Text S1. Resolution tests for isotropic and anisotropic models

Two checkerboard test models (Basu, 2019) were inverted at each period with varying checkerboard sizes of 1.0, 1.5, and 2.0. The first model (isotropic model) was perturbed by ±4% from the reference velocity and no anisotropy variations. The second test model (anisotropic model) has no velocity perturbations but contains anisotropy with a magnitude of 0.04 km/s in alternating NS-EW directions. The input a priori data and model errors are the same as those used for the real data. Random noise with the same magnitude as\(\sigma_{d}\) (data error at each period) was added to the synthetic models. The tradeoff between azimuthal anisotropy and heterogeneity was minimal.
A final checkerboard resolution test was conducted with a checkerboard test model containing ±4% velocity perturbations from the reference velocity and anisotropy with a magnitude of 0.04 km/s in alternating NS-EW directions. Resolution at each period is dependent on the ray-path coverage. For periods of 20-60s, the best resolution was obtained with a 1.5×1.5 checkerboard size, and 2.0×2.0 checkerboards give the best solution for the higher periods of 75-100s. Figure S1 shows the inversion results for the simultaneous phase velocity and anisotropy test model for all periods for the chosen checkerboard sizes. Once again, this result indicates that the tradeoff between azimuthal anisotropy and phase velocity is minimal.