DYNAMICS, VELOCITY AND RUN-OUT OF THE GIANT STOREGGA SLIDE F. V. DE BLASIO1, D. ISSLER1,2, A. ELVERHØI1, C. B. HARBITZ2, T. ILSTAD1, P. BRYN3, R. LIEN3 , F. LØVHOLT2 1 Institutt for geologi, Universitetet i Oslo, P.O. Box 1047 Blindern, 0316 Oslo, Norway 2 NGI, POB 3930 Ullevaal Stadion,N-0806 Oslo, Norway 3 Norsk Hydro, N-0246 Oslo, Norway Abstract A huge slide (volume of 2400 km3 and run-out 450 km) was released in the Storegga area off the western coast of Norway during early Holocene, followed by numerous smaller debris flows. We perform numerical simulations of the giant slide using a Bingham model for the clay material. Agreement with present deposit distribution and run-out is found by assuming that the shear resistance between the debris flow and the seabed decreases during the flow, and we suggest sediment remolding or hydroplaning as possible explanations. Debris velocities are predicted and possible applications to the associated tsunami event are investigated. Keywords: Submarine slide, debris flow, turbidity current, Storegga 1. Introduction The giant Storegga slide that occurred along the western coastline of Norway in the late Holocene is one of Earth’s largest known gravity mass flows. Due to the plans for gas field developments, the Storegga slide area has been intensively investigated, especially in the upper parts close to the headwall (the region of the Ormen Lange gas field). Detailed sea floor surveying combined with high resolution seismic, sediment coring and drilling have provided an extensive database and detailed maps of the various slide phases (Haflidason et al. 2002). Lobes from more than 70 individual slide events have been identified, ranging from the first major phase of 2400 km3 with a run-out distance of 450 km (termed phase 1), to minor slides of about 0.01 km3 or less located along the headwall. In a companion article (Issler et al., 2003) we present numerical simulations of the debris flows in the Ormen Lange area, where the typical run-out distance of debris flows is 15–20 km. The present paper focuses exclusively on the large Storegga slide “phase 1” as defined by Haflidason et al. (2002), see Fig. 1. The aim is to simulate the dynamics of the flow and velocity based on the only constraint of post-slide deposit location, thickness and run-out. As for the debris flows in Ormen Lange (Issler et al., 2003), basis for our simulation is the BING model for a Bingham visco-plastic fluid (Huang and Garcia, 1999; Imran et al., 2001), supplemented by a front and a surface drag. Some variants of the basic model are also introduced. 223 224 De Blasio et al. Initial deposit 20 kPa 0.5 kPa Present deposit 10 kPa seabed profile Figure 1. Simulated profiles of the final deposit. The material is modelled as a viscoplastic (Bingham) fluid with the indicated yield stress. The present deposit profile is also shown for comparison. 2. Simulations of Storegga slide, phase 1 In a first approximation we use a Bingham (visco-plastic) model with yield stress of 10 and 20 kPa. The Bingham rheology rather than a granular model was selected because of the high clay content of the sediments (Huang and Garcia, 1998, 1999; Coussot, 1996). Fig. 1 shows the final deposits from some calculations. The results are at variance with the field data: the calculated run-out distance is much smaller than the observed one and the deposit shapes do not reflect the present distribution. Such values of the yield stress result in an overloading of the upper part of the slope located between 100 and 200 km, much more than observed, and absence of deposition at larger distances (400–450 km) where material from phase 1 is observed to be abundant. In order to understand the flow of the phase 1 slide, one needs to identify a mechanism capable of increasing the run-out distance, i.e., decreasing the bottom friction. We consider three candidate processes: remolding, wetting and hydroplaning. Geotechnical measurements of sediment static resistance in Storegga indicate a shear strength of the order of 50–100 kPa, at least five times larger than the dynamic yield stress needed to give an accurate prediction in minor debris lobes in the Ormen Lange area. This marked decrease of the sediment resistance can be attributed to a sudden rearrangement of clay particles during the flow in these sensitive clays. A crucial question is: Could this process have led to a further reduction of strength in the phase 1 slide? A reduction of yield stress to 1–2 kPa due to strength loss is indeed observed in another region within the Storegga scar, called the northern flank, where sediments have been Dynamics and material properties of the giant Storegga slide 225 extensively remoulded and mobilised (Haflidason et al. 2002). Assuming the material of phase 1 and of the northern flank to be similar, we simulated the slide with a very small yield stress and found that a value below 0.5 kPa is necessary for the run-out distance to be compatible with field data (see Fig. 1). The plateau between 200 and 400 km plays an important role in that only simulated debris flows with yield stresses lower than 0.5 kPa can traverse it. In fact, even for a value of 2 kPa we find that the debris flow stops before the plateau. However, such a small yield stress (<0.5 kPa) results in too little deposition in the upper part of the basin. There are also sedimentological problems for this explanation. Phase 1 and North Flank sediments should differ strongly. The latter involve materials from the upper 30–50 m of the column, while phase 1 includes sediments that have been buried much deeper. An even more efficient mechanism for a progressive decrease of yield stress during the flow is wetting, i.e., water incorporation into the soil, a process promoted by the high shear rates at the interface between the flowing sediment and water. The difference with respect to the previous case is that wetting leads to a progressive (i.e., increasing with time) loss of strength. This process is not sufficiently understood at present to develop a well-founded model; rather, an empirical approach combining back-calculations of comparable slides and qualitative arguments to determine the direction of deviations from other slides is called for. In the following, we loosely refer to remoulding as the combined effect of clay particle rearrangement and wetting. We tentatively adopt a simple model for the eYROXWLRQRI WKHUKHRORJLFDOSDUDPHWHUVDOORZLQJWKH\LHOGVWUHVV y(t) to vary as τ y ( x, t ) = τ y (∞) + [τ y (0) − τ y (∞)] exp[−Γγ ( x, t )] , (1) where •y(0) and •y(•) are the initial and residual (i.e., completely remoulded) yield stresses, • is the total shear deformation, • is a dimensionless remoulding efficiency and x is the co-ordinate parallel to the sea bed. This model is reminiscent of the KomamuraHuang rheological model in soil mechanics (Komamura and Huang, 1974). The effect proposed here, however, results from wetting rather than rearrangement of soil grains. The total shear deformation at the base of the debris flow at time t is calculated using a steady state approximation for the Bingham fluid as t γ ( x, t ) = ∫ dt ’ 0 ∂U ( x, y = 0; t ) U ( x, y = D; t ) = 2∫ dt ’ ∂y D 0 t (2) where D is the depth of the shear layer and U(x,y;t) is the flow velocity parallel to the seabed as a function of the height y. As shown in Fig. 2, progressive remoulding leads to long run-out distances even for a high initial yield stress. The simulation also indicates that the front of the debris flow can potentially cross the plateau. Sediment accumulates more uniformly along the flow path, both on the steeper slopes and on gentle slopes in the distal part. In fact, in the upper part of the basin the mobilised material still has a relatively high yield stress, resulting in thick deposits, in agreement with our simulations for the Ormen Lange area, 226 De Blasio et al. which require a yield stress of 10 kPa or higher. When reaching the deeper parts of the basin, the material becomes softer and can easily cross the plateau. In short, a more homogeneous distribution of the deposit is obtained, in better agreement with the observations. Initial deposit 10 kPa with remoulding to 500 Pa 10 kPa with remoulding to 100 Pa 5 kPa with hydroplaning Seabed profile Figure 2. Simulated deposits with remoulding and with hydroplaning. In Fig. 3, the yield stress at the front is shown as a function of the position of the VOLGHIURQW:LWKDYDOXHRI Ú–4 in Equation (1), the yield stress is substantially decreased after about 50 km and the effect is significant already after only 20–25 km. A YDOXHRI LQWKLVUDQJHZRXOGH[SODLQZK\RQO\WKHODUJHVWOREHVRI2UPHQ/DQJHZRXOG be partially affected. However, this value is purely empirical and cannot presently be confirmed or refuted on the basis of laboratory measurements or theory. Hydroplaning. As the average velocity U of the debris flow front increases, the dynamic 2 pressure at the front grows rapidly as U . At a critical velocity, the pressure is sufficient for a thin water layer to be intruded underneath the debris flow, producing a lubrication effect. So far, hydroplaning has been observed only in experiments (Mohrig. et al., 1999). However, some puzzling features of submarine debris flows like out-runner blocks are strong evidence that hydroplaning can occur naturally in the sub-aqueous environment, under conditions still poorly understood. In Fig. 2 we also show the final deposit from a simulation with hydroplaning. During hydroplaning, the sediment flows in a plug-like fashion above a lubricating water layer even at low slope angles (Harbitz et al., 2002) and may be transported over a large distance until either the debris flow decelerates below the critical velocity for hydroplaning, or water is expelled from underneath the debris flow or is absorbed by the sediment. Fig. 2 shows that including hydroplaning, most of the debris stops on the plateau, although the front part may hydroplane Dynamics and material properties of the giant Storegga slide 227 a little further. Unfortunately, the model for hydroplaning depends on unknown parameters, such as the minimum thickness of the water layer for an efficient lubrication, Figure 3. Change in the yield stress for the same model as in Fig. 2. The material starts with a yield stress of 10 kPa and is remoulded to a minimum value of 100 Pa or 3D:LWKDFRQVWDQW = 0.0005, remoulding becomes very pronounced after about 50 km of flow. or the stress between the water layer and the debris flow. Even apparently innocuous quantities such as the initial water velocity profile can affect the results of the computation. We conclude that a simple model for hydroplaning can reproduce the run-out distance of phase 1 with a yield stress one order of magnitude higher than for a non-hydroplaning simulation. In principle, both remoulding and hydroplaning may have played a role during the Storegga phase 1 debris flow. They might be two closely related rather than distinct, mutually exclusive processes. Considering the high pressure in the water layer and the shear stresses between the water layer and the debris flow (on the order of 1 MPa and 1–10 kPa, respectively) one concludes that water incorporation into the sediment at the bottom of the flow must be very effective. Hydroplaning might be favoured at the beginning of the flow, when the material is still sufficiently compact to prevent water from seeping into the sediment. During the flow, water will be incorporated in the sediment with increasing efficiency, partly because of the pores in the sediment and partly because of cracks and rupture planes created by the large shear stresses. Note also that for the dynamics, one does not need the whole debris to be fully wetted. From Newton’s equations, it follows in fact that the resistive forces determining the change in the average momentum of the debris flow depend only on the shear at the top (drag force) and at the bottom, while the material properties in between can only redistribute locally the velocity. 228 De Blasio et al. 3. Velocity of the giant Storegga slide and tsunami generation After the initial acceleration, the forces acting on a debris flow are close to equilibrium between the component of the gravity force parallel to the seabed, the drag force exerted by water and the friction at the base. In the high-velocity and low-yield stress regime relevant to the phase 1 slide the drag force becomes the most important resistive force, 1/6 and it follows that the velocity U scales like U ~ V where V is the volume. One can thus expect the flow of the huge mass of Storegga to be associated with very high velocities. The velocities without and with hydroplaning are shown in Figs. 4. Interestingly, the top velocities are approximately the same in all cases (about 60 m/s or more than 200 km/h), but the flow without remoulding stops after only 1.5 hours. The case with remoulding and hydroplaning are more similar to each other, with flow duration of about 4–5 hours and also comparable top velocity. The critical velocity for hydroplaning (which depends on the debris flow thickness and is of the order of 25 m/s) is reached early during the simulation. The case with hydroplaning is sensitive to the details of the seabed. Since the seabed profile is partly hypothetical, the spikes in the velocity were probably absent or smeared out when and if the Storegga phase 1 was hydroplaning. The impact pressure P exerted by the front of the debris flow of density ρ on a static object 2 can be estimated as P ≈ ρ V and would be of the order of 5–8 MPa. Tsunami surface elevations of about 5 m were probably produced in the source area during the slide (Harbitz, 1992). Sedimentary traces of tsunami deposits contemporary with the Storegga slide have been tracked for example in Scotland and off the western coast of Norway (Dawson et al., 1988, 1993; Long et al., 1989; Bondevik et al., 1997). The Norwegian Geotechnical Institute (2002a, 2002b, 2002c) has produced a set of new simulations for tsunamis generated by submarine gravity mass flows in the Storegga/Ormen Lange area. Owing to the uncertainties related especially to volume and dynamics of potential future slide events, a series of 19 potential mass flow volumes moving with simple and prescribed velocity profiles have been studied for a parameter sensitivity analysis. The simulations reveal complex wave dynamics with wave amplification due to bottom topography. They show prevailing large-scale longitudinal oscillations in the fjord systems, rather insensitive to local run-up effects. While the large slide scenarios have a regional effect, the smaller slides may generate the same maximum surface elevations, but only with local effects. The maximum surface elevations correlate best with the mean kinetic energy of the slide or with the product of initial acceleration and volume of the slide (i.e. mean kinetic energy normalised with respect to run-out and density). Hence, the results show that both the initial acceleration and the volume are important for determining the maximum surface elevation, but that the run-out is of minor importance. For the largest slide volume (2400 km3), the simulated surface elevations vary from about 50 m in the most exposed inner fjord locations to about 10 – 15 m along the coast. The wave current speeds are large (up to 2.6 m/s) for the open sea locations close to the wave generation area. The wave current speeds are about 1 m/s and less in the open sea locations outside the wave generation areas. The wave current speeds can be signifi- Dynamics and material properties of the giant Storegga slide 229 cantly larger in the coastal and fjord domains. The dominating wave periods are above 2 hours for the largest slide volume. For the smaller slides producing shorter waves, the maximum surface elevations are larger at the coastal locations than in the fjords. =10 kPa with remoulding =20 kPa with remoulding =5 kPa Bingham =5 kPa with hydroplaning Figure 4. Front velocity of the debris flow as a function of the position with the different models explained in the main text. Average velocities are approximately 60 % of the value at the front. Slide volumes less than 5 km3 with initial acceleration of 0.033 m/s2, maximum velocity of 10 m/s, and run-out distance of less than 6 km do not give significant impact on land, i.e. the tsunami inundation level added to the mean tidal high water does not surpass the highest recorded sea level. The velocity profiles are calibrated versus retrogressive slide run-out analyses (Norwegian Geotechnical Institute 2002c). 4. Conclusions Our study of Storegga slide phase 1 uses the BING model for a Bingham visco-plastic fluid including mechanisms such as high-degree remoulding and hydroplaning. Our simulations indicate a decrease of the resistive shear forces between the debris flow and the seabed during flow, and we suggest remoulding and hydroplaning as possible mechanisms. At present, our understanding of the physics of remoulding and hydroplaning and a more strict determination of the relevant parameters is far from complete. The most significant gaps in our present understanding of the Storegga phase 1 and similar slides are: (i) the break-up rate of overconsolidated clays, leading to a dramatic decrease of strength, (ii) the rate of mud entrainment and mixing with water in the shear layer, (iii) hydroplaning and (iv) the relation between hydroplaning and wetting. To produce more reliable flow simulations, dedicated experiments and theoretical work on these topics should be carried out. 230 De Blasio et al. 5. References Bondevik, S., Svendsen, J. I., Johnsen, G., Mangerud, J., and Kaland, P. E., 1997. The Storegga tsunami along the Norwegian coast, its age and runup. Boreas 26:29–53. Coussot, P., 1997. Mudflow Rheology and Dynamics. IAHR-AIRH monograph series, Balkema. Dawson, A.G., Long, D., and Smith, D.E. (1988). The Storegga Slides: Evidence from eastern Scotland for a possible tsunami. Marine Geology 82, 271-276. Dawson, A.G., Long, D., Smith, D.E., Shi, S., and Foster, I.D.L. (1993). Tsunamis in the Norwegian Sea and the North Sea caused by the Storegga submarine landslides. In: S. Tinti (ed.) Tsunamis in the world, Kluwer Academic Publishers, Netherlands, 228 pp. Edgers, L., and Karlsrud, K., 1982. Soil flows generated by submarine slides—case studies and consequences. In: Chryssostomidis, C. and Connor, J.J. (eds.), Proceedings of the Third International Conference on the Behaviour of Offshore Structures. Hemisphere, Bristol, pp. 425– 437. Haflidason, H., B., Torgersen, B., Lie, H., and Sulebak, J. R., 2002. Morphological analyses of the seafloor in the Storegga Slide area based on the TOBI and the bathymetrical data sets available. Final Report to Norsk Hydro AS 100-03/02. Bergen, Norway, Dept. of Geology, University of Bergen. Harbitz, C.B., 1992. Model simulation of tsunamis generated by the Storegga slides. Marine Geol. 105:1–21. Harbitz, C.B., Parker, G., Elverhøi, A., Mohrig, D., and Harff, P., 1992. Hydroplaning of subaqueous debris flows and glide blocks: Analytical solutions and discussions. J. Geophys. Res., in press. Huang, X. and Garcia, M. H., 1998. A Herschel-Bulkley model for mud flow down a slope. J. Fluid Mech., 374:305–333. Huang, X. and Garc•a, M.H., 1999. Modeling of non-hydroplaning mud flows on continental slopes. Marine Geol., 154:132–142. Imran, J., Harff, P. and Parker, G., 2001. A numerical model of submarine debris flows with graphical user interface. Computers and Geosciences, 27:717–729. Issler, D., De Blasio, F. V., Elverhøi, A., Ilstad, T., Bryn, P. and Lien, R., 2003. Issues in the assessment of gravity mass flow hazard in the Storegga area off the western Norwegian coast. This volume. Komamura, F. and Huang, R. J., 1974. A new rheological model for soil behavior. J. Geotech. Engng. Div., A.S.C.E., 100: 807–824. Locat, J. and Lee, H.J., 2002. Submarine landslides: advances and challenges. Can. Geotech. J., 39:193–212. Long, D., Smith, D.E., and Dawson, A.G. (1989). A Holocene tsunami deposit in eastern Scotland. J. Quat. Sci. 4, 61-66. Mohrig, D., Elverhøi, A. and Parker, G. 1999. Experiments on the relative mobility of muddy subaqueous and subaerial debris flows, and their capacity to remobilize antecedent deposits. Marine Geol., 154: 117–129. Norwegian Geotechnical Institute (2002a). Slope Stability Assessment in the Ormen Lange Field– Tsunami Analyses. NGI report 993016-8. Norwegian Geotechnical Institute (2002b). Slope Stability Assessment in the Ormen Lange Field– Extended Tsunami Analyses. NGI report 993016-16. Norwegian Geotechnical Institute (2002c). Ormen Lange Slope Stability Assessment – Run out Analyses – Upper headwall. NGI report 993016-11.