Dynamics and style transition of a moderate, Vulcanian-driven eruption at Tungurahua (Ecuador) in February 2014: pyroclastic deposits and hazard considerations

Abstract. The ongoing eruptive cycle of Tungurahua volcano (Ecuador) since 1999 has been characterised by over 15 paroxysmal phases interrupted by periods of relative calm. Those phases included one Subplinian as well as several Strombolian and Vulcanian eruptions and they generated tephra fallouts, pyroclastic density currents (PDCs) and lava flows. The 1 February 2014 eruption occurred after 75 days of quiescence and only 2 days of pre-eruptive seismic crisis. Two short-lived Vulcanian explosions marked the onset of the paroxysmal phase, characterised by a 13.4 km eruptive column and the trigger of PDCs. After 40 min of paroxysm, the activity evolved into sporadic Strombolian explosions with discrete ash emissions and continued for several weeks. Both tephra fall and PDCs were studied for their dispersal, sedimentology, volume and eruption source parameters. At large scale, the tephra cloud dispersed toward the SSW. Based on the field data, two dispersal scenarios were developed forming either elliptical isopachs or proximally PDC-influenced isopachs. The minimum bulk tephra volumes are estimated to 4.55 × 106 m3, for an eruption size estimated at volcanic explosivity index (VEI) 2–3. PDCs, although of small volume, descended by nine ravines of the NNW flanks down to the base of the edifice. The 1 February 2014 eruptions show a similar size to the late 1999 and August 2001 events, but with a higher intensity (I 9–10) and shorter duration. The Vulcanian eruptive mechanism is interpreted to be related to a steady magma ascent and the rise in over-pressure in a blocked conduit (plug) and/or a depressurised solidification front. The transition to Strombolian style is well documented from the tephra fall componentry. In any of the interpretative scenarios, the short-lived precursors for such a major event as well as the unusual tephra dispersion pattern urge for renewed hazard considerations at Tungurahua.


Introduction
In comparison to Strombolian or Plinian fallout deposits, Vulcanian fine grained fallout deposits often lack a sufficient preservation potential for extensive studies from proximal to distal portions (Rose et al., 2008). Generally, these eruptions produce 20 show the progressive deepening of the source in the magma conduit, brittle fragmentation of highly viscous magma and high explosivity (e.g., Clarke et al., 2015). The ejected tephra thus help to constrain the source depth and intensity of the eruption.
Vulcanian eruptions can be associated with the generation of pyroclastic density currents (PDCs, e.g., Brown and Andrews, 2015), especially if they trigger moderate andesitic eruptions, which are more prone to produce PDCs than other eruption types (Bernard et al., 2016). Although the transitional behaviors of Vulcanian eruptions from/to Subplinian and Strombolian erup- 10 tions are occasionally observed (Maeno et al., 2013), these changes in the eruptive style are not widely described and claim attention, in particular for volcanic hazard assessment.
This study focuses on the pyroclastic deposits associated with the eruptions from 1-14 February 2014 of Tungurahua volcano (Ecuador). Tephra fallout and PDC erupted material are described and used to infer the style transition, eruption magnitude and source parameters. Special attention is given to explain the unusual rapid evolution from unrest to eruption, and the possible 15 existence of a plug in the conduit. The deposits and flow of PDCs are also described and discussed. The meaning of this eruption is put in perspective with the context of Tungurahua's ongoing cycle.

Sampling and analysis of PDC deposits
The PDC deposits were observed in the field between the 08 and 25 February. Twenty-two samples from these deposits were sieved mechanically in 0.5 φ steps with a shaker from Retsch. The sieving protocol was 15 min shaking in intervals of 20 sec 5 steps with a brief pause between each step for the whole fraction. The fraction < 0.5 mm was additionally shacked for 5 min in 10 sec intervals. This protocol was chosen to ensure the lowest amount of clast-abrasion during sieving. Further, the fine fraction (< 125 µm) was analyzed with a laser-diffraction particle-size-analyzer (LS230 from Coulter). For this, each sample was measured five times with three consecutive runs, with the final result taken as the average of all runs.
3 Deposits of the 01 February 2014 eruption 10

Dynamics characteristics
Over the area of Tungurahua, annual wind directions at an altitude of 15 km a.s.l. mostly blow toward the NW with speed magnitudes of 0 to 20 m/s whereas the January to March tendency is NW for speeds of 0 to 10 m/s, but mostly E for speeds of 10 to 20 m/s (Fig. 3a, 3b). In all the time series, less than 1% correspond to S-SW directions (180 to 210 • azimuth). The Washington Volcanic Ash Advisory Center (VAAC) reported a column height of 13.72 km after 22:45 UTC on 01 February 15 2014 (http://www.ssd.noaa.gov/VAAC/ARCH14/TUNG/2014B020235.html). The tephra cloud was mainly dispersed in a very uncommon pattern toward the SW and S-SW, being divided into two fly levels (<7.6 and >7.6 km a.s.l., Fig. 3c). The column

Distribution and volume
The tephra fallout field study was carried out in mid-February, and thus contain the signature of two weeks of activity (between 01 and 14 February). Fallout distribution is well constrained on land at <20 km from the vent ( Fig. 4a) but scarce data has been obtained in the distal areas (>20 km.) Whole tephra fall deposit mapped covers almost 13 000 km 2 with a nearly elliptic 15 distribution dispersed toward the S-SW. This suggests that most of the distal deposit here studied correspond to the 01 February explosions. The thickest deposit (∼1 cm) was measured in a radius of about 9 km from the crater and following the dispersal axis whereas >1 mm deposit is distributed <30 km around the vent (Fig. 4a). Ash traces (∼0.1-1.0 mm thickness) were identified through the social media geo-referenced photography sent by people to the IG-EPN at distal areas on the main dispersion axis (e.g. Loja, 290 km S). The calculated volume of tephra fall deposit ranges from ∼0.97 to ∼2.37 10 −2 km 3 , 20 depending on the model (Table 2). Using a bulk deposit density supposed to be 760 kg/m 3 , this leads to a total erupted mass between ∼0.74 to ∼1.80 10 10 kg and total Dense Rock Equivalent (DRE) volume from ∼3.0 to ∼7.34 10 6 m 3 (Table 2). In consequence, our mean bulk tephra volume estimate is 1.53±0.35 10 −2 km 3 (4.76±2.23 10 6 m 3 DRE).
PDCs branched into at least 9 valleys from the N, W and SW flanks of the volcano and reached the base of the edifice in several locations, endangering the main road of the area (Fig. 4b). According to Hall et al. (2015) the deposit lengths and   (Table 1). approximate values for channel widths and deposit thickness return bulk volume estimates of ∼1.2 10 6 m 3 , yet this is based on constant thicknesses extrapolated from final lobes and may be exaggerated.

Morphology of PDCs
Typically, sediment morphologies showed levees made of large clasts producing self-channelization with depleted inner channels in running parts, whereas lobes with tongue shapes developed at final local runout distances ( Fig. 5-6). Interestingly, lobes could form even in very steep setting (>25 • slope, e.g. Juive-viejo-minero Fig. 5d), whereas they could overrun long flat distances in other places (e.g. Pondoa). Two final lobes are found <1 m from a cliff and on a very inclined bed (Vascún 5 and Juive-Viejo-minero, Fig. 5a and 5d). The Juive-Grande final lobe is found at almost right angle to the general slope and deviated in an artificial trench created to protect the main road from lahars, proving its efficiency for slow currents (Fig. 5c).
PDCs travelling in the Hacienda valley were confined in a very narrow ravine (<10 m) were they could reach very low parts of the cone (Fig. 5e). They crossed the main road below the bridge without any damages. There, a lobe shape is scarred by a re-incision, the surface of the scar having a distinct color due to the absence of a final ash draping ( Fig. 5e). At Achupashal the 10 most distal deposits consists of two superposed PDC lobes of contrasting componentry and grain size distribution ( Fig. 5f).

Temperature and disturbance from PDCs
Relatively low temperatures (40-170 • C) were measured between 8 and 14 days after the eruption at depth down to 1 m in several valleys (Table 3). No pattern could be recognized in the temperatures regarding PDCs' observed chronology, componentry, or morphology. For all PDC deposits, even when no thermometer was available, a warm temperature was manually checked to 15 ensure the primary and recent nature of the deposits.
In some areas close to the PDC pathways, vegetation was affected and found with dead leaves in the days to weeks following the eruption. This contrasted with early observations (1 day after eruption) documenting leaves covered by ash but not burnt (Yepes, pers. comm.). Zones on outer curves of PDCs pathways were affected with dead leaves on a much wider zone than in inner curves ( Fig. 7). In the steep-sided, narrow valley of Juive-Pampa, PDCs got highly constrained in an edgy curve

Stratigraphy and lithologic components
The composite stratigraphic column of tephra fall deposit consists in three recognizable layers (Fig. 8). Even when this full sequence is observed (e.g. at Palictahua, sample P1), the sequence remains incomplete in most outcrops. The basal layer (1) is a grayish to reddish fine-grained ash deposit (<1 mm in thickness. Above, layer 2 is thicker (1-8 mm) and formed 30 by gray and coarse-grained ash deposits and lapilli sized fragments. Lapilli-sized, porphyric, andesite, dense and dark clasts  (Px). Dark-gray to reddish scoria fragments represents the second most abundant clast type (14.5-16.8 vol.%, Fig. 8). They are blocky, sub-rounded to sub-angular, with a vitreous matrix and high presence of non-elongated irregular-shaped vesicles.
Light-to-gray pumice is also abundant (15.5-15.9 vol.%), being sub-angular in shape, with moderately microvesicular texture made of elongated to sub-spherical vesicles (<1 mm, Fig. 8). The matrix includes nailed Pl phenocrysts reaching 1 mm and minor presence of Px. This layer also contains a subordinate altered lithics fraction (5.5-5.9 vol.%, Fig. 8) and scarce volcanic 5 glass (<1 vol.%, Fig. 8). Ultimately, the top layer (3) is very thin (<1 mm) and consists of Px and Pl crystals and abundant (>50%) juvenile volcanic glass (Fig. 8). It also comprises reddish, non-vesicular, dense altered fragments and white particle aggregates. The same type of samples were obtained in-situ (during the ash sedimentation over plane surfaces) at Pillate, after discrete explosions in mid-February 2014. At greater distances from the vent (e.g. Penipe, 14.3 km southwest), volcanic glass is a major component. It has a curviplanar surface and sharp morphology, in some cases of "shard" type, transparent to semi-     The PDC deposits include blocks and bombs grouped in four types (Fig. 6b): 1) dense, seemingly-glassy (micro-cristalline) clasts with pervasive fracture pattern, 2) dark to greenish, porous, cauliflower-shaped, glassy clasts, 3) light-gray, micro vesicular, bread-crusted clasts with dense, fractured margins up to 3 cm thick, and 4) accidental lithics. According to Hall et al. (2015) 5 the February eruption PDC deposits stratigraphy is segregated into a poorly-developed top layer compound by both sub-angular to sub-rounded clasts of black vesiculated andesite (juvenile andesite), as well as subordinate number of gray dense andesite clasts, and a fines-rich lower layer is dominated by dense andesite clasts whose angular clasts have micro-fractures and chilled margins. Whereas this sequence can occur in some areas, many flows can be superimposed in other zones (e.g. > 4 successive units in Juive-Pondoa). Individual lobes show very variable aspect in terms of largest clast size, surficial fine content, and most 10 frequent type of clast encountered ( Fig. 6c-d). No tendency could be encountered with runout distance or eruption chronology.
The superficial differences in content of fines is however absent in subsurface, where it seems constant between lobes with varying surface signatures. This suggests that the superficial content is a simple effect of local winds during emplacement. The size of the largest block fraction can vary from ca. 5 cm in some lobes to >40 cm in others (Fig. 6c-d).

Grain size 15
Four samples collected from tephra fall deposits were sieved for grain size analysis. Samples B1 and C1 show unimodal distribution, whereas C2 and SJ are bimodal and trimodal, respectively (Fig 9).
Sample B1 is composed of coarse to very fine ash (mode at 2.66 φ), with moderate sorting (according to the classification of Folk and Ward, 1957) or very well sorted (according to Cas and Wright, 1987) and with a coarse-skewed distribution.
Sample C1 is mostly made up of coarse ash (mode at 1 φ), showing the same sorting as B1, and very fine skewed distribution.  to well sorted (depending on the classification applied: Folk and Ward, 1957;Cas and Wright, 1987, respectively), and very coarse skewed.
Sample SJ has a singular trimodality (3.73, 2.5 and -2.16 φ). It contains a low amount (<15 wt.%) of coarse material (medium lapilli to coarse ash), a notable proportion (ca. 60 wt.%) of medium to fine ash, and less very fine ash (ca. 25 wt.%), being poorly sorted and coarse skewed. All the samples are leptokurtic (i.e. better sorted than a standard deviation).
5 Figure 9. Grain-size analysis for samples from tephra fall collected around Tungurahua two weeks after eruption All samples from PDCs present a main mode at 125 µm (3 φ), with lesser secondary peaks at 250 (2 φ) and 500 µm (1 φ) ( Fig. 10). Most samples follow a very consistent trend (black samples), with three samples slightly enriched in their fraction 125 and 90 µm (3.5 φ; cyan), yet this seems to only correspond to a minor measurement deviation and is not reflected in other fractions. Three samples have clearly distinct signatures switched towards the fines (red, blue, magenta). Two of these correspond to overflow sampling areas from co-PDC clouds rather than dense PDCs deposits (Romero T104 and Pampa T10).

15
The variability in the amount of ash visible at the surface of the PDC deposits was previously interpreted as representing two stages of the eruption. Although there may be more fines in basal units compared to the top layer in some particular cases, this is not systematic at all, and seems to be due to wind during emplacement rather than reflecting the eruptive dynamics. This is further supported by the video footage of the eruption showing that co-PDC clouds were rapidly drifted by the wind in some ravines (C.F. Diego, youtube). Sample T110 from the basal layer of Pondoa zone illustrates this further, since this PDC deposit 20 surface was covered by at least 3 further co-PDC clouds, and thus has a co-PDC cloud signature in its superficial grain size.
The variability in componentry and aspect of the final lobes and channels might have a signature from the eruption dynamics, yet the first emitted flows remained deposited on the proximal Achupashal drainage, and were likely re-entrained during the main-PDC flows. It seems that the Vascún and Juive drainage flows have more dense fractured blocks whereas the western deposits are richer in cauliflower and altered clasts. This would go against the eruption dynamics chronology with initial 25 plug destruction, since the first PDCs were emitted in the proximal Achupashal, and then the onset of the third explosion triggered PDCs in the SW ravines (Rea and Romero), and later in the Vascún, Juive, lower Achuapashal and Hacienda almost simultaneously (Fig. 4).
No correlation of the grain-size curves with runout distance, componentry and morphology was observed, neither to any path-dependent abrasion and thus the grain size signature seems inherited from conduit fragmentation only. Two patterns are 30 found from the grain size data. The "dense PDC" pattern is fairly homogeneous, and no trend was identified. It plots close to the fields of "dilute PDCs" from the 2006 eruption of Tungurahua (Douillet et al., 2013b, Fig. 10e). The "co-PDC cloud" pattern has more variability, and plots at the limit of the "fall" field of Walker (1971). Thus, the grain size is mainly a result of the transport process, dominated by elutriation from the dense PDCs and subsequent suspended transport.
It is noteworthy that neither the characteristics of the PDC deposits nor the thermal and color videos from the monitoring network (www.ig-epn.edu.ec) show any presence of lateral blast. Contrarily to Hall et al. (2015), we rather interpret the origin of the PDCs as a result of flank loading from a vertically erupting column, and subsequent destabilization of the material accumulated on the steep upper flanks, a mechanism quite similar to the 2006 PDCs (e.g. Kelfoun et al., 2009).
Volume calculations of PDCs and incandescent bombs over the flanks of the volcano estimated by Hall et al. (2015) are about 5 5.7 10 6 m 3 . These measurements seem to be overestimated since they consider thicknesses at lower (thicker) emplacement areas. In facts, the thickness of the PDC deposits for the 2014 eruption is fairly unequal. Most of the material seems to be accumulated in discrete places and terminal lobes, but the majority of the pathway ravines were almost empty of PDC deposits, and only had levees, some even showed signs of erosion (and thus flow bulking). Field work was carried out before any rain occurred, and thus there was no secondary transport by lahars. The first implication is that previously extrapolated PDC volume 10 based on the thickness of terminal lobes is probably fairly exaggerated.
Most importantly, the February 2014 PDCs illustrate a new danger of the ongoing eruption: even with a small volume, PDCs are able to travel further and further. This character is likely due to the fact that ravines have been stripped and eroded by 15 years of lahars and PDCs. There are no more obstacles to PDCs' flow, no trees or natural dams. Thus even of small volumes, future PDCs at Tungurahua are likely to reach the lower inhabited areas both easier (more frequently) and faster.

A plug-driven onset evolving into an open conduit eruption
Presence of disequilibrium textures in Pl (complex zoning patterns and resorption features) are usually interpreted as changing physical conditions in magmatic systems due to rapid decompression and/or magma mixing/mingling adding heat or mass (Nelson and Montana, 1992). The phenocrystal assemblage of the February eruption samples consisting in Pl and Px represents the liquidus phases in equilibrium with the melt. As no reaction or disequilibrium textures were found, a rapid rise in liquid 20 and crystals is required, in contrast to the 2006 eruption, where clear disequilibrium textures have been documented in Pl and Cpx phenocrysts (Samaniego et al., 2011). This scenario implies that the February magma was not re-heated or mixed. At the ejection level, a quick quenching process, rock fracturing and eruption is able to generate these andesite blocks and bombs included in the PDCs, without any evidence of interaction between two melts inside the magmatic reservoir. The geophysical background of inflation (Vallejo et al., 2014) and striking seismicity 48 h before the eruption onset are probably triggered by the 25 pressurization of the upper chamber due to the presence of a plug in the conduit (Fig. 12a). This rapidly evolved into a volcanic unrest and eruption. According to the direct observations of the eruption, we can infer that the plug failure was progressive, starting with the first explosion at 22:12 UTC (Fig. 12b) and then evolving into a total plug destruction at 22:39 UTC (Fig. 12c) during the paroxysmal phase of explosive activity. This Vulcanian mechanism is also supported by the componentry of tephra fallout, and has already been described for previous eruptions at Tungurahua (e.g. Bustillos et al., 2016;Parra et al., 2016).

Volume and style
We suggest the Weibull volume calculation (Bonadonna and Costa, 2012) as the most realistic, due to its smaller mean relative squared error (∼0.065) if compared to the other methods (Table 2). This method gives a volume estimate of 1.53 10 7 ± 0.35 10 7 m 3 .  (Table 2). Following the physical model of eruptive column in Sparks et al. (1997), the maximum column height (HT) would range from 10.4 to 13.1 km in height above the crater (Table A1), in agreement with the height reported by the VAAC (14 km a.s.l.).
Respective Magnitude and Intensity calculated following Pyle (2000) yield ∼3 and 9.5-11.0 ( Table 2). The released tephra 10 volume and maximum eruptive column height are consistent with a Volcanic Explosivity Index (VEI) of 3 (Newhall and Self, 1982). The column height and intensity suggest that peak activity resulted in a small Subplinian eruption (in the Cioni et al., 2015, classification). In contrast, the components of tephra deposits reflect a Vulcanian style. Although the tephra fall deposit has not been sampled in the current investigation, we suspect that the mechanism of the 04 April 2014 eruption was roughly similar to that shown by the 01 February 2014, based on the event sequence and shortlived unrest period. A value of 2.51 10 7 kg/s MDR is obtained for the HT column of the April eruption with the Sparks and 15 Walker (1977) equation. Applying this MDR for an eruption duration of 9 minutes yields a total erupted mass of 1.36 10 9 kg, which represents between 9 and 14% of the total mass released during the February eruption studied here. Although the April eruption was smaller in size, the repeated Vulcanian mechanism with little warning observed at Tungurahua requires to be assessed because it augurs little forecast timing.

20
Setsuya Nakada and another reviewer have contributed to improve this manuscript. This study is financially supported by the Deutsche Forschungsgemeinschaft grant DO1953/1-1 and Baylat CoCotE grants to GAD.
Appendix A: Calculation of erupted parameters Table 2 gives the results of calculation following three interpolation methods, with details below and used parameters in Table A1:

25
-Exponential thinning volume is calculated through the adjunction of the volumes calculated for the Segments 1 and 2 following the method of Pyle (1989): V = c * e (−mx) -Weibull method used the Weibull function integration (Bonadonna and Costa, 2012): V = 2((Θ * λ 2 )/η) -Power law is following the approach of Bonadonna and Houghton (2005). TPL and m are coefficient and exponent of the power law. Here, we used as C=2 and B=600 as both proximal and distal limit of integration: T (x) = T pl * A (−0.5m) ;