Fundamental and higher-mode Rayleigh waves propagating across northern Eurasia and the western Pacific Ocean have been used to construct a lower bound on zmax, the maximum depth extent of seismic velocity differences between old continents and old oceans. The waveform inversion procedure of Lemer-Lam and Jordan has been modified to search for optimum path-averaged elastic models whose significant structural differences are ''squeezed'' above some chosen depth z*. An inversion with z* unconstrained yield two reference models, EU0 for northern Eurasia and PA0 for the western Pacific. The reference models provide good fits to the vertical component wavefrom data out to the fourth overtone, but they are characterized by substantial velocity differences extending into the lower mantle.

Resolving-power calculations suggest that such deep variations are not required by the data. Using EU0 and PA0 as starting models for additional inversions and their fit to the data to define noise levels, we constructed at a pair of acceptable models, EU2 and PA2, whose shear velocity differences are squeezed above z*=400 km and compressional velocity differences squeezed above z*=220 km. However, we were unable to obtain a successful fit to the data with z*=220 km for shear velocity and thus reject the hypothesis zmax≤220 km in favor of the alternative zmax>220 km. The EU2-PA2 differences in shear velocities are generally consistent with other recent seismological models of continent-ocean heterogeneity, including regional models SNA and ATL of Grand and Helmberger and the global model M84C of Woodhouse and Dziewonski. The average upper mantle shear velocities in our P-SV models are substantially lower than Grand and Helmberger's SH models, however, this discrepancy cannot be explained by simple path differences and points to the importance of polarization anisotropy. The high-velocity zones underlying the ancient cratonic nuclei are inferred to be part of the tectosphere, translating coherently with the continents during plate motions. This and other studies of upper mantle structure are more consistent with a thick plate, compositionally stratified model for the cratons than with the conventional thin plate models of continental thermal structure. ¿ American Geophysical Union 1987