a gravity study of the intermediate-spreading Valu Fa Ridge, Lau Basin

Christine Peirce?, Ian M. Turner?* and Martin C. Sinha?

? Department of Geological Sciences, University of Durham, Science Laboratories, South Road, Durham, DH1 3LE, UK.

email: Christine.Peirce@durham.ac.uk

? School of Ocean and Earth Science, Southampton Oceanography Centre, University of Southampton, European Way, Southampton. SO14 3ZH, UK.

email: Martin.Sinha@soc.soton.ac.uk

* Now at: BG International, 100 Thames Valley Park Drive, Reading, RG6 1PT, UK.

Submitted: 10th July, 2000

Revised: 20th November, 2000

1 Introduction

Investigations of the internal structure of the oceanic crust have been used in attempts to identify the nature of mid-ocean ridge accretionary processes and the associated magma supply from the mantle. Models to date are based largely on information gleaned from seismic studies, which provide details of layering, structure, porosity and the existence and extent of sub-axial magma bodies. In addition, gravity studies provide not only a mechanism of verifying these results but also a means by which lateral crustal thickness variations, and density variation within the mantle associated with underlying accretionary processes, may be investigated. These features have implications for elucidating the nature of the underlying process of melt ascent from the mantle, its ponding within an axial magma chamber and its eruption at the seabed.

To truly understand ridge accretionary processes, a variety of ridges spreading at a variety of rates must be studied. To date, the largest number of investigations have focussed on the faster spreading ridges, particularly so the East Pacific Rise (EPR), where more favourable seabed conditions for seismic imaging exist. The results presented in this paper and that of Turner et al. (1999), aim to supplement these studies by providing details from a ridge system spreading at the slower end of the intermediate classification.

The Valu Fa Ridge (VFR) is an intermediate spreading centre (full rate of 60 mm yr^-1), situated on the eastern edge of the back-arc Lau Basin in close proximity to the Tofua Island Arc, part of the active Tonga-Kermadec Island Arc (Fig. 1). The VFR has been the subject of several recent detailed investigations which have shown it to be underlain by a robust magma chamber and associated melt lens (e.g. Turner et al., 1999; Sinha, 1995; Morton & Sleep, 1985; Collier, 1990; Collier & Sinha, 1990 and 1992) and to exhibit active hydrothermal vent systems (Fouquet et al., 1991).

Along-axis ridge segmentation has long been recognised as a fundamental characteristic of mid-ocean ridges. Previous swath bathymetry studies have shown the VFR to consist of three morphological segments - the Northern, Central and Southern Valu Fa Ridges (NVFR, CVFR and SVFR respectively) - which are separated by short, non-transform offsets (von Stackleberg et al., 1988). This segmentation closely resembles that observed at the East Pacific Rise (EPR) and other fast and intermediate spreading ridges (e.g. Macdonald et al., 1982, 1988 and 1991), and is thought to reflect along-axis variation in crustal accretionary process related to magma supply. Several studies of the EPR have indicated that the centres of individual spreading segments may be the sites of locally enhanced magma budgets (e.g. Macdonald et al., 1988) with magma supply being attenuated towards segment ends. However, beneath the overlap between the Central and Northern Valu Fa Ridges thickened crust has been observed (Turner et al., 1999), together with a wide melt lens reflection (Turner et al., 1999; Collier & Sinha, 1990 and 1992) and a sub-circular negative mantle Bouguer anomaly (Sinha, 1995). These observations imply that this overlapping spreading centre (OSC) is currently the site of an enhanced magma supply, which contrasts with widely accepted models which infer that the location of morphological offsets reflects regions of reduced magma supply (cf. Mid-Atlantic Ridge [MAR] - Kuo & Forsyth, 1988; Lin et al., 1990).

In this paper we describe the results of modelling a network of gravity profiles acquired during a multidisciplinary research cruise to the Valu Fa Ridge in 1995, which aimed to image crust and upper mantle structure both on-, off- and along-axis, and beneath an overlapping spreading centre located in a back-arc, intermediate spreading tectonic environment. These results, in conjunction with those of the associated seismic study (Turner et al., 1999), will then be discussed in relation to the nature of the underlying ridge accretionary processes and associated magma supply, and the mode of development of this back-arc spreading system.

2 Experimental configuration

A multidisciplinary geophysical experiment was conducted aboard the R/V Maurice Ewing (EW9512 - Peirce et al., 1996) in December 1995 at the VFR, centred on 22° 20¢ S, 176° 40¢ W (Fig. 1). Forward modelling of the seismic component of this experiment is described in Turner et al. (1999) (Fig. 2) and 2-D modelling of the controlled source electromagnetic (CSEM) data also acquired during this cruise is currently in progress (MacGregor et al. 2000). A network of gravity profiles was acquired over the entire work area, including profiles coincident with the across-axis, along-axis and axis-parallel seismic lines.

The observed gravity readings, logged at 1 s intervals, have been filtered, Eötvös corrected and interpolated to 60 s intervals and have an associated overall error of ±1-2 mGal. This interpolation interval results in a maximum sample spacing of ~150 m. As Eötvös correction of data acquired while turning is unreliable these data points have not been included. The resulting dataset, consistent at all cross-over points, was finally compared with the 2' x 2' Sandwell & Smith (1997) dataset to ensure a consistent base-level and to investigate the nature of the regional gravity field before modelling commenced.

3 Gravity modelling

Gravity modelling was undertaken for a number of reasons. Firstly, by adopting a 2-D/2¾-D approach, the wide-angle seismic models of Turner et al. (1999) could be tested for validity and uniqueness. Secondly, using the 2-D/2¾-D preferred gravity model parameters, crustal structure in areas of limited, or no, seismic control could be investigated. Thirdly, by adopting a 3-D approach and using the along-axis preferred density model as a reference, structures identified off-axis in the across-axis models could be mapped in an along-axis direction. Finally, having identified and tied like characteristics of the preferred seismic and gravity models, the variation in crustal structure associated with the development of the Valu Fa Ridge and the Lau Basin behind the propagating rift tip could be investigated with respect to underlying accretionary processes.

3.1 Initial 2-D modelling

Turner et al. (1999)?s velocity-depth models were initially converted to crustal density models using the mean velocity-density envelope of Nafe & Drake (1957), together with methods 2 and 3 of Carlson & Raskin (1984), and the method of Christensen & Salisbury (1975).

Modelling of the free-air gravity data was carried out using the programs GRAV2D (2-D - based on the Talwani et al., 1959 algorithm), which is based on the assumption that structures are infinite perpendicular to the model, and GMSYS (2¾-D - Northwest Geophysical Associates Inc., 1995) which accommodates out-of-plane structures. With the 2¾-D approach, structures of varying and limited extent, both laterally and perpendicularly, in front of and behind the plane of the model may be defined and their effect included in the calculated anomaly. The former assumption is more than satisfactory for Line 1 (Seismic South), where topographic features are effectively two-dimensional (Fig. 2). As this profile was effectively the simplest it was modelled first. The 3-D nature of the OSC and the numerous large seamounts close to Line 6 (Seismic North) also have an effect on the gravity field along this profile and hence the 2-D assumption is less valid. Thus Line 6 and the CVFR along-axis profile (Line 4) were modelled in 2¾-D in an attempt to faithfully reproduce the significant off-line seafloor topography.

A brief description of the 2-D modelling of Line 1 is contained in Turner et al. (1999). Here we present a detailed description of the gravity modelling of both across-axis profiles (Line 1 and Line 6) and the CVFR along-axis profile (Line 4). A complete set of gravity modelled sections, including several axis-parallel profiles, can be found in Turner (1998), Offer (1996), Rushforth (1996) and Calvert (1996).

Model densities are quoted in g cm^-3, mainly for reasons of clarity of annotation on the figures - 1 g cm^-3 is equivalent to 1000 kgm^-3.

3.1.1 Line 1 - Seismic South

P-wave seismic velocities of Turner et al. (1999)?s wide-angle seismic model for Line 1 (Fig. 3) were averaged within layers, converted to densities (Table 1) and used to construct the initial density model. The Nafe-Drake (1957) relationship is based on measurements of sediment velocity and density, and thus the density derived from this method was used to assign a density to the limited-extent off-axis sediment ponds. Likewise, method 3 of Carlson & Raskin (1984) attempts to represent large-scale porosity variation in the extrusive layer, hence a density derived using this method alone was considered most appropriate for modelling layer 2A. A density of 1.03 g cm^-3 was assigned to the water column (Kuo & Forsyth, 1988) and the upper mantle a density of 3.30 g cm^-3 - a value commonly used in gravity studies of mid-ocean ridges (e.g. Kuo & Forsyth, 1988; Solomon & Toomey, 1992; Cormier et al., 1995). The velocity-density relationship of Christensen & Salisbury (1975) was used as a final check on the calculated layer densities (Table 1: last column).
 

		P-wave	Nafe-Drake	Carlson & Raskin
		Velocity	r nd	r m2	r m3		Avr			r c&s
		(km s-1)	(g cm-3)	(g cm-3)	(g cm-3)		(g cm-3)			(g cm-3)
	water column	1.49	-	-	-		1.03			-
	sediment	2.15	1.98	1.02	1.74		1.58			1.80
	layer 2A	3.02	2.20	1.83	2.25		2.09			2.04
	layer 2B/C	5.01	2.51	2.61	2.74		2.62			2.60
	layer 3	6.58	2.83	2.90	2.92		2.88			3.04
	mantle	8.08	3.33	3.07	3.03		*3.30*			3.47

Table 1: Initial layer densities for Line 1. Densities are derived from the Nafe & Drake (1957) relationship (r _nd) and methods 2 and 3 (r _m2 & r _m3) of Carlson & Raskin (1984). The penultimate column represents the average of these three methods, in addition to the generally accepted mantle density (denoted by *). Initial density estimates used in the gravity modelling process are shown in bold. The final column shows densities calculated from the relationship of Christensen & Salisbury (1975) (r _c&s) for comparison.

As the 2-D gravity modelling attempts to test the validity and uniqueness of the seismic modelling, the latter of which more accurately locates interface depth and geometry, model layer boundaries were left unaltered. Only minor adjustment of layer densities from initial estimates (<0.04 g cm^-3) were required to generate a free-air gravity anomaly which matched the observed data to within error bounds (± 1-2 mGal). The results of gravity modelling of Line 1 are shown in Fig. 4 and are described below. The only significant misfit to the observed free-air gravity anomaly is centred on ~12 km model offset and is attributed to the contributory effect of the large seamount located to the north of the profile (Fig. 2 - 176° 33¢ W).

The uppermost layer of the model represents the water column with an average density of 1.03 g cm^-3 and sediment cover, restricted to a basin centred on 10 km model offset, was modelled with a density of 1.98 g cm^-3 (cf. Nafe & Drake, 1957). The density of layer 2A was calculated as 2.25 g cm^-3 and was kept constant across the model to reflect the minimal lateral variation in velocity within the seismic model (Fig. 3).

In the seismic model, layer 2B/C has a lateral variation in thickness of the order ± 1 km, and therefore makes a significant contribution to the overall free-air gravity anomaly. This layer was modelled with a density of 2.62 g cm^-3, although east of ~31 km model offset the density was reduced to 2.60 g cm^-3 to reflect the lower P-wave seismic velocities observed within the pre-rift island arc crust (cf. Fig. 3).

A low density block, representing the magma chamber melt lens (cf. velocity-depth model in Fig. 3), was modelled at a depth of ~3 km beneath the CVFR crest and was incorporated into the model for completeness only. This block was nominally assigned a density of 2.50 g cm^-3 (cf. Murase & McBirney, 1973) but has no appreciable effect on the calculated free-air gravity field.

For the region of layer 3 and the Moho constrained by the seismic modelling (see annotation on Fig. 4), the free-air gravity anomaly could not be faithfully reproduced with a constant density in layer 3. This area of layer 3 was, therefore, divided into five distinct regions, coinciding with the two pre-rift regions, the two transition zones and the post-rift region postulated in the seismic model interpretation shown in Fig. 3. The density of layer 3 was modelled as 2.89 g cm^-3 within ± 12 km of the ridge axis, with reduced densities of 2.86 g cm^-3 required in the transition zones to improve the fit. Note particularly the difference between the anomaly calculated from the initial and preferred final models between 10 and 40 km model offset (Fig. 4). Densities of 2.86 and 2.84 g cm^-3 (within the range of possibilities for the modelled velocities based on the methods outlined above) were also included beneath the pre-rift back-arc and island arc crust respectively.

The rationale behind our choice of preferred model for layer 3 (variable density blocks rather than a single constant density block and variable Moho depth) was that for the majority of the width of the model the Moho is constrained seismically and a good fit could be achieved simply by varying the density in the off-axis regions slightly. In light of the tectonic interpretation of the crustal velocity structure (Turner et al. 1999), the variation in density laterally seems more likely. Overall a good fit can be achieved without inclusion of a region of reduced density, corresponding to the low velocity zone (LVZ) within the axial region in layer 3 (within ± 5 km model offset - Fig. 3). However, within the resolution of gravity modelling this is to be expected, as the density contrast inferred from the seismic model is sufficiently small that its gravitational effect would also be small.

A density of 3.30 g cm^-3 is generally acknowledged to be a reasonable estimate for the upper mantle and, due to the lack of evidence to the contrary, i.e. lack of refracted arrivals turning at depth within the seismic dataset, this density was assigned across the profile.

3.1.2 Line 6 - Seismic North

Modelling of the more northerly across-axis profile (Line 6) proved more problematic, largely a consequence of the more variable out-of-plane seabed depth due mostly to numerous off-line large seamounts located close to this profile, and the effect of the topographic expression of the OSC. The addition of lower densities at each end of the profile, to be consistent with the modelling and interpretation of Line 1 (Fig. 4), improved the fit somewhat, although a considerable misfit (up to 7 mGal) was still apparent between -40 to -5 km model offset (Fig. 5: dashed line).

The observed gravity anomaly on Line 6 could not be faithfully reproduced by a 2-D crustal structure alone without the addition of a regional gradient. However, there is no evidence of such a regional gradient in the global 2´ x 2´ gravity grid of Sandwell & Smith (1997). Thus an attempt to account for the gravitational effect of out-of-plane structures was made by modelling in 2¾-D, with this approach also providing a means of accommodating the greater variability in seafloor topography at this end of the CVFR associated with its overlap with the NVFR. Note that 2¾-D model parameters are described here as minimum and maximum y-lengths, corresponding solely to the perpendicular extent out-of-plane (no in-plane adjustments were made) of structures on and adjacent to Line 6 (positive values denote the extent south of the profile and negative to the north). The preferred final gravity model is shown in Fig. 5 and the calculated free-air anomaly from 2-D modelling is shown for comparison (dashed line).

The CVFR was modelled with a minimum y-length of -5 km (i.e. extending 5 km north of Line 6), and having infinite extent south of the profile, likewise the NVFR was modelled with a maximum y-length of 5 km and infinite extent north of the profile. The seamount located at -8 km model offset (Fig. 2: 22° 10´ S, 176° 40´ W) extends to a height of 400 m above the surrounding seafloor. It is clear from the conical shape of this topographic feature that it has limited extent north and south of Line 6. The seamount was thus modelled with minimum and maximum y-lengths of -0.5 and 1 km for the uppermost 200 m and -1 and 1.5 km for the lower 200 m (Fig. 5). To the north of Line 6, at 5 km model offset, a second seamount was modelled with a height of 500 m and minimum and maximum y-lengths of -4 and -1 km (Fig. 2: 176° 32´ W).

No bathymetry data to the north of Line 6, between -40 km and -10 km model offset, was acquired during EW9512. However, SeaBEAM data from three R/V Sonne cruises (SO-35, SO-48 and SO-67) as described in Wiedicke & Collier (1993), identify a large basin to the north of 22° 05´ S and this feature was also incorporated into the 2¾-D model. The average depth of the seabed north of Line 6 is ~2.5 km and therefore any crust above this depth along the profile was modelled in 2¾-D with a minimum y-length of -2 km (i.e. extending only 2 km north of the profile) to simulate the basin described above. Similar parameters are used to describe the layer 2A/2B boundary in an attempt to maintain the approximately constant off-axis thickness of 2A as modelled from the seismic data.

The main layer densities used in 2-D modelling remain in the final 2¾-D density model. The average density of layer 2A was modelled as 2.25 g cm^-3, although the seamounts and ridges modelled in 2¾-D were assigned lower densities (2.15 and 2.20 g cm^-3) as they represent only the uppermost part of this layer. A density of 2.52 g cm^-3 was assigned to the top of layer 2B/C west of -10 km model offset for a similar reason (Fig. 5).

The preferred final 2¾-D across-axis density model provides a far better fit to the observed anomaly between -40 and -5 km model offset and indicates that axis-parallel variation in layer 2 structure contributes significantly to the observed gravity field. The maximum misfit of ~3 mGal is associated with regions where there is greatest variation in the in- and out-of-plane seabed topography.

3.1.3 Line 4 - along-axis (CVFR)

Densities for the CVFR along-axis profile (Line 4) were obtained from the intersections with the two across-axis profiles already described, and extrapolated across the entire model. The four main layers (there is no sediment on the along-axis profile) - layer 2A, layer 2B/C, layer 3 and upper mantle - were thus modelled with densities of 2.25, 2.62, 2.89 and 3.30 g cm^-3 respectively.

The crustal structure in- and out-of-the-plane of Line 4 is clearly not 2-D (Fig. 2) and was thus modelled using the 2¾-D approach described above, in an attempt to simulate off-line topographic features, intra-crustal interface geometry and the limited extent of the CVFR axial high perpendicular to the profile. The preferred 2¾-D gravity model is shown in Fig. 6 with input parameters similar to those described for Line 6 (cf. Fig. 5). The main features incorporated into this model are described below. Note that for 2¾-D modelling of Line 4, positive y-lengths correspond to structures that extend east of the profile, and negative to the west. The maximum error in the fit of the calculated and observed free-air gravity data is ~3 mGal, again associated with areas of significant out-of-plane topography for which there is limited seismic control.

The NVFR, located west of Line 4 at model offsets greater than 30 km, was represented using three separate blocks located at different depths (1.9-2.1 km; 2.1-2.3 km and >2.3 km depth). The shallowest block was modelled with a density of 2.15 g cm^-3 and minimum and maximum y-lengths of -6 and -3 km at 34 km model offset, and -7 and -4 km at model offsets greater than 37 km. Below this block lies a body of density 2.20 g cm^-3, with minimum and maximum y-lengths of -9 and -2.5 km which was modelled at >39 km model offset. The deepest block was modelled with a density of 2.25 g cm^-3 and a width of 10 km with its eastern limit 1 km to the west of Line 4.

Line 4 linearly follows the bathymetric expression of the CVFR, which has limited extent perpendicular to the profile. In an attempt to reproduce this feature, blocks of different density (2.15-2.25 g cm^-3) were incorporated into the model with differing minimum and maximum y-lengths (e.g. Fig. 6: centred on 11 km model offset with a density of 2.15 g cm^-3, extending 2 km west and 1 km east of Line 4).

In maintaining a linear north-northeast-south-southwest profile, Line 4 did not always follow the shallowest bathymetric expression of the CVFR along the entire length of the profile, as the ridge orientation is altered by the two devals at 22° 21´ S and 22° 16´ S. Between 17 and 24 km model offset the ridge crest is located 2 km east of the profile and was thus modelled in 2¾-D as an additional block of density 2.15 g cm^-3, a width of 2 km and a height of ~200 m.

As Line 4 has the most significant out-of-plane bathymetric variation, a test of its contribution to the observed field was undertaken before further out-of-plane modification was made to model intra-crustal layer geometry. Using the bathymetry grid shown in Fig. 2, the gravitational effect of the seabed in three-dimensions was calculated (see next section for details). A profile along Line 4 was extracted and plotted in Fig. 6 for comparison (dotted line). Note that no comparison should be made with respect to absolute values, but merely its shape with respect to the observed. As would be expected the profiles follow a similar, but not identical, trend thus indicating that the seabed significantly contributes to the total observed field. However, the variation between these two profiles also suggests that there must be out-of-the-plane variation in intra-crustal layer thickness and/or density. As can clearly be seen in Fig. 6 the seabed profile is a better reflection of the trend of the observed anomaly than the calculated anomaly obtained from the simple 2-D modelling. However, south of ~20 km model offset there is a distinct deviation between the observed and seabed trends which is suggestive of one or more of the following: along-axis variation in crustal layer density, out-of-plane layer geometry and/or density variation, axis-parallel crustal thickness variation or along-axis mantle density variation. As along-axis (not axis-parallel) variation in crustal thickness has already been incorporated into the model to some extent based on the results of seismic and gravity modelling of Lines 1 and 6, the remaining possibilities were considered and incorporated into the 2¾-D modelling process and a much better fit achieved.

Topographical variations in the layer 2A/2B boundary were modelled perpendicular to the profile with additional blocks centred on 5 km (density of 2.52 g cm^-3 - top of layer 2B/C) and 20 km model offset (density of 2.35 g cm^-3 - base of layer 2A).

The depth of the Moho in this model was constrained solely from the two intersection points with the across-axis profiles (0 and 28.5 km model offset) and hence there was scope for changing the depth of the Moho away from these two points in order to better fit the observed gravity. The Moho on Line 1 decreases in depth away from axial region (cf. Fig. 3) and thus the upper mantle between 0 and 30 km model offset was modelled at depths less than 11 km, with a lateral extent of 10 km east and west of Line 4 (dashed line at Moho depth in Fig. 6b). The slight decrease in depth of the Moho (~0.7 km) at model offsets greater than 30 km (dotted line at Moho depth - Fig. 6b) provides a much better fit to the observed data and was thus included as an additional constraint in the final model.

The main conclusion to be drawn from modelling of Line 4 is that the crust appears to thicken northwards along the CVFR, and this modelling has provided some constraint on Moho depth and geometry in regions unachievable from seismic modelling alone. There is also evidence for crustal thinning, particularly so to the west, between the ridge axis and the transition zone between pre- and post-rift crust (cf. Fig. 3). The modelling of Line 4 provided no direct evidence for along-axis variation in crustal or mantle density.

3.2 Residual mantle Bouguer anomaly

In addition to 2-D and 2¾-D free-air gravity modelling, the entire gravity dataset was used to calculate the mantle and residual Bouguer anomalies (MBA and RMBA respectively) over a region spanning the Northern Valu Fa Ridge southwards to the propagating tip at ~22º 55?N (Fig. 2). The main stages involved in the calculation of these anomalies are summarised below.

3.2.1 Calculation of the mantle Bouguer anomaly

The MBA represents the gravity field arising from lateral variations in crustal thickness and changes in crustal and/or mantle densities. In simple terms the MBA is calculated by subtracting the gravitational attraction of a reference crust, composed of layers of constant thickness and density, from the free-air gravity anomaly (e.g. Prince & Forsyth, 1988; Lin et al., 1990).

Initially, the EW9512 free-air gravity data were compared with regional data from the Sandwell & Smith (1997) 2´x2´ gravity grid to equate the relative base-levels of the two datasets and to check for any evidence of a regional field at the wavelength of the ridge features being studied. Within the area defined by the EW9512 network of profiles, this dataset was merged with the free-air gravity data acquired during RRS Charles Darwin cruise CD34/88 (Sinha, 1995), again after checking the latter for base-level consistency before gridding to produce the final combined free-air gravity anomaly map shown in Fig. 7. Of note here is the magnitude of the background field (60-70 mGal). A large positive anomaly of a wavelength of the order of 700 km is observed over the entire back-arc region in the global 2?x2? satellite gravity compilation of Sandwell & Smith (1997). The free-air gravity anomaly shown in Fig. 7 thus effectively represents medium-short wavelength features superimposed upon the long-wavelength field associated with the descending slab of the Tonga subduction zone adjacent to the east.

The MBA was calculated using an approach based on the method of Parker (1972), which determines the gravitational effect of topography and density contrasts at user-defined boundaries (Kuo & Forsyth, 1988 and Fig. 8). This method requires a rectangular grid of regularly spaced bathymetry data points with the grid spacing set to less than the water depth to avoid aliasing (Navin, 1996). As the EW9512 bathymetry data are included in the Lau Basin compilation bathymetry grid of Zellmer et al. (1998), which provides more extensive off-axis coverage to the south of the study area, an extract from the latter was used as the basis of the MBA calculation (Fig. 9).

The VFR survey area is elongate parallel and perpendicular to the spreading axes and does not trend in a north-south direction (see Fig. 2), hence in a geographical co-ordinate system a suitable rectangular grid cannot be easily extracted. The ?geographic? bathymetry dataset was thus rotated to align the horizontal axis parallel to the spreading direction (i.e. parallel to the two across-axis seismic profiles, Line 1 and Line 6) and resampled into a new x-y co-ordinate system. A grid of 256 samples along each axis, encompassing all ridge segments, the propagating rift tip and offsets up to 32 km off-axis, was then extracted from this dataset. The location of the rectangular grid with respect to the EW9512 and CD34/88 gravity lines is shown in Fig. 7, and in relation to the entire bathymetry grid in geographical co-ordinates in Fig. 9. Where appropriate, the rotated and resampled bathymetry grid has been clipped to show only areas of densest gravity data coverage.

Within this area, gravity anomalies calculated for the predictable components of the gravity field (water column depth and crustal layers of constant thickness and density) were subtracted from the observed field to create the MBA shown in Fig. 10. The choice of layers, their thicknesses and densities for the MBA calculation was based upon the results of modelling described in Section 3.1. Fig. 8 illustrates this uniform layer thickness and density model, and compares it with the seismic modelling results of Turner et al. (1999) for Line 1. Many previous calculations of MBAs (e.g. Kuo & Forsyth, 1988; Lin et al., 1990) assume a constant density for the entire crust and Kuo & Forsyth (1988) state that a stratified crust (a gradual density increase from 2.40 g cm^-3 at the seabed to 2.95 g cm^-3 at the base of the crust) changes the predicted anomaly by less than 1 mGal. However, as knowledge of the density stratification of the crust was readily obtainable from 2-D seismic and gravity modelling it was included for completeness. The predicted gravitational attraction of the layered crustal model (density contrast of 1.22 g cm^-3 at the seabed, 0.37 g cm^-3 at the layer 2A/2B boundary, 0.26 g cm^-3 at the layer 2/3 boundary, and 0.42 g cm^-3 at the Moho) is shown in Fig. 9d. The seafloor topography (Fig 9a) contributes ~70% of the total calculated anomaly, and hence the combined anomaly (Fig. 9d) is dominated by the gravitational effect of the seabed (cf. Fig. 9c). The predicted gravity field shown in Fig. 9d is thus subtracted from the observed free-air gravity anomaly (Fig. 7) to generate the MBA (Fig. 10).

The most dominant features of the MBA are: a) the general decrease in amplitude from south to north (from >90 to <70 mGal); b) the higher amplitude anomaly between the ridge-axis and the pseudofault (marked A on Fig. 10); and c) a slight decrease in anomaly to the east of 176° 32´ W (marked B on Fig. 10), the latter being most likely related to the proximity of the Tofua island arc. Contours are also deviated, mirroring the topography, around large seamounts (e.g. 22° 15´ S, 176° 33´ W) which may be suggestive of a locally thickened upper crust, probably due to an increase in the volume of intrusive and extrusive material as found by Offer (1996) who modelled seismic and gravity data across the large seamount located at ~22° 10'S on Line 5 at the westernmost extent of the survey area.

The general decrease in anomaly amplitude from the southern tip of the CVFR (~78 mGal) to the OSC at 22° 12´ S (~66 mGal) implies thicker crust and/or lower crustal/mantle densities beneath the OSC. An MBA plot calculated from the RRS Charles Darwin cruise CD34/88 gravity dataset alone using the forward modelling algorithm of Talwani & Ewing (1960), is described by Sinha (1995). The general decrease in anomaly towards the OSC shown in Fig. 10 is also observed in the CD34/88 data (inset to Fig. 10). Sinha (1995) notes that areas of relatively low MBA correlate with areas of shallowest seafloor topography, in particular the ridge crest, suggesting a remarkable degree of isostasy. This correlation is not observed in the combined EW9512 and CD34/88 dataset. The Sinha (1995) MBA calculation was based upon wide-beam echo sounder bathymetry data, which tends to underestimate water depths on steep slopes. In the current study this artefact is avoided by the use of swath bathymetry data, which is also available out to greater distances off-axis. The accuracy of seabed bathymetry measurement probably accounts for much of the difference in MBA between the two studies, particularly so over the eastern and western flanks of the VFR. Also, the enclosed basins to the west and east of the CVFR crest are bounded by the central ridge at 176° 52¢ W (Wiedicke & Collier, 1993) and the large north-south seamount chain at 176° 34¢ W respectively (see Fig. 2). However, these shallow bathymetric features lie largely outside the topographic grid used by Sinha (1995) and are thus not included in the CD34/88 MBA calculation. The gravitational attraction of the shallower seafloor external to the CD34/88 grid, and hence more negative MBA either side of the ridge crest, may further explain the difference between the two MBA plots (cf. Fig. 10 and inset).

The most pronounced feature in the MBA is the significant positive anomaly off-axis, which lies largely between the pseudofault and the ridge axis. Comparison of the location of this anomaly with Fig. 4 - 2-D gravity modelling across the southern end of the CVFR - shows a remarkable correlation with the transition regions between pre- and post-rift crust in which significant thinning of layer 2 is observed. The increasing amplitude of this anomaly heading southwards implies that further thinning of layer 2 (and possibly layer 3 - see later) occurs towards, and south of, the propagating rift tip.

3.2.2 Calculation of residual mantle Bouguer anomaly

To further investigate lateral variation in layer thickness within the crust, and density variations in the crust and mantle across-axis, requires removal of the component of the gravity field associated with passive upwelling at the ridge axis. The method adopted is outlined by Phipps Morgan & Forsyth (1988), a development of the approach first proposed by Forsyth & Wilson (1984) - in which passive upwelling is modelled within a triangular region whose apex lies beneath the ridge crest with flow horizontal elsewhere parallel to the direction of plate motion - to accommodate more complicated spreading geometries such as that associated with short transform offsets and multiple short segments. A steady-state thermal model of a ridge-transform-ridge system can thus be converted into a model of density variation by multiplying by an appropriate thermal expansion coefficient (e.g. Prince & Forsyth, 1988). This thermal anomaly may then be subtracted from the MBA to generate the RMBA. A summary of the main input parameters, apart from the ridge geometry which is shown in Fig. 11, is contained in Table 2, and a more detailed description of calculation of the residual mantle Bouguer anomaly can be found in Prince & Forsyth (1988) and Navin (1996).

The approach of Phipps Morgan & Forsyth (1988) calculates the thermal effect of a north-south trending ridge-transform-ridge system. The VFR can be approximated in this manner in the rotated co-ordinate system, as shown in Fig. 11, with linear segments defined based upon interpretation of the swath bathymetry data, GLORIA side-scan sonar and other geophysical datasets from the area (Parson et al., 1990; Cann, 1994; Parson & Wright, 1996). The resulting calculated thermal gravity anomaly is shown in Fig. 11b.

Once calculated the thermal gravity anomaly was subtracted from the MBA to generate the RMBA which was then rotated back into geographical co-ordinates (Fig. 12). The RMBA shows similar features to those observed in the MBA (Fig. 10) and is dominated by the general decrease in anomaly towards the OSC at 22° 12´ S. The anomaly decreases from 86 mGal at the southern tip of the CVFR to 74 mGal, centred on the OSC. When considering the implications of this general decrease in anomaly towards the OSC it should be noted that the ridge-transform-ridge system, used to calculate the thermal anomaly, breaks down at this location because the CVFR and NVFR are not separated by a simple transform and, in reality, overlap by at least 10 km (Fig. 11). However, as the offset and overlap between both ridges is relatively small, this introduces only a relatively small error into the thermal anomaly at the OSC. The proximity of the active Tofua island arc and the presence of the subducting former Pacific plate at a depth of 200 km beneath the VFR axis (Sinha, 1995) should also be borne in mind.

The main observation of an anomaly low above the OSC at 22° 12´ S on both the MBA and RMBA contrasts with the bull?s-eye negative anomalies identified at the centre of spreading segments on the slow spreading Mid-Atlantic Ridge (e.g. Tolstoy et al., 1993) and East Pacific Rise (e.g. Wang et al., 1996). These bull?s-eye anomalies are interpreted as representing thickened crust at segment centres. If the same reasoning is applied to the VFR anomaly then a thickened crust (and/or lower mantle density) beneath the OSC is implied - confirming the seismic results of Turner et al. (1999).

Of note also is the general southward increase in anomaly related to the location of the propagating rift tip in the region of ~22° 55´ S and, in particular, the geometry of this anomaly in relation to the pseudofault location of Wiedicke & Collier (1993), which marks the transition between pre- and post-rift crust. This observation suggests that the region to the south of ~22° 55´ S is a region of stretched and thinned island arc crust not yet undergoing true seafloor spreading. Again a degree of caution should be exercised when considering the implications of the RMBA in the region of the propagating rift tip. Calculation of the thermal anomaly assumes a steady-state constant spreading rate ridge system. This scenario is thus not strictly applicable in the region of, and in front of, the propagating rift tip. However, subtracting the calculated thermal anomaly in this region only has a small effect because of the limited extent of the survey area and the short offsets between the modelled ridge segments. Hence the incorporated error is only relatively small.

The most obvious feature of the RMBA plot is that the largest changes in RMBA occur along, rather than across-axis. This suggests that, broadly speaking, the RMBA assumptions provide a reasonable approximation of across-axis ridge structure; but that the assumption of a uniform crust and mantle structure along-axis is not correct.

In an attempt to investigate this along-axis variation, and distinguish between rift propagation and magma supply versus segmentation effects, a 2-D profile was extracted from the RMBA, running along-axis from south of the propagating rift tip to the northern end of the NVFR. This profile was filtered to remove shorter-wavelength trends of most likely an intra-crustal origin, thus leaving the anomaly primarily associated with crustal thickness and/or density variation and/or mantle density variation along-axis (Fig. 13). The smoothed along-axis trend in the RMBA was then modelled initially using the constant layer thicknesses and densities of the RMBA crustal model (Fig. 8). These layers were perturbed following the known layer thickness variations between the CVFR and OSC as constrained by the seismic modelling of Turner et al. (1999) and Day et al. (2000). Note that the water layer was not included in this modelling as its density is effectively constant throughout the study area and the gravitational attraction associated with its varying thickness is fully accounted for as part of the RMBA calculation. The control on crustal thickness from the seismic modelling (Turner et al. 1999) at two points along the profile was also included to provide some initial constraint on Moho depth and the degree of lateral variation in Moho depth possible along the profile. Fig. 13 shows the results of this modelling, and that the longer-wavelength variation of the RMBA along-axis can largely be explained by crustal thickness variation alone without any need to vary layer density along-axis within the crust or mantle. This is consistent with the 2-D seismic and 2¾-D gravity modelling of the CVFR axial region crust which also shows no evidence of significant lateral variation (Fig. 6). The preferred model shows a remarkable correlation between thinner crust and the propagating rift tip, with thinning of 1-2 km compared with the seismically consistent crustal reference point at the southern end of the CVFR, from which the RMBA crustal model was derived. Comparison of this result with the RMBA anomaly plot (Fig. 12) suggests that the crust is even thinner in the region between the axis and the pseudofault - of the order of 3-4 km, i.e. 30-40% of its normal thickness. Modelling is also suggestive of thickening associated with each of the SVFR, CVFR and NVFR segments. However, distinct thickening associated with each of the latter two segments is overprinted to an extent by the significant thickening associated with the OSC. Hence the results of this modelling suggest that the VFR in generality conforms to the ridge segmentation versus crustal thickness model of Macdonald et al. (1988).

3.3 Summary of modelling results

The main results of the gravity modelling may be summarised as follows:

1. the island arc region to the east comprises a crust which is of slightly lower average density than elsewhere in the study area;
2. lateral density variation within layer 3, with the pre-rift crust generally having a lower density than that of post-rift, implies compositional differences in the parent magma supply;
3. the density change corresponds well with the interpreted transition zones (cf. Fig. 3) between pre- and post-rift crust and with the postulated location of the rift-related pseudofault of Wiedicke & Collier (1993), which marks the lateral extreme of the post-rift crust and maps the progress of the propagating rift tip as it migrates southwards;
4. the density variation sub-axis which corresponds to the magma chamber low velocity zone and associated melt lens (see Fig. 3 and the LVZ of Turner et al., 1999), contributes little to the overall gravity anomaly and thus, these features are largely unconstrained by it;
5. the relatively low density of 2.25 g cm^-3 in layer 2A reflects correspondingly low velocities when compared with other spreading ridges, an observation interpreted to reflect highly fractured surficial material of an unusually high porosity and permeability, perhaps a result of the increased volatile content of constituent andesitic lavas (Vallier et al. 1991; Frenzel et al. 1990; Jenner et al. 1987);
6. layer 2A appears to thicken beneath the larger seamounts;
7. the lower crust, layer 3, off-axis varies in thickness by up to 1-2 km with topography on the layer 2/3 boundary associated with the transition zones between pre- and post-rift crust (Fig. 3);
8. there appears to be significant crustal thinning off-axis to the west in the region between the ridge-axis and the pseudofault location;
9. there is no evidence to support a significant variation in crustal density along-axis;
10. VFR crust generally thickens northwards towards the OSC (Fig. 6) and thins towards, and south of, the propagating rift tip (Fig. 8);
11. crustal thickening is associated with the SVFR, CVFR and NVFR and the latter?s overlapping spreading centre.

4 DISCUSSION

Despite the oceanic crust being generally uniform in terms of its basic layering and velocity structure, detailed geophysical studies have shown there to be considerable variation in the morphological structure and thickness associated with mid-ocean ridges of all spreading rates (e.g. Kuo & Forsyth, 1988; Lin et al., 1990; White et al., 1992; Hasselgren & Clowes, 1995). Along-axis variation in crustal thickness and/or crustal/mantle density has been interpreted to be the result of changes in magmatic budget (e.g. Kuo & Forsyth, 1988; Macdonald et al., 1988; Lin et al., 1990), and thus mapping of such lateral variations may lead to an understanding of the accretionary processes which appear to generate a similar crustal thickness and internal structure no matter the spreading rate, including an indication of how melt ascends to the near surface from deep in the mantle.

Forward modelling of across-axis, along-axis and axis-parallel seismic lines at the VFR have provided detailed images of the crustal structure in two dimensions at distinct locations (Turner et al., 1999; Collier & Sinha, 1990: 1992). Modelling of the coincident free-air gravity data, as described in this paper, has provided a test on the validity and uniqueness of these models.

To investigate the details of crust and upper mantle structure, the supply of magma to the Valu Fa Ridge segments and the nature of the on-going rift propagation, the mantle and residual mantle Bouguer anomalies were calculated as outlined in Section 3.2. The 2-D seismic and 2-D/2¾-D gravity models provide detailed information on layer thickness and density for these calculations. The average crustal model used in these calculations represents layers of constant thickness and density (Fig. 8), any deviation from this average model will result in anomalies in the MBA (and hence RMBA).

As can be seen in Fig. 10, there is minimal topographic signature remaining in the MBA, except associated with seamounts to the east of the study area, thus indicating that, as far as axis-parallel crustal structure is concerned, the average crustal model and the passive upwelling assumptions of the RMBA calculation are largely valid. However, under these seamounts the average model would predict a corresponding shallowing of each layer mirroring the shallowing bathymetry, which explains the negative sub-circular anomalies associated with the larger seamounts to the east of the study area and the large seamount towards the northwest corner and accounts for the majority of the smallest wavelength anomalies. This interpretation is supported by the seismic modelling of Turner (1998) and Offer (1996) for the seamount at the northern end of Line 5 (Fig. 2), located on the westernmost side of the survey area.

Thus the main larger-scale features of interest in the MBA and RMBA are the general decreasing anomaly trend heading towards the OSC and the increase towards the propagating rift tip. The general decrease in anomaly towards the OSC implies that a thicker crust and/or possibly lower mantle density exists. Both the 2-D seismic and gravity models indicate that the crust beneath the CVFR thickens from its southern tip to the OSC by ~1-2 km. This increased crustal thickness suggests that the OSC at 22º 12' S is presently, or has been in the recent geological past, the focus of a enhanced magma budget.

Wiedicke & Collier (1993)?s analysis of bathymetric data at this OSC reveals a complex evolution, with past OSC locations marked by relic ridges now off-axis. Under normal conditions the NVRF would be expected to propagate southwards at the expense of the CVFR. However, reconstruction of past ridge locations has shown a succession of oscillations between southwards propagation of the NVFR and northwards propagation of the CVFR, to the present day where the NVFR is propagating southwards at the expense of the CVFR. The OSC appears locked astride a larger-scale shallowing of the seabed in a zone some 20 km wide, which extends off-axis and contains numerous large seamounts of anomalous geochemical signature (Sunkel, 1990), which Wiedicke & Collier (1993) attribute to the proximity of the ridge system to the Tonga trench to the east. Thus the enhanced magma supply and the effectively stationary position of the OSC may simply be artefacts of the adjacent subducting system (see also Day et al. 2000).

The observation of a general increase in crustal thickness towards the north may additionally be explained in the context of the continual development of a propagating rift system (Peirce et al. 2000). This trend may simply reflect an establishing spreading system over time. For example, as the rift tip propagates southwards the pre-existing crust becomes progressively thinned until rifting occurs and a spreading system becomes established. Initially melt supply is weak, resulting in a generally thin crust accreted at the spreading axis and thinner crust associated with the pre-rift stretching preserved off-axis in a transition zone associated with the bounding pseudofault. As the rift tip propagates southwards over time, the ridge system in its wake becomes progressively more established with a stronger melt supply, in turn, resulting in a thicker crust. Thus for a southwards propagating rift system, a northwards thickening of the crust would be expected. Calvert (1996)?s 2-D modelling of nine gravity profiles located between the CVFR and the current location of the propagating rift tip also supports this general northwards thickening and also suggests an approximately 3 km thinner crust off-axis in the transition zone than on-axis in the region just behind the propagating rift tip. This work and the work of Calvert (1996) suggest that pre-rift crust thins by approximately one third of its thickness before spreading initiates.

Thus the results presented in this paper suggest that the VFR system doesn?t become fully established north of the propagating rift tip until, approximately, the current location of the CVFR, a distance of more than 60 km behind the current rift tip location. Assuming a half spreading rate of 30 mm yr^-1, and the pseudofault located at ~18 km off axis at the latitude of the southern end of the CVFR (see Fig. 4), the ridge system, therefore, has taken ~600,000 years to become fully established, since it is ~600,000 years since the propagating rift tip passed this point.

5 CONCLUSIONS

The results presented in this paper follow largely on from the results of modelling the EW9512 seismic data presented in Turner et al. (1999), with gravity modelling validating the seismic models and providing further constraint on crustal and upper mantle structure and properties. The across-axis gravity modelling confirms the difference in character of the pre- and post-rift crust, which is most likely attributable to a compositional difference between parent magmas, while the along-axis gravity modelling suggests that crustal accretion is largely a 3-D process at the Valu Fa Ridge, as evidenced by thicker crust associated with segment centres. Thus, at the VFR ridge segmentation is postulated to be controlled by magma supply, and present-day enhanced magma supply postulated to be associated with the overlapping spreading centre between the Central and Northern Valu Fa Ridges. In addition, a distinct gravity signature may be recognised associated with the stretching and thinning of the pre-existing crust as the VFR rift tip propagates southwards towards the Havre Trough.

The results presented in this paper therefore suggest that the intermediate-spreading VFR conforms to models of crustal accretion which infer a direct relationship between magma supply and morphological segmentation (e.g. Macdonald et al., 1988), where upwelling magma is largely restricted to segment centres and the lateral extent of magma flow determines segment length. The results also challenge these models as the OSC between the CVFR and NVFR also appears to also be the location of an enhanced magma supply.

Acknowledgements

We would like to thank the officers and crew of the R/V Maurice Ewing and members of the Scientific Party of cruise EW9512 without whose efforts data collection would not have been possible. This research was supported by the National Environment Research Council; via research grant (GST/02/1123), ship time and a Ph.D. studentship (IMT - GT4/95/72/E). Figures were created using the Generic Mapping Tool of Wessel & Smith (1995). We thank Don Forsyth and an anonymous reviewer whose positive and helpful comments greatly improved the clarity of this paper. Further information can be found on the University of Durham DOBS www home page (http://www.dur.ac.uk/christine.peirce/ [all lower case]), or by emailing any of the authors.

References

Calvert, S.E.E., 1996. Southern Valu Fa Ridge, The Lau Basin: A gravity and bathymetric investigation, unpublished M.Sc. thesis, University of Durham, pp.120.

Cann, M., 1994. A magnetic interpretation of the Lau Basin, Southwest Pacific, unpublished M.Sc. thesis, University of Durham, pp.81.

Carlson, R.L. & Raskin, G.S., 1984. Density of the oceanic crust, Nature, 311, 555-558.

Christensen, N.I. & Salisbury, M.H., 1975. Structure and constitution of the lower oceanic crust, Rev. Geophys. Space Phys., 13, 57-86.

Collier, J.S., 1990. Seismic imaging of magma chambers and mud diapirs, unpublished Ph.D. thesis, University of Cambridge, pp.231.

Collier, J.S. & Sinha, M.C., 1990. Seismic images of a magma chamber beneath the Lau Basin back-arc spreading centre, Nature, 346, 646-648.

Collier, J.S. & Sinha, M.C., 1992. Seismic mapping of a magma chamber beneath the Valu Fa Ridge, Lau Basin, J. geophys. Res., 97, no. B10, 14 031-14 053.

Cormier, M.-H., Macdonald, K.C. & D.S. Wilson, 1995. A three-dimensional gravity analysis of the East Pacific Rise from 18° to 21°30'S, J. geophys. Res., 100, no. B5, 8 063-8 082.

Day, A.J., Peirce, C. & Sinha, M.C., 2000. Three-dimensional crustal structure and magma chamber geometry at the intermediate-spreading, back-arc Valu Fa Ridge, Lau Basin - results of a wide-angle seismic tomographic inversion, Geophys. J. Int., Accepted.

Forsyth, D.W. & Wilson, B., 1984. Three dimensional temperature structure of a ridge-transform-ridge system, Earth Planet. Sci. Lett., 70, 355-362.

Fouquet, Y., von Stackleberg, U., Charlou, J.L., Donval, J.P., Foucher, J.P., Erzinger, J., Herzig, P., Mühe, R., Wiedicke, M., Soukai, S., & Whitechurch, H., 1991. Hydrothermal activity in the Lau back-arc basin: sulphides and water chemistry, Geology, 88, 2154-2181.

Frenzel, G., Mühe, R. & Stoffers, P., 1990. Petrology of the volcanic rocks from the Lau Basin, Southwest Pacific, Geol. Jahrb. Reihe D, 72, 395-479.

Hasslegren, E.O. & Clowes, R.M., 1995. Crustal structure of the Northern Juan de Fuca plate from multichannel reflection data, J. geophys. Res., 100, no. B4, 6 469-6 486.

Jenner, G.A., Cawood, P.A., Rautenschlein, M. and White, W.M., 1987. Composition of back-arc volcanics, Valu Fa Ridge, Lau Basin: Evidence for slab-derived component in their mantle source, J. Volc. & Geotherm. Res., 32, 209-222.

Kuo, B.-Y. & Forsyth, D.W., 1988. Gravity anomalies of the ridge-transform system in the South Atlantic between 31° and 34.5°S upwelling centres and variations in crustal thickness, Marine geophys. Res., 10, 205-232.

Lin, J., Purdy, G.M., Schouten, H., Sempere, J.C. & Zervas, C., 1990. Evidence from gravity data for focused magmatic accretion along the Mid-Atlantic Ridge, Nature, 344, 627-632.

Macdonald, K.C., 1982. Mid-ocean ridges: Fine scale tectonic, volcanic and hydrothermal processes within the plate boundary zone, Ann. Rev. Earth planet. Sci., 10, 155-190.

Macdonald, K.C., Fox, P.J., Perram, L.J., Eisen, M.F., Hayman, R.M., Miller, S.P., Carbotte, S.M., Cormier, M.-H. & Shor, A.N., 1988. A new view of the mid-ocean ridge from the behaviour of ridge-axis discontinuities, Nature, 335, 217-225.

Macdonald, K.C., Scheirer, D.S. & Carbotte, S.M., 1991. Mid-ocean ridges: discontinuities, segments and giant cracks. Science, 253, 986-994.

MacGregor, L.M., Sinha, M.C. & Constable, S.C., 2000. Electromagnetic images of the Valu Fa Ridge, Lau Basin, Geoscience 2000, Geol. Soc. London, 75.

Morton, J.L. & Sleep, N.H., 1985. Seismic reflections from a Lau Basin magma chamber, In: D.W. Scholl and T.L. Vallier (eds.), Geology and offshore resources of Pacific Island Arcs - Tonga region, Earth Sci. Ser., vol. 2, Circum-Pacific Council for energy and mineral resources, Houston, Texas, 441-453.

Murase, T. & McBirney, A.R., 1973. Properties of some common igneous rocks and their melts at high temperatures, Geol. Soc. Am. Bull., 84, 3 563-3 592.

Nafe, J.E. & Drake, C.L., 1957. Variation with depth in shallow and deep water marine sediment of porosity, density and the velocities of compressional and shear waves, Geophysics, 22, 523-553.

Navin, D.A., 1996. Seismic investigation of crustal accretion at the slow-spreading Mid-Atlantic Ridge - The Reykjanes Ridge at 57° 45' N, unpublished Ph.D. thesis, University of Durham, pp.309.

Northwest Geophysical Associates Inc., 1995. GMSYS users guide, pp.175.

Offer, A.R., 1996. Geophysical investigation of the off-axis crustal structure of the Lau Basin (SW Pacific), unpublished M.Sc. thesis, University of Durham, pp.119.

Parker, R.L., 1972. The rapid calculation of potential anomalies, Geophys. J. R. astr. Soc., 31, 447-455.

Parson, L.M. & Wright, I.C., 1996. The Lau-Havre-Taupo back-arc basin: A southward-propagating, multi-stage evolution from rifting to spreading, Tectonophysics, 263, 1-22.

Parson, L.M., Pearce, J.A., Murton, B.J., Hodkinson, R.A. and RRS Charles Darwin Scientific Party, 1990. Role of ridge jumps and ridge propagation in the tectonic evolution of the Lau back-arc basin, Southwest Pacific, Geology, 18, 470-473.

Peirce, C., Day, A.J., & Turner, I.M., 2000. Spreading vs. stretching: rift propagation and ridge structure in the southern Lau Basin, Geoscience 2000, Geol. Soc. London, 75.

Peirce, C., Sinha, M.C. & Constable, S., 1996. Controlled source electromagnetic and seismic investigation of the Valu Fa Ridge, Lau Basin, SW Pacific, R/V Maurice Ewing 9512 Cruise Report, (unpublished), pp.56.

Phipps Morgan, J. & Forsyth, D.W., 1988. Three-dimensional flow and temperature perturbations due to a transform offset: effects on oceanic crustal and mantle structure, J. geophys. Res., 93, no. B4, 2955 -2966.

Prince, R.A. & Forsyth, D.W., 1988. Horizontal extent of anomalously thin crust near the Vema fracture zone from the 3-D analysis of gravity anomalies, J. geophys. Res., 93, no. B7, 8 051-8 063.

Rushforth, I.M., 1996. A seismic investigation of crustal properties near the Central Valu Fa Ridge, Lau Basin, SW Pacific, unpublished M.Sc. thesis, University of Durham, pp.91.

Sandwell, D.T. & Smith, W.H.F., 1997. Marine gravity anomaly from Geosat and ERS-1 satellite altimetry, J. geophys. Res., 102, no. B5, 10 039-10 054.

Sinha, M.C., 1995. Segmentation and rift propagation at the Valu Fa Ridge, Lau Basin: Evidence from gravity data, J. geophys. Res., 100, no. B8, 15 025-15 043.

Solomon, S.C. & Toomey, D.R., 1992. The structure of mid-ocean ridges, Ann. Rev. Earth planet. Sci., 20, 329-364.

Sunkel, G., 1990. Petrology and geochemistry of submarine lavas from the Lau Basin (S.W. Pacific), Mar. Min., 9, 205-234.

Talwani, M. & Ewing, M., 1960. Rapid computation of gravitational effect of three-dimensional bodies of arbitrary shape, Geophysics, 25, 203-225.

Talwani, M., Worzel, J.C. & Landisman, M., 1959. Rapid gravity calculations for two-dimensional bodies with application to the Mendocino submarine fracture zone, J. geophys. Res., 64, 49-59.

Tolstoy, M., Harding, A.J. & Orcutt, J.A., 1993. Crustal thickness on the Mid-Atlantic Ridge: bull's-eye gravity anomalies and focussed accretion, Science, 263, 726-729.

Turner, I.M., 1998. Crustal accretionary processes at mid-ocean ridges - Valu Fa Ridge, Lau Basin, unpublished Ph.D. thesis, University of Durham, pp.263.

Turner, I.M., Peirce, C. & Sinha, M.C., 1999. Seismic imaging of the axial region of the Valu Fa Ridge, Lau Basin - the accretionary processes of an intermediate back-arc spreading ridge, Geophys. J. Int., 138, 495-519.

Vallier, T.L., Jenner, G.A., Frey, F.A., Gill, J.B., Davis, A.S., Volpe, A.M., Hawkins, J.W., Morris, J.D., Cawood, P.A., Morton, J.L., Scholl, D.W., Rautenschlein, M., White, W.M., Williams, R.W., Stevenson, A.J. & White, L.D., 1991. Subalkaline andesite from Valu Fa Ridge, a back-arc spreading centre in the Southern Lau Basin: petrogenesis, comparative chemistry and tectonic implications, Chemical Geology, 91, 227-256.

von Stackleberg, U. and the Shipboard Scientific Party, 1988. Active hydrothermalism in the Lau back-arc basin (SW Pacific): First results from the SONNE 48 Cruise, Marine Mining, 7, 431-442.

Wang, X., Cochran, J.R. & Barth, G.A., 1996. Gravity anomalies, crustal thickness and the pattern of mantle flow at the fast spreading East Pacific Rise, 9° -10° N: Evidence for three-dimensional upwelling, J. geophys. Res., 101, no. B8, 17 927-17 940.

Wessel, P. & Smith, W.H.F., 1995. New version of the Generic Mapping Tools released, EOS Trans. AGU, 76, 329.

White, R.S., McKenzie, D.P. & O'Nions, R.K., 1992. Oceanic crustal thickness from seismic measurements and rare earth element inversions, J. geophys. Res., 97, no. B13, 19 683-19 715.

Wiedicke, M. & Collier, J.S., 1993. Morphology of the Valu Fa spreading ridge in the Southern Lau Basin, J. geophys. Res., 98, no. B7, 11 769-11 782.

Zellmer, K.E., Taylor, B., Martinez, F. & Goodliffe, A., 1998. Complexities in the seafloor spreading history of the Lau Basin, EOS Trans. AGU, 79, no. 45, 908.

FIGURE CAPTIONS

Figure 1. Location map of the Central Lau Spreading Centre (CLSC) and Eastern Lau Spreading Centre (ELSC) within the Lau Basin showing the Valu Fa Ridge crustal study (after Turner, 1998). The two spreading axes are delineated by parallel lines and divergent arrows, based on magnetic anomaly identification, GLORIA sidescan sonar and SeaBEAM bathymetry data. Note the asymmetric position of the ELSC within the Lau Basin, proximal to the active Tofua Arc (Tonga Ridge). The Valu Fa Ridge marks the southern extension of the ELSC and is also annotated. The main tectonic and morphological features are identified (cf. inset) and the location of the EW9512 survey area is marked by the shaded box. Bathymetric contours are plotted at 1000 m (dashed line) and 2000 m (solid and annotated line) intervals. Inset shows the tectonic and morphological features of the plate boundary between the Pacific and Indo-Australian Plates. Plate boundaries are identified (solid and dashed lines) and the major trenches, ridges and basins annotated. The triangular-shaped Lau Basin is centred on 20º S 177º W and separates the Lau and Tonga Ridges. Pacific plate subduction rates at the Tonga-Kermadec trench are also shown (after Parson & Wright, 1996). Note that the present subduction point of the Louisville Ridge coincides with the southern extension of the Lau Basin. Bathymetric contours are plotted at 2000 m intervals.

Figure 2. The Valu Fa Ridge crustal study (after Peirce et al., 1996; Turner et al., 1999). Experimental configuration overlying the bathymetry (Zellmer et al., 1998) of the study area. Profile lines are shown by solid (Seismic South) and dashed (Seismic North) lines and the main 2-D seismic and gravity profiles discussed in this paper are annotated. Seabed depths shallower than 2000 m have been shaded to show the location of the CVFR and NVFR, and ocean bottom seismometer (OBS) and sonobuoy locations are marked by triangles and stars respectively.

Figure 3. Preferred across-axis velocity-depth model for Line 1, Seismic South (after Turner et al., 1999). Solid lines indicate the location and geometry of the seafloor, sediment, the layer 2A/2B boundary, the layer 2/3 boundary, the low velocity block and the seismic Moho (constrained between -30 and 30 km model offset - thicker grey line). The model is also divided into five lateral regions (see text for details). Dotted lines represent isovelocity contours at 0.5 km s^-1 (layer 2) and 0.25 km s^-1 (layer 3) intervals. OBS positions are marked by triangles and intersection points with the along-axis profile (Line 4) and two axis-parallel profiles are indicated by vertical dashed lines. Velocity-depth profiles are calculated at ~10 km intervals. Note the lower velocities within layer 3 beneath the axial region and the presence of a low velocity block at a depth of ~5 km (cf. notch in the velocity-depth profile at 0 km model offset). Vertical exaggeration x2.

Figure 4. Preferred across-axis 2-D free-air gravity model of Line 1, Seismic South (after Turner, 1998; Turner et al., 1999), using average P-wave seismic velocities, extracted from the velocity-depth model shown in Fig. 3, to estimate initial densities. Note that the layer boundaries remain unchanged from the velocity-depth model. (a) Comparison of the observed (dots) and calculated (solid line) free-air gravity anomalies. Dot size gives an indication of the error bars (±1.5 mGal). The calculated gravity anomaly of the initial model (dashed line) is shown for reference and displays the necessity of inclusion of reduced densities at either end of the profile. Note the general good fit of the main anomaly peaks and the slight misfit centred on 14 km model offset is attributed to the large seamount north of the profile. (b) Preferred density model with layer densities shown in g cm^-3. The model extends down to a depth of 15 km and has a semi-infinite half-space at either end of the profile. Note the presence of lower densities beneath the island arc region towards the eastern end of the model and higher densities in layer 3 within the post-rift region. The thicker grey shaded line shows the extent of Turner et al. (1999)'s seismic control on Moho depth. Vertical exaggeration x2.

Figure 5. Preferred across-axis 2¾-D free-air gravity model of Line 6, Seismic North. Display parameters are described in detail in Fig. 4. (a) Comparison of the observed (dots) and calculated (solid line) free-air gravity anomalies. The calculated anomaly from 2-D gravity modelling (dashed line) is included for comparison. Note the better fit of the calculated anomaly generated by 2¾-D modelling west of -5 km model offset. (b) Preferred density model with layer densities shown in g cm^-3. 2¾-D input parameters are also described with minimum and maximum y-lengths, shown by subscripts (on-line structures) and superscripts (off-line structures) respectively. Stars (-*/*) denote infinite lateral extents.

Figure 6. Preferred along-axis 2¾-D free-air gravity model of Line 4. Display parameters are described in Figs. 4 and 5. (a) Comparison of the observed (dots) and calculated (solid line) free-air gravity anomaly. Note the greatly improved fit of the calculated anomaly generated by a 2¾-D approach. The calculated anomaly from 2-D gravity modelling (dashed line) and the calculated contribution of the seabed in three dimensions (dotted line) are also included for reference. (b) Preferred density model with layer densities shown in g cm^-3. Note that Turner et al. (1999)'s seismic modelling provides constraint on crustal layer thickness at two locations (dotted-dashed vertical lines).

Figure 7. Free-air gravity anomaly map calculated from the EW9512 and CD34/88 data overlying the 2000 m seabed bathymetry contour to show the location of the Valu Fa Ridge segments. The location of the Wiedicke & Collier (1993) pseudofaults are also shown for reference. Contour interval is 5 mGal. Note the general agreement between the topography and observed free-air gravity anomaly. Inset shows the location of the gravity profiles used to create the free-air gravity grid. The current location of the propagating rift tip is shown by the dotted line centred at 22° 55?S.

Figure 8. Model used as input for calculation of the gravitational attraction of a layered crust of constant density and thickness (dotted lines). Layer thicknesses and densities are shown in km and g cm^-3 respectively. The preferred density model for Line 1 (solid lines) is shown for comparison.

Figure 9. Calculation of the gravitational effect of a layered crust of constant thickness and density. The axes of the rotated x-y co-ordinate system are annotated in km and seabed depths shallower than 2000 m have been shaded to show the location of the Valu Fa Ridge segments. Bathymetric contours are plotted at 250 m intervals and gravity anomaly contours are plotted at 2 mGal intervals. (a) Rotated extract from the bathymetry dataset of Zellmer et al. (1998) used as a basis of the MBA calculation. (b) Zellmer et al. (1998)'s Lau Basin bathymetry grid in geographic co-ordinates showing the location of the region extracted for the MBA calculation. (c) Calculated gravitational attraction of the seabed. A profile taken from this anomaly along Line 4 is shown in Fig. 6. (d) Combined gravitational attraction of the seabed, layer 2A/2B and layer 2/3 boundaries and the Moho. The attraction due to the seabed (c) contributes to ~70% of the total calculated field. Note the large gravitational effect of the OSC centred on -2, 26 grid co-ordinates.

Figure 10. Mantle Bouguer anomaly calculated by subtracting the gravitational effect of a layered crust of constant thickness and density (Fig. 9d) from the observed free-air gravity anomaly (Fig. 7). Contour interval is 5 mGal. The 2000 m bathymetric contour is plotted for reference. Note the general decrease in anomaly towards the OSC (22° 12´ S) and the general increase towards the propagating rift tip (dotted line). Of note also are the increase in anomaly associated with the pseudofault (marked A) and the decrease in anomaly to the east of the CVFR (marked B). Inset shows the MBA calculation of Sinha (1995) using the forward modelling algorithm of Talwani & Ewing (1960) and the CD34/88 free-air gravity dataset alone for comparison. Note that the Sinha (1995) MBA still contains a topographic component, which is not apparent in the MBA of this study.

Figure 11. Calculation of the thermal effect of a north-south ridge-transform-ridge system assuming passive upwelling (Phipps Morgan & Forsyth, 1988 as developed from Forsyth & Wilson, 1984). Display parameters are described in Fig. 9. (a) Location of the ridge-transform-ridge system (thick solid lines) used to calculate the thermal anomaly overlying the bathymetry which is shaded in regions shallower than 2000 m. Ridges are modelled with an east-west half-spreading rate of 30 mm yr^-1 and are identified from the EW9512 bathymetry data and from previous GLORIA side-scan sonar data (Parson et al., 1990) collected in the region. (b) Thermal anomaly due to passive upwelling (see text for details). Contour interval is 0.2 mGal. Note the inherent periodicity in the thermal anomaly as shown by the general s-shape of the anomaly pattern.

Figure 12. Residual mantle Bouguer anomaly calculated by subtracting the thermal effect of passive upwelling at the Lau Basin ridge-transform-ridge system (Fig. 11b) from the mantle Bouguer anomaly (Fig. 10). Contour interval is 5 mGal. The 2000 m bathymetric contour is also plotted for reference. The propagating rift tip is shown by the dotted line. Note that the removal of the thermal component does not remove the general decrease in anomaly towards the OSC (22° 12´ S) and the increase towards the propagating rift tip as observed in the MBA (Fig. 10).

 Figure 13. Along-axis 2-D gravity profile sampled from the RMBA anomaly plot of Fig. 12. (a) Comparison of the observed (long dashed line), smoothed observed (dots) and calculated (solid line) gravity anomalies. Dot size gives an indication of error bars. Modelling shows that the 2-D trend can simply be explained by variations in crustal thickness along-axis as shown in b). The anomaly (short dashed line) calculated assuming a constant thickness layered crust is included for reference. (b) Preferred density model with layer densities shown in g cm^-3. The model extends down to a depth of 15 km and has a semi-infinite half-space at either end of the profile. The RMBA calculation takes account of the crustal attraction of the water column and hence this layer is not included. The four remaining layers - layer 2A, layer 2B/C, layer 3 and the upper mantle - were modelled with densities equivalent to those in the RMBA crustal model. Layer thicknesses were allowed to vary north and south of reference points obtained from the across-axis and along-axis seismic modelling of Turner et al. (1999) in 2-D and 3-D by Day et al. (2000) (i.e. between the vertical dashed lines). The latter study has shown that the thickness of layer 2A varies in proportion to the variation in layer 3 and thus this constraint has also been applied along the entire length of the model. The location of the propagating rift tip, Southern, Central and Northern Valu Fa Ridges, together with the OSC are shown for reference. Note that thicker crust is associated with these ridge segments and the OSC, and that thinning is observed towards, and south of, the propagating rift tip. Of interest also is the observation that, beneath the northern half of the Northern Valu Fa Ridge, the crust thins again towards its northern ridge tip. This observation suggests that the general northwards trend of increasing crustal thickness away from the propagating rift tip extends only as far as approximately the southern end of the Central Valu Fa Ridge, whose location appears to mark the onset of a mature and stable spreading ridge system.