SUMMARY
Sp receiver functions have been widely used to detect the lithosphere–asthenosphere boundary (LAB) and other mantle discontinuities. However, traditional common conversion point (CCP) stacking can be biased by the assumption of horizontal layers and this method typically underestimates scattering amplitudes from velocity boundaries with significant dips. A new pre-stack migration method based on recently developed Sp scattering kernels offers an alternative that more accurately captures the timing and amplitude of scattering. When calculating kernels, Sp-S times are estimated with the fast-marching method, and scattering amplitude versus direction, geometrical spreading and phase shifts are accounted for. To minimize imaging artefacts with larger station spacing, Sp receiver functions are interpolated to more closely spaced pseudo-stations using either compressive sampling or spatial averaging algorithms. To test the kernel-based stacking method, synthetic Sp phases were predicted using SPECFEM2D for velocity models with a flat Moho and a negative mantle velocity gradient with a ramp structure. The kernel-based stacking method resolves horizontal interfaces equally well as CCP stacking and outperforms CCP stacking when imaging boundaries with dips of more than 8°, although dip resolution is still limited. Use of more vertically incident phases such as SKSp improves retrieval of dipping discontinuity segments. A second approach is to down-weight the portions of the kernels that have the greatest positive interference among neighbouring stations, thus enhancing scattering from dipping structures where positive interference is lower. With this downweighting, the kernel-based stacking method applied to Sp data is able to continuously resolve LAB discontinuities with dips up to 15° and to partially resolve continuous LAB discontinuities with dips of ∼20°. The intrinsic properties of teleseismic Sp phase kernels limit their ability to resolve LAB structures with dips of ∼20–35°, but still larger dips of ∼40–50° are resolvable with dense and appropriately placed stations. Analysis of Sp scattering kernels also explains the effectiveness of CCP stacking for quasi-horizontal interfaces.
Crustal and mantle velocity models of southern Tibet from finite frequency tomography
