Western Tibet – Massive collapse of 2 Glaciers in 2016 following Surge-like Instability

Australia: The Land Where Time Began

Surges and glacier avalanches are expressions of glacier instability, and are among the most dramatic phenomena that are known from the mountain cryosphere. The catastrophic collapse of a glacier, that combines the large volumes of surges and mobility of ice avalanches has, until now, been reported only for the detachment in 2002 of the Kolka Glacier in the Caucuses Mountains, totalling 130 x 10⁶x m³, which has been considered to be a globally singular event. In this paper Kääb et al. report on the similar detachment of the entire parts of 2 glaciers adjacent to each other in western Tibet in July and September 2016, which led to an unprecedented pair of giant ice avalanches of low-angle with volumes of 68 ± 2 x 10⁶ x m³ and 83 ± 2 x 10⁶ m³. Kääb et al. found, based on satellite remote sensing, numerical modelling and field investigations, that the cause of the 2 collapses was climate-driven and weather-driven external forcing, which acted specifically on specific poly-thermal and soft-bed glacier properties. Surge-like enhancement of driving stresses and massively reduced basal friction which was connected to subglacial water and fine-grained bed lithology, factors that converged to produce the enhancement, to eventually exceed collapse thresholds in resisting forces of the tongues frozen to their bed. It was found by Kääb et al. that large catastrophic instabilities of low angle glaciers can happen under rare circumstances without historical precedent.

Sudden mass failures of glaciers, that were gravity driven, have been observed over a wide range of magnitudes, from ice falls at steep glacier fronts, to large ice avalanches when hanging glaciers, typically with angles steeper than 30^o, detach (Faillettaz, Funk & Vincent, 2015; Huggel, 2009). Impacts are mostly felt within a couple of kilometres, with the exception of when the ice transforms to a mass flow, which is highly mobile, by ingestion and production of meltwater and the incorporation of sediments (Evens et al., 2009; Evans & Delaney, 2015). Topographic, atmospheric and ice-thermal conditions (Faillettaz, Funk & Vincent, 2015; Huggel, 2009),and instabilities of underlying bedrock or seismic events (Evens et al., 2009; van der Woerd et al., 2004) are typically included among the factors which lead to pre-failure conditions and triggering.

According to Kääb et al. glacier surges are a 2^nd process of instability of a glacier, which refers to events that last for weeks to a few years that have flow speeds that are abnormally high, that can reach up to 10s of 100s of metres per day over large parts of glaciers (Harrison & Post, 2003; Yasuda & Furuya, 2015; Harrison et al., 2015; Sevestre & Benn, 2015; Murray et al., 2003). Such surge-type glaciers are known of in clusters from mountainous regions around the world, including the Tibetan Plateau (Yasuda & Furuya, 2015; Sevestre & Benn, 2015; Jiskoot, 2011).

Included among the processes that are involved in the lowering of basal friction surging are water pressure that is abnormally high (Harrison et al., 2015; Jiskoot, 2011; Fowler, Murray & Ng, 2001; ), change in the thermal regime (Murray et al., 2003; Jiskoot, Sevestre & Benn, 2015;; 2011; Frappe & Clarke, 2007) and complex responses of the rheology of subglacial till to increasing shear stress and input of water (Harrison & Post, 2003; Harrison et al., 2015; Jiskoot, 2011; Fowler, Murray & Ng, 2001; Kamb, 1987; Truffer et al., 2000; Clarke et al., 1984).

A 3^rd type of glacier instability involves larger sections that detach from low-angle valley glaciers. The Kolka Glacier event, the Kazbek massif, Caucuses is the only such process that had previously been documented in 2002, when 130 x 10⁶ m³ formed an ice/rock avalanche that travelled 18 km down the valley claiming 120 lives (Evans et al., 2009; Huggel et al., 2005; Haeberli et al., 2004). This picture was changed by the massive glacier collapses in 2016 in Tibet (Tian et al., 2017) and opened up critical questions about the causes of detachments and the potential for similar events to occur elsewhere. In this paper Kääb et al. describe the twin events in Tibet and reconstruct the evolution of the collapsed glaciers since the 1960s, based on remote sensing and mass-balance modelling. Also, for more recent years they analysed glacier dynamics and modelled the thermal conditions to infer details of the stress on the glacier and frictional regime. Here they discuss the influences of melt and precipitation, lithology of the bed and geometry of the glaciers on the collapses.

From glacier collapses to high-speed avalanches

A massive volume of glacier ice detached from the lower part (5,800-5,100 m above sea level) of a glacier that had not been named in the Aru Range (Rutok County, China) in the western Tibetan Plateau, on 17 July 2016, 11:15 Beijing time, which has been termed Aru-1 for the glacier and the collapse. The fragmented ice mass ran out 6 km beyond the terminus of the glacier, and killed 9 herders and hundreds of their animals, reaching the Aru Co lake (Tian et al., 2017) (⁓4,970 m above sea level). The avalanche ran for 8.2 km and its vertical path of 800 m yield an angle of reach that is surprisingly low of only 5-6^o (mobility index fahrböschung (Heim, Bergsturz & Menscheneben, 1932)), which indicates very low basal friction in the movement. An area of 8-9 km² was covered by the debris and Kääb et al. used pre/post collapse differencing of digital elevation models (DEMs) to calculate a volume of the detached part of the glacier of 68 ± 2 x 10⁶ m³. An impact wave was generated by the avalanche which inundated the opposite shore of Aru Co over a stretch of shoreline measuring 10 km, and extended up to 240 m inland and 9 m above the level of the lake.

A second glacier (Aru-2) detached 2.6 km south of the July event on 21 September 2016. This second detachment occurred at 5,800-5,240 m above sea level in 2 main flows at about 5:00 and 11:20 Beijing time. The glacier mass fragmented and transformed into a mass flow that ran out 5 km beyond the terminus of the glacier, reaching 4,965 m above sea level in a similar manner to the July event. The vertical height of the path was 830 m over a horizontal distance of 7.2 km, which gave a similarly low fahrböschung of 6-7^o. In the 2^nd collapse the glacier debris covered 6-7 km² with a volume of the detached glacier of 83 ± 2 x 10⁶ m³.

The geomorphology of the avalanche paths and deposits, based on eyewitness reports, and observations of seismic waveforms that were generated by the mass movements all suggest that the events occurred suddenly and rapidly, with a duration of 2-3 minutes at a mean speed of 30-40 m/sec). The analysis of the 3-D forces inverted from the glacier avalanche seismic signals reveals basal friction coefficients that were exceptionally low, of 0.11 ± 0.07 for Aru-1 and 0.14 ± 0.05 for Aru-2 (ratio between frictional and normal force). According to Kääb et al. these values are exceptionally smaller than values on the order of 0.2-0.6 that had previously been found for landslides. Kääb et al. found average flow speeds of ⁓20 m/sec, using different avalanche flow models (Christen, Kowalski & Bartelt, 2010; Hungr & McDougall, 2009), which peaked at 70-90 m/sec in the gorge sections of the paths. The presence of a grassy understorey vegetation on the lee side of hills in the avalanche path of Aru-1 that had not been destroyed also indicated that the avalanche speeds were high, which suggests that the fast-moving mass had partially jumped over it (Huggel et al.,2005). Processes of ice liquefaction from frictional heating during the collapse and flow are suggested by the high speeds of Aru-1 and Aru-2 avalanches (Schneider et al., 2011). Kääb et al. suggest that up to 2-3% of the ice could have liquefied if all potential energy was consumed, based on simple energy conservation, with the water that was generated being concentrated at the bottom of the flow.

Crevasses that were very similar to the ones preceding the Aru-1 event were detected on satellite imagery prior to the collapse of Aru-2, the potential avalanche run-out geometry and general pattern of deposition were modelled from the Aru-1 simulation, and an alert was issued to Chinese authorities.

Climate and glacier changes 1961 to 2016

Kääb et al. investigated recent climate and meteorological records in the Aru region in order to identify what possible pre-conditions and triggers could have caused the 2 glacier collapses. The closest long-term record was provided by the Shiquanhe meteorological station, and these records revealed strong warming of 1.7 ± 0.5^oC since 1965, which is in line with length of the glaciers in the western Tibet Plateau (Yao et al., 2012; Ye et al., 2017). A long term trend in precipitation at Shiquanhe has not been found, though in 2013 and 2015 there were exceptionally large precipitation sums, and in 2016 at the Nagri station, which was established in 2010.

Kääb et al. found by using Corona satellite imagery for the period 1961-1980, Landsat from 1982 onwards, and high resolution optical and radar satellites from 2011 to date, that the glaciers in the Aru Range retreated from the 1970s onwards; Aru-1 glacier retreated by 520 m and the Aru-2 glacier by 460 ± 15 m between 1970 and 2015 or 2016. Contrasting with this frontal retreat, it has been found by analysis of DEMs from STRM-X (2000), TanDEM-X (2011-2016) and Advanced Spaceborn Thermal Emission and Reflection Radiometer (ASTER) (Brun et al., 2017), as well as ICESat laser altimetry (Kääb et al., 2015) (2003-2008), that glaciers in the wider Aru region are part of the Karakorum, Kulun Shan and Eastern Pamir anomaly (Brun et al., 2017; Brun et al., 2017) and experienced a slight increase in thickness of between 0.20 ± 0.16 m/yr in the ASTER and 0.28 ± 0.15 m/yr in the ICESat, water equivalent (w.e.) since the early 2000s. Steepening of the surface of the glacier is caused by simultaneous thickening at high elevation and the thinning at low elevation, which in generally not typical. For glaciers of the non-surge type, however, it has been observed when there is an increased temperature and snowfall simultaneously (Berthier et al., 2010) that causes increased ice and melting and accumulation. Specifically, non-surging glaciers in the Aru region thickened by up to 0.40 ± 0.15 m/yr above an elevation of about 5,650 m above sea level and thinned up to 1.20 ± 0.15, m/yr below that elevation. In the range north of the Aru Range for at least 5 of 16 glaciers this overall climate-driven pattern of surface elevation changes was interrupted by mass-type redistributions from 2000 to 2016.

Mass-balance modelling of the Aru glaciers confirms positive mass balance between 1995 and 2008, and steepening of the mass-balance gradient, despite regional warming. Kääb et al. attributed this to an increase in precipitation since the mid-1990s (Tao et al., 2014; Kapnick et al., 2014) captured by the ERA-interim reanalysis that drives their model. Widespread growth of the endorheic lakes of the region during the same period confirms this increase in precipitation (Zhang et al., 2017).

Searching the archives from Corona (from 1961) and Landsat (from 1972) Kääb et al. found no indication of glacier collapses having occurred earlier in the Aru Range as well as its wider region (radius of about 300 km). Kääb et al. noted however, that early images from Corona and 2015/2016 satellite data for the 2 Aru glaciers and some of their close neighbours show similar crevassing that is strong and bulged tongues; i.e., features that commonly presage surge-like instabilities or rapid advance.

Pre-collapse mass distribution within polythermal glaciers

In the period immediately before the collapse, 2011-2015, patterns of elevation changes in the glacier surface are measured from repeat TanDEM-X DEMs and optical satellite stereo DEMs.

The sections immediately above the eventual detachment zones of both Aru glaciers at ⁓5,800-5,400 m above sea level bulged upwards. As early as 2011, at least, and until 2014, the sections immediately above the eventual detachment zones at ⁓5,800 m of both Aru glaciers subsided. Simultaneously glacier sections ⁓5,800-5,400 m above sea level bulged upwards. Combined with a loss of thickness at the termini ⁓5,400-5,200 m above seas level, the bulge resulted in steepening by 5-6^o at its front, and therefore to locally increasing driving stresses. It is indicated by the changes in elevation rates derived for 2011-2014 that a down-glacier mass transfer had begun already 2003 ± 3 yrs. However, this internal mass movement did not result in surging of Aru-1 before mid-2015 and an advance of only 200 ± 15 m up to collapse, whilst no advance had been detected in Aru-2 at all. It was indicated by these observations that strong resistive forces to sliding at the termini of the glaciers, which suggests that they were frozen to their beds (Frappe & Clarke, 2007).

It was revealed by offset-tracking that the medial bulge of Aru-1 was associated with increased velocities of ice, from 0.18 ± 0.03 m/day in late 2013, to 0.50 ± 0.04 m/day in the spring of 2016. According to Kääb et al. these are 3-10 times higher flow speeds compared to the flow of 0.05 m/day that had been modelled. They found that average central speeds were unchanged in the medial bulge of 0.12 ± 0.03 m/day between July 2013 and April 2016.

Crevassing that was strongly enhanced developed on both glaciers in the weeks to months prior to the detachments. Along the lateral margins and across the glaciers crevassing was particularly evident at an elevation of about 5,800 m above sea level, which increased displacement where the glaciers later detached. Cracks across the entire Aru-1 glacier had already evolved in September 2015. On Aru-2 crevasses that were developing rapidly were discovered on satellite images prior to its collapse.

It is indicated that enhanced basal sliding occurred in the central parts of the glaciers, based on the high initial acceleration of the avalanches, the pattern of changes on the surface of the glaciers, thermo-mechanical modelling of the glaciers, and the formation of crevasses. It is also suggested that there were large amounts of water by a fan that was mud-flow like of basal fines that originated at about 5,500 m above sea level from crevasses, which is observed on satellite images of Aru-1 obtained 2 days prior to its collapse. No such observation was made for Aru-2.

It is indicated by thermo-mechanical modelling that the interface between the ice and the bed was probably temperate (thawed) in the central part of the glaciers though cold (frozen) in the remaining zones, which suggests a structure for the Aru glaciers that is polythermal. The late (Aru-1) or missing (Aru-2) advances of the fronts, significant sticking of the ice masses at the margins of the glaciers following detachment, and the location of the Aru Range within a semiarid region of permafrost support the thermal patterns that are model-indicated (Gruber, 2012). The polythermal structure of the glaciers led to conditions for:

· Margins and fronts of the glacier tongues to be frozen to the underlying bed;

· Infiltrating melt water to reach the bed of the glacier at the top of the detachment zones;

· The retention of this water upslope of the cold-based plugs leading to the precursory decrease in friction within the medial bulges (Clarke et al., 1984);

· And the development upstream of the cold-based fronts of a progressively steepening geometry.

Critically, the geometry of the glacier was prevented by the cold-based front from adapting quickly enough to the reduced friction, as would be the case in typical surges. The accumulation of water at the bed of the glacier through ingress of rainfall and summer melt in particular was probably accelerated after 2010 when there was an increase in the sum of both contributions by about 50%.

Causes and implications

Kääb et al. found that there were no earthquake events that were associated with either of the Aru events. It was indicated by Global Precipitation Measurement Integrated Multi-Satellite Retrievals (GMP IMERG) and the meteorological stations in the region that there was significant amounts of precipitation through the summer of 2016 of up to a total of 200 mm or more. From 10 to 25 mm of rainfall was observed for 2 days immediately prior to the Aru-1 event. Significant reduction of snow cover was revealed by optical satellite images in the weeks leading up to the collapse. It is indicated by very low back-scatter in the Sentinel-1 radar data from 1 July through to 31 August 2016 that continued melting conditions up to the tops of the mountains of the Aru Range (about 6,100 m above sea level). Several rainfall events occurred during the weeks leading up to the collapse, though there was no similar strong precipitation observed for the days immediately preceding the Aru-2 event. Therefore, at least for Aru-1, unusually high liquid water from melting snow and precipitation (highest positive degree-day summer since 1979 found for 2016) increased the content of water in the glacier system, which therefore puts it among the possible triggering factors for the collapse.

The deformed bed (Harrison & Post, 2003; Harrison et al., 2015; Truffer et al., 2000; Boulton & Jones, 1979; Cuffey & Paterson, 2010) of the Aru glaciers is an important factor linked to the fine-grained sandstone and siltstone, that is possibly lightly metamorphosed, that was mapped beneath the collapsed Aru glaciers, and till rich in clay, which was sampled. Kääb et al. suggest that on commencement of basal thaw and sliding, the till that was formed from these lithologies, combined with the high input of water, may have formed basal slurries locally that had low shear strength under high pore-fluid pressure (Kamb, 1991), and may possibly have destroyed any subglacial drainage system that was present (Clarke et al., 1984). In those areas the limited strength of the till would have induce d a rheology prone to trigger instabilities (Tulaczyk, Kamb & Engelhardt, 2000; Iverson, Hooyer & Baker, 1998).

The 2 tongues of the glaciers might have also influenced their stability by causing lateral resistance because of their curved shape in plan geometry. This is indicated by strong crevassing and maximum flow speeds at the northern, left-lateral glacier margins. Kääb et al. also noted that the surface slopes of 12-13^o (bed slopes 9-10^o) of the detached avalanche sources rank rather high for glacier tongues of surge type, which would therefore exert high driving stresses (Harrison et al., 2015).

The foregoing shows, in sum, that three is no evidence for a single trigger for the twin collapses. The factors that are present in a number of other glaciers on the Tibetan plateau, as well as elsewhere, where:

· Surge-like behaviour;

· Glacier steepening that is driven by climate;

· Polythermal glacier structure’

· Their geometry and slope;

· Liquid water from melting snow and rain during the summer of 2016.

Kääb et al. proposed therefore that the collapse of the Aru glaciers was caused by a transient convergence of factors that acted at different spatial and temporal scales. Important basic conditions are represented by a polythermal glacier regime, glacier morphology and lithology of bedrock. The simultaneous increase in temperature and precipitation over timescales of 15-20 years has acted on the geometry of the glacier and basal friction respectively increasing the slope and enhancing the input of liquid water to the bed of the glacier, and possibly expanding the area over which thawed conditions exist. The glaciers are prevented from adjusting their geometries to changed driving stresses and frictional conditions by the presence of cold-based fronts and margins. This increased continuously the stress on the frozen terminus and margins, until reaching a critical point at which the resisting forces were eventually overcome and the collapse occurred. These events are made unique by the sustained exceptional low frictions over large parts of the glacier, and are probably associated with a particular response of the fine-grained basal till to large amounts of liquid water (Harrison & Post, 2003; Harrison et al., 2015; Clarke et al., 1984; Roering et al., 2015). It is believed that the final trigger is linked to short-term variation in subglacial hydrological system that was induced by unusually high water input from melting snow and rain in the summer of 2016. Potentially, hydro-thermodynamic feedback also played a role – a mechanism by which basal sliding is increased by water reaching the bed of the glacier through crevasses, strain and fracturing that is associated with it by a combination of thermal and dynamic changes, therefore facilitating the input of additional water and modifying subglacial drainage (Fountain et al., 2005).

There are some similarities between the Aru events in 2016 and the Kolka Glacier collapse of 2002 including volume, glacier and avalanche slopes, and speed of the avalanche. In the days leading up to the detachment (Evans et al., 2009; Huggel et al., 2005; Haeberli, 2004), signs of destabilisation were shown by the Kolka Glacier which included heavy crevassing and bulging, as well as unusual hydrological conditions (Fountain et al., 2005) such as supraglacial ponds. The Kolka Glacier surged in 1969/1970 (Huggel, 2005) and had a slope that was similar to that of the Aru glaciers (Evans et al., 2009). The Kolka/Kazbek area is known for its fine-grained rocks and volcanic sediments (Drobyshev, 2006) which cause a variety of mass movements (Chernomorets et al., 2007), and putting forward the possible role of lithology in the rapid and sustained reduction of basal friction in the 3 events. Beneath the Kolka (Drobyshev, 2006) and Aru glaciers unusual geothermal heat fluxes are not certain but they cannot be excluded. The Kolka Glacier, however, contrasts with the Aru glaciers in having been temperate throughout. Also, heavy rock and snow fall activity deposited several million m³ of material on the glacier over a number of weeks before the surge event (Evans et al., 2009; Huggel et al., 2005; Haeberli et al., 2004). When the Kolka and Aru cases are considered together 2 possibilities are suggested to reach critical glacier geometry and an associated increase in driving stresses; i.e., additional loading from surrounding mass movements (Kolka) or steepening of the glacier (Aru).

A new form of glacier instability, the catastrophic collapse of large parts of an entire valley glacier, in this case about 25-30% by area, and up to 40% by volume, has been recognised as a result of the twin glacier collapses of the Aru glaciers. It seems that these collapses are made possible by a rare, though not unique array of factors that coalesce into an anomalous increase in driving stresses, as well as rapid and sustained reduction of basal friction. Kääb et al. say it is spectacular and completely unprecedented that such combined exceedances of instability thresholds, that are highly unlikely for a single glacier, occurred on 2 neighbouring glaciers within 2 months. It doesn’t seem likely that rock-mechanical or hydrologic/hydraulic collateral effects of Aru-1 could have triggered, or preconditioned the collapse of Aru-2, which highlights the role of common external forcing by climate and weather in synchronising the 2 collapses. New light is shed on the occurrence of glacier instabilities that are surge-related; several documented glacier instabilities (Zhang, 1992; Ugalde et al., 2015; Heybrock, 1935; Espizua, 1986; Milana, 2007) should be revisited to clarify their relation to the processes of glacier collapse suggested here. It is implied by the transient nature of some of the factors involved in the collapses of the Aru and Kolka glaciers, and the range of conditions and thresholds that are suggested by the differences between them, that such events can happen without historical precedent. It has been suggested that when the deposits from the Aru avalanche have melted there should be an investigation on their geomorphic and lithologic imprint and a search should be carried out in the wider region for signs of potential previous glacier collapses. The analysis of Kääb et al. has shown that under specific circumstances climate variability and change have a critical potential to contribute to glacier instability on a large scale by changing glacier geometry, thermal conditions and the content of liquid water. Though it seems beyond reach at the present that long-term prediction of similar events could be achieved the pre-event observations and simulations of the Aru-2 collapse demonstrate that available, state-of-the-art ground-based and satellite monitoring, as well as modelling capabilities allow early warnings to be issued even in regions that are very remote. Scientific advances to benefit also the most remote communities would be possible by harnessing these capabilities.

Sources & Further reading