Post-Collapse Gravity Increase at the Summit of Kīlauea Volcano, Hawaiʻi
Abstract
We conducted gravity surveys of the summit area of Kīlauea Volcano, Hawaiʻi, in November 2018 and March 2019, with the goal of determining whether there was any mass change at depth following the volcano's May–August 2018 caldera collapse. Surface deformation between the two surveys was minimal, but we measured a gravity increase (maximum 44 μGal) centered on the caldera that can be modeled as mass accumulation in a region ~1 km beneath the surface. We interpret this mass increase to be mostly magma accumulation in void space that was created during the summit collapse. Caldera uplift was evident by April 2019, indicating that the magma volume had reached a point where pressurization could be sustained. Modeled gravity change suggests a maximum magma storage rate at Kīlauea's summit during November 2018 to March 2019 that is much less than the pre-2018 magma supply rate to the volcano.
Key Points
- A gravity increase following the 2018 collapse of the summit of Kīlauea Volcano, Hawaiʻi, indicates mass accumulation in void space ~1 km beneath the surface
- The maximum magma storage rate determined from gravity and deformation is much less than the pre-collapse magma supply rate to the volcano
- Gravity monitoring provides an important means of assessing subsurface magma storage and supply rates as Kīlauea recovers from its 2018 collapse
Plain Language Summary
Measuring changes in gravity over time at a volcano provides an indication of mass variation beneath the surface, which can offer evidence of whether or not magma is accumulating in, or draining from, subsurface reservoirs. After the 2018 collapse of Kīlauea Volcano, Hawaiʻi, there was uncertainty over the status of subsurface magmatic activity. We found that an increase in gravity occurred between November 2018 and March 2019. The gravity increase was not accompanied by uplift of the surface, which implies that mass, probably magma, was accumulating in void space created during the collapse. The rate of magma accumulation was much less than the pre-2018-collapse magma supply rate. By April 2019 Kīlauea was inflating, indicating that the magma reservoir beneath the summit had filled to the point that pressure could be sustained.
1 Introduction
During May–August 2018, Kīlauea Volcano, Hawaiʻi, experienced a voluminous eruption of lava from its lower East Rift Zone (LERZ) coincident with a major summit collapse. Over 1 km3 of lava erupted from LERZ vents, and the volume of summit collapse was 0.83 km3 with the maximum downdrop exceeding 500 m (Lundgren et al., 2019; Neal et al., 2019). Following the end of this major activity, the status of Kīlauea's magma plumbing system was uncertain. Was continuous supply from the mantle source causing magma to again accumulate beneath the summit? Did mantle supply change—either accelerating or decelerating—as a result of the eruption?
Post-collapse magmatic activity beneath Kīlauea's summit caldera may be difficult to detect, given that void space was probably created due to the chaotic nature of the 2018 caldera subsidence. If void space were present, magma could accumulate without causing a pressure increase that would deform the surface or result in significant seismicity, but there should be a gravity signal due to the change in subsurface mass. Past studies of gravity change (also called microgravity) at Kīlauea have documented mass decreases and increases at depth without commensurate surface deformation, interpreted as reflecting the formation and filling, respectively, of void space beneath the summit (e.g., Dzurisin et al., 1980; Johnson et al., 2010). In the aftermath of the 2018 collapse, microgravity may be the only strong surface manifestation of renewed magma accumulation below the summit of Kīlauea, so gravity monitoring is an important means of assessing the magma supply rate and how it might change through time.
We conducted microgravity surveys at Kīlauea in November 2018 and March 2019. During that time period, deformation of the summit caldera was minimal, yet we measured a significant microgravity increase. These results indicate mass accumulation in void space that was created during the 2018 collapse and is located within ~1–2 km of the surface, suggesting increased storage of magma and/or groundwater. Inflation of Kīlauea's summit became clear shortly after the March 2019 gravity survey, supporting the interpretation of magma accumulation and implying that by April 2019, void space had filled to the point that pressurization could be sustained enough to cause surface deformation.
2 Background and Previous Work
The microgravity network at Kīlauea was established in 1975 and expanded significantly in the following decades, ultimately comprising a network of about 50 stations (Figure 1). Microgravity surveys that spanned the 1975 Mw7.7 earthquake identified a mass loss that far exceeded the volume decrease determined from deformation, suggesting the creation of void space within the volcano (Dzurisin et al., 1980). Subsequent surveys between 1975 and 2008 identified a microgravity increase centered 1–2 km beneath the center of the summit caldera but without significant surface uplift, implying magma accumulation in void space (Johnson et al., 2010). During the same time period, gravity decrease and subsidence characterized the south part of the caldera, probably due to gradual magma drainage from a deeper storage area about 3–5 km beneath the surface (Johnson et al., 2010; Kauahikaua & Miklius, 2003). These microgravity variations reflect changes within the two primary magma reservoirs beneath Kīlauea's summit—the so-called Halemaʻumaʻu reservoir, located 1–2 km beneath the center of the caldera, and the south caldera reservoir, at 3–5 km beneath the south part of Kīlauea caldera (Poland et al., 2014).
Following the onset of Kīlauea's summit eruption in 2008, nearly annual surveys during 2009–2012 revealed continued mass increase associated with the Halemaʻumaʻu reservoir. Although some uplift occurred during that time, it was minor compared to that which would be expected from the measured mass increase. Magma filling void space at shallow levels and/or compression of bubble-rich magma was apparently an ongoing process even during the formation and evolution of the 2008–2018 summit lava lake (Bagnardi et al., 2014).
These previous microgravity results offer critical insights into magmatic activity beneath the summit, but they were never used in an operational way—that is, results were never available to aid with interpreting unrest, and interpretations were always made in hindsight. For example, microgravity increase without accompanying uplift during 1975–2008 was not interpreted as magma accumulation in void space at shallow levels beneath Kīlauea caldera until after the onset of the summit eruption in 2008; a more timely result might have helped to anticipate the summit activity (Johnson et al., 2010). To be useful for monitoring, forecasting, and hazard assessment, microgravity results must be interpreted soon after data collection; this was our aim with the November 2018 to March 2019 surveys.
3 Data Collection and Reduction
We conducted two post-2018-collapse microgravity surveys at the summit of Kīlauea (Poland et al., 2019). In November 2018, we occupied 14 stations over the course of 4 days (14–16 and 18 November). In March 2019, we occupied 51 stations over 7 days (19–23 and 25–26 March). Both surveys utilized station P1, northeast of Kīlauea caldera (Figure 1), as the base, consistent with past work (Bagnardi et al., 2014; Johnson et al., 2010). Two of the days in November 2018 were double looped, with the base station (or a subbase) occupied three times during the day to identify any potential tares (sudden offsets in the gravimeter sensor), while all other stations were occupied twice to assess gravimeter drift. The remaining 2 days were single loops but occupied the same stations each day (except for station HVO35, which was done on only one of the days). Double loops were done on all days of the March 2019 survey. During both surveys, all stations, except for sites on the downdropped block (see Figure 1), were accessed by car or walking, using P1 as the thrice-occupied base. Sites on the downdropped block required helicopter access combined with walking and utilized CALM as the thrice-occupied base, with P1 occupied at the start and end of those days.
Data analysis followed Poland and de Zeeuw-van Dalfsen (2019), with tilt and tide corrections provided by the instrument and daily drift rates for each gravimeter calculated using gTOOLS (Battaglia et al., 2012). For the single-looped days in November 2018, we calculated the drift rates based on a linear fit to repeat occupations of the base station and then averaged the 2 days, taking the standard deviation of the measurements as the uncertainty in the gravity values. For station HVO35, which was measured on only 1 day, we assigned an uncertainty of 20 μGal, which is expected to be a worst-case scenario (Poland & de Zeeuw-van Dalfsen, 2019).
Both surveys utilized the same two gravimeters—Scintrex CG-5 units 578 and 579. The meters functioned well, and the calibration factors, which have been remarkably steady through time (Battaglia et al., 2018), probably did not change, given that the meters were not used, shipped, serviced, or otherwise handled between the two surveys. We can therefore compare measurements from a single meter over the two surveys without applying any calibration factors. Temporal changes in gravity determined from each meter can be compared without the need for calibration as well, since gravity changes are not significantly influenced by instrument calibration.
Gravity change is caused not only by variations in subsurface mass, but also vertical displacements due to the “free-air effect,” with a theoretical impact of −3.086 μGal/cm (e.g., Carbone et al., 2017). Surface displacements are measured by both Global Navigation Satellite Systems (GNSS) and interferometric synthetic aperture radar (InSAR) at Kīlauea (Figure 2). Between the November 2018 and March 2019 gravity surveys, GNSS and InSAR indicated minimal vertical deformation except in the south caldera (Figures 2a and 2b). GNSS stations in the north (UWEV) and central (BYRL and CALM) parts of the caldera had net changes of no more than 1 cm—effectively the same as the uncertainty in the measurement (about 9 mm)—so we did not apply any corrections to gravity stations in these regions. GNSS stations in the south part of the caldera (CALS, CRIM, and OUTL) are characterized by subsidence of 3–5 cm over the time spanned by the gravity campaigns. This subsidence, which is also apparent in interferograms (Figure 2b), would cause a free-air gravity increase of up to 15 μGal, which is close to the expected repeatability (Poland & de Zeeuw-van Dalfsen, 2019). For station 112YY (Figure 1), we used the 5-cm subsidence of collocated GNSS site OUTL to determine a free-air effect of 15 μGal, which lowers the observed gravity change from 23 ± 9 to 7 ± 13 μGal. Station HVO35 is midway between GNSS sites CALS (3-cm subsidence over the time spanned) and CRIM (4-cm subsidence), prompting us to use a subsidence magnitude of 3.5 cm—a 10.5-μGal free-air effect—and lowering the gravity change from 40 ± 29 to 29 ± 33 μGal. The uncertainties are slightly higher for the free-air-corrected values to account for uncertainty in the GNSS displacements. We did not apply a correction to stations HVO10 or 113YY, which are east and south, respectively, of the main area of south caldera subsidence. Although the local free-air gradient has been measured at Kīlauea, −3.303 μGal/cm (e.g., Kauahikaua & Miklius, 2003), the difference from the theoretical gradient we used is not significant given the uncertainty in vertical deformation, and therefore, we followed previous workers in using the theoretical gradient (e.g., Bagnardi et al., 2014; Johnson et al., 2010).
4 Results
The gravity differences at each station are slightly different in magnitude depending on the instrument (Table 1), but they show the same overall pattern of change (Figure 3)—gravity increases in the caldera, especially on the downdropped block, with respect to base station P1. The maximum gravity increase during November 2018 to March 2019 is at station CALM and amounts to 44 ± 14 μGal when combining measurements from both gravimeters. The increase decays with radial distance from that site, approaching zero ~2 km away. Four of five stations inside the caldera saw gravity increases of 25 ± 18 μGal or more, whereas only one of six stations outside the caldera saw comparably large increases (Figure 3). Although the network is sparse, the overall pattern of gravity change indicates the result is robust and not an artifact of any one station.
| 2018 | 2019 | 2018-2019 | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 578 | 579 | 578 | 579 | 578 | 579 | Combined | ||||||||
| Station | Grav (mGal) | SD (mGal) | Grav (mGal) | SD (mGal) | Grav (mGal) | SD (mGal) | Grav (mGal) | SD (mGal) | Grav (μGal) | SD (μGal) | Grav (μGal) | SD (μGal) | Grav (μGal) | SD (μGal) |
| 21YY | 13.108 | 0.005 | 13.133 | 0.004 | 13.089 | 0.003 | 13.116 | 0.006 | −20 | 6 | −17 | 7 | −18 | 9 |
| 110YY | 70.328 | 0.009 | 70.424 | 0.016 | 70.363 | 0.00 | 70.450 | 0.004 | 34 | 10 | 26 | 16 | 30 | 19 |
| 112YY | 27.951 | 0.003 | 27.983 | 0.004 | 27.971 | 0.007 | 28.009 | 0.004 | 20 | 7 | 26 | 6 | 7a | 13a |
| 113YY | 28.955 | 0.003 | 28.988 | 0.005 | 28.961 | 0.006 | 28.999 | 0.003 | 6 | 7 | 11 | 6 | 8 | 9 |
| 118YY | 16.928 | 0.003 | 16.946 | 0.005 | 16.944 | 0.006 | 16.966 | 0.003 | 16 | 7 | 20 | 6 | 18 | 9 |
| CALM | 72.208 | 0.007 | 72.320 | 0.012 | 72.264 | 0.002 | 72.352 | 0.003 | 56 | 7 | 32 | 12 | 44 | 14 |
| HVO10 | 21.078 | 0.009 | 21.101 | 0.001 | 21.086 | 0.007 | 21.115 | 0.004 | 8 | 11 | 15 | 4 | 11 | 12 |
| HVO28 | −0.812 | 0.003 | −0.813 | 0.005 | −0.810 | 0.007 | −0.813 | 0.003 | 2 | 7 | 0 | 6 | 1 | 9 |
| HVO31 | 23.190 | 0.003 | 23.21 | 0.005 | 23.205 | 0.003 | 23.246 | 0.006 | 15 | 4 | 32 | 8 | 24 | 9 |
| HVO35 | 24.325 | 0.020 | 24.365 | 0.020 | 24.365 | 0.007 | 24.405 | 0.004 | 40 | 21 | 40 | 20 | 29b | 33b |
| HVO47 | 74.496 | 0.009 | 74.596 | 0.016 | 74.514 | 0.002 | 74.606 | 0.004 | 18 | 9 | 11 | 16 | 14 | 18 |
| HVO48 | 74.870 | 0.009 | 74.980 | 0.015 | 74.903 | 0.003 | 74.997 | 0.004 | 32 | 9 | 18 | 16 | 25 | 18 |
- Note. Differences (2018–2019) are given in microgals. Raw data are available from Poland et al. (2019).
- a Corrected for a free-air effect of −15 μGal (which increases SD by 4 μGal to account for GNSS uncertainty).
- b Corrected for a free-air effect of −10 μGal (which increases SD by 4 μGal to account for GNSS uncertainty).
We modeled the 2018–2019 change using the Geodetic Bayesian Inversion Software of Bagnardi and Hooper (2018) modified for use with gravity data and assuming a point source, as in Bagnardi et al. (2014). Our modeling assumes uniform prior probabilities between arbitrary bounds, which we select to be wide enough to contain what we consider to be geologically realistic values of source location and strength (see supporting information for details). We found an optimal (maximum a posteriori probability) source location about 1 km southwest of station CALM at a depth of about 720 m (440–3,728 m as 95% credible interval), with a mass change of 1.5 × 1010 kg (0.9–2.8 × 1010 kg as 95% credible interval). This location is very similar to that of the pre-2018-collapse Halemaʻumaʻu reservoir (e.g., Poland et al., 2014).
5 Discussion
We interpret the cause of the gravity increase to be mostly magma accumulation in void space that was created during the chaotic collapse of 2018 at a depth of ~1 km beneath the caldera floor—the location of the Halemaʻumaʻu reservoir that drained during the 2018 LERZ eruption and summit subsidence (Neal et al., 2019). The only other potential cause of gravity increase is groundwater accumulation. This has been discounted as a likely source by previous studies of gravity change at Kīlauea because of the unrealistic scale of the level change in the ~500-m-depth water table (Kauahikaua, 1993; Keller et al., 1979) needed to explain observed signals. In the current situation, water as a source of mass increase is more reasonable, given that water is infiltrating the area vacated by the magmatic conduit that fed the summit lava lake. A small water pond was identified at the lowest point of the collapse pit, which is about 70 m below the water table, in late July 2019, and had grown to a depth of about 20 m by mid December 2019.
Three lines of evidence suggest that the mass increase is mostly due to magma accumulation and not water. First, the mass flux is greater than expected for water. The modeled mass accumulation over November 2018 to March 2019 (see section 4) gives a mass flux of 1,266 kg/s (814–2,532 kg/s as 95% credible interval). This is 2 orders of magnitude higher than the modeled likely water influx, even assuming high permeabilities (Hsieh & Ingebritsen, 2019). Observations of pond growth during July–October 2019 suggest an influx of less than 10 kg/s, consistent with the modeling of Hsieh and Ingebritsen (2019). Second, the modeled location and depth of the mass increase is very similar to that of the pre-collapse Halemaʻumaʻu reservoir (e.g., Poland et al., 2014). Third, starting in April 2019, unambiguous uplift became apparent in interferograms and at GNSS stations within the caldera, and especially on the downdropped block (GNSS stations CALS and CALM; Figures 1 and 2), where the gravity increase was strongest. This uplift is difficult to model owing to the inelastic nature of the heavily faulted caldera floor, but it has a spatial pattern that resembles previous episodes of deformation that were attributed to changes in pressure within the shallow Halemaʻumaʻu reservoir. In addition, the rate of caldera uplift—tens of centimeters per year—far exceeds other examples of groundwater-related uplift, which typically occurs at rates of millimeters per year to a few centimeters per year (e.g., Bell et al., 2008; Schmidt & Bürgmann, 2003). Although water influx probably contributes to the measured mass change, based on the above evidence we believe this contribution to be small compared to that of magma accumulation.
Our preferred conceptual model (Figure 4) of post-2018-collapse gravity change and deformation involves the formation and subsequent filling of void space beneath Kīlauea caldera. Void space was created during drainage of the Halemaʻumaʻu reservoir in May–August 2018 due to the chaotic collapse of the caldera floor (Figures 4a and 4b). The presence of void space is suggested by the scale of the collapse, numerous open fractures that now exist throughout the caldera, and analogy with past large summit subsidence events, like that associated with the 1975 earthquake (Dzurisin et al., 1980). During November 2018 to March 2019, magma was gradually filling the void space, causing the gravity increase recorded at the summit (Figure 4c). By April 2019, sufficient void space had been filled for pressure to be transmitted to the surrounding rock, causing surface uplift to accompany continued magma accumulation (Figure 4d).
The gravity increase can be explained by a mass accumulation of 1.5 × 1010 kg (0.9–2.8 × 1010 kg as 95% credible interval; see section 4 and the supporting information). Assuming a density of 2,300–2,800 kg/m3 for basalt, this mass implies a volume of 5.0–6.1 × 106 m3 (2.9–12.2 × 106 m3 as 95% credible interval for the given density range). Over November 2018 to March 2019, this works out to a magma storage rate of 0.014–0.018 km3/year (0.008–0.035 km3/year as 95% credible interval). If some of the mass increase is due to groundwater influx, this rate of magma volume storage can be viewed as a maximum. Previous studies of overall magma supply to Kīlauea have documented a long-term rate of 0.1 km3/year, with a near doubling of that rate during a surge in the middle-late 2000s, a slight lull below the long-term rate in the early 2010s, and a return to the long-term average by the mid-2010s (Anderson & Poland, 2016; Dzurisin & Poland, 2018; Poland, 2014; Poland et al., 2012, 2014). The range of maximum post-collapse magma accumulation rates in Kīlauea's Halemaʻumaʻu reservoir is thus an order of magnitude smaller than the long-term magma supply rate to the volcano. Although our modeling does not account for material and rheological heterogeneity, and magma is a compressible fluid (which complicates the conversion of mass to volume), the order-of-magnitude difference between magma accumulation rates is unlikely to be caused solely by these complexities.
These results provide no evidence for a post-collapse surge in magma supply to Kīlauea's summit—a possibility that has been proposed previously based on models of the pressure difference between the shallow and deep magma plumbing system following major summit drainage (e.g., Dvorak & Okamura, 1987). This may indicate that magma supply overall to the volcano is reduced relative to pre-2018 levels. Alternatively, it is possible that magma was being supplied to other storage areas, including the rift zones, that cannot be detected by the limited gravity data set we collected. The south caldera was subsiding during November 2018 to March 2019 (Figures 2a and 2b) so was probably not a locus of magma accumulation during that time. In contrast, uplift and spreading of the middle East Rift Zone, just east of Puʻu ʻŌʻō, has been observed since the end of major eruptive activity in August 2018, suggesting the possibility of magma storage in the area. In addition, seaward motion of the south flank has accelerated since 2018, raising the possibility of magma storage in deeper parts of the rift zone as well (Poland et al., 2014). With regard to Kīlauea's Halemaʻumaʻu magma reservoir specifically, however, gravity and deformation in the immediate aftermath of the 2018 collapse both suggest that magma storage is not occurring at a level commensurate with pre-2018 magma supply rates. Future gravity surveys will be critical for better constraining magma storage and supply rates at Kīlauea's summit and rift zones and possible changes over time.
Our work emphatically demonstrates the importance of gravity monitoring at Kīlauea and elsewhere. Only gravity measurements were able to detect mass accumulation at a time of great uncertainty following the 2018 summit collapse. Preliminary results were available to the Hawaiian Volcano Observatory immediately upon completion of the March 2019 survey, allowing them to be used in assessments of summit activity and planning for additional monitoring (both gravity and otherwise). The results, when combined with deformation data, also provide constraints on magma storage and supply that would not otherwise be possible. Magma supply from depth and eruptive activity at the surface are directly related at Kīlauea (e.g., Poland et al., 2012, 2014), so assessing the rate of supply, including any potential changes, is of vital importance for monitoring, forecasting, and hazard assessment. While deformation might also indicate changes in magma storage and supply, the presence of subsurface void space and the compressible nature of magma mean that deformation measurements alone are insufficient.
6 Conclusion
Following the 2018 summit collapse at Kīlauea, we detected a gravity increase at the volcano's summit caldera during November 2018 to March 2019 that we interpret as mass increase due to magma accumulation in a reservoir at 1- to 2-km depth. The gravity increase occurred in the absence of significant surface uplift, arguing that magma accumulation was occurring in void space created by the collapse. By April 2019, caldera inflation was evident in GNSS data and InSAR, indicating that void space had been filled to a point that the magma reservoir could sustain pressurization. The maximum magma storage rate determined from the November 2018 to March 2019 mass increase, coupled with overall subsidence of the summit, indicates that magma supply to Kīlauea's summit magma plumbing system following the collapse was low compared to previous years. The massive volume loss from the summit reservoir system apparently did not trigger an immediate increase in supply from a deeper magma source, as had been hypothesized. Continued gravity and deformation monitoring is needed to assess the status of this condition over time, aid with hazard assessment, monitor the state of the volcano, and forecast future activity.
Acknowledgments
Gravity data analyzed in this work are available from Poland et al. (2019): https://doi.org/10.5066/P9L1K6SJ. GNSS data are archived at UNAVCO. SAR data are courtesy of the Agenzia Spaziale Italiana (ASI) and were provided to the Hawaiʻi Supersite via the Geohazards Natural Laboratories and Supersite initiative. M. B. was supported by an appointment to the NASA Postdoctoral Program at the Jet Propulsion Laboratory, administered by the Universities Space and Research Association (USRA) through a contract with NASA. We are grateful to David Okita for his exceptional piloting skills, which enabled access to gravity sites on the downdropped block, and to Hawaiʻi Volcanoes National Park for their support in conducting the work. The Hawaiian Volcano Observatory provided critical support to all phases of the project, and we are particularly grateful to Sarah Conway, Ashton Flinders, and Carolyn Parcheta. Our thanks to Jim Kauahikaua, Dan Dzurisin, Jo Gottsmann, and an anonymous reviewer for their valuable input.
