Reservoir Sedimentation. Jolanda Jenzer Althaus and Giovanni De Cesare

Transcription

1 Reservoir Sedimentation Jolanda Jenzer Althaus and Giovanni De Cesare

2

3 ALPRESERV Sustainable Sediment Management in Alpine Reservoirs considering ecological and economical aspects Volume 3 / 2006 Reservoir Sedimentation Jolanda Jenzer Althaus and Giovanni De Cesare Neubiberg 2006

4 Die Deutsche Bibliothek CIP-Einheitsaufnahme ALPRESERV Sustainable Sediment Management in Alpine Reservoirs considering ecological and economical aspects Neubiberg, 2006 Publisher: Institut für Wasserwesen Universität der Bundeswehr München Neubiberg Germany Tel.: +49-(0) , Fax: +49-(0) Editor: Dr.-Ing. Sven Hartmann Dr.-Ing. Helmut Knoblauch Dr.-Ing. Giovanni De Cesare Dipl.-Bauing. Jolanda Jenzer Althaus Dipl.-Math. Christiane Steinich ISSN ISSN Alpreserv (Print) Alpreserv (Internet) 2006 All rights reserved Print: Universität der Bundeswehr München (Germany)

5 Contents I Contents 1 The Problem of Reservoir Sedimentation 1 2 Run-of-River Installations Reservoir and Sedimentation The Morphology of Reservoir Sedimentation - the Phenomenon Influencing factors The Mechanics of Sedimentation Dead-Water Zones Case study Experience gathered in Reservoir Flushing on the River Drau in Austria Plant Description Sediment Management Reservoir Flushing Experiences Discussion of Flushing Results as obtained at the Rosegg Dam Summary 14 3 Deep, seasonal storage reservoir Turbidity current - main cause of sediment transport in deep reservoirs Propagation of sediments to dam Theoretical Background Flow over an obstacle Flow through a screen Experimental Studies Experimental set-up at EPFL-LCH (De Cesare, 1998) Experimental results Experimental set-up at EPFL-LCH (Oehy, 2003) Flume description Properties of the sediment materials Measuring Instrumentation Flow velocity measurements Front Velocity and Time Measurements Density and Temperature Measurements Auxiliary Measurements Deposition Measurements Experimental Procedure Mixture and Flume Preparation Preparation of Measuring Instruments Experimental Run Numerical Modelling Two buoyancy-extended k-ε models or the standard k-ε formulation Results Conclusions 43

6 II Reservoir Sedimentation 3.5 Case Studies Submerged Dams in Lake Grimsel Generalities Turbidity current simulation of Flood Event in October Turbidity current passing over submerged dams Conclusions Luzzone Numerical simulation Selected numerical results of the physical model Selected numerical results from Lake Luzzone Grid Generation Boundary Conditions Simulation of a Flood Event Results Links to Observations Conclusions Livigno Scope of the study Lake topography Hydrologic and operation data Sediment characteristics Numerical Simulations using CFX Current situation Conclusions and recommendations 75 4 Outline of the Historical Development of Reservoir Sedimentation Scope of investigations Evolution of knowledge regarding reservoir sedimentation Evolution of management competence to master reservoir sedimentation Statistical analysis of publications on reservoir sedimentation Conclusions 82 5 Bibliography References for chapter Papers published before Papers published between 1950 and Papers published between 1960 and Papers published between 1970 and Papers published between 1980 and Papers published after Contact 107

7 Figures III Figures Figure 1.1-1: Classification of impounding facilities 1 Figure 1.1-2: Inventory of known measures against reservoir sedimentation (Schleiss and Oehy, 2002) 3 Figure 2.1-1: Flow and sedimentation patterns (Westrich, 1988) 5 Figure 2.2-1: Flow conditions and sedimentation process in a river sand trap (Mertens, 1987) 7 Figure 2.3-1: Location map of the Rosegg power plant 10 Figure 2.3-2: Duration and erosion volume for different reservoir flushings 11 Figure 2.3-3: Effects of reservoir flushing in autumn 1993, differences in area [m 2 ] before/ after flushing 11 Figure 2.3-4: Effects of reservoir flushing in autumn 1998, [m 2 ] before/ after flushing 12 Figure 2.3-5: Cross section changes in the Rosegg reservoir 12 Figure 2.3-6: Effects of reservoir flushing in Nov. 2004, differences in area [m 2 ] before/ after flushing 13 Figure 2.3-7: Cross section changes 13 Figure 2.3-8: Effects of reservoir flushing in Oct. 2005, differences in area [m 2 ] before/ after flushing 14 Figure 2.3-9: Cross section changes, Profiles 2, 7 and 19/1 14 Figure 3.1-1: Sediment-laden river entering a reservoir - plunging flow phenomenon and turbidity current formation. 15 Figure 3.1-2: Areas affected by sedimentation in the surroundings of a reservoir 17 Figure 3.1-3: Maximal transportable grain sizes dependent on the flow velocity of the turbidity current according to Fan (1986) 17 Figure 3.2-1: Types of turbidity currents 18 Figure 3.2-2: Turbidity current flowing at the bottom 19 Figure 3.2-3: Schematic dimensionless velocity profile for turbidity currents (Graf and Altinakar, 1996) 19 Figure 3.2-4: Flow over an obstacle (Oehy, 2003) 23 Figure 3.2-5: Flow regimes of shallow-layer flow over an obstacle (Oehy, 2003) 25 Figure 3.2-6: Flow over an obstacle in a laboratory flume (Oehy, 2003) 25 Figure 3.2-7: Flow through a screen (Oehy, 2003) 26 Figure 3.2-8: Ratio of heights down- and upstream of the screen, H p =h 3 /h 2, as function of the effective porosity, f (Oehy, 2003) 26 Figure 3.2-9: H j =h 2 /h 1 as function of the porosity f and the upstream Fr d1 (Oehy, 2003) 27 Figure : Proportion of the incoming flow that is predicted to continue through the screen as a function of the effective porosity f and the upstream Fr d1 (Oehy, 2003) 27 Figure : Flow through a screen in a laboratory flume (Oehy, 2003) 28 Figure 3.3-1: Schematic drawing of the experimental installation : 1) mixing tank, 2) upstream tank, 3) recirculation pump, 4) free surface weir, 5) inflow gate, 6) turbidity current, 7) experimental flume, 8) ultrasonic probes, 9) sharp crested weir, 10) flexible duct, 11) UVP instrument, 12) control computer 29 Figure 3.3-2: Arrangement with 8 transducers looking with an angle of 60 against the main flow 30 Figure 3.3-3: Axial disposition with 8 transducers looking straight against the main flow 30 Figure 3.3-4: Square grouping with 4 transducers on each side looking straight at and perpendicular to the main flow in the spreading part just after the inflow gate 31 Figure 3.3-5: Expanding turbidity current in the experimental flume 25 s after opening of the gate; the current spreads out almost radially, large eddies developing at the current front; these eddies give the characteristic surface appearance of turbidity currents like clouds, 125 mm x 125 mm grid on PVC bottom. 31 Figure 3.3-6: Figure 3.4-1: Figure 3.4-2: Figure 3.4-3: Figure 3.4-4: Figure 3.4-5: Measured velocity values compared with theoretical vertical velocity distribution, u(z). Values from run n 2, 80 ms between two succeeding measurements. 32 Scheme of numerical model with inflow conditions, discretized reservoir topography, bottom user interfaces and erosion-deposition map. 38 Comparison of velocity and density profiles of buoyancy-extended (Model 1: Burchard and Petersen, 1999; Model 2: Rodi, 1980) and the standard k-ε models against experimental data (SALT08; Altinakar, 1988). 41 Comparison of the propagation of the turbidity current head for the buoyancy-extended (Model 1: Burchard and Petersen (1999); Model 2: Rodi, 1980) and the standard k-ε models against experimental data (SALT08; Altinakar, 1988). 42 Comparison of velocity and density profiles of buoyancy-extended (Model 1: Burchard and Petersen, 1999; Model 2: Rodi, 1980) and the standard k-ε models against experimental data (TK1305; Altinakar, 1988). 42 Comparison of velocity profiles of buoyancy-extended (Model 1: Burchard and Petersen (1999); Model 2: Rodi, 1980) and the standard k-ε models against experimental data (NOVA7; Garcia 1993). 43

8 IV Reservoir Sedimentation Figure 3.5-1: Overview of the investigated obstacles in a 1:25'000 map 44 Figure 3.5-2: Sedimentation after the flood in October 2000 (40 m 3 /s and 15 g/l) with a single obstacle at 1845 m a. s. l. 46 Figure 3.5-3: Luzzone Reservoir during emptying in 1985, looking upstream. 47 Figure 3.5-4: Picture of the 225 m high Luzzone arch dam and its reservoir. 48 Figure 3.5-5: Measured precipitation, discharge, temperature, optical turbidity and suspended sediment samples for a typical small flood 49 Figure 3.5-6: Location and vertical disposition of current meters of the underwater measuring network consisting of 3 stations A, B and C with 2 levels at each station. 50 Figure 3.5-7: Observed directions at the bottom of the reservoir at station C, 4 meters above ground, classified by current velocity. The flow is well oriented along the longitudinal axis of the lake51 Figure 3.5-8: Computed and measured 2D flow field close to the bottom and limits of the spreading turbidity current a) 5 and b) 10 seconds after opening of the gate. Numerical simulation : black velocity vectors, turbidity current as a blue surface; Physical model : white velocity vectors, limits of the turbidity current as a bold line 53 Figure 3.5-9: Method of decomposition of the real reservoir geometry a) to a simplified physical space without shore and islands b) and then to the rectangular computational space c) 53 Figure : Location of stations s11 to s61, used for extraction of local values of the significant parameters for data analysis. Also shown are the directions of the turbidity current and the generated single grid mapped on the reservoir bottom 54 Figure : Measured and adjusted non-dimensional time evolution of inflow (, ) and sediment concentration (+) 55 Figure : Pictures a) to f) in plane view and along the axis of Luzzone Reservoir showing the advancing turbidity current 10, 20, 30, 40, 50 and 210 minutes after the rise of the hydrograph 56 Figure : Evolution in time and space of the horizontal turbidity current velocity along the axes of the reservoir at station s11 to s61 in the bottom cell 57 Figure : 3D view of the bottom of the Luzzone Reservoir showing the numerically simulated turbidity current flow, 30 minutes after the rise of the hydrograph 58 Figure : Calculated sediment depth change due to a simulated 1000-year flood on the bottom of the reservoir 58 Figure : Qualitative representation showing the location and the measured sediment deposits magnitude after 31 years in service 59 Figure : Map showing the location of the Livigno Reservoir on the Swiss-Italian border. 61 Figure : Overview of the EKW hydraulic scheme 62 Figure : 3D topography of the Livigno Reservoir 63 Figure : October 2000 flood event 64 Figure : Simulated flood events 65 Figure : Log-normal fit for the peak flood discharge at the Livigno Reservoir 65 Figure : Unit inflow hydrograph and unit concentration curve 66 Figure : Scenarios of simulations 67 Figure : Computational domain for the study 68 Figure : Final condition for the annual flood with maximum inflow concentration 15 g/l 69 Figure : Final condition for the October 2000 flood event with maximum inflow concentration 15 g/l (left) and 30 g/l (right) 69 Figure : Final condition for the 1960 flood event with maximum inflow concentration 15 g/l (left) and Figure : 30 g/l (right) 69 Final condition for the 100 years of return period flood event with maximum inflow concentration 15 g/l (left) and 30 g/l (right) 70 Figure : Evolution of the current in the cross section of the foreseen obstacle. 71 Figure : Longitudinal cross section for time step 9 h without the effects of obstacle and screen 71 Figure : Evolution of the current in the cross section of the foreseen screen. 71 Figure : Longitudinal cross section for time step 16 h without the effects of obstacle and screen 72 Figure : Zoom of the body-fitted grid around the obstacles 72 Figure : Final condition for the October 2000 flood event with a maximum inflow concentration of 15 g/l for the present situation (up left) and obstacles with 4 (up right), 8 (down left) and 12 m height (down right) 73 Figure : Longitudinal cross section at time step 9 h with the effects of obstacles of 4, 8 and 12 m height (up to down) 74 Figure : Deposition profile along the longitudinal cross section of the reservoir 75 Figure 4.4-1: Evolution of the total number of publications related to reservoir sedimentation 79 Figure 4.4-2: Evolution of the number of publications separated into five major categories 80 Figure 4.4-3: Parts of number of publications separated into the five major categories 81

9 Tables V Tables Table 3.4-1: Experimental conditions of data sets used for computations (D sg geometric mean diameter; ρ a density of ambient water). 41 Table 3.5-1: Boundary conditions used in the numerical simulation 54 Table 3.5-2: Average and peak discharge values 64 Table 3.5-3: Summary of boundary conditions used in the numerical model 67 Table 3.5-4: Characteristics of the discretisation of the grid for the current situation (SYM G.P = symmetric geometric progression; G.P = geometric progression) 68 Table 3.5-5: Ratio between the depositions upstream of the foreseen section for the alternatives and the total deposition in the full west arm of the reservoir at the end of the simulations 70 Table 3.5-6: Characteristics of the discretisation of the grid for simulations with obstacles 72 Table 3.5-7: Ratio of the depositions upstream of the foreseen sections for the alternatives and the total deposition in the full west arm of the reservoir at the end of the simulations for different obstacle heights 73

10 VI Reservoir Sedimentation

11 The Problem of Reservoir Sedimentation 1 1 The Problem of Reservoir Sedimentation Reservoir sedimentation is a problem that will keep those responsible for water resources management occupied more than usual during the decades to come. Although the aim behind the efforts to create reservoirs is storing water, other substances are carried along by the water and are usually deposited there. Lasting use of reservoirs in terms of water resources management involves the need for de-sedimentation. The alterations of the flow behaviour due to dam constructions lead to transformations in a fluvial process, where deposition of solid particles transported by the flow can be cited (Chella et al., 2003). Each reservoir, independent of its use (water supply, irrigation, energy or flood control), can have its capacity decreased due to deposition over the years. In an extreme case, this may result in the reservoir becoming filled up with sediments. A reservoir, like a natural lake, silts up more or less rapidly. In actual fact, they may completely fill with sediments even within just a few years, whereas natural lakes e.g. in our Alpine foreland, may remain as stable features of our landscape for as much as 10,000 or 20,000 years after they were formed during the last Ice Age. Impounding structures include large dams, in the form of fill or concrete dams, as well as river barrages comprising weirs, power plants, locks, impounding dams and dykes (Figure 1.1-1). The artificial lakes formed by such closure structures may be called reservoir lakes and backwater reservoirs, respectively. In addition, there are flood-retention basins, pumpedstorage basins and sedimentation basins. Impounding facility Large dam Run-of-River dam Structure Reservoir Backwater area Storage basin Reservoir basin Backwater reservoir Lake Figure 1.1-1: Classification of impounding facilities Impounding facilities are always costly, but this is justified by their various potential uses. Reservoir sedimentation, however, reduces the value of or even nullifies this investment. The use for which a reservoir was built can be sustainable or represent a renewable source of energy only where sedimentation is controlled by adequate management, for which suitable measures should be devised. All lakes and reservoirs created on natural rivers are subjected to reservoir sedimentation. There are no accurate data on the rates of reservoir sedimentation worldwide, but it is commonly accepted that about 1 2 % of the worldwide capacity is lost annually (Jacobsen 1999). Analysis of data of 14 reservoirs in Switzerland showed that this percentage is only

12 2 Reservoir Sedimentation about 0.2 % in Alpine reservoirs. The lower filling rates are results of geologic characteristics of these basins at high altitude. Nevertheless, sedimentation is also a subject of major importance in Alpine reservoirs and is in big reservoirs - mainly related to the phenomenon of sediment transport by the means of turbidity currents. The sediment discharge of the inflowing rivers is usually significant during flood events. Turbidity currents are often the governing process in reservoir sedimentation by transporting fine materials in high concentrations and following over long distances the reservoir bottom along the thalweg through the impoundment down to the deepest point in the lake normally near the dam. At the dam the eroded and transported sediments settle down and can cover the bottom outlet adversely affecting the operation of the power intake. The turbidity currents belong to the family of sediment gravity currents. These are flows of water laden with sediment that move downslope in otherwise still waters like oceans, lakes and reservoirs. Their driving force is gained from the suspended matter (fine solid material), which renders the flowing turbid water heavier than the clear water above. If the difference in density between the lake water and inflowing water is high enough, it may cause the flow to plunge and turbidity current can be induced. Turbidity currents are encountered in fluvial hydraulics, most prominently if a sediment-laden discharge enters a reservoir, where, during the passage, it may unload or even resuspend granular material. Thus sediment deposition in reservoirs not only reduces storage capacity, but also increases the risks of blockage of intake structures and sediment entrainment in waterway systems and hydropower schemes. Sediments removed by flushing have a detrimental impact on the downstream river. The design of a sustainable reservoir requires the accurate prediction of sediment transport, erosion and deposition. For existing reservoirs, deeper knowledge of the phenomenon involved is needed to better understand and solve sedimentation problems for improvement of reservoir operation. Accepted practice has been to design and operate reservoirs to fill with sediment, generating benefits from remaining storage over a finite period of time. The consequences of sedimentation and project abandonment are left to the future. These consequences can be summarized as: sediments reaching intakes and greatly accelerating abrasion of hydraulic machinery, decreasing their efficiency and increasing maintenance costs; blockage of intake and bottom outlet structures or damage to gates that are not designed for sediment passage (Boillat and Delley, 1992), etc. Considering reaches downstream of the dams, one important problem can be the augment of the erosions risk in the river since the sediment equilibrium was affected. This future has already arrived for many existing reservoirs and most others will eventually experience a similar fate, thereby imposing substantial costs on society (Palmieri et al., 2001). Normally, the measures against reservoir sedimentation can be divided in three groups, after Schleiss and Oehy, 2002: measures in the catchment area, in the reservoir and at the dam. These measures are schematically presented in Figure

13 The Problem of Reservoir Sedimentation 3 Measures against reservoir sedimentation In the catchment area In the reservoir At the dam in the reservoir Soil conservation Settling basins Slope and bank protection Bypassing structures Off-stream storage reservoir Dredging Dead storage Flushing Hydrosuction, air lift Sluicing Turbidity current venting Turbining suspended sediments Dam heightening Heightening of intake and bottom outlet structures Figure 1.1-2: Inventory of known measures against reservoir sedimentation (Schleiss and Oehy, 2002) For Alpine reservoirs, sedimentation phenomenon due to turbidity currents represents a large part of the problem and has been extensively studied in the recent years. Sediment deposition in reservoirs causes mainly loss of water storage capacity (Graf, 1984, Fan and Morris, 1992), the risk of blockage of intake structures as well as sediment entrainment in hydropower schemes (Boillat et al., 1994, Schleiss et al., 1996, De Cesare, 1998). Finally the sediment ends to some extent in the downstream river during flushing (Rambaud et al., 1988). The planning and design of a reservoir requires the accurate prediction of sediment transport, erosion and deposition in the reservoir. For existing artificial lakes, more and wider knowledge is still needed to better understand and solve the sedimentation problem, and hence improve reservoir operation.

14 4 Reservoir Sedimentation

15 Run-of-River Installations 5 2 Run-of-River Installations 2.1 Reservoir and Sedimentation Run-of-river installations can be classified into various basic types according to their flowthrough characteristics (Figure 2.1-1). Figure 2.1-1: Flow and sedimentation patterns (Westrich, 1988)

16 6 Reservoir Sedimentation Developed bodies of running water with relatively regular flow cross-sections generally show a unidimensional sedimentation profile. By contrast, backwater reaches with complex crosssections (channels with groyne fields, river beds with flooded washlands) show a clear differentiation with respect to the transport and sedimentation processes across the main flow direction. The parameters determining reservoir sedimentation are sediment delivery and discharge. Starting from the end of the upstream reach, flow velocity and bottom shear stress decrease as the cross section of discharge increases, and the solids begin to settle. Sedimentation within the reservoir is structured according to grain size. The coarser bedload material (gravel) is deposited at the end of the upstream reach, developing a sediment tongue over time, which gradually moves towards the barrage. The finer, suspended particles, such as fine sand and silt, are swept far into the reservoir and down to the dam structure, and diffuse transverse transport carries them at low flow velocities to border zones and peripheral shallow water zones, where they are allowed to settle. Only the very fine suspended particles are discharged from the reservoir. Fine material can be deposited wherever the flow velocity falls below the respective limit shear-stress velocity for the sedimentation range. Natural remobilisation of fine sediment will take place as soon as the shear-stress building up during a flood exceeds the appropriate velocity for the size of sediment. A factor of particular importance is cohesive inorganic sediment. Deposition is followed by a major consolidation phase, after which the erosion shear-stress may have increased by more than a factor of 10. Substances of man-made origin may intensify this effect. Such sedimentation zones will rarely be washed out even by major floods. 2.2 The Morphology of Reservoir Sedimentation - the Phenomenon Influencing factors The sedimentation process in a reservoir is governed by a wide variety of highly complex factors. Sediment influx, reservoir geometry and flow are considered as the determining factors. These in turn depend on a number of parameters and conditions such as catchment, reservoir management and climate. In addition, the upstream river morphology is important; in particular, other reservoirs, dams, retention basins or lakes may substantially reduce sediment delivery to the reservoir under study. Other man-made measures, such as sewers, may also have a direct effect on the sedimentation process The Mechanics of Sedimentation Flow conditions change substantially as the water travels from the end of the upstream reach in the direction of the dam. As the cross-section increases, the bottom shear-stress as a factor governing sediment transport decreases, and the solids start settling. Bedload is deposited near the upstream end of the reservoir, while the suspended particles settle further downstream. Another important factor influencing the sedimentation process, in particular its morphology, is reservoir geometry. In a long, narrow backwater reach (above a river barrage) and under idealised constant conditions (water level, inflow, sediment influx), the bedload deposit progresses relatively evenly from the upstream end of the reservoir in the direction of the dam. By contrast, the sedimentation process is irregular in reservoirs of major width. Even small bedload bars, sudden widenings of the channel (such as lakes), etc. may generate unexpected sedimentation conditions. Other factors of some importance are the changes over time in inflow, water level and sediment supply, resulting in a constant alternation between

17 Run-of-River Installations 7 sedimentation, a state of equilibrium and erosion. Stream bends intensify the difference in transport capacity of the current between the inner and the outer bends, the sedimentation tendency being higher on the inner side. An experimental study on a sand trap some 50,000 m³ in capacity (Mertens, 1987) showed a narrow main current to flow first along the left bank of the basin, while by far the largest portion of the sand trap volume was taken up by an extensive recirculation zone (Figure a). A small bedload bar in the intake area, however, altered the flow pattern substantially (Figure b). Flow over the bar was uniform and fanned out over the whole basin. Later, a wide main current developed (Figure c and Figure d), which slowly swung between the two banks, alternating between deposition and erosion of sediment. The hydraulic and sedimentological processes are even more complex and irregular in wide reservoirs (lakes) with rapidly widening inflow cross-sections. The bedload is deposited in delta-shaped formations (Mangeldorf, Scheuermann, Weiss, 1990) which, at a later stage, become traversed by several distributary streams transporting the settling particles to the channel edges. These distributaries keep migrating, causing the delta to spread approximately radially from the inlet (Mertens, 1987). In rivers developed by a series of dams, the location of the reservoir under study is another influencing factor, and so is the commissioning year, which is important for the sedimentation condition (initial sedimentation followed by erosion and then sedimentation in the downstream reservoirs, state of equilibrium). In practice, sedimentation processes are also influenced by changes in time of inflow, water level (in the basin) and sediment supply, which results in an alternation of sedimentation, state of equilibrium and erosion of already deposited solids. While the transport behaviour of noncohesive sediment obeys the limit curves of the diagram by Shields (1936) or others, considerable changes in the critical erosion and sedimentation shear stresses may result for inorganic fine material with aggregation and flocculation properties. Thus, the sedimentation shear stress for fine sediment with a high mineral percentage in the clay fraction may be 100 times higher than suggested by the Hjulström criterion (1935) or even above the erosion limit. After settling, such fine sediment undergoes a major consolidation phase during which the erosion shear stress may increase by more than a factor of 10. In consequence, in reservoirs with a high inflow of inorganic and organic fine sediment, once this is consolidated, the sediment may fail to be carried along even by floods. Figure 2.2-1: Flow conditions and sedimentation process in a river sand trap (Mertens, 1987)

18 8 Reservoir Sedimentation Dead-Water Zones In running waters, the presence of water-retaining structures, groynes and training structures usually involves the formation of zones where flow discharge is very low. Such dead-water zones include groyne fields, dead river branches and harbours. They often develop into biotopes of ecological importance, which in the long run ought to be protected from detrimental sedimentation. Depending on size and configuration, groyne fields and harbours act as sediment traps where major amounts of sediment (usually suspended solids) accumulate. Where backwater reaches above barrages have flooded washlands, the coarse bedload material is deposited within the river channel, while the finer suspended solids settle on the washlands. Lateral transport of suspended sediment is greatly influenced by water exchange between main channel and washland. A cross-current from the main channel to the washland (e.g. washland widening) favours sedimentation on the washland, while a cross-current in the opposite direction (e.g. narrowing washland) reduces sediment deposition. Washlands with shallow water will tend to fill up with sediment over the years. In conjunction with vegetation, such areas may even end up again as dry land. Fine sediment may be resuspended during major flood flows. Convective and diffuse water exchange between the dead-water zone and the river is responsible for suspended sediment delivery, which is mainly a function of the size of the exchange area, flow velocity and suspended load concentration in the river as well as inner circulation determined by the shape and size of the swirl area. Of the suspended material supplied by the main stream, the coarser fraction is deposited, while the very fine particles are kept in suspension by turbulent movement, their residence in the dead-water zones thus being temporary. The relationship between the shape and size of dead-water zones of simple configuration, such as groyne fields, on the one hand, and water exchange on the other hand is as follows: the smaller the exchange coefficient, ε, (dispersion coefficient), the larger the half-life time, t 1/2, within which water exchange takes place. Unsteady water-level fluctuations (ship traffic, flood waves, tide), density flows (mainly in tidal waters) and turbulence-generating factors (ship propulsion) may substantially increase the delivery of suspended sediment to such deadwater zones. Ship traffic not only whirls up sediment and keeps it longer in suspension, but also produces substantial wave-induced cross-currents, which increase the lateral transport of suspended sediment and bedload towards the dead-water zones. Studies conducted on the River Main have demonstrated the influence of navigation on the transverse transport of suspended material towards near-bank water zones. Water exchange and, hence, suspended load delivery increases considerably under the influence of water-level fluctuations (Westrich, 1988).

19 Run-of-River Installations Case study Experience gathered in Reservoir Flushing on the River Drau in Austria Plant Description The River Drau between the Austrian border with Slovenia and almost up to the town of Spittal forms a practically continuous series of 10 power plants. Its total length is about 134 km with a total head of approximately 176 m. The individual heads all exceed 20 m at the larger dams. The damming stages have not been constructed by proceeding in one direction, but as dictated by economy. The damming stages with the high heads have usually wide reservoirs and are equipped with bay power stations, while those with the lower heads tend to have canal-shaped backwater reaches and pier power stations. The series of power stations is operated by Austrian Hydro Power (formerly Österreichische Draukraftwerke AG) Sediment Management It was decided at the outset that no coarse material that is, bedload should enter the reservoirs. This is mainly achieved by dredging in bedload traps and sedimentation basins where and when necessary. The sediment management strategies selected depend on the size of the respective reservoir. No de-sedimentation measures are considered necessary in minor reservoirs. Sedimentation is allowed to progress until the basin fills up. Floods will wash out the accumulated material without the need for previous water-level drawdown, so that even large floods may be routed through the reservoir without risking dam overtopping. By contrast, silting in the larger reservoirs is not allowed to exceed a certain level. When this is reached, the reservoirs must be flushed with the help of partial drawdown during discharges corresponding to not less than about 0.7 x HQ1. In addition, weir operation rules dictate what water level must be maintained, depending on the inflow. Water-level drawdown starts at approximately HQ1 and varies according to the degree of sedimentation and the respective reservoir. These strategies were already included in the application for the Water Right Permit and received approval by the Austrian Supreme Water Right Authority Reservoir Flushing Experiences Meanwhile the reservoirs have advanced in years, and this strategy which, as mentioned above, was made part of the design, has proved efficient. The uppermost of the wide reservoirs, Rosegg, has been flushed consistently for 20 years. In contrast to the 5.50 m drawdown provided in the Water Right Permit, the water level has been lowered by not more than 2.50 m from 1984 on. Hydraulic model analyses and extensive monitoring after flushing events have proved so far the smaller drawdown to be sufficient. In connection to reservoir Rosegg, partial water-level drawdown has also been practised at the downstream Feistritz and Edling developments over the past few years, but so far by no more than 1.50 m (Edling) and 2.50 m (Feistritz).

20 10 Reservoir Sedimentation The drawdown process is not started until the installed hydrological forecast model predicts an inflow larger than 700 m³/s. Water-level drawdown at the downstream reservoirs is controlled by their respective hydraulically determined drawdown plans and weir operation rules Discussion of Flushing Results as obtained at the Rosegg Dam Rosegg is characterised by a number of detail problems resulting from its specific features. Rosegg is a derivation-type power station, (located at the end of a river derivation where it discharges into the main river). That means that any surplus water runs through the original riverbed. Vegetation that has developed there "combs out" part of the passing suspended solids, especially during reservoir flushing. More than 1/2 Mio m 3 of material has thus been dredged and hauled from this river section since the plant was commissioned in Furthermore, the design provided for a hydraulic constriction in reservoir width by means of dyke structures. The meander cutoff provided to isolate a river bend upstream of the plant had to be improved in terms of sediment hydraulics after about 15 years of service. A large tributary, the River Gail, joins the Drau at the upstream end of the backwater reach. The Gail occasionally transports large amounts of sediment. Normally, this material is dredged out in a sediment trap directly above the junction of the two rivers. During major floods, however, some of the sediment finds its way to the Drau, and this must then be removed by means of a pontoon dredger. Villach Power Station Tailwater Erosion Meander cutoff Weir Derivation Headrace channel Rosegg Power Station Figure 2.3-1: Location map of the Rosegg power plant The erosion problems in the tailwater of the Villach power station, within the town of Villach, will not be discussed in greater detail here. Suffice it to mention that even minor floods have caused severe riverbed degradation, which has called for extensive stabilisation measures. By far the largest proportion of accumulated sediment is removed by reservoir flushing. Since 1991, with the use of the above mentioned partial drawdown processes which transport the sediments through the reservoir, an additional siltation of the Rosegg reservoir was avoided. These partial drawdown events have taken between a few hours and a few days, depending on the nature of the respective hydrological event. To assure the absolute safety

21 Run-of-River Installations 11 against floods according to the city of Villach, in 2006 and 2007 substantial dredging activities (800,000 m 3 ) have to be performed. To perform the desilting process in the future by means of flushing, substantial approving processes are necessesary in a medium term. Figure 2.3-2: Duration and erosion volume for different reservoir flushings The volume changes over the length of the reservoir for four events are shown below: Figure 2.3-3: Effects of reservoir flushing in autumn 1993, differences in area [m 2 ] before/ after flushing

22 12 Reservoir Sedimentation The accretions immediately upstream of the weir are certainly due to re-sedimentation after erosion during the desiltation according to the drawdown process (flushing). The above graph clearly suggests that this particular desiltation process was stopped relatively early. The whirled-up sediment accumulated and settled near the weir during reservoir refilling. In general, it may be assumed that the greater part of the fine sediment eroded during flushing tends to be carried along near the bottom. Measurements of suspended load concentration over the whole series of dams has shown that the amounts delivered to the chain are not much different from those discharged from the flushed reservoir and further downstream. Figure 2.3-4: Effects of reservoir flushing in autumn 1998, [m 2 ] before/ after flushing The desiltation processes represented above in autumn 1998 show degradation extending over the whole reservoir length up to the junction with the River Gail. As demonstrated by the graph, this is a result of the long flushing period and the relatively high peak discharge. By way of example, the following graphs demonstrate the changes in cross section profile resulting from reservoir flushing. Figure 2.3-5: Cross section changes in the Rosegg reservoir The change in cross section shown in Profile 2 is mainly a result of landfill measures having moved the channel line to the left.

23 Run-of-River Installations 13 Figure 2.3-6: Effects of reservoir flushing in Nov. 2004, differences in area [m 2 ] before/ after flushing By way of example, the following graphs demonstrate the changes in cross section profile resulting from reservoir flushing. Figure 2.3-7: Cross section changes

24 14 Reservoir Sedimentation Figure 2.3-8: Effects of reservoir flushing in Oct. 2005, differences in area [m 2 ] before/ after flushing By way of example, the following graphs demonstrate the changes in cross section profile resulting from reservoir flushing. Figure 2.3-9: Cross section changes, Profiles 2, 7 and 19/1 On the whole, the consistent desiltation strategy has ensured de-sedimentation. Failure to draw down the water surface to a lower level even during a short flood event has proved to result in substantial sedimentation. Consequently, the intention is to provide for adequate drawdown even during short events, so as to ensure the passage of incoming fine sediment Summary Consistent de-sedimentation management has been practised according to the desedimentation plan on the River Drau in Austria since This includes water-level drawdown of the Rosegg reservoir by 2.50 m, resulting in adequate suspended loads while achieving a sufficiently large washing-out effect in downstream reservoirs. The riverbed is thus maintained at a state of equilibrium within the permitted range of water-level variation, so as to comply with requirements of regulatory authorities. The water surface of the next downstream reservoirs Feistritz and Edling are also partial drawn down. Likewise, the progress of sedimentation will require these partial drawdowns to be extended to the next two reservoirs downstream, Ferlach-Maria Rain and Annabrücke.

25 Deep, seasonal storage reservoir 15 3 Deep, seasonal storage reservoir 3.1 Turbidity current - main cause of sediment transport in deep reservoirs If water of higher density (ρ w +Δρ) flows over a bottom slope into stagnant water, such as oceans, lakes and reservoirs with a smaller density ρ w, the inflow pushes the ambient water until it reaches a balance of forces, at which point the denser water sinks beneath the ambient water. This point is referred to as plunge point (see Figure 3.1-1). The parameters influencing the plunging behavior are slope, geometry, inflow and density distribution. The plunging phenomenon has been observed in the field and in the laboratory. The plunge phenomenon can be defined as the transitional flow from a homogeneous open channel flow to stratified, entraining underflow, which is commonly referred to as density current, or more specifically turbidity current. Plunge point Sediment-laden inflow Clear stagnant water Erodible sediment deposits Turbidity current Head Figure 3.1-1: Sediment-laden river entering a reservoir - plunging flow phenomenon and turbidity current formation. Turbidity currents are driven by the density difference in the turbidity current and the ambient water. The difference in density is caused by suspended fine solid material, which renders the flowing turbid water heavier than the ambient water above. To travel long distances, the velocity of a turbidity current must be sufficient to generate the turbulence required to maintain its sediment load in suspension, thereby maintaining the density difference between the gravity-induced current and the ambient water. During passage the turbidity current may unload or resuspend fine granular material. The sediment exchange at the bottom of the reservoir can be described by a flux between the bed and the current, separated into a sediment entrainment and a sediment deposition term evaluated at a reference height slightly above the real bed level (Parker et al., 1986, De Cesare, 1998). Suspended sediment is constantly falling out of the current at a rate given by the sediment fall velocity and the mean volumetric concentration of suspended particles near the bed. The motion of the turbidity current exerts a stress on the bed and is capable of entraining sediments from the bed into suspension. If the entrainment rate is less than the depositional rate, then the turbidity current experiences a net loss of granular material, and the sediment concentration in the current decreases. Consequently, the driving force acting on the current decreases, causing the current to decelerate and eventually to vanish. On the other hand, a higher flow velocity of the turbidity current can produce a rate of sediment entrainment from the bed that is greater than the depositional rate. The density of the current increases, and the turbidity current accelerates. As bed stress increases further and more sediment is entrained, a self-reinforcing cycle is created which allows the development

26 16 Reservoir Sedimentation of a self-sustaining turbidity current that can gradually reach high speeds. Only the availability of bed sediment for entrainment, the reservoir geometry or eventually the damping of turbulence at high concentrations will limit the growth of such a gravity-driven flow. When a turbidity current reaches a barrier such as a dam, the forward velocity is converted back into potential energy as the current rises up against the face of the dam, and subsequently falls back down, initiating the formation of a muddy layer. The top surface of this highly sediment charged layer will extend along a nearly horizontal profile upstream from the dam. The volume of the muddy layer will increase and the interface will rise, as long as turbid inflow exceeds losses by the upward seepage of clear water from within the muddy layer due to sedimentation and compaction of solids. Sedimentation within the highly sediment charged layer causes the fluid density to increase at the bottom of the layer. Subsequent turbid inflows will spread across the top of the higher density fluid in the sediment charged layer rather than mix. The inflowing density current is thus converted from underflow to interflow in areas affected by muddy lake accumulation. With time the muddy lake creates a horizontal deposit of fine-grained material extending upstream from the dam, which typically consists of material significantly finer than the delta deposits Propagation of sediments to dam Depending upon the sediment supply from the watershed and flow intensity in terms of velocity and turbulence, river flows usually carry sediment particles within a wide range of sizes. Bed load consists on the coarser components which are transported near the river bed. The suspended sediments are generated by superficial erosion as well as by smashing and abrasion of coarser components. In the events of floods the fraction of sediments smaller than sand reaches 80 to 90 % of the total sediment carried by the river. When a river flows into a reservoir, the coarser particles deposit gradually and form a delta in the headwater area of the reservoir that extends further into the reservoir as deposition continues. Finer particles, being suspended, flow through the delta stream and pass the lip point of the delta, first entering a quasi-homogeneous (non-stratified) flow region and subsequently being deposited along the path due to a decrease in flow velocity caused by the increased cross-sectional area (Figure 3.1-3). The quasi-homogeneous flow is shorter in cases of smaller discharge and/or higher sediment concentration. On the contrary, the region of quasi-homogeneous flow is longer in cases of larger discharge and/or lower sediment concentration. As a consequence close to the dam the deposition of the finest particles takes place.

27 Deep, seasonal storage reservoir 17 Figure 3.1-2: Areas affected by sedimentation in the surroundings of a reservoir Due to bigger erosive forces and transport capacities with increasing discharge, the sediment content in a river gets especially high during floods. The flow with high concentration on sediments is carried into the artificial lake by means of the turbidity currents. This phenomenon takes place several times a year. The sediment-laden river, driven by a density difference, plunges into the reservoir, where it follows the thalweg of the lake to the deepest area, which is normally close to the dam. It can move for several kilometres (Figure 3.1-2). Average Mittlere velocity Geschwindigkeit Max. velocity Geschwindigkeit Geschwindigkeit (m/s) Flow Velocity Figure 3.1-3: Maximal transportable grain sizes dependent on the flow velocity of the turbidity current according to Fan (1986)

28 18 Reservoir Sedimentation 3.2 Theoretical Background Turbidity currents are flows driven by density differences caused by suspended fine solid material in an ambient fluid. They can appear as different forms, depending on the density of the mixed (water/sediments) fluid. If the density of the mixed fluid (ρ m ) is superior to the density of the ambient fluid (ρ w ), the turbidity current will be formed on the bottom. If the density of the mixed fluid is less than the ambient fluid, the current will be propagated at the surface. Sometimes, the reservoir presents a stratification of density, and the mixed fluid can flow like as an intrusion. Generally, it is the first type of current which can transport the largest quantities of sediments downstream. ρ m ρ ρ w1 w ρ w ρ m a) ρ m > ρ w b) ρ m < ρ w c) ρ w1 <ρ m <ρ w2 ρ m ρ w2 Figure 3.2-1: Types of turbidity currents Turbidity currents consist of two regions of a steady flow, one region dominated by the momentum and another dominated by the buoyancy. For the currents at the bottom, at a position determined by the balance between the momentum of the inflow and the baroclinic pressure resulting from the density difference between the inflow and the receiving water, the inflow plunges below the surface. The physical parameters that control the plunging depend on inflow conditions as well as the geometrical characteristics of the reservoir. For a channel with a constant width and lateral slopes of the trapezoidal section (m) between 0.2 and 0.8, Savage and Brimberg (1975) propose a depth of plunging (h p ) given by h p 1 = F 2 p q 0 0g ε 1 3 Eq where q 0 is the initial inflow per unit width, ε 0 is a relative density difference between inflow (mixed fluid) and ambient fluid, g is the gravitational acceleration and F p is the densimetric Froude number at the plunge point. The density of the mix fluid can be expressed by ρ m = ρ ρ ρ )C Eq w ( s w where C is the volumetric concentration. The term F p, after Savage and Brimberg (1975) is empirically defined by F p S = 1 m C Eq d In this formula S is the bottom slope and C d a total friction coefficient of the underflow (0.01 < C d < 0.09).

29 Deep, seasonal storage reservoir 19 After plunging, turbidity currents normally can be separated in 2 parts. The front or head, which has as driving force essentially the pressure gradient due to the density difference between the front and the ambient fluid ahead of it and the body, with the driving force the gravitational force of the heavier fluid (mixed water/sediment). The flow in the front is unsteady while for the body, the flow can be considered as being steady. In turbidity currents, the quantity of the suspended sediment is not conserved and it is free to exchange with the bed sediment by means of bed erosion and deposition. It can cause selfacceleration of the turbidity by entrainment of bed sediment. Figure illustrates the typical turbidity current. Figure 3.2-2: Turbidity current flowing at the bottom The current moves in the longitudinal direction x, over a bottom slope S, and under a deep layer (H>>h) of an ambient stagnant fluid (U a o) with a density ρ w. This density is less than the density of the turbidity current, ρ w <ρ m. Altinakar et al. (1996) compared the body of the current to a wall jet with 2 regions (Figure 3.2-3), the wall and the jet region. The height h max attended where the velocity is maximum, u=u max, is the separation of these regions. Water entrainment Jet region z/h max z = h max Maximum velocity Wall region Erosion / Deposition Figure 3.2-3: Schematic dimensionless velocity profile for turbidity currents (Graf and Altinakar, 1996)

30 20 Reservoir Sedimentation Wall region : z<h max, turbulence is created at the wall and entrainment of sediments can take place; In the free region, z>h max, turbulence is created by friction and by entrainment of the ambient fluid. Experimentally they found that U ax h max 0.3 h = m = 1. U 3 h t = 1. 3 Eq h In these expressions, h is the average height of the current and h t is the height of the current at which u 0. A simple hydraulic approach using the Chezy-type relationship between the front velocity (U f ) and height of the front (H f ) can be described after Turner (1973) for a large range of slopes and roughness as U = l g' H Eq f 1 f where l 1 is a constant that can vary between 0.63 (Altinakar et al.,1990) and 0.75 (Middleton,1966 and Turner,1979), being 0.63 more appropriate for small slopes and g is a reduced gravitational acceleration expressed as ρ = m ρw g' g ρw Eq After Britter and Linden (1980), the front velocity can also be expressed by f U = l B Eq where l 2 is a constant depending on the bottom slope and the Reynolds number (Re). For slopes less than 5%, Altinakar et al. (1990) found values varying linearly between 0.7 and 1.0. Choi and Garcia (1995) proposed l 2 =1, based on numerical experiments and laboratory data of Altinakar et al. (1990). BB0 is the initial buoyancy flux per unit width and can be defined as B = 0 = g' 0 h0u 0 g' 0 q 0 Eq with g 0 as initial reduced gravitational acceleration, h 0 is the initial depth and U 0 is the initial velocity. For the description in a simple model, the body of a turbidity current can be considered twodimensional and plane, and the flow turbulent and incompressible. The velocity of the body is normally greater than the velocity of the head. This difference grows as the angle of the bottom slope increases and in order to maintain the flow continuity, the height of front will always be larger than the height of the body. The height of the front increases with entrainment of ambient fluid depending on the distance covered. After Parker et al. (1986), the description of the flow in the turbidity currents with entrainment can be based on a conventional three-equation model, conservation of mean

31 Deep, seasonal storage reservoir 21 momentum, fluid mass and sediment. This model is an extension of the unidirectional steady gravidity flows developed by Ellison and Turner (1959). conservation of the mean momentum d dt d 2 1 d 2 2 (Uh) + (U h) = Rg (Csh ) + grcshs u*b Eq dx 2 dx where t is time, R is the submerged specific density, (ρ s ρ w )-1, and u *b is the shear velocity, that can be defined as, u sb = α k Eq k with α k as a dimensionless constant and k the mean kinetic energy of turbulence. The conservation of fluid mass (mixed water/sediment) dh dt + d (Uh) dx = W h Eq where W h is the entrainment velocity of the ambient fluid into the current often assumed to be proportional to the velocity of the turbidity current, U. W h = E W U Eq with the constant of proportionality, E w, being the entrainment coefficient of ambient fluid and depending on the Richardson number, Ri, that represents the ratio of inertia to reduced gravity forces. g' cosα 1 Ri = = 2 U Fr 2 d Eq with Fr d as a densimetric Froude number. The parameter E w is given after empirical relation as 2.4 E W = 0.075( Ri ) 0.5 Eq Conservation of mass of sediment d d (Csh) + (CsUh) = SE SD Eq dt dx where S E is the sediment erosion and S D is the sediment deposition. The equation of continuity for the solid phase in a steady two-dimensional case approximates the equation of diffusion of granular material. (uc x s ) (wc + z s 2 cs 2 ) cs = υss + ε s Eq z z

32 22 Reservoir Sedimentation In this formula, the horizontal and vertical velocities are designed by u and w. The local volumetric sediment concentration is designed by c while υ ss represents the settling velocity and it can be calculated with different methods. For very fine particles the Stokes equation (Graf, 1971) can be used υ ss ρs ρw 1 2 = g d Eq ρ 18ν w with d as the sediment particle diameter and ν the kinematic viscosity of water. The ε s is the diffusion term and can be expressed by, 2 cs ε s (c' s w' s ) Eq z z The terms w and c represents the fluctuations of the vertical velocity and volumetric sediment concentration. The term c' s w' expresses the vertical Reynolds flux of sediments. The sediment s mass exchanges at the bottom can be found integrating vertically the expression ( c' s w' s ) over the entire height (h t ) and analyzed close to the bottom at z = b = 0.05 h t. Thus, S E = (c' s w' s ) z= b = υ E ss s Eq This represents the erosion of sediments per unit area. E s is the entrainment coefficient of sediments from the bed and can be found after an empirical relation, E s 7 5 Zu 7 5 Zu 1.3 * 10 = Eq * 10 Where Z u is a dimensionless parameter defined by Z u u* = υ b ss f ( ), with f ( Re ) Re p = Re 0.6 Re p p p Re Re p p 3.5 < 3.5 Eq The term Re p represents the particle Reynolds number and is based on the characteristics of the characteristic sediment size, D s, and the kinematic viscosity of water. RgDs Ds Re p = Eq υ For the sediment deposition term, expressed by deposition of sediments per unit area on the bed, S D = υ sscs z = b = υ c ss s b Eq where c sb is the local volumetric sediment concentration at z=b. The numerical solution for this three-equation model under certain initial conditions leads to high acceleration of the current and high consumation of the turbulent energy, more than

33 Deep, seasonal storage reservoir 23 available. This led to introducing a fourth equation in which the turbulence production rate is specifically dealt with through the conservation of the mean value of kinetic energy of turbulence. d dt d (Kh) + (KUh) = u* b + U Ew 0 h RgCshυss RgChEw Rghυss (EsE S csb ) Eq dx where K is the mean kinetic energy of turbulence and 0 represents the mean rate of turbulent kinetic energy dissipation due to viscosity. Parker et al. (1986) proposed 3 1 Cd* K 2 Ew (1 Ri 2 ) + Cd 0 = β with β = 2 a Eq h 3 Cd 2 a where, β is a volume porosity and a is a reference level for equilibrium concentration. The reference concentration c sb is calculated close to the bed at b = 0.05h t Flow over an obstacle When a turbidity current meets an obstacle some of the denser fluid may flow over the obstacle while a hydraulic jump or bore travels upstream as shown in Figure Sometimes, it can be partially or totally blocked by the obstacle. In this figure, there is a fluid traveling with constant depth h 1 and velocity U 1 and a moving hydraulic jump propagating upstream with the velocity U j as well as the conjugated depth h 2 with a smaller velocity U 2. The obstacle has a height h m. Figure 3.2-4: Flow over an obstacle (Oehy, 2003) The flow over an obstacle is characterized by two independent variables: the densimetric Froude number of the approaching flow Fr d1 and the ratio H m between the obstacle height h m and the height of the incoming flow h 1. If motion of the upper fluid can be ignored (U a 0), the flow has many features in common with free-surface flow and the shallow-water approximation is valid (Rottman et al., 1985). Neglecting the water entrainment and the friction (E w = 0; u *b = 0), as well as the erosion and deposition, the inflow can be described by

34 24 Reservoir Sedimentation and dh d + (Uh) = 0 Eq dt dx du d d + U U + g' (hm + h) = 0 Eq dt dx dx Steady solutions to these equations are sought by setting the partial derivatives with respect to time equal to zero. The equations then reduce to and Uh= U 1 h 1 Eq U U + hm + h = + h1 Eq g' 2g' With U 1 and h 1 defined as the upstream values of U and h. It can be shown that for each value of Fr d1, a limit H m =H mc exists at which the flow on the crest becomes critical. Applying the Bernoulli s equation between sections 1 and 3 (Figure 3.2-4), this limit can be found by H mc 2 d1 Fr 3 2 = 1 + Frd1 3 Eq In order to define limits for the flows over an obstacle, equation Eq and the next two can be combined, as shown is Figure given by Baines (1995) and 2 2 H m + 1 Fr d1 = (Hm 1) Eq Hm H m 2 d1 2 d1 3 2 (8Fr + 1) = Frd1 3 Eq Fr 4 2 if considered that U j =0 (Long, 1970).

35 Deep, seasonal storage reservoir 25 Figure 3.2-5: Flow regimes of shallow-layer flow over an obstacle (Oehy, 2003) In Figure 3.2-5, steady flow is only possible in the left part of the curve from the equation Eq In the regions B and C, the flow is supercritical whereas in region A, subcritical. In region D, the flow will be partially blocked when H m is increased above H mc. In this case, an internal bore, i.e. a moving hydraulic jump is then formed and propagates upstream. This internal bore dissipates energy in order to match this steady solution to the upstream flow condition. For the calculation of the hydraulic jump, the Belanger s equation for single phase open-channel flow can be utilized, considering the densimetric Froude number. h2 2 h 1 1 = 1 + 8Frd1 1 2 Eq In the region E, the obstacle is high enough to block the approaching flow completely. Figure 3.2-6: Flow over an obstacle in a laboratory flume (Oehy, 2003)

36 26 Reservoir Sedimentation Flow through a screen The interaction of a gravity current with a screen, after the transient effects have disappeared, can be solved by a similar method to that used in the case for the obstacle, as shown in Figure and depends mainly on its porosity. Since the screen is permeable, the current does not climb as high as in case of an obstacle (Rottman et al., 1985). Figure 3.2-7: Flow through a screen (Oehy, 2003) The theory of the flow through a screen presented in the following is presented in the thesis document of Oehy (2003). In these analyses, the deposition was not considered. In Figure 3.2-7, h 1 represents a constant upstream flow depth with a velocity U 1, U j the upstream propagation velocity of a hydraulic jump, h 2 and U 2 the depth and the velocity of the current immediately upstream of the screen and h 3 and U 3 the conditions of the current, depth and velocity downstream of the screen. Also in this figure, f represents the porosity of the screen. The ratio H p between down- and upstream depth of the screen can be determined as a function of the porosity as shown in Figure Figure 3.2-8: Ratio of heights down- and upstream of the screen, H p =h 3 /h 2, as function of the effective porosity, f (Oehy, 2003)

37 Deep, seasonal storage reservoir 27 The relationship H j between h 2 and h 1 can be found as a function of the porosity of the screen and Fr d1 and is illustrated in Figure Figure 3.2-9: H j =h 2 /h 1 as function of the porosity f and the upstream Fr d1 (Oehy, 2003) The proportion η of the flow which continues over the screen is illustrated in Figure also as a function of the porosity of the screen and the upstream Fr d1. Figure : Proportion of the incoming flow that is predicted to continue through the screen as a function of the effective porosity f and the upstream Fr d1 (Oehy, 2003) This proportion can be calculated by U2h2 η = Eq U h1 1h 1 The force F exerted on the screen can easily be calculated by

38 28 Reservoir Sedimentation 1 f 2 2 F = ( )(1 f ) ρg' h2 Eq f Figure : Flow through a screen in a laboratory flume (Oehy, 2003) 3.3 Experimental Studies Turbidity currents are flows driven by density differences which are caused by fine solid material suspended in the fluid. They belong to the family of sediment gravity currents. These are flows of water laden with sediments that plunge in a mass of stagnant water such as an ocean, a lake or a reservoir. The turbidity currents have been observed by experimental studies Experimental set-up at EPFL-LCH (De Cesare, 1998) Figure shows the general schematic view of the flume, two adjacent mixing and storing tanks and the measuring equipment. The flume used in this investigation is 8.4 m long, 1.5 m wide and 65 cm deep. On the bottom a 6 m PVC plate is placed, which makes it possible to vary the slope between 0 and 6%. Water and sediments are mixed in a separate tank (2 m 3 ) by a propeller-type mixer. This tank is connected to an upstream tank by a recirculation pump. The turbid water returns to the mixing tank over a free surface weir that controls the water level in the upstream tank. A gate with variable width and opening allows the controlled release of the turbidity current into the flume.

39 Deep, seasonal storage reservoir v = v ( t, y ) Figure 3.3-1: Schematic drawing of the experimental installation : 1) mixing tank, 2) upstream tank, 3) recirculation pump, 4) free surface weir, 5) inflow gate, 6) turbidity current, 7) experimental flume, 8) ultrasonic probes, 9) sharp crested weir, 10) flexible duct, 11) UVP instrument, 12) control computer Before passing the open gate, water is constrained to flow through a set of 20 cm long horizontal tubes, with increasing diameter from the base to the top of the opening. This set-up is mainly used so that there is a one directional horizontal velocity field at the open gate. Thus, when the turbidity current enters the experimental flume it is perfectly perpendicular. The arrangement of more than 40 tubes gives an almost uniform velocity over the total height of the gate. A reinforced PVC wall with a sharp crested weir that controls the water level in the flume closes the downstream end of the flume. All the experiments were conducted with fine homogenous clay as suspended matter. The density of the sediments is ρ s = 2'740 kg/m 3. The particle size distribution ranges from d 10 = mm to d 90 = 0.1 mm, with a mean particle diameter of d 50 = 0.02 mm. The corresponding settling velocity calculated using Stokes law is v ss 0.4 mm/s for the representative particle size in calm water. In all experiments the clear water from the main reservoir of the hydraulic laboratory was used as the ambient fluid. The water-sediment mixtures were prepared in the mixing tank by adding the dry clay to the clear water. The density of the water-sediment mixture, ρ m varied between 1'002 and 1'005 kg/m 3, and the mixture was considered to be a Newtonian Fluid. The measuring section is located above the PVC bottom. This allows the monitoring of the spreading turbidity current and its uniform flow over the total width of the flume. The turbidity currents in the laboratory were monitored by means of ultrasound probes functioning with the Doppler Method (UVP), thus giving a complete velocity profile along the ultrasound beam in a very short time. The spatial resolution is smaller than a millimetre and it takes less than one tenth of a second to measure and compute a whole profile (Takeda, 1995). Measurements were made in the flume with three different configurations of the UVP transducers. Vertical and frontal velocity profiles were taken, and 2D flow mapping close to the bottom was performed. Because the UVP instrument allows only one transducer to be connected at a time, the eight transducers used in the experimental set-up were connected to the UVP via a multiplexing unit.

40 30 Reservoir Sedimentation The beam directions and the penetration length were chosen in order to cover the interior of the advancing turbidity current. As the model is symmetric, the profiles were taken on the axis of symmetry and the flow-mapping region was situated on one side of the flume only. The arrangement of the transducers are described as follows: Vertical arrangement with 8 transducers looking with an angle of 60 against the main flow. The measurements give the projected vertical velocity profiles over 2 m flow length from the gate, the distance between the transducers was 25 cm. α Figure 3.3-2: Arrangement with 8 transducers looking with an angle of 60 against the main flow Axial disposition with 8 transducers looking straight against the main flow. The measurements give the horizontal velocity profiles over the whole 3 m flow length from the gate; the distance between the transducers was 50 cm. The probes were installed 12 mm above the bottom and were slightly set off laterally in order to reduce interference by reflection of US signal from different transducers. Figure 3.3-3: Axial disposition with 8 transducers looking straight against the main flow Square grouping with 4 transducers on each side looking straight at and perpendicular to the main flow in the spreading part just after the inflow gate. The side of the square plane where the flow mapping took place was 62.5 cm long, the distance between transducers was 12.5 cm close to the inflow gate, and 25 cm elsewhere. The transducers were installed with plastic clamps 12 mm above the bottom on an aluminium frame.

41 Deep, seasonal storage reservoir 31 Figure 3.3-4: Square grouping with 4 transducers on each side looking straight at and perpendicular to the main flow in the spreading part just after the inflow gate The echo of the flowing turbidity current was strong enough to allow rapid measurements, only 4 successive profiles were taken with each transducer, and the averaged profile is used as velocity profile at one location. The temporal resolution was therefore less than ½ second per profile. The duration to sweep all transducers was around 3 seconds and the cycle was repeated every 5 seconds, thus giving a quasi-instantaneous velocity information every 5 seconds. The vertical profiles were obtained with 8 measurements per profile, thus doubling cycle time; repetitions were made every 10 seconds Experimental results Figure shows a photograph of the spreading turbidity current 25 seconds after its initial release. After its spreading to the total flume width, the current adjusted itself rapidly to a uniform flow advancing steadily within the tank. When the current reached the downstream end of the flume, the turbid water was evacuated by opening the bottom gate. During the total duration of the experiment a constant turbid water flux was maintained through the inflow gate. Figure 3.3-5: Expanding turbidity current in the experimental flume 25 s after opening of the gate; the current spreads out almost radially, large eddies developing at the current front; these eddies give the characteristic surface appearance of turbidity currents like clouds, 125 mm x 125 mm grid on PVC bottom. Turbidity currents can be separated into two characteristic regions of flow behaviour:

42 32 Reservoir Sedimentation A bottom surface layer region, where turbulence is created from the bottom surface roughness and where sediment entrainment may occur. The velocity distribution is assumed to be logarithmic. A jet region above the nose, where main turbulence comes from mixing with surrounding clear water. The presumed velocity distribution is half the normal (Gaussian) distribution. The separation of these two behaviours is located at the point of maximum velocity, often referred to as the nose of the turbidity current head. The measured velocity profiles were compared with the distributions described above; the result fits well as shown in Figure z / h m jet region Gaussian distribution z = h m bottom layer region logarithmic distribution u / U m Figure 3.3-6: Measured velocity values compared with theoretical vertical velocity distribution, u(z). Values from run n 2, 80 ms between two succeeding measurements. The measured velocity profiles were compared with a theoretical distribution; the result agrees well as shown in Figure 3.3-6: A bottom surface layer region, where the velocity distribution is logarithmic, and a jet region, where the velocity is the half Normal (Gaussian) distribution Experimental set-up at EPFL-LCH (Oehy, 2003) Flume description The experiments were carried out in a modified multi-purpose tilting flume with a total length of 8.55 m. A second flume was inserted into the multi-purpose flume, and the sides of the inner flume consisted of 8 mm thick transparent PVC plate to allow the observation of the underflow. The inner experimental flume had a section of 27.2 cm in width and 90.0 cm in depth. The flume can be tilted in a slope range S between 0% and +5% (2.86 ). The entire flume was divided into two sections of unequal length by a sliding gate. Adjacent to the experimental flume, a mixing tank with a maximum capacity of 1,500 liters was used to prepare and store the dense fluid mixture. The mixing tank was equipped with a propellertype mixer to create a homogeneous sediment mixture. After filling the experimental flume

43 Deep, seasonal storage reservoir 33 with tap water, the dense fluid (water-sediment mixture) was pumped up into the stilling box through a 60 mm diameter flexible plastic pipe, passing a calibrated electromagnetic flowmeter. The water-sediment mixture entered the stilling box through the bottom turning horizontally behind the entrance gate into a slotted pipe. During preparation the sluice gate was closed and the flow returned through a 30 mm orifice and a PVP pipe back into the mixture tank. This circulation ensured a uniform mixture in the stilling tank and an accurate regulation of the pump. The shorter upstream section served as a stilling box and head tank for the mixture to be released by opening the sliding gate to create the turbidity current. The movement of the sliding gate was controlled by a lever mounted on the flume wall. Downstream of this gate a box containing rectangular tubes of 1.5 cm diameter reduced the scale of the initial turbulence. The long downstream section simulates a reservoir in which a turbidity current is flowing. A compartment at the end of the flume bottom trapped the turbidity current for withdrawal. The rate of the outflow is controlled by a drainage valve where a rotameter allowed the regulation of the water level. To prevent a contamination of the hydraulic system of the laboratory with fine sediment, a filter was installed to retain the sediment particles Properties of the sediment materials Various materials can be used to create turbidity currents in the laboratory. They include different types of clay, bentonite, quartz or marble powder obtained by grinding. For the present study a cohesionless, light weight homogeneous material was chosen, which had a well known and relatively narrow particle size distribution, as well as a small settling velocity. Specifically, the material was ground polymer with a density of 1135 kg/m 3. The grain size distribution of the sediment material was determined with a Mavern Mastersizer (Laser refraction method). The material had a fairly narrow grain size range, but the frequency histogram was skewed towards large grain sizes, which is typical for ground particles. The particles have a median diameter, d 50, of 90 μm. With a standard deviation, σ g, of 1.6 the grain size distribution cannot be considered uniform, therefore, some grain sorting effects may occur. The selection of a characteristic grain size of the sediment is important for the calculation of the representative settling velocity of the sediment, which greatly affects the sediment transport. Altinakar (1988) proposed that the grain size which has a settling velocity equal to the average settling velocity of the sediment material to be the representative particle size for the sediment material. The settling velocity, v ss, can be calculated by means of different methods described in the literature. Here, Stokes' law has been used, considering the small, almost spherical particles and the low Reynolds number, Re=dv ss /ν < 0.2. It is expressed as v ss ρ ρ 1 g d ρ 18υ s w 2 = Eq w To prepare the water-sediment mixtures the particles first had to be wetted in a small pot with a mixer before they were added to the mixing tank. Due to the low sediment concentrations of up to 5% by volume, no corrections for the concentration are made on the settling fall velocity and the viscosity of the water. The mixture is considered to be a Newtonian fluid.

44 34 Reservoir Sedimentation Measuring Instrumentation Flow velocity measurements An ultrasonic velocity profiler (UVP) was used to measure flow velocities. With this device an instantaneous velocity profile in a liquid flow can be measured along the ultrasonic beam axis Front Velocity and Time Measurements The front velocity of the head of the turbidity current is determined from video recordings by reading the time at which the head passes predetermined positions. In turbidity current experiments, the interface between the current front and the ambient fluid could be easily observed. Since the camera was only 4 m away from the flume, a parallax correction was made; the actual positions of the head were measured from scales drawn on both the front and back walls of the flume. Unfortunately, no measurements could be taken on the first 2 meters because of the opaque wall of the multi-purpose flume. A digital chronometer was used to link the internal clock of the digital camera with the start of the experiment. The frame rate of the camera is 25 frames per second and the screen resolution is 720 x 576. The digital video records were analyzed to determine the position of the turbidity current head as a function of time. Generally, 35 to 40 measuring points were taken. In all the experiments the front velocity was nearly constant over the whole length, and was determined from a leastsquare fit. The relative change of the front velocity measurements was less than 1%. At the end of each experiment the watch was stopped with the closing of the gate and the duration of the experiment was read from the chronometer Density and Temperature Measurements In all experiments clear tap water was used as the ambient fluid. During the experimental work it was observed that the temperature of the water shows diurnal variations. The density of the sediment mixture and the clear water was measured before and after the experiment by means of a hydrometer, which allowed determination of the density to a precision of ± 0.1 g/l. Care was taken to obtain a turbidity current mixture of approximately the same temperature as the clear water in the flume, such that the density difference between turbidity current mixture and the ambient fluid was only due to the presence of sediment. Prior to the experiment the temperatures of the mixture prepared in the mixing tank and the temperature of the clear water in the flume were measured using a mercury thermometer with a resolution of ± 0.1 C Auxiliary Measurements The flowmeter of the inflowing pump is monitored during the experiments, which allowed checking the stability of the inflow. During each experiment the depths of the free surface at the upstream end of the flume, and in the stilling tank were measured by means of ultrasonic level probes. This allowed a verification that the boundary conditions were constant during the experiments Deposition Measurements It is greatly desirable to measure the deposition thickness during physical experiments of turbidity currents. Most of the experimental techniques described in the literature are designed to measure the deposition after the experiment, but give no information on the variation in

45 Deep, seasonal storage reservoir 35 time (Bonnecaze, 1993, Garcia, 1994, Altinakar, 1988). They often used a suction method, by which they siphoned up the particles deposited on a known area at several locations in the tank, dried them and weighted the samples. The advantage of this method is that by means of a sieve analysis the sample can be analyzed and the segregation of the particles can be investigated, see Altinakar (1988) for example. In the present work, a new device to measure the local sediment layer thickness was developed based on work done by De Rooij et al. (1999). The technique is based on the fact that the electrical resistance of a layer of particles depends on its thickness. In the present experiments a polymer powder was used. The electrical resistance of this material is much higher than that of clear water, which implies that the current passes through liquid phase, when one applies a voltage. When a large volume fraction is occupied by particles, the resistance increases and the current is reduced. These high concentrations are present in the sediment layer that accumulates on the bottom. The thickness of the layer can thus be determined by measuring its resistance. It is important to note that the particle size and shape affect the packing density, and therefore the resistance of the layer. The apparatus thus has always to be calibrated for the specific particles used Experimental Procedure Mixture and Flume Preparation Prior to each experiment, the mixing tank and the flume were cleaned and then filled with tap water. The mixing tank was filled from the flume through the pipe system which allowed flushing and cleaning of the system from sediment depositions. After this, sediment was added to the mixing tank until the desired fluid density was achieved. The mixer guaranteed a homogenous mixture in the tank. Difficulties were encountered wetting the relatively light sediment. A kitchen blender was used to prepare a highly concentrated mixture, which then was added to the mixing tank. Once the dense fluid was ready, it was recirculated with a pump through the stilling box. The electromagnetic flowmeter and a valve were then used to adjust the desired inflow rate for a given experiment. Standard mercury thermometers were used to measure fluid temperatures in the mixing tank and in the flume. This was done to ensure that the only source of density difference was the sediment. A temperature difference of ~1 C corresponds to a density difference in the order of g/cm 3, which is only 2% of the density difference generated during the experiments. This estimate indicated that the difference in water temperatures would have to be considerable to have any significant influence on the dynamics of the underflow. The temperature differences between clear water and the injected mixture was at most ~2.9 C and the difference in the density due to this difference in temperature was less than ~ g/cm 3. The density of the mixture was measured by means of a hydrometer when the installation was ready for the experiment. At the end of the experiment a second density measurement was performed to verify that no change in density had occurred during the experiment. Generally, the density of the mixture slightly decreased, by at most 5 %, during the experiment due to the sedimentation in the inflow tank. The average density was then retained. As the preparation of the mixture and flume was always carried out immediately before the experiment, the change in temperature due to the mechanical energy supplied by the mixer was negligible and only one temperature measurement was necessary.

46 36 Reservoir Sedimentation Preparation of Measuring Instruments UVP probes were placed at selected measuring stations at defined inclination angles. For the deposition measurement the reference electrode was put in place and an initial measurement of the resistivity of the clear water was carried out. All the runs were filmed with a digital video camera to view the experiment, and determine the propagation of the front and reflected bores Experimental Run Generally, each run involved the participation of two people, one in charge of the opening of the inflow sluice and the other one in charge of the drainage valve. In the case of water or air injection a third person was needed to manipulate the valve for dye injection. A typical run was started when the water levels in both sections of the flume were equal to each other. The sliding gate was slowly opened to a predetermined height of ~4.5 cm to let the mixture enter into the flume. Simultaneously, the valve for the return flow was closed and a sediment-laden bottom current was delivered at a specified discharge. The level of the water in the flume was kept at a constant level by means of the drainage valve, which also released the turbid water. The drainage discharge was kept close to the inflow discharge to minimize the variation of the water surface in the flume. The variations stayed within 1-2 cm for a water depth of 70 cm during an experiment of 10 minutes duration. Hence, no adverse influence of these manipulations of the flow on the general behaviour of the current was observed. Automatic data acquisition started before the gate was opened. The digital video camera recording was started after a measured time delay at the moment when the turbidity current became visible in the test section. The head of the turbidity current was then followed to later determine its velocity. A chronometer time measurement allowed the time calibration between the different measurements. When the underflow reached the end of the flume, a reflected wave started to propagate upstream. For the continuously-fed turbidity current experiments, the inflow sliding gate was closed at the moment when the head reached the downstream end of the flume. For the other series the experiment was stopped when the reflected wave reached the test section. The maximum acquisition time of the UVP and deposition measurement device was 600 s. Typically, the experiments lasted about 5 to 10 minutes. 3.4 Numerical Modelling Turbidity currents have been observed in artificial reservoirs (De Cesare, 1998 and De Cesare et al., 2001) and have extensively been studied in the laboratory (Altinakar, 1988 and Garcia, 1993). Numerical simulations of turbidity currents were performed by means of 2D depthaveraged formulations which were solved by finite-volume or finite-element methods (Bradford and Katopodes, 1999 and Choi, 1999). The conservation equations for mass and momentum of the turbidity currents were solved, where the mixing between the turbidity current and the ambient water was taken into account by introducing a water entrainment factor depending on the Richardson number. Only recently, some authors have considered the problem by solving the whole set of the 3D Navier-Stokes equations using some kind of turbulence model to close the system (De Cesare et al., 2001 and Olsen and Tesakar, 1995). A numerical two-phase flow model has been developed and applied to simulate river-induced turbidity currents in real reservoir geometry. The numerical model used is based on the general Navier-Stokes solver code CFX from Computational Fluid Dynamics Services (CFX- 4.2, 1997). It has been validated with laboratory experiments and compared with in-situ

47 Deep, seasonal storage reservoir 37 measurements of turbidity currents in a reservoir collected during two summers (De Cesare, 1998, De Cesare et al., 2001). The CFX (Computational Fluid Dynamics) code is based on a conservative finite-volume method. It provides a numerical solution for the Navier-Stokes equations in three dimensions and offers several possibilities for their extension. All variables defined are introduced at the centre of the control volumes (non-staggered grid) which fill the physical domain considered. Each equation is integrated over each control volume to obtain a discrete equation that connects the variable at the center of the control volume with its neighbors. The program uses a non-orthogonal, structured multi-block grid. The coordinate system and the computational grid are entirely distinctive entities. The coordinates in the physical space are often referred to as a body-fitted coordinate system. After a Jacobian matrix coordinate transforming applied to the physical space, the converted space is exactly rectangular with a uniformly distributed finite-difference grid. It is also referred to as the computational space, as it is in this space in which the equations are solved. Suspension is directly treated in the CFX code through its homogeneous two-phase flow advection-diffusion model with a continuous Eulerian description. Separate mass balances for each phase are solved, but a common momentum balance for the mixture takes into account the changing mixture density due to the different volume fractions. The following assumptions apply: mass and momentum transfers from one phase to the other are negligible, the process is isothermal, phases are mixed on length scales larger than the molecular length scales and smaller than the numerical grid, water is defined as a continuous phase occupying connected regions of space, sediment particles are defined as a solid disperse phase distributed uniformly in a control volume, velocity differences between particles and fluid are negligible; hence one single flow field is calculated. User defined sediment deposition and erosion modules were added to take into account sedimentation and the interaction between the turbidity current and the reservoir bed material (De Cesare, 1998, De Cesare et al., 2001). The description of the sediment exchange between the turbidity current and the bed follows the ideas of Parker et al. (1986). For the inflow boundary condition a generic Maxwell distribution is used for both discharge and concentration evolution in time. A calculation time of one hour has been used for all the results discussed in this paper. Boundary conditions must be set for each variable at every boundary on the domain created. The boundary conditions and the information on the erosion-sedimentation interaction model are stored in dummy cells glued around the computation domain. The finite volume is invariable in time, since the amplitude of the newly formed local sediment depositions and erosion is negligible compared to flow depth. The reservoir bottom interactions, which lead to sediment transport within the reservoir were analyzed with erosion-deposition maps as presented in Figure All the turbidity current simulations showed that the erosion takes place in the upper steeper part of the reservoir, whereas the fine sediments deposit in the vicinity of the dam.

48 38 Reservoir Sedimentation Discretized reservoir geometry Erosion-deposition map Sediment transport Inflow velocity [m/s] Inflow concentration [%] Time [s] Luzzone Dam Bottom bed interaction Erosion Deposition Figure 3.4-1: Scheme of numerical model with inflow conditions, discretized reservoir topography, bottom user interfaces and erosion-deposition map. Routines were added to the program to take into account the settling character of the suspended sediments and the erosion-deposition at the bottom. These implementations were realized with help of FORTRAN routines and have been developed in LCH by De Cesare (1998) and extended by Oehy (2003). The reader can find more details on these routines in the mentioned references. Turbidity currents are strongly stratified flows with high density gradients and may be characterized by an active bottom layer in which turbulence is produced or maintained by ground shear, a sharp density interface, and a relatively uniform non-active layer representing the still water body. The stratification influences the structure of the turbulence, since vertical motions are directly affected by buoyancy forces. Turbulence in a stratified fluid is typically anisotropic, with reduced vertical velocities and vertical length-scales, and a reduced correlation between the density perturbation and the vertical velocity component. On the basis of this description, the accuracy of computational predictions in sharply stratified flows is expected to depend mainly on the treatment of turbulence in the vicinity of the interface and on the modelling of diffusive transport. To properly account for these additional effects of stratification, two buoyancy-extended k-ε models and a standard k-ε formulation were compared with experimental data of turbidity and gravity currents described in the literature Two buoyancy-extended k-ε models or the standard k-ε formulation The fluid is considered as incompressible and the buoyancy effects are introduced through the density variations due to sediment concentration or salinity by assuming the Boussinesq approximation. The water flow computation is based on the Reynolds averaged Navier-Stokes Equations with density depending on sediment concentration or salinity. The transport of the diluted matter is calculated with the convection-diffusion equation for the sediment concentration or salinity c:

49 Deep, seasonal storage reservoir 39 c + t c T c ( U j w j ) ν = ( SE SD b x j x j σ c x j ) Eq where U j represents the velocity components and w j, the sediment fall velocity. S E and S D represent the source or sink terms at the bed due to erosion or deposition of sediments, respectively. The erosion-deposition model proposed by Garcia and Parker (1993) was used. For the turbidity current test cases three sediment concentration transport equations were solved for grain diameters d 16, d 50 and d 84. For the salinity current w j, S E and S D were set equal to zero. The turbulent Prandtl number is defined as σ c = ν T / ΓT. It is considered constant in the standard k-ε model and is chosen equal to unity. The turbulent eddy viscosity ν T must be modeled with a turbulence model. The k-ε model is still the most common turbulence closure model and is used throughout the study, where k is the turbulent kinetic energy and ε denotes the dissipation rate of k. The equations for the standard k-ε model are k + U t ε + U t j j k x j x ε x x j j j ν T σ k k x ν ε T ε = σ ε x j k j = P + G ε ( C P + C G ε ) 1ε 3ε C2ε Eq where P, the shear production and G, the buoyancy production of k, are defined as U U i j U i P = ν + T and x j xi x j ν T G = σ c gi ρ 0 ρ x i Eq respectively. So, the eddy viscosity ν T = C μ (k 2 /ε) based on k and ε is calculated by solving the two-coupled differential equations for the turbulent kinetic energy k and its dissipation ε. The parameters σ k and σ ε denote the turbulent Prandtl numbers related to the diffusive transport of k and its dissipation rate ε, respectively. The modeling constants, C μ, C 1ε, C 2ε, C 3ε, σ k and σ ε, are respectively taken to be equal to 0.09, 1.44, 1.92, 1.0, 1.0 and 1.3. For stable flows (i.e. if G is a sink term), the constant C 3ε is given the value 0 (Rodi, 1987). The buoyancy extension proposed by Burchard and Petersen (1999) introduces stability functions acting on the eddy viscosity, ν T and diffusivity, Γ T and are defined as ν f ( cμ ) 2 k ε T = and T = f ( cμ ) 2 k ε Γ Eq respectively. The values of the parameters in the k-ε equations stay the same except c μ 0 and σ ε which are equal to and 1.08, as well as C 3ε which becomes equal to 0.4 in stable flows. The role of the stability functions is to correct the eddy viscosity and diffusivity for further effects of stratification (stratification effects are already included in the buoyancy production term G on the right-hand sides of the k and ε equations). Stability functions generally damp turbulent exchange for stable stratification and enhance turbulent mixture for unstable stratification.

50 40 Reservoir Sedimentation The stability functions f 1 and f 2 depend only on two non-dimensional parameters for stratification and shear, represented by α N. f 1 c 0 μ α α α N N = Eq N f α = Eq N 2 6 g ρ k = and α N 0.56 ρ x ε 0 i with α N ( cμ ) 2 0 i The standard k-ε equations may be obtained by using f 1 and f 2 with α N = 0, and thus neglecting the influence of stratification on the stability functions (Burchard et al., 1998). The second buoyancy extension to the k-ε model was proposed by Rodi (1980) and used by Olsen and Tesaker (1995) for the simulation of turbidity currents. In this model the effect of the density variations on the water field is taken into account by introducing a modified eddy viscosity. The eddy viscosity from the k-ε model is multiplied with a factor depending on the local Richardson number, Ri. So, the general effect of increasing the Richardson number is to suppress eddy viscosity. ν = ν (1 + β Ri) T mod T α Eq g ρ / z with Ri = ; α = 0.5 and β = 10 2 ρ ( U / z) At the entrance section of height h 0, a uniform distribution is chosen for the inflowing boundary velocity U 0 and concentration c 0. Along the inclined bottom and the walls, the classical log-law velocity distribution is applied and the turbulence is assumed to be at equilibrium. This is a standard approach and may be found in the reviewed literature and will not be further explained (Rodi, 1980). At the water surface, the rigid lid assumption has been adopted. It implies that the free water surface is treated as a surface of symmetry for all variables. Zero gradient boundary conditions are used for all variables at the outflow boundary. To solve the governing equations a control volume method was chosen for discretization, using higher-order schemes. For time stepping an implicit backward difference procedure has been used. The numerical methods are described in more detail in CFX-4.3 (CFX-4.3, 1997) Results Numerical results derived from the two buoyancy-extended k-ε models and the standard k-ε model are compared with experimental data of turbidity and salinity current experiments documented in the literature. Two turbidity current flume experiments with different grain size distributions and a saline gravity current from Altinakar et al. (1990) were considered. Furthermore, a turbidity current experiment of a hydraulic jump presented by Garcia (1993) was used for comparison. Table summarizes the experimental conditions, whereas the experimental configurations may be found in the references.

51 Deep, seasonal storage reservoir 41 Table 3.4-1: Experimental conditions of data sets used for computations (D sg geometric mean diameter; ρ a density of ambient water). Author Exper. No. Type D sg w S h 0 U 0 C 0 ρ a ρ 0 g' 0 T [10-6 m] [10-3 m/s] [-] [m] [m/s] [10-3 ][kg/m 3 ] [kg/m 3 ] [m/s 2 ] [ C] Altinakar (1990) SALT08 Saline Altinakar (1990) TK0609 Sediment Altinakar (1990) TK1305 Sediment Garcia (1993) NOVA7 Hydraulic Jump / For the experiments by Altinakar the first density profile was used to adjust the inlet density to better fit the measurements, whereas the inlet velocity was kept unchanged. Figure shows the three density and velocity profiles at stations 2.2 m, 4.0 m and 8.0 m along the flume for the salinity current experiment SALT08. The model 1 represents the buoyancy extension proposed by Burchard and the model 2 the eddy viscosity modification presented by Rodi. z [m] Station 220 Model 1 Model 2 Standard Experiment Station 400 Model 1 Model 2 Standard Experiment Station 800 Model 1 Model 2 Standard Experiment ρ [kg/m 3 ] z [m] Model 1 Model 2 Standard Experiment Model 1 Model 2 Standard Experiment Model 1 Model 2 Standard Experiment U [m/s] Figure 3.4-2: Comparison of velocity and density profiles of buoyancy-extended (Model 1: Burchard and Petersen, 1999; Model 2: Rodi, 1980) and the standard k-ε models against experimental data (SALT08; Altinakar, 1988). The head propagation of the current in the flume is presented on Figure for the different models as on Figure

52 42 Reservoir Sedimentation Model 1 Model 2 Standard Experiment 8 x [m] t [s] Figure 3.4-3: Comparison of the propagation of the turbidity current head for the buoyancy-extended (Model 1: Burchard and Petersen (1999); Model 2: Rodi, 1980) and the standard k-ε models against experimental data (SALT08; Altinakar, 1988). z [m] Station 220 Model 1 Model 2 Standard Experiment Station 400 Model 1 Model 2 Standard Experiment Station 800 Model 1 Model 2 Standard Experiment ρ [kg/m 3 ] z [m] Model 1 Model 2 Standard Experiment Model 1 Model 2 Standard Experiment Model 1 Model 2 Standard Experiment U [m/s] Figure 3.4-4: Comparison of velocity and density profiles of buoyancy-extended (Model 1: Burchard and Petersen, 1999; Model 2: Rodi, 1980) and the standard k-ε models against experimental data (TK1305; Altinakar, 1988).

53 Deep, seasonal storage reservoir 43 Figure shows the results of the computations of the turbidity current experiment TK1305 by Altinakar (1988). For this case, no density adjustment was necessary at all. Finally, the computation of the hydraulic jump experiment by Garcia with a supercritical flow at station 300 and a subcritical flow at station 800 is presented on Figure Station 300 Model 1 Model 2 Standard Experiment Station 800 Model 1 Model 2 Standard Experiment z [m] U [m/s] U [m/s] Figure 3.4-5: Comparison of velocity profiles of buoyancy-extended (Model 1: Burchard and Petersen (1999); Model 2: Rodi, 1980) and the standard k-ε models against experimental data (NOVA7; Garcia 1993). In all the numerical computations both buoyancy-extended k-ε models showed almost identical results with very good agreement to the measured velocity and density profiles. Nevertheless, a significant difference can be observed between the standard k-ε model and the experimental data. In the standard k-ε model, velocity is systematically underestimated because the vertical diffusions are too high. As a consequence turbidity current height is overestimated with this model Conclusions The turbulent regime of a turbidity current is governed by two factors: production of turbulence (mainly by shear of the turbidity current at the bed) and suppression of turbulent intensities by density stratification. The results clearly indicate that the vertical mixing and the spreading of the shear layers are reduced by the stable stratification and confirm the importance of accounting for buoyancy effects in numerical modelling of turbidity currents. The performance of the two buoyancy extended k-ε models in the case of turbidity currents has been shown to be similarly good. The velocity and density profiles measured in the laboratory are well reproduced. Using unmodified standard k-ε model, the eddy viscosity is too high in the entrainment region between the turbidity current and the clear water and diffusion is too high. With taking into account the density stratification, either by using stability functions as proposed by Burchard and Petersen (1999) or a damping function as introduced by Rodi (1980), the standard eddy viscosity is corrected and results form the numerical model agrees much better with the physical model studies performed by Altinakar (1988 and 1990) and Garcia (1989 and 1993).

54 44 Reservoir Sedimentation 3.5 Case Studies Submerged Dams in Lake Grimsel Generalities In Lake Grimsel, in Switzerland, an ongoing design project consists of heightening the two existing dams by 23 m (Spitallamm Arch Dam 114 m; Seeuferegg Gravity Dam 42 m). The excavation and demolition works necessary for the planned heightening generate approximately 150'000 m 3 of rock material. This large amount of materials has to be stored somewhere near the construction site. This led to the idea of building some kind of obstacle in the form of a submerged embankment dam to prevent sediment deposition due to the turbidity currents in the area near the intake structures. A case study is presented to investigate the occurrence and impact of turbidity currents on the reservoir sedimentation and to check the efficiency of such submerged obstacles to retain the sediments (Figure 3.5-1; see also Oehy und Schleiss, 2001; Schleiss and Oehy, 2002). Single obstacle at 1845 m a.s.l. Two obstacles at 1860 m a.s.l. Figure 3.5-1: Overview of the investigated obstacles in a 1:25'000 map Based on bathymetric measurements of the reservoir and on the initial, topographical map of the valley, the physical domain was discretized using a body-fitted, non-orthogonal, structured grid. The shoreline has no direct effect on the behaviour of the turbidity current, flowing only at the bottom of the lake. The reservoir geometry was, therefore, slightly simplified to get a topologically rectangular physical space. The grid consisted of 250 cells in the streamwise direction, 30 in the lateral, and 20 cells in the vertical directions, resulting in a total number of 150'000 cells. The cells had a mean horizontal mesh size of about 25 by 25 m 2 and in the vertical direction the cell height was growing in size in a geometrical progression from 0.3 m at the bottom to 6.0 m at the free surface. The reservoir is approximately 5.5 km long and 300 m wide. The depth is regularly increasing from the inflow to the middle of the lake where a hollow upstream of the canyon exists. The intake and bottom outlet structures are located in the deepest area, approximately 90 m deep, downstream of the canyon. To the east of the deepest area, the reservoir bottom increases by more than 50 m to a branch of the lake. The outflow boundary was placed in this branch of the lake where no turbidity current flow could occur. At the river mouth a uniform spatial distribution was chosen for the inflowing boundary conditions of velocity and concentration. By the examination of floods in mountainous regions, Hager (1984) suggested a hydrograph in the form of an asymmetrical bell corresponding to the statistical distribution of Maxwell. It is commonly accepted that during flood events high suspended sediment concentrations can occur in Alpine rivers. De Cesare (1998) analyzed concentration measurements in an Alpine

55 Deep, seasonal storage reservoir 45 river during a whole summer and recorded several times values of more than 15 g/l. In Lake Grimsel no measurements for the concentration at the inflow were available; therefore it was assumed that the concentration evolution is proportional to the discharge and that the peak concentration had a value of 15 g/l. Because no stratification due to temperature difference is present in the Alpine lake, the clear water density was assumed constant in the whole reservoir. At the water surface, the rigid-lid assumption was adopted, which implies that the free water surface is fixed to a given level. This assumption was justified as the water level in the deep lake changed only slightly (2-4 m) during the simulation period. A single transport equation was solved for the median grain-size diameter of d 50 = 40 mm, determined from field samples. The erosion-deposition source and sink terms at the bottom of the reservoir are introduced through the erosion-deposition model proposed by Van Rijn (1984). Along the bottom and the walls, a log-law velocity distribution was applied changing continuously from smooth to rough wall conditions as proposed by Wu et al. (2000). It was assumed that no additional roughness due to bed forms exists at the bottom, because no measurements or data were available. It is commonly accepted that in large-scale simulations the turbulence model k-ε needs modification to take into account the stratification. Olsen and Tesakar (1995) used a modified turbulent viscosity depending on the local Richardson number combined with the standard k-ε turbulence model to model turbidity currents in a reservoir. Rodi (1987) gives a good review of examples of calculation methods for flow and mixing in stratified flows. Here, the buoyancy extension of the standard k-ε model proposed by Burchard and Petersen (1999) was included to properly account for these additional effects of stratification. To get reasonable computation times for the flood event, a time step of 60 s was chosen, which resulted in CPU times of hours on a DEC alpha station Turbidity current simulation of Flood Event in October 2000 The results of the high flood event occurring in October 2000 revealed that a turbidity current develops and propagates to the deepest area of Lake Grimsel close to the dam. During such an event considerable sediment deposits are created in the area of the intake and bottom outlet structures. The canyon with the negative slope causes a slowing down of the current and deposition takes place upstream of this place. After passing the ridge the current accelerates again and finally dies out in the deepest area of the lake with maximum deposit heights of approximately 0.10 m. The turbidity current develops with increasing sediment and water discharge. After the peak the driving force of the current decreases and the turbidity current starts to die out until it finally disappears after 84 hours. The concentration decreases continuously from the inflow to the deepest area Turbidity current passing over submerged dams To prevent reservoir sedimentation in the deepest area close to the intake and bottom outlet structures, the efficiency of submerged obstacles was numerically investigated. Two possible configurations for these obstacles were evaluated. The first configuration consisted of a dam, 15 m high and 150 m long, situated upstream of the canyon in a counter-slope of the lake. The second configuration consists of two submerged dams placed in the middle of the lake one after each other in a displaced manner, so that the current needs to turn around or overflow

56 46 Reservoir Sedimentation them. In this case, the height of the two dams was 10 m with a length of 210 m each. Both configurations do not extend over the whole width of the valley to keep a free passage for the water flow during emptying of the reservoir. To simulate the obstacle, a thin surface has been introduced in the grid. This surface is assumed to be located between control volumes and allows setting wall boundary conditions on this patch. The obstacle clearly blocks the flow and reflects the major part of the turbidity current. A considerable amount of sediment deposits occurs, therefore, upstream of the obstacle. Some of the fluid of the turbidity current flows over the obstacle while a bore will be reflected. The same observations were made in the numerical simulation of the flow over the submerged dams (Figure 3.5-2). Bores travel upstream as was observed in the physical experiments and their numerical simulation. The retention volume upstream of the obstacle is slowly filled with the turbid fluid until it starts overtopping. Because the dam does not block the valley completely a small part of the current continues also flowing around the obstacle. Figure 3.5-2: Sedimentation after the flood in October 2000 (40 m 3 /s and 15 g/l) with a single obstacle at 1845 m a. s. l Conclusions The turbidity current flow over the complex topography of Lake Grimsel shows that significant amounts of suspended sediments are transported to the deepest area of the reservoir, where they settle down. For a single flood event these deposits attained approximately 10 cm in height. The effect of an embankment dam, built of demolition and excavation materials from the heightening of the Grimsel dams, was investigated. In agreement with the physical experiments, it is found that the height should at least extend to twice the height of the approaching turbidity current to block the flow efficiently. A height of the dam of 15 m is sufficient and ensures that the elevation of the dam crest is below the minimum operation level of the reservoir. It is estimated that the retention of sediments behind the dam lasts for at least 20 to 50 years. It can be concluded that the recycling of the demolition and excavation materials to build a submerged embankment dam gives an excellent example to control reservoir sedimentation due to turbidity currents.

57 Deep, seasonal storage reservoir Luzzone The Luzzone arch dam of the Blenio Hydropower Company (OFIBLE) was built from 1958 to 1963 near the village of Olivone in the southern part of Switzerland in the Canton Ticino. The maximum crest height is 208 m and the crown is 530 m long. It is equipped with a power intake, a multilevel outlet structure with a bottom and an intermediate outlet as well as a crest overfall spillway. The bottom outlet capacity is 55 m 3 /s, and its entrance is situated approximately 150 m from the dam on the right bank of the valley. From 1995 to 1998 the dam crest was raised by 17 m from 208 m to 225 m to bring the water level 15 m higher which increased storage from 87 Mio. m 3 to 107 Mio. m 3, allowing additional production of 60 Mio. kwh of energy in the wintertime. The reservoir drains directly an area of 36.5 km 2, with three main tributaries coming from the north, east and south respectively. Their pre-construction confluence is located on the old Luzzone Alp, submerged by the present lake. Nine water intakes and the corresponding galleries enlarge the total drained area to approximately 100 km 2. The initial geometry of the reservoir in the deeper part is characterized by a V-shaped valley that has accumulated sediment for more than 30 years. The mean bottom width is now around 50 m. A trapezoidal section approximately characterizes the reservoir geometry with side slopes between 1:1 and 1:2. The bottom shape is nearly symmetrical in the reach near the dam. The average longitudinal slope along the reservoir bottom is about 4 %. The annual mean sediment inflow is approximately 38,000 m 3 (based on measured deposit volume), and the total sediment volume in the lake is 0.90 Mio. m 3 or 1.03 % of the pre-heightening total storage capacity in Major sediment deposits cover approximately 0.1 km 2 of the lake bottom, or around 8 % of the total lake surface. During 1985, the reservoir was emptied (see Figure 3.5-3), allowing the release of approximately 0.3 Mio. m 3 of alluvial deposits through the bottom outlet over a period of 7 weeks. At present, regular short-term flushings keep the intake of the bottom outlet free from sediments. These flushings create an almost 18 m deep cone in the sediment deposits around the bottom outlet structure. Dredging in 1995 removed sediments from the upper part of the power intake. The intake has been recently raised to minimize potential blockages by sediments. Figure 3.5-3: Luzzone Reservoir during emptying in 1985, looking upstream.

58 48 Reservoir Sedimentation In order to clarify the flow mechanism of river-induced turbidity currents in an artificial lake, field observations of turbidity currents were carried out in the Alpine reservoir of Luzzone in Southern Switzerland and its main inflow river in the Val di Garzora (see Figure and Figure 3.5-4). In 1992 and 1995, extended measuring campaigns took place in the mentioned test reservoir. Figure 3.5-4: Picture of the 225 m high Luzzone arch dam and its reservoir. The National Hydrological and Geological Survey (SHGN), in close partnership with the Blenio Hydropower Company (OFIBLE) and the Laboratory of Hydraulic Constructions (LCH), placed a fully equipped measuring station on the inflow river about 500 m upstream of the lake. The following measurements were made during two summer periods in 1995 and 1996: discharge Q [m 3 /s]; temperature T [ C]; photo-optical turbidity measurements [NTU]; automatic suspended sediment sampling during floods C s [g/l] beyond a certain water level, for calibration of the optical sensor. During the two years of on site investigations in 1995 and 1996, no significant floods were observed. It was possible, however, to show the relationship between precipitation, water and sediment flow and turbidity current in the reservoir even for minor events. The inflow measurements showed large variations in water discharge, sediment concentration and water temperature (see Figure 3.5-5).

59 Deep, seasonal storage reservoir 49 Figure 3.5-5: Measured precipitation, discharge, temperature, optical turbidity and suspended sediment samples for a typical small flood A simple relationship describing the inflow hydrograph and the sediment transport for small floods in the incoming river was established as a set of non-dimensional equations. These relations described hereafter were implemented in the numerical flow code as an upstream boundary condition. By the analysis of floods in mountainous regions (Hager, 1985) proposes a hydrograph in the form of a "bell" corresponding to the statistical distribution of Maxwell, a relationship entirely defined by three parameters: Q(t) = Q p t t p n t 1 t p e Eq where : Q p peak discharge [m 3 /s] t t p time [s] time to peak [s] n form factor (n 2) [-] The determination of these parameters was based on field measurements. Time to peak and the peak discharge can be obtained by hydrologic analysis of measured floods, the form factor

60 50 Reservoir Sedimentation was obtained by best-eye fitting the curve over some observed typical flood hydrographs. The study of the hydrographs and the corresponding evolution of water turbidity of eighteen minor flood events indicate that the same distribution can be applied to characterise the progress of suspended load during a flood. It has its own three parameters as described in Eq Its time to peak can be related to the one of the hydrograph by the following relationship: t / = 0.85 ± 0.16 Eq sp t p where: t sp time to peak for the suspended load [s] Eq shows that the maximum suspended load during floods appears regularly before the hydrograph reaches its peak value. The relationship between optical turbidity measurements and suspended sediment samples didn't show a good correlation, less than 0.3. Even the analysis of each event separately did not indicate a valuable connection between these two types of turbidity evaluation. As part of the overall study of sedimentation, continuous monitoring of the occurrence and flow behaviour of turbidity currents in an artificial lake was carried out using an underwater current meter network. The system is able to detect and track density underflows. It consists of six current meters located at the bottom of the lake at three stations A and B, 150 m and 800 m upstream of the dam and C at the confluence of the three major tributaries. Figure 3.5-6: Location and vertical disposition of current meters of the underwater measuring network consisting of 3 stations A, B and C with 2 levels at each station. Two instruments were stacked vertically at each station 2 and 4 meters above the bottom, see Figure Each of the verticals were anchored on the lake floor and kept tight with a subsurface buoy, connected to a surface buoy for easy recollection. The recorded information was stored on data loggers inside waterproof casings on each individual instrument. The current meters survey at a rate of one record every thirty minutes the following parameters: mean average flow velocity during the 30 minute interval, precision of 1.0 cm/s instant direction of the flow associated to the magnetic north, precision of 5 instant water temperature, precision of 0.1 C water pressure, precision of 1 m

61 Deep, seasonal storage reservoir 51 The current meters located at station C showed the most sensitive reaction on inflow variations. The instrument at 2 m from the bed recorded more than 700 occurrences during the five-month monitoring period from May to September Most of the covered currents came from the north watershed. The instrument at 4 m from the bed recorded five times fewer occurrences, showing that most of the occurrences at station C were low dimension bottom currents with heights of less than 4 meters. Figure represents the statistically analysed directions of flow at station C, 4 meters above the bottom as a function of the mean velocity. The main flow direction indicates that the currents originate predominantly from the east tributary. No flow came from the south catchment area. The highest recorded velocity 2 meter above the bed was 80 cm/s and 82 of all records were lower than 20 cm/s. The maximum velocity 4 m above bed level was 55 cm/s. Figure 3.5-7: Observed directions at the bottom of the reservoir at station C, 4 meters above ground, classified by current velocity. The flow is well oriented along the longitudinal axis of the lake The directions at station B in the middle of the reservoir are fairly well aligned with the axis of the lake. Maximum velocity 2 m above bed level was still 73 cm/s. Only very few occurrences of flow were recorded at station A. They could just once be related to observe currents at the stations B and C both located upstream. The only event where a current ran through the whole lake of around 2.5 km length lasted seven hours in total. Its activity came to a rather abrupt end at each station and can be associated to a surge type event. Unfortunately, the upstream gauging station was not yet fully functional by that time, in early spring 1995, and no inflow measurements are available for the event. Similar observations were made during a storm in 1992, when an earlier measuring campaign in the same reservoir took place (Sinniger et al., 1994). Some small return currents moving upstream from the dam were observed in 1992 and The measurements indicate that density underflow activities are common close to the inflow rivers, whereas, further down in the reservoir, only a few currents could be observed. The presented measurements, the sediment samples taken on the lake bottom and the analysis of the shape of the reservoir floor suggest that the main carrying media for sediments are turbidity currents.

62 52 Reservoir Sedimentation Numerical simulation A general purpose Navier-Stokes solver CFX-4 (CFDS-CFX-4, 1995) was used to predict turbulent flow in the physical model. The numerical model equations are solved in the Boussinesq approximation and the effect of density variations is neglected, except in the buoyancy term. The standard equations of continuity, momentum and energy conservation are used in the model. A constitutive equation relates enthalpy to temperature and pressure. The model provides closure for turbulence. The equations are solved on a non-staggered grid by means of a finite volume approach. Suspension is directly treated in the CFX-4 code through its homogeneous two-phase flow model with a continuous Eulerian description in which interactions between phases are ignored. The following assumptions apply: phases are mixed on length scales larger than the molecular length scales and smaller than the numerical grid; each phase is treated as an interpenetrating continuum; every phase is present in each control volume and assigned a volume fraction; water is defined as a continuous phase occupying connected regions of space; sediment particles are defined as a solid disperse phase distributed uniformly in a control volume; transported quantities are the same for each phase; one single flow field is calculated. User defined sediment deposition and erosion modules are added to take into account sedimentation and the interactions between the turbidity current and the reservoir bottom deposits. The flow in the physical model flume was simulated with exactly the same boundary conditions. A uniform inlet velocity profile was introduced in the numerical model; the inlet velocities are the one observed in each of the three completely analysed experiments with one extra UVP transducer located at the open gate Selected numerical results of the physical model Figure shows the velocity distribution in the turbidity current expansion. Both calculated and measured values are plotted on the same graph. The results of the numerical modelling are presented by an iso-surface of concentration where the maximum concentration gradient can be found. This surface fits well to the position of the interface between turbidity current and the surrounding water. The velocity vectors are given on a fine grid in a plane parallel to the bottom. This plane is situated at the same level as the one where the transducers were installed. The figure also shows the measured limits of the advancing current as a continuous line. The numerical calculation and interpolation of the interface result in a theoretical smooth surface; this is not the case for real turbidity currents. The interface looks more or less like a cloud, see also Figure The velocity distribution inside the expanding flow is very well predicted by the numerical modelling. The velocity vectors in the measured flow field confirm the direction and the magnitude of the computed velocity vectors.

63 Deep, seasonal storage reservoir 53 a) b) Figure 3.5-8: Computed and measured 2D flow field close to the bottom and limits of the spreading turbidity current a) 5 and b) 10 seconds after opening of the gate. Numerical simulation : black velocity vectors, turbidity current as a blue surface; Physical model : white velocity vectors, limits of the turbidity current as a bold line The validated numerical two-phase flow code was applied to simulate river-induced turbidity currents moving in the real geometry of an artificial reservoir. Lake Luzzone in the Southern Alps of Switzerland was selected, since in situ measurements were available (Sinniger et al., 1994) Selected numerical results from Lake Luzzone Grid Generation Based on the pre-construction topographic map and on bathymetric measurements of the reservoir bottom from 1982 to 1994, a mathematical model of the reservoir topography was prepared. The model is a schematization of the contemporary state after the last complete bathymetric measurement in The reservoir has two small islands along its left shore with an intermediate dip; see Figure a). The shoreline has no direct effect on the local behaviour of the turbidity current, which flows only along the bottom of the lake. Therefore the real reservoir geometry was simplified to produce a topologically rectangular physical space with four nodes in a cross section cutting off shore and islands (see Figure b)). The physical space was discretized by 100 cells in the streamwise direction, 20 laterally in a symmetrical progression growing from the centre to the border, and 18 vertically also growing in size from the bottom to the top, giving a total of 36,000 cells (see Figure ). The bottom cells in the centre have a size of about 25 x 5 x 2 m 3. reservoir turbidity current reservoir turbidity current reservoir turbidity current a) b) c) Figure 3.5-9: Method of decomposition of the real reservoir geometry a) to a simplified physical space without shore and islands b) and then to the rectangular computational space c)

64 54 Reservoir Sedimentation One feature of the applied flow solver code CFX-4 is that the coordinate system and the finite difference grid are entirely distinctive entities. The coordinates in the physical space are often referred to as a body-fitted coordinate system. After a Jacobian Matrix coordinate transformation is applied to the physical space, the converted space is exactly rectangular with a uniformly distributed finite difference grid (see Figure c)). It is also referred to as the computational space, as it is in this space that the equations are solved. The boundary conditions and the information on the current-floor interaction model are stored in 4,300 dummy cells glued around the computational domain. The finite volume mesh is invariable with time. Since the newly formed local sediment deposit and scour do not influence the overall flow behaviour during one simulated event, their amplitude of a maximum of 0.50 m per event are negligible compared to the reservoir size and depth of more than 160 m Boundary Conditions The location and extent of the surfaces on which boundary conditions are to be set are specified using the concept of patches. The tributary is an inlet patch, the bottom of the reservoir and the dam are wall boundary patches, and the surface spillway is an outlet patch. The variables are either value given (Dirichlet) or gradient given (Neumann). The boundary conditions and the variables to be defined are summarized in Table '500 local coordinate Y [m] 1'000 N s21 s31 s41 s51 s s11 0 entry patch local coordinate X [m] '000 1'500 2'000 2'500 3'000 Figure : Location of stations s11 to s61, used for extraction of local values of the significant parameters for data analysis. Also shown are the directions of the turbidity current and the generated single grid mapped on the reservoir bottom Table 3.5-1: Boundary conditions used in the numerical simulation BOUNDARY bottom of reservoir inflow river spillway lake surface VARIABLES velocity (discharge) volume fraction (concentration) no slip extrapolated from upstream temperature - Value given (Dirichlet) Value given (Dirichlet) Value given (Dirichlet) gradient given (Neumann) gradient given (Neumann) free slip - - pressure not needed not needed p = 0 atmosphere (p=0) -

65 Deep, seasonal storage reservoir Simulation of a Flood Event Based on existing hydrologic analysis, a hypothetical 1000-year flood event was prepared as input to the numerical model. The peak flow is Q p = 137 m 3 /s and time to peak is t p = 60 minutes. The shape factor in Eq was put to n = 30 for the hydrograph and to n s = 60 for the sediment concentration with its reduced time to peak value of t sp = 51 minutes according to Eq The maximum concentration, based on extrapolation from in-situ measurements, was set at C sp = 10 % vol or 265 g/l, giving a peak solid discharge of Q sp = 10.6 m 3 /s or 28,090 kg/s occurring some 3 minutes after the sediment concentration peak. Volume fractions and not mass fractions must be given in the flow solver CFX-4. The selected single sediment size was d 50 = 0.02 mm, mean value based on samples taken from the inflow river, from the reservoir bottom, and from flushing. The rise of the hydrograph starts at about half the time to peak in Eq , in this case 30 minutes later (see Figure ). The numerical simulation starts only at that time, which becomes the new reference time in all that follows Q / Q p Q s / Q sp 0.75 Discharge Suspended matter t sp t p 0.00 t / t p Figure : Measured and adjusted non-dimensional time evolution of inflow (, ) and sediment concentration (+) Results The plunging of the tributary occurs just after the inflow patch of the computational domain (see Figure ). The maximum mean inflow velocity perpendicular to the patch is only 0.09 m/s for the maximum discharge. The underflow turbidity current then accelerates downstream along the bed, guided by the banks of the reservoir (see Figure ). It reaches a maximum velocity of more than 2.5 m/s in the narrow canyon just below the inlet at station s11, and still has a velocity of 1.5 m/s in the larger part of the reservoir from station s31 to s61. The flow along the reservoir bed as a function of travel time can be seen in Figure a) to Figure f) at 10 minute interval and after 3.5 hours. After about 40 minutes the current arrives at the dam (see Figure d)). It is reflected and returns upstream, interacting with the still downstream moving body of the turbidity current. The returning current travels upstream over a distance of about two thirds of the total reservoir length (see Figure f)). It reaches a velocity of 0.3 m/s. The global motion inside the lake becomes insignificant after approximately 4 hours while sediment inflow stopped already after

66 56 Reservoir Sedimentation 1.5 hours (see Figure ). A sediment-laden underwater "muddy lake" is formed, which will then settle its granular material over several hours or even days. a) 10 min b) 20 min c) 30 min d) 40 min e) 50 min f) 210 min Figure : Pictures a) to f) in plane view and along the axis of Luzzone Reservoir showing the advancing turbidity current 10, 20, 30, 40, 50 and 210 minutes after the rise of the hydrograph

67 Deep, seasonal storage reservoir velocity [m/s] current front reaches dam s11 s31 s21 s s51 s return current time [h:mm] 0:00 0:15 0:30 0:45 1:00 1:15 1:30 1:45 2:00 2:15 2:30 Figure : Evolution in time and space of the horizontal turbidity current velocity along the axes of the reservoir at station s11 to s61 in the bottom cell Figure shows, in a 3D view, the surface of the numerically simulated turbidity current 30 minutes after the rise of the hydrograph flowing along the axis of Lake Luzzone. The figure illustrates a surface of equal concentration at the location of maximum concentration gradient along the vertical axes, a method commonly used to characterize the interface between the turbidity current and the clear water. Due to particle entrainment from the existing sediment deposits, concentration increases as the current moves on. The current is globally erosive and thus becomes stronger during the first two hours, but becomes depositing later on. The volume of sediment entrained from the bottom is around 35,000 m 3, compared to 9,000 m 3, contributed by the inflow river. The maximum erosion depth of 0.35 m takes place in the centre of the lake; and the maximum deposition of 0.50 m is located close to the dam. Figure illustrates the location of the global erosion and deposition for the simulated turbidity current over a total duration of 4 hours and 10 minutes.

68 58 Reservoir Sedimentation Figure : 3D view of the bottom of the Luzzone Reservoir showing the numerically simulated turbidity current flow, 30 minutes after the rise of the hydrograph ] [m y 0001 t=end: cm x[m] Figure : Calculated sediment depth change due to a simulated 1000-year flood on the bottom of the reservoir

69 Deep, seasonal storage reservoir Links to Observations Observation and numerical simulation show that turbidity currents are not only the main transport medium for the incoming fine granular material, but can also redistribute sediments inside the reservoir by entraining bed material and transporting it closer to the dam. After stopping, the current will deposit its entire sediment load almost uniformly across the deepest part of the lake. The surface covered with sediments brought in by turbidity currents can be seen in the detailed bathymetric map in Figure showing the deepest deposits near the dam. Ultrasonic bathymetric measurements of the reservoir bottom made in 1994, after 31 years in service, showed deposit depth increasing from 5 m in the upstream part of the lake, to 30 m in front of the dam. 5 m depth 30 m depth m depth Figure : Qualitative representation showing the location and the measured sediment deposits magnitude after 31 years in service The numerical simulation showed that turbidity currents from a large flood could entrain a substantial amount of bed sediment into suspension. The entrained sediment increased the density and accelerated the velocity of the current, causing even more entrainment to occur. Similar observations were made during experiments by Garcia and Parker (1993). Smaller floods have also been simulated, special attention has been paid on one observed flood occurred in 1992, that had an estimated statistical return period of one year. Velocity values for the simulated and observed one-year flood were of the same magnitude. The measured average velocities over a 30 minute interval at station B were 0.51 m/s and 0.48 m/s at station A close to the dam. The numerically simulated flow reached a velocity of 0.45 m/s at the same downstream location. A significant return flow was measured for that event with a velocity of 0.18 m/s at station A and 0.11 m/s at station B two meters above the bed. No upstream-directed velocity could be attained by numerical simulation for that event, only a slight reduction of the main downstream flow velocity. Sedimentation prevails over entrainment for minor turbidity currents, the density of the turbidity current decreases in the downstream direction. Therefore turbidity currents induced by small floods and snowmelt seldom reach the dam, and when they do then only at low speeds and concentrations. They will deposit their granular material in the upper part of the reservoir in the form of temporary deposits, which can be eroded by larger flood events and carried into the deepest part of the lake. This phenomenon is similar to snow drift by powder snow avalanches (Hermann, 1990).

70 60 Reservoir Sedimentation Conclusions Turbidity current flow in a laboratory flume as well as field measurements during two summer seasons at the Luzzone Reservoir in the Swiss Alps were used to validate a 3D numerical model. User-defined erosion and deposition modules that take into account the interaction between the current and the existing sediment deposits were used to simulate the balance between sediment deposition and erosion in the model. The results of the computer modeling and field observations showed that turbidity currents caused by a large flood could entrain substantial amounts of bed sediment and transport this sediment within the current to a zone of deposition near the dam. Field observations made in the Luzzone Reservoir illustrated that as long as the main tributary carries a sediment concentration, the inflow plunges and the induced turbidity current travels along the bed at velocities up to 0.80 m/s. Based on the numerical simulation, not only general conclusions can be drawn, but the precise behavior of turbidity currents can also be predicted. Although the distribution of the calculated velocity and concentration fields could not be compared with observations, because measurements in the reservoir were available at only three monitoring stations, the characteristics of the simulated turbidity currents in the Luzzone Reservoir were in reasonable agreement with observed conditions. The numerical model can be used as a strategic evaluation tool for reservoir management. In the future the model will be applied to analyze various technical solutions to prevent sedimentation in the most vulnerable parts of the reservoir, the bottom outlet and the water intake. Based on this simulation the optimal timing of the opening of the bottom outlet can also be determined to pass an important part of the sediment yield beyond the dam during floods Livigno Scope of the study A simple analysis of the problem using 1D calculation was done to get a first insight of the problem, afterwards a 3D numerical simulation was performed in order to validate the results for the studied measures. Two types of alternatives were considered for the reduction of the effects of sedimentation studied for the Livigno Reservoir: The implantation of pervious obstacles with three different heights located 3.0 km downstream of the inlet of the reservoir for the maximum operation level. A measure consisting of a geotextile screen placed approximately 2.5 km upstream of the Punt dal Gall dam Lake topography The Livigno reservoir, created by the Punt Del Gall dam, is mainly located on Italian territory with the dam half in Italy and half in the Canton of Grisons, in Switzerland. The capacity of the reservoir is million m 3, with the maximum operational level at m a. s. l.. The minimum operational level is at m a. s. l.. Currently, the water surface is never lower than m a. s. l. The total catchment area is 295 km 2. Figure illustrates an overview of the reservoir.

71 Deep, seasonal storage reservoir 61 Figure : Map showing the location of the Livigno Reservoir on the Swiss-Italian border. The construction of the Punt Dal Gall arch dam was completed in The dam, 130 m high and 540 m of crest length forms a lake with two arms. The principal west arm of the reservoir, the object of this study, is approximately 9 km long and is formed by the Spöl River. The spillway is gated and has a capacity of 280 m 3 /s. The length of the east arm is approximately 4.5 km. The first main stage of energy production consists of the Ova Spin power station in the Spöl Valley. With a capacity of 47 MW, the powerhouse turbines water from Lake Livigno in winter and pumps water to the lake (50 MW) in summer. The Ova Spin power station, with its two units is located birds-nest-like below the spillway chute of the dam, forming the compensation basin with a storage capacity of 6.2 million m 3. The main inflow is water from the Inn River which is captured at the S-chanf intake and directed through a freeflow tunnel over 15 kilometers. The water from the Ova Spin compensation basin pass through the turbines in the Pradella power station (280 MW), which represents 75% of the gross energy produced. The water flows through a 23 kilometer pressure tunnel between Ova Spin and Pradella and after the turbines, is collected in a compensation basin with m 3 capacity, which also receives water from the Inn River. This basin's contents are then further processed through another 14.3 km long pressure tunnel to the Martina power station (74 MW). Figure illustrates the full hydraulic scheme of the EKW. The entire hydraulic scheme has an average yearly energy production of some 1'400 GWh. It is property of Engadiner Kraftwerke AG (EKW), a shareholders company, owned by EGL, ATEL, BKW, CKW, NOK, the Canton of Grisons and the concessionary municipalities. A detailed description can be found on the homepage of the Engadiner Power Company under

72 62 Reservoir Sedimentation Figure : Overview of the EKW hydraulic scheme The topography for the whole reservoir is based on a map (1:2 000) with elevation curves each 2 m. This bathymetry map corresponds to the pre-impoundment topography of the submerged valley (1970). The elevations were digitalized with elevation curves every 10 m. In the vicinity of the dam, the new bathymetry, obtained in November 2003 by Straubag Ingenieure + Geoinformatiker and having a 1:500 scale was used. Both bathymetries were supplied by Engadiner Kraftwerke AG. The longitudinal average slope along the west arm of the reservoir is about 1.2 %. Figure shows a 3D view of Livigno Reservoir.

73 Deep, seasonal storage reservoir 63 Punt dal Gall dam Gallo River Spöl River Figure : 3D topography of the Livigno Reservoir Hydrologic and operation data The hydrological data for this study is based on four representative flood events for the reservoir: The annual flood, the highest flood ever measured in 1960 in the catchment area and the 100 years return period flood (floods taken from the review of the initial basic hydrologic study performed by Elektrowatt in 1992 and the October 2000 flood event). The unitary hydrograph gives a flood with a peak time of 8 h and the total time of 48 h. The base flow was considered as 10% of the peak discharge, except for the 2000 October event, where this ratio was 8%. For the annual, highest and 100 years of return period events, the maximum discharges are taken from the Elektrowatt study. For October 2000 flood event, the basic data was supplied by the client and is the result of the flood routing in the reservoir for that month where the daily values given are the average daily inflows. The maximum average value attained is 55 m 3 /s on the 14 th October. The results for the design floods with 100, and years of return period are presented in Table for average and peak values.

74 Reservoir Sedimentation Table 3.5-2: Average and peak discharge values Period: Return period (years) Discharge (m3/s) peak average Qav./Qmax Flood oct Based on the ratio of average and maximum discharge, the peak flood value for the October 2000 event is estimated to be 90 m 3 /s. An inflow hydrograph corresponding to a uniform spatial rainfall distribution was chosen. This hydrograph, suggested by Hager (1984) by the examination of the floods in mountainous regions has the form of an asymmetric bell corresponding to a modified statistical distribution of Maxwell. Q(t ) = Q b + (Q p t Qb ) e t p t (1 ) tp n Eq where Q is the inflow discharge as a function of time, Q p is the peak discharge, Q b is the base flow, t is time, t p is the time to the peak and n is a form factor (n 2). The time of the flood was adopted as 48 hours and for the peak time of 8 hours. The adopted value for the form factor n was 2, comparing the hydrograph proposed by Hager (1984) and the unitary hydrograph presented in the Eletrowatt study. Figure shows the measured hydrograph for the October 2000 flood, as well as the maximum hydrograph for this flood. Daily average discharge (m 3 /s) Day 19 Daily average discharge Peak discharge Peak discharge (m 3 /s) Figure : October 2000 flood event

75 Deep, seasonal storage reservoir 65 Figure presents the simulated hydrographs for the considered floods. 180 Discharge (m 3 /s) 160 Discharge (m 3 /s) 140 Flood Base Maximum annual years Time (h) 100 years return period 1960 flood event October 2000 flood event Annual flood Figure : Simulated flood events The data from the period between 1951 and 1968 was used to estimate the return period of these floods. A hydrological analysis based on a log-normal fit provided a return period between 5 and 10 years for the 2000 event and 30 and 40 years of return period for the maximum flood, measured in 1960 (Figure ) flood Discharge (m 3 /s) flood Period: Design floods Return period (years) Figure : Log-normal fit for the peak flood discharge at the Livigno Reservoir Sediment characteristics Since there are no measurements for the concentration of the sediments in this reservoir and it is commonly accepted that during flood events in Alpine rivers high suspended sediments concentration can occur, a concentration peak of 15 g/l was adopted. According to De Cesare (1998), who analyzed concentration measurements in an Alpine river during one entire summer, this value can be easily surpassed. In order to estimate the influence of the maximum

76 66 Reservoir Sedimentation concentration in turbidity currents for the Livigno Reservoir, an exceptional value of 30 g/l was also simulated for all hydrographs studied, except for the annual flood. The time evolution for the concentration is also based on the proposition of Hager (1984). After analyses done by De Cesare (1998) of several events, the peak time of the concentration can be considered as 85% of the peak time of the flood. These assumptions lead to a unit time evolution of the concentration as shown in Figure This figure shows also the unit hydrograph of the inflow for comparison purposes Ratio Q/Q max Inflow Concentration Ratio C/C max time (h) 0.0 Figure : Unit inflow hydrograph and unit concentration curve The density of sediments was adopted as ρ = kg/m 3 and the average diameter of grains, d 50 was assumed as 21 μm. The average diameter of the grains is presented in the study Möglichkeiten zur Abführung von Feststoffen aus dem Gebiet zwischen Fangdamm und Staumauer Punt dal Gall; Vorprojekt, realized by Philipp Huwyler (2003) Numerical Simulations using CFX-4 For the numerical simulation of the problem, some scenarios (Figure ) were defined in order to estimate the behavior of the turbidity currents in the present topography of the reservoir and the performance of the alternatives. For the current situation of the reservoir, seven simulations were performed: the 1960 and 2000 floods as well as the 100 years design flood were calculated with 15 and 30 g/l of maximum concentration, while the annual flood was only simulated with a sediment concentration of 15 g/l. The performance of the alternatives was only studied for the hydrograph corresponding to the 2000 October event and the maximum inflow concentration of sediments equal to 15 g/l. For the final analysis, the time step corresponding to 33 h 20 min ( seconds) after the beginning of the floods are utilized. At this moment, the inflow concentration is practically 0 g/l and the inflow discharge is already the base flow at the inlet.

77 Deep, seasonal storage reservoir 67 Numerical simulations Annual flood October 2000 flood 1960 flood event 100 years return flood Cmax = 15 g/l Cmax = 15 g/l Cmax = 30 g/l Cmax = 15 g/l Cmax = 30 g/l Cmax = 15 g/l Cmax= 30 g/l Obstacle 4m high Obstacle 8m high Obstacle 12m high Geotextil Screen Figure : Scenarios of simulations The boundary conditions for the variables of velocity, pressure, concentration, K and ε utilized for the simulations are presented in the Table At the inlet, constant values for all variables (Dirichlet boundary condition), except pressure are specified. The bottom is treated as a wall with a no-slip condition (velocity in all directions is 0). The water surface and reservoir banks are simulated as a plane of symmetry, where the free-slip condition for the velocity components tangential to the surface is applied and the velocity component normal to the boundary is set to 0. For the outlet, a pressure boundary condition is used. A constant pressure of 0 is specified at the top of the outlet. Zero normal gradients are applied to all other variables (Neumann boundary condition). Table 3.5-3: Summary of boundary conditions used in the numerical model Boundary conditions U i p k ε c s Symmetry (water surface and valley slopes) Free-slip Neumann Neumann Neumann Neumann Wall (bottom) No-slip Neumann Neumann Dirichlet Neumann Inlet (inlet) Dirichlet Neumann Dirichlet Dirichlet Dirichlet Pressure (outlet) Neumann Dirichlet Neumann Neumann Neumann The computational domain for the calculations is shown in Figure The discretisation of the grid is different for the simulation of the current situation and the simulations with technical measures. These grids are presented in the corresponding chapters.

78 68 Reservoir Sedimentation Figure : Computational domain for the study Current situation For the simulation of the current situation the generated grid consists of 325 cells in the inflow direction ( i ), 30 in the lateral ( j ) and 12 in the vertical directions ( k ), resulting in a total number of cells. In order to refine the grid in the invert and at the bottom, the lateral and vertical directions are not an uniform distribution. The specification of the grid is presented in Table Table 3.5-4: Characteristics of the discretisation of the grid for the current situation (SYM G.P = symmetric geometric progression; G.P = geometric progression) Direction i j k Number of cells Distribution factor Distribution Uniform SYM. G.P G.P The main analyses are made in Sections 1 and 2 (see Figure ). The principal variable considered was the deposition of sediments at the bottom of the reservoir. The final condition of the deposition for the present situation of the reservoir is illustrated hereafter for different studied scenarios (Figure to Figure ).

79 Deep, seasonal storage reservoir 69 Figure : Final condition for the annual flood with maximum inflow concentration 15 g/l Figure : Final condition for the October 2000 flood event with maximum inflow concentration 15 g/l (left) and 30 g/l (right) Figure : Final condition for the 1960 flood event with maximum inflow concentration 15 g/l (left) and 30 g/l (right)

80 70 Reservoir Sedimentation Figure : Final condition for the 100 years of return period flood event with maximum inflow concentration 15 g/l (left) and 30 g/l (right) The turbidity currents formed by the floods simulated can be considered subcritical and nonconservative (deposition > erosion) in all cases. In fact, there is no erosion in this reservoir due to turbidity currents. As shown in the Table 3.5-5, approximately 60% of the total deposited sediments are in the first 3.0 km of the reservoir for all studied floods, except for the annual flood where this deposition is approximately 74%. The quantities of settled sediments at the hydraulic structures are not relevant. This can be explained by the weak slope of the inlet bottom (1.2%) and the width of the inlet trumpet, which both decrease the velocity of currents and facilitate the deposition. The length of the reservoir is another factor reducing sedimentation close to the hydraulic outlet structures and the Punt dal Gall dam. The current loses its forces while flowing downstream due to water entrainment. The maximum height of sediment depositions varies from 3 cm for the annual flood with maximum inflow concentration 15 g/l and 10 cm for the 100 years of return period with maximum inflow concentration equal to 30 g/l. The volume deposited upstream of each location of alternatives (sections 1 and 2) was compared to the total volume deposited after 33 h 20 min for the whole west arm of the reservoir. Table shows the behavior of the deposition in the reservoir for the simulated scenarios. Table 3.5-5: Ratio between the depositions upstream of the foreseen section for the alternatives and the total deposition in the full west arm of the reservoir at the end of the simulations Flood Event Concentration (g/l) Sediment inflow (m 3 ) Deposition upstream cross section 1 Deposition upstream cross section 2 Annual % 98% October years % 94% % 92% % 92% % 91% % 91% % 90%

81 Deep, seasonal storage reservoir 71 The evolution of the current at the foreseen section obstacle for the October 2000 flood event and maximum inflow concentration equal to 15 g/l is showed in Figure The peak of the current occurs after 9 h of flood and with a maximum concentration of sediments equal to 2.4 g/l and a velocity of 0.25 m/s. Elevation (m asl) t=8h t=9h t=10h t=12h t=16h t=20h t=24h t=30h Elevation (m asl) t=8h t=9h t=10h t=12h t=16h t=20h t=24h t=30h Concentration (g/l) Velocity (m/s) Figure : Evolution of the current in the cross section of the foreseen obstacle. A longitudinal cross section at the invert of the reservoir shows the distribution of concentration for the time step 9h as illustrated in Figure Figure : Longitudinal cross section for time step 9 h without the effects of obstacle and screen At the envisaged location of the screen, the peak of the current occurs 16 h after the start of the flood and attains maximum concentration of 0.54 g/l with a maximum velocity of 0.30 m/s (Figure ) t=10h 1800 t=10h Elevation (m asl) t=12h t=16h t=20h t=24h t=30h Elevation (m asl) t=12h t=16h t=20h t=24h t=30h Concentration (g/l) Velocity (m/s) Figure : Evolution of the current in the cross section of the foreseen screen.

82 72 Reservoir Sedimentation Figure presents a longitudinal cross section at the invert of the reservoir with the distribution of concentration for the time step 16h. Figure : Longitudinal cross section for time step 16 h without the effects of obstacle and screen According to this analysis of the present situation, the alternative of installing a geotextile screen was dismissed, since more than 90% of the sediments are settling upstream of its proposed location for all cases and the current is almost completely dissolved in the reservoir at this section. Pervious dam For the simulation of the obstacles, the grid has to be deformed in order to refine the vicinity of the structure. Thus, a symmetric geometric progression is realized in the region of the obstacle (Figure ). In this case, the formed grid consists of 300 cells in the inflow direction ( i ), 20 in the lateral ( j ) and 12 in the vertical directions ( k ) resulting in a total number of cells (Table 3.5-6). Figure : Zoom of the body-fitted grid around the obstacles Table 3.5-6: Characteristics of the discretisation of the grid for simulations with obstacles Direction i j k Number of cells Distribution factor Distribution Uniform SYM. G.P G.P

83 Deep, seasonal storage reservoir 73 For the simulation of a pervious obstacle in the reservoir, three heights are considered. The obstacles with 4, 8 and 12 m height are simulated based on the October 2000 flood event with the maximum concentration of sediments reaching 15 g/l. Figure presents the situation of deposition for the three studied obstacles at the end of the simulations. The final situation without any obstacle is also presented in this figure. Figure : Final condition for the October 2000 flood event with a maximum inflow concentration of 15 g/l for the present situation (up left) and obstacles with 4 (up right), 8 (down left) and 12 m height (down right) Table shows the ratio of sedimentation for the different studied heights of obstacle as well as the volume of material needed to construct the obstacles. Table 3.5-7: Ratio of the depositions upstream of the foreseen sections for the alternatives and the total deposition in the full west arm of the reservoir at the end of the simulations for different obstacle heights Alternative Deposition upstream of the Section 1 Deposition upstream of the Section 2 Volume of the rockfill dam (m 3 ) Obstacle 4 m 69% 95% 1'700 Obstacle 8 m 76% 97% 12'300 Obstacle 12 m 87% 99% 34'600

84 74 Reservoir Sedimentation The performance of an obstacle against turbidity currents starts to be significant for the obstacles higher than 8 m. In this case, the sedimentation upstream of the location of the obstacle increases from 63% to 76% for the obstacle with a height of 8 m and from 63% to 87% for the obstacle with a height of 12 m. A longitudinal cross section in Figure at the invert of the reservoir shows the distribution of concentration at the peak of the current in the Section 1 for the different studied heights. Figure : Longitudinal cross section at time step 9 h with the effects of obstacles of 4, 8 and 12 m height (up to down) Concerning the sedimentation along the reservoir, Figure shows a longitudinal cross section at the invert of the reservoir with the maximum deposition for the October 2000 event around 5 cm upstream of the obstacle.