Peter D. Clift

Department of Geology and Petroleum Geology, School of Geosciences, Meston Building, King's College, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom

pclift@abdn.ac.uk

Introduction

Evidence that the Earth’s mantle circulates is well documented, yet the influence of deep-seated mantle plumes, some possibly originating at the core-mantle boundary, is still hotly contested (e.g., Anderson, 2003; Foulger & Natland, 2003). Areas of elevated topography linked to magmatism greater than normal in the ocean basins have been historically attributed to greater melting above a steady-state plume or to the impacting of a newly initiating plume head on the base of the lithosphere (e.g., Campbell & Griffiths, 1990). New models, however, attribute excess melting to lithospheric processes above essentially normal upper mantle (Anderson, 2003). Resolution of this debate is important for understanding the major controls on terrestrial magmatism and the geochemical evolution of the Earth.

Although different plume models propose different magnitudes of thermal anomaly, ranging up to 350°C (Farnetani & Richards, 1994), a common characteristic of deep-seated mantle plumes is their excess heat relative to the surrounding ambient asthenosphere. Greater heat results in plume mantle being less dense than normal, allowing it to rise to the base of the lithosphere. This same excess heat and buoyancy is also inferred to cause the long-wavelength uplift and shallowing of the seafloor which can be seen in modern bathymetry around active hotspots or superswell areas (Crough, 1978; Detrick & Crough, 1978; McNutt & Judge, 1990; see also Superswell page).

Ancient plume-driven uplift can also be preserved in the sedimentary rocks deposited in hotspot regions (e.g., White & Lovell, 1997). As a result of plate motion, crust that is affected by plume activity is removed from over the mantle thermal anomaly. Continued upper-mantle convection then removes the hotter, less dense material and allows the crust to subside to normal depths over periods of tens of millions of years (Olson, 1990; Griffiths & Campbell, 1991). Although the sedimentary records from continental margins, most notably in the NE Atlantic and the Kerguelen Plateau, have been interpreted as showing the effects of plume activity (Brodie & White, 1994; White & Lovell, 1997; Clift et al., 1995; Coffin & Gahagan, 1995), anomalous subsidence possibly linked to plume activity is most readily identified in oceanic crust because the subsidence behavior of oceanic lithosphere is relatively simple and well characterized (Parsons & Sclater, 1977; Stein & Stein, 1992; see also Heatflow page). Although the absolute depth of any given piece of oceanic crust will vary depending on the thickness and density of the crust, the rate at which subsidence occurs subsequent to crustal accretion can be used to quantify mantle buoyancy because this is controlled by cooling and thickening of the mantle lithosphere (Parsons & Sclater, 1977).

Isolating the effect of hot asthenosphere on plate subsidence is simplest if the crust was emplaced close to the crest of a mid-ocean ridge, as this allows a simple comparison of the subsidence reconstructed from the sedimentary record with that of normal oceanic crust. Clift (1997) used this approach to deduce anomalously shallow depths within the oldest oceanic crust adjacent to the volcanic rifted margins of the NE Atlantic shortly after continental break-up in the Eocene (see also Iceland pages). If all the anomalous depth is assigned to temperature, rather than compositional buoyancy or dynamic flow, then the degree of misfit with the “normal” oceanic model can be used to estimate maximum temperature anomaly.

The subsidence of lithosphere affected by plume activity has been modeled in a variety of ways, treating the plume as either an isostatic anomaly or as a region of dynamic flow (e.g., Ribe & Christensen, 1994; Sleep, 1987, 1990). Dynamic models generate uplift through the viscous normal stresses that would be imposed on the base of the lithosphere by an upwelling column of material. They fit modern bathymetry closely with no requirement for a temperature anomaly. However, results consistent with the evolving bathymetry along the Hawaiian seamount chain have also been generated by simple isostatic, conductive cooling models (Sleep, 1987) such as that of Ito & Clift (1998) which uses a one-dimensional thermal diffusion approach. That model allows the hot cushion of plume material to diffuse away after the initial emplacement resulting in gradually reducing thermally induced buoyancy with time.

The predicted subsidence history is slightly more complex if the emplacement of hot mantle material and associated hotspot volcanism occured away from an oceanic spreading center. In this case the water depth shallows as equivalent age crust away from the hotspot as a result of crustal thickening and the buoyancy of the mantle plume material. The rate of subsidence of the crust after removal from the hotspot area and its swell is faster than crust of similar age that was not influenced by the hotspot. Detrick & Crough (1978) noted such behavior in the Hawaiian seamount chain, a phenomenon that they attributed to thermal erosion of the Pacific lithosphere after it passes over the Hawaiian plume. In this model, although the Hawaiian Islands are much shallower than the surrounding abyssal seafloor they subside at a faster rate after they are removed from the immediate influence of the thermally erosive plume. Alternatively Phipps Morgan et al. (1995) explained the bathymetry of the Hawaiian Islands as being the result of melt extraction in the asthenosphere producing a buoyant pillow of mantle residuum centered beneath the hotspot. This buoyant mantle is partially entrained with the lithospheric plate as the latter moves away from the hotspot.

In the case of old hotspot provinces it is not always clear whether the hotspot lavas were emplaced on axis or not. This may be investigated by comparing basement ages with lithospheric ages. In the present study I infer the age of the surrounding oceanic lithosphere on the basis of the nearest identified magnetic anomaly, while the age of hotspot volcanism is taken from the radiometric age of basement lavas, or from the biostratigraphic age of overlying sediment if the former is not available.

In order to assess the magnitude of mantle temperature anomalies associated with hotspot magmatism I have reconstructed the subsidence history of a series of oceanic plateaus, ridges or seamounts using their sedimentary covers. The sites studied include all the major ocean basins and a wide period of geologic time (Figure 2). The analyses were made possible through coring by the Deep Sea Drilling Project (DSDP) and the Ocean Drilling Program (ODP) which recovered sediments overlying the volcanic sequences. The sediments and fossils found in the cores allow the depositional environment, and especially the water depth of deposition, to be constrained. The data were compiled from cruise reports and associated postcruise papers, particularly those describing benthic foraminifer fauna, which are sensitive water depth indicators.

Figure 2: The locations of ODP and DSDP drill sites considered in this study. They span all major ocean basins and long periods of geologic time.

Water depth estimation is the principle uncertainty in the reconstruction of vertical motions of igneous basement and is generally better where water depths are shallow (< 200 m) as this is the range where sedimentary structures and benthic foraminifers show the greatest resolution. Nonetheless, some resolution is possible at greater depths. For the purpose of the present analysis I assumed that the current depth of features is the maximum during their history. The stratigraphy at each drill site was “backstripped” in order to isolate the tectonic component of the subsidence history. This involves the correction for sediment loading and restoration of the seafloor to the estimated depth in the past for each dated interval, assuming local isostasy (Sclater & Christie, 1980). In doing so the vertical motion of the igneous basement subsequent to the end of volcanism can be reconstructed at each drill site. No correction was made for sea-level variations because of the current lack of clarity concerning the timing and magnitude of this effect. However, the uncertainties introduced by this omission are small (~ 100 m) compared to the much greater values of total subsidence and the general uncertainties in the water depths.

Examples

Figures 3 – 5 show the results of backstripping and compare each reconstructed history with that of normal oceanic crust (Stein & Stein, 1992). In each case the normal oceanic subsidence curve was shifted to match the modern depth to basement in order to compare the rates and not the total depth of subsidence. It is the rate of subsidence that is affected by the mantle thermal state. Absolute depth is influenced by crustal thickness and thus the degree of melting, which is partially controlled by mantle temperature, and also by the chemical state of the mantle and the nature of upwelling under the hotspot (Mutter et al., 1988). In the case of features emplaced off-axis a second oceanic subsidence curve is shown indicating the predicted rate of subsidence if the lithosphere had been totally reset to zero age at the time of eruption of the basement.

Some of the drill sites show clear evidence for depths early in their history shallower than predicted for emplacement above mantle of normal temperature. Sites in the NE Atlantic, the Hatton Bank, SE Greenland and the Vøring Plateau and even SE Newfoundland show early depths that are anomalously shallow, i.e., subsidence curves that are steeper than expected, similar to that seen in modern Hawaii. Modeling by Clift (1997) indicated that the moderate nature of the depth anomaly was consistent with a temperature anomaly of around + 100°C. Outside this region definitive evidence for shallower depths following eruption is scarce as many of the sites fall within the predicted trend of crust overlying normal upper mantle temperatures. Where water depth uncertainties are large the possibility of hot mantle exists but is not required. There appear to be as many sites, if not more, where the reconstructed depth to basement after emplacement is deeper than predicted for normal mantle temperatures. This raises the possibility that the asthenosphere under the sites has actually heated, not cooled since emplacement, though this is perhaps unlikely.

A number of models have been proposed to account for the lack of subsidence observed in old hotspot regions. The relative lack of subsidence and initial submarine eruption of lavas in the Ontong Java and Manihiki Plateaus led Ito & Clift (1998) to propose growth of these features in two or more stages. In practice they argued that the plateaus are inflated through time by underplating spanning the period 120 to 90 Ma. In this case uplift caused by crustal growth counteracted the subsidence driven by lithospheric cooling. This model is practical in that particular setting because these plateaux remained approximately stationary over the proposed plume source during that time. However, it is not appropriate to many other igneous provinces, which moved rapidly relative to their sources, e.g., the NinetyEast Ridge.

The Mid Pacific Mountains and MIT Guyot

The Mid Pacific Mountains and MIT Guyot (ODP site 878) are good examples of features that appear to represent excess melting, moved rapidly in the hotspot reference frame after emplacement, and are not associated with a hotspot track. Their subsidence histories are slower than predicted for normal oceanic crust of their age, the opposite of that predicted if they were emplaced over hot asthenosphere. Long-term crustal accretion is not a reasonable explanation for the slow subsidence of this type of hotspot. Instead mantle processes must be inhibiting subsidence.

Most likely the presence of buoyant, depleted mantle residue from melting is acting as a cushion beneath them. Mantle from which melt has been extracted has a lower density than normal depleted upper mantle. If this material was not advected away it could inhibit subsidence (Robinson, 1988). Such a compositional-derived density model for hotspots was invoked by Phipps Morgan et al. (1995) for the Hawaiian swell and may be applicable to many Pacific and Indian Ocean hotspot features. What is surprising is that although Phipps Morgan et al. (1995) consider depleted residue to be more viscous that normal asthenosphere, they predict that it disperses over long periods of geologic time, resulting in a reduction in the height and width of the Hawaiian Swell. That prediction is at odds with the long-term lack of subsidence found by many of the reconstructions presented here, which require much of the buoyant residue to remain under the hotspot edifice until the present day. This implies that it has a high viscosity.

Shatksy Rise and the NinetyEast Ridge

The subsidence patterns of hotspot features in the modern oceans only occasionally fit simple models for rapid crustal generation over hot mantle followed by subsidence faster than normal after the ridge or seamount was removed from the thermal anomaly by plate motion. The normal or slower-than-normal subsidence seen at many hotspots may signify the presence of a buoyant mantle root. Such subsidence patterns do not preclude mantle hotter than normal at the time of emplacement, but they do not require it. Many hotspots are compatible with emplacement over mantle of normal temperature, followed by conductive cooling of the type associated with regular oceanic crust. See also Shatsky page.

The North Atlantic margins: Discussion

The North Atlantic margins where subsidence anomalies do exist are consistent with mantle temperature anomalies of ~ 100°C, but preclude very hot plume models. This conclusion is supported by inversion modeling of rare earth element data from hotspot volcanic rocks that predict temperature anomalies not exceeding 150°C (e.g., Watson, 1993; Kent & McKenzie, 1994; Kerr, 1994) and with modeling of the bathymetry of the North Atlantic assuming a plume-type model (Ribe et al., 1995). If any of the anomalous shallow depths seen after emplacement are caused by buoyancy resulting from mantle composition or dynamic mantle upwelling then the temperature anomalies inferred from the subsidence anomalies would be lower.

The subsidence behavior of the North Atlantic margins and the Hawaiian seamount chain appears to differ from many other hotspots. This may be linked to a different origin associated with hot mantle that is not required in many other areas. Fast subsidence in these regions begs the question of why they are not affected by a buoyant mantle residual root. Perhaps vigorous mantle circulation of the type proposed by Mutter et al. (1988) to increase melting was responsible for the removal of any residue that did form. There is little evidence to support the current presence of a buoyant root in the NE Atlantic related to Tertiary magmatism. Temporary uplift of the NW European shelf away from the immediate vicinity of the volcanic margins during their formation is consistent with the moderate thermal anomaly derived from the oceanic crust subsidence histories (Clift et al., 1998), modern heatflow (Stein & Stein, 2003) and the modeling of rare earth element data (Kerr, 1994).

Summary

The sedimentary records of many hotspot provinces worldwide do not require the mantle to be hotter than normal at the time of magmatic crustal growth. In the few examples where hotter mantle is required the temperature anomaly is no more than ~ 100°C and could be much less if dynamic flow and compositional buoyancy contribute to driving temporary uplift.

Other pages on mantle temperature include Heatflow, Temperature and Mantle temperature.

References

Discussion

Mon Jun 7, 2004: Edward Winterer
I understand why this paper of Peter Clift got into trouble with reviewers (I was not one of them). The subsidence curves based on benthic forams are so poorly constrained as to be almost meaningless. I refer you to a paper I wrote about the limitations and pitfalls of all this in Winterer, E.L. (1998) Paleobathymetry of Mediterranean Tethyan Jurassic pelagic sediments. Memorie, Società Geologica Italiana, 53, 97-131. Also, I actually calculated a subsidence curve at Resolution Guyot in the Midpacs, based on the best data we got from drilling,. He should at least have looked at that one, in ODP volume 143, published in 1993.

Mon Jun 7, 2004: Peter Clift
Thanks for your comments on the web page and the reference to your paper in Memorie, Società Geologica Italiana.  I agree of course that the uncertainties in deep water sediments are pretty poor, although we can at least be certain that they are deep water sediments and not shallow.  The web page actually only shows a small number of the sites that I have backstripped.  I have in fact looked at ODP Site 866 on Resolution Guyot within the Mid Pacs and it too shows no clear evidence for anomalous subsidence rates that would require a hot asthenosphere.  As you know this curve is constrained by the presence of oolitiic limestone of Barremian age at the base so my confidence is pretty good with that result.  In practise this is often the case that the depth of eruption is well known but is overlain by deep water poorly known sediments.  Nonetheless, the total amount of subsidence is thus well constrained and can be used to compare asthenospheric temperatures at the time of eruption and the modern day.

Mon Jun 7, 2004: Edward Winterer
One other thing to keep in mind is that seamounts, especially big ones, go through a two-stage subsidence. Hawaii, for example, is still in the flexural, isostatic stage, which can be very rapid (2.5 m/ky) but as soon as volcanism ceases, the rate shifts to that of the underlying oceanic lithosphere, which subsides (thermally) much more slowly. Dave Clague and his MBARI gang and I have been working and publishing on this and Ed Purdy and I have a (submitted) paper on barrier reefs that exploits this behavior. The 1600-m Resolution Guyot shallow-water carbonate succession records only the slower regime, while the older, faster rates are in drowned fringing reef down the flanks, neither exposed (buried under talus) nor sampled.

A second cause of complication is the moat-and-swell effect on nearby seamounts, exemplified by islands like Oahu, which record a bobbing motion as younger volcanoes are emplaced to the south and as huge landslides are offloaded. As you point out, you are looking only for the “big” effect (hot asthenosphere or not), but it would be nice to find really clean examples.

Mon Jun 7, 2004: Peter Clift
I agree that subsidence is a complex thing and that flexure around hotspots can complicate the subsidence history but generally speaking I would think that this would be a moderate amplitude, short duration ripple in subsidence histories that last 50-120 Ma. I am focusing only on the longer-duration thermal subsidence in this work, as I am not sure that the syn-volcanic rapid subsidence is even preserved in much of the sediment cover, so I like your idea that the fringe reefs may be the key to that process.

I am intrigued to see how very infrequently the model for rapid emplacement over a plume seems to be even closely satisfied, although it does seem to work in the Hawaiian islands and in the NE Atlantic to some extent. My feeling is that these areas are likely somewhat hot.

Thursday Jun 10, 2004: Ian Campbell
Peter Clift has made an assumption in his modeling; he assumes that the old, hot material in a plume below the lithosphere is dissipated by a combination of mantle convection and the plate moving away from the plume.  This is not what is expected.  Because the old plume material remains hot for a long time it remains buoyant and continues to try to rise.  It sticks to the overlying plate and follows plate when it move away from the plume.  A plume head takes ~1,000 Myr to cool.  We are less certain how long a plume tail takes to cool because it diameter is less well constrained. 

I have always been puzzled by the “documented” lack of uplift before the Siberian plume.  Gerry Czamanske's contribution makes this clearer [Ed: See Siberia page].  He describe a basin developing prior to the eruption of the Traps, which he correctly interprets as evidence for subsidence, but then goes onto describe erosional features below the Traps.  The plume hypothesis makes no prediction of what will happen 10 to 50 Myr prior to eruption of the basalts so it does not preclude the formation of a sedimentary basin.  However it does predict uplift in the 1 to 5 Myr before eruption and the most likely way this will manifest itself is in erosion below the flood basalt, just what Gerry describes.

You will have seen Montelli's paper [Ed: see A variety of plumes in the mantle?] in which she claims to have been able to trace 20 plumes to the CMB.  However, some such as Iceland stop at the bottom of the upper mantle which I would not expect.  Plumes must come from a thermal boundary layer and it is unlikely that there are two in the mantle.

Thursday Jun 10, 2004: Peter Clift
Ian Campbell is right that the modelling that predicts faster subsidence of plume influenced lithosphere requires dissipation of the plume head asthenosphere through time. The lack of this faster subsidence in many old hotspots does need some long-term buoyancy in the mantle to explain it. I had favored the model of a depleted mantle root, but a permanent hot cushion would do the job as well. The problem in both cases is that the sediment record in these scenarios does not provide any information about mantle temperature, neither supporting or disproving hot plume models.

There are exceptions in Hawaii and the NE Atlantic where the cooling models I used do seem to work. Its not clear to me at the moment why a buoyant plume cushion or a depleted root would not remain attached to the base of the plate in these areas alone, especially as these are the most powerful modern plume candidates. Dynamic uplift in the NE Atlantic is seen at the onset of hotspot volcanism though the NW European area for sure, though it's magnitude is more in tune with a 100°C anomaly plume rather than a 300°C one.

One thing that still puzzles me if the hot cushion model is correct is why so many of the hotspots fall so close to the Stein & Stein (1992) regular depth-age curve. It would seem remarkable that 90 East Ridge, Hess Rise, Mid Pacific Mountains, MIT Guyot, Manahiki Plateau, Walvis Ridge and the Rio Grande Ridge for example all show curves that agree well with the normal cooling curve. They don't seem to be either subsiding quickly as I would have guessed from my model or being buoyed up in the fashion that Ian describes. Either the two processes are cancelling each other out or they are cooling and subsiding above mantle of close to normal temperatures.

Friday June 11th, 2004: Don Anderson
The plume model does not predict rapid emplacement (or rapid termination of magmatism). Courtillot & Richards in their plume heads-tails paper summarized the known properties of LIPs and pointed out that they were generally erupted in a million years or less. Since these (e.g. LIPs) were known to be caused by plumes, this became a property of plumes, rather than an observation to be explained. In subsequent papers, Richards and others–with fluid dynamic modeling – could not come up with such short times (emplacement times being limited by the viscosity of the lower mantle rather than local stress changes). On the other hand, the lithospheric stress-valve mechanism permits rapid onset and shut-off of magmatism, but can also allow long durations. The numerous very short volcanic chains, and the rapid change in magma flux – even in Hawaii – strongly support the stress control hypothesis. The main remaining question, then, is whether the asthenosphere is close to the melting point or whether it is so cold and uniform that deep heat and material is required to melt it and replenish it, as in the plume hypothesis.

The residual plume-head thermal effect that Anderson, Tanimoto and Zhang looked for in their surface-wave tomographic studies–the absence of an upper mantle LVZ under LIPs was one of the reasons that have been used to discount thermal plumes (plus low heat flow and no uplift). It is often stated that tomography does not have the resolution to detect plumes but this shallow effect, plus the predicted ponding beneath other barriers, such as the 650 km endothermic boundary, is easy to detect with surface waves and other techniques. Plume heads, and the flattened pancakes that plume heads become, have not been found. Plume tails have not been found either except with first arrival P-waves where the coverage is poor (or even non-existent) and source, streaking and smearing artfacts cannot be cancelled out. Skepticism (“documented” – in quotes – lack of uplift), puzzlement, and “not expected”s are common when a favorite paradigm is in distress.

These, plus apparent paradoxes and unexplained coincidences (association of flood basalts with deep sedimentary basins, sutures and craton edges) have motivated some Earth scientists to explore mechanisms that do not involve transfer of heat and material from the core mantle boundary into the upper mantle. A large plume head in the upper mantle must cause uplift, low-velocity and, eventually, high-heatflow, and anomalous subsidence. The lack of support for any of these predictions at any proposed plume site should give one pause, and sufficient motivation to investigate alternative models, including those that do not require core-mantle type temperatures under LIPs. Both hot-cushions and fast-residues (mutually exclusive predictions and post-dictions of the plume-head hypothesis), and lithospheric rejuvenations, are easy to spot with surface waves and receiver functions, but are not required in alternate hypotheses. Although one might be able to rationalize the lack of precursory uplift and anomalous subsidence in a few places, the fact that all continental flood basalt (CFB) provinces, as far as I know, are on top of deep sedimentary basins suggests a tectonic origin (stretching, thinning, extension and subsidence). The fact that all CFB are emplaced in sutures – involving one or more cratons – also suggests, if not demands, a tectonic, rather than a CMB, explanation.

Modern research has shown that the mantle is hotter than assumed in the plume model, that the melting point is lower, and that the melting point and chemistry of the shallow mantle are highly variable. We also know that the stress in the lithosphere is highly variable and that the stress-valve mechanism can turn magmatism on and off. Recent amendments to the plume hypothesis involve large distance lateral transport in the shallow mantle, rapid motions of hotspots, and the use of eclogite to reduce the melting point. These are features of models that do not involve plumes at all. Altogether, this means that plumes are not necessary to explain variations in bathymetry and melting rates or the locations of melting anomalies. Arguments based on uplift and subsidence, as well as those based on fixity, chemistry, eruption duration and heatflow, seem to favor lithospheric and asthenospheric (basically non-thermal) explanations, without the numerous surprises and paradoxes that have accompanied the plume model since its inception. The plate model predicts long-wavelength variations in mantle temperature and heat flow, but abrupt changes in lithospheric stress and thickness, mantle composition and melting point; this seems to be what the data is telling us. Peter’s work, when added to magma temperatures, LIP durations, heatflow, absence of uplift and lithospheric thinning, surface wave tomography and mantle petrology, seems to close the book on localized high-temperature explanations for melting anomalies.

The lack of fixity and parallelism of chains, and the large number of isolated volcanoes (no associated chains, no LIPs, no heads, no tracks) with “hotspot” chemistry, has motivated some workers to seriously look at the stress and crack ideas again (even before the uplift/subsidence paradoxes were raised) . Physics (pressure and Prandtl number considerations) also is not on the side of the plume model. Athermal and tectonic explanations – involving fertility, stress and ponding, rather than high temperature – need to be put back on the table. The alternative is to keep modifying a favorite hypothesis – and rationalizing the data and the failure of predictions – in ways that Karl Popper warned us against, at least in the practice of good science. More recently, philosophers of science have discredited Popper’s idea of falsification because scientists have a facility for propping up failing theories and rationalizing contrary data-falsification appears to be no more viable than proof.

last updated 4th July, 2006