Introduction

Hekla, located at the southwest flank of the East Volcanic Zone (Fig. 1a), is one of the most active volcanic systems in Iceland. It has featured a series of pre-historic large Plinian eruptions (e.g. Larsen and Thorarinsson 1977; Larsen and Eiríksson 2008b; Gudmundsdóttir et al. 2011), notably the H3, H4, and H5 eruptions at about 3000 BP, 4200 BP, and 7000 BP, respectively (Larsen and Eiríksson 2008a; Thordarson and Höskuldsson 2008), and 18 historical subplinian to smaller Plinian eruptions (Thordarson and Larsen 2007). Due to the frequent historical activity, the explosive behaviour of Hekla is well characterised. In general, the eruption volume and the silica content of the erupted melt increases with increasing repose time (Thorarinsson 1967). Explosive activity at the central volcano may be of two types: (1) a short, intense purely Plinian eruption lasting hours, or (2) a hybrid eruption starting with a Plinian or subplinian phase followed by prolonged Strombolian activity and lava flows (Thordarson and Larsen 2007). Type 1 activity may be confined to the summit area, whereas type 2 activity typically starts there and then expands to the larger vent system (Fig. 1b). All the well-constrained historical Plinian and subplinian phases were short, typically less than a few hours, and typified by high eruption rate in the range from ∼1000 to ∼25,000 m3 s−1 (Gronvold et al. 1983; Höskuldsson et al. 2007; Thordarson and Larsen 2007).

Fig. 1
figure 1

a Map of Iceland with the three main active volcanic regions shown in grey (NVZ North Volcanic Zone, WVZ West Volcanic Zone, EVZ East Volcanic Zone). Location of Hekla indicated by the black triangle. b Map of Hekla with location of the type section for the 1158 eruption (black dot: 64° 1.466′ N, 19° 36.080′ W). Red dashed line shows the position of the Hekla fissure, red star the summit vent and orange stars the shoulder vents. Topographic contour interval 100 m. Isopach lines (Janebo et al. 2016) in green with thickness values in cm

Full size image

Only two historical eruptions have had bulk magma compositions with >66 wt% SiO2: the Plinian eruption in ad 1104 (also known as the H1 event) and the smaller, and to date, less well-characterised eruption in ad 1158. The initial magma compositions were rhyolitic (70–73 wt% SiO2), and dacitic (65–68 wt% SiO2), respectively (Larsen et al. 1999).

The eruption of 19 January 1158 (Thorarinsson 1967) deposited an estimated 0.33 km3 (unconsolidated, 0.2 km3 consolidated) tephra over parts of eastern Iceland (Larsen 1992, 2002). There are no constraints on the duration, but the eruption likely lasted less than a few hours to maximum a day. Based on the isopach map (Larsen 1992), there is no indication of a change in wind direction during the eruption. The ad 1158 eruption is atypical for Hekla. It both contrasts with and is similar to the older purely Plinian eruptions and the younger subplinian opening phases. The eruption was subplinian in dispersal, and the 0.1 km3 lava flow Háahraun, which has the same composition as the ad 1158 tephra, is believed to be associated with the ad 1158 eruption (Sigmarsson et al. 1992). The magma composition is relatively evolved for the quite short 54-year repose time since ad 1104, and the juvenile ejecta contain white, grey, and banded pumices of contrasting vesicularity and crystallinity (Fig. 2).

Fig. 2
figure 2

Photo of pumice types from the Hekla 1158 eruption: white, grey, and banded clasts. Dashed yellow line marks the contact between a grey band in a white pumice clast

Full size image

The presence of white and grey (and banded) pumices has been observed in numerous eruptions at Hekla and elsewhere. Such a difference between the two pumice types at other volcanoes is either compositional (Lirer et al. 1973) or textural caused by differences in the microlite content and vesicularity (Klug and Cashman 1994; Polacci et al. 2001), or a combination of both (Wright et al. 2011).

Several of the pre-historic Hekla eruptions have two distinct pumice populations of contrasting composition, namely, white rhyolitic pumice and dark grey andesitic pumice. In the deposits, a white pumice layer is overlain by a greyish brown pumice layer. Especially in the zoned Hekla H-A to H-Z eruptions (Larsen and Vilmundardóttir 1992; Róbertsdóttir et al. 2002), the shift from yellow-white or greyish-yellow to dark grey pumice is abrupt, without any significant break in grain size and sorting (thereby indicating that these are from a fairly continuous steady eruption). For the ad 1158 eruption, which was much smaller in volume and lower in intensity than most of these zoned eruptions, there is no significant compositional difference between the white and grey pumices (see supplementary Table 1) and both were erupted, in varying proportions throughout the entire eruption. We suggest below that the white (glassy) and grey (microlite-rich) pumices are more likely derived from a single, initially chemically and texturally homogeneous melt by contrasting patterns of ascent and degassing/outgassing in the shallow conduit, rather than resulting from the tapping and orderly evacuation of a chemically zoned magma body.

Methodology

A type section was defined about 5 km downwind from the summit vent (Fig. 1b), where the ad 1158 deposit is the thickest, a little over 1.5 m thick. To investigate changes over time, the deposit was divided into eight equal intervals (Fig. 3a) that were sampled for grain size and four larger equal intervals (Fig. 4a) that were sampled for clast density/vesicularity analysis. Grain size analysis was conducted, using standard sieving techniques, down to 4 ϕ (63 μm, with ϕ = −log2(d) and d being the particle diameter in mm). Median diameter (Mdϕ) and sorting coefficients (σ ϕ) were calculated following Inman (1952). Componentry analysis was done on all clasts >2 mm in diameter, corresponding to a minimum of 85 wt% of each grain size sample.

The density of 100 juvenile clasts, with diameters of 16 to 32 mm, from each sample interval was determined following Houghton and Wilson (1989). A dense rock equivalent (DRE) value of 2350 kg m−3 was used when converting from measured density to calculated vesicularity. Quantitative vesicularity analysis was performed on nine juvenile clasts, representative of the range of pumice types: one pair of modal density white and grey pumices from each stratigraphic level and one additional highly vesicular white pumice, following Shea et al. (2010). Nested image sets were constructed from 1200 dpi scans of the thin sections combined with back-scattered electron (BSE) images at four magnifications (25×, 100×, 250× and 500×) collected using a JEOL-5900LV scanning electron microscope (SEM) operating at 15 kV accelerating voltage and 1 nA beam current. All images were converted to binary format and rectified in Adobe Photoshop, and quantitative vesicle analysis was performed using the ImageJ software (Schneider et al. 2012). The 2-D areal vesicle number density (N A) was adjusted for the presence of phenocrysts and converted to 3-D volumetric vesicle number density (N V) using the method of Sahagian and Proussevitch (1998). N V is referenced to melt volume (N Vm) in order to account for the presence of the vesicles (Klug et al. 2002). In total, 23 images were analysed for each clast. The lower size limit of analysed vesicles included in the vesicle-size distribution was set to 4 μm, equivalent to a cut-off value of 20 and 10 pixels in the 500 (white pumice) and 250 (grey pumice) magnification images, respectively. The effect of the choice of lower size limit of analysed vesicles on the resulting vesicle number density (VND) is discussed in supplementary document 1.

Results

Deposit characteristics

At the type section, the deposit consists of two parts: a lower, coarser part and an upper, slightly finer part (Fig. 3). The juvenile material comprises two types of microvesicular pumice (white and grey in colour), which are present both as discrete clasts and as mingled or banded clasts (Fig. 2). The white pumice is dominant in all samples, and the mingled clasts are generally white with mm to cm-thick grey bands. The lithic component consists of grey and red lava clasts, probably derived from the conduit walls.

Fig. 3
figure 3

a Stratigraphic log of the Hekla 1158 deposit with intervals sampled for grain size and componentry analysis. Samples are numbered from the top to the bottom of the deposit. b Results from grain size analysis. Median diameter is showed in filled circles and sorting in open circles. c Lithic abundance. d Proportion of pumice types (normalised to 100 %) with white pumice in blue, grey pumice in orange, and banded pumice in green

Full size image

All grain size samples are unimodal, with a median diameter in the range of −4.4 to −4.1 ϕ (22 to 17 mm) in the lower half of the deposit, fining to −3.7 to −2.4 ϕ (13 to 5 mm) in the upper part. The deposit is very well to well sorted (σ ϕ 0.83 to 1.4) throughout (Fig. 3b).

The abundance of white pumice increases upwards slightly in the lower section and then decreases in the upper section, whereas the grey and banded pumices show the opposite trend (Fig. 3d). White pumice ranges from 81 to 93 wt% of the juvenile fractions. Wall-rock lithic clasts range in abundance from 3 to 8 wt%. The lithic abundance is high in the lowermost sample, decreases slightly towards the middle of the deposit and then increases towards the top of the deposit (Fig. 3c).

Clast density

All density samples show similar positively skewed unimodal distributions with a denser tail (Fig. 5). The main peak is dominantly white pumice, the denser tail grey pumice, and the banded pumice predictably spans the range of the two endmembers. The mean density ranges from 480 to 600 kg m−3 for the total samples and 450 to 540 kg m−3 for the white pumice only (Fig. 4), corresponding to a mean calculated vesicularity of 74 to 80 vol.% (total) or 77 to 81 vol.% (white pumice only). There is no clear temporal trend in the density data, but the widest distribution within the white pumice (with a range of 310 to 900 kg m−3, 62 to 87 vol.% vesicularity) occurs in the upper middle part of the deposit. The mode of the density distribution also alternates between 400–500 and 500–600 kg m−3 with successive samples (Fig. 5).

Fig. 4
figure 4

a Stratigraphic log of the Hekla 1158 deposit with intervals sampled for vesicularity analysis. Samples are numbered from the top to the bottom of the deposit. b Mean density of the white pumice is shown by open square, and range in density of white pumice is shown by the thick bar. c Range in density of all pumice

Full size image
Fig. 5
figure 5

Pumice density histograms for the Hekla 1158 deposit, with clast types overlain for sample ρ0506

Full size image

Quantitative vesicularity and textural analysis

The nine clasts chosen for detailed textural analysis are all internally heterogeneous, containing populations of large (>150 μm to mm sized) vesicles surrounded by intermediate to small (150 to <25 μm) vesicles (Fig. 6). In thin section, vesicle shapes range from round or elongate to convoluted polylobate. Evidences of coalescence (Fig. 7), such as wall rupture, thin planar melt films, and small vesicles enclosed inside larger vesicles (described as donuts or ‘donut-like’ features in previous studies; Klug et al. 2002; Adams et al. 2006), are present in all clasts. The crystal content was not quantified, but all clasts contain both phenocrysts and microlites.

Fig. 6
figure 6

Range of textures observed in pumice from the Hekla 1158 eruption. Decoalesced images at four different magnifications (a–d 25×, e–h 100×, i–l 250×, and m–p 500×). Vesicles are shown in black, glass in white and phenocrysts and microphenocrysts in grey. Images with blue outline are from white pumice, and images with orange outline are from grey pumice

Full size image
Fig. 7
figure 7

Coalescence features observed in the Hekla 1158 white pumice: wall rupture (R), donuts (D), wrinkling (W), and planar walls (P)

Full size image

The vesicle volume distributions (VVDs) are mostly unimodal with tails of vesicles up to about 3 mm in diameter superimposed on broad modes of approximately 125–200 μm (Fig. 8). The VVDs for the four modal white pumice clasts are very similar. VVDs for the grey clasts are less homogeneous, and the clasts from the stratigraphically lowest (ρ0708) and highest (ρ0102) samples are weakly bimodal. All clasts have sigmoidal cumulative vesicle volume distributions (CVVDs) (Fig. 9), and the median vesicle diameter is between 40 and 100 μm (Table 1).

Fig. 8
figure 8

Vesicle volume distribution for white (left column) and grey (right column) pumice clasts from the Hekla 1158 eruption. Geometric bins for the equivalent diameters were constructed using a scale factor of 100.1

Full size image
Fig. 9
figure 9

Cumulative vesicle volume percent plot for white (top) and grey pumice (bottom). Each line represents an individual clast. For each pumice type, colour scheme from dark to light corresponds to stratigraphic position of sample from lowest to highest, respectively. For the white pumice, the black line is the maximum vesicularity clast

Full size image
Table 1 Vesicularity data for the Hekla 1158 eruption
Full size table

The cumulative vesicle-size distributions (CVSDs) for all clasts are shown in Fig. 10 as log plots of melt-corrected number density (N Vm) of vesicles greater than diameter L. In every sample, the moderate to large vesicle population can be best fit by a power law with an exponent of −3.5 to −4. The much larger number of the smaller vesicles (<25 and <40 μm in the white and grey pumice, respectively), however, is better described by an exponential fit.

Fig. 10
figure 10

Cumulative vesicle-size distributions for white (left) and grey (right) pumice, shown as number of vesicles greater than a given size versus equivalent diameter on a log–log plot. Larger vesicles follow a power law trend, whereas smaller vesicles (<25 and <40 μm for white and grey pumice, respectively) follow an exponential trend. EXP exponent of power law trend

Full size image

The melt-corrected vesicle number densities (N Vm) range from 1.51 × 108 to 1.04 × 109 cm−3 (Table 1). Decreasing the lower size limit of analysed vesicles included in the vesicle size distribution to 2 μm increases the measured N A by a factor of 1.03 to 1.09 but does not significantly change the derivative N Vm values (average increase <10 %; supplementary document 1). Although all are within one order of magnitude, the white pumice clasts have higher VNDs than the grey pumices (Table 1). On plots of the volume ratio of vesicles to melt (V G/V L after Gardner et al. 1996) versus vesicle number density (Fig. 11), the white and grey pumices cluster as two separate groups. The white pumice is characterised by fairly uniform VND despite slight variations in vesicularity, whereas the grey pumice displays a wider range in VND and more uniform vesicularity. There are no temporal trends in the VND, either between or among pumice types. The variation within any one sample is larger than the differences over time or between pumice types.

Fig. 11
figure 11

Volume ratio of vesicles to melt (V G/V L) against vesicle number density referenced to melt (N vm) for selected clasts from the Hekla 1158 eruption. White pumice clasts are shown as blue squares, grey pumice as orange circles. The blue diamond is representative of the higher end vesicularity among the white pumice

Full size image

In addition to the differences in VND and VVD, the white and grey pumice types display contrasting textures (Figs. 6 and 12). The white pumice generally has no fabric, with only small regions where the vesicles are stretched and oriented in a common direction. Its vesicle population is slightly more homogeneous, clustering around an equivalent vesicle diameter of 50–80 μm (Figs. 6e, f and 8) and separated by thinner (2–10 μm) melt films. The grey pumices, in contrast, are less vesicular and have more mature textures, i.e. more complex vesicle shapes and higher abundances of stretched and oriented vesicles, as well as more regions characterised by trains of vesicles (Fig. 6l, o). Vesicles in grey pumice are separated by thicker (5–15 μm) melt films than in the white pumice, and clusters of vesicles in the grey pumice are separated by 20–40 μm thick glass (Fig. 12). These contrasting textures are also accompanied by differences in crystal content. The grey pumice contains greater abundance of macrocrysts and microlites than the white pumice (Fig. 13b, c).

Fig. 12
figure 12

Comparison of rectified images at three magnifications from a pair of white (left column) and grey (right column) pumice clasts from the Hekla 1158 eruption. The clasts are from sample ρ0708 (lowermost sample range in Fig. 4a). Vesicles are shown in black, phenocrysts in grey, and melt in white. Microlites are shown as melt. In the lowest magnification images (top row), the smallest vesicles below the inclusion limit are depicted as melt

Full size image
Fig. 13
figure 13

a Rectified images of contrasting textural domains within grey pumice clasts at two different magnifications. The less mature domain is shown in the middle column and the more mature domain in the right column. Vesicles are shown in black, phenocrysts in grey and melt in white. Microlites are included as melt. For comparison, left column shows images from a white pumice clast at the same stratigraphic level. b Non-rectified BSE images of domains within one grey pumice clast, showing contrasting microlite content (examples highlighted with red arrows) between less mature domain on the left and more mature domain on the right. Both images are at the same scale. c Non-rectified BSE images of white pumice clast, at two magnifications, showing the difference in microlite content compared to grey pumice

Full size image

The grey pumices display further complexity, containing regions characterised by textures of contrasting maturity on length scales of millimetres (Fig. 13). The dominant texture, by area, is more mature, having irregularly shaped vesicles and significantly higher microlite content. The subordinate texture is less mature, with dominantly round and larger vesicles, and fewer microlites, similar to the textures in the white pumice (Fig. 13b). The more mature domain has a higher VND, smaller vesicle median diameter, and image vesicularities <50 %, whereas the less mature domain has a lower VND, larger vesicle median diameter, and image vesicularities >50 %. The two domains are intermingled in every clast; the contacts and distinction between the two are most distinct in the two stratigraphically highest samples.

Interpretations and discussion

Implications of grain size, sorting, and componentry

Small upward fining of grain size (decrease in Mdϕ) and uniform sorting with stratigraphic height (i.e. over time) indicate that the ad 1158 eruption was dry (magmatic) and relatively steady and sustained. After initial rapid onset, the intensity was steady and then declined slightly over the course of the eruption. The observed trend in relative abundance of lithic clasts, together with the grain size data, is interpreted to represent three eruptive stages, namely, initial wall-rock-rich conduit opening and clearing (sample 08), followed by a more stable, steady part of the eruption (samples 07–04) and ending with late-stage vent-wall destabilisation and collapse of the conduit (samples 03–01) as the eruption intensity decreased. The grain size data, as well as the low overall abundance of the wall-rock lithic component (Fig. 3), thus indicate fairly simple eruptive dynamics, similar to that of many subplinian and Plinian eruptions worldwide.

Interpretation of vesicle number and size distribution

A more complex picture of pre- and syn-eruptive dynamics than the one presented above is apparent from the vesicularity data and we explore this in more depth below.

Bubble nucleation

Vesicles start to nucleate at a critical level of volatile supersaturation. Numerical studies and experiments on silicic melts have shown that the VND scales with the degree of volatile supersaturation and thus increases with increasing rate of decompression, i.e. ascent rate (Toramaru 1995, 2006; Mourtada-Bonnefoi and Laporte 1999, 2004; Mangan and Sisson 2000; Mangan et al. 2004b; Hamada et al. 2010). At high supersaturation, nucleation is homogeneous and continuous (Mourtada-Bonnefoi and Laporte 1999, 2002; Mangan and Sisson 2000; Mangan et al. 2004b), resulting in exponential size distributions, where the dominant small vesicle subpopulation represents those vesicles that nucleated last. To a first approximation, the VND thus reflects nucleation conditions and the ascent rate immediately prior to fragmentation.

The white pumice clasts have a very narrow range of VND that suggests a common history of late-stage bubble nucleation under very similar conditions for the sampled portion of the melt. Unimodal volume fraction histograms, dominated by small bubbles (Fig. 8), and smooth sigmoidal cumulative vesicle volume curves (Fig. 9) in the white pumice indicate that most of the vesicle nucleation occurred in a single, relatively short stage close to fragmentation, probably during rapid, non-linear acceleration of magma up the conduit. The high VNDs (108 to 109 cm−3) are comparable to those of more powerful and sustained Plinian phases, e.g. the 1912 Novarupta eruption (Adams et al. 2006) for which Gonnermann and Houghton (2012) inferred that the smallest vesicle population formed within a few 10−1 s prior to fragmentation, at decompression rates of 107 Pa s−1.

In comparison, the grey pumice shows a significantly wider range of VND values, suggesting either that nucleation rates varied more widely or that the signature of nucleation was variably overprinted by other processes. We present evidence in the “Bubble collapse” section that supports the latter interpretation. The grey pumice clasts have broader unimodal (or weakly bimodal) VVDs (Fig. 8), in which the primary mode is less pronounced and shifted towards slightly larger vesicle sizes. The small (<25 μm) vesicle population, however, amounts to 10 to 30 % of the vesicle volume, indicative that nucleation was still occurring within the melt at the time of fragmentation.

Bubble growth and coalescence

During magma ascent, vesicles increase in size by free growth and by coalescence (Gonnermann and Manga 2007 and references therein). Vesicle growth skews the population towards larger vesicles, and coalescence reduces VND.

The VSD trends (Fig. 10), with larger vesicles following a power law relationship and smaller (<25 μm) vesicles following an exponential trend on log–log plots, are similar to those observed in pumice from the ad 1912 Novarupta eruption (Adams et al. 2006). In general, a single pulse of sustained nucleation and growth results in simple exponential trends (Sarda and Graham 1990; Mangan et al. 1993), whereas power law relationships can be produced by a range of processes, including coalescence and protracted or multiple stages of nucleation (e.g. Cashman and Mangan 1994; Gaonac’h et al. 1996; Simakin et al. 1999; Blower et al. 2001, 2002; Mangan et al. 2004a). The combination of exponential and power law trends in all white pumice clasts suggests that a subpopulation of the earliest nucleated vesicles had sufficient time to interact and undergo limited coalescence, while new, small vesicles were still nucleating. The complex shapes observed in the larger vesicle population, however, indicate that coalescence occurred on time scales shorter than the characteristic viscous relaxation time scale (Larsen and Gardner 2000), preventing vesicles from returning to their spherical state prior to fragmentation.

The grey pumice clasts display VSD trends similar to those of the white pumice (Fig. 10); for two of the clasts, however, the exponential trend ends at larger vesicles (30 and 40 μm) than in the white pumice. The shift of the VVD (Fig. 8) towards a weaker mode at slightly larger vesicle-sizes indicates that the grey melt experienced more prolonged time for vesicle growth and coalescence to occur.

VND is not affected by free growth, but coalescence reduces the VND by creating a few larger vesicles at the expense of smaller vesicles. The trends in Fig. 11 suggest that variability within the dominant subpopulation of small bubbles in the white pumice reflects differing degrees of free growth of originally very similar bubble populations. The grey pumice clasts, however, have more complex vesiculation histories. The range in vesicularity of the white pumice is not accompanied by a significant range in VND, and the diversity within samples is larger than the diversity between samples. The very similar numbers of small bubbles suggest that latest-stage ascent of the magma resulting in the white pumice, immediately prior to fragmentation, occurred under very uniform conditions. The single analysed more vesicular clast has a markedly higher vesicularity (diamond in Fig. 11), reflecting an extended role for free growth of bubbles, but a similar VND to the modal samples, consistent with a relatively simple vesiculation history for all the melt that formed the white pumice. This is in stark contrast to the more complex vesiculation history of the grey pumice (see below).

Bubble collapse

As vesicles continue to increase in size and interact, outgassing can occur via formation of permeable vesicle networks (Eichelberger et al. 1986; Mueller et al. 2005) or by elongate pathways formed by cracking and brittle deformation of the magma (Gonnermann and Manga 2003; Rust et al. 2004). The textures observed in the grey pumice (Figs. 6o, p and 13) support the interpretation that the magma experienced partial outgassing and vesicle collapse. Evidence of vesicle collapse includes thickened interstitial melt between vesicles and pinched appearance of vesicles, features similar to those observed in pumice from the ad 1912 Novarupta (Adams et al. 2006) and the ad 181 Taupo (Houghton et al. 2010) eruptions. Similar textures with thickened interstitial melt and lack of small vesicles, resulting from collapse of micropore networks during outgassing, have also been observed in experiments (Kennedy et al. 2016).

Derived vesicle data supports the conclusion that the grey pumice formed from magma that experienced partial outgassing and vesicle collapse. The volume ratio of vesicles to melt (V G/V L) is largely independent of VND, because the small vesicles that drive the VND make a negligible contribution to the total gas volume. Volatile loss through outgassing, however, can collapse large vesicles and thereby decreases V G/V L, whereas free growth increases V G/V L. The lower V G/V L of the grey pumice compared to the white pumice is interpreted to result from onset of vesicle collapse and partial outgassing that is not observed in the white pumice.

Temporal changes during the eruption

The white pumice clasts display several trends with time (i.e. from the stratigraphically lowest to highest sample). VND increases slightly through the first three samples then decreases up section. One interpretation of this change is that the ascent rate, and hence nucleation rate, of the sampled portion of the melt increased slightly over time, followed by a minor decrease towards the end of the eruption. The VVD broadens with time (left side of Fig. 8), and there are corresponding shifts of the CVVD curve (Fig. 9) towards less steep slopes, as well as an increase in the median diameter from 40 to 70 μm with time. We interpret these relatively smooth and progressive changes, both in the VND and in the CVVD, to reflect increasing time for growth and coalescence during the course of the eruption, associated with a decrease in ascent rate. The textural changes are consistent with the notion that the magma for the white pumice formed a coherent batch that gradually and subtly evolved with time during the eruption.

In contrast, the grey pumice shows no clear temporal trends. Rather than the systematic shift in the CVVD curves observed for the white pumice, the entire curve shifts seemingly randomly back and forth between samples (Fig. 9). The VND also varies more widely from lower and higher values (1.5 to 1.6 × 108 versus 3.1 to 4.7 × 108 cm−3; Fig. 11; Table 1). Care should be taken not to infer too much regarding any temporal trends, or lack of trends, from the grey pumice clasts because of the relatively low number of clasts that were analysed. However, we think it is likely that the magma from which the grey pumice formed was unsystematically or accidentally incorporated, by the co-eruptive magma resulting in the white pumice, on short time scales during ascent.

Pumice diversity and conduit heterogeneity

Textural differences between the two pumice types, and their vesicle-size distributions and number densities stipulate that the more mature grey pumice textures could have evolved from a magma resembling that of the texturally less mature white pumice but not the other way around.

The two populations do not merge to form a smooth trend on any vesicularity plot (e.g. Fig. 11). Instead, the textural separation between the white and grey pumice in terms of V G/V L versus VND (Fig. 11), and their existence as banded clasts, suggests that the two experienced contrasting histories in the shallow plumbing below Hekla and came together at a comparatively late stage in order to mingle and then fragment. The principal distinction between the two populations is a marked difference in the V G/V L values (Fig. 11). The grey pumice clasts have markedly lower V G/V L values and slightly decreased VND, consistent with a phase of bubble collapse and partial outgassing not experienced by the magma which formed the white pumice.

The limited textural diversity of the white pumice can be explained by varying amounts of free growth of bubble populations formed under very consistent conditions of bubble nucleation throughout the eruption. The diversity in the grey pumice, on the other hand, can be explained by varying degrees of renewed bubble coalescence after a significant amount of bubbles had already been lost by outgassing. The two pumice types thus represent melts with different histories. The grey melt experienced an extended opportunity for crystallisation and bubble coalescence, accompanied by variable amounts of open-system degassing, manifested as a decrease in VND and broadening of the VSD.

Of the grey clasts for which quantitative textural analysis was performed, sample ρ0304-B02 has the highest vesicularity (58 vol.%; Table 1). In addition, it has the largest median vesicle diameter (100 μm) of all analysed clasts, including the more vesicular white pumice clasts. The more mature domain has generally thicker (5–40 μm) interstitial melt than do the other grey clasts, but the less mature domain is significantly more vesicular than the other clasts. The median vesicle diameter of the more mature domain is comparable to that of the other clasts (∼40 μm), whereas the median vesicle diameter of the less mature domain is distinctly larger than in the other clasts (90 versus 40–60 μm). We interpret the range in vesicle textures observed in the grey pumice as resulting from a complex protracted intermingling process in which melts of contrasting vesicularity and maturity interact in the conduit. These melts continue to mature up until fragmentation, preserving their original integrity.

Eruption model

The textural diversity observed in the pumice population of the Hekla ad 1158 eruption cannot be explained by a traditional Plinian eruption model in which relatively uniform melt rapidly ascends and vesiculates evenly in the shallow conduit. It requires simultaneous fragmentation of melt of contrasting textural maturity and thus residence time in the shallow conduit. We propose that the endmember pumice types correspond to two texturally distinct portions of the melt with different ascent histories that mingled and fragmented together during the eruption. The banded pumice captures the mingling of the two melts during ascent, displaying either sharp interface between white and grey bands or stretched pockets of grey melt in the white melt.

The relatively uniform white pumice is always dominant within the stratigraphy. It represents the newest and more rapidly ascending melt that interacted with melt that had slowed during ascent and experienced increased crystallisation as well as vesicle coalescence and outgassing (Fig. 14). The grey pumice formed by not only slower but also more variable ascent rates. We propose that small packages of more mature melt, along the conduit margins, were overtaken and entrained (close-up ii, Fig. 14b) by the ascending fresh melt over a wide-depth range, thus accounting for the textural diversity and range in VND of the resulting grey pumice. The entrainment process was likely not continuous but occurred repeatedly. As a consequence, some of the rising new melt, together with entrained older melt, slowed sufficiently for coalescence and outgassing to begin, thus evolving into texturally more mature melt at the expense of fresh melt. This incremental process can account for the textural diversity of the grey pumice. In this model, the textural domains of contrasting maturity within many grey pumice clasts resulted from the early mixing of melts of contrasting maturity (Fig. 14b). As the eruption continued, the difference lessened between contrasting grey domains, gradually homogenising the melt filling the conduit.

Fig. 14
figure 14

Cartoon of proposed model for the 1158 Hekla eruption with two stages of ascent prior to fragmentation. a Mature texture (‘grey melt’, close-up ii) developing progressively axisymmetrically along the conduit margins through bubble coalescence and degassing of melt (‘white’, close-up i) that stalled or slowed down during ascent. b Injection of fresh ‘white melt’ (i) which ascends faster than the more mature melt. Entrainment of pockets of fresh melt into mature melt (ii) leading to the domains of contrasting maturity within the grey pumice

Full size image

The relative abundance of white pumice scales inversely with lithic content. The lowest proportion of grey pumice correlates with the inferred peak intensity of the eruption based on grain size (see Fig. 3b–d). We interpret this to reflect a dominant role of the fresh melt during conditions of highest ascent and discharge rate in the middle stages of the eruption. Slower ascent accompanying conduit opening permitted slightly longer residence times during which the melt texture evolved, resulting in the grey pumice. Similar conditions prevailed towards the end of the eruption, when discharge and ascent rates waned.

Conclusions

Based on superficial deposit characteristics, the ad 1158 subplinian eruption of Hekla appears similar to the preceding Plinian eruptions and subplinian opening phases of succeeding eruptions. The variability in the juvenile material, however, indicates a more complex story perhaps linked to the relatively low mass discharge and hence ascent rates. Despite its sustained and relatively steady but short duration, the ad 1158 eruption shows evidence of complex vesiculation and partial outgassing of magma in the shallow conduit. A dominant, newly arrived and erupted (‘white’) melt was physically homogeneous and characterised by a narrow range of VNDs. A smaller portion of the melt (‘grey’) developed more evolved textures along the conduit margin. This melt was then entrained by, and mingled with, the dominant fresh melt.

Similar bimodal assemblages of compositionally identical white and grey pumices are common in subplinian–Plinian deposits elsewhere. White pumice is always dominant, microlite free and characterised by larger populations of bubbles that preserve evidence for late-stage coalescence. Grey pumice clasts have thicker bubble walls, round bubble shapes and generally abundant microlites. Such grey pumice is inferred to be derived by heterogeneous flow and degassing but under diverse conditions that include (i) slowed or stalled ascent, either on conduit margins or across the conduit (Hammer et al. 1999; Adams et al. 2006) and (ii) rapid ascent of lower viscosity melt along conduit margins due to marginal viscous heating (Polacci et al. 2001). During (i), protracted degassing, bubble collapse and gas separation can occur during repose intervals of days (Hammer et al. 1999), whereas dynamic processes of bubble shearing, onset of connected permeability, foam collapse and even melt fracturing during (ii) occur on time scales as short as hours (Polacci et al. 2001).

The magma resulting in the grey pumice is, except in the case of the Pinatubo climactic deposit (Hammer et al. 1999), inferred to develop under conditions of extended residence time in the shallow conduit. This permits the development of local domains, first of extended connectivity and then of bubble collapse and outgassing. This is plausibly postulated to occur under conditions that range from relatively static conditions during significant periods of repose (28–262 min before the climactic phase at Pinatubo; Hammer et al. 1999) to ongoing dynamic degassing and outgassing during sustained, although sometimes unsteady eruption. Not surprisingly, relatively static conditions appear more common with low mass eruption rate, such as during many subplinian eruptions. In the case of the Hekla ad 1158 eruption, where a sustained discharge rate is clearly indicated, we favour the latter scenario. This type of eruption therefore lie in the middle ground between more powerful Plinian events characterised by relatively few and very closely spaced eruptive phases, e.g. the first two Plinian phases of the Novarupta ad 1912 eruption (Adams et al. 2006) and the Taupo ad 181 eruption (Houghton et al. 2010) and weaker eruptions in which the explosive interval is divided into relatively short pulses separated by repose intervals of less than 1 day (Hammer et al. 1999). During the more powerful phases, the extent of bubble nucleation, growth, and coalescence remains relatively uniform in time and space. More texturally mature melt is therefore not developed, resulting in a uniform population of white pumice clasts. In contrast, during the weaker eruptions, the wider range of ascent and storage conditions creates a more diverse clast assemblage that still contains white pumice but also contains a population of grey pumice.