Report

A Great Earthquake Rupture Across a Rapidly Evolving Three-Plate Boundary

See allHide authors and affiliations

Science  10 Apr 2009:
Vol. 324, Issue 5924, pp. 226-229
DOI: 10.1126/science.1167476

Abstract

On 1 April 2007 a great, tsunamigenic earthquake (moment magnitude 8.1) ruptured the Solomon Islands subduction zone at the triple junction where the Australia and Solomon Sea–Woodlark Basin plates simultaneously underthrust the Pacific plate with different slip directions. The associated abrupt change in slip direction during the great earthquake drove convergent anelastic deformation of the upper Pacific plate, which generated localized uplift in the forearc above the subducting Simbo fault, potentially amplifying local tsunami amplitude. Elastic deformation during the seismic cycle appears to be primarily accommodated by the overriding Pacific forearc. This earthquake demonstrates the seismogenic potential of extremely young subducting oceanic lithosphere, the ability of ruptures to traverse substantial geologic boundaries, and the consequences of complex coseismic slip for uplift and tsunamigenesis.

Great earthquakes typically involve sudden sliding between two tectonic plates, and the largest events are located in subduction zones where an oceanic plate thrusts into the mantle below an overriding plate. In a few locations, a boundary between two oceanic plates impinges on a subduction zone, causing both plates to descend beneath the overriding plate, but at different rates and directions. This is the situation that led to the great moment magnitude Mw 8.1 Solomon Islands earthquake of 1 April 2007 [8.47°S, 157.04°E, 20:39:58.7 UTC (1)], which ruptured the megathrust between the overriding Pacific plate (PaP) and the independently subducting Australia (AuP) and Solomon Sea–Woodlark Basin (SWP) plates (Fig. 1). The inferred rupture area along the San Cristobal and New Britain trenches, as indicated by aftershocks (1), patterns of uplift/subsidence (24), and preliminary rupture analysis (5, 6), straddles the down-dip extension of the Simbo transform fault that separates the SWP from the AuP; thus, the earthquake ruptured across the SWP-AuP plate boundary. The rupture generated a large local tsunami, and about 50 lives were lost and more than 9000 people displaced.

Fig. 1.

Plate tectonic setting and recent tectonic evolution of the rupture area of the 1 April 2007, Mw 8.1 earthquake (epicenter shown by star). Plate boundary structures at ∼0.5 Ma, including the incipient Simbo transform fault and the slab window produced by the subduction of the actively spreading Woodlark spreading ridge, are shown by gray lines. Black lines show present plate boundary structures. The shaded triangular region (Ghizo Fragment) maps the extent of subducted oceanic lithosphere that may have been transferred from the SWP to the AuP with the formation of the Simbo transform at ∼0.5 Ma. Following Tregoning et al. (17), we consider the Woodlark Basin and Solomon Sea to comprise a single tectonic unit, the Solomon Sea–Woodlark Basin plate (SWP). The arrows show plate motions of the SWP and AuP relative to the PaP. The inset shows the plate motion velocity triangle in the earthquake rupture area. Locations labeled on map: SCT, San Cristobal trench; NBT, New Britain trench; B, Bougainville; M, Mbava; V, Vella Lavella; R, Ranongga; and G, Ghizo.

Before the 2007 event, this triple junction region, where the three plates meet, had low seismic activity and no record of large interplate events (7); thus, pre-event seismicity, or other available geologic and tectonic data, provided limited constraint on the subduction zone geometry. The region above the down-dip extension of the Simbo Fault is a localized region of rapid Holocene uplift (8). The age of lithosphere currently subducting along the trench varies from ∼0.5 to 3.5 million years old, and it has been speculated that such young, hot lithosphere will not produce large earthquakes. Here we describe the 2007 earthquake rupture to address how, before the earthquake, strain was distributed between the subducting and overriding plates.

Before ∼ 0.5 million years ago (Ma), the easternmost segment of the Woodlark Basin spreading ridge was subducting beneath the western margin of the Solomon Islands (Fig. 1), and the ridge-trench triple junction migrated northwesterly at ∼110 to 120 mm/year. The differences in plate subduction rates and directions produced a slab window, which today lies beneath the southern New Georgia Islands. The SWP subducts at 135 mm/year (N45°E), whereas the AuP subducts at 97 mm/year (N70°E) (8, 9). The near-total cessation of spreading on the most trenchward section of the Woodlark Basin ridge at ∼ 0.5 Ma (8, 9) and the formation of the right-lateral strike-slip Simbo transform fault caused the Ghizo lithospheric fragment (Fig. 1) to be transferred from the SWP to the AuP. The Simbo transform now accommodates ∼60 mm/year plate motion between SWP and AuP, and the new triple junction formed by the Simbo transform and the trench migrates southeastward at ∼110 mm/year (fig. S1).

Teleseismic body waves and short-arc Rayleigh waves (R1) were collected to characterize the mainshock rupture. Azimuthal variations of the R1 source time functions (STFs) (10) constrain the average rupture length and rupture velocity. Figure 2 shows the R1 STFs, which are narrow (∼50 s) in the rupture direction and broad (∼180 s) in the opposite direction (11). The systematic duration variation is consistent with unilateral rupture parallel to the trench axis with an average rupture speed of 2.5 ± 0.4 km/s. The STFs suggest a seismic moment of 2.5 × 1021 N·m (Mw = 8.2), somewhat larger than the global centroid-moment tensor (GCMT) estimate, which has an unusually large (37°) dip (5). A large aftershock is evident in the STFs ∼7 min after the mainshock. This Mw = 6.6 event initiated ∼200 km to the northwest of the mainshock hypocenter (7.17°S, 155.78°E, 20:47:31.3 UTC).

Fig. 2.

R1 Rayleigh wave effective source time functions (STFs) plotted as a function of directivity parameter, Γ (18), assuming a rupture propagation azimuth of 305°. STFs along the direction of rupture (positive Γ) are shorter in duration, indicating predominantly unilateral rupture in the direction 305°. Fitting a line along the ends of the STFs, projected to limiting values of Γ (±0.25 s/km), allows one to estimate the minimum (52 s), maximum (182 s), and average (117 s) STF durations; each of these values has about ±10 s uncertainty due to scatter in the STFs. Lines indicate the onset, average, and end times of the STFs. STFs for a large aftershock about 7 min after the mainshock are apparent.

We computed finite-fault models for relatively uniform azimuthal distributions of teleseismic P and SH waves, using a least-squares inversion (12) with prescribed fault orientation, specified rupture velocity Vr, and variable subfault rake and source time function (11). Our preferred solution is for a fault strike of 305° ± 5°, dip of 25° ± 5°, and Vr = 2.5 ± 0.4 km/s (Fig. 3). The main effect of variations in rupture velocity is a simple stretching of the slip zone (fig. S3). The basic solution is stable with respect to changes in fault dip (fig. S4), strike (fig. S5) and data subsets (fig. S6).

Fig. 3.

Summary of the finite slip model for the mainshock obtained from inversion of P and SH waves. The model shows a source focal mechanism for the average fault orientation; the rupture surface with slip vectors indicating the sense of motion of the PaP relative to the SWP and AuP, with 1-m slip contour intervals; and the moment rate function, which indicates seismic energy release as a function of time. The seismic moment estimate, Mo, gives Mw = 8.1. The rupture velocity assumed in the inversion, Vr, is 2.5 km/s.

The slip model shows two main patches of slip at shallow depth along strike, with a systematic 36° difference in rake between the patches (the near-epicentral patch has an average rake of 49 ± 11°, and the northwestern patch has a rake of 85 ± 15°). A third region of deeper slip with rake varying between values for the two shallow patches is located ∼100 km from the hypocenter. The body waves also indicate predominantly unilateral rupture, with a total duration of about 100 s and seismic moment Mo = 1.87 × 1021 N·m (Mw = 8.1). The basic slip attributes are consistent with those indicated by rapid studies (5, 6).

The distinct directions of coseismic slip in the major slip regions define separate domains of interseismic strain accumulation. The position of this transition on the slab interface (Fig. 4) is consistent with the expected down-dip extent of the Simbo fault if it developed ∼0.5 Ma. The consistency between the shallow earthquake slip directions and current relative plate motions for the two plate pairs delineates the region of the subducted Ghizo fragment that has been transferred from the SWP to the AuP. The gradual transition in fault slip between AuP-PaP and SWP-PaP relative motion directions down-dip of the subducted Simbo fault suggests that the subducted slab is still at least partially connected in that region, leading to complexity in the strain accumulation and coseismic release.

Fig. 4.

Comparison of the earthquake slip model with the plate tectonic setting of the region. (A) Map view of the rupture model, which spans the region from the subducted (extinct) Woodlark spreading center to the initial location (at 0.5 Ma) of the Simbo transform triple junction. Inset shows kinematic relations between the slip along the two main patches of moment release, resulting in approximately 2.5 to 3.0 m of shortening within the overriding Pacific plate in the region of maximum observed uplift. (B) Three-dimensional view of rupture along the subduction interface. Coseismic slip on the southern patch of moment release suggests that region is fully connected to the Australia plate. Slip within the shallow northwestern patch suggests SWP affinities. The location of expected intraplate shortening during the coseismic slip within the Pacific plate is shown.

The limited depth extent of rupture in the southeastern main rupture patch (AuP-PaP interaction) is compatible with the extreme youth of the subducted plate and the extent and location of the slab window that developed during ridge subduction before 0.5 Ma (Fig. 1). The relatively small coseismic slip along the megathrust near the subducted Simbo fault is enigmatic, particularly given the appreciable uplift associated with this earthquake (24) in the region and the similar patterns of Holocene uplift (8) above the subducted transform fault.

There are two end-member scenarios for interseismic and coseismic elastic strain accumulation and release in subduction zones. In the “slab deformation” model, all interseismic elastic deformation occurs within the subducting slab—i.e., there is no deformation of a rigid upper plate. Under this assumption, all coseismic recovery would involve displacement of the slab. In the “upper plate deformation” model, the overriding plate accumulates all of the interseismic deformation, and thus coseismic displacements all occur within the upper plate. This latter case would produce more surface uplift and potentially be more tsunamigenic than the former. Subduction zone behavior likely falls somewhere between these two end-members, and normally seismologic observations are unable to distinguish in what domain a particular subduction plate boundary resides. If the Solomon subduction regime lies on the slab deformation end of the spectrum, then the coseismic slip behavior would require sympathetic or triggered slip on the subducted Simbo fault during the event and/or loading of the Simbo Fault. Neither appears to be the case (fig. S8).

Rather, the seismic cycle behavior of the Solomon subduction regime is consistent with that expected in the upper plate deformation model. Before the event, Global Positioning System displacements for the AuP, just to the south of the rupture region, indicated unimpeded motion of Australia with respect to the Pacific (13), implying little if any strain accumulation in the subducting Australia plate. Patterns of uplift associated with the earthquake (2, 3) and the locations of maximum local tsunami run-up (3, 4, 14) are consistent with coseismic shortening in the upper plate in the transition zone between the two domains of interplate displacements (Fig. 4). The uplift, which may have been coseismic or rapid postseismic deformation, was observed on a suite of islands in the arc-trench region (24)—Ranongga, Mbava, and Vella Lavella—that overlie the location of the subducted Simbo transform and Simbo Ridge. Previously this near-trench uplift was attributed to uniform slip of 5 m over a narrow rectangular fault (2), but the more complex fault rupture pattern shown in Fig. 3 and in particular the relative minimum in fault slip right in the vicinity of the subducted Simbo transform suggest that an alternative mechanism, other than subduction of the bathymetry feature along the Simbo fault, may be needed to account for the uplift pattern.

Models of uplift driven by the observed pattern of coseismic slip show an increase in local uplift of >30% in that region (15). Similarly, the localization of maximum tsunami run-ups in that region (3, 4) may represent the amplifying effects of the localized uplift (if coseismic) and/or reflect a constructive interference of waves produced by the two distinct patches of moment release. The lack of near-trench islands elsewhere along the margin hampers our ability to assess whether the uplift and tsunami run-ups observed on Ranongga are typical along the rupture zone. The existence of Ranongga (and associated islands) and the longer-term patterns of rapid uplift, >3 mm/year in the Holocene (8), for nearby islands implies that this region represents a relative maximum in near-trench uplift. As shown schematically in Fig. 4, under the assumption that this event represents an upper plate deformation case, our rupture model for the 2007 event indicates that this rapid uplift is a region of local along-arc convergence within the Pacific forearc between that overlying the southeastern rupture patch and the forearc overlying the northwestern rupture zone. This complex rupture pattern coupled with an upper plate–dominated deformation regime increases the seismic hazard in such areas.

Supporting Online Material

www.sciencemag.org/cgi/content/full/324/5924/226/DC1

Methods

Figs. S1 to S11

Table S1

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article