Superdomes and Plumes
|Location||International Geological Congress,oslo 2008|
|Author||Helmberger, Donald; Gurnis, Mike|
|Holding Date||21 September 2008|
Current tomographic models of the Earth display perturbations to a radial stratified reference model. However, if these are chemically dense structures with low Rayleigh numbers, they can develop enormous relief, perhaps with boundaries closer to vertical than radial. Several new methods have been developed to simulate 3D synthetics for such structures that involve both analytical and numerical techniques. The method we use approximates 3D effects by adding out-of-plane contributions from virtual receivers at neighboring azimuths with two related to the inner Fresnel zone and two longer-period contributors sampling the outer Fresnel zone. The four responses are scaled by diffraction operators that are defined by the source duration and travel time from the sharp edge structures. Here, we develop a new tool for processing array data based on such a decomposition referred to as a multi-path detector which can be used to distinguish between horizontal structure (in-plane multi-pathing) vs. vertical (out-of-plane multi-pathing) directly from processing array waveforms. We demonstrate the usefulness of this approach by processing samples of both P and S data from the Kaapvaal Array in Southern Africa. The strength difference between patterns produced by S-waves vs. P-waves for the same event (same ray paths) further validates the nature of these chemical boundaries. A detailed SKS wavefield is assembled for a strip along the southern boundary by combining multiple events from the Kaapvaal Array. Applying this technique to this composite data set, we locate a prominent ultralow velocity zone at the edge of a 1000 km high jagged wall. We present evidence for a narrow plume with a diameter less than 100 km extending upward another 500 km located near the center of the superdome structure, in agreement with recent tomographic images and thermo-chemical convection modeling.