Experiments in migration processing of SS precursor data to image upper mantle discontinuity structure

Long-period SS precursors result from underside reflections off upper mantle discontinuities. By grouping and stacking global seismic data by SS bounce point location it is possible to map lateral variations in depths to the 410- and 660-km discontinuities, a process analogous to common midpoint (CMP) stacking in reflection seismology. Because this method assumes horizontal reflectors, energy arriving from dipping or intermittent reflectors may not be correctly imaged. To address this possibility, we experiment with techniques based on migration processing of shallow seismic reflection data. The problem is complicated by the uneven distribution of sources and receivers for the SS precursor observations, but the data are sufficiently dense beneath the northwest Pacific Ocean that reasonably good coverage can be obtained for this region. We parameterize the model as a grid of point scatterers in latitude, longitude, and depth (from the surface to 1000 km depth) and compute travel times from each grid point to the source and receiver locations. These times are used to construct a matrix equation that yields predicted SS precursor waveforms from the assumed scatterers. To recover the model, we experiment with both simple back projection and full inversions using a conjugate gradient method. Tests on noise-free synthetic data (generated using the same source-receiver distribution as the actual data) suggest that detailed resolution of discontinuity structure is possible, at horizontal scales much smaller than the Fresnel zone. However, the real data do not produce coherent results unless some degree of horizontal smoothing is imposed, at least partially defeating the purpose of this approach. Results for the northwest Pacific find structure on the 410- and 660-km discontinuities and hints of intermittent reflectors at other depths. Random resampling tests, however, suggest that most of these features are not reliably resolved, with the exception of a depression on the 660-km discontinuity seen in the northwest Pacific. Our experiments show that it is unlikely that small-scale structure on the 660-km discontinuity near subducting slabs causes significant bias in maps of the large-scale 660-km topography derived from long-period SS precursor observations.