| The amplitudes of seismic spectra decay with increasing frequency because of anelastic attenuation along the ray path and earthquake source characteristics. We have developed a method of stacking compressional wave spectra to separate these effects. At higher frequencies, attenuation effects dominate. We estimate the differential attenuation for thousands of paths through the Earth's mantle from the slope of the log spectrum between 0.16 and 0.86 Hz. First, we invert the attenuation estimates for a two-layer, frequency-independent Q model, which shows that the uppermost ∼220 km of the mantle is approximately five times more attenuating than the lower mantle. To reconcile these results with longer period studies, which indicate a smaller contrast in attenuation between the two layers, we construct a Q model with independent absorption bands for the uppermost and lower mantles. The data require significant frequency dependence in the lower mantle but not in the uppermost mantle. Next, we map lateral variations in attenuation in the uppermost mantle. The patterns we see generally agree with surface tectonics: the most-attenuating regions tend to occur beneath young, warm oceanic regions while the least-attenuating areas are usually found beneath old, stable continental interiors. At longer periods, source characteristics have a greater effect on variations in spectral decay. We interpret differences in spectral slopes as variations in apparent rupture duration. For 66 intermediate-and deep-focus earthquakes, we compare our observed azimuthal variations in rupture duration with the patterns predicted for a simple model of unilateral rupture to determine the rupture direction of each earthquake. We use the rupture direction to distinguish the fault plane from the auxiliary plane of the focal mechanism. |
