Common conversion point stacking with beam-formed multitaper receiver functions: Images of the upper mantle beneath Kamchatka and the northern Apennines

In many regions of the world, standard common conversion point stacking of receiver functions computed from time-domain or water-level deconvolution methods is problematic for a variety of reasons. For example, some locales lack sufficient station coverage to produce robust stacks or strong variations in shallow structure make imaging deeper structure difficult. To improve our ability to image discontinuities in the upper mantle under these conditions, we developed a novel stacking technique that stacks "beams" of multitaper receiver functions using a common conversion point framework. In this method, we compute receiver functions for beams formed by grouping observed earthquakes on the basis of their location. Each beam is defined by a center and varies by backazimuth and epicentral distance. Receiver functions from individual sources within a beam are combined using the multitaper spectral correlation algorithm of Park and Levin (2000). A resulting receiver function is associated with a ray path connecting the center of the beam and the station. We use a set of "beams" overlapping in both backazimuth and epicentral distance by 50% (i.e. the distance between beam centers is equal to half of the beam width). Our method resembles that employed by Bostock (1999), but adds spatial overlap of the beams. The resulting beam-formed RFs are subsequently converted from time to depth using to associate the amplitudes of the beam-fromed receiver functions with depths of P-S conversion points. We then build a database of P-S conversion points and their corresponding amplitudes beneath the region every 5 km in depth. We determine this information for rays from a grid of sources located every 15° in backazimuth, and every 10° in epicentral distance range 20°-90°. Once all amplitudes in the beam formed receiver functions are associated with a particular point at depth, we are able to produce images of the discontinuity structure in the upper mantle. We present preliminary images of the upper mantle computed using this technique beneath two regions: Central Kamchatka and the Northern Apennines, where we were successfully able to image complex structures in the subsurface ~ 20 km in size.