ScS wave splitting of deep earthquakes around Japan

ScS wave splitting of five deep earthquakes in subduction zones near Japan is investigated using horizontal seismograms recorded at JMA stations. For each earthquake, we clearly observe uniform ScS wave splitting in all stations over Japan, especially for the events located south of Honshu in 1982, 1984 and 1993. However, the directions of fast-polarized waves of these events differed by a maximum of about 50 degrees from one another. The orientation of fast-polarized waves in the 1982 event was NNW-SSE; those in the two later events WNW-ESE. We also recognize this discrepancy in the results of the analysis of the 1971 Sea of Okhotsk event reported by Fukao (1984). The Sakhalin Islands event in 1990 reveals a linear particle motion without such a change in direction of the second arrivals, implying no anisotropy. These observations are interpreted as indicating an anisotropic region within the slab near the earthquake sources but not beneath the receivers, since the orientations of fast-polarized waves recorded at each station are not common to all the earthquakes. Furthermore, we consider that anisotropy exists non-uniformly within the slab. The event in 1982, which occurred in almost the same area as those in 1984 and 1993, showed a fast direction different from the events in 1984 and 1993. The 1982 event was 179 km deep, but the two later events were at 398 km and 360 km, respectively. The fast direction observed from the 1982 event is parallel to the fossil plate motion, whereas those from the events in 1984 and 1993 are parallel to the compression axis within the subducting slab. The depth of 400 km is a phase boundary, where olivine changes to beta spinel. We consider that the most likely cause of the change in anisotropy direction is the re-orientation of crystals associated with the phase change of olivine to beta spinel due to subduction of the slab.