Seismic and geodetic progression of the 2018 summit caldera collapse of Kīlauea volcano

https://doi.org/10.1016/j.epsl.2020.116250Get rights and content

Highlights

  • Analyzed cyclic deformation and earthquake swarms from stress loading and faulting.

  • Evidence for asymmetric failure of the magma reservoir roof observed.

  • Identified 3 phases of caldera collapse, including decoupling/recoupling of roof.

  • Observed tangential motion of central piston and movement of peripheral block.

  • Caldera collapse mostly a passive result of magma reservoir drainage.

Abstract

The 2018 eruption of Kīlauea volcano, Hawai'i, resulted in a major collapse of the summit caldera along with an effusive eruption in the lower East Rift Zone. The caldera collapse comprised 62 highly similar collapse cycles of strong ground deformation and earthquake swarms that ended with a magnitude 5 collapse event and one partial cycle that did not end with a collapse event. We analyzed geodetic and seismic data to better understand how the caldera collapse progressed over 3 months of activity, focusing on the cyclical activity. We identified 3 main phases of collapse: initial ring-fault activation and small explosions (Phase 1), an eastward shift in activity and freeing of the central piston (Phase 2), and a recoupling of the piston to the reservoir followed by relatively steady behavior until the eruption's end (Phase 3). Additionally, we observed geodetic evidence of tangential motion from the localization of the main ring fault (Phase 2) and the formation of a major peripheral ring fault on the eastern side of the collapse caldera during Phase 3. Both geodetic and seismic parameters suggest that the collapse may have had an eastward-component of motion after the ring fault system had formed. The cyclical seismic and geodetic parameters show no obvious signs that the collapse was coming to an end, with the only notable change being a significant increase in the ratio of cyclical displacement to co-collapse displacement observed during the last complete cycle on GNSS stations outside the caldera region.

Introduction

During the 2018 eruption of Kīlauea volcano, the summit crater and surrounding regions of the caldera floor slowly subsided and repeatedly collapsed over the course of 3 months due to the magma reservoir below the summit draining into the lower East Rift Zone (ERZ) and erupting at a network of fissures there (Neal et al., 2019). The caldera collapse comprised 62 complete cycles that each ended in a collapse event, wherein the caldera floor rapidly subsided with a moment release approximately equivalent to a magnitude 5 earthquake, and a final partial cycle which did not result in a collapse event. Each cycle comprised widespread ground deformation and a swarm of earthquakes that increased in rate leading into the collapse event. The first 12 cycles had a distinct behavior compared to the later cycles, and those events are thought to be combined collapses and explosions (e.g., Crozier et al., 2018; Neal et al., 2019). The caldera collapse ultimately affected an elliptical region roughly 2.5 km by 2 km. The summit eruption and caldera collapse were well-recorded by seismometers and geodetic instruments both inside and outside the collapse region.

Prior to the eruption, the summit of Kīlauea was defined by the large, elliptical Kīlauea Caldera (approx. 3km×4.5km) and the smaller Halema'uma'u (HMM) crater (1km in diameter) in the southwestern portion of the caldera (Fig. 1a). Kīlauea Caldera, HMM crater, and several other craters have formed at the summit of Kīlauea throughout its history as a result of collapse processes associated with eruptions (e.g., Macdonald, 1965; Swanson et al., 2014). Starting in 2008, HMM crater hosted a small, persistent lava lake on the eastern side of the crater (Patrick et al., 2018). The 2018 eruption began with an overflowing of the lava lake in late April before it began to quickly drain on May 1 following the collapse of Pu'u ‘Ō’ō crater in the middle ERZ (Neal et al., 2019). On May 4, a Mw 6.9 earthquake occurred on the south flank of Kīlauea, releasing compressive stress along the magma conduit to the lower ERZ and accelerating the pace of the eruption. Shortly after, strong subsidence was recorded at the summit, and the caldera collapse process described above began, with the first collapse event recorded on May 17 (Fig. 1b). By the last collapse event on August 2, HMM crater had expanded to 15 times its pre-eruption volume, deepening by over 500 m in some areas (Neal et al., 2019).

Similar caldera collapses to the one of Kīlauea in 2018 have been recorded at Fernandina, Galapagos in 1968 (Simkin and Howard, 1970; Filson et al., 1973); Miyakejima, Japan in 2000 (Kumagai et al., 2001; Geshi et al., 2002; Kobayashi et al., 2003); Piton de la Fournaise (PdF), La Reunion in 2007 (Michon et al., 2007; Staudacher et al., 2009); and Bardarbunga, Iceland in 2014 (Gudmundsson et al., 2016). All of these collapses occurred in a series of large amplitude, cyclic collapse events originating from beneath the summit with ground deformation and seismicity observed during each cycle as the available instrumentation permitted. The Fernandina collapse had the most distant observations, primarily seismic, and the Miyakejima collapse had more proximal but still relatively little geodetic or seismic recording of the collapse. The PdF eruption was recorded by several geodetic and seismic stations surrounding the area of collapse (Staudacher et al., 2009). The Bardarbunga collapse was also well-recorded; however, the area of collapse is buried under 700–800 m of ice, and only one geodetic station was located near the collapse region (Gudmundsson et al., 2016). All of these collapses generated a sequence of large amplitude (M45.5) seismic events, typically with strong very-long-period components. The general interpretation of these collapses is that they occurred due to the emptying of a large magma body beneath the volcano's summit that induced collapse through a piston-like mechanism which caused the floor to drop in steps corresponding to the large-amplitude seismic events. Additionally, the Fernandina and Miyakejima collapses were accompanied by summit explosions preceding or in the early stages of collapse (Simkin and Howard, 1970; Geshi et al., 2002), similar to Kīlauea.

Various analog (e.g., Marti et al., 1994; Roche et al., 2000) and numerical (e.g., Chery et al., 1991; Hardy, 2008; Holohan et al., 2015) studies have further explored the details of how caldera collapses occur from initiation to conclusion. Most of these studies start from an intact, largely homogeneous magma reservoir roof that collapses to form a caldera as the reservoir drains. They suggest that the process is asymmetrical with the faults forming alternately on opposite sides of the collapsing region and with the progression and ultimate morphology dependent on the height-to-width aspect ratio of the roof. Two limitations of these studies are that they typically start with an intact roof and that they offer little detailed explanation for the cyclical collapse steps observed during caldera collapses at volcanoes. In most cases, such as those previously described, caldera collapses occur within regions where previous collapses of varying size occurred. Ruch et al. (2012) performed analog experiments to analyze the kinematics of collapse and also considered the effect of pre-existing ring faults for cases with high roof aspect ratios. They found 3 types of collapse – continuous, incremental, and sudden – and showed that incremental collapse only occurs after ring fault structures have fully formed or if they are pre-existing. Michon et al. (2011) took an observational approach to understanding the cyclical collapse processes, comparing data from the Fernandina, Miyakejima, and PdF collapses. As previously noted, however, the data from these collapses was relatively limited and provided no geodetic observations from within the collapse region.

In this study, we analyze geodetic and seismic data from the 2018 Kīlauea eruption to better understand the processes involved in the caldera collapse and its progression throughout the eruption. In contrast to the aforementioned collapses, geodetic data were recorded both within and outside the Kīlauea collapse region, allowing for a more detailed examination of the deformation (Fig. 1b). The geodetic displacement data vary in progression during each cycle based on station location and reflect the stress loading conditions. These spatial and temporal variations in behavior provide insight into the formation of the central piston and its coupling to the deflating magma reservoir. The seismic data are dominated by swarms of earthquakes that initially accelerate in rate during each collapse cycle (Fig. 1b), reflecting stress loading and inelastic damage in the build-up to each major collapse event. The swarm earthquakes were primarily located in the eastern portion of the collapse region and appear to be more closely related to eastern fracturing and fault activity than to the M5 collapse events of the central piston.

Section snippets

Data & analysis

In this section, we provide an overview of the seismic and geodetic data and methods used for our study and present the results of our analysis. More details about the data and methods can be found in Appendix A.

Conceptual model of caldera collapse

Analog (e.g., Roche et al., 2000) and numerical (e.g., Holohan et al., 2015) studies have explored the dynamics of caldera collapse, typically beginning with completely intact roof rock above the magma reservoir. They show that caldera collapse evolves differently depending on the aspect ratio (height to width, r) of the roof. An aspect ratio of r1 divides the potential morphologies into two broad regimes which are further dependent on the strength of the roof rock. Those roofs with r>1 form a

Conclusions

The 2018 Kilauea caldera collapse is arguably the best seismically- and geodetically-recorded basaltic caldera collapse. We analyzed earthquake swarms and ground deformation that occurred between cyclical collapse events to understand how the collapse progressed. From our results, we divided the collapse into 3 phases after the initial sagging stage:

  • Phase 1: The ring fault system has been formed or reactivated enough that collapse events begin to occur and are accompanied by explosions. During

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We thank Eoghan Holohan, Laurent Michon, and Emily Montgomery-Brown for their very helpful and constructive reviews that greatly improved the manuscript. G. Tepp and A. Hotovec-Ellis were funded through the USGS Mendenhall Fellowship Program. We acknowledge the use of imagery from the NASA Worldview application (https://worldview.earthdata.nasa.gov/), part of the NASA Earth Observing System Data and Information System (EOSDIS). The early July LiDAR map was downloaded from the OpenTopography

References (40)

  • J. Chery et al.

    Numerical modelling of caldera dynamical behaviour

    Geophys. J. Int.

    (1991)
  • J.A. Crozier et al.

    Hindcasting May 2018 Kilauea summit explosions with remote sensing, geophysical monitoring, and eruption simulations. Part 1: seismic source inversions and self-consistent initial conditions for plume models

  • H.R. Dietterich et al.

    Lava flow hazard modeling and the assessment of effusion rates and topographic change with UAS and lidar during the 2018 Kilauea lower East Rift Zone eruption

  • A.M. Dziewonski et al.

    Determination of earthquake source parameters from waveform data for studies of global and regional seismicity

    J. Geophys. Res.

    (1981)
  • J. Filson et al.

    Seismicity of a caldera collapse: Galapagos Islands 1968

    J. Geophys. Res.

    (1973)
  • F.R. Fontaine et al.

    Very- and ultra-long-period seismic signals prior to and during caldera formation on La Réunion Island

    Sci. Rep.

    (2019)
  • N. Geshi et al.

    Caldera collapse during the 2000 eruption of Miyakejima Volcano, Japan

    Bull. Volcanol.

    (2002)
  • M.T. Gudmundsson et al.

    Gradual caldera collapse at Bárdarbunga volcano, Iceland, regulated by lateral magma outflow

    Science

    (2016)
  • S. Hardy

    Structural evolution of calderas: insights from two-dimensional discrete element simulations

    Geology

    (2008)
  • E.P. Holohan et al.

    Effects of host-rock fracturing on elastic-deformation source models of volcano deflation

    Sci. Rep.

    (2017)
  • Cited by (22)

    • Are the earthquake locations reliable in caldera structures?

      2022, Journal of Volcanology and Geothermal Research
      Citation Excerpt :

      Caldera collapse events are relatively common phenomena in earth sciences, in recent years, several major basaltic eruptions have occurred beneath caldera structures; deciphering their collapse mechanism is paramount to assessing volcanic hazards. Bárðarbunga (Gudmundsson et al., 2016; Riel et al., 2015; Sigmundsson, 2019), Piton de la Fournaise (Michon et al., 2007; Peltier et al., 2009), Sierra Negra (Jónsson et al., 2005; Bell et al., 2021), Kilauea (Anderson et al., 2019; Tepp et al., 2020), among others, are recent examples of such processes. This endeavour provides insight into how the locations (hypocentres and centroids) are influenced by the curvature of the ring rupture; however, it does not intend to provide a definite solution for all cases, since there are several parameters I did not consider, but I address in this section.

    • Material failure and caldera collapse: Insights from the 2018 Kilauea eruption

      2021, Earth and Planetary Science Letters
      Citation Excerpt :

      This observation suggests that the ground motion at those locations became disconnected. The separation began as the surface expression of the eastern collapse boundary became apparent (June 22-30) along with other changes that Tepp et al. (2020) interpreted as the formation of a secondary ring fault that defined a separate block on which CALS was located. The independent tilting motion of this block could explain the differences in the CALS and VO46 displacements after July 1 and the relatively high variation in CALS displacement a-values.

    View all citing articles on Scopus
    View full text