Sediment discharge division at two tidally influenced riverbifurcations
M. G. Sassi,1 A. J. F. Hoitink,1,2 B. Vermeulen,1 and H. Hidayat1,3
Received 10 August 2012; revised 22 March 2013; accepted 25 March 2013; published 24 April 2013..
[1] We characterize and quantify the sediment discharge division at two tidally influencedriver bifurcations in response to mean flow and secondary circulation by employing a boat-mounted acoustic Doppler current profiler (ADCP), to survey transects at bifurcatingbranches during a semidiurnal tidal cycle. The ADCP collecting flow velocity andacoustical backscatter data was used to quantify suspended sediment discharge, adopting arecently introduced calibration procedure. Measured profiles of flow velocity and sedimentconcentration allowed us to compute spatiotemporal distributions of the shear velocity, theroughness length and the Rouse number. Spatiotemporal distributions of the settlingvelocity were obtained by combining the Rouse number and shear velocity estimates with insitu measurements from a laser particle size analyzer. Bed-load transport rates were inferredfrom shear stress estimates. The concentration field shows a direct response to bed shearstress, stressing the alluvial context of the system. The flow in the bifurcation regions ischaracterized by counter rotating secondary-flow cells, which stretch over the full width anddepth of the cross sections in the downstream branches, and persist throughout the entiretidal cycle. The pattern of secondary flow suggests the flow approaching the bifurcation isconcentrated in two independent threads. A two-cell structure inhibits the exchange ofsediment that would occur in case a single cell would stretch over the full channel width.The division of suspended sediment primarily depends on the upstream transverse profile ofthe suspended sediment concentration, which is in turn dependent on geometrical factorssuch as upstream curvature.
Citation: Sassi, M. G., A. J. F. Hoitink, B. Vermeulen, and H. Hidayat (2013), Sediment discharge division at two tidally influencedriver bifurcations, Water Resour. Res., 49, 2119–2134, doi:10.1002/wrcr.20216.
1. Introduction
[2] Channel junctions are key features in tidally influ-enced delta channel networks, and control the division ofwater and sediment discharge over downstream channels.Sediment dispersal, which affects the morphology and theecology of the delta, is controlled by these junctions, whichcan be regarded as river bifurcations fed by alluvial flowsunder the influence of tides. Attempts to route sediment dis-charge in multichannel networks [Fassnacht, 1997] haveoften challenged modelers, mainly because the mechanismsof sediment transport through bifurcations are poorlyunderstood [Fassnacht, 2000]. In tidally influenced sys-tems, the division of water at a bifurcation may not entirelybe controlled by the geometry and hydraulic roughness in
the downstream branches, because tides introduce time-varying Stokes transports and affect the tidal-mean watersurface topography [Buschman et al., 2010; Sassi et al.,2011b, 2012a]. At present, prediction of the partitioning ofwater and sediment under different discharge and tidalregimes is not well understood. This contribution aims tocharacterize the local dynamics of water and sediment dis-charge division at two tidally influenced river bifurcationsin the Mahakam delta in East Kalimantan, Indonesia.
[3] Local hydraulic conditions at bifurcations are influ-enced by secondary flow patterns related to local topogra-phy [Richardson and Thorne, 1998], affecting suspendedload sediment transport. Both in tropical meandering riversand in deltaic systems, the suspended-load is often domi-nant over bed load. Processes driving suspended sedimentdischarge division at bifurcations extend over the full waterdepth [Slingerland and Smith, 1998], and may particularlybe governed by secondary flows causing transverseexchange of sediment. In a confluence-diffluence unit inthe Paran�a River [Parsons et al., 2007], the division of flowis initiated close to where the depth of the central scourreduces, well upstream of the downstream diffluence.Dargahi [2008] found that the existence of multiple second-ary flow fields generated upstream, such as in a river bend,may cause deposition in one of the bifurcating branches. Inmeandering rivers, details of the geometry in the upstreamchannel section as well as secondary circulation can exert a
1Hydrology and Quantitative Water Management Group, WageningenUniversity, Wageningen, Netherlands.
2Department of Physical Geography, Institute for Marine and Atmos-pheric Research Utrecht/IMAU, Utrecht University, Netherlands.
3Research Centre for Limnology, Indonesian Institute of Sciences(LIPI), Cibinong, Indonesia.
Corresponding author: M. G. Sassi, Hydrology and Quantitative WaterManagement Group, Wageningen University, Droevendaalsesteeg 4,Wageningen 6700 AA, Netherlands. ([email protected])
strong control on the division of sediment discharge atbifurcations [e.g., Kleinhans et al., 2008; Hardy et al.,2011; Kleinhans et al., 2013]. In tidal systems, patterns ofsecondary circulation may be continuously adjusting to thetidal surface level gradients. This paper reveals the dynam-ics of secondary circulations in the downstream branches oftwo bifurcations, showing the secondary circulation patternsto be consistent within the cross sections, and persistentthroughout the tidal cycle.
[4] Secondary circulation patterns vary with respect totheir sense of rotation, strength and extent. In an experi-mental river bifurcation, which was not subject to the effectof streamline curvature in upstream bends, Thomas et al.[2011] showed that single secondary flow cells were gener-ated in each distributary, with water flowing toward theinner side of the bifurcation at the surface and toward theouter side at the bed. The flow structure was not affectedby the bifurcation angle and arose as a direct consequenceof the curvature induced by the bifurcation, right after thebifurcation point. In the presence of upstream curvature,the effects of secondary flow induced by upstream curva-ture can overwhelm the effects of local bifurcation charac-teristics [Kleinhans et al., 2008]. Hardy et al. [2011]showed that flow structures generated by the flow in a hy-pothetical meander bend upstream of a bifurcation can belarger than those generated by any geometric configurationof the bifurcation tested, which influences the discharge ra-tio between the two bifurcating channels. Recent experi-mental evidence combined with numerical modeling[Miori et al., 2012] has shown that when bedforms are notpresent, counter-rotating secondary circulation cells maydevelop upstream of the apex of the bifurcation and moveinto the downstream channels, with water converging at thesurface and diverging at the bed. In the same study, second-ary circulation cells extending over the full water depth didnot form when bedforms were dominant [Parsons et al.,2007]. These studies suggest that secondary circulationmay be an important factor determining sediment dischargepartitioning at bifurcations. The results presented hereshow two examples of bifurcations in a sandy environment.In both cases, secondary circulation cells stretch over thefull cross sections in the downstream branches, but with anopposite sense of rotation, reducing their influence on sedi-ment partitioning.
[5] While secondary circulations are a primary agentgoverning advective exchange of suspended sediment, tur-bulence causes diffusive transport. It is often assumed thatthe sediment and momentum diffusivities are equal and in-dependent of depth [van Rijn, 1984]. Experiments in flumes[e.g., Hill et al., 1988, 1998; Cellino and Graf, 1999; Grafand Cellino, 2002; Nikora and Goring, 2002; Muste et al.,2005], however, indicate a wide range of variation around� ¼ 1, where � denotes the ratio of sediment and momen-tum diffusivities. Nikora and Goring [2002] showed that astrong depth-dependence may exist. Graf and Cellino [2002]conducted flume experiments with movable beds and con-cluded that � < 1 for experiments with a flat bed and � > 1for experiments with bedforms, corroborating the conclu-sions drawn by van Rijn [1984] for the case of real rivers.Recent field observations in coastal environments [e.g.,Whitehouse, 1995; Amos et al., 2010] have shown that �may be as low as 0.2 and as high as 3.5, and suggest a
positive correlation with sediment size. Nielsen and Teakle[2004] proposed a finite-mixing-length theory that explainsthe observed trends for � > 1. The model fails to explainthe occurrence of values of � < 1 often found in flat bedexperiments, because the appropriate scale for mixing ofparticles may be somewhat smaller [Muste and Patel, 1997]than the one considered in that model. The proper determina-tion of � plays a crucial role in suspended sediment transportcomputations, because settling velocities derived based onthe gradients in sediment concentration are scaled by a factor�, which leads to a systematic source of uncertainty in thecalculations. Here we present results from an extensive dataset showing that � is systematically greater than unity, indi-cating that sediment diffusivity generally exceeds momentumdiffusivity, in sandy environments with a moving bed.
[6] A description of the field site and instrumentation ispresented in section 2. Section 3 analyses the flow structureat the bifurcations and provides estimates of the bed-loadsediment transport rates. Suspended sediment concentrationand settling velocity results are presented in section 4. Sec-tion 5 shows the division of water and sediment dischargeat bifurcating branches of the two bifurcations. Section 6offers a discussion of the main results. We finalize this con-tribution with the conclusions in section 7.
2. Methodology
2.1. Hydrographic Surveys
[7] Hydrographic surveys were carried out at two bifur-cations in the Mahakam delta, East Kalimantan, Indonesia(Figure 1). Salinity intrusion generally reaches to about 10km seaward from the delta apex. Only during extremelylow flows, such as the El Ni~no-related drought in 1997, sa-linity intrusion can reach beyond the delta apex. The study
Figure 1. Map of the Mahakam delta in East Kalimantan,Indonesia.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
area is therefore generally subject to freshwater conditions.Due to the gentle slope of the river, the tidal wave can propa-gate up to 190 km from the river mouth, depending on theriver discharge. A boat-mounted ADCP surveyed 13 h trans-ects, collecting velocity and backscatter data at the first andthe second bifurcation in the delta (Figure 1), labeled ‘DA’and ‘FB’, respectively. The research boat was equipped witha 1.2 MHz RDI Broadband ADCP measuring in mode 12, amultiantenna Global Positioning System compass operatingin differential mode (D-GPS) and a single-beam echo-sounder. The ADCP measured a single ping ensemble atapproximately 1 Hz with a depth cell size of 0.35 m. Eachping was composed of 6 subpings separated by 0.04 s. Therange to the first cell center was 0.865 m. The boat speedranged between 1 and 3 ms�1. ADCP surveys covered springand neap tidal conditions at both locations, during a floodwave. A summary of the tidally averaged quantities duringthe moving-boat ADCP surveys is presented in Table 1. Allsurveys lasted for approximately 13 h, to cover a semidiurnaltidal cycle. Due to technical difficulties, the survey at thefirst bifurcation during spring tides lasted for about 7 h, cov-ering the rising tide period only.
[8] Prior to the 13 h ADCP surveys, calibration measure-ments were taken to translate the echo intensity measuredby the boat-mounted ADCP and particle size distributionsmeasured with a Laser In Situ Scattering and Transmissom-etry (LISST) to Suspended Sediment Concentration (SSC).We performed the calibration by navigating betweenanchor stations representative of each bifurcating branchduring neap tides. The deepest location represents the entirecross section. In the southern branch, we performed the cal-ibration at two stations corresponding to the main watercourses. The calibration procedure consisted of ADCP sam-pling of echo intensity using the four transducers, while theOBS, the LISST and a Niskin Bottle were winched down toa certain level, where they measured for approximately 2–3min. Within that period, a water sample was taken. Thisprocedure was repeated for different levels at each vertical.Details of the calibration procedure can be found in Sassiet al. [2012b]. After calibration, spatiotemporal distribu-tions of echo intensity measured with the shipborne ADCPwere converted into SSC.
2.2. Determination of the Velocity and SedimentConcentration Fields
[9] Vertical profiles of velocity and SSC were trans-formed to relative height above the bottom according to:
� ¼ z
H þ � ; (1)
where H is mean water depth, z is the height above thebottom, and � is water level variation. Water levels wereobtained from a pressure sensor located nearby. Meanwater level was defined as the mean over a semidiurnaltidal cycle. The variation around mean water level by othercauses than tides (subtidal fluctuations) was negligible(< 0.2 m) in the time-span between ADCP surveys. Along-channel (s) and cross-channel (n) coordinates for eachADCP campaign were defined on the basis of bed morphol-ogy following Hoitink et al. [2009] and Sassi et al. [2011a].Easting and northing coordinates of the depth map wererotated systematically in steps of 0.5�. For each rotationstep, the root mean square deviations from mean values inthe potential s direction were averaged. Depth variationalong the s-coordinate in the first bifurcation was found tobe minimal when it deviated 75� and 140� from the North,for the northern and southern branch, respectively. Simi-larly, depth variation along the s-coordinate in the secondbifurcation was found to be minimal when it deviated 125�
and 185� from the North, for the northern branch and south-ern branch, respectively. The n-coordinate points perpen-dicular to the s-coordinate, with its origin at the inner sideof the bifurcation. We normalized all transects with thetotal width obtained from the intersection of the n-coordi-nate with the riverbanks, to yield a normalized spanwisen-coordinate, �.
[10] In general, the orientation of the s-component coin-cides with the direction of the depth-mean flow. We definea velocity component u0 which is aligned with the depth-mean flow vector, according to:
UffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiU2 þ V 2p : (4)
[12] Variations over the depth of the spanwise velocitycomponent can then be considered as the secondary veloc-ity field. This is justified by the fact that the depth-meanflow (not shown) does not exhibit a significant amount of
Table 1. Summary of the Hydrographic Surveysa
Location Date Tide W mð Þ A m2ð Þ Q m3s�1ð Þ
The first bifurcation North 4 Jan 2009 Neap 540 4480 3470The first bifurcation South 4 Jan 2009 Neap 1040 7040 4920The first bifurcation North 26 Dec 2008 Spring 560 4610 2730The first bifurcation South 26 Dec 2008 Spring 1040 6970 3550The second bifurcation North 3 Jan 2009 Neap 410 2930 1940The second bifurcation South 3 Jan 2009 Neap 660 5280 3290The second bifurcation North 27 Dec 2008 Spring 440 3050 1770The second bifurcation South 27 Dec 2008 Spring 640 5080 2760
aThe hat denotes tidal averaging. The average computed over a time span smaller than a semidiurnal tidal cycle is given in italics.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2121
variation in direction along the n-coordinate. For simplic-ity, we will denote the along-channel and the cross-channelvelocity components as u and v, respectively.
[13] Profiles of u and SSC were projected onto a uniformgrid in (�, �) space. The grid spacing is typically 0.5 m and5 m in the vertical and spanwise directions, respectively.ADCP velocity measurements have contributions of meanflow, turbulence and error components [e.g., Szupianyet al., 2007; Hoitink et al., 2009; Sassi et al., 2011a]. SSCprofiles derived from backscatter strength have similarcontributions. To isolate the mean flow component fromrepeated transect measurements, we assumed the mass fluxthrough (�; �) grid cells to be constant in the streamwisedirection, within the measurement range. Therefore, theproduct of u and H þ �, and the product of SSC, u andH þ �, are both independent of s. Although the latterassumption may be hampered by erosion/deposition proc-esses, we assume that the suspended sediment flux remainsconstant over the limited time-span and along the short spa-tial span of ADCP measurements in the s coordinate (typ-ically< 50 m). The resulting water mass flux time series,multiplied by H þ �, were filtered with a cutoff frequencycorresponding to 1.5 h, subsequently divided by H þ �, andfinally averaged in the s direction over the range that wascovered during the measurements. Hereinafter, u denotesthe mean flow component in the s-direction, obtained fol-lowing the latter procedure. The suspended sediment massflux time series were subjected to a similar procedure. Onaverage, one complete transect covering both bifurcationstook about 30 minutes. Thus, approximately four transectswere navigated every hour at each bifurcating branch.
2.3. Bathymetry Mapping and Analysis of BedSediments
[14] To produce the bathymetric map of the region of in-terest (Figure 2), depth data across the river, collected witha single-beam echosounder, were projected on a curvilineargrid using linear interpolation [Legleiter and Kyriakidis,
2007]. The bathymetry upstream of the first bifurcationshows a meandering thalweg, which continues in the north-ern branch. At the southern branch, an elongated depositio-nal area in the middle of the channel extends over fourkilometers, dividing the channel in two well defined watercourses. About halfway between the first and the secondbifurcation, the southernmost water course splits the elon-gated bar in two parts, marking the start of the northernbranch of the second bifurcation. Transects during springand neap tides were navigated along a predefined line (seeFigure 2).
[15] Bed samples were obtained with a Van Veen grab-ber at locations in the first and the second bifurcation. Sam-ples from 30 cross transects, consisting of five bed sampleseach, were sieved into eleven size classes to obtain theGrain Size Distributions (GSD). Figure 3 shows an interpo-lated map of the median grain size D50 in micrometers. Thespatial distribution of D50 indicates that the river bed ismainly composed of fine to medium sands (D50¼ 200–400�m), with medium sands in the main channels (D50 > 250�m). Riverbanks comprise fine sands and large amounts ofsilt and clay. Patches of irregularly distributed coarser sandare present in the middle of the section, whereas somepatches of very fine material are found along the elongatedbar. The GSD of bed samples collected at the deploymentlocations indicated in Figure 3 are centered in the mediumto fine sand fraction (see Figure 4). Wider distributions,with a large content of very fine sand and silt, characterizesamples from the northern branch of the first bifurcation(site DAN) and the second bifurcation (site FBN), respec-tively. The GSD from a sample taken at DAN is slightlymore concentrated in the coarse fraction, indicating signifi-cant amounts of medium to coarse sand. No significant dif-ference is found between the samples obtained at the twolocations in the southern branches.
[16] Sediment sorting at the upstream bend [Frings andKleinhans, 2008; Kleinhans et al., 2008] may explain therelatively coarse fraction transported toward the northern
Figure 2. Bathymetry (in meters below mean sea level)of the study area (modified after Sassi et al. (2011b)). Navi-gated cross transects to obtain discharge estimates in north-ern and southern channels of the first bifurcation (DA) andthe second bifurcation (FB), respectively, are depicted withthe full lines.
Figure 3. Map of median grain size D50 (�m) of thestudy area. Deployment locations to perform the ADCPbackscatter calibration are depicted with the circles. Fulllines represent remotely sensed riverbanks.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2122
branch at the first bifurcation. Bend sorting typically actson bed-load transport [Parker and Andrews, 1985], whichis likely represented by the bottom sediments that we haveused to construct the D50 map. The GSDs from samples atDAN and FBN confirm this. The GSDs of DAN and FBNsamples also suggest that a relatively fine fraction is trans-ported toward the northern branch at the first and the sec-ond bifurcation, respectively. Samples from other locationsat the southern branches all feature the same GSD, showinga uniform along-channel spatial distribution.
2.4. Estimation of Bed Shear Stress and Bed-LoadTransport Rates
[17] In sand bed rivers such as the Mahakam, bed shearstress can be inferred from the velocity profiles providedthat variation in flow velocity is gradual in time and space.A preliminary analysis showed that effects of spatial andtemporal gradients in flow velocity were visible only dur-ing slack water. For the largest part in the tidal cycle, bedshear stress was therefore estimated from a log-law fit ofthe velocity profile [Hoitink et al., 2009], according to:
u �; �; tð Þ ¼ u� �; tð Þ�
ln �ð Þ þ 1ð Þ þ U �; tð Þ; (5)
where u� is the shear velocity, � � 0:4, and U is the depth-averaged velocity. Estimates of U and u� were obtainedfrom the linear regression of u against ln �ð Þ þ 1ð Þ=�. Withthese estimates, the roughness length z0 results in
z0 ¼H þ �
exp �Uu�þ 1
� � : (6)
[18] Volumetric bed-load transport rates qb were com-puted based on the bed shear velocity estimated asdescribed above, and using the formulation provided byvan Rijn [2007]:
qb / �� � ��;cr
� �1:5; (7)
where �� is the dimensionless bottom shear stress equal tou2� and ��;cr is the critical Shields stress at the threshold of
motion, obtained using the curve published in Parker et al.[2003], both dependent on D50. The bed shear stress calcu-lated from u� represents both form drag and skin frictiondrag, whereas only the skin friction part results in sedimenttransport. Consequently, the estimated bed-load transportrates calculated herein can be considered upper bounds.
2.5. Analysis of Suspended Sediment
[19] To investigate settling velocity, a Rouse functionwas fit to the SSC profiles, assuming the diffusive and ad-vective sediment flux in the vertical direction to be largelyin balance. Even with lag effects, the SSC profile mayexpected to be dominated by the latter balance. Accord-ingly, suspended sediment concentration (c) profiles can beapproximated by the power law:
c zð Þ ¼ crz
H � z
H � zr
zr
� ��p
; (8)
where c is the mass concentration (kg m�3) at a height z (m)above the bottom, cr is a reference concentration at zr, withzr typically assumed to coincide with the upper boundary ofthe bed-load layer, and p is the Rouse number. The Rousenumber is defined as
p ¼ ws
�u�¼ wf
��u�; (9)
where ws is the settling velocity obtained from the fit(m s�1), wf is the fall velocity of the particles in suspension(m s�1) and � is the inverse of the turbulent Prandtl-Schmidt number, defined as the ratio of sediment diffusivityks to momentum diffusivity kz. The momentum diffusivityis given here by:
kz ¼ �u�z 1� z
H
� �: (10)
[20] Equation (8) can be rewritten in terms of the nor-malized depth, such that:
ln c �; �; tð Þð Þ ¼ �p �; tð Þln �
1� �
� �þ f ; (11)
where f is a function that depends on the reference concen-tration cr, the relative reference height �r, and p. Estimatesof p, the Rouse number, are then obtained from a linearregression of ln cð Þ against ln �=1� �ð Þ.
[21] From the estimates of p and u�, spatiotemporal dis-tributions of the settling velocity ws were readily derived.The magnitude of the settling velocity depends crucially onthe value of �. Using data from the ADCP calibration sur-veys, in situ determinations of the mean particle size withthe LISST instrument were used to compute the particlefall velocity [Cheng, 1997; Camenen, 2007]:
Figure 4. (left) Grain Size Distribution (GSD) of bedsamples obtained at locations indicated in Figure 3, in�-scale (� ¼ �log2 D=D0ð Þ where D is the size class andD0 corresponds to 1 mm). (right) D50 as a function of nor-malized width �, along the transects depicted in Figure 2.� varies from 0 to 1 corresponding to the riverbanks in theinner side and the outer side of the bifurcations, respec-tively (see section 3).
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2123
where is the kinematic viscosity of water (10�6 m2 s�1),d is the sediment diameter (m), a ¼ 32;B ¼ 1;M ¼ 1:5,and d� is the dimensionless particle size. The latter isdefined as
d� ¼g�
2
� �1=3
d; (13)
where g is the gravitational acceleration (9.81 m s�2), isthe density of water (1000 kg m�3) and � ¼ s � is theapparent density, with s the density of the suspended ma-terial. Volume concentration V measured by the LISST canbe used with in situ determinations of mass concentrationto compute the under water apparent density [Gartner andCarder, 1979; Mikkelsen and Pejrup, 2001]:
� ¼ s � ¼Ms
V; (14)
where Ms is mass concentration in water samples. Equation(12) is based on the two asymptotic functions of the drag coef-ficient for low- and high-Reynolds numbers, and is valid forparticles of different shape and roundness [Camenen, 2007].
3. Flow Pattern
3.1. Three-Dimensional Velocity Field
[22] Figure 5 shows the flow structure at the first bifurca-tion during spring tides. The figure covers a time span
between two moments of high tide. The velocity core islocated on the surface at both bifurcating branches, however,its position slightly differs between the branches. The second-ary velocity field in both bifurcating branches exhibits thesame pattern as in the first bifurcation (Figure 6). In the south-ern branch of first bifurcation, the secondary velocity fielddepicts also a zone at about � ¼ 0:2 with reversed orientationof the flow, and complex velocity profiles, during part of thetidal cycle. This three-dimensional velocity pattern may beassociated with the complex morphology along the cross sec-tion. During neap tides (not shown), the flow structure is notsignificantly different, but a bi-directional flow pattern arises inthe southern branch, which lasts for about three hours. At bothlocations, secondary flow fields are consistent with the curva-ture of the riverbanks. The persistent secondary flow fields inresponse to local channel curvature may explain the cross-chan-nel bed level profiles, which are deeper in the outer bends.
3.2. Vertical Profiles of Along-Channel Velocity
[23] Spatial distributions of the tidally averaged shear ve-locity u� (Figure 7) are similar between neap and springtides at both bifurcating branches, and are consistent withthe location of the velocity core. The magnitude of the ve-locity peak is smaller during spring tides, which may beassociated to the lack of a sufficiently long averaging period,since this survey lasted for about 7 h. The tidal-mean rough-ness length z0 was obtained by integrating equation (5) overtime [Hoitink et al., 2009], assuming z0 does not vary
Figure 5. Temporal sequence of u (m s�1) with superposed secondary velocity field at (left) North and(right) South transects at the first bifurcation during spring tides as a function of depth and normalizedwidth. The downstream flow direction is pointing into the paper. Dashed lines indicate water levelwhereas full lines indicate the bottom. Bottom profiles have been linearly extrapolated toward the bank.The vertical coordinates indicate z – H (m).
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2124
significantly throughout a tidal cycle. This is supported bythe fact that flow reversed only for a short period during themeasurements. The similarity in spatial distributions of z0 -profiles between neap tide and spring tide over the northernbranch is high. z0 profiles also depict an increase toward the
outer side of the channel, which resembles the distributionof bed sediments shown in Figure 3. In the southern branch,z0 in the range 0:2 < � < 0:5 is lower during spring tidethan during neap tide, consistent with a decrease in u� from� ¼ 0:1� 0:5.
Figure 6. Temporal sequence of u (m s�1) with superposed secondary velocity field at (left) North and(right) South transects at the second bifurcation during neap tides as a function of depth and normalizedwidth. The downstream flow direction is pointing into the paper. Dashed lines indicate water levelwhereas full lines indicate the bottom. Bottom profiles have been linearly extrapolated toward the bank.The vertical coordinates indicate z – H (m).
Figure 7. Tidally averaged shear velocity u� obtained from the fit to each u profile, and roughnesslength z0 obtained by integrating equation (5) over time, both during neap (black) and spring (red) tides,at bifurcating branches in the first bifurcation (DA) and the second bifurcation (FB). Smooth linesremove spatial variations of the order of � ¼ 0:1.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2125
[24] Profiles of the shear velocity at both branches of thesecond bifurcation are similar at neap tide and spring tide(Figure 7). In the northern branch, u� in the region 0:3 <� < 0:5 is lower in magnitude during spring than duringneap tide. In the southern branch, values of u� at about0:1 < � < 0:4 are higher during spring tide than duringneap tide. Spatial distributions during neap tide and springtide reveal a region 0:1 < � < 0:4ð Þ with low magnitudesof z0 in the northern branch, which can be related to thegrain size distribution of bed sediments (see Figure 3). Inthe southern branch, z0 during spring tide is higher thanduring neap tide. Finally, for 0:4 < � < 0:6; z0 duringspring tide is one order of magnitude lower than duringneap tide, consistent with the decrease in u� .
3.3. Bed-Load Transport Rates
[25] Figure 8 shows spatiotemporal distributions of qb
(kg m�1 s�1) at bifurcating branches at the first and the sec-ond bifurcation, during neap tide. Temporal variations in qb
are mainly controlled by the semidiurnal tide, with verylow transport rates during high water slack, and higher ratesduring the onset of rising tide. The width-integrated bed-load transport rate (Qb) varies in between nearly zero up toa maximum of about 100 kg s�1. Estimated bed-load trans-port at the second bifurcation is more equally divided overthe distributaries than at the first bifurcation. Spatial distri-butions of the tidally averaged bed-load transport rate, qb ,resemble the spatial distributions of u� , despite the spatialvariation in D50 across the channel.
4. Suspended Sediment Concentration
4.1. Spatiotemporal Distribution
[26] Figure 9 shows spatiotemporal distributions ofSSC (in mg l�1) at bifurcating branches at the first bifurca-tion during spring tides. Temporal variations in SSC are
primarily controlled by the semidiurnal tide, with concen-trations below 20 mg l�1 during high water and up to 180mg l�1 during low water. The largest concentrations aretypically found near the bottom, and values of SSCdecrease toward the surface. The spatial distribution ofSSC in the southern branch shows relatively high concen-trations near the banks. In the northern branch, high con-centrations are also found in the middle and toward theouter side of the channel. The spatial distribution of SSC islinked to the general secondary circulation pattern (seeFigure 5). During neap tides (not shown), due to the smallertidal range, temporal variations are highly reduced, withrelatively high concentrations (> 50 mg l�1) at high water.The spatial distribution at neap tides remains the same as dur-ing spring tides, although in the southern branch it features anearly uniform high-concentration region near the bottom.
[27] Figure 10 shows SSC plots at bifurcating branchesfor the second bifurcation, during neap tides. The largestconcentrations appear toward the inner side of both bifurcat-ing channels. In the southern branch, a region near the outerside of the channel also shows high concentrations near thebottom. The region 0:4 < � < 0:6 shows very low concen-trations. During the surveys, at this part of the cross section,we observed a strong decrease in the echo intensity of theADCP, as well as the presence of oil at the water surface. Itappears that the oil, or some other component in the water,may have had a significant influence on the absorption ofthe ADCP acoustic waves, which has not been accountedfor during the computation of the backscatter. Observationsduring spring tide (not shown) indicate a similar spatial dis-tribution of SSC, except for poorly defined concentrationprofiles during the 3 h that the bi-directional flow lasts.
4.2. Concentration Profile Fitting
[28] In the southern branch, the Rouse number p remainsrelatively constant throughout the cross section and decreases
Figure 8. Spatiotemporal distribution of bed-load transport rate qb (kg m�1 s�1) at (left) northern and(right) southern branches in the first bifurcation (DA) and the second bifurcation (FB) during neap tides.Also shown are the width-integrated bed-load transport Qb as a function of time, and tidally averagedbed-load transport qb as a function of normalized width. The vertical coordinate indicates time since thestart of the measurements.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2126
toward the outer side of the channel (Figure 11). Similar val-ues of p are obtained in the northern branch at 0 < � < 0:6,peaking at approximately � ¼ 0:8, and decreasing towardthe outer side of the channel. p-values typically lie in theinterval between zero and one, which is consistent with theresults reported by van Rijn [1984]. The tidally averaged rel-ative Root Mean Square Deviation (rRMSD) between the ob-servation and the reconstructed profiles using equation (11),indicates that in both channels, on average, the error is 5%–10% of the depth-averaged value, except for the profilesaround � ¼ 0:8 in the northern branch, where the error mayincrease up to 20%. At the second bifurcation, p valuesincrease toward the outer side of the northern branch, whiledecreasing toward the banks. At the southern branch, profilesof p show variations, which are consistent with variations inthe bottom topography. The maximum error at both crosssections amounts to 15%, but in general, rRMSD valuesremain below 10%.
4.3. Settling Velocity Estimates
[29] Figure 12 shows time series of ~wf , computed within situ LISST data obtained at three locations at the firstbifurcation (see Figure 2). The mean particle size rangesfrom 100 to 150 �m and the apparent density ranges from900 to 1500 kg m�3. Values of wf show a significant varia-tion in time that is well correlated with the tidal cycle. Thisconfirms that tidal resuspension processes govern the
dynamics of suspended sediment transport at the locationsunder study.
[30] Bulk estimates of � as inferred from the LISSTinstrument can be obtained from a direct comparisonbetween wf and the corresponding settling velocity ws
retrieved from the fit to the profiles, using equation (9) with� ¼ 1. Figure 13 shows the bulk estimates of � and timeseries of wf and hwsi, obtained by averaging over a suitablewidth over which the in situ measurements were taken.Bulk estimates of � yield values greater than one and differbetween channels at the first bifurcation: 2.36 for the north-ern branch, and 1.26 and 1.59 for the two locations in thesouthern branch. We performed the same analysis with themeasurements at the second bifurcation (not shown): esti-mates of � yield 2.69 for the northern branch, and 1.29 and1.46 for the two locations in the southern branch.
[31] We computed spatiotemporal distributions of settlingvelocity derived from the concurrent fit to velocity and SSCprofiles, using the estimates of � as described above (Figure14). For the southern branches, the two estimates of � wereaveraged. During neap tides, estimates near the outer side ofthe northern branch are well correlated with flow strength.Maximum values of ws reach 30 mm s�1. For � < 0:6, ws
remains relatively constant throughout the tidal cycle. In thesouthern branch, spatial variations in ws depict a zone of rel-atively low magnitude around � ¼ 0:4, whereas temporalvariations are rather limited. Tidally averaged values depictcontrasting spatial variations between bifurcating branches.
Figure 9. Temporal sequence of SSC (mg l�1) at (left) North and (right) South branches in the firstbifurcation during spring tides as function of depth and normalized width. Dashed lines indicate waterlevel whereas full lines indicate the bottom. Bottom profiles have been linearly extrapolated toward thebank. The vertical coordinates indicate z – H (m).
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2127
Figure 10. Temporal sequence of SSC (mg l�1) at (left) North and (right) South branches in the secondbifurcation during neap tides as function of depth and normalized width. Dashed lines indicate waterlevel whereas full lines indicate the bottom. Bottom profiles have been linearly extrapolated toward thebank. The vertical coordinates indicate z – H (m).
Figure 11. Tidally averaged Rouse number p obtained from the fit to each SSC profile and relativeRoot Mean Square Deviation rRMSD between the observation and the reconstructed profile using equa-tion (11), both during neap (black) and spring (red) tides. Smooth lines were used to remove spatial var-iations over � window of about 0.1.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2128
At regions of relative constancy through time, ws averagesto about 5–7 mm s�1 in both channels, which correspondsto fine sands [Cheng, 1997]. At the second bifurcation dur-ing neap tides, ws estimates generally show a clear responseto flow strength. ws distributions show relatively large val-ues across the northern branch, whereas at the southernbranch, a zone of relatively low values arise at about�¼ 0.4–0.5.
5. Sediment Discharge Division
[32] Total sediment discharge Qs was calculated by add-ing Qb to the product of u and c integrated over depth andwidth. Qb is typically within 10% of the suspended sedimentdischarge. Water discharge was computed by integrating u
over depth and width. We neglect the unmeasured areas nearthe surface and the channel boundaries, because extrapola-tion may introduce error and it is unlikely the blanking areasinfluence the overall dynamics. Figure 15 illustrates the rela-tion between Qs and Q for northern and southern branches atthe first and the second bifurcation, during neap tides andduring spring tides. The nearly closed loop in all dischargerelations stems from the semidiurnal tide. During springtide, the loop is stretched, although the opening of the hys-teresis loop seems to remain the same, suggesting that thephase lag between Qs and Q remains constant between neaptide and spring tide. At the second bifurcation, tidal effectsare clearly stronger as both Q and Qs approach zero duringspring tides, in contrast to the first bifurcation.
[33] The nonlinear behavior between Qs and Q can becaptured in the relation [Bagnold, 1966]:
qs tð Þ / UA t � �ð Þð Þb; (15)
where qs stands for specific sediment discharge, calculatedas Qs per unit width, UA ¼ hUi denotes the section- anddepth-averaged flow velocity, � is a time-lag function, andb is an empirically derived exponent. The time lag in equa-tion (15) may be explained by the relaxation model ofGroen [1967], which states that the rate of increase ordecrease of the suspended load at any time is proportionalto the deficit or excess of the load with respect to an equi-librium value. Table 2 shows a summary of the parametersin equation (15), determined empirically by fitting a line tothe log-transformed variables. We computed the best-fitlines for time lags varying from �3 to 3 h. The time-lag �corresponds to the best-fit, which was based on the coeffi-cient of determination R2 (greater than 0.95 in all cases).All derived exponents exceed 2.5. b is consistently greaterduring spring tides, indicating that tides enhance the degreeof nonlinearity in the sediment transport relation (equation(15)). The time lag remains nearly constant in the northernbranches, whereas it alternates between neap and springtides in the southern branches.
[34] Figure 16 illustrates the intratidal variation in divi-sion of q, qs, and qb over the bifurcating branches at thefirst and the second bifurcation. At the first bifurcation,more water and sediment discharge per unit width isdirected toward the northern branch, which is most pro-nounced during high discharges. The division at the secondbifurcation is more symmetrical, and depicts some cross-overs regarding the branch that receives the largest specificdischarge: for low discharges (high tide) a relatively largershare is directed to the northern branch and for high dis-charges (low tide) a larger share goes to the southernbranch. The division functions of water and suspendedsediment are qualitatively similar, being different from thedivision function of bed-load sediment transport.
[35] The differences between locations cannot simply beexplained by the ratio between the areas of the bifurcatingchannels, which are similar (Table 1). Upstream conditionsmay exert a strong control on the division functions. Thefirst bifurcation is subject to flow from a curved channelwith a nonuniform depth. A deep section in the northernpart of the channel increases the local transport capacity,enhancing the supply of water and sediment to the northernbranch. Upstream conditions at the second bifurcation are
Figure 13. (left) Time series of wf derived from LISSTmeasurements and corresponding hwsi derived from the fitto the profiles and averaged over a suitable width that spansthe location of the in situ measurements, for (top) North,(middle) Center, and (bottom) South locations. Smoothlines were used to remove temporal variations of the orderof 1.5 h. (right) Bulk estimates of � from a direct compari-son between the estimated settling velocities. The dottedline indicates the best-fit line forced through the origin,with a slope given by �.
Figure 12. Particle fall velocity averaged over depth (wf),computed using equation (12) at locations in the first bifur-cation (see Figure 2). Error bars depict one standard devia-tion around the depth-mean value.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2129
different, featuring a smaller curvature of the channel and acomplex bed topography shaped by the intersected mid-channel bar.
[36] The crossovers in dominance of one branch over theother may be partly related to intratidal variations in waterdischarge and sediment transport, and small lag effects.Intratidal variations are related to complex nonlinear inter-actions between the flow and the bathymetry, driving tidalasymmetry, which can exist in individual channels. In eachbranch at a bifurcation, the phase of tidal constituents can
be slightly different, which will be manifest as a periodicasymmetry of the sediment discharge division. A 13 h pe-riod observation is too limited to distinguish between peri-odic and permanent asymmetries in the division ofsediment discharge, because the period of diurnal and fort-nightly tidal constituents is much longer.
6. Discussion
[37] The flow in the bifurcations under analysis is char-acterized by counter-rotating, secondary-flow cells, whichpersist throughout the tidal cycle. The short length overwhich the secondary flow patterns of the downstreambranches develop, which can be inferred from the channelcurvature, imply that the parallel flows reaching the bifur-cation are largely independent. If the secondary circulationwould have the same orientation in both branches, the sec-ondary circulation could cause sediment exchange betweenthe parallel flow lanes in the region immediately upstreamof the bifurcation apex. The two-cell structure inhibits suchexchange, and causes the division of suspended sediment tobe strongly dependent on local flow processes, largely gov-erned by the tide. The adaptation length of the secondaryflow, defined as the distance over which the secondary flowdevelops in response to the driving force due to inertial
Figure 14. Spatiotemporal distribution of settling velocity ws (mm s�1) at (left) northern and (right)southern transects at the first bifurcation (DA) and the second bifurcation (FB) during neap tides. Alsoshown the width-averaged depth-mean velocity hUi as a function of time, and tidally averaged settlingvelocity ws as a function of normalized width. The vertical coordinate indicates time since the start ofthe measurements.
Figure 15. Total suspended sediment discharge Qs as afunction of water discharge Q at (left) northern and (right)southern branches at the (top) first bifurcation and (bottom)second bifurcation during neap and spring tides.
Table 2. Summary of Parameters Using Equation (15)a
aThe variability denotes the standard error in the linear regression.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2130
effects [Johannesson and Parker, 1989] appears to bewithin the range between the onset of the local flow curva-ture and the bifurcation apex. Although laboratory experi-ments showed that the adaptation lengths for secondaryflows are typically within a quarter of the meander length[Zhou et al., 1993], Johannesson and Parker [1989] arguedthat in natural meandering rivers, the adaptation length issignificantly reduced to about a tenth of a meander length.This suggests the secondary flow observed in the northernbranch at the first bifurcation has adapted long before enter-ing the northern branch. It would be interesting to monitorthe transition from a one-lane to a two-lane flow structure,which is where an exchange process of suspended sedimentmay occur. This transition may be gradual, and it is uncer-tain where this region exactly occurs.
[38] A persistent feature in the bathymetry of the south-ernmost distributary channel is an elongated depositionalarea at about the middle of the channel, which separates twowater courses along that channel. The map of the D50 (seeFigure 3) shows that patterns of medium to coarse sands fol-low the thalweg of the river, advancing through the firstbifurcation to the North and to the South, and leading to twowell-defined sediment pathways along the southern channel.Despite that bed-load transport rates are significantly smaller
than the suspended sediment transport rates, these amountsof bed-load transport can be substantial for the developmentof bars, affecting the morphology of the bifurcations.Crosato and Mosselman [2009] developed a simple physics-based predictor to discriminate between river patterns on thebasis of the mode m of the wavenumber describing the trans-versal oscillation of the river bed. Based on the linear modelof Struiksma et al. [1985], they derive m from the ratiobetween the adaptation length for perturbations in the trans-verse profile of depth-averaged streamwise flow velocity andthe adaptation length for perturbations in the cross-sectionalriver bed profile. Theoretically, m ¼ 1 represents channelswith alternate bars, whereas m ¼ 2 indicates the presence ofcentral bars; m > 3 would imply more complex transversebed level profiles. The transverse mode is given by
where is the width to depth ratio of the channel, b is thesame exponent as in equation (15), with b > 3, and
� ¼ s � ð Þ=. Using Cf ¼ u2�
U 2 and width-averaged valuesfor H and D50; and all other terms in Eq. (16) are readily
Figure 16. Division of specific water discharge q (m2s�1), total sediment discharge qs (kg m�1s�1),and bed-load sediment transport qb (kg m�1s�1) between bifurcating branches at the first bifurcation(DA) and the second bifurcation (FB) during neap and spring tides. The dashed line indicates the line ofperfect agreement.
Table 3. Summary of Transverse Modes m Obtained Using Equation (16)a
aThe brackets denote width averaging, the variability is given by one standard deviation around the width-averaged value.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2131
computed, and the results from section 3.2 can beemployed to estimate m. Table 3 shows a summary of theresults for each location during spring tides (b > 3), con-firming the theory by Crosato and Mosselman [2009] forthe first bifurcation. The discrepancy at the second bifurca-tion may be partly related to the location of the transect,which is further downstream of the actual section where themidchannel bar occurs. The nonlinear theory by Blanckaertand de Vriend [2010] may predict lower values for the adap-tation length of the flow, consequently affecting the outcomeof equation (16). However, the ratio between channel widthand radius of curvature, central in the approach by Blanck-aert and de Vriend [2010], can be considered too small fornonlinear curvature effects to be of any significance.
[39] Although field estimates of the ratio between sedi-ment to momentum diffusivity are scarce, several studiessuggest considerable variation around unity [e.g., Hill et al.,1988, 1998; Whitehouse, 1995; Amos et al., 2010]. In gen-eral, a positive correlation is found between � and sus-pended sediment size, which critically depends on the flowconditions [Nielsen and Teakle, 2004]. van Rijn [1984]showed that a functional dependency exists between � andthe ratio ws=u�, where � always exceeds ws=u�. The expres-sion by van Rijn [1984] was later corroborated by Graf andCellino [2002] in experiments with a moving bed and bed-forms. Graf and Cellino [2002] also showed that for experi-ments with a flat bed, � was typically lower than unity.Field determinations of the relation between � and the ratiows=u� can be cast in power law relations with a wide rangeof exponents [e.g., Whitehouse, 1995; Kawanisi and Yokosi,1997; Hill et al., 1988, 1998; Rose and Thorne, 2001]. Ourobservations in a moving bed, sandy environment, providefield evidence showing � can exceed unity significantly.Neglecting lag effects in the procedure to estimate settlingvelocity from concentration profiles can only slightly influ-ence the estimates of �. The estimated high values of �qualitatively agree with findings in the existing studies. Thecorrelation between � and the ratio wf =u� was, however,only weak (R2 ¼ 0:21 based on six observations). To someextent, our bulk estimates of � can be impacted by the inex-act collocation of the measurements. Current research is fo-cusing on rigid deployments with upward looking ADCPs[Vermeulen et al., 2011], which may provide refined fieldestimates of �, as the ADCPs yield simultaneous, collocatedestimates of flow velocity and suspended sedimentconcentration.
[40] Despite the process of suspended sediment transportbeing intrinsincally three-dimensional, modeling this pro-cess in rivers and estuaries can be accomplished by depth-integrated, two-dimensional models [e.g., Galappatti andVreugdenhil, 1985; Wang, 1992]. The increased computa-tional cost in adding the vertical dimension is often regardedas too high, in particular when the morphology is allowed tocoevolve with the flow [Wang and Ribberink, 1986; Tal-mon, 1992]. In river bends and in bifurcations, the increasedcomplexity of the flow due to secondary circulations maycall for a fully three-dimensional approach. In the two bifur-cations under analysis, we show that even though secondarycirculation arises as a consequence of the curvature of thebifurcating branches, the rapid adaptation of the secondaryflow causes the parallel flows at the bifurcation to act nearlyindependently. This flow pattern, inhibiting the exchange of
sediment between flow lanes, suggests that in the Mahakamthe three-dimensional effects of the suspended sediment arelimited, and restricted to an upstream region.
[41] From a long-term perspective on alluvial morphol-ogy, local conditions at bifurcations adjust such that the localdischarge division at the bifurcation is compatible with thegeometry and roughness conditions in the downstreambranches. There are several reasons why the discharge divi-sion may not be predictable from roughness and geometry inthe downstream branches, including dynamic effects bytides. In tidal channel networks, the instantaneous divisionof discharge depends on the local surface level topographyupstream of the bifurcation, and so does the tidal-averageddischarge division. Tides have a nonnegligible control onsurface topography [Buschman et al., 2009, 2010; Sassiet al., 2011b, 2012a], which in turn depends on the geometryboth upstream and downstream of a channel bifurcation.From numerical modeling results obtained in a previousstudy (see Figure 12 in Sassi et al. [2012a]) we can concludethat the ratio between surface level elevation differences towater depth is approximately between 5% and 10%, takingmean channel depth as the reference. Consequently, hydrau-lic theory on gradually varied flow does not directly apply,and existing work on river bifurcation stability, such as initi-ated by Wang et al. [1995] and extended by Bolla-Pitalugaet al. [2003], cannot readily be extended to the case withtides. Results from the bifurcations studied in this manu-script suggest it can be promising to use a depth-averagedmorphological model for the latter purpose, in which settlinglag and scour lag effects are being accommodated implicitly[Galappatti and Vreugdenhil, 1985]. Such a model wouldallow to upscale results from measurements over a tidalcycle to time periods relevant to bifurcation stability.
7. Conclusions
[42] Transects surveyed with a boat-mounted acousticDoppler current profiler (ADCP) at bifurcating branchesduring a semidiurnal tidal cycle were used to characterizeand quantify the sediment discharge division at two tidallyinfluenced river bifurcations. The ADCP collecting flowvelocity and acoustical backscatter data was used to quan-tify bed-load and suspended load transport, adopting arecently introduced calibration procedure to transformacoustic backscatter into mass concentration of suspendedsediments. We draw the following conclusions:
[43] (1) The concentration field at the bifurcatingbranches shows a direct response to the bed shear stress,indicating that washload is subordinate and the system canbe considered alluvial. The lack of complex hysteresis loopsor phase lags allows to determine spatiotemporal distribu-tions of the settling velocity by combining the Rouse numberand the shear velocity, determined from a fit to the concen-tration and the velocity profiles (respectively). Temporal var-iations in settling velocity are strongly correlated with theflow strength whereas spatial variations can be readily linkedto variations in median grain size of the bottom sediments.
[44] (2) In spite of lag-effects, settling velocity can beestimated from vertical profiles of suspended sediment con-centration, which feature a Rouse distribution. Althoughthe settling velocity is predictable, complexity is introducedby the fact that the ratio between momentum diffusivity
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2132
and sediment diffusivity differs significantly from the valuethat is commonly assumed for Rouse profiles. The bulkestimates of the latter quantity qualitatively agreed withresults reported in studies focusing on moving bed, sandyenvironments, which show that the ratio is generally greaterthan unity, and increases with the ratio between settling ve-locity and shear velocity.
[45] (3) The flow in the bifurcating branches is character-ized by counter-rotating, surface-convergent secondaryflow cells, which persist throughout the entire tidal cycle.The secondary flow structure suggests the parallel flowsapproaching the bifurcation act largely independently. Thistwo-cell structure inhibits the exchange of suspended sedi-ment that would occur in case the cell would stretch overthe full width of the feeding channel in the region upstreamof the bifurcation apex. The division of suspended sedimentprimarily depends on the upstream transverse profile of thesuspended sediment concentration, which is in turn depend-ent on details of the tidal motion and geometrical factorssuch as upstream curvature. The three-dimensional effectsof the suspended sediment dynamics on sediment dischargedivision at the two tidally influenced bifurcations presentedhere are shown to be limited.
[46] (4) Bed-load transport rates estimated on the basisof the shear velocities and the median grain size remainednearly within 10% of the suspended sediment load. Thetotal suspended load depicts a nonlinear relation with theflow, with the degree of nonlinearity increasing duringspring tides. Time lags occur in all cases. The division oftotal sediment discharge per unit width follows closely thedivision of specific water discharge, which is different fromthe division of bed-load sediment. In general, a greater spe-cific discharge was directed toward the northern branch atthe first bifurcation, whereas the division remained rela-tively equal at the second bifurcation. These results wereattributed to the characteristics of the flow inherited fromthe upstream region. The division of water, in turn, doesnot only depend on the geometry and roughness of thedownstream branches, but also on the tidal impact on subti-dal surface level profiles and Stokes transports.
[47] Acknowledgments. This study is part of East Kalimantan Pro-gramme, supported by grant number WT76-268 from WOTRO Science forGlobal Development, a subdivision of the Netherlands Organisation for Sci-entific Research (NWO). We thank the Ministry of Public Works (Rijkswa-terstaat) for the lending of the LISST-100 instrument. We thank JohanRomelingh and Pieter Hazenberg (Wageningen University) for excellenttechnical support. David Vermaas, Unggul Handoko, Fajar Setiawan andthe captain of the research vessel Tahang are acknowledged for their helpduring data collection. Yohannes Budi Sulistioadi and Wawan Kustiawanare thanked for facilitating the fieldwork campaign in many ways.
ReferencesAmos, C. L., et al. (2010), The measurement of sand transport in two inlets
of Venice lagoon, Italy, Estuar. Coast. Shelf. Sci., 87(2), 225–236.Bagnold, R. A. (1966), An approach to the sediment transport problem
from general physics, U.S. Geol. Surv. Prof. Pap., 422(1), 1–37.Blanckaert, K., and H. J. de Vriend (2010), Meander dynamics: A nonlinear
model without curvature restrictions for flow in open-channel bends, J.Geophys. Res., 115, F04011, doi:10.1029/2009JF001301.
Bolla-Pitaluga, M., R. Repetto, and M. Tubino (2003), Channel bifurcationin braided rivers: Equilibrium configurations and stability, WaterResour. Res., 39(3), 1046, doi:10.1029/2001WR001112.
Buschman, F. A., A. J. F. Hoitink, M. Van Der Vegt, and P. Hoekstra(2009), Subtidal water level variation controlled by river flow and tides,Water Resour. Res., 45, W10420, doi:10.1029/2009WR008167.
Buschman, F. A., A. J. F. Hoitink, M. Van Der Vegt, and P. Hoekstra(2010), Subtidal flow division at a shallow tidal junction, Water Resour.Res., 46, W12515, doi:10.1029/2010WR009266.
Camenen, B. (2007), Simple and general formula for the settling velocity ofparticles, J. Hydraul. Eng., 133(2), 229–233.
Cellino, M., and W. H. Graf (1999), Sediment-laden flow in open-channelsunder noncapacity and capacity conditions, J. Hydraul. Eng., 125(5),455–462.
Cheng, N. (1997), Simplified settling velocity formula for sediment parti-cle, J. Hydraul. Eng., 123(2), 149–152.
Crosato, A., and E. Mosselman (2009), Simple physics-based predictor forthe number of river bars and the transition between meandering and braid-ing, Water Resour. Res., 45, W03424, doi:10.1029/2008WR007242.
Dargahi, B. (2008), Mitigation of sedimentation problems in the lowerreach of the river Klaralven, J. Hydraul. Res., 46(2), 224–236.
Fassnacht, S. R. (1997), A multi-channel suspended sediment transportmodel for the Mackenzie Delta, Northwest Territories, J. Hydrol., 197,128–145.
Fassnacht, S. R. (2000), Flow modelling to estimate suspended sediment traveltimes for two Canadian Deltas, Hydrol. Earth Syst. Sci., 4(3), 425–438.
Frings, R. M., and M. Kleinhans (2008), Complex variations in sedimenttransport at three large river bifurcations during discharge waves in theriver Rhine, Sedimentology, 55, 1145–1171.
Galappatti, G., and C. B. Vreugdenhil (1985), Depth-integrated model forsuspended sediment transport, J. Hydraul. Res., 23(4), 359–377.
Gartner, J., and K. Carder (1979), Method to determine specific gravity ofsuspended particles using an electronic particle counter, J. Sediment.Pet., 49, 631–633.
Graf, W. H., and M. Cellino (2002), Suspension flows in open channels; ex-perimental study, J. Hydraul. Res., 40(4), 435–448.
Groen, P. (1967), On the residual transport of suspended matter by an alter-nating tidal current, Neth. J. Sea Res., 3(4), 564–574.
Hardy, R. J., S. N. Lane, and D. Yu (2011), Flow structures at an idealizedbifurcation: A numerical experiment, Earth Surf. Process. Landforms,36(15), 2083–2096.
Hill, P. S., A. R. M. Nowell, and P. A. Jumars (1988), Flume evaluation ofthe relationship between suspended sediment concentration and excessboundary shear stress, J. Geophys. Res., 93(C10), 12,499–12,509,doi:10.1029/JC093iC10p12499.
Hill, P. S., J. P. Syvitski, E. A. Cowan, and R. D. Powell (1998), In situobservations of floc settling velocities in Glacier Bay, Alaska, Mar.Geol., 145(1-2), 85–94.
Hoitink, A. J. F., F. A. Buschman, and B. Vermeulen (2009), Continuousmeasurements of discharge from a Horizontal ADCP in a tidal river,Water Resour. Res., 45, W11406, doi:10.1029/2009WR007791.
Johannesson, H., and G. Parker (1989), Velocity redistribution in meander-ing rivers, J. Hydraul. Eng., 115(8), 1019–1039.
Kawanisi, K., and S. Yokosi (1997), Characteristics of suspended sedimentand turbulence in a tidal boundary layer, Cont. Shelf Res., 17(8), 859–875.
Kleinhans, M., H. Jagers, E. Mosselman, and C. Sloff (2008), Bifurcationdynamics and avulsion duration in meandering rivers by one-dimensionaland three-dimensional models, Water Resour. Res., 44, doi:10.1029/2007WR005912.
Kleinhans, M. G., R. I. Ferguson, S. N. Lane, and R. J. Hardy (2013), Split-ting rivers at their seams: bifurcations and avulsion, Earth Surf. Process.Landforms, 38(1), 47–61, doi:10.1002/esp.3268.
Legleiter, C. J., and P. C. Kyriakidis (2007), Forward and inverse transfor-mations between cartesian and channel fitted coordinate systems formeandering rivers, Math. Geol., 38, 927–958.
Mikkelsen, O., and M. Pejrup (2001), The use of a LISST-100 laser particlesizer for in-situ estimates of floc size, density and settling velocity, Geo-Mar. Lett., 20, 187–195.
Miori, S., R. J. Hardy, and S. N. Lane (2012), Topographic forcing of flowpartition and flow structures at river bifurcations, Earth Surf. Process.Landforms, 37(6), 666–679.
Muste, M., and V. C. Patel (1997), Velocity profiles for particles and liquidin open-channel flow with suspended sediment, J. Hydraul. Eng., 123(9),742–751.
Muste, M., K. Yu, I. Fujita, and R. Ettema (2005), Two-phase versus mixed-flow perspective on suspended sediment transport in turbulent channelflows, Water Resour. Res., 41, W10402–1–W10,402–22, doi:10.1029/2004WR003595.
Nielsen, P., and I. A. L. Teakle (2004), Turbulent diffusion of momentumand suspended particles: A finite-mixing-length theory, Phy. Fluids,16(7), 2342–2348.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
2133
Nikora, V. I., and D. G. Goring (2002), Fluctuations of suspended sedimentconcentration and turbulent sediment fluxes in an open-channel flow, J.Hydraul. Eng., 128(2), 214–224.
Parker, G., and E. D. Andrews (1985), Sorting of bed load sediment by flowin meander bends, Water Resour. Res., 21, 1361–1373, doi:10.1029/WR021i009p01361.
Parker, G., C. M. Toro-Escobar, M. Ramey, and S. Beck (2003), Effect offloodwater extraction on mountain stream morphology, J. Hydraul. Eng.,129(11), 885–895.
Parsons, D., J. Best, S. Lane, O. Orfeo, R. Hardy, and R. Kostaschuk(2007), Form roughness and the absence of secondary flow in a largeconfluence-diffluence, Rio Paran�a, Argentina, Earth Surf. Process. Land-forms, 32, 155–162.
Richardson, W., and C. Thorne (1998), Secondary currents around braidbar in Brahmaputra river, Bangladesh, J. Hydraul. Eng., 124(3), 325–328.
Rose, C. P., and P. D. Thorne (2001), Measurements of suspended sedimenttransport parameters in a tidal estuary, Cont. Shelf Res., 21(15), 1551–1575.
Sassi, M. G., A. J. F. Hoitink, B. Vermeulen, and H. Hidayat (2011a), Dis-charge estimation from H-ADCP measurements in a tidal river subject tosidewall effects and a mobile bed, Water Resour. Res., 47, W06504,doi:10.1029/2010WR009972.
Sassi, M. G., A. J. F. Hoitink, B. de Brye, and E. Deleersnijder (2012a),Downstream hydraulic geometry of a tidally influenced river delta, J.Geophys. Res., 117, F04022, doi:10.1029/2012JF002448.
Sassi, M. G., A. J. F. Hoitink, B. de Brye, B. Vermeulen, and E. Deleersnijder(2011b), Tidal impact on the division of river discharge over distributarychannels in the Mahakam Delta, Ocean Dyn., 61, 2211–2228, doi:10.1007/s10236-011-0473-9.
Sassi, M. G., A. J. F. Hoitink, and B. Vermeulen (2012b), Impact of soundattenuation by suspended sediment on ADCP backscatter calibrations,Water Resour. Res., 48, W09520, doi:10.1029/2012WR012008.
Slingerland, R., and N. D. Smith (1998), Necessary conditions for a mean-dering-river avulsion, Geology, 26, 435–438.
Struiksma, N., K. W. Olesen, C. Flokstra, and H. J. de Vriend (1985), Beddeformation in curved alluvial channels., J. Hydraul. Res., 23(1), 57–79.
Szupiany, R. N., M. Amsler, J. Best, and D. Parsons (2007), Comparison offixed- and moving-vessel flow measurements with an ADP in a largeriver, J. Hydraul. Eng., 133(12), 1299–1309.
Talmon, A. M. (1992), Bed topography of river bends with suspended sedi-ment transport, Ph.D. thesis, Commun. on Hydraul. & Geotechn. Eng.,Delft, The Netherlands: Delft Univ. of Technol.
Thomas, R. E., D. R. Parsons, S. D. Sandbach, G. M. Keevil, W. A. Marra,R. J. Hardy, J. L. Best, S. N. Lane, and J. A. Ross (2011), An experimen-tal study of discharge partitioning and flow structure at symmetricalbifurcations, Earth Surf. Process. Landforms, 36(15), 2069–2082.
van Rijn, L. C. (1984), Sediment transport, part II : suspended load trans-port, J. Hydraul. Eng., 110(11), 1613–1641.
van Rijn, L. C. (2007), Unified view of sediment transport by currents andwaves. I : Initiation of motion, bed roughness, and bed-load transport, J.Hydraul. Eng., 133(6), 649–667.
Vermeulen, B., A. J. F. Hoitink, and M. G. Sassi (2011), Coupled ADCPscan yield complete Reynolds stress tensor profiles in geophysical surfaceflows, Geophys. Res. Lett., 38, doi:10.1029/2011GL046684.
Wang, Z. B. (1992), Theoretical analysis on depth-integrated modelling ofsuspended sediment transport, J. Hydraul. Res., 30(3), 403–421.
Wang, Z. B., and J. S. Ribberink (1986), Validity of a depth-integratedmodel for suspended sediment transport, J. Hydraul. Res., 24(1), 53–67.
Wang, Z. B., R. Fokkink, M. de Vries, and A. Langerak (1995), Stability ofriver bifurcations in 1D morphodynamic models, J. Hydraul. Res., 33,739–750.
Whitehouse, R. (1995), Observations of the boundary layer characteristics andthe suspension of sand at a tidal site, Cont. Shelf Res., 15(13), 1549–1567.
Zhou, J., H. H. Chang, and D. Stow (1993), A model for phase lag of sec-ondary flow in river meanders, J. Hydrol., 146(C), 73–88.
SASSI ET AL.: SEDIMENT DISCHARGE DIVISION AT BIFURCATIONS
