The reverberative interval of the seismogram, defined as the portion following the surface wave train propagating along the minor arc and ending with the first body wave arrivals from the major arc, provides an excellent window in which to observe subcritical reflections from mantle discontinuities. On an SH-polarized seismogram the intervals between ScSn and sScSn wave groups (zeroth-order reverberations) are approximately 15 min long and consist almost entirely of waves reflected one or more times from mantle discontinuities (first- and higher-order reverberations).

We have developed signal enhancement and waveform inversion schemes to progressively extract information pertaining to mantle layering from the waveforms of zeroth- and first-order reverberations and have applied them to a data set of long-period digital seismograms drawn from a number of tectonic regions in and around the western Pacific. The results place important constraints on the nature of the transition zone discontinuities. Measurements of the two-way, vertical shear wave travel time from the free surface to the 400-km and 650-km discontinuities (&tgr;400 and &tgr;650 respectively) yield estimates of transition zone travel time, &tgr;TZ, largely uncontaminated by outside heterogeneity. Variations in the path-averaged estimates of &tgr;TZ are of the order of 5.4 s, indicating significant lateral heterogeneity in the transition zone.

Correlations of &tgr;TZ with &tgr;400 and &tgr;LM (travel time through the lower mantle) are consistent with, but do not require, the thermal deflection of an exothermic phase transition near 400 km and an endothermic transition at 650 km, with ~12 km of long-wavelength (500--5000 km) topology on each. The amount of topography estimated for the 650-km discontinuity is significantly less than predicted for a rigorously stratified mantle (>100 km). In a search for other mantle discontinuities, we find evidence for a strong reflector possibly associated with the onset of partial melting in the upper mantle (average depth of 86 km). Due to ambiquity inherent to the method, this feature can be interpreted as an impedance increase in the D' region of the lower mantle (70--170 km above the core-mantle boundary). No evidence is found for a 220-km discontinuity or D' discontinuity proposed by Lay and Helmberger (1983), suggesting that if such features are present in the western Pacific, they are either small (normal incidence, SHreflection coefficient R<1.5%) or intermittent or have lateral depth variations well in excess of 50 km. ¿ American Geophysical Union 1989