Reports

Inflation-predictable behavior and co-eruption deformation at Axial Seamount

See allHide authors and affiliations

Science  16 Dec 2016:
Vol. 354, Issue 6318, pp. 1399-1403
DOI: 10.1126/science.aah4666

Volcano monitoring goes into the deep

Axial Seamount is a large and active submarine volcano along the Juan de Fuca midocean ridge off the coast of the western United States. Eruptions in 1998 and 2011 were followed by periods of magma recharge, making it an ideal location to include in the Ocean Observatories Initiative Cabled Array. Wilcock et al. present real-time seismic data from the most recent eruption in April 2015 that allow the tracking of magma before and during eruption. Nooner and Chadwick show that eruptions are predictable on the basis of deformation data. As magma pools underneath it, Axial Seamount inflates and erupts when the inflation hits a threshold. Both studies elucidate the dynamics of submarine volcanoes, which vastly outnumber their aboveground counterparts.

Science, this issue p. 1395, p. 1399

Abstract

Deformation of the ground surface at active volcanoes provides information about magma movements at depth. Improved seafloor deformation measurements between 2011 and 2015 documented a fourfold increase in magma supply and confirmed that Axial Seamount’s eruptive behavior is inflation-predictable, probably triggered by a critical level of magmatic pressure. A 2015 eruption was successfully forecast on the basis of this deformation pattern and marked the first time that deflation and tilt were captured in real time by a new seafloor cabled observatory, revealing the timing, location, and volume of eruption-related magma movements. Improved modeling of the deformation suggests a steeply dipping prolate-spheroid pressure source beneath the eastern caldera that is consistent with the location of the zone of highest melt within the subcaldera magma reservoir determined from multichannel seismic results.

Successful volcanic eruption forecasting is traditionally based on short-term (minutes to hours) increases in seismicity, surface deformation, or both during the time that magma is already moving toward the surface (1, 2). Successful forecasts made days to weeks in advance are much rarer because the patterns of geophysical signals are generally not clear or repeatable enough. However, some notable successes at volcanoes such as Mount St. Helens and in Iceland have been documented (3, 4). Seven months in advance of an April 2015 eruption at Axial Seamount, we made a successful forecast that it would occur within a 15-month time window, on the basis of long-term deformation monitoring. The deformation measured during the 2015 eruption also provides important constraints on the location and depth of magma reservoirs and conduits.

Axial Seamount is a heavily instrumented submarine volcano that is part of the Ocean Observatories Initiative (OOI) Cabled Array (5, 6). Axial Seamount is distinguished from other submarine volcanoes in that it has a long-term volcano deformation time series that spans three eruptions. A combination of bottom pressure recorders (BPRs) and mobile pressure recorders (MPRs) (711) provided measurements of vertical deformation from 2000 to 2015. Both methods use changes in the overlying water pressure to detect vertical displacements of the seafloor with a resolution of ~1 cm. BPRs record continuously to capture sudden events (such as eruptions) over minutes to days, but these instruments are not ideal for longer-term measurements because of sensor drift. MPR campaign-style surveys require a remotely operated vehicle (ROV) to deploy the instrument at seafloor benchmarks (fig. S1) and, after correcting for sensor drift, can document gradual deformation over months to years (9, 10). MPR data can constrain BPR drift where the two are colocated. Autonomous, battery-powered BPRs have been used at Axial since the mid-1980s (12, 13) and, in September 2014, the OOI Cabled Array began providing real-time data from three BPR-tilt instruments (5, 6). At the time of the 2015 eruption, three autonomous BPRs, three cabled BPRs, and 10 MPR benchmarks were deployed at Axial (Fig. 1 and fig. S2).

Fig. 1 Map of the summit caldera of Axial Seamount.

Locations of MPR benchmarks (white circles) and BPR instruments (red and blue circles) are indicated. An additional benchmark located 10 km south of the caldera center is not visible. Numbers show vertical displacements in centimeters at each of the MPR benchmarks between 14 September 2013 and 25 August 2015, a period that included pre-eruption inflation, co-eruption deflation, and post-eruption inflation. Numbers in parentheses show subsidence in centimeters during deflation only, as measured by the BPRs. BPRs on the OOI Cabled Array (red dots) include tiltmeters (data shown in Fig. 3). The map also shows locations of 2015 lava flows and eruptive fissures [white outlines and red lines, respectively (24)] and 2011 lava flows and eruptive fissures [gray outlines and yellow lines, respectively (26)]. The + symbol denotes the centroid of the best-fit prolate-spheroid deformation model (Fig. 4). AXID, Axial International District.

After the 2011 eruption at Axial Seamount (14), a time- or inflation-predictable model was proposed in which the volcano erupts at or near a threshold level of inflation (15, 16). The 2015 eruption provided a test for this model and its usefulness in forecasting. The average linear rate of inflation measured at the caldera center between 2000 and 2010 was 15 ± 0.2 cm/year (Fig. 2), with higher rates measured in the months after the 1998 eruption and before the eruption in 2011 (14). After 2011, we initially expected the next eruption to occur in 2018 if the pattern of deformation repeated itself (14). However, from continued monitoring we found that the rate of inflation increased substantially after the 2011 eruption. We observed that the average inflation rate at the caldera center was 61 ± 1.4 cm/year between August 2011 and September 2013 (Fig. 2). A marked increase in the magma supply may explain the fourfold increase over the 2000–2010 inter-eruption rate (11). Continuation of this higher rate of inflation during 2013–2014 was observed after a Monterey Bay Aquarium Research Institute autonomous underwater vehicle (AUV) collected repeat high-resolution bathymetry (17) and with data recovered from a prototype self-calibrating BPR (18) in August 2014. Thus, in September 2014, we revised our forecast that Axial would erupt sometime during 2015 (19, 20). The expected eruption began on 24 April 2015, detected in real time by the OOI Cabled Array (21).

Fig. 2 Deformation time series at the caldera center.

Long-term time series of inflation and deflation at the center of the caldera at Axial Seamount (to 19 May 2016). Purple dots represent MPR measurements (error bars indicate 1 SD); blue curves show BPR data (drift-corrected after 2000). The relative depth of data before and after the 1998–2000 gap in measurements is unknown.

This long-term forecast was unusually successful for any volcano (1, 2). The level of inflation as the 2015 eruption began was only 30 cm higher than in 2011 (Fig. 2). This observation supports the model of a pressure threshold in the shallow magma reservoir above which diking events are triggered, but it also suggests that the threshold may increase with time because of accumulating tectonic stress, as observed in Iceland and Ethiopia (22, 23). Nevertheless, the volcanic system at Axial may be unusually repeatable due to the continuous magma supply and the thin ocean crust in a mid-ocean ridge setting. The increase in magma supply rate documented by the inflation data led to a marked decrease in the eruption recurrence interval (1998–2011 versus 2011–2015), as well as a change in lava composition erupted at the summit (24).

Pre-eruption inflation changed abruptly to co-eruption deflation at ~06:00 on 24 April 2015 (all times GMT), more than an hour after seismicity began to increase (21), due to magma intruding out of the summit reservoir (fig. S4). The 2015 eruption extended >20 km from the northeastern edge of the caldera and along the north rift zone (24), in contrast to the last two eruptions, which were on the south rift (25, 26). The total subsidence at the caldera center was similar in 2015 and 2011 (–2.45 and –2.43 m, respectively), which suggests that roughly the same volume of magma was removed from the magma reservoir. However, the rate of deflation was noticeably higher in 2015 than in 2011. For example, the total subsidence in the first 24 hours amounted to –2.22 m in 2015 (91% of the total), compared with only –1.57 m in 2011 (65% of the total). In 2015, the rate of deflation decreased quasi-exponentially until 5 May and was followed by a transitional period of alternating minor inflation and deflation until 19 May, when rapid re-inflation resumed (fig. S5). Thus, the deflation lasted much longer in 2015: 25 days compared with only 6 days in 2011. Notably, this is similar to the duration of the impulsive seismoacoustic events detected on the north rift zone (21) that could be interpreted as steam explosions during lava flow emplacement (24).

We captured co-eruption deformation during the submarine dike intrusion and eruption with in situ tiltmeters at two sites on the OOI Cabled Array (11). The tilt signals can be closely related to the seismicity generated by the initial dike intrusion (21). Tilt magnitudes and directions began to change at 05:25 (Fig. 3), soon after the seismic crisis began. At this time, most of the earthquakes were located beneath the northeastern edge of the caldera (21), where the dike intrusion initiated and the southernmost eruptive fissures are located (24). Between 06:00 and 08:00, the earthquakes propagated 3 to 4 km southward along the eastern edge of the caldera. The tilt signals changed substantially (Fig. 3) during this time period in a way that is consistent with a dike intruding southward but not reaching the surface in this area (21).

Fig. 3 Co-eruption tilt.

Results show east-west (blue) and north-south (red) components and net tilt direction (green) from OOI Cabled Array instruments at stations AXCC (A and B) and AXEC (C and D), shown in Fig. 1. Tilts at station AXCC are consistent with north-to-south propagation of a dike along the eastern edge of the caldera, with the dike tip passing by the station at 07:10 on 24 April. Tilts on 25 to 27 April mainly reflect deflation during the eruption, which lasted until 19 May.

We observed >100 microradians (μrad) of downward tilt toward the south in ~1 hour, which then abruptly reversed at 07:10 (Fig. 3A) in the central caldera tiltmeter [Axial caldera center (AXCC)] record. This behavior is consistent with modeling of lateral dike propagation past the instrument (27), because the tilt component parallel to the direction of dike propagation (north-south at Axial) is most sensitive to the lengthening of the dike. The initial tilt is in the direction of dike propagation, and the reversal in tilt direction occurs when the dike tip passes the tiltmeter. At the same time, the east-west component of the AXCC tiltmeter also reversed direction from slow eastward tilt (toward the dike axis) to a rapid westward tilt (away from the dike axis) (Fig. 3, A and B). Dike modeling (27) shows that the tilt component perpendicular to the dike axis (east-west at Axial) is most sensitive to the depth to the top of the dike, and the initial tilt is downward toward the dike axis, which reverses to tilt away from the dike axis when the top of the dike nears the surface. This tilt record suggests that the dike propagated to the south for 3 to 4 km before stalling out and continuing to the north along the north rift zone.

We observed several large and nearly instantaneous easterly jumps in tilt magnitude between 06:13 and 07:10 from the eastern caldera tiltmeter [Axial eastern caldera (AXEC)] located closer to the dike axis. By 07:10, the net tilt amounted to >1000 μrad before it reversed and started gradually decreasing (Fig. 3, C and D). In contrast to the AXCC tiltmeter, which recorded smoothly varying tilt signals, we interpreted the large sudden offsets in the AXEC tilt record as either inelastic deformation (perhaps cracking or faulting near the dike axis) or slight movements of the instrument caused by earthquake shaking during the peak of the seismic swarm. This makes the AXEC tilt signals more difficult for us to interpret in the context of dike models. However, after 08:00, both tilt records were again smooth and became dominated by deformation due to the ongoing deflation rather than from the dike intrusion (fig. S6).

We used the MPR results from September 2013 to August 2015 for the net vertical displacement at seafloor benchmarks in and near the summit caldera to model the source of the inflation and deflation signals (Fig. 1). This time period includes some pre-eruption inflation, the co-eruption deflation, and some post-eruption re-inflation (Fig. 2). The 10 MPR stations in 2015, compared with only 6 in 2011, provided better constraints for ground deformation models. We fit the MPR data from all 10 MPR stations to a suite of models, including a point-source (28), a penny-shaped sill (29), and a prolate spheroid (30) (figs. S7 to S9 and table S1). The best-fitting source for the 2013–2015 time period Embedded Image is a prolate spheroid with the major axis dipping at 77° in the direction of 286°, with major and minor axes of 2.2 and 0.38 km, respectively, and a depth to center of 3.81 km (Fig. 4 and fig. S8). The fit of the observed data to this model is much better than the fit to previous sill or point-source models (9, 10, 14).

Fig. 4 Deformation modeling results.

(A) Map view of vertical displacements expected from the best-fit prolate-spheroid model (color contours and blue vectors) and comparison with the 2013–2015 MPR data (red vectors). (B) North-south and (C) east-west depth profiles show orientation of the steeply dipping prolate-spheroid model. Yellow circles in (A) to (C) show the surface projection of the prolate-spheroid centroid. (D) Plot of vertical displacement versus radial distance (R) from the model centroid, comparing data (red crosses; data are mean ± SD) and model calculations (blue squares).

The best-fitting deformation source for the inter-eruption period between the previous MPR surveys (an inflation-only period with data from six stations from July 2011 to September 2013) is very similar, a steeply dipping prolate spheroid with the major axis dipping at 75° in the direction of 290°, major and minor axes of 2.2 and 0.33 km, and a depth of 3.77 km (fig. S7). We suggest that the source of the deformation is the same for time periods dominated by inflation or deflation. The location of the 2013–2015 source is east-southeast of the caldera center, but its steep dip to the west-northwest causes the maximum uplift or subsidence to be observed near the caldera center (Fig. 4). The prolate spheroid shape approximates a nearly vertical conduit, and the location of its top (at 1.6 km depth beneath the eastern edge of the caldera) is almost the same as the shallowest part of the magma body imaged by multichannel seismic (MCS) surveys (31). The southern half of the caldera was also where the MCS data showed the highest percentage of melt, interpreted as the locus of magma supply from the underlying hot spot (31). The eastern edge of the caldera was the source area of the dikes that fed the 2015 eruption (21, 24), as well as the two previous eruptions (25, 26). The best-fitting deformation model does not necessarily show the geometry of the entire magma body, which is approximated by a horizontal ellipsoid underlying the caldera from MCS data. The deformation model instead shows where the greatest volume change occurred during inflation and deflation. Inflation and deflation at Axial Seamount appears to be concentrated in a nearly vertical conduit that feeds the high-melt core of the magma body imaged by Arnulf et al. (31), and the surrounding mush zones of the magma body have less of an influence on surface deformation.

To calculate the volume of magma removed from the subcaldera reservoir during the 2015 eruption, we used the prolate spheroid source parameters obtained from modeling the 2013–2015 MPR data to solve for the volume change due to the co-eruption subsidence only (11), as recorded by BPR instruments at five locations (fig. S9 and table S1). The volume change was 2.88 × 108 m3, which is 1.95 times the volume of lava erupted in 2015 (24). This implies that 1.40 × 108 m3 of magma was in the dike that intruded along the north rift zone. The length of the dike was ~24 km from seismicity and the location of 2015 eruption sites (21, 24), so with an average dike height of 2 to 3 km, as was assumed for the previous eruptions at Axial (14), the required dike thickness would be 2 to 3 m to match the estimated dike volume. This is much greater than the 1-m dike thickness from similar calculations for the 1998 and 2011 eruptions. A thick dike could explain the high deflation rate and the longer duration in 2015 compared with 2011. Axial did not have an eruption on its north rift zone in decades or more, whereas at least the last three eruptions have been on the south rift zone (32). Larger accumulated extensional stress on the north rift may have resulted in the intrusion of a wider dike that accommodated a higher rate of magma transport from the summit reservoir and took longer to solidify (22, 33, 34). Inflation resumed as soon as co-eruption deflation stopped (Fig. 1), with 33 cm of uplift at the caldera center from the 25 August 2015 MPR survey and another 80 cm as of 19 May 2016 from OOI data, indicating that more than one-third of the magma volume lost in 2015 was recovered in the first year (table S1).

We successfully forecast the 2015 eruption of Axial Seamount from seafloor deformation data and captured the eruption in real time from tilt and pressure instruments on the OOI Cabled Array. The tilt data showed that the dike initially propagated 3 to 4 km south before stalling out and continuing to the north, consistent with the seismic data (21). Improved modeling of the magma source for the eruption revealed a steeply dipping magma conduit beneath the eastern caldera, in agreement with the location of a high-melt zone from recent seismic results (31). Uplift rates have gradually decreased with time since the 2015 eruption, so it is unclear whether the higher magma supply rate evident from 2011–2015 will continue. If reinflation continues at ~60 cm/year, the next dike intrusion could occur as early as 2019 but will occur later if the inflation rate slows or if higher tectonic stresses from previous dike intrusions need to be overcome (22). Variations in the magma supply rate complicate forecasts but may be at least partially explained by a deeper magma body hydraulically connected to the shallow magma reservoir (35). These complexities could be overcome using a generalized time-predictable model after several more eruption cycles have been observed (16). However, for the first time, we will be able to continuously update the next eruption and/or intrusion forecast for Axial Seamount with real-time data from the OOI Cabled Array.

Supplementary Materials

www.sciencemag.org/content/354/6318/1399/suppl/DC1

Methods

Supplementary Text

Figs. S1 to S9

Table S1

References

References and Notes

  1. Methods, supplementary text, figures, and a table are available as supplementary materials on Science Online.
Acknowledgments: The BPR and tilt data presented here are archived at the Integrated Earth Data Applications Marine Geoscience Data System (12, 13) and at the OOI Data Portal (6). This work was supported by NSF awards OCE-1356216 and 1546616 and by the National Oceanic and Atmospheric Administration, Pacific Marine Environmental Laboratory (NOAA-PMEL), Earth-Ocean Interactions Program. We thank M. Fowler and the crews of R/V Thomas G. Thompson, ROV Jason, and AUV Sentry for logistical support at sea during expedition TN327. This work would not have been possible without the support of the NOAA-PMEL Engineering Division and the University of Washington OOI Cabled Array team, led by J. Delaney and D. Kelley. The paper benefited from helpful input from two anonymous reviewers. This is PMEL contribution number 4509.
View Abstract

Navigate This Article