Assessment of global phase velocity models

SUMMARY We construct new Love and Rayleigh wave phase velocity models based on measurements made from an aspherical starting model and strict data quality control derived from cluster analyses. These new models are in good agreement with previous ones and the question arises whether the slight changes show an improved capacity to explain the data. To this effect, we propose an objective method to compare different phase velocity models published in recent literature. The method is based on comparing calculated synthetics to raw seismograms. We find a reassuring convergence, between all the models we tested, at the longest periods and more scatter at the shorter periods. At 40 s, the different models show gains of up to 3.5 cycles over PREM. Generally, the higher the gain over PREM, the smaller the period considered, which confirms that the Earth's heterogeneity is strongest in the uppermost parts of the Earth. Apart from assessing different models against each other, our method gives an estimate, comparable to cluster analyses, of the underlying data errors that went into the construction of the models themselves. Moreover, ray coverage is still far from perfect for constructing phase velocity models. As a result, we find that without precaution, degree zero is biased through spectral leakage by 0.1 to 0.2 per cent with respect to PREM.