Ongoing eruptive activity at Kīlauea volcano’s (Hawai‘i) summit has been controlled in part by the evolution of its vent from a 35-m-diameter opening into a collapse crater 150 m across. Geologic observations, in particular from a network of webcams, have provided an unprecedented look at collapse crater development, lava lake dynamics, and shallow outgassing processes. These observations show unequivocally that the hundreds of transient outgassing bursts and weak explosive eruptions that have punctuated the vent’s otherwise nearly steady-state behavior, and that are associated with composite seismic events, were triggered by rockfalls from the vent walls onto the top of the lava column. While the process by which rockfalls drive the explosive bursts is not fully understood, we believe that it is initiated by the generation of a rebound splash, or Worthington jet, which then undergoes fragmentation. The external triggering of low-energy outgassing events by rockfalls represents a new class of small transient explosive eruptions.


Ongoing eruptive activity at Kīlauea volcano’s (Hawai‘i) summit (Wilson et al., 2008; Houghton et al., 2011) has been dominated by outgassing from a lava lake in a pit within Halema‘uma‘u Crater (Figs. 1 and 2). Since the onset of the eruption, the generally steady outgassing from the vent has been punctuated sporadically by dozens of small explosive eruptions, and hundreds of even smaller outgassing bursts (Patrick et al., 2011). Associated with these bursts are composite seismic events, which we define as being characterized by an emergent high-frequency (HF) onset that transitions into long-period (LP) seismicity superimposed on a distinctive very-long-period (VLP) oscillation (Fig. 3; Fig. DR1 in the GSA Data Repository1). The VLP oscillation is often large enough to be seen in the raw velocity waveforms without the aid of filtering.

Continuously recording webcams have documented the eruption since its onset in 2008, and the video observations of lava lake dynamics and explosive activity are unparalleled. The video record, enhanced by near-daily field observations, shows that the explosive eruptions and outgassing bursts, which have occurred sporadically since the vent opened, were triggered by rockfalls from the vent walls onto the top of the lava column. Here we look in detail at rockfalls that occurred during January–March 2011, a period of slowly rising lava level within the vent. These were among the best observed rockfalls recorded to date, and they provide direct evidence that rockfalls can trigger explosive outgassing of low intensity at Kīlauea.


After a 25 yr hiatus in eruptive activity at the summit of Kīlauea, a new eruptive vent opened on 19 March 2008, forming a 35-m-wide opening on the southeastern wall of the 1-km-wide Halema‘uma‘u Crater (Fig. 1) (Wilson et al., 2008). By March 2011, the vent had evolved into a 150-m-diameter collapse crater that extended onto the relatively flat floor of Halema‘uma‘u (Figs. 1 and 2A). This widening was the result of the piecemeal collapse of the vent walls and overhanging portions of the vent rim into an underlying cavity holding an actively circulating lava lake (Figs. 1B and 2A).

The lava lake first became visible in September 2008. Since then, its level has varied from a high of 65 m below the vent rim to a low of at least 210 m below the rim, rising and falling within a roughly cylindrical vertical conduit (Fig. 1B). On a few occasions, the lava surface has dropped out of sight below 210 m, revealing a floor with a complex and changing arrangement of one to three smaller openings. Based on these observations, the minimum volume for the entire collapse crater indenting the floor of Halema‘uma‘u is ∼4 × 106 m3. The amount of lithic material erupted from the vent, however, totals a mere 1.1 × 103 m3 (as of March 2011), accounting for <0.1% of the collapse crater volume. Vent enlargement has occurred as a result of the collapse of wall rock into the vent and the transport of this material into the deeper magma system (Swanson et al., 2009).

Such collapses have been common throughout the eruption, and thousands of rockfalls have been observed, ranging in size from tiny fragments to massive slices of the vent rim with volumes exceeding 1 × 105 m3. The largest several hundred of these rockfalls were followed by short-lived ash emissions from the vent (Fig. 2B). Dozens of times, these rockfalls triggered small explosive eruptions that deposited ash- to bomb-sized juvenile scoria and ash- to block-sized lithic clasts as much as several hundred meters from the vent. The scoria pyroclasts are coated in lithic dust, and lithic fragments are distributed throughout their juvenile matrix (Fig. 2D; Wooten et al., 2009), showing a thorough mixing of lithic and juvenile components during eruption.

In mid-November 2010, the lava lake began to rise slowly as Kīlauea’s magmatic system pressurized. By January 2011, the lava lake surface had reached to within ∼100 m of the rim of the vent. Over the following 2 months, the lava level rose an additional 35 m, bringing it to within 65 m of the vent rim by early March (Figs. 1B and 2A). On 5–6 March 2011, the lava lake dropped ∼145 m over a period of about a day as summit magma intruded Kīlauea’s east rift zone, where it fed a new eruption. We focus on this period from January to 5 March 2011, when the lava lake was at its highest level.


The foundation of this study is the wealth of observational data collected during the eruption. In particular, we analyze video of the lava lake collected at ∼4 frames per second using a StarDot Netcam SC webcam operating in a near-infrared mode. This camera was positioned on the edge of Halema‘uma‘u Crater, 85 m above the erupting vent (Fig. 2A). Video from this camera was used to time rockfalls and their associated explosive events and to elucidate eruption mechanisms. The video was telemetered to the Hawaiian Volcano Observatory, where it was time stamped to Hawaii-Aleutian Standard Time (Coordinated Universal Time [UTC] minus 10 h) by the acquisition software (Milestone XProtect Professional) based on the clock of the acquisition computer, which was synched to a network time server. The image time stamp is precise to 0.01 s, but was found to be ∼0.4 s slow. The image time stamps shown in Figure 3, and in Movies DR1–DR5 in the Data Repository, have not been adjusted for this delay.

We compare the video data to seismicity from the vertical channel of station NPT (North Pit), where the seismic signal is time stamped. NPT, one of Kīlauea’s 10 summit seismic stations (Fig. 1), is a broadband station (60 s—50 Hz sensitivity) located in an underground vault ∼800 m north of the active vent. The composite seismic events share the same frequency content across all of Kīlauea’s summit stations, and thus the frequencies are an effect of the source, not the path the waves travel through or the seismometer site. The seismic traveltime between NPT and the vent is ∼0.4 s (assuming shear wave velocity = 2.3 km s−1; Chouet et al., 2010). Thus, the timing of the video data matches the timing of the seismicity, and no correction is required to directly compare the two. A 0.5 Hz low-pass minimum phase shift filter was applied to the seismic data to effectively reduce the noise and allow us to focus on the frequencies of interest.


Video and field observations during early 2011 showed, at times, an almost constant rain of lithic material from the vent walls into the lava lake. This disintegration process was accelerated by vent wall heating due to a slowly rising lava lake level. The collapse of larger vent wall sections often took place as a sequence of failures that occurred progressively higher up the wall in a process similar in style to that of roof stoping above an underground cavity (e.g., Waltham et al., 2005, p. 57). These sequences of larger rockfalls, which typically spanned a period of several hours, were interspersed with many smaller rockfalls and usually culminated in catastrophic collapses of the vent rim. Five rockfall sequences occurred in 2011: one that spanned midnight on 17–18 January, and one each on 14 and 15 February and 2 and 3 March (Fig. DR1). In every case, substantial rockfalls were associated with HF seismicity, and the largest of these with LP and VLP seismicity. This is apparent in the video record, which shows a one-to-one correlation between rockfalls and seismic events during periods when visibility was adequate, which includes most composite seismic events in early 2011 (e.g., Movies DR1 and DR2).

In all of the larger rockfalls, the wall rock disaggregated into loose debris, enveloped by a cloud of dust, during descent (Fig. 2C; Movies DR1–DR5). The impact of the rockfalls with the lava lake surface then resulted in the release of billowing ash clouds (Fig. 2B; Movies DR1–DR5) and, in the case of the largest collapses, the visible ejection of juvenile tephra (e.g., Fig. 2D; Movies DR1 and DR4). The ash cloud and eruption of juvenile tephra always originated from the point of impact (e.g., Movie DR3). There was also a general correspondence between the size of the rockfall and the resulting emission of pyroclastic material. Only the largest rockfalls triggered explosive eruptions that dispersed juvenile lapilli and bombs beyond the rim of the vent. The preexisting condition of the lava lake (i.e., whether capped by a crust or energetically spattering) seemed to have little bearing on the eruptive response following rockfalls. This is apparent during rockfall sequences where later rockfalls were just as likely to trigger an explosive eruption as earlier rockfalls, despite the lava surface being completely disrupted by the earlier rockfalls (e.g., Movie DR1).

The start of the rockfalls for all visible events coincided closely with the onset of the emergent HF seismic energy (e.g., Movie DR5). In many instances, steaming cracks, marking the line of eventual failure, opened on the floor of Halema‘uma‘u as much as several minutes in advance of the rockfall (e.g., Movie DR5). A sudden increase in HF seismicity was coincident with the main body of the rockfall striking the lava lake surface (Fig. 3; Movies DR1–DR5). Ash emission and outgassing followed this impact and corresponded to the onset of LP and VLP energy. The lava lake, when not completely obscured by the ash cloud, sloshed visibly with a 7–12 s period that was intermediate to the dominant periods of the LP (1–2 s) and VLP (20–40 s) energy.


The configuration of the vent in Halema‘uma‘u in early 2011 (and through much of the eruption), with vertical to overhanging walls poised above a circulating lava lake, meant that rockfalls from the vent walls nearly always impacted the lava surface. When large enough, these collapses were accompanied by a composite seismic event. The webcam record shows a striking one-to-one correlation between rockfalls and composite seismic events, where each discrete composite seismic event is linked to a visible rockfall, and each rockfall resulted in the emission of a short-lived ash plume. Such plumes have been a common feature of Kīlauea’s ongoing summit eruption. Moreover, the size of the rockfall, in a qualitative sense, scales with VLP duration and amplitude.

Chouet et al. (2010) proposed an alternate model for the composite seismic events. In their model, the emergent HF signal is related to the near-surface expansion and bursting of a gas slug, which then stimulates the LP and VLP response. Applying a two-pole, zero-phase Butterworth filter to a selection of composite seismic events that occurred during the first year of the eruption, each of which are also associated with rockfalls, Chouet et al. (2010) reported that the VLP onset precedes the HF seismicity by ∼20 s; they cited this as evidence of an ascending gas slug. This method of filtering, however, is known to introduce acausal signal artifacts that produce erroneous onset times (Scherbaum and Bouin, 1997; Haney et al., 2012). By using a minimum phase shift filter, which is causal and preserves onset times (Haney et al., 2012), we find that the rockfalls start at or before the onset of HF seismicity, and that HF seismicity increases abruptly upon rockfall impact, indicating that the rockfall is the source of the seismicity (Fig. 3; Movies DR1–DR5). The transfer of a rockfall’s momentum to the lava column during impact likely induces a pressure transient that is transmitted through the conduit (Patrick et al., 2011). We believe that the pressure transient generates the decaying VLP oscillation, and speculate that the LP energy is related to the explosive outgassing process.

Gas slugs would be expected to burst at the point of lava upwelling, which has dominantly been at the northern edge of the lava lake; the two relatively large (∼10 m diameter) gas slugs that have been identified in the video record over the past 4 yr did just that. Only one of those gas slugs was associated with VLP energy, and neither was associated with appreciable HF seismicity or an explosive eruption. Instead, we see that the outgassing and explosive bursts associated with the composite seismic events originate from the rockfall impact point, regardless of which point on the lava lake surface the impact occurs (e.g., compare back-to-back rockfalls in Movie DR3). This further excludes gas slugs as the cause of the outgassing that accompanies rockfall impact.


While the observational data show unequivocally that rockfalls trigger outgassing, and in some cases weak explosive eruptions, the mechanism responsible for this response remains obscure. Several possibilities exist. For example, an active hydrothermal system is visible on some portions of the vent wall. Other evidence of this hydrothermal system is found in the eruption of hexahydrite [MgSO4·6H2O] spherules from the vent following periods of rainfall (Hon and Orr, 2011), and in the ejection of sparse gelatinous lapilli of pure MgSO4 during one explosive eruption. It is possible that water within the body of the rockfall may play a role in initiating an explosive response.

Most lapilli and bombs erupted from the vent during explosive events are scoria with vesicularities as high as 89%, and pyroclasts lofted from the vent at other times have vesicularities as high as 95% (R. Carey, 2012, personal commun.). These products indicate a significant amount of dissolved and/or exsolved gas stored at a shallow level in the lava lake. A massive disruption to the top of the lava lake could abruptly release this gas, and may even trigger a runaway decompressive vesiculation that helps drive the explosive response. This may be further enhanced by the addition of a multitude of vesiculation surfaces provided by the rockfall. However, the observation that the explosive response is largely unaffected by the preexisting condition of the lava lake argues against the release of gas trapped beneath a capping crust as the driving mechanism behind rockfall-triggered bursts.

Another possibility, which we favor, is that the explosive eruptions are a manifestation of a rebound splash caused by the collapse of a rockfall’s impact cavity. Such splashes, called Worthington jets, can occur in both liquids and granular materials (e.g., Worthington, 1908; Lohse et al., 2004; Gekle and Gordillo, 2010, and references therein) and are seen in everyday settings as simple as raindrops impacting a puddle, or a rock being tossed into a pool of water. Studies have shown that Worthington jets can achieve velocities 20–30 times higher than the velocity of the impacting object (Thoroddsen et al., 2004; Gekle and Gordillo, 2010), and that jet height in some instances can exceed the initial drop height (Worthington, 1908; Lohse et al., 2004). Furthermore, higher impact velocities result in higher jets (Harlow and Shannon, 1967; Bergmann et al., 2006). Small jets, formed by tiny rockfalls impacting the lava lake surface, are seen often at Kīlauea (e.g., Movie DR2 during time period 09:41:31–09:41:41). As the size of the rockfall scales up, however, the formation of a larger Worthington jet is not always directly observable because of the introduction of other factors such as occlusion by ash, fragmentation of the jet, and saturation of the webcam video from incandescence of the freshly exposed lava. In general, though, the largest rockfalls triggered the most energetic eruption response, perhaps by creating a larger impact crater and a correspondingly larger Worthington jet. The coherence of the rockfall probably also plays a role in determining the height of the jet, with rockfalls that form a consolidated mass producing higher jets than rockfalls composed of loose debris. The collapse of the impact cavity could also temporarily trap air beneath the surface of the lava lake, similar to what has been observed in some Worthington jet studies (Lohse et al., 2004; Gekle and Gordillo, 2010). This air, especially when heated, could then act as a supplementary driving mechanism.

In the volcanologic scenario presented here, a rockfall plunging into the surface of the lava lake creates a deep impact crater in its wake, compressing the lava column beneath the impact site while decompressing that surrounding the impact site. The size of this crater should depend on the ratio between the rockfall’s inertial force and the surface tension of the lava lake (i.e., the Weber number). The crater then collapses and, if the Weber number is high enough (Hsiao et al., 1988), forms a Worthington jet, which, when it is composed mostly of lava from beneath the lake surface, undergoes immediate decompression and fragmentation. Some smaller lithic clasts are also swept back up in the jet, as seen in jets formed during water impacts (Worthington, 1908; Harlow and Shannon, 1967). The rising fragments that compose the jet then travel through the lithic ash plume, incorporating additional rockfall material. Evidence of these interactions between the jet and the lithic rockfall is preserved in the tephra, which shows a thorough mixing of lithic and juvenile components and a coating of lithic ash to fine lapilli (Fig. 2D; Wooten et al., 2009). The formation of the Worthington jet is probably coupled with several mechanisms, including decompressive vesiculation and the release of trapped gas, that together drive the transient explosive eruptions triggered by rockfalls. We believe, however, that the formation of a rebound splash plays the primary role in this complex process.


Geologic observations have been crucial in developing our understanding of vent evolution and shallow degassing processes at Kīlauea. These observations show that rockfalls within the vent, when they impact the top of the lava column, are capable of triggering transient outgassing bursts and weak explosive eruptions, possibly by generating a Worthington jet that may be closely coupled with other processes such as decompressive vesiculation and release of trapped gas. This type of triggering mechanism has not been recognized before, and it seems likely that other small, enigmatic explosive events reported at Kīlauea over the past century, and possibly at other open-system basaltic volcanoes, could have been triggered in a similar fashion.

Network-enabled webcams have become an important monitoring tool at volcano observatories around the globe in recent years and provide another way to study volcanic activity and its associated hazards (e.g., Paskievitch et al., 2010). The study here highlights the importance of using observational data, when available, to constrain eruption mechanisms based on geophysical interpretations.

We thank the staff and volunteers at the Hawaiian Volcano Observatory (HVO) for observations and discussions related to this research. HVO’s technical staff worked tirelessly to ensure continued operation of the webcam network used in this study. Constructive reviews by R. Hoblitt, M. Ripepe, and S. Valade improved this manuscript immensely.

1GSA Data Repository item 2013051, Figure DR1 and Movies DR1–DR5, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.