Volcano-ice interactions - Brown University · Modeling Volcanic Processes: The Physics and...

VOLCANO-ICE INTERACTIONS: Wilson, Smellie and Head page 1

Volcano-ice interactions

Lionel Wilson1,3, John L. Smellie2 and James W. Head3

1Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK

2British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, UK

3Department of Geological Sciences, Brown University, Providence, RI 02912, USA

Submitted to: Modeling Volcanic Processes: The Physics and Mathematics of Volcanism

edited by

Sarah A. Fagents, Tracy K.P. Gregg and Rosaly M.C. Lopes

Cambridge University Press

Abstract: We review basic physical processes that control the interactions between silicate

magmas and surface ice and snow layers, focussing on subglacial, englacial and supraglacial

interactions. Where possible we have linked theoretical considerations with observations of the

lithofacies and sequence characteristics of the deposits expected as a result of these various

interactions, with a particular focus on the products of basaltic eruptions. The range of possible

interactions is large, resulting in a correspondingly diverse group of resulting landforms. These

predictions are made for the environment of the Earth, but with suitable changes to atmospheric

temperature and pressure and acceleration due to gravity are readily applicable on Mars.

Numerous putative Mars examples have already been documented and this paper provides a

comprehensive unifying theoretical framework for their further interpretation.


1. Introduction

Magma-ice interactions can occur in a number of ways and can produce a range of

products and landforms (e.g. Lescinsky & Fink, 2000; Mee et al., 2006; Komatsu et al., 2007;

Larsen & Eiriksson, 2008; Smellie, 2009), the details depending on the geometry and time scale

of the interaction. Ice, in the form of glaciers, snow-fall or frozen lakes, commonly occurs as a

sheet emplaced onto a silicate rock surface, and magma rising through the shallow crust always

travels through brittle fractures, i.e. in dikes. Thus, possible initial geometries of the interaction

between rising magma in dikes and ice are (a) injection of a magmatic dike into a layer of ice;

(b) formation of what is initially a sill-like magma intrusion beneath ice at the ice-rock interface,

though this may ultimately develop into what might be thought of as a sub-glacial lava flow; and

(c) emplacement of a lava flow or a pyroclastic deposit on top of ice from a vent not within the

ice layer (Wilson & Head, 2002, 2007, 2008).

Case (a), injection of magma into an ice layer as a dike, is predicted to be a possibility on

theoretical grounds (Wilson & Head, 2002). The inherent instability of the resulting geometry

means that a fragmental deposit surrounding the vent (fissure) at the base of the ice sheet is

likely to be produced at the end of the interaction. If exposed after removal of the ice, such

deposits may potentially be confused with those formed by other mechanisms, particularly parts

of some hyaloclastite-bearing ridges generated by eruptions under ice.

In case (b), melting of ice overlying an intrusion at the ice base can produce a large volume

of water (Höskuldsson & Sparks, 1997; Wilson & Head, 2002, 2007; Tuffen, 2007). The water

may accumulate relatively slowly, but in the case of a large body of water accumulated under a

glacier it may eventually escape very rapidly in the form of a jökulhlaup. If the water escapes at

the glacier edge and is replaced by air, and the overlying ice does not deform too quickly, the sill

may evolve into what is effectively a subaerial lava flow in an ice cave, resulting in a range of

possible features (Wilson & Head, 2002, 2007). Key issues then are the relative values of the

rate of advance of the magma sheet beneath the ice, the rate of advance of thermal waves

penetrating both the unmelted ice and the magma itself, and the rate of drainage of the melt water

produced.


In case (c) the presence of the ice plays no part in the magmatic eruption (except in so far

as glaciation in general may influence crustal loading and stresses, but that is a secondary issue

not addressed here). If the eruption involves deposition of a pyroclastic fall deposit onto ice, the

consequences depend on the rate of thickening of the deposit and its final total thickness (e.g.,

Wilson & Head, 2007, 2008). Except very close to the vent, fall deposits are likely to consist of

clasts with temperatures close to the ambient atmospheric temperature, and so immediate heating

of underlying ice will be minimal. Longer term, the fact that the pyroclasts will almost certainly

have a lower albedo than the ice, and hence will reach higher daytime temperatures during the

diurnal solar heating cycle, may be an important factor. If the volcanic material is hot on

emplacement, e.g. a lava flow or a pyroclastic density current deposit, the effects on the

underlying ice will depend on both the rate of advance and thickness of the flowing material, and

the rate at which water can migrate beneath the advancing deposit to escape at its edges.

We now examine the above processes theoretically, and give suggested and/or actual

examples, principally for basaltic deposits, of glaciovolcanic deposits that might form from those

processes, to facilitate the kinds of field observations needed to constrain the models and to

identify subtleties so far overlooked.

2. Englacial dyke emplacement

2.1. Theoretical issues

Magma propagation at shallow depths in planetary bodies takes place in brittle fractures

(dikes and sills), held open by a combination of magma pressure and the stress field in the host

rocks (Pollard, 1987; Rubin & Pollard, 1987). Dikes propagate both vertically and laterally

when the least principal stress is horizontal. Some reach the surface to feed eruptions, but others

stall as intrusions, and these may cause surface manifestations such as bedrock graben or ice

fractures (crevasses) if their tops are sufficiently close to the surface (Mastin & Pollard, 1988;

Rubin, 1992; Gudmundsson et al., 2004). If the least principal stress changes from horizontal to

vertical a dike will cease to propagate upward but may still propagate laterally to form a sill.


The fact that a dike approaching the surface in a location where an ice layer of substantial

thickness is present is likely to encounter such a stress change, coupled with a discontinuity in

material properties, may encourage sill injection at the rock-ice interface. Such sills are expected

to evolve rapidly into sub-glacial lava flows (Wilson & Head, 2002; Smellie, 2008), the subject

of Section 3.

The density structure of the crust in volcanic provinces is commonly such that mafic

magma generated in the mantle is positively buoyant in its source zone but negatively buoyant

near the surface. Magma reservoirs form from the accumulation of multiple stalled intrusions at

intermediate levels of neutral magma buoyancy (Ryan, 1987) and dikes are driven upward from

these reservoirs by excess pressures within them. The excess pressures arise due to the positive

buoyancy of the melts feeding them from below and are typically at least several MPa (Parfitt,

1991), and so for reservoir depths of ~3 km the pressure gradient acting on the magma is ~1000

Pa m-1. In cases where magma reaches the surface directly from lower crustal or mantle depths,

a condition that requires the positive buoyancy at great depths to more than balance the negative

buoyancy at shallow depths, net density differences between magma and host rocks are generally

~100 kg m-3 (Parfitt et al., 1993), and this value, combined with the acceleration due to gravity of

~10 m s-2, also leads to pressure gradients of order 1000 Pa m-1. Widths of dikes propagating

from shallow reservoirs are expected to be ~1-3 m (Parfitt, 1991), and under pressure gradients

of this order mafic magma flows at speeds of ~1 m s-1 (Wilson & Head, 1981). The strain rates

near the dyke tips implied by these speeds are thus ~1 s-1, about seven orders of magnitude larger

than the strain rates at which the surrounding ice can flow plastically given the rheological

models (a pseudo–plastic power–law fluid with a yield strength) proposed by Glen (1952), Nye

(1953) and Paterson (1994). Thus a dyke can easily overshoot an ice–rock interface because the

ice appears to the propagating crack as a brittle, low–density rock with elastic properties similar

to those of the basalt substrate (Wilson & Head, 2002). We show below that the amount of ice

melting that takes place on the timescale of dyke emplacement may be small enough for the

initial emplacement process to be stable, though subsequent, more extensive ice melting may

lead to collapse of the dyke.


Wilson & Head (2002) described models of mafic dike propagation in crustal rocks

overlain by ice sheets that can be used as the basis of an illustration of the relevant issues.

Consider a dike that has propagated from a basaltic magma reservoir whose top is at a depth of z

= 2 km below a rock-ice interface, the ice thickness being y (see Figure 1; all symbols used are

defined in Table 1); the top of the dike has come to rest having penetrated a distance p into the

ice. We assume that y will lie in the range 0.5 to 2 km, based on ice-cap thicknesses in Iceland

measured under current conditions (up to ~900 m, Sigmundsson & Einarsson, 1992; Einarsson,

1994) and estimated for glacial periods (1000–1500 m, Einarsson & Albertsson, 1988;

Geirsdóttir & Eiríksson, 1994; Bourgeois et al., 1998) conditions. These are probably typical

thicknesses for most terrestrial subglacial eruptions. Greater thicknesses of ice overlying

volcanoes are possible, of course (e.g. 2-4 km in the volcanically active West Antarctic Rift

System: Blankenship et al., 1993; Behrendt et al., 2002; Corr & Vaughan, 2008), but as they do

not materially affect the geological processes and products we are considering, they are ignored

here, for simplicity. The crustal rock consists of accumulated volcanics with a bulk density of ρr

= 2300 kg m-3 and the ice density is ρi = 917 kg m-3.

For the moment, the pressure in the magma reservoir is assumed to be close to lithostatic,

any initial excess pressure having been relaxed by the propagation of the dike. This condition

will be relevant when a relatively large-volume dike leaves a relatively small-volume reservoir

and provides a base-line for the analysis. Using the above ice and rock depths and densities, the

reservoir pressure will then lie between ~50 and 63 MPa. The pressure distribution in

propagating dikes adjusts to maximise the pressure gradient driving flow of the magma against

wall friction (Rubin, 1992). As a result the pressure in the dike tip is buffered at a low value

determined by the solubility of the most soluble magmatic volatile, essentially always H2O. If

the magma in the reservoir contains, say, 0.25 mass % water, a plausible value for mafic magma

(Gerlach, 1986) the solubility as a function of pressure (e.g. Harris, 1981; Dixon, 1997) defines

this buffering pressure as ~3.3 MPa. With magma reservoir pressures in the above 50-63 range,

the solubility of the other common magmatic volatile, CO2, is such (Dixon, 1997) that it is likely

to be supersaturated in the reservoir, and with a plausible CO2 content of 0.2 mass % (Gerlach,

1986), 0.16-0.17 mass % will have exsolved. Assuming a reservoir magma liquid density of,


say, ρm = 2700 kg m-3, numerical evaluation of the average bulk density β of the magma in the

dike using equations 5(a) and 5(b) in Wilson & Head (2002) yields a value within ~3% of 2480

kg m-3.

When the tip of the dyke comes to rest, magma flow within the dike ceases and no pressure

gradient is needed to balance frictional losses. As a result the pressure in the gas in the tip cavity

can increase from the low, buffered value maintained during the dike growth to reach a final

value Pt. This pressure can be found by equating the lithostatic stress at the reservoir roof to the

weight of the magma in the dike. The lithostatic stress at the reservoir roof, Pl, is given by

Pl = Pa + (g [z ρr + y ρi]) (1),

where Pa is the atmospheric pressure, ~ 0.1 MPa. The pressure due to the weight of the magma

in the dike, Pm, is given by

Pm = (Pt + g β [z + p]) (2).

Equating Pl to Pm and simplifying we have

Pt = Pa + g [ρi y + (ρr - β) z - β p] (3).

Using the densities adopted above, we find (ρr - β) = -180 kg m-3. The relatively small

absolute value of (ρr - β) relative to both ρi and β means that Pt is only weakly dependent on the

magma reservoir depth, z, and is controlled mainly by the ice thickness, y, and the distance that

the dyke tip penetrates into the ice, p. Note that conditions when the dike tip comes to rest are

not likely to lead to Pt becoming less than the 3.3 MPa buffered pressure during dike

propagation: any decrease in pressure below this value would lead to additional exsolution of

water from the magma. We can therefore use eq. (3) to find the maximum distance pmax that the

dike tip can penetrate into an ice layer of a given thickness:


pmax = {ρi y + (ρr - β) z - [(Pt - Pa) / g]} / β (4).

The line labelled 0.25% in Figure 2 shows how pmax varies with y in the above case where the

magma water content is 0.25 mass % and the CO2 content is 0.2 mass %. Unless the ice is at

least 760 m thick the dike tip does not cross the rock-ice interface; instead the dike stalls as an

intrusion in the rock. If the ice is thicker than this threshold value, penetration of the ice occurs,

the penetration reaching ~450 m when the ice is 2000 m thick, so that the top of the dike is

~1550 m below the ice surface. Also shown in Figure 2 are the corresponding results when the

CO2 content is kept constant and the water content is made larger by a factor of 2 (i.e. 0.5%,

corresponding to a dike tip buffering pressure of ~9.0 MPa), is made smaller by a factor of 2

(0.125%, corresponding to a tip pressure of ~1.2 MPa), and is negligible (the line labelled

"small"). If the water content increases, dike penetration into the ice is only possible for larger

ice thicknesses and the maximum amount of penetration of a 2000 m thick ice layer is smaller, at

~200 m. Conversely a lower water content leads to a much larger amount of ice penetration,

with penetration beginning at a smaller ice thickness. In the limiting case of negligible water

magma content about 630 m penetration of a 2000 m thick ice layer is possible. Even so, none

of the dikes modelled in Figure 2 is expected to come close to breaking through to the upper

surface of the ice. Thus dikes from magma reservoirs in which there is no excess (i.e. super-

lithostatic) pressure should not penetrate through glaciers and ice–caps.

However, it will be shown below that an excess pressure is virtually always required to

allow a subglacial intrusion to occur, even though it is not essential for a dike to intrude an ice

layer. If an excess pressure Pe is present, it must be added to the lithostatic load Pl in defining

the magma reservoir pressure; then examination of eqs. (3) and (4) shows that the extent of

possible penetration of a dike into an ice layer is increased by an amount [Pe / (g β)]. For

plausible excess pressures of at least 3-5 MPa (Parfitt, 1991), this implies extra penetration

distances of at least ~120-205 m. The equivalent of Figure 2 with these additional penetration

distances added then shows that complete penetration of thin ice layers by low-volatile content

magmas begins when the excess pressure in the reservoir exceeds ~3.3 MPa. Excess pressures

of ~5 MPa would result in dikes with low volatile contents being able to penetrate nearly half


way through an ice sheet. Magmas with higher volatile contents will require greater excess

pressures than these examples, but nevertheless complete penetration of ice layers up to ~300-

400 meters in thickness would be possible with excess pressures up to 10 MPa. Though the

above calculations are developed for basaltic magmas, the results will not change greatly with

magma type; thus Tuffen & Castro (2009) describe a rhyolitic dike that penetrated a 35-55 m

thick glacial ice layer to build an 80 m high obsidian ridge.

2.2. Phreatomagmatic activity from englacial dikes

Although, as shown earlier, circumstances must exist in which dikes cannot penetrate

completely through surface ice layers, secondary phreatomagmatic activity may develop if the

dike tip approaches sufficiently close to the ice surface (Wilson & Head, 2002). More important

in these cases than the value of the dike tip pressure Pt is the difference between Pt and the

ambient stress in the surrounding ice, defined as Pd. If the stress in the ice is isotropic this stress

difference is equal to (Pt - g ρi [y - p]), and so using equation (3)

Pd = g [(ρr - β) z - (β - ρi) p] (5).

Note that this expression does not involve the ice layer thickness y: for a fixed reservoir depth z

the stress acting across the walls of the dike tip depends only on the distance that the dike

penetrates into the ice, p. Figure 3 shows the values of Pd for the same four magma water

contents used in Figure 2. Pd is in all cases negative: the pressure in the dike tip is less than the

external compressive stress. When the penetration is a minimum, Pd is about -3 to -4 MPa, the

exact value depending on the magma water content; at the maximum penetration the values

varies from about -7 MPa for 0.125% water to about -12.5 MPa for the negligible water content

case.

The timescale for the dyke emplacement process can now be obtained from the typical

magma rise speed, ~1 m s-1, quoted earlier. Penetration distances p range up to ~600 m and so

the time t required to reach this distance at 1 m s-1 is ~600 s, i.e. 10 minutes. In practice

somewhat longer would be required because the dike tip would decelerate to rest rather than


stopping abruptly. Temperature changes caused solely by thermal conduction would penetrate a

distance d of order (κ t)1/2 where κ is the thermal diffusivity of the ice or chilling dyke magma.

Thermal diffusivities of both ice and basalt are ~10-6 m2 s-1 and so even if t were, say, 20

minutes, d would be at most 3-4 cm. Thus minimal thermal exchange would occur by

conduction alone during the dike emplacement process.

However, the stress conditions shown in Figure 3 render it very likely that the ice around

the dike tip will initially fracture in tension or shear, with blocks of ice falling through the gas-

filled cavity onto the top of the magma column and initiating thermal and mechanical

interactions. If progressive collapse occurred to the extent that a pressure pathway to the surface

of the ice was formed, the excess water vapour pressure in the dyke tip would be vented to the

atmosphere and the consequent unloading of the magma would lead to further magma

vesiculation as both H2O and additional CO2 were exsolved, with the possible onset of explosive

activity reaching the surface of the ice. Such activity would almost certainly be

phreatomagmatic because of the intimate contact between magma, water and spalled blocks of

ice. The most extreme activity that could be expected would occur if intimate mechanical

mixing of ice blocks and magma took place so rapidly that no change in the total volume was

possible. This process would necessarily involve, on purely geometric grounds, the mixing of

equal volumes of ice and magma (and would therefore be slightly less extreme than some

interactions between surface lava flows and surface water, where the maximum energy release is

known to occur at a water:melt ratio of 3:7 (Wohletz, 1986)). The magma would cool from its

initial temperature to some equilibrium temperature, and the ice would melt to form a fluid

heated to the same equilibrium temperature. The fluid could be liquid water, water vapor, a

mixture of the two, or, if the equilibrium pressure and temperature were above the critical point,

a single-phase supercritical fluid.

When equal volumes V of ice and magma mix, they reach equilibrium by sharing the total

available thermal energy. If we neglect the mass of the water vapor in the dike tip cavity, the

magma in the uppermost part of the dike consists of silicate liquid and bubbles of CO2 gas at the

pressure Pt. We assume that the ice is initially at the triple point Ti (273.15 K), the magma is at


temperature Tm, and the final equilibrium temperature is Tf. Let the CO2 mass fraction in the

magma be nm. The solubility function for CO2 in mafic magma (Harris, 1981) can be

approximated as a function of pressure P by

n(P) = 3.4 × 10-6 + 6.0 × 10-12 P (6),

where the pressure is in Pascals and n is expressed as a mass fraction. Thus the mass fraction ne

exsolved at any tip pressure Pt is given by nm - n(Pt). The density of this gas is ρg where

ρg = (m Pt) / (Q Tm) (7).

Here m is the molecular mass of CO2, 44 kg kmol-1, and Q the universal gas constant, 8.314 kJ

kmol-1 K-1. The partial volumes of the CO2 gas and magmatic liquid are proportional to (ne / ρg)

and [(1 - ne) / ρm], respectively, and so the actual volume of gas is

Vg = {(ne ρm V) / [ne ρm + (1 - ne) ρg]} (8)

and of liquid is

Vm = {[(1 - ne) ρg V] / [ne ρm + (1 - ne) ρg]} (9).

The heat energy lost by the magmatic liquid is [Vm ρm sm (Tm - Tf)], where sm is the specific heat

of the magma, close to 900 J kg-1 K-1 averaged over the relevant temperatures using mineral data

compiled by Dobran (2001). The heat energy lost by the CO2 gas (which we assume then

dissolves into the water fluid) is [Vg ρg sg (Tm - Tf)], where sg is the specific heat of the CO2 gas,

~950 J kg-1 K-1 at the relevant temperatures (Kaye & Laby, 1995). This total heat lost by the

magma components can be equated to the energy required to heat the water fluid to Tf, equal to

[V ρi (Li + E(Tf, Pf)], where Li is the latent heat of fusion of the ice, 3.33 × 105 J kg-1, and E(Tf,

Pf) is the heat energy (enthalpy) of the resulting H2O fluid at the final equilibrium temperature

and pressure, resulting in


Li + E(Tf, Pf) = [ρm ρg (Tm - Tf) / ρi] {[ne sg + (1 - ne) sm] / [ne ρm + (1 - ne) ρg]} (10).

Enthalpy values as a function of pressure and of the temperature in excess of the triple point can

be obtained from standard tables (ASME, 2000), but there is a problem in that the final pressure

cannot be found from thermal arguments alone. A unique solution can be found by taking trial

values of the final pressure and using the densities of the water fluid phases, also taken from

tables (ASME, 2000), to find the total volume of fluid (liquid, vapour, liquid-vapour mixture, or

supercritical phase) produced, given that the total mass of water fluid derived from the melted ice

must be (V ρi). The fluid volume plus the magmatic liquid volume must, of course, sum to the

total available volume (2 V).

As in the analysis by Wilson & Head (2002), we assume an initial basaltic magma

temperature of 1473 K (i.e. 1200 ºC), and adopt a magma CO2 content nm = 0.002 (i.e. 0.2%).

Figure 4 then shows the values of the final equilibrium temperatures, Tf, and pressures, Pf, when

the pre-mixing dike tip pressures, Pt, are 1.2, 3.3 and 9.0 MPa, corresponding to the magma

water contents used in Figures 2 and 3, i.e. 0.125, 0.25 and 0.5%, respectively. The trend seen is

that for a high tip pressure, less CO2 will have exsolved from the magma, the volume fraction of

the magma occupied by low density (and hence low thermal capacity) CO2 gas bubbles will be

smaller, and hence the amount of heat per unit volume available from the magmatic liquid will

be larger. This will drive the melted ice to a higher entropy state, i.e. a higher temperature and

pressure. Tip pressures of 1.2 and 3.3 MPa result in a mixture of subcritical water and steam,

and the 9.0 MPa tip pressure leads to a very high-pressure supercritical fluid. Clearly, if the

magma CO2 content is greater than the value used in this illustration, more will be exsolved at

any given tip pressure, the gas volume fraction will be larger, the liquid magma volume fraction

(and hence heat per unit volume) will be less, and the final state of the water fluid will be less

extreme. Conversely, very low magma CO2 contents have the potential to lead to more extreme

temperatures and pressures after magma-ice mixing, with a greater potential for the explosive

generation of a pathway to the top of the ice layer. However, even if the above mixing process

did lead to phreatomagmatic eruptive activity immediately after dike injection into the ice, this


would probably not be long-lived: Figure 2 shows that even complete relaxation of the pressure

at the top of the magma column to atmospheric pressure would not cause magma to rise more

than about 30% of the distance to the surface of the ice, and so the magma at the top of the

column would rapidly be chilled, causing explosive activity to cease.

Whatever the details of the physical interaction between dike magma and surrounding ice,

the initially high rate of heat transfer will decrease with time. The absolute maximum amount of

ice that might eventually be melted by a given dike can be found by assuming that the magma

cools from its initial temperature, Tm, to the ice melting point, Ti, and that the ice just melts to

form water at temperature Ti. In practice some of the water fluid formed early in the heat

transfer process would have been much hotter than this, but convection in the fluid will

eventually transfer virtually all of the available heat from the fluid to the adjacent melting ice.

The total heat lost by the dike (per unit surface area of contact with the ice) is {W β [Lm + sm (Tm

- Ti)]}, where W is the dike width and Lm is the latent heat of fusion of the magma, ~2.09 × 105 J

kg-1, and the heat gained by the ice is (2 X ρi Li), where X is the total width of the ice layer

melted, half on each side of the dike. Equating the heat lost by the magma to that gained by the

ice using the above material property values yields (X/W) = 10.6. Consider one option from

Figure 2: a dike on average 2 m wide intruding 250 m into a 1000 m thick ice layer; intrusion of

the dike strains the adjacent ice elastically. A zone 10.6 m wide is now melted on each side of

the dike. Water at the ice melting point has a density of almost exactly 1000 kg m-3, and so the

total 21.2 m width of melted ice is converted to [(917/1000) × 21.2 =] 19.4 m width of water.

Thus 1.8 m, i.e. 90%, of the 2 m strain induced in the ice by the dike intrusion is relaxed. While

this is happening, cooling cracks in the thickening chilled dike margin allow water penetration,

and the lack of physical support by the surrounding water causes the dike material to lose

coherence and collapse into a breccia pile. In this example, conservation of volume would

suggest that if the breccia pile margins reached a 30º angle of rest, and if there were no pore

space and no volume changes due to chemical alteration, the pile would be ~17 m high and ~59

m wide at its base. In practice is would probaby be ~10-20% higher and wider than these values.

If the dike penetration distance were greater the breccia pile would be proportionally larger; the

500 m penetration of a 3 m wide dike would yield a pile at least 29 m high and 102 m wide.


2.3. Englacial dyke intrusion products

Unambiguous evidence for dikes intruding into glaciers is sparse. However, during the

1996 eruption of Gjálp, Iceland, Gudmundsson et al. (2004) described an en echelon fracture in >

500 m-thick ice overlying the erupting fissure, in a location previously free of crevasses. The

fracture was interpreted to be a consequence of the ice rupturing ahead of an englacial dyke.

Though there are no published examples of the lithofacies that might be created following dyke

intrusion into an ice sheet and its subsequent collapse, nevertheless the characteristics of those

lithofacies can be predicted (Wilson & Head, 2002, 2007) from a consideration of the likely

sequence of events involved (Figure 5). Initially after stalling, the dyke will be flanked on both

sides by a narrow but slowly expanding water-filled layer. Although all the heat available in a 2

m wide dyke is capable, ultimately, of melting up to 9.7 m of ice on either side of the dyke (i.e.

19.4 m in total), that is unlikely to happen since the water cannot support the chilled and

fractured, mechanically weak (thus gravitationally unstable) dyke. The dyke will quickly

collapse by slumping under its own gravitational weight, displacing meltwater and filling the

space available with coarse chilled (glassy) dyke rubble up to a lower elevation than that reached

by the dyke tip when it stalled.

Magma will continue to extrude into the narrow overlying water-filled englacial space

(“vault”), probably as pillow lava. With its coarse pore spaces, some of the dyke rubble will also

be intruded by magma in the form of subsurface breakouts, particularly immediately after

collapse when the original dyke pathway is disrupted. The pillow lava will spread laterally until

confined by the vault ice walls, but continued melt-back of the walls, there and at lower

elevations, will remove support from the tall rubble pile, causing further collapses and repeated

lowering of what is now an unusual type of early-stage volcanic edifice. Collapses will continue

until the pile rests on the underlying bedrock and has stably graded flanks, thus finally forming a

cone or ridge that, for an initial dyke 2 m wide, should be at least 59 m wide and 17 m high,

although probably significantly larger due to dilatancy (see above).


The lithofacies forming the cone or ridge is likely to be quite distinctive, comprising a

heterogeneous pile of coarse, essentially fines-free glassy breccia, fragments of breccia cemented

by chilled dyke material, and numerous fragmented and intact lava pillows. It will also likely

show crude radially outward-dipping bedding and possible grading as a consequence of multiple

avalanching events of the coarse debris (as grain flows) down the cone flanks. Stratification will

probably only affect the latest stages of growth of the breccia pile, when the cone has formed a

free upper surface. Further eruption will result in the basal breccia cone/ridge being draped and

buried by pillow lava, thus forming a small and probably inconspicuous core to the subsequent

pillow volcano.

Whilst distinctive, the recognition in the field of such progressive dyke-collapse

lithofacies might prove difficult since it relies on the exposure of a section eroded fortuitously

directly into the former vent region. Although there are no published descriptions of dyke

collapse-related lithofacies interpreted in this way, a basal facies observed in a Pliocene

glaciovolcanic sequence at Mussorgsky Peaks, Alexander Island (Antarctic Peninsula) might be

an example. There, a poorly stratified outcrop of intrusive “pods” and broken and intact pillow

lavas is mingled with up to 40 % fines-free coarse glassy breccia (Smellie and Hole, 1997;

Figure 6). However, the outcrop is isolated and the field relationships are unclear, although the

lithological characteristics are plausible. In another possible example, Wright (1978) described

unusually highly fractured massive basalt surrounded by a pile of basalt talus in the Royal

Society Range, Antarctica. Because of its situation, she interpreted the outcrop speculatively as

the remnants of an englacial dyke, but again the field relationships are inconclusive.

After the initial interactions described above, and depending on the duration and

discharge volume of the eruption, the subglacial volcano might evolve up into a subaqueous tuff

cone or ridge (tindar; formed of stratified phreatomagmatic tuffs) and ultimately into laterally

prograding lava-fed delta(s), which together comprise a tuya volcano (Mathews, 1947; Smellie,

2000). There are numerous descriptions of the lithofacies and construction of glaciovolcanic

basaltic pillow volcanoes, tindars and tuyas (e.g. Jones, 1969, 1970; Skilling, 1994, 2002;


Smellie & Hole, 1997; Werner and Schmincke, 1999; Smellie, 2000, 2006; Schopka et al.,

2006), and they need not be repeated here.

3. Magma intrusion at the base of an ice layer

3.1. Theoretical issues

The treatment in Section 2 of a dike passing through an ice-rock interface showed that the

pressure in the dike tip region would commonly be less than the external lithostatic load of the

overlying ice. This is possible because the ability of a dike to remain open at any point along its

length is determined by the integrated effect of the stress difference across the dike wall, not the

local values of the internal pressure and external load (Pollard, 1987). Locally the outward stress

difference across the wall can be negative as long as the appropriately weighted average along

the extent of the dike is positive. If a dike reaches an ice-rock interface and feeds a sill

controlled by the elastic properties of the host materials, this criterion must now be applied to the

sill, and clearly the net outward stress across the sill "walls", i.e. the upper and lower surfaces of

the sill, cannot be positive unless the magma pressure at the sill-dike contact is greater than the

stress due to the overlying load of the ice. This sets an additional requirement that can be

evaluated as follows.

When magma rises through a dike to feed a sill, there are finite pressure gradients in both

the dike magma and the sill magma. In general, each pressure gradient consists of two

components: the pressure gradient required to support the weight of the magma, and the pressure

gradient required to overcome the wall friction opposing magma flow. However, if the sill is

intruded near-horizontally, there is a negligible pressure gradient related to the magma weight in

the sill. Furthermore, if the magma flow rate is small, as it will be near the end of the intrusion

process, the pressure gradients due to magma flow will also be small. Under these conditions the

pressure Psi in the magma in the dike at the sill inlet located at the level of the ice-rock interface

will be given by the absolute pressure in the magma reservoir, Pr, minus the pressure due to the

weight of magma in the dike between the reservoir and the sill inlet, Pw. The minimum

requirement for sill injection to occur is that Psi must exceed the weight of the overlying glacial

ice, Pg (cf. Smellie, 2008). Pr is equal to [Pa + g (ρi y + ρr z) + Pe] where, in anticipation of its


importance, we introduce a finite pressure Pe in the reservoir magma in excess of the lithostatic

load Pl. Pw is equal to (g βw z) where βw is the mean density of the magma averaged between the

reservoir and the sill inlet, and Pg is equal to (Pa + g ρi y). The requirement Psi > Pg is therefore

[Pa + g (ρi y + ρr z) + Pe] - (g βw z) > (Pa + g ρi y) (11)

which reduces to

Pe > g z (βw - ρr) (12).

Thus the required value of the excess magma pressure Pe increases with the depth of the

reservoir z and depends on the difference between the densities of the dike magma, βw, and the

crustal rocks, ρr. As noted above, a constant value of ρr = 2300 kg m-3 has been adopted here.

Care is needed in evaluating the magma density βw. This will be a function of the total CO2

content, again assumed constant and equal to 0.2% for illustration, but also of the absolute

pressures in the reservoir and at the sill inlet. We have assumed that the sill inlet pressure is

close to the glacial ice overburden pressure Pg, so this is only a function of the ice thickness, but

the reservoir pressure depends on both the lithostatic load at the top of the reservoir and the

excess magma pressure Pe. Thus βw cannot be evaluated accurately unless Pe is known, but Pe

cannot be found unless βw is known. This problem is resolved by initially assuming Pe is zero,

evaluating a first approximation to βw using equations 5(a) and 5(b) in Wilson & Head (2002),

using this value to find a first approximation to Pe, updating the total reservoir pressure, and

calculating a second approximation to βw which then gives a second approximation to Pe. This

recursive procedure converges to better than 1% accuracy after only two iterations and yields the

minimum excess reservoir pressures shown in Figure 7. Values range from a few MPa for

shallow reservoirs to approaching 10 MPa for deeper ones. The larger of these pressures are

greater than the ~3-5 MPa values deduced for the summit reservoir of Kilauea volcano, Hawai'i

(Parfitt, 1991), but they could readily be generated if a reservoir were fed by dikes joining it to a

deeper magma source located in denser host rocks. Thus, if magma with density ρm = 2700 kg

m-3 is generated by partial melting of host rocks in the upper mantle with density 3300 kg m-3, an


excess pressure of 10 MPa will be present in the magma at the top of the zone of partial melting

provided its vertical extent is only ~1.7 km, whereas it is generally accepted that partial melting

occurs over much more vertically extensive regions (Turcotte & Schubert, 2002).

If the conditions in a magma reservoir feeding a dike reaching an ice-rock interface are

indeed such as to allow a sill-like intrusion to start to grow, a range of possible consequences

exists (Wilson & Head, 2002, 2007). Possible early interactions are the same as for dikes

penetrating upward into ice: rapid generation of highly pressurized water fluid as ice is melted

with no opportunity for a significant volume change. The fact that only one face of the planar

magma body is in contact with ice instead of two is probably of little consequence. Figure 4

shows that significant transient pressures could be generated if ice and magma are intimately

mixed. As long as these local pressures do not initiate fractures that extend to any of the ice

margins, they will be accommodated by deformation of the bulk of the ice (and the underlying

rock) on a time scale corresponding to the propagation of deformation (i.e. seismic) waves in the

solids. For ice thickness or lateral extents of order 1 km and wave speeds in solids of order 1 km

s-1 the time scale will be seconds. While these high pressures, which Figure 4 shows can be from

a few to more than 10 times the stable dike tip pressures, are present, they will in theory act to

reduce the pressure gradient driving the magma up the feeder dike. However, the speed of sound

in vesicular magma is generally at least an order of magnitude less than that in pure liquids

(Kieffer, 1977) and so the effects of the pressure transients will propagate a much smaller

distance through the dike magma than through the host ice and rock; the net effect on the

subsequent transfer of magma through the system will be negligible. Thus the major potential

importance of any initial violent interactions is the possibility that they may create fractures to

the surface. However, as discussed above for englacial dikes, even if they do so the system may

heal itself if the ice layer is very much thicker than the thickness of the sill.

After this short-lived transient stage has established the geometry of the dyke-sill

connection, the system will evolve in a number of possible ways. If the glacier is cold-based, the

ice-rock interface will be strong enough to support tensile stresses and the intruding sill will

propagate in the same way that a sill grows at a rock-rock interface, with both the overlying (ice)


and underlying (rock) materials initially behaving as elastic solids. The eruption may cease

while the system still has this configuration, but heat will continue to be transferred from magma

to ice, and the H2O volume decrease due to ice melting will at least partially relax any excess

pressure in the magma. If the glacier is warm-based the interface between the glacier and the

underlying rock will not support tensile stresses. The requirement that the magma pressure at the

dyke-sill connection exceeds the weight of the glacier means that there is the potential for the

glacial ice to be lifted by the magma. However, this can only occur locally: ice does not have an

infinitely large shear modulus, and an extensive slab of ice (the glacier in this case) supported

over a very small region within it (the dyke-sill connection) will bend, remaining in contact with

the underlying rock beyond some finite distance from the supported region. As a result, in the

early stages of the intrusion the resulting stress in the ice will interact with the magma pressure

in very much the same way as in the cold-based case, and the pressure in the intruding magma

will be equal to or greater than the weight of the overlying ice. However, as the magma in the

intrusion spreads far enough to approach the edge of the glacier, a stage will be reached when the

glacial margin is lifted, providing a pressure connection to the atmosphere and an escape route

for water, water vapour and exsolved volcanic gases. Even in cold-based glaciers this condition

must eventually be reached as the elastically-propagating sill tip approaches the ice margin.

Water loss is likely to happen faster than the overlying ice can deform plastically (Tuffen, 2007),

especially if the water release is on the scale of a jökulhlaup (e.g., Björnsson, 1992), so unless

wholesale disintegration of the glacier occurs, air will replace the water overlying the magma

and the pressure acting on the magma will be much closer to atmospheric pressure than to the

weight of the overlying glacier. Further volatile exsolution will occur in all parts of the magma

except its chilled margins, and the scenario will change from a sill intrusion to a sourceward-

propagating explosive eruption as volatiles are catastrophically released by the suddenly reduced

overlying pressures. In extreme cases enough volatiles may be released, especially from the

least-cooled magma at the point where the feeder dike reaches the rock-ice interface, that a lava

fountain forms over what becomes a subglacial fissure eruption. A sufficiently high fountain

will radiate heat to the overlying ice (indeed, pyroclasts may even come into physical contact

with it), thus greatly enhancing ice melting in this region. We now elaborate on each of these

three stages.


3.2. Elastic intrusion

We argued above that the pressure Psi in a sill intruded beneath an ice layer must be at least

equal to the pressure due to the weight of the overlying ice (see also Smellie, 2008). We also

argued in Section 2 that the pressure in a propagating dike tip would probably not decrease

below the saturation pressure corresponding to the water content of the magma. These criteria

can be used to explore the vesicularities, caused by the presence of exsolved CO2 bubbles, of

magmas intruded as sills. Equation (6) gives the solubility of CO2 as a function of pressure, and

so for any assumed magma CO2 content the amount ne of gas exsolved at any given pressure can

be found and eq. (8) can be used to find the volume fraction (Vg / V) of the magma that it

represents. Figure 8 shows examples of bubble volume fraction (left axis, solid curves) and

magma pressures Psi (right axis, broken line) for ice thicknesses up to 2000 m for the case where

the pressure is just equal to the weight of the overlying ice. Gas volume fractions are given for

total magma CO2 contents of 0.1 and 0.2 %. The horizontal solid lines mark the generally

accepted range (e.g., Sparks, 1978; Jaupart & Vergniolle, 1989) of magma gas volume fractions,

~0.7-0.85, over which magmatic foams become unstable and disintegrate into pyroclasts. Thus

for ice layers less than ~100 m thick we might expect that, rather than a vesicular liquid magma,

what would be intruded beneath an ice layer would be a mixture of gas and pyroclasts - in

essence a subglacial pyroclastic density current. Note, however, that this would only be true if

the magma water content were very low, because otherwise the buffering effect of water

exsolved at the dike tip would require the magma pressure to be significantly higher than the ice

weight: we saw in Section 1 that magma water contents of 0.125, 0.25 and 0.5% correspond to

dike tip buffering pressure, and hence minimum sill inlet pressure, of 1.2, 3.3 and 9.0 MPa,

respectively. Thus the magma water content would need to be less than about 0.1% for this

scenario of a subglacial pyroclastic eruption to be possible. Needless to say, if such an event did

occur, violent mixing of mobile pyroclasts and water produced by ice melting would take place.

The clasts produced by the interaction might be distinguished from those from the collapse of an

englacial dike by their greater degree of fragmentation. However, it should also be noted that if

a water-rich magma is intruded under a thin layer of ice, the magma pressure, say up to 10 MPa,

may be very much greater than the ice weight, which is ~0.5 MPa for a 50 m ice thickness, and


so fracturing of the ice is very likely to occur. The creation of pathways to the atmosphere will

then guarantee magma fragmentation and some kind of phreatomagmatic eruption extending

deposits out onto the ice surface.

Clearly a large range of combinations of magma reservoir pressure and magma water and

CO2 contents are consistent with the injection of vesicular magmatic liquids at ice-rock

interfaces. Various aspects of the resulting growth and evolution of the system have been

explored by Höskuldsson & Sparks (1997), Wilson & Head (2002, 2007) and Tuffen (2007).

Wilson & Head (2002) showed quantitatively that major morphological differences are expected

between subaerial lava flows and elastically-emplaced subglacial intrusions. The most obvious

of these is the relative reduction of overall volatile exsolution due to the high ambient pressure,

in a manner analogous to deep submarine eruptions (Head & Wilson, 2003). Other differences

are more subtle. A subaerial flow has an upper free surface and acquires a cross-sectional shape

(channel-and-levee, sheet-like, compound pahoehoe flow field) dictated by the slope on which it

is emplaced, its volume effusion rate, and its evolving rheological properties. The absence of a

free surface at the top of a subglacial magma body together with the stress applied by the elastic

properties of the ice above and rock beneath will lead to it spreading into a sheet-like structure

likely to be initially much wider, thinner, and more slowly advancing than a subaerial lava flow

erupted at the same volume flux. Other differences between subglacial and subaerial behaviour

are similarly driven by the stress field: thus, a subglacial intrusion will get thicker as its edges

advance, this process offsetting the inward migration of the effects of cooling at its upper and

lower faces and allowing it to advance to a much greater distance from the vent than a subaerial

flow before cooling limits its growth. The predicted thickening process has some similarities to

the inflation of subaerial lavas (Hon et al., 1994) but, whereas after their margins come to rest as

a result of cooling subaerial flow units inflate and thicken, subglacial intrusions thicken

continuously as their edges advance.

Heat transfer rates from subglacial magma to melted ice have been predicted by

Höskuldsson & Sparks (1997), Wilson & Head (2002, 2007) and Tuffen (2007), and measured in

a few well-studied cases (e.g. Jarosch et al., 2008). The theoretical calculations differ somewhat


in the values used for material constants such as the specific heat of the magma, which has to be

averaged over the temperature ranges involved, and the magma density, a function of

composition and volatile content. Wilson & Head (2007) found that the volume of water that

can be produced by ice melting is ~6.5 times as much as the intruded magma volume, to be

contrasted with the volume ratio of ~10.6 if all of the magmatic heat is transferred to the water.

Thus a 1 m thick sill would generate a ~6.5 m deep water layer from the melting of a [(1000/917)

× 6.5 =] ~7.1 m thick layer of ice, the volume change thus accommodating (7.1 - 6.5 =) ~0.6 m

of the 1 m sill thickness. This implies that perhaps as much as half of the excess pressure in the

magma at the time of its intrusion might be relaxed during the ice-melting period. This pressure

relaxation could in principle cause some additional CO2 exsolution and increased vesiculation in

the magma, but given that the magma is progressively cooling while the ice is melting, any

changes would be minor and could only take place near the center of the intrusion where it was

hottest. As suggested by Dixon et al. (2002), the fact that, as long as they do not make pressure

contact with the atmosphere, subglacial intrusions are expected to exsolve a significant fraction

of their pre-eruption CO2 and at most a minor fraction of their pre-eruption H2O suggests a

potential diagnostic tool, in addition to purely morphological examination, for distinguishing

such intrusions from subaerial lavas after all of the overlying ice has been removed by climate

changes.

3.3. Subglacial eruption after atmospheric connection

When the margin of a growing subglacial sill approaches sufficiently close to the edge of

an ice sheet that a significant connection to the atmosphere is made, the first consequence will be

the escape of the pressurized water vapor in the sill tip. This will be followed by progressive

water leakage and replacement by atmospheric air, almost certainly leading to a wave of

decompression advancing into the sill from its distal margin towards the inlet from the feeder

dike. This can only be avoided completely if the ice overlying the sill can deform fast enough to

replace the water and stay in close proximity to the top of the sill magma. For conditions similar

to those proposed here, Höskuldsson & Sparks (1997) calculated an ice deformation rate of ~1

mm s-1, so if the rate of thinning of the water layer exceeds this value, some pressure decrease

will inevitably occur.


A pressure reduction in the sill will cause an increase in the pressure difference between

the magma reservoir and the top of the dyke, and hence an increase in the magma flow rate into

the sill. Typical values found earlier of likely sill pressures (less than 10 MPa) and reservoir

pressures (several tens of MPa) suggest that even if the sill pressure were reduced all the way to

atmospheric pressure (0.1 MPa) the flow rate increase would be modest - at most of order 20%.

However, reduction of the pressure on the upper sill surface from several MPa toward

atmospheric pressure will have dramatic consequences.

Initially, the magma will remain a vesicular foam: existing carbon dioxide bubbles will

expand and new bubbles of both CO2 and H2O will form at a rate which may allow the magma to

stay in physical contact with the overlying ice. If so, this will lead to continuing water

production, and the system will tend toward an equilibrium where the pressure in the water is

greatest near the dyke and least at the edge of the ice sheet, with the resulting pressure gradient

driving the water toward the exit. However, as water drainage becomes more efficient, it seems

inevitable that the pressure at the sill–ice contact will eventually decrease to atmospheric

pressure. The total magma volatile content then determines whether the volume fraction of gas

bubbles in the magma becomes so great that magma fragmentation occurs. For a magma with

the 0.2 mass % CO2 content used earlier, and a fragmentation threshold of 80% by volume gas

bubbles, magma fragmentation would begin at about 0.6 MPa, 1.0 MPa, and 1.9 MPa, i.e. 6, 10

and 19 bars, for magma water contents of 0.125%, 0.25% and 0.5%, respectively. We therefore

infer that such fragmentation will be common.

Magma fragmentation will begin at the margin of the sill closest to the connection to the

atmosphere. As the pressure in the space above the chilled magma crust decreases, the crust may

temporarily prevent any response from the underlying magma. However, the crust is likely to be

pervaded by cooling cracks and have little strength. When the crust fails, an expansion wave

will propagate down into the sill at a large fraction of the local speed of sound which, in a

vesicular liquid, will be ~100 m s-1 (Kieffer 1977; Wilson & Head 1981). Thus for a sill a few

meters thick this time scale will be less than a tenth of a second. The expansion wave will


fragment the magma, and expansion of the released gas will accelerate disrupted magma clots to

impact the overlying ice. Formulae given by Wilson (1980) for transient explosions show that,

even if the magma contains no water, a pressure reduction from several MPa to 0.1 MPa in a

magma with ~0.2 mass % of CO2 will generate speeds in the hot vesiculated pyroclasts up to ~50

m s-1. The impact of these hot magma clots on the overlying ice will locally enhance the ice

melting; sufficiently violent interaction of magma and ice at this time may be again enough to

trigger a sustained, violent fuel-coolant type of interaction (Wohletz & McQueen 1984;

Zimanowski et al. 1991). The products of the explosive mixing would be propelled toward the

exit to the atmosphere as the wave of magma vesiculation and fragmentation propagated back

toward the feeder dyke. The propagation speed of the interaction would be a balance between

the speed of the wave into the as-yet unaffected sill magma (again some large fraction of the

~100 m s-1 local speed of sound) and the speed at which water and fragmented magma could be

discharged from under the ice. This explosive fragmentation process is an excellent candidate

for the origin of sudden jökulhlaup production. Some fraction of the fragmented magma would

be washed out with the escaping water and the rest would be left behind to form a vitroclastic

(phreatomagmatic) deposit.

When the wave of magma disruption reaches the feeder dike, the flow rate in the dike will

increase somewhat as it adjusts to the fact that pressure at the dike outlet has become very close

to atmospheric, and the system will behave in the same way as if the eruption had started

subaerially, with the formation of a chain of lava fountains. Depending on the efficiency with

which water is being drained from the vicinity of what is now the subglacial vent, the lava

fountains may be entirely magmatic or may be phreatomagmatic. Initially these fountains will

"drill" into the overlying ice. A cavity will grow upward until the fountains reach the subaerial

height corresponding to the total magma volatile content (Head & Wilson 1987). Thereafter heat

will only be transferred to the ice by radiation from pyroclasts in the fountains, and magma clots

falling from the fountains will begin to form rootless lava flows (Head & Wilson 1989)

spreading away from the vent over the top of the disrupted sill material. A new balance will

eventually be reached between ice subsidence and ice melting, and if the eruption continues for

long enough the explosive activity at the vent may eventually emerge through the ice. A


complication to this picture is the possibility that the ice layer above the fragmented sill residue

might fracture in a brittle fashion and collapse rather than slowly deforming plastically. The

pressure at the dike vent would still be close to atmospheric provided there was a reasonably

high porosity and permeability in the collapsed ice block pile, but interactions between magma

and ice would be more vigorous because of their greater proximity.

A hybrid situation can be envisaged in cases where the magma volatile content is very low.

If water drainage is inefficient, the pressure reduction rate in the water above the sill will be less

dramatic, but the pressure acting on the upper sill surface will still be greatly reduced. The distal

part of the still-spreading magma body may then begin to evolve into a thicker, narrower

morphology more like that expected for a subaerial lava flow. Melt-water will be channelled

along the side(s) of the flow unit(s) that evolve, and the system will remain stable as long as the

pressure in the water above the flow(s) is maintained high enough to suppress magma

vesiculation to the point of fragmentation. One or more discrete lava flow units may eventually

emerge from beneath the ice. After ice removal by climate change, it should be possible to

identify this kind of flow by its changing shape as a function of distance from its vent, and to

distinguish it from a flow generated by lava fountain activity in an ice cavern after sill disruption.

3.4. Discussion

Eruptions under thin ice, in which magmatic vesiculation is not suppressed, will be

explosive and result in the abrupt removal of overlying ice and construction of pyroclastic cones.

The hypothetical generation of extreme conditions such as formation of a pyroclastic density

current (section 3.3) and its emplacement by intrusion beneath an ice cover is clearly unlikely in

these cases. Once magma fragmentation is achieved by high levels of vesiculation, irreversible

and violent fuel-coolant interaction with meltwater will take place resulting in an abrupt

transition to an explosive phreatomagmatic eruption. This will form stratified tuff cones or tuff

rings composed of interbedded fall and pyroclastic density current tephra. No examples have

been observed, possibly because the edifice, formed of unconsolidated tephra, is easily eroded by

ice. Additionally, some of the edifice will be formed on the ice itself and will be destroyed as

the ice flows down slope and melts (Figures 9 and 10). The pyroclastic cones thus have a very


low preservation potential. However, distinctive glaciovolcanic sequences, known as sheet-like

sequences of Mount Pinafore type, are believed to be the products of eruptions under thin ice

(probably < c. 150-200 m thick; Smellie et al., 1993; Smellie & Skilling, 1994; Loughlin, 2002;

Smellie, 2008). They are outflow deposits (i.e. they accumulated away from the source edifice;

Figure 11) and are characterised by prominent basal beds of fluvially deposited stratified

phreatomagmatic tephra, for which an explosive phreatomagmatic source has been inferred. The

sequences are typically composed of (from base up) glacial diamict, phreatomagmatic tuffs,

hyaloclastite, water-cooled lava and, finally, subaerial lava (Figure 12a). By contrast, if

meltwater drainage is particularly efficient beneath a thin ice cover (e.g. eruptions on relatively

steep bedrock), the vent might dry out and result in magmatic eruptions of Strombolian or

Hawaiian type. This was observed during the 1969 subglacial fissure eruption of Deception

Island, Antarctica, during which a line of cinder cones was generated along a fissure in ice c. 100

m thick (Smellie, 2002; Figure 9).

Eruptions under much thicker ice might have quite different consequences for the

lithofacies formed. Unusual lava—hyaloclastite sequences of Plio-Pleistocene age in southern

Iceland, which were ascribed to lava extrusion in a shallow-marine (shelf) setting (Bergh &

Sigvaldason, 1991), were reinterpreted as products of multiple subglacial sill intrusion and

associated meltwater floods (jökulhlaups; Smellie, 2008; Figure 11b). The Icelandic outcrops,

which were named subglacial sheet-like sequences of Dalsheidi type, are products of voluminous

subglacial fissure eruptions individually up to > 30 km3. The products can be traced up to 30 km

along strike and at least 14 km down-dip and are separated by sharp, undulating, likely glacial

erosion surfaces. Each sequence is formed from the products of a single eruption. A “standard

sequence” has been identified, formed from an association of four major lithofacies, i.e.

diamictite, lava (sill), hyaloclastite and mudstone (from base up; Figure 12b). Although broadly

resembling sheet-like sequences of Mount Pinafore type, the Dalsheidi-type sequences are wider

and usually much thicker (up to 300 m), and they lack basal explosively generated

phreatomagmatic tuffs and subaerial capping lava, all of which are environmentally significant

(contrast Figures 12a and 12b).


In Dalshieldi-type sequences, diamictite occurs above a basal unconformity. It probably

represents a combination of tillite and melted-out glacier bedload, and is succeeded by laterally

extensive sheet lava, called an interface sill by Smellie (2008), showing evidence for basal

loading and peperitic interaction with the diamictite, which must have been relatively soft and

unconsolidated at the time. The sill shows spectacular water-induced columnar cooling joints of

colonnade and (particularly) entablature type. Its upper surface is locally deformed into

structures resembling flow folds, some planed off by coeval erosion, but it is also conspicuously

characterised by prominent apophyses, usually a few tens of meters long, that intrude overlying

massive to faintly planar stratified hyaloclastite. Thick hyaloclastite dominates each eruptive

unit. It is monomict, relatively fine grained (mainly coarse sand to granule grade) and fine-ash

poor/free, with dispersed broken and intact pillows that increase in proportion up-dip toward the

source. The glassy upper surface of the sills has been locally stripped off prior to deposition of

the hyaloclastite. Features of the hyaloclastite indicate that they are mainly hyperconcentrated

flood-flow deposits transported in major meltwater floods (jökulhlaups) towards the end of each

eruptive period. The sequences are capped by thinly stratified mudstone and fine sandstone,

representing finer-grained hyaloclastite detritus deposited by a waning flood. Some of the finer-

grained capping sediments are formed from explosively generated lapilli tuff indicating that

some eruptions were explosive (phreatomagmatic) in the final stages.

A genetic model for the emplacement of the Dalsheidi-type sequences was suggested by

Smellie (2008), comprising: (1) sill emplacement at the base of an ice sheet; (2) sill stagnation

caused by eruptive overpressures becoming insufficient to lift the overlying ice cover; (3)

transformation to pillow lava effusion at the erupting fissure, with construction of a pillow ridge

and simultaneous creation of a meltwater-filled vault; and (4) floating of the overlying ice and

release of meltwater in a major jökulhlaup that destabilised and destroyed much of the pre-

existing pillow edifice, redepositing it as thick hyaloclastite beds; (5) the high-energy flood event

locally eroded the surface of the sill. Finally, (6) finer-grained beds were laid down as the flood

waned. The final stage was also associated with intrusion of apophyses of magma derived from

the still-molten sill interior, which were injected up into the sluggishly moving flow and

associated hyaloclastite pile.


Similar to eruptions of sheet-like sequences of Mount Pinafore type, the source edifices

responsible for Dalsheidi-type eruptions have never been observed. They are thought to have

been pillow ridges that were extensively removed during the late-stage jökulhlaups. Since

explosive activity is absent or else confined to the final stages of eruptions, the sills and pillows

are believed to have been undegassed during emplacement due to volatile exsolution being

suppressed by high ambient pressures associated with unusually thick ice sheet conditions

(empirically calculated to be at least 1000 m). An entirely subglacial setting is also suggested by

the apparent absence of any capping subaerial lavas, indicating that the overlying ice was never

completely melted through.

It was suggested in section 3.3 that a combination of thicker-ice conditions (initially

suppressing significant vesiculation) and decompression caused by sudden late stage pressure

reduction, e.g. as a subglacial sill nears the margins of an ice sheet and connects with

atmospheric or near-atmospheric pressures, might cause major changes within the sill. Many

variations on this theme appear to be possible. Thus Höskuldsson et al. (2006) described basaltic

pillows inferred to have been erupted under 1.5-2.0 km of ice that have outer zones with 15-20%

vesicularity surrounding cores with 40-60% vesicularity, interpreted to be the result of an ~4.5

MPa pressure decrease as a jokulhlaup abruptly removed much of the water that had been

produced by ice melting. In contrast Schopka et al. (2006) described a basaltic "hyaloclatite"

ridge formed under ~500 m of ice where the activity was explosive throughout; pressure

conditions were maintained low enough for magma fragmentation by continuous drainage of

melt water at a high enough rate that the overlying ice could not deform fast enough to maintain

contact with the erupting magma. Finally, Tuffen (2007) has stressed the fine balance that may

exist between magma intrusion rate, water production and drainage rate, and ice deformation

rate. Tuffen et al. (2008) described a rhyolitic eruption that took place beneath ~150 m of ice,

beginning as an explosive event but changing to the intrusion of vesicular magma into the

fragmental deposits from the previous explosive phase. Clearly a low-pressure pathway to the

atmosphere never formed in this case.


The most extreme version of late-stage pressure reduction is proposed (section 3.3) to be

the explosive decompression of an intruded sill. The loci of the explosions will migrate rapidly

back toward source (the vent), generating subglacial pyroclastic density currents that will

simultaneously mix with the abundant ambient meltwater and rapidly transform into

hyperconcentrated flood-flows, resulting in a major jökulhlaup where the floods exit from the ice

sheet. There are currently no known examples of those deposits, although they might easily have

been missed. The deposits will probably be massive to faintly stratified and largely formed of

angular highly vesicular hyaloclasts in abundant fine ash matrix, and there will be evidence for

lateral (down-dip) and vertical transitions to fluvial deposits as floods wane and normal stream-

flow conditions recur. Because they are generated from the destruction of the original sill, they

will not overlie any coeval sill rock, which is a major distinction from “standard” Dalsheidi-type

sheet-like sequences (as is their formation, composed of phreatomagmatic lapilli tuffs, not

mechanically-generated hyaloclastite (sensu White & Houghton, 2006)). Deposits that failed to

escape from the ice will rest on a glacially eroded surface and/or glacial diamict, whereas

deposits of the proglacial jökulhlaup will lie on outwash. This relationship raises the possibility

of using these features to identify the geographical limits of past ice sheets.

4. Supra-glacial eruptions

Eruptions onto glacial ice may involve either the advance onto the ice of a lava flow from

a vent located off the glacier or the deposition onto the ice of pyroclasts from an explosive

eruption (e.g., Wilson & Head, 2008). The vent for the latter may be located off the glacier or

may lie within the glacier in cases where a dyke penetrates into, or at least fractures, the ice

leading to a phreatomagmatic eruption. To advance onto a glacier, an encroaching lava flow

must be thicker than the glacier and, given the likely range of glacial ice thicknesses and lava

flow unit thicknesses, such events are probably rare, though lava flow advance over thin snow

and ice deposits is common. Where a thinner lava encounters a thicker glacier surface, it will

“pond” by thickening against the ice barrier; examples are common, both for basalts and more

evolved lava flows on Earth (e.g. Lescinsky & Sisson, 1998; Lescinsky & Fink, 2000; Mee et al.,

2006; Stevenson et al., 2006; Harder & Russell, 2007) and Mars (Shean et al., 2005; Kadish et


al., 2008). Another way is for lava to spill over onto a glacier from an adjacent ice-free

topographic high, although subsequent shear through flow of the glacier will separate the

bedrock- and glacier-covering lava outcrops as the glacier moves down-valley.

4.1. Lava flowing onto glacial ice

Wilson & Head (2007) treated the advance of a lava flow of a given thickness over ice in

a region where the ambient temperature is slightly below the ice melting temperature. Their

treatment, and that followed here, ignores the presence and likely insulating effects of a basal

autobreccia layer, so the calculations of extent of ice melting (below) should be regarded as

optima. The entire core of the flow is assumed to be initially at the lava eruption temperature

and the upper lava surface is assumed to rapidly cool to the ambient temperature. Thus the ice

melting rates calculated will be maxima, corresponding to a flow encountering ice almost

immediately after leaving the vent. The base of the flow melts ice to water at the triple point

temperature 273.15 K and it is assumed that the water produced can drain efficiently. In that

case the temperature of both the water and the base of the flow remain at 273.15 K. The

evolving temperature profile within the flow can then be evaluated as a function of time using

the series expansion solution given by Carslaw and Jaeger (1959; Article 3.4, equation (1)). The

temperature profile can be differentiated to yield the temperature gradient and this, multiplied by

the thermal conductivity of the lava, provides the rate of heat transfer to the ice. If the heat

transfer rate is divided by the ice density and latent heat of fusion, the result is the rate decrease

of ice thickness beneath the flow, and hence the water production rate per unit area of the flow

base. Figure 13 shows examples from Wilson & Head (2007) of the ice melting rate as a

function of time after the start of melting at any given location for flows 1, 3, 10 and 30 m thick.

Integration of the water production rate over the growing area of the base of an advancing

flow yields the total water production rate as a function of time. After one day, the absolute

thicknesses of ice melted by flows 1, 3, 10 and 30 m thick are respectively 4.95, 4.65, 4.0 and 1.8

m, corresponding to 99%, 31%, 8% and 1.2% of the total melting that is ultimately achievable.

The maximum value of the ratio (total ice thickness melted)/(lava flow thickness) is predicted to

be independent of flow thickness and very close to 5 (Wilson & Head, 2007). The overall time


scale for melting depends very strongly in the thickness of the flow: for flows 1, 3, 10 and 30 m

thick, 90% melting is achieved after about 20 hours, one week, 3 months and 3 years,

respectively. These lengthy delays will undoubtedly be part of the explanation for the

observations (e.g. Einarsson, 1948; Kjartansson, 1948) that advancing flows cause negligible

amounts of snow melting, apparently at odds with the above predictions. The other factors are

likely to be inefficient water drainage from beneath irregular flow bases and the fact that the heat

content of the flow available to cause melting will decrease steadily with increasing distance

from the vent.

4.2. Pyroclasts emplaced onto glacial ice

When a layer of pyroclasts accumulates onto a glacier, there are many possible outcomes

(e.g., Wilson & Head, 2008). If the vent is nearby and the activity is not too energetic, e.g. as in

a mafic lava fountain eruption, it is possible that many of the larger clasts will still be very hot on

landing: Head & Wilson (1989) showed the trends of how the mean deposit temperature will

vary with magma volatile content and discharge rate. In these circumstances rapid melting of ice

may occur, but the consequences will be very localized. Of more wide-ranging importance will

be more energetic (i.e. high mass flux) but relatively volatile-poor eruptions leading to the

emplacement of hot pyroclastic density currents onto glaciers, where a sufficient thickness of

deposit may trigger a vigorous interaction ultimately forming a lahar (Walder, 2000a).

In contrast, a relatively high mass flux, high volatile-content eruption (e.g. a rhyolitic

plinian eruption) will generate an eruption plume from which the bulk of the clasts will have

fallen from a great enough height that they will land at the ambient atmospheric temperature. If

the glacier onto which they fall is stable (i.e. not in the process of melting) then they will not

immediately begin to melt it. However, pyroclasts are likely to have a significantly lower albedo

than ice, and will therefore absorb solar insolation more efficiently than the ice, raising the

surface temperature of the pyroclast layer, possibly above the ice melting point. If this

temperature increase is communicated by conduction through the pyroclast layer to the ice

below, it will initiate ice melting. Informal field observations by the authors of tephra on

glaciers have suggested that a tephra layer < 2 cm thick acts to accelerate ice/snow melting,


whilst thicker layers seem to insulate the snow surface (e.g. Manville et al., 2000 and personal

observations of the authors). However, such observations do not take time into account. On

diurnal and annual timescales, melting is enhanced overall, to varying degrees (i.e. more for

thinner layers), and the entire surface draped in ash is lowered at rates above “normal” (i.e. under

ash-free conditions). However, much thicker ash layers do insulate snow/ice surfaces, as we

show below.

The warming effect can be quantified by considering the balance between the incoming

solar heat flux and the heat flux radiated by the surface, equal to (ε σ T4), where ε is the

emissivity, σ is the Stephan-Boltzman constant, and T is the absolute surface temperature.

Clearly, the mean surface temperature of an exposed surface is inversely proportional to the

fourth root of its emissivity. The albedo of glacial ice ranges up to 0.4 (Paterson, 1994), whereas

mafic silicates have albedoes as low as 0.1 (Farrand & Singer, 1992). The corresponding

emissivities are 0.6 and 0.9, and so the effect of emplacing a layer of pyroclasts will be to

increase the surface temperature by a factor of up to (0.9/0.6)1/4 = ~1.11. Thus ice at a

temperature as low as (273/1.11 =) ~247 K (-26 ºC) could be heated to the melting point by this

process. Melting will not be instantaneous. A thermal wave will penetrate a layer of pyroclasts

of thickness λ in a time τ equal to ~2.32 (κ τ)1/2, where κ is the thermal diffusivity, ~7 × 10-7 m2

s-1. The relevant time scales for diurnal and seasonal temperature fluctuations are one day (~9 ×

104 s) and one year (~3 × 107 s), for which λ = 0.58 m and 10.6 m, respectively. Thus a low-

albedo pyroclast layer much less than half a meter thick overlying ice in a region where the mean

diurnal temperature is only a few degrees below the melting temperature may cause a significant

amount of ice melting beneath it during each daily temperature cycle. Conversely a pyroclast

layer a few meters thick will delay the onset of ice melting for several days, even if the surface of

the pyroclast layer warms above 273 K every day. However, a pyroclast layer more than about

20 meters thick would be needed to protect the ice against an annual heating cycle.

4.3. Geological examples

Detailed descriptions of lava that flowed onto glaciers are rare and focus mainly on the

secondary generation of lahar and meltwater flood events (e.g. summary by Major & Newhall,


1989; also Khrenov et al., 1988; Vinogradov & Murav’ev, 1988). Smellie (2007, 2009)

postulated conceptually that lava emplaced on a glacier surface would be redeposited

immediately down-dip of the source edifice as breccia during rapid in situ downwasting of the

glacier melts (e.g. at glacial termination). With coeval glacier flow, however, it also seems likely

that most of the lava will be carried away from its source by the glacier and deposited as lava

clast-dominated breccia in ice-marginal moraines in much the same way as cool pyroclastic

deposits laid down on moving ice (cf. Figure 10). The only published description of in situ lava

characteristics specifically attributed to snow interaction is by Mee et al. (2006), who described

an andesite lava from a winter eruption in Chile that flowed over snow. The evidence comprised

(1) a flow-front c. 5 m wide and 5 m high composed of blocky glassy breccia, that formed as a

talus apron during post-emplacement gravitational instability; and (2) a 20 m-wide zone behind

the flow-front forming a basal layer several metres thick, that consisted of glassy andesite lava

showing conspicuous distinctive, cross-cutting, curviplanar “pseudopillow” fractures and

abundant perpendicular small-diameter secondary fractures (Figure 14). Both joint sets were

thought to have been caused by the overridden snow melting and flashing to steam. The lava in

the snow-interaction zone was subsequently overridden during the same emplacement event by

crystalline subaerial lava with a blocky surface autobreccia showing no evidence for water

chilling.

Pyroclasts deposited on glaciers, whether derived from fallout or pyroclastic density

currents, are reworked and redeposited by either eolian activity or in lahars generated by melting

related to pyroclastic density currents (e.g. Thouret, 1990; Walder, 2000b). They will also be

advected en masse to marginal locations by ice flow, where they will be dumped in moraines

and/or extensively reworked as further mass flows or by proglacial or later non-glacial

weathering and fluvial processes, leaving a broad zone centred on the erupting vent(s) swept free

of pyroclasts (e.g. Manville et al., 2000; Höskuldsson, 2001). Because the off-ice deposits are

not preserved in their original position, but are remobilised, reworked and redeposited, there are

no primary criteria with preservation potential by which they can be related to their initial

deposition on ice. The reworked deposits are likely to be mixtures of pyroclasts, ice, snow and

water (perhaps as much as 65-90 % snow and ice particles in some cases; Manville et al., 2000;


Figure 15). The snow/ice components will melt out, leaving behind a “lag” deposit comprising

mainly porous ashy material that is very susceptible to being washed away by rain, thus

destroying any original textures. When mixed with coarser accidental material, these lag

deposits will simply resemble deposits of lahars. Of the many post-depositional interactions that

might occur on a snow and ice-covered volcano, slope angle and aspect, ice thermal regime

(temperate, polar), pyroclast grain componentry and distribution, and a variety of local climate

parameters, such as mean temperature, diurnal temperature range, insolation and precipitation,

are believed to be most important (Manville et al., 2000).

5. Summary

We have described some basic physical principles that underlie the nature of volcanic

eruptions taking place under, into, and onto glacial ice deposits, and some candidate examples of

such deposits and edifices encountered in the field on Earth. The likely modes of magma-ice

interaction can be predicted theoretically, and form the basis for further testing these predictions

in the field. Direct observations of magma-ice interactions beneath thick ice covers have not

been made, but could be studied in the future with properly geophysically instrumented sites.

Although the immediate products of such interactions are commonly extensively modified by the

flow of the water that is inevitably produced, these specific theoretical predictions form a

paradigm on which initial conditions can be visualized, and interpretations can be based. The

range of possible interactions is large, and the potential clearly exists for the formation of a very

diverse suite of deposits and landforms, many of which we have illustrated. Furthermore,

although our understanding of the relevant mechanisms has evolved from studies of eruptions

and landforms on Earth, the same basic principles can also been applied to volcano-ice

interactions on Mars, where examples of dike and sill intrusions into glacial ice, as well as

related phreatomagmatic eruptions and tephra deposition have been documented (e.g., Head &

Wilson, 2002, 2006; Shean et al., 2005; Kadish et al., 2008; Wilson & Head, 2008).

6. Acknowledgements


This paper is a contribution to the British Antarctic Survey’s GEACEP programme

(ISODYN Project), which sought to investigate climate change over geological time scales and

to develop appropriate novel climate change proxies, in this case glaciovolcanism. We are

grateful to Malcolm Hole and Katy Mee for permission to use their photographs. JWH gratefully

acknowledges grants from the NASA Mars Data Analysis Program, NNG04G99G and

NNX07AN95G, and the Mars Express High Resolution Stereo Camera Co-Investigator Program

(DTM-3250-05).


References ASME (2000) ASME international steam tables for industrial use. New York, NY: American

Society of Mechanical Engineers, 286 pp. Baker, P. E., McReath, I., Harvey, M. R., Roobol, M. J. & Davies, T. G. (1975). The geology of

the South Shetland Islands: V. Volcanic evolution of Deception Island. British Antarctic Survey Scientific Reports, 78, 81 pp.

Behrendt, J. C., Blankenship, S. D., Morse, D. L., Finn, C. A. & Bell, R. A. (2002). Subglacial volcanic features beneath the West Antarctic Ice Sheet interpreted from aeromagnetic and radar ice sounding. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 337-355.

Bergh, S. & Sigvaldason, G. E. (1991). Pleistocene mass-flow deposits of basaltic hyaloclastite on a shallow submarine shelf, South Iceland. Bulletin of Volcanology, 53, 597-611.

Björnsson, H. (1992). Jokulhlaups in Iceland: Prediction, characteristics and simulation. Annals of Glaciology, 16, 95-106.

Blankenship, D. D., Bell, R. E., Hodge, S. M., Brozena, J. M., Behrendt, J. C. & Finn, C. A. (1993). Active volcanism beneath the West Antarctic Ice Sheet and implications for ice-sheet stability. Nature, 361, 526-529.

Bourgeois, O., Dauteuil, O. & Van Vliet-Lanoe, B. (1998). Pleistocene subglacial volcanism in Iceland: Tectonic implications. Earth and Planetary Science Letters, 164, 165-178.

Carslaw, H. S. & Jaeger, J. C. (1959). Conduction of Heat in Solid, 2nd edn. Oxford: Clarendon Press, 510 pp.

Corr, H. F. & Vaughan, D. G. (2008). A recent volcanic eruption beneath the West Antarctic Ice Sheet. Nature Geoscience, 1, 122-125.

Dixon, J. E. (1997). Degassing of alkalic basalts. American Mineralogist, 82, 368–378. Dixon, J. E., Filiberto, J. R., Moore, J. G. & Hickson, C. J. (2002). Volatiles in basaltic glasses

from a subglacial volcano in northern British Columbia (Canada): implication for ice sheet thickness and mantle volatiles. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 255-271.

Einarsson, T. (1948). The flowing lava. Studies of its main physical and chemical characteristics, Societas Scientiarum Islandica, IV (3), 70 pp.

Einarsson, T. (1994). Geology of Iceland. Rocks and Landscape. Reykjavik: Málog Menning Publishing Company, 309 pp.

Einarsson, T. & Albertsson, K. J. (1988). The glacial history of Iceland during the past three million years. Philosophical Transactions of the Royal Society London Series A, 318, 637-644.

Farrand, W. H. & Singer, R. B. (1992). Alteration of hydrovolcanic basaltic ash - observations with visible and near-infrared spectrometry. Journal of Geophysical Research, 97 (B12), 17393-17408.

Geirsdóttir, A. & Eiríksson, J. (1994). Growth of intermittent ice sheet in Iceland during the Late Pliocene and Early Pleistocene. Quaternary Research, 42, 115-130.

Gerlach, T. M. (1986). Exsolution of H2O, CO2, and S during eruptive episodes at Kilauea Volcano, Hawaii. Journal of Geophysical Research, 91, 2177-2185.

Glen, J. W. (1952). Experiments on the deformation of ice. Journal of Glaciology, 2, 111-114.


Gudmundsson, M. T., Sigmundsson, F., Björnsson, H. & Högnadottir, T. (2004). The 1996 eruption at Gjálp, Vatnajökull ice cap, Iceland: efficiency of heat transfer, ice deformation and subglacial water pressure. Bulletin of Volcanology, 66, 46-65.

Harder, M. & Russell, J. K. (2007). Basanite glaciovolcanism at Langorse mountain, northern British Columbia, Canada. Bulletin of Volcanology, 69, 329-340.

Harris, D. M. (1981). The concentration of CO2 in submarine tholeiitic basalts. Journal of Geology, 89, 689-701.

Head, J. W. & Wilson, L. (1987). Lava fountain heights at Pu'u 'O'o, Kilauea, Hawai'i: indicators of amount and variations of exsolved magma volatiles. Journal of Geophysical Research, 92, 13715-13719.

Head, J. W. & Wilson, L. (1989). Basaltic pyroclastic eruptions: Influence of gas-release patterns and volume fluxes on fountain structure and the formation of cinder cones, spatter cones, rootless flows, lava ponds and lava flows. Journal of Volcanology and Geothermal Research, 37, 261-271.

Head, J. W. & Wilson, L. (2002). Mars: A review and synthesis of general environments and geological settings of magma-H2O interactions. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 27-57.

Head, J. W. & Wilson, L. (2003). Deep submarine pyroclastic eruptions: theory and predicted landforms and deposits. Journal of Volcanology and Geothermal Research, 121, 155-193.

Head, J. W. & Wilson, L. (2006). Heat transfer in volcano-ice interactions on Mars: Synthesis of environments and implications for processes and landforms. Annals of Glaciology, 45, 1-13.

Höskuldsson, A. (2001). Late Pleistocene subglacial caldera formation at Cero las Cumbres, eastern Mexico. Jökull, 50, 49-64.

Höskuldsson, A. & Sparks, R. S. J. (1997). Thermodynamics and fluid dynamics of effusive subglacial eruptions. Bulletin of Volcanology, 59, 219-230.

Höskuldsson, A., Sparks, R. S. J. & Carroll, M.R. (2006). Constraints on the dynamics of subglacial basalt eruptions from geological and geochemical observations at Kverkfjöll, NE-Iceland. Bulletin of Volcanology, 68, 689-701.

Jaupart, C. & Vergniolle, S. (1989). The generation and collapse of a foam layer at the roof of a basaltic magma chamber. Journal of Fluid Mechanics, 203, 347-380.

Jarosch, A., Gudmundsson, M. T., Högnadóttir, T. & Axelsson, G. (2008). Progressive cooling of the hyaloclastite ridge at Gjálp, Iceland, 1996-2005. Journal of Volcanology and Geothermal Research, 170, 218-229.

Jones, J. G. (1969). Intraglacial volcanoes of the Laugarvatn region, south-west Iceland - I. Journal of the Geological Society of London, 124, 197-211.

Jones, J. G. (1970). Intraglacial volcanoes of the Laugarvatn region, southwest Iceland, II. Journal of Geology, 78, 127-140.

Kadish, S. J., Head, J. W., Parsons, R. L. & Marchant, D. R. (2008). The Ascraeus Mons fan-shaped deposit: Volcano-ice interactions and the climatic implications of cold-based tropical mountain glaciation. Icarus, 197, 84-109, doi: 10.1016/j.icarus.2008.03.019.

Kaye, G. W. C. & Laby, T. H. (1995). Tables of physical and chemical constants, 16th edn. Harlow, Essex: Longman.


Khrenov, A. P., Ozerov, A. Yu., Litasov, N. E., Slezin, Yu. B., Murav’ev, Ya. D. & Zharinov, N. A. (1988). Parasitic eruption of Kluchevskoi volcano (Predskazannyi eruption 1983). Volcanology and Seismology, 7, 1-24.

Kieffer, S. W. (1977). Sound speed in liquid-gas mixtures: water-air and water-steam. Journal of Geophysical Research, 82, 2895-2904.

Kjartansson, G. (1948). Water flood and mud flows, in The Eruption of Hekla (1947-1948). Societas Scientiarum Islandica, II (4), 51 pp.

Komatsu, G., Arzhannikov, S. G., Arzhannikova, A. V. & Ershov, K. (2007). Geomorphology of subglacial volcanoes in the Azas Plateau, the Tuva Republic, Russia. Geomorphology, 88, 312-328.

Larsen, G. & Eiríksson, J. (2008). Late-quaternary terrestrial tephrochronology of Iceland - frequency of explosive eruptions, type and volume of deposits. Journal of Quaternary Science, 23 (2), 109-120.

Lescinsky, D.T. & Sisson, T.W. (1998). Ridge-forming, ice-bounded lava flows at Mount Rainier, Washington. Geology, 26, 351-354.

Lescinsky, D. T. & Fink, J. H. (2000). Lava and ice interactions at stratovolcanoes: Use of characteristic features to determine past glacial extents and future volcano hazards. Journal of Geophysical Research, 105 B10, 23711-23726.

Loughlin, S. C. (2002). Facies analysis of proximal subglacial and proglacial volcaniclastic successions at the Eyjafjallajokull central volcano, southern Iceland. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 149-178.

Major, J. J. and Newhall, C. G. (1989). Snow and ice perturbation during historical volcanic eruptions and the formation of lahars and floods. Bulletin of Volcanology, 52, 1-27.

Manville, V., Hodgson, K. A., Houghton, B. F., Keys, J. R. & White, J. D. L. (2000). Tephra, snow and water: complex sedimentary responses at an active snow-capped stratovolcano, Ruapehu, New Zealand. Bulletin of Volcanology, 62, 278-293.

Mastin, L. G. & Pollard, D. D. (1988). Surface deformation and shallow dyke intrusion processes at Inyo Craters, Long Valley, California. Journal of Geophysical Research, 93, 13221-13235.

Mathews, W. H. (1947). “Tuyas,” flat-topped volcanoes in Northern British Columbia. American Journal of Science, 245, 560-570.

Mee, K., Tuffen, H. & Gilbert, J. (2006). Snow-contact volcanic facies and their use in determining past eruptive environments at Nevados de Chillán volcano, Chile. Bulletin of Volcanology, 68 (4), 363-376, doi:10.1007/s00445-005-0017-6.

Nye, J. F. (1953). The flow law of ice from measurements in glacier tunnels, laboratory experiments, and the Jungfraufirn borehole expedition. Proceedings of the Royal Society of London Series A, 219, 477-489.

Parfitt, E. A. (1991). The role of rift zone storage in controlling the site and timing of eruptions and intrusions of Kilauea volcano, Hawai'i. Journal of Geophysical Research, 96, 10101-10112.

Parfitt, E. A., Wilson, L. & Head, J. W. (1993). Basaltic magma reservoirs: factors controlling their rupture characteristics and evolution. Journal of Volcanology and Geothermal Research, 55, 1-14.

Paterson, W. S. B. (1994). The physics of glaciers, 3rd. edition. Oxford: Pergamon Press, 480 pp.


Pollard, D. D. (1987). Elementary fracture mechanics applied to the structural interpretation of dikes. In Mafic Dike Swarms, ed. H. C. Halls & W. F. Fahrig, Geological Association of Canada Special Paper, 34, 5-24.

Rubin, A. M. (1992). Dyke-induced faulting and graben subsidence in volcanic rift zones. Journal of Geophysical Research, 97, 1839-1858.

Rubin, A. M. & Pollard, D. D. (1987). Origins of blade-like dikes in volcanic rift zones. In Volcanism In Hawaii. U. S. Geological Survey Professional Paper, 1350, 1449-1470.

Ryan, M. P. (1987). Neutral buoyancy and the mechanical evolution of magmatic systems. In Magmatic Processes: Physico-chemical Principles, ed. B. O. Mysen, Geochemical Society Special Publication, 1, 259-287.

Schopka, H. H., Gudmundsson, M. T. & Tuffen, H. (2006). The formation of Helgafell, southwest Iceland, a monogenetic subglacial hyaloclastite ridge: sedimentology, hydrology and volcano–ice interaction. Journal of Volcanology and Geothermal Research, 152, 359-377.

Shean, D. E., Head, J. W. & Marchant, D. R. (2005). Origin and evolution of a cold-based tropical mountain glacier on Mars: The Pavonis Mons fan-shaped deposit. Journal of Geophysical Research, 110, E05001, doi:10.1029/2004JE002360.

Sigmundsson, F. & Einarsson, P. (1992). Glacio-isostatic crustal movements caused by historical volume change of the Vatnajökull ice cap, Iceland. Geophysical Research Letters, 19, 2123-2126.

Skilling, I. P. (1994). Evolution of an englacial volcano: Brown Bluff, Antarctica. Bulletin of Volcanology, 56, 573-591.

Skilling, I. P. (2002). Basaltic pahoehoe lava-fed deltas: large-scale characteristics, clast generation, emplacement processes and environmental discrimination. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 91-113.

Smellie, J. L. (2000). Subglacial eruptions. In Encyclopedia of Volcanoes, ed. H. Sigurdsson. San Diego: Academic Press, 403-418.

Smellie, J. (2002). The 1969 subglacial eruption on Deception Island (Antarctica): events and processes during an eruption beneath a thin glacier and implications for volcanic hazards. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 59-79.

Smellie, J. L. (2006). The relative importance of supraglacial versus subglacial meltwater escape in basaltic subglacial tuya eruptions: an important unresolved conundrum. Earth-Science Reviews, 74, 241-268.

Smellie, J. L. (2007). Quaternary vulcanism: subglacial landforms. In Encyclopedia of Quaternary science, ed. S. A. Elias. Amsterdam: Elsevier, pp. 784-798.

Smellie, J. L. (2008). Basaltic subglacial sheet-like sequences: evidence for two types with different implications for the inferred thickness of associated ice. Earth-Science Reviews, 88, 60-88.

Smellie, J. L. (2009). Terrestrial sub-ice volcanism: landform morphology, sequence characteristics and environmental influences, and implications for candidate Mars examples. In Preservation of random mega-scale events on Mars and Earth: influence on geologic history. ed. M. G. Chapman & L. Keszthely, Geological Society of America Special Papers.


Smellie, J. L. & Hole, M. J. (1997). Products and processes in Pliocene—Recent, subaqueous to emergent volcanism in the Antarctic Peninsula: examples of englacial Surtseyan volcanoc construction. Bulletin of Volcanology, 58, 628-646.

Smellie, J. & Skilling, I. P. (1994). Products of subglacial volcanic eruptions under different ice thickness: Two examples from Antarctica. Sedimentary Geology, 91, 115-129.

Smellie, J. L., Hole, M. J. & Nell, P. A. R. (1993). Late Miocene valley-confined subglacial volcanism in northern Alexander Isand, Antarctic Peninsula. Bulletin of Volcanology, 55, 273-288.

Sparks, R. S. J. (1978). The dynamics of bubble formation and growth in magmas: a review and analysis. Journal of Volcanology and Geothermal Research, 3, 1-37.

Stevenson, J. A., McGarvie, D. W., Smellie, J. L. and Gilbert. J. S. (2006). Subglacial and ice-contact volcanism at Vatnajfall, Öraefajökull, Iceland. Bulletin of Volcanology, 68, 737-752.

Thouret, J-C. (1990). Effects of the November 13, 1985 eruption on the snow pack and icecap of Nevado del Ruiz volcano, Colombia. Journal of Volcanology and Geothermal Research, 41, 177-201.

Tuffen, H. (2007). Models of ice melting and edifice growth at the onset of subglacial basaltic eruptions. Journal of Geophysical Research, 112, B03203, doi:10.1029/2006JB004523.

Tuffen, H., McGarvie, D., Pinkerton, H., Gilbert, J. S. & Brooker, R. A. (2008). An explosive-intrusive subglacial rhyolite eruption at Dalakvísl, Torfajökull, Iceland. Bulletin of Volcanology, 70, 841-860.

Tuffen, H. & Castro, J. (2009). The emplacement of an obsidian dyke though thin ice: Hraftntinnuhryggur, Krafla, Iceland. Journal of Volcanology and Geothermal Research (in press).

Turcotte, D. L. & Schubert, G. (2002). Geodynamics, 2nd edn. Cambridge, UK: Cambridge University Press.

Vinogradov, V. N. & Murav’ev, Ya. D. (1988). Lava-ice interaction during the 1983 Kluchevskoi eruption. Volcanology and Seismology, 7, 39-61.

Walder, J. S. (2000a). Pyroclast/snow interactions and thermally driven slurry formation. Part 1: Theory for monodisperse grain beds. Bulletin of Volcanology, 62, 105-118.

Walder, J. S. (2000b). Pyroclast/snow interactions and thermally driven slurry formation. Part 2: Experiments and theoretical extension to polydisperse tephra. Bulletin of Volcanology, 62, 119-129.

Werner, R. & Schmincke, H-U. (1999). Englacial vs lacustrine origin of volcanic table mountains: evidence from Iceland. Bulletin of Volcanology, 60, 335-354.

White, J. D. L. & Houghton, B. F. (2006). Primary volcaniclastic rocks. Geology, 34, 677-680. Wilson, L. (1980). Relationships between pressure, volatile content and ejecta velocity in three

types of volcanic explosion. Journal of Volcanology and Geothermal Research, 8, 297-313.

Wilson, L. & Head, J. W. (1981). Ascent and eruption of basaltic magma on the Earth and Moon. Journal of Geophysical Research, 86, 2971-3001.

Wilson, L. & Head, J. W. (2002). Heat transfer and melting in subglacial basaltic volcanic eruptions: implications for volcanic deposit morphology and meltwater volumes. In Volcano-Ice Interaction on Earth and Mars, ed. J. L. Smellie and M. G. Chapman, Geological Society of London Special Publication, 202, 5-26.


Wilson, L. & Head, J. W. (2007). Heat transfer in volcano-ice interactions: synthesis and applications to processes and landforms on Earth. Annals of Glaciology, 45, 83-86.

Wilson, L. & Head, J. W. (2008). Tephra deposition on glaciers and ice sheets on Mars: Influence on ice survival, debris content and flow behavior. Journal of Volcanology and Geothermal Research, doi: 10.1016/j.jvolgeores.2008.10.003.

Wohletz, K. H. (1986). Explosive magma-water interactions: thermodynamics, explosive mechanisms, and field studies. Bulletin of Volcanology, 48, 245-264.

Wohletz, K. H. & McQueen, R. G. 1984. Experimental studies of hydromagmatic volcanism. In Explosive Volcanism: Inception, Evolution, and Hazards, Studies in Geophysics, Washington DC: National Academy Press, pp. 158-169.

Wright, A. C. (1978). A reconnaissance study of the McMurdo Volcanics north-west of Koettlitz Glacier. New Zealand Antarctic Record, 1, 10-15.

Zimanowski, B., Frohlich, G. & Lorenz, V. (1991). Quantitative experiments on phreatomagmatic explosions. Journal of Volcanology and Geothermal Research, 48, 341-358.


Table 1. Nomenclature.

Symbol Definition Value Units

E enthalpy of H2O fluid J kg-1

Li latent heat of fusion of ice 333 kJ kg-1

Lm latent heat of fusion of magma 2.09 × 105 J kg-1

Pa atmospheric pressure 105 Pa

Pd dike tip pressure minus ambient stress in ice Pa

Pe super-lithostatic pressure in magma reservoir Pa

Pg weight of glacial ice layer Pa

Pl lithostatic pressure at roof of magma reservoir Pa

Pm pressure due to weight of magma in dike Pa

Pr total pressure in magma reservoir Pa

Psi pressure in magma at inlet to sill Pa

Pt final pressure in gas in dike tip cavity Pa

Pw weight of magma in dike between reservoir and sill inlet Pa

Q universal gas constant 8.314 kJ kmol-1 K-1

T absolute temperature of surface materials K

Tf final equilibrium temperature of ice-magma mixture K

Ti triple point temperature of H2O 273.15 K

Tm magma temperature 1473 K

V volumes of ice and magma mixing m3

Vg volume of gas in magma mixing with ice m3

Vl volume of liquid magma mixing with ice m3

W dike width m

X width of ice layer melted on each side of the dike m

d distance of penetation of thermal wave in time t m

m molecular mass of CO2 44 kg kmol-1

ne mass fraction of CO2 exsolved from magma

nm mass fraction of CO2 initially dissolved in magma 0.002

pmax maximum distance of dike tip penetration into ice layer m

p distance dike tip penetrates into ice layer m

sg specific heat of the CO2 gas 950 J kg-1 K-1

sm specific heat of the magma 900 J kg-1 K-1

t time required for dike tip to penetrate distance p s


y thickness of ice layer m

z depth of magma reservoir top beneath rock-ice interface m

β bulk density of magma in dike 2480 kg m-3

βw mean density of magma between reservoir and sill inlet kg m-3

ε emissivity of surface materials

κ thermal diffusivity of ice and dyke magma ~10-6 m2 s-1

λ distance thermal wave propagates into surface m

ρg density of CO2 gas kg m-3

ρi ice density 917 kg m-3

ρm density of magma in reservoir 2700 kg m-3

ρr crustal rock bulk density 2300 kg m-3

σ Stephan-Boltzman constant 5.67 × 10-8 W m-2 K-4


Figure Captions

Figure 1. Geometry of a dike rising from a magma reservoir with its top located at a depth z

below a rock surface on which an ice layer of thickness y is present. The dike tip penetrates to a

distance p into the ice. The densities of ice, crustal rock, and magma are indicated.

Figure 2. Variation of p, the distance that a dike tip can penetrate into a surface ice layer, as a

function of the ice layer thickness, y, for four values of the reservoir magma H2O content labeled

as a mass percentage. "Small" implies a negligible water content. These results assume that there

is no excess pressure in the magma reservoir - see text for description of other conditions.

Figure 3. Variation of the stress difference, Pd, acting across the walls of a dike tip as a function

of the distance that a dike tip penetrates into a surface ice layer, p. Details as for Figure 2.

Figure 4. Variation of the final equilibrium temperature, Tf, and pressure, Pf, in an intimate

mixture of magma and ice after the ice has melted to form a fluid, shown as a function of the

pressure Pt in the propagating dike tip, an indicator of the magma H2O content.

Figure 5. Series of schematic diagrams showing development of englacially-emplaced dyke-

collapse breccias and subsequent edifice growth. (a) dyke stalling after intrusion in ice sheet; (b)

lack of support at the dyke base (by ice melt-back) causes dyke to collapse and form

hyaloclastite-rubble breccia that fills the available space by displacing meltwater; initial ice

fracture caused by dyke injection pinches off by stress relaxation and ice deformation; (c) chaotic

dyke re-injection causes pillow lava effusion on top of dyke-collapse breccia, further melt-back


of supporting ice walls and further collapses as unstable steep pile repeatedly fills any gap

between the breccia mass and the retreating confining ice walls; incorporation of lava pillows

creates pillow-fragment breccia; (d) further collapses of the breccia—pillow lava pile as

confining ice walls melt back; (e) breccia—pillow lava pile now draped entirely by pillow lava,

with breccia core, as the edifice transforms into (f) pillow volcano; the breccia core will rapidly

become volumetrically very minor compared to pillow lava, and hard to discern in outcrop; if the

eruption is long-lived and voluminous, the edifice may ultimately evolve into a subaqueous tuff

cone (tindar) or even a tuya. Not to scale.

Figure 6. Field photograph showing possible outcrop of englacially-emplaced dyke-collapse

breccia, Mussorgsky Peaks, Alexander Island, Antarctica. The outcrop is very crudely stratified,

dominated by orange-coloured coarse hyaloclastite breccia with abundant water-chilled intrusive

lenses of sheet- and pillow lava. Photograph by Malcolm Hole.

Figure 7. Values of the pressure, Pe, in excess of the local lithostatic load at the roof of a magma

reservoir required to allow a sill to be intruded at the rock-ice interface as a function of the ice

layer thickness, y, and the depth of the roof of the magma reservoir below the interface, z.

Figure 8. Variation of the maximum magma CO2 bubble volume fraction (left axis, solid

curves), and corresponding magma pressure Psi (right axis, broken line), in a sill intrusion at the

ice-rock interface as a function of the thickness, y, of the overlying ice layer for the case where

the pressure is just equal to the weight of the overlying ice. Gas volume fractions are given for

total magma CO2 contents of 0.1 and 0.2 %. Solid horizontal lines delimit the range of gas


volume fractions, ~0.7-0.85, over which magmatic foams become unstable and disintegrate into

pyroclasts.

Figure 9. Field photograph showing several coalesced cinder cones constructed on glacier ice

downslope of and partly infilling a large glacier fissure (seen as a prominent ice wall in the

background) formed during the 1969 eruption on Deception Island, Antarctica. Person in lower

left foreground for scale. See Baker et al. (1975) and Smellie (2002) for descriptions of the

eruption.

Figure 10. Conspicuous bright red stratum of weakly welded, oxidised scoria in glacier ice that

was deposited on the surface of the glacier on Deception Island shown in Figure 9 during an

earlier subglacial fissure eruption some time between 1830 and 1927 (unpublished information

of J. L. Smellie). Subsequent glacier flow has resulted in the products being carried to the

glacier snout, where they were being deposited in the sea when photographed in 1968. The ice

cliffs are c. 25 m high.

Figure 11. Schematic diagrams illustrating the emplacement conditions of the two different types

of basaltic glacovolcanic sheet-like sequences. (a) Mt Pinafore-type (eruptions under thin ice (<

200 m); perspective view and transverse section); (b) Dalsheidi-type (eruptions by sill

emplacement under thick ice (> 1000 m); perspective view). Modified after Smellie (2008 and

unpublished). Not to scale.


Figure 12. Schematic vertical profile sections showing “standard sequences” characteristic of (a)

Mt Pinafore- and (b) Dalsheidi-type basaltic subglacial sheet-like eruptions. Products of two

discrete Mt Pinafore-type eruptive events are shown in (a), resulting in a repetition of most

lithofacies, whereas products of just a single Dalshiedi-type eruptive episode are shown in (b).

The Mt Pinafore-type sequences in (a) are both incomplete as they lack capping subaerial

lithofacies. See text for details. Adapted from Smellie et al. (1993) and Smellie (2008).

Figure 13. Variation of the ice melting rate beneath a lava flow near its vent as a function of time

after the onset of melting. Curves are labelled for 4 flow thicknesses, 1, 3, 10 and 30 m. The

curves for 1 and 3 m are indistinguishable at small times.

Figure 14. Close view of an andesite lava flow originally emplaced on snow. The lava shows

distinctive curviplanar primary cooling fractures, as well as smaller secondary fractures

orientated perpendicular to the primary set. Although such fractures are not diagnostic of lavas

emplaced on snow, they are a distinctive and conspicuous feature and indicate thermal

contraction during quenching of the lava probably by steam derived from melted underlying

snow (see Mee et al., 2006; photograph provided by Katy Mee). However, the host rock appears

to be aphanitic rather than glassy, which implies that the fractures were formed at sub-solidus

temperatures and not while the lava was still molten. Thus, there was a finite time lag between

emplacement of the lava and influx of the steam, suggesting a comparatively slow rate of heat

transfer to the snow from the base of the lava.


Figure 15. Lahar deposits composed of dark basaltic pyroclasts (ash and lapilli) produced during

phreatomagmatic and magmatic eruptive activity, clasts ripped up from the local (volcanic)

bedrock, and abundant ice blocks in (a) supraglacial and (b) proglacial locations. Both deposits

are products of the 1969 subglacial fissure eruption on Deception Island photographed about one

month after the eruption. See Baker et al. (1975) and Smellie (2002) for descriptions of the

eruption. The prominent ice blocks in (a) are mainly c. 30-50 cm in diameter but some exceed 1

m. The long-term preservation potential of supraglacial lahar deposits such as those shown in (a)

are negligible, whilst those deposited in a proglacial setting are higher, despite local fluvial

reworking.

Volcano-ice interactions - Brown University · Modeling Volcanic Processes: The Physics and...

Documents

Transcript of Volcano-ice interactions - Brown University · Modeling Volcanic Processes: The Physics and...