Melt volatile content
Groundmass glass volatile analyses evince very low volatile contents in the interstitial melt of the magma plugs, with a higher volatile content in the magma that gave rise to inflated bombs than in the magma that gave rise to scoriaceous and dense bombs. The correlation of glass volatile content with bomb morphology and texture is consistent with the findings of Hoblitt and Harmon (1993) at Mount St Helens, who found that pyroclasts of 1980 blast dacite containing > 0.4 wt% H2O in the groundmass glass vesiculated on eruption timescales, whereas pyroclasts containing < 0.4 wt% H2O showed no evidence of syn-eruptive vesiculation and remained dense. They attributed this effect to the speciation of water in rhyolitic melt at very low water contents. Wright et al. (2007) also studied the H2O content of groundmass glasses in dense bombs and bread-crust bomb rinds from Guagua Pichincha volcano, Ecuador, and attributed the existence of dense bombs to a syn-eruptive bubble nucleation delay. Wright et al. (2007) suggested a higher threshold for syn-eruptive vesiculation to occur on eruptive timescales (0.9 wt% H2O), but noted that they found no dense bombs at > 0.4 wt% H2O. The morphological types of bombs found at Galeras volcano are therefore consistent with these interpretations. The pre-eruptive nature of vesicles in dense and scoriaceous bombs is also supported by the frequent presence of vapour-phase cristobalite growing on vesicle walls and the presence of tuffisite material infilling vesicles in scoriaceous bombs, as the growth of vapour-phase cristobalite occurs within magma plugs and domes (Horwell et al. 2013) and tuffisite veins form due to localised overpressure events within magma plugs and domes (Kendrick et al. 2016).
The only inflated bomb (AB28) with groundmass glass water content < 0.4 wt% H2O also features low CO2 content, the lowest F and Cl of all inflated bombs, elevated S and textural evidence of a cm-scale tuffisite vein in the rind (Online Resource 5, S5). The three dense bombs in which tuffisite veins were identified show similar volatile patterns, with low H2O (0.08–0.2 wt%) and CO2 (2–4 ppm), the lowest F (49–78 ppm) and Cl (63–87 ppm) contents in the sample suite and elevated S (11–16 ppm). These observations suggest an efficient, localised degassing mechanism associated with tuffisite veins, allowing H2O, CO2, F and Cl to become depleted in the interstitial melt surrounding the vein. In contrast, S is apparently re-absorbed into the adjacent melt, possibly due to increased oxygen fugacity and/or temperature (Wallace et al. 2015) and the advection of a S-rich gas phase through the tuffisite vein. As tuffisite veins locally increase permeability until sintering reduces it back to lower values (Heap et al. 2015), a tuffisite vein is likely to remain a preferential gas transfer pathway on a longer timescale than the short time required to create the tuffisite vein by brittle failure, enhancing the opportunity for localised re-equilibration of the melt to occur. The preservation of tuffisite material and this distinctive volatile signature in the rind of an inflated bomb (AB28) suggests that, while the localised degassing effect of tuffisite veins is significant, it remains a local effect that did not impede the ability of this parcel of magma to vesiculate upon eruptive decompression. Hence, the interstitial melt at a distance of a few millimetres from the tuffisite vein must have retained > 0.4 wt% H2O in order for the bomb to inflate. That the effect is localised is also supported by the absence of this signature in other inflated bombs that show preservation of tuffisite veins on a smaller scale, implying that larger-scale tuffisite veins affect a greater volume of adjacent interstitial melt. This is consistent with the findings of Castro et al. (2012), who investigated the effect of tuffisite veins on degassing rhyolitic magma erupted at Chaitén volcano, Chile, and concluded that their effect was limited. Castro et al. (2012) noted that the conditions where tuffisite veins may play a significant role in degassing magma on a large scale probably occur where they intersect existing porous, permeable networks. Tuffisite veins at Galeras volcano may therefore have contributed most to the overall degassing of the magma when they intersected the region of each magma plug that gave rise to scoriaceous bombs.
The only dense bomb that shows evidence of brittle deformation and grain comminution in the form of a cataclasite band (AB38) records a significantly higher CO2 content (60 ppm) in the groundmass glass of the adjacent homogeneous andesite. Cataclasite bands are thought to occur primarily at conduit margins and represent highly anisotropic high permeability zones channelling gas towards the surface (Gaunt et al. 2014). The cataclasite band in this dense bomb may therefore have acted as a permeable pathway for gas fluxing from deeper degassing magma towards the surface. The interstitial melt may have become enriched in CO2 as a result of this gas flux, as CO2 is soluble in water-poor melt at conduit pressures (Wright et al. 2007). Sulphur is also slightly enriched in this bomb sample (9 ppm), suggesting the flux of an S-bearing gas phase through the cataclasite zone.
These observations suggest that tuffisite veins act as degassing pathways advecting locally sourced, S-rich gas, whereas cataclasite veins act as pathways for deeper, more CO2-rich gas. The pre-eruptive porous network identified in scoriaceous bombs is likely to have provided permeable pathways that allowed relatively “volatile-rich” magma akin to that which gave rise to inflated bombs to become almost completely degassed. This degassed magma was the source of scoriaceous bombs, and porous network collapse may have allowed the densification of this magma, giving rise to dense bombs.
Storage pressures and depth
The H2O and CO2 content of groundmass glasses were used in the solubility model of Newman and Lowenstern (2002) for rhyolitic melts to estimate the final storage pressure of each parcel of magma prior to explosion. The best temperature constraint available was obtained by two-pyroxene thermometry from products erupted during 2004–2006 (personal communication, G.P. Cortés, SGC, 21/10/2015). The temperatures indicated by touching pyroxene pair rims were higher than those indicated by the cores and averaged 980 °C. Assuming isothermal ascent and a temperature of 980 °C, the calculated storage pressures range from 0.2 to 12.5 MPa (Fig. 11). Estimates of storage pressure are excessively high for some samples containing excess CO2 interpreted to originate from vapour fluxing (Wright et al. 2007). These samples lie above the overall degassing trend fitted by eye in Fig. 11, but probably degassed along this trend with other samples prior to re-absorption of CO2.
The magma that gave rise to dense and scoriaceous bomb types was stored at the lowest pressures and over a similar range of pressures (Fig. 11): 0.2–3.9 MPa for scoriaceous bombs and 0.5–5.1 MPa for dense bombs (omitting sample AB38 that hosts the cataclasite vein). The magma that gave rise to inflated bombs was typically stored at higher pressures in the range 2–12.5 MPa (omitting sample AB28 that hosts a cm-scale tuffisite vein in the rind). Taking a magmatic temperature of 880 °C reduces these pressure estimates by 14.5–17%.
Assuming a uniform magma density of 2700 kg m−3, calculated storage depths are in the range 9–472 m (Fig. 11). Conversely, taking the maximum observed porosity for scoriaceous bombs (25%), assuming a uniform magma porosity and a gas density of 21 kg m−3 at 980 °C and 12 MPa (assuming pure water vapour and treating it as an ideal gas (Clarke et al. 2002)) yields a bulk density of 2030 kg m−3 and a range of depths of 11–627 m. As the uncertainty in magma density does not significantly affect our conclusions, for the purposes of the following discussion, we proceed with the depth estimates assuming a uniform magma density of 2700 kg m−3, which represent minimum storage depths.
Magma that gave rise to dense, scoriaceous and inflated bombs was stored in the ranges 18–193 m, 9–146 m and 77–472 m respectively. These results show that the magma that gave rise to inflated bombs upon eruption contained more dissolved water in the melt phase and was stored deeper than the magma that gave rise to contemporaneous dense and scoriaceous bombs. In addition, no more than approximately 500 m of stratified magma were ejected in any single explosion.
Storage pressures alone do not demonstrate whether the porous magma that gave rise to scoriaceous bombs underlay or overlay the degassed, dense magma that gave rise to dense bombs. However, most scoriaceous bombs record higher F and Cl content in the groundmass glass than most dense bombs, whereas dense bombs tend to record higher S and somewhat higher CO2 than scoriaceous bombs. In other words, the glasses in scoriaceous bombs contain higher concentrations of the volatile species that are expected to follow a degassing trend and that display this behaviour within the inflated bomb suite. In contrast, glasses in dense bombs contain higher concentrations of the volatile species that are expected to become re-dissolved in highly degassed magma at conduit pressures and that display this behaviour in samples containing remnant degassing pathways (tuffisite and cataclasite veins). Finally, the lower vesicularity of dense bombs, the polylobate shapes of vesicles in dense bombs when they are found and the gradual transition in vesicularity that exists between scoriaceous and dense bombs all support the idea that dense magma formed from densification of porous magma by viscous collapse of the porous network.
The analyses of volatiles in groundmass glasses therefore support a conceptual model of repeated development and destruction of shallow, stratified plugs of magma in the conduit prior to vulcanian explosions. These volcanic plugs consisted of a deeper, dense and relatively water-rich magma that gave rise to inflated bombs, overlain by a largely degassed magma hosting a permeable, porous network that gave rise to scoriaceous bombs and capped by degassed, dense magma that gave rise to dense bombs (Fig. 12a). Prior to the onset of a vulcanian explosion, the permeable, porous network present in the porous magma had efficiently degassed the melt in the shallowest region of the plug and was in the process of viscously collapsing to form the dense magma cap. This process proceeded to various extents prior to individual explosions. This model implies that a pulse of vesiculation, bubble growth and coalescence occurred at a relatively consistent depth in the shallow conduit prior to each explosion. It also implies that the process of porous network closure and collapse in highly crystalline intermediate composition magmas is typically less efficient than the degassing process that the porous network effects, as pore pressures must have been decreasing for bubble collapse to occur.
The development of the overpressure that drove vulcanian explosions is therefore unlikely to have developed in the porous magma region that gave rise to scoriaceous bombs. The necessary gas overpressure is also unlikely to have developed in the magma that gave rise to inflated bombs, as this magma was dense (typically 0–0.5% vesicularity) rather than porous. It is most likely to have developed in a water-rich, porous zone underlying the magma that gave rise to inflated bombs, which was not preserved as ballistic bombs (Fig. 12a). Given the dense nature of inflated bomb rinds, these observations suggest that magma degassing during ascent to shallow levels was accomplished by multiple pulses of bubble nucleation, growth, coalescence and collapse, as deeper porous networks that degassed the magma from typical andesitic arc magma water contents (3–6 wt% H2O, Wallace et al. 2015) at depth to inflated bomb rind water contents (0.4–1 wt% H2O) must have collapsed prior to magma emplacement at the level of inflated bomb storage (< 500 m). The preservation of tuffisite veins in all bomb types suggests that this degassing mechanism was also operating and may have been a significant contributor over time. Given the anticipated impact of the development of porous permeable networks (accompanied by concomitant degassing and crystallisation) on magma buoyancy and rheology, the repeated development and collapse of porous networks is likely to be closely related to the occurrence of multiple cycles of magma ascent and stalling in the conduit, contributing to low average ascent rates.
Residual melt in the shallowest, most degassed region of each magma plug was generally more evolved than residual melt in the deeper, less degassed region. Assuming the magma that fed each plug was initially relatively compositionally homogeneous, the development of each stratified plug must have resulted from contrasting extents of crystallisation of the magma that gave rise to inflated and scoriaceous/dense bombs. Furthermore, variations from one explosion to the next imply that this process occurred repeatedly to form the sequential plugs but with slight variations over the course of the eruption sequence.
Glasses from the latest explosion for which time-constrained samples are available (2 January 2010) cover a more restricted range for all analysed oxides. Contrary to time-constrained samples from other explosions, these samples also feature increasing FeO*, TiO2 and Na2O wt% with increasing SiO2 wt%, and analyses of SiO2 in all samples from this explosion are among the highest in the sample suite (Fig. 4). Inflated bombs from this explosion are texturally unique in that the groundmass contains a higher proportion of glass and feldspar microlites are much smaller with skeletal morphologies. The unique compositional characteristics of the melt phase within this magma plug suggest nucleation-dominated crystallisation at higher degrees of ΔT. This implies that the magma involved in this explosion ascended at a faster rate and that a burst of crystal nucleation produced rapid evolution and degassing of the melt, resulting in magma arrest and rapid plug development in the shallow conduit. The repose time prior to this explosion was 43 days and is the shortest in the studied sequence.
In the haplogranite ternary system, An-corrected normative Qz-Ab-Or compositions track the shift in the feldspar-quartz minimum with decreasing water pressure (Fig. 13) (Cashman and Blundy 2000). As the kinetics of plagioclase crystallisation are likely to be sluggish in high-viscosity, degassed rhyolitic melt due to low rates of diffusion, the high pressures indicated in the haplogranite ternary represent closure pressures where kinetic effects effectively limited further melt evolution and prevented the system from reaching equilibrium, rather than a true storage pressure (Cashman and Blundy 2000).
Glass composition and volatile measurements were used to calculate the dynamic viscosity (μ) of the melt phase using the model of Giordano et al. (2008). Caution must be exercised with respect to the absolute values of viscosity as a result of the uncertainty in melt temperature. However, assuming that temperature gradients between the dense and volatile-rich portions of each plug are small, variations in viscosity may be used to assess the properties of the melt in the stratified magma plugs. Assuming a temperature of 980 °C, μ is in the range 105.9–107.5 Pa.s. μ is consistently lower for inflated bombs (105.9–106.5 Pa.s) than for dense and scoriaceous bombs (106.5–107.5 Pa.s), illustrating the compositional and volatile effects on melt viscosity. The degassed regions of the plugs therefore contained melt that was 1–1.5 log units more viscous than the more volatile-rich region of the plugs.
Feldspar microlite compositions
As noted by Hammer et al. (2000) and Preece et al. (2016) at Merapi volcano, the compositions of feldspar microlites span a large range of An content. As plagioclase microlites evolved towards more sodic compositions, they record an apparent increase in temperatures with respect to isothermal sections of the dry ternary solvus at low pressures (Fig. 5c). Following Hammer et al. (2000) and Preece et al. (2016), we attribute this observation not to an actual increase in temperature but to the changing H2O activity in the melt in response to decompression, the resultant shift in the liquidus position and the compositional evolution of the melt phase due to ongoing crystallisation. Feldspar microlite compositions therefore act as a record of the extent of melt degassing and evolution prior to each explosion. At Galeras, they attest that the magma plug that was emplaced prior to the explosion of 17 January 2008 that experienced the longest repose time (554 days) produced the most “evolved” feldspar compositions, including the only observed alkali feldspar crystal. Unsurprisingly, these most evolved microlites were found in the more degassed scoriaceous bombs from that explosion rather than the contemporaneous inflated bombs (Fig. 5). However, in the sample suite as a whole, the lack of correlation between feldspar microlite composition and bomb type implies that the compositional range down to approximately An30 was achieved through degassing-driven crystallisation deeper in the conduit, during ascent prior to plug emplacement at the shallowest levels (< 500 m).
Feldspar microlite textures
The final groundmass textures of volcanic rocks reflect the relative importance of microlite nucleation and growth rates (Williams et al. 1954; Hammer and Rutherford 2002), which are both functions of ΔT (Fig. 10). Effective undercooling ΔT in the context of ballistics produced by vulcanian explosions such as Galeras samples is controlled by melt degassing (rather than cooling), which depends on average decompression rate and decompression style (single-step or multiple-step decompression, Hammer and Rutherford 2002). Textures characterised by high NA and low ɸ are generally thought to result from nucleation-dominated crystallisation prevailing during higher intensity vulcanian eruptions (higher ΔT), whereas textures characterised by lower NA and higher ɸ are likely to result from growth-dominated crystallisation prevailing during lower intensity eruptions (lower ΔT) (Hammer et al. 2000; Brugger and Hammer 2010b; Preece et al. 2016). Feldspar crystal micro-textures in ballistic samples from Galeras volcano show systematic variations in batch textural parameters NA, mean crystal area and S/L over the course of the eruptive period (Fig. 7a–d), as well as variations in three-dimensional textural parameters n0, Nv and Lc (Figs. 7e–h and 9f–h), which must be driven by variations in ΔT. These textural variations coincided with systematic changes in repose time and ejected volume (Fig. 9), suggesting that variations in average ascent rate controlled the observed changes in ejected volume and repose time.
We compare our textural results with the results of continuous isothermal decompression experiments conducted on hydrous rhyodacite by Brugger and Hammer (2010b) in order to estimate average magma decompression rates during this period. The decompressed samples of Brugger and Hammer (2010b) do not generally cover the same NA-ɸ space as most Galeras samples (Fig. 14). However, one experimental charge held at 880 °C and decompressed continuously at a rate of 1 MPa h−1 from 130 to 5 MPa, then held at the final pressure for 915 h (approx. 38 days) matches Galeras samples more closely (Fig. 14). Continuous decompression experiments approximate multi-step conditions with many small steps, relatively small degrees of ΔT at each step (compared to single-step decompression) and long-time steps between each decompression increment. This ascent path tends to produce melts that are far from equilibrium conditions and consequently record unrealistically high closure pressures, as observed in Galeras samples. The final annealing step reflects a possible stagnation period in a magma plug. The compact/euhedral (rather than skeletal) habit of crystals in most Galeras samples also supports slow decompression under modest degrees of ΔT (Lofgren 1974; Couch et al. 2003). The two continuous experiments that plot closest to sample AB22 along a line of constant mean crystal size through the origin were decompressed from 130 to 5 MPa and from 130 to 45 MPa respectively, at an average decompression rate of 10 MPa h−1 with no annealing time. This sample therefore appears to have experienced a decompression history that is distinct from other samples.
Inflated bombs typically show higher NA and NV, smaller mean crystal area and Lc and lower ɸ than dense or scoriaceous samples produced in the same explosion (Fig. 7). This indicates that spatially variable degrees of ΔT existed within the stratified magma plugs and suggests that textural parameters may be used to compare the crystallisation regimes and degree of ΔT prevailing at different depths, as well as in plugs emplaced prior to different explosions. These textural trends indicate higher degrees of ΔT in the magma that gave rise to inflated bombs than in magma that gave rise to contemporaneous dense and scoriaceous bombs. This is in agreement with the higher extent of melt evolution in the degassed region of the plugs (Fig. 4), as high growth rates under moderate degrees of ΔT are expected to produce the most efficient crystallisation regime. High growth rates are also likely to result in crystals with higher aspect ratios (Lofgren 1974; Holness 2014), which are typically observed in dense and scoriaceous samples compared to contemporaneous inflated bombs (Fig. 9e). Textural trends also indicate higher degrees of ΔT overall in magma plugs emplaced from 2008 onwards, prior to the three latest explosions studied here. The average decompression rates estimated in this section suggest that an increase in average magma ascent rate corresponding to a shift from average decompression rates of 1–10 MPa h−1 was responsible for the changing ΔT and resulting crystallisation conditions.
The concave-upwards shapes of CSDs indicate slowing crystal growth rates prior to vulcanian explosions, leading to a steepening of the CSDs at small sizes. This could result from decreasing ΔT in each magma parcel prior to the explosions and associated microlite coarsening or from viscous limitation of crystal growth accompanying increasing ΔT. Decreasing ΔT could result from equilibrium being approached and a decrease in the driving force for crystallisation. However, the disagreement between the closure pressures indicated in the haplogranite ternary (Fig. 13) and the pressures indicated by the volatile contents of groundmass glasses imply that equilibrium was not reached. Rather, the increase in final nuclei population density n0 and steepening CSD slopes (Table 2, Online Resource 5) strongly argues for an increase in ΔT (Zieg and Marsh 2002) associated with slowing growth rates due to increasing melt viscosity. In general, higher ΔT is expressed texturally as higher NA and NV and more tabular microlites with lower mean crystal areas and Lc as a result of increasing nucleation rates and decreasing growth rates (Fig. 10). ɸ also tends to be lower under conditions of higher ΔT but ɸ is not the best criterion to evaluate ΔT as ɸ can increase under conditions of both high growth rates and high nucleation rates and thus is not linearly related to ΔT. This results in the scattered relationship between ɸ and other textural parameters (e.g. Fig. 7b, f).
Sample AB22 is distinct in that its CSD is only slightly concave and closer to a straight CSD with constant growth rate. This sample hosts large numbers of small skeletal microlites and has high n0 with a shallow CSD slope (Table 2). We interpret the CSD of this sample as reflecting a more rapid rate of ascent from the shallow magma storage region, with a more constant microlite growth rate over the timescale of ascent. This suggests a contrasting ascent path characterised by shorter stagnation periods during multi-step ascent in the conduit, producing a CSD that is closer to a straight CSD that might be expected from a continuous, steady ascent rate. This is consistent with the shorter repose periods between explosions prior to the eruption of this sample on 2 January 2010. We interpret the slight curvature of the CSD as a result of the short stagnation period prior to this explosion during which ΔT increased and growth rates decreased. This contrasts with the dense bomb ejected in the same explosion (AB21), which features a more curved CSD as a result of more prolonged multi-step ascent and a stagnation period in the most shallow, degassed part of the magma plug under lower degrees of ΔT.
n0 is higher and CSD slopes are steeper in samples from the latest three explosions studied here (17 January 2008, 20 February 2009, 02 January 2010), supporting the suggestion that ΔT was higher during this period than in samples from the earliest three explosions (11/12 August 2004, 21 November 2004, 12 July 2006) (Table 2). The volume ejected during an explosion generally increased as n0 in dense and scoriaceous bombs increased (Fig. 9b, f). The largest explosion (2 January 2010) ejected material with high n0 in both dense and inflated bombs, with a steep CSD slope in the dense bomb (AB21) and a relatively shallow CSD slope in the inflated bomb (AB22). This suggests that the largest explosions occur when there is a high nucleation rate throughout the plug under conditions of high ΔT, with a steep viscosity gradient in the interstitial melt. This viscosity gradient attests to the rapid formation of a densified plug at shallow levels with relatively low-crystallinity magma below. This could arise in the case of rapid gas loss from the top of the magma column following the previous explosion, followed by rapid, efficient sealing of the conduit. These conjectures are supported by the short repose time prior to this explosion.
Larger volume explosions typically produced textures characterised by lower S/L (more tabular crystals), lower mean crystal area and Lc and higher NA and NV (Fig. 9). Groundmass plagioclase crystallinity ɸ is more variable for the reasons mentioned previously, but larger explosions tend to eject inflated bombs with relatively low ɸ and dense/scoriaceous bombs with relatively high ɸ. Conversely, the smallest explosions are those characterised by modest ɸ, low NA and NV, low S/L, high mean crystal area and Lc and low n0 with shallower CSD slopes. This implies that small explosions occur when ΔT is low (high growth rates prevail overall) and large explosions occur when ΔT is high (high nucleation rates prevail overall). Whereas ejected volumes vary with n0 in dense and scoriaceous bombs, repose times typically vary with n0 in inflated bombs (Fig. 9a, b, f). We interpret this as evidence that processes in the most degassed region of the magma plugs (e.g. densification) control the ejected volume whereas processes in the deeper, more volatile-rich region of the magma plugs (e.g. outgassing efficiency) control the repose time between explosions.
Implications for magma decompression, conduit processes and eruption dynamics
The magma erupted during vulcanian explosions at Galeras volcano was decompressed in a step-wise fashion, at an average rate of 1 MPa h−1, accelerating to 10 MPa h−1 towards the end of 2009. Crystal nucleation began somewhere between a shallow crustal storage area and the level of shallow emplacement (< 500 m). Spatially variable crystallisation driven by differences in ΔT then proceeded within each magma plug. Crystal growth rates were typically rate limited by increasing residual melt viscosities under conditions of increasing ΔT prior to each explosion, leading to concave-upwards CSDs and increasing n0 over time. Higher volume explosions ejected material with a higher range of melt water content, interstitial melt viscosity, ɸ, mean crystal area, NA, NV and n0, consistent with the ejection of magma that was stratified with respect to these properties.
The higher crystal nucleation rates and lower growth rates apparent in inflated bomb samples must reflect crystallisation conditions triggered by a final outgassing step that was greater in magnitude and produced a higher degree of ΔT than the outgassing experienced by scoriaceous and dense samples. To explain this, we invoke a vesiculation event that allowed a greater amount of degassing to occur in this region, followed by densification to produce the dense magma evinced by inflated bomb rinds (Fig. 12a, b). We envision that the magma arriving at the top of the Galeras conduit experiences multiple vesiculation events of different magnitudes on its path to the surface. These multiple vesiculation events accomplish the process of degassing the melt and create notably different final crystallisation regimes within 500 m of the surface that are recorded in feldspar microlite textures. We have direct evidence of one vesiculation event in the form of the pre-eruptive porous network preserved in scoriaceous bombs, and we hypothesise that the ash fraction produced by these vulcanian explosions may hold key information regarding vesiculation events that occurred at greater depth and that may represent the source of overpressure that eventually drove these explosions. Exceptions to the general pattern of higher ΔT in inflated bombs than in dense/scoriaceous bombs may be explained by lateral variations within the magma plugs or by highly variable ascent steps.
The overall higher rates of ΔT prevailing in the three latest explosions studied here likely result from a higher average magma ascent rate and hence a higher average decompression rate during this period. In the 2004–2006 period leading up to the extrusion of the first dome, repose times became longer and the volume ejected decreased, whereas in the 2008–2010 period, repose times became shorter and ejected volumes increased (Fig. 9a, b). We propose that magma degassing became more efficient throughout the first stage of the eruptive period, with accompanying longer repose times between explosions and smaller volumes ejected. This culminated in the extrusion of the first dome in January 2006. Following this, degassing became less efficient and repose times became shorter with larger amounts of material ejected in each explosion. The lack of dense bombs in the second and third explosions of the sequence (21 November 2004 and 12 July 2006) also supports the suggestion that the conduit was in a comparatively open state during this time. The fourth explosion in this study (17 January 2008) also lacked the presence of dense bombs, whereas the fifth and sixth explosions (20 February 2009 and 2 January 2010) produced dense bombs.
These observations suggest that variations in densification rate and magma ascent rate may be responsible for the shift in behaviour observed over the course of the eruption sequence. For example, a low densification rate during the early part of the eruptive period while magma ascent rates were also low could have allowed the porous network to remain open and the magma to successfully outgas, resulting in the extrusion of a dome and vulcanian explosions that became infrequent and smaller. No dense bombs were ejected, only scoriaceous and inflated bombs were produced. Conversely, higher densification rates in the latter part of the eruptive sequence while magma ascent rate was also high may have reduced degassing efficiency, leading to more frequent vulcanian explosions that ejected larger volumes, with a higher proportion of inflated bombs. Further investigation of this densification mechanism is beyond the scope of this paper, but we surmise that the systematic differences in crystal micro-textures resulting from variations in ascent rate may be responsible for rheological differences that control densification rate in the shallow conduit. For example, large, prismatic microlites that are characteristic of low ascent rates at Galeras may be expected to increase magma bulk viscosity by increasing particle-particle interactions (Mueller et al. 2011) and reduce densification rate, leading to more efficient outgassing. Conversely, small tabular microlites characteristic of higher ascent rates may produce a comparatively lower bulk viscosity that promotes rapid densification. This mechanism will be explored in future work. In terms of broader implications for vulcanian explosion dynamics and products at other arc volcanoes, we would expect small volume explosions (~ 105 m3) to be generally associated with long repose times (hundreds of days) and produce mostly scoriaceous-type bombs, often in the presence of an extruded lava dome. Larger volume explosions (~ 106 m3) may destroy existing lava domes and favour rapid plug formation and should be associated with generally short repose times (tens of days) and larger proportions of dense and inflated bombs.
Finally, the unique geochemical and textural characteristics of samples from the explosion that occurred on 2 January 2010 suggest that it represents a different type of explosion related to higher ascent rates that should be considered separately from the others at Galeras and possibly represents a regime shift that contributed to ending the eruptive sequence. This implies that results from the typical explosions of Galeras volcano may not be transferable to all types of vulcanian explosions at other volcanoes if different regimes exist. For example, the lack of pumice clasts produced by explosions at Galeras volcano contrasts with the presence of these products in vulcanian explosions at other volcanoes (e.g. Soufrière Hills, Clarke et al. 2007; Giachetti et al. 2010). The reason for this is not clear and may be related to syn-eruptive processes that might be recorded in the interiors of inflated bombs or in the ash fraction of erupted products. We suggest that the different regimes of vulcanian activity need to be carefully identified to build further understanding of this eruption style.