The MMVF is situated in northern Victoria Land. It is part of the Melbourne volcanic province, which contains four other volcanic fields (Smellie and Rocchi, 2021). The MMVF has been described by Wörner and Viereck (1987), Wörner and Viereck (1990), Wörner et al. (1989), Wörner and Orsi (1990), and Giordano et al. (2012) and is summarized here. The petrology is described by Armienti et al. (1991), Wörner and Viereck (1990), Beccaluva et al. (1991), Lee et al. (2015), and Gambino et al. (2021). Compositionally, the MMVF is a Na-alkaline series, ranging from tephrite/basanite/alkali basalt to trachyte (Figure 2).

Mount Melbourne has a crater or possibly a small caldera 700 m in diameter. The summit is mainly constructed of trachyte to benmoreite domes, scoria and phreatomagmatic cones, and lavas from which the youngest published ages have been obtained (Giordano et al., 2012; Lee et al., 2015). Tephras are also present, including welded fall, and basaltic bombs scattered around the summit attest to recent mafic activity. The crater and upper slope on the north side also contain areas of heated ground and fumaroles (Lyon and Giggenbach, 1974; Lyon, 1986; Wörner et al., 1989; Gambino et al., 2021). The most recent eruption dates back to c. AD 1892–1922 (Lyon, 1986; Wörner et al., 1989), but numerous cryptotephras found in ice cores extend the recent explosive history of Mount Melbourne back to Eemian time, at least 124 ka, possibly 137 ka (Del Carlo et al., 2015; Narcisi and Petit, 2021). Outcrops at lower elevations, which represent flank and satellite vents, are the focus of this study and are scattered widely surrounding Mount Melbourne. They are generally isolated and small, with the largest snow-free outcrops at Edmonson Point and Shield Nunatak. Apart from a small isolated trachyte dome or lava c. 3 km west-northwest of Edmonson Point, the flank and satellite outcrops are tephrite/basanite to benmoreite in composition. They are mainly scoria cones showing variable degrees of degradation, but there are also outcrops of palagonitized lapilli tuffs described as tuff rings, together with enigmatic outcrops composed of megapillows and large lava lobes. Additionally, our study identified several outcrops composed of ‘a‘ā lava-fed deltas mainly confined to the Cape Washington peninsula (Figure 3).

Radioisotopic ages for rocks in the MMVF are mainly based on K-Ar data (Armstrong, 1978; Wörner et al., 1989; Armienti et al., 1991; Lee et al., 2015), mostly without analytical details. More recently, Giordano et al. (2012) reported ⁴⁰Ar/³⁹Ar data. The previously published results were based on old constants (decay constants and/or the age of the reference mineral) and are therefore expected to be c. ≤ 1% younger, but these variations fall largely within the analytical uncertainties. The radioisotopic ages suggest that volcanism began in the Late Miocene (c. 12.5 Ma) but became widespread from the Pliocene (from c. 4 Ma but mainly after c. 2.5 Ma). It continues today, although no eruptions have been witnessed (Wörner et al., 1989; Giordano et al., 2012; Gambino et al., 2021). Units dated by the K-Ar method are mainly intrusions in tuff cones and range in age from 2.96 ± 0.2 to 0.2 ± 0.04 Ma but including an outlier age of 12.50 ± 0.18 Ma (Supplementary Figure S1). ⁴⁰Ar/³⁹Ar ages published by Giordano et al. (2012) included samples from Edmonson Point, Shield Nunatak, Markham Island, Harrow Peaks, and Random Hills (Figure 3). The dated outcrops vary from scoria cones and associated lava fields to ignimbrite and a lava in one of the megapillow complexes. The localities do not overlap with those dated by K-Ar, and the ages range between 1.368 ± 0.009 and 0.0907 ± 0.019 ka. The 2-σ uncertainties are generally quite high.

The scoria deposits form variably degraded pyroclastic cones constructed mainly of massive to crudely bedded fines-free deposits dominated by oxidized scoria (LP; Figure 4A). They are associated with agglutinate, clastogenic lavas, and ‘a‘ā (sL; rarely pāhoehoe [cL]) lava fields (Figure 4B). The lavas have oxidized autobreccias. Additionally, prominent thick, irregular crystalline intrusions with coarse, crude prismatic joints occur at depth in outcrops on the Cape Washington peninsula (iL; Figure 5). The characteristics of the outcrops are indicative of weakly explosive and effusive magmatic eruptions in the absence of groundwater or surface water (or snow/ice).

At least 12 discrete outcrops formed of lapilli tuff and tuff are present in the MMVF (Figure 3). They are strongly eroded. All the outcrops are dominated by khaki-yellow fine lapilli tuffs with dips often between 18 and 24 but varying to horizontal (Figures 6, 7). Crater-rim unconformities are preserved in strata high in the outcrops at northern Baker Rocks and on the coast northeast of Baker Rocks. Important characteristics include predominant planar, diffuse, discontinuous stratification, rare low-angle cross-stratification, and monomict compositions, with minor accessory lava clasts and scarcity of impact structures beneath outsize clasts (dsLT, dsT). Clasts are blocky and variably vesicular sideromelane and poorly sorted, they often contain abundant fine ash-size matrix, and some contain ash-coated lapilli. These are characteristics shared by deposits formed from dilute pyroclastic density currents (PDCs) during explosive hydrovolcanic eruptions (Branney and Kokelaar, 2002). They result in the construction of tuff cones and tuff rings, but the absence of accidental (bedrock-derived) clasts (except possibly in basal beds at Harrow Peaks; Giordano et al., 2012) is more typical of tuff cones, with the loci of explosions high in the eruptive pile rather than in bedrock. Bedrock-sourced explosions, typically resulting in a high proportion of bedrock clasts (>10 vol%; Sohn, 1996; Ort et al., 2018; Latutrie and Ross, 2020), result in the construction of tuff rings and maars, whereas tuff cones typically form where magma interacts with shallow surface water (lakes or the sea), although they can also form above free-flowing readily recharged aquifers (Sohn and Chough, 1992). The absence of basal pillow lava mounds and pillow lava fragments as clasts in the outcrops indicates that volatile exsolution and explosivity were not suppressed at the outset consistent with low ambient pressures and thus relatively shallow water depths, or possibly relatively thin overlying ice. Although it is thought that powerful explosive MFCI hydrovolcanic eruptions are increasingly unlikely at water depths exceeding 100 m (Zimanowski and Büttner, 2003; Dürig et al., 2020), the depth of the effusive to explosive transition for mafic magmas is not well defined. Estimates typically range between 100 and 200 m (Jones, 1970; Allen, 1980; White et al., 2003; Schopka et al., 2006; Valentine et al., 2014). Explosivity at greater depths is theoretically possible but probably requires special circumstances (IFCI: Dürig et al., 2020), and the relatively high hydraulic pressures at depths greater than 100–200 m seem likely to severely restrict explosivity (cf. Wohletz, 2003; Chadwick et al., 2008; Resing et al., 2011).

The pervasive diffuse stratification of the dominant lapilli tuff lithofacies, with its ill-defined bedding surfaces (Figure 4C), indicates rapid deposition from PDCs during vertical aggradation of the eruption-fed tephra piles, probably accompanying a sustained continuous-uprush style of explosive activity (White, 2000; Smellie, 2001). The deposits are not diagnostic of depositional settings, which can be subaerial or subaqueous (Sohn et al., 2008; Russell et al., 2013; Sohn and Sohn, 2019). However, the preservation of ash-coated lapilli (sensu Brown et al., 2012) is unlikely if the lapilli fell into and sank through water. They occur in outcrops at northern Baker Rocks, Oscar cliff, Shield Nunatak, and east of Willows Nunatak. In each case, they are confined to the uppermost strata, consistent with subaerial exposure of the tuff cone in its final eruptive stages. In a few examples, the final eruptive activity consisted of scoria cones, signifying “dry” subaerial conditions [Baker Rocks (south), Shield Nunatak]. Uncommon occurrences of ash-coated lapilli at lower elevations (e.g., at Oscar cliff) might be reworked during discrete mass flow events in which the lapilli were transported while isolated from the overlying water by relatively concentrated PDCs rich in fine tuff. However, tuff cone eruptions within a water-flooded vent occur through repeated ejection and recycling of water-saturated slurries (Kokelaar, 1983). It is an environment that will probably reduce the formation of ash-coated particles and ash aggregates because of the lack of surface tension effects (Go et al., 2017). Ash-coated particles and aggregates are common in the subaerially constructed summit cones formed after the vent is no longer flooded (Sohn et al., 2008).

By contrast, the associated, less common lithofacies may have been deposited mainly under subaqueous conditions. They include lithofacies glLT, dsT, xT, bT and glT (Table 1). They consist of well-structured laterally continuous beds with sharp surfaces and a variety of structures, including amalgamation, reverse- and normal-grading, syn-sedimentary instability (flame structures, soft-state deformation), ripple bedforms, and planar and wavy laminations (Figures 4D,E), which collectively probably indicate subaqueous transport and deposition. Beds of lapilli tuff with well-defined (rather than diffuse) bedding surfaces have been interpreted as products of intermittent jetting activity in Surtseyan tuff cones resulting in discrete pulses of tephra cascading down the subaqueous flanks as gravity flows (White, 1996; White, 2000; Smellie, 2001; Sohn et al., 2008). Although there is a resemblance to subaerial deposits (e.g., Sohn and Chough, 1992; Sohn, 1996; Solgevik et al., 2007), it is unlikely that fast-moving subaerial pyroclastic density currents are capable of creating the deposits with (admittedly rare) small (<c. 5 cm amplitude) ripples with steep lee faces, whereas they are a common bedform in subaqueous sequences (Skilling, 1994; White, 1996; White, 2000; Sohn et al., 2008; Moorhouse and White, 2016; Sohn and Sohn, 2019). The soft-state deformation is also more likely to occur in a water-saturated substrate. The absence of steep-sided erosion surfaces, such as incised rills and other channels cut by surface wash, and the absence of ash-coated lapilli are also indirect evidence for a subaqueous depositional setting (Sohn et al., 2008). Additional support is provided by the presence of intrusions at high elevations in the tephra piles showing pervasive entablature or closely spaced irregular jointing and intrusive lava pillows (iL; Figure 4F). The style of jointing and development of lava pillows indicates a water-rich environment. Hence, the tephra piles were water-saturated and presumably surrounded by a lake. Entablature jointing in the coeval intrusions typically extends up to c. 200 m above sea level (asl) but may exceed 300 m asl in the outcrop northeast of Baker Rocks. The neck or large dyke in the outcrop north of Willows Nunatak also includes marginal megapillows invading adjacent lapilli tuffs (Figure 4G). Finally, the tuff cone outcrop at Shield Nunatak has a capping unit of sheet-like ‘a‘ā lavas. In part, they are coarsely jointed and finely crystalline consistent with subaerial emplacement, although lacking oxidation of the autobreccias. They were probably sourced from the small summit scoria cone present (Figure 6). However, the lavas transform down-dip into invasive sheets showing the development of aphanitic breccias and pillows (Figure 4H), indicating intrusion into the tephra deposits, which were water saturated and unlithified, and a high coeval water level. Scoria deposits also occur in other lapilli tuff outcrops in the MMVF [east of Willows Nunatak (Figure 7), Baker Rocks (south)]. In summary, the tuff cone outcrops show evidence for subaqueous and subaerial eruptions consistent with an origin as Surtseyan (subaqueous to emergent) tuff cone edifices.

Outcrops composed of interconnected megapillows, lobes, and lava sheets (MpL) are restricted to the east side of the Mount Melbourne peninsula, at Edmonson Point, and in surrounding nunataks (Figures 3, 8, 9A–C). The division into several morphologies is based on size (terminology after Goto and McPhie, 2004). However, the exposure is only in two dimensions, so the full three-dimensional morphologies of the lava bodies are unknown. Although megapillow shapes can reasonably be inferred for some outcrops (Figure 9A), other megapillows might be cross-sections through tubes. All lava masses are characterized by multiple sets of prominent cooling fractures, including sheeting, entablature, sheet-like, blocky, and pseudopillow (Figures 9B,C). Blocky joints (also called cube joints or hackly joints) are a more densely fractured version of entablature caused by interaction with larger amounts of coolant and a faster cooling rate (Forbes et al., 2014). The spatial development of the joint types is the same in each outcrop and is concentric to the lava morphology (Figure 10). All the joints are extensional structures, and the style of fracturing and fracture spacing reflect (1) the position of the fractures relative to the lava margins and (2) the cooling history of the lava, that is, progressive inward cooling (Lescinsky and Fink, 2000). They are caused by the rapid chilling of lava against water or ice, and the fracture spacing is inversely proportional to the cooling rate (Long and Wood, 1986; Grossenbacher and McDuffie, 1995; Lescinsky and Fink, 2000; Goehring and Morris, 2008). The marginal sheeting joints wrap around and mimic the shape of the individual lava bodies. They form after the lava has substantially crystallized and are thought to be related to late-stage shear within the lava (Bonnichsen and Kauffmann, 1987). In the absence of exposed surfaces, they are often the best indicator of the overall morphology of the individual lava masses. Moreover, observations of the sheeting joint orientations have shown that most of the lava masses in the MMVF outcrops are probably interconnected. The multiple bands of vesicles observed in some megapillows and lobes are vesicle zones characteristic of lava inflation (Self et al., 1996; Self et al., 1998). The autobreccia clinkers are also vesicular. Together with the vesicle zones, the features suggest that the lavas were not fully degassed when they erupted.

Lava pillows and megapillows are thought to characterize subaqueous lava emplacement (Batiza and White, 2000; Goto and McPhie, 2004; Hungerford et al., 2014). In the megapillow complexes of the MMVF, the presence of lava pillows and megapillows with glassy rims and lenses of hyaloclastite indicates rapid water chilling. Moreover, entablatures and pseudopillow fractures form by steam and water penetrating down extensional fractures (Long and Wood, 1986; Lescinsky and Fink, 2000; Forbes et al., 2012; Forbes et al., 2014). However, the lava masses typically have dm-thick dark grey rims that are aphanitic or hypohyaline (Figure 9B), and glassy (holohyaline) rims are poorly developed in general. The presence of relatively thin colonnades below some of the lava sheets (Figure 9C) also indicates slow conductive cooling. The observations are inconsistent with the rapid cooling of subaqueous lavas. The sedimentary features of the interbedded lenses of pumiceous lapilli tuff (pLT; Table 1) indicate transport and deposition by tractional and density currents (planar lamination and variable grading; cf. Mulder and Alexander, 2001). The abundant highly vesicular pumice is likely to float, suggesting that transport was fluvial unless the pumice was hot when it fell onto the water and ingested water, causing it to rapidly become waterlogged and to sink (Witham and Sparks, 1986). The abrasion observed in many of the pumices is also consistent with fluvial transport. The aphanitic lava fragments forming the autobreccias, with their fluidal and planar broken faces, resemble the broken crusts seen in many pāhoehoe lavas, but the absence of oxidation suggests that emplacement was not subaerial (i.e., “dry” conditions; Self et al., 1996; Bondre et al., 2004). Thus, the evidence suggests that although water cooling played a prominent part in the emplacement of the megapillow complexes, the individual lava masses may not have been fully submerged (see Section 6.1.3).

Subaqueous lavas are typically categorized on the basis of their morphology as pillowed, lobate, or lava sheets (e.g., Fox et al., 1987; Schmincke and Bednarz, 1990; Gregg and Fink, 1995; Batiza and White, 2000; Goto and McPhie, 2004; Clague and Paduan, 2009; Hungerford et al., 2014). The distinction is based principally on size, although the lavas also become more flattened as they become larger, and, as described above, they display regular cooling joint patterns (Figure 10; Lescinsky and Fink, 2000). The morphological differences have been related to composition, eruption temperature, cooling rate, effusion rate, viscosity, and crystallinity (Schmincke and Bednarz, 1990; Griffiths and Fink, 1992; Gregg and Fink, 1995, and references therein). However, analog models and studies of ocean-floor lavas have demonstrated that the principal link is to discharge rate and bedrock gradient (Gregg and Fink, 1995; Batiza and White, 2000; Clague and Paduan, 2009). Although pillow diameter is known to increase in more evolved magmas (Walker, 1992), studies have shown that a wide spectrum of flow morphologies (i.e., pillow lavas to lava sheets) can occur with negligible compositional differences (Batiza et al., 1989; Batiza and White, 2000) and the rheological effects of composition, temperature, viscosity, and crystallinity are interlinked. There is no compositional dependence evident in the MMVF. Despite the wide range of compositions present in the megapillow complexes (alkali basalt to Si-poor trachyte; Figure 2), similar lava morphologies occur in each outcrop. Moreover, in some (e.g., outcrop 7 km north of Edmonson Point), the lava masses increase in size up through the lava pile despite no obvious compositional variation. Thus, we suggest that the progression of morphologies, from rare pillows through megapillows and lobes to lava sheets, probably reflects increasing discharge rates, with the generally large dimensions due to relatively slow cooling. We suggest that the slower cooling was due to only partial submergence in water; pillow lavas emplaced fully submerged are faster-cooled and, therefore, smaller. Bedrock gradient might have been a contributing factor affecting flow rate, but the observed variations in morphology, which are usually random in the outcrops, would require that the bedrock gradient varied during the course of the individual effusive eruption, which is unlikely unless the lava piled up around the vent, thus increasing the surface gradient over time. Therefore, we suggest that increasing discharge rates at the vent and environmentally imposed slower cooling rates were the likely dominant factors that led to the different lava morphologies in the megapillow complexes. The lowest effusion rates and highest cooling rates were probably responsible for the pillow lavas, the highest effusion rates and slower cooling led to the lava sheets, and intermediate rates were responsible for the megapillows and lava lobes.

A discovery of the present study was the recognition of multiple lava–breccia couplets within the MMVF. They crop out mainly on the Cape Washington peninsula, particularly close to Cape Washington (Figure 11), but there is an additional small isolated outcrop a few kilometers north of Edmonson Point (Figure 3). The combination of massive chaotic aphanitic lava breccia with abundant irregular lava lobes (cB) capped by subaerially emplaced ‘a‘ā lavas (sL) is diagnostic of ‘a‘ā lava-fed deltas (Smellie et al., 2011a; Smellie et al., 2013). The lava-fed deltas have shallow dips of c. 5 consistent with original shield-like landforms. The bedding attitudes of additional outcrops with deltas on the east side of the Cape Washington peninsula indicate that multiple small shield-like volcanoes are present (Figure 12). The upper surfaces of only two lava-fed deltas (#1 and #3) were accessible. They are erosive (Figure 13), resulting in the capping ‘a‘ā lava sequences of both deltas thinning markedly in a southerly direction. The thickest sequence (c. 250 m) is at Cape Washington and comprises four superimposed lava-fed deltas (Figures 5, 11). The lava lobes in the breccia lithosomes show strong evidence for water chilling in the form of diagnostic joint patterns (mainly closely spaced, irregular, or blocky), and the lobes often break up marginally into aphanitic lava breccia. Oxidized clinkers are also dispersed in the breccias and partially encase lava lobes, which are distinctive features of water-chilled lavas and breccias in ‘a‘ā lava-fed deltas (Smellie et al., 2013). The number of lava lobes increases up toward the capping ‘a‘ā lavas, and their proportion often exceeds 90% in the uppermost several meters, corresponding to a passage zone, that is, a transition zone separating subaqueous from subaerial lithofacies (see Jones, 1969; Smellie and Edwards, 2016). Passage zones in ‘a‘ā lava-fed deltas are characteristically crudely defined.

The size, lithofacies, compositional uniformity, and internal architectures of the individual low-elevation outcrops in the MMVF examined in this study indicate that they are the products of small volcanoes (sensu White and Ross, 2011). They might also be called monogenetic, but for ancient eruptive centers, it is often impossible to distinguish the products of multiple eruptions in a single edifice. Studies of similar-sized centers often show that they had long histories of multiple eruptions separated by short breaks in activity. However, the lack of obvious internal erosional surfaces that could have been caused by an eruptive hiatus suggests that they are likely to be monogenetic. The construction and morphology of small volcanic centers are determined by internal and external factors. Internal factors include the composition and physical properties of the erupting magma (e.g., viscosity and volatile content). External factors include tectonic setting (differential stress) and magma input (production) rate, but these factors simply affect the size of an edifice [polygenetic (large) versus monogenetic (small)] (Takada, 1994). They do not determine the style of eruption or type of monogenetic edifice (e.g., scoria cone versus tuff cone). Other external factors are essentially environmental and include the presence and recharge rate of groundwater and surface water (lake, sea). The influence of environmental factors is reflected in the sequence architecture and features of the lithofacies (e.g., clast grading and morphology, clast types, joint patterns, and relative crystallinity).

The range of compositions overlaps for each of the edifice types in the MMVF and, apart from a bias to tephrite/basanites for the lava-fed deltas (Figure 2), there are no obvious compositional differences that explain the markedly different styles of eruption represented. Although many of the scoria cones are at relatively high elevations (e.g., above 1,200 m in Random Hills; Figure 3), they also occur close to sea level. The other types of flank and satellite vents occur at similar elevations to the scoria cones (i.e., mostly <600 m asl), and relative elevation (with respect to potential interaction with marine water) is not the dominant determining factor for eruptive style. A major conclusion of our study is that the principal determinant for volcanic eruptions and edifice construction in the MMVF is environmental, specifically the presence at the eruptive sites of water in all its forms (snow, ice, meltwater, and surface water). Figure 18 schematically shows the relationships between the principal lithofacies, internal architectures, and edifice construction for centers that erupted in various environmental settings. For example, an absence of water or insufficient snow/ice at eruptive sites in the MMVF resulted in scoria cones and their associated lava fields. Most of the tuff cone outcrops visited show strong evidence for eruption within the ice, including gently and sometimes inward-dipping summit strata at high elevations and evidence for water saturation of the tephra piles at high elevations. However, the characteristics of a few tuff cone outcrops are more ambiguous and possibly fit better with eruptions in an unconfined setting (i.e., laterally continuous strata, steep-dipping in summit areas), with a water source potentially provided by the sea. This is particularly true for the Oscar Point tuff cone (Figure 15). Moreover, their ages appear to correspond to warmer periods compared with the marine δ¹⁸O curve (Figure 17; and see Section 6.3) when ice extent would have been reduced and sea levels higher than today. However, independent corroborating evidence for the substantially higher sea levels required is currently lacking. Distinguishing between glacial and non-glacial settings reliably for tuff cones depends on reconstructing the internal architecture and, ideally, the original morphology of the tephra piles, which are not always possible for outcrops that are highly eroded or largely obscured by snow and ice. Effusive eruptions associated with a thin slope-draping glacial cover (<c. 150 m thick for the MMVF) created several small shield volcanoes constructed from multiple ‘a‘ā lava-fed deltas, mainly in the Cape Washington peninsula area. Similar examples also crop out extensively elsewhere in Victoria Land (Smellie et al., 2011a; Smellie et al., 2011b; Smellie et al., 2013; Smellie, 2021). By contrast, effusive eruptions associated with high magma discharge rates and a much thicker glacial cover (several to many hundreds of meters; see Section 6.4) created the distinctive megapillow complexes. Under similarly thick ice but lower magma discharge rates, pillow lava mounds would be formed.

The arguments given above for eruptive settings are based solely on interpretations of the lithofacies present and their architecture. However, considerable ancillary support is also provided by the ages of the outcrops when they are plotted on the marine δ¹⁸O isotope curve (Figure 17). Only two outcrops we visited have no eruptive environment currently assigned due to inconclusive lithofacies and architectures. The remaining outcrops are designated as “glacial” (lava-fed deltas; megapillow complexes; several tuff cones) and “non-glacial” (scoria cones and associated lava fields; possibly some other tuff cones). It is probably significant that despite the high errors on some of the published ages (most determined by imprecise K-Ar), all of the unambiguously “glacial” outcrops have ages corresponding to relatively high δ¹⁸O values on the marine isotope curve, whereas the “non-glacial” outcrops correspond in age to significantly lower δ¹⁸O values. Only one outcrop conflicts with this conclusion: an intrusive neck c. 8 km north of Cape Washington (Figure 17). The constituent “lithofacies unequivocally indicate “dry’ magmatic eruptions and a non-glacial or at least ice-poor setting, but the equivalent value on the δ¹⁸O curve is relatively high (Figure 17). However, the outcrop age corresponds to a steep section of the δ¹⁸O curve. Thus, the simplest explanation is that even a slight change in the age (by only a few ka) could displace it to substantially lower δ¹⁸O values on the curve and remove the conflict. For the few tuff cones in the MMVF that might have erupted within the sea, a similar argument applies, thus emphasizing the ambiguity of the interpretation currently assigned to those outcrops. Finally, it is noticeable that within the constraints of the errors attached to the ages [which are very small (mainly 20–30 Kyr) in our new age dataset], the eruptions of units that experienced “glacial” conditions fall on the steep rising limbs of the δ¹⁸O curve corresponding to periods of rapid ice melt, either glacial to interglacial transitions or a change to interstadials (Figure 17). Such a trend is a predicted consequence of the warmer climatic conditions lowering the superimposed glaciostatic load and triggering eruptions (Jull and McKenzie, 1996; Maclennan et al., 2002; Jellinek et al., 2004; Watt et al., 2013; Wilson and Russell, 2020).

Empirical estimates for contemporaneous ice thicknesses are also shown in Figure 17. For each of the Cape Washington peninsula lava-fed deltas near Cape Washington, calculated ice thicknesses are >55, 55, 60, and 135 m (oldest delta to youngest, respectively). The additional younger delta 9 km north of Cape Washington was associated with ice c. 145 m thick, whereas the isolated delta relict north of Edmonson Point is too poorly exposed for an estimate. Calculations for the deltas are based on the observed thickness of subaqueous lithofacies to which is added c. 30 m to allow for any superimposed snow, firn, or crevassed ice (see Smellie et al. 2011b). The calculated thicknesses are minima as no allowance is made for ice surface sagging, but any sagging would be confined principally to the vicinity of the edifice; see Smellie et al. (2011b) for details of assumptions used in calculations. Other putative ice thicknesses at the eruptive sites are as follows:

1) Megapillow complexes: >200 m (i.e., the individual thickness of most of the megapillow complexes) but probably much thicker; for example, the coeval ice surface for the highest outcrop (west-northwest of Edmonson Point) must have substantially exceeded 400 m asl, which is the elevation of its preserved surface. The thicknesses shown in Figure 17 are minima and based on the maximum elevation reached by the individual outcrops (but see further comments below).

The estimates calculated for the lava-fed deltas and Shield Nunatak are the most accurate and are unlikely to be incorrect by more than a few tens of meters because the bases and original tops of most of the outcrops are exposed and the outcrops are capped by lavas that were emplaced subaerially. Hence, the coeval ice was melted through completely.

Estimates for the other landforms are probably less accurate, particularly for the megapillow complexes emplaced entirely subglacially. Two scenarios are possible for the complexes. (1) In undegassed magmas, the coeval ice thicknesses would have been substantial [i.e., much greater than the preserved thicknesses (mainly c. 200 m) of the individual complexes] to suppress runaway vesiculation triggering explosivity, whether magmatic or hydrovolcanic. Thus, for a complex 200 m thick, the minimum thickness of contemporary ice would have been c. 550 m (comprising 200 m of magma and 200 m of superimposed ice plus c. 150 m to account for ice surface sag toward the complex; see Gudmundsson, 2003; Smellie et al., 2011b). The elevation of the highest megapillow complex surface is c. 400 m asl (nunatak west-northwest of Edmonson Point). It is an eroded surface; hence, it must have been higher originally. Similar reasoning suggests that the coeval ice surface would have had a minimum elevation of c. 750 m asl (i.e., 400 + 200 + 150 m). This is the thickest ice sheet for which volcanic rocks in the MMVF provide evidence. However, for that example, if uplift has subsequently occurred (although undocumented at present), the deduced surface elevation and associated ice sheet thickness would need to be reduced. Since the megapillow complex at Edmonson Point shows an upward transition to explosive activity, the coeval ice thickness is better defined and any overlying ice would not be greater than c. 200 m thick. The exposed thickness of the complex at the locality is c. 200 m; hence, the maximum ice thickness there during effusion was ≤ 550 m.

Alternatively, (2) if the magmas were degassed and emplaced in a partially drained vault, the ambient pressures would have been much less than those imposed by the superimposed ice and might even have been atmospheric if the vault drained to the ice sheet margins (cf. Schopka et al., 2006; Wilson et al., 2013). Low vault pressures (equivalent to less than c. 200 m of superimposed ice) would have promoted pervasive vesiculation and a rapid transition to explosive (hydrovolcanic) activity at an early stage, which is not observed. Hence, some of the magmas (not Edmonson Point) might have been degassed when emplaced. However, the presence of vesicle zones signifying inflation, together with megavesicles, suggest that the magmas were not degassed.

Even if it was assumed (1) that the magmas were degassed and that c. 50 m of ice remained above the cooling complex (to avoid subaerial exposure) and (2) that the ice also sagged toward the cooling magma mass, the megapillow complexes that were c. 200 m thick would have been emplaced below the ice that was at least 400 m thick (i.e., 200 + 50 + 150 m). Whichever assumptions are used, the estimates indicate that substantially thicker ice conditions existed during the emplacement of the megapillow complexes than occurs at those locations today (see also Smellie et al., 2018). However, even the maximum calculated ice thicknesses would be unable to completely drown the topography of the region (Random Hills rise to >1,500 m asl, and the summit of Mount Melbourne is at 2,732 m). The ice cover indicated is, therefore, relatively thin overall, suggesting that it was mainly an icefield during the past 4 Myr. This does not preclude transient episodes of greater ice thicknesses at times when the ice overrode much of the topography (see Smellie et al., 2018), but those episodes are not represented by the volcanic record, and they were atypical for the bulk of the period represented.