Amplitude of the core-mantle boundary topography estimated by stochastic analysis of core phases

The core-mantle boundary (CMB) topography is an important parameter for con straining the models of mantle dynamics and core-mantle coupling. However, the various large wavelength seismological models of the CMB topography, wh ich have been obtained up to now, are poorly correlated, Moreover, their ma ximum amplitudes vary considerably from one model to another, with values r anging from +/-4 to +/-12 km. These large discrepancies may be due to the d ifficulty to separate, in the travel rime anomalies, the contribution of th e CMB topography from that of the highly heterogeneous D" region at the bas e of the mantle. In order to better constrain the amplitude of the CMB topo graphy, we perform a stochastic analysis of the core phases which sample th e CMB as transmitted waves and/or as reflected waves. In particular, we ana lyse underside reflected PKKP phases, which help to discriminate between CM B topography and D" structure, and have in addition a great sensitivity to CMB topography. The other phases used are the upperside reflected waves PcP , and the transmitted waves PKP. The analysis is performed on the travel ti me residual file obtained by Engdahl er al. [Engdahl, E.R., van der Hilst, R,, Buland, R.P., 1998. Global teleseismic earthquake relocation with impro ved travel times and procedures for depth determination. Bull. Seismol. Sec . Am. 88, 722-743.] after earthquake relocation and phase re-identification . After careful travel time data selection, the stochastic analysis allows us to separate coherent signal from random signal in the data at different length scales ranging from 300 to 1500 km. Then the CMB topography variance is obtained at the different length scales from a joint analysis of the co herent signal in the different core phases, in taking into account their di fferent sensitivities to CMB topography. The estimated CMB topography varia nce has a significant signal in the wavelength range that we have investiga ted, showing that 95% of CMB topography amplitude is in the range +/- simil ar to 4.0 km for wavelengths larger than 300 km. This value decreases to +/ - similar to 1.5 km for wavelengths larger than 1200 km, indicating that th e long wavelength CMB topography is significantly lower than previously pro posed, However, it has not been possible to invert the data for deriving a map of the CMB topography. A checkerboard pattern analysis reveals that, de spite the introduction of the PKKP phase, the relative contributions of D" heterogenieties and CMB topography cannot be separated with the presently a vailable data, (C) 2000 Elsevier Science B.V. All rights reserved.