The imprint of crustal density heterogeneities on seismic wave propagation

Abstract:

We present the results of a set of numerical experiments designed to observe the imprint of three-dimensional density heterogeneities on a seismogram. To compute the full seismic wavefield in a three-dimensional heterogeneous medium, we use numerical wave propagation based on a spectral-element discretization of the seismic wave equation. We consider a 2000 by 1000 km wide and 500 km deep spherical section, with the one-dimensional Earth model PREM, altered so that the crust is 40 km thick and all the parameters in the crust are constant, as a background. Onto the uppermost 40 km of the underlying one-dimensional model we superimpose three-dimensional randomly generated velocity and density heterogeneities of various correlation lengths. We use different random realizations of heterogeneity distribution. We compare the synthetic seismograms for three-dimensional velocity and density structure with three-dimensional velocity structure and one-dimensional density kept as PREM, calculating relative amplitude differences and time shifts as functions of time and frequency. The misfits in time shift and amplitude for different frequency bands, epicentral distances and medium complexities are then stacked into histograms and statistically analysed. We observe strong dependency on frequency of density-related amplitude difference. We also conclude potential sensitivity to distant density structures, and that scattering is essential to observe significant density imprint on a seismogram. The possible density-related bias in velocity and attenuation for regional tomographic models is calculated using mean misfit values for given epicentral distances. Whereas the bias in velocity does not exceed 0.5% of the model value, the density-related change in attenuation may be as big as 71% of the model value for the mean amplitude difference in the highest frequency band. The results suggest that density imprint on a seismogram is not negligible and with further theoretical developments direct inversion for density may become feasible.