Observations consistent with an anisotropic upper mantle are seen in
residual histograms of the transverse- and the radial-component S and SS
autopicks at ranges near the upper-mantle triplications
(Figure 9). On
average, the times are offset with SH arriving slightly ahead of SV.
Two properties of the observed travel-time discrepancy suggest that it arises
from upper-mantle anisotropy--the offset is observed only for shallow
turning rays and on average the offset is 3.6 s larger for the SS phase
than for the S phase. In transversely isotropic material, the SH and
SV velocities approach each other as the rays become vertical; thus shallow
anisotropy is hard to detect at long ranges since the rays are steeply dipping
through the anisotropy. At closer ranges (e.g., 17
to 23 for S and
37 to 43 for SS), the rays travel more horizontally through the
anisotropy and are affected more strongly. For example, the difference in
travel-time between SV and SH waves (calculated from PREM [
Dziewonski and Anderson, 1981]) is as large as 5 s for ranges between
15 and 20\; however, at 30 the difference decreases to about 1
s. The SH-SV travel-time difference is greater for the SS phase
than it is for the S phase since SS turns twice in the upper mantle.
SH versus SV travel-time differences were previously noted by Gee and Jordan [1988] in multiple S phases for Eurasian paths, and are seen in the automatic-gain-control waveform stacks of Shearer [1991b]. Transverse isotropy in the upper mantle was originally proposed to explain differences in Love and Rayleigh wave observations [e.g., McEvilly, 1964; Anderson, 1966; Forsyth, 1975]. In addition, body wave studies indicate azimuthal upper-mantle anisotropy, both from travel-time observations [e.g., Hess, 1964; Raitt et al., 1969; Shearer and Orcutt, 1986] and shear-wave splitting observations [e.g., Ando et al., 1983; Vinnik, 1984; Silver and Chan, 1988]. Note that even if upper-mantle anisotropy is locally azimuthal, global averages will still appear transversely isotropic [e.g., Regan and Anderson, 1984].
