Lava lake level as a gauge of magma reservoir pressure and eruptive hazard Matthew R. Patrick, Kyle R. Anderson, Michael P. Poland*, Tim R. Orr, and Donald A. Swanson U.S. Geological Survey, Hawaiian Volcano Observatory, P.O. Box 51, Hawai‘i National Park, Hawaii 96718, USA
INTRODUCTION Magma reservoir pressure is a primary variable in physical models of volcanic systems and forecasts of eruptive activity (Segall, 2013). While ground deformation is typically used as a proxy, quantitative constraints on reservoir pressure have traditionally been elusive, as it is not normally possible to separate changes in pressure from changes in reservoir volume using deformation data alone (Dzurisin, 2007). This uncertainty constitutes a significant barrier to the implementation of physically realistic models of active volcanoes and their attendant hazards. Recent eruptive activity at Kīlauea (Hawai‘i, USA), however, offers an opportunity to address this challenge at one of the most active volcanoes in the world. Volcanism at Kīlauea is driven by magma that ascends from the mantle and accumulates in a reservoir system, from which it may rise and erupt at the summit or flow laterally at shallow depths to intrude or erupt along one of the volcano’s rift zones (Figs. 1A and 1B) (Eaton and Murata, 1960). Since 2008, eruptive activity within Halema‘uma‘u Crater at the summit (Patrick et al., 2013) has been concurrent with the long-lived eruption (since 1983) from the Pu‘u ‘Ō‘ō vent area on the volcano’s east rift zone (ERZ) (Orr et al., 2013a), the first time in recorded history (~200 yr) that prolonged simultaneous eruptions have occurred on Kīlauea. The lava lake in Halema‘uma‘u Crater provides a new opportunity to study Kīlauea’s interconnected magmatic system, and it offers the first modern, long-term, continuous record of lava level in an open vent anywhere on Earth. We demonstrate that lava level measurements can be used to quantify magma reservoir pressure change and to forecast potentially hazardous fluctuations in eruption rate on Kīlauea’s ERZ. METHODS Ground tilt was measured by the Uēkahuna (UWE) tiltmeter, 2 km from the lava lake (Fig. DR1 in the GSA Data Repository1) (Cervelli and Miklius, 2003). We use data with a 1 min sampling interval and record tilt along an azimuth of 327°, which is approximately radial to the lava lake and the inferred shallow magma reservoir located near Halema‘uma‘u *Current address: U.S. Geological Survey, Cascades Volcano Observatory, 1300 SE Cardinal Court, Suite 100, Vancouver, Washington 98683, USA. 1 GSA Data Repository item 2015282, details on methodology and additional figures, is available online at www.geosociety.org/pubs/ft2015.htm, or on request from
[email protected] or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.
A Pāhoa
Halema‘uma‘u Crater
Kīlauea Caldera (summit)
st lava lake we uth one o S ft Z i R
t Eas
Pu‘u ‘Ō‘ō East Rift Zone
Rift
e
Zon
19.4 N
Hawai‘i Island
Kalapana Gardens Pu‘u ‘Ō‘ō flow field (1983-present) 0
B
10 km 155 W
155.2 W
Kīlauea Caldera (summit) Halema‘uma‘u Crater lava lake Eas tR
19.3 N
ABSTRACT Forecasting volcanic activity relies fundamentally on tracking magma pressure through the use of proxies, such as ground surface deformation and earthquake rates. Lava lakes at open-vent basaltic volcanoes provide a window into the uppermost magma system for gauging reservoir pressure changes more directly. At Kīlauea Volcano (Hawai‘i, USA) the surface height of the summit lava lake in Halema‘uma‘u Crater fluctuates with surface deformation over short (hours to days) and long (weeks to months) time scales. This correlation implies that the lake behaves as a simple piezometer of the subsurface magma reservoir. Changes in lava level and summit deformation scale with (and shortly precede) changes in eruption rate from Kīlauea’s East Rift Zone, indicating that summit lava level can be used for short-term forecasting of rift zone activity and associated hazards at Kīlauea.
N
Pu‘u ‘Ō‘ō
ift Z
one
conduits summit magma reservoir complex conduit from mantle
C
High summit lava lake level
D Low level: lake drained,140 m drop)
297 °C
500 °C 5 Mar 2011 07:30
10
7 Mar 2011 06:30
10
Figure 1. Overview of Kīlauea Volcano. A: The southeastern portion of the Island of Hawai‘i. Two eruptions are ongoing: (1) the summit eruption with a lava lake in Halema‘uma‘u Crater, and (2) the East Rift Zone eruption at Pu‘u ‘Ō‘ō that has produced a 138 km2 lava flow field (pink area shows flow field mapped on 30 December 2014). Gray lines show roads. B: Cross section showing the simplified layout of Kīlauea’s magmatic system. Not to scale. C, D: Thermal images of Kīlauea’s summit lava lake bracketing the March 2011 Kamoamoa eruptive event on the East Rift Zone (event 1 in Fig. 2D). The lake is ~150 m in diameter.
Crater (Cervelli and Miklius, 2003). GPS measurements of contraction and extension across Kīlauea Caldera were determined from line-length changes between stations UWEV and CRIM, located on opposite sides of the caldera. Line-length changes have less noise than individual station solutions due to the cancellation of shared errors, and show a clearer picture of deformation over limited time periods, but these data are affected by slow, steady motion of Kīlauea’s south flank. For a long-term (multiyear) comparison of GPS and lava level, we chose to minimize the effect of south flank motion by using vertical displacements recorded at station HOVL, on the east rim of Halema‘uma‘u Crater.
GEOLOGY, September 2015; v. 43; no. 9; p. 831–834 | Data Repository item 2015282 | doi:10.1130/G36896.1 | Published online 5 August 2015
© 2015 Geological Society America. permission to copy, contact
[email protected]. GEOLOGY 43 | ofNumber 9 For | Volume | www.gsapubs.org
831
Lava level was manually measured from visual and thermal camera images (Figs. 1C and 1D; see the Data Repository). The measurements, in image pixels, were converted to elevation in meters using sporadic calibration data from laser rangefinder, lidar, and photogrammetry surface models of the vent. For most data the measurements were made once an hour. The abundance of bright pixels in webcam images of the ERZ flow field, taken as a proxy for ERZ effusion rates, was calculated using nightly composite images (Patrick et al., 2010a) from a webcam on high ground observing a large portion of the lava flow field. Other areas of active flows were present outside of the camera field of view (Fig. DR3), but for most of the study period the camera captured the main area of activity. The pixel footprint area was calculated using the viewing geometry of the camera and the range of distances to the active lava (see the Data Repository). RESULTS Kīlauea’s summit undergoes nearly continuous deformation of the ground surface, which has been inferred to reflect changes in magma pressure in the summit reservoir (Eaton and Murata, 1960). Ground tilt, often used as a proxy for reservoir pressure change (Tilling, 1987; Denlinger, 1997), closely follows the Halema‘uma‘u lava lake level over time scales ranging from hours to weeks (Fig. 2A). This coupling is clearly illustrated by characteristic multiday cycles of ground deformation called deflationinflation (DI) events (Cervelli and Miklius, 2003; Anderson et al., 2015). DI events produce remarkably sympathetic changes in lava lake level (Figs. 2A and 2B), spanning amplitudes of ~20 m. Focusing on a longer time period, cross-spectral coherence of lava level and tilt data from late 2011 to mid-2014 indicates that the correlation is strongest (coherence >0.8) over periods of 3–9 days, which may partly reflect the time scale of DI events. We use a 1 week moving window with 75% overlap to perform
linear regression on tilt and lava level over this 2.8 yr period, and find a mean coefficient of determination, r2, of 0.82 ± 0.16 (Fig. DR4). This correlation between the summit lava lake level and summit tilt is, as expected, much stronger than that previously measured between ERZ lava lake level and summit tilt (Tilling, 1987; Denlinger, 1997), as our data reflect a more direct connection within the summit magmatic system. Limiting the regression to the highest r2 values (>0.90) indicates that for every microradian of radial tilt change at summit station UWE, the Halema‘uma‘u lava level changes 5.0 ± 0.8 m (Fig. 2C; Fig. DR5). The cross-correlation of tilt and lava level indicates that any delay is not appreciably longer than the 5 min sampling interval (Fig. DR6). Lava level is also closely correlated with surface deformation over longer time scales (months), as shown by a comparison with GPS measurements of both cross-caldera line lengths and vertical displacement. Several episodes of rising lava level closely track GPS line length (Fig. 2D), with rising lava level culminating in new ERZ eruptive activity that interrupted the ongoing Pu‘u ‘Ō‘ō eruption (Orr et al., 2015a). Increasing rates of small ERZ earthquakes preceding the eruptive events also indicate an increase in magmatic pressure (Thelen and Patrick, 2012). Focusing on the vertical component of the closest GPS station to Halema‘uma‘u from 2010 to mid2014 (Fig. DR7), the overall correlation with lava level has an r2 of 0.78. Because lava level is linearly correlated with fluctuations in ground tilt and GPS, and no significant time delay exists between changes in deformation and lava level, lava level can be used as a piezometer to directly estimate magma reservoir pressure. In this case, p = rgh1, where p is the reservoir pressure, r is the average bulk density of the magma column, g is gravity, and h1 is the positive vertical distance between the lava lake surface and the reservoir. Assuming a bulk density of 1000–2500 kg m–3 for the potentially gas-rich magma column (see the Data Repository; CarERZ eruptive events: 1
950
950
5
Ground tilt 0
960
B
Jan 15
Jan 24 40
r 2=0.94
Frequency
Lava elev., m
Jan 06
940 920 900
−5
0 Tilt, µrad.
5
Feb 02
C
2
800
0 Summit GPS inflation
mean: 5.0 stdev: 0.8
4 6 8 Slope (m µrad-1)
−0.1
deflation
−5
Feb 12
20
0
0.1
Lava lake level
ate infl
deflation
DI event:
te defla
inflation
D
850
60 earthquakes per day
900
900
3
GPS line-length change, m
925
Lava lake elevation, m
Lava lake level
Ground tilt, µrad
Lava lake elevation, m
A
2
E
40 20 0
Oct
Dec 2010
Feb
Apr
Jun Aug 2011
Oct
Dec
Figure 2. Comparison of summit lava lake level with surface deformation over short and long time scales at Kīlauea Volcano. A: Lava level and summit ground tilt over 5 weeks in 2012. Tilt often follows a pattern called a deflation-inflation (DI) cycle. B: Scatterplot for the period shown in A (elev.—elevation; r 2—coefficient of determination; see text). C: During 2011–2014, 1 microradian of tilt corresponded with 5.0 (±0.8) m of lava level change (stdev—standard deviation). D: Long-term lava level and summit GPS line-length change. Three cycles of rising values culminated in eruptive events on the East Rift Zone (ERZ), each associated with drops in summit lava level. E: Increasing pressure prior to the eruptive events is also shown by the increase in small earthquakes along the ERZ.
832 www.gsapubs.org | Volume 43 | Number 9 | GEOLOGY
Q(t) = KcDh(t),
(1)
where Dh is the height difference between the surface of the summit lava lake and the level at which the summit lava lake is in magmastatic equilibrium with the ERZ vent (see the Data Repository for full derivation), and t is time. Kc is a constant of proportionality that can be thought of as representing magma “conductivity” through the ERZ, that is, the volume flux of magma for each meter of height that the lava lake surface exceeds its magmastatic level (in units of m2 s–1; we make no assumptions about the geometry of the ERZ pathway but assume that magma rheology is Newtonian). If we also assume no significant change in magma storage volume at Pu‘u ‘Ō‘ō, we can equate the rate at which magma exits the summit reservoir with the effusion rate at Pu‘u ‘Ō‘ō. We apply this model to ERZ activity during April–June 2012, when lava erupted from the Peace Day vent at Pu‘u ‘Ō‘ō and supplied a tubefed pāhoehoe flow field (Poland, 2014). The unknown Kc is constrained by reconciling mean effusion rate with mean summit lava level above the magmastatic equilibrium height. The time-averaged effusion rate was estimated by differencing TanDEM-X (TerraSAR-X add-on for Digital Elevation Measurement; www.geo-airbusds.com/terrasar-x/) radar-derived flow field topography and correcting for an assumed 25% vesicularity in the lava (Poland, 2014), yielding an average dense-rock discharge rate of 2.0 m3 s–1. The magmastatic equilibrium height can be determined using observations from brief periods when summit pressure drops sufficiently that ERZ eruptive flux goes to zero. Here we use data from 28 May 2012, when the summit lava lake was at ~922 m elevation (Fig. 3A) and the ERZ vent elevation was ~840 m; the difference of ~80 m is presumably due to differences in magma bulk density between the summit and ERZ (see the Data Repository). The lava lake maintained a mean level of 24 m above this baseline during April–June 2012, when flux averaged 2.0 m3 s–1, implying Kc = 0.08. We calculate that time-dependent flux Q(t) varied from 0 to 3.8 m3 s–1 during this time (Fig. 3B). For every 12 m of summit lava level change, or 2.4 microradians of ground tilt, reservoir pressure changed by 0.1–0.3 MPa, which yielded a change in the ERZ eruption rate of 1 m3 s–1. GEOLOGY | Volume 43 | Number 9 | www.gsapubs.org
Lava lake elevation, m
960
A
Observed summit lava level and tilt
950
5
940 930
0
920
ERZ eruption rate, m3s-1
4
Area of active lava, hectares
MODELING ERZ EFFUSION RATE Variations in magma reservoir pressure control, to a large degree, the volume flux of magma (Q) supplying the ongoing ERZ eruption at Pu‘u ‘Ō‘ō. This relationship has been well established observationally (Cervelli and Miklius, 2003; Orr et al., 2015b); summit deflation is broadly associated with a reduction in ERZ effusion rate, and inflation is broadly associated with increased ERZ effusion rate. Continuous tracking of effusion rate, however, has not been possible due to limitations in the measurement techniques (related in part to their campaign-style nature, often performed weekly; Sutton et al., 2003). The summit lava level presents a potential opportunity to fill this monitoring gap. Assuming that ERZ flux is linearly proportional to the difference between summit reservoir pressure and the ERZ magmastatic pressure,
970
UWE tilt, μrad
bone et al., 2013), each meter of lava level change equates to 0.01–0.02 MPa of reservoir pressure change, and each microradian of tilt equates to 0.05–0.12 MPa of pressure change in the reservoir. Pressure changes controlling lava lake surface height are believed to be due to the changing balance of supply and eruption rate to and from Kīlauea’s shallow reservoir, although deep fluctuations in magmatic gas content cannot be entirely discounted. However, not all lava lake level fluctuations reflect changes in magma reservoir pressure. Very short-term (seconds to hours) fluctuations in the lake level at Halema‘uma‘u relate to shallow gas-driven processes such as small explosive events (Orr et al., 2013b) and gas pistoning (Patrick et al., 2010b, 2014; Fig. DR8). These shallow gas-driven changes can be easily distinguished with seismic data (Orr et al., 2013b; Patrick et al., 2014).
−5
B
Modeled ERZ eruption rate
C
Observed ERZ flow field activity proxy
3 2 1 0 6 5 4 3 2 1 0 04/15
04/23
05/02
05/11
05/19
05/28
06/06
06/15
2012 Figure 3. Fluctuations in summit lava level and East Rift Zone (ERZ) flow field activity at Kīlauea Volcano. A: Observed lava level (black) and Uēkahuna (UWE) radial tilt (gray). B: Modeled eruption rate on the ERZ. C: Measured proxy for activity levels on the ERZ flow field, based on the area of active surface flows imaged by a field webcam. This is taken as a rough indicator of eruption rate, and correlates with the modeled ERZ effusion rates (r 2 = 0.31–0.51, for delays of 23.9 ± 6.4 h; see text).
As an independent comparative proxy for ERZ effusion rates, we quantify the area of active pāhoehoe lava in nightly webcam images of the lava flow field by the abundance of pixels above a specified brightness threshold, as higher effusion rates tend to produce larger areas of active lava (Wright et al., 2001; Mattox et al., 1993). This ERZ effusion rate proxy (Fig. 3C) scales with summit lava level and inferred Q, confirming a proportionality between the summit pressure (and lava level and tilt) and ERZ effusion rate. Thus, the fundamental premise of this model is reasonable. DISCUSSION AND CONCLUSIONS During 2010–2011, lava flows invading the Kalapana Gardens subdivision (Fig. 1A) routinely exhibited synchronized fluctuations between apparent effusion rate and summit pressure; summit DI inflation preceded surges of lava and new breakouts from the lava tube that threatened nearby residences (Orr et al., 2015b). Three homes were ultimately destroyed; two were engulfed by surge-fed lava flows. Summit inflation preceded the increase in ERZ lava flow activity by an average of 24 h (13 events, mean = 23.9 h, standard deviation = 6.4 h), thereby providing a practical hazard forecasting tool (Orr et al., 2015b). Hawaiian Volcano Observatory (HVO) geologists used summit tilt and lava level in a qualitative man-
833
ner in 2010–2011 to anticipate these surges and communicate upcoming hazards to emergency managers. The model we present here is the first step in moving from qualitative assessments to more robust quantitative operational forecasts, with the modeled effusion rates shown here having a lead time of ~1 day. Recent activity on Kīlauea reinforces the need for forecasts of effusion rates. The ongoing Pu‘u ‘Ō‘ō eruption produced the 27 June 2014 lava flow that advanced northeastward toward an area not previously affected by the eruption (McCarter, 2014). In late October 2014, the front of the pāhoehoe flow entered the town of Pāhoa (Fig. 1A), stalling a short distance from the main road. Several episodes of stalling corresponded to DI deflation, low summit lava levels, and reduced effusion rates. Stalling events were followed by breakouts upslope of the flow front, associated with summit inflation and high summit lava levels, which widened the flow or altered its direction. Such widening destroyed a home in Pāhoa in early November 2014. It was not possible to estimate real-time effusion rates with our model at that time owing to the lack of constraint on time-averaged effusion rate and the magmastatic equilibrium level. These problems highlight a current limitation of the model that may be addressed with further work. Nevertheless, this ongoing activity illustrates how fluctuations in effusion rate have a direct impact on lava flow hazards and can be tracked and anticipated by monitoring summit lava level. The direct measure of change in reservoir pressure afforded by the lava lake at Kīlauea provides an important constraint for physical models of the magmatic system (Segall, 2013), such as the volume flux model presented here and models of reservoir volume (Anderson et al., 2015). Lava lake level may also provide critical insight at other open-vent mafic volcanoes. With broad similarity to Kīlauea’s 2011 eruptive episodes (Fig. 2D), rising summit lava levels presaged flank eruptions at Stromboli Volcano (Italy; Calvari et al., 2005, 2009), and at Nyiragongo Volcano (Democratic Republic of the Congo; Tazieff, 1977; Burgi et al., 2014), where lethal, fastmoving lava flows were produced. The lava level in open summit vents therefore provides not only quantitative insight for improved volcano models, but also a practical low-technology monitoring tool for anticipating hazardous changes on a volcano’s flank. ACKNOWLEDGMENTS Funding for the thermal cameras was provided by the American Reinvestment and Recovery Act. Lidar control points on lava level were collected by A. LeWinter, D. Finnegan, S. Anderson, and G. Bawden. We thank R. Denlinger, L. Mastin, and D. Carbone for useful input. Reviews by D. Clague, P. Cervelli, D. Dzurisin, and an anonymous reviewer improved the paper. REFERENCES CITED Anderson, K., Poland, M., Johnson, J., and Miklius, A., 2015, Episodic deflationinflation events at Kīlauea Volcano and implications for the shallow magma system, in Carey, R., et al., eds., Hawaiian volcanism: From source to surface: American Geophysical Union Geophysical Monograph 208, p. 229–250, doi:10.1002/9781118872079.ch11. Burgi, P.Y., Darrah, T.H., Tedesco, D., and Eymold, W.K., 2014, Dynamics of the Mount Nyiragongo lava lake: Journal of Geophysical Research, v. 119, p. 4106–4122, doi:10.1002/2013JB010895. Calvari, S., Spampinato, L., Lodato, L., Harris, A.J.L., Patrick, M.R., Dehn, J., Burton, M.R., and Andronico, D., 2005, Chronology and complex volcanic processes during the 2002–2003 flank eruption at Stromboli volcano (Italy) reconstructed from direct observations and surveys with a handheld thermal camera: Journal of Geophysical Research, v. 110, B02201, doi:10.1029/2004JB003129. Calvari, S., Lodato, L., Steffke, A., Cristaldi, A., Harris, A.J.L., Spampinato, L., and Boschi, E., 2009, The 2007 Stromboli eruption: Event chronology and effusion rates using thermal infrared data: Journal of Geophysical Research, v. 115, B04201, doi:10.1029/2009JB006478. Carbone, D., Poland, M.P., Patrick, M.R., and Orr, T.R., 2013, Continuous gravity measurements reveal a low-density lava lake at Kīlauea Volcano, Hawai‘i: Earth and Planetary Science Letters, v. 376, p. 178–185, doi:10.1016/j.epsl .2013.06.024. Cervelli, P.F., and Miklius, A., 2003, The shallow magmatic system of Kīlauea Volcano, in Heliker, C., et al., eds., The Pu‘u ‘Ō‘ō–Kupaianaha eruption of Kīlauea Volcano, Hawai‘i: The First 20 Years: U.S. Geological Survey Professional Paper 1676, p. 149–163.
Denlinger, R.P., 1997, A dynamic balance between magma supply and eruption rate at Kilauea volcano, Hawaii: Journal of Geophysical Research, v. 102, p. 18,091–18,100, doi:10.1029/97JB01071. Dzurisin, D., 2007, Volcano deformation: Geodetic monitoring techniques: Chichester, UK, Praxis, 442 p. Eaton, J.P., and Murata, K.J., 1960, How volcanoes grow: Science, v. 132, p. 925– 938, doi:10.1126/science.132.3432.925. Mattox, T.N., Heliker, C., Kauahikaua, J., and Hon, K., 1993, Development of the 1990 Kalapana flow field, Kilauea Volcano, Hawai`i: Bulletin of Volcanology, v. 55, p. 407–413, doi:10.1007/BF00302000. McCarter, T., 2014, Scientists engage with public during lava flow threat: Eos (Transactions, American Geophysical Union), v. 95, p. 409–410, doi:10 .1002/2014EO450002. Orr, T.R., Heliker, C., and Patrick, M.R., 2013a, The ongoing Pu‘u ‘Ō‘ō eruption of Kīlauea Volcano, Hawai‘i —30 years of eruptive activity: U.S. Geological Survey Fact Sheet 2012-3127, 6 p. Orr, T., Thelen, W.A., Patrick, M.R., Swanson, D.A., and Wilson, D.C., 2013b, Explosive eruptions triggered by rockfalls at Kīlauea volcano, Hawai‘i: Geology, v. 41, p. 207–210, doi:10.1130/G33564.1. Orr, T., Poland, M.P., Patrick, M.R., Thelen, W.A., Sutton, A.J., Elias, T., Thornber, C.R., Parcheta, C., and Wooten, K.M., 2015a, Kīlauea’s 5–9 March 2011 Kamoamoa fissure eruption and its relation to 30+ years of activity from Pu‘u ‘Ō‘ō, in Carey, R., et al., eds., Hawaiian volcanism: From source to surface: American Geophysical Union Geophysical Monograph 208, p. 393–420, doi:10.1002/9781118872079.ch18. Orr, T., Bleacher, J., Patrick, M., and Wooten, K., 2015b, A sinuous tumulus over an active lava tube at Kīlauea Volcano: Evolution, analogs, and hazard forecasts: Journal of Volcanology and Geothermal Research, v. 291, p. 35–48, doi: 10.1016/j.jvolgeores.2014.12.002. Patrick, M.R., Kauahikaua, J.P., and Antolik, L., 2010a, MATLAB tools for improved characterization and quantification of volcanic incandescence in webcam imagery: Applications at Kīlauea Volcano, Hawai‘i: U.S. Geological Survey Techniques and Methods 13-A1, 16 p. Patrick, M.R., Orr, T., Wilson, D., Elias, T., Sutton, A.J., Fee, D., and Nadeau, P., 2010b, Evidence for gas accumulation beneath the surface crust driving cyclic rise and fall of the lava surface at Halema‘uma‘u, Kilauea Volcano: American Geophysical Union, fall meeting, abs. V21C-2339. Patrick, M.R., Orr, T., Sutton, A.J., Elias, T., and Swanson, D., 2013, The first five years of Kīlauea’s summit eruption in Halema‘uma‘u Crater, 2008–2013: U.S. Geological Survey Fact Sheet 2013-3116, 4 p. Patrick, M.R., Orr, T., Antolik, L., Lee, L., and Kamibayashi, K., 2014, Continuous monitoring of Hawaiian volcanoes with thermal cameras: Journal of Applied Volcanology, v. 3, 19 p., doi:10.1186/2191-5040-3-1. Poland, M.P., 2014, Time-averaged discharge rate of subaerial lava at Kīlauea Volcano, Hawai‘i, measured from TanDEM-X interferometry: Implications for magma supply and storage during 2011–2013: Journal of Geophysical Research, v. 119, p. 5464–5481, doi:10.1002/2014JB011132. Segall, P., 2013, Volcano deformation and eruption forecasting, in Pyle, D.M., et al., eds., Remote sensing of volcanoes and volcanic processes: Integrating observations and modelling: Geological Society of London Special Publication 380, p. 85–106, doi:10.1144/SP380.4. Sutton, A.J., Elias, T., and Kauahikaua, J., 2003, Lava effusion rates for the Pu‘u ‘Ō‘ō–Kupaianaha eruption derived from SO2 emissions and very low frequency (VLF) measurements, in Heliker, C., et al., eds., The Pu‘u ‘Ō‘ō – Kupaianaha eruption of Kīlauea Volcano, Hawai‘i: The first 20 years: U.S. Geological Survey Professional Paper 1676, p. 137–148. Tazieff, H., 1977, An exceptional eruption: Mt. Niragongo, Jan. 10th, 1977: Bulletin of Volcanology, v. 40, p. 189–200, doi:10.1007/BF02596999. Thelen, W., and Patrick, M.R., 2012, A conceptual model for recent seismicity on Kilauea’s upper east rift zone: Proceedings, American Geophysical Union Chapman Conference, Hawaiian Volcanoes: From Source to Surface: Waikoloa, Hawaii, 20–24 August 2012, abs. TU-35. Tilling, R.I., 1987, Fluctuations in surface height of active lava lakes during 1972–1974 Mauna Ulu eruption, Kilauea Volcano, Hawaii: Journal of Geophysical Research, v. 92, p. 13,721–13,730, doi:10.1029/JB092iB13p13721. Wright, R., Blake, S., Harris, A.J.L., and Rothery, D.A., 2001, A simple explanation for the space-based calculation of lava eruption rates: Earth and Planetary Science Letters, v. 192, p. 223–233, doi:10.1016/S0012-821X(01)00443-5.
Manuscript received 14 April 2015 Revised manuscript received 6 July 2015 Manuscript accepted 8 July 2015 Printed in USA
834 www.gsapubs.org | Volume 43 | Number 9 | GEOLOGY
GSA Data Repository 2015282
Lava lake level as a gauge of magma reservoir pressure and eruptive hazard Matthew R. Patrick, Kyle R. Anderson, Michael P. Poland, Tim R. Orr, and Donald A. Swanson U.S. Geological Survey – Hawaiian Volcano Observatory
Additional Methods 1.1 Monitoring network Figure DR1 shows the position of the lava lake relative to summit GPS and tilt instruments, as well as the cameras used for monitoring lava lake level. GPS processing is described in Miklius et al. (2005). 1.2 Lava level measurements The majority of data points in the lava level time series (February 2011-June 2014) were derived from hourly images collected by the HTcam thermal camera on the rim of Halema‘uma‘u Crater (Fig. 1, DR1). The HTcam thermal camera is a Mikron M7500 longwave (7.5-13 microns) camera with an image size of 320 x 240 pixels, having a horizontal field of view of approximately 53° (Patrick et al., 2014). Lava height in image pixels was measured by hand and converted to vertical elevation using a linear fit to sporadic laser rangefinder and LIDAR measurements, and structure-from-motion (SfM) models of the vent geometry (Fig. DR2). The vent crater geometry has evolved due to rim collapses, and we divide the time period into relatively stable epochs which are separated by short periods (on the order of several days) of vent instability. Individual linear regressions were applied to each stable epoch to determine best fit coefficients linking image pixel values to laser rangefinder and LIDAR calibration points, with interpolated values used in the brief periods between these stable epochs. The resulting calibrated lava level measurements from the HTcam images have a root-mean-square residual value of 1.7 m from the calibration points. Most calibration points consist of laser rangefinder measurements of lake level taken from the Halema‘uma‘u Overlook, in which the measurement error is about one meter. Seventeen calibration measurements were from tripod LIDAR measurements of lake level. Five calibration measurements, which are the majority of points in the later portion of 2011, are from structure-from-motion surface models of the vent. The SfM models were constructed using Agisoft Photoscan software using images collected by a handheld thermal camera during helicopter overflights of the vent. The thermal camera used for this purpose was a FLIR Systems SC620 camera with an image size of 640 x 480 pixels. About six waypoints were used to scale and geo-register each surface model, using points taken from an orthorectified WorldView 2 image of Kīlauea’s summit. The error in these SfM measurements is up to a few meters.
For lava level from mid-2010 to February 2011, measurements were made from images collected from a low-light camera (HMcam) also operating on the rim of Halema‘uma‘u Crater (Fig. DR1). These measurements were made once per day, and show the “baseline” lava level (i.e. we excluded images collected during abrupt spikes in lava level due to gas pistoning). The lava level was estimated from these image pixel values using the known field of view of the camera along with a crater geometry taken from tripod LIDAR data collected the previous year. Calibration points during this earlier period (pre-2011) were generally sparse because the crater was narrow and filled with thick fume, and the lava level was lower, making laser rangefinder and conventional theodolite measurements impossible. To check the accuracy of the imagebased measurements during this period, we constructed several SfM models of the crater geometry, again using thermal images collected during helicopter overflights. The thermal camera was effective at “seeing” through the thick fume in the crater. SfM results were within five meters of the image-based estimates, confirming that they are reasonable. In addition, lava level was measured at a finer time increment over limited periods for specific analyses. Lava level was measured from HTcam images as described above for January and February 2012, but at five-minute spacing to examine time delays with summit tilt. During November and December 2010, lava level was measured in a similar manner from HTcam images at 10-minute increments to study gas piston behavior. These 2010 measurements were done when the HTcam camera had a narrower field-of-view lens (21° horizontal field of view), and image pixels were converted to absolute elevation using a combination of the camera viewing geometry and the measured lake and crater geometry inferred from a SfM surface map (described above) constructed from helicopter thermal images collected on December 2, 2010. 1.3. Measuring area of active breakouts with webcam imagery Images were recorded by a Stardot Netcam SC 5-megapixel camera with an 80° horizontal field of view operating in a near-infrared mode during the night hours. Nightly composite images show the maximum brightness detected in a given pixel location throughout the night (Patrick et al. 2010a). A mask was applied to eliminate areas not on the active flow field, and pixels above a 94% gray value were counted for each night. The areas of active lava were ~800 m to 3.3 km from the camera (Fig. DR3), so a simple linear distance correction was applied to the image pixels, based on image row, to correct for the variable viewing distance and estimate pixel footprint size. The webcam was perched on high ground above the coastal plain during our study period (midApril to mid-June, 2012) and captured activity on the lower portion of the East Rift Zone (ERZ) flow field. Two main areas of breakouts were active during this period, with the camera capturing the lowermost, which normally accounted for half, or more, of the total area of breakouts (Fig. DR3). Both areas of breakouts were fed by the same master lava tube, and therefore likely exhibited shared fluctuations in activity. Because of this, we expect that the trend shown in Figure 3C is likely a reliable depiction of fluctuations on the entire flow field.
1.4 Moving-window correlation of tilt and lava level We used a one-week moving window, with 75% overlap, between September 2011 and June 2014, to measure the relationship between UWE radial ground tilt and lava level. Because our focus was on steady-state behavior we used data after September 2011 as they were not affected by large-scale disruptions of the magmatic system, such as ERZ eruptive events (Fig. 2D). We applied linear regression to each window period and only calculated the slope and coefficient of determination (r2) for those windows that had tilt ranges greater than 2 microradians. The r2 values for these results are shown in Figure DR4. To measure the slope of this linear fit, we limited these results further, to those with r2 values >0.90 (Fig. DR5) 1.5 Cross-correlation of tilt and lava lake level Cross-correlation of UWE radial tilt and lava level was accomplished using lava level data that were measured at approximately five-minute intervals during January and February 2012. Tilt data are sampled at one minute intervals. We performed a linear interpolation of both lava level and tilt data to a time vector with uniform five minute sampling. The cross-correlation of these data was performed with MATLAB’s xcorr function (Fig. DR6). 1.6 Bulk density of lava column The bulk density of the lava column is bracketed by two end-members possibilities: 1000 and 2500 kg m-3. The maximum density is limited to that of gas-free liquid tholeiitic basalt (roughly 2600 kg m-3) (Murase and McBirney, 1973), and perhaps representative of magma deeper in the magma column, below the exsolution depth of H2O and SO2. However, recent gravity data suggest that the Halema‘uma‘u lava lake and shallow magma column (upper 150 m) have a very low density, of 950 ±300 kg m-3 (Carbone et al., 2013). The low density suggests that large volumes of gas remain stored in the lake. Tephra ejected from the lake have bulk densities (5002000 kg m-3) (Carey et al., 2012) consistent with this low value; however, it is unclear how deep this very low bulk density region extends into the deeper system. 1.7 Derivation of the magma flux formula We assume that volumetric flux rate Q of magma from the summit to the ERZ is linearly proportional to the difference in pressure Δp between the summit magma reservoir and the ERZ eruptive vent, resisted by the weight of magma in the ERZ conduit: ∆
(2)
Here K is a constant of proportionality, ρ2 is the average magma density in the ERZ conduit, g is gravity, and is the positive vertical distance between the ERZ eruptive vent and the reservoir (Fig. DR9). Note that this model cannot account for an observed time lag of several hours between pressure changes at the summit and then at Pu‘u ‘Ō‘ō, but we believe that it should be correct to first order once these pressure changes fully propagate through the system. Assuming negligible atmospheric pressure so Δ
, where p is reservoir pressure,
(3) Magma reaches the surface in (at least) two locations: the summit lava lake and the ERZ vent at Pu‘u ‘Ō‘ō. The summit lava lake is in magmastatic equilibrium with the magma chamber, while the ERZ vent is not, generally, due to viscous drag in the ERZ conduit. Magmastatic equilibrium in the summit lava lake and conduit can be used to constrain reservoir pressure using p=ρ1g h1, so (4) If magma density is constant throughout the system (ρ1=ρ2 ) then during times of zero flux (magmastatic equilibrium in both the summit and ERZ) the elevation of the summit lava lake should be no higher than the ERZ eruptive vent. However, during brief eruptive pauses it has been observed that the surface height of the summit lava lake maintains a level roughly 80 m higher than that of the ERZ vent. This height difference suggests that the magma in the ERZ conduit is more dense than magma in the summit lava lake, consistent with observed degassing of magma at the summit and the low density of the summit lava lake (Carbone et al., 2013). During times of zero flux, . Rough constraint on 2013) but is largely unknown. Solving for ,
is available (Carbone et al.,
|
(5)
|
where |
denotes the value during the zero-flux condition. Substituting into equation (4), | ∆
(6) (7)
where ∆ is the difference between summit lava lake height and its height at magmastatic | which is valid during times of constant ERZ vent equilibrium, and we have used height. Note that this expression is independent of absolute reservoir depth. Finally, the time-averaged version of equation (7) is given by ∆
(8)
Given time-averaged flux rate and height difference it is thus possible constrain . We use data from April-June 2012, during which time accurate estimates of time-averaged flux rates are available (Poland, 2014), and we compute ∆ from the lava lake time series and an observed magmastatic equilibrium elevation of approximately 922 m. Once Kc is constrained, it can be applied to the equations above for those instances when Q is not known, to translate instantaneous lava level height to instantaneous ERZ flux Q (assuming no major changes in the ERZ conduit or shallow reservoir at Pu‘u ‘Ō‘ō).
1.8 Density difference between summit and ERZ magmas Figure 3a shows that during mid-2012 the magmastatic equilibrium level at the summit was roughly 922 m elevation, or 80 m above that of the ERZ vent. As the magmastatic equilibrium level occurs at times of ERZ pauses when there was likely little to no flow from the summit to Pu‘u ‘Ō‘ō, dynamic effects due to viscosity or conduit geometry cannot account for this height difference. Instead, this 80 m height difference might be explained by density differences in the magma columns. Presumably, much of this density difference is driven by different gas content. We can balance the pressure of the summit and ERZ magma columns using equation (5). We use h1=h2+80, assume that the ERZ conduit and summit are connected at a depth no shallower than the Halema‘uma‘u reservoir at 1.5 km depth (Anderson et al. 2015), and for the summit conduit use a range of bulk densities from 1000 to 2500 kg m-3 (section 1.6). We estimate that magma density along the ERZ is 50-130 kg m-3 higher than that at the summit, which is a relatively minor difference. 1.9 Correlation of modeled and observed ERZ flow field activity Figure 3 shows a comparison of modeled ERZ effusion rates (Figure 3B) with an observed proxy for ERZ effusion rate (Fig. 3C). The latter is based on areas of active lava observed in nightly webcam images. The Figure 3 caption indicates that values for r2 were 0.31-0.51. This range is based on using different delay times between summit and ERZ flow field activity, which Orr et al. (2014) measured as 23.9±6.4 h. To perform the comparison, we assumed that the observed ERZ effusion rate proxy based on composite images depicted nighttime activity (20:00 to 06:00), and then subtracted the time delays. The average value of the modeled ERZ effusion rate was calculated for this delay-corrected time interval, and used to perform the linear regression.
Additional References Carey, R.J., Manga, M., Degruyter, W., Gonnerman, H., Swanson, D., Houghton, B., Orr, T., Patrick, M., 2013, Convection in a volcanic conduit recorded by bubbles: Geology, v. 41, p. 395-398 Miklius, A., Cervelli, P., Sako, M., Lisowski, M., Owen, S., Segal, P., Foster, J., Kamibayashi, K., Brooks, B., 2005, Global Positioning System Measurements on the Island of Hawai‘i; 1997 through 2004: U.S. Geological Survey Open-File Report 2005–1425, 48 p. Murase, T., McBirney, A.R., 1973, Properties of some common igneous rocks and their melts at high temperatures: Geological Society of America Bulletin, v. 84, p. 3563-3592. Patrick, M., Orr, T., Sutton, A.J., Lev, E., Fee, D., 2014b, Episodic outgassing and lava level fluctuations at Kilauea Volcano’s summit lava lake in Halema‘uma‘u Crater. American Geophysical Fall Meeting abstract V33E-08.
Supplemental Figures
Figure DR1. Map of Kīlauea Volcano’s summit region, including relevant monitoring instruments. UWEV, HOVL and CRIM are GPS instruments, and UWE measures ground tilt. HTcam is a thermal camera and HMcam is a visual-wavelength camera. Gray lines are roads.
Figure DR2. Results for conversion of continuous lava level measurements from HTcam (thermal camera) images (in pixels) to elevation using sporadic calibration points from laser rangefinder (most of the points in 2013 and 2014), LIDAR, and photogrammetry measurements.
Figure DR3. Tracking activity levels on the ERZ flow field with a webcam. A. Earth Observing 1 Advanced Land Imager (EO-1 ALI) image (Bands 10-8-9 RGB) of the flow field activity on May 15, 2012. At this time two main areas of breakouts were active, with the webcam capturing the activity on the coastal plain. In general during this study period (midApril to mid-June, 2012), the camera captured half or more of the activity. B. Sample composite image from the webcam, showing activity on the night of April 14-15, 2012. White pixels are active breakouts.
Figure DR4. One-week moving-window linear regression results, relating lava level and radial ground tilt at UWE. In most cases, the linear fits are strong and have very high r2 values, with a mean of 0.82.
Figure DR5. One-week moving-window correlation results, showing the slope (m μrad-1) of the linear regression relating lava level change and radial ground tilt at UWE, limited to those sets with r2>0.90.
Figure DR6. Cross-correlation results for lava level and ground tilt, based on a five minute sampling interval. The peak cross-correlation product is at the five-minute lag position (i.e. lava level follows tilt by five minutes), however, the shape of the curve suggests that the actual peak resides at a lag shorter than five minutes. The precise time lag is not crucial to this study, and our results show that if a time lag does exist it is very short (several minutes) and that the lava level is responsive over short timescales to changes in magma reservoir pressure.
Figure DR7. Comparison of lava level and summit GPS data (vertical component at HOVL station). The scatter plot shows a roughly linear relationship. Lava level data before and after February 2011 were measured in a different manner and are shown by separate colors. Prior to February 2011, lava level was measured once a day for the baseline (non-gas-piston) level, while after February 2011 the level was measured hourly. For those data after February 2011, the average daily lava level is matched to the daily GPS solution. Vertical green lines show ERZ eruptive (solid line) and intrusive (dotted line) events. Lava lake level can be used to determine the time-varying magnitude of persistent eruptive activity on the ERZ, but it can also be useful in forecasting abrupt interruptions to this behavior. This figure shows that lava level rose to unusually high levels prior to the three eruptive events in 2011 on the ERZ. This suggests that rising magmatic pressure, tracked by GPS and lava level, reached critical levels which triggered magma to intrude from its existing ERZ conduit to create new vents. Although the data from 2011 suggest a similar pressure threshold for these events, and would be described as “inflation predictable” (Segall, 2013), later intrusive events in 2012-2014 occurred at higher lava levels, which suggests a dynamic threshold.
Figure DR8. Lava lake level changes due to “gas pistoning”. Gas piston events can last for several hours and involve the lava level abruptly rising above its normal baseline level. Gas piston events occur sporadically (commonly a few per day during 2010-2014) and can have amplitudes of up to 20 m (Patrick et al., 2014, 2014b). They represent relatively abrupt spikes in lava level, and are relatively easy to distinguish from the gradual lava level changes that correlate with deformation data. Gas piston events can also be distinguished based on seismic tremor, as tremor drops to very low levels during the high lava stands of gas pistons. The low tremor levels are associated with a cessation of spattering at the lava lake surface due to gas accumulating in the shallow portions of the lake. The lava level was measured every 10 minutes as described in section 1.2.
Figure DR9. Cartoon of the model geometry.
