Equatorial Pacific gravity lineaments: interpretations with basement topography along seismic reflection lines
The central equatorial Pacific is interesting for studying clues to upper mantle processes, as the region lacks complicating effects of continental remnants or major volcanic plateaus. In particular, the most recently produced maps of the free-air gravity field from satellite altimetry show in greater detail the previously reported lineaments west of the East Pacific Rise (EPR) that are aligned with plate motion over the mantle and originally suggested to have formed from mantle convection rolls. In contrast, the gravity field 600 km or farther west of the EPR reveals lineaments with varied orientations. Some are also parallel with plate motion over the mantle but others are sub-parallel with fracture zones or have other orientations. This region is covered by pelagic sediments reaching ~ 500–600 m thickness so bathymetry is not so useful for seeking evidence for plate deformation across the lineaments. We instead use depth to basement from three seismic reflection cruises. In some segments of these seismic data crossing the lineaments, we find that the co-variation between gravity and basement depth is roughly compatible with typical densities of basement rocks (basalt, gabbro or mantle), as expected for some explanations for the lineaments (e.g., mantle convection rolls, viscous asthenospheric inter-fingering or extensional deformation). However, some other lineaments are associated with major changes in basement depth with only subtle changes in the gravity field, suggesting topography that is locally supported by varied crustal thickness. Overall, the multiple gravity lineament orientations suggest that they have multiple origins. In particular, we propose that a further asthenospheric inter-fingering instability mechanism could occur from pressure variations in the asthenosphere arising from regional topography and such a mechanism may explain some obliquely oriented gravity lineaments that have no other obvious origin.
KeywordsOceanic thermal subsidence Oceanic plate isostasy Free air gravity anomalies Pacific plate deformation Mantle dynamics
Haxby and Weissel (1986) discovered a series of lineaments in the marine gravity field recorded using satellite altimetry, forming elongated troughs and swells of 5–20 mGal amplitude spaced ~ 200 km in the Pacific and Indian Oceans. As the lineaments in the Pacific Ocean were parallel to the present plate movement over Earth’s mantle and cross fracture zones (“cross-grain”), they were initially speculated to be caused by elongated small-scale convection rolls in the asthenospheric mantle, which were expected to develop within the first few million years (Buck and Parmentier 1986; Haxby and Weissel 1986). That convection would provide vertical stresses that initially distort the overlying lithosphere (Fig. 1a), creating topography that becomes “frozen in” by the cooling plate. Subsequently, other authors have characterised and commented on these and larger similar features in satellite altimetry data (Baudry and Kroenke 1991; Cazenave et al. 1992; Fleitout and Moriceau 1992; Maia and Diament 1991; McAdoo and Sandwell 1989; Sandwell et al. 1995; Wessel et al. 1996) and the convection model has received renewed interest more recently (Ballmer et al. 2009; Harmon et al. 2006).
Other explanations for the gravity lineaments have been put forward. Winterer and Sandwell (1987) suggested that extension arising from lithospheric cooling (Turcotte and Oxburgh 1973) could explain en-echelon volcanic ridges found within the Pacific plate that are parallel to these gravity lineaments, with the lithosphere deforming in a boudinage manner (Fig. 1b). Alternatively, extension could arise from far-field stresses caused by slab pull associated with the Pacific plate’s subduction zones (Dunbar and Sandwell 1988). Goodwillie and Parsons (1992) used separations of fracture zones at different locations to assess plate extension in the South Pacific about the East Pacific Rise (EPR), but ruled out significant extension as explaining lineaments there, a conclusion similarly reached by Gans et al. (2003). Gans et al. (2003) suggested instead that flexing of the lithosphere caused by varied cooling rates with depth (Fig. 1c) could explain the gravity lineament spacings and amplitudes. Cormier et al. (2011) presented multibeam data showing faults that support the possibility of such extension of the Cocos plate.
To explain a correspondence between slow seismic velocities beneath seamount chains of an area of the southern EPR, Weeraratne et al. (2007) and Holmes et al. (2007) suggested that a viscous inter-fingering could occur in asthenosphere affected by an off-axis mantle thermal anomaly. Essentially, buoyancy in asthenosphere off-axis causes it to rise towards the axis beneath the dipping lithospheric lid. Fingers of the mobile asthenosphere may explain volcanic seamount chains (Fig. 1a) that do not age systematically as expected of traditional mantle hotspot origins (Ballmer et al. 2009).
One of the most recently released grids of the marine gravity field (Sandwell et al. 2014) shown in Fig. 2b reveals the lineaments originally identified by Haxby and Weissel (1986). The greater resolution of this new grid reveals a finer structure immediately west of the EPR than in the earlier altimeter data, with troughs and swells spaced as closely as ~ 20 km. Figure 3c (described later) shows a high-pass filtered version of the gravity data. In it, the lineaments west of the EPR can be seen giving way westwards into a pattern of lineaments that is more confused though with three or more dominant orientations (red arrows annotated A-F), which we focus on here.
In this study, we take advantage of interpreted seismic reflection data from three research cruises (Bloomer et al. 1995; Dubois and Mitchell 2012; Eittreim et al. 1994; Gnibidenko et al. 1990) and a recent compilation of seafloor spreading anomalies (Barckhausen et al. 2013) to evaluate the extent to which basement relief varies with gravity anomalies across these lineaments, supporting some of the explanations for the lineaments that have been put forward. The model profiles shown in Fig. 1 suggest that, as the upper surface of the plate is a major density contrast, its deformation by the various mechanisms should lead to gravity anomalies varying sympathetically by amounts expected of basement with a high density typical of mantle rocks. This assumes that the crust has uniform thickness so it contributes a nearly uniform amount to the gravity field. In practice, however, the crust may vary in thickness, in particular, where volcanism has been promoted (e.g., at locations marked “v?” in Fig. 1), hence the comparison is not straightforward. As part of the work, we also re-evaluate the Pacific tectonic plate subsidence. Previous subsidence studies (Crosby and McKenzie 2009; Parsons and Sclater 1977; Trehu 1975) have tended to use gridded datasets [e.g., sediment thickness from seismic reflection data (Divins 2003; Ludwig and Houtz 1979; Whittaker et al. 2013)], where errors may have been possible due to incomplete sampling and reflective chert horizons (Mitchell 1998). The new results allow us to verify those earlier subsidence rates and locate where the seabed lies above or below those trends.
Data and methods
Seismic reflection data
Seismic reflection data were obtained during the AMAT03 cruise of RV Revelle in 2004, which was a site-survey cruise for IODP Expedition 320/321. Recordings were made with a 4-channel streamer towed at ~ 10 knots between sites and with a 48-channel streamer towed at ~ 6 knots at the drill sites. The seismic source comprised two 150 cubic inch generator-injector (GI) guns. A segment of the data is shown in Fig. 4, illustrating the typical visibility of the basement reflection. The full dataset is too large to present here, though several images of the data covering > 1000 km were shown by Tominaga et al. (2011) and Dubois and Mitchell (2012), and the full dataset can be obtained from the UTIG Academic Seismic Data Portal (http://www.ig.utexas.edu/sdc/).
The recorded seismic data were interpreted as described by Dubois and Mitchell (2012) (also see Mitchell and Dubois 2014), including the basement reflection two-way time, which is plotted against ship tracks in Fig. 5. To compute accurately the depth of the basement below the seabed from the interval two-way time t, allowance needs to be made for variations of seismic velocity with sediment depth because of compaction and diagenesis in these sediments (carbonate ooze). Mayer et al. (1985) estimated in situ velocity from logged core velocity data from the central equatorial Pacific, correcting mainly for loss of confining pressure. The velocity variation with depth [~ 1500 m s−1 at the surface to > 2000 m s−1 below 400 m below seafloor (mbsf)] was approximated with a linear trend, V = V0 + Bz, where z is depth below seabed, and coefficients V0 and B found by least-squares regression (V0 = 1479 m s−1; B = 1.0286 s−1). The following equation derived by integration was used here to estimate z:
As described by Eittreim et al. (1994), the RV Akademik Selskiy data were collected with a 24-channel, 2400-m streamer with airgun arrays of 23 or 46 L and mostly shot with 24-fold along the line marked in Fig. 2. After processing, the data revealed Moho, crustal and basement reflections. The seabed and basement reflection two-way times interpreted by Eittreim et al. (1994) were digitized and converted to depth using a sediment velocity of 1540 m s−1 typical of ODP Site 1219 measurements (Lyle et al. 2002) and a water velocity of 1500 m s−1. Eittreim et al. (1994) also identified flat-lying Moho reflections and modestly dipping reflections in the lower crust. Two-way times from basement to the Moho reflections were converted to depth with a simplified oceanic velocity structure of an upper crust of constant 2.5 km thickness with velocity (Vp) of 5 km s−1 overlying a gabbro-dominated lower crust of varied thickness and velocity of 7 km s−1 (Mutter and Mutter 1993). A segment of the reflection interpretations are shown in Fig. 6.
Magnetic anomaly identifications
Magnetic anomalies have been re-evaluated in this region by Barckhausen et al. (2013), exploiting the greater coverage provided by recent research vessel transits. One set of transits involved a vector magnetometer, which aided anomaly identifications in this low field area near the magnetic equator. The density of identifications (red crosses in Fig. 2a) was significantly improved compared with previous studies.
Crustal ages corresponding with their magnetic anomaly identifications and some of their isochrons around fracture zones were gridded using a surface-fitting program (Smith and Wessel 1990). Figure 2a shows age contours derived from that grid. The grid was then sampled along the seismic survey tracks to obtain crustal age at each seismic measurement. (Crustal age contours in Fig. 2a do not perfectly honour the isochrons of Barckhausen et al. (2013) over the whole area, but the grid was constructed to match faithfully isochrons where they underlie the seismic ship tracks.) For the Akademik Selskiy data, crustal ages were similarly assigned using the isochrons of Barckhausen et al. (2013) and, where they were not available (< 14 Ma and > 76 Ma), from the age grid of Müller et al. (2008).
Basement depths derived from the AMAT03, Venture-1 and Akademik Selskiy cruises were corrected for isostatic sediment loading using a sediment wet bulk density of 1.4 g cm−3 [a column-average of IODP core measurements (Lyle et al. 2002; Pälike et al. 2010)] and a mantle density of 3.3 g cm−3. Those corrected depths are shown in Figs. 7 and 8 versus the square root of seafloor age. Simple least-squares regressions of the data was used to find the subsidence rates, with that in Fig. 8 calculated from data restricted to < 8 Ma1/2 where basement depths break from the subsidence trend. Basement elevation residuals about the regression lines of Figs. 7 and 8 are shown along-track in Figs. 5 and 3 (these are residual elevations rather than residual depths as positive values lie above the regression line).
Free air gravity anomalies were derived from satellite altimeter measurements (Sandwell and Smith 1997). The data shown in Fig. 2b are from version 23 of the free-air anomaly grid available from http://topex.ucsd.edu/. According to Sandwell et al. (2014), this version has an average accuracy of ~ 2 mGal in deep water, with most of the improvement over earlier versions in the 12–40 km wavelength band and improved resolution of features to as small as 6 km.
For crust that is locally Airy supported, variations in basement topography have little effect on the free-air gravity field because topography is supported by thickened crust of lower density than the underlying mantle (i.e., the increased gravitational attraction from the elevated topography is almost balanced by the reduced attraction from the greater crustal root which displaces dense mantle rocks). However, short-wavelength topography can be supported by the rigidity of the lithosphere so it can affect the gravity field. Free-air anomalies that are high-pass filtered therefore reflect crustal density and topography variations, whereas low-pass filtered anomalies should reflect potential deeper mantle sources (non-Airy supported topography). Relevant to the filter length-scale required, according to Cochran (1979), the elastic thickness of the lithosphere is 2–6 km about the EPR. Modelling by Smith (1998) suggests that little effect of crustal variations appear in the gravity field for spatial scales larger than 200 km for these elastic thicknesses. The gravity field (Fig. 2b) was therefore convolved with a cosine-tapered weight of 400 km full width (Wessel and Smith 1991), which contains most weight over 200 km distance. The result in Fig. 3a was subtracted from the field (Fig. 2) to produce the high-pass filtered gravity field in Fig. 3b. Thus, Fig. 3a is intended to represent deeper (mantle) sources, whereas Fig. 3b should represent mainly lateral variations in basement topography, or crustal densities and thickness.
Line segments V1, A2, A4 and A7 located in Fig. 2a run along crustal isochrons, so any effect of lithospheric thermal subsidence should be uniform along them. For discussion of variations of gravity with basement structure, Fig. 9 shows cross sections of basement and seabed topography along with the unfiltered free-air anomaly sampled from the grid provided by Sandwell et al. (2014).
The bathymetry grid of Ryan et al. (2009) was used for h. The grid has varied resolution depending on the distribution of the tracks of cruises contributing bathymetry, although the multibeam data from both the AMAT03 and Venture-1 cruises have been incorporated in the grid, so resolution is adequate for the Bouguer correction along those lines. Resolution is poorer along the RV Akademik Selskiy line. Bouguer anomalies computed using depths derived from the seismic data differed from those computed with the Ryan et al. (2009) grid with a root-mean-square (RMS) difference of 2.3 mGal. It was therefore decided to use the former Bouguer anomalies for comparison with the basement variations. High-pass filtered anomalies shown by the red line in Fig. 6 were computed by fitting a 5th-order polynomial of Bouguer anomaly versus longitude and calculating residuals relative to that polynomial.
To help assessment of any co-variation between Bouguer anomalies and basement elevations, these data for each line are shown in Fig. 10 (admittance plots), using the residual elevations about the subsidence trends for basement elevation (Figs. 7, 8). In these graphs, Bouguer anomalies are shown after removing the 400-km-filtered field, but the elevations have not been similarly high-pass filtered. In the case of the Akademic Silskiy line, data were restricted to 113°W to 145°W to avoid influence of some features near the EPR visible in Fig. 2b and the Line Islands Swell to the west. Data from the two segments of apparently co-varying basement elevation and Bouguer anomalies marked by horizontal grey lines in Fig. 6 are shown with red and blue colour in this panel (see figure caption for details).
In admittance plots constructed as described above, we expect changes in basement elevation with associated crustal thickness variations (local Airy isostatic support) to form horizontal trends. On the other hand, deflections of basement caused by mechanisms such as shown in Fig. 1 should lead to a local correlation of Bouguer anomaly with basement elevation and the gradient of the correlation trend should correspond with a density contrast of mantle rocks with the sediments, if the crust has uniform thickness and therefore contributes little (also note that the Bouguer correction in effect replaced water with sediment in terms of its effect on the gravity field, hence the contrast is with sediment). If topography of the basement is supported by the rigidity of the lithosphere (e.g., because of off-axis emplacement of seamounts or faulting), other correlations are possible and hence we provide model trends in Fig. 10 for gabbro and basalt lithologies (3.0 and 2.7 g cm−3) as well as mantle lithologies (3.3 g cm−3). Complicating somewhat the interpretation, the long-wavelength variation in basement elevation caused potentially by large-scale mantle convection (Crosby and McKenzie 2009; Crosby et al. 2006) will spread out the data generally parallel to the horizontal axis (any gravity effect having been removed by the high-pass filtering). However, this spreading out of the data is useful as it helps visibility of trends due to basement relief in smaller segments of the data.
In order to confirm the seismically derived structure, while also providing a test for the approach of providing calculations using the Bouguer slab formula, model free-air anomalies (green line in Fig. 6) were calculated for the Akademik Selskiy line where Moho depths are available by adding the gravity contributions (using Eq. 2) of the water column, sediment and crust using densities of 1.0, 1.4, 3.0 and 3.3 g cm−3 for water, sediment, crust and mantle, respectively. The somewhat large 3.0 g cm−3 was used for the crust given that crustal seismic layer 3 (gabbro) dominates variations in seismic estimates of crustal thickness (Mutter and Mutter 1993), whereas the seismic layer 2 is more uniform and therefore variations in layer 2 thickness should have less affect on gravity field variations than variations in the gabbroic layer thickness. Minshull et al. (1998) derived similar densities of ~ 3.0 g cm−3 for lower oceanic crust using their ~ 7 km s−1 velocities from seismic refraction experiments and a relation of Christensen and Shaw (1970), which Minshull et al. (1998) suggested was accurate to within a few percent. As the deeper mantle density structure is unknown here, the resulting model was simply de-trended and then re-trended to the gradient of the observed free-air anomalies. The resulting model and data have a root-mean square difference of 2.3 mGal, which is similar to the ~ 2 mGal accuracy of the gravity grid (Sandwell et al. 2014). The small discrepancy could partly arise from inadequate modelling of the crustal densities [e.g., seamounts may have a lower density (Mitchell 2001)] and/or from the Bouguer slab formula approach, which does not account well for topographic relief and crustal structure away from each point of calculation along the profile.
Results and interpretation
In the following, we first characterize the large-scale systematic variations before examining the variability due to the gravity lineaments and other (e.g., volcanic structures) superimposed on those major trends.
The subsidence rate obtained from the AMAT03 and Venture-1 data (Fig. 7) is 313 m m.y.−1/2, which is similar to that of Marty and Cazenave (1989) who found 329 m m.y.−1/2 for the equatorial Pacific (their corridor 26). It is also similar to the global average rate of 323–336 m m.y.−1/2 of Korenaga and Korenaga (2008), to the 307 m m.y.−1/2 North Pacific rate of Hillier and Watts (2005) and to the 320 m m.y.−1/2 average rate for the Pacific of Crosby and McKenzie (2009). The latter is perhaps surprising, as Crosby and McKenzie (2009) selected depths where they coincided with free air anomaly magnitudes < 5 mGal, to reduce effects of mantle upwelling and downwelling on their subsidence rates. The AMAT03 seismic lines cross a broad gravity low (114°–123°W) but encroach on gravity highs in the west and east ends of the analysed region (Fig. 3a), so any effect of downwelling may be fortuitously compensated by upwellings to leave a similar subsidence trend as Crosby and McKenzie (2009) found.
The subsidence rate derived from the Akademik Selskiy dataset was 343 m m.y.−1/2, which is somewhat more rapid than the previous estimates outlined above. This is perhaps surprising if we expect a dynamic effect associated with the long-wavelength free-air anomalies in Fig. 3, as the dataset spans ~ 30 mGal from a − 20 mGal low in the east to nearly + 10 mGal in the west. If the east of the line were dynamically depressed by mantle downwelling, while the west of the line were elevated by upwelling near the Line Islands swell, a reduced apparent subsidence rate might be expected.
We initially calculated 2σ uncertainties for our regressions using standard methods (Taylor 1982) and obtained values of 2 and 4 m m.y.−1/2 for Figs. 7 and 8. These values are much smaller than for the other subsidence rates quoted above, whose uncertainties were reported to be 61 m.y.−1/2 (Marty and Cazenave 1989), 22–23 m m.y.−1/2 (Korenaga and Korenaga 2008), 21–29 m m.y.−1/2 (Hillier and Watts 2005) and 30 m m.y.−1/2 (Parsons and Sclater 1977). Our small nominal uncertainties arise from the large numbers of data contributing to the regressions. However, adjacent depth values along survey lines obtained by digitising the seismic data are not strictly speaking independent of each other because of the finite length-scales of abyssal hills and other features (Goff 1992). Furthermore, the seismic lines severely under-sample basement across the region. Hence, true uncertainties will be larger. However, these issues highlight problems with the earlier subsidence rates also, as the authors have used interpolated datasets where the underlying variations (e.g., sediment thickness) were not measured continuously and there was no allowance made apparently for the strong averaging of some data that occurs during gridding. We therefore suggest that uncertainties of subsidence trends are not so well constrained in ours and in the other studies. The significance of differences in subsidence rates between regions may in turn be difficult to assess.
The residual elevations of basement corrected for sediment loading derived from the seismic data have similar magnitudes and vary with a similar spatial pattern to those of Crosby et al. (2006), as shown in Fig. 11. Larger positive relief is evident along line V1 and generally larger amplitudes occur along lines A2–A7 due to the new data resolving seamounts and other elevated features.
Relations between basement elevations and gravity anomalies
Basement elevations varying sympathetically with bouguer anomalies
In some segments, the basement rises and falls sympathetically with the Bouguer anomalies, as would be expected from some of the origins of the gravity lineaments (Fig. 1). This can be seen in line A2 (Fig. 9a), where local gravity highs “a”–“e” mimic basement highs. The corresponding admittance plot for this line (see filtered data in red in Fig. 10) contains trends that generally have lower graph-gradients than the trends expected with the various basement densities, although the anomalies in Fig. 9a are small in extent so this may be due to upward continuation effects.
The graph for profile A3 in Fig. 10 contains varied trends that could be compatible with basement densities smaller than 2.7 to larger than 3.3 g cm−3 (i.e., large gravity variations occur in Fig. 9b with modest variations in basement elevation). In Fig. 9a, profile A4 mostly does not show sympathetic behaviour, although anomaly “a” and the further anomaly immediately south of it along that line do show sympathetic variations. In the admittance plot (Fig. 10) some trends for A4 are nearly parallel to the expected trends.
The graph for profiles A5 and A6 in Fig. 10 mostly shows trends that are equal to or shallower than the 2.7 g cm−3 trend. That for A7 is more confused, though the profile in Fig. 9a suggests some sympathetic variation.
The Akademik Selskiy data also contain two segments where there are basement-Bouguer correlations (marked “c” and “d” in Fig. 6 and highlighted by grey bars). The corresponding graph in Fig. 10 contains data segments with the expected gradients for basement densities.
Basement topography with little or no bouguer anomaly variation
In a few segments of the data, we find basement elevated by a few 100 m or more coinciding with subdued gravity anomalies. For example, at “c” in profile A5–A6 (Fig. 9b), there is a basement high of a few 100 m with two extreme local highs (seamounts) that is associated with a weak Bouguer high of only ~ 4 mGal. At 1°N in profile A3, a similarly extreme relief is associated with a gravity low. At 5°N in profile A7, a basement high of > 500 m is accompanied by negligible Bouguer anomaly.
These features are most likely caused by thickened crust underlying seamounts or groups of seamounts. Figure 12a shows seamounts occurring in bathymetry collected with multibeam sonars in the northerly parts of profiles A4–A7. In the Akademik Selskiy data, local basement highs EH1 and EH2 are also associated with subdued Bouguer anomalies (Fig. 6). The Moho imaged seismically beneath EH1 appears to be 500–1000 m deeper than areas a few degrees of longitude either side of the high. There is less clearly local deepening of the Moho beneath EH2, though the Moho is noticeably shallower under the adjacent low EL2. [Although an explanation is unclear, EH2 interestingly coincides with the change in Pacific plate motion indicated by the bend of the Hawaiian-Emperor seamount chain (Clague and Dalrymple 1987).]
In line A4 (Fig. 9a), the basement declines by ~ 300 m on crossing the Galapagos FZ over a distance of ~ 200 km, but the Bouguer anomalies are little changed. This forms a transition to an important low L2 in residual elevation that is ~ 300–400 km across (Fig. 3a). We modelled the change in free-air anomaly going over this fracture zone (arrow in Fig. 9a for A4) using the gravity slab formula (Eq. 2) with two different assumptions. The mean basement depth, bathymetry and free-air gravity anomaly were first calculated for each of − 1° to 0° and for 1.5° to 2.5°N.
First, we assumed the basement variation is locally compensated by crustal thickness variations. The difference in averaged basement elevation between the two areas is 311 m. If that were Airy compensated, the crust is predicted to thin by 2383 m from 0.5°S to 2°N as illustrated in Fig. 13. This value was derived using a simple isostatic calculation using mantle, crustal and water densities (ρm, ρc, ρw) of 3.3, 3.0 and 1.0 g cm−3. The gabbroic layer was assumed to dominate crustal thickness variations so a nearly constant 2.5 km thick layer 2 (Mutter and Mutter 1993) does not contribute to the isostatic calculation and we use a gabbroic 3.0 g cm−3 density for the crust as explained earlier. Sediment isostatic effects were ignored. With these assumptions, the change in Moho depth ΔM is predicted from the change in basement depth ΔB to be
The green bar in Fig. 9a shows the expected free air anomaly (free-air anomaly at 0.5°S plus change) computed using the gravity slab formula from the changes in seabed, basement and Moho density interfaces [a 1.6 g cm−3 density was assumed for the sediments (Mayer and Theyer 1985)]. The predicted anomaly is not much different from the free-air anomaly in the Sandwell et al. (2014) grid (blue circle at 2°N in Fig. 9a).
Second, if the basement depth change instead arises without change in crustal thickness (topography not locally compensated), the effect on gravity is shown by the red bar in Fig. 9a (free-air anomaly at 0.5°S plus change). This calculation involved applying the gravity slab formula to the change in seabed and basement density interfaces, applying a mantle 3.3 g cm−3 density to the basement change as the gravity effect of the crust would be uniform in this case. This model clearly does a worse job of predicting the free-air gravity variations. There appears therefore to be a significant thinning of the crust by ~ 2 km crossing the Galapagos FZ northwards.
At the larger scale (Fig. 3a), the residual elevations correlate poorly with the regional free-air gravity anomalies. For example, a residual elevation high EH1 lies over the lowest free-air anomalies (< 20 mGal), whereas the lowest residual elevation EL2 is far from the gravity low. Crustal thickness variations also affect basement elevations, so analysis of long-wavelength residual elevations also requires a correction for crustal thickness (Hoggard et al. 2016, 2017).
We therefore derived a residual elevation for the Akademik Silskiy data instead using the same global subsidence rate of 324 m m.y.−0.5 of Hoggard et al. (2017) and with their procedure. Basement depths were corrected for the effects of crustal thickness variations using crustal and asthenospheric mantle densities of 3.0 and 3.3 g cm−3. The results offset to a zero mean are shown by the pink line in Fig. 6. Although closer to the large-scale variation observed in the free-air anomalies, this profile still does not reproduce the low in those anomalies. Whatever causes the regional gravity low, it is apparently not so straightforwardly linked here to a surface elevation change.
We have earlier highlighted the ~ 400 km extent of depressed residual elevation L2 along line A4 and that it appears to correspond with crust that is ~ 2 km thinner than to the south of the Galapagos FZ. Such a change is similar to the full variability of crustal thicknesses compiled by Chen (1992) from post-1970 seismic refraction data. This is an unusually large area of thinned crust created at a fast-spreading ridge. If the crust south of the fracture zone has a thickness more typical of average crust [~ 7 km (Mutter and Mutter 1993)], that north of the fracture zone has 5 km thick crust. If this were to have been created by an area of mantle that is unusually cold, modelling suggests the temperature anomaly could be ~ 25°–30° (Brown and White 1994, Su et al. 1994). Such a localised mantle temperature anomaly with an abrupt southern boundary seems unlikely, however. We suggest this was created by a body of mantle that had been depleted by a prior melting event, e.g., lithospheric mantle earlier recycled into the mantle at a subduction zone.
The graphs described above (Figs. 6, 9, 10) suggest that the lineaments observed in the gravity field (Fig. 3c) could potentially have arisen from a local deformation of the plate, such as by the mechanisms illustrated in Fig. 1. Although the apparent densities are not resolved sufficiently well to address their formation mechanisms, some aspects of the spatial pattern made by the lineaments in Fig. 3c support some mechanisms over others.
East of “F” in Fig. 3c, lineaments are finely spaced and oriented WNW, parallel with the recent motion of the Pacific plate over the mantle (Fig. 3b). Mantle small-scale convection likely takes a few million years to initiate (Buck 1985), so their initiation in Fig. 3c very close to the EPR (within 20 km?) suggests another origin. These are probably crustal structures (e.g., seamount chains) created above melting anomalies that are fixed in the mantle (Fleitout and Moriceau 1992). The finest 20 km scale of these lineaments equals the spacing of P-wave velocity anomalies suggestive of mantle upwellings in a large-scale seismic refraction experiment on the EPR between the Clipperton and Siqueiros fracture zones (Toomey et al. 2007).
Immediately west of there, lineaments are oriented N060°E (parallel with arrow marked by “F” in Fig. 3c) and have a wider spacing (> 100 km). These lineaments are shorter (up to ~ 500 km) than others in Fig. 3c. As they lie oblique to flowline directions (Fig. 3b), their orientations are incompatible with small-scale convection (Buck and Parmentier 1986; Haxby and Weissel 1986) or with plate contraction, which is expected to occur parallel to the ridge (Cormier et al. 2011; Dunbar and Sandwell 1988; Gans et al. 2003; Winterer and Sandwell 1987). Given the proximity of the area of elevated topography and seamounts over the westward extension of the Siqueiros FZ (Fig. 2a), we speculate that the static component of pressure in the asthenosphere here could be affected by this topography and the asthenosphere mobilised in a manner somewhat like lower crustal channels proposed on continents (Beaumont et al. 2004; Clark and Royden 2000; Royden et al. 1997). The asthenosphere may then have been affected by viscous inter-fingering as suggested by Weeraratne et al. (2007), though here arising from topographic potential energy rather than the buoyancy from off-axis hotter mantle that they had suggested.
The elevation difference driving such an inter-fingering movement between the elevated region and that ~ 500 km to the south is roughly 200–500 m (Fig. 2a) so the excess lithostatic pressure in the asthenosphere would be 4–10 MPa (200–500 m times 2000 kg m−3 times 10 m s−2). This is small compared with the lithostatic pressure driving lower crustal flow in mountain regions with elevation differences of many kilometres and with crustal rocks contrasting in density with air rather than water (e.g., 100 MPa for the 4 km altitude of the Tibetan Plateau). However, the depth-extent of weak rheology forming any potential asthenospheric channel could be ~ 50–100 km or larger based on seismic velocity profiles (Harmon et al. 2011; Maggi et al. 2006; Weeraratne et al. 2007), compared with < 30 km for lower crustal channels (Beaumont et al. 2004). Buck (1985) assumed an asthenosphere viscosity of 1018 Pa s, compared with lower viscosities of 1018–1019 Pa s suggested to occur in the crustal channels (Beaumont et al. 2004; Clark and Royden 2000). Furthermore, the volume of asthenosphere moved to deform the plate by up to a few 100 m at most (Fig. 9) would be less than that elevating orogens by kilometres.
Some lineaments with trend “C” in Fig. 3c, which parallel fracture zones, may have arisen from spreading centre processes. From the above interpretation of the residual elevations in line A4 and the apparent continuation of elevation lows to line A3 to the east and possibly parts of A6 to the west (Fig. 3b), these may accompanied long-lived processes that led to thinner than normal crust for several million years, perhaps as long as 10 m.y. (Fig. 2a). Further lineaments with this orientation can be observed in this spreading corridor to the west (144˚-135˚W) and south of the Clipperton FZ.
Those lineaments in Fig. 3c appear to be cross-cut by lineaments with orientation “A”, which lie parallel to the motion of the Pacific plate for the past 20 m.y. or longer (see flowline in Fig. 3b). The lineament marked “A” appears to continue to the EPR, where it has a finer structure, presumably in part due to the shallow water depth (smaller upward continuation effect). Further lineaments beneath the Akademik Silskiy line (“B” in Fig. 3c) are oriented clockwise of those at “A”, but are parallel with the direction of motion of the Pacific plate prior to 30 Ma (Fig. 3b). Given this correspondence, small-scale convection may explain both sets of these lineaments (Buck and Parmentier 1986; Haxby and Weissel 1986). Alternative explanations involving crustal construction at the EPR modified by melting anomalies that are fixed in the hotspot reference frame (Fleitout and Moriceau 1992) seem to be ruled out in this area by their continuation across fracture zones and other structures where crustal thickness is reduced.
Returning to the possibility of extension causing these features, unfortunately we have found the seismic reflection records are not adequate to resolve this, despite sediment up to 600 m thick in places (Mitchell and Lyle 2005; Mitchell et al. 2003). As the sediment is pelagic, it commonly drapes the underlying basement (Dubois and Mitchell 2012; Mitchell and Huthnance 2013; Tominaga et al. 2011), making fault or fold displacements difficult to separate from depositional geometry. Faults can also arise because of differential compaction. Nevertheless, pelagic draping means that the seabed relief commonly appears as an attenuated version of the underlying basement relief (Fig. 4). As shown in Fig. 12, multibeam sonar data from this area clearly show the abyssal hills originally created at the EPR. At fracture zones (such as the Galapagos FZ in Fig. 12b), there can be transverse features, but otherwise we have observed no clear evidence for faults crossing the abyssal hills here that would arise if the gravity lineaments were caused by plate extension (Cormier et al. 2011; Dunbar and Sandwell 1988; Gans et al. 2003; Winterer and Sandwell 1987).
Further lineament “E” (Fig. 3c) is neither parallel to adjacent fracture zones, nor to flowline directions. This feature is ~ 1000 km long. Some similarly oriented features lie north of the Clipperton FZ also. If lithosphere was weak along the fracture zone, it is unclear how feature “F” could have formed by extension with this oblique orientation. By elimination, a viscous inter-fingering mechanism (Weeraratne et al. 2007) seems the only viable explanation for these lineament directions at present.
In summary, the diversity of lineament orientations is difficult to explain by any single mechanism (Figurer 1). It seems likely that multiple processes have created the gravity lineaments here, of which we favour spreading centre processes, small-scale convection, viscous inter-fingering and melting anomalies that are fixed in the mantle reference frame. The Pacific plate changed absolute velocity from ~ 0.5°/m.y. prior to 30 Ma to ~ 0.9°/m.y. after 30 Ma (Wessel and Kroenke 2008), though neither this nor the changes in direction of motion marked in Fig. 3b appear to correlate with any of these changes in lineament style. Further progress on these issues would benefit from further deep-seismic data capable of mapping the Moho.
Pelagic sediments up to 600 m thick obscure basement in the equatorial Pacific, preventing a simple comparison of the gravity field with bathymetry data. Depths of basement were therefore derived from these three seismic reflection datasets that cross this region and combined with re-evaluated seafloor spreading magnetic anomalies (Barckhausen et al. 2013). After correction for sediment isostatic loading, these depths reveal a simple plate-cooling trend with a subsidence rate of 313 m m.y.−1/2 south of the Clipperton FZ (AMAT03 and Venture-1 datases) and 343 m m.y.−1/2 immediately north of it (Akademik Selskiy dataset). These rates are broadly consistent with results of previous studies of this area that had been based on interpolated datasets.
Bouguer anomalies were computed from free-air gravity anomalies using the slab formula to correct for the seabed density contrast. The resulting anomalies in some places do not vary with basement topography, suggesting areas of crustal thickness variation, such as around seamounts. In other areas, however, basement topography varies in a similar way to that noted from bathymetry datasets elsewhere over gravity lineaments, forming low relief (100 s of m) swells over ~ 100 km or smaller length-scales. Bouguer gravity anomalies vary over these features with an apparent basement density that varies, but includes 2.7–3.3 g cm−3 densities expected of basalt to mantle lithologies.
The gravity lineaments in the equatorial Pacific have remarkably varied orientations. Some are oriented parallel to Pacific plate flow-lines, changing direction north of the Clipperton FZ, as expected from the different plate direction prior to 30 Ma if a mechanism such as small-scale convection formed them. However, other lineaments are oriented with directions that are not explainable by plate motion, so they require other mechanisms to be considered, such as viscous inter-fingering in the asthenosphere. Despite the thick sediment, the fine-scale abyssal hill morphology is commonly observed because the sediment is pelagic and tends to drape the underlying basement. We note that there are very few indications of faults crossing the abyssal hills that might be expected if the gravity lineaments were produced by extension.
Our detailed investigation has highlighted an area ~ 400 km across where residual basement elevations are unusually depressed. From a simple Airy isostatic argument, oceanic crust in this area is ~ 2 km thinner than outside the area. We suggest that this thin crust was created at the EPR for a period when the ridge overlay depleted mantle.
Nathalie Dubois interpreted the seismic reflection data from RV Revelle cruise AMAT03 and kindly made her interpretations available to us. Udo Barckhausen provided the magnetic anomaly identifications and advice on the isochron mapping. Steve Eittreim provided a paper copy of the seismic data summarized in Eittreim et al. (1994). The free air anomaly grid was provided by David Sandwell and others of the Scripps Satellite Geodesy group. The Lamont group provided the bathymetric grids (Ryan et al. 2009). Figures were created with the GMT software system (Wessel and Smith 1991). The Revelle data acquisition at sea and seismic interpretation was supported in part by Natural Environment Research Council grant NE/C508985/2. Mitch Lyle is thanked for his work in initial processing the seismic data and running the AMAT03 cruise, which was also part-funded by the NSF. We thank the editor Roger Urgeles and three anonymous reviewers for insightful and useful comments that significantly improved this article.
- Bloomer SF, Mayer LA, Moore TC (1995) Seismic stratigraphy of the eastern equatorial Pacific Ocean: Paleoceanographic implications. In: Pisias NG, Mayer LA, Janecek TR, Palmer-Julson A, van Andel TH (eds) Proceedings of the Ocean Drilling Program, Scientific Results, Vol. 138. Ocean Drilling Program, College Station, TX, pp 537–553Google Scholar
- Clague DA, Dalrymple GB (1987) Geologic evolution of the Hawaiian Emperor volcanic chain. In: Decker RW, Wright TL, Stauffer PH (eds) Volcanism in Hawaii, U.S. Geol. Surv. Prof. Paper 1350. U.S. Geological Survey, pp 85–100Google Scholar
- Divins DL (2003) Total sediment thickness of the world’s oceans & marginal seas. Center NNGD, Boulder, COGoogle Scholar
- Dunbar J, Sandwell DT (1988) A boudinage model for cross- grain lineations (abstract). Eos Trans AGU 69:1429Google Scholar
- Gnibidenko HS, Eittreim SL, Helsley CE, McClellan P, Ryan HF (1990) The Central Pacific Transect; cruise report of the research vessel Akademik Selskiy. U.S. Geol. Surv. Open-File Rept 90–532:1–13Google Scholar
- Ludwig WJ, Houtz RE (1979) Isopach map of the sediments in the Pacific Ocean Basin, color map with text. In: Am. Assoc. Pet. Geol., Tulsa, OKGoogle Scholar
- Lyle MW, Wilson PA, et al (2002) Leg 199 preliminary report, Paleogene Equatorial Transect, 23 October-16 December 2001. Ocean Drilling Program, College Station, TXGoogle Scholar
- Mayer L, Theyer F (1985) Initial Reports DSDP, 85. U.S. Govt. Printing Office, WashingtonGoogle Scholar
- Mayer LA, Shipley TH, Theyer F, Wilkens RH, Winterer EL (1985) Seismic modeling and paleoceanography at Deep Sea Drilling Project Site 574. In: Mayer L, Theyer F et al (eds) Initial Reports DSDP. U.S. Govt. Printing Office, Washington, DC, pp 947–970Google Scholar
- Pälike H, Lyle M, Nishi H, Raffi I, Gamage K, Klaus A, the_Expedition_320/321_Scientists (2010) Proceedings of the Integrated Ocean Drilling Program, Vol. 320/321. Integrated Ocean Drill Program, College Station, TX., https://doi.org/10.2204/iodp.proc.320321.320101.322010
- Taylor JR (1982) An introduction to error analysis, the study of uncertainties in physical measurements. Oxford University Press, Mill Valley, p 270Google Scholar