the ﬁnal phase of dead-ice moraine development: processes

The ®nal phase of dead-ice moraine development: processesand sediment architecture, KoÈtlujoÈkull, Iceland

KURT H. KJáR* and JOHANNES KRUÈ GER *Department of Quaternary Geology, Lund University, SoÈlvegatan 13, S-223 62, Lund, Sweden(E-mail: [email protected]) Institute of Geography, University of Copenhagen, éster Voldgade 10, DK-1350, Copenhagen K,Denmark

ABSTRACT

Consecutive phases of de-icing of ice-cored moraines and the formation of

dead-ice moraine were monitored over a 4-year period at the terminus of the

KoÈtlujoÈkull glacier, Iceland. Particularly, the transition from partially ice-cored

moraine with isolated dead-ice blocks to the ice-free landscape receives

attention in this paper in order to link the ®nal melting processes to the

architecture of the sedimentary end product. In the current humid sub-polar

climate of south Iceland de-icing of partially ice-cored moraines results chie¯y

from melting along the bottom surface of ice-cores with an annual average rate

of 25 cm. The ®nal de-icing is associated with an interrelated group of

re-sedimentation processes and surface features. Series of sinkholes evolve at

the toe of dead-ice blocks, which initiate retrogressive rotational sliding or

backslumping of the ice-cored slopes and the formation of distinct edges and

fractures in the adjacent basins. Although backslumping is the dominant

process in this phase of re-sedimentation, structures resulting from this process

are rarely recognized in the ice-free landscape. As ice-cores gradually diminish

the effect of the latest re-sedimentation events will overprint or destroy most

existing sedimentary characteristics. Thus, in the ice-free landscape, structures

mainly related to the formation of sinkholes and fractures remain imprinted on

the sediment succession. Generally, no inversion of the topography occurs

during the ®nal phase of de-icing. The overall topography recognized in the

late phase of the fully ice-cored terrain is merely lowered and the amplitude of

the relief reduced as de-icing progresses. The sediment architecture of the ice-

free landscape is characterized by heterogeneous and often slumped diamict

sediments with variable thickness and lateral distribution; clast orientation is

related to the direction of slopes, and boulders are found in isolated groups or

in linear arrangements.

Keywords Dead-ice moraine, Iceland, processes, sediment architecture,

supraglacial.

INTRODUCTION

The accessibility to the supraglacial environmentat modern glaciers makes it attractive to study thecoupling between current climates and sedimen-tary processes during the development of dead-ice moraines. Qualitative descriptions of the

melting of debris-mantled ice in the terminusregions of modern continental glaciers serve asmodels for development of dead-ice moraines informer glaciated areas (Sharp, 1949; Boulton,1972; Clayton & Moran, 1974; Shaw, 1979; Shaw& Rains, 1981; Paul, 1983; Brodzikowski & vanLoon, 1991). Although, these models demonstrate

Sedimentology (2001) 48, 935±952

Ó 2001 International Association of Sedimentologists 935

the development of dead-ice moraine topographyand the depositional complexity of supraglacialsuccessions, they often fail to identify and quan-tify the processes associated with de-icing andsupraglacial sediment deposition.

Using the terminology of Clayton (1964) forkarst evolution on stagnant ice, the developmentof dead-ice moraines shows three evolutionarystages: (1) the initial or young phase is related todynamically active ice, where an upward com-ponent of ice movement and overthrusting trans-ports debris towards the glacier surface; (2) themature phase is represented by a fully ice-coreddead-ice ®eld where ice disintegration and mass-movement processes rework the sediment coverrepeatedly, leading to a gradual lowering of theice-cored terrain; and (3) the ®nal, or old, phase isrepresented by a partially ice-cored terrain wherethe former coherent ice mass is disintegrated intoisolated dead-ice blocks capped by multiplere-sedimented deposits. The rapidity with whichthe dead-ice ®eld passes through this series ofchanges is controlled by the contemporary cli-mate, thickness of sediment cover and hydrolog-ical processes (KruÈger, 1994; KruÈger & Kjñr,2000).

Detailed sedimentological work including des-criptions and identi®cation of major reworkingprocesses has been provided from modern gla-ciers (Boulton, 1967; Drozdowski, 1977; Eyles,1979; Lawson, 1979; Watson, 1980; Paul, 1983;KruÈger, 1994). Limited work, however, considersin detail the sediment-process-landform relation-ships associated with the transition from the ®nalphase of ice-melt where blocks of buried ice stillexist to the post-melt landscape of hummockymoraines. Sediment may become reworked aslong as buried ice is present, but it is theprocesses associated with the ®nal ice-decaywhich have the overall impact on the sedimentarchitecture within the ice-free landscape (Law-son, 1988; KruÈger, 1994). Thus, an improvedunderstanding of the processes acting during the®nal ice-decay is important to support the geneticinterpretation of supraglacial sediments in formerglaciated areas.

The objective of this paper is to examine theprocess-sediment-landform relationships associ-ated with transition from ®nal phase of ice-meltto the landscape of dead-ice moraine. The studyarea is the KoÈtlujoÈkull outlet-glacier on the eastside of the MyÂrdalsjoÈkull ice cap in central southIceland. This glacier tongue descends from1200 m a.s.l. on to the MyÂrdalssandur plain at220 m a.s.l.; below the 600 m level the glacier

spreads out to form an expanded piedmont outlet-glacier with an ice front nearly 12 km long(Fig. 1).

TERMINOLOGY AND METHODS

Two ablation processes are distinguished: back-wasting de®ned as the sub-horizontal retreat ofnear-vertical free ice-walls, or steep ice-coredslopes; downwasting de®ned as the thinning ofice-cores by melting along the top and bottomsurfaces (Eyles, 1979; KruÈger, 1994). Figure 2Aand B summarize the mode and rate of ablationby these processes in, respectively, the fullyand partially ice-cored moraines (KruÈger & Kjñr,2000). De-icing refers to the close interrelation-ship between the ice-core and the sedimentcover, e.g. de-icing of ice-cored moraines. Also,ice-cored moraine is used for areas with sedi-ment covered stagnant glacier ice, whereasdead-ice moraine is used for ice-free areascreated as a result of de-icing of ice-coredmoraines. As ice-cores gradually diminish thesediment cover is reworked, because of back-slumping and other re-sedimentation featuressuch as fall-sorting, sinkholes and extensionfractures (Fig. 3).

Mapping of areas for process and landformstudies was completed by precision levellingusing a GTS-6 Topcon instrument with an accu-racy of �1 mm. Three-dimensional models of theterrain and contoured diagrams of surface fea-tures are based on an ordinary kriging inter-polation using a GIS (geographical informationsystem) platform.

A selected area with partially ice-cored mor-aine was monitored during a 4-year period torecord the rate of surface lowering and to relatethis to changes in surface features. Surfacelowering was measured by making repeatedstadia surveys from a stable benchmark in theglacier fore®eld to points on the partially ice-cored terrain. The progressive downwastinggives rise to a series of surface features relatedto distinct re-sedimentation processes and asso-ciated deposits. To show the spatial relationshipof this series of surface features such as rota-tional sliding, sinkholes, distinct edges andextension fractures were mapped. A distinctionwas made between active or inactive features atthe time of recording. An area with surfaceactivity is often revealed by the offset or des-truction of the surface vegetation whereas aninactive area has a coherent cover of vegetation.

936 K. H. Kjñr and J. KruÈ ger

Ó 2001 International Association of Sedimentologists, Sedimentology, 48, 935±952

In addition, clusters of boulders and basindeposits were mapped.

The approach presented by KruÈger & Kjñr(1999) for detailed ®eld description of glacialdiamicts and associated sediments is adopted inthis study. Clast fabric data were measuredaccording to the criteria suggested by Kjñr &KruÈger (1998) and evaluated through a three-dimensional eigenvector analysis (Mark, 1973;Woodcook, 1977). The classi®cation of glacialdiamicts largely follows the recommendation of

tills as summarized by Dreimanis (1988) andco-workers. Accordingly, the position of trans-ported debris, the position of deposition inrelation to a glacier and the process of depositionare factors that must be considered, before anyconclusions are drawn on the genesis of diamicts.The position of deposition in relation to a glaciermight for instance be sub-glacial or supraglacial.If sub-glacial, then deposition might occur pas-sively, e.g. basal melt-out till or actively, e.g.lodgement till or deforming bed till. Diamict is

Fig. 1. Location map. (A) Thestudy area at the KoÈtlujoÈkull, anoutlet-glacier of the MyÂrdalsjoÈkullIce Cap in central south Iceland.(B) The marginal position ofKoÈtlujoÈkull in 1904, 1940, 1955and 1987 and associated dead-ice®elds. Based on aerial photo-graphs by Landmaelingar Islands.(C) Geomorphological map of thestudy area in 1996. Boxes indicatethe location of investigated areaswith, respectively, partiallyice-cored moraine (4a) and ice-free dead-ice moraine (4b). Theposition of the KoÈtlujoÈkull glaciermargin in 1955 and 1987 isshown.

Dead-ice moraine development 937


used as an overall non-genetic term for a non-sorted or poorly sorted, unconsolidated sedimentthat contains a wide range of particle sizes (Flint,1971; Frakes, 1978; Eyles et al., 1983). Conse-quently, diamicts might be matrix or clast sup-ported.

THE TERMINUS REGION OFKOÈ TLUJOÈ KULL

About 1á2 km beyond KoÈtlujoÈkull a system of ice-marginal ridges produced by a glacier advancearound 1900 separates the glacier fore®eld fromthe extensive MyÂrdalssandur outwash plain slo-ping gently towards the coast some 15±25 km tothe south and southeast. Behind the outermostmoraine ridges, the glacier fore®eld consists of a

complex pattern of ice-marginal ridges, slightly¯uted ground moraine, hummocky dead-icemoraine and patches of ice-cored morainesrepresenting at least three glacier events (KruÈger,1994). The moraine landscape is dissected byactive and abandoned meltwater channels and iscut by proximal extensions of the MyÂrdalssanduroutwash plain (Fig. 1C).

Previous studies carried out in the terminusregion of KoÈtlujoÈkull have identi®ed three prin-cipal phases (mature, old and fossil) of successivede-icing of ice-cored features and formation ofdead-ice moraine by examining time-dependentstages (KruÈger, 1994). This implies an investiga-tion of areas with ice-cored features and dead-icemoraine with different ages mirroring the ¯uctu-ations of KoÈtlujoÈkull. The terminal 400±500 m ofthe glaciated area consists of a fully ice-cored

Fig. 2. De-icing of ice-cored moraines in the terminus region of KoÈtlujoÈkull. Mode and rates of de-icing progressionin the fully ice-cored dead-ice ®eld (A) and in the partially ice-cored dead-ice ®eld (B). (C) Diagram showing thevariation in lowering of the terrain surface for eight observation points in the partially ice-cored moraine during the1995±99 period. Observation points 4 and 6 were buried in 1997 as meltwater invaded the low-lying part of the studyarea and deposited 2±3 m of sorted sediment, mostly gravel and sand.



moraine (mature phase) of irregular topographybehind a 4- to 7-m-high, well-de®ned frontalmoraine ridge produced in 1987 (Fig. 1B and C).Within this zone, continuous re-sedimentation iscommon, including fall, slump and sediment¯ows combined with lateral backwasting of ice-free faces, or steep ice-cored slopes. Further to theeast, beyond the 1987 end-moraine, a zone ofpartially ice-cored moraine representing the oldphase of de-icing occurs behind ice-marginalmoraine ridges produced by an ice advancearound the mid-1950s (Fig. 1B and C). Reworkingof the sediment cover in the partially ice-coredmoraine is governed by slumping associated withbackwasting of ice-cored slopes and downwa-sting. Outside the 1955 end-moraine, patches ofice-free areas with dead-ice moraine, which haveescaped meltwater erosion, are relicts from icestagnation around 1940 and represent the endproduct of dead- ice moraine development (fossilphase). Thus, advances and retreats of KoÈtlujoÈ-kull generated these successive phases of deicing,as evident from the present landscape, during thepast 60 years.

The terminus region of KoÈtlujoÈkull some 20±25 km inland is dominated by a highly oceanic

sub-polar climate with moderately cold winters,cool summers and high cloudiness. During the1995±98 period, the mean August precipitationwas about 650 mm. Estimated mean annualprecipitation (extrapolated from VõÂk data) isabout 5000 mm, which is a major contributoryfactor to the de-icing process (KruÈger & Kjñr,2000).

STUDY AREAS

Two study areas were selected: one in thepartially ice-cored moraine (old phase) and onein dead-ice moraine (fossil phase; Fig. 1B and C).Figure 4A shows the terrain of the partiallyice-cored moraine. To the east a marginalmoraine ridge, 10±20 m wide, with frontal slopedeposits represents a section of the 1955 end-moraine. Behind this zone a partially ice-coredmoraine is found with sorted basin sedimentsoccupying the central area. Isolated dead-iceblocks from the stagnation are estimated to beup to 5 m thick. Steeply rising hills 10±20 m indiameter characterize the southern part of thispartially ice-cored terrain (Fig. 4A). Calculations

Fig. 3. Diagram summarizing the major re-sedimentation processes and surface features at KoÈtlujoÈkull associatedwith the ®nal phase of de-icing of ice-cored moraine and their corresponding sedimentary characteristics in the post-melt landscape. Not to scale.



based on the three-dimensional terrain modelshow that 20% of the slopes are inclined at morethan 20° and a few more than 60°. The localrelief is 3±5 m, but with a prominent almost ice-free hill to the north rising more than 10 mabove the surrounding terrain (Fig. 4A). Aerialphotographs from 1945 show that a northwest-southeast directed end-moraine formed prior tothe 1955 end moraine crosscuts this part of thestudy area (Fig. 1B). This might account for theoutstanding hill, as stagnant ice from the 1950shas superimposed a remnant of this old end-moraine. Thus, the surface expression of thepreviously deposited, ice-free end-moraine in¯u-ences the present-day topography. The ¯at-lying

area in central part of the study area can beascribed to meltwater activity. Within a few daysof August 1997, after a period of heavy precipi-tation, meltwater invaded the low-lying centralpart and deposited 2±3 m of sorted sediment.The basin sediment consists mainly of medium-to coarse-grained, horizontally bedded sand andgravel.

Figure 4B shows the three-dimensional ter-rain model of the dead-ice moraine. Generally,the terrain slopes gently towards the southeastand has a relief of 1±2 m. A striking surfacefeature is a series of small northeast-southwesttrending sub-parallel ridges, up to 1 m high,often associated with groups of boulders on

Fig. 4. Three-dimensional terrain models illustrating the ®nal phase of dead-ice moraine development. (A) Part of apartially ice-cored moraine with isolated dead-ice blocks. The model is based on an ordinary kriging interpolationover a high density of points (N � 2350) to ensure an acceptable order of accuracy. Ringed letters indicate theposition of observation points used to monitor the gradual lowering of the terrain surface during the 1995±99 period.Lines indicate the location of measured pro®les (A, B, C in Fig. 5). (B) Dead-ice moraine. Sub-glacially and supra-glacially deposited sediments along a well-exposed geological section, 110 m long, are shown. The terrain model isbased on 983 points.



the crest. The inclination of slopes commonlyranges between 2° and 10° and rarely exceeds20°.

FINAL PHASE OF ICE-MELT

In the fully ice-cored moraine, the most importantde-icing process is backwasting of free ice wallsbecause ice is commonly exposed as a result ofcollapse, ¯uvial erosion, or mass-movement afterheavy precipitation (KruÈger & Kjñr, 2000). Also,downwasting by melting along the bottom surfaceof ice-cores contributes signi®cantly to de-icing asheavy precipitation increases the production ofsub-glacial water. In the partially ice-cored mor-aine, however, de-icing is dominated by down-wasting processes, because of a relatively thick(1±3 m), complete sediment cover that preventsexposure of ice. The contribution from backwa-sting of ice-cored slopes is considered to belimited, because of the thick sediment cover.

Eight observation points were chosen within thepartially ice-cored moraine in 1995 at locationswhere the terrain surface is locally horizontal(Fig. 4A). The level of each of these observationpoints was recorded in 1995, 1996, 1997, 1998and 1999. Figure 2C displays the magnitude ofthe surface lowering for the individual observa-tion points. It appears that the reduction of theice-surface by downwasting is most distinct inthe southwest part of the partially ice-coredmoraine area as indicated by points 1, 2, 3 and 5,in comparison with the outstanding hill towardsthe north and the area lying in vicinity of themarginal moraine as indicated by points 7 and 8(Fig. 2C and 4A). During the period 1995±99, theannual reduction in altitude due to downwastingamounted to 28 cm based on points 1, 2, 3, 5 and6 (Fig. 2C).

The data do not allow a differentiation of therelative contribution of top melt and bottom melt.However, from the fully ice-cored moraine, KruÈ -ger & Kjñr (2000) showed that below 1 m ofsediment the contribution from top melt wasapproximately 1 cm. A ®gure for geothermal heat¯ow over a year in the MyÂrdalsjoÈkull region is0á21 Wm±2 (Lee, 1970); this ¯ux will be enough tomelt c. 2 cm of the ice-core. When the bottommelt due to geothermal heat and top melt contri-butions are subtracted from the total annuallowering of the terrain surface, the annual bottommelt due to sub-glacial water drainage averaged25 cm. Thus, with a thickness of the supraglacialdeposits exceeding 1á5 m in the partially ice-

cored terrain, bottom melt is the major contribu-tor to de-icing of ice-cores.

Subsequent to the ¯ooding event in August1997, it was clear that the presence of melt wateraccelerated the decay of the remaining ice blocksand initiated a sequence of new processes. Thethree traverses shown in Fig. 5 were establishedacross the partially ice-cored terrain after the¯ooding event in order to survey the lowering ofthe terrain surface and its morphological effect.The terrain pro®les show that in some areas thesurface was lowered by a magnitude several timeslarger than was apparent from the individualobservation points. For instance, at 46 m inpro®le B-B¢ the surface was lowered more than1 m during 1 year whereas observation point 3some 5 m towards the north was lowered only30 cm (Fig. 2C and 5). Also, between 1997 and1998 several metres of supraglacial sedimentswere removed by lateral erosion in the centralpart of the study ®eld as seen in pro®le C-C¢. Inboth cases the changes detected in the pro®lesmight be ascribed to the in¯uence of ¯uvio-thermal erosion, i.e. warm ¯uvial water penetrat-ing the sediment cover, or lateral erosion of thesediment cover by ¯owing water (Fig. 5).

Figure 6A distinguishes between areas with orwithout surface activity. Clearly, high-lying areasin the southwest part of the study area have thehighest surface activity, indicative of areas whereburied ice still remains. Figure 6B shows thedistribution of the selected features and depositswithin the study area. In combination with theterrain model, it appears that zones of extensionfractures, distinct edges, and sinkholes and areasof rotational sliding, con®ne the perimeter of hills(Fig. 4A). Clusters of boulders are either locatedon crests of hills or in isolated depressions. Inareas of frontal moraine and slope deposits, noneof the surface features or structures were identi-®ed (Fig. 4A and 6).

The zonation of surface features and structuresre¯ect an interrelationship between ablation ofice-cores and re-sedimentation of sediment coverin the partially dead-ice moraine. Isolated depres-sions are created as local melting occurs at the toeof the buried dead-ice blocks. These sinkholes inturn initiate collapse of sediment in the vicinitydue to undermining which results in the forma-tion of distinct edges. In the adjacent basins,extension fractures occur as collapse takes placealong discrete planes. The sinkholes also causeover-steepening of the adjacent slopes and intro-duce retrogressive rotational sliding or backslum-ping, seen as discrete concave, or linear cracks,



arranged stepwise and often producing an arcuateniche (Fig. 7). Evidently, the cornerstone of theseinterrelated re-sedimentation processes is thedevelopment of sinkholes in locations where iceblocks pinch out. Clusters of boulders may resultfrom gravitational sorting along slopes or mayre¯ect remnants of former slope processes wherematerial was accumulated in well-de®ned areas.

Figure 6C shows the areal distribution ofre-sedimentation features and their correspondingactivity or inactivity at the time of recording.Taken together, backslumping constitutes thesingle most important process with the largestareal coverage, followed by sinkholes, extensionfractures and fall sorting. Backslumping covers53% of the area where the selected surfacefeatures are recognized and in most cases thisprocess seems to be active at the time of recording(Fig. 6C). Assuming the recorded activity is rep-resentative (summer season and excluding thepossibility for reactivation of processes) then theactivity of individual processes is perceived as ameasure of signi®cance that the process has on thesedimentary end product. Therefore, it is predic-ted that processes involving well-de®ned collapse

have a major impact on the sediment architectureobserved in the post-melt dead-ice moraine.

PRODUCTS OF FINAL ICE-MELT

In the post-melt landscape beyond the activedead-ice ®elds, well-exposed sections displaysediments associated with an advance of KoÈtlu-joÈkull, subsequent stagnation and de-icing of ice-cored moraines. A sub-horizontal clast pavementseparates the sediment deposited during advancefrom those related to stagnation and de-icing(Fig. 4B and 8). The thickness of sediments abovethe pavement varies between 1 and 4 m. Conse-quently, the surface relief was formed by thesediments accumulated during the time of stag-nation and de-icing (Figs 4B, 8 and 9). This is alsore¯ected in the spatially variable distribution ofthe different sedimentary units. Nonetheless, it ispossible to establish a composite stratigraphy forwell-de®ned areas, representing the order ofevents, which does not necessarily comply withthe law of superposition due to the stagnant ice.This is because sediments might be deposited in

Fig. 5. Terrain pro®les A, B and C across over the partially ice-cored moraine in 1997 and repeated in 1998. Majorre-sedimentation and surface features are marked along the pro®les.



cavities beneath stagnant glacier ice after sedi-ments were deposited on top of ice-cores.

Units 1±3

Description

The lowermost part of the succession consists oftwo diamict units (1 and 3) separated by a thinunit (2) of ®nes and gravel (Fig. 10). Unit 1 belowthe pavement is a grey, compact, massive diamictwith a uniform thickness and a moderate contentof clasts. Most clasts have a high grade ofroundness with a high abundance of striationson their surfaces. Clast fabric, striation on clastsurfaces and stoss-lee morphologies show a spa-tially consistent orientation being parallel withthe local ice-¯ow direction of KoÈtlujoÈkull. Unit 2is characterized by laminated silt and ®ne sandthat drapes the clast pavement at the base. Unit 3is a crudely strati®ed, matrix-supported, friable

diamict with a high proportion of angular clastsand a low proportion of clasts with striationsrelative to the underlying unit 1. Laterally, thethickness of unit 3 is variable; in some parts of thesection it reaches up to 30 cm, in others it isabsent (Fig. 9). Along the pro®les, the sorted ®nesdrape clasts within unit 3 (Fig. 10). Two clastfabrics in unit 3 display a spatially consistentorientation parallel with local ice-¯ow direction.

Interpretation

The absence of collapse structures suggests thatunits 1±3 were accumulated prior to stagnation ofthe glacier or at the bottom of stagnant ice. Thelower diamict unit (unit 1) is interpreted as aslightly deformed lodgement till deposited inresponse to an advance of the KoÈtlujoÈkull glacier.The upper diamict unit (unit 3) is interpreted asdeposited in a sub-glacial environment by in situmelt-out of debris from stagnant ice. The sorted

Fig. 6. Final phase of dead-ice melting. (A) Activity of processes at the time of recording in August 1997. (B) Arealdistribution of re-sedimentation and other surface features in a partially ice-cored moraine. (C) Areal coverage ofre-sedimentation and surface features and their corresponding activity as percentage of area excluding the zonewhere no features and structures were identi®ed.



sediments separating the two diamict beds areinterpreted as water-laid cavity ®ll as a result ofmeltwater activity beneath the stagnant ice(KruÈger & Kjñr, 1999).

Units 4±6

Description

Above unit 3, a succession of both sorted anddiamict sediments constitutes a supraglacial sedi-ment association (units 4±7). The lowermost twounits (units 4 and 6) of sorted sediments, respect-ively, 0á9 and 0á2 m thick, are separated by a0á3-m-thick diamict (unit 5). The sorted sedimentsconsist chie¯y of medium-grained, horizontallybedded sand and laminated mud occasionallywith horizons of gravel. The diamict is a massive,friable, matrix-supported sandy-gravelly unit witha high proportion of clasts. A clast fabric showsmoderately developed elongation with the orien-tation of the principal eigenvector (V1) at NE (53°).The diamict unit appears to lens out towards theeast with a sharp conformable basal contact.

Interpretation

The complex bedding of sorted sediments (units 4and 6) and the diamict (unit 5) is interpreted astrough ®llings caused by meltwater activity along

depressions in the fully ice-cored terrain com-bined with the entry of diamict sediment.

Units 7±8

Description

The sorted sediment (unit 6) grades upwards in toa loose, heterogeneous 1á5- to 3á0-m-thick diamict(unit 7) displaying extreme lateral and verticalvariation in the grain-size distribution and con-tent of clasts. Within a few centimetres, thediamict changes from a clast-supported diamictof sandy gravel to a matrix-supported diamict ofgravelly sand containing abundant sand lensesand numerous groups of boulder-rich material.Throughout the diamict, thin horizons of silt orsand occur, together with lenses of predomin-antly sand. Large-scale normal faults were notedlocally in unit 7 (Fig. 8C). In contrast, slumpstructures are very common at every scale. Slumpstructures or patterns of slump structures arerelated to wedge-shaped accumulations of boul-ders, for instance, at 27 and 86 m along the tran-sect (Fig. 8B). A striking feature discovered in thepro®les is the concentration of boulders in lineararrangements, generally with a 30±35° slope.Commonly, the boulder accumulations are cou-pled to the ridges in the surface morphology with

Fig. 7. View across partially ice-cored moraine showing typical surface features related to the de-icing progression ofisolated dead-ice blocks. Sinkholes develop at the base of slopes and backslumping upslope. Boulders are mainlylocated in depressions or the crest of hills as a result of fall sorting processes in the preceding phases of de-icing.



an opposite direction of tilt on either side of thecrest. Less than half of the clasts have striationson their surfaces, the orientation of striationsbeing highly dispersed between individualclasts. Several clast fabrics sampled upwards inthe diamict show weak to moderate eigenvalues(S1 � 0á48±0á66; Fig. 10). Although most clastfabrics show a preferred orientation, the direc-tion is not consistent in between fabrics.Furthermore, 10 clast fabrics scattered over thehummocky moraine terrain were taken 10±30 cmbelow the surface in heterogeneous diamict (unit7). Generally, the orientation of the 10 fabricsshows weak to moderate (S1 � 0á45±0á67)

clustering around the principal eigenvector andwith a high number of steeply plunging clasts(Fig. 11). Two fabrics indicate that no statisticalpreferred orientation occurs as con®rmed bothby the eigenvalues and contoured diagrams(Kamb, 1959; Anderson & Stephensen, 1971).However, the relative high S2 eigenvalues(S1 � S2 � S3) suggest that many clasts areorientated in the same plane. Bimodal distribu-tion has developed in two fabrics as seen in thecontoured diagrams. Unit 8 is mostly massivegravel interbedded with laminated clay/silt oftenfound in topographical lows with a thickness upto 60 cm.

Fig. 8. Dead-ice moraine. (A) View across the dead-ice moraine with well-exposed section illustrating the distri-bution between sub-glacial and supraglacial deposits. The supraglacial deposits form the relief amplitude of theterrain surface. In the background the fully ice-cored dead-ice ®eld is seen. (B) Slumping of both sorted and diamictassociated with collapse feature. (C) View of normal fault with prominent offset related to rotational sliding (seearrow).



Interpretation

The syn-depositional melting of underlying ice isillustrated by the occurrence of normal faults andslump structures in unit 7, which in turn provethe supraglacial origin for this complex sedimentsuccession. Unit 7 is a product of mass-movementprocesses including interbedding with sortedsediment deposited by meltwater. Backslumpingby rotational sliding overprinted by collapsestructures have resulted in a somewhat chaoticappearance. The lack of large-scale normal faultsis most probably a result of the retrogressivenature of the interrelationship between processesas ®nal melt proceeds and sinkholes and exten-sion features overprint the rotational structures ofsliding. Wedge-shaped accumulations of bouldersare likely to be an impression left by sinkholes,which is probably the only structure of the majorde-icing processes that survive more or lessunaltered (Fig. 3).

In those parts where the ice-degradation is at amore advanced stage, layers of boulders cover thelower part of hill slopes as a result of fall andsliding processes in the fully ice-cored terrain.

As reworking continues, sediments from theadjacent higher area may cover the boulder layersas slope deposits. Once the boulders have accu-mulated they are in general not separatedalthough reworking continues. As a result, thelinear arrangement of boulders seen in thepro®les of the post-melt landscape probablyrelates to boulders covering the surface of formerslopes. The bimodal distribution in some of theclast fabrics might re¯ect separate events ofmass-movement and fracture formation workingperpendicular to each other (KruÈger, 1994). Manyclasts could also be brought into a more verticalposition along with collapse of the sediment.Although individual fabrics appear to show apreferred orientation, the azimuths of individualpreferred orientations are widely dispersed andnot related to ice-¯ow directions as evident fromKoÈtlujoÈkull. Instead, clasts are oriented parallelwith the terrain slope indicating a positivecorrespondence between clast orientation andthe surface topography (Fig. 11). Because meas-urements were performed in the uppermost partof the sediment, post-depositional processessuch as soil creep might naturally be argued for

Fig. 9. Section through the dead-ice moraine at 27±39 and 85±90 m distance along transect. Basal melt-out till, cavity®ll and supraglacial deposits overlie subglacial-deposited lodgement till separated by a clast pavement. Thesupraglacial deposits consist chie¯y of collapsed water-laid sediment and mass-movement deposits, mainly diamict.Note the discontinuous distribution of the basal melt-out till.



as a cause for this trend in orientation. However,the trend in orientation of fabrics is associatedwith ridges that might be linked to largerstructures in the section such as straight align-ments of boulders and collapse features. Thus,the orientation of clasts at the surface re¯ects theorientation of mass-movement processes actingalong discrete planes (Lawson, 1979; Alm &Kleman, 1982). The gravel and laminated siltand sand that caps the supraglacial sediments(unit 8) is interpreted as trough-®lling, washeddown from the hummock side slope after themelting of underlying ice.

DISCUSSION

A scenario for the ®nal de-icing of ice-coredmoraine and the formation of hummocky moraineis pictured in Fig. 12. This sedimentologicalmodel begins at a late phase within the fully ice-cored dead-ice ®eld immediately before theice disintegrates into isolated dead-ice blocks.Apparently, no topographical inversion takesplace during the last phases of ice melting. Theoverall topography seen in the late phase ofthe fully ice-cored terrain is unchanged in thepartially ice-cored landscape even though the

Fig. 10. Data chart comprising a detailed description of the sedimentary succession recognized in the dead-icemoraine landscape. Partly after data chart developed by KruÈger & Kjñr (1999). Lithofacies code for sorted sedimentsafter Miall (1977) and Eyles et al. (1983).



Fig. 11. Terrain surface of a land-scape with hummocky moraine.The surface represents an ice-freeremnant of dead-ice moraine thatsurvived ¯uvial erosion in front ofKoÈtlujoÈkull. The area matches thethree-dimensional model inFig. 4B. The diagram shows thecorrespondence between clastorientation in subsurface diamictsediments and the topography ofhummocky dead-ice moraine.Clast fabrics are presented on thelower hemisphere of a Lambertprojection. Contoured diagramsaccording to Kamb (1959) withindication of E+3s (shaded areas).V1 is the principal eigenvector andS1 the corresponding normalizedeigenvalue, S3 eigenvalue is cor-responding to a V3 eigenvector.

Fig. 12. Sedimentological model showing different phases of de-icing of ice-cored features and the formation ofdead-ice moraine. (A) Lowermost zone of the fully ice-cored moraine. (B) Partially ice-cored moraine. (C) Dead-icemoraine. Letters in italics mark ®nal re-sedimentation processes and surface features.



amplitude of relief is gradually reduced (Fig. 12Aand B). Towards the ice-free terrain the surfacebecomes a reduced re¯ection of the relief seen inthe preceding phases, characterized by small sub-parallel ridges, 1±2 m high, with boulders cover-ing their crests (Fig. 12B and C). This, however, isnot equivalent with the development of uncon-trolled/controlled topography according to Grav-enor & Kupsch (1959) as the scale and signi®cancein terms of debris band distribution is different.

Sedimentological model

Fully ice-cored moraine

In those parts within the fully ice-cored terrainwhere ice-degradation is at a more advanced stage,the ice thickness is estimated to be approximately10 m and the sediment cover between 0á5 and2á0 m (KruÈger & Kjñr, 2000). Most re-sedimenta-tion is related to collapse of sediment such asbackslumping of ice-cored slopes represented byretrogressive rotational sliding. However, few andminor free ice-walls still exist in the most elevatedpositions commonly associated with steep-sidedniches (Fig. 12A). Collapse of the sediment coverresults in fall or sliding of sediment from the top ofice-cores and subsequent re-mobilization by sedi-ment gravity ¯ows. Fall sorting of the dumpedsediment produces distinct boulder patterns withthe direction of coarsening downslope. Often,minor meltwater streams erode into the base ofslopes and remove ®ne components. Meltwater isalso located in small lakes on the ice surface andin tunnels or cavities beneath the ice and inboth cases melting of ice is likely to accelerate(Pickard, 1983).

Partially ice-cored moraine

Within the partially ice-cored moraine the indi-vidual ice-cores are estimated to be up to 5 mthick with a measured sediment cover of 1±3 m(Fig. 12B). Reworking of the sediment cover isinitiated by sinkhole formation at the base ofslopes as a result of melting of the toe of ice blocks.The formation of sinkholes leads to collapse andslumping of adjacent sediment. The dominatingsurface process is backslumping, which dissectsmost of the sedimentary structures produced inthe fully ice-cored terrain (Fig. 6). Boulders arelocated on crests of hills or concentrated at thebottom of sinkholes. Some boulder patterns arefound at the base of slopes as a result of fall sortingprocesses that have survived the transition fromthe fully ice-cored moraine. Directly on top of ice-

cores, melt-out till deposits reach a thickness of20±30 cm, but are laterally discontinuous andhave a low preservation potential due to constantre-sedimentation. The surface is modi®ed bymeltwater that erodes into the sediment cover or®lls topographical lows with sorted horizontalbedded gravel and sand that is occasionallytruncated by troughs with cross-bedding.

Dead-ice moraine

Sedimentary structures and surface features pro-duced continuously during de-icing of ice-coredmoraines are mostly destroyed at the transition tothe ice-free phase as a result of the re-sedimen-tation processes working in the partially ice-coredmoraine. This scenario is included in Fig. 12Band C, which illustrate the retrogressive nature ofre-sedimentation processes. Thus, as ice-coresgradually diminish, the most recent re-sedimen-tation will overprint the results of previous eventsand will consequently destroy most sedimentarystructures. Thus, the sediment architecture of theice-free terrain results from sinkhole formation,backslumping and fracture formation. Some sedi-mentary characteristics recognized within thesupraglacial sediment succession might be tracedback to processes working in the partially andeven in the fully ice-cored terrain. Most abundantare the isolated groups of boulders, which repre-sent accumulation of boulders in former topo-graphical lows (Fig. 3). Also commonly identi®edare concentrations of boulders in straight orconcave alignments coupled with ridges andlinked to fall sorting processes along slopes,which might have been generated in the fullyice-cored moraine. Other common characteristicsare sorted sediment, either collapsed or horizon-tally bedded, related to water-laid sedimentationin cavities at the bottom or in local depressions atthe top of ice-cores. The effect of sinkholes,distinct edges and extension fractures cannot bedistinguished separately. Collectively, however,they are represented by wedge-shaped accumula-tions of boulders and/or patterns of slump struc-tures (Fig. 3). Distinct slump structures might beseen at every scale from few centimetres toseveral metres, occasionally reverse or normalfaulted (Fig. 8b). In contrast, normal fault struc-tures linked to backslumping are rare and nostructures from sediment gravity ¯ows havebeen identi®ed. The signi®cance of the lastprocesses on the sediment architecture has beenacknowledged previously, although, withoutidenti®cation of which ®nal processes remain



imprinted on the sediment architecture (Lawson,1979; KruÈger, 1994). This study clearly identi®esthe signi®cance of sinkholes that initiate back-slumping and formations of extension fracturesand at the same time leave a sedimentary ®nger-print that might be recognized in the post-meltlandscape.

TOPOGRAPHICAL DEVELOPMENT

Current literature often links the melting ofburied ice and lowering of the terrain surface torepeated inversions of the topography (Clayton,1964; Boulton, 1967, 1972; Clayton & Moran,1974; Watson, 1980; Paul, 1983; Benn & Evans,1998). Initially, the distribution of englacialdebris produces an uneven cover of sediment onthe stagnant ice. This allows differential ablation,which produces an irregular topography becausemelting rates diminish where the sediment coveris thick and the insulation high. When the reliefincreases, mass-movement processes will redis-tribute the sediment cover from the ridges intotopographical lows, which eventually becomenew topographical highs. The topographicalinversions are repeated until all ice is wasted.At the margin of KoÈtlujoÈkull reworking of thesediment cover has also been linked to inversionof the topography (KruÈger, 1994). However, asillustrated by the sedimentological model noinversion of the topography takes place in the®nal phase of de-icing (Fig. 12). Only secondarydepressions are found on highs developed in thefully ice-cored terrain. Thus, the model of repea-ted topographical inversion does not adequatelydescribe the ®nal phase of the dead-ice morainedevelopment at KoÈtlujoÈkull. A stepwise loweringof the ice-cored moraine where sediments aremoved laterally is probably a better visualizationof the dead-ice moraine development. Repeatedinversions of the topography only occur in theyoungest phases within the fully ice-coredmoraine or on a secondary scale.

The question arises as to why topographicalinversions are limited at KoÈtlujoÈkull. In the litera-ture the most common mass-movement processesare those of sediment gravity ¯ows, often closelyassociated with meltwater activity (Boulton, 1968,1971, 1972; Eyles, 1979; Lawson, 1979, 1988;Fitzsimons, 1990; Bennett et al., 1996). Sedimentgravity ¯ows are clearly important in the redistri-bution of the sediment cover associated withmodels describing topographical inversion (Clay-ton, 1964; Boulton, 1967, 1972; Clayton & Moran,

1974; Paul, 1983). Therefore, the scarcity of sedi-ment ¯ows at KoÈtlujoÈkull might explain the lack oftopographical inversions (Lawson, 1979, 1981).KruÈger (1994), however, inferred that the com-bined effects of mass-movement processes onaverage rework the sediment cover up to two timesduring an ablation period. This is compatible withthe result from at the Matanuska glacier in Alaska,where ®ner-grained sediments over a wide areawere reworked two to three times under lesshumid conditions (Lawson, 1979). Thus, althoughthe climatic conditions and sediment characteris-tics are different the difference in topographicalinversions is not linked to reworking of thesediment cover, as the re-sedimentation rates arethe same (Lawson, 1979, 1981). The lack of obvioustopographical inversion is more probably a conse-quence of the rates of backwasting relative todownwasting. When ice-cored areas are attackedby downwasting and backwasting along ice-coredslopes or free ice walls, the sediment cover willslide or ¯ow downslope into the adjacent depres-sions. Ideally these depressions are the potentialnew topographical highs. Generally, however,these areas will not become topographical highs,because backwasting of ice-cored slopes across theice-cored terrain consumes the high ground caus-ing ¯attening of the terrain. Sediments are consid-ered to move from one level to the next asbackwasting repeatedly overtakes downwasting.

Implications

A key objective for most sediment-process-land-form studies in modern glacial environments is toprovide analogue models for formerly glaciatedareas. More to the point, the study of dead-icemoraines and their development under well-known climatic conditions, e.g. at KoÈtlujoÈkull, isuseful in the reconstruction of climatic condi-tions in formerly glaciated areas at the time ofdeposition. Clearly, this study represents dead-ice moraine development under extreme climaticconditions. The huge annual precipitation atKoÈtlujoÈkull accelerates the de-icing progressionand favours backwasting over downwasting as thedominant ablation process. This leads to a dis-tinct set of re-sedimentation processes and sur-face features that rework the sediment cover onice-cores. For instance, backwasting of ice-coredslopes increases the abundance of backslumpingand hampers the development of true sedimentgravity ¯ows. Under different climatic conditionswhere downwasting is the dominating ablationprocess, a different set of processes might govern



re-sedimentation of the sediment cover. Also, ifdownwasting is the primary ablation process,then the grain-size distribution of the sedimen-tary end product is probably ®ner grained than ifbackwasting is the predominant ablation process.This is because downwasting ensures that sedi-ments pass through less re-sedimentation cycles.In the future, however, more descriptions fromother contemporary dead-ice ®elds (differentclimate and source material) are needed, if theclimatic signal of dead-ice moraines from for-merly glaciated areas is to be deciphered.

CONCLUSION

Monitoring the ®nal melting of partially ice-coredmoraine and detailed descriptions of sedimentsassociated with dead-ice moraine at the terminusregion of KoÈtlujoÈkull, permits the followingconclusions.

In the current humid subpolar climate, thepartially ice-cored moraine is predominantlyde-iced by downwasting along the bottom surfaceof ice-coreswith average ratesof25 cm a±1.The®nalphase of ice-melt is associated with an interrelatedseries of processes initiated by sinkhole formation.Sinkholes develop at base of slopes in response tolocal melting of the toe of buried ice blocks, causingcollapse of sediment. These sinkholes also initiatethe formation of distinct edges and fractures in theadjacent basins and retrogressive rotational slidingor backslumping on the ice-cored slope.

Generally, topographical inversion is not asso-ciated with ®nal phase of ice-melt. The overalltopography recognized in the fully ice-cored ter-rain is gradually lowered and the amplitude of therelief reduced in consecutive phases. The lack oftopographical inversion is related to the rate ofbackwasting relative to downwasting, whichresults in stepwise lowering of sediment cover asbackwasting repeatedly overtakes downwasting.

As ice-cores gradually diminish, the latest inter-related group of sinkholes, extension fractures andbackslumping processes will overprint and des-troy most previous sedimentary structures. Thus,only the product of local collapse and mass-movement along discrete planes remain imprintedon the resultant sedimentary end product.

ACKNOWLEDGEMENTS

The authors acknowledge the Danish NaturalScience Research Council for ®nancial support

of the ongoing MyÂrdalsjoÈkull project. ZenicaG. Larsen is recognized for invaluable assistancein the ®eld. To Jaap van der Meer and JeroenWijnen, University of Amsterdam, Sùren Hilde-brandt, Michael Houmark-Nielsen, Thaimi Olsenand Niels Richardt, University of Copenhagen,and Knud Erik Klint, Geological Survey of Den-mark and Greenland, we extend our sinceregratitude for valuable co-operation and discus-sions. Finally, we thank the Icelandic NationalResearch Council for giving us the opportunity towork at the MyÂrdalsjoÈkull ice-cap in Iceland. Thearticle was signi®cantly improved by carefulreviewing of Neil Glasser, University of Wales,an anonymous reviewer and Chris Fielding,University of Queensland.

REFERENCES

Alm, G. and Kleman, J. (1982) Egenvektormetoden som

hjaÈlpmedel vid analys av stenorienteringar ± en geometrisk

tolkning. Naturgeogra®sk Institutionen, Stockholms Uni-

versitet. Forskningsrapport, 37, 5±31.Anderson, T.W. and Stephensen, M.A. (1971) Tests for Ran-

domness of Directions Against Equatorial and BimodalAlternatives. Technical Rept. 5, Stanford University.

Benn, D.I. and Evans, D.J.A. (1998) Glacier and Glaciation.Arnold, London.

Bennett, M.R., Hambrey, M.L., Hubbart, D. and Ghienne, J.F.(1996) The formation of a geometrical ridge network by thesurge-type glacier Kongvegen, Svalbard. J. Quatern. Sci., 11,437±449.

Boulton, G.S. (1967) The development of a complex supra-

glacial moraine at the margin of Sùrbreen, NY Friesland,Vestspitsbergen. J. Glaciol., 6, 717±735.

Boulton, G.S. (1968) Flow tills and related deposits on some

Vestspitsbergen glaciers. J. Glaciol., 7, 391±414.

Boulton, G.S. (1971) Till genesis and fabric in Svalbard,Spitsbergen. In: Till, a Symposium (Ed. R.P. Goldthwait),

pp. 41±72. Ohio State University Press, Columbus.

Boulton, G.S. (1972) Modern Arctic glaciers as depositionalmodels for former ice sheets. J. Geol. Soc. London 128, 361±

393.

Brodzikowski, K. and van Loon, A.J. (1991) Glacigenic Sedi-ments. Developments in Sedimentology 49. Elsevier,Amsterdam.

Clayton, L. (1964) Karst topography on stagnant glaciers.

J. Glaciol., 5, 107±112.

Clayton, L. and Moran, S.R. (1974) A glacial process-formmodel. In: Glacial Geomorphology (Ed. D.C. Coates),

pp. 89±119. State University of New York Publications

in Geomorphology, Binghamton.

Dreimanis, A. (1988) Tills: Their genetic terminology andclassi®cation. In: Genetic Classi®cation of Glacigenic Depo-sits (Eds R.P. Goldthwait and C.L. Matsch), pp. 147±169.

A.A. Balkema, Rotterdam.Drozdowski, E. (1977) Ablation till and related indicatory

forms at the margin of Vestspitsbergen glaciers. Boreas, 6,107±114.

Eyles, N., Eyles, C. and Miall, A.D. (1983) Lithofacies typesand vertical pro®le models: an alternative approach to



description and environmental interpretation of glacial

diamict and diamictite sequences. Sedimentology, 30,395±410.

Fitzsimons, S.J. (1990) Ice-marginal depositional processes ina polar maritime environment, Vestfold Hills, Antartica.

J. Glaciol., 36, 279±286.

Flint, R.F. (1971) Glacial and Quaternary Geology. Wiley,

New York.Frakes, L.A. (1978) Diamictite. In: The Encyclopedia of

Sedimentology (Eds R.W. Fairbridge and J. Bourgeois),

pp. 262±263. Dowden, Hutchinson & Ross, Stroudsburg.Gravenor, C.P. and Kupsch, W.O. (1959) Ice-disintegration

features in Western Canada. J. Geol., 67, 48±64.

Kamb, W.B. (1959) Ice petrofabric observations from Blue

Glacier, Washington, in relation to theory and experiment.J. Geophy. Res., 64, 1891±1909.

Kjñr, K.H. and KruÈger, J. (1998) Does clast size in¯uence

fabric strength? J. Sed. Res., 8, 746±749.

KruÈger, J. (1994) Glacial processes, sediments, landforms, andstratigraphy in the terminus region MyÂrdalsjoÈkull, Iceland.

Folia Geogr. Dan., 21, 1±233.

KruÈger, J. and Kjñr, K.H. (1999) A data chart for ®eld des-cription and genetic interpretation of glacial diamicts and

associated sediments ± with examples from Greenland,

Iceland, and Denmark. Boreas, 28, 386±402.

KruÈger, J. and Kjñr, K.H. (2000) De-icing progression of ice-cored moraine in a humid, subpolar climate, KoÈtlujoÈkull,

Iceland. Holocene, 10, 721±731.

Lawson, D.E. (1979) Sedimentological analysis of the western

terminus of the Matanuska Glacier, Alaska. Army ColdRegions Res. Engineering Laboratory, Report, 79±9, 1±112.

Lawson, D.E. (1981) Distinguishing characteristics of diamic-

tons at the margin of the Matanuska Glacier, Alaska. Ann.Glaciol., 2, 78±84.

Lawson, D.E. (1988) Glacigenic resedimentation: Classi®cation

concepts and application to mass-movement processes and

deposits. In: Genetic Classi®cation of Glacigenic Deposits(Eds R.P. Goldthwait and C.L. Matsch), pp. 147±169.A.A. Balkema, Rotterdam.

Lee, W.H.K. (1970) On the global variations of terrestrial heat-

¯ow. Phys. Earth Planet. In., 2, 332±341.

Mark, D.M. (1973) Analysis of axial orientation data, inclu-ding till fabric. Bull. Geol. Soc. Am., 84, 1369±1374.

Miall, A.D. (1977) A review of the braided river depositional

environment. Earth-Sci. Rev., 13, 1±62.Paul, M.A. (1983) The supraglacial landsystem. In: Glacial

Geology (Ed. N. Eyles), pp. 71±90. Pergamon Press,

New York.

Pickard, J. (1983) Surface lowering of the ice-cored moraine bywandering lakes. J. Glaciol., 29, 338±342.

Sharp, R.S. (1949) Studies of supraglacial debris on valley

glaciers. Am. J. Sci., 247, 289±315.

Shaw, J. (1979) Tills deposited in arid polar environments.Can. J. Earth Sci., 14, 1239±1245.

Shaw, J. and Rains, R.B. (1981) Some mechanisms of

controlled moraine development. Antart. J. Glaciol., 27,113±128.

Watson, R.A. (1980) Landform development on moraines of

the Klutan Glacier, Yukon Territory, Canada. Quatern. Res.,14, 50±59.

Woodcook, N.H. (1977) Speci®cation of fabric shapes using an

eigenvalue method. Bull. Geol. Soc. Am., 80, 2342±2352.

Manuscript received 6 March 2000;revision accepted 2 March 2001.



the ﬁnal phase of dead-ice moraine development: processes

Documents