LATERALLY VARYING REFLECTOR AT THE TOP OF D'' BENEATH NORTHERN SIBERIA

We analyse P-wave data from the French 'Laboratoire de Detection et Geophysique' (L.D.G) network of digital short-period seismic stations. Whereas our previous paper (Houard & Nataf 1992) showed a few promising Kuril events and depicted the methods we proposed to apply, an extensive analysis of 32 earthquakes from the Kuril/Kamchatka region has now been carried out. By using the deconvolution technique previously described (Houard & Nataf 1992), clear and quantitative evidence of intermediate arrivals between the P and PcP waves is shown, in the 75-82-degrees distance range. Our approach is complementary to previous studies in the same region, which used broad-band data recorded at a dense narrow-aperture array (Weber & Davis 1990; Weber 1993) or long-period data recorded at the WWSSN stations (Gaherty & Lay 1992). For each station, record sections of deconvolved signals from different earthquakes are built, and lead us to eliminate crustal reverberations beneath L.D.G stations as a major explanation for these secondary arrivals. A compilation of the most conclusive signals has been made. It is perhaps one of the most complete PdP-wave record sections, since the move out with distance can be followed clearly with a good sampling over more than 5-degrees. The move out with distance of this extra 'PdP' arrival, its residual slowness value, eliminates source complexities, diffraction, or multipathing through the Kuril subduction slab as possible explanations. The extra arrivals indicate the existence of a lower mantle reflector. The structure of the 'Lay discontinuity' is investigated, using T(PdP)-T(P) residual traveltimes. Two different residual time versus distance tendencies are found, which correlate to different geographical bounce point regions, indicating the existence of lateral variations of the structure. The D'' region is also sampled further south at distances well beyond the theoretical triplication cross-over by analysing events from the Honshu/Ryukyu Islands region, but only one P-wave branch is observed. The hodochrones and p-DELTA curves of the first P arrival are used as a complementary analysis. The global PREM model (Dziewonski & Anderson 1981), predicts correct P wave slownesses, though its smooth lower mantle structure does not account for PdP observations. This may be an indication that the 'Lay discontinuity' is a global feature. However, the hodochrones of the first P arrival do not allow us to test the existence of the discontinuity in the 82-85-degrees cross-over distance range.