remote sensing

Article Repeat Glacier Collapses and Surges in the Amney Machen Mountain Range, Tibet, Possibly Triggered by a Developing Rock-Slope Instability

Frank Paul

Department of Geography, University of Zurich, 8057 Zurich, Switzerland; [email protected]; Tel.: +41-44-6355175


Received: 19 February 2019; Accepted: 19 March 2019; Published: 24 March 2019


Abstract: Collapsing valley glaciers leaving their bed to rush down a flat hill slope at the speed of a racing car are so far rare events. They have only been reported for the Kolkaglacier (Caucasus) in 2002 and the two glaciers in the Aru mountain range (Tibet) that failed in 2016. Both events have been studied in detail using satellite data and modeling to learn more about the reasons for and processes related to such events. This study reports about a series of so far undocumented glacier collapses that occurred in the Amney Machen mountain range (eastern Tibet) in 2004, 2007, and 2016. All three collapses were associated with a glacier surge, but from 1987 to 1995, the glacier surged without collapsing. The later surges and collapses were likely triggered by a progressing slope instability that released large amounts of ice and rock to the lower glacier tongue, distorting its dynamic stability. The surges and collapses might continue in the future as more ice and rock is available to fall on the glacier. It has been speculated that the development is a direct response to regional temperature increase that destabilized the surrounding hanging glaciers. However, the specific properties of the steep rock slopes and the glacier bed might also have played a role.

Keywords: glacier surge; glacier collapse; rock-slope instability; hazard; Landsat; Sentinel 2; Tibet

1. Introduction Glacier surges have recently been in the focus of several scientific studies [1], largely because remote sensing data with a sufficiently high spatial and temporal resolution allow accurate tracking of surface and morphological changes as well as creation of dense time-series of flow velocities [2–12]. During a surge, large amounts of ice are transported at comparably high velocities (several m/day) from an upper reservoir area downward to a receiving zone, possibly creating a strong advance of the terminus. After a surge, the ice becomes stagnant and melts away while in the reservoir zone new ice accumulates [8]. Whereas an improved understanding of possible surge mechanisms slowly emerges [13–15], collapses of glaciers (where the ice is removed from a flat bed) have so far only rarely been reported. The two most prominent and best-documented examples are the 2002 collapse of Kolkaglacier in the Caucasus [16–19] and the 2016 collapse of two small valley glaciers in the Aru mountain range of Tibet [20–23]. In both regions the collapses caused long-distance (several km) mass movements with very high velocities (200 to 290 km/h) [19,22] that have not been thought possible before they occurred. In contrast to the frequent ice break-offs at steep hanging glaciers [24], the Kolka and Aru glaciers rested on beds of a comparably low slope and should thus have been resistant to mechanical failure. The history of events leading to the failure was different in both cases and they are thus unique on their own. Kolkaglacier is a well-known surge-type glacier [17] that is mostly nourished by avalanche snow and has been removed from its bed during its 2002 collapse, possibly due to a massive

Remote Sens. 2019, 11, 708; doi:10.3390/rs11060708 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, 708 2 of 18 loading with debris from rock fall of its surrounding slopes [25]. Kolkaglacier had a regular surge in 1969/70 and might have also collapsed in 1835 and 1902 [17]. The Aru glaciers were not known as surgingRemote before Sens. 2017 [23, 9,], x FOR but PEER showed REVIEW a typical pattern of surge-related elevation changes2 before of 17 they collapsed [21]. The twin-glacier collapse removed the lower parts of both glaciers, likely due to a surge in 1969/70 and might have also collapsed in 1835 and 1902 [17]. The Aru glaciers were not change in thermal/hydrologic conditions at the glacier bed [22]. known as surging before [23], but showed a typical pattern of surge-related elevation changes before Thethey collapses collapsed reported [21]. The here twin-glacier for a glacier collapse (GLIMS removed ID the G099443E34824N, lower parts of both Inventory glaciers, likely ID CN5J352E0017 due to ) ◦ locateda on change the north-western in thermal/hydrologic slope conditions of the Amney at the Machenglacier bed (A’ny [22]. ê Maqên) mountain range (34.822 N, 99.44◦ E) tookThe already collapses place reported in 2004, here and for again a glacier in 2007 (GLIMS and 2016.ID G099443E34824N, Figure1a shows Inventory the location ID of the mountainCN5J352E0017) range and thelocated glacier on the and north-western Figure1b isslope an obliqueof the Amney perspective Machen view(A’nyê up-glacier Maqên) mountain showing the range (34.822° N, 99.44° E) took already place in 2004, and again in 2007 and 2016. Figure 1a shows study region after the second collapse in 2008. The surges and collapses of this glacier have so far not the location of the mountain range and the glacier and Figure 1b is an oblique perspective view been reportedup-glacier [26 showing] or analyzed the study in region the scientific after the literature,second collapse although in 2008. they The repeatedlysurges and collapses buried aof country road thatthis has glacier been have reconstructed so far not been after reported each event[26] or (Figureanalyzed A1 in a).the Hence,scientific localliterature, authorities although are they aware of the collapsesrepeatedly and buried an information a country road board that describing has been reconstructed the 2004 avalanche after each event had been(Figure installed. A1a). Hence, The board can be seenlocal authorities in a picture are fromaware a of tourist the collapses [27] and and showsan information the advancing board describing glacier the a few2004 weeksavalanche before its third collapsehad been in installed. 2016 (Figure The board A1b). can The be seen small in valleya picture glacier from a (sizetourist about[27] and 1.5 shows km2) the has advancing a homogenous, glacier a few weeks before its third collapse◦ in 2016 (Figure A1b). The small valley glacier (size about gently sloping2 tongue (slope about 13 ) with a length of about 2 km and an elevation range of 450 m 1.5 km ) has a homogenous, gently sloping tongue (slope about 13°) with a length of about 2◦ km and (from 4800an elevation to 5250 range m). Theof 450 tongue m (from is 4800 nourished to 5250 bym). iceThe from tongue a steepis nourished (mean by slope ice from 36 )a andsteep its once connected(mean upper slope part, 36°) and that its reaches once connected as high uppe as 5900r part, m. that reaches as high as 5900 m.

Figure 1.Figure(a) Location1. (a) Location of the of the Amney Amney Machen Machen mountainmountain range range in innorth-eastern north-eastern Tibet, Tibet, China. China. The The collapsedcollapsed glacier glacier is marked is marked with with a red a red square. square. TheThe location of of the the stud studyy region region is marked is marked in the inset in the inset with a yellowwith a yellow arrow. arrow. Image Image sources: sources: Screenshots Screenshots from Google Google Earth. Earth. (b) (Obliqueb) Oblique perspective perspective view view of the collapsedof the collapsed glacier glacier as seen as seen on on a Quickbirda Quickbird imageimage acquired acquired on on 13.1.2008, 13.1.2008, a few a months few months after the after the second collapse. The steep and partly already glacier-free rock walls can be seen in the background. second collapse. The steep and partly already glacier-free rock walls can be seen in the background. The extent of the 2004 collapse is still visible in the foreground, the deposit from the 2007 collapse The extentcan ofbe theseen2004 on top collapse of it. Image is still source: visible Screenshot in the from foreground, Google Earth. the deposit from the 2007 collapse can be seen on top of it. Image source: Screenshot from Google Earth. The Amney Machen mountain range is covered by numerous glaciers (covering about 80 km2), Theof Amney which several Machen have mountain been classified range in is covereda previous by study numerous as surge glaciers type and (covering strongly aboutadvancing 80 km 2), of which severalbetween have 1966 beenand 1981 classified [28]. Time in series a previous of Landsat study images as surge (available type since and 1987) strongly reveal advancing that several between 1966 andof 1981the larger [28]. Timeglaciers series in this of Landsatmountain images range ha (availableve surged since again 1987) during reveal the past that three several decades. of the larger Without temperature measurements, any assumptions about thermal conditions of the ice and rock glaciers in this mountain range have surged again during the past three decades. Without temperature walls are speculative. However, thermokarst features that can be found in the valley floor to the measurements,southwest any of the assumptions mountain range about and thermal the permafrost conditions zonation of the map ice andby Gruber rock walls (2012) are [29] speculative. are However,providing thermokarst evidence features that the thatglaciers can higher be found up co inuld the be valleypoly-thermal floor or to cold-based. the southwest The purpose of the mountainof range andthis the study permafrost is providing zonation an overview map on by the Gruber glacier surges (2012) and [29 collapses] are providing between evidence1987 and 2017 that based the glaciers higher upon the could analysis be poly-thermal of freely available or cold-based. satellite image The time purpose series. A offurther this studyanalysis is of providing the events anusing overview on the glaciernumerical surges modeling and and collapses field data between is worthwhile 1987 but and beyond 2017 basedthe scope on of the this analysis study. of freely available satellite image time series. A further analysis of the events using numerical modeling and field data is worthwhile but beyond the scope of this study. Remote Sens. 2019, 11, 708 3 of 18

2. Datasets

2.1. Satellite Data TheRemote analysis Sens. 2017 of, 9 the, x FOR surges PEER REVIEW and collapses of the glacier is based on optical satellite images3 of 17 from different2. sourcesDatasets (Table A1). Declassified reconnaissance imagery from the Corona KH4-A and B missions acquired in 1964 and 1969 are among the earliest sources of information about the region. With a spatial2.1. Satellite resolution Data of about 3 to 5 m they reveal several details of the glacier tongue and its forefield. However,The analysis sun-lit of the snow surges is and over collapses exposed. of the Landsat glacier is imagery based on covering optical satellite the period images 1987from to 2016 are mostlydifferent taken sources from Landsat(Table A1). 5 (21Declassified scenes) atreconnaissance 30 m resolution imagery (red-band), from the andCorona Landsat KH4-A 7 (14and scenes)B at 15 m resolutionmissions acquired (panchromatic in 1964 and band). 1969 are The among striping the earliest of Landsat sources scenes of information after 2003 about due the to region. the failure of With a spatial resolution of about 3 to 5 m they reveal several details of the glacier tongue and its the scan-line corrector had little impact on this, as the study region is located close to the scene center. forefield. However, sun-lit snow is over exposed. Landsat imagery covering the period 1987 to 2016 Four Landsatare mostly 8 scenes taken acquiredfrom Landsat from 5 (21 2013 scenes) to 2016 at 30 (15 m resolution m resolution) (red-band), and four and Landsat Sentinel 7 2(14 scenes scenes) acquired in 2016at and 15 2017m resolution (10 m resolution)(panchromatic have band). been The used striping for of the Landsat more scenes recent after analysis. 2003 due Collectively, to the failure satellite data provideof the anscan-line image corrector in about had every little yearimpact for on the this, full as period the study (none region in 1992/98),is located close allowing to the ascene continuous interpretationcenter. Four of events. Landsat Several 8 scenes furtheracquired scenes from 2013 have to 2016 been (15 used m resolution) for a detailed and four analysis Sentinel of 2 scenes the collapses, partly alsoacquired including in 2016 scenes and with2017 snow(10 m coverresolution) and clouds.have been Key used images for the of themore surges recent and analysis. collapses are Collectively, satellite data provide an image in about every year for the full period (none in 1992/98), provided as a separate dataset in the Supplemental Material. allowing a continuous interpretation of events. Several further scenes have been used for a detailed Additionalanalysis of analysisthe collapses, was partly performed also including using sc veryenes high-resolutionwith snow cover and imagery clouds. acquiredKey images by of Quickbirdthe on 13 Januarysurges and 2008 collapses and likely are provided one of as the a sepa Worldviewrate dataset satellites in the Supplemental in winter 2016/17. Material. These images were directly analyzedAdditional at maximumanalysis was resolution performed using in Google very high-resolution Earth and maps imagery from acquired bing.com by Quickbird. For the latter, the imageon 13 source January is not2008 provided.and likely one The of Coronathe Worldview and Landsat satellites 5/7 in winter scenes 2016/17. have These been downloadedimages were from earthexplorer.usgs.govdirectly analyzed .at The maximum Landsat resolution 8 and Sentinel in Google 2 Earth images and were maps obtained from bing.com. from For remotepixel.ca. the latter, the image source is not provided. The Corona and Landsat 5/7 scenes have been downloaded from 2.2. Digitalearthexplorer.usgs.gov. Elevation Models (DEMs)The Landsat 8 and Sentinel 2 images were obtained from remotepixel.ca. For2.2. topographic Digital Elevation analysis Models and(DEMs) calculation of elevation changes, the 30 m version of the SRTM DEM (acquiredFor topographic in February analysis 2000) and has calculation been used of inelevation combination changes, withthe 30 the m 8version m High of the Mountain SRTM Asia (HMA)DEM (acquired acquired in in February 2015 [30 2000)]. Hillshade has been versionsused in combination of both DEMs with arethe 8 shown m High for Mountain the study Asia region in

Figure2(HMA). The latter DEM acquired suffers locallyin 2015 [30]. from Hillshade data voids versions but of is both otherwise DEMs are of shown very highfor the quality. study region The SRTM in Figure 2. The latter suffers locally from data voids but is otherwise of very high quality. The SRTM DEM has a more rough or ‘bumpy’ surface but is otherwise looking reasonable. The SRTM DEM was DEM has a more rough or ‘bumpy’ surface but is otherwise looking reasonable. The SRTM DEM was downloadeddownloaded from earthexplorer.usgs.govfrom earthexplorer.usgs.gov and and the the HMAHMA DEM DEM from from nsidc.org. nsidc.org.

Figure 2.FigureVisual 2. Visual comparison comparison of theof the two two DEMs DEMs forfor the re regiongion around around the thecollapsed collapsed glacier glacier (outlines (outlines from RGI6.0 in black). (a) The 30 m SRTM DEM from 2000 (source: earthexplorer.usgs.gov), (b) the 8 from RGI6.0 in black). (a) The 30 m SRTM DEM from 2000 (source: earthexplorer.usgs.gov), (b) the 8 m m HMA DEM from 2015 (source: nsidc.org/data/HMA_DEM8m_CT). HMA DEM from 2015 (source: nsidc.org/data/HMA_DEM8m_CT). 3. Methods 3. Methods 3.1. Satellite Data 3.1. Satellite Data As a base for quantitative assessments, glacier outlines were mapped automatically from the AsSentinel a base 2 for scene quantitative acquired on assessments,4 August 2017 using glacier the outlinesred/SWIR wereband ratio mapped method automatically [31]. The various from the Sentinelextents 2 scene of acquiredthe glacier, on the 4 three August debris 2017 fans, using and th thee lake red/SWIR were created band by ratioon-screen method digitizing [31]. using The various extents of the glacier, the three debris fans, and the lake were created by on-screen digitizing using Remote Sens. 2019, 11, 708 4 of 18

ArcGIS from ESRI. Contrast-enhanced versions of the highest resolution dataset (panchromatic bands fromRemote Landsat Sens. 72017 and, 9, x 8) FOR are PEER used REVIEW for this purpose. To identify the chronology of the surge and4 of collapse 17 events, all images were displayed in chronological order in a Geographic Information System (GIS) and flicker-imagesArcGIS from ESRI. (going Contrast-enhanced back and forth between versions two of scenes)the highest are usedresolution to follow dataset the (panchromatic changes described below.bands The from changes Landsat are 7 difficult and 8) are to used see infor side-by-side this purpose. comparisonsTo identify the of chronology static images, of the sosurge some and of the multi-panelcollapse images events, haveall images been arranged were displayed to facilitate in chronological top-bottom comparison.order in a Geographic For improved Information visibility of System (GIS) and flicker-images (going back and forth between two scenes) are used to follow the the changes the reader is referred to the Supplemental Material. changes described below. The changes are difficult to see in side-by-side comparisons of static 3.2. Topographicimages, so some Analysis of the multi-panel images have been arranged to facilitate top-bottom comparison. For improved visibility of the changes the reader is referred to the Supplemental Material. The HMA DEM provided elevation and slope values of the glacier and the surrounding mountain range.3.2. In Topographic combination Analysis with the digitized glacier extents and deposit regions, elevation ranges have been calculatedThe HMA to determineDEM provided mean elevation slope values.and slope The values related of the horizontal glacier and distances the surrounding were directly measuredmountain in the range. GIS In using combination the Sentinel with the 2 image digitize fromd glacier 2017 extents in the and background, deposit regions, i.e., theelevation geometric referenceranges has have UTM been zone calculated 47N with to determine WGS 1984 mean datum. slope Elevation values. The changes related were horizontal calculated distances by subtracting were the SRTMdirectly DEM measured from thein the HMA GIS DEM,using afterthe Sentinel both were 2 image resampled from 2017 bilinearly in the background, to 16 m resolution i.e., the and co-registered.geometric Horizontalreference has shifts UTM were zone smaller 47N with than WGS the 19 cell84 datum. size, but Elevation the different changes vertical were calculated datum caused by subtracting the SRTM DEM from the HMA DEM, after both were resampled bilinearly to 16 m a bias of about 30 m that was subtracted. Due to artifacts in the difference DEM, elevation changes resolution and co-registered. Horizontal shifts were smaller than the cell size, but the different were only analyzed locally. A manually digitized centerline through the glacier and deposit area was vertical datum caused a bias of about 30 m that was subtracted. Due to artifacts in the difference usedDEM, to derive elevation elevation changes values were from only the analyzed co-registered locally. DEMs.A manually digitized centerline through the glacier and deposit area was used to derive elevation values from the co-registered DEMs. 4. Results: The History of the Surge and Collapse Events 4. Results: The History of the Surge and Collapse Events 4.1. The 1960s Situation The4.1. The Corona 1960s Situation images from 1964 (Figure3) and 1969 (not shown) revealed that the glacier was connectedThe to Corona its eastern images and from southern 1964 (Figure tributary 3) and during 1969 that(not timeshown) and revealed had a wellthat the developed glacier was tongue with someconnected debris to its cover eastern along and its southe northernrn tributary margin. during In the that 1969 time image, and ahad crevasse a well acrossdeveloped its entiretongue width is visiblewith insome its flatdebris upper cover part along (indicating its northern a steep margin. slope In the in the1969 bedrock) image, a andcrevasse a small across rock its outcrop entire (in the followingwidth is visible RO-A) in already its flat upper existed part in the(indicating northern a steep part ofslope its accumulationin the bedrock) region.and a small The terminusrock outcrop (in the following RO-A) already existed in the northern part of its accumulation region. The was flat and retreating from 1964 to 1969 by about 50 m. The river leaving the glacier has eroded an terminus was flat and retreating from 1964 to 1969 by about 50 m. The river leaving the glacier has increasingly deeper trench in the foothill section of the mountain slope before reaching the main river eroded an increasingly deeper trench in the foothill section of the mountain slope before reaching (Qu’ngointhe main He) river draining (Qu’ngoin the regionHe) drai intoning the the Yellow region Riverinto the (Huang Yellow He).River In (Huang its lower He). part, In its the lower outflow is thuspart, spatially the outflow well constrained,is thus spatially whereas well constrained, in its upper whereas part, erosion in its upper was less part, intense erosion and was a potentialless floodingintense would and havea potential been flooding able to leave would the have riverbed been able and to spread leave the out riverbed over a largerand spread region. out Apartover a from this, nothinglarger region. unusual Apart (or from a previous this, nothing surge) unusual can be (or detected a previous on thesurge) images. can be The detected hill-slope on the section images. below the glacierThe hill-slope seems tosection be intact, below i.e., the notglacier impacted seems to by be a intact, previous i.e., not collapse. impacted by a previous collapse.

FigureFigure 3. Corona 3. Corona image image from from 1964 1964 showing showing the the glacier glacier beingbeing connected to to its its accumulation accumulation area area and and a glaciera glacier forefield forefield that is that free is of free any of avalanche any avalanche deposits. deposits. Note: Note: South South isup is up to highlightto highlight the the relief. relief. Image source:Image earthexplorer.usgs.gov source: earthexplorer.usgs.gov..

Remote Sens. 2019, 11, 708 5 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 5 of 17

4.2.4.2. TheThe 1987–19951987–1995 SurgeSurge TheThe fewfew availableavailable MSSMSS imagesimages coveringcovering thethe periodperiod 19721972 toto 19821982 areare tootoo sparsesparse andand tootoo coarsecoarse toto revealreveal any any substantial substantial variability variability in in terminus terminus position. position. It is It thus is thus assumed assumed here here that thethat glacier the glacier had been had aboutbeen about stagnant stagnant during during the 1970s the until1970s 1985. until The 1985. following The following Landsat Landsat TM time TM series time reveals series areveals surging a glaciersurging tongue glacier that tongue advanced that advanced about 700 about m from 700 1987 m fr toom 1995 1987 (Figure to 19954). (Figure The size 4). of The RO-A size increased of RO-A duringincreased this during time and this by time 1997 and it connectedby 1997 it connected to the ice-free to the terrain ice-free further terrain down further and down was thus and no was longer thus anno outcrop.longer an As outcrop. it is unlikely As it is thatunlikely all the that ice all has the been ice has removed been removed by melting by melting (other glaciers (other glaciers at lower at elevationslower elevations are stable), are stable), it can be itassumed can be assumed that a larger that parta larger of it part was depositedof it was deposited on the lower on glacierthe lower in formglacier of in ice form avalanches. of ice avalanches. From 1996 From to 2000, 1996 the to terminus 2000, the had terminus been stationary had been andstationary the lower and part the oflower the glacierpart of tonguethe glacier showed tongue the showed typical post-surgethe typical down-wasting.post-surge down-wasting.

FigureFigure 4.4.Landsat Landsat imagesimages fromfrom ((a)a) 1515August August 19871987 andand ((b)b) 1010 AugustAugust 19971997 showingshowing thethe firstfirst surgesurge ofof thethe glacier. glacier. The The false false colour colour images images are are showing showing (clean) (clean) ice andice and snow snow in cyan, in cy rocksan, rocks and graveland gravel in pink in topink purple, to purple, and vegetation and vegetation in light in green.light green. Image Image source: source: earthexplorer.usgs.gov earthexplorer.usgs.gov..

4.3.4.3. TheThe DevelopingDeveloping SlopeSlope InstabilityInstability FromFrom AugustAugust 20002000 toto AugustAugust 2002,2002, thethe areaarea ofof thethe ice-freeice-free terrainterrain aroundaround thethe formerformer RO-ARO-A furtherfurther increased,increased, particularly particularly in in its its upper upper parts parts (Figure (Figure5 a,b).5a,b). The The 2002 2002 image image also also shows shows substantial substantial darkeningdarkening ofof thethe northernnorthern partpart ofof thethe glacierglacier (in(in itsits flatflat section),section), indicatingindicating thatthat substantialsubstantial rockrock fallfall occurredoccurred before,before, maybemaybe asas aa partpart ofof iceice avalanches, avalanches,maybe maybeindependently. independently. ThisThisdebris-covered debris-covered partpart remainedremained untiluntil 2003.2003. From 20022002 toto 2003,2003, thethe lowerlower partpart ofof thethe glacierglacier diddid notnot changechange muchmuch butbut aa furtherfurther rock-outcroprock-outcrop (RO-B)(RO-B) developeddeveloped inin thethe steepsteep partpart ofof its its eastern eastern accumulation accumulation region. region. WhenWhen alsoalso consideringconsidering thethe laterlater veryvery high-resolutionhigh-resolution satellitesatellite images,images, itit seemsseems thatthat betweenbetween AugustAugust andand SeptemberSeptember 2003, 2003, a a larger larger part part of of the the main main eastern eastern tributary tributary broke broke off andoff and slid slid down down onto onto the lower the lower part ofpart the of glacier. the glacier. Afterwards, Afterwards, the terminus the terminus started started surging. surging.

4.4.4.4. TheThe 2003/42003/4 SurgeSurge andand CollapseCollapse TheThe glacierglacier advancedadvanced byby aboutabout 230230 mm andand collapsedcollapsed atat somesome pointpoint betweenbetween 26.1.200426.1.2004 (Figure(Figure5 5c)c) andand 3.2.20043.2.2004 asas revealedrevealed byby aa largely largely snow-covered snow-covered Landsat Landsat TM TM image. image. TheThe ETM+ETM+ panpan imageimage formform 11.2.200411.2.2004 shownshown inin FigureFigure5 d5d has has already already much much less less snow snow so so that that the the debris debris fan fan becomes becomes visible. visible. The The brightbright trace trace in in the the panchromatic panchromatic band band of of the the ETM+ ETM+ image image and and the the still still snow snow covered covered part part visible visible on the on TMthe imageTM image is indicating is indicating that a largerthat a partlarger of part the collapsing of the collapsing glacier tongue glacier has tongue been deflectedhas been todeflected the south to bythe a doublesouth by moraine a double wall moraine at the end wall of theat shortthe end valley of the and short left the valley river bedand atleft the the ‘potential river bed overflow’ at the point‘potential marked overflow’ in Figure point3. However, marked thein Figure related 3. ice/rock/water However, the mixturerelated ice/rock/water must have been mixture sufficiently must fast/mobilehave been sufficiently to also overflow fast/mobile the c. 40to malso high overflow moraine the wall. c. 40 The m markshigh moraine of this first wall. collapse The marks can stillof this be identifiedfirst collapse in the can very still high-resolution be identified imagein the acquiredvery high-resolution by Quickbird image on 13.1.2008 acquired (cf. by Figure Quickbird1b). on 13.1.2008 (cf. Figure 1b).

Remote Sens. 2019, 11, 708 6 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 6 of 17

FigureFigure 5.5.Landsat Landsat ETM+ ETM+ pan pan images images from from (a (a)) 25 25 July July 2000, 2000, (b )(b) 31 31 July July 2002, 2002, (c) (c) 26 January26 January 2004, 2004, and and (d) 11(d) February 11 February 2004 2004 showing showing in (a )in the (a) active the active terminus terminus in 2000 in compared 2000 compared to the maximumto the maximum surge extent, surge (extent,b) the loading (b) the with loading debris with by 2002,debris in by (c) the2002, maximum in (c) the surge maximum extent in surge 2004 beforeextent thein 2004 collapse, before and the in (collapse,d) the debris and fanin (d) after the the debris collapse. fan after Image the source: collapse. earthexplorer.usgs.gov Image source: earthexplorer.usgs.gov..

TheThe fullfull extent extent of of the the deposit deposi hast has been been mapped mapped using using the 15 the m panchromatic15 m panchromatic band of band the Landsat of the 2 ETM+Landsat scene ETM+ acquired scene acquired on 5 August on 5 August 2004 (Figure 2004 (Figure A2a). A2a). With aWith measured a measured area ofarea 2.16 of 2.16 km kmand2 and an arbitrarilyan arbitrarily assumed assumed mean mean thickness thickness of of maybe maybe 10 10 m, m, the the volume volume of of the the avalanche avalanche would would bebe 20–2520–25 3 millionmillion mm3,, whichwhich isis muchmuch lessless thanthan bothboth avalanchesavalanches inin thethe Aru Aru mountain mountain range range [ [21].21]. TheThe LandsatLandsat 2 imageimage alsoalso showsshows aa lakelake (area(area 0.3530.353 kmkm2)) thatthat likelylikely formedformed asas aa resultresult ofof thethe blockingblocking ofof thethe mainmain 2 riverriver byby thethe avalancheavalanche depositdeposit (Figure(Figure A2 A2a).a). UntilUntil 1414 SeptemberSeptember 20042004 thethe lakelake grewgrewto to0.534 0.534 km km2 andand 2 draineddrained sometimesometime betweenbetween 2929 JuneJune andand 1515 JulyJuly 20052005at atan aneven evenslightly slightlylarger larger size size of of 0.71 0.71 km km2..

4.5.4.5. TheThe 20072007 SurgeSurge andand CollapseCollapse FromFrom 20042004 toto SeptemberSeptember 2007,2007, thethe glacierglacier advancedadvanced byby aboutabout 300300 mm beforebefore itit collapsedcollapsed againagain atat somesome timetime betweenbetween 2323 SeptemberSeptember andand 22 NovemberNovember 20072007 (Figure(Figure6 ).6). Its Its maximum maximum extent extent before before the the collapsecollapse waswas muchmuch smallersmaller thanthan inin 2004,2004, closeclose toto itsits ‘normal’‘normal’ minimumminimum extentextent (cf.(cf. FigureFigure A2 A2a,b).a,b). ThisThis indicatesindicates thatthat the the surge/collapse surge/collapse mechanism mechanism might might have have been been different, different, that lubricationthat lubrication by melt by water melt mightwater havemight played have played a role, a or role, that or the that glacier the glacier has reached has reached a different a different point point of dynamic of dynamic stability. stability. It is wellIt is possiblewell possible that after that threeafter years,three years, the collapsed the collapsed material material was more was a more loose a ice/rock loose ice/rock mélange mélange rather thanrather a homogenousthan a homogenous glacier. glacier. The contrast-stretched The contrast-stretched close-up close-up of the Quickbirdof the Quickbird scene acquiredscene acquired on 13 Januaryon 13 January 2008 (Figure 2008 7(Figure) reveals 7) thatreveals a small that part a small of the part flat, of lower the flat, glacier lower remained glacier inremained its bed. in This its partbed. isThis separated part is separated from highly from crevassed highly crevassed dirty ice higher dirty ic upe higher that is stillup that connected is still connected to the southern to the tributary.southern Intributary. other words, In other the rupturewords, didthe not rupture follow did a potential not follow failure a zone,potential but occurredfailure zone, somewhere but occurred across thesomewhere glacier. across the glacier.

Remote Sens. 2019, 11, 708 7 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 7 of 17 Remote Sens. 2017, 9, x FOR PEER REVIEW 7 of 17

Figure 6. Two Landsat 7 ETM+ pan images showing the surging glacier in 2007: (a) before and (b) FigureFigure 6. 6.Two Two Landsat Landsat 7 ETM+7 ETM+ pan pan images images showing showing the the surging surging glacier glacier in 2007:in 2007: (a) before(a) before and and (b) after(b) after the collapse. Image source: earthexplorer.usgs.gov. theafter collapse. the collapse. Image Image source: source: earthexplorer.usgs.gov earthexplorer.usgs.gov.. The resulting avalanche deposit is well visible on a Landsat ETM+ image from 16 August 2008 The resulting avalanche deposit is well visible on a Landsat ETM+ image from 16 August 2008 (Figure(Figure A2 A2b)b) thatthat hashas beenbeen usedused forfor mappingmapping itsits extentextent (1.46(1.46 km2).). The The very very low low reflectance reflectance in in the the (Figure A2b) that has been used for mapping its extent (1.46 km2). The very low reflectance in the ETM+ETM+ panchromaticpanchromatic bandband (that(that stretches well into the near infrared) infrared) indicates indicates that that the the deposit deposit has has aa highhigh water water content, content, likely likely from from the meltingthe melting ice inside. ice inside. The area The covered area bycovered the deposit by the is considerablydeposit is (32%)considerably smaller than(32%) in smaller 2004, but than a part in 2004, of the but ice–debris a part of mixture the ice–debris flowed mixture again over flowed the doubleagain over moraine, the indicatingdouble moraine, high flow indicating velocities. high The flow oblique velocities. perspective The oblique view in perspective Figure1b shows view thein Figure fresh deposit1b shows on topthe offresh the deposit older, more on top extended of the older, deposit more from extend theed 2004 deposit collapse. from The the nadir-view2004 collapse. of theThe newnadir-view deposit (Figureof the 8new) reveals deposit several (Figure further 8) reveals details. several It shows furthe ther details. larger It extent shows of the the larger 2004 extent collapse of (thatthe 2004 has flattenedcollapse out)(that with hasthe flattened new 2007 out) deposit with onthe top new of it.2007 The deposit latter has on still top considerable of it. The relief,latter indicatinghas still highconsiderable ice content. relief, At indicating the southern high margin ice content. of the At deposit, the southern even crevassesmargin of canthe deposit, be seen, even indicating crevasses that mostcan be of theseen, former indicating glacier that might most be of located the former here. glacier The deposit might is be also located filling here. a large The part deposit of the is trench also andfilling after a large a sharp part turn of the to thetrench south and some after material a sharp turn has beento the moved south outsome of material the trench has and been deposited moved higherout of up.the Antrench additional and deposited stream onhigher top ofup. the An main addi deposittional stream (with aon different top of the surface main pattern) depositindicates (with a thatdifferent possibly surface a second pattern) wave indicates of material that came possibly down a after second the mainwave collapse.of material The came image down also showsafter the the buriedmain collapse. country roadThe image and the also overtopped shows the moraineburied country wall. road and the overtopped moraine wall.

Figure 7. The large image shows a contrast-enhanced close-up of the glacier remnant (between the FigureFigure 7. 7.The The largelarge imageimage showsshows aa contrast-enhancedcontrast-enhanced close-upclose-up of the glacier remnant remnant (between (between the the arrows and the dashed line) after its 2007 collapse as seen on the 13 January 2008 Quickbird image arrowsarrowsand and thethe dasheddashed line)line) afterafter itsits 20072007 collapsecollapse as seen on the 13 January 2008 2008 Quickbird Quickbird image image (image source: Screenshot from Google Earth). The smaller inset shows about the same region after (image(image source: source: Screenshot Screenshot from from Google Google Earth). Earth). The Th smallere smaller inset inset shows shows about about the the same same region region after after the the 2016 collapse on an undated satellite image. Image source: Screenshot from bing.com. 2016the 2016 collapse collapse on an on undated an undated satellite satellite image. image. Image Image source: source Screenshot: Screenshot from from bing.com bing.com..

Remote Sens. 2019, 11, 708 8 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 8 of 17

Figure 8. Close-up showing the avalanche deposit zone as seen on a Quickbird scene acquired on 13 January 2008,2008, i.e.,i.e., about about 3 3 months months after after the the 2007 2007 collapse. collapse. Remnants Remnants from from the 2004the 2004 collapse collapse can becan seen be asseen well. as well. Image Image source: source: Screenshot Screenshot Google Google Earth. Eart Theh. insetThe inset shows shows a close-up a close-up of the of third the third deposit deposit from thefrom 2016 the collapse.2016 collapse. Image Image source: sour Screenshotce: Screenshot from bing.comfrom bing.com.. 4.6. Further Head Wall Degradation after 2009 4.6. Further Head Wall Degradation after 2009 From 2008 to 2009, another larger part of the steep glacier around the second rock outcrop (RO-B) From 2008 to 2009, another larger part of the steep glacier around the second rock outcrop started separating. The beginning of it is already visible on the Quickbird image from 13 January 2008 (RO-B) started separating. The beginning of it is already visible on the Quickbird image from 13 (Figure1b). At some point between August 2010 and April 2011, this glacier part collapsed as well January 2008 (Figure 1b). At some point between August 2010 and April 2011, this glacier part and likely ended on the lower glacier tongue. Interestingly, the failure zone did not follow the major collapsed as well and likely ended on the lower glacier tongue. Interestingly, the failure zone did not crevasses but was located higher up. This means that the further development of the glaciers on this follow the major crevasses but was located higher up. This means that the further development of rock wall is difficult to predict. In the 15 years from 1997 to 2012, the original rock outcrop RO-A the glaciers on this rock wall is difficult to predict. In the 15 years from 1997 to 2012, the original rock developed into a large, heart-shaped region of steep glacier-free terrain with only a small remnant of outcrop RO-A developed into a large, heart-shaped region of steep glacier-free terrain with only a the former eastern glacier tributary left. small remnant of the former eastern glacier tributary left. 4.7. The Surge and Collapse in 2016 4.7. The Surge and Collapse in 2016 From 2011 to 2016, the glacier surged again by about 700 m (Figure9) and the ice-free region aroundFrom RO-B 2011 extended to 2016, furtherthe glacier upwards. surged The again last by phase about of 700 the surgem (Figure and following9) and the collapseice-free canregion be followedaround RO-B on Sentinel-2 extended images further at upwards. 10 m spatial The resolutionlast phase (Figureof the surge 10). The and glacier following was collapse still surging can onbe 30followed July 2016 on (FigureSentinel-2 10 a),images reaching at 10 its m maximum spatial resolution extent shortly (Figure after 10). 28The September glacier was 2016 still (Figure surging 10 onb), and30 July collapsing 2016 (Figure until 1810a), October reaching 2016 its (Figure maximum 10c). extent The inset shortly in Figure after 728 shows September that approximately 2016 (Figure 10b), the sameand collapsing part of the until glacier 18 October remained 2016 in its (Figure bed as 10c). after Th thee inset 2007 collapse.in Figure Whereas7 shows thethat big approximately gap to its upper the easternsame part part of is the still glacier there, thisremained part is in much its bed less as crevassed after the than 2007 in collapse. 2008. Whereas the big gap to its upperA eastern photograph part thatis still was there, taken this on pa 12rt September is much less 2016 crevassed by G. Butler than (Figure in 2008. A1 b) shows the advancing tongueA photograph about three weeksthat was before taken its on collapse. 12 September As one can2016 see by fromG. Butler the image, (Figure the A1b) terminus shows looks the dual-layeredadvancing tongue with anabout upper three layer weeks of brighter before iceits collapse. over the ‘normal’As one can tongue. see from The the image image, also the shows terminus some vegetationlooks dual-layered growth on with top an of theupper 2007 layer avalanche of brighter deposit ice (in over the foreground),the ‘normal’ thetongue. fine grainedThe image material also coveringshows some the mountainvegetation flanksgrowth to on the top north of andthe 2007 south avalanche of the glacier deposit valley (in (middlethe foreground), ground), the and fine the steepgrained and material dark, now covering nearly the ice-free mountain rock flanks wall in to the the background. north and south The of path the through glacier valley the avalanche (middle depositground), shown and the in steep Figure and A1 dark,a reveals now that nearly the ice-free rocky material rock wall is in rather the background. small-grained The and path that through the ice underneaththe avalanche the deposit debris layershown melted in Figure away A1a in the reveals meantime. that the rocky material is rather small-grained and that the ice underneath the debris layer melted away in the meantime.

Remote Sens. 2019, 11, 708 9 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 9 of 17 Remote Sens. 2017, 9, x FOR PEER REVIEW 9 of 17

FigureFigure 9. Time 9. Time series series of theof the 2013–2016 2013–2016 surge surge asas seenseen with the the Landsat Landsat 8 OLI 8 OLI panchromatic panchromatic band. band. The The whitewhiteFigure arrow arrow 9. marksTime marks series the the position of positionthe 2013–2016 of of the the terminus. surgeterminus. as seen Image with acquisition acquisition the Landsat dates dates8 OLI are: panchromatic are: (a) 16 (a )April 16 April 2013,band. 2013, (b) The 6 (b) 6 JuneJunewhite 2014, 2014, (arrowc) 12 (c) Augustmarks 12 August the 2015, position 2015, and and of (d (d) the) 29 29 terminus. July July 2016.2016. Image Image Image acquisition source: source: earthexplorer.usgs.gov. earthexplorer.usgs.govdates are: (a) 16 April 2013,. (b) 6 June 2014, (c) 12 August 2015, and (d) 29 July 2016. Image source: earthexplorer.usgs.gov. The Sentinel-2The Sentinel-2 image image from from 4 4 August August 20172017 shows the the deposit deposit in innadir-view nadir-view (Figure (Figure 10d) 10 andd) was and was usedused to digitize Theto digitize Sentinel-2 its extent.its extent. image It isItfrom is smaller smaller 4 August (area (area 2017 1.251.25 shows km22) the) than than deposit the the one onein fromnadir-view from 2007 2007 and (Figure and the avalanche the 10d) avalanche and washas has 2 this timethisused time seemingly to digitize seemingly its not extent. overtoppednot overtopped It is smaller the morainethe (area moraine 1.25 walls km wa) atthanlls the at the endthe one end of from the of shortthe 2007 short valley.and valley.the Hence,avalanche Hence, all materialhasall this time seemingly not overtopped the moraine walls at the end of the short valley. Hence, all was divertedmaterial was to the diverted south-west to the and south-west the deposit and coversthe deposit a larger covers region a larger here region than inhere 2007; than very in 2007; similar to material was diverted to the south-west and the deposit covers a larger region here than in 2007; 2004very (Figure similar 11). to However, 2004 (Figure it reached11). However, not that it reac farhed down, not that i.e. didfar down, likely i.e. not did cross likely the not main cross river. the The mainvery similarriver. The to 2004inset (Figurein Figure 11). 8 showsHowever, a subset it reac ofhed the notfresh that deposits far down, on top i.e. of did the likely former not ones. cross the insetmain in Figure river.8 Theshows inset a in subset Figure of 8 theshows fresh a subset deposits of the on fresh top deposits of the former on top ones.of the former ones.

Figure 10. Time series of the 2016 glacier collapse as seen on false-colour composite images (near Figureinfrared,Figure 10. Time10. red, Time seriesand series green of ofthe as the RGB) 2016 2016 acquired glacier glacier collapse bycollapse Sentinel-2. as s seeneen (a) onThe on false-colour image false-colour from composite 30 composite July 2016 images shows images (near the(near infrared,glacierinfrared, red, during red, and andits green surge. green as (b) as RGB) MaximumRGB) acquired acquired extent by by on Sentinel-2.Sentinel-2. 28 September (a) (a )The The2016 image imageabout from 1 from week 30 July 30before July 2016 the 2016 shows collapse. shows the the glacier during its surge. (b) Maximum extent on 28 September 2016 about 1 week before the collapse. glacier(c) duringAfter the its collapse surge. on (b) 18 Maximum October 2016. extent (d) onThe 28 deposit September from the 2016 last about collapse 1 week (in darker before shades the collapse. of (c) After the collapse on 18 October 2016. (d) The deposit from the last collapse (in darker shades of (c) Aftergrey) the in summer collapse 2017. on 18 The October destroyed 2016. country (d) The road deposit (that is fromalso used the lastas a collapsepilgrim path) (in darker has been shades re-establishedgrey) in summer over 2017. the fresh The deposit.destroyed Image country source: road Copernicus (that is also Sentinel used asdata a pilgrim2016 and path) 2017. has been of grey)re-established in summer over 2017. the fresh The destroyeddeposit. Image country source: road Copernicus (that is Sentinel also used data as 2016 a pilgrim and 2017. path) has been re-established over the fresh deposit. Image source: Copernicus Sentinel data 2016 and 2017.

In Figure 11, the Sentinel-2 image from 4 August 2017 is shown along with overlays of all three deposit extents and the various glacier extents (minimum and maximum) for comparison. Remarkable Remote Sens. 2017, 9, x FOR PEER REVIEW 10 of 17

Remote Sens. 2019, 11, 708 10 of 18 In Figure 11, the Sentinel-2 image from 4 August 2017 is shown along with overlays of all three deposit extents and the various glacier extents (minimum and maximum) for comparison. areRemarkable the large are maximum the large surge maximum extents surge from extents 2004 andfrom 2016 2004 compared and 2016 compared to the extent to the before extent the before 2007 collapse.the 2007 Thecollapse. growth The of growth rock outcrop of rock RO-B outcrop from RO-B 2007 tofrom 2017 2007 and to the 2017 re-established and the re-established path crossing path the freshcrossing avalanche the fresh deposit avalanche (Figure deposit A1a) (Figure can also A1a) be seen. can also be seen.

Figure 11. Extents of the deposits and glaciers for the three surges shown on the background of Figure 11. Extents of the deposits and glaciers for the three surges shown on the background of Figure6d: Deposits from the 2004, 2007, and 2016 collapses are marked green, yellow, and blue, Figure 6d: Deposits from the 2004, 2007, and 2016 collapses are marked green, yellow, and blue, respectively. Maximum glacier extents (before the collapse) from 2004, 2007, and 2016 are shown in red, respectively. Maximum glacier extents (before the collapse) from 2004, 2007, and 2016 are shown in yellow, and orange, whereas minimum extents (after the collapse) from 2004 and 2017 are shown in red, yellow, and orange, whereas minimum extents (after the collapse) from 2004 and 2017 are green and blue, respectively. Image source: Copernicus Sentinel data 2017. shown in green and blue, respectively. Image source: Copernicus Sentinel data 2017. The close-up from the very high-resolution satellite image in Figure 12 (acquired after the third The close-up from the very high-resolution satellite image in Figure 12 (acquired after the third collapse) reveals that the bedrock is covered with fine-grained material and that the rock cliffs point collapse) reveals that the bedrock is covered with fine-grained material and that the rock cliffs point downwards. There is thus not much resistance offered by the terrain and a hanging glacier that downwards. There is thus not much resistance offered by the terrain and a hanging glacier that might get lubricated at its bed will have a severe stability problem. Furthermore, larger parts of the might get lubricated at its bed will have a severe stability problem. Furthermore, larger parts of the rock outcrops seem to be covered by thin ice and the rocks as well as the fine-grained material are rock outcrops seem to be covered by thin ice and the rocks as well as the fine-grained material are comparably dark (i.e. they absorb much energy). It is thus well possible that the remaining glacier on comparably dark (i.e. they absorb much energy). It is thus well possible that the remaining glacier on this steep slope will degrade further and maybe slide down, thereby providing additional material to this steep slope will degrade further and maybe slide down, thereby providing additional material the lower glacier tongue. This could trigger further glacier surges and collapses in the future. Further to the lower glacier tongue. This could trigger further glacier surges and collapses in the future. Landsat 8 and Sentinel-2 images from 2017 and 2018 indicate that the glacier started advancing again Further Landsat 8 and Sentinel-2 images from 2017 and 2018 indicate that the glacier started in autumn 2017 but was in a stable position over most of 2018 at about the maximum 2007 extent advancing again in autumn 2017 but was in a stable position over most of 2018 at about the shown in Figure 11. maximum 2007 extent shown in Figure 11. 4.8. Elevation Changes and Topographic Analysis 4.8. Elevation Changes and Topographic Analysis Elevation changes between the SRTM DEM and the HMA DEM over the 2000–2015 period are depictedElevation in Figure changes 13 showing between the the difference SRTM DEM grid and and the elevation HMA DEM profiles over along the 2000–2015 a centerline period through are thedepicted deposit in andFigure the 13 glacier showing tongue the difference for both DEMs. grid and Apart elevation from regionsprofiles withalong strong a centerline elevation through gain (blue)the deposit and loss and (red) the that glacier are relatedtongue to for DEM both artifacts DEMs. in Apart regions from of steepregions terrain with and strong low contrastelevation (snow, gain shadow),(blue) and elevation loss (red) changes that are over related glaciers to DEM are alsoartifa wellcts in recognizable. regions of steep When terrain starting and at low the top,contrast the strong(snow, ice shadow), loss in theelevation region changes of the former over glaciers tributaries are andalsolater well rockrecognizable. outcrops AWhen and Bstarting (cf. Figure at the 12 top,) is clearlythe strong visible ice (valuesloss in the are region from about of the− former40 to − trib100utaries m). Despite and later artifacts rock beingoutcrops close, A theand heart-shaped B (cf. Figure regions12) is clearly of the twovisible rock (values outcrops are match from very about well − to40 theto area−100 ofm). elevation Despite loss. artifacts being close, the heart-shaped regions of the two rock outcrops match very well to the area of elevation loss.

Remote Sens. 2019, 11, 708 11 of 18 Remote Sens. 2017, 9, x FOR PEER REVIEW 11 of 17

Figure 12. Close-up of the instable slope with the two large, heart-shaped rock outcrops (RO-A and Figure 12. Close-up of the instable slope with the two large, heart-shaped rock outcrops (RO-A and RO-B) that have merged in the meantime. Some remaining debris-covered ice is visible (partly snow RO-B) that have merged in the meantime. Some remaining debris-covered ice is visible (partly snow covered) in the outcrops that seem to be covered with dark and fine-grained sediment. The former covered) in the outcrops that seem to be covered with dark and fine-grained sediment. The former eastern accumulation accumulation area area is is now now limited limited to toa small a small remaining remaining tongue tongue (cf. Figures (cf. Figures 4 and4 A1).and A1A larger). A largerpiece of piece ice ofin icethe in valley the valley floor floor is still is stillconnected connected to tothe the southern southern tributary tributary but but a alarge large crevasse crevasse is separating it fromfrom thethe restrest ofof thethe glacier.glacier. ImageImage source:source: ScreenshotScreenshot fromfrom bing.combing.com..

Going furtherfurther down,down, the the (former) (former) glacier glacier tongue tongue is also is also sharply sharply depicted depicted in the in difference the difference image. Atimage. the acquisition At the acquisition date of the date first of DEMthe first (SRTM), DEM the (SRTM), tongue the was tongue still close was to still its maximumclose to its extent maximum from theextent first from surge the (Figure first4 b)surge that was(Figure never 4b) reached that againwas never afterwards reached (Figure again 11 ).afterwards Accordingly, (Figure from 2000 11). toAccordingly, 2015, this lower from region2000 to showed 2015, this a pronouncedlower region thickness showed a loss pronounced (about 40 thickness m). When loss the (about second 40 DEM m). wasWhen acquired the second in 2015, DEM the was glacier acquired was again in 2015, surging the (Figureglacier 9wasc) and again the bluishsurging region (Figure just 9c) in frontand the of thebluish maximum region just 2007 in extent front inof Figurethe maximum 13 is depicting 2007 extent the slightly in Figure higher 13 is (about depicting 5–10 m)the elevationslightly higher of the terminus(about 5–10 at this m) positionelevation compared of the terminus to the year at this 2000 position surface. Thecompared advancing to the tongue year can2000 also surface. be seen The in theadvancing black profile tongue shown can also in thebe seen inset in of the Figure black 13 prof. Higherile shown up, in the the glacier inset surfaceof Figure in 13. 2015 Higher is generally up, the lowerglacier than surface in 2000 in 2015 (also is up generally to 40 m), lower and inthan the in zone 2000 of (also the steepup to cliff, 40 m), elevations and in the are zone about of the the same. steep cliff, Muchelevations more are subtle about and the withinsame. the DEM uncertainty are the elevation changes for the deposit region.Much However, more subtle closer inspectionand within reveals the DEM a majority uncertainty of bluish are the cells elevation (elevations changes about for 5–10 the m deposit higher) withinregion. theHowever, limits of closer the 2007 inspection deposit reveals and more a majority variable of changes bluish cells outside. (elevations This would about confirm 5–10 m that higher) this iswithin indeed the a limits region of of the mass 2007 deposit, deposit but and the more uncertainties variable changes are too highoutside. for aThis quantitative would confirm determination that this ofis indeed the deposited a region volume. of mass deposit, but the uncertainties are too high for a quantitative determination of theWhen deposited the study volume. region is separated in four units A: deposition, B: transition, C: flat glacier, D: steepWhen glacier the (see study bottom region of Figureis separated 13), the in elevation,four units length,A: deposition, and slope B: transition, values presented C: flat glacier, in Table D:1 aresteep found glacier from (see the bottom HMA DEM.of Figure The 13), mean the slope elevatio valuesn, length, indicate and that slope the areavalues of presented deposition in is Table indeed 1 ◦ comparablyare found from flat the (<6 HMA) and DEM. that the The flat mean tongue slope of values the glacier indicate is in that the typicalthe area range of deposition for valley is glaciers indeed ◦ ◦ (aboutcomparably 13 ). flat The (<6 transition°) and that zone the is flat somewhat tongue of steeper the glacier and theis in mean the typical slope range of the for rock valley walls glaciers is 36 , ◦ However,(about 13° locally,). The transition values exceeding zone is 50somewhatare found. steeper The overall and the slope mean or Fahrböschungslope of the rock (measured walls is from 36°, However, locally, values exceeding 50° are found. The overall slope or Fahrböschung (measured from the upper point of the collapsed glacier to the lowest point of the deposit) is 10.9° and thus within the range of similar events [18].

Remote Sens. 2019, 11, 708 12 of 18 the upper point of the collapsed glacier to the lowest point of the deposit) is 10.9◦ and thus within the range of similar events [18]. Remote Sens. 2017, 9, x FOR PEER REVIEW 12 of 17

FigureFigure 13. 13.Elevation Elevation difference difference grid grid (SRTM-HMA (SRTM-HMA DEM) DEM) over over the 2000the to2000 2015 toperiod 2015 period showing showing volume lossvolume (red) loss of the (red) glacier of the but glacier also but artefacts also artefacts to the north to the and north south and of south it. Glacier of it. Glacier outlines outlines and the and extent the ofextent the avalanche of the avalanche deposit deposit from 2007 from are 2007 shown are inshown black, in the black, elevation the elevation values for values theinset for the are inset sampled are alongsampled the dottedalong the line. dotted Elevation line. contourElevation lines contour (grey) lines have (grey) 50 m equidistance,have 50 m equidistance, sections A, B,sections C, and A, D areB, describedC, and D inare the described text. Elevation in the text. differences Elevation have differe beennces limited have been to a rangelimited of to± 150a range m. Theof ±150 inset m. shows The elevationinset shows profiles elevation extracted profiles from extracted both DEMs from along both DEMs the center along line. the center line.

TableTable 1.1.Topographic Topographic characteristicscharacteristics ofof thethe differentdifferent zones as derived from the HMA-DEM.

ElevationElevation (m) (m) ◦ ZoneZone Name name Length Length(m) Mean (m) SlopeMean (°) Slope ( ) MinimumMinimum Maximum Maximum Range Range AA DepositDeposit 4250 4250 4450 4450 200 200 2000 2000 5.7 5.7 BB TransitionTransition 4450 4450 4800 4800 350 350 1200 1200 16.3 16.3 C Flat glacier 4800 5250 450 2000 12.7 C Flat glacier 4800 5250 450 2000 12.7 D Steep glacier 5250 5900 650 900 35.8 DOverall Steep slopeglacier 5250 4250 5900 5250 650 1000 900 5200 35.8 10.9 Overall slope 4250 5250 1000 5200 10.9 5. Discussion 5. Discussion The detailed analysis of optical satellite images allowed reconstructing the changes of a surging The detailed analysis of optical satellite images allowed reconstructing the changes of a surging and collapsing glacier in eastern Tibet in much detail. Useful scenes are available for nearly every year and collapsing glacier in eastern Tibet in much detail. Useful scenes are available for nearly every and much denser time series could be used to constrain the timing of the collapses down to a week year and much denser time series could be used to constrain the timing of the collapses down to a when images from Landsat 5 and 7 are combined. Whereas the 30 m spatial resolution of Landsat week when images from Landsat 5 and 7 are combined. Whereas the 30 m spatial resolution of TM is sufficient to follow major changes of the glacier and the rock outcrops (Figure4), the 15 m Landsat TM is sufficient to follow major changes of the glacier and the rock outcrops (Figure 4), the panchromatic bands from Landsat 7 and 8 reveal several details much better (Figures5,6 and9) and 15 m panchromatic bands from Landsat 7 and 8 reveal several details much better (Figures 5, 6, 9) haveand have thus beenthus been used used together together with with 10 m 10 resolution m resoluti imageson images from from Sentinel-2 Sentinel-2 (Figure (Figure 10) 10) to followto follow the changesthe changes and to and map to the map extents the extents of the glacier of the and glacie ther deposits and the (Figure deposits 11 ).(Figure There are11). likely There uncertainties are likely uncertainties in the delineation, but these do not impact on the general interpretation of the events. Further in-depth insights are provided by the very high-resolution satellite images from Quickbird (in Google Earth) and the image of unknown source available in bing.com that show the situation after the second and third collapses, respectively. Both images reveal fine details of the glacier

Remote Sens. 2019, 11, 708 13 of 18 in the delineation, but these do not impact on the general interpretation of the events. Further in-depth insights are provided by the very high-resolution satellite images from Quickbird (in Google Earth) and the image of unknown source available in bing.com that show the situation after the second and third collapses, respectively. Both images reveal fine details of the glacier remnants (Figure7) and deposits (Figure8), as well as of the instable rock slope with its degrading hanging glaciers and incompetent lithology (Figure 12). Several further insights can be derived from these images but they would be more speculative. Additional quantitative information about glacier thickness changes is derived from two DEMs representing the situation after the first (SRTM DEM from 2000) and during the fourth surge (HMA DEM from 2015) of the glacier. Despite large artifacts close to the glacier, its volume loss, the removed tributaries, and a small elevation gain in the region of the deposit are visible, confirming the observations from the satellite images. In particular, the visibility of the 2015 surge front in the difference image (that is also visible in the DEM hillshade of Figure2b) confirms the high quality of the elevation data and reliability of the results. As a further back-up, elevation differences have also been calculated with the ALOS DEM (AW3D30) that was merged from images acquired over the 2007 to 2011 period (not shown here). These fully confirm the description of events presented above. The images for the DEM have likely been acquired shortly after the second collapse (e.g., in 2008 or 2009) and show a completely removed valley glacier, an even clearer elevation gain in the deposit area, a smaller area affected by elevation loss in the region of RO-B, and that the artifacts marked in Figure 13 are due to the SRTM DEM. In contrast, the difference between the ALOS and HMA DEM show a small surface lowering over the deposit area (from ice melting), the fully re-developed glacier tongue up to its 2015 extent, only small elevation changes in the region of RO-A and the strong surface lowering over RO-B. Whereas the glaciologic and geomorphologic observations described above are comparably robust, any reasoning about possible processes leading to the observed events is speculative. Hence, only the more general characteristics of the Kolka and Aru glacier collapses are compared in the following. The most striking difference is that the glacier observed here already collapsed three times (2004, 2007, and 2016) in only 12 years. Repeat collapses might have also occurred for Kolkaglacier as Kotlyakov et al. (2004) [17] wrote that the 1902 surge ended “in a catastrophic outburst of ice and water” similar to the event in 2002 and that “in a few minutes, the ice spread over 9 km through the valley”. However, the Kolkaglacier ”eruption” in 1902 occurred “at the height of a hot summer and after heavy showers” [17] seemingly with large amounts of water, whereas only small amounts of water could have been available for the 2004 mid-winter collapse reported here. The much longer time period between the two collapses (100 years) is also different. Moreover, all collapses reported here occurred during a surge with strong frontal advance. The Aru glaciers advanced only slightly before they collapsed, but the observed elevation change pattern is fully compliant with an ongoing surge [21]. In contrast to Kolkaglacier with its well-known surge history [17], both Aru glaciers had not been known as surging before [23]. On the other hand, Kolka glacier was retreating from 1984 to 2002 and basically stagnant before its 2002 collapse. Whereas the 2002 collapse of Kolkaglacier was likely triggered by the additional load of rock/ice avalanches on its surface [19,26], the Aru collapses seem to be triggered by extreme water pressure in combination with specific properties of the glacier bed [22]. The events reported here have likely been facilitated by the accumulation of ice and debris from the instable rock wall. However, these likely occurred in small portions and more gradually, so that the additional load might have first caused a surge rather than a sudden collapse. A similar surge initiation has been observed for Lysiiglacier in the Ak-Shirak mountain range of the Tian-Shan, where the material excavated from the Kumtor mine has been dumped on its surface [32]. Based on the chronology of events and the DEM differences, it is speculated that the 2003/04 surge was triggered by the extra load of ice and rock from RO-A, whereas the growing RO-A and RO-B triggered the 2007 surge. The third event in Oct 2016 might have been triggered by material mostly excavated from the region of RO-B. Compared to the Kolka and Aru Remote Sens. 2019, 11, 708 14 of 18 collapses, the events reported here were much smaller in volume. Chinese authorities estimated a volume of 36 million m3 for the 2004 event (as written on the information board), resulting in a mean thickness of about 15 m for the deposit. They speculated that the collapse was caused by a rising snow line and freeze-thaw action. Regarding the short time between the first and the second collapse (3 years), one has to consider that the ice mass coming down the second (and third) time was likely more a loose ice/rock mixture rather than a compact glacier. The mechanical properties for the latter two failures might thus have been different and need further investigation. More detailed studies (e.g., on ice avalanche modeling, thermal conditions, climatic trends, bedrock lithology) are also required to understand why and how the glacier has collapsed the first time, particularly when considering that the glacier did not collapse during the first surge although it was much more extended then. Some fieldwork in this region would certainly be helpful to constrain the physical properties of the glaciers and the surrounding environment. Overall, it seems that glacier collapses are not unique events but can occur repeatedly, particularly when there is extraordinary and continuous supply of material from surrounding rock walls. As there is still some ice left to fall on the glacier and the degradation of the steep rock walls continues, the on-going but infrequent input of ice and rock will play an important role in keeping the surges (and maybe also the collapses) alive. As the rock wall and the glacier are still very active, it can be recommended to observe them more closely in the future.

6. Conclusions This study presented a description of four glacier surges and three collapses of a small valley glacier in the Amney Machen mountain range of eastern Tibet that took place from 1987 to 1995 (surge only) and in 2004, 2007, and 2016. Whereas some characteristics of the events resemble similarities of glacier collapses reported earlier (Kolka, Aru), the combination found here is unique, because they already occurred three times with only a few years in-between and—despite the magnitude of the events and their impacts on infrastructure—because they have not been reported so far in the literature. The analysis of satellite images reveals that a developing rock-slope instability might be responsible for the last three surges by infrequently adding mass on the lower glacier. However, other reasons might play a role as well and more theoretical (numerical modelling) investigations as performed for the Kolka and Aru collapses might provide additional insights into the governing processes. The description of events presented here might help in analyzing related details in future studies. Further collapses of this glacier might occur, as the supply of material from the degrading rock wall is continuing. It is thus recommended to observe this glacier more closely in the future and collect some field data to constrain modelling efforts.

Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/11/6/708/s1, Key images of the surges and collapses. Author Contributions: All analysis and the writing of the paper have been performed by the author (F.P.). Funding: This research and the APC was funded by the ESA project Glaciers_cci, grant number 4000109873/14/I-NB. Acknowledgments: This study would not have been possible without the free access to Landsat and Sentinel data, digital elevation models, and very-high resolution satellite imagery visible in Google Earth and bing.com. Conflicts of Interest: The author declares no conflict of interest. Remote Sens. 2019, 11, 708 15 of 18

Abbreviations

ASTER Advanced Thermal Emission and reflection Radiometer DEM Digital Elevation Model ESRI Environmental Systems Research Institute ETM+ Enhanced Thematic Mapper + HMA High Mountain Asia GIS Geographic Information System GLIMS Global Land Ice Measurements from Space MSS MultiSpectral Scanner Remote Sens. 2017, 9, x FOR PEER REVIEW 15 of 17 RO Rock Outrcrop SRTM Shuttle Radar Topography Mission UTM Universal Transverse Mercator TM Thematic Mapper UTM Universal Transverse Mercator

Appendix Appendix

Figure A1.A1. TwoTwo imagesimages from from the the deposit deposit taken taken by by Brandon Brandon G. ButlerG. Butler on 12on September 12 September 2016. 2016. (a) A (a) view A toview the to north the showingnorth showing the avalanche the avalanche deposit deposit to the leftto the and left right and of theright reconstructed of the reconstructed road crossing road it.crossing Image it. source: Image [ 27source:]. (b) [27]. A view (b) toA view the east to the showing east showing the advancing the advancing glacier glacier (the white (the arrow white marksarrow itsmarks front), its aboutfront), 3 about weeks 3 beforeweeks itsbefore next collapse.its next colla Somepse. grass Some is visiblegrass is on visible the deposit on the of deposit the previous of the collapseprevious incollapse the foreground. in the foreground. The dark The rock dark slope rock in the slope background in the background (to the left (to of the the left image of the center) image is thecenter) now is huge the now rock huge outcrop rock RO-B. outcrop Image RO-B. source: Image [27 source:]. [27].

Remote Sens. 2017, 9, x FOR PEER REVIEW 15 of 17

UTM Universal Transverse Mercator

Appendix

Figure A1. Two images from the deposit taken by Brandon G. Butler on 12 September 2016. (a) A view to the north showing the avalanche deposit to the left and right of the reconstructed road crossing it. Image source: [27]. (b) A view to the east showing the advancing glacier (the white arrow marks its front), about 3 weeks before its next collapse. Some grass is visible on the deposit of the previous collapse in the foreground. The dark rock slope in the background (to the left of the image Remote Sens. 2019, 11, 708 16 of 18 center) is the now huge rock outcrop RO-B. Image source: [27].

Figure A2. Landsat 7 ETM+ scenes (pan band) used to mark the extent of the deposits. (a) Scene from

5 August 2004 (green: maximum glacier extent before the 2004 collapse, red: extent of the deposit, light blue/yellow: lake extent from 5 August/14 September 2004. (b) Scene from 16 August 2008 (after the 2007 collapse) showing the new deposit in red. Image source: earthexplorer.usgs.gov.

Table A1. Overview of the satellite images used for the analysis. All Landsat scenes have path-row 133-036 and the Sentinel-2 tile used is 47SNU. Corona scenes used: #1: KH-4A scene DS1015-2164DF117, #2: KH-4B scene DS1108-2184DA110. Sensor names: L5 TM: Landsat 5 Thematic Mapper, L7 ETM+: Landsat 7 Enhanced Thematic Mapper plus, L8 OLI: Landsat 8 Operational Land Imager, S2 MSI: Sentinel-2 Multi-Spectral Instrument.

Nr. Sensor Date Nr. Sensor Date Nr. Sensor Date 1 Corona 30.12.1964 16 L7 ETM+ 16.08.2002 31 L5 TM 23.09.2007 2 Corona 16.12.1969 17 L7 ETM+ 03.08.2003 32 L7 ETM+ 02.11.2007 3 L5 TM 15.08.1987 18 L5 TM 12.09.2003 33 L7 ETM+ 16.08.2008 4 L5 TM 30.06.1988 19 L5 TM 14.10.2003 34 L5 TM 11.08.2009 5 L5 TM 21.09.1989 20 L5 TM 15.11.2003 35 L5TM 14.08.2010 6 L5 TM 06.07.1990 21 L5 TM 18.01.2004 36 L7 ETM+ 24.07.2011 7 L5 TM 11.09.1991 22 L7 ETM+ 26.01.2004 37 L7 ETM+ 12.09.2012 8 L5 TM 31.08.1993 23 L5 TM 03.02.2004 38 L8 OLI 16.04.2013 9 L5 TM 14.05.1994 24 L7 ETM+ 11.02.2004 39 L8 OLI 06.06.2014 10 L5 TM 20.07.1995 25 L7 ETM+ 05.08.2004 40 L8 OLI 12.08.2015 11 L5 TM 24.09.1996 26 L5 TM 14.09.2004 41 L8 OLI 29.07.2016 12 L5 TM 10.08.1997 27 L7 ETM+ 09.09.2005 42 S2 MSI 30.07.2016 13 L5 TM 31.07.1999 28 L7 ETM+ 26.07.2006 43 S2 MSI 28.09.2016 14 L7 ETM+ 25.07.2000 29 L5 TM 20.09.2006 44 S2 MSI 18.10.2016 15 L7 ETM+ 13.08.2001 30 L7 ETM+ 15.09.2007 45 S2 MSI 04.08.2017

References

1. Qiu, J. Ice on the run. Science 2017, 358, 1120–1123. [CrossRef][PubMed] 2. Bevington, A.; Copland, L. Characteristics of the last five surges of Lowell Glacier, Yukon, Canada, since 1948. J. Glaciol. 2014, 60, 113–123. [CrossRef] 3. Quincey, D.J.; Glasser, N.F.; Cook, S.J.; Luckman, A. Heterogeneity in Karakoram glacier surges. J. Geophys. Res. Earth Surf. 2015, 120, 1288–1300. [CrossRef] 4. Yasuda, T.; Furuya, M. Dynamics of surge-type glaciers in West Kunlun Shan, Northwestern Tibet. J. Geophys. Res. Earth Surf. 2015, 120, 2393–2405. [CrossRef] 5. Häusler, H.; Ng, F.; Kopecny, A.; Leber, D. Remote-sensing-based analysis of the 1996 surge of Northern Inylchek Glacier, central Tien Shan, Kyrgyzstan. Geomorphology 2016, 273, 292–307. [CrossRef] 6. Herreid, S.; Truffer, M. Automated detection of unstable glacier flow and a spectrum of speedup behavior in the Alaska Range. J. Geophys. Res. Earth Surf. 2016, 121, 64–81. [CrossRef] Remote Sens. 2019, 11, 708 17 of 18

7. Pitte, P.; Berthier, E.; Masiokas, M.H.; Cabot, V.; Ruiz, L.; Ferri Hidalgo, L.; Gargantini, H.; Zalazar, L. Geometric evolution of the Horcones Inferior Glacier (Mount Aconcagua, Central Andes) during the 2002–2006 surge. J. Geophys. Res. Earth Surf. 2016, 121, 111–127. [CrossRef] 8. Bhambri, R.; Hewitt, K.; Kawishwar, P.; Pratap, B. Surge-type and surge-modified glaciers in the Karakoram. Sci. Rep. 2017, 7, 15391. [CrossRef] 9. Paul, F.; Strozzi, T.; Schellenberger, T.; Kääb, A. The 2015 surge of Hispar Glacier in the Karakoram. Remote Sens. 2017, 9, 888. [CrossRef] 10. Round, V.; Leinss, S.; Huss, M.; Haemmig, C.; Hajnsek, I. Surge dynamics and lake outbursts of Kyagar Glacier, Karakoram. Cryosphere 2017, 11, 723–739. [CrossRef] 11. Strozzi, T.; Paul, F.; Wiesmann, A.; Schellenberger, T.; Kääb, A. Circum-Arctic changes in the flow of glaciers and ice caps from satellite SAR data between the 1990s and 2017. Remote Sens. 2017, 9, 947. [CrossRef] 12. Wendt, A.; Mayer, C.; Lambrecht, A.; Floricioiu, D. A glacier surge of Bivachny glacier, Pamir mountains, observed by a time series of high-resolution digital elevation models and glacier velocities. Remote Sens. 2017, 9, 388. [CrossRef] 13. Sund, M.; Lauknes, T.R.; Eiken, T. Surge dynamics in the Nathorstbreen glacier system, Svalbard. Cryosphere 2014, 8, 623–638. [CrossRef] 14. Sevestre, H.; Benn, D.I. Climatic and geometric controls on the global distribution of surge-type glaciers: Implications for a unifying model of surging. J. Glaciol. 2015, 61, 646–659. [CrossRef] 15. Dunse, T.; Schellenberger, T.; Hagen, J.O.; Kääb, A.; Schuler, T.V.; Reijmer, C.H. Glacier-surge mechanisms promoted by a hydro-thermodynamic feedback to summer melt. Cryosphere 2015, 9, 197–215. [CrossRef] 16. Haeberli, W.; Huggel, C.; Kääb, A.; Zgraggen-Oswald, S.; Polkvoi, A.; Galushkin, I.; Zotikov, I.; Osokin, N. The Kolka–Karmadon rock/ice slide of 20 September 2002: An extraordinary event of historical dimensions in North Ossetia, Russian Caucasus. J. Glaciol. 2004, 50, 533–546. [CrossRef] 17. Kotlyakov, V.M.; Rototaeva, O.V.; Nosenko, G.A. The September 2002 Kolka Glacier catastrophe in North Ossetia, Russian Federation: Evidence and analysis. Mt. Res. Dev. 2004, 24, 78–83. [CrossRef] 18. Huggel, C.; Zgraggen-Oswald, S.; Haeberli, W.; Kääb, A.; Polkvoi, A.; Galushkin, I.; Evans, S.G. The 2002 rock/ice avalanche at Kolka/Karmadon, Russian Caucasus: Assessment of extraordinary avalanche formation and mobility, and application of QuickBird satellite imagery. Nat. Hazards Earth Syst. Sci. 2005, 5, 173–187. [CrossRef] 19. Evans, S.G.; Tutubalina, O.V.; Drobyshev, V.N.; Chernomorets, S.S.; McDougall, S.; Petrakov, D.A.; Hungr, O. Catastrophic detachment and high-velocity long-runout flow of Kolka Glacier, Caucasus Mountains, Russia in 2002. Geomorphology 2009, 105, 314–321. [CrossRef] 20. Tian, L.; Yao, T.; Gao, Y.; Thompson, L.; Mosley-Thompson, E.; Muhammad, S.; Zong, J.; Wang, C.; Jin, S.; Li, Z. Two glaciers collapse in western Tibet. J. Glaciol. 2017, 63, 194–197. [CrossRef] 21. Kääb, A.; Leinss, S.; Gilbert, A.; Bühler, Y.; Gascoin, S.; Evans, S.G.; Bartelt, P.; Berthier, E.; Brun, F.; Chao, W.; et al. Massive collapse of two glaciers in western Tibet in 2016 after surge-like instability. Nat. Geosci. 2018, 11, 114–120. [CrossRef] 22. Gilbert, A.; Leinss, S.; Kargel, J.; Kääb, A.; Yao, T.; Gascoin, S.; Leonard, G.; Berthier, E.; Karki, A. Mechanisms leading to the 2016 giant twin glacier collapses, Aru range, Tibet. Cryosphere 2018, 12, 2883–2900. [CrossRef] 23. Zhang, Z.; Liu, S.; Zhang, Y.; Wei, J.; Jiang, Z.; Wu, K. Glacier variations at Aru Co in western Tibet from 1971 to 2016 derived from remote-sensing data. J. Glaciol. 2018, 64, 397–406. [CrossRef] 24. Faillettaz, J.; Funk, M.; Vincent, C. Avalanching glacier instabilities Review on processes and early warning perspectives. Rev. Geophys. 2015, 53, 203–224. [CrossRef] 25. Drobyshev, V.N. Glacial catastrophe of 20 September 2002 in North Osetia. Russ. J. Earth. Sci. 2006, 8, ES4004. [CrossRef] 26. Ye, Q.; Zong, I.; Tian, L.; Cogley, J.G.; Song, C.; Guo, W. Glacier changes on the Tibetan Plateau derived from Landsat imagery: Mid-1970s-2000-13. J. Glaciol. 2017, 63, 273–287. [CrossRef] 27. Brando Abroad. Available online: http://www.brandoabroad.com/single-post/2017/09/22/The-life-of- Chinese-construction-workers-a-million-miles-from-home-Days-9-10 (accessed on 9 March 2019). 28. Wenying, W. Glaciers in the north-eastern part of the Ch’ing-hai–hsi-tsang (Qinghai–Xizang) Plateau (Tibet) and their variations. J. Glaciol. 1983, 29, 383–391. [CrossRef] 29. Gruber, S. Derivation and analysis of a high-resolution estimate of global permafrost zonation. Cryosphere 2012, 6, 221–233. [CrossRef] Remote Sens. 2019, 11, 708 18 of 18

30. Shean, D. High Mountain Asia 8-Meter DEM Mosaics Derived from Optical Imagery; Version 1, Tile 391; NASA National Snow and Ice Data Center Distributed Active Archive Center: Boulder, CO, USA, 2017. [CrossRef] 31. Paul, F.; Winsvold, S.H.; Kääb, A.; Nagler, T.; Schwaizer, G. Glacier remote sensing using Sentinel-2. Part II: Mapping glacier extents and surface facies, and comparison to Landsat 8. Remote Sens. 2016, 8, 575. [CrossRef] 32. Jamieson, S.S.R.; Ewertowski, M.W.; Evans, D.J.A. Rapid advance of two mountain glaciers in response to mine-related debris loading. J. Geophys. Res. Earth Surf. 2015, 120, 1418–1435. [CrossRef]

© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).