Constraints on mantle melt geometries from body wave attenuation in the Salton Trough and Snake River Plain

Melt can be retained in the mantle at triple junctions between grain boundaries, be spread in thin films along two-grain boundaries, or be organized by shear into elongate melt-rich bands. Which of these geometries is most prevalent is unknown. This ambiguity makes the interpretation of anomalous seismic velocities and quality factors difficult, since different geometries would result in different mechanical effects. Here, we compare observations of seismic attenuation beneath the Salton Trough and the Snake River Plain; two regions where the presence of melt has been inferred. The results suggest that seismic attenuation is diagnostic of melt geometry. We measure the relative attenuation of P waves from deep focus earthquakes using a time-domain method. Even though the two regions are underlain by comparably strong low-velocity anomalies, their attenuation signature is very different. The upper mantle beneath the Salton Trough is sufficiently attenuating that the presence of melt must lower Qp, while attenuation beneath the Snake River Plain is not anomalous with respect to surrounding regions. These seemingly contradictory results can be reconciled if different melt geometries characterize each region. SKS splitting from the Salton Trough suggests that melt is organized into melt-rich bands, while this is not the case for the Snake River Plain. We infer that beneath the Snake River Plain melt is retained at triple junctions between grain boundaries, a geometry that is not predicted to cause seismic attenuation. More elongate geometries beneath the Salton Trough may cause seismic attenuation via the melt-squirt mechanism. In light of these results, we conclude that prior observations of low seismic velocities with somewhat high quality factors beneath the East Pacific Rise and Southern California suggest that melt does not organize into elongate bands across much of the asthenosphere.