To explore upper mantle heterogeneity beneath the North Atlantic we have measured 70 SS-S differential travel times using the waveform cross-correlation method. Both the two-way preliminary reference earth model (PREM) and J-B (Jeffreys-Bullen) residuals exhibit a maximum variation of 17 s, ranging from-10 to 7 and-9 to 8 s, respectively. The lack of correlation between SS-S and S residuals, along with the strong correlation between SS-S and SS residuals, suggests that SS-S differential travel times eliminate source and receiver effects and sample the upper mantle anomalies near the SS bounce points without being strongly affected by lower mantle heterogeneities. We find evidence for a strong local anomaly in the North Atlantic; residuals for bounce points immediately to the north of the Azores-Gibraltar plate boundary average about 4 s faster than for the bounce points to the south of the boundary. We infer that this velocity contrast exists in the lithosphere and continues below a depth of 200 km. A series of linear multiple-regression experiments demonstrate that variation in velocity with increasing age of the seafloor and azimuthal anisotropy are both significant contributors to the PREM and J-B residuals. Assuming that travel time delays are a linear function of the square root of the seafloor age, we find the coefficient for (age)1,2 to be about-1.0 s/(m.y.)1,2.

This is smaller than but not significantly different from the corresponding slope for the S delays of intraplate earthquakes in the Atlantic reported by J. D. Duschenes and S. C. Solomon (1977) and favors the presence of small amounts of water (e.g., 0.1%) in the upper mantle. In modeling anisotropy we carry out an experiment using synthetic seismograms to examine the shear wave splitting phenomena in the anisotropic medium and investigate three possible forms for SH wave delay as a function of azimuth (&thgr;): sinusoidal variations as a function of 2&thgr;, 4&thgr;, and a combination of 4&thgr; (predominant) and 2&thgr; that mimics the delay expected in olivine crystals when the α axis is aligned horizontally. Our regression models strongly prefer the predominance of a 4&thgr; to a 2&thgr; variation, and the last type of anisotropy explains our data best. From the preferred anisotropy model the olivine α axis, or the upper mantle shearing flow, is inferred to be oriented N-S. The magnitude of the azimuthal variation suggests that anisotropy extends over a depth range of several hundred kilometers in the upper mantle. ¿American Geophysical Union 1987