INTRA-SWASH HYDRODYNAMICS AND SEDIMENT FLUX FOR DAMBREAK SWASH ON COARSE- GRAINED BEACHES. Technology, Hong Kong

Transcription

1 P P P INTRA-SWASH HYDRODYNAMICS AND SEDIMENT FLUX FOR DAMBREAK SWASH ON COARSE- GRAINED BEACHES 1* T. O DonoghueP P, G. A. KikkertP P, D. PokrajacP P, N. DoddP P and R. BrigantiP P *PUCorresponding author 1 PSchool of Engineering, University of Aerdeen, King s College, Aerdeen, United Kingdom, AB24 3UE (37Tt.odonoghue@adn.ac.uk37T; d.pokrajac@adn.ac.uk) 2 PDepartment of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Hong Kong (kikkert@ust.hk) 3 PDepartment of Civil Engineering, University of Nottingham, University Park, Nottingham, United Kingdom, NG7 2RD (37TNicholas.Dodd@nottingham.ac.uk37T; Riccardo.Briganti@nottingham.ac.uk) ABSTRACT The paper reports on damreak-type swash experiments in which intra-swash hydrodynamics and sediment flux are measured for swash on a coarse sand each and a gravel each. Flow velocity and depth are measured using PIV and LIF respectively; the intra-swash sediment flux is measured using sediment traps. Comparison of measured hydrodynamics with the immoile, permeale ed experiments of Kikkert et al. (2013) indicate that ed moility impacts on the swash hydrodynamics, reducing the maximum run-up y approximately 8% for oth eaches, compared to the maximum run-up on the corresponding immoile each. The measured intra swash sediment flux at a given location is characterised y high flux at the moment of ore arrival, followed y rapid decay during uprush, ecoming zero at some time efore flow reversal. For the gravel each, the ackwash sediment flux is negligily small, while for the sand each the ackwash flux increases slowly as the flow accelerates down the each, and peaks at aout the time of maximum ackwash velocity. Intraswash sediment flux calculated using the Meyer-Peter and Müller ed load transport formula, with measured hydrodynamics as input and ed shear stress estimated using oth the Swart and Colerook formulae, are within a factor 2 of the measured intra-swash flux. The agreement etween the calculated and measured flux is etter for the sand each than for the gravel each, and etter for uprush than for ackwash. For the sand each there is good agreement etween calculated and measured total uprush and total ackwash sediment volumes. The agreement is less good for the gravel each, for which calculated and measured uprush volumes show a similar trend ut the calculated ackwash volumes over-estimate the (negligile) volumes oserved in the experiments. 1

2 Keywords swash, hydrodynamics, sediment transport, shear stress, laoratory, dam-reak 1.0 INTRODUCTION Swash on steep, coarse-grained eaches is generated y the collapse of wave ores on the each slope, resulting in high flow velocities with potential for sustantial sediment flux and morphological change. Hydrodynamics and sediment dynamics within a swash event are complex ecause of the highly turulent and aerated nature of the collapsing ore and the high unsteadiness and nonuniformity of the flow across the swash zone. In field conditions the complexity is augmented y interactions etween swash events of varying magnitude and duration, caused y the varying period and amplitude of the incident waves, and y the effects of low-frequency water surface oscillations in the surf and swash zones. Field experiments investigating swash hydrodynamics and sediment dynamics have developed sustantially over the last two decades in terms of the sophistication of the instruments deployed and the degree to which detailed processes are captured y the measurements. Regarding sediment flux, sediment trapping has een used to measure total uprush and total ackwash transport volumes (Hughes et al., 1997; Masselink and Hughes, 1998; Austin & Masselink, 2006; Masselink et al., 2009) and high-resolution ed elevation measurements across the swash zone have een used to otain net sediment transport volumes for individual swash events (e.g. Blenkinsopp et al., 2011). These measurements are extremely valuale in terms of quantifying sediment fluxes and morphological change for swash events in the field. However, they do not provide a complete picture since they reveal little (in the case of traps) or nothing (in the case of ed elevation measurements) of the sediment flux during a swash event. In principle, measurements of intra-swash sediment flux can e otained from co-located measurements of velocities and concentrations, ut otaining these measurements sufficiently accurately over the complete water column to give total sediment flux through the full swash cycle remains a significant practical challenge (Blenkinsopp et al., 2011). In the laoratory, small-scale wave flumes have een used to study swash hydrodynamics over immoile, impermeale eaches (e.g. Petti and Longo, 2001; Cowen et al., 2003; Gedik et al., 2005; Shin and Cox, 2006; Sou et al., 2010; Rivillas-Ospina et al., 2012). Large-scale wave flume experiments have studied sand suspension processes in the swash zone (Alsina and Caceres, 2011; Caceres and Alsina, 2012), interactions etween surf and swash ed dynamics (Alsina et al., 2012), the effects of 2

3 long waves, wave groups and random waves on surf and swash ed dynamics (Baldock et al., 2011) and each groundwater effects on swash sediment transport (Masselink and Turner, 2012). These have yielded insights into swash zone sediment processes and morphology, ut, as for field experiments, estimates of swash sediment transport are either inferred from ed elevation measurements or are limited to suspended sediment flux ased on co-located velocity and suspended sediment concentration measurements. More recently, van der Zanden et al. (2015) and Puleo et al. (in press) otained measurements of intra-swash sediment concentrations and velocities within the sheet-flow layer of swash in large-scale wave flume experiments. The studies used conductivity-ased instrumentation for the concentration measurements; for the sheet-flow velocities, Puleo et al. (in press) used an acoustic velocity profiler while van der Zanden et al. (2015) cross-correlated concentration measurements from a pair of concentration proes horizontally separated y 15 mm. Neither method was successful in fully resolving the velocities through the sheet-flow layer and through the whole swash cycle. Nevertheless, Puleo et al. (in press) comines their sheet-flow results with concentration and velocity measurements aove the sheet-flow layer to estimate the relative contriutions of suspended load and sheet-flow load to the total transport. The complexity of processes at work in the field and in large-scale wave flume experiments makes it difficult to isolate and quantify fundamental processes and to provide measurements of the kind needed for the development of swash numerical models. An alternative to wave flumes for laoratory swash experiments is to generate swash via a damreak, wherey a reservoir of water is suddenly released in a flume, leading to a ore that collapses on a each located downstream. The damreak produces a single, highly repeatale, large-scale swash event, with ore depth, ore speed and maximum run-up comparale to that seen in the field under energetic wave conditions. The set-up avoids many of the complexities associated with swash in the field, and indeed with wave-generated swash in laoratory wave flumes, such as the variaility in swash events, swash-swash interactions and the effects of low-frequency oscillations. This reduction in complexity, comined with the aility to repeat the same swash event many times, allows particular fundamental swash processes to e isolated and studied in detail. Moreover, damreak swash experiments provide good enchmark data for numerical models since the oundary and initial conditions are well defined and data are availale with high resolution in time and space. Barnes et al. (2009) used a damreak set-up to directly measure intra-swash ed shear stress using a shear plate; a similar set-up was used y O Donoghue et al. (2010) and Kikkert et al. (2012) to study the detailed hydrodynamics of swash over immoile, impermeale eaches of varying surface roughness, and y Kikkert et al. (2013) and Steenhauer et al. (2011) to measure hydrodynamics over and within immoile, permeale eaches. The present study 3

4 uses the same damreak facility as used for these previous experiments, ut with the each now consisting of moile sediment, the primary ojective eing to measure the intra-swash sediment flux for well-controlled swash conditions. Previous damreak swash experiments involving a moile sediment each are limited to Othman et al. (2014), who used a sloping damreak apparatus to measure swash uprush sediment transport at the end of a truncated slope, their particular focus eing on the influence of grain size and pressure gradient on sediment transport, not on the detailed intraswash sediment flux. This paper reports on damreak swash experiments in which intra-swash flow depth, flow velocity and sediment flux are measured at a numer of cross-shore locations for swash on eaches consisting of moile, coarse-grained sediment. The experiments involve two each types: a coarse sand each and a gravel each. The experimental setup is the same as that used for the permeale, immoile each experiments of Kikkert et al. (2013), which means that for each of the present moile ed experiments, the incident ore, each slope, each material and each permeaility are the same as for the corresponding immoile each experiment reported y Kikkert et al. (2013). Comparing hydrodynamic measurements from the present experiments with the hydrodynamic measurements from Kikkert et al. (2013) therefore enales the effects of ed moility on the swash hydrodynamics to e isolated and quantified. More importantly, the present experiments yield measurements of intra-swash sediment flux for well-controlled, large-scale swash events. To the authors knowledge, intra-swash flux measurements of this kind, comined with detailed depth and velocity measurements, have not een reported previously. The details of the experimental setup are presented in Section 2. Section 3 presents the experimental results for shoreline motion, swash depths and depth-averaged velocities, including comparisons with results from the corresponding immoile ed experiments of Kikkert et al. (2012, 2013) in order to quantify the effects of ed moility on the swash hydrodynamics. The measured intra-swash sediment flux is presented in Section 4, followed in Section 5 y a comparison of the measured flux with the flux calculated using a ed load sediment transport formula. Section 6 concludes the paper with a summary of the main results. 2.0 EXPERIMENTAL SET-UP AND MEASUREMENTS 2.1 Set-up and test conditions 4

5 P for The experiments were carried out using the same facility used y Kikkert et al. (2012, 2013). A reservoir was placed at one end of a 20 m-long, 0.9 m-high and 0.45 m-wide, glass-sided flume (Figure 1). The reservoir is fronted y a gate, which is rapidly lifted y a falling-weight mechanism to produce a damreak-generated ore. The reservoir is m long (inside dimension), m wide and filled with water to a depth of m; the water depth in front of the gate was set at 62 mm. A 1:10 each was located downstream from the reservoir. The initial shoreline on the each, corresponding to the intersection of the water surface with the top of the each roughness, was m from the toe of the each and 4.82 m from the gate. The origin of the x z coordinate system is at the initial shoreline, with the x -axis parallel to the each slope and positive shoreward, and the z -axis perpendicular to the slope. The gate is raised at time t = 0, resulting in a plunging wave, which produces a ore, approximately 0.25 m high propagating with approximate speed 2.0 m/s towards the each. The ore collapses on the each, producing a single, repeatale swash event, with velocity, depth and maximum run-up magnitudes similar to those of full-scale swash in the field. Experiments were carried out on two each types: a coarse sand (CS) each with d = 1.3 mm and a 50 gravel (GV) each with d = 8.4 mm. Constant head permeameter tests (Steenhauer et al., 2012) 50 estalished Forcheimer coefficients a k = 81.2 s/m, k = 3587 sp 2 2 s/m, k = 383 sp P/mP Pfor the GV each (here I au u 2 P/mP 2 the CS each and a k = = k D + k D, where I is hydraulic gradient and D Darcy velocity); the sand is therefore an order of magnitude less permeale than the gravel. The sediments are the same as those used for the corresponding immoile, permeale each experiments reported in Kikkert et al. (2013), for which the top layer of the sediment each was made immoile using a dilute cement mix, without changing the permeaility (Steenhauer et al, 2011). The permeaility of each of the present CS and GV moile eaches is therefore equal to the permeaility of each of the corresponding immoile, permeale eaches reported in Kikkert et al. (2013). u is The sediment occupied the full each, i.e. from the surface of the 1:10 each face to the floor of the flume. Lines corresponding to the required 1:10 each slope were drawn on the glass sides of the flume and the each surface was matched to these lines efore each swash run. The 1:10 lines on the glass were approximately 1mm thick. Between runs the sediment ed was re-shaped y hand so that the top surface of the sediment ed was as close as possile to the top of the line, ut never aove or elow; hence the variation in initial ed level etween runs is estimated to e of the order of 1mm. No sediment was present on the horizontal ed of the flume etween the gate and the each, which means the incoming ore does not arrive at the shoreline already loaded with sediment. This feature of the experiment differs from swash in the field, where ore-advected sediment can make a 5

6 significant contriution to sediment flux on the each (Alsina et al., 2011). The far (downstream) end of the each was supported y a vertical sheet of 30 mm-thick marine plywood, drilled with several large-diameter holes and covered with 1 mm-diameter stainless steel mesh to allow the each to drain. Each experiment started with the each groundwater level equal to the water level in front of the reservoir (62 mm). This was achieved using a weir with 62 mm crest elevation placed approximately 0.5 m eyond the end of the each (Figure 1). Between experiments the each was relevelled to its 1:10 slope and the each was given time to fully drain to the 62 mm groundwater level. Drainage times for the CS and GV eaches were approximately 60 mins and 6 mins respectively. 2.2 Hydrodynamic measurements Simultaneous measurements of velocity and depth were otained at a numer of cross-shore locations using the comined particle image velocimetry (PIV) and laser-induced fluorescence (LIF) system previously descried y Kikkert et al. (2012). The water in the reservoir was seeded with neutrally-uoyant 20 μm titania particles for the PIV measurements and with fluorescent dye (concentration 0.1 mg/l) for the LIF measurements. The PIV/LIF measurements were made at five cross-shore locations on the each. At each location the Nd-YAG laser sheet was introduced to the flow through a window in the flume floor and a highly-polished perspex tower extending from the flume floor to the each surface (Figure 1; see also Kikkert et al., 2013). Each tower was 20 mm wide and 200 mm long (in the cross-shore direction); the top edge had a 1:10 slope to match the each slope and the height was such that the tower was flush with the each surface at the measurement position. Velocities were not measured at the moment of ore arrival ecause of air ules and ecause of high sediment load locking the light sheet. Thereafter, the presence of individual sediment grains in the flow give rise to spurious velocity vectors that were removed y the post processing. During the experiments the tower on the lower part of the each ecame slightly exposed ecause of slight erosion in its vicinity; scour at the other towers was negligile. No tower was completely covered y sediment at the end of a swash run ecause accretion levels at the tower locations was close to zero. DANTEC Dynamic Studio v1.45 software controlled the PIV and LIF cameras, enaling synchronised PIV and LIF images to e recorded at a frequency of 13.5 Hz. The PIV-LIF measurements were triggered at the moment of gate release. For the CS each, simultaneous velocity and depth measurements were otained at five cross-shore locations: x = 0.072, 0.772, 1.567, m and m. Because swash depth decreases with increasing x, the PIV camera view area was smaller higher up the each: 231 mm x 173 mm at x = m, reducing to 114 mm x 86 mm at x = m. The resulting 6

7 instantaneous velocity vector fields have spatial resolution etween 2.5 mm (at x = m) and 1 mm (at x = m), with corresponding random errors etween 14 mm/s and 7 mm/s. The LIFmeasured flow depth has a random error etween 0.3 and 0.1 mm. For the CS each, the simultaneous velocity and depth measurements were repeated 10 times for each of the five measurement locations; more repeats were not possile ecause of the long time to drain the each following each swash run. For the GV each, simultaneous velocity and depth measurements were otained at three cross-shore locations: x = 0.072, and m; the lower maximum run-up on the gravel each precluded measurements at higher x. The PIV camera view area ranged from 241 mm x 181 mm at x = m to 186 mm x 140 mm at x = m. The measured velocity vector fields have spatial resolutions etween 2.4 mm and 1.9 mm, with random errors etween 15 mm/s and 12 mm/s. The LIF-measured flow depth has a random error less than 0.3mm. For the GV each, the simultaneous velocity and depth measurements were repeated 15 times for each of the three measurement positions. The ensemle-averaged depth at location x measurements via and time t was otained from the LIF depth N 1 h xt = h xt (1) (, ) (, ) N i = 1 i where h i is the LIF-measured depth for run i and N is the numer of repeats of the experiment ( N = 10 for CS and N = 15 for GV). Note that depth here refers to the distance from the instantaneous water surface to the top face of the PIV/LIF tower. The ensemle-averaged, depthaveraged velocity at location x and time t was otained via N h 1 1 i u x t = u x z t dz N i= 1 hi 0 (, ) i (,, ) (2) where angle rackets indicate depth averaging. For the remainder of the paper we use simply ut () and ht () to denote the ensemle-averaged, depth-averaged velocity and ensemle-averaged depth respectively. Figure 3 presents example depth and velocity measurements from individual runs and the corresponding ensemle-averaged results. 7

8 In addition to measuring depth at selected x locations, measurements were also made of the swash lens (swash lens eing the term used for the instantaneous cross-shore profile of flow depth). This was achieved y mounting the laser aove the flume in order to illuminate approximately 0.3 m of the water surface from aove and using LIF to otain the time-varying flow depth over the 0.3 m extent of the measurement with frequency 13.5 Hz (Kikkert et al., 2012). By moving the laser and LIF camera etween repeats of the swash event, we otain depth measurements for multiple 0.3 m-long crossshore portions of the lens. Comining these measuremenst with each other, and with the depth measurements from the PIV-LIF locations, yields the swash lens measurement. 2.3 Sediment flux measurements Intra-swash sediment flux was measured using sediment traps. Different traps were used for the uprush flux and the ackwash flux. The uprush trap (Figure 2) comprises a net and timing mechanism that allows the net to e raised at a predetermined time. The net material was chosen to ensure sediment is trapped while keeping net porosity as high as possile: muslin and 2 mm nylon netting was used for CS and GV respectively. To minimise interference with the uprush flow, the net extended eyond the point of maximum run-up. The net was held in place y frames made from 1 mm-thick, 25 mm-wide aluminium ar; the frames were 204 mm high and 451 mm wide. Being slightly wider than the inside width of the flume, the good fit meant that no separate clamps were required, the trap was secure in the flow ut sufficiently free to e rapidly lifted. The trap was placed with the ottom edge of its opening at a depth of 4-5 mm elow the ed surface (corresponding to approximately 3.5d 50 for the sand each and 0.5d50 for the gravel each). In the case of the gravel each, individual particles are entrained y the flow and are ounced higher into the flow y collisions with neighouring stationary particles; the result is that the gravel transport takes place aove the initial mean ed level, not elow. The trap was raised vertically using an electromagnet-controlled falling-weight mechanism connected to the front frame of the trap y steel wire and pulleys. The time of the trap lift was controlled y a pulse generator triggered y the lifting of the reservoir gate. The delay programmed into the pulse generator takes into account the time delay etween the release of the trap weight and the lifting of the net. The delay was measured at each trap position using a reed switch at the trap linked to the gate reed switch and the pulse generator: the average of 10 measurements of the delay was taken. The uncertainty in the time etween the release of the trap weight and the net eing raised is the largest source of timing error; the standard deviation of the 10 measurements was taken as a measure of the uncertainty and was found to range etween and s across the five trapping locations. 8

9 P is The intra-swash uprush sediment flux at a given x was measured y lifting the trap at different times for repeats of the swash event and measuring the (oven-dried) mass of sediment collected over the time interval. The uprush sediment flux for the th i trapping period is then q s ( i) 1 1 mi ( ) mi ( 1) = w ρs tl( i) tl( i 1) (3) where tl ( i ) and L ( 1) qs L ( ) t i are the trap lift times (s) for the th i and previous time interval respectively; 3 i is the mean sediment flux per metre flume width (mp P/s/m) over the time interval ( ) ( 1) t i t i L ; s ρ = 2650 kg/mp 3 the sediment density; w = 0.45 m is the width of the flume; mi ( ) and mi ( 1) are the masses (kg) of sediment collected during the i th and previous time interval respectively; t ( i= 0) = t = the time of ore arrival and ( 0) L a mi= = 0. The shortest time interval, corresponding to the time of ore arrival when flux is highest, was set at 0.2 s in order to ensure the interval time error does not exceed 5% (given a maximum uncertainty of 0.01 s in the timing). The ackwash trap (Figure 2) is ox-shaped, open at the front and faces up the each. It comprises an aluminium frame, steel wire mesh on the ottom and at the ack, and muslin on the sides. The trap is 0.31 m high, m wide (slightly narrower than the flume width) and m long (sufficient to trap sediment entering at the front end). The trap is initially suspended aove and parallel to the 1:10 each face using a support structure attached to the top rails of the flume. Operation of the trap involved lowering to the ed at a prescried time during the ackrush, holding on the ed, and raising at a prescried time. Trap lowering to the ed is achieved via a compressed air-driven actuator (SMC CD85N B, 1.0 MPa max pressure). The actuator has a solenoid-controlled valve to control the direction of compressed air, and a 5-port air control, including 2 exhaust ports, which controls the input pressure and hence the acceleration and deceleration of the trap. The actuator enales the trap to e lowered very quickly; four springs provide cushioning as the trap comes in contact with the each surface. The extent to which the ottom edge of the trap goes elow the ed surface level is not known for the ackwash. In the case of the gravel the ackwash flux was negligily low. For the sand, the lower ackwash velocities (compared to uprush) mean that only the surface layer of sand is moilised in the ackwash. The thickness of the edging on the trap is less than one sand grain size, so even if the trap sits on the each surface with no penetration the sand flux enters the trap. Like the uprush trap, the ackwash trap timing is controlled y a pulse generator, triggered y the lifting of the reservoir gate. At the moment of gate lifting, the reed switch triggers the pulse generator, which sends 9

10 two delayed pulses to the actuator, the first to lower the trap and the second to lift it. The delay programmed into the pulse generator for lowering the trap takes into account the time required for the trap to travel from its suspended position aove the each to the each surface. The uncertainty in the time that the trap is on the each surface is the largest source of error in the ackwash timing. For this reason the trapping times are measured for each experiment individually via a micro-switch attached to one of the guiding rods. The switch is activated when the trap is within 3 mm of the each surface; it therefore sends two pulses to the data acquisition system, one corresponding to the trap reaching the ed and the second corresponding to the lifting from the ed. Estimates for the error due to the variaility in the down-time, as a percentage of the time the trap is on the each surface, varied etween 0.7 and 10.6%. The intra-swash ackwash sediment flux is calculated from q s ( i) 1 1 mi () = w ρs tup ( i) tdn ( i) (4) where tdn ( i ) and tup ( i ) are the times (s) of trap lowering and raising respectively; qs ( ) 3 sediment flux per unit flume width (mp mass (kg) of sediment collected during the time interval. i is the mean P/s/m) over the time interval t ( i) t ( i) and ( ) up dn mi is the For the CS each, uprush and ackwash sediment flux measurements were made at four cross-shore locations, at x = 0.602, 1.447, and m (the locations of the sediment flux measurements were slighty seaward of the PIV-LIF measurements in order to avoid the perspex towers). At each location, uprush and ackwash flux was measured for 5 time intervals (except at x = m for which ackwash flux was measured for 3 time intervals), and measurements were repeated 3 to 6 times. For the GV each, uprush sediment flux measurements were made at x = 0.602, and m; the lower maximum run-up on the gravel each precluded measurements at higher x. At x = and m, the uprush flux was measured for 5 time intervals, with 3 to 4 repeats of each measurement; for x = m, the uprush flux was measured over one time interval only ecause of the short duration of the swash at this position. No ackwash flux measurements were made for GV ecause ackwash flux was negligile at all measurement locations in this case. 10

11 Tale 1 presents the details of the sediment flux measurements: 176 measurements were made in total, covering four x -locations for CS and three x -locations for GV, with typically five intra-swash times at each location and typically four repeats of each measurement. Note that the each was fully drained to the initial 62 mm groundwater level and restored to its 1:10 uniform slope etween each run. 3.0 EXPERIMENTAL RESULTS: HYDRODYNAMICS 3.1 Shoreline motion and maximum runup The swash lens measurements have een analysed to determine shoreline motion through the swash cycle, with instantaneous shoreline position defined as the cross-shore location where the instantaneous lens depth is 5 mm (Kikkert et al., 2012). The ensemle-averaged shoreline trajectories for the moile CS and GV eaches are presented in Figure 4, along with the shoreline trajectory results from the corresponding immoile, impermeale and immoile, permeale each experiments reported y Kikkert et al. (2012, 2013). Shoreline motion for the moile CS each (Figure 4a) is similar to that for the corresponding immoile eaches during the first 1 s following ore arrival. Within this time the shoreline advances approximately 2.3 m. Thereafter shoreline deceleration increases as the shoreline approaches maximum run-up. The deceleration is higher and the maximum run-up correspondingly lower for the moile CS each compared to the immoile eaches: maximum run-up is 3.95 m, reached at 5.33 s, compared with a maximum run-up of 4.33 m and 4.53 m for the immoile permeale and immoile impermeale eaches respectively. There is therefore a 13% reduction in maximum run-up etween the immoile, impermeale each and the moile each, with two thirds of this reduction arising from ed moility and one third from ed permeaility. For the relatively low-permeaility CS each therefore, the effect of ed moility is more significant than the effect of ed permeaility in terms of maximum run-up. Backwash shoreline motion is similar for the three eaches, although the speed of shoreline retreat is a little higher for the moile and permeale eaches compared to that of the impermeale each: an average speed of approximately 1.5 m/s for t 7 s on the moile and permeale eaches compared with an average speed of 1.3 m/s for the impermeale each. The slightly higher retreat speed is likely due to infiltration into the moile and permeale eaches during ackwash (as discussed y Steenhauer et al. (2011) and Kikkert et al. (2013), 33% of the water volume crossing x = m infiltrates the CS permeale each, with 13% of this infiltration occurring during the ackwash). 11

12 Figure 4() shows the measured shoreline trajectory for the moile GV each. The trajectory is similar to that of the corresponding immoile, permeale each (except close to maximum run-up, etween 3.5 and 4.5 s), ut it is very different to that of the corresponding immoile impermeale each. Infiltration is much greater for the GV each compared with the CS each, with 45% of the water volume that crosses x = m infiltrating the GV each during uprush (Steenhauer et al., 2011). Maximum run-up on the moile each is 2.7 m, occurring at 4.5 s, compared with a maximum run-up of 2.93 m and 3.95 m on the immoile permeale and impermeale eaches respectively. For GV therefore, the effects of permeaility on shoreline motion are much more significant than the effects of ed moility: maximum run-up on the moile each is 32% lower than maximum run-up on the corresponding immoile, impermeale each, with 26% of this eing due to the permeaility. 3.2 Swash depths Figure 5 presents swash lenses for twelve intra-swash times on the CS each and Figure 6 presents the ensemle-averaged depth time-series for the five ( uh, ) measurement locations. The lenses and depth time-series for the GV each are presented in Figures 7 and 8. In the case of CS, the lenses and depth time-series for the moile each are similar to those of the immoile permeale and impermeale eaches. Significant differences in the lenses are seen only at times close to maximum uprush (i.e. etween 4.52 and 6 s), when the moile ed lenses tend to e lower than the immoile ed lenses; significant differences in the depth time-series are seen only at locations far up the each slope ( x = m in Figure 6), where moile ed depths are lower than the immoile ed depths. These results are consistent with the results for shoreline trajectory seen in Figure 4, and are also consistent with the hypothesis that ed moility serves as an added momentum sink through an increase in the effective ed roughness and through fluid-particle interactions. In the case of GV (Figures 7 and 8), the high permeaility of the moile and immoile permeale eaches results in the swash lenses for these eaches eing very different to those of the immoile impermeale each: the lenses for the moile and permeale eaches are lower than those of the impermeale each and extend much less far up the each (lower maximum run-up). At the same time, the moile ed and permeale ed lenses are rather similar (ut not the same), underlining the earlier result that the effects of ed permeaility are more significant than the effects of ed moility in the case of GV. Differences in swash depth etween the moile and the immoile, permeale eaches are seen in the region x > 1 m, where depths on the moile each are up to 20 mm higher 12

13 than on the the immoile, permeale each during the later uprush. The reason for the difference is likely related to the fact that gravel entrained lower down the slope tends to deposit in this region during uprush, therey increasing the ed level and forcing the free surface upwards y an amount equivalent to 1 2 sediment grain diameters. 3.3 Swash velocities The depth-averaged velocity measurements are presented in Figures 9 and 10. For the CS each (Figure 9), except far up the each, uprush velocities for the moile each are similar to those of the immoile eaches in terms of the timing and magnitude of the maximum velocity, the flow deceleration and the timing of flow reversal. Noticeale differences occur in the ackwash, with ackwash velocities for the moile each and the immoile permeale each eing lower than ackwash velocities on the impermeale each. Also, the moile each ackwash velocities are lower than the permeale each ackwash velocities for three of the five measurement locations. The difference at x = m reflects the lower maximum runup on the moile each, while the lower ackwash velocities at x = and m may e an effect of an increase in the effective ed roughness and momentum loss to fluid-sediment interactions, as more sand ecomes moile during the later stages of the ackwash. Figure 10 presents the depth-averaged velocity time-series for the three measurement locations on the GV moile each. As for the GV shoreline trajectories, the results are dominated y the effects of permeaility, with ackwash velocities on the moile and immoile permeale eaches eing much lower than those on the impermeale each. Backwash velocities are low in the case of the GV moile each - up to a factor of 2 lower compared with the CS moile each which means there is limited capacity for ackwash sediment moility in the case of the GV each. 4.0 EXPERIMENTAL RESULTS: SEDIMENT FLUX Figure 11 presents the processed flux results. The figure shows the intra-swash sediment flux for the four flux measurement locations on the CS each and the three flux measurement positions on the GV each. All 176 individual flux measurements are shown (although not all are visile ecause of overlapping results); the average flux at each x and t is indicated y the horizontal line, the length of which indicates the sediment trapping time interval for the measurement. The transported masses of sediment are sustantial: the highest uprush mass recorded was approximately 3.5 kg, occurring at x = m on the GV each; this equates to a 7.8 kg/m uprush mass, which is comparale to uprush masses measured in the field (Blenkinsopp et al., 2011). (Note that a suset of the CS data was reported y Briganti et al. (2012) and used thereafter y Hu et al. (2015); the results presented here 13

14 constitute the entire experimental dataset for the two eaches and contain corrections to the ore arrival times and, consequently, to the sediment flux at ore arrival reported in Briganti et al. (2012)). There is generally good agreement etween individual flux measurements for given x and t and the results are consistent in terms of their variation with x and t, and consistent with visual oservations made during the experiments. For each x the sediment flux is highest at the time of ore arrival, decays rapidly with time thereafter and reaches zero at some time efore flow reversal. Peak uprush flux is much higher for GV than for CS a factor 2 higher approximately - consistent with a higher ed shear stress arising from the rougher ed. In the ackwash, the flux is slow to uild up as the flow accelerates down the each. For the GV each, the ackwash sediment flux is negligily small. For the CS each the ackwash flux peaks at aout the time of maximum ackwash velocity; peak ackwash flux is higher and occurs later in time for lower locations on the each, consistent with the ackwash velocities. No sediment was transported off the each onto the horizontal ed etween the gate and the each. This is ecause the supercritical ackwash flow jumps to sucritical elow the initial shoreline location (ut still on the slope) at the end of the ackwash, with the result that particles in suspension settle to the ed, raising the ed elevation in this region y a few millimetres in the case of the CS each and y order of a grain size in the case of the GV each (note that ed elevation was not measured). This is consistent with the creation of a ed-step at a ackwash ore as descried y Zhu and Dodd (2015). In addition to the accretion oserved elow the initial shoreline, accretion also occurred in the mid-to-upper swash, reaching an estimated (not measured) maximum of order of a few millimeters in the case of CS and of one grain size in the case of GV. After each swash run the sediment was re-distriuted to restore the 1:10 each slope y matching the each surface to the lines marked on the glass sides of the flume. 5.0 SEDIMENT TRANSPORT CALCULATION In Section 5.2 we apply a sediment transport formula to the experimental conditions: we take the measured hydrodynamics as input to the transport formula and compare the formula-calculated intraswash sediment flux with the measured flux. Because the sediment transport formula is ed shear stress-ased, we first give some consideration to estimating ed shear stress for swash flow. 5.1 Swash ed shear stress We assume the instantaneous ed shear stress, τ, relates to the instantanteous flow through 1 τ = ρ fuu (5) 2 14

15 3 where ρ is the water density (1000 kg/mp P) and f is the ed friction factor. For steady flows and oscillatory flows, f depends on Reynolds numer and relative roughness, and there are wellestalished methods for estimating f for such flows. Swash ed shear stress is complicated y the flow s high unsteadiness and non-uniformity. Measurements of intra-swash ed shear stress have een otained y Conley and Griffin (2004) using a hot-fim sensor at a field site, y Barnes et al. (2009), Pujara et al. (2015) and Jiang and Baldock (2015) using a shear plate on laoratory impermeale and moile slopes, and y Kikkert et al. (2012, 2013) using log-law fitting to PIV-measured velocity profiles over laoratory impermeale and permeale eaches. Such measurements have yet to result in a predictive formula for intra-swash ed shear stress. In the asence of an estalished predictor, here we estimate f using two methods. The first is ased on the Swart formula (1974) commonly used to calculate friction factor for wave-driven oscillatory flow conditions. The approach was previously used y Othman et al. (2014) to estimate friction factor for their laoratory overwash experiments and y Masselink and Turner (2012) for swash flows in field conditions. The Swart formula is f a = exp k s (6) where a is the amplitude of the oscillatory flow water particle displacement and ks = 2.5d50 is the ed roughness, with d 50 the sediment size. Applied to swash, we estimate a from Ts a= usd (7) 2π where u sd is the standard deviation of ( ) u t at a given x location in the swash zone and Ts = tend ta is the swash period, with t a eing the time of ore arrival at x and t end the time corresponding to the end of the swash at x (when the ackwash velocity is zero). Sustitution for a in equation (6) yields a time-invariant f for the given x location. Of course, the applicaility of Swart to swash flow is questionale: for Swart the near-ed hydrodynamics are estalished over multiple flow cycles and the flow depth is much greater than the oscillatory oundary layer thickness. In contrast, here we 15

16 have a single flow cycle with peak velocity occurring at the start of the cycle and flow depth varying and ecoming very shallow in the late ackwash. A disadvantage of Swart in the context of predictive modelling is the need to know the swash period T s and the velocity standard deviation u sd a priori. The second method used to estimate f is the Colerook formula which calculates instantaneous velocity: f ased on the instantaneous flow depth and 1 k 2.51 s = 2log10 + λ 3.7Dh Re λ (8) uh where λ = 4 f, Dh = 4h and Re = is Reynolds numer, with ν eing the water kinematic ν viscosity. Again, the applicaility of Colerook for estimating swash flow f is questionale for a numer of reasons. First, the Colerook formula is applicale to conditions in which the oundary layer is fully developed, a condition that is unlikely to e met at, and soon after, the time of ore arrival at a location on the each. Second, the Colerook formula applies to turulent flow, a condition that is not met close to the time of flow reversal (in addition, Re = 0 at flow reversal, giving a singularity for f at this time) and towards the very end of the ackwash (when oth flow depth and flow velocity are low); this is a minor concern however in the context of estimating intra-swash sediment flux ecause the percentage of the swash duration for which the flow is not turulent is small (less than 5% in the mid-swash) and velocities are very low at these times. Third, in the late ackwash a fully-developed oundary layer cannot occur ecause of the limited water depth as the flow ecomes increasingly shallow; a Lagrangian-type model as suggested y Barnes and Baldock (2010) may e more appropriate at this stage of he swash. Moreover, in the extreme late ackwash, when the flow depth is extremely shallow, the flow regime ecomes akin to flow over ostacles rather than a oundary layer flow condition. To compare the two methods with each other, and to evaluate them against experimental data, we apply the methods to the experimental data of Kikkert et al. (2013), who measured swash depths and velocities at four crossshore locations on immoile, permeale coarse sand and gravel eaches and otained estimates of the intraswash ed shear stress at each location ased on log-law fitting to the measured velocity profiles. Here we take the measured depth and velocity time-series as input to the Swart and Colerook calculations. At each cross-shore measurement location the measurements give t a and u( t ), ( ) 16 ht for ta < t < tend (u and h cannot e

17 accurately measured at, and immediately following, the moment of ore arrival ecause of air ules in the arriving ore front). The measured data are processed as follows to give u( t ) and ht ( ) spanning the full swash period from t = ta to t tend = : (i) set end t equal to the time of the last (, ) uh measurement plus one measurement timestep; (ii) extrapolate the measured u( t ) to find ( = ) and set ( ) 0 extrapolate the measured ht ( ) to find ( = ) and set ( ) ht t a u t t a u t = t end = ; (iii) ht= t end equal to the last measured depth; (iv) interpolate to get u( t ) and ht ( ) at 200 evenly-spaced times etween t a andt end. An example calculation is shown in Figure 12: it takes Kikkert et al. s (2013) measured u( t ) and ht ( ) at x = m on their coarse sand each as input and calculates f and τ using Swart and Colerook. The Swart f is time-invariant with, for this example, a value of The Colerook f is time-varying and lower than Swart f for most of the swash; the lowest Colerook f for this example is , approximately 0.7 times the Swart f, and occurs around midway during the uprush when the flow depth is a maximum and the Colerook relative roughness ks D h is a minimum. In the late ackwash the flow depth ecomes shallow, leading to a rapid increase in the relative roughness and to Colerook Colerook f reaching twice the value of Swart f are reflected in the calculated ed shear stresses, ( t) Note that the large difference in ecause of the low velocities at this time. f. The differences etween Swart and τ, shown in the ottom panel of Figure 12. f in the late ackwash translates to a relatively small asolute difference in τ Figure 13 shows the Swart- and Colerook-calculated ( t) τ otained using the measured depths and velocities for all four locations on Kikkert et al. s (2013) coarse sand and gravel eaches; Kikkert et al. s log-law-ased estimates of ( t) τ are also shown. A comparison etween the Swart and Colerook ed shear stresses ( τ S and τ C respectively) is presented differently in Figure 14 y plotting τ S versus τ C. The results show that τ S is generally greater than τ C during uprush; the ratio of τ S to τ C in the uprush lies in the range 0.9 < τ τ S C up < 1.7 across all locations on the two each types, with the value of the ratio mainly dependent on cross-shore position. In the early ackwash τ S and τ C are in close agreement, while in the late ackwash τ C ecomes increasingly greater than τ S ; the ratio of τ S to τ C in the ackwash lies in the range 0.5 < τ S τ 0.25 < τ S τ C ack C ack < 1.1 for CS and in the range < 1.1 for GV, with the value of the ratio mainly dependent on time in the ackwash. 17

18 In Figure 13, Kikkert et al. s (2013) experimental results for ( t) τ, ased on log-law fitting to the measured velocity profiles, show the same general ehaviour as Colerook and Swart decreasing τ during uprush, increasing in the early ackwash and decreasing again in the late ackwash ut the results show significant scatter and varying level of agreement with Swart- and Colerook-calculated τ, depending on position on the each and depending on uprush or ackwash. Figure 15 shows the comparison y plotting calculated τ (taken as the mean of τ S and τ C ) against experimental τ. The calculated τ lies mostly within a factor 2 of the experimental values for oth eaches, ut tends to e higher than the experimental τ during uprush and much lower in the late ackwash. Note that the experimental results cannot e taken as ground truth in this comparison ecause the unsteady and non-uniform nature of the flow means that the validity of the log law is questionale and its practical application difficult, which may explain some of the scatter in the experimental results. We note that a previous comparison of log-law-ased measurements of swash ed shear stress over impermeale eds showed reasonale agreement with corresponding shear plate measurements for uprush, ut very poor agreement in the ackwash, with the log-law ackwash estimates eing much higher than the shear plate ackwash estimates (O Donoghue et al., 2010). 5.2 Intra-swash sediment flux Given the relatively large sediment size, the sediment flux is treated as edload with the Meyer-Peter and Müller formula used to estimate the instantaneous flux, i.e. 3 ( ) ( ) 1.5 θ qs = C sg 1 gd50 θ θcr (9) θ where C is a constant, s g = 2.65 is sediment specific weight, g is acceleration due to gravity and θ = ρ τ ( sg 1) gd50 is instantaneous non-dimensional ed shear stress, with water density ρ = kg/mp P; θ c is the non-dimensional ed shear stress corresponding to the threshold of sediment motion and is here calculated in the same way as in Othman et al. (2014), including an adjustment for the ed slope. We process the measured u( t ) and ht ( ) from the present moile ed experiments in the same way as previously descried, calculate ( t) τ via the Swart and Colerook formulae and use equation (9) to otain two estimates of the intra-swash sediment flux, q ( ) s t, one ased on the Swart- 18

19 calculated τ ( t) and the other ased on the Colerook-calculated ( t) τ. In applying Swart and Colerook we make no allowance for the fact that the ed is now moile, even though the presence of a moving sediment layer may increase the ed shear stress somewhat (Jiang and Baldock, 2015). For steady flows, a value of 8 is normally used for the constant C in the Meyer-Peter and Müller formula; here we adopt C = 12, as is often used in the case of unsteady and oscillatory flows (e.g. Nielsen, 2006). Note that the choice of value for C (12 or 8) generally has a similar level of impact on the calculated instantaneous flux as the choice of Swart or Colerook for the friction factor. Figure 16 presents the calculated and measured intra-swash flux for four locations on the CS each and two locations on the GV each (flux measurements at the third flux-measurement location on the GV each give only one point for the time-series). The experimental results shown are the averaged results from Figure 11 and include a horizontal line to indicate the trapping time interval for the measurement. For the purpose of comparing the measured and calculated flux, we ignore the difference (approximately 0.1 m) in the position of the sediment flux and the position of the hydrodynamic measurements: i.e. we assume the flux measured at the location adjacent to the ( uh, ) measurement location can e attriuted to the ( uh, ) measurement location. In Figure 16, uprush sediment flux ased on Colerook is always lower than that ased on Swart, reflecting the differences already seen etween Swart- and Colerook-calculated differences are more apparent for on f for uprush; the q s than for τ ecause of the higher power dependence of 1.5 f ( τ f, q f ). With the exception of x = m for the GV each, there is good s agreement etween the Colerook- and Swart-ased estimates of the peak uprush flux at the time of ore arrival; the large difference at x = m on the gravel each is due to the relatively low value of a at this location, which increase τ S relative to τ C. In the ackwash, the Colerook and Swart estimates are in good agreement in the early ackwash; differences ecome apparent in the late ackwash when high Colerook friction factors result in higher q s compared to the Swart q s. q s As shown in Figure 17, uprush flux calculated using Swart and Colerook generally lies within a factor 2 of the measured uprush flux. In some cases (e.g. CS, x = m) the level of agreement etween the measured and calculated uprush flux is remarkaly good. Agreement is generally poor in the ackwash. For the CS each, there is reasonale agreement in terms of the magnitude of the peak ackwash flux (at least compared to Swart), ut there is discrepancy in the timing of the peak and of the ackwash flux generally: the calculated ackwash flux is generally higher than the measured flux 19