Eropean Journal of Mechanics Fis 45 (2014) 125-143, Contents lists avallablo a ScloneoDiroct European Journal of Mechanics B/Fluids journal homepage: www.elsovier comMlocatalsimtiy Turbulent flow over the trough region formed by a pair of forward-facing bedform shapes K. Sarkar', BS. Mazumder* vl Mecha Laboratory, Pyles an Ape Mateos Une, dan Static tte, Coca - 700 108, nda HIGHLIGHTS ‘Turbulent flow within the rough region constructed by cw forward facing waveforms, ‘© Mean flows, Reynolds shear stress and overall drag are analyzed. { Semi-empirial relations fr turbulent intensities and shear stresses are provide. 1 Spectral analyses ate performed at vasious points along the waveforms, ARTICLE INFO ABSTRACT Frick hon Received § December 2012 Received in read orm ‘Accepted 24February 2014 ‘ese online 1 March 2014 ‘The present paper explores the rbulen low characetsc and overall dag aver he wougheedion ofa pairofaacent-Dforward-laciny dune shaped arf tutus with te diferent stoss-sie slopes Sd aims wo make a commpatative study. Suuctres considered here were of equal base len (2) witha ommon gente slope a6 athe downstream ace. Me toss sie angles Were respectively 50. and 90 Experiments were conducted atthe uv! Mechanics aboratry (FR) of lan tats nse (ST) Kolaca-Te velo data were elected using a3 Mro-Acoustie Doppler Veacimeter (ADV) a the Rewer Aue cenerine analyze te mean lows, Reynolds stresses and the overall drag at a Reynolds number cece f= 144 » 10% One-dimensional profes of urulen ow parameters demonstrate that the Now Forward aos woe Separation bubble and a thin perturbed shear layer between the negative veloc region and ote yer nov thigh velocity are the main sources of trbulence production. A changin turbulence characteristics Fw separation Demeenthe two crestpostionsis dented Areater amount oflow ressancesobservedin thease of emai stenes thestructurehavinghighestossside slope Thespetalarayssreveasthat peak power spel density Spstalanahss general occurs a¢0.25-1 Ie forthe steamewise velocty component ad at 09-30 He forthe ess Steam and vertical veloc components ower spectra showed beter dined peaks nat the seat ayer in the separation cel here i reaches to maximum which s two to thee times greater than ts vas inch upstream ad farther dowasueam eon. Stouhal numbers ae clelted using the requences oedies and vortices with dfetet set of chatacters len and elo scales: and compared Wi the previous ess of oes researchers (© 2014 Elsevier Masson SAS. ll sights reserved, 1. Introduction In the alluvial channels, various kinds of bedforms like rip- ples, dunes, bars or biological organisms like shells, wood frag- ‘ments or pebbles and small steps like structures are frequently observed on the bed. These objects significantly affect the over- all low field inducing separation and recirculation bubbles near ™ Coresponding author, Tel: 9 9831175236, smal eres: soundotphysiks@prailcom (Saka bioy@isicaL cin (asMarumder, ng/g) 10.1016) euromecina.2014.02.013 (987-7545/0 014 Elevier Matson SAS Al ight served the boundary due to the adverse pressure gradient. Detailed tur- bulent statistics, separation of flow and drag over complex ge- comets like wavy hilly surfaces, artificially made dunes or rip- ples, backward- and forward-facing steps with sharp and rounded faces, sudden contraction of pipe under various flow conditions hhave been tested both experimentally and numerically. Problems of lows over two/three-dimensional backward-facing steps with sharp edges, rounded faces, etc. have received much attention in ‘many engineering fields ike hydraulics, aerodynamics, ocean en- gineering (Barkley etal. [1], Poole and Escudier [2]. Casarsa and Giannattasio [3] and many others). Turbulent flows over moving or rigid bedform structures in riverine environment have been studied by many investigatorsSark, 8. Macunder Europea Journal of Mechanics ud (2014) 126-142 wa Notations A,B,C... L = Stream-wise selected locations for the ADV measurements for both DSFFS & DVFFS. P,Q => Extra two locations for the ADV measurements {or DSFES. (M,N, 0,P,Q,R => Extra six locations for the ADV measure- ments for DVFES. . => Profile of mean stream-wise and vertical velocity of DSFFS A _=Profile of mean stream-wise and vertical velocity ‘of DVFFS, £ =P Acceleration due to gravity, h = Flow depth fh =Crest height Re = Reynolds number. Fr = Froude number. u,v, = Instantaneous velocity components. U,V, W. = Mean Velocity components vi, w! => Fluctuating velocity components. uy" = Shear velocity Usp = Spatially averaged shear velocity. Uy = Maximum velocity. 2% Equivalent bed roughness. a =} Wavelength ot base length of the structures, « = Von Karman constant Gu. => Root mean square velocity components. uly => Normalized stream-wise & vertical turbulent intensity. foram = Dimensionless Reynolds shear stress. 8 = Normalized turbulence kinetic energy (TKE), Rye => Coefficient of correlation (U) = Spatial averages of centerline mean velocity. tr =Total boundary shear stress. St =Strouhal number (Lyn [4], Bennett and Best [5], Parsons et al. [6]. Best [7], Poggi etal [8], Ojha and Mazumder [8], Peetet al [10], Stoesseretal.[11] Mazumder etal. [12], Keshavarzi etal [13)) The turbulent Now and related bursting phenomena over an isolated asymmetric ‘waveform structure were studied statistically by Mazumder and Mazumder [14] using the analysis of multivariate normal distribution. Mazumder [15.16] studied the turbulence of fluid ‘ow in the trough region between a pair of adjacent asymmetric waveform structures using a statistical clustering technique based. ‘on geometry and interactions of turbulence bursting rate Recently, Paul [17] made a comparative study of the turbulence and fractional contributions of bursting events to the Reynolds shear stress over the trough regions formed by a pair of scalene and a pair of isosceles triangular-shaped waveform structures, Spectral and co-spectral analyses had been performed on the nature of oscillations that were embedded in the turbulent flow field at various points along the bedform profile. Simpson [18] suggested that the low-frequency signal resulted from wake flapping and the high-frequency signal from vortex shedding. Experiments of Krogstad et al. [19] revealed that the roughness of a structure primarily affected the shear stress, but it did not influence the power spectra. Krogstad and Antonia [20] showed that the co-spectra had a roughness effect very clase to the wall ‘A spectral analysis was also applied to examine the correlation between the velocity components and sediment movements in neat-bed environment. Kostaschuk [21] showed that the frequency spectra of near-bed velocity and optical back-scatter probe (OBS) data from the le-side flow-reversal zone ofa dune were composed of two distinct peaks. Investigations of Venditti and Bennet [22], ‘Venditti and Bauer {23] illustrated thatthe power spectra showed. larger and better defined peaks in the near-bed and recirculation regions. In thei studies, the Stroubal number calculated from the various combinations oflength and velocity scales revealed a broad similarity amongst eddies produced over the flow field and the bedform structures. ‘Ando and Shakouchi [24] conducted experiments on the forward-facing step through a sudden contraction of pipe using the LDA technique, Flows over smooth and rough surfaces of the forward-facing step (FFS) were tested by Camuss etal, [25], Sherry etal. [26], Ren and Wu [27] using different techniques such as LDV and PIV with different Reynolds numbers. These studies verified that the Reynolds number, the roughness of upper surface and the slope of the flow facing surface play important roles on the geometry, shape and size of separation bubble and the position of, feattachment point. There are some data available predicting the effects of hile. the ratio of boundary layer thickness 8 tothe step height h on the reattachment length X.. Sherry etal. [25] studied the flow over the forward-facing step for several values of the ratio 4h, They categorized the study into two groups: for 8/h > 1, the reattachment length X, was strongly dependent on the 3/h and for 5/h < 1. the X, was weakly affected by the rato, and usually situated around 4h to 6h, In spite of all these studies, no experimental studies were performed to examine the mean flow, turbulence and drag over the {rough region ofa pair of steeper stoss-side slope bedforms under the reverse flow conditions, though such a study has the potential tw be useful to the researchers who study the bedforms in natural environments, especially those that experience reverse flow conditions (Dinchart [28], Lefebvre etal [29], Winter etal. [30 How does the flow over the trough region of forward-facing bedlform structures compare to that over the rare-facing ones? ‘Therefore, a substantial investigation is required to understand the basic hydrodynamics experimentally ina flume over a pair of 2-D adjacent bedform structures oriented against the Now. ‘The alm of the present study isto determine the spatial changes Jn turbulence and overall drag over and within the trough re- gion formed by a pair of adjacent 2-D forward-facing dune-shaped structures. Two different stoss face slopes (50° and 90°) of bed- forms oriented against the flow with a common slanted lee face of 6 along downstream were considered separately under iden- tical flow conditions. More precisely, to understand the turbulent flow and its characteristics over the trough regions between two artificial 2-D bedform structures, the two cases of bedforms were chosen witha common gentle lee slope of: (1) asymmetric struc- tures with $0" stoss-side angle, and (2) vertical (80°) stoss-side structures facing against the flow. which were akinto the bedforms ‘measured in the seabed bathymetry in tidal environments (Lefeb- vre et al_[29)), The divergence ofthe stoss-side slopes between the two cases and the existence of the second structure placed adja- cently downstream of the fist, made significant differences in the ‘mean flow and turbulence fields forthe first and second bedforms. The stoss-side slopes of primary bedforms measured by Lefebvre cet al, [29] were approximately 77° and the lee-side slopes were al- most gentle (see Fig. 6 of Lefebvre etal. [28]). Therefore, the use of steep stoss-side slopes (50° and 90°) with a common gentle lee- side slope in the present study was justified because the slopes of the bedforms were of similar order which fall into the range of ‘hose measured in natural environments. The velocity data were analyzed to emphasize the turbulent statistic, overall drag and the power spectral density aver the trough regions of such bedforms With different stoss-side angles, which were not studied earlier Although the choice of two artificial bedform shapes in the Slow was not the actual representation to form the equilibrium flow conditions over many similar bedforms in the natural tidal flow regimes, this study would provide some understanding of the5 Mazumde/ European four of Mecha us 45 (2014) 125-142 To Computer Deedtion A 8 co © Fon 4h KL 8 Case: (nosFes ~s0! a 0.2. The usual log-law is also observed at a distance of 3.0(%/2) from the toe point K for the case of DVFFS (figure rot shown), For cleat visualization, contour plots of normalized stream-wise velocity are presented in Fig, 4(2, b) for both cases Itis clearly seen that for the DVFFS case, there isa flow separation (2et0 U velocity) at the upstream ofthe frst structure, between the location Mf (x/3. = —1.116) and the first crest. The reattachment point is also observed at location F. The flow again separates at G with zero velocity following the positive low velocity between F and G, just before the second crest Subsequently, the measured normalized stream-wise mean velocity profiles for both DSFFS and DVFFS from upstream to downstream including the recirculation region, except two crest locations D and I for the DSFFS case, are best fitted with high coefficient of regression by the equation as U/ty = ae 91-0780) 6)Sark, 8. Macunder Europea Journal of Mechanics ud (2014) 126-142 a aos oa 02 b | “| “| o a GH 10 02 0 os. 08 1 12 EE sated values ferent parameter fr DSS & DVFS using the Gompertz grow function (6) ti steamewise velocity profiles (Fe 3 q = = & a = = E c 1037 0.026 26381 099 toa 0133 387.395, 092 1 1029 00093 1934 095 1.088 0029 5348 098 Where the parameters a, b and care determined from the best fit. This equation is named as Gompertz growth function (Gompertz [33)), This growth function is proposed here mimicking the spatial growth in the turbulent boundary layer beyond the frst test point. The best fitted equations predicted from the data at the referred locations for both the structures are plotted against 2z/h in Fig, 3 as continuous lines, Its interesting to note that the Gompertz growth function fits well for both cases at all locations except the crest locations D and I for DSFFS. The values of the parameters for each case are given in Table 2. Here, a and b indicate, respectively, the maximum normalized velocity of the distribution and the normalized elevation where the normalized velocity reaches the value 1/e. The parameter c signifies the nature ofthe profil, Le. how fast or how steep isthe movement to reach the maximum peak velocity. It clearly signifies the thickness of the bottom shear layer and fluid movement in that region. More importantly, higher values of a and c occur in the recirculation region just downstream of the crest, indicating a high degree of turbulence in that region, Significantly, cis related to the shear in the profile, which is thus tied to turbulence production, but the maximum normalized velocity a varies mostly because of the variation in c. Hence, by comparing the values of a and c, it can be argued that due to the difference in stoss-side slope, a greater amount of turbulence is created for the case of DVFFS, which acquires a larger and steeper peak than DSFFS. ‘A comparative study between the two cases revealed that at ‘measuring stations A and B up to x/ = —1.266. the profiles of, mean velocity U/Uy overlapped each other and showed a standard log-law. Thereafter, at location C, the intersecting point of the velocity profiles from the two cases is observed at a level z/h ‘5 0.15, nd that level o intersecting point increases asi proceeds further downstream up tothe location L(x/ = 1.266).That means the locus of intersecting points of velocity profiles from the two cases increases with distance up to x/ = 1.266. [tis naticed that the heights of intersecting points at C (fist crest position of the VFS case) and D (For the DSFFS case) are much lower than that at corresponding locations H and I (for second crests). Te distance of the reattachment point from the first crest point C fr the case ‘of DVEFS is approximately SH which is in between the locations E and F (at x/i = ~0.48 approximately). “The time-averaged wall-normal velocity profiles (W/U) for both the cases are shown against z/h in Fig. 5 at different stream- Wise locations. It is observed that the mean velocity increases with the vertical height z/h approximately up to z/h =< 0.2 near the boundary up to the first crest positions of the corresponding structures, Non-zero positive vertical velocity profiles for both DSFFS and DVFFS are observed at the crest points. In the trough region for both cases, the wall-normal velocity is affected by the flow separation, ust downstream of the crest, negative velocity occurs (directed towards the bed). Away from the separation bubble, the vertical mean velocity (W/U,) is almost unaffected and shows zero value throughout the entite water-depth. It is notable that, at the leading point C just upstream of the crest position for the DSFFS case, the velocity reaches a maximum value atthe level 2/l = 0.1, whereas for the DVFFS case the maximum vertical velocity occurs at its frst crest position C. Up to some distance in the trough region starting from the first crest, W/Un 1s negative up to z/h_< 0.4 and then it becomes positive near the junction position. For the both cases, the wall-normal velocity sn the vicinity of the first crest point is much greater than that ofstar, tau uropean foural of Mechanls us 45 (2014) 125: ‘A08 B09 cos of 3 oa| oa goa oa ol oa i 02 oa 3 3 4 + of Doe = roa 06| $ os] fos oa| o2| o2| 4 $ro001 os os 4%. Gos 08 06| $ os 04 asl old Joe Kos Los os} os} $ os] Soa] B o4 oa} 2 “if q oa Wu, wu, wu, Fig.5. Normalized walnoralvelcty component W/V] athe laeations A B,C, ita points of DSS and DVS ass. a 06; to forbotnDSFTS and DVETS where the symbols and respectively indeate Fig 6 sueamine pet the second crest point, which provides the most comprehensive results in explaining the entire flow field over the trough region ‘between the two adjacent forward-facing structures. The profiles of vertical velocity show plunging vertical flow over the trough portion and flow directed towards the bed surface, This implies ‘that the reattachment of low towards the bed starts behind the velocy vctrs or) DSFS and (b) DVS; (c) Magid version of the DVFES cae, «test position, Away from the separation bubbles the mean velocity W/Up is almost zero throughout the whole depth for both DSFFS and DVFFS cases. ‘To visualize the neat-bed separation bubble, streamline plots of velocity vectors (U, W) over the complete trough regions of the both structures are presented in Fig. 6(2-c).Fig. 6cis the magnifiedA Srko 8S. Mazur /Eurepeo Journal Mechanics ud (20) 125-148 Aon Boe Care os a6) ool f oa aa ol § o2| o2| o2| - 9 —— o| — o| — 002 oa 08 «02 On rr Dos Eo Foe os os|$ os\t Eos ae oa o2| o2| o2| > 9 ot o| 002 0* 06 0 02 0* 0 020A Hoe 109, a) o8| ak: oa * o2| | tyke aN aos "0 02 0a 002 oa Koa Loe 6) 06) oat oa o2| * o2| ° as 6% oz oa 8% va 06 ig. 7. Normalize streamwise intensity component s/t the locations A.C {ata points of DSFFs and DVERS. Te black cuved ies Cvespond to (11, version ofthe streamline plot for the complete trough region. For the DVFFS case, a significant flow separation is observed after the fist crest C on the upper lee-side (Camussi etal. [25], Ren and ‘Wu [27]), which sustains up to the location F (reattachment point), and again it shows zero velocity at G (Fig. 4b) at the near-bed region before the second crest H. The length of the separation bubble X, ~ 16.5 em with the bubble center at x/i — —0.8, z/h = 0°12 (exactly atthe crest height), For the DSFFS case, no indication of separation and reverse low neat the bed is observed either in upstream or in downstream of the crests D and I. The disappearance of any separation bubble downstream of the crest for the DSFFS case can be thought of as a result of favorable stoss- side slope facing structure. This finding agrees well with Ernstsen etal. [34] and Lefebvre etal [35]. However, the separation region could be quite thin, thus escaping detection by the ADV andjor Separation could be intermittent and thus not appearing in mean flow measurements. Perhaps, the time series measurements at nneat-bed just downstream ofthe crest might include few negative data; however the noise inherent in the ADV signal makes it slfficut to identify intermittent flow separation, 42 Distribution of turbulence intensities In the turbulent flow field, the root-mean-square velocity ‘components a, and oy are defined as w= Xe or Lye wy. @ Oe @) {oor beth SFIS and DFS, where the symbol «and neat rspecivey ‘The normalized turbulence intensity is thus defined as u=B © unk (10) where m is the total number of velocity measurements and Us, is the maximum velocity. ‘The one-dimensional profiles of intensities of turbulence (ly. fs} in stream-wise and wall-normal ditections were computed for both DVFFS and DSFFS along different stream-wise locations. The distributions of normalized horizontal intensity along the stream- wise locations are plotted against z/h in Fig. 7 and the contour _maps are shown in Fig, 8(a, b)-As the low approaches both of the structures, a slight increase in J is visible. For the DVFFS case, 6, shows a maximum value in the vicinity of the first crest position which is about SOX of Up but this is valid only in the internal layer above the crest region. Again. in the layer between the outer fluid velocity and recirculation cell just downstream ofthe first crest point. an increase in [, is observed with maximum value at the level z/h = 0.18. The maximum of a, reaches to about 25% of, Uy, anywhere except at the crest C profile, where 4, reaches 0 about 50% of Up, which agrees well the previous studies by Ren and Wu [27] and Agelinchaab etal. [36], The maximum value of, atthe second crest H is about 60% ofthat atthe fist crest at C. The ‘ase of DSFFS with smaller slope atthe stoss face does nat show any cleat peak ofa, asin the other case at any ofthe measuring stations. It shows that the maximum value of 3, over the first crest D is about 17% of Up and further downstream itis only about 92—11% inthe trough region and at the second crest point itis only about 13% of Up, As we go in the vertically upward direction, after someaos ‘suka, 8, Mazunde/ European journal of Mechs us 45 (2014) 125-142 c 015 D E 02 F 0.25 GHI 03 0.35) 7 04 045 os ig. 8. Normalized steam-vwise intensity contour pet fr Both) DSF and(b) DVS. The Mow erections fom eft tori Acs Bos 6087 as os 08 04] 04] 04] 02 02 02 a a a oc ee Tis a2 pos Eos 0.6| 06) £ oa oun a2 02 0! 7 0! 2 oC ne ec Hoa 08 7 7 0 0.05 01 015 02 °o 0.05 0.1 015 02 °> 0.05 0.1 0.15 02 cag fee oa at oa as 02 02 0 005 01 015 02 % 005 0.1 015 02 > 0.05 0.1 015 02 Ww Ne Noord wor ty enpne hf] he cl 11 bh Sk DS us ee bl «an ee spectively data pent of DSEFS and VES. Tae Back ved lnscerrespond to E1}. increment inf, it decreases gradually towards the free surface. The position of the maximum intensity moves towards the free surface from the point of flow separation to further downstream. ‘A zone of higher turbulence intensity is attributed to a high level of noise from the acoustic reflection of the bed (Nikora and Goring [37)) and is characterized by low mean velocity. It can be Stated that the DVFFS case creates greater turbulence intensity and ‘thus mixing more quickly into the interior ofthe flow than that of the DSFFS case. This difference is due tothe fact thatthe DVFFS case has greater height and creates greater roughness. Based on this difference in the mixing between the two cases, shortly after the ‘low reversal uring flood-tide, the dune would create significantiy| ‘greater mixing than would be created later as the dune's shape Degins to shift from originally backward-facing to forward-facing. “The distributions of normalized wall-normal intensity (Vy) along the stream-wise locations are plotted against 2/f in Fig. 9 and the corresponding contour maps are presented in Fig. 10(ab) Note that this component J, has a similar trend as J,, however much smaller in magnitude, only one-third to one-fourth of ly ‘The intensity i seen to be prominent downstream from the fistSark, 8. Macunder Europea Journal of Mechanics ud (2014) 126-142 ns. a 06 B cD EG F GHI J KEEL: b 3 cD € F GHI J x. 0025 005 0075 0.0125 01S OATS 2 ——— SSE Fig. 10. Novmnazed wal-nrml intensity contour po for bth} DSFS and) DTS The Naw direction om et tg ‘ables Eximated vakes fiferent parameter fr DUETS uring the Gate Pak function (1140 turbulence itenstes (Fis. 7 and 9 Taxatons i z c 003 oes 9555 231 2800 1379 bea F 0156 0159 0205, os? 0.168 out 095 ' Tins ons aise a8 22s eau bea 1 187 ons om oar 0236 oat ost crest and this effect persists for a longer distance than in the case of stream-wise intensity. The Jy, staits to increase prominently in the flow separation region just downstream of the first crest point solely, due to the development of a thin shear layer in between the high velocity at the outer layer and the negative velocity recirculation region. The maximum value of ly occurs in the vicinity of the second crest point. tisinteresting to note tha for both of the structures DSFFS and DVFFS, wall-normal intensity ly is much greater atthe second erest than atthe fist crest position: 33% and 83% greater than the values of the first crest point respectively. However, the maximum value of zis found to have a magnitude of about 17% and 6% of Un for the cases DVFFS and DSFFS respectively, Farther downstream, it diffuses towards both the outer region and the near-bed region. These conclusions can be made easily by visualizing the contour plots, Fr the DSEFScase, the peaks inthe intensity profile fy are more prominent than for the case ofl. Interestingly, the appearance of peaks gradually moves inthe downward direction measuring in the trough region for the DSFFS case and it moves in the upward direction for the DVFFS case. The recovery ofl, beyond the downstream toe of the second structure for DSFFS is much faster than that in DVFFS and at greater distances compared to the recovery length of hy ‘all the measuring locations for the DSFFS case, the measured profiles of I, and fy follow the semi-empirical relations (Ea. (1)) proposed by Nez! and Rodi (31) (not shown in the figure) However, it fails to estimate the intensity profiles for the DVEFS case at the locations beyond the first crest C due to stronger perturbation, For the flow reversal zone just after the fist crest C Of the DVFFS case to the last measuring location I, the measured profiles of both J, and J, can be fitted using the following expression as be =Ao[-05(2=2)'] where the parameters m, s and A, are determined from the best fit. This equation is known as Gauss Peak function. The best fitted equations are predicted from the data at the referred locations and are plotted against 2/h as continuous lines in Fig. 7 for , and in Fig. 9 for j,. The estimated parameters for each case are given in Table 3 When the Gauss Peak function is used to describe a probability density, the parameters m and s are the normalized mean and standard deviation of the distribution, Here they are directly related to the characteristics of the intensity profile. The parameter im identifies the normalized vertical height, where the maximum Intensity occurs. The parameter s denotes the thickness of the region of high turbulence, and A, denotes the magnitude of the turbulence distribution; thus the two together characterize the {ota turbulent energy integrated over a given profile. ap 43, Reynolds shear stress ‘The one-dimensional profiles of normalized Reynolds shear St1€SS (Cazgq, = —WW'/US) ate shown versus 2/t forall stream Wise locations in Fig 11 for both DSFFS and DYES, Itis observed that forthe DVEFS, the shear stress in the viinity ofthe fist crest is comparatively small and it inereases along thestar, tau uropean foural of Mechanls us 45 (2014) 125: Avs: Bos, Cos 06: 06! os| § 04) oa oa 02) o2| o| ° ° 0 05 Ww 15 20 0 5 1 158 20 0S 10 1 20 x10" x10”) x10" Doe Eos, Fo3) oc os os| oa oa oa | 02| 02] 0! o 0 0 8 1 18 2 0 5 0 15 20 0 5 10 18 20 x10" x10? x10" Gos Hos [03) | os os| oa] oa oa 02 02| “ 08 | ° a 0 8 10 15 20 0 5 10 18 20 0 5 10 15 20 x10" x10 | 0 5 10 18 2% 0 5 10 15 20 Tm, «10° Fig. 11. Dimensionless Reyples sea 155 fg] athe leans A. BC paints of DSP and DVEPS. Black curved ies covfespone UE (1), ong, 110° x10" ‘o Lfr beth DSS and DVFS, where the symbols and nate espectively ata Table ‘timate values of ferent parameters or DSFES& OVERS sing the Gaus Pek function (11). the Reyne shear sues (fg 1) Tecatons DSS DS 7 z = = m= 3 ze e m one 0108 aot 085 oa 105 007 089 c ose? oats aot at ou 8008 8003 035 D om cos baat on ona one cri 098 F ars bose ace os? ous ore ons 099 ¢ rr 68 oaz O96 oer 8098 Sais 099 8 O10 bas az 070 09 os: oon os 1 ons 30 00s 098 200 Siee Soe 098 K ost oar aa bse base ont 007 098 ‘entire trough region with a clear peak near the bottom boundary and maximum in the recirculation tegion. Itis observed that the shear stress is much higher at the second crest than that at the fist crest position, which indicates a greater flow resistance at the second crest H profile, following the higher resistance along the ‘ough region. A development ofa strong shear layer is observed between the two overlaying flow within the trough region: flow reversal near the boundary and high velocity away from the boundary. similar effect is observed forthe DSFFS case, the stress is greaterin the region just downstream of the frst crest. However, shear stress is significantly higher at the recirculation region inthe DVFFS than that for the DSFFS case. Reynolds shear stress profiles collapse together in the upper part ofthe flow. z/h > 0.3, which is ‘consistent with the findings of [27] The measured profiles of Reynolds shear stress for both the cases can also be fitted with the same Gauss peak function {(€,(11)) a all the positions A toL as used earlier forthe intensity, profiles. The corresponding parameters A,,s and mare determined from the best fited profiles, and are provided in Table 4. The best fitted equations predicted from the data at the referred locations are plotted against z/ in Fig. 11 as continuous lines. [tis observed from the & values in Table 4 and from the figures thatthe fitted Eq, (11) agrees well for all the locations except the locations C and D for the DSFFS case and locations # and C for DVFFS. It is interesting to note that the single Eq (11) can easily be fited tothe profiles of both the intensity components and the Reynolds shear stress just by adjusting the parameter values 44, Turbulence production and correlation coefiient ‘The magnitude of turbulence production P by vertical shear ‘over the two-dimensional dune-shaped bedform structure [22.23]A Srko 8S. Mazur /Eurepeo Journal Mechanics ud (20) 125-148 a7 Ae Bos, Cos) os} 0. os| Soa! ou oa | | ol B 20 2 4 6 : Dos; F038; os| oa 02| ol }03) 0s| oa o2| B 20 2 4 6 = Ko8, Lo3) 3 oc os 2 | oa 02| 02] ol B 024 6 2 2 6 P (ems sect) x 10° ig. 2. Turbulence production Ph] attheloations A.B. vrs is defined as (12) ‘The turbulence production P is plotted against z/h along. all the locations for both cases in Fig. 12. This plot confirms that turbulence production P is mainly concentrated in the separation ‘zone just downstream of the first crest and in the trough region between two adjacent bedform shapes and up to some distance after the second crest for the DVFFS case at 2/h approximately about 0.18. While for the DSFFS case, maximum turbulence production is concentrated atthe near-bed toe point af the second bedform shape. The turbulence intensity and Reynolds stress plots are greatly complemented by this plot. This illustrates that the region of flow deceleration and intermittent flow reversal dominates the turbulence production over such bedform shapes, ‘A coefficient of correlation is a mathematical measure of the strength of the relationship between two variables. Inthe present ‘ease, it can be thought of as locally normalized Reynolds shear 055 Ryy = —iu"/oyay. Hence, the boundary-layer correlation coefficient (1 = Rye = -+1) is a normalized covariance that expresses the degree of linear relationship between two variables W and w velocity fluctuations. The contour plots ofthe boundary layer correlation coefficient Ry, against 2/ are shown in Fig. 13(2, b) for both DSFFS and DVFES.{n lear water, itis typically around ‘04 over the most fat boundary layer (Hinze [38], Nelson etal [39]) and away from the bed in the outer flow region, it possesses alower vvaluein the range of0.0-0.3 [32] For laboratory observations over artificial dunes, Venditti and Bennet [22] suggest that Ryy < 0.3 within the internal boundary layer, Roy ~ 0.3-0.5 in the outer flow region, and Rye > 0.6 in the wake and separation cell. In our P (on?) sec8) x10 P (on?) sec) x 10° ‘0 Lor both DSFFS an DVEFS, where the symbol and inate respecivey datapoints of DSFFS ane experiment, itisin the range of 0.20-0.35 inthe internal boundary layer, it reaches to maximum in the separation cell which isin the range of 05-065, and in the outer flow region tis approximately 0.05. So, it agrees well with the previous results. Low values in the vicinity ofthe internal boundary layer downstream ofthe low reattachment may be due to the turbulent eddies intermittently Impacting the bed, whereas higher values in the separation cell indicate strong turbulence production and structural coherence. 45. Flow resistance Various approaches were taken by the researchers to deduce the flow resistance, ic, the total boundary shear stress and hence the overall drag, Commonly, vertical profiles of stream-wise velocity measured over the bedform are used tocalculate boundary shear stress. Profiles may be either locally derived or by means of spatial averaging. Researchers preferred the latter procedure eg, Smith and MeLean [40], MeLean and Wolfe [41] over the years We construct the spatial average by averaging the velocity data at constant values of above the bed along the stream-wise direction. Spatial averages of flume centerline mean stream-wise velocity at constant height (2), denoted as (U), are computed for both the structures DSFPS and DVFPS and are plotted in Fig. 142, b). The Spatilly averaged profiles for both the cases are logarithmic with Rt -0,98 and are given by w) and (U) = 5.280 log(z/h) + 53.184 (for DSFES) 4) where the lead coefficients of the above equations have the dimension of cmjs. The shear velocity u,r computed from the 10.506 log(z/h) + 57.056. (for DVFFS) 13)E 00504 015 02 025 09 F CHT 03504, i os 045 05 055 0.65 Fig. 15, Contour plas of corel coetcent fr bth DSFS and) DVFS. The flow deco fom et tig. aos bos 07 07 ‘ or DSFFS or DvFFS 0s os oa 0 oa 03 02 02 ao Sot 0105304080 010" 3530-850 U> ig. 14. Spaally averaged streamwise velocity profes) fitted log-aw and the total boundary shear stress (= pu) are given by 4.202 cm/s and 17.66 dyne/cm? for the DVFFS case, and 3.112 emjsand-4.461 ne cmt for STFS respectively. is notable that mean flow resistance over the DVFFSis mich higher than that of DSTHS as expected 6. Spectral and co-spectral analysis [Extensive research works suggest thatthe turbulent flow over the blunt structure is dominated by eddy-like motions, especially ‘where the turbulence production is maximum, ie. on the lee side fof a dune structure [21-23]. We performed a spectral analysis (DSF and (9) DVS, where the symbol «and indicate respectively data pinta SFIS and fon the velocity-time series with special cate at the interface between the streaming flow over the crest and reciculating fluid in the separation cell, At various points, such as outer layer, ‘wake, separation cell and reattachment point over the trough fegion, velocity signals were analyzed to calculate the power spectra. Velocity data at a point were frst low pass filtered using Butterworth infinite impulse Response (UR) with cut-of frequency 5 Hz to remove noise and to eliminate the possible aliasing effect. Thereafter, to minimize the noise in the higher frequency region of the intended spectra, each of the velocity signal was divided into ensembles of 2048 measurements and hence we obtained 5 ensembles. All of them were detrended about their mean separately, The power spectra were estimated at 0.0195 HzSark, 8. Macunder Europea Journal of Mechanics ud (2014) 126-142 moo | EOL) TF OL Velocity Spectra (f# P(A) em2s-2) iD oa 110 or. 110 01 Frequency (Hz) Flow Regions —u ae re v TEL tral Borday layer eat we | rH OF location Ds consicred in ths ease which les in he outer lar Felon} Thiek sob lines inate the ples of steams spec, thin said nes ace poles of Gros-steam spectra and hick dashed ines indate the intervals from 0 to 20 Hz with 95% confidence limit using the ower spectral density (psd) algorithm available in the Matlab software package with Hamming window. Ifthe spectral density PQ) (cn? 5 Hz!) is plotted against frequency in log-tog scale, it should follow the —5/3 slope. As our intention is to know how much power is contained in each frequency rather than the spectral density, we multiply the P() by the frequency ie, f P() (em? =~) to obtain the power spectra and plot it against f in semilog scale Power spectra of velocity signals at various measuring lations are shown in Fig. 15 forthe DSFFS case and in Fig. 16 for DVFFS. ‘The area under the spectral curve represents the total variance and the peaks correspond to energetic eddies (Boppe and Neu [42]} Power spectra forall velocity-time series by default were seen to be largely dominated by a lower frequency region from 0.25 Hz to 3 Hz, The stream-wise spectra generally possess peak at frequency ranging ©0.25-1 Hz (eg, 2-£ of Fig, 15 and 0.40-D of Fig. 16) ‘with a mean at about 0.625 Hz, whereas the vertical and crass velocity spectra show peaks at ~0.9-3 Hz and 1-3 He (eg. 17- D of Fig. 16) respectively. So the power spectra of v and w have peaks at higher frequencies than w. The occurrence of peaks at different frequencies differs due to the occurrence of dominant eddies of different shapes and sizes, which might be elliptical when they advect and diffuse away from the shear layer at further downstream. Itis observed from the figures that near the bottom region (eg. 2-F of Fig. 15 and 2-£ of Fig. 16), especially in the recirculating zone of DVFFS (eg. 0.60-N and 2-£ of Fig. 16) and at the junction point of the two structures, generally higher peaks are Seen indicating larger spectral power. Comparatively sharper and better defined peaks are observed in the case of DVFFS than that of DSEFS especially in the separated region indicating higher degrees of turbulent eddy generation. There is a clear sign of increment in the maximum spectral power further downstream in the separation zone. In the outer layer (eg. 17-D and 17-Fof ig. 15, and 17-C and 17-F of Fig, 16), not much differences in spectra are visible between the two cases which indicate similar behavior of turbulent flow at the outer layer. As we move downstream and vertically, the magnitude of the spectral power decreases (eg. 0.40-D,7-D and 17-D of Fig 16) but the value of te frequency ff maximum energy remains the same, Further upstream of the structure to the leading crest point, minimum spectral power is seen similar to the behavior in the stoss region over the dune [22,23] which indicates that there isno dominant or preferred eddy frequency, Here, the w spectral variance is observed to be greater than the w spectral variance, So, it can be argued that no “relict” ‘motion s occurring in the present study which isin contrast to the observation of [23] Co-spectral analyses are performed on the time series of u and w components of velocity following the same procedure and are shown in Fig. 17 at selected locations for both of the cases, Crass-spectra have peaks at a slightly higher frequency region than the corresponding u spectra. The coherency-squared spectrum shown in Fig. 18 at some locations can be thought of| a8 a Frequency of specific correlation coefficient as interpreted by Jenkins and Watts [43]. Clear peaks were observed in the case of DVFFS, whereas coherency-squared spectra showed peaks and valleys in the case of DSFFS indicating weakly correlated velocity fluctuations especially in the wake region. In the case of DVFFS, strong correlation is seen with notable decay at frequencies greater than 1 Hz.It showed mote defined and dominant peaks inthe case (of DVEFS in the recirculation region indicating higher correlation. For the case of DVFFS, the co-spectral analysis at 2-E of Fig. 17 showed three strong peaks: at 0.45, 1 and 2 Hz At this location, spectra showed two defined peaks at 0.35 Hz and 15 Hz and .w spectra showed a peak at 2 Hz. The coherency-squared spectra resulted in a good correlation showing a peak at 05 Hz and interestingly the secondary peak isnot recognized [22] ‘The frequency J. of eddies and shedding vortices in turbulent motion can be defined using a non-dimensional frequency parameter named as Strouhal number defined by Levi [44] as fh Uy, sr 3)star, tau uropean foural of Mechanls us 45 (2014) 125: 34 we 7D OL 172 OL TF OL 176 OL 2 ct 5 nN 15 12 414 08 ma, rw Eis £10 5 A A 0 Em 1A, Few m 7 aun se ron 730 RY 7 Oo mw 20 s 20 o Late err 1 10 ora 110 Orr 110 orto ora 110 ar 14 Frequency (Hz) Flow Regions —u owe v feof er fara R Reatachren| ig. 16, spectral anayss ofthe slece vloty-ime series fo the case af DVS (were 1-C OL means that se velocity signal a 17cm eight above the Bed suitce location is considered in tis case which es nthe outer lye ‘osetream petra and ik dashed ne inet the pais of wal-nrmal spect fegion, Thick solid ines indicate the poles of eameise spectra thin solié nes ate pote of a Frequency 2) O14 410014410 O14 4 10 O14 4 1001.14 10 ° 08 4 2 464 ,V 8 ob 2 AW San pw As Sai e ES pacup find fn ei! of q “6 4 cuoo wt een se 2esc 120k 30 FW 040 mL ig. 17, co-specral analysis ofthe selected strea-wise veloc aginst se wall-nemal velocty-time eis fo the cas of 2) DSFS and (9) DVEES. where J, and U, ate the relevant length and velocity scale, and fis the frequency of oscillation. For turbulent flow, one might ‘use flow depth (h), bedform height or crest height (h’), length of separation cell (X) ete fora length scale and fr U,the candidates are maximum stream-wise velocity (Up), depth-average velocity (G) (where standard log-law is maintained), friction velocity (u.) ttc, These scales must be proportionate with the process, where ‘the Strouhal number Sr is tobe characterized, Using the mean for the stream-wise spectra, we intend to calculate the corresponding. Strouhal numbers which are given in Table 5. These Strouhal numbers can be compared with the other investigations. For laboratory studies, Sr generally are in the range of 0.3-0.60 using 1, = h; and 0.1-025 using |, = W (using U, = Up for both the cases) (Nezu et al. 45], takura and Kishi [4], Ikeda and Asaeda [47], For the field studies, these ranges are 0.3-0.4 using I, ‘and 0.10-0.12 using , = ft (Kostaschuk and Church [48], Venditt [49)). The overall consistency of the present Sr data with other investigations is quite good, Tables toa mbers (5) sing the peta analysts of Uvelctysgnals. Meaniequensy 7) _tengthscle(G) _Veoeiysealeiuy) 5 ‘625 te ne Wea eas 625 ie k 054 9525 He X= 165m) oat 0625 te x 029 52s He W335e0)| on Kostaschuk and Church [48] estimated thatthe constant 1/St is ‘equal to 2. Using this from the Strouhal number relation we get fhe F wo. sr (16) ‘This produces a frequency of 0.25 Hz, which is similar to what we ‘ee from our spectral analysis. There seems to be a generally good agreement with [32] that eddies are generated in the separationSark, 8. Macunder Europea Journal of Mechanics ud (2014) 126-142 SB gos 72 FW ew ew row Hw 74 Fw io B02) My a) My yh ily Boo june eae Bos ' goo eM lM by | rc Fw OW 1 08 Ww 1 WwW 1 FW Ly why My go ~ Me 8 Bos go. Wy Frequency (H2) Fig. 18, Squared-coberency spectra the seece velociy-time series othe aso (a DSFES ang (b) DVS zone region. Simpson [18] showed that in the lee side downstres of the eres positions from the shear layer bounding a separation zone the frequency of eddy shedding () can be estimated from, 5-08, a ‘Again, be argued that the ‘flapping’ of the shear layer in the reattachment position could be approximated from cas) Using Xq — 16.5 em (approximately) yields eddy-shedding fre- quencies of 1.74-2.33 Hz (using the maximum velocity Un as the Velocity sale) and 1.24=1,64 Hz (using the depth-averaged veloc- ity as the velocity scale) and shear layer Napping frequencies of (0.29 Hz and 0.21 Hz by a similar procedure, The lower frequency ‘matched well with our experiments in the separation region at 2-E of Fig. 15 and 2-E of Fig. 16 similar to the achievements of [21]. As the measurement ofthe exact reattachment length Xp is uncertain, so the eddy-shedding frequencies which we got might not corre” late with the experimental result; still 2-£ and 3-G of Fig. 16 show comparatively good agreement withthe predicted frequencies. 5. Conclusions ‘The purpose of the present study is to investigate the turbu- lence, overall rag and the power spectrum within trough regions formed by two 2-D forward-facing dune-shaped waveform struc- tures with two different stoss-side slopes = The structure with a vertical stoss face (DVFFS) has a thick sep- aration bubble with a prominent recitculation zone just down- stream ofthe first crest, whereas no evidence of recirculation is found at the downstream of the second crest. However, in the case of DSFFS, there is a clue of flow separation over & small portion inthe vicinity ofthe first crest with nearly zero value of streamwise velocity. Its interesting to note that forthe case of DSFFS the stream- Wise velocity follows the standard log-law at all locations ex- ceptthe crest points D and! along the flow, whereas forthe case of DVEFS, the velocity follows the log-law at locations A, Band [M before the first crest, but it drastically deviates from the log- law after the crest. The recovery ofthe log-law for DVFFS is at about 3¢x/2) from the toe point ofthe second waveform. How ever, an empirical model, known as Gompertz growth model. is propased to fit the normalized stream-wise mean velocity pro- files for both DSFFS and DVFFS from upstream to downstream ‘except two crest locations D and I for DSFFS, The real merit of the growth model is that the fitted parameters of the made! sig- nify the physical meanings associated with the spatial growth of the turbulent boundary layer. This model provides a useful ‘means to smooth measured profiles, but it does not seem 10 hhave any predictive capability. tis interesting to note that the locus of intersecting points of stream-wise velocities of the two cases increases along the flow. ‘Ahigh turbulence intensity is characterized by low flow in the flow reversal region anda low turbulence intensity is character= ized by high flow in the outer flow region. The maximum of 6, reaches to about 25% ofthe maximum velocity Ux forthe DVEFS case and itis about 17% of Un forthe DSFFS case, The values of 4, atthe second crest are 60% for the DVFFS case and 76% for the DSFFS case ofthat at the first crest, However, a isrmuch greater at the second crest than a€ the first crest position, which are approximately 83% and 33% for DVFFS and DSFFS respectively The vertical normalized intensity ly is seen to be more promi- rent downstream of the crest than the stream-wise intensity and the effect persists fora longer distance. Itis worthwhile to ‘mention here that the exponential relations proposed by Nez and Rodi [31] agtee well with the experimental data of I, and 1, at all the locations for the DSFFS case, whereas the Gaussian Peak function agrees well with the observed data for the DVFFS case after the frst crest. In the trough regions for two different cases, a strong enrich- ‘ment in shear stress is observed for DVFES compared withwa, ‘suka, 8, Mazunde/ European journal of Mechs us 45 (2014) 125-142 ‘that of DSFFS at downstream ofthe first crest, which reveals ‘a greater flow perturbation in that region. The recovery of the shear stress is much faster for DSFTS than that of DVFFS. The po- sition of the maximum shear stress is observed to move in the ‘upward direction as we proceed further downstream from the recirculation region. = Turbulence production greatly complements the results of shear stress and intensities. [is verified that the production of ‘turbulence is mainly concentrated in the separation zone just downstream ofthe fist crest and in the trough region between ‘two adjacent bedform shapes and for the DSFFS case, maxi ‘mum turbulence production is concentrated at the near-bed toe point ofthe second bedform shape. Correlation coefficients show very similar values to the previous measurements by the others. Greater values of Ryy ate observed in the recirculation zone in the range of 050.65, which is a direct result of shear layer development in this region, Spectral analysis showed some evidences that the flow at slownstream of the fist crest, especially in the near-bed zone, recirculation area and in the traugitof the two adjacent bedform shapes is dominated by eddy-like motions. As we proceed to- ‘wards the first crest, flow deceleration or flow separation begins and a shear layer begins to grow. Downstream of the frst rest, neat the bed and at approximately the height of 2 em above the bed, the shear layer becomes unstable and spectral energies ‘row tremendously and confined to amore narrow peak than at ‘upstream and above. Stream-wise power spectral density is ob- served tobe higher than that of vertical and cross-stream spec- ‘raat all the locations along the flow. Spectral peaks, occurring athigher frequency values for v and w than u may be due to the different eddy shape, advection and diffusion process in turbu- lence. Power spectra showed better defined peaks inthe sepa- ration cell where it reaches to a maximum which is 2-3 times {greater than ts values in the upstream and further downstream region, No relict motion is observed inthe present study. Acknowledgments ‘The authors express their sincere thanks to Professor Jeremy G. Venditi, Simon Fraser University (SFU), Canada forhis constructive suggestions during the analysis and the preparation of the paper. ‘Thanks to the Department of Science and Technology (DST), New Delhi for financial support (Project No. SERBS4/ES-127|2004) to conduct the research work: and many thanks to anonymous referee and Editor-in-Chief, Professor Frederic Dias for their detailed and constructive input on the paper. References [1] tare, M.Comes.K Henderson three-mensonalintbity now over ‘hacoarfacing sep Fut ech 473 (2002 1-90 [2] Ki Poote te eseuer TurSulent How ef 2 wscoelasesear-thnnng guid ‘hrousnplane sudden expansion efmodent spectra) Non Newton Sid Mech 1122003) 126. [5] Lasts, Cinna, Thze-dimensional fetes ofthe urulent ow ‘ivough 2 pana sudden expansion, Phys Hus 201) (2008) 01510315, pate [a DA-tyn Tubulence measurements in open chanel flows ever atic bed forms) Hydra ng 19,3) (1993) 305-326. (51S Bent Bes Mea ow antec tae oe edo [61 D4 Patsns, CFS. Wiggs, 1} Walker, Ri Fegaton. BG. Garvey, Nemec 19 2008 ssis2 [7] J Bese The fe dynam ef rver dunes: 2 ewew an some furure esearch ties Geophys Hes. 110 (2005) 02502, Inept ore10.1029/2004Fo00218 [a1 D Fogg c-kit. Albertson 1 dof, An experimental vestigation of ‘usbulent flows ver a lly suace Py Huds, 19 (2007) 0365014112). (a1 SP. oj, BS, Mazumder, Tubuence characteris of flow region over Series f2-b dune shaped structres,Adv Water Resour, 31/2008) 964-376 [n0] VPeee sagate¥ Charon. Tublent ag eduction using siusea rblets ‘with angular cos-secon I Sea AAA Pd yam. Cont xh. 3735, Sestle WA. 2008, [11] TStaeser © Bea, Marea W. Red Tsbulence ‘ver erdimeninal dns jHyasa Eng (2008) 42 [na] BS Marumder Pal KcGhachal SP. tka Trbuence tases of ver Felted resene and vrei angular taped bedrrs,) yérl es 47 5},2008) 625-537 [ns] A Kestavarat | fal H. Nabavl, Reguency pattern of tubule ow and Sediment entrahmentover nipples using nae prcesingHytol Eat 95 Ser (2012) 17-38 [14] RMaramdec, 85. Maramder, Stati characterization of cieuation [15] Kc Mazumeéer Clustering based on geometry and interactions of turbulence Dusting rate processes in a tough tepion Enironmetes 18 (4) (2007 a8 [16] Rater Pid tow paten analysis arourh region: nonparametric ppresch App Sat 356) 2008 633-0, [171 DR Pas Investations on flaw ane sediment movement effected by the Jadavpur University JedeyparKelata= 700052 na (20 [18] KL Simpson Turbulent boundary-layer separ. Us] FA Kota fons LW Soe Compasion betwee ugha [20] PA Krogsad A Antonia, surface roughness eects in turbulent boundary Tages ap. is 271999) 450-850, [21] RA Kostsehk A eld study of turbulence and sediment dynamics aver ‘ubagcous ues wi ow separ, Seientaogy 7 (2000) 519-531 [a2] J, Vena Sf Bentet, Spectral anayss of tsbulent few and suspended Sediment wanspor over fed ines | Geaphys. Res 105 (3) 2000 dussea204 a1 eso say een oy ne ee Re Cie [24] T Ano, Shakouch, Flow character ove ferwacd facing step and {hota conacn pe ab rg edicion hes Re Fac Ea Me [2s] R Camuss, M Foi F Petes, Alisa, AD, Mare, Statistica properties 2008) 7511313 pages, [26] i, Shes. DL Jone, | sheridan. An expesimental investigation ofthe Feceuation zone omed dvrstenm of forward fay sep.) Wine Ene In feroey. 962010} 988-494 [27] ken, ¥ Wu, Turbulent Boundary tyes over smeath an rough frward- Fal steps Py lus 23(@) 2000) oir pas) RL Bedfocm moveren recorded by sequel single-beam surveys Inia svers | Hyeral 258 9002) 25098, [29] A tstebire¥ Fee Al Psoberg V' Ersten C Winter. Variation af foe ‘epatstion ovr arge beers during tial eve: Marine sna River Dane Dynams MARID'¥, Broges, Belgium. 15 ane 18 Api 2012p. 169-175. [30] C' Winter, ¥-Ferzel A Lefebve. Vi. Emstsen. Ceometic properties of Iydathearelevane el bedforms, in: Marine an River une Dyas MIAIDIV 154 16 Api Bruges. Belgium, 2013. pp. 337-338 bu Ti Ner fe opece etm wh Wet one 12] 1 Nez. H. Nakagawa, Turbulence n open channel lows. in: AHR monograph. Balkema Rotterdam, The Nethesande, ISBN 95410-1160, 1983, 133] 8 Gerperte On the nature ofthe actin exresve of theta of aman [54] VB. Emsten, R Noormes, C Winter, D. Hebbeln. A. Barthlama, 8)V. Hemming, Hartoey, Quaication of éune dynamic during 3 tal cee Inanunle channe ofthe Daneh Weeden Se, seat Let 383) (208 125] AlLefebvre, V8. Emtsen,C Winter, Estima Separation over compounebedforas ns Natural Taine, Cont Sel Res ste (201s) e811 [35] ML Agelinghaa. ML Tae, PIV study of separated and eatached open ‘han lw ver sutace mounted Secs ASME} lade Eag 130 (6) (2008 [37] Vi. Nikora, BC Goring, ADV measurements of turbulence: can we improve thet interpretation?) Hydra Eng ASCE 12 (2) (1998) 23-133, [30] Jo, Hine Tuulence, McCraw til Companies, 1875 Hatceover, [38] M.Nelson si Melean Sx Woe Mean‘ow angrurbulenc els over we- ‘Simensional bed forms, Water Resour Ret 2913) 1983) 935-9953, oJ. sa Mea py vag oer sya Ce [41] Si McLean, SR Wolfe, Spatially averaged flow over a wary boundary revista Geophys Re (Ocene 10401 998) 19743-15735, [a2] RS opp, WE New Quishcaheent structs ih thease atmesphete Sound ier} Geophys es 100 (1995) 20055-20508, [43] Go enki DG wats pecteal Analysis ands Applications, Holéen-Day SanFranec, 1968A Srko 8S. Mazur /Eurepeo Journal Mechanics ud (20) 125-148 2 [44] Lev Aunversl croubal law. Eng Mec, 109 (3) (1983) 718-727 [43] Lea, Nakagawa, Tominaga Ae Yoshukawa, Visual sty of agescle ‘orci metiens in opet- channel aw Aan Con japanese oc CEng kansaBanch 1960 [45] Tetakara. Tish Open channel fw with suspenéed segment on sand ‘waves nH ikzava¥ivase (2 Prec. edt Symp. Sach eal. Int Ase of Hydeul Res. Toby 1880 pp 589-809. [a7] Sheds. Asaeda, sedsment suspension with ipled bed, yéraul Eng. 109 fous) ioseans [a8) RAC Kestaschk MA church, Macroturbulence generated by dunes: Haser ive Canada Sediment. Geol #5 (1995 25037 149] fe Vendy Spat and tepors turbulence structure over sub-aqueeus fines eld and aberatoryoxperimetis Mase Thess Univesity afSeubecs cafonnla, 199, p17.