Phase arrivals resulting from reflections and conversions at upper-mantle seismic discontinuities have been previously observed in short- and long-period seismograms [e.g., Engdahl and Flinn, 1969; Vinnik, 1977; Paulssen, 1988; Nakanishi, 1988; Davis et al., 1989; Revenaugh and Jordan, 1991]. Observations of these discontinuity phases provide information on the physical properties of the discontinuities [e.g., Lees et al., 1983; Shearer, 1991a]. Short-period observations are particularly valuable since they help to constrain the sharpness of the discontinuities. For example, the existence of short-period reflected phases from the 660-km discontinuity is consistent with a change in seismic velocity occuring over a fairly narrow depth range (less than 4 km) [ Richards, 1972] at the reflection points. The thicknesses of the upper-mantle discontinuities have direct implications in the ongoing debate over the thermal and chemical composition of the upper mantle.
As previously noted, arrivals resulting from P-to-S phase
conversions at the 410- and 660-km discontinuities are visible in
the long-period autopicks; however, we have so far been unable to
identify any upper-mantle discontinuity phases in the short-period picks.
Previous studies have reported the observation of discontinuity phases in the
region between the P and the PP phase arrival. Kato and
Hirahara (1991) examined the short-period phase arrivals contained in the ISC
catalog and observed a distinct cloud of arrivals preceding the PP phase
in the range 90
to 120. They attributed the bulk of the PP
precursors to near surface scattering; however, they suggested a subset of
the picks may have resulted from underside reflections at a localized
upper-mantle discontinuity (i.e. a PdP phase).
The short-period autopicks also produce a cloud of arrivals preceding the PP phase (see Figure 3); however, unlike the results from the ISC data, the cloud observed in the autopicks blends into the PP branch. We attempted to associate the PP precursors with PdP phases by calculating the PdP reflection depths which would produce the observed travel times. Histograms made from these apparent reflection depths showed no distinct peaks. Thus, it is unlikely the PP precursors observed in the short-period autopicks originate from a globally coherent discontinuity. The incoherence of the automatically picked PP precursors can be explained by a combination of false triggering and near-surface scattering.
Other discontinuity phases have been observed in the P-PP window. Stacks
of long-period vertical-component seismograms [Shearer, 1991a] reveal
PdP and Ppdp phases in the range 50 to 120. Ppdp phases result
from a surface reflection and a topside upper-mantle reflection which produces
two additional legs between the surface and the discontinuity [see Shearer
1990]. We attempted to highlight PdP and Ppdp phases in the short-period
picks by reducing the travel times to the observed P phase arrival time. The
autopick closest to the predicted P arrival time was located and its
travel time was subtracted from all the subsequent arrivals at a given station.
Plots of these differential travel times versus range tightened the spread of the
PcP arrivals but did not reveal the existence of PdP or Ppdp phases.
Another method for resolving the possible existence of these phases is the
calculation of apparent reflection depth. We convert the difference between
the calculated PP arrival time and the picked precursor's arrival time
into an apparent reflection depth. No significant trends are seen the
apparent reflection depths.
Short-period precursors to 
have been identified in a
variety of studies including Nakanishi (1988), Davis (1989), and
Vidale and Benz (1993), and are believed to result from underside reflections at
the 410- and 660-km discontinuities. Although can be
seen (faintly) in the image of short-period autopicks (Figure 3), there is no
evidence for coherent precursors in these data. As a check, we tested the
autopicker on a set of high quality short-period seismograms (from the Calnet
stations in northern California) which contained
precursors [John Vidale, personal communication]. The autopicker was able to
identify discontinuity phases on the majority of these records. The lack of
clear precursors in the global data set indicates that
observations of these phases are fairly unusual. This is consistent with the
large number of null observations reported by Davis et al. (1989), and
suggests that the amplitude of precursors is typically at
or below the noise level on short-period seismograms.