Mass-gravity flows pose major hazards for deepwater pipelines

Nov. 20, 2000
Two recent ultra-deepwater pipeline projects have encountered mass-gravity flows as significant geohazards.

Two recent ultra-deepwater pipeline projects have encountered mass-gravity flows as significant geohazards.

The pipelines are the Shell Malampaya Pipeline in the Philippines and the Gazprom Blue Stream Pipeline across the Black Sea.

These and other deepwater projects point out the need for methods quantitatively to assess these risks in a wide variety of deep seafloor terrains.

Rare, unpredictable flows

On this relief map of the seafloor off the coast of Mindoro Island, deep incised channels and levee systems are the result of turbidity currents (Fig. 1).
Click here to enlarge image

In some deepwater and continental slope environments (Fig. 1), high-resolution bathymetry data have revealed steep and rugged topography with evidence of seafloor instabilities and mass-gravity flows. Although the bathymetric data can be used to aid in routing and site-selection studies, they are not sufficient to assess the risk of future mass-gravity flow events.

The key pieces of information needed for a risk analysis are the frequencies at which the events can be expected to occur and their magnitudes. They can be characterized in terms of speed, width, height, and density. The principal difficulty in quantifying these risks is that mass-gravity flows are rare and generally unpredictable, making direct measurements impractical.

This contrasts with more classical oceanographic design criteria studies in which it is possible to quantify the future risk by analyzing a large number of previous occurrences.

For example, either direct analysis of long records of wave characteristics or similar long records computed from wind data can be used to establish the design wave heights and periods. Both the likelihood and the magnitude of these hazards are well documented.

Despite the lack of direct measurements of mass-gravity flow events, there is a wealth of information in their seafloor deposits. USR Corp. has developed an approach for establishing engineering design criteria for mass-gravity flows that exploits the information available in those deposits.

This is done by combining measured data (bathymetery, sub-bottom profiling, and sediment cores) with detailed process-based models of the mass-gravity flow events. The models act as transfer functions, taking information about the mass-gravity event deposits and associating the speed, width, height, and density.

These models represent the balances of inertia, dissipation, and gravitational acceleration and have been developed to reproduce the entire flow episode, from initiating conditions through an unsteady flow down the slope and ultimately the development of the final deposits.

When the models have been suitably constrained to reproduce the measured deposit characteristics, the predicted velocities and densities of the flow are combined with frequency data to develop the design criteria.

In most study areas, the seafloor deposits have formed from the action of a large number of individual events whose properties can be sampled and dated. The development of recurrence rates from the dating analysis provides the frequency data.

In conjunction with the age-dating analysis, or when dating analysis is not available, the mass-gravity flow source or 'triggering' events may be used to estimate recurrence rates.

For instance, if seismic activity is expected to be the source of debris flows, then the established frequency of seismic events can be combined with the slope-stability analysis, measured data, and modeling to develop the design criteria.

Flow models

The mass-gravity flows that are modeled can generally be divided into two categories, debris flows and turbidity currents, and therefore two different flow models have been developed.

Debris flows, which include mud flows, are mass movements in which the source sediment travels downslope, coming to rest after the initially stored potential energy is dissipated by friction.

During the flows, the source sediment is remolded and reconstituted; the degree to which this occurs determines the rheological properties and flow type.

Generally, the soil mass travels as a visco-plastic material, with distinct stress-strain rate characteristics. Turbidity currents represent a significantly different form of mass-gravity flow. Suspended sediment provides the density contrast with the ambient water in these turbulent currents and gives rise to the gravitational energy that drives the flows.

These flows erode on the steep upper slopes and form deposits further downslope. The density of a turbidity current is on the order of 2-4% greater than that of the surrounding ambient water. These flows could travel as fast as 22 m/sec,1 but more recent measurements indicate speeds ranging between 10 m/sec to a fraction of 1 m/sec.2 3

Debris flows are simulated with the debris flow model BING, which was developed at the University of Minnesota and is based on the Jiang and LeBlond model that uses a Bingham rheology (yield strength and viscosity).4

The model is time-dependent and simulates the deformation of an initial block of sediment as it flows down slope and eventually comes to rest, providing predictions of flow speed, flow thickness, and run-out distance.

The model is based on numerical solutions to differential equations for the conservation of mass and the conservation of momentum. Five model parameters requiring specification are the initial source or "failure block" geometry, the bathymetric profile slope, sediment density, sediment yield strength, and sediment viscosity.

Once the model's initial conditions (that is, failure block) are specified, the model implementation proceeds with the numerical solution to the governing equations. The numerical solution provides the speed and geometry of the debris flow, starting with the initial conditions and following it down slope until it eventually comes to rest.

Click here to enlarge image

Output includes time-dependent information on the height, length, and speed as a function of time and position down the slope. The debris-flow model has been executed for a typical flow scenario to demonstrate its operation (Fig. 2).

The graph shows the evolution of the flow from an initial failure block to the final deposit. The front of the turbidity current is predicted to approach 20 m/sec about a km away from the failure area. The debris flow model is constrained by comparing the predicted final deposit geometry and position with measured data.The initial failure block is shown as a parabolic mound on the seafloor. The initial block deforms into a debris flow, travels down slope, eventually coming to rest. Typically, velocities at the head of the debris flow are predicted to be in excess of 20 m/sec.

Turbidity currents are simulated with a width-averaged, two-dimensional model representing the vertical and longitudinal (down slope) coordinates. This model consists of three major components: hydrodynamics, sediment transport, and bed evolution.

The hydrodynamics component is based on a numerical solution to the Reynolds-averaged Navier-Stokes equations. The model represents inertial and gravitational forces, advection, turbulent diffusion, density-stratification effects on mixing, erosion, deposition, and bed armoring.

The sediment transport component consists of two parts: a suspended sediment algorithm, based on a numerical solution to the two-dimensional scalar transport equation, and a bedload transport algorithm.

The last component of the model is the bed composition-tracking algorithm. This keeps track of the erosion and deposition at the bed surface and adjusts the bed sediment distribution to conserve mass.

Application of the model requires specification of the bathymetry, sediment-grain size classes (multiple grain size classes are simulated), and the source or initial conditions. The initial conditions consist of specifying the flow speed, height, and suspended sediment concentrations at the upstream end of the current.

Click here to enlarge image

The model simulations can best be seen in contour plots of the suspended sediment (Fig. 3).

The turbidity current model simulation shows the evolution of a flow. The current thickness grows as the current flows down slope. Vertical profiles of the current at 1, 5, and 10 km show that as the current grows, the peak speed remains fairly constant but the suspended sediment load is mixed upward over the height of the current.The flow emanates from the upstream end, expanding as it flows in the downslope direction. The concentration is reduced in the upper portions of the current due to sediment settling and mixing with ambient fluid. The vertical profiles show a high-speed region of the turbidity current close to the bed.

Above the high-speed region, the speed decreases, eventually to the ambient flow speed (that is, zero). The concentration profile reveals the upward mixing of sediments from the high concentration region near the bed. The stress profile shows the highest stress occurring at the bed and a decreasing stress upward characteristic of gravity-driven flows.

Click here to enlarge image

The deposit created by the simulated turbidity current shows the downslope sorting of sediment particle sizes, reflecting the different mobilities of the grain sizes (Fig. 4). Proper simulation of this characteristic is critical because it is used to constrain the model to the measured data.

The clay material content of the sediment deposit is minimal because its settling velocity is so low. In real flows, these fine sediments may drift with ambient currents after the turbidity current has ceased, and run out over a large area of the abyssal plain. The fine sand is transported far downstream and is much more mobile than the gravel class, which travels only as bedload.

Five-step approach

The approach for developing design criteria for mass-gravity flows is divided into five steps. The first step, which involves collection, review, and analysis of field data, provides insight into the controlling processes and the likelihood of various event-triggering mechanisms. This step provides the basis for developing event scenarios for use with the models.

In addition, the data are used to constrain the models and, after specific specialized laboratory analysis of field samples, also provide ranges (i.e., upper and lower bounds) of values for model parameters.

Next, the models are implemented for the event scenarios developed in the data analysis. Generally, the initial conditions and parameters are varied in a sequence of model simulations until the measured deposit characteristics are reproduced by the model. This process is essentially a model calibration and is completed for a number of events that have been identified from the field measurements and data review. When completed, flow speeds, densities, and heights and their variations along the flow path are predicted for each of the events considered.

The third step in the approach is age dating of the mass-gravity flow events, which is necessary to develop recurrence rates. Field samples associated with the deposits (i.e., within, above, or below) are dated to determine the age of each event.

Ideally, the range of events for which measurements were obtained represents a sufficient range of event magnitudes. The recurrence interval for events of different magnitudes can then be determined.

Selecting the number and location of sediment cores is one of the most important steps in planning the field survey because they must be located in areas where the chance of encountering a stacked sequence of mass gravity flow-event beds with intervening "quiescent" periods is good.

Core sites selected for age-dating purposes are most often located on the periphery of the depositional zones because the event beds tend to be thinner and there is a better chance that pelagic sediment deposited between events has been maintained.

The fourth step in the approach, prognostic modeling, is only necessary if the result of the diagnostic modeling and age dating do not span the range of return intervals necessary to develop the design criteria. For instance, if the oldest event measured is on the order of 1,000 years and the design required consideration of a 10,000-year event, additional analysis is required.

In this case, the magnitude of the 10,000-year event would be extrapolated from the age-dating analysis. The conditions associated with that event are then simulated using the mass gravity-flow models to determine the associated flow speed, density, and height profiles.

Click here to enlarge image

The final step in the approach consists of summarizing the age-dating results and modeling data. This usually consists of curve plotting a flow parameter (i.e., speed) against the return period for the associated event (Fig. 5).

For turbidity currents, the speed is the most critical parameter, as indicated in Fig. 5. For debris flows, the interest is usually the farthest distance that the flow will extend relative to the position of the pipeline.

Such curves then can be used in a risk analysis for the pipeline design.

Blue Stream

A shaded map of the seafloor slope off the Russian coast of the Black Sea reveals the dendritic canyon structure. Major canyons extend from headwalls at the shelf break to about 2000m water depths. Smaller canyons are imbedded in the sidewalls of the major canyons and have formed by a sequence of headwall failures (Fig. 6).
Click here to enlarge image

The routing of the Blue Stream pipeline (Fig. 6) from the bottom of the Black Sea to the shoreline in southern Russia near the town of Djubga necessitated mapping a 20 km by 35 km area parallel and perpendicular to the shore, respectively.5

In the first phase of the Blue Stream geophysical survey, combined MAK-1M side scan sonar and sub-bottom profiler data were collected. Swath bathymetry was obtained using a SeaBat multibeam echo sounder. The second phase of the geophysical survey consisted of using an ROV as a platform with the following instruments mounted on it: a SeaBat 8101 multibeam system, Geoacoustics "Chirp" 3.5-11 kHz sub-bottom profiler and side scan sonar, and a video system for seafloor inspection.6 The geotechnical survey consisted of vibra-coring, box coring, piston coring, and cone-penetrometer tests. Additional characterization consisted of detailed grain-size analysis, age-dating samples, and hydrometer tests.

A dendritic complex of submarine canyons was found over an area extending from the shelf edge at about the 80 m water depth to the beginning of the abyssal plain near a depth of 2,000 m. A few large canyons extend up to the shelf break and are associated with major fault sequences; however, many of the smaller canyons emerge from the side walls of the major canyons.

Click here to enlarge image

Evidence indicates that these canyons have grown upwards by sequential failures of their head walls, probably triggered by earthquakes. In some cases, the survey data distinguished the scar of the mass soil failure that became the debris flow (Fig. 7).

In Fig. 7, a debris-flow deposit, mapped at about 1,600 m depth, was evident in bathymetric, sub-bottom profile and core data. The source area for the flow was also apparent, and the data were used to constrain the debris flow model.This allowed estimates of the height and volume of the initial 'failure block' which are needed as initial conditions in the debris-flow modeling.

A diagnostic modeling scenario for this feature was derived directly from the field data.

Rheological tests on debris-flow deposit samples provided a relationship between the water content and the three parameters: sediment bulk density, yield strength, and viscosity.

Click here to enlarge image

Although the precise values that occur during actual flows are not known, the test results provide a range of conditions that can be used to constrain values. The diagnostic modeling was conducted by implementing the model for a variety of parameter values (bulk density, viscosity, and yield strength) until the predicted final deposit was in good agreement with the measured deposit (Fig. 8).

The debris-flow model parameters have been calibrated so that the simulated final deposit closely matches the location and geometry of the measured debris-flow deposit. Predicted flow speeds at the front of the debris flow are in excess of 20 m/sec.

Diagnostic modeling was conducted for each of the features for which sufficient field data were available to constrain the model. One of the most notable results of the modeling analysis was the need to use a much larger viscosity than the values obtained in the laboratory tests to obtain good agreement with the measured deposit.

When the laboratory values were used, the model grossly over-predicted the down-slope extent of the debris flow deposit. This is likely due to differences in the sediment composition assumed in the laboratory rheology test and that which occurred in the field. The laboratory tests assumed a homogeneous liquefaction, whereas in the field, large undeformed clasts existed within a nonhomogeneous liquefied matrix. This composition probably increased the effective viscosity of the debris-flow material.

The design criteria addressed debris flow speeds and ranges for events with return intervals of 500 and 10,000 years. The age dating of the sampled debris-flow deposits did not provide sufficient data on return intervals to address this range, and an alternate approach was adopted.

The failure block parameters were derived from a parallel study of earthquake-induced failures. This study provided failure block characteristics and associated return intervals for failures in selected canyons on the Russian slope.

The failure block dimensions from the earthquake study and the model parameter values obtained in the diagnostic modeling simulations were used for prognostic modeling. When the modeling analysis was completed, the debris-flow parameters (speed and run-out distance) were tabulated, along with the associated return interval.

As the predicted debris-flow speeds were quite large, on the order of 20 m/sec, the primary focus of this analysis was on the run-out distance of the debris flow. The downslope extent of debris flows on this slope was relatively short (1 or 2 km) when compared to turbidity currents. The analysis focused on determining which events would trigger a debris flow that would cross the proposed pipeline route.

Malampaya

Survey operations for the Malampaya pipeline consisted of swath bathymetry, sidescan sonar, sub-bottom profiling, and high-resolution digital seismic. Geotechnical evaluation consisted of vibra-coring, box coring, piston coring, hydraulic coring, hydrostatic coring, cone-penetrometer tests, and an ROV for video observations of the seafloor.7

Additional characterization consisted of seismic velocity whole-core logging, detailed grain-size analysis, and age-dating samples. The erosion characteristics of a number of samples were determined in special laboratory flumes.

A system of submarine canyons was discovered along the pipeline route (Fig. 1) off the coast of Mindoro Island. Because the pipeline route paralleled the coast, it crossed a number of individual canyons. One of the larger submarine canyon systems is located off the mouth of the Bongabong River.

The top of the submarine canyon is at the mouth of the river and about 2 km offshore at a water depth of 80 m. The system continues to a depth of 600 m, and overall is 22 km long. The upper portion of the canyon system has deeply incised and steep-walled valleys that join and extend down to 350 m. We have designated this portion of the system the 'chutes' because they are clearly erosional and have only small deposits of coarse sand, gravel, and cobbles at their bottoms.

Beyond the depth of 350 m, the seafloor is a system of shifting channels and coarse sediment resembling terrestrial systems of coalesced alluvial fans. These extend to a depth of 450 m, where the channel system becomes simpler and better organized into an individual channel with subsea levee banks to either side.

The channel bottom continues to have coarse sand and gravel sediments, but the maximum size of the gravel progressively decreases until gravel no longer occurs near a depth of 500 m. A little further offshore, at a depth of 550 m, the channel and levees flatten out to a relatively smooth run-out zone.

Click here to enlarge image

The grain-size data have been mapped along the profile of the main canyon (Fig. 9). Event bed grain-size data from numerous cores taken along the turbidity current flow path have been analyzed and used to create this composite picture showing the general trends of progressive sorting in the downstream direction. These data represent a composite of data from several individual events so that the maximum downslope extent should be considered the furthest downslope the turbidity currents exert enough bottom stress to entrain grains of these sizes.

The grain size and the thickness of each bed were measured. Beds as thick as 0.5 m were found in cores taken at the 585-m water depth. However, it was discovered later that bed thickness was not a good parameter to equate to flow intensity.

The turbidity currents were likely related to long-duration hyperpycnal flows emanating from the river mouth. Thus, in this case, the bed thickness was associated with flow duration rather than intensity.

Sensitivity analyses were conducted prior to obtaining diagnostic modeling results that conform to the grain size data in the measured event beds. The relevant conclusions from these analyses are:

  • The turbidity current speeds could not be maintained down the entire channel system unless a significant portion of very fine material (< 0.032 mm) was present in the source material. Larger grain sizes settled out too rapidly, causing the current to lose its density difference relative to the ambient water.

    The larger grain sizes, therefore, could be considered "passengers" in the current, that were moved by the hydrodynamic forces, but are not significant in providing the density contrast.

  • The bed thicknesses found in the field data were only obtainable through long-duration simulations consistent with the view that hyperpycnal flows were the likely source of turbidity currents in the area.

The diagnostic modeling was conducted by reproducing the measured bed deposits along the canyon. A single core was selected near the base of each canyon and the event bed grain size distributions. However, the grain size data were not used directly, but rather the hydrodynamic stress associated with the event bed grain size distribution was used.

Each event bed has a characteristic grain size (the 90th percentile, D90, was used) that characterizes the larger grain sizes in the bed. Since most of the material at the D90 size travels as bed load, the critical hydrodynamic stress required to mobilize those grain sizes must have just been met by the flow.

Higher stresses than those associated with D90 can be discounted, since the next larger class would have to be present. To obtain simulated turbidity currents that matched the characteristic bed stresses, the turbidity current model was implemented for a range of source conditions, until the predicted bed stress levels at the location corresponding to the core locations would match each of the critical stresses associated with each of the event beds.

The predicted speeds at the proposed pipeline crossing were obtained from each simulation. Then a frequency analysis was conducted on the event bed age dates which provided a return interval for each event bed. It was found that the range of return intervals determined from age dating of the event beds was sufficient to develop the design criteria and therefore additional (prognostic) modeling of other conditions was not necessary.

Speeds predicted from the diagnostic model simulations corresponding to each event bed were used with the return intervals to provide a curve of turbidity current speeds vs. return interval. This curve could then be used to develop design criteria for the pipeline.

Acknowledgments

The authors express their appreciation to Shell International, as well as PeterGaz and Gazprom, for their support of this work.

References

  1. Heezen, B.C., and Ewing, M., "Turbidity Currents and Submarine Slumps, and the 1929 Grand Banks Earthquake," American Journal of Science, pp. 849-73, December 1952.1. 2. Dengler, A.T., Wilde, P., Noda, E.K., and Normark, W.R., "Turbidity currents generated by Hurricane Iwa," Geo-Marine Letters, v. 4, 1984, pp. 5-11.
  2. Hay, A.E., "Turbidity currents and submarine channel formation in Rupert Inlet, British Columbia 1: Surge observations," Journal of Geophysical Research, v. 92, 1987, pp. 2875-81.
  3. Jiang, L., and LeBlond, P.H., "Numerical modeling of an underwater Bingham plastic mudslide and the waves which it generates," Journal of Geophysical Research, pp.10303-317, 1993.
  4. Niedoroda, A.W., Reed, C.W., Parsons B.S., Breza, J., Forristall, G.Z., and Mullee, J.E., "Developing Engineering Design Criteria for Mass Gravity Flows in Deepsea Slope Environments," proceedings of the Offshore Technology Conference, paper 12069, May 2000, Houston.
  5. Bucklew, S., "Gazprom's Bluestream pipeline project using the ROV as a geophysical platform in 2,200 m in the Black Sea," proceedings of the Offshore Technology Conference, paper 11046, 1999, Houston.
  6. Schneider, T., and Groenveld, J., "The technical and engineering challenges faced in the design of the Malampaya pipeline" proceedings of the Annual Pacific Pipeline Conference, Kuala Lumpur, Malaysia, Apr. 2-3, 1998.

The authors

Click here to enlarge image

Christopher W. Reed has been working as a senior engineer for URS Corp. since 1995. Before joining URS, he worked independently as a consultant in the defense and environmental engineering industries. Chris received his bachelors degree (1982) in engineering sciences from the Georgia Institute of Technology, Atlanta, and a masters (1984) and PhD (1987) from the University of Florida. He is a member of the American Geophysical Union, the ASME, and the American Society of Civil Engineers.

Click here to enlarge image

Alan W. Niedoroda has been a vice-president of URS Corp. since 1998, having also been a vice-president with Woodward-Clyde Consultants since 1995; Woodward-Clyde became a part of URS in 1998. He also served as a senior consultant for Woodward-Clyde Consultants (1991-95), senior oceanographer, Environmental Science and Engineering (1987-1991), a senior oceanographer for R.J. Brown & Assoc. (1983-1987), among other positions in his career. Niedoroda holds a BA (1966) from Queens College of the City University of New York and MS (1968) and PhD (1968) degrees, both from Florida State University. He is a member of the American Geophysical Union and ASME.

Click here to enlarge image

George Z. Forristall is a research advisor with Shell International Exploration and Production, Houston. Since 1972, he has held several positions with Shell in Houston and Holland. Forristall received all of his academic degrees from Rice University, Houston, finishing with a PhD (1970) in mechanical engineering. He is a member of the American Geophysical Union.

Click here to enlarge image

John E. Mullee is a senior project manager with INTEC Engineering Inc., Houston. He has 30 years' experience in engineering and construction supervision of pipelines and facilities worldwide. Mullee holds a bachelors degree in civil engineering from the National University of Ireland and is an associate member of the American Society of Civil Engineers.