ISLAND AND ANABRANCHING GENERATION PROCESSES – A COMPARATIVE REVIEW IN THE UPPER PARANÁ RIVER PROCESSOS DE MULTICANALIZAÇÃO E GERAÇÃO DE ILHAS – UMA REVISÃO BASEADA NO CASO DO ALTO RIO PARANÁ

Although the anabranching channel pattern is frequent in rivers of diff erent sizes and climates, understanding the processes involved in its genesis requires further research, especially focusing on island formation. This article discusses the processes involved in anabranching patterns and island formation, focusing on the environmental characteristics and sedimentary records in a stretch of the Upper Paraná River PR. We compare the results with concepts and defi nitions in the literature. Islands in the Upper Paraná River form in two ways: from 1) the stabilization of central bars (in-channel processes) and 2) the fl oodplain cutoff (off -channel processes). The islands of the fi rst case are generally smaller and younger than those formed by off -channel processes. Both islands can expand their surface by attaching lateral bars. Multi-channeling, island formation, and permanence can be explained through the concept of maximum fl ow effi ciency. Informações sobre o Artigo Recebido (Received): 01/05/2020 Aceito (Accepted): 03/09/2020 Palavras-chave: Sedimentação Fluvial; Multicanalização; Barras de Canal; Rio Paraná; Ilha Fluvial.


Introduction
Interest in research on multiple channel rivers has increased since Brice (1982Brice ( , 1983 included the anastomosed pattern as a new end-member in the original tripartite classifi cation proposed by Leopold and Wolman (1957) for alluvial channels. Contrary to the braided channel, in which an unstable sandy bar divides the river fl ow at the barfull level, the anastomosed pattern is essentially multichannel. In this case, the river consists of secondary channels defi ned by stable islands at the bankfull level. First, the multichannel river was considered a geomorphological rarity, exclusive to arid areas in Australia (STEVAUX and LATRUBESSE, 2017), being studied only from the architectural point of view of its sedimentary record (MIALL, 1977(MIALL, , 1980. Currently, multichannel patterns are recognized in the most varied landscapes and climates on the planet (STEVAUX and LATRUBESSE, 2017).
Although multichannel rivers are extremely common, most of their classifi cations or descriptions have developed from studies in medium and small rivers as well as in fl umes. Latrubesse (2008) presented an initial classifi cation for large multichannel river systems based on the degree of anabranching (number and channels per section), fl ow magnitude, width/depth ratio, sinuosity, and slope. The author also mentioned that the anabranching pattern is predominant in mega-rivers (medium discharge > 17 000 m 3 s -1 ). In this study, we use the term multichannel in the same manner as anastomosing or anabranching, as mentioned by Stevaux and Latrubesse (2017). Huang and Nanson (2007) show that anabranching is the product of the slope adjusting with a consequent increase of fl ow effi ciency. On the other hand, the authors also demonstrate that fl ow effi ciency can be signifi cantly increased by channel width reduction through the formation of vegetated islands: the presence of islands decreases channel width. Other studies on multichannel formation processes also present the hydraulic slope, fl ow, and bed sediment input as essential characteristics of the system for the development of the anabranching pattern (NANSON and KNIGHTON, 1996;STEVAUX and LATRUBESSE, 2017). Besides, the infl uence of human activities is added, such as the construction of dams, which alter the balance between the supply of sediments and their transport capacity, which generally impacts the pattern of the river channel (SLOWIK et al., 2018).
In this context, islands are considered an essential component in multichannel pattern physiography. However, these forms have not received much attention regarding the study of their formation until the mid-1990s. Nanson and Knighton (1996) and Nanson (2013) demonstrated that islands can evolve from the interaction between the stabilization of sandy bars and vegetation settlement. Once formed, the island increases by attachment of the lateral bar and frontal bars (NANSON and KNIGHTON, 1996;LELI et al., 2018). Islands are formed by in-and off -channel processes (LELI et al., 2020a). According to these authors, during the in-channel process, islands are formed by fi xing central bars to an existing island or bar. Those from the off -channel process result from channel avulsion and fl oodplain cutoff .
In studies of large rivers in the Amazon, Latrubesse and Franzineli (2005), , and Latrubesse and Stevaux (2013) characterized the large lake islands of the Negro River originating from the formation of sandy bars during the Late Holocene, when the river carried a large amount of bed sediment. Lake-island is a rare type of in-channel island practically unknown in the literature (LELI et al., 2016). In the Upper Paraná River, Leli (2015) and Leli et al. (2018) observed the occurrence of islands formed from central bars (in-channel processes) and derived from the avulsion process (off -channel process).
Although the recognition of the multichannel pattern is a consensus in the literature, the understanding of its formative processes, morphological resilience, and sedimentology remains the subject of ongoing research (CARLING et al., 2014). The objective of this study is to present a review of the processes involved in island formation and multi-channeling development based on the results from the Upper Paraná River. The Paraná River is one of the ten largest rivers in the world (LATRU-BESSE, 2008). In the last 40 years, it has been studied in diff erent fi elds, mainly ecology (e.g., THOMAZ et al., 2004;IRIONDO et al., 2007) and geomorphology (e.g., ORFEO et al., 2020). Concerning fl uvial geomorphology, these studies provided signifi cant knowledge on channel morphology, pattern, hydrology, hydraulics, and sedimentology (STEVAUX, 1994;DRAGO and AMSLER, 1998;ORFEO and STEVAUX, 2002;ALARCÓN et al., 2003;STEVAUX and SOUZA, 2004;MARTINS and STEVAUX, 2005;SANTOS et al., 2017). These facts contributed to the use of the Paraná River as a comparative model for other river systems.

Method and study area
In this study, we used local and general references. We tried to compare the concepts and defi nitions with the results obtained in the study area. The greatest diffi culties emerged in terms of the diff erent magnitudes of the systems: in general, the basic references in the literature concern river systems smaller than that of the study area.
The Upper Paraná River extends for approximately 600 km, entirely in Brazilian territory, from the confl uence of the Grande and Paranaíba rivers (20° 04' 49″ S 51° 00' 08″ W) up to the Sete Quedas Falls (currently, Itaipu Dam's lake), the largest knickpoint in the basin (Fig. 1). The Upper Paraná River Basin is one of the most dammed large river basins in the world (STEVAUX et al., 2009), with more than 150 large dams in operation. Only 235 km of the upper reach, from Porto Primavera and Itaipu dams, are in natural non-dammed conditions. The river in this reach is multichannel and formed by 265 islands, with nodal sections of 1100 m in width up to 12 500 m in the six-channel section of Porto 18 (LELI, 2015;LELI et al., 2018). The Upper Paraná presents the lower slopes typical of large rivers between 3.0 and 6.0 cm km -1 . Two Late Pleistocene terraces occurs at the right bank of the river at 20 m (Taquaruçu Terrace) and 10 m (Fazenda Boa Terrace) above the river average water level (STEVAUX, 1993(STEVAUX, , 1994(STEVAUX, , 2000. The fl oodplain develops on the right bank. It presents a very complex morphology generated by a long history of avulsions, island incorporations, and secondary channel developments. Stevaux and Souza (2004) classifi ed this fl oodplain as a meandering-anastomosed alluvial plain, according to the conception by Nanson and Croke (1992). The left bank is formed mainly by a 20 m tall sandstone wall of the Caiuá Formation (K).
The average temperature in the region is 22 °C and the annual rainfall is 1200 mm. At the Porto São José gauge station (upstream), with a historical series of 60 years, the average discharge varies from 8400 m 3 s −1 during the dry season (June-August) to 13 000 m 3 s −1 during the fl ood period (November-March). The extreme discharges recorded at this station were 33 740 m 3 s −1 in 1983 and 2 550 m 3 s −1 in 1969. At the Guaíra -PR gauge station (downstream), in operation since 1910, the average discharge is 10 800 m 3 s −1 , with an extreme of 2 490 m 3 s −1 in 1944 and 39 870 m 3 s −1 in 1983 (SOUZA FILHO, 1993).

Sediment transport
The construction of the Porto Primavera hydroelectric dam determined a new confi guration of hydrology and load transport in the studied stretch of the Upper Paraná River (CRISPIN, 2001;MARTINS and STEVAUX, 2005;STEVAUX et al., 2009;SOUZA FILHO, 2016). Although eff ectively completed in 1998, since 1991, the construction of the dam began to interfere in the canal due to the installation of coff erdams for the construction of concrete structures (SOUZA FILHO, 2016). Souza Filho et al. (2010) observed changes in the dissolved load in measurements made in Porto São José after dam closure. The authors identifi ed a reduction in P, Ca, Fe, Sr, Si, Mg, Zn, Pb, Al, Cd, Cu, Co, and Ba, and an increase in Na, K, Cl, and Ni as well as an increase in water temperature and a decrease in pH during the fl ood season.
Analyses from diff erent sources have shown that, during the construction of the Porto Primavera Dam, signifi cant changes in suspended sediment discharge occurred in the region of Porto Rico (Fig. 1). The values of 214.7 kg s -1 in the pre-dam period increased to 222.2 kg s -1 during construction and drastically reduced to 56.4 kg s -1 after dam closure (SOUZA FILHO, 2016). Other studies have confi rmed the gradual retention of suspended sediment by the dam. The suspended sediment concentration reduced from 35.0 mg L -1 to the current 0.5 mg L -1 (CRISPIN, 2001;STEVAUX and TAKEDA, 2002;MARTINS and STEVAUX, 2004). This concentration partially recovered downstream of the confl uence of the Ivaí and Piquiri rivers (Fig. 1). The Ivaí River supplies 2 097 292.6 t y -1 of suspended sediments (LELI et al., 2017), increasing the concentration to 5.0 mg L -1 (FRANCO et al., 2008). The Piquiri River introduces 1 593506 t y -1 , increasing the concentration of the trunk river to 16.2 mg L -1 (BENNERT, 2018).
Analyzing the bed load is more diffi cult, owing to the lack of data obtained directly in the fi eld before dam construction, but it is assumed that the bedload transport has also changed (SOUZA FILHO, 2016). Martins and Stevaux (2005)  In general, Souza Filho (2016) suggested that the closure of the Porto Primavera Dam reduced, in the reach near Porto Rico, 74.6% of suspended load and 58.7% of bedload discharges, with a total reduction of 71.7% in sedimentary transport. Such information is essential for understanding some current forms and processes that occur in the Paraná River.

Bedforms
The bedforms of the Upper Paraná River (region of Porto Rico -PR) are lower and upper fl ow regime fl at beds, ripples, dunes, and sand waves (STEVAUX, 1994;STEVAUX et al., 2009). Sand waves are bed forms constructed by the accumulation and overlapping of smaller forms formed during large fl oods (Fig. 2). In general, sand waves emerge at medium-low water levels, constituting sand bars. Stevaux and Takeda (2002) and Martins and Stevaux (2005) measured dunes up to 100 m in length, up to 2.20 m in height, with a linear movement of 45 meters per month for periods of lower discharge and 67 meters per month during the fl ood periods (Fig. 3). Antidunes are rare in the Upper Paraná River that normally present low fl ow regime (Froud number = 0.9, according Martins, 2004). However, Leli (2015) reported the occurrence of antidunes during fl oods near the Piquiri River mouth.
There were no signifi cant changes in bedload displacement speed after the closure of the Porto Primavera Dam. However, Stevaux et al. (2009) found that the average height of dunes reduced from 2.2 to 1.5 m after dam operation. The authors associate this morphological alteration as a result of the alteration in the bedload texture that changed from fi ne/medium to medium/coarse sand at several points in the channel.
Such changes in the bedload are due to changes in the river regime introduced by the operation of the dam. Santos (2010) and Santos et al. (2017) estimated that the movement of the coarser bedload occurs during discharges greater than 10 853 m 3 s -1 . However, as the water is normally stored in the dam during the rainy season, it signifi cantly reduces the fl ood peak downstream. In this way, the fl ows of greater competence for the mobilization of the coarser sediment are rare or disappear, and only the transport of the fi ner load is maintained. In such a condition, the fi ner material is removed from the bedload and not replaced because the dam prevents its replacement from upstream. The coarser material does not move and forms lag deposits that pave the channel (armoring eff ect) with a consequent reduction in the bedform height and steepness (STEVAUX et al., 2009;STEVAUX and LATRUBESSE, 2017).

Bars
Bar morphologies are formed from the accumulation of bedforms that emerge at the mid-water level of the river. Generally, sandy bars are ephemeral and can be eroded during river fl oods, but they can also stabilize and evolve into islands (STEVAUX, 1994;ORFEO and STEVAUX, 2002;LELI, 2015;LELI et al., 2018). The factors that contribute to the stabilization and fi xation of the sandy bars are, mainly, the absence of extreme fl oods (fl oods with above average fl ow) for a relatively long period and vegetation cover development (STEVAUX, 1994). Although its surface consists mainly of sand, when the barfull level is reached during ordinary fl oods, the bar can receive a deposition of fi ne material (clay and silt), which forms thin layers of mud on the surface. This situation protects the surface of the bar from wind erosion and serves as a substrate for the implantation of pioneer grass and herbaceous vegetation. The more this type of fl ood occurs, the greater the chances of a bar being preserved in eventual larger fl oods. In this way, the bar starts functioning as an in-channel fl oodplain, with aggradation of fi ne sediments (FERNANDEZ et al., 1993;STEVAUX, 1994).
Bars can be classifi ed according to their morphology (SUNDBORG, 1956;COLLINSON, 1970), size (SMITH, 1974), stability (JACKSON, 1975), occurrence mode (isolated or in the group) (SMITH, 1974(SMITH, , 1982, and based on the channel pattern (READING, 1978;KELLERHALS et al., 1976). The highly accepted classifi cation of Smith (1976) uses the morphology and position in the channel to classify a bar as longitudinal, transverse, point or diagonal. Suguio and Bigarella (1990) contributed to the classifi cation of the bars considering the position in the channel (lateral and longitudinal bars) and the formation process. Santos et al. (1992), Souza Filho (1993, Stevaux (1994) and Turra et al. (1999) presented the fi rst classifi cation of the bars of the Upper Paraná River (Fig. 4, Tab. 1) that included the central, lateral, attachment, frontal (or island head bar), and confl uence bar. Leli (2015) and Pereira (2016) identifi ed, in the Paraná River, the Moa bar, a special type of lateral bar formed during fl ood events, described early in the Amazon Basin rivers (LATRUBESSE, 1992;STEVAUX and LATRUBESSE, 2017).

Islands
According to the defi nition provided by Brice (1964), islands diff er from the bars in that they keep the river fl ow separate at the bankfull level and persist in the system from 10 1 to 10 3 years in the Upper Paraná River (LELI, 2020a). These characteristics give the islands a fundamental role in the fl ow hydrodynamics of multichannel rivers (Fig. 5, Tab. 1). Islands are formed by processes developed inside (in--channel) or outside (off -channel) the channel. The island formation process is an important and determining factor in age, size, surface morphology, and characteristics of the vegetation cover (LELI et al., 2018;LELI et al., 2020a).

Central bar island (in-channel processes)
A stable and partially vegetated sandy bar is the precursor stage for central bar island formation (Fig. 5A, B, C and D). Leli et al. (2018Leli et al. ( , 2020a found that the presence of vegetation in the bar favors the deposition of fi ne sediment on its surface. Thin sediments, in turn, increase fertility and help plant development in a positive feedback system. With the vertical accretion of fi ne sediments, the surface of the island bar progresses gradually, rising in relation to the water level, which allows for the fi xation of larger plants until, fi nally, it is covered by tree vegetation (STEVAUX, 1994). The process continues until the surface of the island reaches the approximate level of the natural levee system. Stevaux (1994) established, based on the sedimentary composition, that the separation between the bar and island occurs in the contact of the basal sand with the top mud.
The sedimentary sequence of this type of island demonstrates the channel phase composed of sand from the initial bar, followed by the phase of island formation with the vertical accretion of fi ne deposits during the fl oods (Fig. 6A). In cores from Mutum Island, the contact between sand and the island environment is dated 8200 y BP (LELI et al., 2020b).
Lateral bar (Attachment bar) Elongated morphology, with a high length/width ratio. Deposited 10 to 20 m from the island (or channel) banks due to the low-velocity zone occurrence. When they join the island (or channel), banks are called attachment bars. Santos et al. (1992) and Santos et al.
Island head bar Defi ned as modifi ed forms of the central bars, with an emergent extension of up to 45 ha and a length greater than 1500 m. They are formed when a central bar is attached to the upstream part of an island and gains greater stability.

Confl uence bar
Formed by the combination of two diff erent fl ows (rivers confl uence). The hierarchical dynamics between the fl ows allows for the formation of low-speed zones with reduced transport capacity, forming the deposition of the sedimentary load at the confl uence.

Flood bar (Moa bar)
Relatively narrow forms anchored on the high banks of the river channels. They are formed during a fl ood when the concentration of suspended sand load is higher than in average fl ow. Moa bar occurs due to increased roughness and decreased fl ow velocity near the channel banks.

Latrubesse (1992)
Island Diff erent from bars, islands are channel deposits that stay emergent at the bankfull level. In general, islands are formed by fi ne overbank deposits. They are depositional, vegetated forms of varying sizes and can be in any part of a channel. They can be formed by the evolution of a central bar (in-channel processes) or by fl oodplain cutoff (off -channel processes). Both types of islands can have the area expanded by the process of bar attachment (in-channel processes) (Fig. 5, 6, and 8).
Stevaux (1994) Figures 1 and 4. B) Floodplain cutoff island (off -channel processes). Floodplain deposits are very old (12 430 y BP) when compared with the current natural levee deposits (2600 y BP) in the island border. Location information is presented in Figures 1 and 8. Source: Leli (2015).
A fl oodplain cutoff island is formed by the avulsion and incision of a channel on the fl oodplain and its junction with the main channel downstream (Figs. 7 and 8). Thus, the deposits of this type refer to the processes in which the fl oodplain environment was formed. It is a very common process in avulsion river systems in several rivers in the Pantanal, as mentioned by Assine (2003). Avulsion is favored in rivers with water levels higher than the surface of the fl oodplain, generating a hydraulic diff erence that favors the fl ow into the plain. This fl ow is usually fast and highly erosive. Ramonell et al. (2011) reported speeds of up to 2 m s -1 in an avulsion on the Paraná River near Santa Fé, Argentina. The nature of the fl oodplain bank is also a factor in the installation of a crevasse. Sandy banks, for example, are more susceptible to erosion than muddy banks and are potentially subject to crevassing, enabling the formation of fl oodplain channels. The longitudinal and transverse fl oodplain slopes and the magnitude of the avulsion channel determine the size and shape of the cutout sector. As part of the fl oodplain, the islands formed by the fl oodplain cutoff preserve its inherited features as fl ood channels, paleo levees, and paleochannels (Fig. 8). In the case of the Paraná River, islands of this type are generally much larger than those of the central bar at approximately 100 km in length, as in the case of the Bandeirantes and Grande islands (Figs. 1, 7, and 8). However, it is very diffi cult to establish the age of these islands because deposit dating refers to their deposition in the fl oodplain and not their incision due to the avulsion of the channel (Fig. 6B). Stevaux and Souza (2004)

reported a very important phase of generalized avulsion in the Upper
Paraná River about 2800 yr BP. It is possible to see in the core IG -Grande Island (Fig. 6B): the formation of the current island natural levee began 2600 yr BP, after the fl oodplain cutoff .
Sedimentary studies of the floodplain cutoff islands in the Upper Paraná River are similar to the depositional model of the fl oodplain in the area studied in the upper section. Leli (2015) compared the deposits of Grande Island with those from the fl oodplain and found similar processes and ages (Fig. 6B). It consists of basal sandy channel deposits overlaid by a typical fl oodplain sequence of mottled sandy mud intercalated with layers of fi ne muddy sand. At the top, fi ne deposits from crevasse splay can occur. Floodplain deposits of Grande Island were dated at 12 430 yr BP. An impressive other example of fl oodplain cutoff island is the Bananal Island formed by the avulsion of the Araguaia River (Fig. 9). Gradually, the original channel (the current Javaes River) was abandoned be-coming a secondary channel and transferring the major of its fl ow to the avulsion channel: the current Araguaia River. The Bananal is probably the largest fl uvial island in the world with 330 km in length and 70 km in width.

Superfi cial expansion of the islands
After its formation, both the central bar island and the fl oodplain cutoff island can reduce or expand in area, depending on the hydrological and hydraulic conditions, and transported load availability (STEVAUX and LATRUBESSE, 2017). Leli et al. (2020b) defi ned composite islands as islands that increased in size. The island surface expands due to the attachment of lateral and frontal bars (STEVAUX, 1994;LELI et al., 2018). The presence of the island in the channel separates the fl ow, generating a hydraulic disturbance called a low-speed fl ow zone (Fig. 10) (SANTOS et al., 2017;LELI, 2015;LELI et al., 2020b). When the bedform downstream entrainment approaches the low-speed fl ow zone, the deposits form a lateral and/or an island head bar. In general, the lateral bar forms 10 to 20 m away from the island, generating a bar island (LELI, 2015;LELI et al., 2018). Mutum Island, a central bar island, was formed by this process. On the surface of the island, it is possible to observe the long longitudinal scars of the ancient bar-island channels that were formed during the island evolution (Fig. 4G).

Geomorphological approach
The anabranching pattern of the Upper Paraná River in this study has more than 200 islands of diff erent dimensions, morphologies, and geneses. Offchannel process islands form via channel avulsion and fl oodplain cutoff . It is a complex mechanism involving natural levee crevassing, crevasse channel formation on the fl oodplain, and crevasse channel re-junction to the river channel. In turn, the central bar island forms from the stabilization of a central bar followed by vertical aggradation of fi ne sediments deposited during the fl oods. Diff erent islands formed by the in-and offchannel processes indicate that the channel has gone through periods of diff erent hydrological conditions in the anabranching system formation. Island age varies when comparing the two forming processes. Absolute dating processed in diff erent island types (STEVAUX, 1994;LELI, 2015;LELI et al., 2020b) suggest that the island construction by in-channel processes is more active and has been present in the system since the beginning of the Holocene. Such stability and permanence are observed in the Mutum Island formed 8200 y BP, Porto Rico Island formed 920 y BP (ZVIEJKOVSKI et al., 2017), and the Três Ilhas Archipelago (Fig. 4G and 5) with islands that are 60 years old. Although the study section is aff ected by the Porto Primavera Dam, it is relevant to consider that the system still has a minimal condition for the inchannel process functioning. This was confi rmed by the occurrence of a large central bar upstream of the Três Ilhas Archipelago, which has been stable for 17 years and colonized by grassy and shrubby vegetation (Fig. 5).
The ages of the fl oodplain cutoff islands (formed by off -channel processes) varied between the Late Pleistocene and the Late Holocene. However, these ages are related to fl oodplain deposition and not that of the fl oodplain incision, which indicates island formation. Studies carried out by Stevaux and Souza (2004) reported a major avulsion event in Upper Paraná River in the Upper Holocene, around 2800 y BP. This event may be responsible for the generation of many fl oodplain cutoff islands in the river system. This was the last event that formed this type of island. Currently, the rare avulsions occurring in the system do not have suffi cient strength or magnitude to excavate secondary channels in the plain (Fig. 7). Fortes et al. (2004) suggested that the occurrence of neotectonic events in the Holocene may have triggered major avulsion in the river system.
In both cases of island formation, the bedload supply and lateral bar formation are important for island expansion. Although the construction of dams in the upper basin, especially the Porto Primavera Dam, has reduced the transport of suspended sediment and bedload, the system continues to form lateral and central bars. The annexation of lateral and frontal bars to both types of islands is currently observed. However, it is impossible to determine the velocity of these processes and compare them with the pristine condition.

Hydraulic approach
Since the 1990s, multi-channeling has been studied more closely by river geomorphologists. The pioneering work of Nanson and Knighton (1996), Tooth and Nanson (2000), Huang and Nanson (2000), and Huang and Nanson (2007) not only proposed a new classifi cation for multichannel rivers, but also asked about the need for a river to promote multi-channeling. At that time, Huang and Nanson (2000) stated that the channeling could improve specifi c stream power (ω). According to the stream power equation (ω = γQs/w), ω is increased by increasing s and/or reducing w at the discharge (Q) and water density (γ) constants.
Based on this premise, the authors proposed the principle of maximum fl ow effi ciency. This theme was later mentioned by Latrubesse (2008), who applied this concept to justify the multiple patterns of the largest river systems in the world. According to the author, mega-rivers (Q> 17 000 m 3 s -1 ) must transport large amounts of water and sediment for hundreds to thousands of kilometers under very low hydraulic slopes (< 0.00005), thus justifying multi-channeling. However, this process occurs in rivers of any magnitude, including the small Australian courses in which Nanson and Knighton (1996) established the classifi cation of multichannel rivers. The iconic inquiry posed by the title of Huang and Nanson's article (2007) "Why some alluvial rivers develop an anabranching pattern" was answered in a detailed theoretical study. Using hydraulic equations, the authors concluded that, in the impediment to adjust the slopes, rivers reduce the width of the channel and increase its effi ciency. Practical evidence was presented by Gon (2012) and Stevaux et al. (2013), who calculated the effi ciency of the multiple channels of Upper Paraná River using the transported load. The authors compared the transport rates in nodal sections with those in multichannel ones (up to six channels) and concluded that as the channel widened, the number or width of the islands directly increased. In truth, as the width increased (from 2 to 6 km), the channel compensated by islands formation, maintaining the eff ective width (sum of the channel widths in one section) of approximately 2.5 km and the specifi c channel power at approximately 3.2 W m -2 throughout the multichannel reach.

Sedimentary approach
The sedimentary records of the Paraná River deposits are relatively simple, although applying facies analysis is diffi cult because of the few preservations of sedimentary structures (STEVAUX, 1994). A comparison of the depositional architecture model of Upper Paraná River with the classic model for anastomosed rivers proposed by Smith (1986) (Fig.  11) shows that the fundamental diff erence lies in the type of deposit stacking. The classic model, based on rivers under a strong tectonic infl uence, shows vertical stacking, generating vertical tabular bodies of sand wrapped in the mud. Such architectural evidence supports the generation of depositional space by tectonic subsidence. This model was quite popular among sedimentologists linked to oil prospecting to the point of stating that the formation of anastomosed pattern would only be possible in the case of areas subject to subsidence (SMITH, 1986;SHUSTER and STEIDTMANN, 1987). In the case study, it was evident that the generation of space by channel widening generates lateral (horizontal) accretion of deposits, without the need for subsidence. Unlike the classic tectonic-infl uenced model, it is quite likely that the Paraná River model could be the most common in the anabranching rivers of the world. Its architecture consists of a sandy basal lithosome covered by a muddy lithosome.

Conclusion
The Upper Paraná River has developed apparently active anabranching with the same hydrosedimentary characteristics as those estimated since the beginning of the Holocene. Although the basin underwent minor climatic changes during this period, such changes were not suffi ciently intense to signifi cantly aff ect the multi-channeling processes of the system. The changes imposed by the construction of dams throughout the upper basin likely aff ects the current hydrosedimentary processes. However, considering the immense size of the system, the reaction and relaxation times can span decades or even centuries.
Islands derived from the central bar, formed by in-channel processes, are directly related to the hydraulic balance of the system. Island formation is the way the river compensates for the increase in channel width. If the original dimension of the formed island is not suffi cient to maintain fl ow effi ciency, the island grows by the annexation of lateral and frontal bars. In contrast, islands from fl oodplain cutoff s are not related to hydraulic balance, but for such allochthonous reasons as a) occurrence and the dimension of the natural levees, b) fl oodplain transversal and longitudinal slopes, c) fl oodplain surface morphology (occurrence of fl ood channels, paleo-natural levees, paleochannels, etc.), and d) space available for sediment accommodation (more "empty" plains tend to accumulate water and sediment favoring the formation of avulsion channels). Once a fl oodplain cutoff island is formed, it can annex laterally and frontally in the same manner as a central bar island.
Thus, future fluvial geomorphology research, specifi cally in Brazil, should focus on a) reformulating the classifi cation of multichannel rivers given the new results obtained in large river systems, mainly in the tropical belt; b) achieving a more detailed understanding of the hydraulics involved in the formation of multiple channels; and c) establishing sedimentary models, not only for island deposits, but for the fl oodplain itself. These topics suggest the need for new paradigms in fl uvial geomorphology to replace those defi ned in the 1950s and 1960s based on medium and small fl uvial systems in a temperate climate. These suggestions should be included in the objectives of the Oriented Working Group on the Fluvial Sub-system within the scope of the Brazilian Relief Classifi cation System (SBCR).