Anomalous topography in the western Atlantic caused by edge-driven convection



[1] The western Atlantic region contains a long-wavelength intraplate topography anomaly that is defined by the NE-SW trending Bermuda Rise and two adjacent topography lows. Using numerical experiments, we test the hypothesis that the anomalous topography may be the surface response to edge-driven convection. A primary edge-driven convection cell and secondary flow circulation develops at a modeled continent-ocean plate margin and induces subsidence at the continent-ocean margin, an off-shore peak/plateau of high topography on the ocean plate, and distal ocean plate subsidence. Unlike hot spots, the edge-driven convection cell and associated topography migrate with moving surface plates. The flow cell and wavelength of topography is broadened with continent-ward motion of the lithosphere relative to the mantle, whereas a migration in the ocean-ward direction suppresses the formation of the edge-driven convection cell and surface topography. The wavelength of observed anomalous topography in the western Atlantic and estimates of plate motions relative to a fixed hot spot reference frame are consistent with the former.

1. Introduction

[2] Within the interiors of the ocean plates, anomalous topographic features can be superimposed on the long-wavelength ocean floor subsidence associated with plate cooling. The motion of the ocean plate over a fixed mantle hot spot explains some volcano chains [Wilson, 1963]. Recent studies have suggested that fixed mantle hot spots and associated deep plume sources are not capable of explaining certain ocean intraplate topographic anomalies. For instance, Vogt [1991] demonstrates that the topographic swell patterns in the western North Atlantic (Figure 1) can not be reconciled with simple hot spot theory since the swell patterns are almost orthogonal to the predicted fixed hot spot track associated with North American plate motion.

Figure 1.

(a) Bathymetric map of the western North Atlantic region. (b) Profiles of residual topography across the sections shown in Figure 1a and an averaged section (bold solid line). The residual topography is surface elevation corrected for isostatic effects (crustal thickening on the continental plate and lithospheric cooling and subsidence on the ocean plate) and sediment loading. These corrections have been applied separately for continental plate and ocean plate to derive the profiles shown; see text for detailed explanation. Arrow marks the approximate location of the ocean-continent plate transition.

[3] In Figure 1b, ‘residual topography’ profiles across this region are plotted. The residual topography on the ocean plate is based on 2.5° residual depth anomalies corrected for sediment loading and ocean plate cooling and subsidence [Sclater and Wixon, 1986]. For the continental portion the residual topography is calculated as the raw topography corrected for crustal thickness variations. The correction assumes isostatic compensation of the crust and we use the model CRUST2.0 [Bassin et al., 2000], which provides both thickness and density of structures through the crust within 2° × 2° cells.

[4] The average residual topography profile is characterized by an ocean plate elevation high that reaches ∼600 m (Figure 1b). This feature forms the Bermuda rise and related high ocean floor swells which trend northeast-southwest, and approximately parallel to the continental shelf of North America. The residual topography high is flanked by a topography low adjacent at the continental margin, and another deepening further towards the mid-Atlantic spreading centre (Figure 1b). Both of these negative topography anomalies trend in the same northeast-southwest direction as the topography high.

[5] It has been suggested that these anomalous topographic features may be produced by a thermal instability traveling with the North American plate rather than a deep mantle plume [Vogt, 1991]. At a passive continent-ocean margin, the sharp lateral thermal gradient of the lithospheric discontinuity can induce the formation of a small-scale mantle flow cell or ‘edge-driven convection’ [King and Anderson, 1995, 1998]. King and Ritsema [1998] speculate that such convection cells may explain the occurrence of seafloor volcanism in intraplate environments on the South American and African plates. Conrad et al. [2004] proposed that an anomalously deep region off the coast of Nova Scotia may be caused by the downgoing portion of a potential edge-driven mantle convection cell.

[6] Similarly, we speculate that the process of edge-driven convection may be responsible for the observed anomalous mid-plate topography in the western Atlantic region; namely the Bermuda rise and adjacent topography lows.

2. Numerical Experiments

[7] A series of thermo-mechanical numerical plane-strain experiments was designed to study the surface manifestation of edge-driven mantle convection. The set-up of the experiments follows the approach adopted by King and Anderson [1998] where a continent-ocean passive margin is modeled as a step-function lateral change in the thermal and mechanical properties of the lithosphere (Figure 2).

Figure 2.

Set-up of the numerical (arbitrary Lagrangian-Eulerian finite element) experiments. The full solution space is 2100 wide by 1400 km deep; calculations are performed on a non-uniform Eulerian/Lagrangian grids of resolution 200 × 140/400 × 280 with higher element density in the lithosphere and upper mantle. Free-slip boundary conditions are used on the sidewalls and bottom boundary, except along the side along the continental lithosphere which is held fixed. For the temperature-dependent viscosity of the upper mantle, η0 = 5.9 × 1025 Pa·s and c = 121.6 K are used. Density is given by ρ = ρ0(1 − α(TT0)) where T is temperature, T0 = 293 K is the reference temperature, α = 2 × 10−5 1/K is thermal expansivity, and reference density ρ0 varies by material as shown. We assume a specific heat of cp = 1.25 × 103 J/kg/K and a thermal conductivity of k = 3 W/m/K.

[8] The 200 km thick lithosphere on the left side of the model is assumed to be cratonic continental lithosphere; this is flanked by 50 km thick oceanic lithosphere. The viscosity of the ocean lithosphere is held fixed at η = 1023 Pa·s. The cratonic lithosphere is stabilized by compositional buoyancy (ρ0 = 2800 kg/m3) and a very strong rheology (η = 1026 Pa·s). This configuration assumes that the continent-lithosphere discontinuity persists as a compositional boundary between the ocean plate and (cratonic) continental lithosphere. The upper mantle fluid has a temperature-dependent viscosity (Figure 2) and there is a factor 100 increase in viscosity at 670 km depth [Mitrovica and Forte, 1997]. The initial temperature at the base of the lithosphere is 1350°C and this increases linearly to 1700°C at the base of the upper mantle. The top surface of the model is held at 20°C, but mechanically this is a free surface where topography will develop in response to the underlying dynamics.

2.1. Edge-Driven Convection and Topography

[9] The lateral temperature variation across the continent-ocean margin results in the onset of a time-dependent small-scale convection cell with downwelling near the margin and upwelling beneath the ocean lithosphere (Figures 3a and 3b). This convective pattern is essentially confined to the upper mantle owing to the large viscosity increase at 670 km depth. The cell is offset so that downwelling flow occurs approximately 150 km inland from the continent-ocean margin, and the upwelling is ∼300 km outboard. The upwelling portion of the edge-driven convection cell induces a second mantle circulation cell, further out beneath the ocean plate. The downwelling instability associated with this secondary flow develops approximately 750 km away from the continental margin.

Figure 3.

Temperature fields and flow velocity vectors at (a) 22 and (b) 47 Myr. (c) Profiles of surface topography at time intervals as indicated. Abscissa distances are from the left side of the solution space; arrows mark the location of the continent-ocean boundary.

[10] The evolution of dynamic topography across the top surface of the solution space reflects the presence of these dual convection cells (Figure 3c). Topography is characterized by subsidence at the continent-ocean margin, a higher-amplitude positive topography on the ocean plate above the upwelling portion of the edge-driven convection cell, and subsidence further out on the ocean plate. The time-dependence of convection results in slight motion in the positions of the topography peaks and appreciable variation in the amplitude of topography. During vigorous flow at 22 Myr, the maximum and minimum topography reach 200 m and −150 m, respectively. The positions of these topography anomalies correspond with the position of the upwelling/downwelling flow. That is, the uplift on the ocean plate occurs approximately 300 km away from the continent-ocean margin. Although the downwelling flow is focused ∼150 km inland from the margin, the maximum subsidence occurs right at the margin since the thinner oceanic plate is more susceptible to flow-induced topography than the strong continental lithosphere.

2.2. Continent-Ward Plate Motion

[11] King and Anderson [1998] demonstrated that the long-wavelength mantle thermal structure and the related convective flow could significantly alter the development of edge-driven convection cells. This finding motivated us to consider how the character of edge-driven convection and the topographic response would be altered by a leftward (continent-ward) motion of the surface plate (both continent and ocean portions) relative to the underlying mantle.

[12] A counter-clockwise convection cell develops at the edge of lithosphere discontinuity, along with a weaker secondary flow cell (Figure 4a). As the lithospheric plate moves to the left at 1 cm/yr, the center of the edge-driven convection cell also migrates at approximately the same rate. However, the form of the convection cell is altered by the plate motion: The width-to-depth aspect ratio of the convection cell increases as the downwelling portion stays fixed to the moving plate margin and the upwelling portion lags behind.

Figure 4.

(a) Temperature fields and flow velocity vectors at 37 Myr and (b) topography profiles for experiment where lithosphere is being pushed at 1 cm/yr to the left (continent-ward motion of entire surface plate). (c) Profiles of surface topography for experiment where lithosphere is being pushed at 1 cm/yr to the right (ocean-ward). In all frames, the arrows denote the position of the continent-ocean at the specified time.

[13] As in the previous model, topographic uplift above the upwelling portion of the convective cell and subsidence above the downwellings is observed (Figure 4b). However, the topography anomalies migrate as the edge-driven convection cell moves with the continent-ocean margin. The wavelength of the topography (measured across the topography low beneath the continent, and topography high on the ocean plate) increases from 500 km at 13 Myr to 875 km at 37 Myr, and by the late stage, flow-induced topography high forms a broad plateau that is located ∼550 km away from the margin. This is a reflection of the increased aspect ratio of the convection cell as the elongated edge-driven convection cell produces a broader positive topography at a further outboard position from the continental margin.

2.3. Ocean-Ward Plate Motion

[14] In Figure 4c we show the results of an experiment where the direction of the plate motion has been reversed, having an imposed lithospheric velocity relative to the underlying mantle of 1 cm/yr to the right. In this case, the edge-driven convection cell is thrust further beneath the continental margin as the margin moves ocean-ward. The convection roll is approximately centered beneath the continent-ocean transition with the downwelling component of flow located ∼200 km inland, and the upwelling 200 km from the margin (Figure 4c).

[15] The migration of surface topography relative to the fixed model solution space is subtle since the edge-driven convection cell tends to shift beneath the moving lithospheric plate. The wavelength of the topography, approximately 450 km across the continent-low and ocean-high, is reduced by the reversed plate motion, which tends to compress the upper mantle convection cell. The topography high reaches ∼190 m, but the continental topography low is dampened, as the downwelling portion of the convective flow is shifted beneath the strong, thick continental plate. The far-offshore topography low develops in a similar manner to the stationary plate model (Figure 3c), but the nadir of topography migrates with the moving plates.

2.4. Edge-Driven Convection Into the Deep Mantle

[16] In the previous experiments, the edge-driven convection cell is restricted largely to the upper mantle due to the presence of a viscosity increase imposed at 670 km depth (Figure 2). As an end-member case, we conduct an alternative model where there is no impediment to mantle flow through this level in order to consider if a deeper convection cell and longer wavelength topography evolves (Figure 5). The convective flow induced by the lithospheric thermal discontinuity (not shown) is characterized by upwelling and downwelling flow that extends only slightly deeper into the mantle. This is manifested in the style of topography, as it remains similar to preceding experiments, but it is of slightly longer wavelength and higher amplitude (i.e., reaching ∼230 km). The distance from the continental margin to the centre of the outboard topography high is approximately 450 km, compared to 300 km in Figure 3c. However, the experiment demonstrates that even if conditions were favourable, the edge-driven convection and associated topography will essentially develop on an upper mantle lengthscale (<∼700 km), rather than as a deep mantle feature. We speculate that the vertical scale of the continent-ocean lithospheric thermal discontinuity is probably a more important factor in controlling the wavelength of the edge-driven convection cell.

Figure 5.

Topography profiles for an experiment where the viscosity jump at 670 km depth has been removed (constant mantle viscosity of 1021Pa·s). Otherwise, the set-up of the model is identical to the experiment in Figure 3.

3. Conclusions

[17] The numerical experiments show a primary edge-driven convection cell and secondary flow circulation developing at a passive continental margin. This mantle flow will induce a characteristic signal of intraplate dynamic topography defined by subsidence at the continent-ocean margin, an off-shore peak/plateau of high topography on the ocean plate, and distal subsidence further out on the ocean plate. This pattern of topography is consistent with profiles of residual topography across a portion of the western North Atlantic and North American margin.

[18] Unlike a hot spot mantle plume, the thermal anomalies associated with edge-driven convection will migrate with moving surface plates. Furthermore, the experiments show that the plate motion relative to the mantle may be important for controlling the style of edge-driven convection and associated dynamic topography. With a migration of the coupled continent-ocean plate in the continent-ward direction, there is a tendency for the mantle convection cells to be broadened horizontally. (Although stable edge-driven convection will eventually be destroyed by the long-wavelength component of mantle flow associated with progressively higher plate velocities [King and Anderson, 1998].) The broadening of the cell leads to an increase in the wavelength of the associated dynamic topography. Alternatively, a migration of the plate in the ocean-ward direction tends to suppress the formation of the edge-driven convection cell, resulting in lower amplitude and shorter wavelength surface dynamic topography. These results on a regional scale agrees with global-scale models that similarly show a potentially strong influence of plate motions on dynamic topography [Cadek and Fleitout, 2003].

[19] The western Atlantic region is consistent with a continent-ward motion of the plate margin since the horizontal wavelength of observed residual topography is broad compared to the depthscale of upper mantle edge-driven convection. This is also supported by estimates of current plate motion which indicate a general westward motion of the western Atlantic portion of the North American plate in a hot spot reference frame [DeMets et al., 1990].

[20] The magnitudes of the peak topography anomalies in our experiments are lower than in the observed residual topography (Figure 1b). This is primarily a consequence of imposing a homogeneous ocean lithosphere in our models, although the Bermuda Rise is a site of high heat transport/volcanism and complex ocean crust structure. Localized heating related to igneous activity and pre-existing strength discontinuities can significantly alter lithospheric strength and amplify the surface topographic response [Pysklywec and Beaumont, 2004]. Furthermore, it should be recognized that dynamic topography induced by edge-driven convection is only one component that may contribute to anomalous surface elevation in the region. Our predicted anomalies, for example, may be superimposed on another signal of dynamic topography related to a broader, deeper mantle convective flow beneath North America and the Atlantic Ocean [Conrad et al., 2004]. We also note that an approximation is being made in treating the system in a two-dimensional geometry. A more complex 3D evolution may account for the decrease in topography away from the Bermuda Rise and parallel to the North Atlantic margin, (Figures 1a and 1b). As a result of these combined uncertainties, the intent of the experiments is not to attempt an exact match to the averaged signal of residual topography, but to demonstrate that the surface response to edge-driven convection is a feasible mechanism for explaining the observed pattern of anomalous topography in the western Atlantic region.


[21] The research was funded by the Natural Sciences and Engineering Research Council of Canada. Numerical calculations used software originally developed by Philippe Fullsack. We thank two anonymous reviewers for constructive comments on the original manuscript.