Presentation

This work is a collaboration with Stuart Henrys and Stephen Bannister of the Institute for Geological and Nuclear Sciences in Wellington, New Zealand. We thank Russell Robinson of IGNS for providing his earthquake relocations and velocity model. Louie enjoyed sabbatical support as well from the Victoria University of Wellington.

We examine a west-directed, northwest-dipping subduction zone at the Hikurangi margin, where Pacific oceanic crust subducts under the Australian plate continental fragment of New Zealand's North Island. The Hikurangi is at the south end of the Tonga-Kermadec system.

It was examined to the northeast at East Cape (the Raukumara peninsula) recently by Eberhart-Phillips and Reyners of IGNS, who found a 1-2 km thick plate interface with a high Vp/Vs ratio of 2, from S-P converted waves.

We look closer to Wellington, at a sequence of M6 events and aftershocks near a town called Weber.

We use a reflection imaging technique, which has resolution only where we have an array of both sources and receivers.

We have the New Zealand national network, mostly more than 100 km from Weber, courtesy of Terry Webb of IGNS. Thanks to Russell Robinson we also have data from a 10-station portable deployment by IGNS, and 500 events. We will use distances to 70 km.

The first event was the M6.2 Weber I in Feb. 1990, a steep-NW-dipping normal fault in the slab below the 20 km deep plate interface. That was followed 3 months later by the M6.4 Weber II event, which was a shallow-dipping southeast-directed thrust above the interface.

For imaging we use ~200 Weber I aftershocks in the lower plate, and Weber II and about 300 of its aftershocks in the upper plate. All hypocenters are quality-A, done by Russell Robinson. Robinson also found that the data could determine a 1-d velocity model quite well, but did not contain 3-d velocity constraints.

Robinson showed that composite mechanisms of each sequence were very similar to the mechanisms of the mainshocks.

First we will image a 50-km-long dip section across the area of best coverage. Later we will look at a 25-km-wide volume image.

We would prefer to see obvious back-scattered coherent arrivals in this single-receiver gather of 250 Weber I events. There are hints of this, highlighted in brown, that we could probably enhance with velocity filters or with a dip-filtering imaging process. At this point we prefer to let the imaging procedure itself identify the coherent energy that fits the time imaging conditions of our migration.

In this part of the Hikurangi we do not see the strong S-P conversions that Eberhart-Phillips and Reyners saw to the northwest. Such energy would be parallel to and between the blue lines.

We use only the vertical receiver components, although 3 components are available. Data quality from the gain-ranged portables and permanent stations is superb, with no clipping.

We are not locating faults by locating aftershock hypocenters. We are using the events as sources of energy to illuminate other structures, away from the ruptured faults. We examine both forward-scattering and back-scattering of this illumination.

Given an arrival from a certian source at a station at a given time, we assume it diffracted from some point on an ellipsoid with the source and receiver as foci. For the long cigar-like ellipsoids the back-scattering occurrs near the tip and butt of the cigar, with forward scattering occurring around the belly.

Although we include here both forward- and back-scattered waves, it takes many sources, many receivers and many ellipoids to define a structure well. Thus we can only image a limited volume between an event swarm and a receiver array, or below both.

This is a test with synthetic elastic events in the 1-d velocity model from Robinson.

Interfaces below sources only focus where ``depth point'' coverage is excellent, with incomplete source deconvolution resulting in ~2 km depth uncertainty.

The focusing of overlying interfaces from forward-scattered waves is very poor. Only the presence of a scatterer is suggested, with more than 5 km placement uncertainty within the section.

The enhancement procedure involves migrating noise estimated from the data by resampling, as well as the data. Harlan, Claerbout, and Rocca's Bayesian statistics compute where the section is determined from the most coherent events in the receiver gathers. Then we de-emphasize the parts of the migrated section without good pre-migration coherency. We also apply Claerbout and Lumley's migration operator antialiasing criterion, with an assumption of shallowly dipping structure.

These 4 dip sections show the multicomponent migration of P-P, P-S, S-P, and S-S scatterers. These are mostly back-scattered images using the shallow Weber II sources. Below are mostly forward-scattered images using the deep Weber I sources.

We've superimposed a northeastward view into the focal sphere. The refracted nodal plane of the Weber I events is clear between the blue negative and red positive reflectivities in the forward-scattered image at the lower left. We have not corrected for phase shifts around the focal sphere.

Resolution is best for P-P backscatter below the Weber II sequence, illuminating the slab interface in some detail. These details are similar in the S-S image. The S-P backscatter, using only arrivals between the P and S arrival, does not get enough data to focus well geometrically.

The P-P forward scattering is stunning, especially considering how much less S-S forward scattering takes place. To avoid direct arrivals we set a one-second minimum time between image points and the sources. Forward scattering is strongest in a wedge above and just northwest of the Weber I sequence.

Overlaying the event locations on the sections leads to an interpretable structure for the plate interface from the back-scattered images. The double-reflector plate interface is 3-5 km thick, and dips 5 degrees NW except where it is offset above the Weber I event by 3-4 km downdip.

The forward-scattered images suggest strong diffractions on the downdip side of the offset. Our images cannot resolve the downdip extent of this ``bright spot'', or its exact nature. The strong P-P scattering (both forward and back) without strong S-P or S-S scattering suggests porosity rather than rigidity variation.

This result would not agree with the increased Vp/Vs ratio found within the plate interface by Eberhart-Phillips and Reyners to the northeast, if rigidity is held approximately constant. Here Lame's lambda parameter must decrease significantly under constant rigidity, which would lower the Vp/Vs ratio.

Perhaps up to 3% increase in porosity, at grain triple junctions but not at intergranular faces, could decrease lambda without decreasing rigidity. Such parameters would be similar to the Cascadia interface modeled by Cloos and Shreve, with less lithified and higher-porosity sediments trapped in the plate interface.

This is a duplex thrust interpretation of the effect of the downdip step in the plate interface.

The Weber I fault, with significant normal offset before subduction, has bulldozed a wedge of sediment downdip that has not dewatered as completely as elsewhere in the interface.

The frontal ramp of the lower-plate step has raised a duplex thrust out of the upper 2-3 km of the interface. The frontal ramp of the duplex is visible in the P-P back-scattering reflectivity. A previous duplex, aligned with the frontal ramps, could explain the mechanism of the Weber II event. Fault-bend thrusting above the duplex is another possible explanation of that event.

Higher topography along the Wairarapa coast above the duplex thrust system is in agreement with this hypothesis.

Here we tested the effect of including headwave diffractions on our images. On top are traveltime contours from a surface source through Robinson's velocity model that contains sharp refractors. The diffractive, low-energy headwaves are the straight ramps in the contours.

We convert the headwaves to diving waves simply by smoothing away the sharp boundaries in the velocity model. The ramps have become, at middle, higher-energy spherical waves.

Warm colors at the bottom show where headwave diffractions would be advanced over diving spherical waves. The maximum advancement is only 0.1 second, and only appears within 1 km above the refractor.

These sections compare a migration imaging condition using diving waves, left against imaging including diffracted headwaves.

The differences are minimal and do not affect the interpretation. Since we are migrating 2-8 Hz waves, the maximum headwave advance of 0.1 sec is immaterial.

With our Kirchhoff-sum imaging technique, true 3-d geometry is always taken into account. Thus migration into a 3-d image volume instead of an arbitrarily oriented section is a simple though time-consuming operation (1-2 hours here on a Sun Ultra 10). The image volume here is 25 km on a side and covers a 20-km depth range from 10-30 km depth.

The oblique slice of the volume at the top presents just the central part of the dip section we showed already. It has the same features as it should, although we have not statistically enhanced this section and it has more visible artifacts.

Interpreting a depth slice at 20 km depth, middle, has the same problems as interpreting a sectional slice. With our method similar artifacts appear on sections in both directions.

At 20 km depth we can see the 5-km-thick plate interafce curving away from the Weber I fault (which is deeper) to a more westerly strike. The plate interface is above the Weber I fault only at its largest offset, in the middle of the volume.

A couple kilometers deeper the upper plate (on the left) still diverges from the Weber I normal fault in the lower plate. The high-porosity bright spot is entirely within the upper plate.

These 3-d images suggest the bright spot and the duplex-thrusted plate interface are geometrically related. They are both deepest where the Weber I fault offset is greatest.

This cartoon, looking northeast, suggests how the duplex-thrust system may transition from a double duplex at the Weber I frontal ramp, to a single duplex where offset on the Weber I fault dies out.

A wedge of sediment is trapped only in front of the Weber I fault's total offset and pinches out to the southwest. The pinchout of the wedge creates the divergence of the reflective plate interface and the Weber I fault.

I'll conclude by suggest that this natural-source ``reflection survey'' was successful in showing active subduction interface geometry at 20 km depth.

• The Weber I is a slab-bend normal event, maintaining a downdip step in the plate interface. It is not a seamount as are found subducting to the northwest under East Cape.

• Subduction of the fault-offset step has trapped sediment and some fluids, in an anomalous high-porosity wedge. Porosity is limited to a few percent as bulk rigidity is not much affected.

• The Weber II event was either on an active duplex thrust, or on a fault-bend thrust above the duplex.

• The duplex, possibly doubled at the Weber I fault's offset, is consistent with the coastal mountain range and inland plain of the Wairarapa region. Although the plate interface itself has unknown seismicity, the ramp and the duplex increase the area's seismic potential.