Robustness of Global Radial Anisotropy Models of the Upper Mantle
Abstract:Radial anisotropy provides important constraints on mantle deformation. While its presence is well accepted in the uppermost mantle, large discrepancies remain among existing models, even at depths well sampled by seismic data, and its presence at greater depths is highly uncertain. Surface wave phase velocity dispersion measurements are routinely used to constrain lateral variations in mantle S-wave velocity (dlnVS) and radial anisotropy (ξ=VSH2/VSV2). Here, we employed the fundamental and higher mode surface wave phase velocity maps of Visser et al. (2008) that have unprecedented sensitivity to structure down to 800-1000km depth, and we adopted a probabilistic forward modeling approach, the Neighbourhood Algorithm, to quantify posterior model uncertainties and parameter trade-offs.
We investigated the effect of prior crustal corrections on 3-D ξ and dlnVS models. To avoid mapping crustal structure onto mantle heterogeneities, it is indeed important to accurately account for 3-D crustal anomalies and variations in Moho depth. One approach is to solve the non-linear problem and simultaneously constrain Moho depth and mantle anomalies (Visser et al., 2008). Another approach, taken here, is to calculate non-linear crustal corrections with an a priori crustal model, which are then applied to the phase velocity maps before inverting the remaining signal for mantle structure. In this work, we also determined laterally varying sensitivity kernels to account for lateral changes in the crust. We compare models obtained using CRUST2.0 (Bassin et al., 2000) and the new CRUST1.0 (Laske et al., 2012) models, which mostly differ under continents. Our preliminary results show strong differences (ΔdlnVS>2%) between the two models in continental dlnVS for the upper 150-200km, and strong changes in x amplitudes in the top 200km (Δξ>2%). Some of the differences in ξ persist down to the transition zone, in particular beneath central Asia and South America. Despite these discrepancies, inferences on the depth of continental roots (~200-250km) based on either the extent of the dlnVS>0 anomalies or the depth at which ξ changes sign remain independent of the crustal model employed. We also note that VSV>VSH dominates the deep upper mantle except in central Pacific, which is characterized by VSH>VSV down to the transition zone.