Finite‐frequency effects in global surface‐wave tomography

SUMMARY We compare traditional ray‐theoretical surface‐wave tomography with finite‐frequency tomography, using 3‐D Born sensitivity kernels for long‐period, fundamental‐mode dispersion measurements. The 3‐D kernels preserve sidelobes beyond the first Fresnel zone, and fully account for the directional dependence of surface‐wave scattering, and the effects of time‐domain tapering and seismic source radiation. Tomographic inversions of Love and Rayleigh phase‐delay measurements and synthetic checkerboard tests show that (1) small‐scale S ‐wave velocity anomalies are better resolved using finite‐frequency sensitivity kernels, especially in the lowermost upper mantle; (2) the resulting upper‐mantle heterogeneities are generally stronger in amplitude than those recovered using ray theory and (3) finite‐frequency tomographic models fit long‐period dispersion data better than ray‐theoretical models of comparable roughness. We also examine the reliability of 2‐D, phase‐velocity sensitivity kernels in global surface‐wave tomography, and show that phase‐velocity kernels based upon a forward‐scattering approximation or previously adopted geometrical simplifications do not reliably account for finite‐frequency wave‐propagation effects. 3‐D sensitivity kernels with full consideration of directional‐dependent seismic scattering are the preferred method of inverting long‐period dispersion data. Finally, we derive 2‐D boundary sensitivity kernels for lateral variations in crustal thickness, and show that finite‐frequency crustal effects are not negligible in long‐period surface‐wave dispersion studies, especially for paths along continent–ocean boundaries. Unfortunately, we also show that, in global studies, linear perturbation theory is not sufficiently accurate to make reliable crustal corrections, due to the large difference in thickness between oceanic and continental crust.