Tephra fallout from the long-lasting Tungurahua eruptive cycle (1999-2014): Variations through eruptive style transition and deposition processes

The Tungurahua volcano (Northern Andean Volcanic Zone) has been erupting since 1999, with at least four eruptive phases up to present. Although a dozen of research focuses in tephra fall deposits during this period, none of them cover the full eruptive cycle. We investigated the eruptive mechanisms and tephra fall deposition processes at Tungurahua between 1999 and 2014, through systematic analyses of tephra samples collected westward of the volcano using mechanical sieving grain size analysis, lithology, scanning electron microscopy, X-Ray f luorescence and X-Ray diffraction. Tephra is compounded by varying amounts of scoria (black and brown), lithics, hydrothermally altered fragments, pumice, glass shards and free crystals. Textural analyses of juvenile grains (scoria, pumice and glass shards) revealed a diversity of features concerning to their vesicularity, shape and surface/perimeter. Initially, tephra was characterized by hydrothermally altered fragments related to a phreatic phase which then evolved to a pure magmatic activity with Strombolian eruptions. A homogeneous andesitic composition was observed between 1999 and 2003; however silicarich compositions occurred later in 2006. Similarly, the mineral assemblage contained plagioclase, pyroxene and olivine, but magnetite and akermanite were then included during 2006, thus indicating the eruption of a new, probably mixed magma. As consequence, Plinian activity occurred in August 2006. Further activity in 2007 ejected notable amounts (4065%) of recycled material during Vulcanian eruptions. New eruptions occurred between 2008 and 2010, and juvenile ash revealed the interplay between brittle and ductile fragmentation through ash explosions, jetting events and Strombolian activity. The activity between 2010 and 2012 incorporated hydrothermally altered material at time that eruptive silences became longer and frequent, thus suggesting the development of a sporadic hydrothermal system. Finally, between 2013 and 2014 a series of Vulcanian events occurred. Observed grain size distributions allow us to propose three different processes occurring during tephra deposition: 1) deposition of multiple ash plumes, 2) contributions from elutriated pyroclastic density currents or grain size mixing due to major eruptions, and 3) the aggregation of particles due to rain and/or lighting. From mineralogy and grain size we infer that exposition to ash may produce acute human health effects.


Introduction
Studies of geochemical, mineralogical and lithological properties of tephra fall deposits are currently used as robust analytical tools for the complementary description of eruptive styles and processes (Cioni et al., 2008;Tsunematsu and Bonadonna, 2015).As inherited characteristics from these eruptive processes, the shape and nature of the main constituents of volcanic ash reflect explosive mechanisms of volcanic eruptions, and their transition in time contribute to forecast the changes of ongoing eruptive events (Dellino and Volpe, 1995;Cannata et al., 2014;Clarke et al., 2015;Taddeucci et al., 2015).Besides, determination of tephra grain size distribution (GSD) provide important insights into fragmentation mechanisms, eruptive conditions, tephra deposition and volcanic hazards at active volcanoes (e.g., Walker, 1971;Carey and Sigurdsson, 1982;Bonadonna et al., 2002;Houghton and Carey, 2015).In fact, understanding the eruption mechanisms is critical for improving hazard assessment during volcanic crises.This is especially important at longlasting eruptions for the recognition and mitigation of evolving volcanic hazards through time.
Tungurahua volcano (01º28' S; 78º27' W, 5,019 m a.s.l., Fig. 1) is an active andesitic stratovolcano at the Northern Andean Volcanic Zone (NAVZ; Kley et al., 1999), distant 120 km south of Quito (Ecuador), with a record of five post-Columbian eruptive cycles (1641, 1773, 1886, 1916-18 and 1999 to the time of writing).Since 1999 and up to 2014, Tungurahua volcano has erupted ca.0.13 km 3 of bulk tephra, frequently dispersed towards its west flank, through various types of volcanic activity in four major eruptive phases (Bustillos et al., 2016).Several studies with different techniques have included the analysis of these tephras in very specific time periods such as those of 1999-2001, 2006, 2010, 2012, 2013, and 2014(e.g., Le Pennec et al., 2004;Ruiz et al., 2004;Troncoso et al., 2006;Bustillos, 2010;Bustillos et al., 2011;Le Pennec et al., 2012;Eychenne et al., 2012;Bustillos et al., 2013;Bernard et al., 2013;Eychenne et al., 2013;Parra et al., 2015;Romero et al., 2017).Currently, there is not a continuous database of tephras between 1999 and 2013 in Tungurahua, doing harder to establish their variations through eruptive-style transitions and their related feeding mechanisms.In the current contribution, we study the temporal variations of tephra fallouts as consequence of changing eruption mechanisms between 1999-2014 at Tungurahua volcano.We also look at the amount of fine ash ejected by the volcano for health hazard considerations.Overall, our work provides a comprehensive dataset on the characteristics of ash erupted by Tungurahua during fifteen years of eruptive activity.

Phase I (1999-2005)
The reawakening of Tungurahua was progressive between August and October 1999, with a phreatic vent-clearing onset followed by magmatic activity in mid-October, and then by alternating phases of gas and ash emissions, Strombolian to violent Strombolian eruptions and many canon-like shots associated with short-lived "Vulcanian-like" explosions (Ruiz et al., 2006;Le Pennec et al., 2012).After seven months of quiescence, deep LP seismicity announced a new eruption beginning in late May 2001, which developed small-scale lava fountains and explosions in June and July (Le Pennec et al., 2002).Prior to the August 4 th 2001 eruption, precursory activity was not recorded and surface phenomena consisted of Strombolian to violent Strombolian activity, including a lava fountaining event by August 16 th before the waning of eruption in August 21 st (Le Pennec et al., 2012).New activity dominated by Strombolian phases produced sub-regional to regional ash falls that persisted until 2005 (Mothes et al., 2015).Samaniego et al. (2011) proposed that episodic injections fed magma to a modest reservoir 10 km bellow Tungurahua's crater, which in turn supplied magma to the surface between 1999 and 2005.However, Wright et al. (2012) concluded that variable magma supply rates better explain the transition between Vulcanian and Strombolian styles through an almost persistent eruptive activity.

Phase III (2007-2009)
Beginning in August 2007, heightened seismic activity associated with explosion events and jetting activity was recorded up to the February 2008 eruption, and even a ground uplift (17.5 cm in vertical axis) originated in the upper western flank was visible in interferograms that span the period December 26 th 2007 to March 27 th 2008, probably triggered by a magma intrusion of 1.2 Mm 3 (Biggs et al., 2010).Following the February 2008 Vulcanian eruption, the volcano experienced a period of relative quiescence with a few small explosions and jetting tremors recorded (Biggs et al., 2010).Since 2008, short-to medium-lived Strombolian eruptions have occurred, spanning from a few days to a few weeks, separated by periods of quiescence ranging from two to six months (Bernard et al., 2013).

Phase IV (2010 to the present)
Since 2010, the activity has been interspersed by Vulcanian eruptions with varying intensity, often accompanied by explosions as well as small PDCs, some of them reaching the volcano's base (Mothes et al., 2015;Hall et al., 2015;Parra et al., 2015;Romero et al., 2017).One of the largest Vulcanian events recorded at Tungurahua during the current eruptive cycle occurred in February 1 st 2014, with a minimum erupted volume of about 1.2 Mm 3 of PDCs and 4.5-6.4Mm 3 of air fall tephra (Hall et al., 2015;Romero et al., 2017).These Vulcanian events are preceded by abrupt onsets or little precursory activity (e.g., Kim et al., 2014;Hall et al., 2015;Parra et al., 2015;Romero et al., 2017).This is in agreement with volcano degassing, which was more or less continuous between 1999 and late 2008, but then changed to episodic activity without significant degassing during quiescence periods (Arellano et al., 2008;Hidalgo et al., 2015).

Tephra samples
For this study we analyzed 47 samples collected between 1999 and 2013 (Fig. 1; Table 1).The ash samples from 1999 to 2005 were systematically collected by members of the Instituto Geofísico of Escuela Politécnica Nacional (IG-EPN) at distances of 5-20 km from the crater.The samples have been preserved in cool and dry conditions.During the catastrophic 2006 explosive activity, tephra sampling was carried out as soon as possible after the eruptive activity (<24 h).Since 2007 and up to 2012, the IG-EPN members used a series of "ashmeters" which consist of plastic recipients 20 cm diameter and 20 cm height, installed at a distance <20 km from the crater (Bustillos, 2010;Bustillos and Mothes, 2010).These ash-meters were installed below the most frequent dispersal pattern of the ash, to the west of the volcano (Fig. 1).This instrumental monitoring, the reports of volcanic activity, and sample collection have been also enhanced by the help of volunteer community members since 2000, locally known as vigías (watchmen) (e.g., Stone et al., 2014;Mothes et al., 2015).

Tephra lithologies and textures
The lithological and textural descriptions of tephra components are based on the standard classification of Heiken and Wholetz (1985).We used two main procedures for the lithology and texture characterization of ash samples: 1) Representative tephra samples (homogenized, quartered and <2 mm diameter; coarse to very fine ash) were analyzed using a binocular microscope.The relative proportions (%) of each particle type were determined through hand picking using fine forceps.They were distributed over glass plates, allowing the identif ication of their components.2) By using the scanning electron microscope (SEM), a technique used since the last decades for description of ash particles (e.g., Heiken and Wohletz, 1985;Sherdian and Marshall, 1983;Wholetz, 1986;Dellino and Kyriakopoulos, 2003).The SEM analyses were carried out using a Tescan-Vega (Bruker) instrument operating at 15.0 kV.The ash grains, previously selected under the binocular microscope, were placed inside a circular slide (1 cm diameter) on double-sided coal tape, as suggested by Lautze et al., 2012.The samples were coated with a thin metallic layer (200A gold in 20 seconds) prior to analysis in the SEM.

Grain size
Tephra was mechanically sieved from 63 to 2,000 µm (-1 to 4Ф) at regular steps of 0.5Ф (Ф=-log 2 D/D 0 , with D being the particle diameter and D 0 a reference diameter as 1 mm), using a Fritsch analysette instrument, during a period of 3-5 minutes in order to avoid the over-fracturing of tephra.Respective grain size fractions were weighted and their mass fraction (wt%) was determined using the total mass of each sample.The sample statistics (M d =median; M z =mean; σ 1 =sorting or standard deviation; S k =skewness; K G =kurtosis) were then calculated using the Gradisat package (Blott and Pye, 2001) which uses the Method of Moments in Microsoft Visual Basic programming language.Linear interpolation is used to calculate statistical parameters by the Folk and Ward (1957) graphical method, and to derive physical descriptions (such as "very coarse sand" and "moderately sorted") (Blott and Pye, 2001).We also compared these descriptions with the classification of Cas and Wright (1987) which is commonly used for the study of volcanic deposits.After measuring the <63 µm (4Ф) ash fraction, we used the method of Horwell (2007) to estimate both "thoracic" (<10 µm; 6.7Ф) and "respirable" (<4 µm; 8Ф) ash amounts particles.

Chemistry and mineralogy
The major element chemical characterization of bulk ash (particles finer than 4Ф) samples was analyzed by Energy Dispersion Spectrometer (EDS) for SEM with a Qantax EDS (Bruker instrument).The procedure consisted in the mapping and averaging of 100 points in order to avoid heterogeneity effects from individual particles.These particles were not separated prior geochemical analyses, thus the results from samples exclusively compound by fresh juvenile particles are discussed.Work distance was 26 mm, with Secondary Electron Detector (SE) and the software Spirit 1.8 with a detection limit of 1% for the recognized elements.
Mineralogy was characterized from pulverized samples of ash (particles finer than 5Ф) at the X-Ray diffractometer (XRD) D8-Advance.The qualitative and quantitative identification of mineral phases were carried out with the software Diffrac plus (EVA) and these phases were compared with the XRD spectrums of the International Center of Diffraction Data (ICDD) database.In the case of minerals with continue compositions (e.g., plagioclases, from albite to anorthite) these are named as the amount of one of these two end-members changes (e.g., anorthite), producing different X-Ray powdered diffraction patterns and intensities (e.g., Goodyear and Duffin, 1954).Additionally, a semi-quantitative analysis of amorphous material was done with Topas 4.2 software.

Lithological and textural characteristics of tephra
Based in lithology, we have distinguished five types of particles within the tephra samples.

Pumice
Particle of a clear to intermediate gray-color, highly vesicular glassy material whose vesicles are sub-spherical (Fig. 3C).

Lithics
lithics correspond to dense angular rocks fragments, black or reddish in color, blocky, porphyritic or microcrystalline and non-vesicular (Fig. 3D).Even if part of these particles might be juvenile, they are considered as lithics because they were crystallized pior to their fragmentation (in older lavas or hipabisal intrusions).

Glass shards
Glass shards are transparent to semi-transparent, uncolored or clear brown, dense and non-vesicular material (Fig. 3F).Sharp perimeters and conchoidal fractures are typically observed.White spherical aggregates of finer fragments are ocassionally found (Fig. 3E).Thus, we considerate as fresh juvenile constituents of tephra only scoria, pumice, free crystals and glass shards as they directly result from the primarly magma fragmentation.In contrast, lithics and altered fragments such as reddish scoria are interpreted to be accidental.
Quantitative variations of these components are reported in figure 4. Also, we qualitatively describe the juvenile (mostly glassy) particles in function of their morphology (vesicularity, shape and surface;

Phase II
In Phase II, black scoria, brown scoria, free crystals and glass shards were observed.Since April 2006 and up to August 2006, the amount of black scoria decreased from 60 to 45%, while the participation of brown scoria increased from 20 to 35% (Fig. 4).The free crystals remained constant in time (15%) as similarly to glass shards (5%) (Fig. 4).Glassy particles collected in 2006 (before the August 2006 eruption) had none, low or moderate vesicularity (vesicles were, in most of cases subspherical), with blocky shape and broken or jagged perimeters (very scarce molten perimeters) (Table 2).After the paroxysmal eruption of 16 August 2006, the morphology of ash grains evolved in 24 August to moderately or highly vesicular (vesicles were subrounded or slightly elongated), with blocky shapes and the perimeters of these particles were smoothed, molten and jagged (Table 2).

Grain size
We observed unimodal (Fig. 6A; 11.9%), bimodal (Fig. 6B; 76.2%) and trimodal (Fig. 6C; 11.9%) grain size distributions (GSD) within the tephra samples (Table 3).Between 1999 and 2009, deposits    Folk and Ward (1957), 37% of the samples are poorly sorted, 50% are moderately sorted, 12.5% are moderately well sorted and 5% are very well sorted (Table 3).The classification of Cas and Wright (1987) is currently the most used for volcanic deposits, and with their criteria 62.5% of the samples are very well sorted and 12.5% are well sorted.In general, average values of M d =2.48Ф (medium to fine ash), M z =2.54Ф (medium to fine ash), σ 1 =0.93 (well sorted)Ф S k =0.07Ф (asymmetrically positive) and K G = 0.96Ф are observed for the whole dataset (Table 3).All samples are well classified as fall deposits according to the scheme of Walker (1973) (Fig. 6D).
The particles smaller than 6.7Ф ranged from 0.07 to 13.64 wt% in all samples, with an average of 4.9 wt%, and the <8Ф fraction ranged from 0.01 to 5.9 wt% with an average of 1.14 wt% (Table 3).

Bulk ash geochemistry
Due to the coexistence of different particle classes together in ash (including lithics and altered particles), only 21 of 43 geochemical analyses are considered as useful for describing the bulk geochemistry erupted materials, as they are exclusively compound by fresh juvenile particles (scoria, pumice, glass shards and free crystals) as shown in table 4. According to the total alkali-silica (TAS; Le Maitre, 1984) classification scheme (Fig. 7A), the samples from the whole eruptive style (Table 3) plot in the andesite compositional field, with a SiO 2 range between 57.9 and 62.3 wt%.
Harker diagrams are shown in figure 7B-H.Linear trends are only identified in FeO, CaO, and K 2 O versus SiO 2 variation diagrams, while a major scattering is observed in TiO 2 , Al 2 O 3 and MgO contents.In general, samples collected in 2012-2013 are slighly enriched in CaO and K 2 O (Fig. 7F and 7H), but depleted in MgO and Na 2 O (Fig. 7E and 7G) if they are compared with the samples that represent the 2001-2010 eruptions.CaO versus MgO (Fig. 7I), molar K 2 O/K 2 O+CaO versus MgO (Fig. 7J) and the pseudoternary system defined by Opx-An-Or (Fig. 7K) (Díaz-Alvarado et al., 2011;Castro, 2013) are particularly significant for visualizing linear trends in co-genetic igneous rocks.However, Tungurahua samples display a significant scattering for a scarce SiO 2 variation.The higher CaO and K 2 O and lower MgO contents observed in the 2012-2013 samples regarding to the rest of the studied ash deposits separate two groups of samples in these diagrams.

Phase I
The vent-cleaning phase of reawakening at Tungurahua in 1999 (Ruiz et al., 2006;Le Pennec et al., 2012) was recorded within the first tephra fall sample (October 1999) which involved a notable proportion of hydrothermally altered scoria (30%) and small amounts (5%) of glass shards.The recognizance of hydrothermal alteration at juvenile particles is a good tool to assess the existence of juvenile material recycling (e.g., Houghton and Smith, 1993).These particles are expected to contribute little, if any, thermal energy to the explosions (Hougton and Carey, 2015).The jagged perimeter of some of these particles indicates a brittle fragmentation, thus probably of preexistent deposits.There were not observed particles with quenching cracks, commonly associated to the fast passage (a few milliseconds) of newly fragmented particles through a domain of liquid water (Büttner et al., 1999), thus a phreatomagmatic origin is unlikely.However, there are inherent limitations of particle observation associated to the grain size fractions here studied, thus that phreatomagmatic should not be totally discarded, and both lithologic and textural patterns here observed have been frequently associated to the initial phreatic-phreatomagmatic transitional stages of volcano reawakening, such  (1984).B-H.Harker diagrams, and I-K.variation diagrams.
In A: A alkaline, Ca+T Calcoalkaline+Tholeiitic; In K: FM Ferromagnesian minerals (Fe-Mn-Mg), An Anorthite, Or Orthoclase.as the case of Cotopaxi in 2015 (e.g., Gaunt et al., 2016;Troncoso et al., 2017).The disappearance of altered scoria in the samples collected in July 2000, and the increasing participation of glass shards, scoria and microcrystalline lithics is attributed to the rising of the magma to the surface and its progressive eruption through an open-vent system.
In consequence, the onset of Phase I is related to the progressive cleaning of the vent due to rising of a renewed magma which was finally erupted since 2001 with an open-vent style (Fig. 8A and B).The low vesicularity and blocky shape of the juvenile fragments has been suggested by Cioni et al. (2014) as consequence of the shuttering of syn-eruptively degassed, crystallized, more rigid and volatile-rich magmas explosively fragmented, as it was observed during the 2010 Eyjafjallajökull eruption.Further tephra emissions in 2002 and 2003 were produced by Strombolian and Vulcanian-like eruptions consisted of two varieties of juvenile scoria (black and brown) accompanied by higher amounts of glass shards (10-15%) which is consistent with an open vent system.The two different types of scoria ejected may be attributed to textural variations (such as crystallinity/ vesicularity) or to the eruption of an heterogeneous melt (i.e., magma mingling or an stratified magma column).However, the bulk ash geochemistry of these samples was quite stable (57.9-60.2wt% SiO 2 ), similarly to the mineral phases identified via XRD (Table 5).This geochemical and mineralogical stability may be explained by the eruption of juvenile particles with different vesicularity and groundmass crystal content, which may suggest the fragmentation of a vertically heterogeneous magma column (Cioni et al., 2014), or as result of varying magma supply rates as demonstrated by Wright et al., (2012) within this period (Fig. 8B).From textural observations, can be suggested that both rheology and magma supply rates favor this juvenile textural bimodalities.

Phase II
The largest eruptions during 2006 showed long-lasting unrest periods in the months prior to the eruption (e.g., Champenois et al., 2014).There occurred a progressive decrease of black scoria in time, whereas the brown scoria increased.As previously suggested, these bimodal juvenile particles may be correlated to the stratification of the magma column in the conduit.This agrees with the increasing vesicularity of juvenile grains between April and August 2006, which suggest the progressive eruption of a volatile-saturated magma.In fact, for a given viscosity, increasing density of bubbles correlates with higher bubble nucleation rates that have been attributed to progressively later onset of bubble nucleation at higher degrees of supersaturation with volatiles (Houghton and Carey, 2015).Thus, the predominance of highly vesicular fragments suggests a prominent and active role of degassing-related magmatic fragmentation during the eruption (Cioni et al., 2014).An additional fact is the shape of these vesicles.Elongated and fibrous vesicles, such as these observed in the 24 August samples are indicative of magma fragmentation due to a rapid acceleration of the system (Cashman et al., 2000) or as representative of the central part of a magmatic column at the conduit during Pliniantype eruptions, where vesicles are free to grow only subjected to elongational stresses (Polacci, 2005).All these observations are coherent with the direct observations of the volcanic activity during 2006, which produced a stratospheric column (Fig. 8C).However, in contrast to the geochemical data obtained from samples collected from the rest of the time series, the bulk rock geochemistry at August 2006 reveals a silica-rich andesitic composition (62.3 wt% SiO 2 ).Additional changes are noticed from mineralogy by the first recognizance of akermanite and magnetite, higher amounts of plagioclase (>60%) and lower amounts of pyroxene (32-33%) together permanent olivine, all of them recognized via XRD as characteristic for this period.Thus, the compositional heterogeneity should indicate mixing of two magmas (Fig. 8C).This is in agreement with the observations made by several authors (e.g., Fee et al., 2010;Steffke et al., 2010;Samaniego et al., 2011;Eychenne et al., 2013;Myers et al., 2014).In fact, melt inclusion composition paired with host phenocryst zonation made by Myers et al. (2014) indicated mixing of two distinct magmas: a volatilerich (~4.0 wt% H 2 O and ~1,800 ppm S) basaltic andesite containing olivine phenocrysts and a degassed (~1.0 wt% H 2 O and 100-500 ppm S) andesite with plagioclase and pyroxene phenocrysts that contain andesitic to dacitic melt inclusions.

Phase III
During the initial stage of Phase III in 2007, there was a new increase of hydrothermally altered reddish scoria and lithics (total amount 40-65%), reaching similar proportions to juvenile particles (black scoria, free crystals and glass shards).There are not textural indicators of phreatomagmatic interactions in juvenile particles (there were only observed particles with blocky and f luid shapes), however the hydrothermal alteration, the amount of recycled material and the existence of fresh juvenile particles suggest a Vulcanian mechanism (Fig. 8D).For comparison, during the 1 st February 2014 Tungurahua eruption, an average of 49.34% corresponded to non-juvenile material in tephra (38.16% lithics and 11.18% altered scoria), thus a minimum limit of 40% of recycled material will be here considered as indicative of Vulcanian eruptions.Several f luid, filamentous and blocky shaped glassy particles were identified within the ashes of 5-6 February 2008.Melt droplets, preserved as fluidal clasts (filaments) are correlated to magmas that fragment in a hot f luid state (Houghton and Carey, 2015).These particles are most typically associated with small subaerial explosive eruptions of low-viscosity (<10 2 Pa s) basaltic melts, due to the deformation of low-viscosity lava in the air (Moune et al., 2007).Thus, we can associate these particles to the alternation between ductile and brittle fragmentation of a hot magma, as result of varying rheology and/or sypply rate of the magma, similarly to Phase I.This agrees with the direct observation of alternating episodes of jetting and explosions (Biggs et al., 2010).
The short-to medium-lived eruptions occurred since 2008, which erupted exclusively blocky juvenile material with low-to-moderate vesicularities is typical from Strombolian eruptions, and reflects the brittle fragmentation of volatile loaded magma (Fig. 8B).

Phase IV
Hydrothermally altered fragments, represented by reddish scoria was frequent in most of the samples since 2010 up to 2012, but always at lower amount than 30%, much lower than in Phase III.These tephras enriched in altered fragments were interspersed with some infrequent, purely juvenile tephras and a particular tephra fall full of lithics in December 27 th 2011.At surface level, Vulcanian eruptions with little geophysical warning are inferred from eyewitness observations within this period and also in the following years (e.g., Hall et al., 2015;Mothes et al., 2015;Parra et al., 2015;Kim et al., 2014;Romero et al., 2017).However, Vulcanian events are characterized by a series of physical properties and distinguishable eruptive products (e.g., Clarke et al., 2015;Morrisey and Mastin, 2000) such as >40% of recycled particles within tephra, as we previously established for Tungurahua (Phase III and 1 st February 2014) which was not observed between 2010 and 2012, with the only exception of the December 2012 event.In addition, vesicularity of juvenile particles was zero or moderate-to-high, not spanning a wide range as frequently observed during Vulcanian eruptions (e.g., Clarke et al., 2015).These particles also exhibited ductile fragmentation features such as f luid and filamentous shapes, despite fewer blocky shapes were observed.Pure magmatic dynamics, especially during Strombolian activity is more prone to produce these particles (Fig. 8B).This is also supported by the eruption of dominant pumice in 2013.Thus, at this point the origin of hydrothermally altered grains remains unclear.Then, was Tungurahua able to develop an active hydrothermal system (necessary condition to provide hydrothermally altered particles) during the paucity of magmatic eruptions?.Even when the heat transfer from shallow magma batches is not favorable for the existence of an hydrothermal system, the development of a reduced or sporadic hydrothermal system is possible, as indicated by geochemical and isotopic characterization of springs and bubbling waters sampled at Tungurahua, which are Cl-SO 4 earth-alkaline, medium salinity and significant He-mantle signature (around 60%) fluids (Inguaggiato et al., 2010).Thus, the lesser amounts of hydrothermally altered particles between 2010 and 2013 are likely related to a harder involvement of water in the system during the non-eruptive stages, and not to conduit plugging.This condition was favored by the episodic frequency of eruptions, which permitted the infiltration of meteoric water into the conduit (Fig. 8E).Even more, the episodic frequency of these eruptions was probably not associated to the conduit plugging, but yet correlated to a change in the magma supply rate or an increase in the viscosity of magma intrusions.This is well constrained with the geochemical variations observed in the year 2012-2013 which may be interpreted as the intrusion of a "renewed" (deeper, volatile-rich magma) triggering magma mixing and eruption, as it is also noticed by Myers et al. (2014) in 2010.Even though a single Vulcanian eruption is identified using tephra between 2010 and 2013, further eruptions at the end of Phase IV (2013)(2014) were well recognized as Vulcanian (e.g., Hall et al., 2015;Parra et al., 2015;Romero et al., 2017).The transition in time from open to a plugged conduit suggest a progressive decrease of the magmatic ascent rates or the increase in magma viscosity through time.

Magmatic processes feeding eruptive transitions
Within its geological and historical eruptive history, Tungurahua products range from basaltic andesites to dacites, with subordinate rhyolites, and a typical plg + cpx + opx ± hbl ± ol ± Fe-Ti oxides mineral assemblage (Hall et al., 1999;Bustillos, 2008;Samaniego et al., 2011).The rock compositions from ancient and modern Tungurahua edifices (I, II and III) (Hall et al., 1999;Bustillos et al., 2011) turn out to be similar to the composition of tephra emitted between 1999 and 2013, as they are both medium-to high-K andesites (54-58 wt% SiO 2 ).If we assume that analytical issues are avoided during the sampling and analytical processes, the absence of linear trends in variation diagrams and the separation between the 2012-2013 and the rest of the samples may be related to different temperature, pressure and/ or water content conditions at the source of these magmas and reject a differentiation or fractionation process from a similar magmatic source.This could be complementary to that proposed by Myers et al. (2014) on the co-existance of several magma reservoirs at different depth (shallow, 1-3 km under the base of the volcano and a deep reservoir at >7 km depth).

Depositional processes inferred
According to Eychenne et al. (2012), the eruption of August 2006 at Tungurahua had bimodal GSD deposits controlled by two synchronous processes: lapilli deposition from the main plume and fine ash elutriated from PDCs.This situation was also shown by Bernard et al. (2016).At least 5 eruptive phases produced PDCs from 2007 to 2013 (March 2007, February 2008, May and December 2010, December 2012, May and July 2013) (Hall et al., 2015;Bustillos et al., 2016), which is in good agreement with the increase of bimodal and trimodal GSDs from 2009 to 2013.For example, the major eruption occurred in May 28 th 2010, which produced an eruption column of 10 km above the crater level (a.c.l.) and a series of PDCs, corresponds to deposits with bimodal GSD characterized by two opposite modes at fine lapilli (-0.74Ф) and very fine ash (4.23Ф) (Table 3).These features are not unique for purely-magmatic, large scale eruptions, as they are also observed in moderate Vulcanian eruptions occurred later in 2013 and 2014 (e.g., Parra et al., 2015;Romero et al., 2017).
Particle aggregation has been also associated with bimodal or polymodal GSDs in the literature (e.g., Scasso et al., 1994;Durant et al., 2009).We suggest that trimodal GSD in sample 18B is related to particle aggregation, as evidenced by the SEM analyses which show abundant aggregates that correspond to coated particles, bound by hydro-bonds and electrostatic forces, and in most cases are poorly preserved due to the particle impact during fall (particle type PC2 in Brown et al., 2012).As Tungurahua area annually receives a notable amount of rainfall (about 3000 mm; Jones et al., 2015) we suspect that the humidity provided by these atmospheric processes could favor the formation of aggregates during explosive eruptions, especially when no phreatic eruptive mechanism is suspected through ash analyses.Although thunderstorms/lightning are not often reported in this area (e.g., Rollenbeck and Bendix, 2011) but they may occasionally promote the formation of aggregates.Nevertheless, these two explanations do not account for all the bimodal and trimodal deposits and should be considered as alternative mechanisms of ash aggregation in this area.
In the case of samples not generated during PDC-forming eruptions or ash aggregation, a third deposition mechanism is required.Possible explanations are: 1) the nature of Vulcanian events that may erupt both fine and coarse grained tephra including ballistic rocks (e.g., Fierstein et al., 1997;Cashman et al., 2000;Houghton et al., 2004;Clarke et al., 2015), which may be deposited together, especially in the proximal zones as our sampling stations; or 2) the plume dispersal patterns, including plume bifurcation, wind direction variability and deposition of long-lasting (tens of days) eruptions that produce millimeter-thick deposits with complex dispersals and non-elliptic isopachs (Bernard et al., 2013).This would generate modes with different, but not extreme grain sizes (i.e., very fine grained or very coarse grained), as seen in table 3 for all the samples prior to 2006.
We suggest that, during the first 7 years of eruptions (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006), processes like plume bifurcation and sedimentation of multiple plumes were responsible of most polymodal GSD (Fig. 9A), while during and after 2006 there was a combination of deposition processes that generated bimodal/ trimodal GSD deposits, such as interaction between plume and co-PDC deposition (Fig. 9B), Vulcanian fragmentation (Fig. 9B), and the variability of tephra plumes.As an accessory and marginal process, we suggest that ash aggregation could be produced during strong thunders or lighting in the area of Tungurahua (Fig. 9C).The combination of all of those products may be responsible of polymodal GSD in tephra samples; C.During "weak" eruptions (VEI 1-2) atmospheric anomalies (i.e., rain fall and/or electrical storms) may induce particle coating leading to the formation of ash aggregates.In areas without these disruptions, tephra fall will probably produce unimodal or bimodal grain size distribution.

Potential health impacts
Acute respiratory manifestations seen after heavy ash falls include irritation of the chest, nose and throat discomfort, but also acute exacerbations of asthma and bronchitis are seen (Gudmundsson, 2011).As shown in table 3, the extremely fine ash particle size, both thoracic and respirable (breathable) (<10 and <4 µm, respectively) have been calculated using the method of Horwell and Baxter (2006).The average proportion of Tungurahua's respirable-sized ash is estimated at ~5 wt% (range from 0.07 to 13.64 wt%) nevertheless no crystalline silica (i.e., quartz, cristobalite or tridimite) has been identified via XRD.In addition, filamentous glassy particles (known for being abrasive; e.g., Horwell and Baxter, 2006) are always observed in larger sizes (>200 µm) than for respirable or alveolar particles.Thus, the ash of Tungurahua has to be considered as potentially harmful for human health as a result of long-lasting exposure, especially because it may trigger acute health problems despite chronic diseases.This is in agreement with direct health observations, as epidemiological map has changed in the villages affected by the volcano while bronchial asthma, cough, throat irritation, eye irritation and respiratory allergies, and altered pulmonary capacity (spirometry) has been reported after Tungurahua ash exposure (Paladines and Zamora, 2011;Cifuentes and Alvarado, 2015) 6. Conclusions Some conclusions derived from our research are listed below: a.During the initial stage (1999)(2000) of the current eruptive cycle at Tungurahua, a progressive conduit cleaning is observed as the eruption elapses.At the beginning of the eruption in 1999 phreatic mechanism has been interpreted, as no evidence of phreatomagmatic activity has been found using the current methods.The ashes then erupted in 2001 revealed the explosive fragmentation of a degassed, crystallized and "rigid" magma.In contrast, the juvenile products erupted after in 2002 and 2003 exhibited bimodal textures and compositional (chemistry and mineralogy) homogeneity, and we suggest they reflect variations in both magma supply rate and its rheology within a stratified magma column, without any evidence of the interaction between two different magma intrusions.
b.The progressive variation of the 2006 Plinian eruption and associated tephra components, in addition to the textural changes observed in juvenile particles suggest that these eruptions were fed by a volatile-saturated magma, thus the fragmentation of magma occurred by its degassing and its rapid acceleration within the conduit.Even more, the evidence from geochemistry and mineral phases indicate a compositional change of the magma if compared to the pre-2006 series, which is in agreement with the previously reported mixing of two magmas (e.g., Samaniego et al., 2011;Myers et al., 2014) of the magmatic ascent rates or the increase in magma viscosity through time.e.From grain size analyses, we propose two first order phenomena during tephra deposition: 1) plume bifurcation and sedimentation of multiple plumes were responsible of most polymodal GSD, mainly occurring between 1999 and 2006; 2) deposits with bimodal/trimodal GSD as consequence of interaction between plume and co-PDC deposition (PDC elutriation), Vulcanian fragmentation, and the variability of tephra plumes during and after 2006; a second order factor could be ash aggregation during strong thunders or lighting in the area of Tungurahua.f.Due to its mechanical properties, the ash of Tungurahua has to be considered as potentially harmful for human health as a result of long-lasting exposure, especially because it may trigger acute health problems despite chronic diseases.Not crystalline silica has been identified.g.This study demonstrates that periodic and systematic tephra fall sampling at long-lasting erupting volcanoes is a valuable tool for understanding the eruptive mechanisms and the processes and impacts related to tephra falls, especially at inhabited areas.

FIG. 1 .
FIG. 1. Location of Tungurahua volcano.The sampling sites in outcrop (brown dots) and the location of ash-meters (green dots) administered by IG-EPN are included in the map.See legend for details on the sampling sites and their correspondent samples.Dispersion of relevant tephra falls are indicated as dashed lines.

FIG. 2 .
FIG. 2. Eruptive phases during the 1999-2014 eruptive cycle at Tungurahua volcano as defined by Bustillos et al. (2016) combining direct observation, available data published for each eruption (see references therein), and column heights (km above the crater).
FIG. 3. Lithologic components identified by optical microscope in ash samples from Tungurahua. A. Black scoria fragment in sample 5; B. Brown scoria in sample 9B; C. Reddish scoria and lithics found in sample 19A; D. Pumice particles with elongated vesicles from sample 13; E. Free crystals (mainly plagioclase, black arrows) found in sample 17.Also, an aggregate is highlighted with a red circle; F. Glass shards in sample 8.For scale purposes, red bars are 1 mm in length.

FIG. 4 .
FIG. 4. Relative proportion of the different components of tephra samples of eruptions between 1999 and 2012 at Tungurahua volcano.

FIG. 6 .
FIG. 6. Grain size distribution (GSD) of tephra fall samples.A. Unimodal GSD (red); B. Bimodal GSDs, where two main bimodal populations are recognized (orange and dark grey); C. Trimodal GSD with also two main populations (green and violet); D. Walker (1971) classification for GSD of pyroclastic deposits.The samples here studied plot in the field of fall deposits.Colours from lines and circles correspond to each cluster of unimodal or polymodal samples.

FIG. 8 .
FIG. 8. Cartoons of the different eruption mechanisms operating in Tungurahua volcano between 1999 and 2013. A. Model of the early Phase I with the development of an initial phreatic eruption.Rising magma transfers heat to the hydrothermal system producing the eruption of recycled rocks, including hydrothermally altered wall rocks; B. Open conduit dynamics as observed in late phase I, late Phase III and partially in Phase IV.Varying magma supply rates or stratified conduit determines the transition between brittle and ductile fragmentation, producing juvenile particles with varying degree of vesicularity and variable shape; C. Magma mixing produces a volatile-saturated melt which triggers Plinian activity during Phase II; D. A decrease of magma supply rates may contribute to the plugging of the conduit and the over-pressurization of the magma column to trigger Vulcanian eruptions as in early Phase III and late Phase IV; E. Similarly to A, during eruptive silence periods in Phase IV the entry of meteoric water favored the formation of an sporadic hydrothermal system and local wall-rock hydrothermal alteration.Low magma supply was not able to trigger eruptions.

FIG. 9 .
FIG. 9. Cartoons of the different fall out mechanisms proposed for tephra fall deposits at Tungurahua volcano. A. When no major eruptions or atmospheric anomalies are occurring, transitions in eruptive style (e.g., phreatic to Strombolian, Vulcanian to Strombolian or similar cases), produce variations in the dispersal pattern of tephra and spasmodic activity produce tephra deposits with polymodal GSD; B. Major eruptions (purely magmatic as in August 2006 or Vulcanian-triggered as in 28 May 2010 or 14 July 2013, VEI ~3) may develop high eruptive columns (>10 km; not to scale in figure) with "heavy" tephra fall, including lapilli-to bomb-sized fragments at ~10 km radius from the vent, together with abundant ash.Pyroclastic density currents (PDCs) f lowing down the f lanks of the volcano might contribute fine material associated with co-PDC clouds.This was shown by Bernard et al. (2016) for the August 2006 eruption.Ballistics at a distance up to 4-5 km are common.The combination of all of those products may be responsible of polymodal GSD in tephra samples; C.During "weak" eruptions (VEI 1-2) atmospheric anomalies (i.e., rain fall and/or electrical storms) may induce particle coating leading to the formation of ash aggregates.In areas without these disruptions, tephra fall will probably produce unimodal or bimodal grain size distribution.

TABLE 1 . PHYSICAL FEATURES OF THE SAMPLES ANALYZED WITHIN THIS CONTRIBUTION. Sample D (km) Date Mass (g) Bulk density (g/cm3) True density (g/cm3)
D is the distance from the crater measured in km.Towns can be tracked at figure 1.

TABLE 2 . SYNTHESIS OF TEXTURE AND STRUCTURE OBSERVED ON JUVENILE PARTICLES ERUPTED BY TUNGURAHUA BETWEEN 1999 AND 2013 US- ING SEM-EDS.
IVW* correspond to inter-vesicular walls.

TABLE 3 . GRAIN SIZE PARAMETERS OF FOLK AND WARD (1957) FOR 42 SAMPLES SIEVED MECHANICALLY. Sample Folk and Ward (1957) statistics (Ф) Modes Fine to extremely fine ash
unimodal and bimodal GSD's were frequently interspersed, but from 2009 to 2012 most of the samples had a bimodal or trimodal distribution.The tephra from the August 2006 eruptions is bimodal, with modes at 2.74 and 4.23Ф (medium to very fine ash).The trimodal GSD's are typical of samples from the2002, 2011, 2012, and 2013eruptions.According to the classification of Md: median; Mz: mean; σ1: sorting; Sk: skewness; KG: Kurtosis.For sample distribution modes 1, 2 and 3 are M1, M2 and M3.Also, for sample sorting (S) the label meaning is defined as follows: P: poorly sorted, M: moderately sorted; W: Well sorted and VW: Very well sorted.a.Calculated using the method (1) ofHorwell (2007): y=0.0009x2+ 0.236x, where y is the cumulative% <10 µm and x is the cumulative% <63 µm fraction.b.Calculated using the method (2) ofHorwell (2007): y=0.0016x2+ 0.043x, where y is the cumulative% <4 µm and x is the cumulative% <63 µm fraction.The correlation (R2) of methods (1) and (2) are 0.853 and 0.88, respectively.The bold data correspond to the three highest values of finest ash fraction of all the samples.with

TABLE 4 . BULK ASH MAJOR ELEMENT CHEMICAL COMPOSITION FOR THE SAMPLES OF TUNGURAHUA VOLCANO COMPOUND EXCLUSIVELY BY JUVENILE PARTICLES, FROM ERUPTION OCCURRED BETWEEN 1999 AND 2013.
FIG. 7. Major element geochemistry of juvenile ash fragments for samples from Tungurahua volcano. A. TAS diagram of Le Maitre

TABLE 5 . MINERALOGICAL COMPOSITION OF ASH SAMPLES COLLECTED BETWEEN 1999 AND 2013 AT TUNGURAHUA VOLCANO. ALL THE ABUNDANCES ARE MODAL%.
. c.During early 2007, the eruption of high amounts recycled particles (40-65%) accompanied by juvenile particles are compatible with a series of Vulcanian eruptions.In contrast, the eruptions then occurred in 2008 were magmatic and ejected both fluid and brittle-fragmented juveniles during jetting periods and explosions.This is interpreted as a result of varying rheology and/or supply rate of the magma, similarly as seen in 2002-2003.The short-to medium-lived eruptions occurred since 2008, ejected blocky juveniles as typically seen during Strombolian eruptions, and ref lects the brittle fragmentation of volatile loaded magma.d.Varying amounts of hydrothermally altered particles (always <30%) between 2010-2013 are likely related to a harder involvement of water in the system during the periods of eruptive silence, which certainly permitted the infiltration of meteoric water into the conduit.The textural features of juvenile particles suggest a dominant ductile fragmentation associated to Strombolian activity.Formal Vulcanian events are newly recognized in December 2011 and widely observed since 2013.The transition in time from open to plugged conduit dynamics suggest a progressive decrease