Conclusions

The structure of the upper mantle discontinuities has been studied using PP underside reflections (P$^d$P). Because earthquakes from the western Pacific rim recorded at the Yellowknife array (YKA) in the Northwestern Territories (Canada) are used, the region studied stretches from the Hawaiian Islands to the Sea of Okhotsk. The PP reflection points lie within a $\sim$6000 km long and $\sim$700 km wide corridor.
For the detection of the small amplitude P$^d$P phases in the seismograms, array methods were used. A newly developed sliding-window fk-analysis was successfully applied to study P$^d$P.
The source receiver combination allowed the study of the seismic discontinuities beneath geodynamically interesting structures like the location of the assumed Hawaiian mantle plume and the Kuriles subduction zone south of Kamchatka. The use of short-period array data and the application of array methods allowed a much better resolution of the structure and the depth of the discontinuities in the upper mantle than previously obtained by results from long-period seismograms (Shearer, 1991; Goßler and Kind, 1996; Gu et al., 1998; Flanagan and Shearer, 1999). Furthermore, the use of short-period underside reflections allowed an evaluation of the thickness of the discontinuity, i.e. the depth interval of the gradual density and/or velocity change.
The attempt to resolve the depth structure of the anisotropy with SS precursors failed due to the small dataset available, but other possibilities to resolve this structure were proposed.


The study of the P$^d$P phases revealed evidence for discontinuities at different depths and resulted in indications for some new features of the upper mantle structure. The study detected a positive reflector at a mean depth of 60.2 $\pm$ 15.8 km beneath the whole profile. A positive reflector is defined as a velocity change from fast velocities to slow velocities. The depth and the velocity jump corresponds well to the Hales discontinuity (Green and Hales, 1969; Hales, 1969; Revenaugh and Jordan, 1991b; Simpson et al., 1974). The depth of the discontinuity can be correlated with the phase transition of spinel $\rightarrow$ garnet in a lherzolitic mantle (Hales, 1969).
Another reflector has been found at shallow depths. This reflector shows a velocity change from slow to fast, in contrast to the Hales discontinuity. The mean depth of this reflector is 99 km. The depth and the velocity jump infers the interpretation of this reflector as the Gutenberg or 8$^{\circ}$-discontinuity. The depth and velocity contrasts fit the values determined by previous studies (Thybo and Perchuc, 1997; Kanamori and Press, 1970; Knopoff, 1983; Revenaugh and Jordan, 1991b).
Both discontinuities show, that topography and the depths are positively correlated. This is plausible for the phase transition defining the H, but not for all explanations for the Gutenberg discontinuity discussed so far. Therefore, this study favours the onset of partial melt defining a low-velocity-layer as a mechanism (Thybo and Perchuc, 1998), rather than the explanation as a boundary between dry and water saturated material due to the onset of partial melt at the ridge (Gaherty et al., 1999).


The global existence of the Lehmann discontinuity at depths around 200 km is still under debate. For the first time this study confirms the existence of a reflector at this depth beneath oceanic regions. The depth found here is in good agreement with other results (Vidale and Benz, 1992; Woods et al., 1991). The reflector shows unusually strong depth variations and this study cannot prove that the reflections found by the PP underside reflections are really correlated to a continuous reflector. The scattered reflections can also be related to a layer with embedded heterogeneities (Rost and Weber, 2000). The subduction zone does not have any effect on the depth of the Lehmann discontinuity indicating a non-temperature controlled mechanism for this discontinuity.
Synthetic tests using the reflectivity method (Müller, 1985) have been performed to study the impedance structure and the thickness of the discontinuity. These tests indicate an impedance contrast larger than 5.5% - 6.5%. This minimum impedance contrast is larger than changes found by Gaherty and Jordan (1995), but it is in agreement with the change suggested by PREM (Dziewonski and Anderson, 1981). The thickness of the transition can be estimated to be sharper than 7 km. This indicates a very sharp discontinuity.
The results of this study cannot discriminate between different mechanisms generating the Lehmann discontinuity and the scenarios of a continuous reflector and an inhomogeneous layer. This must be tested using other methods, e.g. surface wave studies in this region.


The 410-km discontinuity has clearly been detected in this study. The mean depth of 404 km is in good agreement with other studies (Flanagan and Shearer, 1999; Benz and Vidale, 1993; Melbourne and Helmberger, 1998). The reflection points form two separated groups, one near the Hawaiian Islands and another one between the bend of the Hawaiian-Emperor seamount chain and the Sea of Okhotsk. The gap between both groups correlates perfectly with a rise of the 410 (Flanagan and Shearer, 1999). The reasons for this correlation have been discussed.
The 410 shows a depression of $\sim$30 km approximately 100 km northeast of the Hawaiian Islands. This depression can be interpreted as the piercing point of the Hawaiian plume through the 410. The plume temperature in relation to the temperature of the surrounding mantle has been estimated to be 300 K - 400 K in fairly good agreement with numerical plume studies (Ribe and Christensen, 1994). The plume radius is difficult to estimate, but seems to coincide with suggested radii of $\sim$150 km. This interpretation must be further validated by more data points in this region.
The subduction zone is not sampled densely enough to study the discontinuity at the piercing point of the slab, but one reflection point samples the interaction of the cold subducting slab with the phase transition. This reflection point shows an extreme elevation of $\sim$80 km, which is much higher than in previous studies. This elevation has been explained as the result of the phase transition uplift and the special ray configuration with the parallelism of the slab and the great circle path. The uplift corresponds to a slab temperature of approximately -1000 K compared to the surrounding mantle. This is a reasonable temperature difference for an old (119 million years) and fast subducting slab. This example shows the high resolution which is obtainable using short period array data.


The analysis of the dataset yields no indication for reflections from the 520-km and 660-km discontinuity. The lack of the 520 is not surprising since the suggested phase transition at this depth is extended over a depth interval of at least 25 km (Shearer, 1990) with a small impedance increase. This gradient zone is unresolvable with short-period data.
The non-detection of the 660-km discontinuity offers the possibility to set constraints on the maximum impedance contrast and the minimum thickness of the discontinuity using synthetic seismograms. These tests indicate an impedance increase of less than 10%. This is in agreement with recent models of the 660 indicating 7% - 9% impedance contrast (Shearer and Flanagan, 1998; Estabrook and Kind, 1996). These models assumed a first-order discontinuity because of the use of long-period data. The synthetic tests show a minimum thickness of 12 km for the 660-km discontinuity, which is in disagreement with other seismological studies (Xu et al., 1999; Herrmann et al., 1999). The study of the short-period P$^{660}$P fills a gap between long-period PP studies and short-period P'P' studies of the 660 and it is postulated that the structure of the 660 km discontinuity is more complicated than inferred by the simple phase transition model.


This study demonstrates the possibility to use short-period data of small aperture arrays to study the structure of the upper mantle discontinuities. Some of the results open new questions, which could not be solved entirely. Most of the available data of YKA are used for this study, but in the Pacific different source-receiver combinations exist to study the same or adjacent regions to the corridor investigated here. Possible combinations are Alaskan and South American events recorded at the Warramunga array (WRA) in Australia, an array comparable in size and configuration to YKA, or events from the same source regions used here recorded at the Large Aperture Seismic Array (LASA) in Montana. The Indian ocean can be studied using earthquakes in the New-Zealand and Tonga-Fiji region recorded at the Gauribidanur array (GBA) in southern India (similar configuration to YKA). Best suited for the study of upper mantle discontinuities would be an array with a large aperture ($\sim$100 km) equipped with broad-band seismometers and an interstation spacing of $\sim$3 km in a well defined configuration. Unfortunately, this 'ideal' array does not exist.
Nevertheless, the analysis of data from the small aperture arrays (WRA, LASA, GBA) will improve the knowledge of the upper mantle discontinuities.