Appraising the Reliability of Scattered Wave Imaging: Application to the 410 km and 660 km discontinuities

Passive array wavefield imaging methods used to date suffer from a lack of methods to appraise the reliability of the results. In this paper we explore two methods for error appraisal: (1) simulations of the data from simple test models and (2) statistical approaches commonly used in seismic reflection processing. This simulation program first computes synthetic seismograms using the same source and receiver geometry as the data being evaluated. In this paper we apply this technique to data from the USArray imaged by our 3D, prestack, plane wave migration technique. To date we have implemented point scatterer and radially symmetric models that simulate the actual data geometry. The synthetic data are loaded into the imaging program to generate a 3D imaging volume which contains the model reconstruction. Point resolution is estimated by conducting numerical experiments with different point scatters to yield an error appraisal comparable to checkboard tests in tomography. Flat layer synthetics provide a useful parallel test to evaluate a radially symmetric hypothesis. The statistical error appraisal approach we use is appropriate because our approach is a prestack method. We can thus appraise data consistency by the standard signal processing technique of coherence but applied in a 3D volume. We apply these techniques to appraise the results of a new 3D imaging of the Western US that we obtained using waveform data from the Earthscope Automated Receiver Survey. Our results suggest that the 410 km discontinuity has scattering properties more like a rough surface than a mirror.