FLUVIAL PROCESSES IN ATTACHMENT BARS IN THE UPPER PARANÁ RIVER, BRAZIL PROCESSOS FLUVIAIS EM BARRAS DE SOLDAMENTO NO ALTO RIO PARANÁ, BRASIL

Bars are semi-submerged fluvial forms associated with the availability of sediments and a temporal dynamic, whose dimensions are controlled by the fl ow and depth of the channel. Attachment bars are very common in large anabranching river systems and play an important role in island formation and ecology. The Upper Paraná River exhibits an anabranching pattern characterized by channels of diff erent sizes, separated by islands and bars. The objective of this work is to present the processes involved in the formation and development of attachment bars in Santa Rosa Island, situated in Porto Rico, State of Parana, Southern Brazil. Acquisition campaigns were performed to obtain data on channel hydraulics (ADCP equipment), morphometry (Echo-sound profi les) and textural parameters (grain-size analyses) at high and medium water levels. Santa Rosa Island divides the fl ow into two channels of distinct hydraulic and sedimentary dynamics. Flow diversion produces a decrease in fl ow velocity and consequent sediment deposition near the upstream end of Santa Rosa Island. The formation and maintenance of attachment bars in Santa Rosa Island is related to fl ow competence reduction and the occurrence of divergent currents. Vegetation cover and fl ow regime control its permanence. Informações sobre o Artigo Recebido (Received): 14/09/2016 Aceito (Accepted): 25/04/2017


Introdution
In spite of a controversial and complex terminological discussion (SMITH, 1974;BRIDGE and TYE, 2000), fl uvial sand bars, as used in this study, are in-channel deposits emerged at medium water level (BRIDGE, 2003). Bars are probably the most studied of the fl uvial forms, and their wide classifi cation is based on diff erent parameters involving morphology (SMITH, 1978), genesis (ASHLEY, 1978;STEVAUX, 1994), grain size (KOSTER, 1978) and position in the channel (CHURCH and JONES, 1982).
The term 'anastomosing' was discussed by many authors (SCHUMM, 1985, KNIGHTON and NANSON, 1993, BRIDGE, 1993, and for a long time, this type of rivers has often been confused with braided rivers, which roughly have a comparable planform (MAKASKE, 2001). Nanson and Knighton (1996) enlarged the scope of the term anabranching, a former synonym of anas-tomosed, providing a genetic connotation concerning stream power and sediment size for multiple channel rivers. Latrubesse (2008) showed that many large rivers seem too complex in pattern to be simply categorized other than 'anabranching'. According to Lewin and Ashworth, (2014) many large rivers are simultaneously anabranching and plural systems (braided, meandering or straight), being the result of diff erent process sets. Concerning the importance of bars on anabranching formation and maintenance, Dunne and Aalto (2013) stated that rivers with relatively high sediment: water discharge ratio, such as the Ganga, tend to form large amounts of mid-channel bars and, consequently, tend to be anabranching. As opposed to the case of braided rivers, studies on anabranching rivers have to be conducted in natural systems, since its complex relation between channels and stable vegetated islands, and its mixed suspended sediments and bedload involves time scale and fl ow conditions impossible to be reproduced in fl umes.
The anabranching pattern is characterized by stable vegetated islands, which do not fi t to fl ow seasonal variability. Furthermore, they present unconsolidated sediments that form river sand bars. However, anabranching rivers can also be characterized by bars as the Brahmaputra and the Orinoco, which had numerous sand bars that resulted in a local braiding pattern (LATRUBESSE, 2008).
The Upper Paraná River is defi ned as anastomosing by Stevaux and Souza (2004) and anabranching by Latrubesse (2008). The bars in this region were classifi ed by Santos et al. (1992) and Souza Filho and Stevaux (2000) according to their morphology, evolution and position in central, lateral, tributary mouth, point bar and attachment bar. The latter can be considered a modifi cation of the lateral bar, more commonly found in large anabranching river systems. The term attachment comprises the fact that it usually joins an island or channel bank. Junction can occur laterally (lateral attachment bar) and on the upstream face of an island (island-head attachment bar) (DRAGO et al., 2013). The scars resulting from the attachment processes remain evident on the surface of the island.
The objective of this paper is to present the processes involved in the formation of attachment bars, as well as to discuss their development on an anabranching reach of the Upper Paraná River.

Field Setting
The Paraná River is the principal drainage of the La Plata River Basin and it drains 2.5 million km². The river runs for 3,965 km from the source of the Grande River (one of its formers) in the Mar Mountains, only 17 km of the Atlantic Ocean (S 21 o ), to the La Plata River Estuary in Argentina (S 34 o ), near Buenos Aires (ORFEO and STEVAUX, 2002). Its upper reach is almost totally man-controlled by a series of reservoirs on both trunk and main tributaries. The only "natural" condition is a 255 km reach between Itaipu and Porto Primavera Dams (Figure1).
The studied bars are formed along the Santa Rosa Island (SRI) (22°46'15"S, 53°17'48"W), near the town of Porto Rico, PR,Brazil (Figure1). This reach is directly infl uenced by the upstream dams of Porto Primavera, located at a distance of 48 km, and Rosana, located on the Paranapanema River, 53 km (STEVAUX et al., 2009). In this reach, the river channel has 3,200 m in width and is separated into three anabranching channels by the Mutum Island and SRI The SRI is a relatively small island (2,500-m long and 150-m wide), which divides the left channel of the Paraná River into two channels of 600 and 900 m in width. In this reach, the thalweg is asymmetrically shifted to the left and ranges from 10 to 15m in depth. The secondary anabranching channel on the right of the SRI is 600 m wide with average depth of 4-5 m. The water slope in the reach is 0.00007 with local variation of 0.00002 to 0.00004 (LELI, 2015).
The Porto São José Fluvial Station (Figure1), in operation since 1964 -located in the upstream sector of the study reach -gauged a medium discharge of 8,912 m 3 s -1 , with a minimum and a maximum recorded annual average discharge of 7,089 m³ s -¹ 10,853 m³ s -¹, respectively, and extremes of 2,551 m³ s -¹ in 1969 (probably associated to La Niña event) and 34,912 m³ s -¹ in 1982-83 ENSO. In this reach, the periods of high and low water levels are well defi ned. The largest discharges occur between the months of December and March (summer), followed by a period of low water levels between the months of April and November (STEVAUX et al., 2009).
Along the study reach, the Paraná River's bed channel is essentially sandy, formed by mega-ripples and dunes. In large fl oods, it is possible to develop a large bedform called "sand wave", composed of the accretion of ordinary bedforms. Sand waves generally emerge at medium water levels and form mid-channel bars (STEVAUX, 1994).
Although it is non-damned, this reach suff ers the impacts produced by the dam upstream such as discharge control with (1) reduction of fl ood tension (diff erence between minimum and maximum water level); (2) signifi cant decrease in suspended sediment concentration (from 35 to 0.3 mg L -1 ); (3) formation of armor layers (increase in bed channel grain-size); and (4) reduction in bedform size and steepness (STEVAUX et al., 2009). According to Martins (2008), bedload discharge measured by bedform migration is 1,152,325 ton. y -1 . However, this value tends to drop over time because of the bedload fl ow interruption by the Porto Primavera Dam.
The bars of this study ware chosen due to their spatial position and previous studies by Santos et al. (1992). These authors described such bars as deposits, which rapidly change their morphology over time.

Data collection
Two fi eld observations at high (Jan. 2007) and medium (Feb. 2009) water levels were carried out for discharge determination, bottom profi les, fl ow velocity and directions, and bedload sampling. The fi rst fi eld observation -the Porto São José Fluvial Station -gauged 632 cm (over fl uvial station zero) and a discharge of 17,730 m 3 s -1 (Bankful discharge). In the second fi eld observation, water leveled at 351 cm and daily discharge of 9,240 m 3 s -1 (Figure2). Flow measurement was obtained using the ADCP (Acoustic Doppler Current Profi ler of the RD Instru-ments®, Rio Grande, 600 KHz and software Winriver software (Acquire mode). Longitudinal (15) and trans-versal (19) profi les were performed at high (2007) and medium (2009) water level by ADCP device along the Santa Rosa Island (Figure3). The data collected by the ADCP were processed by Winriver (Playback mode). Bathymetrical surveys were conducted using Furuno echo sound (GP 1650-F model) with GPS acquired and stored by the Fugawi 3 software and processed by the ARCGIS 10.1 software for the generation of bathymetric maps through the geostatistical procedure of kriging.
Grain-size analyses were processed in 76 samples (n =17 medium water, n = 59 high water), collected during the two water-level moments for attachment lateral and head-island bars, main and secondary channels (Figure3). The large diff erence between the sample collection for the fi rst and second fi eld work is due to the emersion of many sectors of bedforms during low water. The samples were analyzed by the sieving method to assess the particle size distribution, afterward it was calculated mean particle diameter (D50) and the coarse fraction of 10% (D90).

Hydraulic parameters and equations
Some basic hydraulic parameters such as stream power, specifi c stream power, critical shear stress and the Shields parameter can reveal the characteristics and changes in the fl ow and, consequently, the fl uvial processes involved in bar formation: a) Specifi c stream power () is the stream power ( by channel width unity and it "determines the capacity of a given fl ow to transport sediment" (Knighton, 1999). It is expressed by the following equations: where ρ (kg m -3 ) is the constant water density, g (ms -2 ) is the gravitational, Q (m³.s -¹) is the water discharge, S is the slope and w (m) is the cross-section width. b) Critical shear stress ( ) characterizes the interaction between the fl ow-channel bed and the beginning of the movement of a certain particle size (D). It is defi ned by Richards (1982) as where ρ s and ρ (kg m -3 ) are the sediment and water specifi c mass, and D 50 (mm) is the grain-size median.
c) The Shields parameter (Shields, 1936) relates a dimensionless critical shear stress (s) with the particle Reynolds number (R * ) weight immersed in water and is expressed by where u (m s -1 ) is the fall velocity (or friction) and ν (m 2 s -1 ) is the kinetically water viscosity.

Flow distribution, direction and channel energy
The area between the channel and the island is a complex of morphologies and fl ows that generates seven sub-environments: (1) main (left) and (2) secondary (right) channels defi ned by fl ow separation by SRI (Figure 1 and Table 1); (3) main and (4) secondary channel attachment bar; (5) head-island bar, formed in the upstream face of the island; (6) island-bar channel to the left and (7) the right side of the island (Figure 3). SRI divides the fl ow into two -main (left) and secondary (right) -channels, with distinctive hydraulic characteristics (Figures. 4, 5). The main channel leads the largest portion of fl ow and has a high fl ow velocity, especially in the thalweg. On the other hand, the secondary channel has lower values for the parameters mentioned above. Flow division generates two low-velocity and divergent "shadow" zones (secondary currents) on each side of the island ( Figure 5).
In studies of the anabranching river systems, is usual the calculation of stream power (eq. 1) and specifi c stream power (eq. 2), considering the entire section  (margin to margin). In this study, these parameters were calculated for each branch (single, main and secondary channel). Therefore, it is possible to better understand the transport and morphology of each channel. Specifi c stream power in this study reach varied according to the space and time (Table 1). The highest values were found in the main channel both at high and medium water levels.

Bar morphology
On the bathymetrical surveys conducted during the high water periods, the attachment bars (lateral and head-island) completely submerged, making it diffi cult to recognize their morphology, considering that in fl oods, there is a larger reworking of river forms. ( Figure  6). In the medium level period, the bars emerged and spread across a large channel area, keeping a more stable morphology (Figure 7).  The secondary channel bars developed downstream and laterally. Its morphological stability was such that allowed the development of a sparse grass vegetation cover. The main channel bar was not as developed as the secondary one and remained submerse in its majority, even in the medium period. Active erosive features could be found in many parts in both periods. The sharp outer border of this bar is parallel to the thalweg of channel, with the depth downing abruptly from 2 to 10m. With bar formation, an island-bar channel was formed, generating a lentic to semi-lotic environment separated from the channel fl ow. The island-bar channel was also more developed in the secondary than in the main channel. Although the head-island bar was also submersed during the fl ood, its morphology did not suff er major changes. In this type of bar, vertical accretion is more eff ective than in the attachment bar. In medium water, the head bar practically connected with the island and presented sparse patches of grass vegetation.

Bar composition and sediment entrainment
Particle diameter changed according to water level. Comparing the frequency curve for both water levels ( Figure 8) for samples of all sub-environments, it is possible to see that, during the fl ood, the bedload presented a high sorting degree for D 50 with a small variation from 1.8 to 1.0 mm (Φ -0.9 to 1). The condition of fl ow homogenization, that normally occurs during the fl ood, might be responsible for the very similar texture of all sub-environments. During the fl ood, all environments are almost under the same fl ow energy, in a situation quite similar to the one that occurs in braided channels. Rev. Bras. Geomorfol. (Online), São Paulo, v.18, n.3, (Jul-Set) p.483-499, 2017 In medium water, the sub-environments are relatively disconnected and present diff erent fl ow velocities. Under this condition, sediment texture presents a lower sorting degree with D 50 ranging from 1.8 to 1 mm. Still in medium water, some samples from island-bar channel deposits presented fi ner (silt and clay) texture, because at some moments the isolation of this environment from the river channel is almost complete with lotic to semi-lentic fl ow conditions. Samples generated four groups when plotted in a grain-size diagram for all sub-environment samples and the two water levels (Figure 9). This diagram assumes that sediments arriving at the stream have diff erent sizes and they suff er diff erent transport processes according to local fl ow conditions. The forces acting on the particles can keep them in suspension or on the river bottom (rolling and traction). This is due to the size, weight and shape of the particle, as well as the function of fl ow type (whether it is laminar or turbulent), fl ow velocity, bed obstacles and various other interrelated functions, such as slope bed and channel shape, among others. Passega (1957Passega ( , 1963 proposed a diagram (Figure 9) to establish the relationship between the particle size characteristics of sediments and deposition processes. In this diagram, the abscissa corresponds to the average grain diameter of a sample (D 50 ), and the ordinates corresponds to the fi rst percentile of the sample or the grain diameter that is only surpassed by 1% of the sample (D 90 ). The latter value indicates the competence of carriers. Finer sediments were found in places of greater depth and fl ow velocity (thalweg) in both water level periods. Coarse and medium grains were distributed regularly to all environments. In medium water level periods, coarse and medium sand was found in shallow environments ( Table 2).
The relation between critical shear stress -ϴs -(Eq. 4) and particle Reynolds number -R * -(Eq. 5) is presented in the Shields diagram (Shields, 1936) ( Figure 10). These parameters were obtained in diff erent sub-environments both at high and medium water levels. According to the diagram, the values obtained are in an intermediary zone, in which particles are covered by a sub-laminar layer with an initial movement at ϴs > 0.6 at both water levels. Critical shear stress and particle Reynolds number are higher in the main channel, with no great variations in both periods analyzed.

Discussion
Earlier studies FRANZI-NELLI, 2002, 2005;SARMA, 2005, LATRUBESSE, 2008ROZO et al., 2012;NICHOLAS et al., 2013;REESINK et al., 2014) indicate that fl uvial bars in anabranching rivers are deposited in sites with: (i) divergent secondary fl ows (ii) fl ow competence reduction (specifi c stream power and fl ow shear stress) and (iii) sand availability. These conditions were present in the study site. The fl ow separation caused by SRI induces the formation of a low-energy zone, which is not only very well defi ned by fi eld equipment (ADCP), but it is visible on the water surface as well. Flow separation generates double fl ow lines with secondary fl ows diverging 47° to 87° from the mainstream direction, generating a "shadow zone" around the island. Velocity in this zone ranges from zero, near to maximum in the limit with normal channel fl ow (Figure4).
Three conditions concerning shear stress relation were found: τcr:τ0 > 1 near the island, τcr:τ0 = 1 at zone boundary with normal channel fl ow and τcr:τ0 < 1 in the channel. Bedload moves along the channel through Rev. Bras. Geomorfol. (Online), São Paulo, v.18, n.3, (Jul-Set) p.483-499, 2017 mega-ripple and dunes with their crests perpendicular to fl ow direction at a τcr:τ0 < 1 condition. When transported, bedload reaches shadow zone boundary τcr:τ0 = 1, particle deposits and an attachment bar begins to be constructed. This condition also generates the bar-island channel in the τcr:τ0 > 1 sector. Unlikely the origin of other bars, as postulated by Ashworth et al. (2000), Sarma (2005) and Njenga et al. (2013) and others, attachment bars do not arise from mega bedforms such as dunes and sand waves. A rather complex process is involved though. In-transit bedforms in the study area move about 2-3 md -1 at medium water levels. To obtain this displacement, fl ow velocity has to be around 0.8 to 1.2 m s -1 for an average depth of 4-5 m over fi ne to medium sand grains (Gon, 2012).
Under this condition, bedforms normally migrate as mega-ripples and dunes with approximately straight crests perpendicular to higher velocity water fl ow. When an end of a bedform crosses the borderline of the low velocity zone, part of the bedform deposits while the rest continues its migration downstream. However, the crest shifts diagonally with the higher water fl ow line by anchor eff ect. In the course of time, it tends to form a shoestring sand bar grossly parallel to the island (Figure11). The continuous deposition of the attachment bar propagates downstream the shadow zone and, consequently, stabilizes the attached bar. Although these bars are reworked annually during fl oods, they also tend to increase vertically by aggradation, which improves their morphological stability. In secondary channel bars, the stability is such that allows grass vegetation to develop. This tendency refl ects reports by Nicholas et al. (2013), which show that bar and island stability may be sensitive to hydrologic regime, especially because greater variability in fl ood magnitude encourages the formation of emergent bars that can be converted into stable islands by vegetation colonization.
Bedload material is predominantly sand, except for the island-bar channel where silt and clay are also found. According to the grain-size relation (Figure9), it is possible to observe that fi ne particles are distributed in places of higher velocity and depth (thalweg) in both water level periods. On the other hand, medium and coarse sand is distributed along both secondary channels during high water level periods and in shallow sub-environments (bar and right channel) at the medium water level. Rocha and Souza Filho (2005) suggested that during the fl oods in the study reach, large bedforms (dunes and mega-ripples) composed of fi ne to medium sand are formed and overlain the coarse bed deposits. In medium water periods, the fi ner grains continue to be transported and the coarser ones remain as lag deposits. This may explain why in regions of higher fl ow velocity, at medium water levels, it is possible to fi nd fi ner sediments (Figure9). As previously seen, medium and coarse grains are being deposited in the head of RSI (head-island bar of Drago et al., 2013) and in the secondary channel, but they are not being removed.
Therefore, the formation and maintenance of attachment bars in SRI are related to the reduction of fl ow competence and occurrence of divergent currents. The bar-island channel keeps fl ow active up to its upstream closing. Vegetation cover and fl ow regime control the permanence of the attachment. Depending on sand availability, the attachment bar can be formed laterally and in front of the island.
The Upper Paraná River anabranching channel pattern is formed by long and narrow islands, whose formation is controlled by attachment bar processes. In this type of channel pattern, a straight and relatively narrow channel forces the formation of long islands and, therefore, the attachment bars as seen in this case. Dunne and Aalto (2013) have emphasized the importance of bar formation and maintenance of anabranching channels, especially in rivers with high sediment: water discharge such as the Ganges and the Upper Paraná. Actually, most islands that compose the studied anabranching reach are formed by processes involving attachment bars.

Conclusions
Results showed that attachment bars are lateral and/or head-island bars formed by fl ow division by the Santa Rosa Island, so this type of bar can be more expressive in anabranching rivers. Bar deposition is caused by a zone of low fl ow velocity developed by fl ow separation. In this zone, fl ow can diverge 47° to 87° from original direction, and velocity reaches gradually to zero from the zone limit to the Island's bank.
Bar sedimentation occurs within shadow zone, where the flow velocity decreases at the transport particle limit. When the bed-form entrainment enters in the shadow zone deposition occurs forming the attachment bars.
Bar morphology is directly related to the Paraná River regime. During the fl ood, bar is usually reworked and remain stable in medium/low water level. Major modifi cations in bar morphology are found in the deepest places while deposition is more intensive in shallow parts.
The fl ow diversion generated secondary channels with diff erent specifi c stream power and shear stress, being the left one that presents major values.
Bed load material is predominantly sand, except for the island-bar channel, where silt and clay also be found. According to the grain size relation, it is possible to identify that fi ne particles are distributed in places of higher velocity and depth (thalweg) in both water level periods. Although it seems a paradox, medium and coarse sand is distributed along in both secondary channels during the high water level period and in the shallow sub-environment (bar and right channel) in the medium water level. We attribute this fact to a lag eff ect that carries out fi ner particles, resting the large one.
Based on stream power and critical shear stress it was possible to defi ne that erosion and transport are more active in the left channel in both periods. In the right channel, these values are lower, with much more active deposition generating a larger bar in it.