Resolving power of surface wave polarization data for higher‐order heterogeneities

The resolving power of polarization data compared with that of phase data is investigated by employing both synthetic and observed data sets, using the linear relationship between the phase velocity perturbation and the phase or polarization anomaly. In order to investigate the intrinsic differences between phase and polarization data, a synthetic test is first undertaken using a white noise model with sufficiently uniform coverage of ray paths. This test shows that polarization data can retrieve higher‐order heterogeneities of degrees up to 20 almost completely, despite damping and smoothing effects, whilst phase data can only retrieve those of degrees lower than 8 with reasonable damping. Next, about 4000 phase and 2500 polarization records are collected for minor‐ and major‐arc Rayleigh waves (R1 and R2) in the frequency range 4–12 mHz. To correct the Rayleigh wave polarization data, the misorientation of each station is estimated from the polarization of long‐period P waves propagating mainly in the lower mantle. The phase and polarization data are then inverted for the phase velocity distribution in spherical harmonics with degrees up to 15. The phase velocity maps derived from the phase data are quite consistent with previous studies, whilst those from the polarization data show some discrepancies. For example, the correlation between the phase and polarization models is quite good for low even degrees such as 2, 4 and 6, but not for low odd degrees or degrees higher than 8. The gradients of ampli‐tude spectra from the polarization data are smaller than those from the phase data, especially at degrees higher than 6, which suggests a slightly higher sensitivity of the polarization data to higher‐order heterogeneities. Nevertheless, the overall spectral characteristics of both models are similar; that is, low‐order heterogeneities are dominant whilst higher orders are clearly reduced. Further investigation using a synthetic test with the same uneven paths as the observed data shows the suppression of higher‐order heterogeneities. Since the synthetic test with even paths retrieves higher‐order heterogeneities sufficiently, this result strongly suggests the path averaging effect of uneven ray paths that is intrinsic in the ray theoretical approach used in this study as well as almost all the global inversions. Although inversions based on geometrical ray theory have some limited resolving power with the current status of global records, polarization data are indeed helpful in resolving higher‐order lateral heterogeneities with the dense and uniform path coverage that is becoming available.