A worldwide comparison of predicted S-wave delays from a three-dimensional upper mantle model with P-wave station corrections

This paper is a comparison, on a worldwide scale, between P-station corrections deduced from ISC residuals (Dziewonski and Anderson) and synthetic S-station corrections computed using a three-dimensional upper mantle model obtained from mantle wave data (Woodhouse and Dziewonski). The upper mantle S-velocity model is described by a spherical harmonic expansion up to degree 8; the P-station corrections are smoothed using a similar expansion, in order that the two data sets can be compared.Correlations between P-station corrections, δt_P, and synthetic S-station corrections, δt_S relevant to various depths of integration indicate that a station correction contains information about structures down to at least 670 km. For this depth of integration, the correlation coefficient of the two data sets is 0.59; the slope ‘a’ of the relation δt_S = aδt_P + b, obtained for the worldwide distribution of stations, is in good agreement with results of previous regional studies using direct readings of P and S arrival times (a = 3.61 ± 0.13).An analysis of regional variations of the relation δt_S = aδt_P + b is carried out on the basis of two published global tectonic patterns (Okal; Jodan). Results for oceanic regions are not reliable, due to the lack of data. On continental areas, a significant difference appears between mountains (a = 2.7 ± 0.3) and shields (a = 4.5 ± 0.4 for Okal's pattern, a = 5.7 ± 1.5 for Jordan's pattern). The largest a-value for shields rules out an explanation by partial melting, as proposed in previous studies. Thermal heterogeneities lead to low a-values; undulations of the lithosphere-asthenosphere boundary appear to be the most feasible explanation of the high slope beneath shields; they are also able to explain the range of variation of the station corrections; the lowest values of the station corrections correspond to a total vanishing of the low velocity zone beneath the oldest shields. For mountains, the mean values of the station corrections as well as the low a-value can be accounted for by a slight increase of Poisson's ratio together with a significant density increase.