Seismic reflection and refraction data from Hikurangi Plateau (southwestern Pacific Ocean) require a crustal thickness of 10 ± 1 km, seismic velocity of 7.25 ± 0.35 km/s at the base of the crust, and mantle velocity of 8.30 ± 0.25 km/s just beneath the Moho. Published models of gravity data that assume normal crust and mantle density predict 5–10-km-thicker crust than we observe, suggesting that the mantle beneath Hikurangi Plateau has anomalously low density, which is inconsistent with previous suggestions of eclogite to explain observations of high seismic velocity. The combination of high seismic velocity and low density requires the mantle to be highly depleted and not serpentinized. We propose that Hikurangi Plateau formed by decompression melting of buoyant mantle that was removed from a craton root by subduction, held beneath 660 km by viscous coupling to slabs, and then rose as a plume from the lower mantle. Ancient Re-Os ages from mantle xenoliths in nearby South Island, New Zealand, support this hypothesis. Erosion of buoyant depleted mantle from craton roots by subduction and then recycling in plumes to make new lithosphere may be an important global geochemical process.


Hikurangi, Manihiki, and Ontong Java Plateaus (southwestern Pacific Ocean) were emplaced coevally and possibly contiguously at ca. 120 Ma (Mahoney et al., 1993; Coffin and Eldholm, 1993). If they formed together, they constitute the largest igneous province preserved at Earth’s surface (Davy et al., 2008; Taylor, 2006). Chemical and petrological signatures of basalts from the plateaus show similarities and differences, indicating a compositionally heterogeneous plume source (Fitton and Godard, 2004; Golowin et al., 2018; Hoernle et al., 2010; Mortimer and Parkinson, 1996; Timm et al., 2011). The “Greater Ontong Java Event” included volcanism in the Nauru, East Mariana, Lyra, and possibly northwest Central Pacific Basins, and thus covered ∼1% of Earth’s surface area (Coffin and Eldholm, 1994; Mahoney et al., 1993), but there is no consensus on what caused the event: mantle plumes (Mahoney et al., 1993), mid-ocean ridges, eclogite recycled by mantle circulation (Anderson, 2005), and bolide impacts (Ingle and Coffin, 2004) have all been invoked. We present new geophysical data from Hikurangi Plateau, the southernmost part of this large igneous province (LIP), where we are able to resolve crustal and upper-mantle structure, and hence provide new light on the origin of the plateau (Fig. 1).

Hikurangi Plateau is actively subducting beneath North Island, New Zealand, and can be recognized at depth as a geophysical anomaly on seismic tomography images (Reyners et al., 2011). Seismic reflection images (Fig. 2) and consideration of the stratigraphic and magmatic history of New Zealand reveal that it was previously subducted beneath the Chatham Rise, before 105–85 Ma, which is when Gondwana margin subduction ceased in this region (Bland et al., 2015; Bradshaw, 1989; Davy et al., 2008; Sutherland and Hollis, 2001).


We collected wide-angle seismic data along a 260-km-long transect during the Seismic Array Hikurangi Experiment (SAHKE) (Henrys et al., 2013). The initial data set was acquired in 2009–2010 aboard M/V Reflect Resolution with a 98 L airgun source. Multichannel reflection data were acquired as part of the PEG09 survey (SAHKE-1) with 10 km streamer length. The SAHKE line was then re-acquired with 100 m shot spacing and with 16 ocean-bottom seismometers (OBSs) deployed at 5 km spacing across the trench slope and undeformed part of the Hikurangi Plateau. A further 10 OBS records were acquired in 2017 aboard R/V Marcus G Langseth using a 108 L airgun source, as part of the Seismogenesis at Hikurangi Integrated Research Experiment (SHIRE).

OBS data were processed using a minimum-phase Butterworth frequency filter (ramp frequencies 2–4 and 8–20 Hz) and spatial trace amplitude balancing. In the wide-angle data, we recognized three different phases that we interpret as (1) refractions through sedimentary (Psed) and crustal layers (Pg), (2) refractions through the mantle (Pn) of Hikurangi Plateau, and (3) a reflected phase (PmP) from the base of Hikurangi Plateau crust (Moho). The collective data set comprises 26 receiver-gathers, from which ∼34,000 travel-time interpretations were made. The picking uncertainties associated with each of these phases were visually estimated to be 40–80 ms (Psed and Pg), 100 ms (Pn) and 120–150 ms (PmP) respectively.

We constructed our model of crustal structure by first performing first-arrival travel-time tomography (Fujie et al., 2006) along the western portion of the transect (0–80 km) where OBSs were closely spaced. We matched the depth section of the model to the two-way time of the reflection section, so that we assigned a velocity to each layer bounded by a significant interface picked on the seismic reflection section. We extended each interface east along the seismic reflection profile (Fig. 2A) and converted two-way time into depth assuming that similar velocities exist along each interface. Using a forward modeling approach (Zelt and Smith, 1992; Zelt, 1999), the layers were adjusted to replicate first arrivals interpreted from OBSs deployed across Pegasus Basin. After each adjustment, our forward model was converted to two-way time and compared with the multi-channel seismic (MCS) section to ensure consistency with the structural constraints imposed (Fig. DR1 in the GSA Data Repository1). Our final step was to use reflected phases from the base of the crust and refracted phases through the mantle to develop an optimized model for the thickness of Hikurangi Plateau crust and the seismic velocity of the underlying mantle (Zelt and Smith, 1992; Zelt, 1999). The final seismic velocity model fits 30,000 observations with root-mean-square (RMS) misfit of 55 ms (Figs. DR3–DR8).

The trade-off between crustal velocity, crustal thickness, and mantle velocity was analyzed by calculating how RMS misfit varied with adjustments to parameters. The crust of Hikurangi Plateau was modeled as two layers, with a high velocity gradient (P wave velocity, Vp, of 5.6–6.5 km/s) in the upper crust overlying a more gradual velocity gradient through the lower crust. Crustal thickness and crustal velocity were altered by adjusting the Moho depth and seismic velocity at the base of the crust. The crustal velocity shown in Figure DR2 is the value just above the Moho. Mantle velocity was defined immediately beneath the Moho, with a velocity gradient prescribed by linearly increasing mantle velocity by 0.1 km/s between the Moho and 30 km depth. Slices through the parameter volume near the optimal solution are shown in Figure 2D and Figure DR2. RMS misfit values are normalized by dividing by the minimum value. The 95% confidence region (determined by an F-test) shows that the crustal thickness is 10 ± 1 km, crustal velocity is 7.25 ± 0.35 km/s, and mantle velocity just beneath the Moho is 8.30 ± 0.25 km/s.


The crustal structure of Hikurangi Plateau has previously been inferred from modeling of gravity and topography, assuming that it is a basaltic large igneous province underlain by mantle with normal density (Davy and Wood, 1994). Initial results suggested that crust of the plateau near the Chatham Rise (Fig. 1) was ∼15 km thick, and subsequent work using constraints from high-quality seismic reflection data indicated that it could be as thick as 20 km (Davy et al., 2008). The crustal thickness we find of 10 ± 1 km is significantly less than that of these models, even if 1–2 km of low-velocity (∼4.5 km/s) overlying material (beneath horizon H in Fig. 2) is included. Hence, combined gravity and topography data require the negative mass anomaly, modeled as crustal root by Davy et al. (2008), to be located in the mantle. Our geophysical estimate of crustal thickness is better constrained, but consistent with estimates for Hikurangi Plateau subducted beneath southern North Island (Henrys et al., 2013; Tozer et al., 2017).

High velocities of 8.3–8.9 km/s are found at relatively shallow depth (30–60 km) beneath onshore New Zealand (Galea, 1992; Haines, 1979; Kayal and Smith, 1984; Reyners et al., 2006). On this basis, it has been proposed that the base of Hikurangi Plateau is composed of eclogite, and hence the base of the crust may reach as much as 65 km depth beneath central South Island (Reyners et al., 2011), where Hikurangi Plateau was subducted during the Cretaceous. Our results are not consistent with the hypothesis of a high-Vp eclogite lower crust. We observe moderately high Vp values, and it may be that Vp is higher at greater depth, but eclogite cannot be the explanation for two reasons: (1) the density of eclogite (∼3.6 g/cm3) is higher than that of normal mantle (∼3.3 g/cm3), which is inconsistent with gravity models (Davy et al., 2008); and (2) the model depth to the top of high-Vp material (20–24 km) is shallower than that predicted for the basalt-eclogite phase transition (>1.1–1.5 GPa; >30–40 km) (Hacker et al., 2003).

We interpret the combination of high Vp and low density, which is what exists beneath the part of Hikurangi Plateau that we have surveyed, as mantle that is highly depleted by melt extraction. A similar pattern of buoyant high-Vp mantle is observed beneath continents (Poupinet et al., 2003) and understood from xenoliths and laboratory experiments to be depleted peridotite, i.e., harzburgite or dunite. High degrees of depletion beneath ancient cratons is likely because the mantle was hotter and degrees of melting were greater earlier in Earth history (Carlson et al., 2005; Pollack, 1986). Xenoliths from nearby South Island are predominantly of peridotite rather than eclogite composition (McCoy-West et al., 2013; Scott et al., 2014), and depleted mantle is inferred beneath the eastern margin of Manihiki Plateau from basalt geochemistry (Ingle et al., 2007).


We propose that buoyancy associated with ancient depleted mantle played a role in driving Cretaceous mantle upwelling, and that adiabatic decompression caused melt extraction that was limited by previous depletion to form the observed thickness of Hikurangi Plateau.

Hikurangi Plateau crust was subducted beneath Chatham Rise and southeastern South Island during the Cretaceous (Bradshaw, 1989), and hence we expect that lower crust and mantle in eastern South Island may be similar to what we observe on our geophysical transect. The oldest continental rocks in nearby eastern South Island are Carboniferous (<360 Ma) (Mortimer, 1995, 2004), but Re-Os isotope ages from mantle xenoliths are 1.6–2.3 Ga (McCoy-West et al., 2013). A province of craton-like lithospheric mantle is recognized beneath South Island (Scott et al., 2014). We propose that it is this material that has anomalously high seismic wave velocities (Reyners et al., 2011).

How did ancient cratonic mantle get beneath South Island if the crust was only assembled after 360 Ma? Geochemical analyses of mantle xenoliths from Antarctica, Australia, and Zealandia have been used to suggest that depleted subcontinental lithospheric mantle can be eroded from ancient cratons, circulated through the asthenosphere, and then returned back to form new subcontinental lithospheric mantle (Liu et al., 2015).

We suggest that subcontinental cratonic mantle was eroded from Gondwana lithosphere by subduction processes during the Phanerozoic. It was held beneath 660 km depth by viscous coupling to subducted slabs, and then when a break in subduction occurred, it rose as a Cretaceous plume that variably impacted one or more spreading ridges (Taylor, 2006) and created the LIP that includes the Hikurangi Plateau fragment (Fig. 3). The mantle reservoir apparently did not mix substantially geochemically or isotopically with surrounding asthenosphere.

There is evidence from seismic tomography and the geoid that breaks in subduction can give rise to upwelling mantle plumes rooted at 660–1200 km depth (Spasojevic et al., 2010b); and there is evidence from dynamic topography, the geoid, and seismic tomography that such a Cretaceous system existed beneath southeastern South Island, and persists today beneath Antarctica (Sutherland et al., 2010).

We propose that subduction can erode melt-depleted subcontinental lithospheric mantle, but that the inherent buoyancy of this material stops it from reaching the base of the mantle. Instead, it accumulates beneath the 660 km phase transition, where viscosity increases by a factor of 80–100 (Spasojevic et al., 2010a; Steinberger and Calderwood, 2006), and viscous coupling to adjacent slab remnants is sufficient to hold it there. The buoyant mantle is released when there is a break in the supply of subducted slab (Spasojevic et al., 2010b; Sutherland et al., 2010). We suggest that this was a primary factor involved in the formation of Hikurangi Plateau, which involved a plume rooted at 660–1200 km depth. Interaction between the plume and an active ridge produced the thick crust of Ontong Java Plateau and thinner crust of Hikurangi Plateau, which was further from the spreading center. Erosion of buoyant depleted mantle from craton roots by subduction and then recycling in plumes to make new lithosphere may be an important global geochemical process and may create instabilities that influence large-scale mantle flow.


We appreciate the efforts of the captains and crew of the M/V Reflect Resolution, the R/V Marcus Langseth, and the R/V Tangaroa during the SAHKE and SHIRE cruises. We also thank the technical staff from the Earthquake Research Institute (ERI) at the University of Tokyo, GNS Science (New Zealand), and the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) for their contribution to these projects. This work was funded by the governments of New Zealand, Japan, and the United States (U.S. National Science Foundation grant OCE-1615815 to Van Avendonk and Bangs), and by individual institutions of JAMSTEC and ERI. We thank the editor, Chris Clark, and anonymous reviewers for constructive feedback.

1GSA Data Repository item 2019287, illustrated model fits to representative ocean-bottom seismometers, is available online at http://www.geosociety.org/datarepository/2019/, or on request from editing@geosociety.org. All SAHKE (http://dx.doi.org/10.21420/S3H9-PQ92) and SHIRE (http://dx.doi.org/10.21420/TQ67-8F60) data presented in this study are available upon request.
Gold Open Access: This paper is published under the terms of the CC-BY license.