Fractures in a trachyandesitic lava at Öræfajökull, Iceland, used to infer subglacial emplacement in 1727–8 eruption

Journal of Volcanology and Geothermal Research 288 (2014) 8–18

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research journal homepage: www.elsevier.com/locate/jvolgeores

Fractures in a trachyandesitic lava at Öræfajökull, Iceland, used to infer subglacial emplacement in 1727–8 eruption A.E.S. Forbes a,⁎, S. Blake a, H. Tuffen b, A. Wilson c,1 a b c

Department of Environment, Earth and Ecosystems, The Open University, Milton Keynes, Bucks MK7 6AA, UK Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK School of Earth, Atmospheric and Environmental Sciences, University of Manchester, Williamson Building, Oxford Road, Manchester M13 9PL, UK

a r t i c l e

i n f o

Article history: Received 3 August 2014 Accepted 2 October 2014 Available online 14 October 2014 Keywords: Pseudopillow fracture systems Columnar jointing Platy fractures Brittle fracture Ductile fracture Lava–ice interaction 1727 Öræfajökull eruption

a b s t r a c t We present detailed field observations of cooling fractures in a small-volume trachyandesitic lava, informally named the Slaga lava, on the south west flank of Öræfajökull volcano, south east Iceland. Columnar joints, pseudopillow fracture systems, and curved platy jointing occur in the lava, whose exposed section is approximately 600 m in length and generally 2–3 m in thickness. Columnar jointing may occur at the base of flow lobes, whereas pseudopillow fracture systems occur throughout the lava in an outer, glassy, fractured carapace, and curved platy fractures occur in the centres of larger flow lobes. Pseudopillow fracture systems, composed of a single master fracture and multiple subsidiary fractures formed normal to the master fracture, are of two types: G-type pseudopillow fracture systems have very narrow striae (chisel marks) on their master fractures, indicative of rapid cooling and brittle fracture; SR-type pseudopillow fracture systems display alternating smooth and rough master fracture surface textures, evidence of alternating brittle and ductile fracture propagation mechanisms. Subsidiary fractures in both types show curved striae on their fracture surfaces, which enable the determination of fracture propagation directions. Pseudopillow fracture systems are thought only to form in the presence of water, including water caused by the melting of ice and snow. The curved platy fractures display prominent river lines and may have resulted from cooling contraction, post-emplacement degassing, flow deflation or shearing in the flow against the outer solid crust of the flow during inflation. Due to recent advances in the understanding of the formation mechanisms of pseudopillow fracture systems they, and the other fractures present in the flow, can be used to reconstruct the cooling environment. The lava is inferred to have been emplaced within subglacial drainage channels incised into or beneath a thin alpine-type glacier, with coolant infiltrating the lava radially from all sides and ponding or draining of water at or along the lava base. We link this flow to the 1727–8 eruption of Öræfajökull volcano on the basis of historical records, which describe lava effusion in this approximate location and striking chemical compositional similarity between 1727 tephra and the Slaga lava. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Cooling fractures can be used to infer the cooling environment and emplacement mechanisms of a lava flow. Three main types of fractures are the focus of this study: columnar jointing; pseudopillow fracture systems; and curved platy fractures. Columnar jointing is a common fracture type in lava, which can form in many environments ranging from arid to subglacial. In some instances columnar jointing may be associated with a particular environment, such as in felsic lavas where it is often associated with lava–ice interaction (e.g. Lescinsky, 1999; ⁎ Corresponding author at: Chemostrat Pty, Unit 12, 33 Delawney Street, Balcatta, Perth, WA 6021, Australia. E-mail addresses: [email protected] (A.E.S. Forbes), [email protected] (S. Blake), [email protected] (H. Tuffen), [email protected] (A. Wilson). 1 Present address: Chemostrat Pty, Unit 12, 33 Delawney Street, Balcatta, Perth, WA 6021, Australia.

http://dx.doi.org/10.1016/j.jvolgeores.2014.10.004 0377-0273/© 2014 Elsevier B.V. All rights reserved.

Lescinsky and Fink, 2000; Tuffen et al., 2001, 2002; Stevenson et al., 2009). Pseudopillow fracture systems are a combination of two distinct fracture types: a single curvi-planar master fracture and multiple evenly spaced subsidiary fractures that are arranged perpendicular to the master fracture (Lescinsky and Fink, 2000; Forbes et al., 2012). Pseudopillow fracture systems form only in close association with water, in any form (Watanabe and Katsui, 1976; Lescinsky and Fink, 2000; Mee et al., 2006; Lodge and Lescinsky, 2009; Tucker and Scott, 2009; Forbes et al., 2012). Platy fractures are more enigmatic in that they do not have a strong environmental association and may have a number of different causes depending on the circumstances of their formation (e.g. Bonnichsen and Kauffman, 1987; Walker, 1993; Spörli and Rowland, 2006; Tuffen et al., 2013). A long-standing advantage of using columnar jointing to interpret lava emplacement environments had been that fracture propagation directions can be easily deduced from striae (chisel marks) and these have been understood for some time (Ryan and Sammis, 1978;

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

9

DeGraff and Aydin, 1987); however this was not the case, until recently, for pseudopillow fracture systems (Forbes et al., 2012). The aims of this study are twofold: firstly to describe in detail a lava flow field (Slaga, Öræfajökull, Iceland) and its features related to cooling and emplacement; secondly to use recent advances in knowledge gained through the study of cooling fractures, particularly pseudopillow fracture systems, to interpret its emplacement environment. This ‘proof of concept’ study demonstrates that detailed investigations of cooling fractures can give pertinent information about the emplacement environments of lava flows. This is necessary as we so rarely see lava interacting with ice or snow (e.g. Edwards et al., 2012a, 2013). Additionally, we show that this lava was likely emplaced during the 1727–8 eruption of Öræfajökull, using historical records and geochemical data. 2. Geology of the Slaga lava The lava in this study is located on the south western flank of the Öræfajökull central volcano, Iceland (Fig. 1). It has no name that we can discern but occurs on the north eastern edge of a hill called Slaga (approximately 63.944290° N, 16.746323° W), so the lava has been informally named the Slaga lava (Fig. 1a,b). XRF data (Table 1, supplementary material) shows the flow to be trachyandesitic in composition. The lava is highly fractured, displaying pseudopillow fracture systems, curved platy fractures and some columnar jointing. The full extent of the flow is difficult to determine as a glacier tongue to the north, and a large ravine to the east, limit the accessible study area. The lava is locally poorly exposed, as scree and glacially deposited debris, including large glacially striated boulders, are scattered across the top of the flow. Some material has clearly been eroded from the flow surface and margins, enabling the detailed study of flow lobes in cross-section. The lava only recently emerged from the receding Kotárjökull glacier, as determined from historical photographs (Guðmundsson et al., 2012; Guðmundsson, 2014). The Slaga lava sits on now lithified polymict sediment, which was probably glacially deposited. No vents were found, but as the lava appears to have flowed downhill away from the glacier (flow direction determined by the topography) it is likely that the vent is currently concealed by the glacier. The lava occurs as multiple lobes across a shallow area with a mapped maximum width of 130 m, and varying thickness due to the lobate nature across the mapped area. Single lobes can be between 0.5 and 5 m thick, but in general the flow is not more than ~2–3 m thick on average, across the area. Lobe widths were more difficult to measure due to erosion, but are generally ≤6 m. Lobes are vertically stacked (Fig. 2), and also laterally offset, indicating that the lava flowed over and around existing lobes (e.g. Fig. 2a). At its furthest downslope extent the flow is funnelled into a narrow valley defined by the polymict sediment, becoming just ~2 m wide. At this location the flow is eroded, revealing a cross section of a lava lobe. Only a few isolated outcrops are preserved downslope (south) of this point, demonstrating that the flow originally continued beyond the last exposure of the main flow body. The trachyandesite lava forming these lobes is variably vesicular, with thinner lobes being more vesicular and larger lobes usually displaying a vesicle-free outer carapace of fractured lava with a central or upper vesicular part. Vesicles range from spherical to highly elongate and in size from sub-millimetre to ≤4 cm in diameter, with the elongate vesicles generally being the largest in size. Crystallinity also varies greatly, with the central parts of lobes being holocrystalline, and other areas being glassy. 3. Description of fracture systems in the Slaga lava There are four different fracture types/systems in this flow: (1) smooth–rough (SR), and (2) G-type pseudopillow fracture systems, (3) columnar jointing and (4) curved platy fractures.

Fig. 1. a) Iceland with outline of the Vatnajökull glacier, the box on the southern side of the glacier expands to Fig. 1b. b) Outline map of the Öræfajökull glacier. The hill Slaga is outlined and labelled, next to the Kotárjökull glacier. Boxed numbers show the main road, route 1. The coastline is labelled on the land side. The box over the northeast corner of Slaga expands to Fig. 1c. c) Dotted lines show the mapped lava area, white dots to the south of this show isolated outcrops of the Slaga lava flow. Flow direction is southwards. Image from Google Earth.

3.1. Pseudopillow fracture systems in Slaga Pseudopillow fracture systems are an association of two fracture types: each system comprises a single curviplanar master fracture and multiple subsidiary fractures that form perpendicular to the master fracture (e.g. Watanabe and Katsui, 1976; Lescinsky and Fink, 2000; Forbes et al., 2012). Subsidiary fractures can occur as either polygonal joint sets or as long, planar, subparallel collections of fractures (Fig. 3).

10

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

the Slaga lava flow (Tangahraun, Snæfellsnes, west Iceland; Forbes et al., 2012). In that study pseudopillow fracture systems were divided into three types based on the characteristics of their master fractures: (1) F-type have master fractures that form parallel to flow banding and show a dimpled surface texture; (2) X-type have master fractures that cross-cut flow banding and show a transition in master fracture surface texture from smooth to rough; and (3) G-type have closelyspaced striae across the glassy master fracture surfaces, and are not defined by their association with flow banding. The Slaga flow contains two types of pseudopillow fracture systems: G-type, and a previously undescribed type that has a smooth–rough master fracture surface texture. While the latter resemble X-type master fractures, visible flow banding is scarce in the Slaga flow, so that X- and F-type terminology cannot be employed. Instead they have been termed smooth–roughor SR-type master fractures. These SR-type pseudopillow fracture systems are defined solely on the basis of master fracture surface texture.

Fig. 2. a) Upper lobe, on top right of image, flowing over lower lobe, person standing on the lower lobe. b) Three lobes stacked on top of one another, dashed line shows lobe boundaries. Two thinner, more densely fractured lobes occur at the base with a thicker lobe on top, showing basal crude columns on upper lobe. Ruler 50 cm.

Pseudopillow fracture systems are present throughout much of the observable lava at Slaga. Pseudopillow fracture systems have been previously studied in detail in a trachyandesite lava of a very similar composition to that of

3.1.1. SR-type pseudopillow fracture systems SR-type pseudopillow fracture systems occur in all the observed lobes of the Slaga flow (Fig. 3). SR master fractures display a sharp, generally straight, transition between smooth and rough surface textures, with river lines (e.g. Pugh, 1967) commonly occurring on the smooth part of the fracture surface (Fig. 4a). These terminate abruptly against the rough portion of the fracture surface. The master fracture itself may be either fairly planar or strongly undulatory and may intersect with other master fractures. Glassier zones a few centimetres thick sometimes occur on either side of the master fracture demonstrating more rapid cooling along the plane of the master fracture (Fig. 4b). Fragmentary material, composed of mm-sized clasts of juvenile and other lithic material and small amounts of clay, is commonly seen trapped in SR-type master fractures. The subsidiary fractures that form in conjunction with SR-type master fractures occur in two distinct geometries: polygonal arrays of fractures and long planar subparallel or fanning fracture sets (Fig. 3). Both types of subsidiary fractures form normal to the master fracture on each side, and propagate a similar distance from the master fracture. In a cross section of a pseudopillow fracture system, taken normal to both the master and subsidiary fractures, the two subsidiary fracture types are indistinguishable. Both subsidiary fracture types are common in SR-type pseudopillow fracture systems in this flow. Long planar subparallel subsidiary fractures (Fig. 3b) display striae (chisel marks; Fig. 5a, b) on their fracture surfaces, demonstrating their formation by incremental advance of brittle fracturing (Ryan and Sammis, 1978; DeGraff and Aydin, 1987). The striae on these subsidiary

Fig. 3. SR-type master fracture surfaces showing: a) Polygonal subsidiary fractures, with a very clear smooth to rough transition on the master fracture face indicating propagation from left to right. b) Subsidiary fractures fanning outwards in an upwards direction, the master fracture here has a rough surface texture.

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

11

Fig. 5. a) Pseudopillow fracture system showing face-on subsidiary fractures with curved striae on subsidiary fractures. The master fracture of this pseudopillow fracture system is marked as a dashed line across the centre of the figure. b) SR-type master fracture and subsidiary fractures with curved striae. Fig. 4. a) River lines on SR-type master fractures at the smooth–rough boundary. River lines start as small closely spaced steps on the fracture surface and merge to form fewer but larger steps. They show fracture propagation downwards here. b) Subhorizontal SRtype master fracture at the bottom of the figure. The lava closest to the master fracture is very glassy and becomes less glassy within a few centimetres upwards. This demonstrates that master fractures are likely conduits for water or steam, causing extensive cooling where they flux through the fractured lava.

fractures are curved which appears to be a common feature on subsidiary fracture surfaces in pseudopillow fracture systems from all compositions of lava (Forbes et al., 2012; Fig. 5). Hackle (plumose structures), a common feature on striae and other brittle fracture surfaces (e.g. Ryan and Sammis, 1978; DeGraff and Aydin, 1987), show fracture propagation direction in a convex forward direction as previously described by Forbes et al. (2012). Both master and subsidiary fractures show propagation directions, always in the same direction. 3.1.2. G-type pseudopillow fracture systems G-type master fractures in the Slaga flow are characterised by very closely spaced (b 1 mm) striae on the master fracture surface (Fig. 6a and b) and are identical to those described by Forbes et al. (2012). Subsidiary fractures in G-type pseudopillow fracture systems are always of the subparallel type. G-type pseudopillow fracture systems do not occur in every lobe, they feature only in very glassy parts of lobes and always at the base. G-type master fractures occur together in pairs, or as single curved fractures that form elongate lobate, or sausage-like, bodies at the base of some lobes (e.g. Mee et al., 2006; Fig. 6 c and d), and only one example was found on a lobe side, where it had propagated inwards. Where G-type master fractures occur in pairs to form lobate bodies they have an open gap between the two master fracture surfaces, appearing to form as pull-apart structures (Fig. 6b). This is not generally

seen on the SR-type master fractures. G-type master fractures occasionally transform into SR-type master fractures as they propagate inwards into the flow (Fig. 6a). The striae on G-type master fracture surfaces show hackle indicating fracture propagation generally inwards from the outside of the lobe. Striae on G-type subsidiary fractures are curved, like those in SR-type systems. They are also very closely spaced, ≤1 mm apart, like those on G-type master fractures. Hackle occur on these curved striae and indicate the same direction of fracture propagation in relation to the shape of curved striae as in SR-type pseudopillow fracture systems and in G-type systems; this is always inwards and upwards from the base of lobes. 3.1.3. Measurements from subparallel subsidiary fractures The widths of curved striae and the spacing between subsidiary fractures were measured from SR-type pseudopillow fracture systems at various locations around the entire flow. The spacing of curved striae and subsidiary fractures were measured as close to the master fracture as possible (Fig. 7). These measurements were not obtained for G-type systems as striae were too small to measure accurately with callipers in the field. A positive correlation between subsidiary fracture spacing and striae widths was found (Fig. 8). These data are plotted with similar measurements of pseudopillow fracture systems from the Forbes et al. (2012) study of the Tangahraun lava. The Slaga lava pseudopillow fracture system data plot in the same region as this previous data, and fall along a similar trend. 3.2. Columnar jointing Crude, stubby columnar jointing or polygonally arranged fractures occur at the base of some lobes (Fig. 9). These tend to form in conjunction

12

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

Fig. 6. a) G-type master fracture surface at the bottom of the figure with closely spaced striae. This transforms into an SR-type master fracture showing a smooth to rough transition, at the top of the image. b) G-type master fracture surfaces showing straight, very closely spaced striae on the fracture surface, indicating rapid brittle incremental fracture. The two sides of the master fracture have pulled apart, leaving gap between the master fracture surfaces. c) G-type master fractures in elongate lobate forms at the base of a lava lobe. Person for scale. d) Close up view of G-type lobate forms, showing extremely glassy margins where the G-type master fractures occur, ruler is 25 cm.

with upward propagating SR-type master fractures so that subhorizontal striae can be seen on some fracture surfaces and smooth–rough transitions are seen on others. Columnar jointing often forms at the base of lava emplaced on top of older lobes, the top of the lobe with a columnar jointed base commonly displaying downward propagating pseudopillow fracture systems.

Fig. 7. Line diagram of half a pseudopillow fracture system (the other half would be a mirrored image above the master fracture) showing the locations of striae measurements (S) and subsidiary fracture spacing measurements (W), plotted in Fig. 8.

In some larger lobes, columns may be pervasively cut by subhorizontal platy fractures after the columns have formed (Fig. 9a). The fracture order can be determined by cross-cutting relationships, whereby the platy

Fig. 8. Subsidiary fracture spacing and striae widths from SR-type pseudopillow fracture systems from the Slaga lava flow (filled diamonds), and from F-, X- and G-type pseudopillow fracture systems (open circles, squares and triangles respectively) of the Tangahraun lava flow (Forbes et al., 2012). They show a positive correlation and a close overlap in the data from pseudopillow fracture systems in these two different lava flows.

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

13

fractures associated with them. These are later cross-cut by the curved platy fractures in the centre of the lobe, and so must precede the platy fractures. 4. Generalised description of fractures and features within lobes

Fig. 9. a) Columnar jointing at the base of an approx. 6 m thick lobe in the northernmost accessible part of the flow. The columns are cross-cut by platy fractures. b) Columns with clear striae at the base of a lobe that rests on top of another lobe from this same flow. Ruler is 1 m.

SR-type pseudopillow fracture systems generally occur around the dense outer margins of lobes and in smaller lobes they occur throughout the entire lobe. The propagation directions of master and subsidiary fractures show that pseudopillow fracture systems generally propagated inwards towards the lobe centres in a radial fashion, whereby they propagated up from the base, inwards from the sides and downwards from the top. The thickness of the pseudopillow-fractured outer carapace of lobes varies, being commonly thinner on the top, usually only a few tens of centimetres thick, and occasionally absent. However, it is unclear to what extent this reflects variable initial lobe thicknesses, rather than differential erosion. At the sides and towards the base, the pseudopillow fractured outer carapace is considerably thicker, up to 1 m thick, surrounding the platy-fractured lobe core. The outer fractured carapace on lobes is generally glass-rich, but the proportion of glass is spatially variable. G-type pseudopillow fracture systems are not found in all lobes but where they do occur they form at lobe bases. Where many lobes have been superimposed, the G-type pseudopillow fracture systems occur in greatest abundance in the lowermost lobes, and tend to be entirely absent in the upper lobes. They occur only in very glassy lava, and hackle on master and subsidiary fractures demonstrates that fractures propagated inwards, and sometimes upwards, from the base. Columnar jointing does not occur in all lobes, but is more common at the base of upper lobes where it may transform into a region of SR-type pseudopillow fracture systems or curved platy jointing above. Curved platy fractures occur in the cores of the larger lobes in lava that is more crystalline than the outer pseudopillow fractured carapace. They are often associated with highly vesicular lava in their upper portions and can terminate against earlier-formed SR-type master fractures, which lack subsidiary fractures. 4.1. Other lobe features

fractures terminate against the fractures that define the columns. So the platy fractures must have formed after the column-bounding fractures. These fractured column zones then transition upwards into zones dominated by pseudopillow fracture systems and irregular blocky jointing. This type of columnar jointing, cut by platy fractures, occurs more commonly in the northern part of the Slaga flow. 3.3. Curved interior fractures The centres of larger lobes are commonly dominated by more massive lava with curved platy joints spaced 1–10 cm apart (Fig. 10a and b). Smaller lobes (generally b 1 m in diameter in cross section) do not show this platy jointed interior. The platy joints form concentric layers in the centre of lobes, being approximately parallel with the lobe edges. Their fracture surfaces commonly show extensive river lines spread across the fracture surface (Fig. 10c), indicating these were brittle fractures. River lines here are much more pronounced and expansive than those seen on SR-type master fractures. Platy jointed lava in lobe centres is less glassy and coarser grained than more marginal lava. The centre is sometimes highly vesicular in the upper and outer parts of the platy jointing zones, with large, elongate, stretched vesicles (Fig. 10d). Platy jointed lava generally has a reddish-brown colour rather than the black or dark-grey of the outer part of lobes. These curved platy fractures are very much like those described by Mee et al. (2006) from the zone 2 andesite lavas of upper Santa Gertrudis valley, at Nevados de Chillán volcano, Chile. A few SR-type master fractures propagate from the base upwards, into the interior parts of large flows. They do not have any subsidiary

A number of other features locally appear at the flow base, including large rounded pillow-like lobes ≤1 m in diameter, which generally have glassy, vesicular margins and more vesicular interiors (Fig. 11a and b). They may occur directly beneath a platy fractured lobe centre or beneath pseudopillow fractured lava. Glassy peperite is commonly associated with these pillow-like lobes, often occurring beneath them (Fig. 11d) in a polymict host. Smaller lobate bodies also occur at the base of flow lobes, commonly as “fingers” of lava that can be seen to have intruded into the underlying sediment (polymict), these show chilled margins (Fig. 11c). Other small lobate bodies were observed beneath a larger lobe, surrounded by coarse-grained breccia containing a fine grained muddy matrix. The clasts in this breccia are small blocks of the lava generally b10 cm across. This may represent a basal flow breccia, a coarse peperite-type facies or a mixture of both, formed by interaction between underlying polymict sediment and the lava base. Extensive breccia deposits, consisting of angular lava blocks ≤15 cm in diameter, occur at the base of some lobes. That this breccia is not ubiquitous may reflect either only local formation and preservation, or poor exposure. However, breccia at lava bases and tops is generally absent where multiple lobes are superimposed. Small isolated areas of vesicular frothy textures and vesicular features showing ductile tearing of the lava occurred in some parts of lobe margins around the base and sides (Fig. 12). These are not linked with any particular type of fracturing. In some lobes highly fractured lateral zones are observed, approximately 1–3 m long and 5–50 cm high, consisting of more densely

14

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

Fig. 10. a) End on view of concentric curved platy fractures in the centre of a lava lobe. Notebook is 20 cm long. b) Oblique view of curved platy fractures in the centre of a large lobe, surrounded by pseudopillow fractured outer carapace. c) River lines on a platy fracture from the centre of a lobe. Notice how they merge and their topography become larger from right to left, indicating fracture propagation from right to left. d) Close up of large stretched vesicles from the top of a platy fractured interior, ruler is 20 cm.

fractured, glassier lava within the fractured outer carapace (Fig. 13). These fracture zones may show extremely glassy selvages on the lava, ≤ 1 cm thickness (Fig. 13b), and usually have a densely fractured, brecciated appearance (Fig. 13c). The fractures in the lava surrounding these features are mostly SR-type pseudopillow fracture systems, but crude columnar joints also occur. These fractures always propagate away from these densely fractured and glassy areas, and so may propagate contrary to the overall expected propagation directions (Fig. 13a).

5. Discussion 5.1. Interpretation of fractures and features River lines, as seen on SR-type master fractures, occur as a result of mixed mode I and III fracture and can be used to demonstrate a fracture propagation direction (e.g. Hull, 1993). In all cases observed at Slaga (and in the pseudopillow fracture-bearing Tangahraun lava Forbes

Fig. 11. a) Large rounded lobate pillow-like bodies with glassy and vesicular margins at the base of a lava lobe, notebook is 20 cm long. b) Close up on the glassy and vesicular margin of these same lobate features, ruler 10 cm for scale. c) Glassy rimmed lava lobe intruded into polymict beneath the lava flow, ruler 50 cm. d) Peperite, fragments of glassy lava in sediment, formed by the interaction of the Slaga lava with polymict sediment beneath, ruler 6 cm.

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

15

Fig. 12. a) Ductile drooping and/or tearing textures. b) Frothy vesicular lava.

et al., 2012) this fracture propagation was from the smooth to the rough part of the fracture surface (Fig. 4). River lines are a feature of brittle fracture, the fact that they terminate abruptly on the smooth part of the fracture surface at the smooth–rough boundary indicates that the rough portion of the master fracture may be formed by ductile fracture as postulated previously (Forbes et al., 2012). Both master and subsidiary fractures show propagation directions, always in the same direction, allowing general fracture forming directions to be deduced. They show that SR-type pseudopillow fracture systems generally propagate upwards from the base, downwards from the top and inwards from the sides of flow lobes. In other words, they always propagate inwards, towards flow lobe centres. Fragmental material in SR-type master fractures has been observed previously where it was postulated as the result of fluxing of particlebearing coolant through the master fracture (Lescinsky and Fink, 2000; Mee et al., 2006). In this case the trapped material appears to be caused by a combination of fluxing of particle-bearing coolant through master fractures, and from later deposition of glacial sediment during glacier retreat and melting. Lescinsky and Fink (2000) observed basal breccias to be either thin or absent in ice-interaction lavas of intermediate to silicic compositions. Therefore the general lack of basal breccias between lobes is not unexpected. Lateral densely fractured zones (Fig. 13, Section 4.1) demonstrate a local perturbation of the cooling regime with substantially more rapid cooling in these regions. They may have been caused by ice blocks becoming entrained within the flow during its emplacement (e.g. Lescinsky and Fink, 2000; Skilling, 2009; Graettinger et al., 2012). However large open cavities of the type described by Skilling (2009) or Lescinsky and Fink (2000) were not observed.

Fig. 13. a) Side of a lava lobe showing pseudopillow fractured outer carapace and some platy fracturing at the top. There are two particularly interesting areas in this figure: one higher up shows extremely glassy selvages in an area of highly fractured lava (marked as ‘Glassy’); the other below this shows a lateral ‘Fracture zone’ of highly fractured, almost brecciated, lava containing curved lobe shapes. Double headed arrows show fracture propagation directions, which are perturbed by these features and are always away from these areas. These are interpreted as ice entrainment features. b) Close up of the glassy area, showing selvages of glass on the lava. c) Close up of part of the lateral fracture zone, showing the brecciation and curved shapes within this zone.

5.2. Curved platy fracture formation mechanisms Platy fracturing has been observed previously in lava flows that have interacted with ice (e.g. Lescinsky and Sisson, 1998; Mee et al., 2006; Spörli and Rowland, 2006; Tuffen and Castro, 2009). The curved platy fractures in the Slaga flow appear to be identical in morphology and

16

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

occurrence to those described by Mee et al. (2006), including their absence in the smaller lobes. However, platy fracture is not confined to ice-contact lavas, being also found in the crystalline interior of subaerial rhyolitic lava (Tuffen et al., 2013). Curved platy fractures form in the much more coarsely crystalline core of the flow, where the glassy areas or pseudopillow fracture systems that characterise the other parts of the flow lobes are absent. This more coarsely crystalline centre may be caused by an inability of coolant to infiltrate into the cores of larger flow lobes, resulting in much slower cooling and a ‘dry’ environment. This could result from the emplacement of another lobe on top of the cooling lobe, blocking off the supply of coolant to the top of the lobe beneath. Alternatively this may reflect a gap melted into the ice above the flow, retarding coolant supply to the top of the flow with a lack of direct contact with the ice; or late stage inflation when flow margins are quenched and stalled but tube-fed lava supply continues (e.g. Tuffen et al., 2013). Spörli and Rowland (2006) suggested that platy fractures may be formed by detachment of the outer from the inner parts of the flow due to deflation, while Tuffen et al. (2013) observed this mechanism but during flow inflation. In this case SR-type master fractures crosscut the platy fractures and so must have formed first. SR-type master fractures form in lava that is still very hot, resulting in ductile fracture textures across their surfaces (Forbes et al., 2012). These SR-type master fractures are undeformed and fairly planar indicating that they have not been affected by continued flow of the lava. The flow interior would then have had to cool further before the curved brittle fractures could form, without deforming plastically or flowing away. Therefore the curved platy fractures are unlikely to be the result of inflation or deflation. For the same reason platy fractures are also unlikely due to late stage shear of the lava flow (Bonnichsen and Kauffman, 1987). Platy fractures in this lava may instead result from cooling contraction, likely focussed into their current geometry by subtle variations in flowbands, or by microlite orientation alignment (e.g. Walker, 1993), perhaps due to shearing in the flow before it solidified. River lines are very prominent on platy fracture surfaces (Fig. 10c). River lines result from mixed mode I/III fracturing and consist of a series of steps on the fracture surface aligned in the direction of crack propagation (e.g. Pugh, 1967; Sommer, 1969). These river lines are a series of closely-spaced steps that merge to form larger, more widely-separated steps. They are caused by the crack trying to overcome its inability to twist, so it forms steps which function as individual crack planes, with a very small rotation of the fracture plane across the boundary (e.g. Hull, 1999). The river lines likely form due to the strongly curved nature of these fractures, as they allow the fracture plane to effectively twist. 5.3. Columnar jointing Columns at the base of a flow lobe indicate that fairly even cooling occurred from the base upwards in a uniform stress environment. Locally in this lava, columnar jointing occurs in the higher lobes within a stack of superimposed lobes. This may result in a drier, more slowly and evenly-cooled lobe base, as coolant cannot pond within porous, fractured underlying lava flow lobes, creating viable conditions for columnar joint formation. These columns may occur in a flow that has pseudopillow fracture systems propagating downwards from the top of the flow, indicating a coolant supply to the upper parts of the flow (Forbes et al., 2012). In some cases striae-bearing column-bounding fractures occur in conjunction with SR-type master fractures. This has been observed previously (Forbes et al., 2012) and may indicate some degree of coolant infiltration and/or non-uniformity in the stress field where this occurs. Columnar jointing locally transforms upwards into pseudopillow fractured lava. This may indicate an upward increase in the input of coolant to the cooling lava flow, and possibly a related increase in the cooling rate, or perhaps a change to a non-biaxial, disordered stress field.

5.4. Pseudopillow fracture systems — palaeo-environmental indicator Pseudopillow fracture systems have previously been linked to the presence of an additional coolant in the emplacement environment: either ice, snow or liquid water (Watanabe and Katsui, 1976; Lescinsky and Fink, 2000; Mee et al., 2006; Lodge and Lescinsky, 2009; Tucker and Scott, 2009; Forbes et al., 2012). SR-type pseudopillow fracture systems are the dominant fracture type in the Slaga lava. They occur in every lobe in the outer fractured, glassier carapace and SR-type master fractures propagate upwards into the core of the lobes. Their propagation directions, taken from both master and subsidiary fractures, indicate a coolant source entirely surrounding the lobes, which would indicate either a subaqueous or a subglacial eruption environment. However, a subaqueous eruption or interaction of lava with water ponded at the glacier base might be expected to form pillow lavas (e.g. Lescinsky and Fink, 2000). SR-type pseudopillow fracture systems commonly have glassy zones on either side of the master fracture that extend a few centimetres into the adjacent lava (Fig. 4). This demonstrates that SR-type master fractures act as pathways to transport coolant (water or steam) through the fractured lava, while causing rapid cooling of the lava surrounding the master fracture. No master fractures with a dimpled fracture surface texture, like those of F-type (Forbes et al., 2012) or like those from basaltic entablature (Forbes et al., 2014), were observed in the Slaga flow. This may be related to the lack of vesicles in the parts of the flow where pseudopillow fracture systems form. In particular, there is a lack of bands of vesicles which might induce cavitation ductile fracturing (e.g. Eichhubl and Aydin, 2003) rather than the ductile fracture morphology seen on X-type/SR-type master fractures. G-type pseudopillow fracture systems are indicative of very rapid cooling of lava, as demonstrated by their glassy nature and extremely narrow striae. Forbes et al. (2012) proposed that they reflect cooling by liquid water rather than steam. This model is supported by evidence from the Slaga lava, as G-type pseudopillow fracture systems occur only at the base of lobes, where melt water would preferentially collect or form a downhill-flowing layer. Where lobes are superimposed, G-type fractures are concentrated in the base of the lowermost lobes. This is consistent with ready availability of meltwater at the lava–bedrock interface and a decreasing availability of meltwater upwards into the overlying lava. A subaqueous lava flow might be expected to show G-type pseudopillow fracture systems around the entirety of the flow rather than solely at the bases of lobes. This implies that the supply of coolant was not uniform, strongly suggesting that the flow was emplaced in a subglacial rather than subaqueous environment. 5.4.1. Subsidiary fractures and striae widths Striae widths on subsidiary fractures in this flow show a strong linear correlation to the spacing of long planar subparallel subsidiary fractures (Fig. 8), demonstrating that subsidiary fracture spacing is likely linked to cooling rate (e.g. Forbes et al., 2012). The relationship between stria width and subsidiary fracture spacing coincides with that from the Tangahraun lava, Snæfellsnes, Iceland. This may reflect a fixed ratio between striae widths and subsidiary fracture spacing, similar to that found for striae widths and column sides in columnar-jointed lava (e.g. DeGraff and Aydin, 1993; Grossenbacher and McDuffie, 1995; Goehring and Morris, 2008; Phillips et al., 2013). Alternatively, coincident values may be caused by the similar compositions of the two lava flows, resulting in similar physical properties. SR-type Slaga data plots at the lower end of the graph demonstrating a generally more rapid cooling in the Slaga flow compared to the lava–seawater interaction of the Tangahraun flow (Fig. 8). 5.5. Environmental implications of fractures and flow morphology The rapid radial cooling of lobes in this lava, demonstrated by small striae widths and abundant glassy lava, implies that the lobes were

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

surrounded by coolant. The high aspect ratio of lobes strongly suggests lateral confinement of the lava, probably by ice. The superposition of numerous lobes may indicate that lobes preferentially flowed along a conduit previously melted into the ice, rather than creating a new one itself. Glassier lava featuring G-type pseudopillow fracture systems at the lowest lobe bases demonstrates the uneven nature of the cooling of lobes. Coolant, probably water, ponded or drained along the base of the flow but had less influence on lobes higher up in the pile. The uneven nature of cooling in this flow, such as suggested by G-type pseudopillow fracture systems at the base of lobes, curved platy fractures in slower cooled lava in the core of lobes, and a general lack of pillow lava indicates a subglacial rather than subaqueous lava flow environment. The flow and drainage of water indicated by the glassy lobe bases would also suggest that the ice here was thin, probably an alpine-type glacier ≤100–150 m thick (Smellie, 2000; Tuffen and Castro, 2009). The holocrystalline nature, lack of pseudopillow fracture systems and glassy lava in cores of larger lobes indicate that they were more slowly cooled than the rest of the flow, possibly because not enough coolant was produced to reach the centre of these large lobes. Thicker outer fracture carapaces at the lobe bases indicate that greater cooling was progressing from the base upwards than from the top downwards. If the lava flow melted a sufficiently large volume of ice to result in the formation of a cavity between the upper surface of the lava and the base of the ice, cooling would be retarded at the top of the flow where it was no longer in direct contact with the overlying ice. The more rapidly cooled flow bases, demonstrated by the glassy basal margins, in the Slaga flow indicate an open system with melt water escape. This could cause an ice cavity above the lava (e.g. Tuffen et al., 2002; Tuffen and Castro, 2009). The insulated cores of larger lobes may also be explained by the successive emplacement of lobes on top of one another, blocking coolant access to the lobes below. Subglacial meltwater drainage channels may be incised upwards into the ice (Rothlisberger channels, also known as R-channels), or downwards into the bedrock (Nye channels). If the Slaga lava did advance within subglacial channels these are likely ice-incised Rchannels, given the absence of observed channelised erosion features in the substrate. R-channels typically develop in spring, are typically broad smooth-walled arches in cross-section, and range from less than 5 m to at least 10 m in radius (Fountain and Walder, 1998). They form branching networks that are either arborescent, and mostly aligned sub-parallel to the glacier flow direction, or non-arborescent, with more variable orientations. R-channels close to the glacier snout are typically only partially water-filled, and at atmospheric pressure (Hooke, 1984; Fountain and Walder, 1998). We note that the dimensions of the Slaga lava lobes correspond well with those of R-channels, and propose that the initial splitting of the Slaga lava flow into several lobes reflects the exploitation of linked cavities within an R-channel system. As lava orientation broadly follows the valley alignment, and thus the glacier flow direction, it is likely that the initial flow lobes advanced within a number of arborescent R-channels, assisted by local melting around the lava margin and any meltwater being flushed from closer to the eruption site. This model is applicable only if the lava were emplaced between spring and autumn, as R-channels typically collapse in winter (Fountain and Walder, 1998). A similar mode of emplacement in subglacial channels/tunnels was observed for the initial part of the Gígjökull 2010 lava flow (Edwards et al., 2012b). In summary, inspection of fracture features in the Slaga lava indicate that it was emplaced as a series of lobes beneath thin ice, probably a relatively thin alpine/valley type glacier. The lava flowed over and locally interacted with a soft, non-lithified polymict sediment. It was cooled rapidly on all sides and lobes were likely confined by ice while they were being emplaced, initially within partially-filled R-channels. Water flowed along the base of the lava, causing particularly rapid cooling there. The flow most likely flowed within a series of preexisting subglacial meltwater channels cut upwards into the overlying ice, and potentially enlarged by eruption-produced meltwater. The

17

lava has since had glacial debris deposited on top of the flow, and been eroded by fluvial processes. 6. The 1727 eruption of Öræfajökull An eruption started at Öræfajökull on August 8th 1727 and continued until April 1728 (Thorarinsson, 1958). The Slaga lava is presently undated, however the evidence, presented above, for sub-glacial emplacement in an area that was until the 20th century under ice (Thorarinsson, 1943; Chenet et al., 2010; Guðmundsson et al., 2012) and its location adjacent to the area from which jokulhlaups emerged during the 1727 eruption of Öræfajökull (Thorarinsson, 1958) permit the possibility that the Slaga lava was emplaced during the 1727 eruption. Some aspects of the eruption are described in the eyewitness account of Rev. Jón Thorláksson, reproduced in translation by Thorarinsson (1958): “The noise and reports continuing, the atmosphere was so completely filled with fire and ashes, that the day could scarcely be distinguished from night, by reason of the darkness which followed, and which was only lighted up by the glowing of the fire that had broken through five or six fissures in the mountain. In this manner the parish of Öræfi was tormented for three days together by fire, water and falling ashes.”

“The mountain continued to burn night and day, from the 8th of August, as already mentioned, till the beginning of spring, in the month of April the following year, at which time the stones were still so hot, that they could not be touched; and it did not cease to emit smoke till the near end of the summer.”

“On the first day of summer 1728, I went with a person of quality to examine the rifts in the mountain, along which it was for the most part possible to creep. I found here a quantity of salpetre, and could have collected it, but did not choose to stay long in the excessive heat.” This account mentions ‘fires’ multiple times, by which is usually meant lava or fire fountaining eruptions but the most explosive phase of the eruption appears to have been restricted to a few days at the beginning. Persistently ‘hot stones’ are also mentioned and may well be a reference to a lava flow of some kind, as it is unlikely to be describing tephra from this eruption, Although tephra has been recovered, any lava linked to this eruption has not been identified previously. Compositional data from the tephra of the 1727 eruption and the Slaga lava (Table 1, supplementary material) show close similarities in both major and trace elements. We propose that the subglacial Slaga lava flow was produced during this historical eruption. Emplacement of the lava in the late summer is also consistent with the presence of R-channels at the glacier base, as inferred in the previous section. 7. Summary This study demonstrates that cooling fractures in lavas, particularly pseudopillow fracture systems, can be an important source of information when trying to establish and understand the eruption environments of lava flows. A number of conclusions can be drawn from the cooling fractures in this study:

(1) This lava likely flowed subglacially, being confined and cooled rapidly from all sides. (2) The environment was probably a thin alpine-type glacier with ice b 150 m thick. (3) The lava arguably advanced within partially water-filled Rothlisberger channels melted into the overlying ice.

18

A.E.S. Forbes et al. / Journal of Volcanology and Geothermal Research 288 (2014) 8–18

(4) This study has strengthened the link between G-type pseudopillow fracture systems and water, rather than steam, as a coolant, due to their formation only at the base of lobes. An improved understanding of cooling fractures, such as pseudopillow fracture systems, has enabled the use of cooling fractures to derive information on the emplacement environment of this lava. However it is important to gather all relevant observations relating to a lava body as the presence or absence of any fracture type can only provide limited information in itself. The subglacial Slaga lava may have been emplaced during the 1727–8 eruption of Öræfajökull volcano. Supporting factors include corresponding chemical compositions of the 1727 tephra and the Slaga lava, the right location for the lava flow, and the recognition from an historical source that lava was likely produced during this eruption. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jvolgeores.2014.10.004. Acknowledgments We would like to thank John Watson for help with XRF analyses. AESF was supported by a NERC PhD studentship and The Open University, HT is supported by a Royal Society University Research Fellowship. The authors wish to thank J Smellie and S White for helpful and constructive reviews which have improved this manuscript. References Bonnichsen, B., Kauffman, D., 1987. Physical features of rhyolite lava flows in the Snake River Plain volcanic province, southwestern Idaho. In: Fink, J. (Ed.), The Emplacement of Silicic Domes and Lava Flows. Geological Society of America, Special Paper 212, pp. 119–145. Chenet, M., Roussel, E., Jornelli, V., Grancher, D., 2010. Asynchronous Little Ice Age glacial maximum extent in southeast Iceland. Geomorphology 114, 253–260. DeGraff, J., Aydin, A., 1987. Surface morphology of columnar joints and its significance to mechanics and direction of joint growth. Geol. Soc. Am. Bull. 99 (5), 605–617. DeGraff, J.M., Aydin, A., 1993. Effect of thermal regime on growth increment and spacing of contraction joints in basaltic lava. J. Geophys. Res. Solid Earth 98 (B4), 6411–6430. Edwards, B., Magnússon, E., Thordarson, T., Guđmundsson, M.T., Höskuldsson, A., Oddsson, B., Haklar, J., 2012a. Interactions between lava and snow/ice during the 2010 Fimmvörðuháls eruption, south-central Iceland. J. Geophys. Res. Solid Earth 117. http://dx.doi.org/10.1029/2011JB008985 (B04302). Edwards, B., Oddsson, B., Gudmundsson, M.T., Rossi, R., 2012b. Field constraints for modeling the emplacement of the 2010 Gigjökull lava flow, southern Iceland: interplay between subaqueous, ice contact and subaerial lava emplacementEGU General Assembly Conference Abstracts 14, p. 11679. Edwards, B.R., Karson, J., Wysocki, R., Lev, E., Bindeman, I., Kueppers, U., 2013. Insights on lava–ice/snow interactions from large-scale basaltic melt experiments. Geology 41 (8), 851–854. Eichhubl, P., Aydin, A., 2003. Ductile opening-mode fracture by pore growth and coalescence during combustion alteration of siliceous mudstone. J. Struct. Geol. 25 (1), 121–134. Forbes, A.E.S., Blake, S., McGarvie, D.W., Tuffen, H., 2012. Pseudopillow fracture systems in lavas: insights into cooling mechanisms and environments from lava flow fractures. J. Volcanol. Geotherm. Res. 245–246, 68–80. Forbes, A.E.S., Blake, S., Tuffen, H., 2014. Entablature: fracture types and mechanisms. Bull. Volcanol. 76 (5), 1–13. Fountain, A.G., Walder, J.S., 1998. Water flow through temperate glaciers. Rev. Geophys. 36, 299–328. Goehring, L., Morris, S.W., 2008. Scaling of columnar joints in basalt. J. Geophys. Res. 113 (B10). http://dx.doi.org/10.1029/2007JB005018 (B10203). Graettinger, A., Skilling, I., McGarvie, D., Höskuldsson, Á., 2012. Intrusion of basalt into frozen sediments and generation of Coherent-Margined Volcaniclastic Dikes (CMVDs). J. Volcanol. Geotherm. Res. 217–218, 30–38.

Grossenbacher, K., McDuffie, S., 1995. Conductive cooling of lava: columnar joint diameter and stria width as functions of cooling rate and thermal gradient. J. Volcanol. Geotherm. Res. 69 (1–2), 95–103. Guðmundsson, S., 2014. Reconstruction of late 19th century geometry of Kotárjökull and Breiðamerkurjökull in SE-Iceland and comparison with the presentMS Thesis University of Iceland. Guðmundsson, S., Hannesdóttir, H., Björnsson, H., 2012. Post-Little Ice Age volume loss of Kotárjökull glacier, SE-Iceland, derived from historical photography. Jökull 62, 97–110. Hooke, R.L., 1984. On the role of mechanical energy in maintaining subglacial water conduits at atmospheric pressure. J. Glaciol. 30, 180–187. Hull, D., 1993. Tilting cracks: the evolution of fracture surface topology in brittle solids. Int. J. Fract. 62, 119–138. Hull, D., 1999. Fractography: observing, measuring, and interpreting fracture surface topography. Cambridge University Press, UK. Larsen, G., Dugmore, A., Newton, A., 1999. Geochemistry of historic silicic tephras in Iceland. The Holocene 9 (4), 463–471. Lescinsky, D.T., 1999. Lava flow morphology: the roles of external confinement and lava– ice interactionPh.D. thesis Arizona State University, Tempe. Lescinsky, D., Fink, J., 2000. Lava and ice interaction at stratovolcanoes: use of characteristic features to determine past glacial extents and future volcanic hazards. J. Geophys. Res. 23, 711–723. Lescinsky, D., Sisson, T., 1998. Ridge-forming, ice-bounded lava flows at Mount Rainier, Washington. Geology 26, 351–354. Lodge, R., Lescinsky, D., 2009. Fracture patterns at lava–ice contacts on Kokostick Butte, OR, and Mazama Ridge, Mount Rainier, WA: implications for flow emplacement and cooling histories. J. Volcanol. Geotherm. Res. 185, 298–310. Mee, K., Tuffen, H., Gilbert, J., 2006. Snow-contact volcanic facies and their use in determining past eruptive environments at Nevados de Chillan volcano, Chile. Bull. Volcanol. 68, 363–376. Phillips, J.C., Humphreys, M.C.S., Daniels, K.A., Brown, R.J., Witham, F., 2013. The formation of columnar joints produced by cooling in basalt at Staffa, Scotland. Bull. Volcanol. 75 (6), 1–17. Prestvik, T., 1985. Petrology of Quaternary volcanic rocks from Öræfi, southeast Iceland. Geologisk Institutt, Norges Tekniske Hojskole, Trondheim, Report 21,p. 81. Prestvik, T., Goldberg, S., Karlsson, H., Gronvold, K., 2011. Anomalous strontium and lead isotope signatures in the off-rift Öræfajökull central volcano, south-east Iceland. Evidence for enriched endmember(s) of the Iceland mantle plume? Earth Planet. Sci. Lett. 190, 211–220. Pugh, S., 1967. The fracture of brittle materials. Br. J. Appl. Phys. 18, 129. Ryan, M., Sammis, C., 1978. Cyclic fracture mechanisms in cooling basalt. Bull. Geol. Soc. Am. 89, 1295–1308. Skilling, I.P., 2009. Subglacial to emergent basaltic volcanism at Hlöðufell, south-west Iceland: a history of ice-confinement. J. Volcanol. Geotherm. Res. 185 (4), 276–289. Smellie, J.L., 2000. Subglacial eruptions. In: Sigurdsson, H. (Ed.), Encyclopaedia of Volcanoes. Academic Press, San Diego, California, pp. 403–418. Sommer, E., 1969. Formation of fracture lances in glass. Eng. Fract. Mech. 1, 539–546. Spörli, K., Rowland, J., 2006. ‘Column on column’ structures as indicators of lava/ice interaction, Ruapehu andesite volcano, New Zealand. J. Volcanol. Geotherm. Res. 157, 294–310. Stevenson, J.A., Smellie, J.L., McGarvie, D.W., Gilbert, J.S., Cameron, B.I., 2009. Subglacial intermediate volcanism at Kerlingarfjöll, Iceland: magma–water interactions beneath thick ice. J. Volcanol. Geotherm. Res. 185 (4), 337–351. Thorarinsson, S., 1943. Vatnajökull. Scientific results of the Swedish–Icelandic investigations 1936–1937–1938. Chapter XI. Oscillations of the Iceland glaciers in the last 250 years. Geogr. Ann. 25, 1–54. Thorarinsson, S., 1958. The Öraefajökull eruption of 1362. Acta Natur. Island. 2, 1–99. Tucker, D., Scott, K., 2009. Structures and facies associated with the flow of subaerial basaltic lava into a deep freshwater lake: the Sulphur Creek lava flow, North Cascades, Washington. J. Volcanol. Geotherm. Res. 185, 311–322. Tuffen, H., Castro, J.M., 2009. The emplacement of an obsidian dyke through thin ice: Hrafntinnuhryggur, Krafla, Iceland. J. Volcanol. Geotherm. Res. 185, 352–366. Tuffen, H., Gilbert, J., McGarvie, D., 2001. Products of an effusive subglacial rhyolite eruption: Bláhnúkur, Torfajökull, Iceland. Bull. Volcanol. 63, 179–190. Tuffen, H., Pinkerton, H., McGarvie, D., Gilbert, J., 2002. Melting of the glacier base during a small-volume subglacial rhyolite eruption: evidence from Bláhnúkur, Torfajökull, Iceland. Sediment. Geol. 149, 183–198. Tuffen, H., James, M.R., Castro, J.M., Schipper, C.I., 2013. Exceptional mobility of an advancing rhyolitic obsidian flow at Cordón Caulle volcano in Chile. Nat. Commun. 4. http:// dx.doi.org/10.1038/ncomms3709. Walker, G.P.L., 1993. Basaltic-volcano systems. Geol. Soc. Lond., Spec. Publ. 76, 3–38. Watanabe, K., Katsui, Y., 1976. Pseudo-pillow lavas in the Aso caldera, Kyushu, Japan. J. Miner. Petrol. Econ. Geol. 71, 44–49.