Introduction

While the structural emplacement of basaltic lava flows is well-constrained (e.g. Pinkerton and Sparks 1978; Lipman and Banks 1987; Self et al. 1997; Calvari and Pinkerton 1999; Cashman et al. 1999; Kilburn 2000; Jerram 2002, Passey and Bell 2007), such studies of more silicic lava flows are comparatively few (e.g. Fink 1983; de Silva et al. 1994; Fujii and Nakada 1999; Uia et al. 1999; Castro et al. 2002; Harris et al. 2002; Cioni and Funedda 2005; Tuffen et al. 2013; Magnall et al. 2016, 2017; Latutrie et al. 2017). This is due in part to a lack of observations of active flows. Prior to 2008, no rhyolite-forming eruption had been witnessed and documented (Tuffen et al. 2013; Schipper et al. 2013; Castro et al. 2014). As a result of this, our understanding of the mechanics of basaltic lava flow emplacement is confirmed by real-time observations, while our knowledge of rhyolitic lava flow emplacement has generally been inferred based on the structural configuration of preserved flows (e.g. Fink 1983; Manga 1998; Tuffen et al. 2013; Magnall et al. 2016).

The Rocche Rosse rhyolite lava flow on the island of Lipari (Aeolian Islands, Italy; Fig. 1) shows exceptional preservation of complex small to large-scale structures. Previous work on the Rocche Rosse has focused on fragmentation and transport mechanisms of the associated pyroclastic sequence (Dellino and La Volpe 1995), the thermal history of flow frontal ramps, i.e. ridges of fold structures extending downward into the flow by tens of metres (Gottsmann and Dingwell 2001), magmatic history (Gioncada et al. 2003, 2005; Davì et al. 2009, 2010) and textural heterogeneity (Gonnermann and Manga 2003; Clay et al. 2013; Shields et al. 2016). Ramp structures of the Rocche Rosse were also documented by Hall (1978) and Cas and Wright (1987), inferring a compressive regime and build-up of lava behind a brecciated flow front, hindering flow advance. Well-exposed surface and flank structures resulted from surface crust development, and sequential emplacement and deformation history. Such structures are often evident and have been studied in viscous intermediate lava flows (e.g. Cerro Chao, Chile—Fink 1980b; de Silva et al. 1994; Puy de Cliergue, France—Latutrie et al. 2017; San Pietro, Sardinia—Cioni and Funedda 2005; Santiaguito, Guatemala—Harris et al. 2002, 2004), and other rhyolitic lava flows worldwide (Fink 1980a, b 1983; Manley and Fink 1987; Carn et al. 2009; Lara 2009; Bernstein et al. 2013; Pallister et al. 2013; Tuffen et al. 2013; Magnall et al. 2017).

Fig. 1
figure 1

a The Aeolian Islands, southern Tyrrhenian Sea (Italy), with Lipari in the centre of the archipelago, straddling the Tindari-Letojanni fault system (Ventura 2013), and b the Rocche Rosse lava flow in NE Lipari. c Google Earth image of the Rocche Rosse (view to the south) (Map data: SOI, NOAA, U.S. Navy, NGA, GEBCO, TerraMetrics, ©2018 Google)

Full size image

Previous studies of rhyolite lava flow crust development and emplacement mechanisms have centred on United States Holocene flows (Big and Little Glass Mountain, Medicine Lake; Fink 1980a, b, 1983; Manley and Fink 1987; Manga 1998; Big Obsidian Flow in Newberry Caldera; Castro et al. 2002). Early conceptual models of rhyolite lava flow emplacement depict slowly spreading lavas overriding blocks of surface crust, creating an upper and basal breccia (Fink 1980a, 1983; Manley and Fink 1987), before stopping when supply has ceased. Processes such as rising pumice diapirs, compression, extension and fracture propagation produce a range of complex surface structural relations (Fink 1983). For instance, the development of a cooled surface crust promotes crustal folding during flow advance (Fink 1980a; Castro and Cashman 1999) and crease structure development (Anderson and Fink 1992; Lescinsky and Merle 2005). Crusts influence lava flow growth (Griffiths and Fink 1992; Fink and Griffiths 1998), and thermally insulate the flow, facilitating prolonged mobility and advance (Swanson 1973; Manley 1992; Fink and Griffiths 1998; Keszthelyi and Self 1998; Harris and Rowland 2009; Tuffen et al. 2013). This has been previously observed in dacitic lava flow emplacement at Santiaguito (Harris et al. 2002, 2004), and active rhyolitic flow emplacement at Puyehue-Cordón Caulle (Tuffen et al. 2013; Magnall et al. 2017).

Since 2008, real-time observations of active rhyolitic flow and dome emplacement at the Chilean volcanoes of Chaitén (Carn et al. 2009; Lara 2009; Bernstein et al. 2013; Pallister et al. 2013) and Puyehue-Cordón Caulle (Castro et al. 2013; Schipper et al. 2013; Tuffen et al. 2013; Bertin et al. 2015; Farquharson et al. 2015; Magnall et al. 2017) have further developed our knowledge of rhyolitic lava emplacement. The Cordón Caulle eruption is an example of a flow that moved away from the vent, while Chaitén was a dome-forming eruption. Observations of the 2008 Chaitén eruption and associated composite rhyolite dome highlight simultaneous explosive and effusive activity, exogenous and endogenous growth phases and complex lava lobe and spine formation (Carn et al. 2009; Lara 2009; Bernstein et al. 2013; Pallister et al. 2013). These resulted from multiple flow paths (flow directions of areas of active dome growth) within the lava dome, and different extrusion rates (Carn et al. 2009; Lara 2009; Bernstein et al. 2013; Pallister et al. 2013). Morphological observations of the 2011–2012 Cordón Caulle eruption highlight the role of thermal insulation by a surface crust in prolonging flow-horizontal transport, channelled flow, surface crust formation, development of breakout lobes and flow frontal stalling (Castro et al. 2013; Schipper et al. 2013; Tuffen et al. 2013; Bertin et al. 2015; Magnall et al. 2017). In these studies, comparisons are made to basaltic lava flow fields, with common controlling mechanisms that work towards a unified model of effusive eruptions (Castro et al. 2013; Schipper et al. 2013; Tuffen et al. 2013; Bertin et al. 2015; Magnall et al. 2017).

While active observations can act as benchmarks for future rhyolite effusive events, they can also improve interpretations of past eruptions. Specifically, morphological features observed in real-time eruptions can help to decipher how structures were formed and modified in older flows. Observations of active rhyolite lava effusion at Chaitén and Cordón Caulle (Carn et al. 2009; Lara 2009; Castro et al. 2013; Tuffen et al. 2013) provide a new motivation to revisit the emplacement mechanisms of older rhyolite lava flows such as the Rocche Rosse flow. This study aims to focus on Rocche Rosse flow morphology, satellite observations and structural measurements, using detailed qualitative and quantitative structural measurements to formulate new interpretations of the flow dynamics associated with emplacement of the Rocche Rosse flow. Surface structures and flank (edge of flow) characteristics are examined to provide further insights into the internal architecture of the Rocche Rosse, and the important role played by progressive deformation and channelled flow morphology in rhyolite lava flow structural evolution.

Geological setting

The Aeolian archipelago (Fig. 1) is the result of the complex collision between the African and Eurasian plates (Ventura 2013). The historically active island of Lipari is located in the centre of the archipelago, and is the largest island (Fig. 1). The NNW-SSE striking Tindari-Letojanni fault system defines the alignment of volcanoes (Barberi et al. 1994, Fig. 1). Re-activation of the fault during the middle- to late-Pleistocene caused recent volcanism in the central arc (Barberi et al. 1994; Gvirtzman and Nur 2001; Ventura 2013). Lipari rises ~ 600 m above sea level and lies on 20-km-thick continental crust (Finetti and Morelli 1973).

The most recent activity of Lipari comprised explosive pyroclastic deposits and rhyolitic lava flows, culminating in the extrusion of the Rocche Rosse and Forgia Vecchia rhyolite lava flows (Forni et al. 2013). The pattern of this recent activity included a hydromagmatic phase, explosive eruption of a vesicular magma, accompanying fall and surge deposits and degassed viscous lava extrusion (Cortese et al. 1986; Dellino and La Volpe 1995; Lucchi et al. 2010; Forni et al. 2013). The Rocche Rosse represents a renewal of activity, following the initial Monte Pilato-Rocche Rosse explosive phase. The age of the Rocche Rosse, as determined by archaeomagnetic methods, is 1230 ± 40 AD (Arrighi et al. 2006).

Lithofacies of rhyolitic lava flows

Rhyolitic lava flows show a complex internal sequence with variably vesicular, devitrified and deformed textures, and complex surface deformation patterns (Fink 1978, 1983; Manley and Fink 1987; Shields et al. 2016). The traditional view of rhyolitic flow interiors includes (from the surface downwards) an upper breccia, thick obsidian, foliated rhyolite, thinner obsidian and a basal breccia (Fink 1978, 1983; Manley and Fink 1987). This idealised cross section has been observed in Obsidian Dome core drills in California, the Banco Bonito flow in New Mexico (Manley and Fink 1987) and Little Glass Mountain (Medicine Lake) (Fink 1983). Spherulitic and crystalline lava forms the central part of the flow, with the population density of crystals and spherulites depending on the length and thickness of the flow (Manley and Fink 1987). Spherulites are radial fibrous crystallites, considered to progressively nucleate as a down temperature continuum during syn- and post-emplacement cooling (Lofgren 1971; Swanson 1977; McArthur et al. 1998; Gardner et al. 2012; Breitkreuz 2013; Clay et al. 2013; Bullock et al. 2017). Differences in spherulite abundance (often manifested as trails), crystallinity, vesicularity and glass chemistry can all define flow foliation in rhyolite (Gonnermann and Manga 2005; Clay et al. 2013; Bullock et al. 2017).

The structural arrangement of flow foliations preserves a record of deformation associated with flow emplacement (Gonnermann and Manga 2003; Cioni and Funedda 2005), a record of brittle-ductile processes (Castro et al. 2008) and the analysis of foliation patterns contributes to the understanding of the structural evolution undergone during polyphase deformation of rocks (Fink 1978, 1983; Manley and Fink 1987; Manga 1998; Gonnerman and Manga 2005). The determination of a preferred orientation of foliations (and lineations) can also provide a record of strain, lava flow and deformation. Fink (1983) identified three deformational processes that disrupted the foliation of the flow during advance: (1) low density coarsely vesicular pumice diapirs, which rise and cause surface doming and plunging anticlines; (2) flow-parallel compressional stresses near the flow front, resulting in flow folding oriented transverse to flow direction; and (3) an extensional regime, resulting in inward propagating surface fractures.

Records of flow emplacement and deformation in the Rocche Rosse are recognisable due to textural variations within the flow. Shields et al. (2016) identified four textural units in the Rocche Rosse: type 1—glassy obsidian, which is spherulitic and has a low vesicularity (< 15%); type 2—‘frothy’ obsidian, which is vesicular (30–60%) and glassy, with large elongated vesicles perpendicular to banding; type 3—pumiceous lava, which is highly vesicular (40–60%), with variable vesicle sizes; type 4—shear banded lava, characterised by a low-medium vesicularity (20–40%), and highly deformed and coalesced vesicles. Glassy obsidian and shear banded lavas occur as massive and/or platy outcrops, pumiceous lava is observed throughout the flow and ‘frothy’ obsidian makes up fold hinges, tension gashes, micro folding and extensional fractures (Shields et al. 2016).

Textural variations such as glassy and devitrified obsidian as well as vesicular units can be identified at middle- and upper-flow regions, and brecciated zones are evident across the entire flow. Flow stratigraphy is exposed along the flanks of the Rocche Rosse, visible from the town of Acquacalda to the west and Porticello to the east (Fig. 2). The Rocche Rosse sits on top of older tephra deposits (Forni et al. 2013). A thick (up to 10 m) basal breccia made up of angular obsidian blocks is evident at the base of the flow (Fig. 2). Pumiceous lava (grey-white, variable sizes and shapes of vesicles, low density) and obsidian (black, glassy or devitrified, poorly vesicular and high density spherulitic lava) lithofacies sit on top of the basal breccia.

Fig. 2
figure 2

Digitisation of a NE and b NW flow flanks and flow divisions. Flow thickness and textural variations shown, as well as dominant observed foliation attitude. General attitude of foliations indicated by dashed lines, vesicular zones shown in grey and basal breccia in brown

Full size image

Methodology

In this study, the Rocche Rosse is characterised by mapping based on high spatial resolution satellite imagery, flank/surface observations and structural measurements. Large-scale surface structures were observed and measured from Infoterra GeoEye-1 satellite imagery (obtained from Airbus Defence and Space) at a 0.5 m resolution. The GeoEye-1 image covers an area of 5010 km2 at a scale of 1:10,000. Large-scale lineaments were identified, traced and measured using ArcMap 10 geo-referenced measurement tools. Here, ‘lineament’ is used as a descriptive term for a linear or curvilinear surface feature identified in GeoEye-1 imagery analyses. Measurements of large scale folds were cross-checked with kinematic data collected in the field. The flow stratigraphy and thickness of the Rocche Rosse is visible along the eastern and western flank exposures (Fig. 2). Flow thickness was calculated at various points around the flanks using a comparable scaling technique (comparing with an object of known height, e.g. a 1 m scale bar).

Flow foliation was identified by the planar arrangement of spherulites, microlites and vesicularity. Foliation attitude was recorded and mapped as strike, dip and dip direction. Folds were measured and described based on their fold axial plane (strike and dip) and fold axis (orientation and inclination of hinge-line measured as a lineation). Wavelength and amplitude of folds were also measured. Stretching lineations, evident as a three-dimensional parallel alignment of flattened spherulites, were measured in terms of plunge angle and azimuth. Stereographs, density plots and rose diagrams were created using Stereo32 software, projected as equal area lower hemisphere plots. Stereoplots were used to graphically represent measured surface structures, with density plots highlighting the density concentrations of data and rose diagrams graphically summarising directional data. All of the data are summarised based on the average agreement of data to the most common orientation. For instance, if foliation measurements were entirely random across flow, we infer that the flow has no degree of preferred orientation. If there is some agreement of the data in a single direction, we infer that the flow shows a weak, moderate or strong degree of orientation. More specifically, no degree of preferred orientation is deemed as being associated with a correlation of less than 20%, a weak degree of preferred orientation is 20–50%, a moderate degree of preferred orientation is 50–70% and a strong degree of preferred orientation is > 70%.

Results

Overall flow morphology

Most of the flow surface is exposed, except in flow frontal regions where exposures are restricted by a man-made pumice heap. The flow extends from an elevation of 310 m to sea level. The maximum flow length measured by GeoEye-1 imagery (Fig. 3) is 2060 m, and the maximum width of the flow is 1130 m. There are three key structural features evident (Fig. 3): (1) lineaments aligned roughly SW-NE, sub-parallel to flow direction from the Monte Pilato vent to the SW, (2) curvilinear features roughly parallel to the flow front and (3) two prominent flow-front headlands. Lineaments extend from near-vent to mid-flow regions (extending over tens to hundreds of metres). Less prominent, smaller scale lineaments are nested within larger scale lineaments (thin red lines within bolder lines; Fig. 3).

Fig. 3
figure 3

a Google Earth image (Map data: SOI, NOAA, U.S. Navy, NGA, GEBCO, TerraMetrics, ©2018 Google) and b GeoEye-1 image of the Rocche Rosse rhyolitic lava flow. c Annotations of the Rocche Rosse lava flow from the satellite image, with the Monte Pilato crater vent region (dashed lines), flow margins, prominent lineaments, curvilinear features and flow lobes

Full size image

The flow can be subdivided into sections based on flow morphology, such as changes in the trend of prominent lineaments, topographic changes identified in GeoEye-1 imagery and flank observations (Figs. 2, 3 and 4). Two flow front lobes have been identified from GeoEye-1 imagery. The eastern lobe extends continuously from the vent, and the western lobe branches to the north (Fig. 3). Most of the prominent lineaments in the eastern lobe trend SW-NE, while smaller curvilinear features verge north eastwards. Conversely, in the western lobe, lineaments and curvilinear structures verge northwards, and the flow extends to the north. Topographic contours indicate that the western lobe responded to a change in topography, with a central flow valley dividing the lobes and topographic contours dipping NNW on the western side of the valley. Mid-flow, the Rocche Rosse sits on a continuous, steep (16–25°) topography compared to flatter (0–12°) flow front and vent regions (Fig. 2). Lineaments are continuous in mid-flow regions, before spreading radially outwards at the NE flow front. Stacks of flow ramps (Cas and Wright 1987; Gottsmann and Dingwell 2001) are evident in flank exposures in the vent region and at the break in slope of the flow front, with lava stacking up towards the mid-flow. Large-scale folding (Fig. 5) is also evident in flank exposures of the flow front.

Fig. 4
figure 4

Flow divisions based on changes in underlying slope (see contours inset and Fig. 3) and orientation/vergence of prominent lineaments and curvilinear features

Full size image
Fig. 5
figure 5

Structural features of the Rocche Rosse lava flow. a Small-scale (parasitic) folds (< 0.2 m wavelength and amplitude) superimposed upon medium-scale folds (0.2–2 m wavelength and amplitude) near NE flow front (f.a.p fold axial plane, connecting the hinge lines of the folds as a planar surface). b Recumbent medium-scale folding near NE flow front. c Large-scale (> 2 m wavelength and amplitude), inclined mid-flow folding, with medium- and small-scale parasitic folds. d Large-scale, recumbent fold near NW flow front (GPS for scale). e–g Examples of sheath folding, identified by characteristic eye shape. h Fractured obsidian lava. i, j Stretching lineations in mid-flow regions

Full size image

Flow foliations

The arrangement of spherulites, microlites and vesicles defines the foliation in the obsidian lithofacies of the flow (Clay et al. 2013; Bullock et al. 2017), making it possible to recognise and measure planar structures (Figs. 5, 6 and 7; Table 1, Electronic supplementary material 1). Flow foliations are generally continuous across several metres, although they may be truncated or disrupted by small brittle faults that have a few centimetres of displacement (Clay et al. 2013). Throughout the flow, foliations are disrupted by flow-wide conjugate fracturing and tension gashes, both up to 10 cm in length. Overall, planar structural data are randomly distributed, as evidenced by high variability of strike and dip measurements (Figs. 6 and 7). Flow foliation dips vary across flow from near vertical to horizontal, though predominance of a weak degree of preferred orientation (40%) indicates overall steeply dipping (80°) foliations towards the vent (Table 1). There is also evidence for dips preferred in a north easterly direction, away from the vent region (Fig. 7). Flow foliations dip steeply from vent towards the NE flow front (81–85°) and dip more gently towards the NW flow front (68°). The western lobe shows a moderate degree (55%) of preferred orientation (Table 1 and Fig. 7). The strike orientations and dip magnitudes vary significantly at the flow fronts, from 12 to 14° in localised areas, but more commonly > 60°. In mid-flow and near-vent regions, foliations typically dip at 30–50°, locally dipping as gently as 2–4°. There are more uniform steeply dipping mid-flow foliations in a SW-NE orientation (Fig. 7), coinciding with the positions of lineaments identified in GeoEye-1 imagery (Fig. 8). At the NE flow front, the ramping foliation typically dips 60–70° south-westerly towards the vent. At the NW flow front, ramping foliations dip from near-horizontal (16°SW) up to near-vertical (84°SW) over a few metres.

Fig. 6
figure 6

Stereographic plots of flow foliations, fold axial planes, fold hinges and stretching lineations (equal area projection, lower hemisphere). Data divided into a entire flow, b NW flow front, c NE flow front, d mid flow and e vent region (see Fig. 4)

Full size image
Fig. 7
figure 7

Planar and linear data projected as density plots and rose diagrams (see divisions in Fig. 4). Statistical summary data available in Electronic supplementary material 2

Full size image
Table 1 Summary statistics of planar and linear structural data of the Rocche Rosse lava flow (entire flow and flow segregations)
Full size table
Fig. 8
figure 8

Linear flow channel ridges mid-flow. a Interpretation of GeoEye-1 image with traced flow channels and radial spreading beyond channel. b Original GeoEye-1 image. c Structural measurements of stretching lineations in mid-flow region, the majority of measurements trending approximately NE-SW (flow direction)

Full size image

Folding

Multi-scale folds are widespread across the entire lava flow, on a scale of decimetres to tens of metres (Fig. 5). Fold asymmetry increases towards the flow fronts, with an increase in inclined axial planes (Fig. 5a–d). Sheath folding is also evident at the flow fronts (Fig. 5e–g). In particular, there are numerous folds at the NE flow front (Table 1). Small- and medium-scale flow folding (< 0.2 and 0.2–2 m wavelength and amplitude, respectively) is extensive, while larger scale folding (2–15 m wavelength and amplitude, accompanied by parasitic folds) is concentrated at the flow fronts. Up to three generations of folds have been identified, with smaller scale folds superimposed upon medium folds, and medium folds superimposed on larger folds (Fig. 5a, c, d). This is particularly evident at the break in slope at both flow fronts. Stereoplots of fold axial planes highlight an apparently random distribution of orientations and dips (Figs. 6 and 7). Density plots suggest that fold axial planes dip steeply (mean 33–60°), with the exception of horizontal trends in the vent region (mean 1°). Rose diagrams show that mid-flow fold axial planes generally dip to the NE, while folds in the vent region dip N-S (Fig. 7). Stereoplots of fold hinges highlight a range of attitudes, but the density plots (Fig. 7) show a high concentration of plunges above 60°. Fold hinges trend SE-NW, plunging 85° to 120° at a strong degree of preferred orientation of 71%, perpendicular to the prominent mid-flow lineaments identified in the GeoEye-1 image.

Lineations

Stretching lineations, formed by shearing and resultant alignment of heterogeneities in the flow, are commonly observed mid-flow (Figs. 5 and 8), forming a parallel alignment of elongate spherulites (Bullock et al. 2017). The plunge of the stretching lineations varies from near-horizontal (2°) to vertical, with most near vertical (86° to the NE), roughly parallel to mid-flow prominent lineaments (Table 1). The majority of stretching lineations plunge steeply, with sporadic examples of near-horizontal plunges at the NW flow front and mid-flow. Mid-flow, the stretching lineations plunge more shallowly, commonly less than 30°, and in some instances, plunge back towards the vent region, orientated NE-SW. Stretching lineations are evident in flow front regions, with more dispersed orientations. These plunges are more variable in flow front regions, but typically plunge ~ 30–40°. The NW flow front exclusively hosts steeply plunging stretching lineations.

Discussion

Crust formation

Silicate liquids of rhyolitic composition exhibit Bingham-like (non-Newtonian) rheologies below liquidus temperatures, with a non-linear dependence of strain rate on applied stress (e.g. Shaw et al. 1968; Walker 1971; Fink 1983; Webb and Dingwell 1990; Dragoni et al. 1992; Manley 1992; Fink and Griffiths 1998; Manga et al. 1998; Castro and Cashman 1999; Gottsmann and Dingwell 2001; Hess et al. 2008; Castro et al. 2013). It has been previously reported that the Rocche Rosse behaved as a Bingham fluid during emplacement (Clay et al. 2013), which can promote strain localisation and folding. Strain localisation affects the rheological behaviour of lavas, and rheological contrasts between layers, such as differences in viscosity, capillary number and yield strength (Castro and Cashman 1999; Gonnerman and Manga 2005; Shields et al. 2016), can promote shear zonation. Due to the high viscosity of flowing rhyolitic lavas, the last increments of flow-related strain are recorded in the structures that are evident in the final solidified flow (Clay et al. 2013).

The model of Giordano et al. (2008) has been used here to estimate the viscosity of the Rocche Rosse rhyolite. This model predicts the non-Arrhenian Newtonian viscosity as a function of temperature and melt composition in naturally occurring silicate melts (Giordano et al. 2008). The presence of crystals and changes in crystallinity can influence viscous flow and the rheological properties of magmas (e.g. Bagdassarov and Dingwell 1992; Smith 1996; Stevenson et al. 1996; Manga et al. 1998; Griffiths 2000; Pistone et al. 2016). However, the Rocche Rosse lava flow is generally a crystal-free glass (with some exceptions; Davì et al. 2009, 2010), thus crystal content is excluded from calculations. The amount of dissolved gases, preserved as bubbles in the lava flow, can also affect viscosity and flow behaviour (e.g. Pinkerton and Sparks 1978; Lipman et al. 1985; Manga et al. 1998; Rust and Manga 2002; Llewellin and Manga 2005). However, knowledge of these influences on the apparent viscosity of a bubbly lava is limited (Stein and Spera 1992; Quane and Russell 2005; Pistone et al. 2012; Chevrel et al. 2013), and they are therefore not included in viscosity calculations here. The estimated flow surface viscosity was 108 to 1010 Pa s, using rhyolite glass chemistry from Bullock (2015); eruptive temperature range estimates (750 to 900 °C) of Cas and Wright (1987), Gottsmann and Dingwell (2001) and Davì et al. (2010); and the water content (0.08–0.25 wt%) of Shields et al. (2016). Deformation is heterogeneous and multi scale, varying greatly and non-systematically across flow (Shields et al. 2016). During emplacement, the lava probably evolved from pumiceous, through shear-banded to outgassed, glassy obsidian (Shields et al. 2016). Vesicular magma densified through bubble collapse and fracture healing, and may have re-vesiculated to produce vesicular lava (Shields et al. 2016). As the flow continued to advance, cooling led to the development of a surface crust, which may become sufficiently strong to inhibit flow advance (Cashman et al. 1999; Blake and Bruno 2000; Hon et al. 2003; Magnall et al. 2017).

As lava flows cool, they develop a solid crust that interacts with the underlying fluid by imparting shear on the internal lava; flow advance can be inhibited as the crust strengthens (Fink and Fletcher 1978; Fink 1980a; Cashman et al. 1999; Blake and Bruno 2000; Griffiths 2000; Lescinsky and Merle 2005; Castruccio et al. 2013; Farquharson et al. 2015; Magnall et al. 2017). Rhyolitic lava crusts are typically stronger, thicker and more rigid than basaltic crusts. Holocene obsidian flows in the western US preserve evidence of a cooled surface crust in folding, crease structures and large gas cavities (Fink 1980a; Anderson and Fink 1989; Castro and Cashman 1999; Castro et al. 2002; Lescinsky and Merle 2005). Crustal development increases the strength of the flow (Hon et al. 1994; Griffiths and Fink 1997; Griffiths 2000; Lyman et al. 2005), insulates the flow and reduces heat loss (Swanson 1973; Fink and Griffiths 1998; Keszthelyi and Self 1998; Harris and Rowland 2009; Tuffen et al. 2013). When the crust fractures during lava movement, blocky lava is formed (Finch 1933; Macdonald 1953; Fink 1980a; Harris et al. 2017), defined as being made up on distinct blocks without some of the key characteristics of an ‘a’a flow, such as a spiny or clinker surface (Finch 1933).

We propose that the Rocche Rosse, with the surface characterised by blocky lava, folding, ramp structures and an overridden (basal) breccia, also was dominated by a surface crust. Early development of a surface crust is suggested by folding at the flow front and in the vent region (up to 5 m in wavelength and amplitude), formed in response to flow stagnation and buckling. Following Gregg and Fink (2000), the shallow slope underlying the Rocche Rosse vent (Fig. 2) may have aided crust formation. The surface crusts may have been strong enough to prevent breakout formation, if the internal pressure was less than the crustal yield strength (Magnall et al. 2017). Uniformly dipping SW-NE orientated foliations in mid-flow regions, where the flow is up to 50 m thick and is therefore likely to require more energy to deform, were formed by 16–25° slope-aided advance and less crustal control. Where cooling did not greatly hinder flow advance, extensional forces were promoted, evident by uniform stretching lineations in mid-flow regions (51% degree of preferred orientation; Table 1 and Fig. 8), forming perpendicular to the principle plane of shortening.

Channel development

GeoEye-1 imagery reveals a channelled flow geometry in mid-flow regions of the Rocche Rosse, with both lineaments and stretching lineations often trending parallel to flow direction (Fig. 8). This changes to lateral spreading (radial expansion) and more sporadic stretching lineation attitudes at the flow front (Fig. 8). Stretching lineations on linear ridges formed due to strike-slip shear. Similar features have been identified in other rhyolitic lava flows (Fink 1980b; Tuffen et al. 2013), and in analogue modelling of silicic lava extrusions (Fink and Griffiths 1998; Merle 1998; Gregg and Fink 2000; Griffiths 2000). During flow in a channelled regime, simple shearing is limited to the edges of the channel (Borgia et al. 1983; Dragoni et al. 1986, 1992; Harris et al. 2002). As a result, curvilinear ridges connect the stalled channel margins to the mobile central flow zone, also observed in other preserved flows (Fink 1980b; Lipman and Banks 1987; Favalli et al. 2018). Large-scale structures, including folds (height and width > 2 m), ramp structures and flow channels control local deformation, with channels hosting stretching lineations and smaller scale folds. Lateral friction along margins generated a wrenching component which, together with a viscosity contrast, can induce folding at the flow margins (Fink and Fletcher 1978). Compressional stress concentration is also promoted by deceleration of the flow front. Stretching lineations change from parallel to concentric where lateral spreading causes flow divergence (consistent with experiments of Merle 1998; Fig. 8). Channels are not evident near the vent, at the flows or in the western flow lobe, all of which lie on lower slopes. This suggests that the steeper slope produced the channelled morphology, contributing to greater stress on the surface crust, more deformational structures and a thicker flow in this area, caused by the formation of flow ramps. Thicker flow configuration mid-flow and in near-vent regions is the product of cooling-limited flow and flow build-up with continual extrusion.

Role of slope

Structural development of the Rocche Rosse relates to the underlying slope, with associated complexity of structural trends attributed to changes in slope angle and uniformity (Fig. 3). Foliations locally dip as shallow as 2–4° in near-vent regions, coinciding with a shallow slope, dipping 30–50° mid-flow over steeper topography (Figs. 6 and 7). At the NE flow front, the ramping foliation typically dips 60–70° back towards the vent (Figs. 6 and 7), suggesting crustal control and compression over a shallow topography. The variable underlying slope angle also created multi-size folds, with compression at breaks in slope superimposing earlier-formed smaller scale folds upon larger folds (Fig. 5). Analogue experiments show that the increasing extrusion rate and slope angle forms rifted, folded and leveed flows that solidified at the margins only (Gregg and Fink 2000). However, analogue experiment material behaviour may vary differently to rhyolite, and real-time rhyolite flow observations give a better indication of lobe development. Long prominent flow lobes at Cordón Caulle also form by overtopping a topographic barrier and flowing down a steepening slope (Tuffen et al. 2013). We suggest that similar processes may have been responsible for the development of the prominent western lobe in the Rocche Rosse.

On the neighbouring island of Vulcano (Fig. 1), the compositionally similar Pietre Cotte rhyolite lava flow was emplaced on a steep underlying slope of 30–35° (Frazzetta et al. 1984). The flow is characterised by a single lobe with a uniform structural arrangement of foliations and lineations, which are the result of a single flow direction over a continuous topography (Bullock 2015). Observations from the Porri andesitic lava flow on the neighbouring island of Salina (Fig. 1) indicate a steep (10–30°), uniform slope with no abrupt changes (Iezzi and Ventura 2000; Ventura 2001). Structural observations of the Porri flow indicate a decrease in fold axial plane dip away from the vent due to increasing contribution of simple shear to bulk deformation, with gravity-driven mechanisms controlling the deformation patterns (Iezzi and Ventura 2000). These observations highlight the relation between simple structural arrangement and a uniform topography. At Little Glass Mountain, CA, extension and breakout lobes from flow margins coincide with a slope increase of ~ 9° (Fink 1978), suggesting that a change in slope angle to this degree promotes structural complexity.

Progressive deformation

Structural complexity of the Rocche Rosse suggests that the flow underwent a sequential deformational history (Fig. 9). Fold asymmetry increases away from the vent, with more folds at flow fronts, evidence for progressive deformation (Flinn 1962; Ramsay 1979). Additional evidence for sequential deformation includes small-scale structures superimposed upon large-scale structures (Fig. 5), brittle structures cross-cutting ductile structures and the variable distribution of planar and linear orientations (Figs. 6 and 7). Progressive shear modified early-formed folds, producing sheath folds (Farrell 1984; Smith 2002; Branney et al. 2004; Andrews and Branney 2011). The time period in which the Rocche Rosse progressively deformed was likely to be weeks following effusion (Gottsmann and Dingwell 2001), and it has been suggested that deformation in rhyolitic lava flows can span a year (Lu et al. 2004; Tuffen et al. 2013).

Fig. 9
figure 9

Model for the emplacement of the Rocche Rosse lava flow, emplaced sequentially over five stages: (1) Initial extrusion (moving NE), crustal development and small-scale folding. (2) Extensional strain, stretching lineations and channel development over steeper topography. (3) Compression at topographic break, autobrecciation, lobe development and further (medium scale) folding. (4) Progressive deformation due to stagnation and continual extrusion, large-scale folding and re-folding. (5) Brittle deformation following flow termination. Typical structural signature associated with each stage also shown

Full size image

The identification and measurement of multi-scale structures across the Rocche Rosse allows us to reconstruct emplacement and deformation mechanisms (Fig. 9). Initially, lava extruded from the Monte Pilato vent and advanced downslope towards the NE (stage 1). The first phase of compressed folding produced small-scale folds, now evident occurrence as parasitic folds on larger scale folds at the flow front (Fig. 5a–d). With advance over a steeper slope, the flow developed a channelled morphology and stretching lineations (Fig. 8; stage 2). As the topography shallowed in more distal flow regions (Fig. 2), the flow spread and compressional forces became prevalent, producing larger folds (Fig. 5a, c, d; stage 3). Cooling of the upper surface, coupled with increased strain from flow advance over a variable topography, promoted the onset of autobrecciation, which the advancing flow overrode to form the top and basal breccia. With continued extrusion, previously formed structures were reorientated by progressive deformation (stage 4), resulting in a complex arrangement of structures. For instance, the sheath folds (Fig. 5e–g) likely formed due to stretching and amplification of pre-existing curvilinear folds and structures (Reber et al. 2013). As the lava continued to advance downslope, individual flow units advanced and overlapped, forming flow ramps (Cas and Wright 1987; Gottsmann and Dingwell 2001). Continual extrusion, repetitive cycles of breaking, flowing and healing (Tuffen et al. 2003; Tuffen and Dingwell 2005; Vasseur et al. 2013; Cabrera et al. 2015) and progressive deformation produced a third generation of largest scale folding. Small-scale folds were superimposed upon the younger medium- to large-scale folding (Fig. 5a, c, d). As the lava flow cooled during emplacement, the flow probably advanced more slowly due to talus and breccia development, as well as compression against a stalled flow front (Griffiths and Fink 1992). With variations in strain rate and eventual cooling below the glass transition, more areas of the flow deformed in a brittle manner (stage 5), resulting in faults, fracturing and autobrecciation. Once the flow had cooled sufficiently, a brittle deformational phase dominated, and the flow no longer behaved as a viscous or ductile flow.

Implications for emplacement processes

Since 2008, new observations at Chaitén and Cordón Caulle have provided an opportunity to witness emplacement of rhyolitic lava flows (Carn et al. 2009; Lara 2009; Castro et al. 2013; Schipper et al. 2013; Tuffen et al. 2013). Recent active observations also give new perspectives on the formation of older rhyolitic lavas, such as the Rocche Rosse. This is particularly true of the Cordón Caulle eruption, which flowed away from the vent (Tuffen et al. 2013), as opposed to the dome-forming eruption of Chaitén (Carn et al. 2009; Lara 2009).

Direct comparisons of structures can be made between the Rocche Rosse and Cordón Caulle obsidian flows. First, a prominent flow lobe formed at Cordón Caulle due to overtopping a topographic barrier and flow down a steepening slope (Tuffen et al. 2013; Magnall et al. 2017). We suggest the same origin for the prominent western lobe in the Rocche Rosse. The Cordón Caulle flow developed due to thermal insulation promoting flow advance, the underlying slope creating channel and compound flow formation, and the roles of a crust in controlling flow configuration and brittle domination as flow cooled (Tuffen et al. 2013; Magnall et al. 2017). These processes are all inferred in the Rocche Rosse based on observations made at Cordón Caulle, and the role played by these variable factors may be understated in the reconstruction of other older lava flows. Recent studies (Carn et al. 2009; Lara 2009; Castro et al. 2013; Schipper et al. 2013; Tuffen et al. 2013; Bertin et al. 2015; Farquharson et al. 2015; Magnall et al. 2017) of active flows and domes also highlight processes which may not be evident in older flows due to overprinting or lack of exposure.

Conclusions

The structurally complex Rocche Rosse rhyolitic lava flow preserves evidence for multi-phase emplacement and deformation stages. Due to the high viscosity of the advancing lava, indicators of emplacement processes and flow-related strain are recorded in the preserved structures of the final solidified flow. The Rocche Rosse is characterised by a channelled, compound lava flow morphology, made up of two main flow frontal lobes small (< 0.2 m) to large scale (> 2 m) folds, foliations and stretching lineations. These planar and linear structures trend and dip/plunge at highly variable orientations and magnitudes. Based on GeoEye-1 and surface observations and measurements, five concomitant stages of emplacement have been identified: (1) initial extrusion, (2) extensional strain onset, (3) compressional forces at topographic break, (4) progressive deformation and (5) a brittle deformational regime. By traditional structural measurement techniques on ancient flows and new observations made on active flows, the roles of down-flow varying dynamic processes and their role in the overall configuration development of this flow and other older rhyolitic lava flows can be defined. For instance, the role of cooling and crust formation are important in the development of multi-scale folds, flow ramps and autobrecciation and thermal insulation and underlying slope can prolong flow propagation and promote prominent lobe formation. The structural complexity observed in the Rocche Rosse flow can be likened to the witnessed actively deforming Cordón Caulle flow, thus providing a structural framework for the evaluation of flow process emplacement interpretations.