What makes a volcano dangerous? Clearly, the severity of any eruption plays a role. So does the presence of people nearby. But it is not always the best known volcanoes that are the most dangerous. Tseax is hardly world-renowned, but it caused a major volcanic disaster in Canada. And sometimes a volcano can be dangerous without actually erupting. Lake Nyos in Cameroon is a well known -and feared- example.
Over time, we have become much better at managing volcano hazards. Accidents can still happen, but timely evacuations have helped tremendously, and sometimes even mitigation is possible. Among European volcanoes Vesuvius and Etna are known to be particularly dangerous, but there have been few casualties since the 17th century. But one event is invariably overlooked in these statistics. Few people have heard about it, even though it happened as recent as this century, and in spite of this being the largest volcanic disaster in Europe since 1669, and the second-largest world-wide since Pinatubo. And we still do not fully understand what happened, how it happened, and what role the volcano played.
It is an impressive mountain in an impressive region. The peak lies at the centre of the Caucasus range, at the border between Russia and Georgia. At 5047 m (16,560 ft), it is the second tallest volcano in Europe. According to a Georgian legend, a cave high on the mountain contains various relics, including Abraham’s tent and the manger of baby Jesus. The shape of the stratovolcano is rugged, reflecting a long history of erosion, a volcano past its prime. Ancient lava flows have spread some 15 kilometer, mainly to the south and filling in the ancient river valleys. Like Elbrus, it is an extended volcanic centre which has been in the same location for half a million years. And like Elbrus, this is not just a relic: there is still life left in it.
Dating of the various flows and volcanic centres show four phases of activity, separated by long times of dormancy. The oldest phase was 400-430 thousand years ago: the lavas from that time are mafic. The Caucasus upthrow fault appears to have provided a magma conduit from the mantle at that time. The next phase was not until 200-250 thousand years ago; now it produced lavas that were andesitic. The faintly visible caldera, 5 kilometers across and surrounding the current peak, dates from this time. The third phase occurred 90-120 thousand years ago, with andesite and dacite lavas. The magma was evolving over time, and the direct conduit to the mantle was closing. At this time the mountain reached its largest size. The current phase started 50 thousand years ago and appears to be still continuing. Distant ash covering Neanderthal sites in Russia has been chemically identified as coming from Kazbek, indicating a large eruption happened here 40,000 years ago, perhaps at the start of this phase.
The central peak seems to be inactive: it hasn’t erupted for perhaps 100,000 years. Instead, the eruptions are coming from satellite peaks, which form an arc to the south of the summit. The Lesser Tkarsheti peak erupted some 6000 years ago and this is most recent confirmed eruption. One lava flow is dated 750 BC but this is unconfirmed. Widespread hot springs show that the heat is still there, underground. These are mostly situated on the flanks, around 3 kilometers altitude.
The summit region is covered by glaciers, but the mountain is rather steep to hold on to its snow and ice. Glacier collapses are not unusual. The Kazbek area has 12 main glaciers. Of these, three are known to cause collapse flows, the Devdorak, Abano and Kolka glaciers.
The first recorded example is from the 18th century. To the south of Kazbek runs the Georgia military road, later used for the Russian conquest of the region. It runs along the Terek river, and has suffered several times from blockages caused by ice flows coming down the mountain. On June 19, 1776, a thunderstorm occurred after a period of hot weather, and during the storm a large ice-debris flow came down, which blocked the Terek River for three days. A temporary lake formed, flooding several villages and drowning both villagers and their cattle. When the Terek River broke the ice dam, it flooded the valley downstream, destroying bridges, dwellings and crops.
Similar episodes occurred during the 19th century, but mainly as mudflows: the ice-debris flows did not recur. The reason appears to be the retreat of the glacier. It no longer reaches the site where the glacial collapses occurred.
Glaciers are known to surge, when the ice flow suddenly speeds up, and the glacier covers distances in days that otherwise would have taken months. The surges can bring the ice forward by hundreds of meters or more. The speed at which a glacier advances is set by the pressure from behind, caused by the weight of the snow and ice, and by the friction between the ice and the ground. If melt water lubricates the bottom of the ice, or worse, so much water collects that it physically lifts the ice, the pressure from behind is suddenly pressing against an open door. Surges are unstoppable, and can move ice walls tens of meter high. But they are slow enough that people can safely evacuate. But if the ice were to break up, and mix with the water, and especially if this happened on a steep slope, the results could be more dangerous. Perhaps this is what happened on that day in 1776. Certainly the thunderstorm and the previous hot weather suggests that water, both rain and melt, played a role.
The Kolka glacier lies in between Kazbek itself and Mount Dzhimarai-Khokh, the second highest peak in the Kazbek formation, 4780 meters tall and 10 kilometers to the west of Kazbek. The blackened glacier starts at a little outside the old caldera, in a cirque: a semicircular basin with near vertical edges. Cirques are a common feature at the head of a glacier: they form by erosion of the rock sides, forming deep basins with vertiginous sides.
The cirque at the head of Kolka is located at an altitude of about 3500 meters. Far above the cirque are the glaciers coming down from the peak of Dzhimarai-Khokh: they forming overhangs which seem in perpetual danger of collapsing unto Kolka. From the cirque, the Kolka glacier follows a short valley before it joins the Maili glacier, in a larger valley which descends towards the north. This becomes the Genaldon river. There is a small village down the valley, Karmadon. Shortly after the village, the river enters a narrow canyon with the appropriate name of Gates of Karmadon.
The Kolka glacier has history: there is something inherently unstable about it. In 1969, the glacier began to surge. It pushed forward meters to tens of meters each day. Initially, it had been 3 kilometers long. By the time the surge ended, in 1970, it had grown to 7.5 kilometers. During the surge, it did not detach from its source: the surge was fed by thinning of the ice up-stream.
But Kolka can also do much worse. In 1835, the original town of Genal, sharing its name with the river, in the valley above Karmadon, was destroyed. It is not known what happened, but in hindsight we can make a guess. In 1902, a double disaster occurred. On July 3rd, an ice/rock slide came down the valley which reached a distance of 11 kilometer. It started when four hanging glaciers fell on the glacier below. Apparently, the entire distance only took four minutes! This indicates a speed of flow of 180 km/hr: this was no mere surge. Three days later, an even larger slide came, caused by two further glacial falls. 36 people died, both at Karmadon, and at Werkhni Karmadon where a thermal bath house had been constructed. The flow had reached 100 meter depth, with blocks thrown up 140 meters into the air.
The 2002 disaster
The first indication of the looming trouble was on Sept 10. Mountain climbers found that there had been mudflows along the upper river valley, with trail and the river banks washed out. At the top of the glacier, they saw evidence for rock and ice falls: the steep wall of Dzhimarai-Khokh was unstable over a kilometre of length. Three lakes had formed along the upper river valley. All this indicated that there was a considerable amount of water building up. It should be noted that the climbers had not been there for two years, and the mudflows and rock falls could have happened at any time in that period.
The disaster struck on Friday night, Sept 20, at about 8 pm. Images taken a few weeks later showed that a section of the steep wall of the cirque had come down. The scar had a volume of four million cubic meters, a mix of rock and ice. In the years before the fall, these rocks had been covered by hanging glaciers. They were no longer there. The material had fallen on to the Kolka glacier from a height of 900 meters above it. The Kolka glacier has a gradient of 5-10% over a length of 3 kilometer. The force of the impact and rebound detached the glacier from the rock wall and from the ground. A volume of 80 million cubic meter of ice began to slide downhill, containing up to 75% of the total volume of the glacier with a thickness of 50 meters.
From the subsequent events, we can infer that the glacier had been held in place by static friction. The impact briefly detached the glacier from the ground and when it re-established contact, within seconds, it was a broken mass, beginning to move sideways. Moving material is governed by dynamic friction which is less than static friction. Because the glacier was already lubricated by extensive melt water, the friction was now insufficient to stop the movement. The glacial mass began to accelerate down the steep slope. This is already unique. Rock falls of this volume and height on glacier are rare but not unique: such falls have occurred elsewhere. But those falls didn’t make their glaciers detach and move. And if this was already unique, what happened next was extraordinary.
The broken ice mass splashed over the moraine opposite the wall, but most was redirected by the moraine down the glacial valley. The wave moved from left to right, leaving marks high on the surrounding rocks, before reaching the Maili glacier. The flow kept accelerating as a wave. The wave front came to resemble a flash flood of mud, ice and rock; it was followed by a second wave with even more debris.
The flow now had to make a 90 degree turn. At this place, it left a mud mount 250 meters above the glacier. Seeing how fast the flow would have had to go to reach that kind of height in the turn gives a minimum speed of 250 km/hr, and possibly up to 300 km/hr.
How could the flow reach such enormous speeds, and maintain them, on slopes no steeper than that found in a road tunnel? Over the glacier, it can have been the effect that makes ice skaters reach high speeds on a solid surface: high pressure on ice can create a thin, almost frictionless layer of pressure melt. In the valley, it must have been the sheer mass of the mud which carried the flow. A quick calculation shows that a speed of 300 km/hr can be reached in free-fall when falling 700 meters. The detachment occured at 3500 meters, and the joining with the Maili glacier is at 2700 meters. That means there is only just enough vertical fall to reach those speeds, assuming the flow was frictionless. The original impact on the glacier may have contributed but it had a volume of only 10% of the total, so it cannot have contributed very much. Perhaps only a fraction of the flow reached the fastest velocities needed for the height of the mud mounds. Air resistance could have been overcome by the sheer mass of the flow, or the U-shape of the valley caused the air to be pushed ahead, forming a blast wave.
From here on, the Genaldon valley is almost straight but the flow kept moving from side to side, still reaching 200 meters above the valley. The flow now slowly decelerated: when it reached Karmadon, the speed was down to 100 km/hr. But it remained dense: nowhere were trees thrown up by the flow, as happens in snow avalanches. Imagine what this would be like. The first thing to hit you is the gale-force air blast wave, almost immediately followed by a category 4 hurricane, where the wind is made of mud and water and the rain is a horizontal waterfall consisting of rocks and ice.
The seismographic record shows that from the collapse of the glacier to the destruction of Karmadon 18 kilometers down-stream, the event took only 5.5 minutes, which makes out the average speed over this distance as 200 km/hr. This includes the initial acceleration and the deceleration in the Genaldon valley! The numbers are absolutely frightening. And it maintained these speeds even though the gradients aren’t that high. This was not a steep mountain slope: it was river valley where people could live.
At the entrance of the gorge, the flow became severely restricted. The ice dropped out and formed a black barrier. The residual went on as a mud and fine-grained debris flow. It continued for another 15 kilometers. There had been flooding earlier that summer, from rain storms, and the debris from that became entrained in the flow. The flow was now down to 10 meter thickness and the speed was a leisurely 40 km/hr. Finally, it petered out, perhaps 35 kilometers from its origin. Over the next days, a shallow lake formed behind the ice dam which flooded much of the area.
There is little in the way of eyewitness reports. One couple mentions a ‘white cloud had come through the dark’. Another person heard a noise like heavy construction machinery. But otherwise, it seems to have come out of the night without warning (although it was almost full moon), and for such an enormous event, it was remarkably silent. The noise was absorbed by the flow itself. It took time for the news of the devastation to come out. The destruction of the power lines would not have helped. Authorities in Moscow were informed only on Saturday morning.The search-and-rescue started shortly after but there was little left to be rescued.
The big danger in snowy mountains comes from avalanches, which can reach speeds almost as high. But this was no avalanche; the density was ten times higher than that of snow and the slope much less than that needed for avalanches. The closest analogue is a pyroclastic flow, when a frictionless mass of fragmented rock, hot ash and possibly water is blown out at high speed. Of course, another name is needed when the mix includes cold ice rather than hot ash: let’s call this a cryoclastic flow. It is just as deadly as its hot counterpart.
The Kazbek event killed around 70 people around Karmadon, and a further 50 died in the mud flood below the canyon. One of the casualties was a actor of national fame, Sergey Bodrov Jr., who had been filming in the canyon. All but four of his film crew also died. Only about 20 bodies were ever recovered from the disaster. The actual death toll may have been higher than what is reported: there are reports of refugees having settled in the upper valley, unregistered. But even a total of 125 would be the second largest number of fatalities from a volcano since Pinatubo, and it could be the largest number for an European volcano since the 17th century.
Or no collapse?
An alternative view of the disaster is that the initial collapse did not occur, at least not on the Friday night. This was argued by Stephen Evans et al. (see Sources, below), who used a satellite image taken 8 hours before the collapse to show that the rock had already gone by that time. They point out that the seismograph had shown no clear signal of an initial impact, and argue that the rock wall had fallen down over months, and that the slide started because of the increasing weight pushing on the water underneath the glacier which pushed the front over the constraining toe of the ice. Apart from the onset, their reconstruction of the event is similar to that presented here. They find a larger volume of moving ice (130 million cubic meters) and a much thicker initial glacier (over 100 meters in depth) but similar velocities. The initial acceleration is found to be 0.8 m/s2. For the slope of the glacier, that corresponds to almost frictionless acceleration. They fit a friction coefficient of 0.05, which is only a little higher to that obtained in the olympic sport of curling, and a factor of ten above that of ice skates.
Their points are well made but not fully conclusive. The continuing rock falls reported several days later suggests that the wall had been severely destabilised, consistent with a larger fall. The impact could have been muted by the water under the glacier. The splashing sideways over the moraine is more consistent with a more energetic start. it is also difficult to see how water can lift ice over a toe while maintaining a very high pressure: one would expect to water to spurt out, leaving the glacier grounded. But neither is their model excluded. The truth may be in the middle: much of the rock had already collapsed, but the final push came from a further rock fall. However, their assessment of a ‘standing start’ (no forward push from the impact) agrees with my own conclusion, I was happy to see.
What makes the Kolka glacier so unstable? Notably, it is one of three glaciers on Kazbek to show slides and collapses, although not all do. But no other glacier in the Caucasus, other than on Kazbek, shows this behaviour. What is unique about Kazbek? Did volcanic activity play a role? This is disputed, but the fact that these cryoclastic flows are only seen in a volcanically active region ,and start at an altitude where hot springs occur, is suggestive.
The name of the village contains a clue. Karmadon means ‘warm water’. There are a variety of hot springs in the region. When the Kolka glacier began to retreat in the late 1800’s, one such spring surfaced from underneath the tip of the glacier. The opportunistic locals quickly build a bath house and huts around the sulphurous site. This was the bath house that was destroyed in 1902. In 2002, the first team to land (by helicopter) in the source region after the event had to leave in a hurry, overcome by the overpowering sulphurous smells. There clearly was strong geothermal activity in the region. The smoke seen in satellite images has also been interpreted as due to this, although this may also have been due to continuing rock falls.
However, heat underneath the Kolka glacier cannot have caused the collapse of the rock wall and overhanging glacier 900 meters above it. The geothermal regions are mainly around 3-3.5 kilometer, while the summit of Dzhimarai-Khokh is considerably higher than this and is itself not known to be volcanic. So the situation is more complex than this. A plausible suggestion is that the collapse was a natural, inevitable event, caused purely by the build up of the overhang, perhaps aided by the warming during the 20th century weakening the rock. But the completely unexpected effect on the Kolka glacier was made so disastrous because of melt underneath it. The whole glacier was already being lubricated, and the purchase to withstand the impact from above just wasn’t there. Sloping ice and ground heat are an unstable combination in any case. The unstable overhang became a sword of Damocles hanging over a powder keg. The collapse came just at the wrong time, when the glacier was already primed to go into surge mode.
If a glacier grows in cold conditions, where the bottom is ice, the friction can carry a considerable amount of weight. If now the bottom begins to melt, the weight carrying capacity is less, and the glacier may suddenly become unstable. This has been seen in ice falls in the Alps, although such glacier collapses there happen on much steeper slopes (some 35 degrees) than seen on Kolka (10 degrees).
Could it happen again? The satellite images show that the newly-grown overhanging glaciers have some deep crevasses. The danger that they will fall is certainly still there. The rock wall is a kilometre long and a collapse could happen at any location along it. The Kolka glacier was almost entirely lost in the event. The cirque has been refilling with snow and ice and we don’t know whether the glacier is beginning to reform. This region of the Caucasus is perhaps not particularly safe for scientists to visit and that could be hindering the assessments of the situation. It is an uncomfortable situation, with a clear danger but too little knowledge of the current state of affairs.
Could it happen somewhere else? Perhaps: with the current warming, meltwater can make both glaciers and rocks unstable. Rock falls have become common in the Alps, especially around the level of declining permafrost, at altitudes around 3 kilometer. They have reached sizes well over 100,000 cubic meters. An event similar to the Kolka collapse is not impossible, and Kolka has shown that a moderate slope does not preclude a cryoclastic flow.
In fact, there have been events with notable similarities. In 1962, part of the ice cap of Huascaràn in Peru broke off, and fell down a kilometer. It set off an avalanche of ice, snow and rock which traveled at 100 km/hr over 15 kilometers into the valley of the Rio Sante where 4000 people died. 8 years later, it recurred and this time the flow reached a city, causing 20,000 deaths. The Kolka flow was three times faster and came as a wall of ice instead of snow. If such a cryoclastic flow would happen in a densely populated area, the magnitude of the disaster could be inconceivable. The chance may be small – but it is non-zero. And the dangers may be growing: have a look at this ice fall. We don’t yet know how to recognize precursors. If it begins with increasing stress, the best way may be with monitoring ice quakes, but seismologists tend to remove those from their data!
One thing is clear: volcanoes do not need to erupt in order to be hazardous. An unstable environment can turn any dormant volcano into a clear and present danger. And that is well worth remembering.
Albert Zijlstra, February 2017
Collapse of the Kolka glacier, Rebecca Lindsay, 9 Sept 2004
Glacier and debris flow disasters around Mt. Kazbek, Russia/Georgia. S.S. Chernomorets et al. 2007. http://www.glacier-hazard.narod.ru/pdf/Ch11_1.pdf
Major events in evolution of the Kazbek neovolcanic center, Greater Caucasus: Isotope-geochronological data. V.A. Lebedev et al. 2014, Doklady Earth Sciences 458, pp 1092–1098
The Kolka-Karmadon rock/ice slide of 20 September 2002. W.Haeberli et al. 2004, Journal of Glaciology, Vol. 50, 553
Catastrophic detachment and high-velocity long-runout flow of Kolka Glacier,
Caucasus Mountains, Russia in 2002. S. Evans et al. 2009, Geomorphology Vol. 105, pp 314–321