BOOK OF ABSTRACTSruif/fluvial_habitats_project/workshops/... · João Fernandes Uniform and...

3RD INTERNATIONAL WORKSHOP ON

RIVER AND RESERVOIR HYDRODYNAMICS AND MORPHODYNAMICS

Current trends in experimental research and numerical modeling

Organized within the scope of project PTDC/ECM/099752/2008, supported by FCT (Portuguese

Foundation for Science and Technology), by IMAR-CMA, FCTUNL, CEHIDRO-IST, and LNEC.

http://www.civil.ist.utl.pt/~ruif/fluvial_habitats_project/workshops3.html

Department of Civil Engineering, Faculty of Sciences and Technology

New University of Lisbon, Caparica, Portugal

9th

– 10th

July 2012

BOOK OF ABSTRACTS

LIST OF CONTRIBUTIONS

page

Ana Margarida Bento Characterization of dam breaching following overtopping 1

Keywords: dam break, overtopping, outflow hydrograph

Ana M.C. Ricardo Hydrodynamics of flows within arrays of rigid emergent stems 3

Keywords: TKE, vegetated flow, PIV measurements, dissipation rate

Anxo Barreiro Aller Sediment Transport with DualSPHysics 5

Keywords: meshless method, shields criterion

Daniel Conde Mathematical modelling of a tsunami propagating over the mobile bed

reaches of the Tagus-estuary 7

Keywords: numerical modeling, tsunami, Tagus-estuary

Edgar Ferreira Mitigation of turbidity currents in reservoirs with passive retention

systems: preliminary CFD results. 9

Keywords: reservoirs, density currents, CFD, algebraic slip model

Gensheng Zhao Hydraulics and erosion process in the embankment breach 11

Keywords: flow erosion, embankment breach

Gregor Petkovsek Challenges in modelling long term reservoir sedimentation processes 13

Keywords: reservoir, sedimentation, flushing, management

Helena I.S. Nogueira Experimental investigation on gravity currents 15

Keywords: gravity currents, PIV measurements, entrainment

João Fernandes Uniform and non-uniform flows in compound channels 17

Keywords: compound channel, floodplain, overbank flow

Marina Filonovich Numerical modelling of compound channel flow 19

Keywords: compound channel, EARSM, energy balance

Moisés Brito CFD modelling of flows over rough floodplain 21

Keywords: CFD, rough bed, porous bed, compound channel flows

Mona Jafarnejad Failure risk of flood protection measures due to modified sediment

transport under climate change 23

Keywords: probabilistic simulation, failure mechanisms, sediment transport,

riprap

Ricardo Azevedo Experimental characterization on compound-channel turbulent field

without and with vegetation 25

Keywords: asymmetric compound-channel, laser Doppler velocimeter, turbulent

scales, dissipation rate

Ricardo Canelas SPH-based numerical simulation of the velocity field in a dam-break flow 27

Keywords: SPH, dam-break, boundary layer, viscosity

Sergi Capapé Limiting concentration of transported fine sediment 29

Keywords: wash-load, deposition, limiting concentration, suspended sediment

transport

Silvia Saggiori Anysotropy in vegetated natural flow 31

Keywords: turbulent flows, vegetated flow, anisotropy, field measurements

Valentina Lombardi Analysis of a gravity current moving on an unsloping bed 33

Keywords: gravity currents, lock exchange flows, PIV measurements

3RD INTERNATIONAL WORKSHOP ON RIVER AND RESERVOIR HYDRODYNAMICS AND MORPHODYNAMICS

Department of Civil Engineering, Faculty of Sciences and Technology, New University of Lisbon Caparica, Portugal 9th – 10th July 2012

1

Characterization of dam breaching following overtopping

Ana Margarida BENTO & Rui M. L. FERREIRA

Instituto Superior Técnico, Lisbon, Portugal, [email protected]

The large quantities of water suddenly released and the consequences of dam failure, such as

the property damage and the loss of life, are all of prime importance to our society. Thus, in

order to prevent us from its devastating potential, it is necessary an accurate prediction of

breach parameters. An example of that was the sudden collapse of Malpasset dam, which gave

added visibility to the problem leading to the development of dam safety programs in most

countries of the world as a result.

The main objective is to study the changes in characteristics of dam materials, for its stability,

peak discharge and breach development time. In particular, this research attempts to

characterize the influence of variable size and sedimentology on the breach development

process and the resultant peak outflow.

To accomplish this objective, an existent numerical dam breach model is being modified,

essentially to include more geotechnical processes. This model is applied to a simplified failure

of homogeneous embankments caused by overtopping. The model consists of the dynamics of

water within the reservoir, the hydraulics of the flow through the breach, and the erosion of

the embankment crest and along the downstream face of the embankment.

The Equation 1 reflects, in a simply way, the strong interactive processes which occur between

the water flow and the bed form, namely in movable boundary hydraulics.

�� = �� − �� (1)

where DAb is the variation of the cross section, FV is the vertical flux of sediment and Elat is

the lateral erosion. When DAb is negative, a scour or erosion takes place. If DAb equals to zero,

the sediment load is compatible to the flow transport capacity. For a DAb positive value, an

increase of channel bed elevation, known as accretion or deposition, occur.

In order to study the progression rate, needed to evaluate the role of the geotechnical issues

on the breach formed, the Darcy law is incorporated in the numerical model. In particular the

Darcy law to unsatured states, to determine the flux of water, given by Equation 2.

�� = �×�� × (�� − �� × �) (2)

where ql is the discharge, �� is the liquid pressure, �� is the liquid density, g is gravity, μ� is the

liquid viscosity, !"� is the percentage in area of the voids filled with the fluid and ! is saturated

hydraulic conductivity given by Equation 3.

� = !# × ∅%(&'∅)( × (&'∅))(

∅)% (3)

The model will be calibrated with laboratory and field data. The former concerns experiences

carried out at Laboratório Nacional de Engenharia Civil (LNEC) and the latter data available in

the literature.

An accurate comparison between experimental and numerical results is the ultimate aim of

this work.



2



3

Hydrodynamics of flows within arrays of rigid emergent stems

Ana M. RICARDO1, M.J. FRANCA2, Anton SCHLEISS3 & Rui M.L. FERREIRA4

1CEHIDRO - IST - TULisbon, Portugal & LCH - EPFLausanne, Switzerland, [email protected]

2Faculty of Sciences and Technology & IMAR- CMA, New University of Lisbon, Caparica, Portugal

3Laboratory of Hydraulic Construction (LCH) - ENAC – EPFL, CH - 1015 Lausanne, Switzerland

4CEHIDRO - Instituto Superior Técnico - Techical University of Lisbon, Portugal

Within the bioengineering framework, designing a non-erodible channel is a complex fluid

dynamics problem as it involves knowing the drag exerted on the boundary, the drag exerted

on the plant stems and the overall friction slope. Most of the existing design criteria employ

resistance formulas such as Manning’s, calibrated ad hoc. Moving toward physically based

design criteria, progresses have been made in the characterization of 3D flows over irregular

boundaries, mainly due to the application of double-averaging methods (DAM). Those

methods are a particular form of upscaling, in both time and space: the conservation equations

of turbulent flows are expressed for time-averaged quantities which, in case of unsteady flow,

are defined in a time-window smaller than the fundamental unsteady flow time-scale, and for

space averaged quantities, defined in space windows larger than the characteristic wavelength

of the boundary irregularities. Such methods are especially pertinent for the characterization

of the flow within and in the near vicinity of plant canopies.

In what concerns vegetated boundaries, the so obtained Double-Averaged Navier-Stokes

equations (DANS) have been successfully employed to calculate the drag exerted on arrays of

identical stems or the drag on plants with dense foliage. DANS are especially pertinent, at the

adequate scales, to calculate the drag partition and the bulk flow resistance of a stem array.

Form induced stresses have been seen to be relevant in flows within moderately to dense

stem arrays and they are key features in DANS and in Double-Averaged Turbulent Kinetic

Energy equation (DATKE), representing the mean effect of advection at scales smaller than

that of the space averaging operator. Closure models for form induced stresses and for the

production terms associated to these stresses are not available for most flows with vegetated

boundaries. Determining these closures is a demanding research challenge and represents a

decisive step in the path towards physically based design criteria for vegetated channels.

The general objective of this PhD program is to devise a closed conceptual model for the flow

over vegetated boundaries at scales of the order of magnitude of the larger wavelength. To

attain this general objective a detailed description of the flow within the inter-stem space is

needed, once the development of modeling tools for larger scales requires a good

understanding of the flow behavior at small scales. The immediate progresses and

methodologies of this program pertain to fluid mechanics domain; however it should be

emphasized that the results will eventually mature into bioengineering design approaches,

what will positively impact river engineering practices.

The methodological proposal is mainly experimental, reproducing in laboratory, flows within

boundaries covered by rigid, cylindrical and emergent vegetation. Two different tests were

carried out: a test with a constant density of vegetation elements of 400 stems/m2 and a test

where two stem densities (1600 and 400 stems/m2) were alternated with a fixed wavelength,

in order to simulate spatial patchiness. The experimental work was carried out in a 12.5 m long

and 0.408 m wide tilting recirculation flume of the Laboratory of Hydraulics of IST. The flume



4

has glass side walls, enabling flow visualization and laser measurements. The flume bottom

was covered with a thin horizontal layer of gravel and sand and arrays of rigid, vertical and

cylindrical stems were randomly placed along of a 3.5 m long reach simulating emergent

vegetation conditions. The stems have a diameter of 1.1 cm. Downstream the vegetation-

covered reach, a coarse gravel weir controlled the flow. The experimental data acquisition

consisted in horizontal and vertical instantaneous velocity maps obtained with a 532 nm, 30

mJ 2D Particle Image Velocimetry (PIV) system operated at 15 Hz. The PIV is an optical

technology that allows obtaining the fluid velocity by measuring seeding particle velocity.

The PIV measurements allow the analysis of instantaneous velocity and vorticity fields, both in

time and space, within the production and eventually the first scales of the inertial sub-range

of the spectral space. A spatial analysis of this kind of flows is particularly interesting since it

avoids the approximation of the frozen turbulence theory. The characterization of the flow is

done by means of velocity and vorticity time average maps, mean Reynolds and form-induced

stresses, autocorrelation functions, second and third order structure functions, spectral

analysis, statistics of occurrence, permanence and intensity of the coherent turbulent

structures and terms of convection, flux, production and dissipation of turbulent kinetic

energy.

One of the main characteristic of the flow through arrays of rigid stems is the great spatial

variability in the inter-stem space, appearing lower velocities in the wake of the stems and

higher in regions between two stems. Thus, it will be possible to observe a shear layer dividing

regions of relative low and high velocity where one can expect to find instabilities akin to

Kelvin-Helmholtz’s. A von Kármán vortex street is clearly visible in the wake of the stems

showing a quasi-symmetric high-vorticity pattern around the stems independently of the stem

density. However, stem density impacts the length of the vortex street: for lower densities the

inter-stem space is larger allowing the development of the vortex street. Concerning Double-

Averaged variables, a strong correlation between stem density and the magnitude of the flow

variables was found. Both turbulent and the form induced stresses increase with the density of

the vertical elements. Form-induced stresses cannot be neglected compared with Reynolds

stresses; they have the same order of magnitude. Normal longitudinal form-induced stresses

are the main expression of flow heterogeneity and spatial anisotropy. Flow complexity

increases towards the bed, expressed by the increase of form-induced shear and normal

vertical stresses. Spectral analysis showed that, in the reach of a specific stem density, spectral

signature from upstream patches, with different stem densities are still visible. The patchiness

of the array of stems does not hide the specific contribution of each stem density to the

energy budget.

Although a considerable amount of work has already been performed, especially concerning

flow characterization, it should be noticed that there are still some questions to be solved in a

work to understand the quite complex interaction of different wavelengths in flows over

vegetation covered boundaries. Advances on spatial heterogeneity are also expected.

The study was funded by the Portuguese Foundation for Science and Technology research

project PTDC/ECM/099752/2008.



5

Smoothed Particle Hydrodynamics modelling of ship induced sediment transport

Anxo BARREIRO, Alejandro J. C. CRESPO, Jose M. DOMINGUEZ & Moncho GOMEZ-GESTEIRA

EPHYSLAB, University of Vigo, Campus as Lagoas s/n, 32004, Ourense, SPAIN, [email protected]

Harbor traffic has increased significantly during the last decades, in particular the presence of

ocean liners has intensified in some areas. They arrive at the innermost parts of the harbors,

where they try to carry the docking and undocking maneuvers in the shortest time possible.

Sediment scours constitute an important problem in harbor activity since it can significantly

weaken coastal structures. The fast movement of ships in shallow water can induce important

scour rates due to the extreme currents generated by thrusters, propellers and by the hull

itself. For example, ships roll-on/roll-off or ships with lateral propellers have been referenced

(Sumer and Fredoe, 2002) as sources of dangerous erosion problems at harbours (Figure 1).

Figure 1. Sketch of roll-on/roll-off ships (left). Sketch of ships with lateral propellers (right).

An approach to quantify scour rates is the use of laboratory experiments, but their cost is high

and models can suffer from important scale-effects. Laboratory tests should be complemented

with numerical simulations, which are complex due to the multi-physics nature of the problem

where water, soil, moving ships and coastal structures are involved. Classical models, which

rely on grid-based techniques, cannot be efficiently used for complex problems like this, where

not only different phases are involved but also the bathymetry of the basin changes due to

erosion and accretion. Among the meshfree methods, Smoothed particle Hydrodynamics (SPH)

(Gingold and Monaghan, 1977) is one of the most well-established models. During the last

decade Smoothed Particle Hydrodynamics (SPH) has become increasingly popular to study

free-surface flows (Gomez-Gesteira et al., 2010). The conceptual simplicity of these Lagrangian

models lies in the fact that the fluid is decomposed into a series of points, the particles, which

interact with each other according to the fluid conservation laws.

The particular SPH code selected for the present work is the so called DualSPHysics, which was

developed as a joint effort by researchers at the University of Vigo (Spain) and the University

of Manchester (U.K.) and it is available as open-source at (www.dual.sphysics.org).

DualSPHysics is designed to be run on either multicore CPUs or GPUs and allows important

speedups in execution time in such a way that the model can be used to simulate complex and

realistic scenarios (Crespo et al., 2011).

The use of the model will assure a good description of the flow velocity, even under extreme

condition like those involving the fast rotation of the moving parts of the ship (e.g. the

thrusters) or strong collisions between the fluid and the structures. Apart from an accurate

description of the fluid movement, the problem under scope implies the existence of different



6

phases (water, soil and suspended sediment). Sediment transport has been previously

considered by several authors using SPH methods. Zou and Dalrymple, 2006 analyzed fluid-

sediment interaction. Other authors (Bui et al., 2008) also investigated soil-water dynamics. In

addition, (Manenti et al., 2012) used SPH to simulate sediment flushing induced by a rapid

water flow. Finally, (Ulrich et al., 2011) modeled water/soil flows using a variable resolution

SPH scheme. These authors also investigated harbour-flow problems similar to the ones

described here, showing that 3D simulations should be considered to obtain realistic results.

Different phases were considered in most of the studies described above: (a) the fluid, which

can be described using standard SPH formulation; (b) the erodible sediment, which is

constituted by particles that can be eventually re-suspended, in such a way that it is allowed to

move according to the equations of movement of a pseudo-fluid; (c) the hidden bed, which

does not interact with the rest of the sediment or the fluid as long as there are no water

particles in its neighborhood. This “frozen” behavior can change in time when the erodible

sediment is removed and the hidden bed becomes exposed. In addition, the problem implies

the existence of different interfaces which can change both in time and in space.

Different modules are under development in DualSPHysics to simulate sediment transport

based both on classical sediment theory (van Rjin, 1993) and on previous studies (Ulrich et al.,

2011; Manenti et al., 2012). Preliminary simulations of sediment transport generated by ship’s

propellers are shown in Figure 2 where an oversimplified bathymetry with an initially flat canal

has been considered.

Figure 2. Sediment transport caused by the ship's propeller.

To sum up, the main goals of the proposed work are: (a) creating a numerical tool to analyze

ship induced sediment transport; (b) validating the numerical tool by means of laboratory

experiments and in situ measurements; (c) designing counter measures to palliate the scour on

coastal structures.

References Bui, H.H., Fukagawa, R., Sako, K., Ohno, S., 2008. Lagrangian mesh-free particles method (SPH) for large deformation and failure

flows of geomaterial using elastic-plastic soil constitutive model. Int. J. Num. Anal. Meth. In Geomechanics, 32(12), 1537-1570.

Crespo, A.J.C, Dominguez, J.M., Barreiro, A., Gómez-Gesteira, M., Rogers, B.D., 2011. GPUs, a new tool of acceleration in CFD:

Efficiency and reliability on Smoothed Particle Hydrodynamics methods. PLoS ONE. 6(6), e20685.

Gingold, R.A., Monaghan, J.J., 1977. Smoothed particle hydrodynamics: Theory and application to non-spherical stars. Monthly

Notices of the Royal Astronomical Society 181, 375–389.

Gómez-Gesteira, M., Rogers, B.D., Violeau, D., Grassa, J.M., Crespo, A.J.C., 2010. SPH for free-surface flows. Journal of Hydraulic

Research 48, 3-5. doi: 10.3826/jhr.2010.0014

Manenti, S., Sibilla, S., Gallati, M., Agate, G. And Guandalini, R., 2012. SPH Simulation of Sediment Flushing Induced by a Rapid

Water Flow. Journal of Hydraulic Engineering, doi:10.1061/(ASCE)HY.1943-7900.0000516.

Sumer M. and Fredsoe J., 2002. “The mechanics of scour in the marine environment”. Advanced Series on Ocean Engineering –

Volume 17. World Scientific Publishing Co. Pte. Ltd.

Ulrich, C., Koliha, N., Rung T., 2011. SPH Modelling of water/soil-flows using a variable resolution scheme. Proc. 6th Int. Spheric

Workshop, June 07-10 Hamburg (De), 101-108.

van Rijn, L.C., 1993. Principles of sediment transport in rivers, estuaries, and coastal seas, Aqua Publications.

Zou, S. and Dalrymple, R.A., 2006. Sediment Suspension Over Ripples under Oscillatory Flow. In Proceedings 26th International

Conference on Coastal Engineering.



7

Mathematical modeling of a tsunami propagating over the mobile bed reaches of the Tagus

estuary

Daniel CONDE, Ricardo CANELAS & Rui M. L. FERREIRA

TU Lisbon, Instituto Superior Técnico, Civil Engineering and Architecture, Portugal,

[email protected]

A recent revision of the catalog of tsunamis in Portugal (Baptista and Miranda, 2009) has

shown that Tagus estuary has been affected by catastrophic tsunamis numerous times over

the past two millennia. One of the most relevant features of the propagation of the tsunami

over solid boundaries is its ability to incorporate debris, either natural sediment incorporated

from the bottom or remains of human built environment. Often, these waves generate

important geomorphic impacts, changing the shoreline and the morphology of estuarine

regions (Dawson, 1994).

The earthquake of 1755 generated a tsunami whose impacts on the Tagus estuary are well-

documented. It is known that the region of Palhais, Barreiro, suffered severe morphological

changes that led to the destruction of Vale do Zebro Royal Complex, namely the factories and

ovens for wheat processing.

The objective of this work is to simulate the propagation of a tsunami whose wave crest is

compatible with that observed in 1755 at Bugio in the Tagus-Coina estuary in today’s

bathymetry and altimetry conditions. The emphasis is placed in the analysis of the

morphological changes occurring in the Vale do Zebro region, where, presently a military

facility is located.

The model employed for the simulations is based on a 2DH shallow-flow solver applicable to

discontinuous waves over complex time-evolving geometries. A finite-volume discretization

scheme is employed, based on a fluxsplitting technique incorporating a reviewed version of

the Roe Riemann solver. The model remains conservative, provided that source terms are

properly formulated and the stability domain reevaluated accordingly. It has been validated in

a benchmark test featuring a mobile bed, clearly showing its main advantages and

shortcomings.

References

Baptista M.A. Miranda, J.M. (2009). Revision of the Portuguese catalog of tsunamis. Nat.

Hazards Earth Syst. Sci., 9, 25–42.

Dawson, A.G. (1994). Geomorphological effects of tsunami run-up and backwash,

Geomorphology, 10, 83–94.



8



9

Mitigation of turbidity currents in reservoirs with passive retention systems: preliminary CFD results

Edgar A. C. FERREIRA1, Elsa C. T. L. ALVES1 & Rui M. L. FERREIRA2

1LNEC, Avenida do Brasil 101, 1700-066 Lisboa, [email protected]

2CEHIDRO – IST – TULisbon, Avenida Rovisco Pais, 1 — 1049-001 Lisboa

Sediment deposition by continuous turbidity currents may affect eco-environmental river

dynamics in natural reservoirs and hinder the maneuverability of bottom discharge gates in

dam reservoirs. In recent years, innovative techniques have been proposed to enforce the

deposition of turbidity further upstream in the reservoir and away from the dam, namely,

among others, the use of solid and permeable obstacles such as water jet screens and

geotextile screens.

In this presentation CFD preliminary data and impressions so far from the project

PTDC/ECM/099485/2008 - ControlSed - Control of sedimentation in reservoirs induced by

turbidity currents - funded by the Portuguese Foundation for Science and Technology (FCT) will

be presented. The purpose of this research project is to clarify, based on experimental and

numerical modeling work, the complex interactions between turbidity currents and passive

retention systems, designed to induce sediment deposition.

Recently [1] has assessed the ability of the Algebraic Slip Model in the 3D CFD software ANSYS-

CFX to simulate hyperpycnal turbidity currents in reservoirs. In this presentation, major

conclusions will be revisited and perspectives for the near future research plans will be

launched.

REFERENCES

Ferreira, E.A.C.; Alves, E. & Ferreira, R.M.L (2012) Mitigation of turbidity currents in reservoirs

with passive retention systems: validation of CFD modeling. Geophysical Research Abstracts,

Vol. 14, EGU2012-13522, 2012, EGU General Assembly, Vienna, April 2012.



10



11

Breach Growth in Cohesive Embankments

Gensheng ZHAO, Paul J. VISSER & Han K. VRIJLING

Department of Hydraulic Engineering, Delft University of Technology stevinweg 1, Delft, 2628CN, the

Netherlands, [email protected]

Embankments, including dikes and dams, are of large benefit to people all over the world.

Since human civilization thousands of years ago, embankments have been playing a vital role

in the development of human being. The history of embankments is epitome of the rise and

fall of human civilization, especially regarding the flood defenses and irrigations from rivers

and lakes. The magnitude and extent of the losses depend highly on the rate of the breaching

of embanks, which in turn determines the discharge through the breach and the speed and

rate of inundation of the polder, the areas outside the embankments or downstream of the

breach. Therefore, the modeling of breach evolution in embankments, the predicting of the

breach parameters (e.g. depth, width, discharge) and the breach flow rate, is of significant

interest to damage assessment and risk analysis. It is also important for the development of

early warning systems for dike and dam failures and for evacuation plans of people at risk.

An embankment dam breach is a complex process, which can be divided into initial erosion,

headcut erosion, breach widening erosion and breach deepening erosion. In the different

process of breach, the breaching flows have their own characteristics. The flow is a kind of

converged flow in the breach. The characteristics change with the development of the breach,

i.e., the breach erosion process affects the hydraulic characteristics (See Figure 1 and Figure 2),

and the hydraulics provides feedbacks to the breach erosion.

The erosion of undisturbed clay by flowing water is an important issue in the breach

development of dikes and dams. Due to many parameters involved, the identification and

prediction of undisturbed clay erosion is a complicated problem in geomorphology and

hydraulic engineering. While the pickup of non-cohesive sediment through discrete particle

entrainment may be quantified (if the flow velocities are not too large) by the magnitude of

shear stress and particle size, however, undisturbed clay is eroded through entrainment of

aggregates. The cohesive strength between and within aggregates makes the erosion complex.

According to the former study analysis, the headcut erosion plays a significant role in the

breach process of cohesive embankments, however, the mechanism of headcut erosion still

needs to be understood. Although various breaching experiments have been conducted in the

past, most of them focused on the mechanism of the breach generally. As part of the breach,

the mechanisms of the headcut erosion (See Figure 3) are not fully understood to study the

breach process and simulate with mathematical models.



12

Figure 1. Water Level Distribution in the Breach of Flume Experiment

Figure 2. Depth-averaged Velocity Distribution in the Breach of Flume Experiment

Figure 3. Headcut in the Field Breach Experiment in Lillo-Fort, 6 May, 2012



13

Challenges in modeling long-term reservoir sedimentation processes

Gregor PETKOVSEK & Marta ROCA COLLELL

HR Wallingford, Howbery Park, Wallingford, Oxfordshire, OX10 8BA, United Kingdom,

[email protected]

Sediment deposition in reservoirs is a serious problem in regions where rivers carry significant

sediment loads. Every year, sedimentation causes an estimated 1% reduction in the total

capacity of all reservoirs worldwide. For sustainable reservoir management, the rates at which

existing storage capacity is lost must be assessed, and prediction of future storage losses and

effectiveness of sediment management measures must be made.

Ideally, 3D models should be applied to reservoir modeling to capture all processes in a

reservoir (flow distribution, stratification, flow at dam structure). However due to significant

number of simulations that are typically required for practical planning, as well as to assess the

uncertainty, this is often not a viable option. Possible simplifications of the models include

quasy steady assumption; usefulness of 2DV models may also be discussed to simulate flow of

sediment towards intakes/outflows at different elevations.

When using 1D models, the distribution of bed changes across cross section must be assumed.

For deposition, usually a “flat bed” is assumed or deposition proportional to flow depth. For

erosion this is less clear, and is a matter of discussion. This is true in particular during sediment

flushing operations, when a flushing channel forms. The width of this channel must be

assumed. This is usually based on regime formulae, either empirical or based on some

extremal hypothesis principles. A formula for flushing channel width has been developed

based on discharge only, however the majority of data used for its derivation comes from one

reservoir only. Discussion of influence of other factors is interesting, as well as alternative

approaches to cross-distribution of erosion.

Figure 1. Flushing channel development (after White 2000).

In reservoirs, transport, deposition and erosion of fine sediment is of particular importance.

Cohesive sediment plays important role. Several approaches exist to model flocculation and

hindered settling. It is commonly understood that settling velocity under the mentioned

conditions depends on concentration and turbulence parameters, usually described with shear

stress. Models of various degrees of complexity exist. Usually, the output of such a model is



14

average, or 50% settling velocity of the settling flocs. For reservoir deposition progress,

however, distribution of settling velocity of flocs (which depends on their type and size) should

also be known. The key question for discussion is therefore the most appropriate practice for

selection of deposition model that can output both average settling velocity of the flocs and

their distribution.

In particular for flushing operation, model should be applicable to high shear stresses as well,

which is another possible point for discussion. High shear stress causes flocs to break up at

some point.

Another topic is erosion of cohesive material, as well as cohesive – non-cohesive mixtures. The

rate of erosion is proportional to excessive shear stress. The selection of critical shear stress is

considered to be dependent on density of deposited material.

Density currents occur in reservoirs. For sediment transport, turbidity currents are most

important and can have a significant role in transporting sediment towards the dam, including

venting turbidity currents through low level outlets. Modeling of density currents can either

take a simple form of velocity computations based on friction and concentration, slope, steady

flow analogy, or a full system of equations for momentum and fluid and sediment mass

conservation. Plunge point, stability criteria, importance and modeling of clear water

entrainment, as well as most efficient method for simulation of long-term multi-fraction

turbidity current transport, all provide points for discussion.

Figure 2. Turbidity underflow in reservoir (after Simoes & Yang 2006).



15

Experimental investigation of gravity currents

Helena I. S. NOGUEIRA1, Claudia ADDUCE2, Elsa ALVES3 & Mário J. FRANCA4

1Faculty of Sciences and Technology & IMAR, University of Coimbra, Pólo II – 3030 Coimbra, Portugal,

[email protected] 2Department of Civil Engineering, University of Rome “Roma Tre”, 00146, Rome, Italy

3LNEC, Avenida do Brasil 101, 1700-066 Lisboa

4 Faculty of Sciences and Technology & IMAR, New University of Lisbon, 2829-516 Caparica, Portugal

Gravity currents, driven by buoyancy differences between two contacting fluids, encompass a

wide range of geophysical flows which can be originated by temperature, dissolved substances

or particles in suspension. In the water one may refer oceanic fronts, resulting from

differences in temperature and salinity, and turbidity currents caused by high concentration of

suspended particles. The release of pollutant materials into rivers, oil spillage in the ocean and

desalination plant outflows are a few examples of anthropogenic gravity currents frequently

with negative environmental impacts. Thus, the study of this phenomenon is important in

engineering sciences, namely in what concerns industrial safety and environmental protection.

The dynamics of gravity currents, i.e., the entrainment of ambient fluid into the current and

the mixing processes involved, have not been totally explored, namely in what concerns the

influence of the bed roughness. Therefore, the present work aims at contributing to the

understanding of the phenomena involved through experimental work, where both lock-

exchange and continuously fed density currents, induced by salinity differences, are

reproduced under controlled conditions on an open channel.

Four runs were performed with the PIV system where the roughness of the bed, ε, is the

varying parameter, maintaining the water depth h0 = 0.2 m and initial density of the saline

mixture ρ1 ≈ 1015kgm3, leading to runs D1 (ε = 0), R1 (ε = 2.9 mm), R2 (ε = 4.6 mm) and R3 (ε =

24.6 mm).

A double pulsed Nd:YAG laser was used to measure the 2D instantaneous flow velocity, being

the pair of images acquired at 3 Hz with 30 ms between pulses. The laser of the PIV system

was placed at the right end wall of the channel, being the laser sheet adjusted to illuminate the

vertical centerline of the flow. A CCD camera was kept at a fixed perpendicular position to the

sidewall capturing a field of view of 0.33 m long and 0.18 m deep, starting from x = 1.5 m, as

shown schematically in Fig. 1 (right). Both dense and ambient fluids were seeded with

polyamide particles with a mean diameter of 100 μm and mean density of 1016 kgm3.

Figure 1 shows a vorticity map obtained for the run performed with smooth bed for the instant

when the current reaches the limit of the visualization window.

Typically, the vorticity field shows a region of positive vorticity in the interface between the

two fluids, arising from the shear between current and ambient fluid, while negative vorticity

is observed near the flow bed as a result of the no-slip boundary condition. Increasing bed

roughness reduces the overall vorticity magnitude, which was expected from the more

homogenized distribution of the streamwise velocity in the vertical direction.

Experiments of continuously fed gravity currents are being performed to investigate the effect

of bed roughness on velocity, vorticity and turbulence in steady gravity currents. PIV results

will be used to characterize the space and time intermittency of the current upper fluctuating



16

boundary, including eventual extra contribution on momentum transfer through dispersive-

type processes.

Figure 1. Instantaneous velocity map superimposed on the vorticity field. Run performed with smooth

bed.





17

Flow in a straight compound channel

João Nuno FERNANDES1,3, João B. LEAL2 & António HELENO CARDOSO3

1LNEC, Av. do Brasil, 101, 1700-066 Lisbon, Portugal, [email protected]

2CEHIDRO & Faculty of Sciences and Technology, New University of Lisbon, Caparica, 2829-516, Portugal 3CEHIDRO & Dept. of Civil Eng. and Architecture, Instituto Superior Técnico, Lisbon, 1049-001, Portugal

In this presentation the main aspects of the research on the compound channel flows is

shown. After some tests conducted in a single channel in order to test the equipment and to

characterize the roughness of the bottom, the work in compound channels comprises three

different phases. The first phase consisted on the study of the uniform flow with both smooth

and rough floodplains. In the second phase an inlet non-uniform flow was imposed and the

longitudinal evolution of the flow was studied. The last phase was devoted to the study of the

influence of the vegetation in the main channel/floodplain interface. A schematically

representation of each experimental phase is presented in Table 1.

Table 1. Schematically representation of each experimental part.

Main channel

bed Floodplain bed Flow

Vegetation in the

floodplain edge

Schematic

representation

Polished

concrete -

Uniform -

Artificial grass - -

Polished

concrete

Polished concrete

Uniform -

Artificial grass

Polished

concrete Polished concrete

Non-

uniform -

Polished

concrete Artificial grass Uniform

Rods

Artificial shrubs

The flume is 10 m long, 2 m wide and it is made of polished concrete with a bottom slope of

1.1 mm/m. The inlet is made separately for the main channel and for the floodplains in

accordance with the recent literature recommendations. The symmetric cross-section is

composed of a trapezoidal main channel (0.6 m wide at the bank full level, with a side slope of

45º) and of two lateral flood plains (0.7 m wide each). The bank full height is 0.1 m.

For uniform flow the flow structure and the influence of the momentum transfer in the

channel conveyance was studied for nine different water depths (six with smooth floodplains

and three with rough floodplains). Taking into account the measurement of the velocities

along the flume, the longitudinal variation of the mixing layer has been analyzed (e.g. the

variation of the lateral shear stress is presented in Figure 1).



18

x – distance to upstream section

y – distance left bank

Bf – floodplain width

z – local vertical distance

hm – water depth in the main channel

Um – main channel velocity

Uf – floodplain velocity

Figure 1. Distribution of the normalized lateral shear stress, under uniform flow conditions.

The non-uniform flow was achieved by disequilibrium in the inlet flow distribution identified

by the percentage of variation in the flood plain inflow, compared to uniform flow conditions.

Three cases of overfeeding of the floodplain and one case of underfeeding were studied for

two water depths. The influence of this non-uniform flows in the depth averaged lateral shear

stress is presented in Figure 2.

Figure 2. Cross sectional distribution of the depth averaged lateral shear stress, τxy for several non-

uniform flows.

In the last part of the present work, tests with vegetation elements in the interface between

the floodplain and the main channel were done (cf. Figure 3).

Figure 3. Presence of the vegetation in the interface between the main channel and the floodplains.

After the calibration of the uniform flow, the influence of the vegetation elements on the

velocities and turbulence fields was analyzed.



19

Numerical modeling in compound channel flow

Marina FILONOVICH1, Luis R. ROJAS-SOLORZANO2 & João B. LEAL1

1CEHIDRO & Faculty of Sciences and Technology, New University of Lisbon, Caparica, 2829-516, Portugal,

[email protected] 2Department of Energy Conversion and Transport, Universidad Simon Bolivar, Caracas, Venezuela

This study is related to numerical modelling in compound channel flow, where the fast flow in

the main channel (MC) is retarded by the slower flow on the floodplain (FP), causing lateral

momentum transfer. The shear layer that develops at the interface of the MC and the FP by

the difference of velocities affects turbulence structures, and streamwise and vertical vortices

are developed (Figure 1).

Figure 1. Three dimensional description of compound channel flow by Shiono and Knight (1991)

In the simulation of the velocity field in compound channel flow using different closure models

a comparison of three turbulence closure models with experimental data was performed. For

this purpose k-ε model, Shear Stress Transport (SST) model and Explicit Algebraic Reynolds

Stress Model (EARSM) were employed. The k-ε model and SST model, both isotropic models,

did not produce secondary flows, while EARSM, anisotropic model, was able to simulate

secondary flows caused by turbulence anisotropy.

Verification and validation of a turbulence closure model was performed for an experimental

compound channel flow using Explicit Algebraic Reynolds Stress Model (EARSM) simulations.

The Grid Convergence Index (GCI) approach was adopted to evaluate the uncertainty

associated to grid resolution. The GCI results presented low values for u velocity component,

but higher values in what concerns v velocity component, w velocity component (represent

secondary flow) and for Reynolds stresses RSxy and RSyz. This indicates that the mean flow has

converged but the turbulent field and secondary flows still depend on grid resolution.

Additionally to GCI analysis correlation analysis were performed for estimating the mesh

quality in what concerns small value variables. Comparison of numerical and experimental

results showed good agreement.

The budget of mean kinetic energy (MKE) for an asymmetric compound channel flow is the

present research study case. Each term from equation 1 for the budget of the MKE is

calculated separately averaging transient results obtained using the Baseline Explicit Algebraic

Reynolds Stress Model (BSL EARSM). A high resolution mesh consisted of 12.5 million elements



20

in total. The mesh spacing in terms of wall units are ∆x+ = ∆y

+ = 10 in streamwise and spanwise

directions, respectively, and ∆z+ ≈ 2-90 in the vertical direction.

**+ ,-.///

(0 1 + 345 *

*67 ,-.///(0 1899999:99999;

"<+=#?@A<BC=#?DEF= �G3H5I

=B="CJGB@#K=?"#KC"<LG+J

−

**67

MNNNO

&P 3H5 Q̅I

S"=TT-"=+"<BTS#"+

+ 3H53HU34U//////8:;+-"�-�=B++"<BTS#"+

− V **67 ,

-.///(0 1899:99;

LGT@#-T+"<BTS#"+ WXXXY+ 3HU34U////// *-.///*6789:9;

=B="CJ+"<BT?="+#+-"�-�=B@=(ZEFS"#[-@+G#B)

− V \*-.///*67]0

89:9;[GTTGS<+G#B

#?K=<B?�#^=B="CJ

(1)

The flow is assumed to be statistically homogeneous in the streamwise direction, thus periodic

boundary conditions are applied. The free surface boundary is treated as a rigid lid where a

free slip condition is applied. The mean flow is driven by a constant pressure gradient ∆p.

The objective of this study is to determine the dominating terms in the total mean kinetic

energy (MKE) equation and to investigate their effect on the mean flow and on the turbulent

structures in compound channel flow.

The simulation of the turbulent field of the compound channel flow with one-line emergent

vegetation near the floodplain edge is future research case study. The objective is to simulate

the turbulent field of the compound channel flow and to investigate the influence of the

vegetation. The turbulent field will be calculated for several cases with different diameters of

the rods and different distances between them. The numerical results will be compared to

experimental data.



21

CFD modeling of flows over rough floodplain

Moisés BRITO1 & João B. LEAL2

1Faculty of Sciences and Technology, New University of Lisbon, Caparica, 2829-516, Portugal,

[email protected] 2CEHIDRO & Faculty of Sciences and Technology, New University of Lisbon, Caparica, 2829-516, Portugal

During floods, the conveyance of river main channel is exceeded and the lateral floodplains are

inundated by the flow. The simultaneous flow in the main channel and in the floodplains

originates a compound cross-section configuration. Usually, the floodplains are occupied by

constructions, vegetation and obstacles that increase the bottom roughness. In general, that

increase leads to an increase of the flow resistance, altering the mean velocity and flow depth

(Tanino and Nepf, 2008). It also originates a reduction in velocity and shear stress near the

bottom (Yen, 2002), which depends essentially on the relation between absolute roughness

height (ks) and the flow depth (H).

Normally, the floodplain roughness presents high geometrical and spatial heterogeneity that

turns difficult its numerical discretization. A practical way to simulate flows over such type of

bottom is to use a wall function valid for rough bed, avoiding the discretization of the

roughness domain. Other alternative is to model the roughness domain as a porous media. In

this work both approaches are tested using an Explicit Algebraic Reynolds Stress Model to

close RANS equations. This model has showed good results for smooth bed compound-channel

flows (Filonovich et al., 2010; Brito et al., 2012), reproducing the turbulent anisotropy

observed in the experiments.

The numerical results are compared against experimental data collected at LNEC. The

roughness is introduced by an artificial grass carpet placed on the floodplains. Two relative

depths are studied, 0.15 and 0.3. The numerical results using wall function show that the

model simulates the secondary flow, which increases with the relative depth. Nevertheless,

the secondary cells obtained numerically are less intense than the ones observed

experimentally. The model also underestimates the mixing-layer width, leading to higher

lateral velocity gradient and, consequently, to higher Reynolds stresses in the interface region.

The wall function approach gives good results for the higher depth, where ks/H << 1. However,

for the lower depth, corresponding to ks/H ≈ 0.5, the wall function leads to poor results, which

justifies the use of the porous media approach.

For the porous media approach, the flow inside the porous media is, like the flow outside, also

described by RANS equations but with an additional source term, _5, that represents the time-

averaged drag force per volume unit, including the viscous and form (pressure) drag. This

additional term can be estimated by the extended Darcy-Forchheimer model (Lemos, 1960):

_5 = `ab5cE − @daP|bc|bc

√E (1)

The application of this model to the porous media induces a discontinuity of flow properties in

transition between the porous media and the free flow (Lemos, 2006), due to the abrupt

variation of porosity and permeability of the media. To avoid this, it is necessary to specify

appropriate contour conditions at the interface. In this study, continuity conditions for

velocity, intrinsic pressure and shear stress were used. The vertical velocity and Reynolds

stress profile (Figure 1) shows an inflection point at the top of the porous media (z = 5.6 mm).



22

a) b) Figure 1. Profile in the middle of the floodplain for relative depth 0.3:

a) vertical velocity; b) Reynolds stress.

The numerical profile obtained with the porous media approach is closer to the experimental

one, than the profile obtained with the wall function. Although the porous media approach

allows achieving good results, it should be mentioned that involves a higher calibration

complexity for parameters like porosity or permeability.

X

X

X

X

X

X

X

X

X

U [m]

z[m

]

0 0.1 0.2 0.3 0.40

0.01

0.02

0.03experimental

wall function

porous media

X

X

X

X

xz[Pa]

z[m

]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80

0.01

0.02

0.03experimental

wall function

porous media

X

τ



23

Failure risk of flood protection measures due to modified sediment transport under climate change

Mona JAFARNEJAD, A. SCHLEISS & E. BRUHWILER

Laboratory of Hydraulic Construction (LCH) - ENAC – EPFL, CH - 1015 Lausanne, Switzerland,

[email protected]

Evaluation of potential failure in river bank protection measures such as ripraps, and walls is

the main issue of their stability and safety assessment. Moreover, a changed sediment

transport in rivers, as a possible result of climate change, influences the failure risk of flood

protection measures. This bank failure can lead to uncontrolled erosion and flooding with

disastrous consequences in residential areas or damage of infrastructures. Thus, probabilistic

analysis of their failure mechanisms due to flood events and sediment transport is a principal

step to assess embankment stability.

The goal of this research is failure risk assessment of flood protection measures, considering

changed hydraulics and sediment transport due to climate change. This should allow

identifying the most efficient strategies in order to deal with the changed sediment transport

in a sustainable way. The goal can be achieved by studying response and behavior of flood

protection measures in different failure mechanisms. Furthermore, the consequences of

failure on the scale river reach and catchment area should be learnt. Determination of failure

risk as a function of sediment supply and transport for different protection measures, analysis

of risk evolution in selected mountain rivers and extrapolation for similar catchment areas in

Switzerland will be the next steps.

As the first step, the probability of failure in different modes such as direct block erosion, toe

scouring and overtopping in riprap has been defined based on Monte Carlo Simulation and

Moment Analysis method. The changed bedload transport due to a probabilistic function of

the design discharge has been taken into account. Sensitivity analysis is performed by varying

slope, block size, bedload characteristics, geometry of the cross section and hydraulic

parameters. The failure probability of ripraps is assessed by a probabilistic function of the

design safety factor. This simulation method can be implemented in water surface and

bedload transport calculation models. This allows applying the method on other rivers for

computing the probability of failure based on prevailing sediment transport regime.

For further steps, theoretical failure analysis of other flood protection measures will be

investigated and dominant parameters will be defined. The other risk analysis methods will

also be applied. Physical experiments are going to be carried out as the next step of the

research project. The experiments will be arranged to study the stability of large block made

ripraps, considering the thickness for one layer of blocks as well as two layers. We expect the

results bring a new modification of riprap design relation with large blocks.



24

Figure 1. Selected trapezoidal section showing bed and water level variation due to change of

sediment and different failure modes

Table 1. Probability of safety factor SF and failure modes based on Monte Carlo simulation

Figure 2. The probability values of safety factor and failure modes in selected trapezoidal section

Figure 3. Experimental approach by comparing 1 layer and 2 layers of large individually placed riprap



25

Characterization on compound-channel turbulent field without and with vegetation

Ricardo AZEVEDO1, Luis ROJAS-SOLÓRZANO2 & João B. LEAL1

1CEHIDRO & Faculty of Sciences and Technology, New University of Lisbon, Caparica, 2829-516, Portugal,

[email protected]

2Department of Energy Conversion and Transport, Universidad Simon Bolivar, Caracas, Venezuela

During floods, the interaction between the floodplain flow and the main channel flow

generates a complex 3D behavior of the turbulent field, where the 2D flow assumption is not

applicable. Further, the flow in straight compound channel is even more complex when

vegetation exists in the floodplain region. In this work, in order to observe the behavior of the

flow with vegetation, rods were placed at the edge of the floodplain. The results were

compared in four vertical profiles, one at the middle of the main channel, two at the interface

region (lower interface and upper interface), and other one at the floodplain.

The measurements were carried out using a Laser Doppler Velocimeter (LDV) Innova 70C

Argon series. To improve the number of particles detected by the LDV, particles of aluminum

oxide with a diameter less than 10 µm were placed into the fluid.

Table 1 shows the experimental conditions where measurements were carried out.

Table 1.Experimental Conditions.

Test Q Hmc Hr Ucs U* Re Fr

(m3/h) (m) (m/s) (m/s) (x10

4) (-)

Without rods 83.38 0.1033 0.50 0.430 0.023 11.78 0.43

With rods 82.99 0.1034 0.50 0.435 0.023 11.07 0.43

Figure 1 shows a general view of the behavior of the flow, where the mean velocities U and W

are presented. The turbulent structure known in straight smooth compound-channel is

abruptly modified with the placement of the rods, where new turbulent structures are

generated due to the interaction between the flow and the rods. Schanauder & Moggridge

(2009) described the behavior of the flow around impermeable rods. Part of the flow

approaching to the rod is deflected downward until the rod base, generating a horseshoes-

vortex system at both sides of the rod. Further, a recirculation of the flow behind the obstacle

was observed. These new currents explain the bulging at the vicinity of the rod.

To understand the 3D behavior, a comparison of the turbulent scales and of the dissipation

rate obtained experimentally and computed with 2D fully developed flow relations (e.g. Nezu

& Nakagawa) was made at four vertical profiles. In the case of the longitudinal integral scale,

the Taylor's frozen-field hypothesis was adopted to transform the time record into a spatial

record, using as convection velocity the time-average velocity U of each record. The spatial

record allows the computation of the longitudinal autocorrelation function by the equation (1)

(e.g. Pope, 2000).

�(g) = -h(")-h(6i")///////////////////-U(///// (1)



26

To calculate the vertical distribution of the longitudinal integral length scale, longitudinal

microscale and dissipation rate, the follow equations were used, respectively:

j6 = k �(g)lgmn (2)

o6 = pqnrsU(t (3)

u = �&& vw&x^'y/q{⁄ (4)

where, V =kinematic viscosity; }′ =turbulent intensity; �&& =velocity power spectrum;

w& =universal constant equal to ≈ 0.53; and x^ =wave number.

The results show that the longitudinal integral scale is almost constant in the studied verticals,

indicating a strong influence of the wakes generated by the rods. For the turbulent microscale

and dissipation rate cases, they acquire a longitudinal variation. While the microscale increases

near the rod, the dissipation rate decreases. However, in the downstream direction the

microscale decreases and the dissipation rate increases, presenting an inverse behavior

between them.

Figure 1. Isovels of mean velocity U and vectors of mean velocity W: a) without rods, X=9.00 m; b)

with rods, X=8.77 m; c) with rods, X=9.25 m; d) with rods, X=9.73 m.





27

SPH-based numerical simulation of the velocity field in a dam-break flow

Ricardo CANELAS1, Rui ALEIXO2 & Rui M. L. FERREIRA1

1 CEHIDRO, Instituto Superior Técnico, Lisbon, Portugal, [email protected]

2 Department of Civil and Environmental Engineering - Institute of Mechanics, Materials and Civil

Engineering, Université catholique de Louvain, Place du levant 1, 1348 Louvain-la-Neuve, Belgium

Should the collapse of a dam be idealized as an instantaneous removal of a vertical barrier

initially separating two water masses with a free surface that extends indefinitely on both up

and downstream directions, the dam-break problem is, mathematically, a Riemann problem.

Under ideal conditions, namely purely 2DV flow, hydrostatic pressure distribution, horizontal

channel and frictionless bottom, the longitudinal profiles of the free surface and of the velocity

admit self-similar solutions. If these ideal conditions are not met, the description of the flow

should take into account the vertical distribution of pressure and of velocity. In particular, the

quantification of the development of a boundary layer and the proper parameterization of the

friction factor requires the detailed knowledge of the vertical distribution of velocity.

The main objective is to simulate the velocity field originated by a dam-break flow over a flat

smooth boundary. Virtual dam-break tests are carried with a SPH-based model implemented

on C++ CUDA for GPU DualSphysics. The numerical data is validated with laboratory results by

Aleixo et al. (2010), allowing for a discussion of the near-bed flow, vis-a-vis the employed

formulations for viscous stresses.

SPH relies heavily on integral interpolant theory fluid domain is represented by a set of nodal

points where physical properties such as mass, velocity, pressure and vorticity are known.

These points move with the fluid in a Lagrangian manner and their properties change with

time due to the interactions with neighboring particles. The Navier-Stokes system can be

discretized in a weakly compressible form by

( ) ( , ) ii j i j ij

j

dm W h

dt

ρρ= − − ∇∑ v v r

(0.1)

2 2 ( , ) ,

ji ij ij ij

j i j

pd pm W h

dt ρ ρ

= − + + Π ∇ +

∑

vr g

(0.2)

where ijΠ is a viscous term resembling

2µ∇ = ∇τ v . The presented results are produced

using a classical artificial viscosity term, found to be dissipative regarding shear and vorticity.

The used equation of state, together with the simple discretization of the continuity equation

(0.1), provide little control of the volumetric evolution of the particles, proving to be a

potential source of error for detailed simulations, even increasing the accuracy by increasing

the smoothing length and using XSPH to advect the particles.

The simulation is set up with a particle spacing of 0.008 m. Using non-dimensional scalings as

0Z z / h= ,

0X /x h= ,

0/T t g h= and

0/U u c= , with

0 0c gh= , the velocity profiles can

be seen in Figure 1.



28

0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

U

0 200 400 600

E

0.4 0.6 0.8

U

0 200 400

0

0.1

0.2

0.3

0.4

0.5

0.6

E

0.4 0.6 0.8

U

0 200 400 600

E

0.4 0.6 0.8

U

0 200 400

E

Apart from the gradient corresponding to a boundary layer effect the experimental and

numerical profiles show a very good agreement. As expected, the velocity is below the

theoretical, frictionless case thus rendering the flow depth superior. Near the free-surface and

next to the bed the measurements present possible artifacts due to the nature of the seeding

and the PTV algorithms, rendering comparisons difficult. Observing the enstrophy field in

Figure 2 a richer view of the structure of the flow can be discussed.

The Reynolds number for the bed-influenced region is superior to 104, and should therefore

exhibit turbulent flow. The relatively coarse size of the particles, for the relevant turbulent

scales, but essentially the lack of an adequate sub-particle model for the dissipation of

turbulent kinetic energy does not allow for the correct representation of a turbulent field.

However, there is a simple viscous formulation that provides the inflection of the velocity

profiles. Such inflection is not present in all the laboratory measurements, indicating that there

is a possibility that the energy dissipation mechanism that provides a compatible numerical

solution is of a different nature than the one present in the flume, not disregarding

measurement errors and limitations. SPH-LES models like the Sub-Particle-Scale model,

together with wall models to overcome the necessity of a resolved region at the wall, were TKE

dissipation is intense, will hopefully provide a closer related model to the actual phenomena

regarding such mechanisms.

The GPU implementation renders the 2x106 particle simulation over 70x faster than a parallel

CPU implementation, with over 6 time steps per second.

0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

U

Z

h0=0.325 m

0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

U

Z

h0=0.325 m

0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

U

Z

h0=0.325 m

0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

U

Z

h0=0.325 m

Figure 1 - Squares-measured, Dots-numerical, Line-Ritter solution. X=-0.234, X=0.0, X=0.38,X=1.41. T=10.

Figure 2 - Dots-velocity, Line-Enstrophy profiles. X=-0.234, X=0.0, X=0.38,X=1.41. T=10.



29

Limiting concentration of transported fine sediment

Sergi CAPAPÉ & Juan P. MARTÍN VIDE

Technical University of Catalonia, Campus Nord, D1-204, c/ Jordi Girona 1-3, 08034 Barcelona,

[email protected]

Wash load – commonly identified with the fine sediment - is ignored in most of the sediment

transport equations (Yang, 1996). However, it constitutes the greatest contribution to the total

sediment load in some rivers (e.g. Pilcomayo in South America (Orfeo, 2007) or the Yellow in

Asia (Yang, 1996)) and it is found in large quantities in reservoir deposits. Likewise, wash load

has been traditionally characterized by both its independence from hydraulic conditions and

its dependence on fine material supply that originates in the basin, although there is criticism

about this (e.g. Yang (2005) and Khullar (2010)).

Equations used to estimate sediment transport present the results at capacity (i.e. assuming

that available material susceptible of being transported is unlimited). A necessary condition,

but not sufficient, for the sediment transport to be at capacity is that suspended load

transport and bed-load transport are related (Graf, 1999). The opposite implication does not

always hold true since wash load transport, which is not related with the bed material, might

be at capacity.

The conditions for the suspended sediment laden flow to reach its capacity have been

investigated and include stream flow power available to transport sediment (Bagnold, 1966),

turbulence damping (Itakura & Kishi, 1980) and regression analysis using non-dimensional

parameters (Nalluri & Spaviliero, 1998), among others. However, there is not a conclusive

result and the list of equations available to estimate the limiting concentration is short.

Furthermore, the uncertainties increase when the sediment is fine (i.e. silt and clay, with a size

less than 0.063 mm) and some of the expressions explicitly exclude fractions of fine sediment

in its applicability range.

Figure 1. Left: Deposit and bed-forms of fine sediment (particle size of the deposit D50 = 40 µµµµm). Right:

Lateral view of the bed-forms while carefully emptying the flume.

Table 1. Hydraulic conditions at the experiment

Q s h B D50 ρs C

(m3/s) % (m) (m) (µm) (kg/m

3) (kg/m

3)

0.012 0.02 0.09 0.37 25 2650 14.06



30

In a prior experiment (Pineda, 2010), the addition of non-uniform fine sediment (D50 = 0.025

mm, σg = (D84/D16)0.5

= 2.5, where Dx means the particle size which the x% of the material is

smaller and σg is the grain standard deviation) in a clear water flow in a fixed and smooth wall

flume caused the sedimentation of part of the material and the appearance of bed-forms (Fig.

1) (Table 1). Flow conditions fulfilled the requisites of Nordin (1985) to define the sediment as

wash load, so the deposition was not expected to develop:

s�� ≥ 1.25�� ≥ �@ (1)

Where Uf means shear velocity, Vs is the fall particle velocity, θ is the Shields’ dimensionless

shear stress and θc is the critical Shields’ dimensionless shear stress for the incipient motion.

Specifically, the data obtained from the experiment was:

s�� = 19; � = 0.46 ≥ �@ = 0.15 calculated from Cao et. al. (2006) (2)

The measured suspended sediment concentration was 14 g/l - the 56% from the predicted

maximum. This value does not match the results obtained using different equations. The

deposition of fine sediment gradually transforms itself into bed material, thus the applicability

of the equations (only valid for bed material load) is partly justified. Complementary, literature

data from Cellino (1998) and Khullar (2010) of suspension flows at capacity is used for: 1) to

compare the correctness of the predictions from different equations and 2) to seek dominant

variables related to the limiting concentration.

As a result of the foregoing facts, new experiments with fine sediment (D50 = 0.006 mm) to

simulate wash load will be carried out. Some of the posed questions are: Is it possible to define

a criterion to estimate the wash load limiting concentration? Is it possible to foresee the

sedimentation of fine sediment on the stream bed? These two questions are answered

confirming or rectifying some already existing expressions of limiting concentration. Are the

processes of wash load transport and bed material transport the same once the capacity of the

former is surpassed? In other words, is the existing knowledge of bed material load transport

extendable to the wash load transport once the latter is partly deposited on the stream

bottom? How is the fine material transport affected by the roughness of the bed? In other

words, if the stream bed is coarser than the fines, does the limiting concentration change?

It is remarkable that the combination between the study of limiting concentration of

transported sediment (capacity) and the use of fine material is not the most frequent. Sand

size is the most common one. To single out the smaller sizes will contribute to a better

understanding of the wash load traditionally avoided.



31

Anysotropy in vegetated natural flow

Silvia SAGGIORI1, Rui M. L. FERREIRA2, & Mário J. FRANCA1

1Faculty of Sciences and Technology & IMAR, New University of Lisbon, 2829-516 Caparica, Portugal,

[email protected] 2CEHIDRO - Instituto Superior Técnico - Techical University of Lisbon, Portugal

The presence of vegetation and its distribution along the water depth induce extra complexity

on the turbulent flow structure due to such local effects as vegetation stems and vegetation

leaves interacting with the flowing water. Defining Reynolds stresses as -u�u�/////, the relation

between velocity fluctuations on the three different Cartesian directions can be examined in

terms of anisotropy, allowing the classification of different turbulence states. The invariant

technique is the basis of this classification.

Lumley and Newman (1976) based their work on the analysis of the state of turbulence as

specified by II and III invariants (Figure 1) defining a dominium within which all the

homogeneous turbulence must be found (Chassaing, 2000).

The objective of this work has been the characterization of turbulence states in a natural

vegetated open-channel (Figure 1). To accomplish the proposed objective, Acoustic Doppler

Velocimetry (ADV-Vectrino) measurements were performed along 10 vertical profiles on a

natural river populated by submerged and flexible vegetation. The sampling frequency was set

to 100 Hz and the further characteristics concerning the field measurements are shown in

Table 1.

Figure 2 is an example of turbulence states inside the vegetation (profile Pr8). Generally close

to the bottom and at vegetation stems, 2D and 1D isotropy were found whilst along the

vegetation leaves layer, mainly 2D-isotropy and pancake-shaped turbulence states were

observed. Free flow layer was characterized namely by cigar-shaped turbulence, expected

given the increasing of streamwise velocity above the vegetated layer. Measurements

conducted on the dune were characterized by 2D-isotropy and cigar-shaped state.

This kind of analysis, conducted in parallel with further physical applications, may have

practical implications on the evaluation of depuration capacity of vegetated channels, and

more specifically, for the further design of wastewater treatment facilities using, for instance,

fitodepuration techniques, identifying the behavior of substances concentration through

vegetated patches.

Table 2: Measurements main characteristics (FS: vegetation reaching the free surface)

nr_profile 1 2 3 4 5 6 7 8 9 10

acq_time 5 5 5 10 10 10 10 10 10 10

nr_points 12 12 12 12 12 19 16 16 16 16

location dune dune dune boundary boundary vegetation vegetation vegetation vegetation vegetation

veg_high - - - 80% 70% 90% FS 65% FS FS



32

Figure 3: (left hand side) field picture highlighting velocity measured profiles; (right hand side)

anisotropy invariant map.

Figure 4: example of states of turbulence plot at profile Pr8





33

Preliminary analysis of a gravity current moving on an upsloping bed: velocity measurements and numerical simulation

Valentina LOMBARDI, Claudia ADDUCE, Giampiero SCIORTINO & Michele LA ROCCA

Department of Civil Engineering, University of Rome “Roma Tre”, Via Vito Volterra 62, 00146, Rome,

Italy, [email protected]

The aim of this work is the investigation of the dynamics of a gravity current moving on an

upsloping bed by measurements of the instantaneous velocity field and a numerical

simulation. The lock exchange release experiment was performed in a Perspex tank of

rectangular cross-section divided into two parts by a vertical gate. The right portion was filled

with a fluid of density ρ2 (i.e. ambient fluid), while the left part was filled with a heavier fluid

(i.e. lock fluid) with higher initial density ρ01. Both in the right and in the left part of the tank

the depth of the fluid was h0. A quantity of dye was dissolved into the lock fluid in order to

allow the visualization of the gravity current during the experiment. The experiment begins

when the sliding gate is suddenly removed and the heavier fluid collapses flowing under the

lighter one generating the gravity current and it stops when the current’s front reaches the

right end wall of the channel.

Instantaneous velocity measurements are obtained by Particle Image Velocimetry. Both the

dense and the ambient fluid were seeded with polyamide particles and a PIV system with a

double pulsed Nd:YAG laser is used. In order to avoid problems associated with the variations

in refractive index with the local value of density the RIM (Refractive Index Matching) strategy

is applied. A solution of glycerol and water as less dense fluid an aqueous solution of

potassium dihydrogen phosphate (KH2PO4) as the heavier one are used. Velocity and vorticity

fields show the main features of a typical gravity current. Interfacial instabilities between the

two layers at the rear of the head of the current can be recognized. A gradual decrease of the

horizontal velocity can be observed from the head to the tail of the current. Vorticity field

shows a strip of negative vorticity near the bottom due to the no slip condition and an area of

positive vorticity at the interface due to the velocity shear between the two layers. From

contour plots of the horizontal velocity and vector maps obtained for the experiment

performed on an upsloping bed a backflow, that is an area in which the flow direction is

reversed to the current’s direction (i.e. the dense fluid is moving downslope) can be observed

near the lock of the tank.

Numerical simulations were performed using a two-layer, 1D, shallow-water model developed

by Adduce et al. 2012. The model takes into account the free-surface and the mixing occurring

at the interface between the two fluids. The time histories of the front’s position show a good

agreement between experimental data and numerical simulations. Moreover both velocity

measurements and numerical results show a backflow near the lock of the tank while the tail is

visible in the domain of analysis.



34

BOOK OF ABSTRACTSruif/fluvial_habitats_project/workshops/... · João Fernandes Uniform and...

Documents

Transcript of BOOK OF ABSTRACTSruif/fluvial_habitats_project/workshops/... · João Fernandes Uniform and...