Banana-doughnut kernels and mantle tomography

Abstract

Theoretical analyses suggest that finite-frequency effects should be considered in tomographic inversions of seismic phase ‘arrival times’ measured with waveform cross correlation at relatively low frequencies. This has led to the development of so-called ‘banana-doughnut [sensitivity] kernels’—hereinafter BDKs. Here we address a practical question: has the use of these kernels produced tomographic images of global mantle heterogeneity that are significantly better than those based on ray theory? A simple model comparison suggests that the answer is ‘not yet’. The effect of BDKs on both the pattern and the amplitude of mantle wavespeed perturbations appears to be smaller than that of practical (and subjective) considerations (such as the level of damping, the weighting of different data sets, and the choice of data fit) and does not exceed realistic estimates of image uncertainty due to, for instance, errors in the data. By itself, the fact that the better theory has not yet resulted in significant model improvements does not imply that models based on BDKs are incorrect. Deep ‘plumes’ may very well exist, but the benificial effect of BDKs on the tomographic images (and on ‘plume’ identification in particular) has been overstated.