Variability of P660s phases as a consequence of topography of the 660 km discontinuity

Records from the broad-band NARS-NL array (deployed in the Netherlands, western Germany and Belgium) have been analyzed for P to S converted phases from the 660 km discontinuity (P660s). No individual seismogram from this array shows a clear P660s phase. Stacks of (up to 113) crosscorrelograms show P660s energy that is weak in comparison with similar stacks elsewhere. Very high amplitude observations of P660s in single seismograms from the same region have been made in a previous study (Paulssen, 1985, Geophys. Res. Lett., 12: 709-712; Paulssen, 1988, J. Geophys. Res., 93: 10489-10500). Such variations may be due to relief on the 660 km discontinuity. In this paper we quantify the effects of different topographies on the time delays, amplitudes, and waveforms of the direct P wave as well as on the P660s converted wave. Synthetic P and P660s waveforms are calculated by applying a Kirchhoff-Helmholtz integration over the wave fields at the 660 km discontinuity interface. For topography on scales of a few hundreds of kilometers, the effects on the P wave are very small compared with those on the P660s phases. Time delays, focused and defocused amplitudes, and waveform distortion of P660s phases influenced by such topography diminish the coherence of P660s in (synthetic) stacks. Our results show that topography of the 660 km discontinuity with dominant scale lengths of a few hundreds of kilometers and variations in depth of ±15 km to ±25 km, is consistent with the pattern of P660s observations in single seismograms and stacks.