plugged with degassed magma. The eruption was
likely fed by the rapid aseismic rise of fresh magma
over a short period of time. During ascent, the
magma decompressed and degassed, triggering
explosive fragmentation. A high-velocity flow of
particles and gas eroded the conduit walls and
vent, producing tremor and driving a jet of ash
and gas into the atmosphere. Tremor amplitudes
strongly correlated with plume height during this
waxing phase of eruption (stage 1), while extensive vent erosion was occurring. Once the vent was
cleared, tremor amplitudes leveled off (stage 2)
and then decreased more rapidly (stage 3) as
flow and particle impacts against the vent walls
diminished because the shallow conduit had
been cleared and widened.
Multiple lines of evidence support our model.
Before and after images from the eruption show
extensive erosion of Pavlof’s summit (fig. S2). After the eruption, the summit crater was ~130 m
wide and 110 m deep, whereas there was no well-defined crater before the eruption. The crater
narrows to a 30- to 40-m-diameter circular vent
at the base. Using scaling relationships and
plume height observations, we estimated the
evolving vent radius during the first two stages
of the eruption (Fig. 3C) ( 13). These calculations
indicate that the vent widened rapidly at the
beginning of the eruption and then stabilized,
following a similar trend as the tremor amplitudes. We estimated a maximum vent diameter
of between 30 and 35 m, similar to that observed
after the eruption. We infer that the low-frequency
seismic event originating near the summit at 16: 45
signifies crater collapse or rapid widening, because this event corresponds to the inflection
point in the hysteresis pattern (Fig. 4). Sudden
vent widening induces magma fragmentation
at shallower depths ( 25), potentially leading to
enhanced bubble growth and ash production ( 26).
Indeed, a major increase in volcanic lightning
occurred around this time (Fig. 3C), supporting
the idea of increased fine ash content in the
volcanic plume ( 27). In stage 3, tremor ampli-
tudes continued to decline, likely because the
crater had sufficiently widened so that the erupt-
ing jet was no longer strongly coupled to the
vent walls, thereby leading to a different rela-
tionship between tremor and plume height. Our
proposed vent evolution is consistent with labor-
atory experiments of vent erosion, which show a
rapid widening followed by stabilization and oc-
casional sudden expansion ( 28).
Our findings are also consistent with the generic three-stage tremor model proposed by McNutt
and Nishimura ( 6). Their analysis of tremor and
conduit and vent features from 24 eruptions
worldwide revealed that these stages represent
the most common temporal evolution of eruption tremor. They found that a gradual increase
in tremor amplitude is common for the first
eruption in a sequence of eruptions, perhaps because of the gradual breaking up and erosion of
the conduit. After a leveling-off phase, an exponential decrease in tremor amplitude during the
waning stage was observed in 92% of the eruptions that they examined.
Comparison of the seismic and infrasonic
tremor from Pavlof Volcano reveals a high similarity between amplitudes and temporal evolution, suggesting a linked seismo-acoustic source
located within the shallow conduit or crater.
These observations, along with the tremor and
plume height hysteresis, guide our multistage
conceptual model, which is consistent with those
from fluvial seismology and eruption observations
from around the world. Future monitoring of
volcanic eruptions and interpretations of vol-
canic tremor, including its relationship to plume
height and eruption intensity, should take into
account the stage of the eruption and state of the
upper conduit and vent, as well as insight from
other studies of flows and the vibrations that
REFERENCES AND NOTES
1. S. De Angelis, D. Fee, M. Haney, D. Schneider, Geophys. Res.
Lett. 39, L21312 (2012).
2. C. F. Waythomas et al., J. Geophys. Res. Solid Earth 115,
3. D. Fee, R. S. Matoza, J. Volcanol. Geotherm. Res. 249, 123–139
4. B. A. Chouet, R. S. Matoza, J. Volcanol. Geotherm. Res. 252,
5. S. G. Prejean, E. E. Brodsky, J. Geophys. Res. Solid Earth 116,
6. S. R. McNutt, T. Nishimura, J. Volcanol. Geotherm. Res. 178,
7. A. M. Jellinek, D. Bercovici, Nature 470, 522–525 (2011).
8. R. S. Matoza et al., Geophys. Res. Lett. 36, L08303 (2009).
9. H. Kanamori, J. Mori, D. G. Harkrider, J. Geophys. Res.
Solid Earth 99, 21947–21961 (1994).
10. S. R. McNutt, Acta Vulcanol. 5, 193 (1994).
11. D. Fee, M. Garces, A. Steffke, J. Volcanol. Geotherm. Res. 193,
12. M. Ripepe et al., Earth Planet. Sci. Lett. 366, 112–121 (2013).
13. Materials and methods are available as supplementary
14. R. S. J. Sparks, M. I. Burski, S. N. Carey, J. S. Gilbert,
L. S. Glaze, H. Sigurdsson, A. W. Woods, Volcanic Plumes
(John Wiley and Sons, 1997).
15. L. G. Mastin et al., J. Volcanol. Geotherm. Res. 186, 10–21
16. M. Ichihara, M. Takeo, A. Yokoo, J. Oikawa, T. Ohminato,
Geophys. Res. Lett. 39, L04304 (2012).
17. R. S. Matoza, D. Fee, Geophys. Res. Lett. 41, 1964–1970
18. K. T. Walker et al., J. Geophys. Res. Solid Earth 115, B12329
19. A. Burtin, L. Bollinger, J. Vergne, R. Cattin, J. Nábělek,
J. Geophys. Res. Solid Earth 113, B05301 (2008).
20. L. Hsu, N. J. Finnegan, E. E. Brodsky, Geophys. Res. Lett. 38,
21. D. L. Roth et al., Earth Planet. Sci. Lett. 404, 144–153 (2014).
22. V. C. Tsai, B. Minchew, M. P. Lamb, J. P. Ampuero, Geophys.
Res. Lett. 39, L02404 (2012).
23. F. Gimbert, V. C. Tsai, M. P. Lamb, J. Geophys. Res. Earth Surf.
119, 2209 (2014).
24. T. C. Bartholomaus et al., Geophys. Res. Lett. 42, 6391–6398
25. L. G. Mastin, Geochem. Geophys. Geosyst. 3, 1– 18 (2002).
26. C. Klug, K. V. Cashman, Bull. Volcanol. 58, 87–100 (1996).
27. C. Cimarelli, M. Alatorre-Ibargüengoitia, U. Kueppers, B. Scheu,
D. B. Dingwell, Geology 42, 79–82 (2014).
28. S. A. Solovitz, D. E. Ogden, D. Kim, S. Y. Kim, J. Geophys. Res.
Solid Earth 119, 5342–5355 (2014).
We thank P. Dawson and V. Tsai for early reviews of the
manuscript; C. Szuberla for discussions on signal processing;
L. Mastin for discussions on vent widening and fragmentation;
R. Wessels and M. Kaufman for imagery of Pavlof Volcano before and
after the eruption; H. Schwaiger for the atmospheric modeling;
and the reviewers for their helpful comments. Seismic and
EarthScope TA data are available from the Incorporated Research
Institutions for Seismology Data Center ( www.iris.edu). Observations
of volcanic activity were made by AVO and are detailed on its
website ( www.avo.alaska.edu). Webcam data are available from the
FAA Aviation Weather Cameras website ( http://avcams.faa.gov).
DLL infrasound data are available from D.F. and the Wilson Alaska
Technical Center. The World Wide Lightning Location Network
( http://wwlln.net) provided the lightning data. The authors
acknowledge support from AVO and NSF grants EAR-1331084,
EAR-1614323, and EAR-1614855. A.R.V.E. acknowledges a U.S.
Geological Survey Mendenhall Fellowship.
Materials and Methods
Figs. S1 and S2
References ( 29–37)
21 July 2016; accepted 7 December 2016
48 6 JANUARY 2017 • VOL 355 ISSUE 6320 sciencemag.org SCIENCE
Fig. 4. Tremor amplitudes and plume height as a function of time. (A) Seismic and (B) infrasound
amplitudes are plotted against the corresponding plume height and are colored as a function of time. Two
distinct relationships are evident and correspond to the waxing and waning portions of the eruption and
changing source conditions. The clockwise hysteresis is indicated by the two arrows. Error bars for the
satellite measurements of plume height were determined using different pixel averaging techniques.