John N. Louie,* Stuart Henrys,† Stephen Bannister,† and Russell Robinson†

Earthquake seismic images show three-dimensional spatial and temporal properties of a subduction fault interface. A seismograph array¹ recorded an unambiguous forward-scattered P-P phase from the interface of the Hikurangi subduction zone in New Zealand. P-P scattering was most prominent within days of the Feb. 19, 1990 Weber I M6.2 earthquake in the lower plate below the 20-km-deep interface, and less prominent three months later. S-P phase conversion at the sediment-laden interface was absent although it had been widely prominent previously.^2-5 We suggest that dilatation of fluid-filled pores by the Weber I normal-faulting earthquake caused mineral precipitation by depressurisation, sealing thin cracks to yield more spherical pores,⁶ and lowering Poisson's ratio^7-8 in the interface. The forward-scattering arose in a wedge of sediment, 3-5 km thick, bulldozed into the interface by a subducted fault offset. Activation of a duplex thrust^9-10 by the temporary sealing and locking of the interface¹ may have triggered a second earthquake, the May 13, 1990 Weber II M6.4 event, within the upper plate.

As the agents of accretion and continental formation, Earth's subduction zones exhibit a wide variety of geometric configurations and appear to have both accretionary and erosional modes. Scraping and seamount pinning may produce stiff barriers or asperities that limit the rupture processes of giant subduction earthquakes.¹¹ Such barriers may also cause uplift and large earthquakes within both the upper and lower tectonic plates. Water released at great depth may pond against rocks sealed by precipitated minerals.^6,12 Aside from forming strong seismic reflectors, the trapped water injects itself into the subducting slab causing seismicity within the lower plate. Observation of subduction-zone fluids, whether ponded⁶ or in motion¹³ by remote, high-multiplicity geophysical measurements would allow direct constraints on models of both subduction processes and their seismic potential. The erosional processes of frontal seamount impact and basal scraping have been inspected by deep seismic-reflectivity imaging,¹⁴ and by seismic-velocity tomography.¹¹ Previous evidence of seamount impacts, scraping, and tunnelling, as well as erosive basal megalenses help reveal the properties of subduction zones in an erosive mode. Accretionary processes imaged by these techniques include tectonic underplating^9-10,15 and accretion folding.¹⁶

We produce detailed three-dimensional seismic reflectivity images characterising, for the first time, accretionary processes at intermediate depths within a subduction interface. The Hikurangi margin east of the North Island of New Zealand (fig. 1) is undergoing shallow, west-directed subduction at 4-5 cm/yr.^2-4,17 Depths of 10-50 km are too shallow for magmatic underplating, and deeper than simple accretionary-wedge thrusts. These images show the large-scale accretion processes that tend to have been preserved geologically in ancient, exhumed subduction interfaces.^9-10 The seismic images, derived from intraplate earthquakes¹⁸ of the 1990-1992 Weber sequence,¹ show as well the dynamic process of fluid entrapment and fracture sealing within the interface. Events below the subduction interface usually produce prominent conversions of shear (S) waves to compressional (P) waves.^2-5 The phase conversions and low shear velocities¹⁹ suggest that the interface is unusually rich in sediment and fluid, at least in front of seamounts or other structures on the plate.

The seismic-reflectivity images show (fig. 2) how a subducted normal-fault offset of the sea floor entraps a thick wedge of sediment. The accretionary margin is underplated with this sediment wedge through a process of duplex thrusting,^9-10 that effectively accretes oceanic sediment to the bottom of the continental wedge at a depth of about 20 km. Thickened low-velocity crust suggests similar accretionary thickening in the Celebes Sea.¹⁵ Duplex thrusting and wedge thickening by subduction of normal-faulted "washboard" oceanic crust has been imaged along the more steeply dipping subduction zone off Nicaragua.¹⁴ Where seamounts subduct along that margin, erosion and wedge thinning occur. Along the Hikurangi margin, velocity tomography at East Cape,¹⁹ northeast of Weber, shows thickly underplated sediment at 20 km depth pushed in by a high-velocity seamount¹¹ and interpreted as accretionary. The images in fig. 2 show structural details most consistent with accretion and thickening, but resulting from a faulted subducted plate rather than a seamount.

The fact that earthquake sources strongly emit shear waves as well as compressional waves means that these images can reveal the elastic properties of the subduction interface. The plate interface at Weber sequence scatters P waves much more efficiently than S waves or P, S conversions. Petroleum exploration work^7-8 shows that this type of scattering is characteristic of rocks with increased pore fluid content in contrast to surrounding rocks, where the pores have at least a 0.1 thickness:width aspect ratio.⁶ Previous images of seismic reflectivity in subduction zones^6,12 could not assess shear-wave reflectivity, as they were derived from compressional sources of seismic energy.

The strongest P-P scattering, shown in fig. 2e, may be limited to a few-day period following the Weber I earthquake in the lower plate. Strong S-P conversions from the plate interface near Weber had been observed decades before the 1990 earthquakes,² and farther north along the Hikurangi subduction zone.^4-5 However, such conversions do not appear in the 1990 data (e.g., fig. 1). This result suggests the 1990 earthquake altered the physical conditions of the interface by dilating the wedge of trapped sediment and temporarily changing the geometry of the pores. Dilation decreased and the properties reverted over the subsequent months. The P-P bright spot is less prominent three months later (fig. 2a) when imaged using earthquakes in the upper plate.

The seismograms in fig. 1 are representative of the Hikurangi margin, although this type of travel path had in the past shown a strong S-P arrival at 8 or 9 seconds time.² The forward-scattered P-P arrival (fig. 1), following the P arrival by just 0.2 sec, is prominent on most of the 164 Weber I seismograms recorded at station WTA. Very few seismograms suggest S-P conversions. The Weber I P-P phase is similar in origin to forward-scattered "late phases" recorded from the Philippine Sea plate below Japan.²⁰ Unlike the events from the Philippine Sea plate, the forward-scattered S-S image (fig. 2h) shows that the Weber I sequence produced a much weaker S-S phase.

P-P arrivals are not prominent on Weber I recordings from other stations such as TEU (fig. 1), where the wave front does not pass through the plate interface at the bright spot just west of the Weber I events (grey diamonds on fig. 1). Fig. 3, at the bottom, shows the location of the bright spot on a horizontal slice through the imaged reflectivity at 21.4 km depth. Tests individually migrating the 164 Weber I seismograms from WTA demonstrate that the bright spot arises from the P-P forward-scattered arrival (such as P-P on fig. 1).

The P-P back-scattered reflectivity in fig. 2 suggests a specific geometry for the plate interface at the Weber sequence. A preexisting normal-fault of the plate offset has protected a 3-5 km section of marine sediment from accretion to the overthrust complex's leading edge,²¹ and pushed or "bulldozed" it to 20 km depth below the North Island. Locally, the 15-km-long Weber I fault offset has formed the frontal ramp of a duplex thrust^9-10 (fig. 2a) and pushed some of the sediment wedge up to accrete it to the base of the upper plate.

The basal accretion by duplex thrusting may have significantly affected the local topography. The 1.5-km-high backstop thrust ridges forming the spine of this part of the North Island are interrupted by a gorge at the Manawatu River, which flows westward from the isolated Puketoi Range (fig. 4). The Puketoi Range, southwest of the Weber I fault, conceivably has its origin in the underplating by duplex thrusting 15 km below. In this hypothesis, the topography of the Puketoi Range should have developed within the last 1-2 million years, and should be progressing westward. This topographic development would be analogous to the uplift of accretionary wedges by underthrust seamounts.^11,14,19

The wedge of sediment bulldozed under the margin by the fault offset appears as a reflectivity "bright spot" in all of the image sections (fig. 2). "Bright spots" (5-20% reflectivity) are prominent in reflection imaging of other subduction zones, and have led to the hypothesis of trapped pore water at lithostatic pressure.^6,12 Using the Weber I earthquake sources and a 2-8 Hz frequency band, the strength and sign of the bright-spot reflection coefficient are apparent in the unambiguous forward-scattered P-P phase of fig. 1. Migration of the earthquake seismograms places the bright spot in the headwall of the Weber I normal fault (fig. 3).

The strong P-P phase, absence of any significant S-P phase, and the lower P-S, S-P, and S-S reflectivity of the bright spot in the multi-component images (fig. 2) all suggest that the bright spot must have anomalous P-wave but not S-wave velocities. Both velocity types depend on rigidity but only P velocity depends on the Lamé elastic parameter l. The multi-component images suggest that the bright spot is a variation in P velocity (l), without significant S-velocity (rigidity) variations. Furthermore, l variations have the special isotropic property that the P-wave forward scattering has the same sign and strength as the P-wave back-scattering, and there is no S-wave scattering.²⁵ In other words, there is no distinction between transmission and reflection if there are no rigidity (or density) variations. Because the P-P reflection coefficient at the bright spot appears to be negative (fig. 1), the plate interface there must have a lower P velocity than its surroundings. Because any S-velocity variations must be much smaller, the interface must be a zone of reduced Poisson's ratio.

Most elastic modelling of subduction interfaces^5,12,19 conclude that free pore water reduces interface rigidity (i.e. S velocity) and increases Poisson's ratio relative to surrounding rocks. Pore geometry (thickness:width aspect ratio) can control whether increased porosity will produce an elevated or reduced Poisson's-ratio.⁶ Spheroidal pores with aspect ratios of 0.1 (oblate) to 1.0 (spherical) will cause decreases in Poisson's ratio while thin, crack-like, pores with ratios <0.03 will cause increases in Poisson's ratio.

Laboratory samples of sandstones and shales often show Poisson's ratios below 0.25.^7-8 Poisson's ratio decreases with increasing porosity in such rocks, which have pore aspect ratios between 0.1 and 1.0. Dense, lithified rocks such as limestones with thin fracture porosity yield Poisson's ratios above 0.25. Hyndman's models⁶ suggest a porosity approaching 5% and a pore aspect ratio of 0.1 to 1.0 for the Weber I headwall bright spot. These properties are evident from the clear decrease in Poisson's ratio.

The bright spot may be a temporary effect of the nearby Weber I rupture. No S-P phase was observed after the event, which indicates an unusual reduction in Poisson's ratio and large pore aspect ratio. The bright spot, on the headwall of the Weber I normal fault and above it, is in the dilatational quadrant (white in fig. 2e) of the earthquake's focal sphere. The Weber I observations were all within a few days of the mainshock, on Feb. 20-22, 1990. Three months later, the bright spot is not nearly as prominent in the same region, using back-scattered P-P reflections from the Weber II events.

Static dilatation of the sediment wedge in the Weber I headwall may have lowered the pore fluid pressure there enough to cause precipitation of minerals out of pore fluids,^6,12,26 sealing cracks and drastically increasing pore aspect ratios. The temporary decrease in Poisson's ratio raised the wedge's S velocity to nearly equal that of the basement rocks bounding the plate interface, preventing the S-P phase conversion usually observed.^2,5 The simultaneous decrease in P velocity allowed observation of the strong forward-scattered P-P bright spot (figs. 2e and 3). Fluid migration into the wedge over the next three months may have increased fluid pressures gradually, leading to more equal P-P, S-S, and S-P scattering (fig. 2a,d,c). The depressurisation sealing of cracks and increase in the rigidity of the plate interface could have temporarily locked the main subduction fault immediately after the Feb. 19, 1990 Weber I event. It is conceivable that any aseismic slip deeper on the interface¹ may have led to the transfer of shear stress around the locked wedge, to trigger the Weber II thrust above.

Correspondence and requests for materials should be addressed to Louie: louie@seismo.unr.edu; http://www.seismo.unr.edu/ftp/pub/louie/weber/.