T. Lay $^{a}$ and E. Garnero $^{b}$
$^{a}$ Earth Sciences Department, University of California, Santa Cruz, USA $^{b}$ Department of Geological Sciences, Arizona State University, Tempe, USA
The lowermost mantle structure beneath the Caribbean is investigated using shear waves from intermediate and deep focus earthquakes in South America recorded by broadband stations in the United States and Canada. Arrival times, waveforms, and polarizations are considered for SKS, S, and ScS phases. The primary goal is to consider the relationships between differential travel time anomalies, shear-wave splitting, and lower mantle triplications in the region beneath Central America and the Caribbean. A very large data set, comprised of shear wave recordings from over 50 earthquakes recorded at about 100 stations (IRIS, TERRAscope, BDSN, NSN, CSN, PASCAL) is utilized to achieve data coverage and redundancy that allows separation of upper mantle, mid-mantle and deep mantle contributions to the signals. As in earlier work, there is clear evidence of waveform complexity consistent with triplication of shear waves as a result of a several percent velocity increase about 250 km above the core-mantle boundary west of Central America and extending into the Caribbean. Lateral variations in the timing and strength of the triplication arrivals indicate the presence of topography and/or lateral pinching out of the velocity increase. This feature is most clearly observed for events beneath Argentina at the dense network of broadband stations in California, but there are numerous compatible observations speckled across North America. Events beneath Bolivia show little direct waveform evidence for a triplication at northern California stations, but careful waveform stacking and deconvolution indicates that there is waveform complexity compatible with one, or possibly two triplications. ScS-S and S-SKS differential times indicate that the regions with triplications correspond to faster than average zones at the base of the mantle. Mid-mantle features are suppressed by applying corrections for high resolution tomographic models, and this proves important, as it reduces lateral heterogeneities in the lowermost mantle structure. Analysis of shear wave splitting supports earlier work favoring the existence of contributions from the deep mantle, but the uncertainty of lithospheric anisotropy corrections presents significant challenges. In some cases, application of station corrections based on studies of SKS splitting for azimuthally distributed events degrades SKS or S linearity. This casts doubts on the reliability of the corrections and on the interpretation of shear wave splitting in the presumably corrected data. Numerous observations of ScS splitting are found, with the transverse (SH) component peaking 1-2 s earlier than the radial (SV) component. While some of this behavior is accounted for by the predominantly E-W orientation of fast S polarization directions for lithosphereic anisotropy beneath North American stations, it appears some of the signal arises from the deep mantle. This is supported by the complexity of S waveforms (apparently triplicated) at distances where the turning point grazes the lowermost mantle. However, the latter phases indicate that the lower mantle anisotropy is not radial (i.e. transverse isotropy with vertical symmetry axis) as proposed in earlier work, but is azimuthal (albeit with SH being close to the fast S polarization direction). Relationships between anisotropy, volumetric shear velocity heterogeneity, and triplication effects are considered.