YKA dataset

Using YKA, the aim was to find P$^d$P phases and use them to study the structure of the upper mantle in the central Pacific. The Pacific region was chosen because the topography of the discontinuities might reveal the origin of the Hawaiian hot spot. The Hawaiian hot spot is interpreted as the source of the Hawaii-Emperor seamount chain, a more than 6000 km long seamount chain in the central Pacific (Wilson, 1971). The use of the short period YKA data will enable a higher resolution of the upper mantle discontinuities than previously obtained and might resolve whether the Hawaiian volcanism is fed by a mantle plume, and whether this plume originates in the lower mantle beneath the 660 or in the upper mantle.
From the PDE (Preliminary Determination of Earthquakes) catalogue events from September 1989 until March 1997 have been selected. The epicentral distance of the events is between 80$^{\circ}$ and 120$^{\circ}$ with magnitudes m$_b$ $\ge$ 5.5. To study the central Pacific backazimuths of 200$^{\circ}$ to 300$^{\circ}$ are necessary. The depth of the events was not restricted, although the depth phases (pP and sP) of deep events arrive within the same time window as P$^d$P. Even though the detection of deep (210 km - 660 km) discontinuities with deep events is difficult, these events can still be used to detect shallow discontinuities.
This search results in approximately 1000 events meeting the limitations mentioned above.
Tests with synthetic seismograms and combinations of different array techniques were used to develop thresholds for event selection from this first large dataset and to develop methods to detect P$^d$P phases. The synthetic seismograms were computed using the reflectivity method (Müller, 1985) and the gauss-beam method (Weber, 1988). These tests resulted in three criteria to select events from the large dataset with the best chance to detect P$^d$P phases. Only (i) strong events (m$_b \ge$ 5.8) with (ii) a short P coda and (iii) a strong PP onset are selected. To avoid the interference of depth phases pP and sP with the P$^d$P phases shallow events (h $\le$ 100 km) were preferred. The P coda should have decayed before the precursors from the deep discontinuities arrive. The reflections from the discontinuities are weak signals. Strong events have enough seismic energy to push the P$^d$P phases out of the noise. Due to the interaction of PP with the crust at the surface reflection point, the amplitude of PP is reduced. Additionally, PP travels on a maximum travel time path along the great circle path. The PP coda generated at the surface reflection point arrives before the main PP onset, in a way the "PP coda" arrives before PP. Due to the reduction of the PP amplitude and the arrival within the PP coda the PP onset is difficult to detect. The exact measurement of the PP travel time is essential for a precise calculation of the reflector depth beneath the surface reflection point. Strong PP arrivals help to reduce errors in the PP travel time determination.
The final dataset consists of 124 events. The event parameters (source coordinates, origin time, depth and magnitude) are listed in Table A.3. The source locations of the events are shown as stars in the map of Figure 4.4. The thin lines exemplarily show the great circle paths from some events of the different source regions (North Island New-Zealand, Kermadec, Tonga-Fiji, Solomon, New-Guinea, Indonesia and Philippines) to YKA.

Figure 4.4: Location of the events selected to study P$^d$P. The source parameters are listed in Table A.3 The source regions are North Island New-Zealand, Kermadec, Tonga-Fiji, Solomon, New-Guinea, Indonesia and the Phillipines. The thin lines mark the great circle paths to YKA for selected events.

The location of the PP surface reflection points are displayed in Figure 4.5 and are listed in Table A.4.

Figure 4.5: The locations of the PP surface reflection points of the events displayed in Figure 4.4 are marked by the grey ellipses. The ellipses indicate the approximated size of the first Fresnel zone of the 1 Hz PP phases. The Hawaii-Emperor seamount chain is marked by the thick line. The reflection points fill a corridor from the Hawaiian Islands to the Sea of Ochotsk with the best coverage near the Hawaiian Islands and south of Kamchatka.

Figure 4.6: Isochrones for PP at an epicentral distance of 95$^{\circ}$. The isochrones are displayed in a linear projection and 1$^{\circ}$ corresonds to $\sim$111 km. The travel time along the great circle path is a maximum travel time, whereas travel times off the great circle path are minimum travel times. The geometrical reflection point is located at 0$^{\circ}$ longitude and latitude. Isochrones for 0.25 s, 0.5 s, 1 s, 5 s and 10 s are displayed.
The outline of the 0.25 s isochrone defines the first Fresnel zone for 1 Hz data. The Fresnel zones indicated in Figure 4.5 are an approximation of the central part of the first Fresnel zone as shown by the ellipse. The approximation neglects the fringes of the saddle shaped Fresnel zone.

The source-receiver combination results in reflection points (light grey ellipses) within a corridor from the Hawaiian Islands to the Sea of Okhotsk. Some reflection points lie beneath the Hawaii-Emperor seamount chain (thick black line), especially at the tip of the chain. Particularly, the line of reflection points originating from events in New-Zealand, Kermadec and Tonga-Fiji, across the tip of the Hawaiian chain represents a unique possibility to study the influence of a postulated mantle plume on the structure of the upper mantle discontinuities. Aside from the reflection points near the Hawaiian Islands, the Kurile subduction zone south of Kamchatka is a geodynamically interesting area, which can be studied with this dataset. In contrast to the points at the tip of the Hawaiian chain, these reflection points can resolve the interaction of a cold slab with the discontinuities.
The size of the ellipses, indicating the location of the reflection points, represent approximately the size of the main part of the first Fresnel zone. The Fresnel zone describes the maximum spatial resolution of a wave with a certain frequency. The first Fresnel zone is defined as the area around the ray (in geometric ray theory) where the waves originating from each point within this area (following Huygens principle) interfere constructively.
Isochrones, lines with equal travel times, for PP for an epicentral distance of 95$^{\circ}$ are presented in Figure 4.6. The typical saddle shaped form of the PP Fresnel zone occurs because the travel time of PP along the great circle path is a maximum travel time (i.e. every point on the great circle path other than the geometrical bounce point shows a shorter travel time). On the other hand, every path off the great circle path has a longer travel time. The first Fresnel zone is defined as the area within the T/4 isochrone (T is the period of the wave). For the 1 Hz data used here this means the 0.25 s isochrone. The fringes of the Fresnel zone extend to large distances. These fringes complicate the interpretation of PP waveforms and travel times, because features far away from the geometrical reflection point can contribute to PP. The ellipses in Figure 4.5 approximate the size of the Fresnel zone as the area of the central part of the real Fresnel zone as indicated by the dashed ellipse in Figure 4.6. The size and the shape of the first Fresnel zone of PP is only changed negligible for the P$^d$P phases.
The use of the French nuclear tests at the Muroroa atoll with reflection points beneath probably undisturbed lithosphere south of Hawaii would provide a possibility to compare the altered lithosphere near Hawaii with ''normal'' oceanic lithosphere. Unfortunately, this comparison is impossible, because the recordings of the nuclear explosions at YKA do not show PP arrivals, although they show clear first P onsets.