Slab and Plume Morphology in the Transition Zone and Below: a Comparison of Images From Recent P and S Velocity Models

Wednesday, 17 December 2014
Laura M Salmi1, Scott W French1 and Barbara A Romanowicz1,2, (1)University of California Berkeley, Berkeley, CA, United States, (2)Institut de Physique du Globe, Paris, France
Resolving subduction zones in the shallow upper mantle using global shear velocity tomography has long been a challenge, likely due to the rather narrow signature of the slabs down to ~400 km depth compared to the wavelength of fundamental mode and overtone surface waves, on which resolution of Vs at these depths often relies. On the other hand, models based on P wave travel times exhibit higher resolution in subduction zone regions, owing to both the higher frequencies of the P waves as well as an optimal illumination geometry. Conversely, the global Vs models typically have better resolution near the CMB, because of constraints provided by Sdiff and multiple ScS phases. Here we compare the morphology of subducted slabs throughout the mantle, as imaged by both a recent Vp model (GAP_P4, Fukao and Obayashi, 2013) and a new Vs model (SEMUCB-WM1, French and Romanowicz, GJI, in revision). The latter model was developed by inverting body (to 32s) and fundamental and overtone surface (to 60s) waveforms, with the forward seismic wavefield computed using the spectral element method. While the S velocity model is still "fuzzier" than the Vp model, it tracks the behavior of slabs trapped in the transition zone, and those ponding around 1000 km depth. We quantify the high correlation of the region of fast Vp and Vs anomalies, and thus derive a robust estimate of the R=dlnVs/dlnVp ratio as a function of depth in regions of faster than average velocity. We compare these results with estimates obtained with other combinations of available P and S models, as well as theoretical values from mineral physical calculations. Estimating R in slow velocity regions is more difficult, as resolution varies more among models. Here we compare slow velocity images in SEMUCB-WM1 with those of other recent Vs and Vp models and attempt to estimate R in those regions as well. Interestingly, we note that, in the SEMUCB-WM1 model, some of the columnar, lower than average velocity regions "rising" from the CMB through the lower mantle appear to be deflected horizontally at ~1000 km depth. This observation suggests that whatever mechanism causes the resistance to downward flow in subduction zones at this depth may also affect upwellings.