Intraplate volcanic ridges

³Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York, USA, scw@ldeo.columbia.edu

Intraplate seamount chains and volcanic ridges are the surface expression of the dynamic interaction between the oceanic asthenosphere and a mobile tectonic plate. One class of such volcanic features is associated with gravity lineations west of the East Pacific Rise that are aligned in the direction of Pacific plate motion in the hotspot coordinate frame [e.g., Sandwell et al., 1995]. A group of such seamount chains and ridges lies in the south Pacific on 3 - 8 Ma seafloor. They include the well-developed Sojourn ridge and the actively forming Hotu Matua volcanic complex (Figure 1). These two linear volcanic trends, and the Pukapuka ridge to the south [Ed: see also Pukapuka page], are parallel to plate motion and spaced about 200 km apart, roughly consistent with the dominant wavelength observed in free-air gravity lineations [e.g., Haxby & Weissel, 1986; Cazenave et al., 1992]. Seamounts and seamount chains are direct indicators of melt in the asthenosphere and its transport through the lithosphere, but the origin of the gravity lineations, the mechanism of melt generation, and the relationship of the volcanic activity to the gravity lineations are debated.

Figure 1: Bathymetric map of the study area, primarily from the GLIMPSE and MELT experiment surveys. Inset indicates the experiment location in the south Pacific. The Southern Cross seamount (SC Smt) was the tallest volcanic seamount we surveyed. The Garrett Fracture zone intersects the EPR at about 13.25° and extends to the west with only subtle relief. OBS deployment sites are shown which recovered data for the refraction and microearthquake studies (open diamonds), long term deployment (filled triangles) and those which did not return useful data (small open circles).

The Gravity Lineations and Intraplate Melting Petrology and Seismic Expedition (GLIMPSE) was designed to test competing models for the origin of these volcanic ridges and gravity lineations. Methods used included petrological and geochemical analyses of dredged basalts, refraction/wide-angle reflection measurements for crustal thickness variations, shipboard measurements of gravity, bathymetry and side-scan reflectivity, and probing of upper mantle structure using seismic tomography. The 12-month duration of the deployment from November 2001 to December 2002 and the low instrument noise enabled us to record 155 earthquakes which had Rayleigh waves with good signal-to-noise ratio and azimuthal distribution for measurement of dispersion, anisotropy, attenuation, and derivation of the 3D velocity structure of the oceanic mantle.

We dismissed the suggestion that mini-hotspots are responsible for the chains of ridges and seamounts because the age progression found by dating basalts dredged from the Pukapuka ridge indicates propagation that is several times faster than Pacific plate motion [Sandwell et al., 1995; Janney et al., 2000].
Small-scale convective rolls, proposed to be responsible for the gravity lineations [Haxby & Weissel, 1986] form as negatively buoyant instabilities drip from a cooling and thickening lower thermal boundary layer [Jha & Parmentier, 1997; Ed: see also Small-scale convection page] or as instabilities within a melt-rich asthenosphere [Schmeling, 2000], that are organized into linear rolls by shear in the asthenosphere between the plate and the deeper mantle [Richter & Parsons, 1975]. Recent numerical work on convective cooling, however, suggests that convective instabilities may form at young ages but require mantle viscosities as low as 10¹⁷ Pa s [Zaranek & Parmentier, 2004].
Tectonic mechanisms such as lithospheric extension or boudinage (Figure 2b; Ed: see also Pukapuka page) are expected to allow pre-existing melt to percolate up through lithospheric cracks on the seafloor [e.g., Winterer & Sandwell, 1987]. Lithospheric extension should thin the plate creating somewhat lower velocities. The amount of extension required to produce the gravity lineations is of the order of 10%. However, recent studies indicated that the possible extension is much less than this [Goodwillie & Parsons, 1992; Gans et al., 2003].
Thermoelastic bending stresses can produce lithospheric failure in the troughs of plate undulations [Gans et al., 2003; Sandwell & Fialko, 2004]. In this case, a downward deflection of lithosphere and underlying asthenosphere of a few 100 m is predicted beneath the ridges but no upwelling of the mantle matrix (Figure 2c).
Viscous fingering instabilities may form in the asthenosphere when hot, volatile-rich mantle material, such as may be produced from an off-axis mantle plume, travels to the spreading ridge through the asthenosphere [Weeraratne et al., 2003]. Higher water or volatile content and higher temperatures should cause lower viscosities, lower seismic velocities and higher melt production as the return flow from a source such as the south Pacific Super Swell [e.g., McNutt et al., 1998; Montelli et al., 2004; Ed: see also Superswell page] rises under a thinning lithosphere [Weeraratne et al., 2003]. This model is consistent with geochemical anomalies observed along spreading ridge axes at hotspot locations [e.g., Mahoney et al., 1994; Schilling et al., 2003].

These models not necessarily mutually exclusive.

Figure 2: Models to explain the formation of seamount chains and gravity lineations in the south Pacific. a) Small scale convection [Haxby & Weissel, 1986], b) Lithospheric extension and cracking [Winterer & Sandwell, 1987], c) Thermoelastic bending and cracking [Gans et al., 2003], d) Viscous fingering instabilities involving anomalously low viscosity mantle intruded into a higher viscosity asthenospheric channel [Weeraratne et al., 2003]. Each model makes unique predictions for lithospheric deformation, asthenospheric flow, and melt generation.

There is a strong negative shear-wave velocity gradient with its top at about 25 km depth, beneath which is a minimum shear velocity of 3.95 km/s at 70 km (Figure 3). We associate the negative velocity gradient with the base of the lithosphere averaged over the study area at 40 ± 15 km depth. The minimum velocity at 70 km is underlain by a sharp positive velocity gradient extending to ~125 km. We suggest that the onset of decompression melting causes the change in shear velocity gradient at ~125 km. Dehydration associated with increased melting and melt removal may be responsible for this reversal in gradient.

Figure 3: a.) Average phase velocity as a function of depth (period). These results are from an inversion that also solves simultaneously for average anisotropy. Error bars indicate 2 standard deviations. b.) Dispersion curves from previous studies for 0 - 4 Ma (solid) and 4 - 20 Ma (dashed).

Anomalously low seismic phase velocities were observed below the Sojourn Ridge and Hotu Matua seamount complex at periods up to at least 67 s (corresponding very roughtly to 67 km depth). High velocities observed between the chains (Figures 4 & 5) are consistent with active geodynamic models for seamount formation such as convective rolls or viscous fingering instabilities.

Figure 4: Examples of two dimensional phase velocity maps for periods of 16 - 45 s. The starting model is the 1D average phase velocity result. Standard error contours are shown in the bottom row and are roughly the same for all models. Phase velocity variations include trade offs with anisotropy. Gaussian interpolation (smoothing) was used with an 80-km length scale. Low velocities are in red and high velocities in blue. The data are masked to show information only in the region considered reliable, based on the distribution of standard errors shown in the bottom row. White triangles indicate the temporary refraction and seismic stations which provide data from the first month for periods up to ~ 50 s. Black triangles are station locations for the long-term deployment. Bathymetry in the top row is contoured in black. White lines in panel at top right indicate lines of section shown in Figure 5.

Figure 5: Shear wave velocity cross sections parallel to the East Pacific Rise. Locations are shown in Figure 4, top right panel. The depth scale from 1 - 4 km is exaggerated to show the bathymetry. SJ: Sojourn ridge, SC: Southern Cross seamount, WHM: Western Hotu Matua, BR: Brown ridge.

Our results do not fit predictions of passive tectonic models such as lithospheric cracking or plate bending that invoke the tapping of pre-existing, widespread melt in the asthenosphere. If the convective roll hypothesis is to fit our data it must be modified to include some mechanism such as mantle return flow currents that sweep instabilities that form earlier, back to the EPR, where low seismic velocities, low Bouguer gravity anomalies, and small seamounts are observed. Viscous fingering in the asthenosphere could introduce compositional anomalies that would vary in enrichment from west to east and continue all the way to the EPR spreading center.