HINLOPEN–YERMAK LANDSLIDE, ARCTIC …folk.uio.no/anelverh/Papers/Vaneste_etal_SP95_inpress.pdf ·...

1HINLOPEN–YERMAK LANDSLIDE, ARCTIC OCEAN

HINLOPEN–YERMAK LANDSLIDE, ARCTIC OCEAN—GEOMORPHOLOGY,LANDSLIDE DYNAMICS, AND TSUNAMI SIMULATIONS

MAARTEN VANNESTENGI – International Centre for Geohazards (ICG), P.O. Box 3930, Ullevål Stadion, N-0806 Oslo, Norway

e-mail: [email protected] B. HARBITZ

NGI – International Centre for Geohazards (ICG), P.O. Box 3930, Ullevål Stadion, N-0806 Oslo, NorwayFABIO V. DE BLASIO

Department of Geosciences, University of Oslo – International Centre for Geohazards (ICG),P.O. Box 1053 Blindern, N-0316 Oslo, Norway

SYLFEST GLIMSDALNGI – International Centre for Geohazards (ICG), P.O. Box 3930, Ullevål Stadion, N-0806 Oslo, Norway

JÜRGEN MIENERTDepartment of Geology, University of Tromsø, Dramsveien 201, N-9037 Tromsø, Norway

AND

ANDERS ELVERHØIDepartment of Geosciences, University of Oslo – International Centre for Geohazards (ICG),

P.O. Box 1053 Blindern, N-0316 Oslo, Norway

ABSTRACT: Swath bathymetry data from the glacier-fed northern Svalbard margin reveal geomorphological details of a large submarinelandslide, the Hinlope–Yermak Landslide. Multiple planar escarpments have several hundreds of meters of relief, with a maximumheadwall height exceeding 1400 m at the mouth of the Hinlopen cross-shelf Trough. Within the slide-scar area, this landslide created arugose seabed geomorphology, with little mass-transport deposition in the immediate vicinity of major escarpments. Beyond a pronouncedconstriction, occurrence of semitransparent acoustic units on seismic profiles indicates that mass-transport deposits are likely theaccumulation of remolded and/or fluidized debris flows that are in places hundreds of meters thick. The surface expression of the mass-transport deposits is hummocky with flow structures, arcuate pressure ridges, and rafted blocks. Smaller debris lobes close to landslidesidewalls are the result of secondary, marginal failures. At the outer rim of extensive mass-transport deposits, numerous rafted blocks risefrom the semitransparent sediment unit, and tower hundreds of meters above the surrounding debris. Maximum remobilized volume fromthe slide-scar area, estimated from pre-landslide bathymetric reconstruction, is approximately 1350 km3. Headwall heights, the ratio ofexcavated volume and slide scar area, and the height of rafted blocks are large, compared to other landslides documented on siliciclasticmargins.

The position, thickness, and shape of the mass-transport deposits illustrate high mobility of sediments involved in submarinelandsliding. Their dimensions require numerical modeling to understand landslide dynamics and potential to generate tsunamis. Insimulations of sediment dynamics, large blocks are rafted by a loose debris flow, derived from disintegrating landslide material in theheadwall area. The main failure process finishes after approximately 1 hour. The upper slide scar is probably not the source area for largerafted blocks.

The Hinlopen–Yermak landslide most likely created a significant tsunami, considering that remobilized sediment volume, initialacceleration, maximum velocity, and possible retrogressive development govern landslide-generated tsunamis. Steep waves, implyingdispersive and nonlinear effects, probably were more pronounced than for most other tsunamis induced by submarine landslides. Thesefeatures exist because of the combination of high speed and substantial thickness of mass transport. Propagation and coastal impact of thetsunami is simulated by a weakly nonlinear and dispersive Boussinesq model. Close to the landslide area, simulations return sea-surfaceelevations exceeding 130 m, whereas sea-surface elevations along coasts of Svalbard and Greenland are on the order of tens of meters.

KEYWORDS: Hinlopen–Yermak landslide, geomorphology, landslide dynamics, tsunami modeling, mass-transport deposits, formerlyglaciated continental margin, Svalbard

Mass-Transport Deposits in Deepwater SettingsSEPM Special Publication No. 95, Copyright © 2010SEPM (Society for Sedimentary Geology), ISBN 978-1-56576-287-9, p. xxx-xxx.

INTRODUCTION

The advent of swath-bathymetry and side-scan sonar map-ping techniques in combination with high-resolution seismicsurveying have greatly improved existing knowledge of theinterplay between sedimentary and erosional processes in shap-ing continental margins. These techniques reveal an abundanceof submarine landslides features, from subtle seabed cracks to

escarpments several hundreds of meters high, and a diversity ofmass-transport deposits, such as debris-flow deposits, raftedblocks, and slumps (Embley and Jacobi, 1977; Hampton et al.,1996; McAdoo et al., 2000; Krastel et al., 2001; Locat and Mienert,2003; Mienert and Weaver, 2003; Canals et al., 2004). Mass-wasting processes affect a significant proportion of the continen-tal margins globally and are efficient in redistributing largeamounts of sediments (Prior and Coleman, 1978; Moore et al.,

M. VANNESTE, C.B. HARBITZ, F.V. DE BLASIO, S. GLIMSDAL, J. MIENERT, AND ANDERS ELVERHØI2

1989; Hampton et al., 1996; Locat and Lee, 2002; Canals et al.,2004). The largest landslides occur mainly in two settings: (1) on(typically gentle) glacially influenced open continental margins;and (2) on (typically steep) oceanic island flanks (Masson et al.,2006). Often, continental margins show a long history of recurrentlandsliding, for example the Norwegian margin (Solheim et al.,2005a), the Canary Islands (Masson et al., 2002), and off Brunei onNorthwest Borneo (Gee et al., 2007). The post-failure stage char-acteristically involves transformation to a remolded and lique-fied state of at least part of the landslide mass prior to mass-transport deposition. Slabs that detach and subsequently disinte-grate into debris avalanches or debris flows may generate turbid-ites and can attain extreme run-out distances (Heezen and Ewing,1952; Mulder and Cochonat, 1996; Piper et al., 1999; Canals et al.,2004; Bryn et al., 2005a). Need for integrated, multidisciplinarystudies gained additional momentum, now that hydrocarbonexploitation is moving into deeper waters. Additionally, mass-transport processes and their deposits may present a drillinghazard, impede field development, or, where infrastructure ex-ists, further induce slope failure to occur (Solheim et al., 2005b).Submarine landslides can create devastating tsunamis, with se-vere consequences for coastal lowlands and their communities.This adds a socioeconomic aspect to the importance of under-

standing submarine mass-transport processes and mass-transportdeposits (Hampton et al., 1996; McSaveney et al., 2000; Tappin etal., 2001; Ward, 2001; Bardet et al., 2003; Bondevik et al., 2005).

The Norwegian–Barents–Svalbard continental margin (Fig. 1)is sculpted by numerous submarine landslides, such as Storegga(Bugge et al., 1988; Haflidason et al., 2003; Haflidason et al., 2004;Bryn et al., 2005a; Kvalstad et al., 2005; Solheim et al., 2005b),Trænadjupet (Laberg and Vorren, 2000; Leynaud and Mienert,2003), Andøya (Laberg et al., 2000), Nyk (Lindberg et al., 2004),Bear Island (Laberg and Vorren, 1993; Kuvaas and Kristoffersen,1996), and Hinlopen–Yermak (Vanneste et al., 2006; Winkelmannet al., 2006; Winkelmann and Stein, 2007). The Storegga landslideis probably the best-studied submarine slope failure, highlight-ing cause, initiation, and mobility from a geomorphologic pointof view to sediment breakup, flow dynamics, and tsunami poten-tial (Solheim et al., 2005b). Equally important, the long-termhistory of slope failures is addressed off mid-Norway. Sedimen-tation and slope failure appear closely related to climatic cycleson (formerly) glaciated margins (Solheim et al., 2005a). Duringpeak glacial times, ice streams deliver large volumes of sedimentsto the upper continental slope, where they deposit extensiveprograding fans, thereby rapidly loading slope sediments (Vorrenand Laberg, 1997; Vorren et al,. 1998). During interglacials, sedi-

Figure 2Figure 2

0º

5º10º

15º 20º 25º30º

35º

79º

80º

81º

82º

83º

HY

50 km

YBYB

NSBNSB

WS

CW

SC

Yermak PlateauYermak Plateau

SpitsbergenSpitsbergen

Nordaust-landet

Nordaust-landet

HTHT

Hinlopen StraitHinlopen Strait

Nansen BasinNansen Basin

Fram StraitFram Strait

KvitøyaKvitøya

200 km

Greenland

Scandinavia

SvalbardSvalbard

BarentsBarentsSeaSea

Fram StraitFram Strait

NansenNansenBasinBasin

Northern Northern North North Atlantic Atlantic OceanOcean

YPYP

TS

NS

Storegga

BIS

AS

1

0

3

2

4

5

6

Wat

er d

epth

(km

)AB

FIG. 1.—A) Overview map of the northern North Atlantic ocean margins, with the outlines (black) of large-scale submarine landslides(solid, Holocene; dotted, pre-Holocene; AS = Andøya Slide; BIS = Bear Island Slide; NS = Nyk Slide; TS = Trænadjupet Slide), theapproximate position of the ice margin during Weichselian glaciations (dashed white line), and major ice-stream pathwaysdraining the continental ice sheets during peak glacial times (blue arrows). The white box marks the area shown in Part B. B)Bathymetric map of the northern Svalbard continental margin, bordering the Yermak Plateau and opening toward the deepNansen Basin in the Arctic Ocean from the International Bathymetric Chart of the Arctic Ocean (IBCAO) (Jakobsson et al., 2000).The Hinlopen–Yermak slide scar (HY, bold black line) lies off the mouth of the Hinlopen cross-shelf Trough (HT). The yellowarrow marks the evacuation direction of the mass-transport deposits mobilized in the Hinlopen–Yermak landslide. The inflowof temperate water masses (dark red arrows), the North Svalbard Branch (NSB), and the Yermak Branch (YB), both derived fromthe West Spitsbergen Current (WSC), enter the study area along the upper continental slope. The blue arrows represent ice-streampathways. Dashed black lines represent fault systems on Svalbard and its shelves. Two faults intersect the eastern part of the slidescar. White box marks the area shown in Fig. 2.


mentation rates are subdued, and sediments consist mainly offine-grained hemipelagic deposits that commonly are reworkedby contour currents (Eiken and Hinz, 1993; Damuth and Olsen,2001; Bryn et al., 2005b). These (glacio-)marine clays are com-monly geotechnically weaker and sensitive; overloading of thesesediments by low-permeability, glacigenic sediments may resultin strain softening and development of excess pore pressure,which significantly reduces slope stability (Laberg et al., 2003;Haflidason et al., 2004; Bryn et al., 2005a; Gauer et al., 2005;Kvalstad et al., 2005). This correlation between submarine slopefailures and glacial processes is in fact not particular to theNorwegian–Barents–Svalbard continental margins but also analo-gous to, for example, the Canadian east coast (Piper and McCall,2003; Piper et al., 2003; Mosher et al., 2004) and likely to otherglacially influenced continental margin settings.

Pinpointing pre-conditioning factors and actual triggeringmechanisms of paleo-landslides is complex, but these most likelyinvolves excess pore pressure and/or weaker layers within spe-cific stratified sequences (Canals et al., 2004). Ultimate triggers forslope failure on (formerly) glaciated margins may amongst otherfactors be seismicity, which relates to crustal deformation due toice loading and unloading cycles (glacio-isostasy). Studies ofsubmarine landslides also revealed surprisingly high mobility ofmass-transport deposits. Rafted blocks with volumes up to 106 m3

and debris flows have regularly been observed at tens to hun-dreds of kilometers downslope of their source area (Bugge et al.,1988; Canals et al., 2004; De Blasio et al., 2004; Ilstad et al., 2004;Lastras et al., 2005; De Blasio et al., 2006), which reflects degree ofdisintegration of a failed mass as well as dynamics of mass trans-port on gently dipping slopes. Hydroplaning or lubrication couldfacilitate these processes (Harbitz et al., 2003; De Blasio et al., 2006).

This paper documents an integrated, multi-disciplinary studyof the geomorphologic characteristics, mobility, and implicationsof mass-transport deposits remobilized from the northern Svalbardmargin in a massive submarine landslide, named the Hinlopen–Yermak landslide. The Hinlopen–Yermak landslide’s escarp-ment heights (hundreds of meters to more than 1400 m) and theoccurrence of large blocks (up to 2 km3), rising hundreds ofmeters above the seabed, present challenges to understandingrelease, mobility, and dynamics of mass-transport deposits. Di-mensions of landslides in combination with the results of land-slide dynamic modeling (e.g., high velocity and wide frontal areaof the landslide) introduce nonlinearity and frequency disper-sion. Nonlinearity causes wave steepening, breaking, or boreformation and is more important in shallower water due toamplification, whereas frequency dispersion is more pronouncedin deep waters and over longer distances. These effects need to beincorporated into tsunami simulations to be able to investigatethe impact of the landslide-induced sea-surface elevation in theArctic and northern North Atlantic Oceans. Study of recent mass-transport deposits, their geomorphology, and consequences mayfurthermore help in interpreting deposits of paleo-slides recog-nized on geophysical data sets.

GEOLOGICAL SETTING

The northern Svalbard margin (Fig. 1) is bordered by theYermak Plateau and the Nansen Basin, and is part of the passivecontinental margin of the Eurasia Basin, which opened approxi-mately 60–55 Ma (calendar years). Evolution of this margincoincided with separation of the Lomonosov Ridge from theSvalbard–Barents margin and subsequent spreading along theultra-slow Gakkel Ridge (Jokat, 2005). The presence of fault zonessubparallel to the coast (Fig. 1) provides evidence of this riftingprocess (Eiken, 1994). However, it is ice-sheet dynamics during

Plio-Pleistocene glaciations that shaped the architecture of thenorthern Svalbard margin to a large extent, affecting both sedi-ment deposition and erosion (Vorren, 2003; Ottesen et al., 2005).An increase in ice-rafted debris recovered in deep boreholesindicates that glaciers advanced to open seas around Svalbard asearly as approximately 2.5 Ma, as a response to northern hemi-sphere cooling. Extensive glaciations have occurred frequentlysince approximately 1.6 Ma (Butt et al., 2000). Weichselian (LatePleistocene) ice sheets reached the shelf breaks around Svalbard(Lambeck, 1996; Svendsen et al., 2004), with fast-flowing icestreams actively discharging the greatest part of ice and sedimentthrough cross-shelf troughs (Bennett, 2003). This was the case forthe Hinlopen Trough (Ottesen et al., 2005). Sediment transportthrough Hinlopen Strait and Trough endured during interglacialtimes (Koç et al., 2002). Total sediment thickness on the northernSvalbard margin ranges from several hundreds of meters to ap-proximately nine kilometers. Upper Plio-Quaternary cover con-sists mainly of stacked glacigenic debris lobes and (glacio-) marinedeposits partly reworked by contour currents (Geissler and Jokat,2004). The partly sea-ice-covered northern Svalbard margin re-ceives temperate-water inflow of the North Svalbard and YermakBranches (Fig. 1). The source of this inflow is the West SpitsbergenCurrent, which flows along the upper slope (100–200 to 600–800m) and supplies the main portion of Atlantic Water to the ArcticOcean (Myhre and Thiede, 1995; Slubowska et al., 2005).

Swath bathymetry data recently revealed details of a largesubmarine landslide on the northern Svalbard margin (Fig. 1)(Vanneste et al., 2006), a landslide first described in 1999 (Cherkiset al., 1999). The Hinlopen–Yermak slide scar at the mouth of theHinlopen Trough (Fig. 1) is an example of margin collapse in aglacier-fed siliciclastic environment (Vanneste et al., 2006;Winkelmann et al., 2006; Winkelmann and Stein, 2007). Despiterelatively small drainage and slide-scar areas, the Hinlopen–Yermak landslide is exceptional in volume (see below), headwallheight (exceeding 1400 m), and dimensions of rafted blocks (up to1.89 km3), compared to other landslides in similar environments.The total area affected by the landslide may well be over 10,000km2, with run-out distance exceeding 300 kilometers (Winkelmannet al., 2006). The Hinlopen–Yermak landslide is the first mappedmega-landslide in the Arctic Ocean (Vanneste et al., 2006). Basedon the amphitheater-shaped slide-scar area with a composite setof linear escarpments and roughly planar slip surfaces, theHinlopen–Yermak landslide is a translational, multi-phase slopefailure that developed retrogressively (Vanneste et al., 2006).Whether the Hinlopen–Yermak landslide was one single eventwith different phases, occurring in near succession (like theStoregga landslide), or in several episodes well spread over timeis not clear, although the former is believed to be more realistic.Seismic data provide an indirect dating of the landslide up to theLast Glacial Maximum, by the presence of semi-transparent,stacked lobes on the headwall, interpreted as glacigenic debris-flow deposits, originating from ice-stream activity during the LastGlacial Maximum (Vanneste et al., 2006). Sedimentological analy-sis and AMS radiocarbon dating on planktonic foraminifera sug-gest an age of approximately 30 ka, and strengthens the geophysi-cal interpretation. At 30 ka, sea level fell and glacier ice volumeincreased, which presumably led to increased glacio-tectonic ac-tivity (Winkelmann et al., 2006; Winkelmann and Stein, 2007).

DATA AND METHODS

Multi-Beam Swath-Bathymetry Data

With a Simrad-Kongsberg EM300 system, approximately 4795km2 of seabed was mapped, mainly covering the Hinlopen–


Yermak slide scar area between 80.60°–81.65° N and 13.0-17.5° E(Fig. 2), during the Euromargins survey onboard R/V Jan Mayen(University of Tromsø) in October 2004. The motion-compen-sated echo-sounder system operates at 30 kHz. Maximum angu-lar coverage is 150 degrees, with 135 individual beams. Use of theequiangular acquisition mode has the advantage of maximizing

spatial coverage but comes at the expense of lower samplingdensities with increasing water depths. Beams are converted intowater depth on-site, using accurate sound velocities through thewater column, measured with CTDs (conductivity-temperature-depth recorders) during the survey. Data processing consisted ofcleaning and filtering navigation data, noise reduction, and data

81.2º

81.0º

80.8º

80.6ºN

14.0ºE 15.0º 16.0º 17.0º

81.6º

81.4º

10 km

2.8

2.4

2.0

1.6

1.2

0.8

0.4

Wat

er d

epth

(km

)

Figure 4

Figure 5

1.8 2.0 2.22.11.9

2 km

Water depth km)VE: ~7x

Figure 6A

Figure 6B

FIG. 2.—Illuminated EM300 swath bathymetry of the slide scar area of the amphitheater-shaped Hinlopen–Yermak landslide. Theswath bathymetry data cover an area of approximately 4795 km2. Blue arrow south of the headscarp represents the sedimentinflow from the Hinlopen Trough. Slide-scar area is characterized by multiple, linear escarpments that are several hundreds ofmeters high with fresh-looking slip planes, as well as a pronounced constriction. The black box downslope from the constrictionmarks the blocky mass-transport depositional area, highlighted in Fig. 4. The encircled rafted block is shown in more detail onthe pseudo-3D view in the inset above (vertical exaggeration (VE): ~ 7x). This 450-m-high rafted block contains an estimatedvolume of 1.89 km3. The dashed black box shows the area used in the landslide reconstruction (Fig. 5).


editing using Neptune software. Final gridding using a near-neighbor algorithm, imaging, and analysis utilized generic map-ping tools (GMT) (Wessel and Smith, 1998), with bin size deter-mined by the beam density and areal coverage, and search radiusset for improving signal/noise ratio. Analysis of beam density asa function of mean water depth (Fig. 3) was done by dividing 3600km2 of the surveyed area into 1 km2 subregions and countingbeams in each of these subregions. The results indicate that a cellsize of 75 m by 75 m (corresponding to approximately 180 beams/km2) is acceptable for the entire surveyed area. Finer binning (50m by 50 m, or 400 beams/km2) was possible for the area shownin Figure 4. For better visualization, data from the rafted blockwere resampled (5 m by 5 m) (Fig. 2, inset).

In order to reconstruct pre-slide bathymetry, an optimalDelauney triangulation and gridding method in Cartesian units(UTM Zone 33, generic mapping tools) was used (Wessel andSmith, 1998), after removing swath-bathymetry data that fallwithin the slide scar. This procedure uses only data points that arenot affected by the landslide, and assumes convergence of bothdata sets (boundary condition) where the slide scar terminates.Difference between pre-slide and present-day bathymetry repre-sents an estimate of volume as well as shape of the mass exca-vated from the slide-scar area (Fig. 5). Prior to subtracting thesetwo grids, reconstructed pre-landslide bathymetry is smoothedusing a Gaussian filter. Note that the bathymetry data covers onlythe slide-scar area, and no other data are available for such areconstruction.

Seismic Reflection Profiles

During the same geophysical survey, surface-towed single-channel seismic reflection profiles were recorded, mainly in the

area of the Hinlopen–Yermak slide scar (Fig. 6). The acousticsource consisted a double-sleeve gun (0.65 l or 40 in3 each), withfrequency content between 30 and 450 Hz. Shots were time-triggered (10 s), which corresponds to variable shot spacingbetween 20 and 24 m. The data shown in this study are processedwith amplitude scaling (automatic gain control) and bandpassfrequency filtering.

Modeling I: Landslide Dynamics and Mobility

A numerical simulation of landslide dynamics and mobilitywas performed, based on evidence from slide-scar geomorphol-ogy as well as the location of the landslide and the extent andthickness of mass-transport deposits (Fig. 7). The simulation usesreconstructed pre-slide bathymetry to define the top of the de-bris, whereas the slip plane is subparallel to present-day bathym-etry and is currently draped with mass-transport deposits (con-firmed observations from geophysical data, Fig. 6). In the model,debris pushes against a solid block initially at rest on a slip plane.This block is initially completely buried by debris. The debrisbehaves as a deformable viscoplastic Bingham fluid, and itsdynamics follows the BING model (Imran et al., 2001). In otherwords, the landslide moves as a plug flow riding on top of a shearlayer with a thickness determined by flow height and yield stressof the material. The model assumes that landslide material disin-tegrated and liquefied instantly upon failure. Impact and shearforces acting on loose debris are calculated, following a previ-ously suggested model for the BIG’95 debris flow on the Ebromargin, northwestern Mediterranean Sea (Lastras et al., 2005).Total force on the block Ftotal is then written as

with FG the gravity force; FC the Coulomb friction of the blockwith shear surface; FI the impact force exerted by disintegratedmobile material on the rafted block; FEP the static earth pressureforce due to lateral pressure of mobile material in contact with theblock; FD the viscous drag of sea water; and FS the shear force ofdebris flow on the block. The block itself is simplified with its longaxis oriented parallel to flow. The model takes into accountArchimedean buoyancy and acceleration of water at rest (added-mass effect). The shape of debris is estimated from reconstructedbathymetry (see below). The model yields both frontal velocitiesand accelerations (averaged over a representative foremost partof the debris flow) as a function of time and position, which canthen be used as input for tsunami simulations (see below). Note,however, that velocity estimates are hampered by uncertainties,because contemporary properties of the sliding masses are un-known. For further details on the modeling approach, refer toLastras et al. (2005). Parameters used are: yield stress = 20 kPa;viscosity = 30 Pa; internal friction angle of the block = 1°; blockdensity = 1.9 kg/m3; density of the slurry = 1.8 kg/m3. Ultimately,a sensitivity analysis of several parameters was performed.

Modeling II: Tsunami Simulations

Numerical tsunami simulations were conducted to evaluatethe extent to which a tsunami generated by the Hinlopen–Yermaklandslide might have impacted coastal areas surrounding theArctic Ocean. Velocity results from landslide dynamic modeling(see above) are input parameters fed into tsunami simulations.The depth matrix for tsunami simulations uses the InternationalBathymetric Chart of the Arctic Ocean (IBCAO) data (Jakobssonet al., 2000), resampled in a 2 km x 2 km grid in polar stereographicprojection with true scale at 75° N. At the time of slope failure,

75 m x 75 m

50 m x 50 m

100

1000

10000

Bea

m D

ensi

ty (

km-2)

500 1000 1500 2000

Mean Water Depth (m)

FIG. 3.—Beam density as a function of mean water depth for 3600subregions of 1 km2 of the surveyed area. Red horizontal linesrepresent the beam densities that correspond to acceptablegridding cells used (75 m by 75 m or approximately 180beams/km2 for the entire grid shown in Fig. 2; 50 m by 50 mor 400 beams/km2 for the distal area shown in Fig. 4).

SDEPICGtotalFFFFFFF +++++=


1.8 1.9 2.0 2.1 2.2 2.3

Water depth (km)N

sidewall/

escarpment

sidewall/

escarpment

raftedblocksrafted

blocks

slidedebrisslide

debris slidedebris

slidedebris

dist

ance

to h

eadw

all:

63 k

m

arcuateridges

arcuateridges

5 km

5 km

20 m

5 km

20 m

20 m

20 m

20 m

5 km

5 km5 km

5 km

W E

NW SE

VE: 100x

VE: 100x

VE: 100x

VE: 100x

VE: 100x

VE: 100x

A

B

C

D

D

E

E

F

A

B

C

F

RB

RB

side

wal

l

shoulder

shoulder

RB

DL DLDLDL

DL DL

DL

DLDL

R R

RB

RB

RB

RB

R R

Abbreviations:RB = Rafted BlockDL = Debris LobeR = Pressure Ridge

FIG. 4.—Detailed swath-bathymetry image of the part of the mass-transport deposits of the Hinlopen–Yermak landslide(location shown in Fig. 2). Lobes consisting of smaller mass-transport deposits (thin dotted line) accumulated close to thesidewalls, and may represent collapse following the main failure phase. The main lobe (thick dotted line) traveled throughthe constriction to the south (Fig. 2), and then the sediment spread out farther downslope. The mass-transport depositconsists of pronounced arcuate pressure ridges and large rafted blocks. Several bathymetric profiles show characteristicsof the debris lobes (Profiles A through C). As a reference, Profile D runs outside the main mass-transport deposit where—despite some irregularities in the order of only a few meters—the seabed is more or less gently dipping. Relief of the mass-transport deposit is hummocky, with ridges rising between 10 and 60 m high, and wavelengths less than 1 km (Profile E).Profile F runs across some large rafted blocks, including the block shown in the inset of Fig. 2. All profiles are displayed withsame scales and vertical exaggeration (100x).


approximately 30 ka (Winkelmann et al., 2006), eustatic sea levelwas approximately 81 m lower than today (Fleming et al., 1998),which left large parts of the Siberian shelves exposed as well asthe northwestern corner of the Barents Sea margin. Accordingly,the bathymetry grid was reduced by 81 m to comply with contem-porary eustatic sea level. For a computational domain spanningthe Arctic Ocean, the stereographic mapping factor ranges from0.98 to 1.05 and was ignored in the calculations.

The potential influence of sea ice in the Arctic Ocean is nottaken into account in the tsunami simulations. Indeed, a rigid icelid above the landslide area would most probably affect bothlandslide dynamics and tsunami generation mechanisms. Thereare, however, no published data on either the effect of sea ice ontsunamis or variations in thickness of Arctic sea ice during thePleistocene. Occurrence of planktonic foraminifera (e.g., N.pachyderma) in sediment cores from the Arctic Ocean, spanning

both glacial and interglacial periods (Hebbeln et al., 1998), im-plies that the Arctic Ocean was far from a permanently sea-ice-covered area even at peak glaciations. The Arctic Ocean wasinstead characterized by at least seasonally open-water condi-tions. This suggests that ocean ice cover was not very thick(maybe not much thicker than today) and that parts of the ArcticOcean may well have been ice-free at time of failure. If so, sea icewas either already broken into pieces by wind, waves, andcurrents or intact, but thin (newly frozen) as seen today. In orderto affect the typical long gravity waves—e.g., tsunamis—a sub-stantially thick and continuous ice cover is paramount. Break-upand wave energy dissipation associated with long waves (i.e.,small sea-surface curvatures) is very small. Moreover, continen-tal ice shelves or glaciers did not extend farther than the shelfbreak on the Norwegian–Barents–Svalbard margin (Ottesen etal., 2005), and they reached the shelf break only during peak

Reconstructed Pre-landslide Water Depth (m)

500 1000 1500 2000 0 200 400 600 800 1000 1200

Landslide Thickness (m)

10 km

10 k

m

A B

FIG. 5.—A) Reconstructed pre-landslide bathymetry from optimal Delauney triangulation and subsequent Gaussian filtering on theswath bathymetry data outside of the slide scar. Bold white curve represents the slide scar. Color code is the same as in Fig. 2, tofacilitate comparison. B) Shape of the total volume (1220 to 1350 km3) excavated from the slide-scar area (2375 km2), calculatedas the difference between the grids presented in Figs. 2 and 5A. Contour lines are 500 m (bold), 100 m (medium), and 20 m (thin)for both figures. Data are projected as UTM Zone 33.


glaciations (Nygård et al., 2007), thus not at the time of failure.These considerations and facts indicate that there was most likelynot a rigid ice lid above the landslide area when the landslideoccurred. Finally, a no-ice-cover assumption provides the mostconservative estimates of sea-surface elevations from engineer-ing and geohazard perspectives. It is nevertheless important tostress that further advanced research and numerical modeling isneeded on this topic.

In tsunami models, the landslide is simplified as a flexible boxwith a prescribed velocity progression. This box is smoothed toavoid numerical noise due to sharp edges. Landslide volume isconstrained by field observations and bathymetric reconstruc-tion at 1.32 x width (28) x height (0.610) x length (60) km3 orapproximately 1350 km3. A factor of 1.32 comes from the appliedsmoothing. The height of the flexible box in the tsunami simula-tions exceeds average landslide thickness by approximately 20%

escarpmentescarpment

MTDMTDMTDMTDglacigenicglacigenic

debrisdebrisflowsflows

rafted blocks

seabedseabed

2.8

3.0

3.2

Profile 04JM056S5 km

2.8

3.0

3.2

3.4

3.4

seabedseabed

MTDMTD

MTDMTD

MTDMTD

rafted blocks

Two

-way

Tra

vel T

ime

(s)

Two

-way

Tra

vel T

ime

(s)

West-Southwest East-Northeast

A

B

Profile 04JM057S5 km

FIG. 6.—Single-channel seismic profiles across mass-transport deposits downslope from the constriction (see Fig. 2 for location of theprofiles). A) Profile 04JM056S approaches the eastern sidewall of the Hinlopen–Yermak landslide, and crosses both medium tolarge rafted blocks. The thick, semitransparent unit in the landslide area is interpreted as the main Hinlopen–Yermak mass-transport deposit (labeled MTD). This unit overlies a weak reflection. A secondary mass-transport deposit with hummocky reliefoccurs towards the east-northeast of this profile. B) Profile 04JM057S crosses several terraces of the western escarpment. A similarsemi-transparent unit characterizes the uppermost sedimentary succession within the landslide area. On the terraces, stackedglacigenic debris flow deposits occur, with acoustic character different from that of the mass-transport deposits.


to compensate to some extent for the irregular shape of thelandslide with maximum thickness of up to 1400 m in the headwallarea. In the model, landslide propagation follows a straight line.In light of the overall uncertainty (landslide parameters, etc.),these models give reasonable results for tsunami propagation.The initial phase of landslide progression determines wave char-

acteristics (Løvholt et al., 2005). Coriolis effects are neglected,because it is known from other studies that they are of minorimportance during computational real time of this study (6hours). For further details on landslide representation, see Harbitz(1992) and Løvholt et al. (2005).

Most tsunami simulations are accomplished in two horizontaldimensions (2HD) for sea-surface elevation and depth-averagedhorizontal wave current speed, by solving linear, non-dispersive,shallow-water equations for conservation of mass and momen-tum (Mei, 1989). In these equations, it is assumed that the charac-teristic sea-surface elevation of waves (i.e., deviation from equi-librium water level) is much less than typical water depth. Inaddition, the characteristic wavelength is much longer than thetypical water depth. For further details on numerical 2HD tsu-nami modeling, see Harbitz and Pedersen (1992).

To assess nonlinearity and dispersion, simulations based onthe weakly dispersive, weakly nonlinear Boussinesq equations(Peregrine, 1972) are performed in one horizontal dimension(1HD) along specific profiles. The 1HD model is nested with the2HD model, by using data recorded as time series in 2HDsimulations as boundary conditions for 1HD simulations alongspecific profiles. These profiles are aligned roughly parallel tothe direction of wave propagation. The 1HD model facilitatesdevelopment of undular bores, but breaking waves and hy-draulic jumps are not included. The 1HD equations of motionare solved numerically on a staggered grid, with resolutiondepending on water depth, which allows finer resolution closeto shorelines. Grid refinement tests were performed to ensureconvergence for each run. 2HD effects, including reflectionsfrom surrounding islands, refractions, focusing, and interfer-ence, are neglected in 1HD simulations. 1HD simulations mayoverestimate sea-surface elevations, inasmuch as radial spread-ing of waves is not taken into account. The 1HD model can alsobe run with a landslide (corresponding to landslide configura-tion in the 1HD model), instead of nested with the 2HD model.The shoreline in both tsunami models is represented by avertical and impermeable wall (i.e., no-flux boundary condi-tion), providing a doubling of sea-surface elevation upon reflec-tion. Bathymetry for the 1HD simulations is interpolated fromthe 2HD matrix.

RESULTS AND DISCUSSION

Hinlopen–Yermak Landslide Mass-Transport Deposits

On the swath bathymetry data set, there is little evidence ofmass-transport deposits adjacent to the main escarpments. Sig-nificant parts of the slide-scar area appear virtually free of land-slide deposits (Fig. 2), which suggests a highly efficient evacua-tion mechanism for the Hinlopen–Yermak landslide. Varioustypes of mass-transport deposits occur north of approximately80.8° N, and are distinct in the intermediate part of the landslidearea, beyond a pronounced constriction (Fig. 2). Two types ofmass-transport deposits are identified downslope of this con-striction. The first type consists of randomly distributed seafloorfabric and small blocks (approximately 25 m wide). These mass-transport deposits lack flow structures and occur close to thesidewalls. This suggests that they probably originated from rockavalanching or sidewall collapse, and are therefore secondary innature. The second type represents the upper part of the mainmass-transport deposits. These deposits contain mainly sedi-ment lobes with material released from the headwall area, andthey traveled through the constriction. From the illuminatedswath bathymetry image (Fig. 2), it appears as though part ofthese sediment lobes can be traced back to an area close to a major

4

3

2

1

0

Wat

er D

epth

(km

)

4

3

2

1

0

Wat

er D

epth

(km

)

4

3

2

1

0

Wat

er D

epth

(km

)

4

3

2

1

0

Wat

er D

epth

(km

)

0 50 100 150 200 250

4

3

2

1

0

Wat

er D

epth

(km

)

0 50 100 150 200 250

Distance along Track (km)

after 25 seconds

after 378 seconds

after 1000 seconds

after 3798 seconds

Present-Day

FIG. 7.—The top four profiles show snapshots of the landslideevolution and the position of the rafted block at discrete times(in seconds) of the simulation. Profile starts at the 1400-m-high headwall and follows the yellow arrow shown in Figure1. The cobalt line represents the slide scar, the downslope baseof the mass-transport deposit, and the assumed flow path; thecarmine line is the top of the debris flow; the olive-shadedgreen area represents the intact rafted block. The rafted blockis entirely buried by the landslide debris during much of thelandslide event. Bottom panel shows present-day bathymetryalong the path used in the calculation for comparison, andillustrates a nice match for both topography and position oflarge rafted blocks.


escarpment. The terrain is hummocky, with high relief (Fig. 4).Several profiles across the mass-transport deposits (Profiles Athrough C in Fig. 4) reveal their positive relief and lobate form.The thickness of the lobate part of the mass-transport deposits ishighly variable. From the curvature of its surface, the lobate partthickens and widens downslope. The minimum thickness in-creases from approximately 20 m to over 80 m (Profiles A throughC and E in Fig. 4), whereas its width increases from a fewkilometers upslope of the constriction (Fig. 2) to more than 20 kmin the most distal part (Profile C in Fig. 4). The geomorphology ofthe lobe is dominated by transverse (i.e., normal to the landslidedirection) undulations, which form arcuate ridges up to severalkilometers long. These features suggest sediment flow (Fig. 4).Heights range from 10 to 60 m, and wavelength between ridgecrests is typically less than 1 km (Profile E in Fig. 4). Relief acrossridges is very different, compared to a profile outside the debrislobe, where irregularities are only meter-scale in height (Profile Din Fig. 4). These transverse undulations may represent pressureridges, formed by compressional deformation associated withcontinued sediment flow piling up (Prior et al., 1982; Posamentierand Kolla, 2003).

In addition to debris lobes, sediment blocks lie scatteredacross the seabed (Fig. 2). Bathymetric data image at least twelvelarge blocks and numerous smaller ones. These blocks are ran-domly oriented, and have variable sizes and shapes, but all havesteep walls (> 30°). Medium-sized blocks have semicircular foot-prints, whereas larger blocks have elongated to triangular orpolygonal footprints. Some occur as irregular features, whileothers resemble tabular blocks (Fig. 2). Larger blocks are sur-rounded by debris, which may have originated either fromerosion of blocks or by compressional deformation upon trans-portation and deposition. Blocks rise high above the seabed,typically more than 100 m. The largest of these blocks (Fig. 2 inset)climbs 452 m above the surrounding debris and stands on anelongated footprint of 8.7 km2 (maximum dimensions of 5.4 kmby 2.1 km). It has an estimated volume of at least 1.89 km3. Thesedimensions and shape suggest that blocks consist of consolidatedor lithified sediment. If not, they would probably have beeneroded, or internally deformed and remolded upon transporta-tion. Attempts to sample sediments from these blocks resultedonly in the corer bouncing off without retrieving any sample(Winkelmann et al., 2006). This result indirectly confirms thatthese blocks are likely at least partially lithified sediments. Blocksmay erode the seabed during downslope transport, therebycreating glide or debris tracks along their path (Ilstad et al., 2004;De Blasio et al., 2006). Clear evidence is absent from this data set(Fig. 4), further suggesting that these features may be buried bymass-transport deposits. Identifying glide tracks would mostlikely require 3D seismic data (Prior et al., 1982; Posamentier andKolla, 2003).

An important question to address is whether these blocks areallochthonous, remobilized by the Hinlopen–Yermak landslide,or in-place topographic highs (e.g., seamounts). The intact-blockexplanation is preferred, for the following reasons. Groupedblocks lie at the outer rim of debris lobes, which suggests thatthese features are related. Rafted blocks are often observed intranslational landslides (Mulder and Cochonat, 1996; Canals etal., 2004) and often go together with debris flows (Ilstad et al.,2004). As such, they are expected to occur in the Hinlopen–Yermak landslide. The nearby Mosby Seamount (Geissler andJokat, 2004; Winkelmann et al., 2006) is significantly larger thanthe blocks observed here. No significant volcanic activity isreported from this part of the margin, which makes groupedseamounts unlikely. As such, the large blocks appear to be raftedas part of the Hinlopen–Yermak landslide.

Considering the dimensions of these rafted blocks and thepresumed lithified nature of their constituent sediments, theyprobably originate from deeper parts of the excavated slabs.Several escarpments of the Hinlopen–Yermak landslide are morethan 600 m high (maximum headscarp height of the Hinlopen–Yermak landslide exceeds 1400 m). Therefore, the landslide cutsthrough consolidated sediments. Recent 3D seismic investiga-tions from the southwestern Barents Sea revealed buried mega-blocks related to fast-flowing ice-stream activity (Andreassen etal., 2004). These glacigenic mega-blocks (areal extent > 1 km2) andsmaller blocks (areal extent < 1 km2) have areal extent similar tothat of blocks identified in the Hinlopen–Yermak landslide area,but with lower relief. A similar genesis of the landslide blocks inthe Hinlopen Strait could be considered, given its glacial historyand location with respect to the Hinlopen cross-shelf trough.

Only a few seismic reflection profiles cross the mass-trans-port deposits downslope of the constriction (Fig. 6). Debris lobeswith hummocky seabed relief as described above form theuppermost part of a relatively thick (roughly 200 ms two-waytravel time) semitransparent sedimentary unit. This unit ex-tends across the entire width of the depression carved out by thelandslide, and is not present in adjacent parts of the margin (Fig.6). Rafted blocks have their roots in this semi-transparent unit(Fig. 6A). Lack of internal character, which could indicate thatthe material is to a large degree remolded and/or fluidized,discriminates this unit from stacked glacigenic debris-flow de-posits on the terraced western landslide flank (Fig. 6B). Thissemitransparent unit appears bounded by a weak, slightlyundulating reflection. These observations suggest that this unitis the main mass-transport deposit from the Hinlopen–Yermaklandslide in this area. The occurrence of this massive mass-transport deposit furthermore suggests that the main phase ofsliding was one large event. Nevertheless, mass-transport de-posits stacked on top of the semitransparent unit (Fig. 6A) andimaged by swath bathymetry (Fig. 4) indicate that subsequentsmaller failure phases took place as well. Additional high-quality seismic reflection data across the main mass-transportdeposits are necessary to elucidate the internal character—ifany—of the mass-transport deposits as well as the underlyingsedimentary environment.

Pre-Landslide Bathymetry Reconstructionand Mobilized Sediment Volume

Figure 5 presents the reconstructed bathymetry before thelandslide took place (A) as well as the shape and volume ofexcavated volume (B). The slide-scar area is approximately 2375km2. From this area, an estimated 1220 km3 of sedimentaryvolume was evacuated. Therefore, average thickness of theseremobilized sediments is approximately 510 m. The bulk of thisvolume was removed from the central headwall area in front ofthe Hinlopen Trough, adjacent to major escarpments (compareFig. 1 with Fig. 5). From the limited seismic data available, mass-transport deposits occur within the area used for reconstruction,thus total volume is probably larger (Fig. 6). In addition, recon-structed bathymetric contours bend southwards inside the west-ern corner of the slide scar, which is a consequence of theinterpolation method, and will underestimate the total volume.Considering these two factors, total estimated volume could wellbe approximately 10% larger. Therefore, the upper limit on thevolume excavated by the Hinlopen–Yermak landslide is approxi-mately 1350 km3.

Volume estimates, using distal mass-transport deposits thataccumulated beyond the area covered by the swath bathymetry,give a similar volume of 1100 to 1250 km3 for the inner landslide


area (Winkelmann et al., 2006), which is the same order ofmagnitude. Total run-out of these distal deposits exceeds 300 km,while the mass-transport deposits attain a mean thickness ofapproximately 200 m (Winkelmann et al., 2006). Discrepancies involume estimates are partly the result of the following issues:

(1) The bathymetric coverage of the landslide-affected area isincomplete.

(2) A detailed, high-quality grid of seismic reflection profiles thatcould be used to map these mass-transport deposits—both itslateral and vertical distribution throughout the slide-scar areaand the distal depositional area—in more detail is not avail-able. At the same time, accurate subsurface velocity profiles toconvert two-way travel time to depth within the mass-trans-port deposits also are needed.

(3) As with all models and interpolations, there are inaccuraciesinherent in the method used (triangulation), as mentionedabove.

(4) It is not uncommon that sediments are incorporated into theflow during sliding processes and transformation, whereasmass-transport deposits that accumulated within the slidescar are not taken into account in the total volume estimate bythis approach here.

(5) Finally, glacigenic post-landslide sediment infill derived fromthe Hinlopen Strait as evidenced from the swath bathymetryand seismic data (Vanneste et al., 2006) reduces the volumeestimate by pre-landslide bathymetry reconstruction. It wasimpossible to correct accurately for this volume. The thick-ness of the post-slide sediment infill varies laterally, andexceeds 150 m onto and immediately adjacent to the headwall.The sediment wedge pinches out at some 12 km fartherdownslope. This sediment infill corresponds, however, toonly a minor correction in volume.

Landslide Dynamics and Mobility

From analysis of the geomorphology of landslide materialdiscussed above, it appears that large landslide blocks origi-nated from the deeper layers of the headwall area. Hence, theyare denser and much stiffer than loose debris material, whichconstitutes the bulk of remobilized and possibly partially re-molded or fluidized landslide volume. It is also known that: (1)rafted blocks are scattered on the seabed at more than 60 kmfrom the headwall area and rise hundreds of meters abovesurrounding mass-transport deposits (Fig. 2); (2) total run-outdistance of debris flows reached several hundreds of kilome-ters, with typical thicknesses of approximately 200 m(Winkelmann et al., 2006); (3) the main phase of landsliding wasprobably one large event, with multiple phases, which hap-pened retrogressively in near succession; and (4) a smaller-scale, secondary landslide occurred. A concern in landslidedynamic modeling is to clarify how these large blocks weretransported over such large distances (Fig. 2), and how theyrelate to the mass-transport deposit. Hydroplaning, which ex-plains well the mobility of rafted blocks and debris flows insome cases (Mohrig et al., 1998; Elverhøi et al., 2002; Harbitz etal., 2003), cannot be invoked for rafted blocks of the Hinlopen–Yermak landslide; their unusual height requires a velocity of atleast 95 m/s for hydroplaning to take place. Therefore, a land-slide generation and transport mechanism is envisioned wherebyfailure of the headwall results in a loose debris flow, while

rafted blocks remained intact. This loose material sets the blockin motion and is responsible for transporting blocks to theirpresent-day location.

To test the validity of this scenario, a numerical simulationwas performed, in which the debris pushes against the solidblock initially at rest. Figure 7 shows the shape of disintegratinglandslide material, as well as the position of the rafted block atdiscrete times after slope failure. Simulation started from recon-structed bathymetry, with a rafted block completely covered bythe debris flow resting on the slip surface. Between 25 s and 378s, the rafted block traveled about 10 km, whereas the front of thedebris flow moved over a longer distance. After 1000 s, thedebris flow traveled nearly 40 km, and fanned out. Also, theblock moved roughly 30 km from its initial position and the crestof the block became uncovered. Resisting forces exceed drivingforces after approximately 4000 s, and movement comes to ahalt. At this time, the frontal part of the debris flow is located at190 km from the headwall, and the block is roughly 50 km fromits original position. This model is in reasonable agreement withfield observations (Figs. 2, 7). According to this modeling sce-nario, the block is at present partly buried by the debris flow,which fits interpretation of the seismic reflection profiles (Fig.6). It is worth noting that the rafted block has not surpassed thedebris lobe. Therefore, it is not an outrunner block in the strictsense of the term.

The velocity profile at the front of a debris flow is shown in Fig.8A as a function of position and in Fig. 8B as a function of timesince failure. In the initial phase, the front of the debris lobe isfound at roughly 55 km from the headwall (Fig. 5). Initial increasein acceleration results in a frontal velocity peak of 54 m/s (aver-aged over a representative foremost part of the debris flow), dueto the extreme height (approximately 1400 m) of the headwall offthe Hinlopen cross-shelf trough (Fig. 2). Accounting for the finitetime necessary for disintegration of the landslide material couldlower the initial acceleration. A second peak with maximumvelocity of about 64 m/s occurs after 1200 s (Fig. 8B). At this time,the front of the debris flow traveled more than 50 km and reacheda run-out distance of more than 100 km, where the increase inseabed dip adds to the velocity (Figs. 7, 8A). These results arequalitatively robust and mostly independent on the parameterset used. The calculated velocities are of the same order ofmagnitude as velocities in the Storegga area (peak frontal veloci-ties exceeding 60 m/s) (De Blasio et al., 2005), and the Big’95debris flow in the Mediterranean Sea (peak frontal velocitiesaround 45 m/s) (Lastras et al., 2005).

In the simulation, the rafted block moves in phase with thedebris flow, although with lower velocities (Fig. 8B). Its maxi-mum velocity is about 44 m/s, which is double the velocity ofrafted blocks in the Big’95 area (Lastras et al., 2005). After approxi-mately 1000 s, the rafted block towers above the debris, yet itmoves at approximately 30 m/s, or roughly half the velocity ofthe debris front. By then, the rafted block decelerates. The raftedblock reaches its final position after about 2400 s. Rafted blockspresumably lose momentum because of change in dip directionoff northern Svalbard, bending from roughly a south–northorientation to a southwest–northeast orientation (Fig. 1). For arafted block to reach its final position as observed on bathymetricdata (Figs. 3, 7), the initial distance of the block from the headwallcannot be less than the one assigned in this model (Fig. 7). Thereason is that the debris flow—at least in the Bingham modeladopted here—can be roughly envisaged as stretching of anelastic spring kept fixed at one end. Displacement of a point at thefront of the spring is twice the displacement of the midpoint, andmuch more than the displacement near the fixed end. Because arafted block is initially embedded in landslide debris and set in


motion by the debris flow, it cannot move farther than the debrisflow itself. This result, which is independent of properties of therafted block and of the debris flow, suggests a minimum initialdistance of about 25 to 30 km from main headwall for thelandslide block. This condition implies that provenance of raftedblocks does not coincide with the main headwall, but probablyoriginated farther downslope.

Many parameters enter the simulations. The most importantones are densities of the debris flow and block, yield stress,viscosity, friction coefficient at the block–terrain surface, anddrag coefficient. In addition, geometrical uncertainties such asthe initial position of the block and the initial shape of theexcavated volume may affect simulation results significantly.Unfortunately, all of these parameters are poorly constrained.Limited knowledge of these parameters necessitates a sensitiv-ity study, performed by running simulations with differentnumerical values of parameters varying within a certain plau-sible range. Consistency of these results can then be checked inretrospect, based especially on final block position, elevation ofthe block above the surrounding debris, and thickness of thedebris lobe in the distal part. Table 1 presents results obtainedfrom sensitivity analysis, on varying some of the parametersinvolved. The first simulation corresponds to results presentedin Fig. 8. Comparing simulations 1 and 2, for example, indicatesthe importance of the internal friction angle for run-out of therafted block. A decrease of yield stress, on the other hand, haslittle influence on block run-out (compare simulations 1, 3, and5). Interestingly, block run-out increases with decreasing yieldstress. This is because the debris flow then travels slower, andkeeps on pushing the block for a longer time, albeit with lessthrust because of reduced velocity. Another simulation illus-trates that viscosity has no effect on results (compare simula-tions 1 and 6).

Dependence of the peak velocities on the rheological param-eters of the debris flow was also investigated. A change in theyield stress by a factor of ten (from 5 kPa to 50 kPa) results in avelocity drop by 12% for the first velocity peak and 21% for thesecond peak. Viscosity needed to be changed by at least twoorders of magnitude to give a velocity decrease of 1%. Althoughsignificant, sensitivity of velocity on the yield stress is not dra-matic, which provides confidence in the simulated results.

Although results are necessarily dependent on the param-eter set chosen, patterns emerging from simulations may rea-sonably represent the true succession of events, and are avaluable tool in estimating tsunami properties. Additionally,landslide dynamic modeling with the rafted block pushed bythe debris adequately explains the final position of the largerafted block, the long run-out distance, and the lack of glidetracks within the debris lobe.

0

10

20

30

40

50

60

Fro

nta

l Velo

city

(m

/s)

0.0

600 1200 1800 2400 3000 3600 4200 4800

Time (s)

Time (hr)1.00.5

0

BINGSlide ASlide BSlide CBlock

30

35

40

45

50

55

60

65

Fro

nta

l Velo

city

(m

/s)

100 150 200

Distance along Track (km)

A

B

FIG. 8.—A) Velocities of the debris-flow front as a function ofdistance from the headwall. B) Velocities of the debris-flowfront (red curve) and the large rafted block (green curve) as afunction of time after failure, which was calculated by theBING model. Velocities are averaged over a representativeforemost part of the debris flow. Blue curves are the velocityfunctions for the three different landslide scenarios (A throughC) used in the tsunami modeling, as simplifications of theBING results. Profiles for scenarios B and C are shifted tomatch the peak velocity of the debris front.

TABLE 1.—Results obtained from sensitivity analysis of some of the parameters entering the BING landslide model(i.e., yield stress, internal friction angle, and viscosity). These results reported are the block run-out and the

elevation of the block above surrounding mass-transport deposits (debris flow).

Simulation Yieldt

Internal Frictionl

Viscosity Block Run-out Debris Thickness(kPa) (degrees) (Pa·s) (km) (m)

1 20 1 30 52 3752 20 2 30 24 2603 10 1 30 51 4104 10 3 30 15 3505 30 1 30 54 3356 20 1 300 52 375


Tsunami Simulations of the Hinlopen–Yermak Landslide

Geomorphologic analyses and landslide dynamic modelingprovide important evidence to investigate the tsunami potentialof the Hinlopen–Yermak landslide. Released volume, velocity,and initial acceleration of the landslide mass are key parameters,but also thickness of the sediment slab as well as time lag betweendifferent phases affect wave structure and, ultimately, sea-sur-face elevation. Based on landslide simulations using the BINGmodel, three scenarios for landslide velocity profiles are investi-gated (Fig. 8B). Landslide scenario A mimics rapid initial accel-eration of the BING model, a maximum velocity of 54 m/s, anda run-out distance of 200 km. The second, slower passing phaseof large acceleration of the landslide after about 600 s is modeledby landslide scenarios B and C. In these scenarios, the maximumlandslide velocity is 64 m/s and run-out distances are 175 km and200 km, respectively.

Fig. 9A through 9F shows six snapshots of the wave patternand propagation from 2HD simulations for landslide scenario A.Landslide motion (approximately to the north) leads to a strong

buildup in front of the landslide where water is displaced, whereasat the rear part a huge depression is formed. This creates apolarity reversal between waves traveling through the ArcticOcean (wave crest first) and those traveling south, entering theFram Strait (wave trough first) (Fig. 9). The northeastern coast ofGreenland is first hit by the tsunami crest between one and one-and-a-half hours after the landslide (Fig. 9B, C). After six hours,the pattern in the Arctic Ocean is chaotic (Fig. 9F). At the northernSvalbard coastline, the tsunami reaches shore in less than half anhour first by a depression, followed by large waves formed at thefrontal part of the landslide (Fig. 9A–C). The tsunami sweeps thenorthern Svalbard coast for at least several hours, during whichtime sea-surface elevation is significant (several meters). It takesabout one-and-a-half hours for the tsunami to travel through theFram Strait and enter the northern North Atlantic Ocean (Fig. 9C,D), whereas the northern Norwegian coast is hit after about threehours by a modest depression (Fig. 9E). The fairly shallow BarentsSea (contemporary water depths less than 250 m) is virtuallyunaffected by the landslide-generated tsunami during the firstfew hours (Fig. 9A–F), although sea-surface elevations rise in

Arctic Ocean

FramStrait

GreenlandGreenland

IcelandIceland

Svalbard

BarentsSea

Siberia

n A

rcticSib

erian

Arctic

Scan

dina

via

Scan

dina

via

0.5 hr0.5 hr 1.0 hr1.0 hr 1.5 hr1.5 hr

2.0 hr2.0 hr 3.0 hr3.0 hr 6.0 hr6.0 hr

A B C

D E F

0°E

70°N

90°ENorth Pole

−2

0

2

4

6

8

10

Sea–

surf

ace

Elev

atio

n (m

)

FIG. 9.—Snapshots of the tsunami sea-surface elevation (m) for landslide scenario A in the Arctic Ocean and northern North AtlanticOcean after 0.5 hour, 1.0 hour, 1.5 hour, 2.0 hour, 3 hour, and 6 hour (time in hours indicated in the lower right corner of each panel).Simulation is for landslide scenario A (see Fig. 8B for the landslide-velocity profile). Color coding for sea-surface elevation isidentical for all panels, with values clipped between -3 m and +10 m to facilitate comparison of wave patterns and sea-surfaceelevations at different times. Maximum elevations exceeding 130 m occur at the slide scar (in the eye of the spreading patternobserved in A). Gray areas represent landmass at the time of failure, with the black line the present-day coastline. Grid resolutionis 2 km by 2 km in polar stereographic projection with true scale at 75° N.


later stages (Fig. 9F). Reflection of the tsunami north and west ofSvalbard back into the Arctic and northern North Atlantic Oceansexplains relatively limited sea-surface elevations in the BarentsSea, of which the northwestern corner was contemporary dryland (Figs. 9, 10).

Fig. 10A shows maximum sea-surface elevation during thefirst six hours after slope failure. In the immediate vicinity of thelandslide evacuation area, landslide scenario A gives sea-surfaceelevations exceeding 130 m. Landslide scenarios B and C returnmaximum sea-surface elevations in the same area of approxi-mately 100 and 75 m, respectively. Radial spreading causes sea-surface elevation to decrease farther away, until the tsunami isamplified towards the coast because of seabed shoaling effects(see Fig. 10B for comparison with bathymetry). As anticipated,the largest impact from a landslide-induced tsunami encroachescoastal areas of northeastern Greenland and northern Svalbard.It should be kept in mind that equilibrium eustatic sea level in thesimulations is 81 m lower (30 ka) than today’s sea level.

The speed at which the tsunami travels across the oceans (Fig.10C) depends on water depth (Fig. 10B), and is given by

with g as gravity acceleration and H(x, y) as water depth. In thedeep Arctic Ocean, the tsunami travels with a speed mainlybetween 140 m/s and 220 m/s, whereas in the shallower Barentsshelf sea propagation speed is typically less than 55 m/s (Fig. 10C).For the deepwater passage from the central Fram Strait, wavevelocities are approximately 160 m/s. However, when a tsunamienters shallow waters, the wave slows down, becomes shorter, andincreases in height. For linear (hydrostatic) waves, the relationbetween sea-surface elevation (A) and water depth (H) is

which explains higher sea-surface elevations along the northeast-ern coast of Greenland and along the northern part of Svalbard,compared to waves traveling northwards into the Arctic Oceanand southwards through the Fram Strait. Higher sea-surfaceelevation along the northern part of Svalbard is also an effect ofreduced radial spread due to shorter distance to the landslidearea.

The development of the landslide (Fig. 7) illustrates that large,rafted blocks move as intact sediment blocks within the debrisslurry, and that it takes about 1000 s (approximately 17 minutes)before the block rises above the debris. This implies that largerafted blocks do not contribute to the tsunami in the initial phaseof landsliding. After 1000 s, when the block starts playing a rolein tsunami generation, it still moves at approximately 30 m/s buthas started to decelerate. Because initial acceleration is critical fortsunami generation, results of the landslide dynamic modelingindicate that rafted blocks—despite their remarkable height andvolume—do not contribute significantly to landslide-inducedtsunami generation.

This dataset presents unique possibilities to investigate wave-forms generated by submarine landslides. The 2HD model (Figs.9, 10) is a linear, hydrostatic wave model, thus it neglects bothdispersion and nonlinearity. In order to investigate coastal im-pact in more detail, 1HD simulations with the Boussinesq modelwere performed along profiles towards Greenland and Svalbard,with data from 2HD simulations used as boundary conditions atthe end points of profiles (Fig. 11 inset). Spatial, non-uniformresolution of simulations range from 3 to 189 m (Greenlandprofile) and 8 to 55 m (Svalbard profile). A conclusion from the1HD simulations is that the scenarios chosen are not strongly

0 5 10 15 20 25 30 35 40

Sea–surface Elevation (m)−4000 −2000 0 2000 4000

Topography/Water Depth (m)0 40 80 120 160 200 240

Wave Propagation Speed (m/s)

A B C0°E

0°E

70°N70°N

90°E90°ENorth PoleNorth Pole

FIG. 10.—A) Maximum sea-surface elevation (m) of the Hinlopen–Yermak tsunami in the Arctic Ocean and northern North AtlanticOcean, for landslide scenario A (see Fig. 8B for landslide velocity profile) during the first 6 hours after the landslide. Sea-surfaceelevation values are truncated at +40 m. Peak amplitudes in immediate vicinity of the landslide reach up to 130 m. B) IBCAObathymetry map used in the calculations for comparison, which is corrected for contemporary sea level. C) Linear shallow-waterspeed of wave propagation controlled by water depth. Gray areas represent landmass at time of failure, with the black line thepresent-day coastline. Grid resolution is 2 km by 2 km in polar stereographic projection with true scale at 75° N.

25.0~

−HA

( ) ( )yxHgyxv ,, ⋅=


affected by dispersion. However, when the tsunami enters shal-lower water, nonlinearity may contribute substantially to tsu-nami propagation (top panel in Fig. 11).

For simulation with the Boussinesq model, the leading waveevolves into a so-called undular bore, and fissions into (a train of)solitary waves if the tsunami travels over longer distances ofshallow water. Not only are these features due to weak nonlineareffects, but they are possible only in combination with weakdispersion and when a very steep front developed while theamplitude is still less then 0.3 times depth (Peregrine, 1966).Generally, they grow to about double the height of the originalbore. Additionally, solitary waves start to break if ratio of sea-surface elevation to depth exceeds a threshold value of 0.72(Kataoka and Tsutahara, 2004).

Sea-surface elevation-to-depth ratios higher than approxi-mately 0.5 fall outside the limitations of the Boussinesq model.Modeling of the three landslide scenarios also results in sub-stantially different sea-surface elevations, and hence the nonlin-ear effects on the tsunami differ (middle panel of Fig. 11). Sea-

surface elevation-to-depth ratio for the Greenland profile (Fig.11) is about 0.55 (the leading peak of scenario A obtained withthe Boussinesq model) and is probably too high for this model.For such high ratios, the model may to some extent be inaccu-rate; nevertheless, simulations indicate that undular bores formwith sea-level elevations exceeding 120 m. In order to describethese high and steep tsunamis more precisely, fully nonlinearmodels with shock capturing must be invoked. For the Svalbardprofile, sea-surface elevation-to-depth ratio for leading peaks inscenario A is approximately 0.47, thus it falls within the limita-tions of the Boussinesq model. It should be noted that theleading peak ratio is strongly dependent on the preceding sea-surface depression, causing reduced water depths. However,such refined investigations of the complex run-up phenomenonlie outside the scope of this paper.

Introducing the landslide in the 1HD wave model providedthe possibility of crosschecking whether or not nonlinearity anddispersion have an effect on initial sea-surface elevation of thetsunami (Fig. 12). Dispersion alters wave pattern slightly, since

0

100

200

300

400 Wat

er D

epth

(m

)

0 10 20 30 40 50 60

Distance (km)

−60

−40

−20

0

20

40

Sea

–sur

face

Ele

vatio

n (m

)

30 40 50 60

−120

−100

−80

−60

−40

−20

0

20

40

Sea

–sur

face

Ele

vatio

n (m

)

30 40 50 60

Distance (km)

0

1

2

3

4Wat

er D

epth

(km

)

0100200300400

Distance (km)

−60

−40

−20

0

20

40

60

80

100

120

Sea

–sur

face

Ele

vatio

n (m

)

280290300310320330

−60

−40

−20

0

20

40

60

80

100

120

Sea

–sur

face

Ele

vatio

n (m

)

280290300310320330

Distance (km)

Greenland Profile Svalbard Profile

No

rth-

eastSo

uth

wes

t

So

uth

No

rth

Linear HydrostaticLinear DispersiveBoussinesq

Linear HydrostaticLinear DispersiveBoussinesq

Slide CSlide BSlide A

Slide CSlide BSlide A

340˚

0˚

20˚

78̊

80˚

Gre

enla

nd

Svalbard

FIG. 11.—Top panels—1HD tsunami simulations towards Greenland (left side) and Svalbard (right side) using different mathematicalmodels (blue = linear hydrostatic model; olive = linear dispersive model; dark red = Boussinesq model). Middle panels—1HDtsunami waveform simulations towards Greenland (left) and Svalbard (right) using the Boussinesq mathematical model on threedifferent landslide scenarios. Frontal velocities of these scenarios are shown in Figure 8B. Bottom panels—IBCAO bathymetryalong the profiles used in the modeling. Gray shade in the bottom panels mark the area displayed in the top two panels. Waterdepths in the gray-shaded areas vary between 153 and 263 m (Greenland profile) and 7 to 304 m (Svalbard profile), respectively.Inset shows locations of profiles towards Greenland and Svalbard, with the green curve marking the outline of the Hinlopen–Yermak slide scar.


smaller waves are introduced on wave crests and troughs. Land-slide parameters in this test are similar to those used in the 2HDmodel, except for landslide height. The 1HD model lacks damp-ing due to radial spreading. This effect is simulated by reducingthe height of the landslide.

CONCLUSIONS

Integration of slide-scar geomorphology and characteristicsof its mass-transport deposits from detailed swath bathymetryand limited seismic reflection profiles along with landslide dy-namic modeling and tsunami simulations has improved under-standing of landslide processes. This study addresses processesinvolved in generating the mega-scale Hinlopen–Yermak land-slide as well as its implications to produce a tsunami.

• The Hinlopen–Yermak Landslide has evacuated an estimated1350 km3 from the northern Svalbard continental margin into

the Arctic Ocean. Little mass-transport deposits accumulatedin vicinity of a major escarpment. Beyond a distinct constric-tion, mass-transport deposits consist of extensive, remoldeddebris lobes with numerous large, scattered rafted blocks. Thesurface of debris lobes is hummocky, with arcuate pressureridges. The major escarpment and the largest rafted blocks areorders of magnitude larger than in other submarine land-slides on siliciclastic continental margins. The maximum run-out of landslide deposits is not well constrained but is at leastseveral hundreds of kilometers.

• Geomorphology combined with numerical simulation oflandslide dynamics (following the BING model) suggests alandslide development in which rafted blocks are intact andmove within a rapidly disintegrating debris slurry. Debrisflows and rafted blocks move with peak velocities of 64 m/s and 44 m/s, respectively. The main failure phase fromheadwall collapse over disintegration, transport, and re-

4

3

2

1

0

Wat

er D

epth

(km

)

0 50 100 150 200 250 300 350 400

Distance (km)

-60

-40

-20

0

20

40S

ea-s

urfa

ce E

leva

tion

(m)

0 50 100 150 200 250 300 350 400

No

rthSo

uth

1HD Linear Hydrostatic1HD Linear Dispersive1HD Boussinesq2HD simulation

A

B

320˚

340˚

0˚

20˚

40˚

40˚

78˚

80˚

82˚

FIG. 12.—A) 1HD simulation illustrating the effect of nonlinearity and dispersion in the tsunami generation phase, compared to aprofile extracted from the 2HD simulation at identical times, occurring 17 minutes after the landslide initiation. B) Bathymetryalong the profile, with the dotted line representing the landslide moving northwards indicated by the arrow. The landslidethickness is reduced by a factor of 0.25 (about 150 m) m to mimic the radial spread, which is not included in the 1HD model. Thisreduction is determined by comparing the wave heights of the 1HD and 2HD simulations. Inset shows the location of a north–south-trending profile located north of Svalbard. The green curve marks the Hinlopen–Yermak slide scar.


deposition took just over one hour. Overconsolidated raftedblocks originated from deeper layers in the slide scar area atapproximately 30 km from the headwall. Their mobility wasprobably not caused by hydroplaning, due to the unusualheight of the rafted blocks, which would require extremevelocities.

• Due to a large volume and especially wide frontal area of theHinlopen–Yermak landslide combined with a high maximumvelocity, sea-surface elevation of the simulated landslide-gener-ated tsunami is extremely high. Maximum values exceed 130 min vicinity of the headwall in the most extreme scenario. Atapproximately 20 km outside the northern Svalbard coast,maximum sea-surface elevation tops 40 m (i.e., about 40 mbelow present-day sea level, considering glacio-eustatic correc-tion applied). This suggests that on-land tsunami depositsoriginating from the Hinlopen–Yermak landslide are unlikelyto be discovered. Landslide development, in which the blocksare covered by debris during the first 1000 s, implies that raftedblocks did not contribute significantly to the generation of atsunami, despite their height, volume, and velocity.

• Refined simulations with a nonlinear and dispersive model(1HD Boussinesq) along specific profiles towards bothGreenland and Svalbard prove that waves are highly nonlin-ear near the shore. There, shoaling water depths amplify thesewaves, causing the front of the tsunami to evolve into anundular bore. These high sea-surface elevations lead to ex-treme run-up also at larger distances from the landslide area.No attempt to describe the run-up of these huge waves wasconducted. Due to steepness, the leading bore probably breaksfar off the coastlines. However, since waves in the shoalingpart are from one kilometer to tens of kilometers long, reduc-tion in run-up height due to frontal breaking is probablysmall.

ACKNOWLEDGMENTS

The paper is part of the European Science Foundation (ESF)project EUROMARGINS, funded by the Norwegian ResearchCouncil (NFR grant 158733/V30), and is contribution No. 168 ofthe International Centre for Geohazards (ICG). We are grateful toS. Iversen, S. Bünz, R/V Jan Mayen captain, and crew for theirsupport during data acquisition. Special thanks to C.F. Forsberg, P.Gauer, S. Guidard, D. Issler, T.J. Kvalstad, J.S. Laberg, F. Løvholt,G.K. Pedersen, A. Solheim, and P. Vanhauwaert for fruitful discus-sions and feedback. We appreciate the constructive reviews by J.Damuth, D. Mosher, and C. Shipp that improved this manuscript.Figures were made with the GMT (Wessel and Smith, 1998).Seismic data processing was performed with SeismicUnix.

REFERENCES

ANDREASSEN, K., NILSSEN, L.C., RAFAELSEN, B., AND KUILMAN, L., 2004, Three-dimensional seismic data from the Barents Sea margin reveal evi-dence of past ice streams and their dynamics: Geology, v. 32, p. 729–732.

BARDET, J.-P., SYNOLAKIS, C.E., DAVIES, H.L., IMAMURA, F., AND OKAL, E.A.,2003, Landslide tsunamis: Recent findings and research directions:Pure and Applied Geophysics, v. 160, p. 1793–1809.

BENNETT, M.R., 2003, Ice streams as the arteries of an ice sheet: theirmechanics, stability and significance: Earth-Science Reviews, v. 61,p. 309–339.

BONDEVIK, S., LØVHOLT, F., HARBITZ, C., MANGERUD, J., DAWSON, A., AND INGE

SVENDSEN, J., 2005, The Storegga Slide tsunami—comparing field

observations with numerical simulations: Marine and PetroleumGeology, v. 22, p. 195–208.

BRYN, P., BERG, K., FORSBERG, C.F., SOLHEIM, A., AND KVALSTAD, T.J., 2005a,Explaining the Storegga Slide: Marine and Petroleum Geology, v.22, p. 11–19.

BRYN, P., BERG, K., STOKER, M.S., HAFLIDASON, H., AND SOLHEIM, A., 2005b,Contourites and their relevance for mass wasting along the Mid-Norwegian Margin: Marine and Petroleum Geology, v. 22, p. 85–96.

BUGGE, T., BELDERSON, R.H., AND KENYON, N.H., 1988, The Storegga slide:Royal Society of London, Philosophical Transactions, v. A325, p.357–388.

BUTT, F.A., ELVERHØI, A., SOLHEIM, A., AND FORSBERG, C.F., 2000, Decipher-ing Late Cenozoic development of the Western Svalbard Marginfrom ODP site 986 results: Marine Geology, v. 169, p. 373–390.

CANALS, M., LASTRAS, G., URGELES, R., CASAMOR, J.L., MIENERT, J., CATTANEO,A., DE BATIST, M., HAFLIDASON, H., IMBO, Y., LABERG, J.S., LOCAT, J.,LONG, D., LONGVA, O., MASSON, D.G., SULTAN, N., TRINCARDI, F., AND

BRYN, P., 2004, Slope failure dynamics and impacts from seafloor andshallow sub-seafloor geophysical data: case studies from the COSTAproject: Marine Geology, v. 213, p. 9–72.

CHERKIS, N.Z., MAX, M.D., VOGT, P.R., CRANE, K., MIDTHASSEL, A., AND

SUNDVOR, E., 1999, Large-scale mass wasting on the north Spitsber-gen continental margin, Arctic Ocean: Geo-Marine Letters, v. 19, p.131–142.

DAMUTH, J.E., AND OLSEN, H.C., 2001, Neogene–Quaternary contouriteand related deposition on the West Shetland Slope and Faeroe–Shetland Channel revealed by high-resolution seismic studies:Marine Geophysical Research, v. 22, p. 369–398.

DE BLASIO, F.V., ENGVIK, L., HARBITZ, C., AND ELVERHØI, A., 2004, Hydro-planing and submarine debris flows: Journal of Geophysical Re-search, v. 109, doi:10.1029/2002JC001714.

DE BLASIO, F.V., ELVERHØI, A., ISSLER, D., HARBITZ, C.B., BRYN, P., AND LIEN,R., 2005, On the dynamics of subaqueous clay rich gravity massflows—the giant Storegga slide, Norway: Marine and PetroleumGeology, v. 22, p. 179–186.

DE BLASIO, F.V., ENOK, L., AND ELVERHØI, A., 2006, Sliding of outrunnerblocks from submarine landslides: Geophysical Research Letters, v.33, no. L06614, doi:10.1029/2005GL025165.

EIKEN, O., 1994, Seismic Atlas of Western Svalbard: Oslo, NorskPolarinstitutt, 73 p.

EIKEN, O., AND HINZ, K., 1993, Contourites in the Fram Strait: Sedimen-tary Geology, v. 82, p. 15–32.

ELVERHØI, A., DE BLASIO, F.V., BUTT, F.A., ISSLER, D., HARBITZ, C., ENGVIK, L.,SOLHEIM, A., AND MARR, J., 2002, Submarine mass-wasting on gla-cially-influenced continental slopes: processes and dynamics, inDowdeswell, J.A., and Ó Cofaigh, C., eds., Glacier-Influenced Sedi-mentation on High-Latitude Continental Margins: Geological Soci-ety of London, Special Publication 203, p. 73–87.

EMBLEY, R.W., AND JACOBI, R.D., 1977, Distribution and morphology oflarge submarine sediment slides and slumps on Atlantic continentalmargins: Marine Geotechnology, v. 2, p. 205–228.

FLEMING, K., JOHNSTON, P., ZWARTZ, D., YOKOYAMA, Y., LAMBECK, K., AND

CHAPPELL, J., 1998, Refining the eustatic sealevel curve since the LastGlacial Maximum using far- and intermediate-field sites: Earth andPlanetary Science Letters, v. 163, p. 327–342.

GAUER, P., KVALSTAD, T.J., FORSBERG, C.F., BRYN, P., AND BERG, K., 2005, Thelast phase of the Storegga Slide: simulation of retrogressive slidedynamics and comparison with slide-scar morphology: Marine andPetroleum Geology, v. 22, p. 171–178.

GEE, M.J.R., UY, H.S., WARREN, J., MORLEY, C.K., AND LAMBIASE, J.J., 2007,The Brunei slide: A giant submarine landslide on the North WestBorneo Margin revealed by 3D seismic data: Marine Geology, v. 246,p. 9–23.

GEISSLER, W.H., AND JOKAT, W., 2004, A geophysical study of the northernSvalbard margin: Geophysical Journal International, v. 158, p. 50–66.


HAFLIDASON, H., SEJRUP, H.P., BERSTAD, I.M., NYGÅRD, A., RICHTER, T., BRYN,P., LIEN, R., AND BERG, K., 2003, A weak layer feature on the northernStoregga Slide escarpment, in Mienert, J., and Weaver, P.P.E., eds.,European Margin Sediment Dynamics: Side-Scan Sonar and SeismicImages: Berlin, Springer-Verlag, p. 55–62.

HAFLIDASON, H., SEJRUP, H.P., NYGARD, A., MIENERT, J., BRYN, P., LIEN, R.,FORSBERG, C.F., BERG, K., AND MASSON, D., 2004, The Storegga slide:architecture, geometry and slide development: Marine Geology, v.213, p. 201–234.

HAMPTON, M.A., LEE, H.J., AND LOCAT, J., 1996, Submarine landslides:Reviews of Geophysics, v. 34, p. 33–59.

HARBITZ, C.B., 1992, Model simulations of tsunamis generated by theStoregga Slide: Marine Geology, v. 105, p. 1–21.

HARBITZ, C.B., AND PEDERSEN, G., 1992, Model theory and analytical solu-tions for large water waves due to landslides: University of Oslo,Oslo, Norway, Department of Mathematics, Preprint Series, v. 4.

HARBITZ, C.B., PARKER, G., ELVERHØI, A., MARR, J.G., MOHRIG, D., AND HARFF,P.A., 2003, Hydroplaning of subaqueous debris flows and glideblocks: analytical solutions and discussion: Journal of GeophysicalResearch, v. 108, no. 2349, doi:10.1029/2001JB001454.

HEBBELN, D., HENRICH, R., AND BAUMANN, K.-H., 1998, Paleoceanography ofthe Last Interglacial/Glacial cycle in the Polar North Atlantic: Quater-nary Science Reviews, v. 17, p. 125–153.

HEEZEN, B.C., AND EWING, M., 1952, Turbidity currents and submarineslumps, and the 1929 Grand Banks Earthquake: American Journal ofScience, v. 250, p. 849–573.

ILSTAD, T., DE BLASIO, F.V., ELVERHØI, A., HARBITZ, C.B., ENGVIK, L., LONGVA,O., AND MARR, J.G., 2004, On the frontal dynamics and morphology ofsubmarine debris flows: Marine Geology, v. 213, p. 481–497.

IMRAN, J., HARFF, P., AND PARKER, G., 2001, A numerical model of submarinedebris flow with graphical user interface: Computers & Geosciences,v. 27, p. 717–729.

JAKOBSSON, M., CHERKIS, N., WOODWARD, J., COAKLEY, B.J., AND MACNAB, R.,2000, A new grid of Arctic bathymetry: a significant resource forscientists and mapmakers: EOS, Transactions of the American Geo-physical Union, v. 81, p. 89, 93, 96.

JOKAT, W., 2005, The sedimentary structure of the Lomonosov Ridgebetween 88°N and 80°N: Geophysical Journal International, v. 163, p.698–726.

KATAOKA, T., AND TSUTAHARA, M., 2004, Transverse instability of surfacesolitary waves: Journal of Fluid Mechanics, v. 512, p. 211–221.

KOÇ, N., KLITGAARD-KRISTENSEN, D., HASLE, K., FORSBERG, C.F., AND SOLHEIM,A., 2002, Late glacial palaeoceanography of Hinlopen Strait, northernSvalbard: Polar Research, v. 21, p. 307–314.

KRASTEL, S., SCHMINCKE, H.-U., JACOBS, C.L., RIHM, R., BAS, T.P.L., AND ALIBES,B., 2001, Submarine landslides around the Canary Islands: Journal ofGeophysical Research, v. 106, p. 3977–3997.

KUVAAS, B., AND KRISTOFFERSEN, Y., 1996, Mass movement in glaciomarinesediments on the Barents Sea continental slope: Global and PlanetaryChange, v. 12, p. 287–307.

KVALSTAD, T.J., ANDRESEN, L., FORSBERG, C.F., BERG, K., BRYN, P., AND WANGEN,M., 2005, The Storegga slide: evaluation of triggering sources andslide mechanics: Marine and Petroleum Geology, v. 22, p. 245–256.

LABERG, J.S., AND VORREN, T.O., 1993, A late Pleistocene submarine slide onthe Bear Island Trough Mouth Fan: Geo-Marine Letters, v. 13, p. 227–234.

LABERG, J.S., AND VORREN, T.O., 2000, The Trænadjupet Slide, offshoreNorway—morphology, evacuation and triggering mechanisms: Ma-rine Geology, v. 171, p. 95–114.

LABERG, J.S., VORREN, T.O., DOWDESWELL, J.A., KENYON, N.H., AND TAYLOR, J.,2000, The Andøya Slide and the Andøya Canyon, north-easternNorwegian–Greenland Sea: Marine Geology, v. 162, p. 259–275.

LABERG, J.S., VORREN, T.O., MIENERT, J., HAFLIDASON, H., BRYN, P., AND LIEN,R., 2003, Preconditions leading to the Holocene Trænadjupet slideoffshore Norway, in Locat, J., and Mienert, J., eds., Submarine Mass

Movements and Their Consequences: Dordrecht, Kluwer AcademicPublishers, Advances in Natural and Technological Hazards Re-search, 19, p. 247–254.

LAMBECK, K., 1996, Limits on the areal extent of the Barents Sea ice sheet inLate Weichselian time: Global and Planetary Change, v. 12, p. 41–51.

LASTRAS, G., DE BLASIO, F.V., CANALS, M., AND ELVERHØI, A., 2005, Conceptualand numerical modeling of the Big’95 debris flow, western Mediterra-nean Sea: Journal of Sedimentary Research, v. 75, p. 784–797.

LEYNAUD, D., AND MIENERT, J., 2003, Slope stability assessment of theTrænadjupet slide area, offshore the mid-Norwegian margin, inLocat, J,. and Mienert, J., eds., Submarine Mass Movements and TheirConsequences: Dordrecht, Kluwer Academic Publishers, Advancesin Natural and Technological Hazards Research, 19, p. 255–266.

LINDBERG, B., LABERG, J.S., AND VORREN, T.O., 2004, The Nyk Slide—mor-phology, progression, and age of a partly buried submarine slideoffshore northern Norway: Marine Geology, v. 213, p. 277–289.

LOCAT, J., AND LEE, H.J., 2002, Submarine landslides: advances and chal-lenges: Canadian Geotechnical Journal, v. 39, p. 193–212.

LOCAT, J., AND MIENERT, J., 2003, Submarine Mass Movements and TheirConsequences: Dordrecht, Kluwer Academic Publishers, 540 p.

LØVHOLT, F., HARBITZ, C.B., AND HAUGEN, K.B., 2005, A parametric study oftsunamis generated by submarine slides in the Ormen Lange/Storeggaarea off western Norway: Marine and Petroleum Geology, v. 22, p.219–231.

MASSON, D.G., WATTS, A.B., GEE, M.J.R., URGELES, R., MITCHELL, N.C., LE BAS,T.P., AND CANALS, M., 2002, Slope failures on the flanks of the CanaryIslands: Earth-Science Reviews, v. 57, p. 1–35.

MASSON, D.G., HARBITZ, C.B., WYNN, R.B., PEDERSEN, G., AND LØVHOLT, F.,2006, Submarine landslides: processes, triggers and hazard predic-tion: Royal Society of London, Philosophical Transactions A, v. 364, p.2009–2039.

MCADOO, B.G., PRATSON, L.F., AND ORANGE, D.L., 2000, Submarine land-slide geomorphology, US continental slope: Marine Geology, v. 169,p. 103–136.

MCSAVENEY, M.J., GOFF, J.R., DARBY, D.J., GOLDSMITH, P., BARNETT, A.,ELLIOTT, S., AND NONGKAS, M., 2000, The 17 July 1998 tsunami, PapuaNew Guinea: Evidence and initial interpretation: Marine Geology, v.170, p. 81–92.

MEI, C.C., 1989, The Applied Dynamics of Ocean Surface Waves: Singapore,World Scientific Publications, Advanced Series on Ocean Engineer-ing, 760 p.

MIENERT, J., AND WEAVER, P.P.E., 2003, European Margin Sediment Dy-namics; Side-Scan Sonar and Seismic Images: Berlin, Springer-Verlag, 305 p.

MOHRIG, D., ELLIS, C., PARKER, G., WHIPPLE, K.X., AND HONDZO, M., 1998,Hydroplaning of subaquous debris flows: Geological Society ofAmerica, Bulletin, v. 110, p. 387–394.

MOORE, J.G., CLAGUE, D.A., HOLCOMB, R.T., LIPMAN, P.W., NORMARK, W.R.,AND TORRESAN, M.E., 1989, Prodigious submarine landslides on theHawaiian Ridge: Journal of Geophysical Research, v. 94, p. 17,465–17,484.

MOSHER, D.C., PIPER, D.J.W., CAMPBELL, D.C., AND JENNER, K.A., 2004, Near-surface geology and sediment-failure geohazards of the central ScotianSlope: American Association of Petroleum Geologists, Bulletin, v. 88,p. 703–723.

MULDER, T., AND COCHONAT, P., 1996, Classification of offshore massmovements: Journal of Sedimentary Reseach, v. 66, p. 43–57.

MYHRE, A.M., AND THIEDE, J., 1995, North-Atlantic–Arctic gateways, inMyhre, A.M., Thiede, J., and Firth, J.V., eds., Proceedings of the OceanDrilling Program, Initial Reports: College Station, Texas, Ocean Drill-ing Program, vol. 151, p. 5–26.

NYGÅRD, A., SEJRUP, H.P., HAFLIDASON, H., LEKENS, W.A.H., CLARK, C.D., AND

BIGG, G.R., 2007, Extreme sediment and ice discharge from marine-based ice streams: New evidence from the North Sea: Geology, v. 35,p. 395–398.


OTTESEN, D., DOWDESWELL, J.A., AND RISE, L., 2005, Submarine landformsand the reconstruction of fast-flowing ice streams within a largeQuaternary ice sheet: the 2500-km-long Norwegian–Svalbard margin(57°–80°N): Geological Society of America, Bulletin, v. 117, p. 1033–1050. doi: 10.1130/B25577.1.

PEREGRINE, D.H., 1966, Calculations of the development of an undularbore: Journal of Fluid Mechanics, v. 23, p. 321–330.

PEREGRINE, D.H., 1972, Equations for water waves and the approximationsbehind them, in Meyer, R., ed., Waves on Beaches: New York, Aca-demic Press, p. 95–122.

PIPER, D.J.W., AND MCCALL, C., 2003, A synthesis of the distribution ofsubmarine mass movements on the eastern Canadian margin, inLocat, J., and Mienert, J., eds., Submarine Mass Movements andTheir Consequences: Dordrecht, Kluwer Academic Publishers,Advances in Natural and Technological Hazards Research, 19, p.291–298.

PIPER, D.J.W., COCHONAT, P., AND MORRISON, M.L., 1999, The sequence ofevents around the epicentre of the 1929 Grand Banks earthquake:Initiation of debris flows and turbidity current inferred from sidescansonar: Sedimentology, v. 46, p. 79–97.

PIPER, D.J.W., MOSHER, D.C., GAULEY, B.-J., JENNER, K., AND CAMPBELL, D.C.,2003, The chronology and recurrence of submarine mass movementson the continental slope off southeastern Canada, in Locat, J., andMienert, J., eds., Submarine Mass Movements and Their Conse-quences: Dordrecht, Kluwer Academic Publishers, Advances in Natu-ral and Technological Hazards Research, 19, p. 299–306.

POSAMENTIER, H.W., AND KOLLA, V., 2003, Seismic geomorphology andstratigraphy of depositional elements in deep-water settings: Journalof Sedimentary Research, v. 73, p. 367–388.

PRIOR, D.B., AND COLEMAN, J.M., 1978, Disintegrating retrogressive land-slides on very-low-angle subaquous slopes, Mississippi Delta: Ma-rine Geotechnology, v. 3, p. 37–60.

PRIOR, D.B., BORNHOLD, B.D., COLEMAN, J.M., AND BRYANT, W.R., 1982,Morphology of a submarine slide, Kitimat Arm, British Columbia:Geology, v. 10, p. 588–592.

SLUBOWSKA, M.A., KOÇ, N., RASMUSSEN, T.L., AND KLITGAARD-KRISTENSEN, D.,2005, Changes in the flow of Atlantic water into the Arctic Ocean sincethe last deglaciation: Evidence from the northern Svalbard continen-tal margin, 80°N: Paleoceanography, v. 20, no. PA4014, doi:10.1029/2005PA001141.

SOLHEIM, A., BERG, K., FORSBERG, C.F., AND BRYN, P., 2005a, The StoreggaSlide complex: repetitive large scale sliding with similar cause anddevelopment: Marine and Petroleum Geology, v. 22, p. 97–107.

SOLHEIM, A., BRYN, P., SEJRUP, H.P., MIENERT, J., AND BERG, K., 2005b, OrmenLange—an integrated study for the safe development of a deep-water gas field within the Storegga Slide Complex, NE Atlanticcontinental margin; executive summary: Marine and PetroleumGeology, v. 22, p. 1–9.

SVENDSEN, J.I., ALEXANDERSON, H., ASTAKHOV, V.I., DEMIDOV, I., DOWDESWELL,J.A., FUNDER, S., GATAULLIN, V., HENRIKSEN, M., HJORT, C., HOUMARK-NIELSEN, M., HUBBERTEN, H.W., INGOLFSSON, O., JAKOBSSON, M., KJÆR,K.H., LARSEN, E., LOKRANTZ, H., LUNNKA, J.P., LYSÅ, A., MANGERUD, J.,MATIOUCHKOV, A., MURRAY, A., MÖLLER, P., NIESSEN, F., NIKOLSKAYA, O.,POLYAK, L., SAARNISTO, M., SIEGERT, C., SIEGERT, M.J., SPIELHAGEN, R.F.,AND STEIN, R., 2004, Late Quaternary ice sheet history of northernEurasia: Quaternary Science Reviews, v. 23, p. 1229–1271.

TAPPIN, D.R., WATTS, P., MCMURTRY, G.M., LAFOY, Y., AND MATSUMOTO, T.,2001, The Sissano, Papua New Guinea tsunami of July 1998—offshoreevidence on the source mechanism: Marine Geology, v. 175, p. 1–23.

VANNESTE, M., MIENERT, J., AND BÜNZ, S., 2006, The Hinlopen Slide: A giant,submarine slope failure on the northern Svalbard margin, ArcticOcean: Earth and Planetary Science Letters, v. 245, p. 373–388.

VORREN, T.O., 2003, Subaquatic Landsystems: Continental Margins, inEvans, D.J.A., ed., Glacial Landsystems: London, Arnold Publishers,p. 289–312.

VORREN, T.O., AND LABERG, J.S., 1997, Trough Mouth Fans—Paleoclimateand ice-sheet monitors: Quaternary Science Reviews, v. 16, p. 865–881.

VORREN, T.O., LABERG, J.S., BLAUME, F., DOWDESWELL, J.A., KENYON, N.H.,MIENERT, J., RUMOHR, J., AND WERNER, F., 1998, The Norwegian–Greenland Sea continental margins: Morphology and Late Quater-nary sedimentary processes and environment: Quaternary ScienceReviews, v. 17, p. 273–302.

WARD, S.N., 2001, Landslide tsunami: Journal of Geophysical Research, v.106, p. 11,201–11,215.

WESSEL, P., AND SMITH, W.H.F., 1998, Improved version of the genericmapping tools released: EOS, Transactions of the American Geo-physical Union, v. 79, p. 579.

WINKELMANN, D., JOKAT, W., NIESSEN, F., STEIN, R., AND WINKLER, A., 2006,Age and extent of the Yermak Slide north of Spitsbergen, ArcticOcean: Geochemistry, Geophysics, Geosystems, v. 7, no. Q06007,doi:10.1029/2005GC001130.

WINKELMANN, D., AND STEIN, R., 2007, Triggering of the Hinlopen/Yermakmegaslide in relation to paleoceanography and climate history of thecontinental margin of Spitsbergen: Geochemistry, Geophysics,Geosystems, v. 8, no. Q06018, doi:10.1029/2006GC001485.

HINLOPEN–YERMAK LANDSLIDE, ARCTIC …folk.uio.no/anelverh/Papers/Vaneste_etal_SP95_inpress.pdf ·...

Documents

Transcript of HINLOPEN–YERMAK LANDSLIDE, ARCTIC …folk.uio.no/anelverh/Papers/Vaneste_etal_SP95_inpress.pdf ·...