We present a new tool for efficient incoherent noise reduction for array data employing complex trace analysis. An amplitude-unbiased coherency measure is designed based on the instantaneous phase, which is used to weight the samples of an ordinary, linear stack. The result is called the phase-weighted stack (PWS) and is cleaned from incoherent noise. PWS thus permits detection of weak but coherent arrivals. The method presented can easily be extended to phase-weighted cross-correlations or be applied in the z-p domain. We illustrate and discuss the advantages and disadvantages of PWS in comparison with other coherency measures and present examples. We further show that our non-linear stacking technique enables us to detect a weak lowermantle P -t o 4 conversion from a depth of approximately 840 km on array data. Hints of an 840 km discontinuity have been reported; however, such a discontinuity is not yet established due to the lack of further evidence.

[1] Regional seismic tomography provides valuable information on the structure of shields, thereby gaining insight to the formation and stabilization of old continents. Fennoscandia (known as the Baltic Shield for its exposed part) is a composite shield for which the last recorded tectonic event is the intrusion of the Rapakivi granitoids around 1.6 Ga. A seismic experiment carried out as part of the European project Svecofennian-Karelia-Lapland-Kola (SVEKALAPKO) was designed to study the upper mantle of the Finnish part of the Baltic Shield, especially the boundary between Archean and Proterozoic domains. We invert the fundamental mode Rayleigh waves to obtain a three-dimensional shear wave velocity model using a ray-based method accounting for the curvature of wave fronts. The experiment geometry allows an evaluation of lateral variations in velocities down to 150 km depth. The obtained model exhibits variations of up to ±3% in S wave velocities. As the thermal variations beneath Finland are very small, these lateral variations must be caused by different rock compositions. The lithospheres beneath the Archean and Proterozoic domains are not noticeably different in the S wave velocity maps. A classification of the velocity profiles with depth yields four main families and five intermediate regions that can be correlated with surface features. The comparison of these profiles with composition-based shear wave velocities implies both lateral and vertical variations of the mineralogy.

The fine structure of the upper mantle discontinuities is investigated using observations of converted waves on short‐period and broadband seismograms. Using a stacking technique to analyze the P wave coda of teleseismic records, evidence is found for coherent near‐receiver P‐to‐S converted phases generated by the 400‐ and 670‐km discontinuity beneath a number of the stations used in this study. Variations in travel time, slowness, and amplitude of these phases as observed among the stacks for stations of the Regional Seismic Test Network in the United States and of the Network of Autonomously Recording Seismographs in western Europe are very likely the expression of upper mantle heterogeneity. Observations of coherent converted phases from the 400‐km discontinuity are fewer in number than of phases converted at the 670‐km discontinuity, suggesting that the latter is more pronounced. Some of the seismograms, especially of station RSCP, show extremely high‐amplitude P‐to‐S converted phases from the 670‐km discontinuity. These seismograms allow a detailed waveform comparison of the converted phase with the direct P phase and present evidence for a sharp 670‐km discontinuity.

[1] We measured interstation fundamental mode Rayleigh wave phase velocities using data from the NARS-Baja seismic network located around the Gulf of California. A region-average, shear velocity model and a set of azimuthally anisotropic phase velocity maps are obtained from these data. The average shear velocity structure shows a strong low-velocity zone underlying a thin lid and the data are suggestive of low velocities down into the transition zone. The phase velocity maps display signatures of sedimentary layers, crustal thickness variations, upwelling under the plate boundary, and the presence of the subducted Farallon microplate remnants beneath the Gulf. The upper mantle features inferred from this study provide new seismic evidence on the tectonic evolution of the region.

New broadband seismic data from Botswana and South Africa have been combined with existing data from the region to develop improved P and S wave velocity models for investigating the upper mantle structure of southern Africa. Higher craton‐like velocities are imaged beneath the Rehoboth Province and parts of the northern Okwa Terrane and the Magondi Belt, indicating that the northern edge of the greater Kalahari Craton lithosphere lies along the northern boundary of these terranes. Lower off‐craton velocities are imaged beneath the Damara‐Ghanzi‐Chobe Belt, and may in part reflect thinning of the lithosphere beneath the incipient Okavango Rift. Lower velocities are also imaged to the north and northwest of the Bushveld Complex beneath parts of the Okwa Terrane, Magondi Belt, and Limpopo Belt, indicating that cratonic upper mantle in some areas beneath these terranes may have been modified by the 2.05‐Ga Bushveld and/or 1.1‐Ga Umkondo magmatic events.

S U M M A RYAlthough the Earth's inner core has long been thought to be solid, there have not, as yet, been unequivocal observations of inner core shear waves. Here we present observations of the phases pPKJKP and SKJKP for the Flores Sea event of 1996 June 17. The observations support the value of approximately 3.6 km s {1 for the inner core shear wave velocity and open up new possibilities for exploring the anisotropic structure and the attenuation properties of the inner core. The analysis and validation of the observation is based on a new method that can be used in the search for and identi¢cation of inner core shear waves. In this method normal mode synthetics for a solid inner core are compared with those for a £uid inner core. It is only by such comparisons that it is possible to be certain that inner core shear waves, rather than interferences of other phases, are responsible for observed energy maxima in stacked seismic records.

We present the first nationwide crustal thickness and Vp/Vs maps for Botswana based on the analysis of new P wave receiver functions of NARS‐Botswana network integrated with previous receiver function results in Botswana. Using H‐K analysis, we found crustal thickness values ranging from 34 km for the Okavango Rift Zone to 49 km at the border between the Magondi Belt and the Zimbabwe Craton. For stations with significant sediment cover, a sediment correction was applied based on sequential H‐K stacking. We observed distinct differences for the Kaapvaal craton. The eastern part has a high Vp/Vs ratio typical of a predominantly mafic composition, suggesting lateral extension of the Bushveld mafic complex. On the other hand, the western part with a Vp/Vs ratio of 1.67 is felsic, probably as a result of delamination caused by Proterozoic rifting processes. Further to the west of the Kaapvaal Craton, we found a crustal thickness of 42 km and a Vp/Vs ratio of 1.76 for the Nosop Basin. These values are similar to other cratonic regions in Botswana, suggesting the presence of a buried craton as proposed by previous studies. We confirm a relatively thin crust (∼34–39 km), compared to the rest of Botswana, and high Vp/Vs ratio (∼1.84) underneath the Okavango Rift Zone found by previous receiver function studies. Notably, we also found a relatively thin crust (37 km) and high Vp/Vs ratio (1.84) in central Botswana underneath the Passarge Basin.

Splitting of SKS waves caused by anisotropy may be analyzed by measuring the splitting intensity, i.e., the amplitude of the transverse signal relative to the radial signal in the SKS time window. This quantity is simply related to structural parameters. Extending the widely used cross-correlation method for measuring travel-time anomalies to anisotropic problems, we propose to measure the SKSsplitting intensity by a robust cross-correlation method that can be automated to build large high-quality datasets. For weak anisotropy, the SKS-splitting intensity is retrieved by cross-correlating the radial signal with the sum of the radial and transverse signals. The cross-correlation method is validated based upon a set of Californian seismograms. We investigate the sensitivity of the SKS-splitting intensity to general anisotropy in the mantle based upon a numerical technique (the adjoint spectralelement method) considering the full physics of wave propagation. The computations reveal a sensitivity remarkably focused on a small number of elastic parameters and on a small region of the upper mantle. These fundamental properties and the practical advantages of the measurement make the cross-correlation SKS-splitting intensity particularly well adapted for finite-frequency imaging of upper-mantle anisotropy.