Roadmap | The review process | Home
   Hotspot reference frame

The hotspot reference frame and the westward drift of the lithosphere

Carlo Doglioni & Marco Cuffaro

Dipartimento Scienze della Terra, Università La Sapienza, Roma, Italy

Hotspots or no Hotspots?

Where do hotspots come from? What is their source depth? Do they provide a fixed reference frame? Hotspot tracks have been used to compute the motion of plates relative to the mantle. For this purpose it is fundamental to know whether hotspots are i) fixed relative to the mantle, ii) fixed relative to one another, and iii) from what depth they originate. However hotspots have often been used uncritically, without regard to their real nature. Volcanic tracks at the Earth's surface may be a result of intraplate plumes (e.g. Hawaii), retrogradation of a subducting slab, migration of back-arc spreading, along-strike propagation of a rift (e.g. East Africa), or propagation of a transform fault with a transtensive component (the Chagos ridge? see also Deccan page). All these volcanic tracks may have different depths of their mantle sources and they must be differentiated (Figure 1). Plate boundaries by definition move relative to one another and relative to the underlying mantle. Therefore any hotspot located on a plate boundary cannot be used for a fixed-hotspot reference frame.


Figure 1. The main volcanic chains at the Earth's surface may have different origins and depths. The red arrows indicate the direction of migration of volcanism with time. Filled triangles represent the youngest volcanic products. Volcanic trails originating on ridges may be wetspots (in sensu Bonatti, 1990) and sourced from a fluid-rich asthenosphere. The hotspots located on plate boundaries are not fixed by definition, since both ridges and trenches move relative to one another and with respect to the mantle. Pacific hotspots, regardless their source depth, are located within the plate and are virtually the only ones that can be considered reliable for a hotspot reference frame.

Deep Plumes or Shallow Upwellings? Fixed or Unfixed?

Accumulating evidence suggests that hotspots are mostly shallow features (Bonatti, 1900; Smith & Lewis, 1999; Anderson, 2000; Foulger, 2002; Foulger et al., 2005). For example, Atlantic hotspots may be interpreted better as wetspots (Figure 2) rather than hot lines, as suggested by Bonatti (1990). An asthenospheric source richer in melting-point-lowering fluids can account for magma overproduction. Propagating rifts (hot-lines, etc.) are shallow phenomena which are not fixed relative to the deeper mantle. The only hotspots that are relevant to a fixed-hotspot reference frame are those located within plates. For example, Norton (2000) grouped hotspots into three main families, each of which has very little internal relative motion (the Pacific, Indo-Atlantic and (single-hotspot) Iceland families). He concluded that a global hotspot reference frame is inadequate because Pacific hotspots move relative to Indo-Atlantic hotspots and to Iceland. Since Indo-Atlantic hotspots and Iceland are located on ridges, they do not satisfy the requirement for fixity. In the analysis of Norton (2000), Pacific hotspots have been reasonably well fixed relative to one another during the last 80 My. As a result, the screening of volcanic tracks that may be used for the fixed-hotspot reference frame leaves a very limited number of hotspots and only the Pacific ones satisfy the requirements.

Decoupling in the Asthenosphere

The origin of intraplate Pacific magmatism is also obscure, and the source depth and mechanism of melting it is still under discussion (Foulger et al., 2005). Since the Pacific is the fastest-moving plate, shear heating along the basal decollement has been suggested as a potential mechanism for generating localized hotspots tracks (Figure 3). Areas with viscosity higher than normal in the asthenospheric decollement should generate greater shear heating.

Figure 2 Hypothetical reconstruction of South Atlantic type migrating volcanic ridges. An anomalously water-rich asthenospheric mantle, or wetline, oblique to absolute African plate motion and Mid-Atlantic Ridge (MAR) migration could generate a SW-ward rejuvenating volcanic track. Similar, mirror-image-like tracks (NW-trending, SE-propagating) could form in the South American plate. This model could explain why age progressive volcanic trails are oblique to transform faults.

Figure 3 The Hawaiian volcanic track indicates that there is decoupling between the magma source and the lithosphere, which is moving relatively toward the WNW. If the source is below the asthenosphere (e.g., in the sub-asthenospheric mantle, option 1), the track records the entire shear between lithosphere and mantle. In the case of an asthenospheric source for the Hawaiian hotspot (option 2), the volcanic track does not record the entire shear between the lithosphere and sub-asthenospheric mantle, since part of it operates below the source (deep missing shear). Moreover the larger decoupling implies larger shear heating, which could be responsible for the scattered, punctiform Pacific intraplate magmatism (after Doglioni et al., 2005). [see also Hawaii and Mantle Temperature pages.]


Kennedy et al. (2002) showed how mantle xenoliths record shearing possibly located at the lithosphere-asthenosphere interface. This supports the notion of flow in the upper mantle and decoupling at the base of the lithosphere, which is also supported by seismic anisotropy (Russo & Silver, 1996; Doglioni et al., 1999; Bokelmann & Silver, 2000). The fastest plate on Earth in the hotspot reference frame (i.e., the Pacific) is also the one affected by the most widespread intraplate magmatism. It is noteworthy that the fast Pacific plate overlies the asthenosphere with the lowest mean viscosity (5 x 1017 Pa s; Pollitz et al., 1998), and possibly the least depleted mantle, and therefore the most prone to melting. Because of the melting characteristics of peridotite with minor carbon + hydrogen (the lherzolite–(C+H+O) system), the asthenosphere is already partially molten (e.g., Schubert et al., 2001) and at a temperature of about 1430°C (e.g., Green & Falloon, 1998; Green et al., 2001; see also Mantle Temperature page). A rise in temperature of only a few tens of degrees will increase the degree of melting and this melt, in a deforming material, will migrate toward the surface. We postulate that locally, the viscosity of the asthenosphere can also increase (e.g., up to 1019 Pa s) due to compositional anisotropy. Shear stress could be irregularly distributed in such inhomogeneous material, and consequently higher shear heating (Shaw, 1973) may develop locally and generate punctiform magmatism (Figure 4).


Figure 4. If locally the viscosity of the asthenosphere is higher than normal, the shear stress and shear heating are also higher, providing an increase in asthenospheric temperature. Variations in lithosphere velocity (100 or 200 mm yr-1) and local increases of the viscosity (4 x 1019 or 1020 Pa s) can determine different excess temperatures ranging between 12°K and 120°K. The highest excess temperatures could result in extra melting and possibly uprising intraplate magmatism (after Doglioni et al., 2005).

The No-Net-Rotation Reference Frame

The NUVEL 1 model (DeMets et al., 1990) and the Space Geodesy ITRF2000 and NASA databases (Heflin et al., 2005) provide information on plate motions. They are based on the artificially imposed assumption of a no-net-rotation (NNR) reference frame, i.e., plates are imagined to move relative to a fixed Earth center, and the lithosphere does not move relative to the underlying mantle, or the sum of its movements is zero. However we know the lithosphere does move relative to the sub-asthenospheric mantle, and this is not only suggested by the Hawaiian and similar volcanic tracks, but it is also kinematically required by the relative migration of plate margins. The ITRF2000 reference frame (e.g., Heflin et al., 2005) is excellent for describing relative plate motions, and it is also considered to be an absolute reference frame (relative to the GPS satellite constellation and the Earth’s center of mass). However, when magmatic sources are included in the kinematic analysis, the movement of the lithosphere relative to the mantle must be taken into account in an "absolute" plate motions analysis, and the NNR must be abandoned.

Westward Drift of the Lithosphere

When plate motions are measured in the hotspot reference frame, the lithosphere shows a net “westward” drift (Bostrom, 1971; O'Connell et al., 1991; Ricard et al., 1991). This “westward” drift persists also when plate motions are computed relative to Antarctica (Le Pichon, 1968; Knopoff & Leeds, 1972), which lies on a plate with almost no subducting slab and is thus often considered to be probably stationary, or moving only slowly, with respect to underlying mantle. However most of the hotspots used are neither fixed, nor do they represent a fixed reference frame, because they are located on plate margins such as moving ridges (e.g., Galapagos, Easter Island, Iceland, and Ascension), transform faults (e.g., Reunion) or continental rifts (e.g., Afar), all of which are features that are moving relative to one another and relative to the mantle. Nevertheless, Gripp & Gordon (2002) computed a net rotation of the lithosphere of 49 mm yr-1 (0.44 ± 0.11 deg Myr-1 about a pole of 56°S, 70°E). The WNW-motion of the Pacific plate relative to the underlying mantle is inferred from the Hawaiian and other major intraplate hotspot tracks (e.g., Marquesas, Society, Pitcairn, Samoan and MacDonald), which show an average velocity of about 103-118 mm yr-1. They also move along the same trend (290°-300°, WNW) and therefore they are the only hotspots that appear to be coherently fixed relative to one another, representing an apparently reliable reference frame for absolute plate motions.

Providing that the velocity of the Pacific lithosphere in the ESE (110-120°) direction is lower than that of the underlying sub-asthenospheric mantle Vm (Vm > Vl), the relative motion corresponding to the WNW delay of the lithosphere is Vm - Vl = 103 mm yr-1. However, if the shear is distributed throughout the asthenospheric channel (Figure 3), and the Hawaiian melting spot is within the asthenosphere (Figure 3, option 2) instead of the lower mantle (Figure 3, option 1) then since only the shear above the hotspot is recorded in the hotspot trace, the total displacement between the lithosphere and mesosphere will be greater than suggested by the surface volcanic chain. Therefore if the location of Hawaii magmatism is within or in the upper part of the asthenosphere, there should be a further deep missing component of the shear, which would increase the total relative velocity. This higher velocity has two basic consequences, 1) it increases the global westward drift of the lithosphere, and 2) it increases the shear heating released within the asthenosphere. If the source of Pacific hotspots is located in the middle of the asthenosphere, half of the lithosphere-sub-asthenospheric mantle relative motion is unrecorded, which means that the total relative displacement would amount about 200 mm yr-1. With this velocity, the net rotation of the lithosphere toward the "west", increases to about 9 cm/yr. In this reference frame, no one plate moves "eastward" relative to the mantle (Figure 5, option 2).


Figure 5. Simplified kinematic relationship of the Pacific–Nazca–South America plates. Relative motions vectors (above) after Heflin et al. (2005). Option 1 indicates the “absolute” motions relative to the Hawaiian hotspot moving at about 103 mm yr-1 (Gripp & Gordon, 2002). Option 2 (below) is the case where the hotspot source is located in the asthenosphere and the relative motion between Pacific plate and sub-asthenospheric mantle is assumed to be ≥ 200 mm yr-1 (see Figure 1). In this last configuration all three plates move “westward” relative to the mantle.

Crazy Tracks

There is evidence that the propagation rate of Pacific “hotspots” or seamount tracks has varied with time, even with jumps back and forth and oblique propagation relative to the “absolute” plate motion, casting doubts on both the notion of absolute plate motion computed in the hotspot reference frame, and the nature of the magmatism itself (deep plume, or rather shallow upwellings generated by cracks or boudins of the lithosphere, Winterer & Sandwell, 1987; Sandwell et al., 1995; Lynch, 1999; Natland & Winterer, 2003), sourced by a mantle with compositional heterogeneity and no demonstrable thermal anomaly in hotspot magmatism relative to normal mid-oceanic ridges (see also Mantle Temperature page). Janney et al. (2000) described the velocity of the Pukapuka volcanic ridge (interpreted as either a hotspot track or a leaky fracture zone), and located in the eastern Central Pacific, between 5 and 12 Ma of about 200-300 mm yr-1. They also inferred a shallow mantle source for Pacific hotspots based on their geochemical characteristics.

Relative plate motions can now be estimated with great accuracy using space geodesy (e.g., Robbins et al., 1993; Heflin et al., 2005), refining the earlier NUVEL 1 plate motions model (DeMets et al., 1990). The East Pacific Rise (EPR), separating the Pacific and the Nazca plates, spreads at rates of 128 mm yr-1 just to the south of the equator (e.g., Heflin et al., 2005). At the same latitude the shortening along the Andean subduction zone, where the Nazca plate subducts underneath South America, has been computed to be about 68 mm yr-1. These relative motions, when input into a reference frame where the Hawaiian hotspot is considered fixed and positioned in the sub-asthenospheric mantle, imply that the Nazca plate moves eastward relative to the sub-asthenospheric mantle at about 25 mm yr-1 (Figure 5, option 1). If we assume that the source of Pacific intraplate hotspots is rather in the middle asthenosphere and half of the lithosphere – sub-asthenospheric mantle relative motion is missing in the Hawaiian track (Figure 3), this motion could increase to 200 mm yr-1, as also suggested by some segments of the Pukapuka volcanic ridge (Janney et al., 2000). Note that in this configuration Nazca would instead move west relative to the mantle at 72 mm yr-1 (Figure 5, option 2) and therefore all three plates would move “westward” relative to sub-asthenospheric mantle.

Shallow Hawaii Upwellling

Another effect of hypothesising a shallow source for Hawaiian magmatism is to increase estimates of the westward motion of the Pacific plate to a velocity that is faster than the spreading rate of the EPR, where the Nazca plate also moves westward relative to the sub-asthenospheric mantle (Figure 5, option 2). A shallow, intra-asthenospheric origin of Pacific hotspots provides a kinematic frame in which all mid-oceanic ridges move “westward”. As a consequence, ridges migrate continuously over fertile mantle (Figure 6). They generate melting and the increase in viscosity of the residual mantle provides a mechanism for slowing the plate motion.


Figure 6. Assuming a fixed mantle, the Pacific lithosphere moves “west” faster than the Nazca plate because the underlying asthenosphere is less viscous and the decoupling is more efficient. Due to the increase in viscosity and decrease in temperature along the rifting area, which is also moving westward, the asthenosphere below the eastern plate is more viscous, causing stronger coupling and a steady state lower velocity of the Nazca plate. These kinematics provide continuous new fertile mantle underneath the oceanic ridge (after Doglioni et al., 2005).


The deep and shallow hotspot interpretations generate two hotspot reference frames. In the case of a deep mantle source for the hotspots, a few plates still move “eastward” relative to the mantle (Figure 7), whereas in the case of shallow sources, all plates have a “westward” component, although with different velocities (Figure 8).


Figure 7. Current velocities with respect to the deep-hotspot reference frame, option 1 of Figure 3. Data from HS3-NUVEL1A (Gripp & Gordon, 2002).


Figure 8. Present-day plate velocities relative to the shallow-hotspot reference frame, option 2 of Figure 3, incorporating the NUVEL1A relative plate motions model. Note that in this frame all plates have a westward component.


These results are in accord with anisotropy measured using shear-wave splitting (Russo & Silver, 1994) and the low dip of the Andean slab, which both suggest relative eastward mantle flow. Similar eastward mantle flow was proposed for the North America plate (Silver & Holt, 2002). The low dip of the Andean slab has also been attributed to the age of the subducting lithosphere. However the oceanic age has been shown to be insufficient to explain the asymmetry between “westerly”-directed (steep and deep) vs. “easterly”-directed (low dip and shallow) subduction zones (Cruciani et al., 2005). In fact the geographically related asymmetry persists even where the same lithosphere (regardless whether it is oceanic or continental) subducts in both sides, such as in the Mediterranean orogens (Doglioni et al., 1999).

Global Tectonic Asymmetries – Earth’s Rotation?

Geological and geophysical signatures of subduction and rift zones show a global pattern, suggesting “eastward” motion of the mantle relative to the lithosphere, regardless the hotspot reference frame, in support of a shallow origin for hotspots. Plates move in a sinusoidal patern (Figure 9), which is largely confirmed by space-geodesy-derived plate kinematics (e.g., Heflin et al., 2005). Along the flow lines, westerly dipping subduction zones are steeper than those that are E- or NE-directed, and associated orogens are characterized by lower structural and topographic elevation, and backarc basins, or by higher structural and morphological elevation and no backarc basins, respectively (Doglioni et al., 1999). The asymmetry is striking when western and eastern Pacific subduction zones are compared, and has usually been interpreted as related to the age of the downgoing oceanic lithosphere, i.e., older, cooler and denser in the west. However these differences persist elsewhere, regardless the age and composition of the downgoing lithosphere, e.g., in the Mediterranean Apennines and Carpathians vs. the Alps and Dinarides, or in the Banda and Sandwich arcs, where even continental or zero-age oceanic lithosphere is orientated almost vertical along westerly directed subduction zones. Rift zones are also asymmetric, with the eastern flanks greater elevated by about 100-300 m worldwide (Doglioni et al., 2003).

Westward drift of the lithosphere implies that plates have a general sense of motion and are not moving randomly. If we accept this postulate, we conclude that plates move at different velocities toward the "west" relative to the mantle along the flow lines shown in Figure 9. In this view, plates are more or less detached with respect to the mantle as a result of the decoupling at their base. The degree of decoupling is mainly controlled by the thickness and viscosity of the asthenosphere. Lateral variations in degree of decoupling could control the variable velocity of the overlying lithosphere (Figure 10). When a plate moves faster westward compared with an adjacent plate to the east, the resulting plate margin is extensional; when a plate moves faster westward compared with an adjacent plate to the west, their common margin will be convergent (Figure 10).


Figure 9. Connecting the directions of absolute plate motions that we can infer from large-scale rift zones or convergent belts from the past 40 Ma, we observe a coherent sinusoidal global flow field along which plates appear to move at different relative velocities in the geographic coordinate system (after Doglioni, 1993).

Figure 10. Cartoon illustrating that plates (cars) move along a common trail (e.g. the lines of Figure 9) but with different velocities toward the west, as indicated by the "westward" drift of the lithosphere relative to the mantle. The differential velocities control the tectonic environment and result from different viscosities in the decoupling surface, i.e., the asthenosphere. There is extension when the western plate moves westward faster with respect to the plate to the east, while convergence occurs when the plate to the east moves westward faster with respect to the plate to the west. When the car in the middle is "subducted", the tectonic regime switches to extension because the car to the west moves faster, e.g., the Basin & Range (after Doglioni, 1993).


The kinematic framework of shallow Pacific hotspots (Figure 8) constrains plate motions as entirely polarized toward the west relative to the deep mantle. This framework predicts a fundamental observation along E- or NE-directed subduction zones. In fact, with this reference frame, slabs tend to move out of the mantle, but are overridden by the upper plates. This detracts from slab pull as the mechanism for driving plate motions since the slab then does not sink into the mantle. In this view slabs are rather passive features (Figure 11).

The global scale asymmetry of tectonic features and the westward drift of the lithosphere supports a rotational component for the origin of plate tectonics (Scoppola et al., 2006). The westward drift of the lithosphere may be the result of three processes combined: (1) tidal torques act on the lithosphere generating a westerly directed torque decelerating Earth’s spin; (2) downwelling of denser material toward the bottom of the mantle and in the core slightly decreases the moment of inertia and speeds up Earth’s rotation, only partly counterbalancing tidal drag; and (3) thin (3–30 km) layers of very low viscosity hydrate channels occur in the asthenosphere. It is suggested that shear heating and mechanical fatigue self-perpetuate one or more channels of this kind, which provide the necessary decoupling zone of the lithosphere (Scoppola et al., 2006).


Figure 11. Cartoon assuming a Pacific plate (A) moving at 16 cm/yr. When plate motions are considered relative to the hotspot reference frame, the slabs of E- or NE-directed subduction zones may move out of the mantle. This is clearly the case for Hellenic subduction and, in the shallow hotspot reference frame, also for Andean subduction. This kinematic evidence of slabs moving out of the mantle casts doubt on slab pull as the driving mechanism of plate motions.


last updated 1st October, 2005