Assessment of tomographic mantle models using spectral element seismograms

We investigated the agreement between real seismograms and those predicted by long wavelength mantle models by looking at phase and amplitude differences. We computed full synthetic seismograms using a spectral element method together with 3-D mantle models and the appropriate crustal model on top. We selected differently damped mantle models to see the effect of regularization on the computed seismograms. To check the phase agreement, we measured time-shifts between the real and synthetic surface waves and body wave phases such as P, S and SS using a cross-correlation technique. We also compared the amplitudes of the real and synthetic seismograms to understand how well the models explain not only the phases, but the whole waveforms. 3-D mantle models improve the phase agreement of surface waves in particular. The remaining misfit, however, is still so large that we cannot distinguish between different tomographic models. We suggest that this is mainly due to an imperfect modelling of the crust, and/or source location if body waves are included, which have to be addressed in future inversions. Amplitude mismatches are large, regardless which 3-D mantle model is used. We observe that 3-D scattering or focusing/defocusing effects can only explain half of the surface wave amplitudes whereas body wave amplitudes are dominated by scattering effects. 1-D Q models, particularly in the crust and upper-mantle, strongly affect surface wave amplitudes and have to be modelled properly.