No abstract

SUMMARY Iceland represents one of the most well-known examples of hotspot volcanism, but the details of how surface volcanism connects to geodynamic processes in the deep mantle remain poorly understood. Recent work has identified evidence for an ultra-low velocity zone in the lowermost mantle beneath Iceland and argued for a cylindrically symmetric upwelling at the base of a deep mantle plume. This scenario makes a specific prediction about flow and deformation in the lowermost mantle, which can potentially be tested with observations of seismic anisotropy. Here we present an investigation of seismic anisotropy in the lowermost mantle beneath Iceland, using differential shear wave splitting measurements of S–ScS and SKS–SKKS phases. We apply our techniques to waves propagating at multiple azimuths, with the goal of gaining good geographical and azimuthal coverage of the region. Practical limitations imposed by the suboptimal distribution of global seismicity at the relevant distance ranges resulted in a relatively small data set, particularly for S–ScS. Despite this, however, our measurements of ScS splitting due to lowermost mantle anisotropy clearly show a rotation of the fast splitting direction from nearly horizontal for two sets of paths that sample away from the low velocity region (implying VSH > VSV) to nearly vertical for a set of paths that sample directly beneath Iceland (implying VSV > VSH). We also find evidence for sporadic SKS–SKKS discrepancies beneath our study region; while the geographic distribution of discrepant pairs is scattered, those pairs that sample closest to the base of the Iceland plume tend to be discrepant. Our measurements do not uniquely constrain the pattern of mantle flow. However, we carried out simple ray-theoretical forward modelling for a suite of plausible anisotropy mechanisms, including those based on single-crystal elastic tensors, those obtained via effective medium modelling for partial melt scenarios, and those derived from global or regional models of flow and texture development in the deep mantle. These simplified models do not take into account details such as possible transitions in anisotropy mechanism or deformation regime, and test a simplified flow field (vertical flow beneath the plume and horizontal flow outside it) rather than more detailed flow scenarios. Nevertheless, our modelling results demonstrate that our ScS splitting observations are generally consistent with a flow scenario that invokes nearly vertical flow directly beneath the Iceland hotspot, with horizontal flow just outside this region.

Determinations of seismic anisotropy, or the dependence of seismic wave velocities on the polarization or propagation direction of the wave, can allow for inferences on the style of deformation and the patterns of flow in the Earth’s interior. While it is relatively straightforward to resolve seismic anisotropy in the uppermost mantle directly beneath a seismic station, measurements of deep mantle anisotropy are more challenging. This is due in large part to the fact that measurements of anisotropy in the deep mantle are typically blurred by the potential influence of upper mantle and/or crustal anisotropy beneath a seismic station. Several shear wave splitting techniques are commonly used that attempt resolve seismic anisotropy in deep mantle by considering the presence of multiple anisotropic layers along a raypath. Examples include source-side S wave splitting, which is used to characterize anisotropy in the deep upper mantle and mantle transition zone beneath subduction zones, and differential S-ScS and differential SKS-SKKS splitting, which are used to study anisotropy in the DD” layer at the base of the mantle. Each of these methods has a series of assumptions built into them that allow for the consideration of multiple regions of anisotropy. In this work, we systematically assess the accuracy of these assumptions. To do this, we conduct global wavefield modelling using the spectral element solver AxiSEM3D. We compute synthetic seismograms for Earth models that include seismic anisotropy at the periods relevant for shear-wave splitting measurements (down to ∼5s). We apply shear wave splitting algorithms to our synthetic seismograms and analyze whether the assumptions that underpin common measurement techniques are adequate, and whether these techniques can correctly resolve the anisotropy incorporated in our models. Our simulations reveal some inaccuracies and limitations of reliability in various methods. Specifically, explicit corrections for upper mantle anisotropy, which are often used in source-side direct S splitting and S-ScS differential splitting, are typically reliable for the fast polarization direction φ but not always for the time lag δt, and their accuracy depends on the details of the upper mantle elastic tensor. We find that several of the assumptions that underpin the S-ScS differential splitting technique are inaccurate under certain conditions, and we suggest modifications to traditional S-ScS differential splitting approaches that lead to improved reliability. We investigate the reliability of differential SKS-SKKS splitting intensity measurements as an indicator for lowermost mantle anisotropy and find that the assumptions built into the splitting intensity formula can break down for strong splitting cases. We suggest some guidelines to ensure the accuracy of SKS-SKKS splitting intensity comparisons that are often used to infer lowermost mantle anisotropy. Finally, we suggest a new strategy to detect lowermost mantle anisotropy which does not rely on explicit upper mantle corrections and use this method to analyze the lowermost mantle beneath east Asia.