Volcanic systems have many ways of recording and responding to variations in their eruptive environments. As early as the 1800s pillow basalts were speculated to form during eruptions of lava in water (aquatic environment; cf. Lewis, 1904)—an assumption that was not confirmed by observation until the 1970s (Moore, 1970; Walker, 1992). In the early 1900s, volcanic deposits with abundant palagonite and associated with glacial sediments were suggested to have formed within a cryospheric environment (the Palagonite/Mobérg Formation of Iceland: Peacock, 1926; the Tuya Formation in British Columbia, Canada: Kerr, 1948, 1940; Mathews, 1947). In contrast, the first written detailed description of an explosive subaerial eruption dates back to at least 79 AD (Pliny the Younger, 2014), and the majority of our observations are from this environment, for which abundant diagnostic criteria exist. For example, deposits of airfall tephra typically drape the entire landscape and lavas flow unimpeded except by local topography.

Forensic volcanological studies are increasingly able to use volcanic stratigraphy to track dynamic changes in environmental conditions attending volcanic eruptions. For example, Russell et al. (2021) mapped variations in the elevations of passages zones preserved in 1.9 Ma glaciovolcanic deposits that recorded transitions from aquatic to subaerial depositional environments. They used these stratigraphic relationships to reconstruct transient fluctuations in the height and depth of the syn-eruption englacial lake that occurred on time-scales of days to weeks to months. Future studies of eruptions and deposits from the cryospheric environment will increasingly be key for reconstructing and tracking changes in Earth’s climate that have occurred over the last 3 Ma or more (Smellie and Edwards, 2016; Smellie, 2018).

All three environments (i.e. air, water, ice) dictate styles of eruption, edifice morphologies, and the distribution and nature of volcanic lithofacies. For the purposes of this work, we consider: the subaerial environment as dominated by air at Earth’s surface under standard conditions (25 C, 0.101 MPa; Table 1); the aquatic environment as dominated by sea water or freshwater with essentially uniform properties (e.g., Cas and Simmons, 2018; Table 1); and the cryospheric environment as dominated by glacial ice but also including permafrost and seasonal snow pack (Figure 1; Edwards et al., 2020; Curtis and Kyle, 2017). While much of what is reviewed here applies to those parts of the cryosphere, the focus of this contribution is on cryospheric ice. The first two environments (i.e. subaerial/aquatic) are well understood and have been the subject of numerous other reviews (see Cas and Simmons, 2018, for the aquatic environment). Thus, the focus of our work here is summarizing how the cryospheric environment affects volcano-magmatic systems. We provide a framework for understanding the influences of the cryosphere on (glacio)volcanic systems, landforms and deposits in contrast to the other two environments. We have expanded the review of environmental physical properties for subaerial and aqueous environments of Cas and Simmons (2018) to include the cryospheric environment. Our compilation and analysis highlight the unique impacts of the cryosphere on igneous systems, including the potential for: 1) modulating the timing and rates of magmatism, 2) creating unique edifice-scale morphologies, 3) controlling eruption processes, and 4) restricting emplacement and deposition of unique volcanic products.

The most obvious physical difference between the subaerial, aquatic and cryospheric environments is the medium into which an eruption occurs: air, water, or ice (and resulting meltwater), respectively. A summary of key physical parameters characteristic of these three environments (Table 1) shows that the cryospheric environment is the most extreme and the most dynamic. Melting of the enveloping ice can cause rapid changes in eruptive pressure at the vent and melting of ice and heating of meltwater allows for efficient heat transfer and loss from the volcanic system.

The rheological properties of the enveloping media for the three environments are very different. Viscosities vary by at least one order of magnitude (air versus water), and up to 17 orders of magnitude (air versus ice). On the time scales of volcanic eruption, however, the cryospheric environment presents a viscoelastic envelope that can melt and/or flow, or fail brittley. This rheological difference in environment directly controls the distribution of explosive and effusive deposits resulting from glaciovolcanic eruptions. For example, whereas subaerial and aquatic environments allow for relatively unimpeded flow of lavas (e.g., Gregg and Fornari, 1997), cryospheric environments limit the aerial distribution of lavas by imposing physical barriers, even though snow and ice can support the mass of flowing lava over short time-scales (e.g., 24 h or less; Edwards et al., 2012, 2013, 2015). In explosive eruptions, the enveloping media strongly controls the distribution of volcanic ash; large subaerial eruptions can produce global distributions of tephra and water can distribute tephra rafts across ocean basins, while cryospheric environments can, although do not always, restrict ash dispersal because ash frequently ash is deposited in cavities within the ice or in the englacial lake (e.g., Jude-Eton et al., 2012), and transport on the ice surface will be slow. For eruptions through very thin cryosphere, such as snowpack, dispersal may not be restricted at all.

Density differences in the surrounding media translate into substantial differences in environmental vent pressures (EVP). Depending on the depth of water or thickness of overlying ice EVP can be ∼2 orders of magnitude higher in aquatic and cryospheric environments than subaerial ones (Table 1), which potentially suppresses expansion of volatile phases and subsequently influences fragmentation and the distribution of tephra relative to subaerial environments (cf. Cas and Simmons, 2018). For subaerial and aqueous environments, EVPs are essentially static during eruptions while for cryospheric environments EVP can be dynamic and can vary by one to two orders of magnitude during the eruption as the enclosing ice melts and the resulting meltwater drains (Figure 2; Table 1).

The three environments also dissipate heat in very different ways and with different efficiencies. Conductive heat loss to the enveloping media depends on the associated physical properties. Thermal conductivity (K) reflects the materials’ ability to conduct heat in a thermal gradient. Air is the most insulatory (lowest K) and is one order of magnitude lower than water and 2 orders of magnitude less than ice. Thermal diffusivity (κ) values dictate characteristic rates of heat transfer through the media, and values of κ for air are 2 orders of magnitude higher than for water and 1 order higher than for ice (Table 1). The lower viscosities of air and water, relative to ice, mean that heat transfer in those environments is likely to be dominated by advective heat transfer. The heat transfer coefficient, which is a simplified proxy for the complexities of advective heat transfer processes, is significantly higher for water than for air (Table 1). In the cryospheric environment, meltwater is rapidly produced by melting of the enclosing ice at which point advective heat transfer will dominate.

Heat in the volcanic system is also dissipated by endothermic phase transitions that attend aquatic and cryospheric eruptions. As an example, for an eruption to convert its enveloping medium to a gas phase at 100°C requires 0.12 × 10³ kJ m⁻³ for air, 5,000.0 × 10³ kJ m⁻³ for water, and 5,300.0 × 10³ kJ m⁻³ for ice, including the phase transitions from ice to water, and from water to steam. In the aqueous environment, about half of the required energy is used to heat water from initial temperatures, here assumed to 0°C although they could be as low as −4°C for abyssal marine environments, to phase transition temperatures for the liquid:gas/fluid transition that depend on pressure (i.e. water depth). In cryospheric environments the temperatures in the vent area will be close to the temperature for ice:water phase transition (∼0°C); even polar ice is generally at temperatures above −60°C (Cuffey and Paterson, 2010). However, the environmental medium (i.e. ice) can pass through two phase transitions via melting and vaporization, consuming the largest amount of energy per cubic meter of erupted material (Table 1). The differences in thermal properties are responsible for many of the macro- and microscopic differences seen in deposits from the water- and ice-dominated eruption environments, relative to subaerial volcanic deposits. As developed below, the physical conditions of the cryosphere, relative to the subaerial and aquatic environments, produce signatory characteristics at volcanoes and in their deposits, which has led to a new subdiscipline known simply as “glaciovolcanism” (Kelman et al., 2002; Smellie, 2009; Edwards et al., 2014; Russell et al., 2014; Smellie and Edwards, 2016)

The cryosphere (Figure 3) perturbs magmatic systems mainly during interstadial and interglacial periods, when ice thickness is changing and influencing lithostatic stress conditions (Andrew and Gudmundsson, 2007; Geyer and Bindeman, 2011). Static ice is no different than any other kind of lithospheric load, although compared to typical sediments or other surficial deposits, ice is about 50–70 percent less dense and is susceptible to higher rates of removal. Land-based accumulation of ice, glaciers and ice sheets operates on time scales of tens of thousands of years based on Pleistocene proxies for global sea level and ocean temperatures (e.g., Lisiecki and Raymo, 2005). However, those same records are interpreted to suggest that deglaciations happen much more rapidly. The Laurentide, Cordilleran and Svennoscandian ice sheets are all thought to have been at least 2 km thick (e.g., Clague and Ward, 2011; Lambeck et al., 2006), while at the Last Glacial Maximum the Icelandic ice sheet was at least 1 km thick on average (Hubbard et al., 2006; Le Breton et al., 2010). Those thicknesses were achieved over a period of ∼90 ka suggesting an average accumulation rate of about 0.02 m per year. To generate a similar lithostatic load from sedimentation requires accumulation rates of about 0.01 m per year, or from volcanic activity about 0.009 m per year. In contrast, full destruction of those ice sheets seems to take place in less than 20 ka, and likely closer to 10 ka (Lambeck et al., 2014). Those time-scales suggest ice sheet “erosion” rates about one order of magnitude larger than the accumulation rates (∼0.1 m per year). These “ablation rates” are significantly higher than those typically assigned to rates of erosion for most terrestrial environments (e.g., 0.0004 m/yr; Ferrier et al., 2007), or for rates of sea level rise during deglaciations (e.g., 0.01–0.04 m per year after the Last Glacial Maximum; Lambeck et al., 2006). After some lag time following rapid interglacial sea level rise, high strain rates combine with increased lithospheric loads to induce crustal stress regimes that suppress dike initiation and cause eruptions to cease in marine-dominated environments (Satow et al., 2021). Thus, the characteristic slow loading rate and more rapid unloading rate of ice sheets can also be expected to impact magma source regions in glacial environments by altering the lithospheric stress conditions, influencing the depths and extents of melting, magma transport and storage processes, and pressure and temperature conditions of magma evolution (e.g., crystallization, devolatilization).

For areas where decompression melting is prominent like Iceland, lithospheric unloading has been suggested to change conditions within the mantle melting region and to produce distinctive geochemical signatures (Jull and McKenzie, 1996; Gee et al., 1998; Slater et al., 1998; Maclennan et al., 2002; Sinton et al., 2005; Sims et al., 2013; Eason and Sinton, 2015; Eksinchol et al., 2019). Compared to lavas erupted during isostatic ice conditions, lavas erupted during deglaciation tend to be depleted in radiogenic isotopes and incompatible trace elements (Figure 3). Geochemical modeling has shown that the extent of isotopic and incompatible element depletion observed in interglacial lavas cannot be explained by changes in magma chamber processes or mantle source composition (e.g., Maclennan et al., 2002). Instead, the geochemical trends are consistent with increased mantle melting and magma ascent rates from the shallow, isotopically and incompatible element-depleted part of the mantle (Harðarson and Fitton, 1991; Jull and McKenzie, 1996; Slater et al., 1998; Eksinchol et al., 2019). In some locations, the shift in melting conditions is also recorded in major element variations (Eason and Sinton, 2015). An additional impact of deglaciation on major elements, and other parameters controlled by crustal magmatic differentiation processes, can be expected because of the effects of unloading on crustal magma transport (Wilson and Russell, 2020). Rapid deglaciation creates local stress fields favorable for dike propagation, with the potential for erupting variably evolved magmas from a range of crustal depths (Wilson and Russell, 2020).

In volcanic arcs, the effects of glaciation on eruptive chemistry and volcanic output are more likely to be driven by changes in crustal stress and rheological conditions (Geyer and Bindeman, 2011). Several different groups have suggested that hydrothermal alteration weakens host rock and glacial loading and unloading produce changes in the crustal lithosphere stress field that inhibit or drive magma storage and transport (e.g., Jellinek et al., 2004; Geyer and Bindeman, 2011; Praetorius et al., 2016; Rawson et al., 2016; Wilson and Russell, 2020). Over long time-scales crustal loading may impede dyke formation and enhance sill production, while unloading favors dyke formation and enhances magma transport into the eruption region (Gudmundsson, 1986; Sigvaldason et al., 1992; Gee et al., 1998; Glazner et al., 1999; McLeod and Tait, 1999; Jellinek and DePaolo, 2003; Jellinek et al., 2004; Andrew and Gudmundsson, 2007; Albino et al., 2010; Wilson and Russell, 2020). The cyclic loading and unloading of ice ages can lead to “glacial pumping in magma-charged lithosphere” (Edwards and Russell, 2002; Wilson and Russell, 2020). Over short time-scales, the local stress field induced by the overlying ice sheet may interact with dyke orientation to enhance or reduce eruptive activity, thereby influencing crustal magma storage capacity (Hooper et al., 2011). Extended crustal residence times enhance magmatic differentiation by crystallization, assimilation, and mixing processes, thereby recording glaciation-driven events in lava geochemistry (Edwards et al., 2002; Asmerom et al., 2005; Wilson and Russell, 2020).

For magmas that are already close to a state of volatile saturation, rapid changes in lithospheric pressures during interstadial periods and deglaciations are of a magnitude that could drive vesiculation (Bindeman et al., 2010; Geyer and Bindeman, 2011). Sudden load removal caused by volcanic edifice destabilization and catastrophic flank collapse can force the system into a two-phase state, (Capra, 2006; Tormey, 2010; Capra et al., 2015). Rapid gravitational failure and volatile exsolution can potentially trigger an eruption or at the very least increase the eruptibility of the magma (Bindeman et al., 2010; Geyer and Bindeman, 2011; Sparks and Cashman, 2017). Separation of a volatile phase also can change the stability of crystallizing phases, and so the dynamicity of the cryospheric environment has the potential to indirectly alter the liquid line of descent for pre-existing magmas.

The imprint (Figure 4) of the cryospheric eruption environment is most visible and iconic at the edifice scale (Figure 4). Entire volcanoes in many parts of the world owe their distinctive morphologies to eruptions within ice while others have a more subtle morphological imprint including: anomalously steep cliffs of ice-dammed (i.e. overthickened) lavas, inverted topography caused by lava emplacement within ice-confined valleys, and characteristic resurfacing and reshaping of edifices due to ice erosion and buttress-destabilization (Figure 4). The morphology of tuyas and other cryospherically constrained volcanic edifices has been discussed in detail by several authors (Hickson, 2000; Smellie, 2009, 2018; Pedersen and Grosse, 2014; Russell et al., 2014; Pedersen, 2016), so here we only review the main elements that can be diagnostic of cryospheric environments.

Tuyas, less commonly referred to as “stapi” and “table mountains,” are volcanoes whose overall morphology is a direct result of eruptions within confining ice (e.g., Mathews, 1947; Smellie, 2009; Pedersen and Grosse, 2014; Russell et al., 2014). These volcanic landforms have three distinct morphologies: 1) flat-topped platforms comprising laterally continuous, subhorizontal lavas commonly capping subaqueous lava-fed deltas (Figure 4A), 3) conical masses comprising palagonitized tephra and intrusions (Figure 4B), or 3) elongate ridges of subaqueous tephra and pillow lava (Figure 4C). A limited number of compound tuyas have been identified, particularly in British Columbia (Figure 4D). There, Kima’Kho tuya features a tephra cone onlapped by a platform of dipping beds of pillow lava breccias capped by horizontal sheets of lava (e.g., Mathews, 1947; Russell et al., 2021). Russell et al. (2014) advocated using the term “tuya” as a generic term to encompass all three morphologies and their combinations, while other authors have more nuanced views (e.g., Smellie, 2009). Recognition of these morphological elements as being unique to the cryospheric eruption environment has been critical for inferring the thicknesses and extents of paleo-ice-sheets on Earth and on Mars (Smellie and Edwards, 2016; Edwards et al., 2020). However, the unexamined assumption that a flat-topped volcano defines a cryospheric environment is problematical especially on not-yet-visited planets. At least in shallow water, the aquatic eruption environment can produce similar morphologies, with the best example being Surtsey volcano, located off the coast of south Iceland.

Longer-lived volcanoes have the capacity to span glacial-interglacial transitions such that only select morphological elements serve to record cryogenic periods of eruption (Figures 4E–G). At Hoodoo Mountain, a volcano in northwestern British Columbia that had eruptions spanning 100 k.y. (Edwards et al., 2002), the prominent cliffs that circumscribe much of the base of the edifice as well as its broad, gently rounded summit that still hosts a small ice-cap form a distinct, squat morphology (Figure 4E). Mount Rainier is a Cascade volcano with a morphology that is typical of andesitic stratovolcanos, however, its extended (500 k.y.) eruption history includes periods of ice-confined volcanic eruption which have not created obvious morphological elements reflecting its glaciovolcanic history (Figure 4F; Sisson et al., 2014). Detailed work by Lescinsky and Fink (2000), however, identified a number of lava ridges which they established as resulting from lateral impoundment by bounding glaciers. At more mafic long-lived volcanoes like Villarrica and Osorno, in central Chile, fluid lavas largely obscure the morphological influences of the cryogenic environment, where young volcanic cones and lavas are likely built upon older, glaciovolcanic deposits (Figure 4G).

The complex (Figure 5) influences of the cryospheric environment on volcanic eruptions has been well documented during several historic eruptions (e.g., 1980 Mount St. Helens, 1996 Gjálp, 2004 Grimsvötn, 2009 Redoubt, 2010 Eyjafjallajökull, 2012–13 Tolbachik; cf. Smellie and Edwards, 2016). Syn-eruption melting of the enclosing ice leads to dynamic pressure changes at the vent and in the environment as coherent icesheet transitions to fracturing and collapsing of the icesheet, and to a rising and falling column of meltwater. Observations on modern eruptions in the cryospheric environment show that the ice melting and formation of ephemeral englacial lakes can be rapid but are areally-limited so that volcanic deposits are largely confined by the enclosing ice (e.g., Jude-Eton et al., 2012; Magnusson et al., 2012; Oddsson, 2016). The linked 2010 Eyjafjallajökull and Fimmvorðuhals eruptions in Iceland are a clear example of this complexity during an essentially single, continuous volcano-magmatic event within the cryosphere (Figure 5).

The initial flank eruption of basalt at Fimmvorðuhals from mid-March to mid-April produced lavas that flowed across the top of a 3–5 m snowpack but eventually melted through snowpack to produce ice-confined lavas (Edwards et al., 2012). Locally however, a combination of thicker basal lava breccia and thicker snowpack resulted in several lavas crossing and remaining on top of the snow, even months after the eruption stopped. Just 1 day after the initial eruption ended on 13 April, the eruption shifted to the volcano summit and vented through the ∼200 m thick summit glacier and produced at least three explosive craters (Magnusson et al., 2012; Figure 5B). One of these eventually enlarged to become the main vent for the eruption and was partly filled with scoria and course tephra. These initial explosive eruptions produced jökulhlaups that predominately passed down the north side of the volcano, through Gigjökull, which is the largest outlet glacier on the north side of Eyjafjallajökull. These initial floods likely exploited and enlarged the pre-existing subglacial drainage N- and R-channels. Between 18 April and 4 May, continued summit explosive eruptions were accompanied by lava effusion. The lavas initially flowed beneath the surface of Gigjökull, exploiting the same R- and N-channels that had been enlarged by meltwater (Oddsson et al., 2016; Figure 5C). Later “subaerial” lava flowed down a narrow, ice-confined channel through Gigjökull (Figure 5D). Distinctive textures of the lava emplaced beneath the ice, including thick outer vitric rinds and highly fractured surfaces, are interpreted to be indicative of an external “coolant” (Oddsson et al., 2016). This single event produced: 1) proximal and distal tephra airfall deposited on land, neighbouring glaciers, and in the ocean, 2) tephra syn-eruptively entrained and transported by meltwater flowing beneath Gigjökull glacier and by the Markarflöt River, 3) lava emplaced on top of snowpack, 4) lava emplaced within tunnels at the base of Gigjökull, and 5) lava emplaced subaerially within an ice canyon.

Outstanding questions concern how the cryospheric environment impacts magma fragmentation and explosivity. Recent work on eruptions in aqueous (englacial) environments by Rowell et al. (2021) suggests that magma-water interactions can produce more energetic eruption columns. As discussed above, the presence of ice and water can increase rates of heat transfer from the magma (and its fragments) to its environment. The accelerated cooling causes rapid increases in melt viscosity to the point of the glass transition temperature where brittle behaviour dominates over ductile. The thermal stresses within volcanic glass raised by rapid heat extraction provides a means for efficient particle fragmentation. Liquid water trapped within low permeability deposits at high-temperature is easily vaporized causing explosive fragmentation driven by expansion of the gas phase (e.g., Dürig et al., 2020).

The cryospheric (Figure 6) eruption environment exerts a pronounced control on the properties, geometries and distributions of volcanic lithofacies. Salient summaries and descriptions of characteristic glaciovolcanic lithofacies are provided by Smellie (2000), Edwards et al. (2014), Russell et al. (2014), and Smellie and Edwards (2016), as well as, in many deposit-specific studies. Here we limit our discussion to the main elements that are diagnostic of eruptions in the cryosphere (Figure 6). These features include anomalous deposit morphologies and properties that reflect confinement, anomalous high rates of cooling (i.e. quenching), and, especially, the presence of a standing body of meltwater in a physiographically implausible location. Most of these deposits show evidence for complicated cooling histories involving variable rates and directions of heat transfer.

Evidence for confinement of the distributions of lavas and volcaniclastic deposits is the signatory of the cryospheric eruption environment. This impoundment can be recorded by deposit geometry, distribution, and other properties. While snow and ice can temporarily support the mass of an overlying lava (Figure 6A), heat transfer from the base of the lava eventually leads to the lavas sinking into the cryospheric cover (Figure 6B). This has been well documented at historic eruptions in Iceland (Edwards et al., 2012) and in Russia (Edwards et al., 2015). Horizontal to shallowly-dipping sequences of lava perched near the summits of stratovolcanoes have been uniquely interpreted as resulting from ice confinement (Lescinsky and Fink, 2000; Figures 6C,D). Impoundment or confinement commonly results in overthickened lavas, which is especially evident for low viscosity lava (i.e., phonolite, basalt, nephelinite, basanite; Harder and Russell, 2007; Figures 6E,F) but also found in intermediate composition glaciovolcanic lavas. The extreme case of ice-confined effusive volcanism leads to the formation of lava-dominated tuyas (Kelmen et al., 2002; Wilson et al., 2019; Hodgetts et al., 2021). The key to interpreting these features as having formed in a cryospheric environment depends on there being no logical present-day physiographic explanation for confinement of the flow of lavas.

On a per volume basis most silicate melts release enough heat through cooling to melt about 3–10 times an equivalent volume of ice. Regardless, ice can effectively impede the flow of lava for two main reasons. Firstly, the continuous unimpeded flow of lava would require the volumetric rate of ice melting to match the eruptive flux. Although the heat content of lava is more than sufficient to melt the ice, the time scales of heat transfer from lava to ice are too slow relative to most effusive fluxes. Edwards et al. (2012) showed that lavas can flow onto and across the top of snow or ice with little melting for all but the very slowest-moving flows. This has also been demonstrated experimentally (Edwards et al., 2013), and observed at a number of eruptions (e.g., 1947 Hekla, Einarson, 1949; 2010 Fimmvorðuhals, Edwards et al., 2012; Edwards et al., 2013 Tolbachik, Edwards et al., 2015). Secondly, cooling the fronts and sides of advancing lava causes a reduction in temperature and a marked increase in viscosity causing a decrease in velocity and thickening of the lava front. The cooler lava is an insulator preventing efficient heat transfer from the hot lava interior to ice (Wilson and Russell, 2020). The main difficulty in using “overthickened” lavas as indicators of ice-confinement is that we lack coherent global databases that document the range of thicknesses for lavas associated with subaerial eruption environments.

Volcaniclastic deposits can also show the effects of confinement by their geometry and distributions and, thus, can also be indicative of cryospheric environments. The main features that are likely to be preserved from volcaniclastic/pyroclastic deposits are found proximally. More distal or dispersed deposits that populate the ice surface at the time of eruption are likely to be lost, or to be transported within ice from the eruption site. Deposits that accumulate within the cauldron overlying the vent will have limited distributions (<1 km; Figure 6G); they may have subhorizontal bedding even when present topography is steeply sloped (Figure 6H) or presumably over-steepened bedding if deposited abutting an ice wall. Either way, the preserved deposits will have geometries that show little thinning with distance from the vent and terminate with anomalous thicknesses and/or geometries (e.g., Smellie, 2000).

Polygonal/columnar jointing is an important volcanological feature that occurs in a wide variety of deposits (e.g., lavas, dykes, and ignimbrites) and provides a record of paleo-cooling history (e.g., Long and Wood, 1986; Goehring and Morris, 2008; Figure 7). Cooling-induced joints inform on the transient rates and directions of heat flow; column diameter is inversely proportional to cooling rate and the column orientation (length) indicates heat flow direction.

In subaerial environments, cooling columns are commonly well-organized into a lower and upper colonade that are oriented perpendicular to cooling surfaces (i.e., ground, valley walls, and etc.). The colonnades are sometimes separated by an entabulature comprising subvertically-oriented columns with a limited range in sizes (Figure 7D). In cryogenic environments, columnar jointing is much more varied and, in many cases, can be diagnostic, or provide supporting evidence, of that environment (e.g., Lescinsky and Fink, 2000; Edwards and Russell, 2002; Lodge and Lescinsky, 2009; Forbes et al., 2014). Lavas cooled in glaciovolcanic settings commonly feature fine-scale columns (cube shaped or small diameter columns; Figure 7A) that are indicative of rapid cooling. Column orientations are also commonly diagnostic. In many glaciovolcanic deposits the axes of columns populating the cliff faces of lavas are subhorizontal indicating efficient and sustained heat transfer outwards to a surface that is no longer there (Figure 7B). Given the settings and geography of these deposits, at relatively high elevations and/or latitudes, the obvious candidate for the missing confining material is a mass of ice, which explains the anomalous thickness of lava and orientations of columnar joints. In other situations, glaciovolcanic lava masses have radiating and curving clusters of columnar joints (i.e., lobes or channels; Figure 7C). Commonly the columns also show a well-organized gradual increase in diameter moving from the exterior surface inwards. The size gradation is a record of relative rates of heat transfer at the moment of columnar joint propagation and shows that heat loss is most efficient on the exterior surface and significantly slower in the interior of the lava mass. The radiating cluster of columns indicates that the exterior surface lava mass was cooling rapidly in all directions and heat was transferred efficiently to an enveloping medium (Figure 7E). The sometimes observed inward-curving character of the columns shows how the heat transfer direction changed as the columns propagated inwards from the exterior surface.

Eruptions within ice sheets lead to rapid melting and collapse of the overlying ice and commonly produce an englacial lake whose depth is dictated by the amount and rate of melting, the flux of volcanic material, the rate of sub-ice leakage, and the thickness of the confining ice (e.g., Russell et al., 2021). Eruptions into a standing body of water result in distinctive volcanic lithofacies that can be diagnostic, while not definitive, of the cryospheric environment including: pillow lava, pillow-lava breccias, hyaloclastite, peperitic dykes, lava-fed deltas, and pervasive and variable palagonitization of the deposits (cf. Smellie and Edwards 2016 for summaries and references). Pervasive palagonitization requires abundant vitric material, water and heat. Melting of enclosing ice produces abundant heated water, and, because the deposits are largely confined to the vent area by enclosing ice, cooling is less efficient. Additionally, dykes are commonly found within palagonitized glaciovolcanic deposits, and their emplacement into the tephra is another source of heat as the dykes cool and solidify, helping to maintain higher temperatures within the tephra pile (e.g., Edwards et al., 2009; Harris Russell, 2022).

Taken individually, most of these lithofacies are not exclusively diagnostic of glaciovoclanism but they clearly demand an aquatic environment and are strong evidence for a standing body of water. The key, as seen at many tuyas, is having multiple lines of evidence for a substantial (deep) standing body of water in situations where there is clearly no obvious means to create and sustain such a lake. Collectively these observations provide the evidence for a cryospheric eruption environment (e.g., Smellie and Edwards, 2016; Smellie, 2018).

When compared to subaerial and subaqueous environments, the cryospheric environment is the most dynamic and complex. This is well-illustrated in two examples. Firstly, by an examination of the stability and dynamics of syn-eruptive englacial lakes of melt water formed during glaciovolcanic eruptions (cf. Smellie and Edwards, 2016; Russell et al., 2021). Secondly, by an analysis of how changes in pressure during eruptions can impact vesiculation and vesiculation induced fragmentation.

Volcanic eruptions within cryospheric environments are controlled by all of the variables and conditions found in subaerial and subaqueous eruption environments save, perhaps, the conditions attending deeper submarine eruptions (Cas and Simmons, 2018). Vent systems can be repeatedly flooded by meltwater, and then drained. Initial stages of eruption are most influenced by environmental confinement but sustained larger eruptions can evolve to support the more long-distance transport of tephra generally attributed to explosive subaerial eruptions (e.g., Eyjafjallajökull 2010). The unique characteristics of these deposits are key to recognizing their glaciovolcanic origins and provide the means to constrain the syn-eruptive cryospheric environmental conditions (e.g., depth of water, ice thickness). Given the wide-spread distribution of these deposits throughout the Plio-Pleistocene, they remain one of the largest land-based archives available for reconstructing past ice extents, thicknesses and hydrologies on Earth, as well as, other planetary bodies (Smellie and Edwards, 2016; Smellie, 2018; Edwards et al., 2020; Wilson and Russell, 2020). Because the it can change over relatively short time-scales, its englacial lakes can be highly dynamic, and eruptions within the it can transition to subaqueous and subaerial environments repeatedly, the cryospheric environment is the most complex eruption environment for volcano-magmatic systems.