Kevin Kiernan


In 1900 the "father of speleology", E.A. Martel, recognised that "there truly exists a speleological sous-glaciare, wide open for future discoveries, but more dangerous than the souterraine...". Few speleologists have yet focussed serious attention on caves formed in snow or ice, although cave systems up to 25km in length have been documented and it may ultimately prove that the longest and deepest caves on earth are developed in glacier ice rather than in limestone. Nevertheless, in some parts of the world caves of this kind are enjoyed by thousands of visitors annually. The impact of visitors on such caves can be significant, but the morphology of the caves is dynamic and the speleothems are generally more rapidly self-renewing than those within karst caves. The safety of visitors may, however, be a greater issue than in conventional karst caves. Deaths have occurred due to accidents involving cave instability and drowning. In addition, the fact that glacier caves are generally located in alpine situations can imply additional hazards on access routes - here too visitors have lost their lives. Hence, management of access to the caves needs to be integrated with management of the caves themselves. The problems that may arise where attempts are made to facilitate the safe entry of visitors to glacier caves are far from insurmountable; the reward can be a very worthwhile and quite unique form of visitor experience. This paper reviews the nature of caves and melt-karst in snow, firn and ice, before discussing some of the issues that require consideration by management agencies.


References in the scientific literature to karst-like phenomena in glaciers date back over a century, there being mention of glacier caves, for instance, in Tyndall's Glaciers of the Alps (1860). Russell (1893) recorded of Alaska's Malaspina Glacier:

"a deep roar coming from the hidden chambers to which moulins [streamsinks] lead frequently tells that large bodies of water are rushing along the ice caves beneath. In the southern portion of the glacier, where the ice has been deeply melted, and especially where large crevasses occur, the abandoned tunnels made by englacial streams are sometimes revealed. These tunnels are frequently 10 or 15 feet high, and occasionally one may pass through them from one depression in the glacier to another. In some instances they are floored with debris, some of which are partially rounded. As melting progresses this material is concentrated at the surface as moraine" (Russel 1893:18).

In his 1900 publication La Speleologie E. A. Martel, the "Father of Speleology", included a chapter entitled 'La Speleologie Glaciare' which he concluded by observing:

"All this permits the affirmation that there truly exists a speleological sous-glaciare, wide open for future discoveries, but more dangerous than the souterraine..."

Caves formed in ice and snow can be large and stunningly beautiful in their form, sometimes their speleothems, and often in the extraordinary ice-blue light that pervades them. To date they have attracted little attention from speleologists. However, they can offer a wonderfully rewarding experience to visitors. In various parts of the world, caves of this kind have for many decades delighted thousands of visitors annually, most notably perhaps in the Rhone Glacier in Switzerland and in the Stephens Glacier on Mt Rainier in Washington State, USA. For this and other reasons, the existence of glacier caves can be of significance in management terms.

Perhaps the most celebrated and visited glacier caves have been the Paradise Ice Caves on the slopes of Mt Rainier. Photographs published by H. L. Toles in 1908 and reprinted in the innaugral Bulletin of the International Glaciospeleological Survey show large parties in these spacious caves, in one case numbering a least 22 people, including ladies in long dresses, in a passage about 5-7m high and 12-14m wide. Another depicts an apparent descent by rope to cave entrances in a sinkhole. Guided trips were popular by the late 1920s. Evidence of cave management initiatives having been taken at an apparently early date have been turned up by parties from the National Speleological Society, including a decaying "Danger" sign and a rotting ladder. A visitor's guide to these caves was published in 1972 (Halliday and Anderson 1978). Glacier cave entrances in the Rocky Mountains were depicted on railroad postcards issued in British Columbia and Alberta in the 1920s and 1930s, while a cave in the snout of the Mendenhall Glacier from the Juneau Icefield in the Alaska-British Columbia border area was depicted on postcards issued in the 1930s and 1940s. In New Zealand too, caves in accessible glaciers have been a tourist attraction in earlier years though they are less so today for reasons we shall address later. One description of a glacier cave near Mt Cook published as late as 1971 still reeked of the literary style of earlier years when it enthused that "one overseas tourist compared the setting with its azure roof to the 'magnificant scenery of a London Christmas pantomine'" (Howard 1971). That the mystery of such places kindles the interest of the general public is amply evident from reaction to the artificial glacier cave constructed in the public display area of the new International Antarctic Centre in Christchurch, New Zealand, as the entrance to a darkened audio-visual theatre.

"Scrambles Beneath the Alps"

While glacier caves have yet to attract much attention from speleologists and cavers the glaciological community has some familiarity with them. Among the boldest of initial attempts to explore a glacier cave were those of Louis Agassiz, one of the founders of the sciences of glacial geomorphology and glaciology. The volume Glaciers of the Alps and Mountaineering in 1861 (Tyndall 1897) records that as part of his researches:

"Agassiz finally determined to descend into the heart of the glacier itself, and, against the advice of his companions he was lowered by ropes into a glacial well, nearly ending his career by his temerity."

Although the stream had been diverted from the entrance, in the crystal blue light that pervades glacier caves Agassiz found himself lowered into a bath of ice-blue still water 40m below the glacier surface, and his cries were initially misinterpreted as instructions to lower him further. As he was hoisted to the surface he was drawn through the clustered points of numerous very large and unstable icicles that had presented a lesser obstacle on his descent. His important observations regarding structures in the ice exposed in the cave walls nevertheless highlighted the potential for scientific insight to be offered by glacier caving. However, Agassiz later reflected of his explorations "I should not advise anyone to follow my example" (Baily 1982). In 1898 Vallot explored a shaft in the Mer de Glace that branched at a depth of 55-60m into a horizontal system (Stenborg 1968). In later years H. Carol (1945) descended into a moulin in a bid to reach the glacier bed but was thwarted by a constriction and water. He later indicated that he was no longer in a position to continue his efforts, but would be delighted if others did so.

More recently, Antonini (1991) has recorded the exploration of a number of moulins up to 140m deep in Pakistan, some of which contained individual vertical pitches of up to 100m length: exploration was again halted by water or constrictions. The most concerted efforts have been in the USA where members of the Cascade Grotto of the National Speleological Society (NSS) have made many trips to document and study the Paradise Ice Caves on Mt Rainier, the stalwart of these ventures being Charley Anderson who was involved in more than 200 visits. Up to 25km of passage has been mapped in this system (Halliday 1979).

Several reasons probably underlie the limited attention given to glacier caves by speleologists. One is probably a perception that the caves are such transient features as to hardly be worth bothering about. However, most landforms are transient, viewed over one time frame or another. Various factors dictate that some components of a subsurface glacier drainage system remain relatively fixed in space; they merely wax and wane in magnitude in response to changes in meltwater discharge, confining ice pressures, and the glaciological and climatic conditions that in turn influence them. A second impediment may be that the exploration of glacier caves tends to demand alpine skills and experience that relatively few cavers possess - and faced with a fine day and a choice between a mountain and a cave many would chose the former. Perceptions regarding the safety of glacier caves have probably also inhibited the desire to explore. One caver who noted the presence of glacier caves near Mt Cook in New Zealand cautioned that:

"you are pelted with rocks at the entrance, get exposure and/or drown if you slip into the river, drown if rain melts the ice and virtually drown if it's fine and the ice melts"

After observing the outflow cave at Franz Josef Glacier on the opposite side of the Southern Alps he added that there "additional forms of sudden death" included "the prospect of being skittled by ice-bergs" (Shannon 1972). Another caver suggested that the Franz Josef outflow was "the most dangerous cave I have ever set foot in" and summarised the exploration of glacier caves generally to represent a "death wish" (Dunkley 1972). However, the latter writer went on to acknowledge that one stream-sink cave at Franz Josef "looked no more dangerous than many a limestone cave in Tasmania".

The Nature Of Caves And Karst In Ice And Snow

Freshly fallen snow has a density of ~0.1 Mg m-3 and a porosity of 90%; compaction and densification due to the refreezing of meltwater that penetrates the snowpack leads progressively to the transformation of snow into firn (density 0.5 Mg m-3, porosity 50%) and then ice (>0.8 Mg m-3 and 10%). Unless specified otherwise, I shall in this paper use the term "glacier cave" as shorthand to describe caves in all these materials but some important differences exist. The caves that form in snow are generally the smallest while those in glaciers are probably the most extensive, the reasons including the greater mechanical strength of ice compared to snow, differences in permeability and the fact that snow is likely to be transformed to firn or ice if it persists in the landscape for long enough for a substantial cave to develop. Notwithstanding the latter factor, caves probably form more rapidly in snow than in ice, which may partly account for the sometimes disproportionately large size of caves formed in seasonal snow avalanche debris, in conjunction with the mechanical strength of this material attained through firnification. It is important to note that the term "ice cave" has generally been restricted to karst caves, lava caves or other caves in which ice persists for all or much of the year (Halliday 1977; McKenzie 1971); it does not imply a glacier cave. Some confusion is ensured by the fact that the largest glacier caves mapped to date have historically been known as the Paradise Ice Caves! This paper is not concerned with ice caves, but rather with caves actually formed in snow and ice which in this context can be considered to be a metamorphic rock.

The morphological and hydrological analogies between features in glaciers and in true karst can be very close. The main difference is that the glacier features develop through the physical processes entailed in the ablation (melting or sublimation) of ice rather than the chemical processes involved in limestone dissolution (Kiernan 1978a). Perhaps the most comprehensive description of melt-karst in a glacier is provided by Clayton (1964) who described from the Martin River Glacier, Alaska, features analogous to sinkholes, tunnels, caves, sinking streams, blind valleys, large springs, natural bridges, lapies, hums, and residual "soils" or ablation tills. Other general descriptions of melt-karst in glaciers have been provided by Watson (1976) and Kiernan (1980a). Among the most awe-inspiring features are major outflow caves at the snout of glaciers, and moulins, the natural shafts which engulf meltwater streams that flow across the surface of many glaciers.

The few explorers who have taken up this branch of speleology have found that with caution, common sense and experience the hazards are manageable. Despite the very limited exploratory efforts, large cave systems have been explored and documented. The density of cave passage development in snow and ice can be high. For example, the longest glacier cave system so far mapped, the 25km long Paradise Ice Caves, occur in a glacier only 1 km long (Halliday 1976, 1977). For comparison, the longest karst cave so far documented in Australia, Exit Cave, comprises perhaps 20km of passages packed into a 2.3km linear distance between the outflow of the cave stream and the intake of its most distant tributaries, which may represent a reasonable average passage density for well-developed limestone karst. Given the extent of the world's glacier-covered landscapes, the density of cave development that tends to occur within glaciers, the extent of many individual glacier systems and their altitudinal range down the flanks of some of the world's highest mountains, it may ultimately prove that the longest and deepest cave systems on earth are developed in glacier ice rather than in limestone.

McKenzie (1971, 1972), Kiver and Steele (1975) and Kiernan (1978a) have each suggested schema for the classification of glacier caves. In broad terms, the predominant modes of cave initiation and enlargement fall into one of four categories. Caves may exist due to: (1) movement of the host material; (2) melting of the host material due to a flow of heat from surrounding rock materials (generally in response to differential heating by solar radiation or due to the prevailing geothermal heat flux); (3) a flow of meltwater or air of sufficiently high temperature to induce melting; or (4) due to biogenic mechanisms. Air currents and to a lesser extent thermo-erosional wash by water and abrasion by debris entrained in streams are the principle mechanisms whereby proto-caves are enlarged to form cave systems of sufficient size to be penetrated by humans. Caves in the first three categories are of greatest size and are reviewed in the following paragraphs. Caves of biogenic origin apear to be relatively minor and will not be considered further. They include small shelters excavated by mountaineers, and tunnels excavated for scientific research. Some organisms produce small tunnels or tubes within or beneath a snowpack. Such features may serve as foci for the evolution of larger caves by other agencies. Passages in one Tasmanian snow cave appear to have been initiated at a point where soil water under pressure rose up the burrows of yabbies (freshwater crayfish) to the base of the snowpack (Kiernan 1980b).

(1) Caves formed predominantly by flow of the host material

Of those caves initiated by movement of the host material, the simplest result from the bridging of gaps between boulders by falling snow, or where cavities are formed in a host mass through its collapse, analogous to boulder caves. However, probably of greatest interest in this group are caves formed where the host material flows over a protrusion in the bed surface, leaving a cavity on the downstream side. Caves of this latter kind have been reported by several writers, and speleothems deformed by glacier flow have sometimes been observed within them (Vivian and Bocqet 1973, McKenzie and Peterson 1975, McKenzie 1971, Thompson and McKenzie 1979, Masataka Izumi 1982). These three types of caves might be called, respectively, bridging caves, collapse caves and obstruction caves but composites occur, as in the case of crevasses that are roofed by snowfall. In our part of the world caves formed in this manner exist in the glaciers of New Zealand, Antarctica and Heard Island, and form seasonally in the mountains of Tasmania and elsewhere in southeastern Australia.

(2) Caves formed predominanly by heat flow from rocks

Slot-like voids commonly develop along the margin of a host mass due to the warmer temperature of the enclosing rocks, most usually in response to differential solar heating and thermal conductivity. The extent to which the geothermal flux operates as a speleogenetic agent under normal circumstances is open to question. However, in one particular class of caves there is little doubt. Caves formed in volcanic craters have been recorded from a number of sites. The best documented have developed in firn in the craters of Mt Rainier in Washington state, USA, where 2.2km of mostly horizontal passage 7-9m wide has been mapped around the perimeter of the crater, connected to the surface by a number of ascending entrances. The largest chamber recorded was 52m long, 40m wide and 8m high (Kiver and Steele 1975). Some 800m of cave passages have been recorded in the crater of Mt Sherman, also in Washington state, while chambers up to 100m long and 10m high have been explored at Mt Wrangell in Alaska (Kiver and Steele 1975). Caves into which geothermal water discharges or that contain geothermally warmed meltwater constitue an important variant of this class of caves. A small lake has been recorded from one crater firn cave in USA, a loadly roaring fumarole fills one passage with steam and elsewhere there is a spectacular 10m meltwater rapid, hence, warmed air and meltwater flows are also involved in cave enlargement (Kiver and Steele 1975). In Iceland a tunnel 50km long produced by geothermally heated water at the base of an ice cap has been postulated (Nye 1976) and massive flooding has periodically occurred due to the collapse of ice barriers blocking a lake in the crater of one volcano (Thorarinson 1953). Warm air currents are probably the principal mechanism whereby both these interface warming caves and geothermal ablation caves are enlarged.

In this part of the world small interface warming caves are common in areas of long-lying snow, with larger but still simple examples present on the margin of some glaciers. Geothermal ablation caves are known from two Antarctic volcanoes, Mt Erebus and Mt Melbourne. Some 400m of passage has been explored at Mt Erebus where an ice tower 4m high forms at the cave entrance due to condensation (Lyon and Giggenbach 1974). A major cave is known in New Zealand where the warm Mt Ruapehu crater lake discharges into the ice cliffs that surround it (Odell 1955; Gregg 1959; Kiernan 1978b, 1979). Probably the only Australian site where geothermal glacier caves may exist is Heard Island.

(3) Caves formed predominantly by air or meltwater flows without strong geothermal influences

These are the most common glacier caves but only in some cases is their speleogenesis by meltwater unequivocal. Halliday (1979) has suggested that subglacial solifluction may sometimes be involved in the initiation of channels that are then enlarged upwards into the ice. Stream caves fed by meltwater or by streams from tributary valleys develop at the base of glaciers, and often just inside and parallel to their margins (Rothlisberger 1968, Shannon 1972). Most of the drainage channels form in ice but a small proportion are incised into the rocks beneath glaciers (Davies 1961).

Where ice is moving downvalley and the need exists for a relatively permanent drainage route, due to such factors as the location of tributary valleys, channels are cut in bedrock in a quest for permanence (Shreve 1972). These atmospheric ablation caves are enlarged primarily by air currents.

Where the slope of the glacier bed is concave, ice is pushed upover the obstacle on the bed behind it. This compressive flow of the ice closes the joint systems and inhibits their penetration by water and air currents, favouring surface drainage. Features such as ridges and runnels similar in form to some karst karren commonly occur on the ice surface. Highly symmetrical meandering stream channels normally form where structural influences are weak and there is little rock debris. The water flow is rapid, and gorges many metres deep form where compressive ice flow occurs over extensive areas, especially if large tributary streams from valley sides can flow onto the ice surface rather than plunge beneath its margins. Where convexities occur in the bedrock topography, extending flow of the ice accelerating downslope away from the ice behind it causes the opening of joints and promotes crevasse formation. In such situations there is very little surface drainage because streams readily sink into the glacier surface. Because the fractures are very frequent the meltwater catchments are small, and because ice flow is relatively rapid, integrated meltwater caves within the ice are probably very limited. The effects of sun and wind, and melting caused by the differential heating of rocks on the ice, surface add to the complexity of surface forms in both compressive and extending situations. From a caving perspective intermediate conditions are probably optimal.

In addition, in very broad terms major valley glaciers can perhaps be considered as potentially comprising three different zones from a caving point of view. The development of karst-like landforms tends to increase downvalley as melting occurs and the proportion of water flowing as solid ice decreases while the proportion of flow as liquid water increases. At the upper end of a glacier is the neve, essentially a snow-mantled area where the snow undergoes transformation into firn and then glacier ice, and much of the surface is blanketted by a relatively uniform snow cover. Downstream lies the white ice, where for much of the year the hard glacier ice occurs at the surface, broken to varying degrees according to flow characteristics of the glacier. Finally, in some cases where rock debris is brought to the centre of a trunk glacier on the margins of tributary glaciers, where rockfall debris collapses onto the glacier from valley sides, or where collapsing moraine walls surround rapidly downwasting glaciers as around Mt Cook, there may be a zone of debris-covered ice. In the upper neve zone any crevasses or cave entrances are covered by snow for much of the year; entrances are more readily accessible but are subject to greater streamflows in the white-ice zone; in the debris-covered zone there may be many sinkholes and entrances, but blockage by debris is the norm.

The hydraulic gradient of glacier cave drainage systems is controlled by the slope of the ice surface rather than by the bedrock topography beneath the glacier, hence the phreatic loops that occur in some karst caves also have their counterparts in glaciers. Evidence for this is seen in the pattern of eroded channels cut by subglacial meltwater that once flowed up and over rock ridges beneath glaciers and have been exposed after glacier recession. The pattern of subglacial meltwater channels cut in rock in the King River Valley, Tasmania, bears remarkable resemblance to a map of some conventional karst cave (Kiernan 1981). All this has major implications where glaciers flow across limestone, superimposing their hydraulic gradients and powerful water flows, and having the potential to form karst cave drainage systems that are inconsistent with the topography after the glaciers have vanished. Some sinuous ridges of sediment known as eskers that remain in some deglaciated landscapes represent sediments deposited in meandering subglacial caves; micro-versions have also been recorded in caves at shallow depth wholly within glacier ice (Kiernan 1979b). The downvalley movement of glaciers produces joint and fracture systems in the ice that serve as foci for penetration by water and air and resultant cave development in much the same manner as do joints in limestone. Moulins that engulf meltwater from the glacier surface appear to form predominantly at joint intersections, their form often being highly asymmetric where the flow of the glacier is rapid (Stenborg 1968, 1969, Kiernan 1980). Tributary streams from valley margins contibute to the water moving through and beneath a glacier, and impact significantly on the thermal and sediment budgets; the Victoria Falls swallet beside New Zealand's Fox Glacier being one spectacular example. Some subglacial cavities are probably very large — Sara (1974) records the collapse of a 400m wide hole that exposed the bed of the Franz Josef Glacier, just below the main ice-fall, in March 1967.

Segments of horizontal cave passage sometimes occur only a few metres beneath a glacier surface. Gently angled thrust planes that form as the ice flow first begins to be impeded by slowing or stagnant ice ahead, upward stress release through unloading, and/or melting adjacent to darker coloured rock debris on the glacier surface may be involved in the development of caves in this position (Kiernan 1979b). Sinkholes and cenotes are particularly common in the debris-mantled lower reaches of some glaciers (Shannon 1972), the depressions commonly being less asymmetric than further upvalley. Closed depressions more than 1 km long have been recorded (Kiernan 1980a).

Active stream caves in glaciers exhibit many features typical of karst caves, including sinuously meandering stream passages developed wholly in ice, river niches and related features. Many caves exhibit magnificant large-scale scallops up to 1 m across, graphically described as "thumbprint ice" by Charlesworth (1957). These are analogous to the scallops produced by streams in karst caves but in glacier caves are the product of ablation by slowly moving air currents. Some glacier caves exhibit passage cross-profiles that in a carbonate rock karst would be regarded as a classical phreatic tube. Although some of these caves may now carry small streams the impact of stream erosion appears to be greatly overwhelmed by cave enlargement through atmospheric circulation. The zonation and sequential development of cave passages in a glacier aquifer is in a sense reversed relative to a karst aquifer. Whereas karst caves may originate under phreatic conditions beneath the water table as water-filled tubes and be modified and acquire vadose features as drainage of the rock mass occurs, phreatic conditions for a glacier cave that is dependant upon aerogenic processes awaits such time as the cave emerges above the "air table" and becomes air-filled: the atmosphere is effectively the phreas for such a cave, and it is at this stage that phreatic tube-like caves develop in glaciers. Provocative misuse of terminology perhaps, but I think it conveys the situation fairly clearly. Hence, the real implications of the "typical phreatic tube" described by Grinstead (1975) from the ceiling of a very active river cave in Antarctica are unclear. Similarly in question are the speculations about glacier cave passage cross-sections, based on analogy with limestone karst, advanced by Sugden and John (1976).

One demonstration of the activity level of glacier melt-karst has been provided over the last two years beneath Mt Cook (3764 m) on New Zealand's Tasman Glacier. Kiernan (1980a) has previously provided photographs and descriptions of melt-karst in this part of the glacier, the features including runnel networks, surface stream channels, caves, small moulins, and a major surface stream that had formed a small gorge that terminated in a spectacular and intimidating streamsink. On 14 December 1991 a 700m high buttress on the east face of Mt Cook collapsed, initially involving ~14 million m3 of rock but rapidly entraining additional volumes of snow, ice and rock debris. The resultant avalanche descended 2700m at speeds of 300-600kph, and ran for 7.3km. A 2km broad lobe of debris crossed the 2.5km wide Tasman Glacier and surged 70m up the moraine wall on the opposite side of the Tasman Valley. A dust cloud rose 700m and an air blast was felt 5km upvalley at Beetham hut. About 55 million m3 of material was spread across ~4 km2 of the Tasman Glacier, ~20% of the deposit being rock debris of which ~40% was pulverised to silt and sand, forming a carpet across the ice much smoother than the usual debris cover predominant just downvalley (Chinn et al 1992). This effectively swept the slate clean as far as surface melt-karst features were concerned, but inspection in February 1993 revealed the formation within the new debris mantle of conical sinkholes, sparsely scattered but in some cases up to 20m wide and 5m deep.

The exploration of atmospheric ablation caves has been described from several parts of the world, including Washington State (Halliday and Anderson 1970); Alaska (Ubach 1978); Canada (Matti Seppala 1972, Saul 1978); New Zealand (Kiernan 1978c, 1979); Pakistan (Antonini 1991) and Antarctica (Grinstead 1975). It is sobering to reflect that during the late Cainozoic glaciations much larger areas of the temperate world were covered with ice than is presently the case and many major rivers flowed from glacier caves. In many cases the extent of melt-karst and caves in glaciers probably far exceeded the extent of present-day karst and caves in limestone. For example, in Tasmania some 7000 km2 of the Central Highlands alone was ice-covered during the late Cainozoic ice ages (Kiernan 1990) and, asuming just that part of the ice mass below the firn line to be the most karstic, that implies an area of melt-karst at least twice the present extent of limestone and dolomite. At that time, all the state's major rivers emanated from glacier caves, the Derwent River, for instance, discharging from the snout of a glacier 70km long (Kiernan 1985,1992).

Management Issues

Large scale natural hazards

Major floods sometimes result from the rapid drainage of ice-dammed water bodies (Glen 1954). The management significance of melt-karst in glaciers has therefore been firmly driven home by events such as the flood from Tete Rouse Glacier, Switzerland, on 12 July 1892 that killed 150 villagers in their sleep. Major blow-outs from glaciers after prolonged heavy rain have also been reported from New Zealand (Sara 1974). Understanding glacier cave drainage systems may provide an important facet of responding to such natural hazards.

Geothermal ablation caves can also be involved in large scale natural hazards, the caves at Mt Ruapehu in New Zealand providing one example. A barrier of debris formed across the crater outflow during an eruption in 1945 and led to a higher water level than was previously the case when the lake reformed. Some outflow occurred through a small ice cave and probably by seepage under the barrier, which collapsed at 8pm on Christmas Eve 1953. Williams (1986:86) includes a photograph of the huge new cave passage that rapidly developed, its entrance about 50m wide and 30m high. The discharging water picked up massive amounts of ash, rock and other debris to form a lahar with a discharge of about 900 cumecs. This swept away parts of the railway bridge at Tangiwai, 27km downstream, including a 125 tonne concrete pier which was moved 64m. When the Christmas Eve express train from Auckland to Wellington arrived a few minutes later 151 people died (Williams 1986).

Although the broad morphology of geothermal ablation caves can be relatively constant, their volume commonly changes in response to variations in the geothermal flux. Because of this relationship, study of these caves can provide early warning of potentially hazardous levels of volcanic activity (Kiver and Steele 1975). Melt features in the snow within the summit of Mt Taranaki (Mt Egmont) in New Zealand have on at least one occasion been suggested to indicate renewed activity in this volcano, although my impression at the time was that they were conventional atmospheric ablation features and no other evidence of renewed activity appears to have been suggested since Taranaki last erupted about 240 years ago (Neall 1976).

Potential Impacts Of Visitors On Glacier Caves

Many glacier caves are relatively permanent features but their detailed morphology is in a constant state of change due to fluctuating climatic and glaciological conditions. Hence, damage inflicted by visitors may be overwhelmed by natural changes. There are of course limits to the laissez-faire approach to management that might flow from this perspective - heat introduced by fixed lighting systems or large numbers of visitors, or artificial changes to entrance or passage configurations all have the potential to cause major disruption. Halliday (1977) has documented some relevant issues that have arisen in the management of ice-bearing karst caves and lava caves that are equally relevant in glacier cave management. However, in comparison to the sorts of damage that can be inflicted upon karst caves by casual visitors glacier caves are relatively forgiving.

As Halliday (1977) observes, the beauties of great ice caves are in some ways even more vulnerable than those of calcite. However, in many cases there is little reason to restrict visitor numbers for the sake of the caves. Yet even seasonal regeneration may seem little compensation to a visitor who arrives on a once-in-a-lifetime holiday to find that those who were present the previous day have destroyed the speleothems.

Halliday (1977) has proposed that the guiding principle in the management of ice caves and glacier caves should be orientated towards speleo-meteorological considerations, since it is the climate of these caves, notably those factors that give rise to ablation, that is the major speleogenetic process and the cause of ice speleothem deposition and loss. This is of major significance if facilities of any sort are established underground - for instance, in the Paradise Caves, Halliday reports that ice screws placed by the National Parks Service have at times needed to be checked daily due to the rate of ablation.

Potential Impacts Of The Caves On Visitors

If the caves are relatively forgiving of the impacts of visitors the converse may not be the case. Important questions of visitor safety arise. Some of these involve access to the caves as much as safety within them: access management and in-cave management need to be closely integrated.

Management issues on the access route

Both safety and expense can be major considerations in the provision of access routes to glacier caves and careful planning and costing is required. The need to traverse sometimes steep and relatively unstable terrain can imply risks. Rockfalls, landslides, snow avalanche or ice-cliff avalanche may threaten the track or those who use it. Problems can even sometimes arise where routes cross fairly easy terrain such as broad braided stream beds. For example, visitors to the Franz Josef Glacier terminal in New Zealand have sometimes been cut off from returning safely to their cars due to changes in the position of active stream channels, necessitating rescue and/or the provision of emergency by-pass tracks along the valley sides. The stability of the terrain crossed by an alpine track may be such that the need for the cost of constant track repairs must be included in the management budget-those responsible for maintaining tracks in alpine areas prone to rockslide or avalanche will be only too familiar with this situation.

To judge from historical photographs and anecdotal evidence, caves and grottoes in the glaciers of New Zealand appear to have figured in the visitor experiences of places like Mt Cook far more prominently in the past than is the case today. The rapidity of topographic change in the glacierised landscape has probably played a major part in this changing emphasis. In the early 1890s visitors to the Tasman Glacier from the area of the old Ball Hut had to ascend about 10m from the moraine wall onto the ice of the glacier, but such has been the extent of glacier downwasting that a descent of around 100m is now required, down a highly unstable moraine wall that is no longer supported by the ice and is in a state of very active collapse. Similar downwasting has occurred at the Mueller Glacier nearer the Hermitage, restricting easy and safe access by visitors to the ice. Descending these moraine walls can be a hazardous affair and there have been numerous serious injuries; at least one would-be glacier caver has been knocked unconscious by rockfall. Such is the rapidity of moraine wall collapse that huts have had to be removed and the old Ball Hut road, which I travelled by bus on my first visit to the Tasman Glacier in the mid 1970s, later proved unmaintainable, has now totally vanished in places, and elsewhere is dislocated by vertical scarps 30m high. In such circumstances it is hardly viable to establish a safe and economic track or road onto the ice. On the other hand, the imminent advent of ground-based operations by helicopter tourism operators in the Mt Cook National Park will, for better or for worse, greatly change the access situation. With the long-established ski-planes restricted to landing on the soft snows of higher neves, it may well be that areas of ice-surfaced glacier further downvalley represent a market opportunity that will be specifically targetted by helicopter operators - and it is there that attractive and safe caves are most likely to be found.

Similarly, glacier recession has influenced visitor use of Mt Rainier's Paradise Ice Caves. In the early days the caves lay only a couple of kilometres from the Paradise Inn, but the situation has been drastically altered by recession and the caves subsequently documented by NSS were, strictly speaking, not the Paradise Ice Caves but rather caves developed in the now separate Stephens Glacier (Halliday and Anderson 1978). One consequence of recession therefore may be that visitors have to travel further through potentially hazardous terrain in order to reach the caves.

Most glacier caves are located in alpine regions that are subject to rapid weather changes including the speedy onset of severe storms and blizzards. Inexperienced parties that set out in fine weather may not appreciate the risk of hypothermia if inadequately clad, the perils of avalanches, or the hazard of becoming lost should dense fog suddenly descend. Even experienced parties may be tempted to leave behind emergency equipment in fine conditions. The death from hypothermia on the tourist track from Paradise Inn to the Paradise Ice Caves of National Speleological Society member Edith Anderson, wife of the doyen of glaciospeleologists Charley Anderson, highlights the reality of the weather hazard. Particular perils face the inexperienced, and in some cases an alpine guide may be warranted. Only half of the winter trips attempted by the NSS groups managed to reach the Paradise Caves. One of the hazards of utilising glacier caves as tourist assets may be that management authorities may face difficult decisions about facilitating access into areas where the natural hazards are such that under normal circumstances they would prefer visitors did not venture, and visitors may not realise this fact.

Additional hazards may face visitors once they reach the glacier surface. Where terminal outflow caves are the target rather than englacial caves further upglacier it may not be necessary for visitors to ascend onto the ice, but many will probably do so anyway. The usual glacier travel possibilities of falling into crevasses or risks associated with collapsing seracs, ice avalanches, snow avalanches and rockfall from adjacent slopes also apply. Glaciers are made of ice and can be slippery, so falls are a real risk. Should a fall occur into the smooth and slippery channel or gorge of a fast-flowing supraglacial metwater stream it may be almost impossible to avoid being swept into one of the deep waterfall shafts in which many such channels terminate. The downwasting of the glaciers at Mt Cook has been accompanied by a change in the character of the glacier surface over extensive areas. In particular, areas of white ice that were readily accessible in the past are now mantled by considerable thicknesses of rock debris - on the Tasman Glacier, for instance, it is now neccessary to travel some kilometres from the old Ball Hut site before white ice is reached. The moraine-mantled ice surface is tedious and hazardous to traverse, with rocks of sometimes massive size often hazardously tottering on a slippery ice base or jammed with questionable stability across active sinkholes and crevasses. Although caves are present the rocks around their entrances can be a real hazard. Visiting glacier caves under such conditions represents a far more serious undertaking today than was the case in the past, even had the difficulties of access to the glacier in this area not increased over the years.

In addition, the karstified nature of the glaciers may produce weak spots in the glacier surface that are not predictable from crevasse patterns. Kiver (1975) reports the existence of thin snow crusts over vertical drops of up to 50m at entrances to caves in the crater of Mt Baker.

At the Paradise Ice Caves at least two visitors have died after falling through the roof of caves when the glacier surface collapsed underfoot. I have had recent personal appreciation of this hazard when, while looking up at an obviously very thin section of roof of a firn cave nearly 1 km long in the Bealey Glacier, New Zealand, and reflecting on how easy it would be to fall through, I glimpsed through a small hole above my head two tourists who had strayed onto the ice surface from a nearby track. They gave me quite a fright, but probably not as great a fright as they themselves received when in the silence of this lonely alpine valley a voice from deep within the glacier began to shout at them with great urgency!

Finally, it is well to remember that even minor injuries can be a serious problem in remote situations, especially given fickle alpine weather, and the injuries may not be so minor where the glacier surface is steep and perhaps terminates in bluffs. A multitude of particularly alpine hazards may be added to those we are more familiar with in any area visited by tourists - ill-equipped visitors may fall prey to snowblindness for instance - and such problems may create demand for the manager to assist.

Hazards within the caves

The issue of snow avalanche or rockfall on access routes has previously been noted, but it warrants restating in the context of the cave itself. One means of entry to caves beneath ice and snow is via bergshrunds or slots adjacent to valley walls, both of which may be open to falling debris. Entrances commonly also exist where streams from tributary valleys plunge into glaciers. Tributary valleys are also commonly avalanche paths in winter, the very time when the prospect of reduced melting and streamflow may make cave exploration seem most viable. Aside from the risk of a party actually being engulfed by an avalanche at the entrance, or returning from beneath the ice to find their exit blocked, the air blast from an avalanche may well be a real issue. The airblast of a fast moving avalanche can extend 100m beyond the margins of the moving snow mass and exert pressures up to about 0.5 tonnes/m2; low density snow dust may be entrained. Airblast has the capacity to push in doors or remove poorly designed roofs, and has been recorded as causing lung injuries, without any impact of the snow mass itself; channelled through a confined glacier cave it may pose a significant hazard. The impact pressure of an avalanching snow mass itself may range from 5-50 tonnes/m2 (Perla and Martinelli 1976) and probably has the potential to collapse thin-roofed caves beneath glaciers hit by avalanche.

The fact that air currents are the major factor in glacier cave enlargement probably underlies a unique caving hazard posed by glacier caves. In conventional karst caves unloading and exfoliation may produce sheets of rock that parallel the roof and walls of a cave and which eventually collapse. The collapse process is slow, internal stresses in the host mass and gravity being the principal agents. However, in glacier caves the development of a fracture facilitates the entry of air currents which provide an additional and very active agent that encourages collapse. This process may be equivalent to the phenomenon of "flakefall", a term used by Halliday (1977) to describe the development and collapse of long thin slabs of ice that may weigh many tonnes and which gradually separate from the ceiling or walls of caves, eventually peeling off and collapsing to the floor. Their presence can be difficult to detect from some angles and the slabs may give little warning of their precarious stability - although sometimes a creaking or groaning noise precedes their collapse. Primarily as a result of experience in the Paradise system flakefall has come to be regarded as perhaps the major hazard of glacier caves. A death due to flakefall in the Paradise Ice Caves was recorded as early as 1915. Today the National Parks Service declares the Paradise Caves open when it is considered that conditions are safe. In addition, the rockfall hazard around cave entrances can be extreme where supraglacial moraine is present. In many instances volleys of small rocks (and sometimes not so small) slither and fall over the lips of cave entrances every few minutes. A careful watch is generally needed to ascertain the probable location and trajectory of the next volley of rocks before entering caves, but this can be frighteningly impossible to spot from inside the cave when it is time to exit.

The vigour of subglacial streamflows also warrants mention, particularly when melting rates are high in summer. One section of the Paradise Ice Caves has been termed "Suicide Passage" due the stream discharge and the highly mobile car-sized boulders that are swept along it. On the slippery ice floors of englacial caves smaller volumes of water may be quite hazardous.

Rapid changes in water level may also be an issue in some cases. Cave stream discharge commonly rises dramatically after heavy rainfall onto the bare ice of a glacier surface. Far more dramatic and unpredictable rises have been observed. Halliday (1976) records an incident in 1963 when a university party witnessed the rapid draining of some circular ponds on the Martin River Glacier in Alaska, exposing a moulin 100m deep and promoting a jet of water more than 6m high at the glacier snout. The moulin and ponds suddenly refilled at such a rapid rate (estimated at 3.8 million litres/min) that the party had to run for safety. The process was subsequently observed on two other occasions. In contrast to this terrifying tale, despite more than 200 visits by the Cascade Grotto of NSS to the Paradise Caves no sudden water level rises of more than 30cm have been observed. However, even a rise of this magnitude may be hazardous to cavers. Halliday (1976) records the impressions of a witness to flakefall into water just inside a cave in Alaska's Mendenhall Glacier: "a terrific surge of water ... as if a giant were throwing a bucket of water out the entrance".

The rapid evolution of caves in snow and ice means that the configuration of passages is liable to speedy change, new passages developing and others collapsing, hence visitors should not expect any glacier cave to conform totally to previous descriptions or especially to maps. For this reason the prospect of visitors becoming lost may be greater than in karst caves or in lava caves and may constitute a significant management consideration. Squeezing through tight constrictions is often also very much easier where the walls are smooth ice rather than knobbly limestone, but where it includes a steep descent the return journey can sometimes be extremely difficult.

Halliday (1977) has drawn attention to the significance of hypothermia in glacier caves where three factors contribute: wind chill; dripping water and running water. Special thermal protection is required for even minor injuries. On the other hand, glacier caves can also offer life-saving shelter from adverse weather outside. The first party to climb Mt Rainier in 1870 owed their survival to the caves in the summit crater:

"Never was a discovery more welcome! ... a deep cavern, extending into and under the ice, and formed by the action of heat ... Its roof ... a dome of brilliant green ice ... Forty feet within its mouth we built a wall of stones, inclosing [sic] a space 5 by 6 feet around a strong jet of steam and heat ... our clothes froze stiff when turned away from the heated jet ... we passed a most miserable night, freezing on one side, and in a hot, steam-sulfur-bath on the other."

However, Mitchell (1969) cautions that some who have since sought the shelter of these caves have avoided potential hypothermia outside only to suffer second degree burns from the steam within the caves!

Kiver (1975) has reviewed other hazards that face visitors to the Mt Baker crater caves. They include rockfall into entrances on the crater margin, pools of boiling water, flakefall and the collapse of large blocks of ice, and floor sediment that liqifies when trodden upon. Oxygen levels have been found to be very low on some occasions, CO2 levels high, and SO2 levels of up to 57.3ppm have been recorded (Kiver, Snavely and Snavely 1977). Atmospheric conditions in these caves become still more hazardous in winter when some entrances become plugged by snow. My own limited exploration of the main outflow cave from the crater lake of Mt Ruapehu revealed another potential hazard (Kiernan 1979a). The lake temperature is generally in the range 20-40°C but extremes of 10-60°C have been recorded, and it is known to have frozen on two occasions. The water is highly acidic, with a pH of 0.8-1.5 (Williams 1986). After having been stopped by a short vertical cascade of warm, acid water in an ice passage 4m high and 9m wide, I returned the following day to find that the previously still, grey waters of the crater lake were all but obscured by steam, but for occasional huge, yellow upwellings that spread from the centre of the lake across its surface; a hydrothermal event just over two weeks later saw material hurled 600m above lake level. Clearly flow surges and major water chemistry changes are probably the norm at times and conditions in the cave have the potential to change fairly quickly.


It is now 100 years since publication of Russell's classic description of karst in the Malaspina Glacier. Over that century many karst caves have been explored and we have gained much scientific insight into the evolution and management of karst landscapes. That advance has not been matched by improvements in our knowledge of glacier caves and melt-karst. Much remains to be learned that will not only be of intrinsic interest but will provide glaciological insights and the knowledge upon which glacier cave management can be based. It may well prove that the greatest cave systems on earth are in glaciers rather than in the carbonate rocks with which most cavers and speleologists are preoccupied. In some cases inproved knowledge of glacier cave systems can be of significance in responding to some large scale natural hazards.

Notwithstanding the limited interest from cavers, in various parts of the world glacier caves are a focus for park users and tourists, and a body of management knowledge already exists. The difficulties that may arise in facilitating visitor use of glacier caves are far from insurmountable. Despite some alarmist attitudes, many glacier caves are safe and the reward for venturing beneath glaciers can be a very worthwhile and quite unique form of visitor experience. Some visitors to alpine areas will enter glacier caves whether managers want them to or not - how many management agencies have the knowledge and expertise to respond to an accident as effectively as they may wish, or the public might expect them to? In this part of the world visitors are presently dissuaded from entering the potentially hazardous outflow caves at places like the Franz Josef and Fox glaciers, but other glacier caves have been tourist attractions in the past and will be again. New opportunities will be identified, including perhaps helicopter access to areas where attractive and safe caves are likely to occur. The continuing development of adventure tourism, and perhaps even the development of tourism in Antarctica, are other possible strands in the evolution of glacier cave utilisation that will demand expertise in the management of these unique and magnificent caves.


ANTONINI, d.G 1991 I "mulinelli" del Biafo. Speleologia 24: 28-32

BAILEY, R. H. 1982 Glacier. Planet Earth Series, Time-Life Books, Amsterdam.176pp

CAROL, H. 1945 Uber einen versuch den Gletscher unterground mittelst Einstiegs durch ein Strudelloch zu erreichen. Die Alpen 21 Bern

CHARLESWORTH, J. K. 1957 The Quaternary Era With Special Reference to Glaciation. Edward Arnold, London. 2 vols

CHINN, T.J., McSAVENEY, M.J. & McSAVENEY, E.R. 1992 The Mount Cook Rock Avalanche of 14 December, 1991. Information brochure, DSIR Geology & Geophysics. 6pp

CLAYTON, L. Karst topography on stagnant glaciers. J. Glaciol. 5(37): 107

DUNKLEY, J. 1972 Postscript on glacier caves. Aust. Speleol. Fed. Nl. 56: 78

FORD, D.C. (ed.) 1983 Castleguard Cave and karst, Columbia Icefields area, Rocky Mountains of Canada: A symposium. Arctic & Alpine Research 15 (4). 149pp

GIGGENBACH, W.F. 1976 Geothermal ice caves on Mt Erebus, Ross Island, Antarctica. NZ J. Geol. & Geophys. 19(3): 365-372

GLEN, J.W. 1954 The stability of ice-dammed lakes and other water-filed holes in glaciers. J. Glaciol. 2(15): 316-318

GREGG, D.R. 1959 Volcanoes of Tongariro National Park. NZ Geological Survey

GRINSTEAD, M.J. 1975 Caving in Antarctica. NZ Speleo. Bul. 5: 413-414

HALBERT, E. & HALBERT, S. 1972 Meltwater caves in the Snowy Mountains. J. Syd. Speleol. Soc. 16(8): 223-224

HALLIDAY, W.R. 1976 Beneath the Ice: The story of glacier caves. Chap 17 [in] Depths of the Earth. Harper & Row, New York. 432pp

HALLIDAY, W.R. 1977 Basic considerations in the management of ice caves and glaciaer caves. National Cave Management Symposium Proceedings, Big Sky, Montana: 81-84. Adobe Press, Alberquerque, Montana

HALLIDAY, W.R. 1979 Glaciospeleology. Caving International 4: 31-34

HALLIDAY, W.R. & ANDERSON, C.H. 1970 Glacier caves: a new field of speleology. Stud. Speleol. 2(2): 53-59

HALLIDAY, W.R. 7 ANDERSON, C.H. 1972 The Paradise Ice Caves, Mount Rainier National Park, Washington. published by the authors. 27pp