<p>A shallow aftershock sequence in the Hawkes Bay region of the North Island, New Zealand (May 1990) was recorded with high quality on an L-shaped, 7-station array of 3-component, short-period seismographs at Wellington, such that the seismic waves travelled almost along strike of the subducted Pacific plate in this region. The arrival times at the stations of the Pn wave pulse from a number of aftershocks could be picked sufficiently accurately for a least-squares inversion to be carried out for wavefront speed, c, and incident azimuth, [theta]. The results show a high apparent velocity, 8.7 [plus or minus] 0.2 km/s, and an azimuth which is shifted by 6.0 [plus or minus] 2.5 degrees east of the true epicentre - station azimuth. The azimuthal anomaly, [delta][theta], has been interpreted as due to lateral refraction of Pn off the subducted slab. The effect of different geometries of the slab on the Pn wavefront characteristics (c and [delta][theta]) at Wellington have been explored through both simple geometrical considerations (in the case of a plane or cylindrical slab) as well as through 3-dimensional ray tracing (in the case of irregular curvature of the slab). It has been shown that a plane or cylindrical slab would require P-wave velocities of about 9.0 km/s to exist within it in order to fit both c and [delta][theta], whereas a model of the slab which departs from a regular cylinder and has a small updip component along strike can fit the observations with P-wave velocities of 8.75 km/s in the high velocity medium. This model has been proposed by Ansell and Bannister (1991) after detailed consideration of the shallow seismicity that defines the slab surface in the lower North Island. Information about the nature of the high velocity medium has been obtained by modeling the waveforms through generation of synthetic seismograms by the reflectivity technique of Kennett (1983). The large number of aftershocks within a small source region, and the sampling of much the same wavepath, meant that a sufficient number of seismograms had very similar and characteristic features that could be modelled. The typical seismogram of the data set had a simple Pn wavepulse, followed immediately by a complex. high frequency (up to 15 Hz) phase (here referred to as Phf) and a high amplitude, lower frequency phase that dominated the P-wavetrain (here referred to as P). A velocity profile that contained a layer of 8.75 km/s material at least 4 km thick, underlying "normal" mantle material of P-velocity 8.2 [plus or minus] 0.2 km/s. and whose surface lies approximately 18 km below the slab surface reproduced the observed seismogram features well. The presence of velocity gradients above and below the layer is not excluded. A gradual decrease in velocity below the layer in fact gives a better fit of the Pn pulse shape. By breaking down the synthetic seismogram into simpler versions. using Kennett's wavefield approximation technique, it has been shown that the Pn wave propagates through the high velocity layer, the Phf phase through the overlying layers as a sequence of reflections and refractions, and the P group as a reverberatory phase in a crustal waveguide, with its energy mostly in the form of free surface reflections and S to P conversion. These results have also been confirmed by ray tracing. Waveform modelling has also clearly shown that a low velocity layer (representing subducted sediment) on the top of the subducted slab produces a highly characteristic imprint on the synthetic seismogram, in the form of an energetic, reverberatory, lower frequency signal late in the P-wavetrain. Wavefield approximations show that this is also a crustal waveguide effect, with a strong component of mode conversion at the free surface, but P - S conversion appears to be the dominant mechanism. Seismograms very similar to such synthetic ones have been observed for the Weber aftershocks recorded at stations along the northern East Coast. The presence of such a low velocity layer in the East Coast region is thus implied, consistently with previous proposals. The petrological implications of the high velocity layer in the subducted Pacific plate are discussed. The most likely explanation is that it represents the maximum P velocity of an anisotropic layer within the Pacific upper mantle. It is proposed that the conditions of stress orientation, pressure and temperature at approximately 36 - 50 km depth in this region induces a strong realignment of olivine crystals with their fast direction along strike of the slab, normal to the maximum compressive stress axis. The upper mantle of the segment of the Pacific ocean just east of the Tonga - Kermadec trench and the North Island has been shown in this study to possess P-wave anisotropy, with the P-velocity reaching a maximum of 8.37 km/s in a direction N60 degrees E. This result was obtained by analysing a large set of ISC travel times from earthquakes along the Tonga - Kermadec - New Zealand subduction zone recorded at stations Niue, Rarotonga and the Chatham Islands. It is suggested that an enhancement of this anisotropy, accompanied by some re-orientation, takes place as the upper mantle medium is subjected to the new stress conditions in the initial stages of subduction.</p>
A sporadic low-velocity layer atop the 410 km discontinuity beneath the Pacific Ocean