Sedimentology (2009) 56, 13461367

doi: 10.1111/j.1365-3091.2008.01037.x

Flow patterns, sedimentation and deposit architecture under a hydraulic jump on a non-eroding bed: dening hydraulic-jump unit bars
ROBERT G. MACDONALD*, JAN ALEXANDER*, JOHN C. BACON* and MARK J. COOKER *School of Environmental Sciences, University of East Anglia, Norwich, Norfolk NR4 7TJ, UK (E-mail: robert.macdonald@uea.ac.uk) School of Mathematics, University of East Anglia, Norwich, Norfolk NR4 7TJ, UK
ABSTRACT

This paper presents results from two ume runs of an ongoing series examining ow structure, sediment transport and deposition in hydraulic jumps. It concludes in the presentation of a model for the development of sedimentary architecture, considered characteristic of a hydraulic jump over a non-eroding bed. In Run 1, a hydraulic jump was formed in sediment-free water over the solid plane sloping ume oor. Ultrasonic Doppler velocity prolers recorded the ow structure within the hydraulic jump in ne detail. Run 2 had identical initial ow conditions and a near-steady addition of sand, which formed beds with two distinct characteristics: a laterally extensive, basal, wedge-shaped massive sand bed overlain by cross-laminated sand beds. Each cross-laminated bed recorded the initiation and growth of a single surface feature, here dened as a hydraulic-jump unit bar. A small massive sand mound formed on the ume oor and grew upstream and downstream without migrating to form a unit bar. In the upstream portion of the unit bar, sand ner than the bulk load formed a set of laminae dipping upstream. This set passed downstream through the small volume of massive sand into a foreset, which was initially relatively coarse-grained and became ner-grained downstream. This downstream-ning coincided with cessation of the growth of the upstream-dipping cross-set. At intervals, a new bed feature developed above and upstream of the preceding hydraulic-jump unit bar and grew in the same way, with the foreset climbing the older unit bar. The composite architecture of the superimposed unit bars formed a fanning, climbing coset above the massive wedge, dened as one unit: a hydraulic-jump bar complex. Keywords Backset beds, cross-bedding, ume, hydraulic jump, submerged wall jet, unit bar.

INTRODUCTION Hydraulic jumps occur naturally in a variety of subaerial environments such as in rapids and cascades, at levee breaches during oods and where steep channels enter lakes or reservoirs. These hydraulic jumps are characterized by a marked increase in free water surface level between a supercritical incident ow and a deeper, slower, subcritical ow. Hydraulic jumps form spontaneously in natural ows and are subject to a 1346

nely balanced interaction between inertia, pressure gradient and gravity. The Reynolds number, Re, is the ratio of inertial to viscous forces (Re = qud/l) where q is the density and l the dynamic viscosity. Re can occur over many orders of magnitude (Chanson, 2007a). In natural water ows, the Froude number, Fr = u/(gd)1/2 has rarely been observed to exceed 4 and is usually lower even in steep mountain streams (Jarrett, 1984). In unconned (and shallow) ow over steep noncohesive slopes, Fr rarely increases much above

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists

Dening hydraulic-jump unit bars critical value (i.e. 1) (Grant, 1997). Comparison of sediment behaviour in natural settings with that in engineered settings (e.g. Fr = 5 to 9 achieved with undershot sluice gates; Chanson, 1999) could be misleading. Here, ow structure (Fr = 273), sediment transport and deposition and deposit character associated with a hydraulic jump are described towards a model that can be used in interpretation of ancient deposits. A range of sedimentary features in alluvial fan, fan delta and jokulhlaup deposits contain backset beds, attributed to hydraulic jumps (e.g. Postma & Roep, 1985; Massari & Parea, 1990; Nemec, 1990; Russell & Knudsen, 2002; Cassidy et al., 2003; Breda et al., 2007). Steep-sided scours, rapidly lled with massive or diffusely graded deposits, have also been attributed to hydraulic jumps, notably in sites downstream of glacial debouchment (e.g. Gorrell & Shaw, 1991; Russell et al., 2003). Normally graded and cross-stratied gravels with scoured bases have also been described at such sites (e.g. Hornung et al., 2007). Despite the fairly common occurrence of hydraulic jumps in natural ows, the deposits associated with them are poorly documented or quantied and identication in ancient deposits is problematic. In one of the few previous laboratory studies of sedimentation under hydraulic jumps, Jopling & Richardson (1966) demonstrated how backset bedding formed under a hydraulic jump in a small ume (02 02 m cross-section, 06 slope, nonrecirculating). Alexander et al. (2001) described

1347

how sedimentary structures (termed chute and pool structures after Fralick, 1999) formed in a well-sorted sand bed by the rapid upstream movement of trains of hydraulic jumps in a ume. Neither of these studies could present detailed measurement of the associated ow but subsequent technological advancements have determined turbulent structure (e.g. Liu et al., 2004) and aeration (e.g. Murzyn et al., 2005; Lennon & Hill, 2006; Chanson, 2007b) in sediment-free settings. The two companion ume runs presented here stand independent to: Run 1 monitor the detailed ow structure; Run 2 (i) monitor the sediment transport behaviour and its effect on the ow behaviour; and (ii) document the deposit architecture. The constant ume slope and width make the nal model ideal for application in environments of ow expansion, gradually changing slopes, slope breaks and where water bodies interact, to generate hydraulic jumps.

Experimental procedure
The two runs are from an ongoing series, examining ow structure and sedimentation in subaerial hydraulic jumps using the tilting ume at the University of East Anglia. This ume recirculates water and sediment in a closed pumped loop (Fig. 1). The near-stationary hydraulic jump (Fig. 2) in Run 1 was the identical starting condition for Run 2, which involved the continuous addition of sand.

Fig. 1. The ume used for these experiments (not to scale). The inset box shows the pattern of UDVP probes, used in the two different arrays along the ume centreline.
2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1348
A

R. G. Macdonald et al.

Fig. 2. Still video images of the hydraulic jump in Run 1 at two time intervals, taken from a consistent distance, through the ume side wall. The blue sidewall vertical supports are 080 m apart. (A) Image of surface splash. (B) Image taken 032 sec after (A) without splash. The toe of the hydraulic jump is in a more upstream position.

The 10 m, straight, test channel has a square (1 1 m) cross-section, glass sidewalls and a at stainless steel oor. Channel tilt was set at 425 to the horizontal. A 0100 m high overshot weir was used at the downstream end of the channel (Fig. 1). The water discharge was kept constant at 026 m3 sec)1 throughout both runs (measured with electromagnetic ow meters in the recirculation pipes). The co-ordinate system used is x downstream (parallel to the ume oor), y crossstream and z normal to the oor; 0, 0, 0 was set at the point of entry to the test channel, on the rightlooking downow and at the ume oor. Throughout this paper, ow is viewed left to right. The x and z components of velocity are u and w, respectively. Repetition of runs with the same conditions reproduced a hydraulic jump in the same mean position, with the same free surface shape and internal ow pattern that remained steady for prolonged periods (many hours). For the two runs presented here, the hydraulic jump was set sufciently far down the ume for the supercritical jet free surface to be uniform across-stream. The hydraulic jump was sufciently far up the ume for sediment accumulation downstream of the jump in Run 2 to be largely unaffected by the ume end-wall conditions. There is a relationship between any inow characteristic and its devel-

opment length and this has a control on the character of the ow or ow structure (here the hydraulic jump structure) and the behaviour of sediment. A parabolic prole of streamwise velocity takes some time and space to develop and, once developed, the shape of the prole is maintained downstream, assuming no change in channel conditions. Kirkgoz & Ardiclioglu (1997) determined the distance, L, of development of a parabolic shape in their eq. (5), to be: L Res 76 00001 ds Frs 1

where ds is the supercritical ow depth and Frs and Res are the Froude and Reynolds of the inow, respectively. In many natural settings, the boundary conditions vary in space and time such that a fully parabolic prole will not develop. A hydraulic jump will form upstream of the position where a parabolic velocity prole would develop. Hydraulic jumps with a parabolic inow velocity prole may be considered a special case attainable only in unnatural settings. The experiment was set up with a partially developed inow to the hydraulic jump, which may be more common in nature. Flow depth downstream of the hydraulic jump was determined at the downstream end of the test

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars channel, dx=1000m, a position downstream of the vertical water surface uctuations associated with the behaviour of the hydraulic jump. Ultrasonic Doppler velocity proler (UDVP) probes, deployed in arrays, measured high-frequency time series of beam-parallel velocity components in transects away from the probes. Vectrino acoustic Doppler velocimeters recorded time series of all component velocities at isolated points (Fig. 3). Outputs were monitored in real time to ensure that aliasing was prevented and that the received signals were acceptably clean. Data were recorded for periods of at least 540 sec, to allow an accurate assessment of the time-averaged velocity allowing for turbulent uctuation (the average was taken over at least 500 readings at each spatial point). In order to maintain wet contact, UDVP probes were submerged by at least 005 m under the oscillating free surface. The ow was recorded on a video camera through a sidewall.
B A

1349

Run 1: Sediment-free hydraulic jump This run was undertaken to determine the water ow structure within a particular hydraulic jump. Flow depths upstream and downstream of the hydraulic jump were 009 and 025 m, respectively. A UDVP array of 10 probes, parallel to the ume oor and pointing upstream (array 1; Fig. 1), was deployed at 13 positions along the channel centreline to measure streamwise velocity (u) just upstream, within and downstream of the hydraulic jump. Two-dimensional (u and w) turbulent characteristics were assessed with data from a second UDVP array (array 2; Fig. 1) at four positions. Velocities in the supercritical ow were measured with a single probe and mean velocities are the average of 12 000 readings at each point. Run 2: Hydraulic jump with sand addition This run was undertaken on an initially at inclined ume oor so that deposition would be controlled by the ow, not the channel topography. Well-sorted, quartz-dominant sand was added to the ume at approximately 4000 kg h)1 over the duration of the 75 min run by conveyor feed (the duration was limited by the volume of sand available and the input rate). Sand was loaded by hand onto a conveyor at as steady a rate as could be maintained by shovelling, from ve crates. There was a short hiatus in sand loading when each full crate was manouvered into place. The hiatus was 5 min before crate 2 and crate 4 but only 2 min before crate 3 and crate 5. Sand added to the discharge tank mixed with sand leaving the downstream end of the test channel (Fig. 1) and travelled

Fr = 024 Re = 270000 Frs = 273 Re s = 282000 d = 025 m ds = 009 m

Fig. 3. Photograph and denition diagram of the hydraulic jump. (A) Oblique downstream view onto the surface of the hydraulic jump. Two acoustic Doppler velocimeters are deployed into the subcritical ow. (B) Diagram showing the sense of rotation in vortices above and below the detached jet (indicated by the thick black line). The general ow characteristics are given for the conditions on either side of the hydraulic jump.

through more than 15 m of recirculation pipes and pumps before entering the test channel. This process acted to damp any uctuation in sand ux. After initiation of sand input, the sand ux to the hydraulic jump did not cease until the run was stopped and no abrupt change in sediment ux was observed within the test channel. The input sand was well-mixed and unimodal (modal class was 250 to 300 lm), with a mean grain-size of 368 lm and standard deviation of 222 lm. Ultrasonic Doppler velocity proler array 2 was xed in position at x = 610 m throughout the run. Because the jump moved slowly upstream, the UDVP recorded u and w velocity components at increasing distances downstream of the hydraulic jump. The shape of the sand deposit was recorded at 23 instants and the position of the water-free surface was recorded at 13 instants, both by tracing onto the sidewalls. The appearance

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1350

R. G. Macdonald et al. hydraulic jump (Hornung et al., 1995; Chanson, 1999). Despite the steady ow conditions in the supercritical ow and at some distance downstream of the hydraulic jump, within the hydraulic jump a lot of the internal structures were transitory, repeatedly forming, changing and dissipating at consistent locations within the ow. The terminology used for the components of the ow is illustrated in Fig. 2. The supercritical ow into the hydraulic jump maintained the form of a steady submerged wall jet for 067 m downstream of the toe (the most upstream point of abrupt ow thickness change). The term wall jet is used here to describe the unidirectional component of ow that is bounded on the underside by the ume oor and that is overlain at some point by further uid region(s).

and developing deposit of the ows were recorded by videography. The nal deposit was photographed. At the end of 75 min the pumps were rapidly slowed to stop recirculation, without seiches or signicant sediment reworking and the water was drained from the ume. The deposits were examined in sections cut at intervals along and across the channel.

RUN 1: HYDRAULIC JUMP IN A SEDIMENT-FREE FLOW Observations of the ow through the glass sidewall (e.g. Figs 2 and 4A) indicated a more complicated time-dependent ow structure than that usually illustrated in theoretical models of a

Fig. 4. Flow characteristics of Run 1. (A) Flow structure. The blue free surface line indicates the maximum recorded splash heights, above the maximum and minimum instantaneous free surface positions when a splash was not present (two brown lines). The larger arrows represent components of the ow structures whose streamwise direction did not alter. The arrows are red in the roller, green in the detached jet and blue within the tailwater. The small blue arrows indicate the path of bubbles in the tailwater. (B) Proles of streamwise velocity component, u, averaged over 540 sec of measurement with UDVP probes at y = 0500 m. Prole position is indicated by the down-pointing grey arrows in (A). Dashed lines in Proles I and J are the velocity proles with outlying velocity data removed. The insert graph shows velocity proles measured in the supercritical ow (dark blue) and the toe of the hydraulic jump (pale blue) with an expanded vertical scale. (C) Vertical proles of Tke and syx calculated from UDVP array 2 data at four locations along the ume centreline for Run 1. The Tke proles are drawn in pink and the syx proles are drawn in black. The lowest position of the free surface is also drawn (compare with Fig. 8).
2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars The submerged wall jet thinned downstream (from 009 to 006 m). At 067 m downstream of the toe, the submerged wall jet expanded vertically (maintaining contact with the ume oor) and the fastest core rose to follow the highest position of the expanding jet, as a detached jet (Fig. 2). The detached jet approached the free water surface and thickened downstream to span the entire ow depth (Fig. 4A). This unidirectional ow spanning the entire ow depth downstream of the hydraulic jump is described as the tailwater. In the volume of uid below the lower boundary of the detached jet, there was no obvious ow movement except when bubbles (described below) highlighted discrete eddies. Above the submerged wall jet and detached jet was a recirculating roller with uid in its lower part travelling downstream, adjacent to the jet and in its upper part travelling upstream, below the free water surface (Fig. 2 and green arrow of Fig. 4A). The thin, highly aerated upstream ow along the top 005 m of the roller is termed the reverse ow strip (Fig. 2) and persisted throughout Run 1. Above the roller, the water surface oscillated with a 005 m vertical amplitude (Fig. 4A). Localized free surface peaks, typically spanning less than half the ow width, rose above this oscillation and spilled upstream in seconds (Fig. 2A and B). These are termed splashes and were observed at 5 to 45 sec intervals throughout Run 1. Splash height and position varied, with maximum heights recorded in Fig. 4A. The hydraulic jump toe oscillated in the streamwise direction between x = 438 and 488 m (total excursion of 050 m) around a near-stationary mean streamwise position. The toe oscillation was a local feature and had a frequency not simply related to the vertical, free surface oscillation of the roller. Video analysis showed that splashes were particularly high when the toe was in its most upstream position. The continuous breaking of the sloping front of the hydraulic jump admitted air bubbles into the ow via the toe. Incident supercritical ow was not aerated but bubbles constituted a considerable fraction of the roller and the ow downstream of it. The highest concentration of bubbles was at the turbulent shear layer between the upper surface of the submerged wall jet and the roller. The submerged wall jet and the core of the detached jet contained few bubbles compared with the water above and below. Immediately downstream of the toe there was a zone of vortex generation: the upper surface of the wall jet as it entered the deeper water ran under

1351

relatively stationary uid. The high shear in this region corresponded to: (i) high vorticity; and (ii) a thickening of the zone of transition from highspeed uid below and slow uid above. Vortices were generated within the shear layer and tightened to discrete entities 005 to 009 m in diameter. Some vortices followed the upper edge of the detaching jet. Others were deected downwards and brushed the ume oor, upstream of jet detachment. Throughout Run 1, further vortices were shed from the lower boundary of the detached jet and these rotated slowly without tightening as they moved into the quiescent region below the detached jet. These vortices generally were larger than, contained more diffuse bubbles than and rotated in the opposite sense to those generated in the shear layer. All vortices had axes of rotation across-stream. Groups of bubbles surrounded by uid without bubbles (bubble sets) exited the upper portion of the downstream edge of the roller in pulses approximately every 5 sec. These bubble sets followed a curved trajectory within the expanding jet and tailwater and burst at the free surface within the ume length (Fig. 4A). The bubble sets were restricted to the top quarter of the ow just downstream of the roller, extended down to three quarters of the ow depth about 11 m downstream (at x = 750 m) and had all escaped by 19 m.

General ow characteristics
Supercritical ow upstream of the hydraulic jump is dened by Frs (Fr > 1), where Frs is the Froude number for mean velocity Us and ow depth ds. As the wall jet enters the jump, Frs = 273, Res = 282 000 and Ls = 591. Downstream of the jump, Frx=640m = 024 and Rex=640m = 270 000. Such high Reynolds numbers conrm that the ow was fully turbulent throughout Run 1. Boundary layer separation from a wall occurs for ows of high Reynolds numbers (Re > 100 000 for ow around a bluff body) travelling from low to high pressure.

Velocity data
At 05 m upstream of the hydraulic jump (x = 420 m), the time-averaged streamwise velocity, u, reached a maximum (274 m sec)1) at ZUmax 0033 m above the ume oor (Fig. 4B inset). Supercritical ow was steady between upper and lower bounds and additionally had instances of slower ow (minimum 118 m sec)1,

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1352
A

R. G. Macdonald et al.
2000 1000 0 Profile G, Z = 027 m

1000 2000
B

Profile I, Z =

022 m

027 m

2000 1000 Streamwise velocity (m sec1) 0 1000 2000

C

2000 1000 0 1000 2000

D

Profile K, Z =

022 m

027 m

Supercritical profile, Z = ZUmax

3000 2000 1000 0 0 200 Time (sec)

Fig. 5. Time series of u at positions downstream through the jet and tailwater. (A) 12000 datapoints from a single probe, at z = ZUmax of the supercritical ow. Compare inset in Figure 4B. (B) to (D) 500 datapoints from array 1 at: (B) within the detached jet; (C) two positions within the tailwater near the upstream limit of the bubble pulses; and (D) further downstream within the tailwater. Note that in (D) the bubble pulses reach down to Z = 027 m more regularly than in (C).

400

600

half the lower bound) of up to 01 sec duration, with no apparent periodicity (Fig. 5D). As the supercritical wall jet became submerged, Umax slowed by 85% in 09 m (to 041 m sec)1 in Prole F). ZUmax moved down over a horizontal

distance of 067 m to within 0001 m of the ume oor and moved up to 003 m by Prole D (Fig. 4B). In Proles A to D, u decreased smoothly upwards, across the shear layer and became negative within the recirculating roller. The time-averaged jet expanded downow of Prole D and ZUmax moved upwards, following the upper limit of the expanded jet. A further peak in u occurred below ZUmax in Proles F and G (at z = 009 m). Underneath the length of the roller, the upper jet edge was abrupt. The lower edge of the detached jet was less abrupt. Although not material surfaces, jet edges are dened by the abrupt transition in velocity from the jet to the outer ow. The upper edge of the detached jet rose to 075 of the ow depth within 06 and, downstream of this, the jet became indistinct. Above the time-averaged jet, there was a minimum in u at z = 020 m; this was negative at the upstream end of the roller ()076 m sec)1) and increased to 003 m sec)1 downstream through the roller (Proles A to F). The integrated mean u at velocity Prole A is negative, indicating nett upstream volume ux (Fig. 4B); this resulted from the greatly aerated state of the recirculating ow within the roller; the negative (up-ume) volume ux of water (rather than water and air bubbles) was less than the down-ume volume ux of the submerged wall jet (that contains no bubbles). The highest magnitude of negative u in each of Proles A to D coincided with the stream of vortices above the detached jet. The ow below the time-averaged detached jet was directed downstream. Downstream of the jet detachment point and upstream of the point where the tailwater was fully developed (Prole G), the water near the ume oor was nearly static (less than 004 m sec)1 below z = 6 mm) and the near-oor velocities were greater both upstream and downstream of Prole G. In Proles F and G, two peaks in time-averaged u were recorded because the jet core position changed repeatedly between a jet-up and jetdown state. The u time series at ZUmax (z = 027 m) shows that most data were spread uniformly around a steady mean (Fig. 5A) and additionally u minima each lasting for less than 2 sec were recorded, at times when the whole detached jet was in the jet-down state and instantaneous ZUmax was below z = 027 m. The detached jet is transient in vertical space, through time. The time-averaged tailwater velocity is drawn above and below an offset, as two unjoined lines (Proles H to K; Fig. 4B). The height of offset followed the trajectory of the bubble sets (which

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars highlighted pulses from an upper region of the roller). The velocity proles below the offset remained unchanged throughout, with a symmetrical spread of data about a steady mean (black lines in Fig. 5B and C). At each position above the offset, the u time series were similar for most of the recording period but with slightly wider bounds (grey lines in Fig. 5B and C) and this recorded the non-offset pattern displayed by the dashed lines of Fig. 4B. In addition, there were repeated short periods (up to 5 sec long) when velocity varied by two to three times the excursion of the steady ow. At any point in space above the offset, the large excursions were only above or below the upper boundary of steady u but not both. By Prole J, the time-averaged u prole was approaching a complete parabola. All bubbles had disappeared. The tailwater continued to slow downstream.

1353

Turbulent kinetic energy and Reynolds shear

Turbulent kinetic energy, Tke, and Reynolds shear, szx, were calculated from coincident u and w data generated by the UDVP array 2 (Fig. 6). In the xz plane, turbulent kinetic energy per unit mass is dened as: 1 02 2 u w 02 2 where u and w are deviations from the timeaveraged velocity components. Reynolds shear per unit mass is expressed as: Tke szx u0 w 0 3

for incompressible Newtonian uids. Localized viscous stresses within the ow are expressed by velocity uctuations, mathematically described in the turbulent acceleration term of the Navier Stokes equations: the gradient of szx accelerates the local uid. At the jet detachment site, Tke was highest at the upper and lower edges of the jet (the lower edge was the ume oor) because of high shear uctuations (Fig. 4C, Prole (i); Fig. 6). The Tke minimum in the middle of the detached jet was approximately 20% of that at the upper jet edge. Tke was low in the downstream half of the roller and the reverse ow strip. At Prole (ii), the detached jet had Tke maxima at the upper and lower edges and a mid-jet minimum, all higher than at jet detachment. A Tke minimum at z $ 005 m underneath the detached jet is associated with the centre of diffuse vortices, shed from

the lower jet edge at consistent locations. Where the pulses were nearest to the ume oor, two distinct regions were recorded, above and below the height at which bubbles were seen [Prole (iii)]. Below the pulses [below z = 014 m in Prole (iii)], Tke was virtually zero. Above z = 014 m, Tke rapidly increased, peaking at the height of the favoured path of bubble sets within the pulses (z = 030 m). The tailwater downstream of the pulses exhibits a smooth Tke curve, peaking at the height of maximum shear which was below the break in the velocity prole [Prole (iv)]. A mid-jet minimum in Reynolds shear, szx coincided with the Tke minimum [cf. Proles (i); Fig. 4C]. Rapidly accelerating sheared ow at the upper jet edge is associated with a strong positive vertical gradient in szx. An abrupt decrease in stress occurred between sheared layers and the upper ow region and this is attributed to the presence of the upwardtravelling stream of vortices. The ow above z $ 0180 m did not deviate abruptly in speed or direction and exhibited low values of Reynolds shear. The detached jet prole at x = 610 m highlights the jet retardation in the negative szx/z around z = 0150 m. The shorter negative gradient below (around z = 005 m) marked the jet-down position. At this streamwise location, the jet-up state produced a more diffuse pattern in u, spanning more of the ow depth than the jet-down state. Strong positive values of szx/z existed at the upper and lower jet edges, despite diminished direct acceleration by the retarding jet to the ow surrounding it. The region of diffuse vorticity below the detached jet had comparatively low values of szx (Fig. 4C). The szx peak, lower in the ow, marked the mean vortex centre point. In this case, szx/z was negative above the vortex centre, so the upper part of the vortex circumference is detected to have rotated downstream, while the lower part rotated upstream. The reverse ow strip had a low value of szx, consistent with the zero-shear stress boundary condition which exists on the free surface ow within this ever-present part of the ow structure and consistent with the low Tke values. Two modes of the ow were detected at x = 750 m [Prole (iii); Fig. 4C]; a low-szx tailwater, below a region where szx had increased three-fold. The transition between these regions was at the base of the observed favoured path of bubble sets, between z = 0120 and 0180 m. The distribution of szx/z within the transition was positive: a nett acceleration of the immediately surrounding water by the lowest reaching pulses

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1354

R. G. Macdonald et al.

Fig. 6. Plots of the instantaneous deviation from mean streamwise velocity, u, against the corresponding instantaneous deviation from mean vertical velocity component, w, for Run 1 at Proles (i) to (iv) (Fig. 4C). u > 0 denotes downstream and w > 0 denotes away from the ume oor. These plots give a visual impression of the variation in turbulence structure in space.

which were still travelling downstream. Above z = 0180 m in Prole (iii), a strong negative gradient in szx was observed because the uid had comparable values of u and w. Because of their buoyancy, the w of bubbles may be greater than that of water. The tailwater further downstream [Prole (iv)] exhibited low negative shear of slowing steady ow except for the marked peak around z = 018 m, consistent with the data for Tke and velocity.

Turbulent velocity uctuations

Instantaneous u and w of necessity have a mean of zero and the data scatter indicates the predominance of horizontal or vertical uctuations, shedding light on the time-transient ow character (Fig. 6). u and w both generally decreased as the ow slowed downstream. Jet ow generated high instantaneous velocity uctuations. The jet thickness was 012 to 023 m as it rose from the

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars ume oor [Proles (i) and (ii); Fig. 6] (ds was 009 m). The detached-jet centre was less spread in u and w than the jet-up edges (Fig. 6). In addition, there was less spread in u and w at the lower edge of the jet while it was in the jet-down state and still less spread between the two states. x = 580 m was the location of the widest spread in u and w and was along the path on which vortices passed downstream. w in the recirculating roller (at z = 0186 m) was twice as widely spread as u. Close to the bed at x = 580 m, where the wall jet was thinnest, a few outlying values negative in u and positive in w indicated slowing, rising bursts when the jet began to detach from the ume oor. Downstream of jet detachment, turbulence was reduced at the ume oor but increased within the water column. Transport of suspended sediment and bedload would be promoted by turbulence because of the jet edges at jet-up and jet-down conditions. The slowly rotating vortices which spanned the ow column below the detached jet at x = 610 m accord with a small spread in u and w compared with the surrounding uid. Throughout the prole at x = 840 m and the lower region at x = 750 m, u and w were clustered particularly close to the origin. u and w were more spread above z = 0186 m and less spread below this, in accordance with the minimum height that pulses reached down to. At z = 0246 m, data were more spread in u than in w because the pulses which travel deeper also travel further downstream before rising towards the free surface. The additional turbulence which the pulses caused within the downstream water column could cause suspended sediment to travel further downstream before settling, than if the vortices had not passed to downstream of the roller.

1355

DISCUSSION OF SEDIMENT-FREE HYDRAULIC JUMP

Flow patterns
Despite the high Froude number in most published experimental studies, there are similarities in the Run 1 ow pattern. The shape of the hydraulic jump is consistent with the Murzyn et al. (2005) photograph for Frs = 30, as well as the sketch by Chanson & Brattberg (2000) of hydraulic jumps of Frs 633 and 848 and was only marginally steeper than the free surface proles measured by Sarma & Newnham (1973)

for Frs 110 to 379 with fully developed inow downstream of an undershot sluice gate. Oscillation of the streamwise position of the toe of the hydraulic jump, on or over the supercritical jet, is thought to result from competition between the upstream pressure gradient force from the thicker subcritical ow to the thinner supercritical ow and the viscous drag on the roller because of the incident wall jet. The jet slowed exponentially with distance downstream from the toe of the hydraulic jump, unlike that observed by Chanson & Brattberg (2000) and this may be due to their thin inow from an undershot sluice gate. It simultaneously thinned because the stream of upward and downward travelling vortices widened downstream. Contrary to the suggestion of Chanson & Brattberg (2000), the turbulent shear layer between the submerged wall jet and the roller did not display triangular u proles with a velocity peak at the layer interface. Here, u proles approached triangular further downstream within the detached jet (Fig. 4B; Proles F and G). In Proles A to E (Fig. 4B) above the jet, there is a velocity prole of the ow, that is increased by the downstream rotation of the lower half of the vortices and decreased above the vortex centre (making possibly negative) by the upstream rotation of the upper half of the vortices, for the time that a vortex exists at a particular point in space. For the time-averaged velocity to show a negative peak, a constant stream of vortices would need to pass along a consistent path, as was observed in this particular hydraulic jump. The Chanson & Brattberg (2000) timeaveraged u proles also showed a decrease in velocity above an increase in velocity compared with the general trend above the jet. This decrease was not prominent in their Frs = 848 case but was prominent for the Frs = 633 case where u approached becoming negative. This observation suggests that the presence of vortices could be the sole cause of the negative velocity peak in natural situations of low Frs. The associated increase in velocity over the lower half of the vortex courses in Run 1 was magnied and the decrease over the lower half decreased, where the jet had slowed (by up to 85%) and ow immediately above the submerged wall jet may have more downstream momentum than that within the jet. Khan & Johnston (2000) noted a streamwise velocity decit within vortex cores and this may contribute to the negative velocity within the roller.

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1356

R. G. Macdonald et al.

Detachment of the jet

Although jet separation from the bed has been described in sites with positive lee-side ramps (McCorquodale & Khalifa, 1980; Mossa & Tolve, 1998; Balachandar et al., 2000; Liu et al., 2004; Yuksel et al., 2004) and at sharp negative steps such as at a headcut (Bennett & Casali, 2001; Alonso et al., 2002), wall jets entering hydraulic jumps above at surfaces have, until recently, been assumed to expand into the tailwater without any separation from the oor. Run 1 conrms that wall jets in hydraulic jumps do separate from a solid oor rather than simply diffuse into a parabolic tailwater (cf. evidence in the gures of Alhamid, 2004; Lennon & Hill, 2006; Dey & Sarkar, 2007). There has been little published discussion of the cause of separation or of the implications of coherent vortices shed from the lower edge of a detached jet and no discussion of the implications for sediment movement. From upstream to downstream through the hydraulic jump, the viscous boundary layer (of sub-millimetre thickness) moves from a region of relatively low pressure to a region of relatively high pressure and it is probable that, in Run 1, the viscous boundary layer in high Re conditions separated from the ume oor. If the boundary layer did not separate, then the ow throughout the tailwater would necessarily have been laminar. Local separation of the viscous boundary layer from the boundary can induce a large-scale change of the direction of the jet as observed in Run 1.

Vortex dynamics Distribution of vorticity With boundary separation from the ume oor, in general there will be a migration of vorticity away from the ume oor and this concentration of vorticity became noticeable as the vortices below the detached jet. The comparatively lower vorticity below the detached jet caused lower-magnitude turbulent uctuations than were due to the vorticity within the roller. Vorticity which enters at the toe, can make its way into the shear layer between the submerged wall jet and the roller by convection as the free surface of the roller overturns and falls onto the supercritical free surface (Hornung et al., 1995). Reconnection of the surface at the streamwiseoscillating toe and spilling splashes occurred repeatedly throughout Run 1.

Vortex motion from the roller Vortex escape from the roller is thought to cause the periodic pulses observed within the upper part of the tailwater. When a pulse was not present, the upper jet edge approached the free water surface (the jet up state), isolating the roller from the tailwater; vortices circulated within the roller and became diffuse adjacent to the reverse ow strip. When a pulse occurred (caused by vortex escape from the roller), the diffuse tip of the jet was lower in the ow and the vortices moved downstream into the tailwater causing the change in mean velocity proles (the offsets to prole patterns in Fig. 4B). There are three possible mechanisms for periodic release of vortices from the roller: (i) tightly rolled vortices may have had strong-enough circulation and streamwise velocity to escape the roller, pushing the upper jet edge downward as they travelled downstream, close to the free surface. Those vortices which maintained outlying positive values in u (u ) 0; Fig. 6) may escape the roller. (ii) The detached jet tip, a short distance upstream of the tailwater, could move back and forth along the path of its core of highest velocity, hypothetically xed in space. As the tip moved upstream and downstream along the path, the vertical distance, H, between the tip and the free surface uctuated and H changed around its time-averaged position at 075 of the total water depth. Vortices were more or less likely to pass to downstream of the jet, depending on the magnitude of H. (iii) At splash condition (Fig. 2A), a free surface peak exists over the roller and, with it, an additional short-lived downstream pressure gradient. It is possible that this pressure gradient forced downward the downstream parts of the detached jet from its usual jet-up state to the jetdown state where its path was lower within the ow. In doing this, a vortex may have been let out of the roller because, in the jet-down state, H would be increased without the path of the jet being shortened. Subsequent upstream spill of the hydraulic jump collapses the free surface peak and so allows the jet to return to the jet-up position. The rst mechanism is least likely to release the vortices which caused the offset to tailwater ow because no vortex was observed to leave the roller as a result of its own force, despite the fact that roller vortices were tightly rolled compared with those observed by Long et al. (1991), were always in motion and did not coalesce. The second mechanism could be effected during the repeated

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars short periods when the supercritical wall jet was slower than the lower bound of the steady ow (Fig. 5A), so shortening the path to the jet tip and increasing H. These periods could be due to the ume pumps (although there is no regular period) or due to the supercritical inow being only partially developed. The third mechanism is also considered likely to produce some of the periodic releases observed: if the constant air intake at the toe and unsteady roller free surface exceeds the bubble ux out of the roller free surface, the volume of the rollers increases; the level of the roller free surface may be raised periodically above the level of the tailwater free surface. The rapid up-ume spill of the splash would rapidly allow the jet to return to jet-up, bound the roller downstream extent, and the cycle restart. The sub-second duration that splashes took to appear and then collapse compares with the durations that the jet was in the jet-down state. The mean frequency of the jet moving into the jet-down state does not account for pulses every 5 sec. The slowing to the supercritical wall jet is much more frequent and the three mechanisms may all operate, resulting in a complex periodicity to the ow within the hydraulic jump.
A

1357

RUN 2: HYDRAULIC JUMP WITH SAND FLUX Run 1 demonstrated the complex internal ow patterns and Run 2 was designed to test: (i) how sediment addition to the ow inuenced the ow behaviour; and (ii) how the ow behaviour created deposits. Starting with ow conditions identical to Run 1, the ow was monitored as sand was added. Upon cessation of the sediment input and ow, the resulting deposits were examined in detail.

Comparison of hydraulic jump behaviour with and without sediment Flow pattern In Run 2, the hydraulic jump moved 272 m upstream in response to sand accumulating downstream of it. The position of the toe of the hydraulic jump in Run 2 oscillated to the same degree as the sediment-free jump. The same internal uid processes were observed, in the same geometry as Run 1: The jet detached from the ume oor with a similar angle, from the same streamwise position relative to the toe of the hydraulic jump and rose to a similar ow depth

Fig. 7. Photographs of Run 2 showing: (A) the roller and the jet beneath, taken 7 min after the sediment addition started. Flow structure was similar to Run 1 (compare Fig. 2). (B) The rst hydraulic-jump unit bar which was prograding over the ne sand sheet, at 13 min. (C) The slip face of the youngest hydraulicjump unit bar prograding downstream over the ne sand sheet at 24 min, after upstream progradation of the upstream part of the bar has ceased. Some slip faces contained distinctly ner sediment than others. Silver bolts are 015 m apart.

downstream; vortices were generated at locations consistent with Run 1, with a similar scale and frequency and passed downstream along similar paths. Fewer bubbles were observed in Run 2 than in Run 1 and none observably persisted downstream of the roller (Fig. 7A and B), making ow components (e.g. vortices) more difcult to see.

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1358

R. G. Macdonald et al.

Fig. 8. Proles of streamwise velocity component, u for Run 2, measured by UDVP and averaged over measuring intervals of 600 sec. The UDVP array was xed at x = 610 m throughout the run. The proles are drawn in positions relative to the contemporary hydraulic jump (compare line showing lowest free surface position with the same line in Fig. 4A and B). Proles Q to S are further downstream of the hydraulic jump than was measurable in Run 1.

Streamwise velocity proles were measured at x = 6100 m at eight time intervals, representing distances increasingly downstream of the hydraulic jump. The progressive upstream movement of the hydraulic jump made it possible to measure velocity further downstream from the hydraulic jump in Run 2 than in Run 1. Despite the same pump settings, as soon as the sediment input started, the ow in Run 2 was generally slower than that in Run 1. The streamwise velocity eld at the most upstream prole within the hydraulic jump in Run 2 was similar to that at the same site in Run 1 (Prole L in Fig. 8 compared with Prole G in Fig. 4B). However, the velocity peak was 02 m sec)1 compared with 095 m sec)1 during Run 1. Velocity decreased more rapidly above the peak than below it in Run 2 while the opposite was true in Run 1. The near-bed velocity peak (in jet-down conditions) was closer to the bed in Run 2 than in Run 1. At the end of the roller the uprole was offset at z = 026 m (Prole M; Fig. 8), above which u markedly increased before decreasing to the free water surface, as in Run 1. Downstream from this position, the velocity eld in Run 2 differed from that in Run 1. No prole offset was apparent. At the more distal sites, the velocity increased towards the bed and was 015 m sec)1 at 001 m above the bed (the lowest measurable height) (Prole Q; Fig. 8).

Sedimentation All sand within the supercritical ow was transported in suspension. As the sand entered the hydraulic jump, some of it moved from the jet into the roller and some moved into the water below the detached jet (Fig. 7A). Most of the sand remained in the jet as it detached from the bed. The sand distribution downstream of the distinct jet appeared uniform throughout the water column, although suspension concentration patterns were not measured. Sand dropping from slowing and reversing ow in the roller was instantly accelerated by the jet underneath. In the jet-up

state, some sand dropped from the detached jet and through the underlying zone of separation onto the bed. When the ow pattern had switched towards the jet-down state, this sand was pushed downstream as a bedload sheet a few grains thick, to form a mound in the tailwater about 08 m from the jet separation point (at x = 690 m). At the same time, ner sand was dropped from suspension downstream of the mound and formed a thin sheet. While the ne sand sheet thickened, the mound grew and prograded downstream over it (Fig. 9A). A small relatively ne-grained set of upstream-inclined laminae formed on the upstream edge of the mound as it grew upstream and a coarser-grained avalanche set formed on the downstream edge (Fig. 9B). As the bedform grew downstream, the topset aggraded and the avalanche face progressively overlapped the distal ne sand sheet (Figs 7B and 9C). The aggradation of the topset was at least 10 times slower than the progradation of the foreset. The small amount of ne sand falling onto the feature surface was transported over the crest and incorporated with the coarser sand in the lee-side avalanche set and topset (Figs 7C and 9D). When this bed feature was fully developed, the upstream growth slowed, at times becoming imperceptible. The ne sand sheet continued to thicken downstream of the avalanche face and a diachronous boundary formed with the overlying set (Fig. 9E). Sediment accumulation caused the free surface of the tailwater to rise, increasing the hydraulic jump free surface gradient. This change made the hydraulic jump unstable and it responded by moving upstream. The sediment accumulation pattern moved gradually upstream in tandem with the hydraulic jump, forcing positive feedback. The bed feature continued to grow with slow upstream progradation, rapid downstream avalanching and slow topset aggradation. Periodically, a new bed feature formed above or upstream of the previous bed feature and grew

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars

A

1359

in a similar way to the previous feature. As a new bed feature developed, the older feature became inactive and the new downstream-dipping avalanche face prograded and climbed over the top of it (Fig. 10). When the avalanche face height produced a strong-enough lee separation eddy, counter-ow ripples formed and these reached up to one-third of the height of the lee slope before being met by an avalanche of downstream-travelling bedload. If a feature reached the downstream end of the previous one, the avalanche faces amalgamated. At the end of Run 2, a bed feature avalanche face was still prograding downstream and had a proto bed feature forming on its upstream edge.

Deposit architecture
In this experiment six discrete beds developed (Fig. 11), numbered 1 to 6 upwards. Beds 2 to 6 developed sequentially and displayed the same general internal characteristics. The pattern of sedimentation associated with the hydraulic jump had formed distinctive bed features with a characteristic anatomy (Fig. 10), here termed a hydraulic-jump unit bar. In general, the hydraulic-jump unit bar, when fully developed, had the form of an elongated bed with a sharp lower boundary, upstream-dipping upstream bed terminus, at upper surface and steep downstream bed terminus. The cross-stream bed shape cannot be dened, as the ume experiments were effectively two-dimensional. The internal architecture of each hydraulic-jump unit bar was, at the upstream end, a small wedge of gently upstream-dipping ()9 to 7 relative to the lower boundary) convex-up laminae with grain-size smaller than the bulk composition of sediment from the crates; this graded downstream, through a small volume of massive deposits, into a downstream-dipping cross-set with coarser grain-size. This foreset passed downstream into a ner-grained foreset, which approached the same grain-size distribution as the bulk sediment load. The brink of the foreset was the highest point of each hydraulic-jump unit bar during Run 2. Above both foreset types, low-angle laminae formed a topset, which locally reached thicknesses similar to the underlying foreset. The oldest (lowest) foreset was underlain by a massive ne layer. Foresets that overrode the oldest foreset were underlain by the ne unit at the distal end. Each hydraulic-jump unit bar had a bidirectional cross-set pattern with greater development of the downstream set(s).

Fig. 9. Model of hydraulic jump unit bar formation and growth. (A) Initial mounding on the non-erodable bed and ne sand falling from suspension, forming a sheet downstream of it. (B) Accretion of relatively ne sand on the upstream side of the newly formed hydraulic-jump unit bar and formation of a coarser grained avalanche face on the downstream side. The massive ne sand sheet continues to thicken to form an elongate wedge. (C) The downstream avalanche face progrades over the massive sand wedge. (D) Reduction in the rate of upstream growth corresponds with more ne sand reaching the avalanche face. As the slip face moves further from the hydraulic jump, ne sediment settling from suspension (previously falling to the massive ne-grained wedge) is incorporated in the foresets. The topsets aggrade as the foreset progrades and the unit appears to climb. (E) The pattern formed in (D) continues with topset aggradation and downstream extension. Inset boxes (i) to (iii) in (B), (C) and (D): detail of change in the shape of the front of the unit bar, explaining the blocking (and deposition) of negrained sand and later bypass.

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1360

R. G. Macdonald et al.
(3) Coarser cross sets Leading edge (transverse) (4) Finer cross sets
Finest sand deposited from suspension Sand finer than the bulk load Sand coarser than the bulk load

Brink
(1) Fine upstreamdipping laminae

(2) Massive deposit

Fig. 10. Diagrammatic representation of the architecture of the unit bars formed downstream of the hydraulic jump, with near-steady sand addition, steady pumping rate and non-erodable ume bed. The large (small) grains depict a location where grain-size distribution is negatively (positively) skewed with respect to the bulk load. The thin lines show the general orientation of laminae within the unit bar and the thick line depicts the suspension fallout deposits.

Fig. 11. Pattern of lamination within the deposits of Run 2 represented in a stream-parallel section. The line representing the water surface corresponds with the nal position of the hydraulic jump at the end of the run. The initial position of the toe of the hydraulic jump was at x = 468 m.

The downstream part of the deposit was a coset of avalanche foresets with set boundaries dipping upstream to give a climbing and upward-thinning coset (Fig. 11). The set boundaries are welldened and grade from convex-up to concaveup. Each bed was longer than its predecessor and pinched out to the ume oor further upstream of it. The set boundaries were less distinct in the area of the upstream-dipping lamination. The transition between the upstream and downstream dipping cross-laminae occurred progressively further upstream in successive beds and contact between the two component sets became sharper. This architecture is termed a hydraulic-jump bar complex.

Details of the deposits After ow stopped, the ume was drained and the deposit examined: adjacent to the ume sidewall, in one section cut parallel to ow (03 m from the ume sidewall) and then in nine sections transverse to ow (Figs 12 and 13). The uppermost 01 m and downstream-most 08 m of Bed 5 had collapsed into the discharge tank (Figs 12C and 13D). Otherwise, the deposit was undisturbed. In

the stream-parallel sections the same features were observed, and at comparable streamwise locations, as seen forming through the ume sidewalls (described above). Bed 4, which terminated within the ume length, pinched out at y = 066 m, at x = 755 m and was thickest at y = 050 m. Most laminae were approximately horizontal in y, or dipped at a low angle towards the sidewalls (Fig. 13). Towards their upstream limit, coarse-grained foresets occurred in three packages a few centimetres thick at the left sidewall (y = 0) that pinched out within 035 m across the ume, at the top of Beds 2 (Fig. 13C), 4 and 5 (Fig. 13B). Towards their downstream limit these foresets occurred across the ume width (Bed 3; Fig. 13C). Upstream-dipping lamination was distinct in the proto bed feature (Fig. 12A), which had been developing only for 3 min. Upstream-dipping laminae appeared amalgamated in more developed hydraulic-jump unit bars, e.g. Bed 4 in Fig. 12B. The contact between the proto bed feature and larger bed feature below was not easy to distinguish (Figs 12A and 13A). In the section shown in Fig. 13A, one set of upstream-dipping

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars

1361

Fig. 12. Photographs taken through the side wall of the ume (i.e. stream-parallel sections) and corresponding line drawings of the deposits of Run 2. The location of each photograph is indicated in Fig. 11. The coin used for scale is a British penny and is 20 mm in diameter. The lighting conditions were varied to highlight the lamination; this gives a false impression of variation in sand colour. The line drawings are labelled with bed numbers as in Fig. 11. The red triangles indicate that grain-size is coarsening-upward within the bed. The asterisk (*) was within Bed 5 before the ume was drained.

laminae directly overlay another and the inter-set contact appeared indistinct except adjacent to the ume sidewall. Both sets consisted of sand ner than the bulk load which ned towards the top. A similar indistinct contact occurred between Beds 4 and 5 (Fig. 13B). Counter-ow ripples also deposited upstream-dipping laminae at the bottom of Bed 5 at the downstream end of the test channel and produced distinctly coarse sets (Fig. 13D). The volume of massive sand (the original mound) was observed in Beds 2 to 6 (Fig. 11) and was visually discernable from the set of ne laminae upstream because of the coarse sand

fraction which it contained. It spanned up to 03 m in the streamwise direction and was preserved with a similar shape to the hydraulic-jump unit bar that developed from it, with upstreamdipping upstream bed terminus and low-angle arcuate upper surface (Fig. 12B; right-hand side of Bed 4). The downstream terminus of the massive sand was arcuate and more abrupt than the upstream bed terminus (though not sharp) (Fig. 12B; left-hand side of Bed 5). Immediately downstream of the massive sand, each coarse-grained foreset was thin (approximately half the thickness of the adjacent massive

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1362
A

R. G. Macdonald et al.
B

Bed 6 Bed 5 Bed 6 Bed 5 Bed 4 Bed 3

Bed 6 Bed 5 Bed 4 Bed 3

Bed 5 Bed 3

Bed 2 Bed 2 Bed 1

Fig. 13. Photographs and corresponding interpretative line drawings of sections cut through the deposit perpendicular to mean ow direction (i.e. stream-transverse sections) with a 014 m long pen for scale. The most upstream section is (A) and the most downstream is (D). The positions of these sections are indicated on Fig. 11. (A) and (B) were photographed looking downstream, (C) and (D) are looking upstream. The thick line on each line drawing is the boundary of the deposit and the ume oor and the laminae above are traced from the photographs in close-up view. Note that in section (D), Bed 5 sits in contact with Bed 3. The asterisk (*) was within Bed 5 before the ume was drained.

deposit) and thickened rapidly downstream, with increased dip angle, as the topset above it thinned. Maximum foreset angle occurred within the coarser foreset and the dip decreased down-

stream with grain ning. Each downstream transition from coarse cross-set to cross-set with granulometry approaching the bulk load, was streaky along the streamwise extent of transition.

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars At the downstream termination of Bed 2, a crossset approaching bulk-load grain-size was preserved (Fig. 12C). The hydraulic-jump unit bar which overrode it (Bed 3) exhibited the patchy transition between the two cross-set types at this streamwise location (Fig. 12C), with the coarse cross-set being interspersed with foresets with bulk load granulometry. The youngest hydraulicjump unit bar (Bed 6) was the only one where the downstream-most metre of the foreset contained no patches of coarser laminae, so Bed 6 displayed the complete transition to bulk-load cross-set. Topsets abruptly overlaid the foresets (Bed 6; Fig. 13B) and generally thinned downstream. Specically, Bed 6 is at a more downstream position than the directly underlying bed (being incepted further upstream of the bed below) and has a thinner and ner-grained topset, thinning downstream (Fig. 12B). The topset of Bed 2 pinched out downstream, upstream of the downstream terminus of the bed (Fig. 12C). The contact with Bed 3 above is not erosional. Many of the individual topset laminae could be traced for orders of metres in the streamwise direction, particularly in the more upstream portion of each bed above the coarser-grained cross-set. The ne sand wedge (Bed 1) gradually thickened downstream from its upstream limit at x = 670 m, to 018 m at the end of the test channel. Containing sediment much ner than the bulk load, it made a distinct contact with the beds which overlay it at different streamwise locations (Fig. 12C). However, this contact was not distinct in stream transverse section (Fig. 13D). No length scale can be dened for this unit as it would have continued downstream if the ume had been longer.

1363

DISCUSSION OF SEDIMENTATION UNDER A HYDRAULIC JUMP A distinct sequence of beds was formed in the ume under conditions of constant water recirculation and near-constant sediment transport. There was no obvious correlation between the timing of initiation of new unit bars and sediment input rate variation. Notably, there was no obvious correlation with the short hiatuses in the sediment addition resulting from changing crates and, indeed, two new unit bars formed during loading of the sand from the second crate. A massive basal unit is thought typical of deposition downstream of a hydraulic jump

because of: (i) the ability of the turbulence within the hydraulic jump to maintain sediment in suspension and lift it into the downstreamexpanding ow depth of the hydraulic jump; and (ii) the decline in turbulent regime downstream of the hydraulic jump where fallout from suspension takes place. The rapid decrease in u above the mid-detached jet maximum within the roller and associated decreased vorticity production compared with Run 1 would lead to decreased vorticity (hence turbulence) passing downstream into the tailwater. This effect provided a hypothetical negative feedback between suspended sediment concentration and the preferential length scale of the massive basal unit. This length scale is also controlled by the segregation of bedload below the detached jet and the grain-size distribution (hence settling velocity) which remained in suspension. In Run 2, the lower jet edge in jet-up conditions was more distinct than that in Run 1 and the velocity peak in jet-down conditions was closer to the ume oor; both these promoted bedload transport underneath the jet and a more distal inception of the initial sediment mound than was anticipated from observations of Run 1. The shape and length scale were different from the massive basal unit (which would form in the absence of a bedload feature). Despite the uniform streamow in the tailwater (affected in an upper region by vortex release from the roller), at no time were downstream-migrating bedforms superimposed on any hydraulic-jump unit bar surface. As each hydraulic-jump unit bar developed, sand was deposited from suspension within the prograding system, particularly at the downstream part of the hydraulic-jump unit bar and this may be the cause of rare mud drapes between coarse laminae described by Massari (1996). The bulkload foreset laminae were fully reverse-graded despite short avalanche face heights (up to 011 m; Fig. 12B and C) which is comparable with the observations of ancient deposits described by Massari (1996). Breda et al. (2007) described 03 m thick cross-beds within Pliocene Gilbert delta deposits, with characteristics similar to those produced in Run 2. A foreset brink climbing a slope greater than 57 raises the deposit height more than the topset aggradation, which was an order of magnitude slower than foresets prograded in the slopewise direction. The upstream movement of the hydraulic jump was affected more by the brink while Bed 2 was growing and more by the topset while Beds 1 and 3 to 6 were growing. After Bed 2

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1364

R. G. Macdonald et al. enced by greater velocities and, consequently, move over the mound. The velocity gradient blocking effect segregates the ner bedload on the upstream face of the mound from coarser bedload downstream. As the upstream edge of the deposit progrades upstream, its dip decreases (Fig. 9C) and the velocity gradient immediately upstream steepens. As the dip decreased, increasingly ner-grained sand is forced over the mound by the ow. The accumulation of relatively ne sand decreases bed roughness and coarse grains more readily traverse the smooth slope in a positive feedback. The blocking effect decreases with time until all bedload is forced across the feature and upstream growth of the unit bar stops (Fig. 9D and E). This effect contributes to the downstream-ning of the foreset (described above) as the ne bedload component is transported to and deposited on the lee. A more signicant cause of downstream foresetning is that as the unit bar extends downstream, more ne sand falls onto it from suspension and is incorporated in the topset and foreset.

was deposited, each new bed made the hydraulicjump bar complex surface shallower and topset aggradation had increasingly more effect than brink progradation on the upstream movement of the hydraulic jump. Local increases in the deposit surface height increased tailwater levels irrespective of their proximity to the hydraulic jump, shown to be true for a change at xed x by Vide et al. (1993). Historically, backset beds are the only facies that have been regularly interpreted as resulting from hydraulic jumps (e.g. Massari & Parea, 1990; Nemec, 1990; Breda et al., 2007). In Run 2, the deposits did not have architecture similar to classic backset beds described from the rock record. The ner-grained proximal part of the hydraulic-jump unit bar did have low-angle upstream-dipping cross-bedding and the set boundaries between the unit bars dipped upstream at a low angle. The absence of relatively steep upstream-dipping surfaces, as described from the rock record, may relate to the absence of bed erosion in a ume; if the bed were erodable, scour upstream of the hydraulic jump would increase the amplitude of bed topography and the deposits would form above a scoured surface (e.g. Massari, 1996). The hydraulic-jump unit bar anatomy and the general architecture of the hydraulic-jump bar complex, produced in Run 2, is considered characteristic of deposit architecture under a hydraulic jump over a noneroding surface.

CONCLUSIONS 1 The ow pattern within a hydraulic jump controls the nature of the resulting deposits and when sediment was added to the system, a deposit developed downstream of the hydraulic jump causing the tailwater to rise and the hydraulic jump to migrate upstream. 2 Most of the ne-grained sediment was deposited from suspension in the slowing tailwater to form a sheet, thickening downstream. In the absence of a bedload, this massive basal unit would be the only depositional record of the hydraulic jump. 3 Coarser sediment dropped to the bed from the ow structure within the hydraulic jump and was pushed downstream to a point where it accumulated as proto bed features which developed into hydraulic-jump unit bars with characteristic streamwise anatomy: passing downstream, a relatively ne-grained wedge of upstream-dipping laminae, a small volume of massive sand, a coarse-grained and a ner-grained foreset. A topset overlaid both types of foreset. The unit bars may progressively override a massive ne-grained basal wedge. 4 The growing hydraulic-jump unit bar inuences the velocity prole of ow approaching it. Coarser grains within the bedload protruded into higher velocity ow on the upstream face and

Grain-size segregation in a hydraulic-jump unit bar

The bedload initially mounds on the ume oor at a position within the tailwater and this mound has sharp upstream and downstream edges (Fig. 9A). The mound builds slowly compared with the ow velocity which encounters the mound as if it were stationary (Needham & Hey, 1991). The presence of the mound was seen to inuence ow velocity in the lower region of the tailwater close to the mound. Well upstream of the mound, the velocity increases rapidly with distance above the ume oor, whereas immediately upstream of the mound the rate of velocity increase above the bed is less, up to the height of the mound. As the bedload approaches the upstream edge of the mound, it slows. Sand at the ner end of the bedload grain-size distribution comes to a halt on the ume oor at the upstream face (Fig. 9B). The coarser grains protrude higher into the ow where they are inu-

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars were washed downstream. Finer grains were blocked at the upstream edge. This blocking effect segregated the ner bedload to form the set of upstream-dipping laminae and, while this was developing, the foreset was coarser than the bulk sediment load. 5 With near-steady ow and only minor uctuations in sediment ux, a series of hydraulicjump unit bars developed, each above and upstream of the previous one and all above the massive basal unit, to form a fanning coset architecture termed a hydraulic-jump bar complex. 6 In conditions where a hydraulic jump forms over a non-eroding bed, steep backsets (as classically attributed to hydraulic jumps when observed in the rock record) are not formed. Rather, gently dipping upstream laminae form the stoss side of the hydraulic-jump unit bar and upstream-dipping set boundaries occur between unit bars in hydraulic-jump bar complexes.

1365

MODEL OF DEPOSITION UNDER A HYDRAULIC JUMP A ve-stage model describes the inception and growth of a hydraulic-jump unit bar (Fig. 10). Initially, there is no sediment in the system. Sand added to the system is all transported in the supercritical ow and into the hydraulic jump. Some of this sand drops to the bed underneath the detached jet at the standard jet-up condition. This bedload is periodically pushed downstream into the slowing tailwater when the jet is deected downward to sit closer to the ume oor in the jet-down state. A single isolated feature (the proto bed feature) begins to build where bedload comes to rest. When the contact of the upper surface and lee slope of this proto feature becomes sufciently abrupt and its height sufcient for a lee slope separation eddy to form, this is when the transition from proto bed feature to non-equilibrium hydraulic-jump unit bar is made. Finer sand falls out of suspension downstream of the proto bed feature to form a wedge (Fig. 9B) with a preferential length scale at which slowing ow, decreasing in turbulence and decreasing suspended sediment concentration combine most effectively (i.e. the length scale is dependent on settling velocity). The non-equilibrium hydraulicjump unit bar grows upstream and downstream over the wedge. The established deposit slows the ow so that ner grains are preferentially trapped at the leading edge because of a boundary

layer-blocking effect [Fig. 9 insets (i) to (iii)]. Downstream progradation continues separately. The upstream face of the hydraulic-jump unit bar becomes less steep, reducing the blocking effect and permitting increasingly ner bedload to traverse the feature. Pulses of ne sand reach the lee avalanche face; a streaky characteristic appears within the coarse foreset (Fig. 9D). Once the leading edge gradient becomes sufciently gentle, all bedload traverses the feature and upstream growth of the unit bar stops. All further bedload traverses the equilibrium hydraulic-jump unit bar to the foreset. As the foreset grows, it receives more sediment which falls out from suspension. Subsequently, the grain-size distribution deposited on the lee avalanche face approaches that of the bulk load. The topset aggrades as the lee side avalanche face progrades and the elevation of the brink point increases (the set climbs; Fig. 9D and E). The associated rise in the tailwater free surface continues to push the jump upstream under a pressure gradient and a new unit bar may be initiated further upstream. The association of a resultant climbing and upward-thinning coset with the underlying massive basal wedge is dened here as one unit: a hydraulic-jump bar complex.

ACKNOWLEDGEMENTS Thanks to Trevor Panter and Gareth Flowerdew for help with equipment construction, Rob Utting for shovelling the sand and James Hodson for assistance with monitoring. Robert Macdonald is in receipt of NERC studentship NER/S/A/2006/ 14111. Thanks also to reviewers Suzanne Leclair and Rick Cheel for positive and stimulating comment on an earlier version of this paper.

NOMENCLATURE ds dx=n D50 Frs Frx=n g H Flow depth upstream of the hydraulic jump Flow depth at streamwise location x = n m The ftieth percentile of the grain-size distribution Froude number upstream of the hydraulic jump Froude number at streamwise location x=nm Acceleration due to gravity Vertical thickness of ow above the detached jet

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

1366 Ls

R. G. Macdonald et al.
Chanson, H. and Brattberg, T. (2000) Experimental study of the air-water shear ow in a hydraulic jump. Int. J. Multiphase Flow, 26, 583607. Dey, S. and Sarkar, A. (2007) Effect of upward seepage on scour and ow downstream of an apron due to submerged jets. J. Hydraul. Eng., 133, 5969. Fralick, P. (1999) Paleohydraulics of chute-and-pool structures in a Paleoproterozoic uvial sandstone. Sed. Geol., 125, 129134. Gorrell, G. and Shaw, J. (1991) Deposition in an esker, bead and fan complex, Lanark, Ontario, Canada. Sed. Geol., 72, 285314. Grant, G.E. (1997) Critical ow constrains ow hydraulics in mobile-bed streams: A new hypothesis. Water Resour. Res., 33, 349358. Hornung, H.G., Willert, C. and Turner, S. (1995) The ow eld downstream of a hydraulic jump. J. Fluid Mech., 287, 299316. Hornung, J.J., Asprion, U. and Winsemann, J. (2007) Jetefux deposits of a subaqueous ice-contact fan, glacial Lake Rinteln, northwestern Germany. Sed. Geol., 193, 167192. Jarrett, R.D. (1984) Hydraulics of high-gradient streams. Proc. Am. Soc. Civil Eng. J. Hydraul. Eng., 110, 15191539. Jopling, A.V. and Richardson, E.V. (1966) Backset bedding developed in shooting ow in laboratory experiments. J. Sed. Res., 36, 821825. Khan, Z.U. and Johnston, J.P. (2000) On vortex generating jets. Int. J. Heat Fluid Flow, 21, 506511. Kirkgoz, M.S. and Ardiclioglu, M. (1997) Velocity proles of developing and developed open channel ow. J. Hydraul. Eng., 123, 10991105. Lennon, J.M. and Hill, D.F. (2006) Particle image velocity measurements of undular and hydraulic jumps. J. Hydraul. Eng., 132, 12831295. Liu, M., Rajaratnam, N. and Zhu, D.Z. (2004) Turbulence structure of hydraulic jumps of low Froude numbers. J. Hydraul. Eng. ASCE, 130, 511520. Long, D., Rajaratnam, N., Stefer, P.M. and Smy, P.R. (1991) Structure of ow in hydraulic jumps. J. Hydraul. Res., 29, 207218. Massari, F. (1996) Upper-ow-regime stratication types on steep-face, coarse-grained, Gilbert-type progradational wedges (Pleistocene, Southern Italy). J. Sed. Res., 66, 364375. Massari, F. and Parea, G.C. (1990) Wave-dominated Gilberttype gravel deltas in the hinterland of the Gulf of Taranto (Pleistocene, southern Italy). In: Coarse-Grained Deltas (Eds A. Colella and D.B. Prior), Int. Assoc. Sed. Spec. Publ., 10, 311331. McCorquodale, J.A. and Khalifa, M. (1980) Submerged radial hydraulic jump. J. Hydraul. Div. ASCE, 106, 355367. Mossa, M. and Tolve, U. (1998) Flow visualization in bubbly two-phase hydraulic jump. J. Fluid. Eng., 120, 160165. Murzyn, F., Mouaze, D. and Chaplin, J.R. (2005) Optical bre probe measurements of bubbly ow in hydraulic jumps. Int. J. Multiphase Flow, 31, 141154. Needham, D.J. and Hey, R.D. (1991) On nonlinear simple waves in alluvial river ows: a theory for sediment bores. Phil. Trans. R. Soc. London, 334, 2553. Nemec, W. (1990) Aspects of sediment movement on steep delta slopes. In: Coarse-Grained Deltas (Eds A. Colella and D.B. Prior), Int. Assoc. Sedimentol. Spec. Publ., 10, 2973.

Development length for the supercritical parabolic velocity prole l Dynamic viscosity q Fluid density Res Reynolds number upstream of the hydraulic jump Rex=n Reynolds number at streamwise location x=nm Tke Turbulent kinetic energy u Streamwise velocity component u Instantaneous deviation from mean streamwise velocity U Depth-averaged streamwise velocity v Cross-stream velocity component w Velocity component normal to the ume oor w Instantaneous deviation from mean vertical velocity x Streamwise coordinate parallel to the channel axis y Stream transverse coordinate parallel to the ume oor z Coordinate perpendicular to the ume oor ZUmax Value of z at which streamwise velocity maximum occurs

REFERENCES
Alexander, J., Bridge, J.S., Cheel, R.J. and Leclair, S.F. (2001) Bedforms and associated sedimentary structures formed under supercritical water ows over aggrading sand beds. Sedimentology, 48, 133152. Alhamid, A.A. (2004) S-jump characteristics on sloping basins. J. Hydraul. Res., 42, 657662. Alonso, C.V., Bennett, S.J. and Stein, O.R. (2002) Predicting head cut erosion and migration in concentrated ows. Water Resour. Res., 38, 391- 3915. Balachandar, R., Kells, J.A. and Thiessen, R.J. (2000) The effect of tailwater depth on the dynamics of local scour. Can. J. Civil Eng., 27, 138150. Bennett, S.J. and Casali, J. (2001) Effect of initial step height on headcut development in upland concentrated ows. Water Resour. Res., 35, 14751484. Breda, A., Mellere, D. and Massari, F. (2007) Facies and processes in a Gilbert-delta-lled incised valley (Pliocene of Ventimiglia, NW Italy). Sed. Geol., 200, 3155. Cassidy, N.J., Russell, A.J., Marren, P.M., Fay, H., Knudsen, O., Rushmer, E.L. and van Dijk, T.A.G.P. (2003) GPR derived architecture of November 1996 jokulhlaup deposits, Skeidararsandur, Iceland. In: Ground Penetrating Radar in Sediments (Eds C.S. Bristow and H.M. Jol), Spec. Publ. Geol. Soc. London, 211, 153166. Chanson, H. (1999) The Hydraulics of Open Channel Flow. An Introduction. Butterworth-Heinemann, Oxford, pp. 5667. Chanson, H. (2007a) Hydraulic jumps: bubbles and bores. 16th Australasian Fluid Mech. Conf., Crown Plaza, Gold Coast, Australia, pp. 3953. Chanson, H. (2007b) Bubbly ow structure in hydraulic jump. Eur. J. Mech. B/Fluids, 26, 367384.

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367

Dening hydraulic-jump unit bars

Postma, G. and Roep, T.B. (1985) Resedimented conglomerates in the bottomsets of Gilbert-type gravel deltas. J. Sed. Petrol., 55, 874885. Russell, A.J. and Knudsen, O. (2002) The effects of glacieroutburst ood ow dynamics on ice-contact deposits: November 1996 jokulhlaup, Skeidararsandur, Iceland. Spec. Publ. Int. Assoc. Sedimentol., 32, 6783. Russell, H.A.J., Arnott, R.W.C. and Sharpe, D.R. (2003) Evidence for rapid sedimentation in a tunnel channel, Oak Ridges Moraine, southern Ontario, Canada. Sed. Geol., 160, 3355. Sarma, K.V.N. and Newnham, D.A. (1973) Surface proles of hydraulic jump for Froude numbers less than four. Water Power, 25, 139142.

1367

Vide, J.P.M., Dolz, J. and Del Estal, J. (1993) Kinematics of the moving hydraulic jump. J. Hydraul. Res., 31, 171186. Yuksel, Y., Gunal, M., Bostan, T., Cevik, E. and Celikoglu, Y. (2004) The inuence of impinging jets on hydraulic jumps. Proc. Inst. Civil Eng. Water Manage., 157, 6376.

Manuscript received 15 February 2008; revision accepted 27 October 2008

2008 The Authors. Journal compilation 2008 International Association of Sedimentologists, Sedimentology, 56, 13461367