Fine sediment dynamics in unsteady open-channel flow studied with acoustic and optical systems

Continental Shelf Research 46 (2012) 2–15

Contents lists available at SciVerse ScienceDirect

Continental Shelf Research journal homepage: www.elsevier.com/locate/csr

Research papers

Fine sediment dynamics in unsteady open-channel flow studied with acoustic and optical systems Fereshteh Bagherimiyab n, Ulrich Lemmin Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), ENAC, Station 2, CH-1015 Lausanne, Switzerland

a r t i c l e i n f o

a b s t r a c t

Article history: Received 1 February 2011 Received in revised form 13 April 2012 Accepted 20 April 2012 Available online 17 May 2012

In order to simulate fine sediment dynamics over an armored bed in a tidal river, unsteady accelerating, then steady open-channel flow over a movable (but not moving) coarse gravel bed (D50 ¼ 5.5 mm) was studied. A layer of fine sediment (D50 ¼120 mm) was placed on the coarse gravel bed. The thickness of the fine sediment layer on the gravel bed was varied between 4 and 6 mm, but it was found that the thickness of the layer had no effect on the results. Quasi-instantaneous profiles of velocity and sediment concentration were taken simultaneously and co-located. An Acoustic Doppler Velocity Profiler (ADVP) was combined with Particle Tracking Velocimetry (PTV) for suspended sediment particle tracking. Measurements resolved turbulence scales. During the final phase of the accelerating flow range, fine sediment suspension from the bed started in packets and rapidly created a ripple pattern that remained nearly stationary. Thereafter, vortex shedding produced most of the sediment suspension into the water column in the form of events or packets, making suspension intermittent. Simultaneously, sediment particles rolled along the bed following the ripple structure, thus slowly advancing the ripple pattern in the direction of the flow without altering ripple geometry. Fine sediment particles and hydrogen bubbles were used individually or combined as flow tracers in the acoustic measurements. When used individually, hydrogen bubbles provided full depth flow and backscattering information, whereas sediment particles traced only the lower layers of the flow, indicating sediment suspension. When both tracers were combined, hydrogen bubbles could only be distinguished from sediment particles when results at two different acoustic carrier frequencies were compared. The intermittency was observed in the backscattering of the acoustic system. The event structure in fine sediment suspension is seen by the PTV method. PTV velocity vectors varied in speed and orientation, were organized in large coherent packets, mainly in the near bed layers, but also extended well above the bed. The two methods provide complementary information. ADVP measurements allow long time series analysis, whereas most of the spatial details seen in the PTV results cannot be resolved in the ADVP measurements. & 2012 Elsevier Ltd. All rights reserved.

Keywords: Fine sediment suspension Ripple formation Tidal open-channel flow Acoustic Doppler method ADVP (Acoustic Doppler Velocity Profiler) PTV (Particle Tracking Velocimetry)

1. Introduction Suspended sediment transport occurs in many geophysical flows and it directly impacts on physical and biogeochemical processes in the whole water column. In tidal channels, estuaries and rivers, sediment erosion and deposition may lead to channel changes and relate to scour. It is therefore of great importance to understand suspended sediment transport dynamics and turbulent sediment fluxes, because geophysical flows are often turbulent. Most geophysical flows are unsteady and flow acceleration and deceleration during the unsteady phase may lead to the initiation or termination of sediment suspension, generate or

n

Corresponding author. Tel.: þ41 216932378. E-mail addresses: fereshteh.bagheri-meyab@epfl.ch (F. Bagherimiyab), ulrich.lemmin@epfl.ch (U. Lemmin). 0278-4343/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.csr.2012.04.014

modify bed forms, and change bed and channel topography. Initiation of sediment motion due to unsteady turbulent water flows is an important aspect of river and coastal engineering (van Rijn, 2005). In order to characterize the dynamics of these flows, it is essential to measure near bed flow quantities (e.g. turbulence), suspension quantities (e.g. concentration profiles) and bedform characteristics (e.g. shape and dimensions). Therefore, instrumentation is needed that ideally can address all three aspects at the same time (Thorne and Hanes, 2002; Hurther et al., 2011). Acoustic and optical methods have recently emerged which may be well suited, because they can provide these combined measurements with high spatial and temporal resolution and they work with a minimum of flow field disturbance. Among the various aspects of the flow field, the interplay between bottom shear velocity, sediment transport dynamics and bedform development can be considered as a good description of

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

the flow and therefore needs to be first identified and visualized and then quantified by suitable measurement and analysis methods. Although bedload transport in unsteady flow has been the subject of much research, less is known about the suspension of sediment under unsteady flow. Suspension of sediment particles occurs when the local bottom shear stress ru2n exceeds the critical value (Shields, 1936). In unsteady open-channel flows over a smooth wall, Nezu and Nakagawa (1993) found that the log law of the mean velocity is still valid and used it to estimate the friction velocity un and the wall shear stress ru2n as a function of time. The log law was confirmed by Afzalimehr and Anctil (2000) who studied spatially accelerating shear velocity in gravel-bed channels. In oscillatory closed-channel flows, Jensen et al. (1989) and Akhavan et al. (1991) observed that the mean velocity obeyed the log law distribution. Baas and Best (2009) indicate that the log law is still valid under conditions of low sediment concentration. Under steady flow conditions, sediment suspension may be caused by coherent structures (Nezu and Nakagawa, 1993; Hurther and Lemmin, 2001; Cellino and Lemmin, 2004; Nezu, 2005) or secondary currents (Nezu and Nakagawa, 1993; Albayrak and Lemmin, 2011). McLean et al. (1994) measured the turbulence structure over sand dunes and pointed out the importance of coherent eddies in lifting up sediment particles behind the dune crest. Turbulent coherent structures are able to penetrate farther into the outer flow and thus induce return flows that are able to exert greater shear stress as they impact on the bed (Best, 2005). In rivers, the entrainment and transfer of suspended sediment within these turbulent structures is an important process (Shugar et al., 2010). These authors documented suspended sediment transport by coherent structures over dunes in a river and found that the height of suspension above the bed does not extend to the full water depth. Among bedforms in loose bed conditions with fine sediments, ripples and dunes are the most prominent small-scale features in the low Froude number range (Graf, 1998; Julien, 2010). A detailed description of the creation, geometry and movement of these bedforms is given in DuBuat (1786). According to DuBuat, sand grains move up a gentle slope, arrive at the summit and fall down a steep slope whose angle is close to the angle of repose of sand. The current erodes the upstream face and deposits the eroded material on the downstream face. By this process of sand grain movement, the whole bedform advances slowly in the direction of the flow. Shugar et al. (2010) suggest that suspended sediment transport may lead to sediment deposition on the lee slope and in the trough, resulting in more gentle lee slope angles. Ripples occur mainly in sands with diameters o0.6 mm and are steeper and shorter than dunes (Allen, 1968). Their dimensions depend on particle diameter. Equations for equilibrium ripple dimensions are suggested by Baas (1999). Best (2005) indicated that differences in unsteady flow conditions may affect the dimensions of bedforms. In the present study, we will focus on flow conditions during the initial development of ripples. Baas (1994, 1999) pointed out that initial dimensions are much smaller than the equilibrium dimensions. He suggests that the process of ripple development passes from incipient ripples (Stage 1) to straight and sinuous ripples (Stage 2) before reaching equilibrium linguoid ripples in Stage 4. Stage 1 and 2 ripples are nonequilibrium bedforms and these patterns are independent of flow velocity. Venditti et al. (2005) investigated the initial development of ripples and observed that for greater flow strength, bedload transport occurred over the whole bed, first forming chevrons from which regular 2D ripples across the channel were spontaneously generated over the whole channel bed. They termed this process instantaneous bedform initiation. These

3

observations correspond to the Stage 1 and 2 steps of ripple development proposed by Baas. In this study, we will apply acoustic and optical methods providing high spatial and temporal resolution profile data in parallel in order to compare velocity profiles and particle dynamics obtained by the two methods and also consider the advantages of each method. Three aspects related to the determination of the flow field dynamics resulting from unsteady sediment-laden flow will be investigated: 1) ripple development in depth-variable accelerating flow will be briefly explored. We will start with a flat bed with sediment at rest and accelerate the fluid to a quasi-steady flow with fine sediment in suspension and ripple formation; 2) in particular, the situation of low-sediment particle concentration suspended over only part of the water column will be addressed. These conditions occur in rivers and in openchannel flow during the initial phase of ripple formation. Therefore, the dynamics of the suspension are related to the presence of ripples. During this initial ripple formation phase, a relatively low number of particles may be suspended in the water column. This may affect the determination of velocity and sediment particle concentration by acoustic methods. Optical methods used in parallel may help to verify the acoustic measurements in this critical situation and in the correct interpretation of the data analysis and 3) in tidal rivers and the coastal ocean, suspended matter acts as a tracer for acoustic and optical studies. This may include gas bubbles, algae, detritus or sediment particles. It is often difficult in field studies to determine a priori the nature of the scatterers in the flow field. Most often they occur as a mixture of several or all of them. However, the transport of sediment particles is frequently of major interest because of its effect on bed and channel morphology. Therefore, the presence, the concentration and the distribution of individual tracers should be known in order to correctly quantify the contribution by sediment particles. Here, we will study the possibility to identify the nature of tracers with the acoustic measurements under controlled conditions by selectively adding one or two tracers to the flow. First, the measurement techniques used and the experimental procedure will be described. The results will be discussed thereafter.

2. Methodology In order to capture the dynamics of unsteady sedimentladen flow, instrumentation is required that can simultaneously measure hydrodynamics, sediment concentration in the whole water column and bed morphology with sufficient spatial and temporal resolution to resolve turbulent scales. Acoustic methods based on the backscattering of sound are well suited to fulfill these requirements. Acoustic Backscattering Systems (ABS) can capture the Doppler phase angle and the intensity of the backscattered signal from which flow velocity and sediment concentration, respectively, can be obtained. Thorne and Hanes (2002) have summarized the development of ABS techniques and instrumentation that allow extracting this information. At present, velocity and sediment concentration are most often obtained separately (Harris et al., 2003; van der Werf et al., 2007; Thorne et al., 2009). By this approach, sediment flux can only be resolved for scales that are larger than the separation of the instruments.

4

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

Velocities and sediment concentration have to be measured in the same scattering volume in order to resolve sediment fluxes on smaller scales. An Acoustic Doppler Velocity Profiler (ADVP) that is capable of taking co-located measurements with resolution of turbulent scales was developed and has evolved over the past twenty years (Lhermitte and Lemmin, 1994; Rolland and Lemmin, 1997; Hurther and Lemmin, 1998, 2008). ADVPs determine velocities from the backscattered echo produced when an ultrasonic signal is scattered by moving targets. The quality of the velocity information is closely related to the nature and the abundance of the scattering targets and their capability to follow the fluid motion. Quasi-instantaneous velocity profiles resolving turbulence scales in space and time are obtained from the Doppler phase angle (Lhermitte and Lemmin, 1994) by applying the pulse-pair algorithm (Lhermitte and Serafin, 1984). Backscattered intensity of the same signal from which the velocity was extracted can be inverted into particle concentration after calibration (Thorne and Hanes, 2002). Most often, an iterative inversion method (Thorne et al., 1993) or an explicit inversion method (Lee and Hanes, 1995) is used to determine the particle concentration. However, both methods suffer from errors propagating through the profile which may be due to attenuation (Shen and Lemmin, 1998). Attenuation compensation, even in flows with high particle concentration, can be provided by combining back and forward scattered profile signals. Integrating this approach into the existing ADVP, an acoustic particle flux profiler was developed (Shen and Lemmin, 1996, 1999) that determines full depth profiles of co-located 3D velocity and suspended particle concentration information. For field applications, sediment concentration can be determined from backscattering intensity at two (or more) emitted frequencies (Hay and Sheng, 1992; Hurther et al., 2006; Bricault, 2006). Two relatively close frequencies (such as 1.25 and 2 MHz) completely resolve the concentration field of fine particles typically found in benthic boundary layer applications. This frequency range can be handled by a single emitter transducer. Therefore, a two-frequency backscattering intensity profiler can easily be integrated into the existing ADVP (Hurther et al., 2011). Smyth et al. (2002) demonstrated the possibility of co-located velocity and concentration profiling using an ABS. Kostaschuk et al. (2005), Kostaschuk and Villard (1996), and Shugar et al. (2010) used an acoustic Doppler current profiler to obtain co-located velocity and backscattering profiles. In low suspended sediment particle concentrations, the number of particles within the scattering volume may not be sufficient to extract the characteristic mean velocity and backscattering information, unless there are other tracers in sufficient quantity in the water. If no other tracers are present and if the sediment particles are suspended and transported intermittently, particles may only be tracked occasionally through two consecutive pulses to obtain a velocity estimate by the pulse-pair algorithm. Mean value estimates over a whole time series calculated from highly intermittent data may then not represent the actual flow mean velocity. We will investigate this situation in the present study. A low particle number density inside the scattering volume does not allow a calibration to obtain sediment concentration from backscattering intensity. Therefore, in this study no sediment flux estimates can be presented. The location of the bed is extracted from ADVP or particle flux profiler data by the strong backscattering signal from the bed or from the level of zero velocity in the Doppler phase. For coarse bed particles of the size discussed in this paper, it is found that the bed level detected from the velocity profiles is located at about 0.2 D50 of the bed particles which corresponds to the definition of reference bed level in rough flows (Bagherimiyab et al., 2008).

In the present study, measurements with the acoustic particle flux profiler were complemented by optical methods using Particle Tracking Velocimetry (PTV). Prandtl and Tietjens (1929) already used particles to study fluid dynamics in a systematic manner. Later, laser Doppler velocimeters were developed using high particle concentrations. With the availability of fast, highresolution digital cameras, Particle Image Velocimetry (PIV) and PTV became standard measurement methods using lower particle concentrations (Adrian, 2005). By the right choice of particle size and specific density it is assured that tracer particles accurately follow the flow. PTV of singly exposed multiple images tracks a particle through sequential images. PTV works well with low particle seeding and can thus avoid pairing ambiguities. Since the particle number density is low and the particle distribution within a frame is rather inhomogeneous during the initial phase of sediment suspension and ripple formation in accelerating flow, this technique is well suited for the present study. PTV algorithms are generally more accurate than correlation-based PIV algorithms, because they are less affected by displacement gradients (Cowen and Monismith, 1997). We will apply PTV in this study. ADVP and PTV are complementary methods. ADVP systems determine velocity and backscattering intensity from a cloud of particles within a scattering volume. The number of particles in the cloud, their scattering characteristics and their location and displacement within the scattering volume affect the quality of the results. When one or all of these parameters reach their lower limit, ADVPs may no longer produce reliable information. PTV systems determine velocity and particle number from the individual particles tracked through two or more images. They do not have the limit of low particle number density in clouds as the ADVP. However, they have difficulties when the number of particles in the flow becomes too high. In low concentration flows which will be investigated here, PTV should be at its optimum and should therefore allow controlling the reliability of ADVP results. Furthermore, the long-term development of flow and suspension can easily be studied from time series obtained with the ADVP in a single profile location. Details of the instantaneous 2D velocity and particle field within the image taken by the PTV provide instantaneous flow field visualization with high resolution, helping in the interpretation of the data and improving the understanding of the underlying processes (Bagherimiyab and Lemmin, 2010).

3. Experimental set-up The measurements were carried out in a glass-walled openchannel which is 17 m in length and has a rectangular cross section 0.6 m wide and 0.8 m deep. The bottom is covered with a 0.1 m thick gravel layer (D50 ¼ 5.5 mm). The channel is operated in closed circuit mode. Discharge is modified by changing the rotational speed of the pump by computer. A shallow weir at the end of the channel controls the water level. The water level in the channel is measured with four ultrasonic limnimeters spaced along the channel axis. The bed of the channel is horizontal. The ADVP, used in this study, measures full depth 2D quasiinstantaneous velocity profiles in the streamwise and the vertical direction. The instrument works optimally at an acoustic carrier frequency of 1 MHz. The transducers still function correctly at 2 MHz; however, with reduced efficiency. Profiles were sampled with a Pulse Repetition Frequency (PRF) of one kHz. A Number of Pulse Pairs (NPP) equal to 32 was used for the estimation of the Doppler shift and subsequently the 2D velocity components from these data, resulting in a velocity profile sampling frequency of 31.2 Hz. The vertical gate height was set at 3.3 mm.

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

The spatio-temporal resolution of the full profiles (3.3 mm and 0.032 s, respectively) is sufficient to quantitatively estimate turbulence parameters in the production and inertial ranges of the spectral space. The emitter and the receivers of the ADVP are placed in a water-filled housing that is installed above the water surface, and that slightly touches the flow. This is done to avoid measurements in the nearfield of the emitter transducer where acoustic wave fronts are perturbed and turbulence measurements cannot be carried out correctly (Lhermitte and Lemmin, 1994). The ADVP follows the surface in the depth-varying region of the hydrograph (Fig. 1) by a computer-controlled system. ADVP profiling was carried out on the centerline of the channel about 15 m from the entrance where turbulence is well developed. A 1 cm thick layer of the water column near the water surface was omitted from the analysis, because the flow in this layer is slightly perturbed by the instrument. This does not affect the present analysis that is focussed on the near bed layer. The acoustic measurements are complemented by simultaneously taken high-speed videos in the center of the channel at the ADVP location. This allowed visualizing the dynamics of particle suspension during the unsteady flow and the formation of bedforms. A laser lightsheet spanning the water column was produced in the same plane as the acoustic transducer beams in the flow direction in the center of the channel by integrating the optical system into the ADVP housing described above (Fig. 1). Thus, ADVP and optical measurements were taken co-located and they were synchronized. A 30 cm long lightsheet was installed just upstream of the ADVP housing. This allowed following flow development over a longer section of the flow. A camera with 640  480 pixels was used with a frame rate of 80 Hz. Images 9 cm high by 11.5 cm wide were taken through the glass wall of the channel for flow visualization. The diameter of the fine sediment particles that serve as flow tracers is near the upper limit of particle size suitable for tracking the flow dynamics. Particles cover at least four pixels in the camera images. PTV analysis was carried out with a subwindow size of 32 pixels. The threshold for rejecting velocity vectors was set at a difference of 3% with respect to surrounding vectors. The capacity of the recording system of the camera did not allow covering the whole hydrograph. Therefore, the beginning of the image recording period was triggered by the ADVP recording system. The point of the trigger can be selected along the whole hydrograph. In the present experiments, image recording was started in the early part of the accelerating flow in order to also cover the period of sediment suspension and ripple formation. Vertically pointing laser distance measurements of the fine sediment bed were made from a carriage during base flow before and after the hydrograph. The carriage traveled a distance of 1.5 m along the channel axis just upstream of the area where ADVP and PTV measurements were made. The vertical resolution was 0.4 mm. For the analysis, the original nearly flat reference bed level was deducted from the final bed measurements when

5

ripples were present, providing the longitudinal cross section distribution of the ripples with respect to the initial bed. 3.1. Experimental procedure The hydrograph for the experiment consists of 5 parts. The flow is first maintained at the base discharge with h¼hb for 30 s, followed by the rising stage of the unsteady flow (accelerating) where the discharge is linearly increased. Then the peak discharge is kept steady at h¼ hp for 180 s. Thereafter, the discharge is linearly decreased during the falling stage of the unsteady flow (decelerating) to the initial base discharge. Two different accelerating and decelerating times, 30 s and 60 s, were investigated. The discharge, ADVP, and limnimeter data are simultaneously recorded during the hydrograph. Since the experiment is computer controlled, the deviations between individual experimental runs were less than 3%. Table 1 gives the range of the discharge, water depth, mean velocity, and Reynolds number at the base and peak flow of the hydrograph investigated here. In order to investigate the suspension of fine sediment particles, a layer of sand with D50 ¼0.12 mm was spread on top of the coarse bed on a surface area of the channel extending about 1.5 m upstream from the location of the ADVP. For the present study, 4 and 6 mm thick layers were installed. They were carefully levelled by a carriage moving along the channel. This layer was thick enough to fully cover the coarse bed and it smoothed out gravel bed roughness in that area. The range of velocities and the diameter of the fine sand of the present study correspond well to the area of ripples designated in the bedform stability diagram of Southard and Boguchwal (1990). The water in the recirculating installation of the channel was permanently filtered in order to minimize the number of floating particles in the water. In recirculating open-channel systems, gas bubbles are often produced by the pump and the water cascade from the weir at the end of the channel. However, it was noticed that in the discharge range used in the present experiments almost no gas bubbles were entrained into the water during the duration of the experiments. Therefore, under conditions of no fine sediment suspension, the water in the channel was nearly free of flow tracers for acoustic and optical flow measurements. Seven sets of experiments were carried out. The experimental conditions for all experiments are summarized in Table 2. In order Table 1 Range of variations of discharge, water depth and Reynolds number during unsteady flow.

Pump discharge Q (l s  1) Water depth h (cm) Mean velocity u (cms  1) Reynolds number

Base

Peak

10 10 17 1.5  104

35 15 39 5.3  104

Table 2 Experimental conditions.

Fig. 1. Schematics of the ADVP instrument in unsteady flow. During the unsteady ranges, the ADVP follows the change in water depth, rising from the base flow position, shown at the left, to the peak flow position, shown at the right.

E1 E2 E3 E4 E5 E6 E7

ADVP frequency (MHz)

Hydrogen bubbles

Rising time (s)

Fine sediment layer thickness (mm)

1 1 1 1 1 1.66 2

No Yes No No Yes Yes Yes

60 60 60 30 60 60 60

No No 4 6 4 6 6

6

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

Fig. 2. Depth variation Dh/Dhp ¼ (h  hb)/(hp  hb) near the ADVP for 4 ranges of the hydrograph. The duration of the different ranges is indicated by vertical lines.

to investigate the performance of the ADVP under these low particle concentration conditions, the acoustic carrier frequency of the emitter was varied and hydrogen bubbles were used as additional flow tracers, as described in Blanckaert and Lemmin (2006). The hydrogen bubble generator was installed about 2 m upstream of the ADVP/PTV. It had previously been shown (Shen and Lemmin, 1997) that small gas bubbles or clusters of gas bubbles are ideal flow tracers, because they follow the fluid motion with negligible inertial lag. In fully turbulent flow, as in the present case, buoyancy forces of the bubbles are sufficiently small in order not to affect the measurements. Hydrogen bubbles and sediment particles are two tracers with completely different backscattering characteristics. This allows investigating whether one can identify what the contribution of each of the two tracers is to the backscattering and the velocity distribution in the water column. The rising time of the hydrograph was also changed in order to determine the effect of flow acceleration on the formation of ripples. For all experimental conditions, the experiments were repeated five times and combined for the analysis. All ADVP data were de-aliased (Franca and Lemmin, 2006) and de-noised (Blanckaert and Lemmin, 2006) to improve data quality. Fig. 2 shows depth variation Dh/Dhp ¼ (h hb)/(hp hb) near the ADVP for four parts of the hydrograph; the initial base flow was omitted. This curve is representative for all experiments with 60 s long accelerating and decelerating ranges discussed here. Even though the pump discharge is varied linearly in the course of the accelerating and decelerating ranges, water depth changes non-linearly during parts of these periods. When the pump discharge was kept constant at peak flow, water depth still slowly increased and did not reach steady state immediately. The discrepancy between the variation of the discharge and the observed water level over time indicates that flow adjustment over the rough bed takes place along the channel. This behavior is different than depth variation in comparable hydrographs reported in the literature (i.e. Nezu et al., 1997; Song et al., 1994). In the present experiments, the channel slope is shallower (0.07) and the velocities are smaller than those in previous studies.

in the flow will be addressed. Finally, we will study the situation of a mix of different tracers in the flow. In order to establish a reference situation, two experiments were carried out over an armoured gravel bed without a layer of fine sediment. Figs. 3 and 4 show backscattering intensity and mean velocity profiles, respectively, against water depth for different experimental conditions. The profiles were obtained by averaging over the whole peak steady flow range. The ADVP measurements were made at a 1 MHz acoustic carrier frequency. In the first experiment, E1 (Table 2), the water was filtered and no tracers were added. It can be seen that the backscattering intensity is small and uniform over the water depth (Fig. 3), because there are hardly any tracers in the water. As a result, the mean velocity profile for E1 is unrealistically small (Fig. 4), considering the peak discharge (Fig. 2) and mean velocity and water depth in the channel (Table 1) for this experiment. With almost no scatterers in the water, these velocities can be considered as an indication of the instrument noise. This suggests that it is useful to analyze the recorded time series in detail when nothing is known about the tracer concentration in the water, in order to avoid a wrong interpretation of the data. In experiment E2 (Table 2), hydrogen bubbles were added as tracers, as described in Blanckaert and Lemmin (2006). It can be seen in Fig. 3 that backscattering intensity increases. Hydrogen bubbles are good tracers in sufficient quantity and a correct

Fig. 3. Backscattering intensity profiles during steady peak flow. For details on the legend, see Table 2.

4. Results and discussion In this paper, the results of the steady peak flow range of the hydrograph for all experiments are compared. In all experiments, the duration of the steady peak flow range was 180 s. First, the generation of ripples and their effect on the flow field will be investigated. Thereafter, the condition of low tracer concentration

Fig. 4. Mean velocity profiles during steady peak flow. For details on the legend, see Table 2.

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

open-channel velocity profile over the whole water depth is obtained (Fig. 4). A log law layer is found in the lower 20% of the water depth. This profile will therefore be the reference profile for the profiles that will be discussed below under different tracer situations. 4.1. Friction velocity Friction velocity un determines fine sediment dynamics, in particular sediment suspension (Shields, 1936). Therefore, it was calculated by means of the logarithmic mean velocity profile method for experiment E2, using profile data in the inner layer. For this calculation, the time series was subdivided into time slices composed of ten consecutive profiles each. A mean value was calculated for each time slice and an average for each time slice was made over all runs of the experiment. Fig. 5 shows friction velocity for all time slices of the hydrograph against depth variation Dh/Dhp ¼ (h  hb)/(hp  hb), in order to highlight the effect of the acceleration. Friction velocity forms a loop over the whole hydrograph and changes differently in the accelerating and decelerating flow ranges. In the accelerating range, which is of importance for the initiation of sediment suspension and ripple development, the peak of the friction velocity is attained before the maximum of the water level, as was found by Nezu et al. (1997). During the later phase of the accelerating range, the friction velocity reaches the value it will maintain during the following steady peak flow. Even though the discharge increases linearly during the accelerating range, the friction velocity does not, thus indicating that the flow is not in equilibrium during the initial phase of the acceleration. It decreased linearly in the decelerating flow range. For comparable mean velocities in the accelerating and decelerating flow ranges, friction velocities are different. This demonstrates that the shape of the mean velocity profile changes significantly during the hydrograph. These calculations do not take into consideration the contribution of form drag which may result from the presence of ripples when fine sediment is put on the bed.

7

bed. The objective was to provide information on the ripple formation and the ripple pattern along the channel. For this experiment, the whole fine sediment bed was covered with a very thin layer of the same sediment particles that had been painted in black. Starting the hydrograph, it was observed that towards the end of the acceleration range, ripples occurred nearly simultaneously along the whole bed and quickly formed a regular 2D pattern, as seen in the top view of the channel (Fig. 6a). Ripples appear as dark transverse bands, with a ripple crest on the left, downstream side of these bands and a steep slope at the angle of repose on the lee side of the ripple. The crest is about 4 to 5 mm high above the initial bed level in the present image (Fig. 6b). The white areas indicate the trough in the lee side of the crest. Areas circled in red show ‘‘remains’’ of the original flat bed reference level. The observations indicate that crests are rapidly piled up above the initial reference bed level with particles that have been moved horizontally from the present trough area onto this part of the downstream ripple. Troughs are dug below that reference level. Over time, during the steady peak flow of about three minutes, the color of the dark area changes little, indicating that not much of the material from the trough area is deposited onto the crest area. It appears that most of the material from the trough area is suspended in packets higher up in the water column and spread out over a wider area. This is confirmed by an excerpt of the laser distance measurements of the bed given in Fig. 6b. It is observed that a ripple is formed by a combination of erosion and deposition with respect to the initial reference bed level before the experiments. Deposition height and erosion depth are nearly equal. In this figure, the dark bands corresponding to deposition in Fig. 6a have been colored by hand in order to help in the understanding of the pattern seen in Fig. 6a. It should be noted that the cross section of the trough

4.2. Ripple development In the next sequence of experiments, the hydrograph of experiment E2 was repeated, but the seeding with hydrogen bubbles was stopped and a 6 mm thick layer of fine sediment was put on the bed, as described above. The experimental conditions are summarized in Table 2 under E3. Two series of video recordings were carried out in order to investigate sediment suspension and ripple formation. In a first video recording, largescale images were taken from above over a long section of the

Fig. 5. Friction velocity u* distribution during the hydrograph, plotted against Dh/Dhp ¼ (h  hb)/(hp  hb).for the unsteady ranges of 30 s (E4) and 60 s (E2).

Fig. 6. Ripple formation in experiment E3. (a) Large scale image, taken from above the channel. Crests are dark, troughs are light. Red circles: level of initial flat bed, and (b) excerpt of laser distance measurements of the longitudinal bed profile. Dark areas in (a) have been colored in (b). For details, see text. (For interpretation of the reference to color in this figure legend the reader is referred to the web version of this article.)

8

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

zones caused by erosion is typically bigger than that of the adjacent deposition zone. This suggests that deposition is mainly a bed-load process and that suspended sediment transport into the higher parts of the water column mainly originates from the erosion zones. During the experiment, the whole pattern slowly moves in the direction of the flow, but the ripple pattern is stable and its dimensions do not change significantly during this movement. In the second experiment, the PTV camera was zoomed in the vicinity of the bed in order to study the details of the particle motion around the ripple structure. In this recording, images were 1.7 cm high by 2.3 cm wide. A recording with a 40 Hz frame rate was made, covering the period from the beginning of the sediment motion on the bed during the acceleration range to the ripple movement during the steady peak flow range. Multiple exposures of individual particles were produced in order to show particle trajectories. The videos made evident that during the initial saltation of particles, the upper layer of the whole bed began to move (Fig. 7a). Particles may rise about 1 cm above the bed. Saltation above the bed level occurs in packets or events. The ripple started forming 2.5 s after this image (Fig. 7b). Most particles roll along the ripple to the crest and then fall down the steep slope into the

trough. However, other particles do not roll, but form curved trajectories when descending from the crest into the trough in the lee side of the crest. Curved trajectories and backward transport of particles indicate that a vortex is formed in the lee side of the crest. This type of particle motion becomes more evident as the ripple grows (Fig. 7c). Particles now also move higher into the water column and those well above the bed follow straight trajectories across the image. A strong event structure of the particle motion around the crest/trough area was observed. The origin of the particle packets ejected into the higher water column is found on the back of the ripple in an area known as the reattachment point (Best, 2005). Two examples are given in Fig. 7d and e. They demonstrate that particles are lifted off along steep, almost vertical, curved trajectories from a limited area of the bed. Therefore, the vertical velocity component is strong in the near bottom layer. Once again, this process is strongly event structured. The curved trajectories that are repeatedly seen in Fig. 7d and e suggest that vortex motion in the water controls particle motion. Vortex motion in relation to coherent structures has been well documented (Nezu and Nakagawa, 1993). It can be expected that coherent structures affect sediment suspension over bedforms, as was observed by Shugar et al. (2010) and discussed in Bennett and Best (1995). Video recordings were also

Fig. 7. Close-up view of sediment particle trajectories related to ripple formation, obtained by multiple exposure of particles in experiment E3. (a) Saltation during the early phase of the accelerating range, (b) early phase of ripple formation, (c) later phase of ripple formation, (d) and (e) area around the reattachment point. For details, see text.

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

taken at the end of the hydrograph when the flow was back to the base flow. Images showed that ripples remained in place even though the flow had decelerated. Particle motion and saltation continued to occur during decelerating flow, gradually advancing the ripple. It is generally accepted in the literature (Best, 2005; Garcia, 2007) that ripples are formed from some instability on the bed. However, the rapid, near simultaneous ripple development observed in our experiments over the whole surface covered with fine sediment was similar to the observation of instantaneous ripple development reported by Venditti et al. (2005). They correspond to the Stage 1 and 2 steps suggested by Baas (1999). In a separate experiment, the steady peak flow was maintained for periods of hours. The 2D sinuous ripples discussed here eventually turned into 3D linguoid ripples comparable to the Stage 3 and 4 steps of Baas (1999). In the present study, the details of ripple formation seen in the video recording occur on small scales which cannot be resolved with ADVP measurements. In field studies, where these processes may take place on larger scales, ADVPs may be able to capture these scales. However, laboratory studies such as the present one have to be carried out first, in order to guide the field studies in data taking and data interpretation. 4.3. Low particle number density flow In this series of experiments, first the hydrograph of experiment E3 (Table 2) was repeated under the same conditions as for the ripple study above. The PTV technique was used to calculate particle velocity and analyze the dynamics of suspended sediment in order to complement the ADVP measurements. The water in the reservoir was again continuously filtered to assure that only sediment particles acted as tracers. This was verified during base flow, when fine sediment particles did not move. In that case, no velocities could be measured with the ADVP and the resultant profiles closely resembled those of E1 (Figs. 3 and 4). The presence of ripples may affect the suspension pattern of the fine sediment and the resultant particle transport dynamics, as will be discussed below. An example of velocity vectors calculated by PTV is presented in Fig. 8, where two images that were taken at an 80 Hz frame rate during the early phase of the peak flow range of experiment E3 are compared. The time interval between these two images is 0.05 s. As seen in Fig. 8, suspension is nearly uniform in a shallow layer above the bed (about 2 cm high). Particle transport remained strong in this near bottom layer, but is rapidly reduced with distance from the bed. At about half the water depth, which

9

corresponds to the upper edge of the PTV images, only a few particles are occasionally seen. This matches the observations of Shugar et al. (2010) who reported suspended load in only the lower part of the water column. Suspension into the water column above the near bottom layer (lower 2 cm) occurs in events and an increase in intermittency of particle patterns with distance from the bed can be seen. A sequence of events can be identified in Fig. 8a and events are strongest just behind the ripple crest that is on the right side of these images. In Fig. 8b, the final event on the right side in Fig. 8a has grown significantly in size, and the shape of the others has changed. The rapid change in event dynamics observed in the PTV images cannot be recorded with ADVP measurements. The finite width of the ADVP beam (between 7 and 10 mm) averages over much of the fine details seen in the PTV images. Considering that the light sheet was placed in the center of the channel and that PTV video images were taken at a distance of 35 cm from the channel center, it is evident that particle number density in this flow was low. This is also confirmed by the images in Fig. 7, where individual particles in the center of the channel can be followed in their trajectories without interference from other particles, even during suspension events. From the PTV images seen above, it is expected that in the present experiments, ADVP measurements will only give results in the lower half of the water column, because sufficiently high tracer concentrations during the whole duration of the experiment are limited to that depth range. The low and intermittent particle number density in the upper part of the water column does not allow reliable pulseto-pulse coherent tracking in ADVP measurements for all emitted pulses. Optical methods such as PIV will also have difficulties in this layer and PTV is the most suitable technique to determine tracer velocities. However, averaging ADVP or PTV velocities over the whole time series may not produce correct mean flow velocities in these layers due to the gaps caused by tracer particle intermittency. First, ADVP time-averaged mean profiles of backscattering and velocity for E3 were calculated and are presented in Figs. 3 and 4, respectively. Backscattering intensity shows a profile which is different than that of E2. This confirms that particles are only suspended in the lower part of the water column, as was observed in the PTV images in Fig. 8. In the upper part of the profile, backscattering falls off to the background level as in E1. Velocity data are only obtained in the lower part of the water column (Fig. 4). In the upper half of the water depth, where the backscattering situation corresponds to E1, no flow velocity is recorded. A strong velocity gradient near the bed with a maximum of the profile is found at around 0.25 h. Velocities then

Fig. 8. Example of PTV results during the early phase of the steady peak flow range for experiment E3. Arrows indicate particle velocity vectors. The time interval between the two images (a) and (b) is 0.05 s.

10

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

rapidly decrease and fall to zero at about 0.6 h. This indicates that no steady sediment transport is detected by the ADVP above the maximum level. A comparison of the results of E3 in Figs. 3 and 4 shows that backscattering intensity decreases more rapidly than mean particle velocity with distance from the bed. This indicates that suspended sediment transport above a water depth of about 0.2 h may no longer be a reliable tracer for mean flow velocity determination. Therefore, the upper part of the mean velocity profile above z ¼2 cm for E3 in Fig. 4 should be interpreted with caution. From the PTV data and the ripple videos, it has to be expected that particle presence in this layer is strongly intermittent. This affects the mean velocity calculation in this layer. The actual velocities in this layer seen in Fig. 8 are higher, but due to particle intermittency, the mean value over the upper part of the profile shown here is not representative, even though the profile appears smooth. In order to compare the results of ADVP measurements with those obtained with the PTV, mean value calculations were carried out on the PTV data. For this purpose, the PTV image was sliced in the vertical into 3 mm thick slices. This slice thickness is comparable to the height of the gates of the ADVP. In the horizontal direction, the image was sliced into 10 mm wide strips. This width is close to the beam width of the ADVP. Averages were obtained for each of these slices in each strip. In Fig. 9 are shown the sediment concentration profiles in six representative positions (strips) in the horizontal direction of the images. These profiles are the average over the slices from 2000 images, corresponding to a 25 s interval, and are calculated from the averaged particle number density in each of the slices at those positions. The highest concentration is found in the position where ripple crests developed (x ¼10.2 cm; Fig. 8) with a gradient towards the trough. The mean backscattering profile recorded with the ADVP for the same section of the hydrograph is similar to the one at x ¼10.2 cm. However, the ADVP cannot reproduce the spatial details seen in the analysis of the video images. Therefore, a combination of the two methods greatly enhances the understanding of the underlying processes. Fig. 10 shows the mean particle velocity profiles at the same positions as the concentration profiles above. The profile form is comparable to the mean velocity profile measured by the ADVP (Fig. 4) for E3. Velocity profiles are similar in all positions with a slight trend of increase from right (above the ripple) to left (in the trough). It has to be recalled that the same averaging procedure that was used for the ADVP measurements was also applied to the present analysis. Therefore, as indicated above, the velocities

Fig. 10. Particle velocity profiles in six positions around a ripple for experiment E3. For position location, see Fig. 8; for details, see text.

Fig. 11. Contour plots for the steady flow range of experiment E2 obtained from ADVP measurements. (a) Backscattering intensity, and (b) horizontal velocity.

above the maximum of the profiles have to be interpreted with caution. In Fig. 8, velocity vectors in the upper part of the suspension events are the strongest, as opposed to the mean profile values in this layer (Fig. 10).

Fig. 9. Particle concentration profiles in six positions around a ripple for experiment 3. For position location, see Fig. 8; for details, see text.

4.3.1. Effects of sediment suspension intermittency The discrepancy between the mean profiles (Figs. 3 and 4) observed by the ADVP when hydrogen bubbles are used as tracers (E2) and when only suspended sediment is present (E3) may be the result of particle suspension dynamics. This will be investigated in some detail below. Contour plots of backscattering and horizontal velocity were established for experiments E2 and E3. In Fig. 11a, backscattering intensity for E2 is presented. High levels of backscattering

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

11

Fig. 12. Contour plots for the steady peak flow range of experiment E3 obtained from ADVP measurements. (a) Backscattering intensity (log scale), and (b) horizontal velocity.

intensity are found throughout the whole water column, mostly in the intermediate depth in the flow. A time-variability and an organization in packets or events is obvious. The corresponding velocity contour plot for the ADVP data of E2 is shown in Fig. 11b. The velocity contour pattern of E2, organized in packets extending over most of the water column, resembles the one previously observed in steady open-channel flow (Shen and Lemmin, 1999) which was related to the presence of coherent structures (Cellino and Lemmin, 2004). The corresponding contour plots for E3 are presented in Fig. 12. For backscattering (Fig. 12a), maxima of the backscattering intensity are comparable to E2 in Fig. 11a. However, they are limited to the lowest 1 cm of the water column. In general, significant backscattering is limited to this near bottom layer. Once again, a structure of packets or events is found. In the case of E3, high velocities are limited to the inner layer (Fig. 12b) and they are organized in packets or events. In the remaining water column above, strongly intermittent smaller packets of velocity are occasionally observed. Since suspended sediment particles are the only tracers in the water, these contour plots reflect the dynamics of these particles and can explain the differences in the mean profiles of E2 and E3 (Fig. 4). The ADVP backscattering for E3 (Fig. 12a) can be compared to the concentration obtained from PTV data, presented in Fig. 13.

Fig. 13. Contour plot of sediment particle concentration (log scale) for the steady peak flow range of experiment E3 obtained from PTV measurements.

Concentration was estimated from the measured particle number density using the mean particle diameter and density. Structure and distribution are comparable to backscattering intensity in Fig. 12a. Thus, it can be assumed that backscattering intensity

12

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

recorded by the ADVP is mainly caused by the distribution of suspended particles in the water column, since other tracers were filtered out. Further details on the flow structure dynamics of E3 can be observed when series of individual 2D vector profiles are plotted. An example is given in Fig. 14a, where it can be seen that scatterer concentration intermittency is strong in the upper layers, above 3 cm from the bed. However, vectors of horizontal velocities which have the same value throughout the water column are observed in most of the packets. In Fig. 14b, vectors for E3 obtained from the PTV data are plotted. Comparable to the ADVP data, this PTV example also shows a continuous presence of full flow velocity vectors in the lower 2 cm of the profiles and intermittency which strongly increases with distance from the bed. In the upper water column, the fluctuating components of ADVP data, which are obtained by subtracting the mean velocity (Fig.4), are often of the same order of magnitude as those of the total velocity vector. The number of non-zero velocity data

recorded in the ADVP time series was investigated. For E3 nonzero velocities were found in all data points of the time series below the maximum in Fig. 4. Above this level, this number dropped, first slowly, but then strongly above h¼5 cm. This indicates that in the upper layer, mean velocity is small and the shape of the mean velocity profile in Fig. 4 is due to high scatterer concentration intermittency in the upper water column. The strong resemblance of the patterns seen in the ADVP data and the PTV data suggests that both methods track the same particle dynamics. It can therefore be concluded that ADVP data correctly present the flow dynamics of suspended sediment particles in the case of low particle number density.

4.3.2. Higher acceleration In experiment E4, acceleration is increased (Table 2), but all other conditions are kept constant (Table 1). The profiles are compared to E3 (Figs. 3 and 4). Velocities are now detected

Fig. 14. Profile plots of 2D total velocity vectors in instantaneous profiles for experiment E3, profile spacing 1 s. (a) Obtained from ADVP measurements, and (b) obtained from PTV measurements.

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

throughout the water column (Fig. 4). However, the velocities in E4 do not correspond to the discharge, because the profile is significantly different from E2, except for the near bottom layer. The velocity profile of E4 is an indication that due to the higher acceleration, sediment particles are transported higher into the water column when compared to experiment E3. This is confirmed by the backscattering intensity profile (Fig. 3). During steady peak flow, mean flow velocity profiles for experiment E4 followed a logarithmic law in the inner layer. This enabled us to calculate friction velocities from these profiles and then compare them to those obtained from experiment E2 with lower acceleration (Fig. 5). Even though flow velocities reach the same value at peak flow in E2 and E4, friction velocities of the unsteady flow ranges in the 30 s hydrograph (E4) are greater than those in the 60 s one (E2). For E4, friction velocities strongly increase during the last two-thirds of the accelerating range, and the maximum values are significantly higher than steady peak flow values. This shows that mean flow adjustment during the peak flow phase is different in the two hydrographs, and that the observed higher backscattering intensity in the upper layers of the water column for E4 is consistent with the dynamics of the friction velocity. The behavior during the decelerating phase is similar for E2 and E4.

13

From the laser distance measurements of the longitudinal bed profile, histograms of ripple length and height were determined. For E3, a most probable ripple length of about 6.3 cm was obtained and, for E4, the most probable length was 5.4 cm. The corresponding most probable ripple heights above the reference bed level were 3.8 mm for E3 and 4.2 mm for E4. This results in ratios of ripple height to ripple length of 0.06 for E3 and 0.078 for E4. Thus, higher acceleration produces shorter ripples with steeper slopes that may affect the development of fine sediment suspension events and result in the observed sediment transport higher up into the water column. The difference in the two ripple profiles once again points to the importance of suspended sediment concentration intermittency in the upper water column as discussed above.

4.4. Distinction of two tracers with the ADVP A layer of fine sediment is added to the bed in experiment E5 and hydrogen bubble seeding is maintained (Table 2). The backscattering intensity profile in this experiment (Fig. 15a) closely follows that of E2 (Fig. 3; using only hydrogen bubbles). Thus, one cannot identify what the contribution is from each of the two

Fig. 15. Mean profiles during steady peak flow for different acoustic carrier frequencies. (a) backscattering intensity and (b) horizontal velocity. For details on the legend, see Table 2.

14

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

different tracers in this profile. The mean velocity profile (Fig. 15b) is again similar to that of E2 (Fig. 4). In order to investigate whether the contribution of each of the two different tracers can be identified, the above experiment E5, was repeated at different acoustic carrier frequencies. The frequency dependence of backscattering has been demonstrated before (Lhermitte and Lemmin, 1994) with gas bubbles being the dominant scatterers at around 1 MHz and sediment particles at higher frequencies. Therefore, the acoustic carrier frequency was increased in two steps, first to 1.66 MHz and then to 2 MHz. The ADVP hardware can work with frequencies up to 3 MHz. The present transducer set was optimized at 1 MHz. A rapid drop in transducer efficiency with frequency increase was observed and this was too strong to investigate frequencies higher than 2 MHz. At 1.66 MHz, in experiment E6, the backscattering intensity fell to low levels (Fig. 15a), except near the bed where suspended sediment particles were observed in the flow. The corresponding velocity profile (Fig. 15b) approaches that of experiment E3 with sediment only (Fig. 4). When the acoustic carrier frequency is increased to 2 MHz, in experiment E7, the backscattering intensity changes little compared to E6 (Fig. 15a). The velocity profile was reduced in the lower half of the water depth (Fig. 15b). As for E6, the maximum velocity occurred at about 20% of the water depth. The maximum velocity was reduced by about 50% of the maximum flow velocity in E6. Thus, due to loss of transducer efficiency, the ADCP underestimates the velocity profile at 2 MHz, when compared to the expected profile of E3 (Fig. 4). It can be expected that a further increase of the acoustic frequency and an optimized transducer system may help to better suppress backscattering from hydrogen bubbles and thus produce a profile that corresponds to the one obtained only from sediment particles even in the presence of hydrogen bubbles. This set of experiments has shown that the ADVP is a good tool for the detection of tracer characteristics due to its potential to vary the acoustic carrier frequency. It also demonstrates the importance of a good knowledge of the nature of the tracers and their distribution in the water column. If the backscattering intensity profile of E5 had been interpreted as resulting from suspended sediment particles, the contribution of sediment particle concentration and transport would have been grossly overestimated.

5. Conclusion In this study, the onset of fine sediment transport and the development of ripples were investigated by first accelerating the flow and then holding it steady. Ripples formed quickly, within about 3 s after fine sediment particle saltation started. The wavelength of the ripples depended on the rate of acceleration. Faster acceleration produced shorter ripples due to a much stronger friction velocity. Sediment suspension occurred in packets and extended up into the intermediate layers of the water column. The height and the intensity of the upward suspension were related to the ripple dimensions. Shorter and steeper ripples produced stronger and higher suspension. Once established, the ripple pattern migrated slowly in the direction of the flow, mainly by sediment particles that rolled up the ramp of the stoss side of the ripple and dropped over the crest. Ripples did not change in appearance or dimension for the duration of the experiments. No difference was found when the thickness of the fine sediment layer was changed from 4 to 6 mm. Intermittency in sediment suspension was observed in the backscattering intensity of the acoustic system when only sediment particles were used as tracers. Velocities also showed much intermittency which increased with distance from the bed. Due to the low particle number density and the high particle concentration intermittency, backscattering could not be translated into

sediment concentration. Thus, in environments with low and strongly variable particle concentrations, particle flux calculations cannot be carried out. Nevertheless, the ADVP is sensitive enough to capture clean signals for a qualitative analysis of the time history of sediment suspension under these low particle number density conditions and valuable information about the sediment transport dynamics can be obtained. The event structure in fine sediment suspension is also seen by the PTV method. PTV velocity vectors varied in speed and orientation, but were organized in large packets, mainly in the near bed layers, but also extending well above the bed, supporting the concept that coherent structure events may contribute to sediment suspension over ripples. The results indicate that sediment suspension in the final phase of unsteady flow and the following steady flow is controlled by large scale turbulence processes as indicated for steady flow in the literature. The low particle number in the images confirmed the backscattering intensity recorded by the ADVP. The spacing of events of the sediment particle suspension and the resultant intermittency in backscattering and velocity signals increased strongly with distance from the bed. Meaningful timemean velocities cannot be established under these conditions. This study shows that under these flow conditions, only the simultaneous analysis of the details of velocity and backscattering, combining acoustic and optical methods, allows determining the physically correct interpretation. Fine sediment particles and hydrogen bubbles were used individually or combined as flow tracers in the acoustic measurements. When used individually, hydrogen bubbles provided full depth flow and backscattering information, whereas sediment particles traced only the lower layers of the flow, indicating sediment suspension. When both tracers were combined, hydrogen bubbles could only be distinguished from sediment particles, when the acoustic carrier frequency was changed. Therefore, in field studies where different acoustic tracers such as algae, gas bubbles and sediment particles may exist at the same time and where all of them may have a certain size range, the detection and quantification of an individual tracer based on acoustic measurements alone may be difficult. It has been shown that multi-frequency systems may help solve this problem. However, transducer efficiency may limit the range that can be covered and therefore the potential of tracer separation. The two methods provide complementary information, particularly when applied simultaneously. Optical methods helped to verify and to interpret the ADVP data and to visualize the physical processes leading to suspension. New and detailed results made possible by combining the ADVP and imaging techniques give valuable insight into the dynamics of fine sediment suspension and ripple formation initiated by unsteady flow conditions that were previously difficult to obtain. ADVP measurements allow long time series analysis. However, the spatial details seen in the PTV results cannot be resolved in the ADVP measurements. The characteristics of the ripple in the present study are too small to be captured by the ADVP, but they are well traced by the PTV. In field studies, ripple dimensions may be large enough to be investigated by the ADVP. However, laboratory studies such as the present one may help guide field measurement strategies and data interpretation. A combination of acoustic and optical methods can be an ideal approach to study sediment dynamics in unsteady or tidal flow conditions, in particular, ripple formation and suspension in low concentration flow.

Acknowledgments This study is supported by the European Commission (FP6; RII3; Contract no. 022441) HYDRALAB III–SANDS. We sincerely

F. Bagherimiyab, U. Lemmin / Continental Shelf Research 46 (2012) 2–15

thank Navid Borhani for providing the optical system and his assistance. The technical support of Claude Perrinjaquet is greatly appreciated. We thank the reviewers and the guest editor for their comments and suggestions of which helped improve the quality of the paper. References Adrian, R.J., 2005. Twenty years of particle image velocimetry. Exp. Fluids 39, 159–169. Afzalimehr, H., Anctil, F., 2000. Accelerating shear velocity in gravel-bed channels. J. Hydrol. Sci 45, 113–123. Akhavan, R., Kamm, R.D., Shapiro, A.H., 1991. An investigation of transition to turbulence in bounded oscillatory Stokes flows. I: Experiments. J. Fluid Mech. 225, 395–422. Albayrak, I., Lemmin, U., 2011. Secondary currents and corresponding surface velocity patterns in a turbulent open-channel flow over a rough bed. J Hydraul. Eng., ASC 137, 1318–1334. Allen, J.R.L., 1968. Current Ripples: Their Relation to Patterns of Water and Sediment Motion. Elsivier, New York. Baas, J.H., 1994. A flume study on the development and equilibrium morphology of current ripples in fine sand. Sedimentology 41, 185–209. Baas, J.H., 1999. An empirical model for the development and equilibrium morphology of current ripples in fine sand. Sedimentology 46, 123–138. Baas, J.H., Best, J.L., 2009. On the flow of natural clay suspension over smooth and rough beds. ERCOFTAC Bulletin 78, 32–37. Bagherimiyab, F., Lemmin, U., 2010. Aspects of turbulence and fine sediment resuspension in accelerating and decelerating open-channel flow. In: Proceedings of Riverflow 2010, 8–10 September 2010, Braunschweig, Germany. Bagherimiyab, F., Albayrak, I., Lemmin, U., 2008. Bed shear velocity estimates in rough open-channel flow obtained from Acoustic Doppler Velocity Profiler data and direct bed shear stress measurements. In: Proceedings of Riverflow 2008, Izmir, Turkey. Vol 2., 151–159. Bennett, S.J., Best, J.L., 1995. Mean flow and turbulence structure over fixed, twodimensional dunes; implications for sediment transport and bedform stability. Sedimentology 42, 491–513. Best, J.L., 2005. The fluid dynamics of river dunes: A review and some future research directions. J. Geophys. Res. 110, F04S02, http://dx.doi.org/10.1029/ 2004JF000218. Blanckaert, K., Lemmin, U., 2006. Means of noise reduction in acoustic turbulence measurements. J. Hydraul. Res 44, 3–17. Bricault, M., 2006. Re´trodiffusion acoustique par une suspension en milieu turbulent: application a la mesure de concentration pour l’e´tude de processus hydro-se´dimentaires. PhD thesis, Grenoble Institut National Polytechnique, Grenoble. Cellino, M., Lemmin, U., 2004. Influence of coherent flow structures on the dynamics of suspended sediment transport in open-channel flow. J. Hydraul. Eng 130, 1077–1088. Cowen, E.A., Monismith, S.G., 1997. A hybrid particle tracking velocimetry technique. Exp. Fluids 22, 199–211. DuBuat, P., 1786. Principes d’hydraulique. L’imprimerie de monsieur, Paris, France. Franca, M.J., Lemmin, U., 2006. Eliminating velocity aliasing in acoustic Doppler velocity profiler data. Meas. Sci. Technol. 17, 313–322. Garcia, M.H., 2007. Sediment transport and morphodynamics. In: Garcia, M.H. (Ed.), Sedimentation engineering. ASCE Manuals and Reports on Engineering Practice no 110, pp. 21–146. Graf, W.H., 1998. Hydraulics of Sediment Transport. Water resources publications, Highlands Ranch, USA. Harris, K., Butman, B., Traykovski, P., 2003. Winter-time circulation and sediment transport in the Hudson Shelf Valley. Cont. Shelf Res. 23, 801–820. Hay, A.E., Sheng, J., 1992. Vertical profiles of suspended sand concentration and size from multifrequency acoustic back-scatter. J. Geophys. Res. 97, 15661–15677. Hurther, D., Lemmin, U., 1998. A constant beam width transducer for 3D acoustic Doppler profile measurements in open channels. Meas. Sci. Technol 9, 1706–1714. Hurther, D., Lemmin, U., 2001. Equilibrium near-bed concentration of suspended sediment. J. Hydraul. Eng 16, 430–433. Hurther, D., Lemmin, U., 2008. Improved turbulence profiling with field adapted acoustic Doppler velocimeters using a bi-frequency Doppler noise suppression method. J. Atmos. Oceanic Technol 25, 452–463.

15

Hurther, D., Lemmin, U., Bricault, M., 2006. Multistatic acoustic Doppler profilers for fine-scale studies of velocity and particle flux processes. In: Proceedings UAM07, Heraklion, Greece. Hurther, D., Thorne, P.D., Bricault, M., Lemmin, U., Barnoud, J.M., 2011. A multifrequency acoustic concentration and velocity profiler for boundary layer measurements of fine-scale flow and sediment transport processes. Coastal Eng. 58, 594–605. Jensen, B.L., Sumer, B.M., Fredsoe, J., 1989. Turbulent oscillatory boundary layers at high Reynolds numbers. J. Fluid Mech. 206, 265–297. Julien, P.Y., 2010. Erosion and Sedimentation. Cambridge University Press, Cambridge, UK. Kostaschuk, R.A., Villard, P.V., 1996. Flow and sediment dynamics over large subaqueous dunes: Fraser River, Canada. Sedimentology 43, 849–863. Kostaschuk, R.A., Best, J.L., Villard, P., Peakall, J., Franklin, M., 2005. Measuring velocity and sediment transport with an acoustic Doppler profiler. Geomorphology 68, 25–38. Lee, T.H., Hanes, D.M., 1995. Direct inversion method to measure the concentration profile of suspended particles using backscattered sound. J. Geophys. Res. 100, 2649–2657. Lhermitte, R., Serafin, R., 1984. Pulse-to-pulse coherent Doppler signal processing techniques. J. Atmos. Oceanic Technol 20, 429–442. Lhermitte, R., Lemmin, U., 1994. Open channel flow and turbulence measurement by high-resolution Doppler sonar. J. Atmos. Oceanic. Technol 11, 1295–1308. McLean, S.R., Nelson, J.M., Wolfe, S.R., 1994. Turbulence structure over two dimensional bedforms: implications for sediment transport. J. Geophys. Res. 99, 12729–124747, http://dx.doi.org/10.1029/94JC00571. Nezu, I., 2005. Open-channel flow turbulence and its research prospect in the 21st century. J. Hydraul. Eng 131, 229–246. Nezu, I., Nakagawa, H., 1993. Turbulence in Open Channel Flows. Balkema, Rotterdam, NL. Nezu, I., Kadata, A., Nakagawa, H., 1997. Turbulent structure in unsteady depthvarying open-channel flows. J. Hydraul. Eng 123, 752–763. Prandtl, L., Tietjens, O., 1929 and 1931. Hydro- und Aerodynamik (2 vol.), Berlin. Rolland, T., Lemmin, U., 1997. A two-component acoustic velocity profiler for use in turbulent open-channel flow. J. Hydaul. Res 35, 545–561. Shen, C., Lemmin, U., 1996. Ultrasonic measurements of suspended sediments: A concentration profiling system with attenuation compensation. Meas. Sci. Technol 9, 1191–1194. Shen, C., Lemmin, U., 1997. Ultrasonic scattering in highly turbulent clear water flow. Ultrasonics 35, 57–64. Shen, C., Lemmin, U., 1998. Improvements in acoustic sediment concentration profiling using an LMS compensation algorithm. IEEE J. Oceanic Eng 23, 96–104. Shen, C., Lemmin, U., 1999. Application of an acoustic particle flux profiler in particle-laden open-channel flow. J. Hydraul. Res. 37, 407–419. Shields, A., 1936. Anwendung der Aehnlichkeitsmechanik und der Turbulenz¨ forschung auf die Geschiebebewegung. Mitt. Preuss. Versuchsanstalt fur Wasserbau, Berlin, Heft, 26. Shugar, D.H., Kostaschuk, R.A., Best, J.L., Parsons, D.R., Lane, N.S., Orfoe, O., Hardy, R.L., 2010. On the relationship between flow and suspended sediment transport over the crest of a sand dune, Rio Parana, Argentina. Sedimentology 57, 252–272. Smyth, C.E., Zedel, L., Hay, A.E., 2002. Coherent Doppler profiler measurements of near-bed suspended sediment fluxes and the influence of bedforms. J. Geophys. Res 107 (C8), 19.1–19.20. Song, T., Lemmin, U., Graf, W.H., 1994. Uniform flow in open channels with mobile gravel bed. J. Hydraul. Res 32, 861–876. Southard, J.B., Boguchwal, L.A., 1990. Bed configuration in steady unidirectional water flows. J. Sediment. Petrol 60, 658–679. Thorne, P.D., Hanes, D.M., 2002. A review of acoustic measurement of small-scale sediment processes. Cont. Shelf Res. 22, 603–632. Thorne, P.D., Davies, J.S., Bell, P.S., 2009. Observations and analysis of sediment diffusivity profiles over sandy rippled beds under waves. J. Geophys. Res. 114, C02023, http://dx.doi.org/10.1029/2008JC004944. Thorne, P.D., Hardcastle, P.J., Soulsby, R.L., 1993. Analysis of acoustic measurements of suspended sediments. J. Geophys. Res. 98, 899–910. van der Werf, J.J., Doucette, J.S., O’Donoghue, T., Ribberink, J.S., 2007. Detailed measurements of velocity and suspended sand concentration over full-scale ripples in regular oscillatory flow. J. Geophys. Res. 112, F02012. van Rijn, L. 2005. Principles of sedimentation and erosion engineering in rivers, estuaries and coastal seas. Aquapublication, Blokzjil, NL. Venditti, J.G., Church, M.A., Bennett, S.J., 2005. Bedform initiation from a flat sand bed. J. Geophys. Res., 110, http://dx.doi.org/10.1029/2004JF000149.