GEOPHYSICAL RESEARCH LETTERS, VOL. ???, XXXX, DOI:10.1029/,

Revisiting Hele-Shaw Dynamics to Better Understand Beach

Evolution

O. Bokhove1,2, A.J. van der Horn3, D. van der Meer3, E. Gagarina2, W. Zweers5,

and A.R. Thornton2,4

Wave action, particularly during storms, drives the evo-lution of beaches. Beach evolution by non-linear break-ing waves is poorly understood due to its three-dimensionalcharacter, the range of scales involved, and our limited un-derstanding of particle-wave interactions. We show how anovel, three-phase extension to the classic “Hele-Shaw” lab-oratory experiment can be designed that creates beach mor-phologies with breaking waves in a quasi-two-dimensionalsetting. Our thin Hele-Shaw cell simplifies the inherentcomplexity of three-phase dynamics: all dynamics becomeclearly visible and measurable. We show that beaches canbe created in tens of minutes by several types of breakingwaves, with about one-second periods. Quasi-steady beachmorphologies emerge as function of initial water depth, at-rest bed level and wave-maker frequency. These are classi-fied mathematically and lead to beaches, berms and sandbars.

1School of Mathematics, University of Leeds, Leeds, U.K.2Department of Applied Mathematics, University of

Twente, Enschede, The Netherlands3Department of Physics, University of Twente, Enschede,

The Netherlands4Department of Mechanical Engineering, University of

Twente, Enschede, The Netherlands5FabLab, Saxion College, Enschede, The Netherlands

Copyright 2013 by the American Geophysical Union.0094-8276/13/$5.00

Figure 1. Schematic of the Hele-Shaw cell. Wedge,waterline, particle bottom, and wave-maker are shown.Parameters are initial water depth W0 = H0 − B0 =(10, 30, 50, 70) ± 3mm, bed level B0 = (20, 50, 80) ±

2mm at rest, and wave-maker frequency fwm =(0.4, 0.7, 1.0, 1.3)Hz with angle θwm.

1. Introduction

Surf-zone dynamics concerns the formation and erosionof beaches by breaking waves. Beach evolution is strongestduring storms, which leads to intermittent coastline changesRoelvink et al. [2009]. Prediction of beach dynamics requiresan understanding of breaking waves, especially how wavespick up, transport and deposit sand particles. The combinedinteractions of water, air and sand are involved, making upa three-phase flow, with several spatial and temporal scales.These include wave action on shorter time scales, which is re-sponsible for the net/overall beach evolution on longer timescales. But beach evolution by non-linear breaking wavesis poorly understood due to its three-dimensional character,the range of scales involved, and our limited understand-ing of particle-wave interactions. In principle, Navier-Stokesand continuum equations are available as models for waterand air motion, and for interactions with and between sandgrains. However, their direct numerical simulation remainsliterally incalculable because the degrees of freedom involvedare simply too large. Models or closures of the unresolvedprocesses such as particle-particle and particle-wave inter-actions are required. Despite great advances in coastal en-gineering [Soulsby , 1997; Calantoni et al., 2006; Garnier etal., 2010; McCall et al., 2010], our understanding of clo-sures of the averaged sand transport and wave breaking re-mains relatively weak. The question we posed was “Howcan we explore the fundamentals of beach evolution underbreaking waves in a simplified laboratory and modelling en-vironment?” Our answer was to design a novel extensionof Hele-Shaw’s classic 19th-century experiment with viscousfluid or Stokes’ flow around obstacles in thin cells [Lamb,1993], as follows.

Imagine an isolated slice of beach that includes air, sandparticles and water waves, and place it between two glassplates. The ’beach-slice’ is just over one-particle diame-ter wide, with all particle and water motion clearly visibleand tractable. Based on analytical and numerical designcalculations, we built a laboratory tank [Bokhove et al.,2010] to demonstrate beach evolution by breaking waves,see Figure 1. It is 1m long, 0.3m high, and 2mm wide,with Gamma Alumina particles 1.75 ± 0.1mm in diameter(density of 1750 ± 30kg/m3), separated from a wave-makerby a submerged wedge to avoid direct impact. We haveused nearly spherical and uniform Gamma Alumina par-ticles. Two joined, metallic rods each with a diameter of1.6mm act as wave-maker, driven by a PC-steered linear ac-tuator with sinusoidal output. We operate the wave-makerfor monochromatic waves in the frequency range 0.4−1.3Hz,with a pendulum angle θwm. It is offset to produce a slightlyanti-symmetric wave, inducing less suction on the beachside. Fresh MilliQ-water was employed to ensure clean op-eration and limit variability in surface tension.

This Hele-Shaw beach configuration allows precise track-ing of the few thousands particles involved, in both timeand in approximately two spatial dimensions, because onlyone layer of particles fits the tank laterally. Similarly, usinghigh-speed cameras the air-water interface is easily tractable

1

X - 2 BOKHOVE ET AL.: BEACH CREATION BY BREAKING WAVES

as a water line. By design, the dynamics in the Hele-Shaw cell is quasi-two-dimensional [Rosenhead , 1963; Wilsonand Duffy , 1998]. Classically, Hele-Shaw hydrodynamics isdominated by side-wall boundary layers leading to a nearlyparabolic flow profile [Batchelor , 1967]. The Navier-Stokesequations can be width-averaged, while ignoring Reynolds-stress and the small in-plane viscous terms. This leads tolinear damping of momentum in the vertical plane, inverselyproportional to the square of the gap width between theglass plates. The crucial insight is that for a gap width2l > 1.5mm a damped breaking wave propagates to the endof a fixed beach of about 0.5m, which we determined numer-ically. Otherwise, the wave dissipates too quickly, and thenthe desired phenomenology of wave breaking and particletransport is too weak or absent.

In Section 2, we confirm that the observed decay of wavemodes in the Hele-Shaw laboratory tank, filled with waterbut without particles, is captured reasonably well by nu-merical simulations using a potential flow model [Lighthill ,1978] with only this linear damping. This comparison sug-gests that other flow profiles or damping of the contact line[Vella, 2005] at the water-air-glass interface may be of sec-ondary importance. Lee et al. [Lee et al., 2007] consid-ered one particle settling in a Hele-Shaw cell, and observedquasi-two-dimensional behaviour for gap widths 2l < 1.05d,with particle diameter d, and three-dimensional behaviourfor 2l > 1.1d. We hence chose a gap width of 2mm, giventhat our particles are 1.75 ± 0.1mm in diameter. While thestronger damping mechanisms due to glass walls and con-tact lines are absent at natural beaches, these simplify thedynamics in the Hele-Shaw cell, and may be readily includedin mathematical models. Anyhow, the Hele-Shaw beachcaptures the complex problem of wave breaking within aquasi-two-dimensional realm, which lends itself very well forfundamental research on wave-particle interactions. Bothrealistic wave breaking and beach dynamics are observed,as we show in Sections 3 and 4. Finally, we conclude anddraw inferences for future research in Section 5.

2. Validity of Linear Momentum Damping

The design of the Hele-Shaw beach experiment was basedon analysis of approximate models for the hydrodynamics.A narrow gap width between the two glass plates, just overone particle, is desirable to enhance the visualization andlimit the dynamics to be nearly two-dimensional. At thesame time, this gap still needs to be wide enough to allownonlinear breaking waves and beach formation. In the clas-sic Hele-Shaw case, the parabolic flow profile in the verticalplane is

(u, w) =3

2(u, w)(l2 − y2)/l2 (1)

with u = u(x, y, z, t) and w = w(x, y, z, t) the velocity com-ponents in the horizontal, x, and vertical, z, directions, lat-eral direction y, gap width 2l and the laterally averagedmean velocities u = u(x, z, t) and w = w(x, z, t), and timet. Under the approximation that the local flow profile isparabolic in the lateral direction, we can substitute (1) intothe three-dimensional Navier-Stokes equations for the hy-drodynamics of water and average these across the gap.Given the anisotropy in the length and velocity scales alongand normal to the vertical plane, the viscous balance acrossthe gap is assumed to be dominant. An approximation isthen made in which slow variations in the vertical plane onscales larger than the gap width are still permitted. Afteraveraging and neglecting Reynolds-stress terms, the result-ing two-dimensional Euler equations with linear momentum

damping become

∂u

∂t+ βu

∂u

∂x+ βw

∂u

∂z= −

1

ρ

∂p

∂x−

3νu

l2, (2)

∂w

∂t+ βu

∂w

∂x+ βw

∂w

∂z= −

1

ρ

∂p

∂z−

3νw

l2, (3)

∂u

∂x+

∂w

∂z= 0, (4)

in which ν is the viscosity of water, g the acceleration ofgravity, p the pressure, ρ the constant density of water, andβ = 6/5 a scaling factor due to the width-averaging. Thefirst two equations are the momentum equations in the x–and z–directions, and the last equation expresses that thiswidth-averaged velocity in the vertical plane is incompress-ible. Partial derivatives are denoted as usual. Higher-ordercontributions to the lateral flow profile (1) are ignored in (4),which especially concerns omissions in the three-dimensionalboundary layers near the free surface, side walls and nearparticles. Nevertheless, for the flow in the bulk away from

Figure 2. Initial conditions. Initial conditions areshown for a) the simulations that start at rest flow with atilted free surface, and b) for the experiments that startat rest with a tilted tank.

Figure 3. Potential energies. Potential energies a) P (t),and b) P (t)exp(3νt/l2), versus time t. Profiles for differ-ent initial angles α = 4.0o, 4.2o5.4o, 6.1o, 6.5o are shownfor the model simulations (thin lines) and the laboratoryexperiments (thick, jagged lines).

BOKHOVE ET AL.: BEACH CREATION BY BREAKING WAVES X - 3

boundaries, these linear damping terms in the momentumequations of (4) are reasonable, as we will see.

Furthermore, when we assume that the velocity has theform (u, w) = (∂φ/∂x, ∂φ/∂z) with velocity potential φ,then the potential flow equations emerge with a linear damp-ing term. To simplify the calculations, we set β = 1 here-after, but this does not change the argument below regard-ing the potential energy. The resulting system of equationscan in principle be derived from the following variationalprinciple

0 = δ

Z T

0

„

Z L

0

φs∂h

∂tdx − (P (t) + K(t))

«

e3νt/l2dt

≡ δ

Z T

0

„

Z L

0

φs∂h

∂t−

1

2g(h − H)2dx −

Z h

0

1

2|∇φ|2dxdz

«

e3νt/l2dt (5)

with water depth h = h(x, t), the potential at the free sur-face φs, mean still water depth H , suitable boundary andend-point conditions for a tank of length L and with a fi-nal time T . Principle (5) is an extension of Miles’s clas-sic variational principle [Miles, 1977], with the exceptionof the additional factor exp 3νt/l2. That factor is the in-verse of the integrating factor due to the linear momentumdamping. The consequence of formulation (5) is that it isbetter to plot the potential energy, P (t), times this factorexp 3νt/l2. When, therefore, we start with the release ofwater from a rest state with available potential energy, thiscorrected potential energy, P (t) exp 3νt/l2, is expected tooscillate around a constant mean value.

A straightforward comparison between the potential flowmodel arising from (5) and an experiment can now be used toassess the validity of linear momentum damping. In the nu-merical discontinuous Galerkin finite element approximationto (5) (cf. Gagarina et al. [2012]), we start at rest with a lin-early slanted free surface with angle α, see Fig. 2a. Instead,in the experiment we start with a tank at rest lifted upwardwith an angle α, see Fig. 2b, and lower it down quickly toa horizontal level such as to obtain, approximately, the stillprofile in Fig. 2.a. The small phase difference is removed byshifting the time profiles of P (t) and P (t)exp(3νt/l2) suchthat the first zero crossings are lined up. For five values ofα, these results are shown in Fig. 3. For the larger anglesα = 6.1o and α = 6.5o, our experimental approximationat the start is less satisfactory, but for the other angles thematch between the simulations (thin lines) and experiments(thick noisy lines) is good for about one second. Within twoseconds, most energy has dissipated, but after one second,the comparison with the model is less good, see Fig. 3b.This discrepancy is presumably due to the neglected three-dimensionality of the profile near the free-surface boundarylayer in combination with effects of surface tension. Thelatter effects have been analysed in the context of a relatedyet different, two-dimensional Faraday experiment by Vega[2001]. We conclude that for driven flows on scales muchlarger than the gap width, linear momentum damping isa good leading-order model approximation, but that mod-elling the fine-structure in breaking waves may require res-olution of three-dimensional free-surface boundary layers.

3. Wave breaking

We found all types of wave breaking [Peregrine, 1983] atnatural beaches, including spilling, plunging, collapsing andsurging breakers. To trace waves, we dyed the water red

to enhance the contrast and facilitate analysis of the mea-surements, obtained by a high speed Photron SA2 cameraat 1000fps. Space-time renderings of the four wave typesare shown in Fig. 4, and discussed in turn. In a plung-ing breaker, the wave front overturns and a prominent jetfalls at the base of the wave causing a large jet, Fig. 4a.We observe two bubbles caught during overturning. For asurging breaker, a significant disturbance and vertical facein an otherwise smooth profile occurs only near the mov-ing shoreline, Fig. 4b. In a spilling breaker, whitish waterat the wave crest spills down the front face sometimes withthe projection of a small jet, Fig. 4c. The grooves indicatethat the presence of (pre-existing) bubbles. In a collapsingbreaker, the lower portion of the wave’s front face overturnsand then behaves like a plunging breaker, Fig. 4d, where thelower portion of the wave is seen to shoot forward after circa

Figure 4. Space-time plots of breaking waves. The wa-ter surface (blue) and bed (red) of measured a) plunging,b) surging, c) spilling and d) collapsing breakers.

Figure 5. Phase diagram of beach morphologies. a)States in the parameter plane of initial water depthW0 = H0 − B0 and wave frequency fwm, for an initialbed level of B0 = 8cm. (For B0 = 2 and 5cm the stateis quasi-static, except for fwm = 1Hz, B0 = 5cm andW0 = 5cm, for which a sand bar developed.) b) Initialwater levels and bed heights (blue and red lines, respec-tively) and final bed height for a: b1) wet beach, b2) drybeach, b3) sand bar, and b4) dune.

X - 4 BOKHOVE ET AL.: BEACH CREATION BY BREAKING WAVES

100ms, and to separate from the top part of the wave. Inthe Hele-Shaw set-up these wave types, see the supplemen-tary videos Supplementary Videos [2013]. are small-scaleversions smoothed by surface tension, when compared to vi-olent three-dimensional breakers at beaches. These typesof wave breaking emerge during various stages of the bedevolution into sandbars, beaches or dunes, observable in theHele-Shaw cell.

4. Quasi-steady beach morphologies

Time-dependent beach morphologies form on a time scaleof minutes to hours, forced by wave action that occurs ona typical time scale of one second. We undertook a pa-rameter study [Horn, 2012] of beach formation by break-ing waves in which we varied the initial water depth W0 =(10, 30, 50, 70)mm and bed level B0 = (20, 50, 80)mm at rest,before turning on the wave-maker in a monochromatic fre-quency range of (0.4, 0.7, 1.0, 1.3)Hz. The state of the bedwas captured every 10s with a Nikon D5100 camera. De-pending on the parameters, several quasi-steady beach mor-phologies were observed to develop after 10 to 30 minutes.They are classified mathematically. When the bed profilehas an interior maximum, its maximum is either submerged(“sandbar”), or dry on top or its onshore side (“berm-duneor beach-dune”). When the bed profile has a boundary max-imum, it is either wet or dry (‘wet or dry beach’). A duneemerges when water is found at both sides of a berm, see themovie on berm formation ([Movie5berm, 2012]). A beach-dune is a dune near the right wall, also with an interior max-imum bed height away from the boundary. A submerged

Figure 6. Bed heights in space-time. Time-space dia-grams of bed height versus spatial coordinate x for a) dryand b) wet beaches (boundary maxima), and c) a sub-merged sand bar and d) a dune (interior maxima). Whendry land emerges the initial water level (black line) isindicated.

sand bar arises when a significant amount of sediment areatransport is taking place, more than 10cm2, with an interior,wet maximum of the bed height, and none of the above def-initions hold. Finally, in the quasi-static state hardly anysediment area transport is taking place (less than 10cm2

and no parts fall dry). Four of these states are displayed inFig. 5, displaying both initial and final states, together witha phase diagram for the case B0 = 80mm. Space-time plotsof the bed height evolution from a nearly flat submergedbed to the dune, dry beach, wet beach or sandbar are givenin Fig. 6, and range from 35 to 60 minutes. It is clear inthe phase diagram in Fig. 5 that the states evolve smoothlyfrom one to the other. Furthermore, wave activity is lessefficacious when: the wave has already lost its energy dueto its breaking over the wedge; the water is very shallow;the frequency is too high such that the viscous damping be-comes too high; or, when the water is too deep such thatparticles do not get picked up. The wet beach state emergeswhen there are insufficient particles available to create a drybeach.

5. Conclusions

All Hele-Shaw beach dynamics presented are robust andreproducible. Details depend on the precise wave-makermotion, its placement and asymmetry, and compaction ofthe initial bed. Bed compaction in the quasi-static state isabout 1%, and about 2% for displaced particles. These smallchanges are, however, known and can be included preciselyas initial conditions and forcing into mathematical predic-tions.

The innovation of our Hele-Shaw beach experiment isthat particle, wave and water motions can be measuredand analysed thoroughly. This will trigger the developmentand precision validation of new forecasting models [McCallet al., 2010; Vega, 2001; Thornton et al., 2006; Cotter andBokhove, 2010; Dumbser , 2011; George and Iverson, 2011].The measurements will form a demanding benchmark forexisting coastal engineering forecast models of sand trans-port ([Operational wave and sediment forecast models, 2013]:Delft3D, Telemac and XBeach). Finally, the scope of ourHele-Shaw research is broad: researching its dynamics fordifferent initial bed slopes, particles of varying size and den-sity, wider gap widths, and more complex wave-maker mo-tion will facilitate breakthroughs in mathematical modellingof near-shore three-phase flows.

Acknowledgments. We thank the Physics of Fluids andMulti-Scale Mechanics groups at the University of Twente andthe Stichting Free Flow Foundation for financial support. Wegratefully acknowledge the careful proofreading by Valerie Zwartand Professor Menno Prins.

References

Batchelor, G.K. 1967: An Introduction to Fluid Dynamics. Cam-bridge University Press. 635 pp.

Bokhove, O., Zwart, V., and Haveman, M.J., 2010: FluidFascinations. Publication of Stichting Qua Art Qua Sci-ence, University of Twente, Enschede, The Netherlands. TheHele-Shaw experiment was first revealed in a Qua Art QuaScience lecture of O.B. and Valerie Zwart with W.Z. onJanuary 17th 2010, in a tribute to the late Howell Pere-grine. http://eprints.eemcs.utwente.nl/17393/ Phil. Trans.Roy. Soc. Lond. A 362,1987–2001.

Calantoni, J. Puleo, J.A., and Holland, K.T. 2006: Simulation ofsediment motions using a discrete particle model in the innersurf and swash-zones. Cont. Shelf Res. 26, 1987–2001.

BOKHOVE ET AL.: BEACH CREATION BY BREAKING WAVES X - 5

Cotter, C. and Bokhove, O. 2010: Water wave model with accu-rate dispersion and vertical vorticity. Peregrine Commemora-tive Issue J. Eng. Maths. 67, 33–54.

Cooker, M. 2010: A commemoration of Howell Peregrine 2007–2010. J. Eng. Math. 67, 1–9.

Dumbser, M. 2011: A simple two-phase method for the simula-tion of complex free surface flows. Comp. Methods Applied.Mech. Eng. 200, 9–12.

Garnier, R., Dodd, N., Falquez, A., and Calvete, D. 2010: Mech-anisms controlling crescentic bar amplitude. J. Geophys. Res.115, F02007.

Gagarina, E., Van der Vegt, J.J.W., Ambati, V.R. and BokhoveO. 2012: A Hamiltonian Boussinesq model with horizontallysheared currents. Third Int. Symp. on Shallow Flows Proc.June 4-6, Iowa, http://eprints.eemcs.utwente.nl/21540

George, D.L. and Iverson, R.M. A two-phase debris-flow modelthat includes coupled evolution of volume fraction, granulardilatancy, and pore-fluid pressure. Italian J. Eng. Geology andEnvironment.

Horn, van der, A.J. 2012: Beach Evolution and Wave Dynamics ina Hele-Shaw Geometry. M.Sc. Thesis, Department of Physics,University of Twente.

Lamb, H. 1993: Hydrodynamics. Cambridge University Press.Lee, A.T., Ramos, E., and Swinney, H.L. 2007: Sedimenting

sphere in a variable-gap Hele-Shaw cell. J. Fluid Mech. 586,449–464.

Lighthill, J. 1978: Waves in Fluids. Cambridge University Press,504 pp.

McCall, R.T., van Thiel de Vries, J.S.M., Plant, N.G., van Don-geren, A.R., Roelvink, J.A., Thompson, D.M., and Reniers,A.J.H.M. 2010: Two-dimensional time dependent hurricaneoverwash and erosion modeling at Rosa Island. Coastal. Eng.57, 668–683.

Miles, J. 1977: On Hamilton’s principle for surface waves. J.Fluid. Mech. 27, 395–397.

Movie5berm 2012: The formation of a berm is shown by putting10s snapshots into a film sequence. Web link at GRL www.

Operational forecasting models 2013: Delft3D. Software plat-form including morphology of Deltares, The Netherlands:http://www.deltaressystems.com/hydro/product/621497/delft3d-suite (2012).

Open Telemac-Mascaret: http://www.opentelemac.org/(2012).XBeach: http://oss.deltares.nl/web/xbeach/

Peregrine, D.H. 1983: Breaking waves on beaches. Ann. Rev.Fluid Mech. 15, 149–178.

Roelvink, D., Reniers, A., van Dongeren, A., van Thiel de Vries,J., McCall, R., and Lescinkski, J. 2009: Modelling storm im-pacts on beaches, dunes and barrier islands. Coastal Eng. 56,1133–1152.

Rosenhead, L. (ed.) 1963: Boundary layer theory. UniversityPress Oxford. 688 pp.

Soulsby, R. 1997: Dynamics of Marine Sands. In: HR Walling-ford. Thomas Telford. 249 pp.

Supplementary videos: Four videos of breaking waves in space-time. Three-dimensional rotating figures of the free surface(in blue on the vertical axis) are shown as function ofspace (15 to 35cm) and time (hundreds of ms) of plung-ing, spilling, collapsing, and surging breakers (any pink sur-face denotes the particle bottom). Axes are not labelled be-cause they obscure the movies. Supplementary video 1 (10Mb):Plunging breaker (bokhoveplungingbreakerV1.mov) Supple-mentary video 2 (10Mb): Surging breaker (bokhovesurging-breakerV2.mov) Supplementary video 3 (9.9Mb): Collaps-ing breaker (bokhovecollapsingbreakerV3.mov) Supplemen-tary video 4 (10Mb): Spilling breaker (bokhovespillingbreak-erV4.mov) Tentative weblinks at GRL www.

Thornton, A.R., Gray, J.M.N.T, and Hogg, A.J. 2006: Three-Phase Model of Segregation in Granular Avalanche Flows. J.Fluid Mech. 550, 1–25.

Vega, J.M., Knobloch E., and Martel, C. 2001: Nearly inviscidFaraday waves in annular containers of moderately large as-pect ratio. Physica D 154, 313–336.

Vella, D. and Mahadevan, L. 2005: The ‘Cheerios Effect’. Amer-ican J. Physics 73, 817–825.

Wilson, S.K. and Duffy, B.R. 1998: On lubrication with compa-rable viscous and inertia forces. Q.J. Mech. Appl. Math. 51,105–124.

O. Bokhove, School of Mathematics, University of Leeds, LS29JT, Leeds, U.K. (O.Bokhove@leeds.ac.uk)