Analysing the Creep of Mountain Permafrost using High...

Analysing the Creep of Mountain Permafrost using High Precision AerialPhotogrammetry: 25 Years of Monitoring Gruben Rock Glacier, Swiss Alps

A. KaÈaÈb,1* W. Haeberli 2 and G. Hilmar Gudmundsson1

1Laboratory of Hydraulics, Hydrology and Glaciology (VAW), ETH, 8092 Zurich, Switzerland2Department of Geography, University of Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerland

ABSTRACT

Aerophotogrammetrical monitoring of Gruben rock glacier over the period 1970 to 1995 resultsin a unique time series documenting the three-dimensional surface kinematics of creepingmountain permafrost. In places, the area under study is a�ected by historical ¯uctuations of thepolythermal Gruben glacier. Changes in elevation and surface velocities were measured over ®veconsecutive ®ve-year periods using an advanced photogrammetric monoplotting technique ofmultitemporal stereo models. The measurements are based on a regular grid with a mesh width of25 metres and have an accuracy of a few centimetres per year. Although surface lifting occurred inplaces and within individual time intervals, surface subsidence predominated at an average rate ofa few centimetres per year in the `periglacial' part of the rock glacier and of a few decimetres peryear in the `glacier-a�ected' part of the rock glacier which still contains some dead glacier ice inpermafrost. Fluctuations in horizontal surface velocities seem to correlate with temporal changesin surface elevation. Analysing ¯ow along principal trajectories and interpreting the advance rateof the front leads to an age estimate of the rock glacier of a few millennia. Dynamic e�ects ofthree-dimensional straining within the creeping permafrost as computed from the measuredsurface velocity ®eld are estimated to potentially contribute to surface heave or subsidence in thesame order of magnitude as the observed vertical changes. Temporal variations of surfacealtitudes at Gruben rock glacier show distinct similarities with mass balance and surface altitudevariations determined on nearby glaciers but at a greatly reduced amplitude. This similarity mayindicate that the same climatic forcing (summer temperatures?) could possibly have a predomi-nant in¯uence on permafrost aggradation/degradation as well as on glacier mass balance inmountain areas. # 1997 John Wiley & Sons, Ltd.

REÂ SUMEÂ

Le suivi aeÂ rophotogrammeÂ trique du glacier rocheux de Gruben pendant la peÂ riode de 1970 aÁ1995 a donneÂ une seÂ rie unique d'observations qui ont permis de preÂ ciser en trois dimensions lacineÂ matique du creep du pergeÂ lisol de montagne. La zone eÂ tudieÂ e a permis aussi de suivre les¯uctuations historiques du glacier Gruben. Des changements en altitude et des vitesses de surfaceont eÂ teÂ mesureÂ es au cours de 5 peÂ riodes de 5 ans, en utilisant une technique avanceÂ e pour eÂ tudierles modeÁ les steÂ reÂ oscopiques multitemporels. Les mesures ont eÂ teÂ e�ectueÂ es avec une preÂ cisionde quelques centimeÁ tres par an selon une grille reÂ gulieÁ re ayant une maille de 25 meÁ tres. Bien quedes souleÁ vements de surface se produisent localement et temporairement, des pheÂ nomeÁ nes desubsidence preÂ dominent. Ils ont une valeur moyenne de quelques centimeÁ tres par an dans la

CCC 1045±6740/97/040409±18$17.50 Received 21 August 1997# 1997 John Wiley & Sons, Ltd. Revised 19 October 1997

Accepted 20 October 1997

Permafrost and Periglacial Processes, Vol. 8: 409±426 (1997)

* Correspondence to: Dr A. KaÈ aÈ b, Department of Geography, University of Zurich, Winterthurerstrasse 190, 8057 Zurich,Switzerland. e-mail: [email protected]

Contract grant sponsor: Swiss National Science Foundation

partie peÂ riglaciaire du glacier rocheux et de quelques deÂ cimeÁ tres par an dans la partie ouÁ de laglace d'origine glaciaire est comprise dans le pergeÂ lisol. Les variations des vitesses horizontales desurface paraissent se correÂ ler avec les variations verticales de celle-ci. L'analyse de l'eÂ coulement lelong des trajectoires principales et l'interpreÂ tation de l'avanceÂ e du front conduisent aÁ estimer l'aÃ gedu glacier rocheux aÁ quelques milliers d'anneÂ es. Les e�ets dynamiques des contraintes en troisdimensions qui deÂ forment le pergeÂ lisol (telles qu'elles sont calculeÂ es aÁ partir du champ desvitesses de surface) peuvent potentiellement contribuer aÁ expliquer des mouvements de souleÁ ve-ment et de subsidence du meÃ me ordre de grandeur que les mouvements verticaux observeÂ s. Lesvariations temporelles des altitudes de surface du glacier rocheux de Gruben montrent dessimilariteÂ s distinctes des variations de masse et des altitudes de surface deÂ termineÂ es sur desglaciers proches mais avec cependant une amplitude consideÂ rablement reÂ duite. Cette similariteÂpeut indiquer que le meÃ me e�et climatique (les tempeÂ ratures d'eÂ teÂ ?) ont une in¯uence preÂ domin-ante sur les pheÂ nomeÁ nes d'aggradation et de deÂ gradation du pergeÂ lisol comme cela est connupour les balances de masses glaciaires des reÂ gions montagneuses.# 1997 JohnWiley & Sons, Ltd.

Permafrost Periglac. Process., Vol. 8: 409±426 (1997).(No. of Figs: 17. No. of Tables: 0. No. of Refs: 31.)

KEY WORDS: rock glacier; photogrammetry; monitoring; creep velocity; deformation

INTRODUCTION

Rock glaciers represent typical phenomena ofsteady ice and debris transport in the permafrostof cold mountain areas. Negative subsurfacetemperatures and generally high ice contentswithin the creeping mountain permafrost lead tocharacteristic spatial surface velocities of centi-metres, decimetres and sometimes up to a fewmetres per year. Temporal changes in ground iceconditions generally involve strong e�ects fromlatent heat exchange coupled with slow and delayedheat di�usion processes and correspondingly weakheat ¯uxes. The resulting thermal inertia with lowrates of freezing and thawing at the permafrosttable and the permafrost base stabilize boundaryconditions, so that the geometry and ¯ow regime ofperennially frozen ground on mountain slopeschange slowly with time. Thus, they are mainlya�ected by long-term climatic trends. Analysis ofsmall permafrost creep rates, and their temporal aswell as spatial variations, requires the applicationof precise high-resolution techniques. Long-termmonitoring is necessary for accurately document-ing slow temporal changes in geometry and ¯ow(Wagner, 1992; 1996). As demonstrated by numer-ous earlier studies, aerophotogrammetry representsa powerful remote sensing tool for measuringgeometrical changes of alpine permafrost surfaces(e.g. Messerli and Zurbuchen, 1968; Barsch andHell, 1976; Haeberli et al., 1979; Evin and Assier,1982; Haeberli and Schmid, 1988; Vonder MuÈ hlland Schmid, 1993; Kaufmann, 1996).

Large-scale aerial photographs have been takenannually by the Federal O�ce of Cadastral Surveyssince 1970 at the Gruben locality in connectionwith the monitoring of dangerous periglacial lakes.A recent investigation as part of the Swiss NationalResearch Programme 31 `Climate Changes andNatural Disasters' successfully applied photo-grammetric techniques for early recognition andprevention of ¯ood and debris-¯ow hazards con-stituted by the growth of these lakes on and aroundthe Gruben rock glacier and the nearby Grubenglacier (KaÈ aÈ b and Haeberli, 1996; KaÈ aÈ b et al.,1996). The availability of excellent photogram-metric material enabled the study of creep pro-cesses and the monitoring of long-term geometricdevelopments within perennially frozen sedimentsand massive ground ice of the large active Grubenrock glacier. Shallow core drilling, geophysicalinvestigations and lichen-cover studies (Barschet al., 1979; Haeberli, 1985; Haeberli et al., 1979;King et al., 1987) have been carried out givingvaluable support to the photogrammetric observa-tions. With its exceptionally long and detailedphotogrammetric record, Gruben rock glacier isnow an important part of the permafrost monitor-ing network in the Swiss Alps (Haeberli et al.,1993), which aims to document changes inmountain permafrost conditions as related toclimatic forcing.

Results from the aerophotogrammetric monitor-ing of the Gruben rock glacier for the period 1970to 1979 were published by Haeberli and Schmid(1988), and for the period 1970 to 1991 by Haeberli

# 1997 John Wiley & Sons, Ltd. Permafrost and Periglacial Processes, Vol. 8: 409±426 (1997)

410 A. KaÈ aÈ b, W. Haeberli and G. H. Gudmundsson

et al. (1993). In this paper, data are presented onthe geometrical development of the creeping rockglacier permafrost from 1970 to 1995. In addition,new results from analysing creep rates and strain®elds are discussed. For that purpose, the wholephotogrammetric time-series was remeasured usinga newly developed, more precise technique.

SITE DESCRIPTION

Gruben glacier, Wallis, Swiss Alps, ¯ows from thepeak of Fletschhorn (3993 m ASL) between theSaas valley and the Simplon pass (Figure 1) downto a well-developed cirque. Within this cirque androughly parallel to the orographic right side of thepolythermal Gruben glacier tongue, Gruben rockglacier developed at an altitude of about 2900 mASL within thick accumulations of periglacialdebris and morainic material in an environmentof discontinuous mountain permafrost (Figures 2and 3). A number of periglacial lakes are located atthe margin, as well as in front of the glacier tongueand on the rock glacier itself. In both 1968 and1970, outburst ¯oods from glacier-dammed lake 3in the contact zone between the glacier tongue andthe rock glacier caused damage in the nearbyvillage of Saas Balen. In order to prevent further¯oods and debris ¯ows, geophysical and geodetic

investigations were carried out, aerophoto-grammetric monitoring was started, a protectiveplan was devised and corresponding constructionwork was undertaken (RoÈ thlisberger, 1979; KaÈ aÈ band Haeberli, 1996; Vonder MuÈ hll et al., 1996).

During its Holocene and historical advances,Gruben glacier must have repeatedly overriddenthe upper part of the rock glacier (Haeberli et al.,1979). During glacier retreat periods, debris-covered dead ice was deposited on top of therock-glacier permafrost, a�ecting the dynamics ofthe rock glacier in its upper and orographic leftpart (Figure 4). In the following, this part of thecreeping permafrost body is called the glacier-a�ected part, whereas the lower and orographicright part, not directly in¯uenced by recent glacieractivity, is called the periglacial part.

Characteristic mean annual surface temperaturesof the rock glacier are close to ÿ1 8C today. Thethickness of the relatively warm permafrost maystill be in¯uenced by ground thermal conditionsduring the somewhat colder past centuries(Haeberli, 1985). The thickness is known to be upto 100 and more metres in the periglacial part, andcharacteristically a few tens of metres in the glacier-a�ected part (Haeberli, 1985; King et al., 1987). Inthe periglacial part of the rock glacier, a transversebedrock riegel was detected at a depth of 30 metresbelow surface (unpublished seismic refractionsoundings, cf. Figure 8).

Figure 1 Gruben rock glacier (left) and Gruben glacier (right) (1985, photo W. Schmid).


Aerial Photogrammetry 411

Figure 2 Gruben rock glacier (top), tongue of Gruben glacier (bottom) and periglacial lakes (aerial photo taken by Federal O�ceof Cadastral Surveys).

Figure 3 Map of Gruben rock glacier and its location in Switzerland. Letters A±E and numbers 1±7 referred to in Figures 8 and 16



In the mid 1960s, two crevasses formed at theupper margin of dead ice occurrences within therock glacier permafrost. These two surface depres-sions ®lled with water and progressively enlarged

by thermal convection. This led to intensi®edmelting of the dead glacier ice and probably alsoof ice contained within underlying permafrost. Oneof these two thermokarst lakes, lake 6, only ®lls upduring winter and spring. In late autumn, freezingwater seals the lake bottom. Subsequently, the lake®lls with snow, water and ice until the next meltingseason starts. When the lake is completely full ofwater, the frozen bottom leaks and the lake slowlyempties by draining through either remaining deadice layers or underlying permafrost. In contrast,the permafrost and dead ice damming the otherthermokarst lake, lake 5, appear to be impermeableand the water level remains nearly stable through-out the year. In 1994, lake 5 had a surface area ofabout 10,000 m2 and a volume (determined bysonar soundings) of about 50,000 m3. Its annualarea and volume growth rate reached values of1500 m2 and 7500 m3, respectively (KaÈ aÈ b et al.,1996; cf. Haeberli, 1980). In view of the increasingrisk of a lake outburst, again endangering thevillage of Saas Balen, most of the water within thisthermokarst lake was pumped out in summer 1995as part of the ¯ood protection plan for the wholeGruben area (KaÈ aÈ b et al., 1996). It is planned tokeep the future water level low by an open ditch.Based on results from aerophotogrammetric ana-lyses of potential hazards from such lakes, addi-tional ¯ood protection measures were taken at theglacier margin. The water level of lake 3, situated ata retreating calving front of the glacier, was loweredand regulated by the excavation of a ditch insummer 1996. The remaining lake volume waslargely ®lled up with the excavated debris. Thearti®cial outlet of the moraine lake in front of theglacier tongue was stabilized by concrete injectionsand deepened in order to increase its waterretention capacity.

PHOTOGRAMMETRIC METHODS

Changes in surface elevation of creeping mountainpermafrost are caused by the sum of (a) sedimentaccumulation/evacuation, (b) three-dimensionaldeformation (extension/compression) and (c)terrain subsidence caused by either thawing ofsupersaturated permafrost or frost heave due tofreezing of water. As a complex function ofsediment accumulation at the surface, ¯ow fromhigher to lower altitudes and cumulative verticalstraining of supersaturated permafrost, thawingand freezing can take place either at the permafrosttable, or within the permafrost body, or at the

Figure 4 Schematic diagram illustrating how Holoceneadvances and retreats of the Gruben glacier a�ect the rockglacier.



permafrost base (cf. Haeberli and Vonder MuÈ hll,1996). Vertical displacements are the result ofspatial/dynamic processes. Thus, determination ofsurface displacements in full three dimensions isnecessary for better understanding the physicalprocesses involved.

Aerophotogrammetric determination of digitalterrain models (DTMs) and subsequent compar-ison of multitemporal DTMs is an e�ective andwell-established technique to exactly de®ne terrainsurfaces and their temporal changes in elevation.The method uses monotemporal stereo models,composed by at least two overlapping photographswhich are taken from di�erent places. The terrainpoint A (see Figure 5) is computed by intersectingtwo spatial rays, each ®xed by the known projec-tion centres and the projections A01�t1� and A02�t1� ofthe selected terrain point. Repeating the procedureat other points gives a DTM of time t1 , andrepeating the procedure using photographs takenat time t2 gives point B respectively a DTM oftime t2 and ± by comparison of two DTMs ± thearea-wide changes in surface elevation.

With respect to surface displacements, a newhigh-precision photogrammetric method wasdeveloped and applied, based on earlier conceptspublished by Finsterwalder (1931), Hofmann(1958), GruÈ n and Sauermann (1977), Flotron(1979), Saksono (1984) and Armenakis (1984)and translated into a programme for a computer-aided photogrammetric compilation system(analytical plotter) (KaÈ aÈ b, 1996). The method

uses multitemporal stereo models composed byaerial photographs taken at di�erent times andfrom di�erent places (e.g. photo 1 (t2) and photo 2(t1) in Figure 5).

Between time t1 and time t2 the point A hasmoved to point C, passively riding on the movingsurface. A block or a stone on the surface of thecreeping permafrost could, for instance, be asuitable target. The projection A02�t1� of such apoint is chosen from the photograph taken at timet1 . Intersecting the spatial ray ®xed by this imagepoint and the known projection centre with theterrain surface represented by the DTM of time t1gives the ground coordinates of point A. Thisprocedure is called monoplotting. The image pointC01 at time t2 , C

01�t2�, which corresponds to A02�t1�,

can now be found using the stereoscopic overlap.The operator thereby cancels the terrain movementwhich has occurred between the times t1 and t2 bydisplacing one image while simultaneously lookingat the multitemporal photographs. This simulta-neous and stereoscopic observation supports theidenti®cation of corresponding points, improvesthe accuracy of the measurements and, addition-ally, indicates whether a local displacement re¯ectsin a signi®cant way its surrounding terrain. In thesame way as for the spatial position of point A, theposition of point C is computed by monoplotting.Thus, spatial displacements of surface points canbe deduced area-wide.

The terrain surface of which the velocity ®eldshall be determined has to ful®l basic requirements:

Figure 5 Monotemporal photogrammetric stereo models are used to determine surface elevation and its changes. Usingmultitemporal stereo models composed by photographs taken at di�erent times from di�erent places, surface displacements can bemeasured as well.



(1) the terrain has to show corresponding featuresat every time of photography; (2) the displacementsmust be larger than the accuracy of the method toobtain signi®cant results; and (3) terrain changesbetween the times of photography, e.g. caused bythawing or terrain slip, should not prohibit theidenti®cation of corresponding points. Require-ments (2) and (3) can be satis®ed by choosing asuitable time interval between the photo missions.The photogrammetric technique described hereworks especially well for determining the creep ofpermafrost surfaces and is also suitable for observ-ing glacier ¯ow and slope movements (KaÈ aÈ b, 1996).Owing to requirement (1) it is not possible todetermine displacements of snow-covered terrain.Maximum accuracy of the method is assumed to beabout +30 mm in image scale or, in the case of thephotogrammetric parameters of the Gruben mis-sions, about +0.2 m on the ground. The actualaccuracy of the displacements measured photo-grammetrically can be deduced by repeatingmeasurements using independent multitemporalstereo models and by comparing the results withgeodetic stake measurements. The error (root meansquare) of the displacement measurements onGruben rock glacier is about +0.4 m or, in the

case of a ®ve-year interval between two photomissions, about +0.08 m per year (m/a).

SURFACE ELEVATION

In order to determine changes in surface elevationof Gruben rock glacier between 1970 and 1995,DTMs of 1970, 1975, 1979, 1985, 1991 and 1995were measured photogrammetrically (time ofphotography was autumn in every case). Ameasuring interval of about ®ve years was chosenon the basis of results from earlier velocity andaltitude change determinations at the rock glaciersurface. Unfavourable terrain and ¯ying con-ditions in 1980 and 1990 prevented keeping exact®ve-year intervals throughout the entire obser-vation period. Each of the six DTM grids has aregular grid distance of 25 m. (Horizontal dis-placements were later measured at the same DTMpoints). The changes in surface elevation deducedfrom the vertical di�erences between identical gridpoints in corresponding DTMs were smoothed bythe Gaussian low-pass ®lter. This procedurereduces the data noise without causing a system-atic elevation shift. Figure 6 shows the average

Figure 6 Changes in surface elevation of the Gruben rock glacier, 1970±95.



changes in surface elevation between 1970 and1995; Figure 7 shows the individual ®ve-yearvariations thinned out for a 50 m grid. The surfacepro®les are shown in Figure 8, and the changesin the vertical displacement rates between theindividual ®ve-year time intervals are presented inFigure 9.

Between 1970 and 1995 the periglacial part of theGruben rock glacier showed a total surface lower-ing of about ÿ0.05 m/a on average with maximareaching ÿ0.10 m/a and more in the central parts.On the other hand, the average loss in surfaceelevation of the glacier-a�ected part containingremains of dead ice is about ÿ0.20 m/a during thesame period. The higher ice content of the dead ice

(about 100%), in contrast to the lower ice contentof the perennially frozen debris (about 50±70%;Barsch et al., 1979), probably causes the observedhigher subsidence rates. Thus, the two parts of therock glacier, i.e. the periglacial and the glacier-a�ected area, clearly di�er with respect to temporalchanges in surface elevation (Figure 6).

Speci®c details of the area-wide changes insurface elevation are illustrated in Figures 6 and 7.The formation and fast growth of the thermokarstlakes described above (cf. Figure 3) brought aboutlarge losses of ice, inducing pronounced surfacelowering at lakes 5 and 6. The retreat of the glacierice margin, which prevailed during the period 1970to 1995 and the years before, caused an important

Figure 7 Vertical velocities at about ®ve-year intervals. Lakes indicated are those at the end of the measurement period.



but decelerating surface lowering to the south-westof lake 3 in the zone of direct contact between theglacier and the rock glacier (Figure 9). Between thisarea and thermokarst lake 5, changes in elevation

show gradual spatial variation. This may indicatedi�erent thicknesses of the insolating debris coverfrom active layer adjustment over increasing timeperiods and, thus, di�erent e�ects from former

Figure 8 Surface pro®les in south-western and south-eastern directions. Two times vertical exaggeration. Start and end pointsA±D, see Figure 3. Bedrock and dead ice layer are schematically indicated.

Figure 9 Variations with time of changes in surface elevation, Gruben rock glacier.



glacier overriding (cf. Figure 8, bottom). Dynamice�ects from compression and extension of the ice/debris mixture and the surface-debris layer due tothe ¯owing of the entire mass towards the topo-graphic depression left by the retreating glacier mayalso play a role. Such dynamic e�ects obviouslytake place in the case of the surface rise to thenorth-west of the glacier-dammed lake 3. Underthe in¯uence of enhanced energy in¯ux from solarradiation and warm lake water, the dead iceoriginally existing in that area may already havemelted away to a large degree as indicated by therelatively stable lake border at the foot of thecorresponding slope. In fact, the remaining debrismay have started slowing down the ¯ow of the ice/debris mixture towards the lake and, hence, causedhorizontal compression and surface uplift. Nearthe rock glacier front, the relatively stable, orslightly rising, surface altitude can be explained bythe advance of the creeping permafrost, as dis-cussed below.

The changes in the rates of height variation(Figure 9) have been calculated as di�erencesbetween the average annual changes in elevation.While individual acceleration values are hardlysigni®cant owing to the accuracy of single elevationmeasurements, the spatial distribution shows aclear pattern. Thus, Figure 9 makes clear the di�er-ences between the changes in surface elevationof the ®ve-year intervals, shown in Figure 7.For instance, between 1970±75 and 1975±79 the

vertical surface velocities were increasing (acceler-ating) in most parts. In other words, surfacelowering diminished from 1970±75 to 1975±79.Also the general decrease of subsidence rates after1979 in the zone of direct contact between theglacier and the rock glacier can be recognized inFigure 9 as vertical acceleration in a mathematicalsense. The increasing subsidence rate at lake 5between 1985±91 and 1991±95 is due to theaccelerating growth of this lake (KaÈ aÈ b et al., 1996).

Although changes in elevation did not varymuch between 1970 and 1995 with respect to thebasic spatial pattern, their variations in timeshow signi®cant trends. Continuous lowering ofthe glacier-a�ected part of Gruben rock glacier asaveraged over this entire area indicates, asexpected, the existence of dead ice in thermal dis-equilibrium (Figure 10, bottom). Annual loss insurface elevation (Figure 13, bottom) was decreas-ing from 1970 to 1979 and increasing again from1979 to 1991. This development may have beenin¯uenced mainly by climatic e�ects as shown bycomparison with variations in surface elevationand net mass balance of the nearby tongue ofGruben glacier (KaÈ aÈ b, 1996), as well as with thevariations in net mass balance of Gries glacier(Funk et al., 1997; Figure 14).

In contrast to the continuous lowering of theglacier-a�ected part of Gruben rock glacier,its periglacial part became thicker from 1975 to1979 (Figure 10, top). The corresponding annual

Figure 10 Cumulative change in surface elevation of the periglacial part (top) and the glacier-a�ected part (bottom). Note thedi�erent scales of the two ®gures.



changes (Figure 13, top) are considerably lowerthan those of the glacier-a�ected part, but showsome correlation with the above-mentioned glaciermass balance variations as well. This indicates thathigh-frequency (annual to multi-year) processes ofice loss and ice formation at the permafrost table ofthe rock-glacier part, and mass balance variationsof the nearby glacier, may both be governed bysimilar climatic parameters.

SURFACE DISPLACEMENTS

The surface displacements of Gruben rock glacierbetween 1970±75, 1975±79, 1979±85, 1985±91 and1991±95 were determined using the simultaneousmonoplotting technique described above. The rawdata were mostly measured at a grid distance of25 m. Average values of 1970±95 are presented inFigure 11. Accuracy of one single vector is about0.08 m/a. Again, the velocity ®eld of Gruben rockglacier clearly illustrates di�erences between theperiglacial and the glacier-a�ected part. From 1970to 1995, the glacier-a�ected part with its dead ice atthe surface has moved `back' towards the glacier

tongue, with surface velocities reaching up toseveral metres per year. The highest velocitiesappear between the thermokarst lake 6 and theglacier-dammed lake 3. The di�erent generaldirection of surface ¯ow within the glacier-a�ectedpart, with respect to the general ¯ow directionwithin the periglacial part, corresponds to thedi�erent directions of maximum surface slope ineach of the two areas. Since reliable information onthe thickness of the viscously deforming layer islacking, the exact cause of the velocity maximaobserved in Figure 11 remains open. Changes inrheology resulting from di�erences in ice contentand thickness, as well as from slope variations or acombination thereof, are possible explanations.Like the above-mentioned changes in elevation,surface displacements between the thermokarstlake 5 and the glacier tongue also show gradualvariations but appear to closely follow gradients insurface slope (Figure 8).

The surface of the periglacial part creeps withvelocities of some decimetres per year, butincreases sharply to about 1 m/a as the permafrostcreeps across and behind the bedrock riegel. This isprobably due to combined in¯uences of changes in

Figure 11 Surface displacements of Gruben rock glacier, 1970±95.



thickness and in surface slope of the frozenmaterial. The lack of deceleration across the littleterrace behind the front (at an altitude of about2820 m; Figure 8) may be the result of stresscoupling or thickness variations of the creepingpermafrost layer. The area above the bedrock riegelseems to function like a reservoir ®lled withperennially frozen debris. The debris has mainlyaccumulated by repeated Holocene glacieradvances, but scree from frost weathering in thewall delimiting the rock glacier to the north-westmay also be a factor. This mass of ice-richpermafrost is now ¯owing over its rock dam,forming the main front and a smaller secondaryfront at the orographic right side.

Figure 13 (bottom) shows the evolution in timeof spatially-averaged surface velocities of the wholeglacier-a�ected part of the Gruben rock glacier.The spatial distribution of the temporal changes inhorizontal velocity (i.e. the horizontal accelerationor deceleration) is inhomogeneous (Figure 12).Average surface velocities and average elevation

changes of the glacier-a�ected part barely corre-late. But in the periglacial part, a clear correlationcan be observed (Figure 13, top). Moreover, thespatial distribution of the temporal variations ofhorizontal and vertical velocities partially corre-lates (Figures 9 and 12). Although the coupling ofchanges in elevation and horizontal surfacevelocities is not surprising, di�erent internal andexternal causes may be involved.Assuming constant permafrost creep rates over

the last few thousand years, ¯ow lines werecalculated for a number of manually selectedstarting points (Figure 15). The time taken for aparticle to travel down the whole length of the rockglacier gives an approximate estimate of the age ofthe permafrost surface. The central ¯ow lines at thefront indicate about 1000±2000 years, but more forthe marginal areas. Glacial overriding of the upperrock-glacier parts may have in¯uenced the velocity®eld and, therefore, the age estimate above.Furthermore, the age of the front surface is notthe age of the rock glacier. Comparing the advance

Figure 12 Changes in horizontal surface velocity with time, Gruben rock glacier.



of the rock-glacier front with the measured surfacevelocities at selected pro®les (Figures 11 and 16)shows that the main front advanced an averagedistance of 3 m from 1970 to 1995. During thesame time interval the surface displacement justabove the front totalled 17 m. If the di�erence isdue to (a) the melting of ice within the advancingfrozen material and (b) the vertical variation ofvelocity, rough estimates can be made of averageice contents of the rock-glacier permafrost byassuming characteristic vertical velocity pro®les.Talus height at Gruben rock-glacier front is about

30%, indicating some type of internal deformationof the creeping permafrost (Haeberli, 1985;Wahrhaftig and Cox, 1959). With the assumptionof 100% basal sliding, the ratio between the mean¯ow velocity ± equalling the surface velocity at thefront ± and the rate of advance gives a maximumpossible average volumetric ice content of up to85%. On the other hand, assuming a linear velocitypro®le with depth and a depth-averaged ¯owvelocity equalling half the surface velocity providesa lower estimate of 60±70% by volume. Annualmelt rates at the permafrost table within the steep

Figure 13 Changes in vertical and horizontal surface velocities of the periglacial part (top) and the glacier-a�ected part (bottom).Note the di�erent scales of top and bottom ®gures.

Figure 14 Net mass balance of Gries glacier (Funk et al., 1997) and change in elevation and modelled mass balance of the tongueof Gruben glacier. The changes in surface elevation of the tongue of Gruben glacier are not only caused by the local mass balance,the direct climate signal a�ecting the rock glacier, but also clearly in¯uenced by changing ice ¯ow. In order to estimate these e�ects,the local mass balance and the vertical ice ¯ow, both together e�ecting changes in elevation, have been modelled using the kinematicborder condition at the surface (KaÈ aÈ b, 1996).



front amount to a few decimetres per year. Suchvalues are considerably higher than the measuredsubsidence rates at the rock-glacier surface but maynot be unreasonable for non-permafrost conditionssuch as they must be expected in the ®ne-grainedrock-glacier front exposed to the sun (south-west).Also, shallow core drillings to a depth of 7 metresand related borehole measurements on the Grubenrock glacier by Barsch et al. (1979) gave a

volumetric ice content of 90% at the permafrosttable, and 50±70% as an average value. Assumingthat the ratio of about 15% between the rate ofadvance and the surface ¯ow velocity has remainedconstant through time, and integrating this rateover today's ¯ow trajectories, leads to an age esti-mate for Gruben rock glacier of roughly 104 years.The right-lateral front of Gruben rock glaciershowed no signi®cant advance between 1970 and

Figure 15 Flowlines interpolated from measured surface displacements assuming a steady state ¯ow ®eld like 1970±95. Smallpoints each represent 10 years, bold points 200 years age di�erence.

Figure 16 Surface displacement and rock glacier front between 1970 and 1995 at selected pro®les (see Figure 3).



1995 (Figure 16). This points to the complexity ofthe evolution with locally occurring inactive partsand/or possible intermittent less active stages. Inany case, it is reasonable to assume that the rockglacier developed since the end of the last Ice Agethrough the entire Holocene period.

SURFACE DEFORMATION

Calculation of Surface Strain Rates

The pattern of surface deformation can best beunderstood by analysing the strain-rate variationat the surface. Strain rates can, in principle,be calculated by applying the expression _eij ��vi;j � vj;i�=2 directly to the raw data, but strainrates derived in this way will be sensitive to noise inthe data. In fact, the error D_eij in the strain rates,will, for a given error Dv associated with eachvelocity determination, increase with the inverse ofthe spatial distance Dr between data points. ForDr� 25 m and Dv� 0.035 m/a, one ®nds, forexample, that D_eij � 0:002 1=a. This error estimateis larger than most of the expected strain-ratevalues. Strain-rate variations having longer wave-lengths can, however, be determined with a muchhigher accuracy but they will not be detectable inthe noise if the amplitudes of the high-frequencyerror components are not eliminated or at leastreduced signi®cantly. Some kind of error reductionand spatial averaging is imperative if strain rates areto be accurately determined.

The Wiener ®lter method was used to reduce thevelocity errors (Vaseghi, 1996). The velocitydistributions vx(x, y) and vy(x, y) were transformedinto frequency space using the 2D forward Fouriertransformation, de®ned as

f �kx; ky� �Z �1ÿ1

Z �1ÿ1

f �x; y� ei�kxx�kyy� dx dy �1�

An optimal Wiener ®lter

W�k� � jS�k� j 2jS�k� j 2 � jN�k� j 2 �2�

where k ��k2x � k2y�

qwas determined by

constructing a signal model S(k) and a noisemodel N(k) from the calculated power spectrumof the velocities. The strain rates _eij�kx; ky� werecalculated from the noise-reduced velocities by

multiplication with wave numbers. As an example,the formula for _exx�kx; ky� was

_exx�kx; ky� � ÿikxW�k�vx�kx; ky� �3�

The resulting Wiener ®lter was e�ectively a low-pass ®lter with a cuto� at a wavelength of about100 m. At this wavelength the errors in the strainrates are estimated to be about 0.001 1/a. Withincreasing wavelengths the errors decrease rapidly.

Surface Strain Rates

Figure 17 shows the pattern of calculated surfacestrain rates. The arrows represent the horizontalcomponents, and the contour lines the verticalstrain rates. Vertical strain rates were calculatedfrom the horizontal ones using the incompress-ibility condition _eii � 0. There are two regions ofhigh surface strain rates: (1) the front of theperiglacial part, in the lower left of Figure 17,and (2) the steep area of the glacier-a�ected part inthe upper right of the ®gure. The vertical strainrates are generally found to be negative (com-pression) along the surface, and horizontal strainrates are to a large extent positive (extension). Thissituation sharply contrasts with the situation onother alpine rock glaciers such as MurteÁ l andMuragl, where horizontal ¯ow in the lower parts ismostly compressive (e.g. Barsch and Hell, 1976;Vonder MuÈ hll and Schmid, 1993).

Focusing on the front of the periglacial part, itcan be seen from Figure 11 that the direction of the¯ow on the orographic right side changes markedlysomewhere close to the point E on Figure 3. Theresulting strain-rate pattern can be understood interms of the general ¯ow characteristics of con-vergent glacier ¯ow (Gudmundsson, 1994), and canbe seen to be the consequence of increasing surfaceslope towards the front. The ¯ow convergence ofthe periglacial part at the height of point E shiftsthe location of the maximum speed towards the leftmargin and results in a sharp increase in surface¯ow velocities. This change in the ¯ow regime is thecause of the observed longitudinal extension andthe much smaller concomitant transversal com-pression along the centreline (Figure 17), as well asthe transverse extension on the orographic rightside. The result is a transverse band of verticalcompression with a maximum at the centre-line. This ¯ow convergence is followed by a ¯owdivergence as the periglacial part ¯ows pastpoint E, leading to a transverse extension and a



corresponding vertical compression. Somewhatsurprisingly, the ¯ow velocities increase at thesame time, presumably related to an increase insurface slope (see Figure 8).

The increase in surface velocities appears toolarge to be caused by the horizontal ¯ow converg-ence alone. The decrease in rock glacier thicknessassociated with the anticipated bedrock riegel mayplay some role in shaping the spatial velocityvariation. It should also be noted that the surfaceheight variations around the riegel resemble astanding wave of the type predicted from a ¯owover basal undulations protruding into a deform-able medium (JoÂ hannesson, 1992).

From the north-west margin along the mean¯ow direction towards the south-east margin of theglacier-a�ected part, the velocities increase at ®rstas the slope increases, but then in general decreaseas the margin is approached. This causes a shiftfrom predominantly extensive to compressive ¯ow.Small deviations from the general pattern(Figure 17) seem to be related to short scale(5100 m) slope variations (see Figure 8).

The gradual spatial variations of strain ratesbetween lake 5 and the glacier margin suggest thatthe spatial variations in surface elevation arepartially a�ected by ice deformation. The strong

vertical extension close to the glacier-dammedlake 3 must be expected to cause vertical thickeningof the dead ice over time. In fact, this is the onlyregion (except for the front of the periglacial part,the thickening of which is due to the advance)where a signi®cant increase in height over time canbe observed (see Figure 6).

Assuming the average thickness of the creepingpermafrost layer to be at least 30 m but not morethan 100 m, and the vertical strain rates to decreaselinearly with depth, the e�ective vertical strain atthe surface can be estimated. The results show thatsurface lowering of some cm/a can be the e�ect ofvertical compression. Owing to the predominantvertical compression of the Gruben rock glacier,surface lifting can hardly be explained by verticalstrain. On the other hand, the relatively smalltemporal variations of vertical strain rates between1970 and 1995 indicate that the changes inelevation observed during the same period maypredominantly be a�ected by climate in¯uenceson freezing and thawing processes at the perma-frost table rather than by dynamic e�ects ofthree-dimensional straining. Freezing and thawingrates at the permafrost base are assumed to have alow frequency and therefore to have little e�ect onthe high-frequency changes in surface elevation.

Figure 17 Horizontal principal strain rates 1970±95 at surface, calculated from measured surface displacements. Vertical strainrates 1970±95 at surface, calculated from horizontal strain rates.



CONCLUSIONS AND PERSPECTIVES

Techniques of analytical aerophotogrammetryhave been applied to monitor the long-termevolution of the Gruben rock glacier, a complexgeomorphological feature with both a periglacialand a glacier-a�ected part. Area-wide velocitiesand changes in elevation of the rock-glacier surfacewere determined in order to analyse the creep ofmountain permafrost. A 25 year time series atGruben rock glacier represents the longest of itskind and quality, and provides the following results:

(1) Growth and degradation of permafrost cantake place simultaneously at di�erent placeswithin the same rock glacier (see also Haeberliand Schmid, 1988) and can vary at yearly timescales.

(2) Changes in surface elevation within the peri-glacial part of Gruben rock glacier, during theobservation period from 1970 to 1995, re¯ectoverall e�ects from sediment accumulation/evacuation, three-dimensional deformation,and freezing and thawing processes. Measuredvalues lie in the range +10 cm/a. Surfacelowering prevails with an average value ofÿ5 cm/a. In the glacier-a�ected part, whichstill has some buried ice, surface lowering isconsiderably higher and averages ÿ0.2 m/a.

(3) Surface velocities reach values of 1 m/a andshow changes of up to 20% within a few years.

(4) For the investigated period, temporal changesin surface elevation do not correlate withtemporal changes in surface strain regime.The latter shows no signi®cant variationsin time. Long-term surface lowering can berelated, however, to the observed overallvertical compression at the surface of aboutÿ0.002 1/a (maxima of up to ÿ0.006 1/a).Thus, only part of the measured surfacesubsidence is assumed to be due to thawing.

(5) Short-term variations of vertical surfacechanges appear to be coupled with synchro-nous variations in horizontal ¯ow velocities.This photogrammetrically documented timeseries shows some similarity with mass balancevariations of the nearby Gruben and Griesglaciers. Possibly, this is a consequence ofcommon climatic forcing (summer tempera-ture?).

(6) Between 1970 and 1995, the front of Grubenrock glacier advanced about 10 cm per year,about 0.15 times the surface velocity. Assum-ing a steady state ¯ow ®eld, the surface of

the rock glacier can be estimated to be 1000±2000 years old. Assuming a constant advancerate of 0.15 times the surface velocity, theGruben rock glacier as a whole can beestimated to have developed and advancedduring the Holocene.

The correlation between the three-dimensionalgeometrical behaviour of Gruben rock glacier anddi�erent glacier mass balance variations shows thatthe long-term observation of rock-glacier kine-matics could provide interesting signals of climatee�ects on mountain permafrost. In order tounderstand these signals, as re¯ected by rock-glacier behaviour, it is necessary to monitor andmodel the energy balance of mountain permafrost.

In view of the inhomogeneous internal structureof the permafrost body and the complex processestaking place at the permafrost base and thepermafrost table, improved physical modelling isrequired to understand the dynamic behaviour ofcreeping mountain permafrost. Combining high-precision and area-wide surface information fromphotogrammetry with depth information fromdrillings and geophysical soundings seems to bemost promising for collecting an adequate database for this purpose.

ACKNOWLEDGEMENTS

The present study was carried out as part ofthe Swiss National Research Programme 31 onClimate Change and Natural Disasters throughproject 4031-34232 with funds of the SwissNational Science Foundation. The photo-grammetric investigations are based on the aerialphotographs taken by the Federal O�ce ofCadastral Surveys (EidgenoÈ ssische Vermessungs-direktion). Special thanks are due to M. Hoelzleand D. Vonder MuÈ hll for critically reading themanuscript.

REFERENCES

Armenakis, C. (1984). Deformation measurements fromaerial photographs. Int. Archives of Photogrammetryand Remote Sensing, XXV(A5), 39±48.

Barsch, D. and Hell, G. (1976). PhotogrammetrischeBewegungsmessung am Blockgletscher MurteÁ l I,Oberengadin, Schweizer Alpen. Zeitschrift fuÈrGletscherkunde und Glazialgeologie, 11(2), 111±142.

Barsch, D., Fierz, H. and Haeberli, W. (1979). Shallowcore drilling and borehole measurements in permafrostof an active rock glacier near the Grubengletscher,



Wallis, Swiss Alps. Arctic and Alpine Research, 11(2),215±228.

Evin, M. and Assier, A. (1982). Mise en eÂ vidence dumouvement sur le glacier rocheux du Pic d'Asti(Queyras, Alpes du Sud, France). Revue de GeÂomor-phologie Dynamique, 31(4), 127±136.

Finsterwalder, R. (1931). Geschwindigkeitsmessungen anGletschern mittels Photogrammetrie. Zeitschrift fuÈrGletscherkunde und Glazialgeologie, 19, 251±262.

Funk, M., Morelli, R. and Stahel, W. (1997). Massbalance ofGriesgletscher 1961±1994: di�erentmethodsof determination. Zeitschrift fuÈr Gletscherkunde undGlazialgeologie, 33(1), 41±56.

GruÈ n, A. and Sauermann, H. (1977). Photogrammetricdetermination of time-dependent variations of detailsof a glacier surface using a non-metric camera. InSymposium on Dynamics of Temperated Glaciers andRelated Problems, MuÈnchen.

Gudmundsson, G. H. (1994). Converging glacial ¯ow ±a case study: the Unteraarglacier. Mitteilungender Versuchsanstalt fuÈr Wasserbau, Hydrologie undGlaziologie der ETH ZuÈrich, 131.

Flotron, A. (1979). Verschiebungsmessungen aus Luftbil-dern. Mitteilungen der Versuchsanstalt fuÈr Wasserbau,Hydrologie und Glaziologie der ETHZuÈrich, 41, 39±44.

Haeberli, W. (1980). Morphodynamische Aspekte ak-tueller Gletscherhochwasser in den Schweizer Alpen.Regio Basiliensis, 21(3), 58±78.

Haeberli, W. (1985). Creep of mountain permafrost.Mitteilungen der Versuchsanstalt fuÈr Wasserbau,Hydrologie und Glaziologie der ETH ZuÈrich, 77.

Haeberli, W. and Schmid, W. (1988). Aerophotogramme-trical monitoring of rock glaciers. In:Proceedings, FifthInternational Conference on Permafrost, Trondheim,Norway. Vol. 1, pp. 764±769.

Haeberli, W. and Vonder MuÈ hll, D. (1996). On thecharacteristics and possible origins of ice in rockglacier permafrost. Zeitschrift fuÈr GeomorphologieN.F., Suppl.-Bd., 104, 43±57.

Haeberli, W., King, L. and Flotron, A. (1979). Surfacemovement and lichen-cover studies at the active rockglacier near the Grubengletscher, Wallis, Swiss Alps.Arctic and Alpine Research, 11(4), 421±441.

Haeberli, W., Hoelzle, M., Keller, F., Schmid, W., VonderMuÈ hll, D. S. and Wagner, S. (1993). Monitoringthe long-term evolution of mountain permafrost inthe Swiss Alps. In: Proceedings, Sixth InternationalConference on Permafrost, Beijing, China. Vol. 1,pp. 214±219.

Hofmann, W. (1958). Bestimmung von Gletschergesch-windigkeiten aus Luftbildern. Bildmessung und Luft-bildwesen, 3, 71±88.

JoÂ hannesson, T. (1992). Landscape of Temperate IceCaps. PhD thesis, University of Washington.

KaÈ aÈ b, A. (1996). Photogrammetrische Analyse zurFruÈ herkennung gletscher- und permafrostbedingter

Naturgefahren im Hochgebirge. Mitteilungen derVersuchsanstalt fuÈr Wasserbau, Hydrologie und Glazi-ologie der ETH ZuÈrich, 145.

KaÈ aÈ b, A. and Haeberli, W. (1996). FruÈ herkennung undAnalyse glazialer Naturgefahren im Gebiet Gruben,Wallis, Schweizer Alpen. Tagungspublikation Inter-praevent 1996, 4, 113±122.

KaÈ aÈ b, A., Haeberli, W. and Teysseire, P. (1996).Entwicklung und Sanierung eines Thermokarstseesam Gruben-Blockgletscher (Wallis). Berichte undForschungen. Geographisches Institut. UniversitaÈtFreiburg/CH, 8, 145±153.

Kaufmann, V. (1996). Der DoÈ sener Blockgletscher ±Studienkarten und Bewegungsmessungen. Arbeitenaus dem Institut fuÈr Geographie der Karl-Franzenz-UniversitaÈt Graz, BeitraÈge zur Permafrostforschung inOÈsterreich, 33, 141±162.

King, L., Fisch, W., Haeberli, W. and Waechter, H. P.(1987). Comparison of resistivity and radio-echosoundings on rock-glacier permafrost. Zeitschrift fuÈrGletscherkunde und Glazialgeologie, 23(1), 77±97.

Messerli, B. and Zurbuchen, M. (1968). Blockgletscher imWeissmies und Aletsch und ihre photogrammetrischeKartierung. Die Alpen, SAC, 3, 139±152.

RoÈ thlisberger, H. (1979). Glaziologische Arbeitenim Zusammenhang mit den SeeausbruÈ chen amGrubengletscher, Gemeinde Saas Balen (Wallis).Mitteilungen der Versuchsanstalt fuÈr Wasserbau,Hydrologie und Glaziologie der ETH ZuÈrich, 41,233±256.

Saksono, T. (1984). Combination of Single Photographsand Digital Height Models for Planimetric and HeightData Extraction. Ohio State University.

Vaseghi, S. V. (1996). Advanced Signal Processing andDigital Noise Reduction. Wiley, Chichester.

Vonder MuÈ hll, D. S. and Schmid, W. (1993). Geophysicaland photogrammetrical investigation of rock glacierMuragl I, Upper Engadin, Swiss Alps. In: Proceedings,Sixth International Conference on Permafrost, Beijing,China. Vol. 1, pp. 214±219.

Vonder MuÈ hll, D. S., Haeberli, W. and KlingeleÂ , E.(1996). Geophysikalische Untersuchungen zur Struk-tur und StabilitaÈ t eines MoraÈ nendammes am Gruben-gletscher (Wallis). Tagungspublikation Interpraevent1996, 4, 123±132.

Wagner, S. (1992). Creep of alpine permafrost, investi-gated on the MurteÁ l rock glacier. Permafrost andPeriglacial Processes, 3(2), 157±162.

Wagner, S. (1996). Dreidimensionale Modellierungzweier Gletscher und Deformationsanalyse von eisrei-chem Permafrost. Mitteilungen der Versuchsanstalt fuÈrWasserbau, Hydrologie und Glaziologie der ETHZuÈrich, 146.

Wahrhaftig, C. and Cox, A. (1959). Rock glaciers in theAlaska Range. Geological Society of America Bulletin,70, 383±436.



Analysing the Creep of Mountain Permafrost using High...

Documents

Transcript of Analysing the Creep of Mountain Permafrost using High...