Beneath Yellowstone: Evaluating Plume and Nonplume Models Using
Teleseismic Images of the Upper Mantle
Eugene D. Humphreys, Department of Geological Sciences,
University of Oregon, Eugene, OR 97403-1272, USA
Kenneth G. Dueker, Department of Geology, University of Wyoming, Laramie,
WY 82071-3006, USA
Derek L. Schutt, Department of Geological Sciences, University of Oregon,
Eugene, OR 97403-1272, USA
Robert B. Smith, Department of Geology and Geophysics, University of Utah,
Salt Lake City, UT 84112-1183, USA
The Yellowstone hotspot commonly is thought to result from a stationary mantle plume rooted in the lower mantle over which North America moves. Yet Yellowstone's initiation and its association with the "backward" propagating Newberry hotspot across eastern Oregon pose difficult questions to those explaining Yellowstone as a simple consequence of a deep-seated plume. Teleseismic investigations across the Yellowstone topographic swell reveal: (1) the swell is held up by buoyant mantle of two typespartially molten mantle (of low seismic velocity) beneath the hotspot track and basalt-depleted mantle (of high velocity) beneath the rest of the swell; (2) an upwarped 660 km discontinuity beneath the Yellowstone hotspot track, as expected for relatively hot mantle at that depth, and an upwarped 410 km discontinuity, indicative of relatively cool mantle at this depth; and (3) anisotropic mantle with a preferred northeast orientation of olivine a axis, consistent with the strain expected for both plate motion and hotspot asthenosphere flow. Imaged mantle velocities can be reconciled with a plume hypothesis only if melt buoyancy within the hotspot asthenosphere drives convection, with melt segregating from the mantle beneath Yellowstone and residuum being deposited adjacent to the upwelling. Once such convection is admitted, an alternative, nonplume explanation for Yellowstone is possible, which has propagating convective rolls organized by the sense of shear across the asthenosphere. This explanation has the appeal that expected asthenospheric shear beneath the northwest United States predicts both the Yellowstone and Newberry hotspots with a single (upper mantle) process.
Recent teleseismic studies of the upper mantle beneath the Yellowstone swell provide insight on the origin of hotspots. The upper mantle beneath this swell now is one of the most seismically resolved regions on Earth, and the physical state of the upper mantle is accordingly well understood. However, interpretation of our findings in terms of hotspot processes remains ambiguous. Where once a plume origin seemed natural, we now consider a nonplume explanation to be at least as attractive. Studies currently collecting teleseismic data in the greater Yellowstone area should answer most questions currently deemed important about this hotspot.
Hotspots are defined by their anomalous surface manifestations, in particular, the time-transgressive propagation of volcanism over hundreds of kilometers, often with geochemically distinct lavas. Because of their inferred association with Earth's "stable interior," hotspots have played an important role in plate tectonic theory (e.g., Morgan, 1971). Also, their presumed role as the actively ascending part of mantle convection (e.g., Davies, 1993), arising from a lowermost mantle thermal boundary layer (e.g., the core-mantle boundary or a boundary in the lower mantle [Kellogg et al., 1999]), gives hotspots special geodynamic significance.
A mantle plume origin of hotspots is widely accepted, on the basis of the relative
fixedness of hotspots, the need for an anomalous heat source, and elevated 3He/4He
values thought to represent long-isolated "primordial" mantle (e.g., Kellogg and
Wasserburg, 1990). These observations combine to support the simple and elegant
model well known to earth scientists: Conduits rooted deep in the stable lower
mantle supply relatively undepleted mantle that feeds the surface expressions
of hotspots. In this model, hotspot magmatic activity begins with the impact of
a large plume-fed head of hotspot mantle, to which many flood basalts are attributed
(Duncan and Richards, 1991), and is followed by supply from the conduit, which
constructs a hotspot track leading away from the site of basalt flooding with
plate motion. However, the actual deep structure of hotspots, and therefore the
actual processes underlying their behavior, are not well understood. Furthermore,
an apparent absence of uplift prior to head impact and the unusual circumstances
under which hotspot magmatism often initiates (e.g., Anderson, 1999; Czamanske
et al., 1998) are difficult to incorporate into a plume model. As a result, alternative
hotspot hypotheses have been suggested with an upper mantle origin (e.g., Anderson,
1994) or a dominance of upper-mantle processes (Saltzer and Humphreys, 1997).
Of the hotspots investigated seismically, Iceland and Yellowstone are the two most thoroughly studied. A plume origin is argued for Iceland based on tomograms of the upper mantle (Wolfe et al., 1997) and imaged deflection of the temperature- sensitive 410 km and 660 km seismic discontinuities (Shen et al., 1998). However suggestive, an absence of seismic information from adjoining regions near Iceland provides little context in which to interpret the imaged structures. The Yellowstone hotspot offers the advantage of broad accessibility compared to oceanic hotspots, but teleseismic arrivals travel through the relatively complicated continental crust. The resulting tradeoff is that, compared with oceanic hotspots, the geometry of the ray set is superior for deep and regional imaging, but the data are degraded by greater amounts of crust-generated noise.
In most ways, Yellowstone is a typical hotspot. Figure
1 shows the Yellowstone-Newberry volcanic-tectonic system in the context of
the western United States. Yellowstone is characterized by a magmatic track and
a southwest-widening topographic swell left in the wake of the northeast-propagating
(relative to North America) hotspot. The topographic swell is thought to result
from plume flattening beneath the southwest-moving lithosphere (Anders and Sleep,
1992), as conceptualized in Figure
2. The swell's margins have been termed the "seismic parabola" (Anders et
al., 1989) for their seismicity (see Fig.
3). The magmatic track is the eastern Snake River Plain, which trends near
the symmetry axis of the swell; it is a topographic depression because basaltic
intrusions have loaded the crust, causing subsidence (Anders and Sleep, 1992).
For Yellowstone, as for some other hotspots, relatively high 3He/4He
values (Hearn et al., 1990) are thought by many to represent a lower mantle source.
The Yellowstone hotspot also is characterized by a strange initiation and a close association with another propagating continental hotspot, Newberry (Fig. 1). Yellowstone-Newberry magmatism began vigorously ca. 17 Ma with the eruptions of the central Nevada rift, Steens Mountains, and Columbia River flood basalts (Christiansen and Yeats, 1992). While often attributed to an impact of a plume head, there is no obvious indication of expected uplift preceding initial magmatism. Furthermore, these magmas erupted from a narrow set of fissures extending roughly north-south for ~700 km along the late Precambrian rift margin of North America (Fig. 1). Magmatic activity continued in this vicinity until ca. 12 Ma before propagating (irregularly) northeast toward Yellowstone and west-northwest toward Newberry. With a west-northwest direction of propagation, the Newberry hotspot cannot be connected to a stationary deep-seated plume
In teleseismic seismology, the distortion of seismic waves is analyzed to infer
the structure of the upper mantle and crust through which the waves propagated
as they arrive from distant earthquakes to an array of seismometers. To address
the structure beneath the Yellowstone swell, we deployed a seismic array occupying
~50 sites in a line trending across the width of the swell (Figs. 1
and 3). Our work follows
that of Evans (1982), who imaged upper mantle P-wave velocity structure by making
use of traveltime delays of the first arriving waves recorded on 1 Hz vertical-motion
seismometers. Our three-component broadband seismometers enabled receiver function
imaging for crust and upper mantle interfaces, S-wave splitting analysis for upper
mantle anisotropy, and P- and S-wave tomographic imaging of upper mantle velocity
variationsmethods now routine in teleseismic seismology.
Receiver Function Imaging of Crustal and Mantle Interfaces
A P wave partially converts to an S wave as it travels across an interface. At Earth's surface, the time delay of the converted S wave relative to the (faster traveling) P wave is proportional to the depth of the interface, and the magnitude of the S wave depends on the seismic contrast of the interface. Using these principals, the receiver function technique was used to image crustal and upper mantle discontinuities beneath our array. Combined with previous reflection-refraction investigations (Sparlin et al., 1982), receiver function analysis allowed Peng and Humphreys (1998) to image the crustal structures shown in Fig. 3: (1) a mid-crustal basalt sill across the width of the Snake River Plain, (2) an ~5 km thick partially molten lowermost crust across the width of the plain, and (3) a Moho that is approximately flat across the width of the seismic parabola but which thickens rapidly southeast of the swell. P-wave velocities (from Sparlin et al., 1982) suggest that the basalt sill is about half basalt and half the granitic country rock that comprises the upper crust away from the eastern Snake River Plain. The ~10 km thickness of the basalt sill therefore implies ~5 km of basalt added to the upper crust across the width of the eastern plain, and the partially molten lower crust suggests an underplating of probably 5 or more km of gabbroic crust. This crustal inflation is not reflected by a greater Moho depth, suggesting that lower crust was squeezed from beneath the eastern Snake River Plain to adjoining regions.
Perhaps the most important result of crustal imaging is the information it provides to model crustal density structure, which allows us to calculate mantle buoyancy across the width of our array. Mantle buoyancy holds the swell about 1 km higher than would normal mantle (e.g., eastern U.S. seaboard mantle), whereas mantle southeast of the swell is of more normal density (Peng and Humphreys, 1998). The highly (and uniformly) buoyant mantle across the width of the swell and the isostatic balance of the crust above it are consistent with standard thoughts on hotspots (e.g., Fig. 2).
Dueker and Sheehan (1997) used P-to-S conversions from the 410 km and 660 km seismic discontinuities to assess if locally hot mantle (e.g., plume-affected mantle) deflects these interfaces. Making use of the fact that interface deflection is of opposite sign on these interfaces for a given temperature anomaly (Bina and Helffrich, 1994), the observed thinning of the intervening layer by ~20 km (Fig. 3) beneath the Snake River Plain suggests a thermal anomaly there of 150200 °C. This result, however, is entirely a consequence of the upwarp in the 660 km discontinuity; the upwarped 410 km discontinuity implies cooler temperatures at this depth beneath the plain.
S-wave Splitting and Upper Mantle Anisotropy
Upper mantle strain via olivine dislocation creep tends to align the olivine
a axis in the finite elongation direction, and even moderate strains (one or more)
can create a significant fabric in this orientation (Ribe, 1992). Much like light
traveling through a crystal, an S wave passing through anisotropic upper mantle
will split into two orthogonally polarized waves, with the faster traveling wave
vibrating parallel to the direction of the a axis. The polarization of SKS waves
in a known direction makes them ideal for anisotropy studies. Figure
3 shows the results of split SKS waves recorded by our array (from Schutt
et al., 1998). The fast-wave polarizations trend approximately N65E, which is
nearly aligned with the hotspot track and North America absolute plate motion.
Waves that were naturally polarized with this orientation are not split, indicating
that anisotropy of a different orientation does not exist at greater depth. The
region of nearly uniform anisotropy orientation ends near the southeast margin
of the swell, and most of the western United States has orientations not aligned
with North America absolute plate motion (Savage and Sheehan, 2000). Thus the
asthenosphere beneath the Yellowstone swell defines a coherent, simple, and distinctive
upper mantle anisotropy domain.
There are two reasonable ways to interpret the observed upper mantle anisotropy. In the first, buoyant mantle beneath the swell is simply sheared by North America as it moves over a more stable interior (causing the a axis of olivine to align preferentially in the direction of plate transport. Another possibility is that a plume supplies buoyant mantle at a high rate, and this buoyant mantle flows to the southwest accommodated by deformation in the previously deposited low-viscosity hotspot asthenosphere (see Fig. 2). In this model, the southwest orientation of the finite elongation direction results from mantle flow driven by the local pressure gradient, and not by passive shear driven by plate motion. The similarity of results from different processes highlights the difficulties in understanding the mechanisms responsible for the mantle structure.
Tomographic Imaging of the Upper Mantle Velocity Variations
Figure 3 shows an
image of the upper mantle P-wave velocity structure (Saltzer and Humphreys, 1997).
Red and blue areas represent areas where waves propagate relatively slowly (red)
and quickly (blue). The blue areas have a seismic velocity that is about average
for mantle beneath continents. The low-velocity anomaly is about as wide as the
Snake River Plain, and is much narrower than the swell. The prominence of the
relatively high-velocity mantle beneath the high-standing swell seems at odds
with simple plume models, which have buoyant mantle distributed beneath the entire
swell (as in Fig. 2).
A nearly universal relation is that seismically fast rock is dense, yet the mantle
is highly buoyant across the width of the swell (as discussed in the "Receiver
Function" section). The only reasonable explanation for mantle that is both buoyant
and relatively fast is that it is significantly depleted in basaltic components.
Such depletion decreases density while increasing seismic velocity (Jordan, 1979),
and this is one of the few cases where density and velocity correlate inversely.
There is only one reasonable explanation for the imaged upper mantle structure:
The slow mantle is partially molten and the fast mantle has been depleted of basaltic
melt and currently is essentially devoid of significant melt. The observed ~7%
contrast in P-wave velocity across the width of the swell requires melt fractions
of up to ~2% in the red areas. The inferred compositional buoyancy of the
blue mantle results from 5%10% basalt segregation, and this compositional
buoyancy accounts for much of the swell's high elevation (Saltzer and Humphreys,
ASSESSING YELLOWSTONE'S ORIGIN
Imaged seismic structures and calculated mantle buoyancy beneath the Yellowstone
swell imply that the swell mantle is anomalously hot to depths of greater than
or equal to 200 km, is not anomalously hot at 410 km, and is hot again at 660
km. The red mantle beneath the Snake River Plain in Figure
3 is 1%2% partially molten, and the blue mantle beneath the adjoining
swell is 5%10% depleted of basaltic components. Mantle is anisotropic beneath
the swell, with finite extension oriented approximately N65E, and this orientation
does not vary with depth. This anisotropy is unique to the Yellowstone swell;
it contrasts with the western U.S. mantle away from the swell, which is more complexly
strained in different orientations.
Plume or no plume, we can make sense of these results only if we include local convection beneath Yellowstone, as illustrated in Figure 4 (Saltzer and Humphreys, 1997). A source of hot and fertile mantle is needed to produce significant basaltic melt upon adiabatic ascent, and the melt buoyancy drives convection (as modeled by Tackley and Stevenson, 1993). Melt release occurs when melt migration rates exceed convective flow rates (probably at melt fraction of ~2%). The escaping melt underplates and intrudes the crust. Convection ceases when the buoyancy of accumulating residuum equals that of the partially molten core. This mantle overturn occurs beneath the active caldera system (currently at Yellowstone). Then, the entire buoyant mass flattens as it is transported by plate motion away from the site of magma release, creating the southwest-widening swell. Mantle strain occurs primarily through southwest-northeastdirected simple shear. This could result from plate motion over a more stable interior, or by flow of Yellowstone asthenosphere away from Yellowstone and confined to the low-viscosity volume of hotspot asthenosphere previously deposited. These conclusions are sound in that they explain the peculiar seismic and density structure observed beneath our array, and they account for the magmatism. They do not specify a source for hot and fertile mantle. In particular, they permit the plume hypothesis for Yellowstone (but require convection to occur within the flattening hotspot asthenosphere).
However, once local upper mantle convection is recognized, there is potential to interpret Yellowstone entirely as an upper mantle phenomenon. Our model for this incorporates the flow interaction of asthenosphere with the volume of residuum created by prior melt release. Because this residuum is buoyant and relatively viscous, it tends to attach itself to the North America plate and move with this plate. The residuum protects the overlying plate by inhibiting subsequent magmatism, and as it is dragged along, asthenosphere flows beneath it and up as it passes the leading edge of the residuum body (as illustrated in Fig. 5). Melting occurs with ascent, driving the local convection that produces focused magmatism (as in Fig. 4) and adds to the residuum body.
Magmatic propagation therefore can be seen as a natural upper mantle process
when hot, fertile mantle is subjected to shear, as in plate transport. Schmelling
(2000) is producing such propagating melt-driven convective instabilities in computer
simulations. An especially attractive feature of this model is that to the west,
near the plate margin, upper mantle flow and shear directions probably are directed
west-northwest, in the direction of Newberry propagation, as a result of subduction-driven
corner flow (Fig. 5).
Hence, both Yellowstone and Newberry magmatism can be explained by a single (upper
mantle) mechanism. Furthermore, the divergence of upper mantle flow drives mantle
ascent between Newberry and Yellowstone (Fig.
5) that can account for the initiation of magmatism over an elongate region.
This could occur if unusually hot or fertile mantle were drawn up in a zone parallel
to the subducting plate, or if the ascending mantle were focused on an area of
fertile North America lithosphere, such as the Precambrian rift margin (where
fertile asthenosphere "froze" onto North America during Paleozoic downwarping
of this margin). And the observed 3He/4He anomaly can be
attributed to drawing up some primordial lower mantle.
One can ask why other Yellowstones and Newberrys are not distributed around the western United States. In fact, there are other magmatic trends oriented northwest (most western Great Basin magmatism) and northeast (e.g., Jemez, St. George) with associated low-velocity mantle trends (Humphreys and Dueker, 1994). The relative vigor of Yellowstone magmatism may result from its tectonic setting (adjacent to the Cascadia subduction zone and where focused northeast-oriented extension occurs), or perhaps it may simply represent the activity of an unusual lithospheric trend (Iyer and Healey, 1972) or relatively hot mantle.
This work was made possible by the PASSCAL instrument pool, and by National Science Foundation grants supporting work across the Yellowstone swell (EAR 9206565 and 9628474), the adjoining Wyoming and Colorado lithosphere (EAR 9418252 and 9614647) and greater Yellowstone area (EAR 9725598). Don Anderson, Norm Sleep, and Karl Karlstrom provided useful reviews.
Anders, M.H., Geissman, J.W., Piety, L.A., and Sullivan, J.T., 1989, Parabolic distribution of circumeastern Snake River Plain seismicity and latest Quaternary faulting: Migratory pattern and association with the Yellowstone hotspot: Journal of Geophysical Research, v. 94, p. 15891621.
Anders, M.H., and Sleep, N.H., 1992, Magmatism and extension:
The thermal and mechanical effects of the Yellowstone hotspot: Journal of Geophysical
Research, v. 97, p. 15,37915,393.
Anderson, D.L., 1994, The sublithospheric mantle as the source of continental flood basalts: The case against the continental lithosphere and plume head reservoirs: Earth and Planetary Science Letters, v. 123, p, 269280.
Anderson, D.L., 1999, A theory of the Earth: Hutton and Humpty Dumpty and Holmes, in Craig, G.Y., and Hull, J.H., eds., James Hutton, present and future: Geological Society [London] Special Publication 150, p. 1335.
Bina, C.R., and Helffrich, G., 1994, Phase transition Clapeyron
slopes and transition zone seismic discontinuity topography: Journal of Geophysical
Research, v. 99, p. 15,85315,860.
Burchfiel, B.C., Cowan, D.S., and Davis, G.A., 1992, Tectonic overview of the Cordilleran orogen in the western United States, in Burchfiel, B.C., et al., eds., The Cordilleran orogen: Geology of North America, Conterminous U.S.: Boulder, Colorado, Geological Society of America, v. G-3, p. 407479.
Christiansen, R.L., and Yeats, R.S., 1992, Post-Laramide geology of the U.S. Cordilleran region, in Burchfiel, B.C., et al., eds., The Cordilleran orogen: Geology of North America, Conterminous U.S.: Boulder, Colorado, Geological Society of America, v. G-3 (plates).
Czamanske, G.K., Gurevitch, A.B., Fedorenko, V., and Simonov, O., 1998, Demise of the Siberian plume: Paleogeographic and paleotectonic reconstruction from the prevolcanic and volcanic record, north-central Siberia: International Geology Review, v. 40, p. 95115.
Davies, G.F., 1993, Cooling the core and mantle by plume and plate flows: Geophysical Journal International, v. 115, p. 132146.
Dueker, K.G., and Sheehan, A.F., 1997, Mantle discontinuity structure from midpoint stacks of converted P to S waves across the Yellowstone hotspot track: Journal of Geophysical Research, v. 102, p. 83138327.
Duncan, R.A., and Richards, M.A., 1991, Hotspots, mantle plumes, flood basalts, and true polar wander: Reviews of Geophysics, v. 29, p. 3150.
Evans, J.R., 1982, Compressional wave velocity structure of the upper 350 km under the eastern Snake River Plain near Rexburg, Idaho: Journal of Geophysical Research, v. 87, p. 26542670.
Hearn, E.H., Kennedy, B.M., and Truesdell, A.T., 1990, Coupled variations in helium isotopes and fluid chemistry: Shoshone Geyser Basin, Yellowstone National Park: Geochimica et Cosmochimica Acta, v. 54, p. 31033113.
Humphreys, E.D., and Dueker, K.G., 1994, Physical state of the western U.S. upper mantle: Journal of Geophysical Research, v. 99, p. 96359650.
Iyer, H.M., and Healey, J.H., 1972, Teleseismic residuals at the LasaUSGS extended array and their interpretation in terms of crust and upper-mantle structure: Journal of Geophysical Research, v. 77, p. 10031027.
Jordan, T.H., 1979, Mineralogies, densities, and seismic velocities
of garnet lherzolites and their geophysical implications, in Boyd, F.R., et al.,
eds., The mantle sample: Inclusions in kimberlites and other volcanics: Proceedings
of the 2nd International Kimberlite Conference, Volume 2: Washington, D.C., American
Geophysical Union, p. 114.
Karato, S.-I., 1992, On the Lehmann discontinuity: Geophysical
Research Letters, v. 51, p. 22552258.
Kellogg, L.H., Hager, B.H., and Van der Hilst, R.D., 1999, Compositional stratification in the deep mantle: Science, v. 283, p. 18811884.
Kellogg, L.H., and Wasserburg, G.J., 1990, The role of plumes in mantle helium fluxes: Earth and Planetary Science Letters, v. 99, p. 276289.
Morgan, W.J., 1971, Convection plumes in the lower mantle: Nature,
v. 230, p. 4243.
MacLeod, N.S., Walker, G.W., and McKee, E.H., 1976, Geothermal significance of eastward increase in age of upper Cenozoic rhyolitic domes in southeastern Oregon, in Second United Nations Symposium on the Development and Use of Geothermal Resources: Washington, D.C., U.S. Government Printing Office, v. 1, p. 465474.
Peng, X., and Humphreys, E.D., 1998, Crustal velocity structure
across the eastern Snake River Plain and the Yellowstone swell: Journal of Geophysical
Research, v. 103, p. 71717186.
Pierce, K.L., and Morgan, L.A., 1992, The track of the Yellowstone hotspot: Volcanism, faulting, and uplift: Geological Society of America Memoir 179, p. 153.
Ribe, N.M., and Christensen, U.R., 1994, Three-dimensional modeling of plume-lithosphere interaction: Journal of Geophysical Research, v. 99, p. 669682.
Ribe, N.M., 1992, On the relation between seismic anisotropy and finite strain: Journal of Geophysical Research, v. 97, p. 87378747.
Saltzer, R.L., and Humphreys, E.D., 1997, Upper mantle P wave velocity structure of the eastern Snake River Plain and its relationship to geodynamic models of the region: Journal of Geophysical Research, v. 102, p. 11,82911,841.
Savage, M.K., and Sheehan, A.F., 2000, Seismic anisotropy and mantle flow from the Great Basin to the Great Plains, western United States: Journal of Geophysical Research, v. 105, p. 13,72513,734.
Schutt, D., Humphreys, E.D., and Dueker, K., 1998, Anisotropy
of the Yellowstone hotspot wake, eastern Snake River Plain, Idaho: Pure and Applied
Geophysics, v. 151, p. 443462.
Schmelling, H., 2000, Partial melting and melt segregation in a convecting mantle, in Bagdassarov, N., et al., eds., Physics and chemistry of partially molten rocks: Dordrecht, Kluwer Academic Publishers, p. 141178.
Shen, Y., Solomon, S.C., Bjarnason, I.T., and Wolfe, C.J., 1998, Seismic evidence for a lower-mantle origin of the Iceland Plume: Nature, v. 395, p. 6265.
Sparlin, M.A., Braile, L.W., and Smith, R.B., 1982, Crustal structure of the eastern Snake River Plain determined from ray trace modeling of seismic refraction data: Journal of Geophysical Research, v. 87, p. 26192633.
Tackley, P.J., and Stevenson, D.J., 1993, A mechanism for spontaneous self-perpetuating volcanism on the terrestrial planets, in Stone, D.B., and Runcorn, S.K., eds., Flow and creep in the solar system: Observations, modeling and theory: Norwell Massachusetts, Kluwer Academic Press, p. 307322.
Wolfe, C.J., Bjarnason, I.T., VanDecar, J.C., and Solomon, S.C., 1997, Seismic structure of the Iceland mantle plume: Nature, v. 385, p. 245247.
Manuscript received August 11, 2000; accepted October 10, 2000.
Return to Current Issue