• No results found

Some giant submarine landslides do not produce large tsunamis

N/A
N/A
Protected

Academic year: 2022

Share "Some giant submarine landslides do not produce large tsunamis"

Copied!
10
0
0

Laster.... (Se fulltekst nå)

Fulltekst

(1)

Some g iant submar ine lands l ides do not produce large tsunam is

Finn Løvholt1 , Stein Bondevik2,3 ,Jan Sverre Laberg3 ,Jihwan Kim1,4 , and Noel Boylan5

1Norwegian GeotechnicalInstitute, Oslo, Norway,2Western Norway University of Applied Sciences,Institute of Natural Science, Sogndal, Norway,3Department of Geosciences, University of Tromsø-the Arctic University of Norway, Tromsø, Norway,4Department of Mathematics, University of Oslo, Oslo, Norway,5Norwegian GeotechnicalInstitute, Perth, Western Australia, Australia

Abstract

Landslides arethe second mostimportant cause oftsunamis after earthquakes, andtheir potentialfor generatinglargetsunamis depend ontheslide process. Amongthe world’slargestsubmarine landslidesisthe Storegga Slidethat generated alargetsunami over an ocean-wide scale, while notraces of atsunami generatedfromthe similar and nearby Trænadjupet Slide have beenfound. Previous models for suchlandslidetsunamis have not been ableto capturethe complexity ofthelandslide processes and are at odds with geotechnical and geomorphological datathatrevealretrogressivelandslide development. Thetsunami generationfromthese massive events are here modeled with new methodsthatincorporate complexretrogressive slide motion. We showthatthetsunamigenic strengthis closelyrelatedtothe retrogressive development and explain,forthe firsttime, why similar giantlandslides can produce very differenttsunamis, sometimes smallerthan anticipated. Becausethese slide mechanisms are commonfor submarinelandslides, modeling proceduresfor dealing withtheir associatedtsunamisshould berevised.

P la in Language Summary

We here study how some ofthelargestlandslidesthat are observed onthe planet may generate destructivetsunamis. The cases ofinterest aretwolandslides, named Storegga and Trænadjupet, which occurred about 8000 and 4000 ago belowthe sea offshore Norway. Their surface areas aresimilartothe area of Scotland. Surprisingly, only one ofthesetwo enormouslandslides, Storegga, was knownto generate a destructivetsunami. Here we use computersimulationsto explain whythe other landslide, Trænadjupet, did not produce a alarge wave. The computersimulationsrevealedthat differences inthe dynamics ofthetwolandslidesledto very differenttsunamis. This has broaderimplications on our understanding of how massive submarinelandslide generatetsunamis and showsthat we need better modelsto explain why apparentlysimilarlandslides generatetsunamis of very differentsize.

1 . Introduct ion

Landslides arethesecond mostimportant cause oftsunamis after earthquakes, buttheir potentialto gener- ate ocean-wide catastrophic tsunamisis not well understood andlimited to the study of a handful of well- documented events[Masson et al., 2006;Tappin, 2010;Harbitz et al., 2014]. The destruction ofthe Fukushima Daishi nuclearreactor[Synolakis and Kanoglu, 2015]in 2011setstheir hazardtoward criticalfacilitiessharply intofocus. Thetype oflandslidefailure and flow process, as well as velocity, acceleration, and volume deter- mine the tsunamigenic strength[Løvholt et al., 2015]. Moreover,landslide tsunamis generally require more sophisticatedsource ortsunami modelsthanthosecommonly usedfor earthquaketsunamis[Ma et al., 2015; Yavari-Ramshe and Ataie-Ashtiani, 2016;Grilli et al., 2017].In this respect, previous models appliedfor giant submarinelandslidetsunamis havefallen short[Bondevik et al., 2005;Harbitz, 1992;Hill et al., 2014], asthey oversimplifythelandslide dynamics and do notincorporatethe complexfailure process.

While some giant landslides produce large tsunamis, others generate ones that are unexpectedly small. The reason for this differenceis due to the slide dynamics, but sofar models have been unable to capture this effectively. Tsunami generationfrom these giant submarinelandslidesis more complex than thosefor other landslide types such as rotational slumps [Lynett et al., 2003;Okal and Synolakis, 2004;Tappin et al., 2008], for which the generating mechanism is mainly controlled by the initial impulsive landslide motion [Løvholt et al., 2015]. We here use new models to study two major submarinelandslides on the Norwegian

RESEARCH LETTER

10.1002/2017GL074062

Key Points:

• New modelsfor complexlandslide emplacement mechanism shed new light on generation mechanisms

• Retrogressivelandslide models explain whytwo giantlandslides produce very differenttsunamis

• Slowretrogression explainsthat one ofthe worldslargestlandslides produced asmalltsunami

SupportingInformation:

• Figure S1

• Figure S2

• Figure S2

• Figure S4

• Figure S5

• Table S1

• SupportingInformation S1

Correspondenceto: F. Løvholt, finn.lovholt@ngi.no

Citation:

Løvholt, F., S. Bondevik, J. S. Laberg, J. Kim, and N. Boylan (2017), Some giant submarinelandslides do not producelarge tsunamis,Geophys. Res. Lett.,44, doi:10.1002/2017GL074062.

Received 5 MAY 2017 Accepted 1 AUG 2017

Accepted article online 7 AUG 2017

©2017. The Authors.

Thisis an open access article underthe terms ofthe Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distributionin any medium, provided the original workis properly cited,the useis non-commercial and no modifications or adaptations are made.

(2)

continental margin,the 8150 BP Storegga andthe∼4500 BP Trænadjupet Slides[Bryn et al., 2005;Haflidason et al., 2004;Laberg and Vorren, 2000;Laberg et al., 2002a;Haflidason et al., 2007]. Whilethe Storegga Slideisthe only known prehistoricalsubmarinelandslide with compelling geological evidencefor an ocean-wide catas- trophictsunami[Bondevik et al., 2005, 1997;Smith et al., 2004],such depositsforthe giant Trænadjupet Slide are sofar absent. To this end, these twolandslides comprise unique examplesfor testing new models and explainingfundamental differencesintsunami generation. First, because oftheir different onshoretsunami footprints and also becausetheirslide deposits are amongthe beststudied worldwide, which allow ustolink thetsunamigenesistolandslide observations.

Examplesfrom differentcontinental margins acrossthe world,including North America and Northern Europe [Canals et al., 2004;Haflidason et al., 2004;Kvalstad et al., 2005;Masson et al., 2006;Lee, 2009;Twichell et al., 2009; Baeten et al., 2014]showthatlargesubmarinelandslides often evolvethroughretrogressivefailure[Kvalstad et al., 2005;Laberg et al., 2002b;Gauer et al., 2005], a processstartingsomewhere downslopefromthe headwall area and retreating backward upslope. This retrogressive processis commonin clay-rich submarine slopes andistherefore a dominant mechanismin associatedlandslides worldwide.Inthis process,landslides often transforminto a debris flow and can also developturbidity currentsfarther downslope[Elverhøi et al., 2005]. These processes affect howthetsunamiis generated, but as notedthey have sofar not been studied using realistic models.

Here we havetestedtwo new alternativelandslide models, aremolding viscoplastic debris flow model, and a retrogressive block model (section 2). These arein turn coupled to tsunami generation and propagation models.Insection 3, we present modelresults usingthe new modelsforthe 8150 BP Storegga andthe∼4500 BP Trænadjupet Slides andcomparethesimulations with fieldinvestigations.Insection 4, we discussthe main differences betweenthesetwo events, andthe broaderimplications ofthestudy, namely, howsome massive landslide events may producetsunamis of only modestsize.

2 . Methods

2.1. Retrogressive Block Model

Theretrogressive block model[Løvholt et al., 2016]is based onthe energy balance approach ofKvalstad et al. [2005] and assumes that thelandslide fails as a continuous sequence of blocks. Each block consists of two triangular subblocks that stayintact during both thefailure and slide motion, and a central parallelogram- shaped bodysubjecttosimpleshear duringthefailure process(see Figure 1).In addition,the modelincludes energy dissipation due to plastic shear work along the seabedinterface. Theinitial block geometryis pre- defined with aninitial variablethicknessH0and a final height ofHf=H0−ΔH. The modelincludes variable topography, variableslidethickness, and hydrodynamic dragterms. Further, a prescribedlandslidetrajectory and aconstantslide width are assumed. Theslidethicknessistapered azimuthally using a quadraticfunction. The modeltakesinto accountthe energy balance betweenthe kinetic energyEk(includingthe block massm andfrontal block speedu), potential energyEp,theinstantaneous plasticinternal(Eint) and basal plasticfric- tion(Ebase) energies of the slide, as well as work caused by the hydrodynamic resistanceforces due to drag (Wdrag) and added water mass(Cm⋅m):

Ek=12(m⋅(1+Cm))u2=Ep−Eint−Ebase−Wdrag (1) In this model, the landslide geometry of deformation is predetermined, with an initial block thickness large degreecontrolled bythesensitivity ofthe basallayerSbase,thisistherefore afree parameter definingthe differentretrogressive Trænadjupet scenarios. A high basal sensitivity valueimplies alow basalfriction, and vice versa—alow basalsensitivity valueimplies a higher basalfriction. Thesimulated block motionfollows a prescribed path. Theretrogressive block model makes use of aseries of additional parametersthat determine thelandslide geometry and soil properties using detailed analytic expressions forEp,Eint,Ebase, andWdrag. These parameters are mainly derivedfromKvalstad et al.[2005][see alsoLøvholt et al., 2016] and arelistedin thesupportinginformation.

2.2. Viscoplastic Debris Flow ModelIncluding Remolding

The viscoplastic debris flow model(BingClaw)is a depth-averaged Herschel-Bulkleyrheological model modi- fiedfromHuang and García[1997] andImran et al.[2001],formulatedin an Euleriansystem.Itsolvesthe mass and momentum balancefor a plug(subscriptp) andshearlayer(subscripts),respectively:

(3)

Figure 1.(top) Principletwo-dimensional sketch ofthefailure pattern using aretrogressive block model(triangles and parallelograms) and aremolding debris flow model superimposedin gray. Theretrogressive slide blocksfail one by one, blocks 1 and 2 are moving paralleltothe slope, blocks 3 and 4 are atrest. The slide materialis mobilized morerapidly inthe debris flow model, and several“blocks” mayfail atthe sametime.(bottomleft) Evolution ofthe Trænadjupet Slideintheremolding debris flow model.(bottomright) Evolution ofthe Trænadjupet Slide usingtheretrogressive block model.

h⃗u)

=0, (2)

and

(

)

(h⃗u) +((

⃗ up⋅∇)

⃗ up +⃗up(

∇⋅(

⃗ up)

+ (

)

gh∇(h+b)

(4)

whereh=hs+hp,isthetotallandslidethickness overtheshear and pluglayers, andup,vp,u, andvare plug layer and average velocities in orthogonal directions. The speed of the plug layer and the mean speed is relatedthroughthefollowing condition:

h |⃗up|, (5)

water depth. Thissystem of equationsissolvedinthreesteps: first, bycheckingthatthe Earth pressure exceeds theshearstrength,then bysolvingfortheterms withoutfriction numerically using a finite volumeformulation arefunctions ofthe Herschel-Buckley parametersn[see, e.g.,Huang and García, 1997].Inthe present analysis, problem[Mangeney et al., 2000]is providedinthesupportinginformation

The debris flow modelincludes remolding.In this way, the material strengthis reduced when subjected to shear stresses. TheremoldingrateΓis afree parameter, which allowslandslide dynamicsto mimic complex

(4)

Figure 2.The simulated Storegga and Trænadjupet Slides. Storeggais simulated withthe debris flow model; shown are(first column) snapshots atthefailure, after 1 h, and after 2 h. Snapshots ofthe Trænadjupet Slide attheinitiation offailure, after 1 h, and after 2 h,(fourth column) usingthe debris flow model and (third column) usingtheretrogressive block model(scenario 2).(second column) The finalrun-out. The observedrun-out ofthelandslide debris are shown as magenta contours.

multistaged failure types such as a retrogressive failure or a top-down failure in a simplified way. When a retrogressive-likefailure developmentisinduced,thefailure mechanismis differentfromthe oneintheret- rogressive block model, distributing the failure across alarger portion of the slide mass at the same time, accordingtothe equation

Theterm⃗fscontains quadratic pressure drag andskinfrictionterms, with coefficientsCP=1.0andCF=0.01, respectively. Theinitial geometries usedforthis modelfor Storegga and Trænadjupet areshownin Figure 2. 2.3. Tsunami Models

Tsunami propagation was simulated using depth-averaged shallow water-type models. Initial simulations (see supporting information) demonstrated that dispersive effects had a limited effect on the tsunami propagation, an observation alsosupported by previous work carried out on voluminouslongrun-outland- slides[Haugen et al., 2005;Løvholt et al., 2005, 2015]. Forthetsunamisimulations employingthe viscoplastic landslide model asthetsunamisource, we used a new version oftheGeoClawmodel[Kimetal., 2017]in non- linearshallow water mode. Forthetsunami generationfromtheretrogressivelandslide motion,the numerical GloBoussmodel[Løvholt et al., 2008] wasruninlinear hydrostatic mode.

The waves aresolongthattheirrun-upis effectivelycaptured bythe doubling ofthetsunami wave amplitude duetocoastalreflection. This assumption was usedin previousstudies ofBondevik et al.[2005],Harbitz[1992], andHill et al.[2014], the same assumptionis used here. Thatis, we assume that the tsunami elevations at offshorecontrol points placedclosetotheshorelinetoincludethe amplitudeincrease dueto wavereflection roughly represents the run-up height. The control points are placedin water depths typically rangingfrom

(5)

50to 100 m, occasionally at asshallow as 10–20 m. Weshowthatthis assumptionisreasonable byconducting localinundationsimulationsfortwo paleotsunamisites(see detailsinthesupportinginformation, andrelated data and methodsinLøvholt et al.[2010] andTitovetal.[2013]). Theseinundation simulations alsoindicate thetypical variability oftherun-upthat can be expected.

For the Storegga simulations, we used the paleobathymetry ofHill et al.[2014], while ETOPO 1 grids were usedforthe Trænadjupet simulations. Forthe GloBousssimulationsthe grids wererefinedto aresolution of 7.5′′×15′′before being employedin the tsunami simulations. To correct for nonhydrostatic effectsin the tsunami generation duetotheshort-scale deformationsfromsmallretrogressive blocklengths, afull potential filter was employed[Kajiura, 1963;Løvholt et al., 2015].

3 . Resu lts

3.1. New Simulation ofthe Storegga Slide Tsunami

The Storegga Slide offshore western Norwayis amongthe world’slargestlandslides, with an estimatedslide volume of 2400–3200 km3[Haflidason et al., 2004] and aslide area of 95000 km2. Thelandslide debrisrun-out reached about 300 km [De Blasio et al., 2005] on a gentle 1–2∘slope, accommodated by the presence of overpressurized baselayering[Bryn et al., 2005;Locat et al., 2014]. Geomorphological and geotechnical data concludethatit evolvedretrogressivelyin asingle major event[Kvalstad et al., 2005]. Theslide deposits below thesteepslopesintheslide escarpmentshow evidence of extensivecrushing and disintegration oftheretro- gressive blocks[Bryn et al., 2005]. Further afield, Storegga produced muddyturbiditycurrentsreaching about 800 km,indicating high slide velocities[Haflidason et al., 2004]. Tsunami depositsfrom Storegga arefound alongthe Norwegian coastline,inthe FaeroeIslands, Shetland, Scotland, andrecentlyin Denmark[Bondevik et al., 2005;Smith et al., 2004;Fruergaard et al., 2015].

Thesimulation ofthe Storegga Slide usingthe debris flow model and atotalslide volume of∼3000 km3agrees well with observedrun-out(see Figure 2). Particularly noticeableisthe match ofthe shape ofthelandslide debris with observations oflandslidelobes[Haflidason et al., 2004], whichis only possible duetothe model’s ability to takeinto account terrain deflection and shear wetting[De Blasio et al., 2005]. Figure 3 shows that simulatedtsunami heightscompare well with field observations ofsedimentrun-upin Norway, FaeroeIslands, and offshore Scotland and Denmark, but underestimates the∼20 m run-up observationsin Shetland, still clearlyimproving all previoustsunami hindcasts[Bondevik et al., 2005;Harbitz, 1992;Hill et al., 2014]. To achieve the model agreement for the Storegga Slide tsunami, applied values of the ultimate andinitial yieldstrengths andshear wettingrates areinthelower end of both previously assumed and measured values. The slide front accelerated rapidly to a speed of 20 m/s, reaching a maximum frontal speed of 30–35 m/s about 1 h afterfailure, comingtorest after 5 h. Oursimulatedslidespeeds comply both with previously pre- scribed block slide speedsintsunami analysis[Bondevik et al., 2005;Harbitz, 1992;Hill et al., 2014], as well as the maximum particle velocity of 60 m/s foundin previous simplified run-out simulations of the Storegga Slide[De Blasio et al., 2005].

Simulations using the retrogressive block model as the tsunami source also provided a reasonably good agreement withthe Storeggatsunami data, butthis model producedlarger offsetsthanthe debris flow model betweensimulations and paleotsunami observations(results notshown). The maintsunami generationtook place after the retrogressive process had ended. At this stage, thelandslide continuedits motion as a solid block.Inconclusion,thesimulationsshowthatthe maintsunami generation ofthe Storegga Slidetook place afterthe majority ofthelandslide was mobilized,following arapidretrogressive process and a high degree ofremolding.

3.2. A Small Tsunami Fromthe Giant Trænadjupet Slide?

The∼4500 BP Trænadjupet Slide[Laberg and Vorren, 2000]involved a smaller slide velocity than Storegga, and arun-out distance of 100–150 km. Terraces nearthe upper headwall[seeLaberg et al., 2002b, Figure 11] indicate that the eventinvolved retrogressive failure, and mostlikelyit also failedin a single major event. Dueto mixing withthe older Nyklandslide deposits[Lindberg et al., 2004], precise estimates ofits volume do not exist butis smallerthan 1000 km3, and probably closerto 500 km3. The slide morphologyindicatesthat thelandslide depositslocated onthe abyssal plain are more blockythan corresponding observations ofthe Storegga Slide [Laberg et al., 2006]. The upper part of thelandslide also reveals traces ofintact blocks that hintthatthe slide progressionretarded and stopped whentheretrogressive slide development progressed

(6)

Figure 3.Maximum water elevationforthe Storegga Slidetsunami, simulated usingthe debris flowlandslide source. Blue-purple bars showthe simulated elevations closetothe field sites, black bars showthe mean observation heights of sedimentrun-up[Smith et al., 2004;Bondevik et al., 2005;Romundset and Bondevik, 2011;Fruergaard et al., 2015].

to the upper headwall. Finally, we note that Trænadjupetis not associated with extensive turbidity current deposits like those of Storegga. Together, these observations indicate smaller slide velocities than for the Storegga Slide.

Wesearchedfor Trænadjupettsunami depositsin coastallakesin northern Norway(Figure 4 andsupporting information Table S1) but did not findconvincingcandidates. However,some ofthelake basins, marked with a green trianglein Figure 4, have distinct sandlaminae(s)in brackish sediments that weinterpret to repre- sent a storm surge or a small tsunami. The sandis traced toward the outlet or seaside of the basin, so they must represent a current entering the basins from the sea. But we have not seen erosion or characteristic tsunamifacies withrip-up clasts andredeposition, asistypicalfor Storeggatsunami deposits[Bondevik et al., 1997]. Wethinkthelaminaea couldrepresent a smalltsunami(lessthan 3 mrun-up)that barely flowedinto thelakes, or a major storm surge. A morein depth discussion of the Trænadjupet depositsis givenin the supportinginformation.

The same landslide parameters used for Storegga were employed for the Trænadjupet debris flow simulation, and the simulatedlandslide thickness and run-out distance agree well with observedlandslide deposits[Laberg and Vorren, 2000]. The simulated meanfrontal velocity usingthe debris flow modelis close to 30 m/s, whichiscomparabletothat ofthe Storegga Slide debris flowsimulation. Thesnapshots oftheland- slidein Figure 2showlargeinitial motionsinthe upperslopethatisreminiscent of atop-downfailure. Figure 4 showsthatthe simulatedtsunamiinduced bythe debris flowlandslide source produced wavesthat clearly exceed 2–3 m at manylocations. Thelack oftsunami depositsshowthatthisscenariois mostlikely unrealistic forthe Trænadjupet Slide based upon ourcurrent understanding ofthe geological evidence. Sufficientlysmall waves,in agreement with the paucity of tsunami deposits, could only be achievedfor scenarios where the landslide could notreachthe observedrun-out.

(7)

Figure 4.Maximum water elevationforthe Trænadjupet Slidetsunamifrom scenario 2 oftheretrogressivelandslide model. Blue-purple bars showthe simulated near-shore water elevations usingthe debris flow model asthetsunami source. Pink bars showthe simulated elevations usingtheretrogressivelandslide source(scenarios 1–3,from topto bottom). Lake basins with atriangle have unique sandlaminae(s) or a zone of coarser sand grains nearthe brackish/lacustrine boundary. Number nexttotriangleis agein ka BP ofthese possibletsunami or storm surge deposits(see supportinginformation Table S1for moreinformation). Thelocations of Vikna and Røst have not beeninvestigated, butrepresent areasthat may be particularly exposedto a Trænadjupettsunami.

Three scenarios were modeled coupling the retrogressivelandslide model to the tsunami model. The first scenario producedthe smallesttsunami, with heightslessthan 0.3 m atthe coast. Wetunedthe baselayer sensitivity in this scenario (S=11.75) to terminate the slide motion when the slide retreated to the upper headwall, mobilizingthe entirelandslide mass(550 km3). Thelandslide first acceleratesrapidlytoreachthe maximum velocity of 15 m/s and then decelerates to rest. The slide velocity is therefore low close to the headwall, wherethe water depth onthe continental slopeis shallowest. Becausethetsunami generationis most efficientforlarge slide speeds at shallow depth,this processisinefficient at producing atsunami. The water elevations arelessthan1moffshore, andthe heightisreducedtowardland dueto geometricspread- ing. When the tsunamireaches the potential paleotsunami sites(see Figure 4) the water elevationis below 0.3 m. This scenario represents the weakest possible tsunami source for the Trænadjupet Slide assuming a retrogressivesingle-eventfailure. However,thislandslidescenarioisin agreement withthelandsliderun-out and evidence atthe paleotsunamiinvestigationsites deposits.

Inthe second andthirdretrogressive Trænadjupet scenarios we employ slightlylarger material sensitivities (S=14–16), whichleadstoincreased maximum velocities of 17–26 m/s,respectively.Inthesescenariosthe

(8)

landslide continuesits motion and slides as a single solid block aftertheretrogressive process hasretreated tothe headwall,in atop-down motion. Scenarios 2 and 3thereforeresultin higher waves. The wavesreach morethan 5 minthetsunami generation area as exemplifiedforscenario 2in Figure 4. Neartheshore, water elevations have afairlylargerange of 1–3 m nearthe fieldsites. Whilethescenario 2simulation givestsunami heightsthat areinvariablysmallerthan2mattheselocations,scenario 3 gives what weconsiderto be exces- sive tsunami heights that we should have discovered at the field sites; thus, we think scenario 3is unlikely. Duetothe NW orientation oftheslide andfocusing andrefraction ofthetsunami, we find higher wavestoward locations such as Vikna (1.9–3.9 m) and Røst (3.7–7.4 m), sites that are not yetinvestigated for geological evidence. This highlights the point that we cannot rule out that the Trænadjupet event produced a mod- erately sized tsunami. Comparing all three scenarios, it is clear that the scenarios 2 and 3, involving a top-down block motionfollowing aretrogressivefailure, are moretsunamigenicthanthe pureretrogressive slide(scenario 1).

4 . D iscuss ion and Conc lus ions

The main reason for the difference in tsunami generation for the Storegga and Trænadjupet Slides lies in their different dynamics. The coupled tsunami-debris flow model that produces excellent results for the Storegga Slide, in agreement with both the slide and tsunami deposits, does not work for Trænadjupet, despitethefactthatthe observedrun-out of bothslides are well matchedinthesimulations. This meansthat in order to quantify the size of the tsunami,itis not sufficient to reproduce the observedlandslide run-out in the model. Using the retrogressive block model for the Trænadjupet Slide, we obtainlower slide veloc- ities compatible with thelandslide depositslike the thin turbiditesin the adjacent deep basins andintact slide blocksin deep water, and the currentlack of geological evidencefor tsunamiinundation.In addition, otherfactors have played aroleresultingin atsunami orthelackthereof: The Storegga Slide volumeis about 5timeslargerthan Trænadjupet, and alarger portion ofitfailedin shallower water while moving at greater speed. The distancefromthe Trænadjupet headwalltoshorelineis also about 3timeslargerthanfor Storegga, meaningthatthe wave was attenuated more beforeitreachedthe shoreline. Moreover,the wave directivity due to Trænadjupet is favorable toward locations that we have not yet investigated for tsunami deposits. As aconsequence,the newsimulationsshowthat afew distantislands alongthe Norwegiancoast might have arecord ofthis event hiddenintheir geological archives.

Inthe present paper, weshow how differencesinfailure mechanisms of giantlandslides produce very different tsunamis. Wefurther demonstrate that the retrogressive processitselfisinefficientin generating tsunamis. This implies that even giant landslides, such as Trænadjupet, may produce tsunamis of only modest size. However, whenthese voluminouslandslides arerapidlytransformedinto afast moving debris flow,they are capable of producing a large tsunami like that of Storegga. Many giant landslides across the world’s con- tinental margins display retrogressive development, and our examples show that some of them may not produce widespread tsunamis, despite having enormous volumes. While otherlandslides tend to produce largelocal waves[Okal and Synolakis, 2004],the newresultsrevealed hereinshowthatsomelandslides, even when exhibiting giant volumes, may belesstsunamigenicthan previouslythought. The presentstudyshows that their diversity and uncertainty, caused by their very different dynamics and emplacement needs to be reflectedinfuturetsunami hazard assessments.

References

Baeten, N. J., J. S. Laberg, M. Vanneste, C. F. Forsberg, T. J. Kvalstad, M. Forwick, T. O. Vorren, and H. Haflidason(2014), Origin of shallow submarine mass movements andtheir glide planes—Sedimentological and geotechnical analysesfromthe continental slope off northern Norway,J. Geophys. Res. Earth Surf.,119, 2335–2360, doi:10.1002/2013JF003068.

Bondevik, S.,J.-I. Svendsen, andJ. Mangerud(1997), Tsunami sedimentaryfacies deposited bythe Storeggatsunamiin shallow marine basins and coastallakes, Western Norway,Sedimentology,44, 1115–1131.

Bondevik, S., F. Løvholt, C. Harbitz, J. Mangerud, A. Dawson, and J. Svendsen(2005), The Storegga slide tsunami comparing field observations with numericalsimulations,Mar. Pet. Geol.,22, 195–208.

Bryn, P., K. Berg, C. Forsberg, A. Solheim, and T. Kvalstad(2005), Explainingthe Storeggaslide,Mar. Pet. Geol.,22, 11–19.

Canals, M., et al.(2004), Slopefailure dynamics andimpactsfromseafloor andshallowsub-seafloor geophysical data: Casestudiesfromthe COSTA project,Mar. Geol.,213, 9–72.

De Blasio, F., A. Elverhøi, D.Issler, C. Harbitz, P. Bryn, and R. Lien(2005), Onthe dynamics ofsubaqueous clayrich gravity mass flows—The giant Storegga Slide, Norway,Mar. Pet.Geol.,22, 179–186.

Elverhøi, A., D.Issler, F. De Blasio, T.Ilstad, C. Harbitz, and P. Gauer(2005), Emerginginsightsintothe dynamics of submarine debris flows, Nat. Hazards Earth Syst. Sci.,5, 633–648.

Acknowledgments

The work has been funded by the Research Council of Norway project Tsunamisinduced bylargelandslides (NFR 231252/F20). We thank Carl Harbitz, Tore Kvalstad, Stefano Lorito, and Carl Fredrik Forsberg, reviewer James Goff, and two anonymous reviewers for their valuable comments that greatly improved the quality of the paper. The source code for viscoplastic debris flow models used to simulate thelandslide dynamicsis avaible at https://github.com/jhkim2/GRL_BingCLAW. The additional models applied in this study are found inKvalstad et al.[2005],Løvholtetal.[2008, 2010], Titov et al.[2013],Løvholt et al.[2016], andKim et al.[2017].

(9)

Fruergaard, M., S. Piasecki, P. Johannessen, N. Noe-Nygaard, T. Andersen, M. Pejrup, and L. Nielsen(2015), Tsunami propagation over a wide, shallow continental shelf caused bythe Storegga Slide, southeastern North Sea, Denmark,Geology,43(12), 1047–1050, doi:10.1130/G37151.1.

Gauer, P., T. Kvalstad, C. Forsberg, P. Bryn, and K. Berg(2005), Thelast phase ofthe Storegga Slide: Simulation ofretrogressiveslide dynamics and comparison withslide-scar morphology,Mar. Pet. Geol.,22, 171–178.

Grilli, S. T., M. Shelby, O. Kimmoun, G. Dupont, D. Nicolsky, G. Ma, J. T. Kirby, and F. Shi(2017), Numerical simulation of subaerial and submarinelandslide generatedtsunami waves—Recent advances andfuture challenges,Nat. Hazards,86(1), 353–391. Haflidason, H., H. P. Sejrup, A. Nygård, J. Mienert, P. Bryn, R. Lien, C. F. Forsberg, K. Berg, and D. Masson(2004), The Storegga Slide:

Architecture, geometry andslide development,Mar. Geol.,213(1–4), 201–234, doi:10.1016/j.margeo.2004.10.007. COSTA- Continental Slope Stability.

Haflidason, H., M. de Alvaro, A. Nygard,J. Sejrup, and H. P. Laberg(2007), Holocene sedimentary processesinthe Andøya Canyon system, north Norway,Mar. Geol.,246, 86–104.

Harbitz, C. B.(1992), Modelsimulations oftsunamis generated bythe Storegga Slides,Mar. Geol.,105, 1–21.

Harbitz, C., F. Løvholt, and H. Bungum(2014), Submarinelandslidetsunamis: How extreme and howlikely?,Nat. Hazards,72(3), 1341–1374, doi:10.1007/s11069-013-0681-3.

Haugen, K. B., F. Løvholt, and C. B. Harbitz(2005), Fundamental mechanismsfortsunami generation bysubmarine mass flowsinidealised geometries,Mar. Pet. Geol.,22, 209–217.

Hill,J., G. Collins, A. Avdis, S. Cramer, and M. Piggot(2014), How does multiscale modelling andinclusion ofrealistic palaeobathymetry affect numericalsimulation ofthe Storegga Slidetsunami?,Ocean Modell.,83, 11–25, doi:10.1016/j.ocemod.2014.08.007. Huang, X., and M. H. García (1997), A perturbation solution for Bingham-plastic mudflows,J. Hydraul. Eng.,123(11), 986–994,

doi:10.1061/(ASCE)0733-9429(1997)123:11(986).

Imran,J., G. Parker,J. Locat, and H. Lee(2001), 1-D numerical model of muddysubaqueous andsubaerial debris flows,J. Hydraul. Eng.,127, 959–968.

Kajiura, K.(1963), Theleading wave of atsunami,Bull. Earthquake Res.Inst. Univ. Tokyo,41, 535–571.

Kim, J., G. Pedersen, F. Løvholt, and R. Leveque(2017), A Boussinesqtype extension ofthe geoclaw model—A study of wave breaking phenomena applying dispersivelong wave models,Coastal Eng.,122, 75–86.

Kvalstad, T., L. Andresen, C. Forsberg, K. Berg, P. Bryn, and M. Wangen(2005), The Storegga Slide: Evaluation oftriggeringsources andslide mechanics,Mar. Pet. Geol.,22, 245–256.

Laberg,J., and T. Vorren(2000), The Trænadjupet Slide, offshore Norway—Morphology, evacuation andtriggering mechanisms,Mar. Geol., 171, 95–114.

Laberg,J., T. Vorren,J. Miener, P. Bryn, and R. Lien(2002a), The Trænadjupet Slide: Alargeslopefailure affectingthe continental margin of Norway 4000 years ago,Geo Mar. Lett.,22, 19–24.

Laberg, J., T. Vorren, J. Miener, D. Evans, B. Lindberg, N. Ottesen, D. adn Kenyon, and S. Henriksen (2002b), Late Quaternary palaeoenvironment and chronologyinthe Trænadjupet Slide area offshore Norway,Mar. Geol.,188, 35–60.

Laberg, J., T. Vorren, N. Kenyon, and M.Ivanov(2006), Frequency andtriggering mechanisms of submarinelandslides ofthe North Norwegian continental margin,Norw.J. Geol.,86, 155–161.

Lee, H.(2009), Timing of occurrence oflargesubmarinelandslides onthe Atlantic Ocean margin,Mar. Geol.,264, 53–64. LeVeque, R. (2002),Finite Volume Methods for Hyperbolic Problems, vol. 54, 258 pp., Cambridge Univ. Press, Cambridge, U. K.,

doi:10.1017/CBO9780511791253.

Lindberg, B.,J. Laberg, and T. Vorren(2004), The NYK Slide—Morphology, progression, and age of a partly buriedsubmarineslide offshore northern Norway,Mar. Geol.,213, 277–289.

Locat, J., S. Leroueil, A. Locat, and H. Lee(2014), Weaklayers: Their definition and classificationfrom a geotechnical perspective, inSubmarine Mass Movements and Their Consequences, edited by S. Krastel et al., pp. 3–12, SpringerInt., Switzerland. Løvholt, F., C. B. Harbitz, and K. B. Haugen(2005), A parametric study of tsunamis generated by submarine slidesin the Ormen

Lange/Storegga area off western Norway,Mar. Pet. Geol.,22, 219–232.

Løvholt, F., G. Pedersen, and G. Gisler(2008), Oceanic propagation of a potentialtsunamifromthe La PalmaIsland,J. Geophys. Res.,113, C09026, doi:10.1029/2007JC004603.

Løvholt, F., G. Pedersen, and S. Glimsdal(2010), Coupling of dispersive tsunami propagation and shallow water coastal response, Open Oceanogr.J.,4, 71–82, doi:10.2174/1874252101004020071.

Løvholt, F., G. Pedersen, C. B. Harbitz, S. Glimsdal, and J. Kim(2015), Onthe characteristics oflandslidetsunamis,Philos. Trans. R. Soc. A, 373(2053), 20140376, doi:10.1098/rsta.2014.0376.

Løvholt, F., G. Pedersen, and C. B. Harbitz (2016), Tsunami-genesis due to retrogressivelandslides on aninclined seabed, inSubmarine Mass Movements and Their Consequences, edited by G. Lamarche et al., pp. 569–578, SpringerInt., Switzerland, doi:10.1007/978-3-319-20979-1_57.

Lynett, P. J., J. C. Borrero, P. L.-F. Liu, and C. E. Synolakis(2003), Field survey and numerical simulations: Areview ofthe 1998 Papua New Guinea Tsunami,Pure Appl. Geophys.,160, 2119–2146.

Ma, G., J. Kirby, T. Hsu, and F. Shi(2015), Atwo-layer granularlandslide modelfortsunami wave generation: Theory and computation, Ocean Modell.,93, 40–55.

Masson, D., C. Harbitz, R. Wynn, G. Pedersen, and F. Løvholt(2006), Submarinelandslides—Processes,triggers and hazard prediction, Philos. Trans. R. Soc. A.,364, 2009–2039.

Mangeney, A., P. Heinrich, and R. Roche(2000), Analyticalsolutionfortesting debris avalanche numerical models,Pure Appl. Geophys.,157, 1081–1096.

Okal, E. A., and C. E. Synolakis(2004), Source discriminantsfor near-fieldtsunamis,Geophys.J.Int.,158(3), 899–912.

Romundset, A., and S. Bondevik(2011), Propagation ofthe Storeggatsunamiintoice-freelakes alongthe southern shores ofthe Barents Sea,J. Quat. Sci.,26(5), 457–462, doi:10.1002/jqs.1511.

Smith, D., et al. (2004), The Holocene Storegga Slide tsunamiin the United Kingdom,Quat. Sci. Rev.,23(23–24), 2291–2321, doi:10.1016/j.quascirev.2004.04.001.

Synolakis, C., and U. Kanoglu(2015), The Fukushima accident was preventable,Philos. Trans. R. Soc. A,373, 2014079. Tappin, D.(2010), Submarine massfailures astsunamisources—Their climate control,Phil.Trans.R.Soc.A,368, 2417–2434.

Tappin, D., P. Watts, and S. Grilli(2008), The Papua New Guineatsunami of 17July 1998: Anatomy of a catastrophic event,Nat.Hazards Earth Syst. Sci.,8, 243–266.

(10)

Titov, V., U. Kanoglu, and C. Synolakis(2013), Development of mostforreal-timetsunamiforecasting,J. Waterw. Port Coastal Ocean Eng., 142(6), 03116004, doi:10.1061/(ASCE)WW.1943-5460.0000357.

Twichell, D. C.,J. D. Chaytor, U. S.ten Brink, and B. Buczkowski(2009), Morphology oflate Quaternary submarinelandslides alongthe U.S. Atlantic continental margin,Mar. Geol.,264(1–2), 4–15, doi:10.1016/j.margeo.2009.01.009,tsunami hazard alongthe U.S. Atlantic coast. Yavari-Ramshe, S., and B. Ataie-Ashtiani(2016), Numerical simulation of subaerial and submarinelandslide generated tsunami

waves—Recent advances andfuture challenges,Landslides,13(6), 1325–1368.

Referanser

RELATERTE DOKUMENTER

This report documents the experiences and lessons from the deployment of operational analysts to Afghanistan with the Norwegian Armed Forces, with regard to the concept, the main

Overall, the SAB considered 60 chemicals that included: (a) 14 declared as RCAs since entry into force of the Convention; (b) chemicals identied as potential RCAs from a list of

[ 58 ] On the basis of thirteen events of in situ Cluster ob- servations from the reconnection region we have examined whether magnetotail reconnection can produce the

Abstract A two-and-a-half-dimensional interactive stratospheric model(i.e., a zonally averaged dynamical-chemical model combined with a truncated spectral dynamical model),

However, a shift in research and policy focus on the European Arctic from state security to human and regional security, as well as an increased attention towards non-military

In a review of US military organizations at war, Roman (1997) found such organizational practices to be obstructing the sharing of information, as well as being an obstacle

[ 2 ] The Residual Geoid (RG) [Hager, 1984] and the Areoid (AR) [Konopliv et al., 2006], the equipotential fig- ures of the Earth and Mars, are characterized by similar features

We show here the input G-buffers, parameters and textures used to produce the styles demonstrated in the paper and in the