Havre Trough Research Articles

Genesis of the Western Samoa seamount province: age, geochemical fingerprint and tectonics

The Samoan volcanic lineament has many features that are consistent with a plume-driven hotspot model, including the currently active submarine volcano Vailulu'u that anchors the eastern extremity. Proximity to the northern end of the Tonga trench, and the presence of voluminous young volcanism on what should be the oldest (∼5 my) western island (Savai'i) has induced controversy regarding a simple plume/hotspot model. In an effort to further constrain this debate, we have carried out geochronological, geochemical and isotopic studies of dredge basalts from four seamounts and submarine banks that extend the Samoan lineament 1300 km further west from Savai'i. 40Ar/ 39Ar plateau ages from Combe and Alexa Banks (11.1 my—940 km, and 23.4 my—1690 km from Vailulu'u, respectively) fit a Pacific age progression very well. The oldest volcanism (9.8 my) on Lalla Rookh (725 km from Vailulu'u) also fits this age progression, but a new age is much younger (1.6 my). Isotopically, these three seamounts, along with Pasco Bank (590 km from Vailulu'u), all lie within, or closely along extensions of, the Sr–Nd–Pb fields for shield basalts from the Eastern Samoan Province (Savai'i to Vailulu'u); this clearly establishes a Samoan pedigree for this western extension of the Samoan hotspot chain, and pushes the inception of Samoan volcanism back to at least 23 my. From geodetic reconstructions of the Fiji–Tonga–Samoa region, we show that the northern terminus of the Tonga arc was too far west of the Samoa hotspot up until 1–2 my ago to have been a factor in its volcanism. Young rejuvenated volcanism on Lalla Rookh and Savai'i may be related to the rapid eastward encroachment of the Trench corner. The Vitiaz Lineament, previously thought to mark a proto-Tongan subduction zone, was more likely created by the eastward propagation of the tear in the Pacific Plate at the northern end of the arc.

From rifting to active spreading in the Lau Basin – Havre Trough backarc system (SW Pacific): Locking/unlocking induced by seamount chain subduction

Associated with Pacific‐Australia plate convergence, the Lau Basin – Havre Trough is an active back‐arc basin that has been opened since ∼5.5 Ma by rifting and southward propagating oceanic spreading. Current back‐arc opening rates decrease from 159 mm yr−1 in the northern Lau Basin to 15 mm y−1 in the southern Havre Trough. Major tectonic changes occur at the transition between Havre Trough rifting and full oceanic spreading of the Eastern Lau Spreading Center (ELSC), where the oblique‐to‐trench, westward subducting Louisville Seamount Chain (LSC) sweeps southwards along the Tonga trench. New swath bathymetry, seismic reflection data, and limited rock sampling in this area constrain a tectonic and kinematic back‐arc model that incorporates the effects of LSC subduction. The ELSC, which extends southward to 24°55′S, forms a deep rift valley propagating southward through older, rifted arc basement. Present‐day seismicity and fresh and fractured pillow lavas at 23°42′S are consistent with rift valley neovolcanism. Conversely, the northern Havre Trough has low seismicity and rifted volcanic basement ridges trending 25–45° oblique to the basin axis consistent with low levels of extensional tectonism and volcanism. This latter structural fabric is interpreted as an early stage of rifting that is now “locked” due to compression on the arc exerted by LSC subduction, while in the Lau Basin such effects have passed as the LSC swept along the Tonga Trench. It is proposed that the Lau‐Havre back‐arc opening is controlled by tectonic constraints exerted at the limits of the system by the LSC subduction, which determines the southward migration of the Tonga Arc pole of rotation and associated Lau Basin opening. A discrete three‐stage back‐arc opening evolution is proposed, comprising: (1) an initial phase of back‐arc rifting along the whole length of the plate boundary, beginning at ∼6–5 Ma; (2) a subsequent phase, mostly present in the southern part of the back‐arc domain and still active in the Havre Trough, of transpression and transtension, starting at ∼4 Ma in the north, as the LSC starts to subduct and sweeps southward along the Tonga trench; and (3) a renewed opening phase in the northern segment of the back‐arc domain, with rifting and spreading, starting at ∼3.5 Ma, as subduction of the LSC along the northern Tonga trench is progressively completed.

Mantle dynamics, element recycling, and magma genesis beneath the Kermadec Arc‐Havre Trough

New geochemical and isotopic data are presented for lavas from three sites in the Havre Trough‐Lau Basin back arc and six volcanoes along the Kermadec arc. The back arc basalts range from MORB‐like to arc‐like in composition and contain a variable contribution from the underlying slab. The least contaminated MORB‐like back arc lavas from 24°–29°S are low degree partial melts of a source with Pacific MORB isotopic characteristics. A transition occurs at 30°S between the strongly depleted northern Kermadec (and Tonga) arc lavas and the mildly depleted southern Kermadec arc lavas. This transition does not correlate with changes in the back arc extension rate or width but may reflect inhibited mantle wedge replenishment behind the shallower‐dipping northern Kermadec‐Tonga slab. Northern Kermadec lavas require mixing between two components: (1) depleted Havre Trough mantle and (2) fluid derived from altered MORB crust with a slight input of sediment lead. Inter‐volcano differences in fluid compositions probably reflect local variations on the subducting slab rather than mineralogical variation in the mantle wedge. Southern Kermadec lavas require an additional component: (3) Pacific sediment melt. This sediment melt is only detected where the subduction rate is <7 cm/year, and requires thermal heating of the slab to >650°C before passing through the sub‐arc melt generation zone.

Structure and structural development of the Havre Trough (SW Pacific)

Northeast of New Zealand the Pacific plate is obliquely subducted beneath the overriding Australian plate. Major differences appear in the oblique‐convergence‐driven structures that occur along the plate boundary. While strike‐slip component of convergence is accommodated in the forearc domain to the south at the eastern North Island margin, we here substantiate that at least part of the strike‐slip component is accommodated within the back arc Havre Trough to the north offshore New Zealand. New swath bathymetry and structural data collected in the Havre back arc domain show that the basin as a whole displays oblique to basin axis structures that allow considering the basin to be the loci of extensive tectonics associated with dextral displacement of the Kermadec Arc Block relative to the Australian plate. Therefore, where the two plates are oceanic, the oblique convergence between Pacific and Australian plates is partitioned at the subduction trench for its normal‐to‐trench component and at the rear of the Kermadec Arc Block for part or all of its along‐strike component where it is intimately associated with extensional rifting within the back arc domain. The kinematic pattern of Australian‐Pacific plate boundary northeast of New Zealand that straddles a limit between continental and oceanic lithosphere allows addressing the specificity of the structural development, the deformation distribution, and the assessment of movement components that occur in a back arc basin associated with oblique convergence.

Late Quaternary sea-level and tectonic changes in northeast Fiji

The islands of northeast Fiji studied are all the emergent parts of the northern end of the Lau–Colville Ridge, a remnant island arc which has been rising for most of the Quaternary. Within the study area, investigations of coastal landforms (primarily emerged notches and shore platforms) and emerged coral reefs allow rates of tectonic movements to be calculated. The Vanuabalavu Island group is shown to have subsided during the late Quaternary while the area around is shown to have risen over the same time period. This disparity may be a result of tensional rifting associated with the continuing rotation of the Fiji Platform.

Tsunami risk assessment in the Marquesas Islands (French Polynesia) through numerical modeling of generic far-field events

Abstract. Earthquakes occurring at the Pacific Rim can trigger tsunamis that propagate across the ocean and can produce significant damages far away from the source. In French Polynesia, the Marquesas Islands are the most exposed to the far-field tsunami hazards, since they are not protected by any outer coral reef and since submarine slopes are less steep than in other islands. Between 1994 and 1996, four tsunamis have reached the bays of the archipelago, among them, the tsunami initiated by the Chilean Mw 8.1 earthquake, produced up to 3 m high waves in Tahauku Bay. Numerical modeling of these recent events has already allowed us to validate our method of resolution of hydrodynamics laws through a finite-difference scheme that simulates the propagation of the tsunamis across the ocean and computes the inundation heights (run-up) in remote bays. We present in this paper the simulations carried out to study potentially threatening areas located at the Pacific Rim, on the seismogenic Aleutian and Tonga subduction zones. We use a constant seismic moment source (that of the Mw 8.1 Chile 1995 earthquake, M0 = 1.2 1021 N.m) located at several potential epicenters, with the fault strike adapted from the regional seismotectonics pattern. Our results show that the sources chosen in the Aleutian trench do not produce large inundations in the Marquesas bays, except for the easternmost source (longitude 194° E). Sources located in the Tonga trench do not produce high amplifications either, except for the northernmost one (latitude 16° S). We also discuss the behaviour of the tsunami waves within the archipelago, and evidence contrasting responses depending on the arrival azimuths. These results show that, for a given initial seismic energy, the tsunami amplification in remote bays is highly dependent on the source location and fault strike.

Crustal structure, accretionary processes and rift propagation: a gravity study of the intermediate-spreading Valu Fa Ridge, Lau Basin

The Valu Fa Ridge is an intermediate‐spreading (full rate of 60 mm yr−1) ridge located in the Lau Basin. In 1995 this ridge was surveyed using a multidisciplinary, geophysical approach to image crust and upper mantle structure, with the aim of investigating the processes of oceanic crustal accretion in a back‐arc tectonic environment. As part of this experiment a network of gravity profiles was acquired, together with seismic, magnetic, swath bathymetry and controlled‐source electromagnetic data. Presented in this paper are the results of forward modelling of a subset of the acquired gravity profiles, two oriented ridge‐perpendicular and one ridge‐parallel, using the preferred seismic models of Turner . (1999) as a basis of initial model construction. In addition, the gravity data set in its entirety has been used to calculate the mantle and residual mantle Bouguer anomalies with the aim of investigating variability in crustal structure, both density and layer thickness, and the nature of the underlying upper mantle. Of particular interest are the overlapping spreading centre between the Central and Northern Valu Fa Ridges, where seismic modelling implies a generally thickened crust and a magma chamber located beneath the overlap basin rather than separate chambers supplying each ridge, and the propagating rift tip and associated basin‐bounding pseudo‐fault. Modelling results suggest that the pre‐ and post‐rift crusts have different compositional origins, with lower densities required > 12 km off‐axis to fit the observed free‐air gravity anomaly. The locations of the transitions into regions of lower density correspond with those of Turner . (1999) derived from seismic modelling, which in turn correspond in location to the rift‐related pseudo‐fault identified by Wiedicke & Collier (1993). Calculation and interpretation of the mantle and residual mantle Bouguer anomalies also confirms the lower off‐axis densities and indicates a general increase in crustal thickness northwards towards the overlapping spreading centre between the Central and Northern Valu Fa Ridge segments. The presence of thicker crust, corresponding to a region of negative mantle and residual mantle Bouguer anomaly, implies that the region surrounding the overlapping spreading centre has been in the recent geological past, or is presently, the site of an increased magma supply. The residual mantle Bouguer anomaly also reveals features related directly to off‐axis lateral variation in density and layer thickness, associated with the southward propagation of the Valu Fa Ridge into the Havre Trough. Modelling of the north–south variation in crustal thickness along‐axis shows that the anomaly trend can be explained simply by thicker crust beneath each of the Valu Fa Ridge segments and the overlapping spreading centre between the Northern and Central Valu Fa Ridges. The observation of segment‐centred crustal thickening is consistent with models of ridge segmentation that suggest that melt influx is located towards segment centres and that segment length is controlled by the maximum lateral extent to which this melt can flow ( Macdonald . 1988 ; Tolstoy . 1993 ). However, the overlapping spreading centre remains anomalous from these models in that it also appears to be associated with thicker crust and an increased melt supply.

Extensional fault kinematics within the Taupo Volcanic Zone, New Zealand: Soft‐linked segmentation of a continental rift system

The Taupo Volcanic Zone (TVZ) is a young (<2 Ma) arc‐backarc system in continental crust at the southern end of the Havre Trough characterised by intense volcanism and geothermal activity. An extensional fault belt (known locally as the Taupo Fault Belt) can be traced semi‐continuously within the TVZ from Ruapehu in the southwest to the Bay of Plenty coast in the northeast. Contained within the belt are numerous active volcanic centres (andesitic, basaltic, rhyolitic) and geothermal systems. Kinematic studies demonstrate that normal faulting predominates within the fault belt, which we interpret as a distributed rift system (here termed the Ruaumoko Rift System), within which rift axes are locally defined by crudely symmetric dispositions of northwest‐ and southeast‐dipping faults. The entire fault belt is partitioned along‐strike into offset rift segments defined by association with discrete rift axes. Fault dips, where measurable, are generally steep (8 > 60°). Systematic fracture sets occur in all rock types in the TVZ and generally show a strong northeast‐southwest trend, indicating a component of extensional strain distributed throughout the rock mass. Extension is predominantly orthogonal to the rift axes except in “accommodation zones” where interaction between offset rifting segments perturbs the stress field, locally enhancing permeability. The varying trends of extensional structures lie perpendicular to least principal stress trajectories in a locally heterogeneous stress field with normal dip‐slip as the predominant fault mechanism.

A three‐plate kinematic model for Lau Basin opening

We present revised compilations of bathymetry, magnetization, acoustic imagery, and seismicity data for the Lau Basin and surrounds. We interpret these data to more precisely locate the plate boundaries in the region and to derive a three‐plate kinematic model for the opening of the Lau Basin during the Brunhes Chron. Our tectonic model includes the Niuafo'ou microplate, which is separated from the Australian Plate along the Peggy Ridge‐Lau Extensional Transform Zone‐Central Lau Spreading Center and from the Tongan Plate along the newly discovered Fonualei Rift and Spreading Center. Our model shows that Australia‐Tonga plate motion between 15.5°S and 19°S is partitioned across the microplate, and it resolves the former apparent conflict between geodetic versus seafloor spreading velocities. Regionally, this implies that the angular rate of opening of the Lau Basin has been faster than that of the Havre Trough to the south and that the rapid current rates of Pacific subduction at the Tonga Trench (∼240 mm/yr at 16°S) have been sustained for at least 0.78 m.y.

Bathymetry and magnetic anomalies in the Havre Trough and southern Lau Basin: from rifting to spreading in back-arc basins

The Havre Trough and Lau Basin are active back-arc basins related to Pacific-Australian plate convergence. These back-arc basins provide insight into the evolution of rifting through various stages, including a final stage of oceanic spreading. We carried out a swath bathymetry and magnetic anomaly survey focused on the less-studied Havre Trough and southern Lau Basin. The western portion of the whole Havre Trough is deeper and more thickly sedimented whereas shallow bathymetry and volcanic edifices are dominant in the east. We suggest that crustal rifting progresses asymmetrically with respect to the trough axis. The active tectono-magmatic zone migrates trenchward during ongoing widening. Rifting along normal faults trending in the longitudinal trough strike occurs in the southern Havre Trough. Volcanism is restricted to a line along the normal faults. Circular or elongated knolls in the northern Havre Trough indicate that en echelon type segmented rifting of short strike length and successive volcanic intrusion takes place. Volcanism in this area is dominated by diapir-like magma intrusion into an extended and thinned arc crust. The southernmost tip of the spreading center of Lau Basin is defined by the deep graben, striking at N20°E, located at 24°00′S, 177°10′W.

Neolepas osheai sp. nov., a new deep‐sea vent barnacle (Cirripedia: Pedunculata) from the Brothers Caldera, south‐west Pacific Ocean

Neolepas osheai sp. nov. is described from collections made at depths of 1200–1700 m within the Brothers Caldera, a site of hydrothermal venting and active volcanism in the southern portion of the Havre Trough. N. osheai sp. nov. is similar to other Neolepas in that it is hermaphroditic, has very specialised cirri, and occupies a “deep water” hydro‐thermal vent habitat. It differs from other species in its trophic structures and the nature of the peduncle. This new record increases the number of described Neolepas species to four, three of which are living. The other, a fossil, is however tentatively ascribed to the genus and is known in rocks of Early Jurassic age. All are from the Southern Hemisphere.

Bathymetry of the Tonga Trench and Forearc: a map series

Four new bathymetric maps of the Tonga Trench and forearc between 14S and 27S display the important morphologic and structural features of this dynamic convergent margin. The maps document a number of important geologic features of the margin. Major normal faults and fault lineaments on the Tonga platform can be traced along and across the upper trench slope. Numerous submarine canyons incised in the landward slope of the trench mark the pathways of sediment transport from the platform to mid- and lower-slope basins. Discontinuities in the trench axis and changes in the morphology of the landward slope can be clearly documented and may be associated with the passage and subduction of the Louisville Ridge and other structures on the subducting Pacific Plate. Changes in the morphology of the forearc as convergence changes from normal in the south to highly-oblique in the north are clearly documented. The bathymetric compilations, gridded at 500- and 200-m resolutions and extending along500 km of the landward trench slope and axis, provide complete coverage of the outer forearc from the latitude of the Louisville Ridge-Tonga Trench collision to the northern terminus of the Tonga Ridge. These maps should serve as a valuable reference for other sea-going programs in the region, particularly the Ocean Drilling Program (ODP) and the National Science Foundation MARGINS initiative.

Crustal structure and seismicity of the Havre Trough at 26°S

An ocean bottom seismographic (OBS) experiment was carried out to investigate the present‐day tectonics at the northern end of the Havre Trough, a backarc basin in the southwestern Pacific. Hypocenter distribution shows no seismicity within 50 km of the OBS array but two active regions outside the Trough; one is to the northeast of the array at a bathymetric high between the Lau Basin and Havre Trough; the other is to the southeast of the array beneath the landward slope of the Kermadec Trench. P wavespeed structure obtained from the refraction survey is characterized by a thinned island arc crust with a thickness of 9–10 km. Many volcanic intrusions shown in reflection profiles that support extension tectonics are covered by undisturbed sediments. These suggest little present activity related to rifting and/or spreading.

Morphology and history of the Kermadec trench–arc–backarc basin–remnant arc system at 30 to 32°S: geophysical profile, microfossil and K–Ar data

Knowledge of the time span of arc activity, essential for correct tectonic reconstructions, has been lacking for the Kermadec arc system, but is supplied in this paper through study of microfossils contained in dredge samples, and K–Ar ages on dredged basalt clasts. The Kermadec system at south latitudes 30 to 32° in the southwest Pacific comprises from west to east the Colville Ridge (remnant arc), Havre Trough (backarc basin), Kermadec Ridge (active arc) and Kermadec Trench (site of west-dipping subduction of Pacific plate lithosphere beneath the Australian plate). Data are presented from two traverses (dredge, magnetic, single-channel seismic) across the whole system. An important transverse tectonic boundary, the 32°S Boundary, lies between the two traverse lines and separates distinct northern (32–25°S) and southern (32–36°S) sectors. The northern sector is shallower and well sedimented with broad ridges and a diffuse backarc basin. The southern sector is deeper with narrow ridges and steep escarpments facing inwards to a little-sedimented, rifted backarc basin. The Kermadec Ridge slopes smoothly trenchward to a mid-slope terrace (forearc basin) with minor sediment fill at 5–6 km water depth. A steeper (10–24°) and more rugged lower trench slope is mantled with New Zealand-sourced rhyolitic vitric mud diamictons containing locally derived mafic volcanic clasts; one clast is of late Miocene age (K–Ar age 7.84 Ma). The arc (Kermadec Ridge) is capped by active volcanoes; very young K–Ar ages (<150 ka) from basalts dredged from the 40-km wide transition zone between the Kermadec Ridge and the Havre Trough, north of the 32°S Boundary, support the concept of arc retreat to the southeast. South of the 32°S Boundary the Kermadec Ridge deepens and narrows, makes a left-step of 10 km, and presents a 2.8-km scarp face to the Havre Trough; K–Ar ages from dredged basalt clasts range from 1.25 to 2.04 Ma and indicate exposure of older arc rocks. The deepest and most sedimented portion of the backarc basin lies on the western side, both north and south of the 32°S Boundary, and the centre of basin opening is inferred to lie on the eastern side. There is foundered arc material and former hydrothermal activity in the centre of the rifted basin. The remnant arc (Colville Ridge) has subsided approximately 700 m. On the northern profile it is broad and has a perched sedimentary basin at 1 km water depth on the eastern flank. On the southern profile it is narrow and presents a 2.5-km scarp face to the Havre Trough. Arc substrate rocks are exposed on both ridges. Derived microfossils, sedimentary clasts with fossil-based depositional ages in the late Miocene and Pliocene, clasts of hypabyssal and plutonic rocks, and dated basalt clasts as old as 2 Ma, together indicate continuing collapse and surficial reworking on both ridges. Derived microfossils establish that the Colville and Kermadec Ridges have existed (initially as one ridge) since at least the earliest Miocene; by inference, ridge volcanism has been active since the same time, about 25 Ma. Rare older microfossils may indicate earlier existence of the ridge. A 25-Ma inception of arc volcanism is synchronous with contiguous arc sectors to the north (Tonga) and south (New Zealand). We have no new data on the age of the age of the Havre Trough, which is generally considered to be less than 5 Ma. A seamount entering the Kermadec Trench is early Eocene or older, and a ridge/seamount in the South Fiji Basin (west of the Colville Ridge) is middle Miocene or older.

Fundamental leaking mode (PL) propagation along the Tonga--Kermadec--Hikurangi--Macquarie margin

High-quality digital data from the temporary Leeds Tararua broad-band array, North Island, New Zealand and the station SNZO at Wellington often record long-period oscillations within the body wave train from large, shallow events in the Tonga–Kermadec and Macquarie seismic zones at regional distances (8°–25°) along the Australasian/Pacific plate boundary. These arrivals are dispersed typically from 40 s period to 25 s and exhibit prograde elliptical polarization, which is diagnostic of the regional leaking mode PL. Theoretical dispersion curves are generated with the simplest structure of a slow layer overlying a faster half-space. We have analysed the group velocity dispersion characteristics of the recorded waveforms and successfully modelled it as purely fundamental mode propagation in a low-velocity waveguide. Our best-fitting structure north of New Zealand consists of a low-velocity layer within the mantle (β=4.1–4.3 km s-1) with a thickness of 32–34 km overlying a typical mantle structure. A seismicity study for the period 1976–1992 at SNZO shows that those events generating PL are shallow (10–60 km), with the highest concentration along the Kermadec arc, suggesting that the low-velocity layer is connected with this feature. We suggest that a small degree of partial melt within the uppermost mantle is responsible for creating a low-velocity channel within the region of backarc spreading in the Havre Trough. Dispersion curves derived from waveforms travelling northwards from events on the Macquarie Ridge give a similar structure with a slow layer 31 km thick with β=3.8 km s-1. This is consistent with the continental crust of South Island, which lies along most of the event–station paths.

The Denham Caldera on Raoul Volcano: dacitic volcanism in the Tonga–Kermadec arc

Denham Caldera, on Raoul Volcano in the Tonga–Kermadec arc, is a simple collapse structure. It is 6.5×4 km 2 across, >0.3 km deep, represents an erupted magma volume of 8–16 km 3, and is considerably larger than previously estimated. Caldera formation was contemporaneous with an eruption of homogeneous dacitic magma at 2.2 ka. Tephra from this event is preserved as a pumice fall and pyroclastic flow sequence up to 120 m thick on Raoul Island. Offshore, this sequence forms a submarine deposit diminishing in thickness from 120 m near the northern caldera rim to 80 m at a distance of 3.5 km, where it rests upon an erosional planation surface of 17–20 ka age incised into older volcaniclastic rocks. Both the caldera volume and ejecta distribution of the 2.2 ka eruption are comparable to the caldera volume and ejecta distribution of the 1883 Krakatau eruption. Resurgent volcanism has built at least six submarine pyroclastic cones along a north–northeast trending lineament crossing Denham Caldera. Redistribution of pumiceous dacite from these cones, including ephemeral islands that emerged during the 1814 and 1870 eruptions, has contributed to recent shoaling in Denham Bay. A smaller eruption occurred in 1964–65, and hydrothermal activity persists at several sites along the lineament. Voluminous felsic volcanism at intra-oceanic arcs is usually associated with crustal extension, and was unexpected in the Tonga–Kermadec arc. However, because the tectonic fabric within the active back-arc Havre Trough is oblique to the Kermadec arc, rifts within the trough apparently propagate to the volcanic front and transfer extensional strain to the arc, thereby promoting the ascent of felsic magma. The orientation of Denham Caldera, as well as that of Macauley Caldera 110 km further south, thus reflect the stress field of the Havre Trough and not the predominant field of the arc.

Crust and upper mantle heterogeneities in the southwest Pacific from surface wave phase velocity analysis

Direct earthquake-to-station Rayleigh and Love wave data observed on high gain broadband records are analyzed in order to improve the lateral resolution of the uppermost mantle in the southwest Pacific region. We used data of nine permanent Geoscope and Iris stations located in the southern hemisphere and nine other stations as part of two temporary networks, the first one installed in New Caledonia and Vanuatu (hereafter named Cavascope network) by ORSTOM and the EOST from Louis Pasteur University in Strasbourg (France) and the second one installed in the Fiji, Tonga and Niue islands (hereafter named Spase network) by Washington University in St. Louis (USA). In order to collect more significant details on the surficial structures, we included the analysis of short period waves down to 8 s. A multiple frequency filtering technique has been used to recover phase velocities of Rayleigh and Love waves for selected earthquakes with magnitude greater than 5.5 and with known centroid moment tensor (CMT). About 1100 well-distributed seismograms have been processed in the period range 8–100 s and corrections for topography and water depth have been applied to the observed phase velocities. The geographical distribution of phase velocity anomalies have then been computed using the tomographic method developed by Montagner [Montagner, J.P., 1986a. Regional three-dimensional structures using long-period surface waves. Ann. Geophys. 4 (B3), 283–294]. Due to a poor knowledge of dense, well-distributed, crustal thickness values and corresponding velocity models, we did not perform or speculate on the construction of an S-wave 3D velocity model; therefore, we limited this study to the interpretation of the phase velocity distribution. The location of phase velocity anomalies are well determined and the deviations are discussed within the framework of the geological context and compared with other tomographic models. At long periods, from 40 s to 100 s, our results agree well with most of previous studies: the tomographic imaging shows a large contrast between low and high phase velocities along the Solomon, New Hebrides and Fiji–Tonga trenches. The lowest phase velocity anomalies are distributed beneath northern and southern Fiji basins and the Lau basin (corresponding to the volume situated just above the dipping slabs), whereas the highest values are displayed beneath the Pacific plate and the eastern part of Indian plate downgoing under the North Fiji basin. At shorter periods, our results show that the phase velocity distributions are well correlated with the large structural crustal domains. The use of local temporary broadband stations in the central part of the studied area gives us the opportunity to observe surface waves showing well-dispersed trains, allowing extended velocity measurements down to 8 s although aliasing due to multipaths become important. The continental regions (Eastern Australia, New Guinea, Fiji islands and New Zealand) show low velocities which are likely due to thick continental crust, whereas the Tasmanian, D'Entrecasteaux, and the Northern and Southern Fiji basins are characterized by higher velocities suggesting thinner oceanic crust. Additional analysis including the anisotropic case and S-wave velocity inversion with depth is in progress.

ORIGIN OF HIGHLY REFRACTORY PERIDOTITES: IMPLICATIONS FOR DIVERSITY OF HIGH-MG ANDESITE MAGMA GENESIS

Highly refractory peridotites are defined here as peridotites more depleted in melt components than abyssal peridotite, i.e. peridotites with spinel showing Cr# (Cr/(Cr+Al) atomic ratio) higher than 0.6. These correspond to the type III peridotite defined by Dick and Bullen (1984) and are reported from forearc regions, such as Izu-Ogasawara, Mariana and Tonga trenches (e.g. Bloomer and Hawkins, 1983; Bloomer and Fisher, 1987; Ishii et al., 1992) and from the Kamuikotan belt, the central axial zone of Hokkaido, northern Japan (e.g. Katoh and Nakagawa, 1986; Makita and Arai, 1997). These peridotites are, however, not common in the upper mantle and their genesis has not been explained thoroughly. We discuss the genesis of highly refractory peridotite from three complexes, the Takadomari, Iwanaidake and Nukabira, in the Kamuikotan belt, Japan, and its bearing on the process of high-Mg andesite magma genesis. The Kamuikotan belt is a tectonic melange zone composed of metamorphic rocks, ultramafic rocks, greenstones and sedimentary rocks. The metamorphic rocks are of typical high-pressure/low-temperature type and ophiolitic rocks which recorded low-pressure ocean-floor metamorphism are distributed in the Horokanai area, northern part of the belt (Ishizuka et al., 1983). Ultramafic rocks form the basal member of the ophiolite in the northern part and are exposed as peridotite complexes of various size in the southern part of the Kamuikotan belt. They are generally in fault contact with surrounding Kamuikotan metamorphic rocks, sediments and mafic volcanic rocks (e.g. Niida and Katoh, 1978). The peridotite complexes mainly consist of harzburgite and dunite which suffered serpentinization to various extent. The Takadomari complex from the northern part, and the Iwanai-dake complex from the southern part of the belt consist of harzburgite, dunite and small amounts of orthopyroxenite. The Nukabira complex (southern part of the belt) consists of lherzolite and harzburgite with dunite and pyroxenites. Podiform chromitite deposits sometimes accompany dunite in the latter two complexes. Primary hydrous minerals are sometimes included in chromian spinel of the former two complexes. The Cr# of chromian spinel and Fo (forsterite) content of olivine in peridotites from the Takadomari complex are high (Cr#=0.64-0.92, Fo=91.9- 94.0), and are higher in dunite than in harzburgite on average. The Cr# and Fo are also high (Cr#=0.43-0.87, Fo=90.8-93.5) in the Iwanai-dake complex and have wider ranges (Cr#=0.18-0.86, Fo=89.0-93.2) in the Nukabira complex. The Cr# of spinel and Fo of olivine in dunite complexes are sometimes similar to or even lower than the values in harzburgite from Iwanai-dake and Nukabira complexes. The mineral chemistry of dunites in the Iwanai-dake and Nukabira complexes depends on their thickness within harzburgite or lherzolite bodies. Thicker dunite layers tend to show higher Cr# of chromian spinel and Fo content of olivine. In addition to this, there are clear systematic variations in lithology across dunite layers and surrounding peridotites. In the case of thick dunite layers, the Cr# of chromian spinel and the Fo content of olivine gradually increase from the surrounding peridotite to the central part of the dunite layer (Cr#>0.6 and Fo>91.5 in dunite). In contrast to this fact, these values in thin dunite layers are similar to the surrounding peridotite, lherzolite or harzburgite (Cr#=0.4- 0.6 and Fo=89-91). These values are generally lower in thinner dunite layers than in the thick ones compared at the center. The harzburgites and dunites in the Takadomari complex are interpreted to be a series of refractory residue after extraction of high-Mg andesite magma; partial melting of the fertile peridotite possibly occurred under hydrous conditions. Harzburgite and lherzolite are mostly simple residue in the Iwanai-dake and Nukabira complexes but their dunites are not simple residue because the Cr# and Fo content are not systematically higher in the dunite than in the harzburgite (e.g. Arai, 1987; 1994). The dunite may be produced by a reaction between wall peridotite and a melt formed at higher pressures (e.g. Quick, 1981; Fisk, 1986; Kelemen, 1990). The differences in thickness and mineral chemistry of dunite layers are due to the difference of the melt/wall peridotite volume ratio. Supplying a large amount of melt caused formation of thick dunite layers and generated high-Mg andesitic magma by the reaction process.

Simplification of the Southwest Pacific Neogene arcs: inherited complexity and control by a retreating pole of rotation

Abstract Neogene arc activity in the Southwest Pacific began simultaneously at 25 Ma on three differently oriented sectors. Two sectors (Norfolk-Three Kings and Colville Ridges), aligned north and north-northeast, were near-orthogonal to subduction, while the intervening Northland-Reinga sector (northwest aligned) was strongly sinistral-oblique and produced twin parallel arcs of differing compositions. This complexity was inherited from a complex late Eocene-Oligocene margin that was convergent but amagmatic, and was due primarily to its proximity to the Pacific-Australia pole of rotation. The inception of arc magmatism at 25 Ma was triggered by a 20° increase in convergence angle between the north moving Australia Plate and the northwest moving Pacific plate (from 44° to 65°), and an increase in convergence rate from c. 20 to 30–40 mm yr −1 . Between 25 and 15 Ma three subduction zones were required. The Pacific Plate was subducted at the Colville Arc, and a South Fiji Basin Plate was subducted at the Norfolk-Three Kings Arc in a manner analogous to the present-day Philippine Sea Plate located between the Mariana Arc and the Japan Trench. During this interval the Norfolk Basin opened as a back-arc basin, while the Three Kings Arc moved eastwards between two fracture zones. Eastward movement of the Three Kings Arc ( c. 350 km) was sufficient to drive a magma-producing subduction rate of c. 35 mm yr −1 independently of Pacific Plate convergence which was therefore taken up at the Colville Subduction Zone. Sinistral-oblique subduction beneath the Northland-Reinga sector was also driven by Pacific Plate convergence of c. 30 mm yr −1 . At 15 Ma the Colville Ridge extended 400 km southwestwards across North Island, New Zealand, to become a Colville-Coromandel-Taranaki Arc, and the two other arc segments died; this was the main simplification event, and it was probably a response to the progressive southeasterly retreat of the pole of Pacific-Australia rotation located to the south of New Zealand and the consequent increase in convergence rate. Between 15 and 5 Ma there was stability and no back-arc basin formed. A big change at 5Ma was probably driven by further movement of the pole, this time to the southwest, and an increase in convergence rate across New Zealand. In response, the arc split to form the still-active Lau-Havre Back-arc Basin and the active Tonga-Kermadec-Taupo Arc, and the New Zealand mountains formed. Factors remaining to be resolved are the time of collision of the oceanic Hikurangi Plateau, which has entered the New Zealand Subduction Zone, and the timing of dextral displacement(s) of the forearc to its present location along the eastern margin of the North Island. Arc simplification was accompanied by further fragmentation and dispersal of New Zealand continental crust which had already been fragmented by late Cretaceous extension. The Vening Meinesz Fracture Zone, 900 km long, originated as an Eocene near-transform sinistral boundary, and had a complex Cenozoic record of strike-slip and transpressional activation and reactivation, with both sinistral and dextral sense.

Modeling the Tonga slab: Can travel time data resolve a metastable olivine wedge?

We present the results of detailed modeling of the Tonga slab with the goals of determining whether high‐resolution travel time data (1) can be fit by simple thermal and petrological slab models and (2) can resolve a metastable olivine wedge at depths greater than the equilibrium olivine‐spinel phase boundary. We model arrival times recorded by a 1000 km line of 23 ocean bottom seismometers (OBS) and island broadband seismic stations extending from the Lau backarc basin, across the Tonga trench and onto the Pacific plate. The data consist of 388 local, P wave travel times from 17 deep and 3 intermediate earthquakes recorded during the 3‐month OBS deployment in late 1994. We locate the events using both local and teleseismic arrival times, and apply a relocation operator to the theoretical travel times to simulate the biases introduced in the data by locating the events with a reference Earth model. The modeling consists of grid searches using a three‐dimensional finite difference algorithm to compute local, first arriving travel times for equilibrium and metastable P wave velocity models constructed from thermal, mineralogical, and morphological constraints. The travel time anomalies are well fit by standard slab thermal models and P velocity temperature derivatives of −0.4 to −0.3 ms−1°C−1. Forward calculations indicate that the presence of a metastable olivine wedge has a subtle effect on the travel times due to the tendency of first arriving waves to avoid the low‐velocity region. Wedge velocity models provide a slightly better fit to the data than equilibrium models, but F tests indicate the improvement is not significant at the 95% level. Our results suggest that providing direct seismological evidence of a wedge of metastable olivine in subduction zones will require either waveform modeling or the observation of later arriving phases created by the depressed phase boundary.