Report

The Yellowstone magmatic system from the mantle plume to the upper crust

See allHide authors and affiliations

Science  15 May 2015:
Vol. 348, Issue 6236, pp. 773-776
DOI: 10.1126/science.aaa5648

Yellowstone's missing magmatic link

Yellowstone is an extensively studied “supervolcano” that has a large supply of heat coming from a pool of magma near the surface and the mantle below. A link between these two features has long been suspected. Huang et al. imaged the lower crust using seismic tomography (see the Perspective by Shapiro and Koulakov). Their findings provide an estimate of the total amount of molten rock beneath Yellowstone and help to explain the large amount of volcanic gases escaping from the region.

Science, this issue p. 773; see also p. 758

Abstract

The Yellowstone supervolcano is one of the largest active continental silicic volcanic fields in the world. An understanding of its properties is key to enhancing our knowledge of volcanic mechanisms and corresponding risk. Using a joint local and teleseismic earthquake P-wave seismic inversion, we revealed a basaltic lower-crustal magma body that provides a magmatic link between the Yellowstone mantle plume and the previously imaged upper-crustal magma reservoir. This lower-crustal magma body has a volume of 46,000 cubic kilometers, ~4.5 times that of the upper-crustal magma reservoir, and contains a melt fraction of ~2%. These estimates are critical to understanding the evolution of bimodal basaltic-rhyolitic volcanism, explaining the magnitude of CO2 discharge, and constraining dynamic models of the magmatic system for volcanic hazard assessment.

The interaction of the North American Plate moving southwestward across a mantle plume created the Snake River Plain, a bimodal basalt-rhyolite volcanic system dating to 16.5 million years ago (Ma) (1). The Yellowstone volcanic field that sits at the eastern end of the plain is the youngest manifestation of the hotspot and is characterized by extensive earthquakes (2, 3), episodic ground deformation (4), high heat flux averaging 2000 mW m−2 (2, 5), and the largest continental hydrothermal system in the world (6, 7). The most recent cataclysmic eruption occurred at 0.64 Ma and created the 40 km × 60 km Yellowstone caldera, which is filled with rhyolitic lava flows as young as 70,000 years (Fig. 1). Earlier teleseismic studies have imaged a west-northwest–dipping plume extending into the top of the lower mantle (811). Local earthquake tomography and waveform modeling studies have revealed an upper-crustal magma reservoir between 5 and 16 km depth (3, 12, 13), of which the shallowest portion correlates with the largest area of hydrothermal activity and extends 15 km northeast of the caldera (3). Even with a large volume of >4000 km3 and a high melt fraction of up to 32% (2, 3, 13), this upper-crustal reservoir cannot account for the large CO2 flux of 4.5 × 107 kg daily and requires additional input of basaltic magma invading the lower to middle crust (6, 7). Moreover, it is unclear how the mantle plume interacts with the crustal volcanic system. The connection between the shallow magma reservoir (above 16 km depth) and the deep mantle plume (below 60 km depth) is therefore a critical component to understand the entire Yellowstone magmatic system but has never been imaged distinctly, despite other geophysical and geologic evidence that hypothesizes the presence of a continuous crustal magma body (2, 5, 7).

Fig. 1 Map of the seismic stations used in this study and the P-wave velocity cross-section locations in the Yellowstone area.

Stations and earthquakes are denoted by blue triangles and red dots, respectively. Black solid and dashed lines outline the Late Quaternary Yellowstone caldera and resurgent domes. Green lines represent the tectonic division of the Eastern Snake River Plain (ESRP). Yellow and thin dotted lines are the border of Yellowstone National Park and the surrounding state borders, respectively. Locations of the cross sections in Fig. 3 are shown by thick black lines with labels. The inset map shows the location of the Yellowstone area (red box) and the major tectonic boundaries (green lines) in the western United States.

Imaging the position and size of the entire volcanic plumbing system is also important to constrain magmatic dynamics modeling for further hazard assessment (14). Many local seismic array experiments have been conducted on volcanoes such as Askja, Iceland (15); Axial, Juan de Fuca mid-ocean ridge (16); Kilauea, Hawaii (17); and Mount St. Helens, Washington (18). A common observation of these experiments is an imaged shallow low-velocity body (LVB) at depths of 5 to 10 km, which is often interpreted as a magma reservoir. Several of these studies also image the top of a second LVB at greater crustal depths (18, 19); however, they usually quickly lose resolution with depth because of the limited array aperture and the shallowly distributed earthquakes (often less than ~10 km depth). Large arrays for teleseismic tomography can only focus on mantle images and poorly resolved crustal structures. A complete framework of the mantle-crust volcanic system under volcanoes has not yet been elucidated.

By combining data from the dense seismic arrays of the Yellowstone, Teton, and Snake River Plain (SRP) regional seismic networks, the NOISY array (20), and the wide-aperture EarthScope Transportable Array (Fig. 1), we present an image of the entire volcanic plumbing system beneath the Yellowstone caldera and reveal a large basaltic magma reservoir in the lower to middle crust by using a joint tomographic inversion of local and teleseismic earthquake data (2123). The seismic data used in this study were compiled from previous studies (3, 24) and consist of 47,815 P-wave first arrivals from 4520 local earthquakes and 4605 relative arrival times from 329 teleseismic earthquakes. A local earthquake inversion is first conducted to obtain a three-dimensional initial crustal model and reduce the dominance of the local data prior to the joint inversion (25).

We found a large east-northeast–west-southwest elongated LVB beneath the Yellowstone caldera (Fig. 2, A to C, and Fig. 3A) at depths shallower than 20 km, consistent with previous studies (3, 25). However, at depths of 20 to 50 km, another larger LVB with >5% P-wave velocity (VP) reduction also emerges in our model (Fig. 2, D and E, and Fig. 3, A and C). From cross section AA′ (Fig. 1 and Fig. 3A), it is clear that this deeper LVB is immediately beneath the shallow LVB. The existence of the lower-crustal LVB and the separation between the two imaged crustal LVBs are validated through characteristic-model synthetic tests and finite-frequency analysis (25), indicating that the deep crustal LVB is a separate magma reservoir in the middle to lower crust, based on a Moho depth of ~45 km in this area (26). The LVB separates into two zones northwest and southeast of the caldera at mantle depths of ~70 km (Fig. 2F). The orientation of the northwestern portion agrees with the extension of the SRP in a northeast-southwest direction, implying the track of the North American Plate across the Yellowstone plume (2, 10). In cross sections BB′ and CC′, this low-velocity zone that dips ~60° northwestward is consistent with the plume geometry determined in previous studies (810). In contrast, the southeastern low-velocity zone is relatively smaller and localized, terminating at a depth of 100 km (Fig. 3C).

Fig. 2 Depth slices of the Yellowstone tomographic P-wave model.

(A to E) Crustal velocity structures; (F) upper-mantle velocity structure. Black solid and dashed lines outline the 0.64 Ma caldera and resurgent domes. The green line is the northeast end of the Eastern Snake River Plain. White lines denote the 5% and 7% P-wave velocity reduction contours. Poorly resolved areas are shaded according to the index of resolvability, R, converted from the checkerboard test results (25).

Fig. 3 Cross sections of the Yellowstone tomographic P-wave model.

(A to C) The crustal magmatic reservoirs and the mantle plume are demonstrated in northeast-southwest (A) and northwest-southeast [(B) and (C)] directions, respectively. Map-view locations of the cross sections are shown in Fig. 1. White lines denote the 5% and 7% P-wave velocity reduction contours. Poorly resolved areas are shaded according to the index of resolvability, R, converted from the checkerboard test results (25).

The VP reduction of >5% within the imaged crustal LVBs is difficult to explain by temperature and composition alone and implies the presence of melts (25). We assume a 5% VP reduction as being diagnostic of partial melt (3) to quantitatively estimate the volume of crustal melt. Weaker anomalies may also be partly explained by partial melt, but tradeoffs with temperature or composition variations and tomographic resolution prevent these weaker anomalies from being robustly interpreted as melt. Using this conservative proposition results in volume estimates of ~46,000 km3 for the lower-crustal LVB and ~10,000 km3 for the upper-crustal LVB. These estimates agree well with the sizes and depths of the basaltic and rhyolitic magma reservoirs interpreted by geochemical studies (6, 7, 27). The melt fraction of the upper-crustal LVB has been previously estimated to range from 5 to 32% (2, 3, 13). With an average VP of 5.21 km/s calculated over the volume of the upper-crustal LVB, we estimate a melt fraction of ~9% (25), based on a velocity-melt fraction relation derived for the Yellowstone granite-rhyolite-melt system (13). For the lower-crustal LVB, we assume similar elastic properties between the lower crust and the uppermost mantle and use previously proposed partial derivatives of VP with respect to melt fraction for a peridotite-basalt-melt system (table S2). Given the calculated average VP reduction of 6.56%, a ~2% fraction of basaltic partial melts is preferred (25). Multiplying the melt fraction of each LVB by its volume gives ~900 km3 of rhyolitic melts and also ~900 km3 of basaltic partial melts. These estimates provide an overall volume estimate that is comparable to the explosive material volumes of the last three Yellowstone giant eruptions at 2.1 Ma (2500 km3), 1.3 Ma (280 km3), and 0.64 Ma (1000 km3) (1). Although lower-crustal basaltic melts are not expected to contribute to the caldera-forming eruptions, and the upper-crustal melts are unlikely to erupt at one time, both melt volumes can feed smaller eruptions. Assuming a CO2 degassing rate of 4.5 × 107 kg per day, 50% of which comes from subsurface magma discharge (28), the addition of a basaltic lower-crustal reservoir can provide a sufficient influx (25) for the reported ~15,000-year history of the intensive hydrothermal degassing system (29).

Our seismic images depict characteristics of the entire Yellowstone magmatic system from the upper mantle to the crust (Fig. 4) in which the west-northwest–dipping plume is the magmatic source that generates the mafic/basaltic partial melts that intrude into the lower crust, fractionate, and melt the crust to produce more silicic magma, and then intermittently ascend to shallower depths to form the dominantly rhyolitic reservoir at depths of 4 to 14 km beneath the Yellowstone caldera. Because volcanic sills act as traps that accumulate upward-migrating magmatic fluids to form a magma reservoir (2, 3, 6, 3032), the two large LVBs observed in our model suggest the presence of two sill complexes in the upper and lower crust that are likely linked by dikes. This layered structure of basaltic intrusions was also suggested for the volcanic crustal structure in the nearby eastern Snake River Plain (33). This model may thus be representative of other bimodal basaltic-rhyolitic volcanoes around the world. The estimates of volume and geometry of crustal magma reservoirs may also be critical for realistic modeling magmatic system dynamics (14) that in turn could provide further information for volcanic hazard models.

Fig. 4 Schematic model for the Yellowstone crust–upper mantle magmatic system.

The orientation of the model is along the cross section AA′ in Fig. 3. The geometry of the upper- and lower-crustal magma reservoirs is based on the contour of 5% VP reduction in the tomographic model. The dashed outline of the lower-crustal magma reservoir indicates the larger uncertainties in its boundaries relative to that of the upper reservoir (25). The white arrow indicates the North American plate motion of 2.35 cm/year.

Supplementary Materials

www.sciencemag.org/content/348/6236/773/suppl/DC1

Materials and Methods

Figs. S1 to S7

Tables S1 to S3

References (3448)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: Data were collected from the Yellowstone seismograph network operated by the University of Utah, the NSF-funded EarthScope project, and a temporary seismic array operated by Stanford University. All waveform data used in this project can be obtained through the IRIS Data Management Center. Supported by the University of Utah, the University of Utah Seismograph Stations, NSF grants CyberSEES-1442665, EAR-1252191, and EAR-03233309 (in support of the EarthScope Transportable Array), and the Brinson Foundation and Carrico funds.
View Abstract

Navigate This Article