Abstract

Understanding the dynamics of ongoing volcanic eruptions is essential for predicting the input and transport of volcanic ash in the atmosphere. To constrain near-vent dynamic processes of explosive Vulcanian events, we used Doppler radar measurements, providing tephra velocities and a proxy of the mass flux, in two field experiments at Volcán de Colima (Mexico) and Santiaguito (Guatemala). We find that explosive eruptions at both volcanoes consist of individual degassing pulses. The analysis of the timing of such pulses shows that both volcanoes have preferred interpulse times of 3 s (Santiaguito) and 2–5 s (Volcán de Colima). The interpulse time during one event may change, but it often returns to the preferred interpulse time. This behavior is similar at both volcanoes and the interpulse time roughly follows a log-logistic distribution indicating the interplay of two competing processes. These could be short-duration degassing of the uppermost conduit versus decreasing permeability due to progressive gas loss and compaction during an eruption.

INTRODUCTION

With the advent of high-speed measurement techniques, like video cameras and Doppler radars, more and more evidence is being gathered that the ejection of volcanic gases and tephra is a highly unsteady process, meaning that it is either transient (short duration) or pulsed (fluctuating) or even both. Pulsed release of tephra and gas has been observed at volcanoes like Etna (Italy; pulse periods of 1.8 s [Tazieff, 1970] and 3.8–5.5 s [Dubosclard et al., 2004]), Stromboli (Italy; Scharff et al., 2008; 0.2–3 s [Taddeucci et al., 2012]), Yasur (Vanuatu; 0.5–3 s [Gaudin et al., 2014]), and Arenal volcano (Costa Rica; 3 s [Donnadieu, 2012; Valade et al., 2012]). Pulses have also been observed during Vulcanian eruptions at Galeras (Colombia; Stix et al., 1997) and Cordón Caulle (Chile; Schipper et al., 2013), yet a systematic study of the timing of such pulses has not been carried out. Video analysis of large-scale (conduit-clearing) Vulcanian events at Soufrière Hills volcano (Montserrat) in A.D. 1997 (Druitt et al., 2002; Formenti et al., 2003) revealed the pulsatory nature of those events; however, a further analysis of pulsed vent exit velocities and pulse timing remained qualitative because the eruption cloud covered the near-vent dynamics of each jet. The eruption clouds from individual pulses merge within the first few hundred meters above the vent(s) and from then on become visually inseparable.

Using a Doppler radar that penetrates the eruption cloud to measure tephra velocities within the column, Scharff et al. (2012) reported that comparatively small-scale (dome-preserving) Vulcanian events accompanying dome extrusion at Santiaguito (Guatemala) also consist of individual explosions and that these pulses occur at an almost regular interval of ∼3 s (Scharff et al., 2014). In this paper, we explore whether the observed periodicity of pulses is specific to the eruption dynamics at Santiaguito or if it also applies to at other volcanoes exhibiting Vulcanian activity during dome growth episodes, such as Volcán de Colima in Mexico.

EXPERIMENTAL SETUP

Volcán de Colima (VdC) is located in the western part of the trans-Mexican volcanic belt (Fig. 1A). Its activity is characterized by near-centennial Plinian or sub-Plinian eruptions and frequently recurring cycles of dome extrusion and destruction (Luhr and Carmichael, 1980; Luhr, 2002). Typically, the extruding magma is significantly degassed, dense, highly viscous, and permeable (Kendrick et al., 2013, and references therein), hence favoring dome activity. Recently, two effusive episodes have occurred with the construction of domes and periods of small-volume rockfalls (Mueller et al., 2013), as well as explosions with highly varying ash contents (Webb et al., 2014). The first episode commenced in January 2007 and was characterized by a very slow extrusion rate (∼0.02 m3/s; Hutchison et al., 2013). It terminated abruptly in June 2011 with an explosion that destroyed parts of the dome (James and Varley, 2012). The second episode started in January 2013, has had a highly variable extrusion rate, and is ongoing (as of August 2015).

During our experiment (March to July 2007) the Doppler radar was installed on the south flank of VdC (Fig. 1B) at ∼2550 m above sea level (asl) at ∼3000 m slant distance from the crater rim, pointing into the sky ∼150 m above the dome (Fig. 1C). The system (Doppler radar and data logger) was also used at the Santiaguito complex (SaC) during the experiment in January 2007 (for more details, see Scharff et al., 2012), where the instrument was placed at the summit of Santa Maria (3772 m asl) enabling an insightful view from above on the near-vent dynamics of the Caliente dome (2550 m asl, slant distance ∼2700 m). The Doppler radar measures velocities of tephra particles moving through the radar beam at a high sampling rate of ∼15 Hz, as well as the amplitude of the reflected radar signal. The resulting velocity spectra (shown in a so-called velocigram) can be interpreted as the amount of material moving at distinct velocities within the field of view (comparable to a single circular ∼80-m-diameter pixel; Figs. 1C–1E). For more details see Scharff et al. (2012) and references therein.

RESULTS

Volcán de Colima, Mexico

Photographic images as well as the velocigram from the radar data of a Vulcanian explosion at VdC on 2 March 2007 are shown in Figures 1D and 1E. The radar data (Fig. 1E) show phases of high signal amplitude (echo power) that represent moving mass. These alternate between high positive and high negative velocities indicating short phases of rising (positive velocities) and falling (negative velocities) particles. The gradual deceleration (transition from positive to negative velocities) is dominated by gravity but also influenced by air drag and wind (Scharff et al., 2012). In contrast, the transition from negative to positive velocities is abrupt without any acceleration phase (white V in Fig. 1E) and thus indicates a new explosive release of particles that have been accelerated below the field of view.

Over the almost continuous measurement between 2 March and 18 July 2007, a total of 91 events comprising a total of 856 pulses were recorded with the Doppler radar. To ensure high accuracy, starting time and corresponding maximum velocity of all pulses were picked by hand.

The number of pulses per event is highly variable (Figs. 2A and 2B) with up to 48 pulses in one event. The majority (64%) of events comprise 2–10 pulses (Fig. 2B). The time intervals between pulses (interpulse time; IPT) are short and mainly vary between 1 s and 5 s (Fig. 2C), but may last for up to 100 s (Fig. 3C). A large number of pulses occurs preferentially after short repose times (<2 days; time between events detected by radar), whereas the number of pulses decreases for longer repose times (>6 days; Fig. 2D). We could speculate on a monthly cycle of high-pulse-quantity events (Fig. 2A).

Pulses occur in 89% of events, and longer events comprise more pulses. In Figure 3A, pulse timing within each event is plotted against the cumulative number of pulses for that event. In this plot, a straight line through several symbols marks a sequence of constant IPT, and the slope of the line gives the recurrence period. With this simple relationship we find pulse recurrence periods between 3 s and 20 s. While the IPT may change during one event, it is often almost constant for several consecutive pulses. The maximum tephra particle velocity in each pulse—a proxy for the gas velocity (Dubosclard et al., 2004)—is lower for the first pulse (Fig. 3C), but no general trend for consecutive pulses exists (not shown).

Santiaguito Dome Complex, Guatemala

The pulse statistics for the SaC data set are shown for comparison in Figures 2 and 3. In that experiment, the Doppler radar recorded a total of 136 events from 9 January to 13 January 2007. Scharff et al. (2014) found a dominant pulse repetition frequency of ∼0.3 Hz using automatic pulse detection based on image cross-correlation. This method provides an average pulse recurrence period for each event, but it does not reflect true IPT (Figs. 2C and 3B). Therefore manual re-picking has been carried out, which revealed 977 pulses with a mean IPT of 4.4 s and a median of 2.9 s.

DISCUSSION

The activity at both volcanoes has been recently characterized by the extrusion of viscous magma and the building of a lava dome, with intermittent Vulcanian explosions that produce plumes of variable tephra contents. At VdC it has been observed that the tephra content can greatly vary between pulses (demonstrated for example by a white plume extruding through a gray cloud). Given that pure gas pulses are transparent to radar, some pulses may have been missed, leading to higher apparent IPT.

The number of pulses per event at SaC (Fig. 2A, inset) varies as much as at VdC, however the time scales of both plots greatly differ: the 136 events at SaC were recorded in 4 days, compared to 91 events in 4.5 months at VdC. The different event rates (34 events per day at SaC versus 0.7 at VdC) may partly be attributed to the slower effusion rate (Massol et al., 2001) at VdC (∼0.02 m3/s; Hutchison et al., 2013) compared to SaC (∼0.43 m3/s; Ebmeier et al., 2012). The less frequent events at VdC are more energetic in terms of jet exit velocity (<110 m/s vertical at VdC versus <50 m/s vertical at SaC; Fig. 3C) indicating more pressure buildup between eruptions. Nonetheless, both volcanoes show a similar activity pattern with regard to number of pulses per event (Fig. 2B), temporal variability (Fig. 2A), and IPT (Fig. 2C). Therefore the general observation of pulsed activity appears to be independent of effusion rate at least at these two dome-building volcanoes.

Vulcanian events at SaC (gray bars in Fig. 2B) comprise on average more pulses, but the distribution is narrower (77% comprise 2–10 pulses versus 64% at VdC). The IPT at SaC is tightly constrained to 3 s (long straight-line sections with constant slope in Fig. 3B). Phases of constant IPT can be interrupted by longer IPT, but the IPT often returns to a preferred 3 s period that has been interpreted as oscillations of the dome (Scharff et al., 2014). At VdC, the IPT is also stable for several consecutive pulses and does often return to the same IPT throughout an event.

The measurement geometry in the SaC experiment differs from that of the VdC experiment in that it was possible to measure velocities directly at the vent from above. Therefore the pulses at SaC must have their origin within the conduit below the surface so that pulse dynamics can be directly related to conduit dynamics (Scharff et al., 2014). At VdC, the radar field of view roughly covers a region between 100 m and 200 m above the dome surface, hence not all measured fluctuations in echo power and velocity may be directly linked to dynamic processes within the conduit. Atmospheric processes like entrainment and convection begin to dominate the jet-like dynamics at some height above the vent (Chojnicki et al., 2014) and may overprint individual pulses. Nevertheless, there are strong similarities in the IPT distributions of SaC and VdC. At both volcanoes, consecutive pulses preferably follow each other within seconds.

Following the regression method of Watt et al. (2007) to fit probability distributions to data (using so-called probability plots), we find that both IPT data sets are best described by the log-logistic distribution (R2 > 0.97, SaC and VdC) instead of a Weibull distribution that describes a simple failure model (R2 = 0.82 at VdC and R2 = 0.78 at SaC). The log-logistic probability distribution describes a system with two competing processes (Connor et al., 2003). At a volcano, these may be related to internal pressure buildup and release or two processes that act to increase (gas venting) or decrease magma viscosity (shear heat generation).

SUMMARY AND CONCLUSION

The Doppler radar data sets of explosive jets at VdC (Mexico) and the SaC (Guatemala) show inter-eruptive pulses, indicating that the mass flux through their vents is not steady. Manual picking of pulse onset times revealed that the IPTs of both volcanoes are often constant for several pulses and that both volcanoes return to their preferred IPT (3 s at SaC and 2–5 s at VdC) after phases of longer IPT.

Compared to our work at SaC (Scharff et al., 2014) where the Doppler radar was used to observe the dome surface from above, we cannot infer a mechanical model for the generation of pulses at VdC. Nonetheless, we interpret the similarities in pulse statistics as an indication for a common pulse generation mechanism: the interplay of pressurization within the dome, related to the formation of tuffisite veins commonly observed in the dome material, and the rhythmic opening and sealing of pathways during explosive degassing events (Kolzenburg et al., 2012; Schipper et al.,2013; Lavallée et al., 2014). The apparent periodicity may originate from some type of oscillation of either solid material within the conduit (e.g., vertical [Scharff et al., 2014] or horizontal [Jellinek and Bercovici, 2011]), or gas pressure–induced opening of a valve, resembling a pressure cooker (Lees and Bolton, 1998) or a clarinet (Lesage et al., 2006).

Pulses have also been observed at other types of volcanoes (e.g., Etna, Stromboli, and Arenal), and their occurrence seems to be independent of magma composition. This strongly suggests that mass flow out of volcanic conduits into the atmosphere is by no means continuous or steady but an inherently time-dependent process. Chojnicki et al. (2014) highlighted the effects of transient mass flux on the dynamics of a single jet. However, the consequences of pulsed mass flux for the dynamics, stability, and rise height of the emerging contiguous eruption clouds still need to be explored.

We gratefully thank all students at the CIIV (Centro de Intercambio e Investigación en Vulcanología, Colima, Mexico; Hester, Erin, Andrew, Gemma, Adi, Laura, and Martin) for their great help in setting up the instruments at VdC. Special thanks go to John Stevenson and Florian Ziemen who maintained the station during the experiment. Scharff acknowledges funding from DFG project Ho 1411/20-1. We thank Yan Lavallée and four anonymous reviewers for their help in improving earlier versions of this article.