonian) viscous behavior is due to diffusion along grain boundaries there can be no gap between the end of the absorption band and onset of viscous behavior.</p><p>Modeling of the planetary response to small strain excitation necessitates consideration of inertia and self gravitation. The phase lag due to solid Earth body tides therefore does not correspond directly to the intrinsic dissipation measured in the laboratory as material property. We have developed a self consistent theory that combines the planetary response with time-dependent anelastic deformation of rocks. Results from a broad range of forward models show that at lower mantle pressures periods of modes fall onto the broad plateau in dissipation at the onset of anelastic behavior. This explains the apparent frequency independence or even negative frequency dependence observed for some normal modes. At longer timescales, solid Earth tides fall on the frequency-dependent absorption band. This reconciles seemingly contradictory results published by seismic and tidal studies. Observations at even longer timescales are needed to constrain the transition from absorption band to viscous behavior.</p>

A Test of Recent Inferences of Net Polar Ice Mass Balance based on Long-Wavelength Gravity

A comprehensive analysis of satellite datasets has estimated that the ice sheets of Greenland, West Antarctica, the Antarctic Peninsula, and East Antarctica experienced a net mass loss of −100 ± 92 Gt yr−1 over the period 1992–2000 and −298 ± 58 Gt yr−1 over the period 2000–11, representing an increase of −198 ± 109 Gt yr−1 between the two epochs. The authors demonstrate that the time rate of change of the degree-four zonal harmonic of Earth's gravitational potential provides an independent check on these mass balances that is less sensitive to uncertainties that have contaminated previous analyses of the degree-2 zonal harmonic [e.g., due to ongoing glacial isostatic adjustment (GIA), solid Earth body tides, and core–mantle coupling]. For the period 2000–11, the signal implied by the ice sheet mass flux cited above is (3.8 ± 0.6) × 10−11 yr−1, whereas the change in the harmonic across the two epochs is (2.3 ± 1.1) × 10−11 yr−1. In comparison, using satellite laser ranging (SLR) data, the authors estimate a GIA-corrected value of (3.8 ± 0.6) × 10−11 yr−1 for the epoch 2000–11 and a change across the two epochs of (5.3 ± 1.6) × 10−11 yr−1. The authors conclude that the former supports recent estimates of melting over the last decade, whereas the latter suggests either that estimated melt rates for the earlier epoch were too high or that the uncertainty associated with the SLR-based inference of during the earlier epoch is underestimated.