3.1 Braided belt width mobility and the implication of composite cohesive banks

In a highly braided river system, lateral migration and bank erosion are related to the complex flow–sediment dynamics and associated geomorphological activities in terms of bar disorderness, thalweg migration and intense water level variation. For rivers such as the Brahmaputra, these braided river dynamics are scaled up by many folds due to the aggravated sediment waves (Fischer et al., 2017; Karmaker & Dutta, 2010), macroturbulence structures (Sharma & Akhtar, 2017), variability in the vegetated landforms (Chembolu & Dutta, 2020), cohesive banks (Karmaker & Dutta, 2015) and nodal points (Chembolu & Dutta, 2018). Generally, in cohesive banks, erosion occurs after the breaking of electromechanical forces between the aggregates. Both shallow water and extreme flood events can be responsible for the shuddering of cohesive banks, leading to accelerated bank erosion. Interestingly, this cohesive bank erosion and timing of peak discharges are weakly related in the Brahmaputra (Sarker et al., 2014). This may establish that the bank erosion process in the Brahmaputra may not depend only upon the high energy expenditure potential of extreme events, rather gradual soil saturation by periodic low-moderate floods, deflection angle, longitudinal slope and local bed material size can affect the extensive bank erosion (Karmaker & Dutta, 2015; Mosselman, 2006; Surian et al., 2009).

The synthesis of literatures pertinent to fluvial bank erosion suggests a varying annual rate across the world's large rivers. For example, rivers like the Sacramento River (US), the Nile (Africa), the Tagliamento River (Italy) and the Yangtze (China) have an annual erosion rate up to 10–50 m/year (Ayman & Ahmed, 2009; Deng et al., 2019; Larsen et al., 2006; Surian et al., 2009). On the other hand, the annual bank erosion rate in the Brahmaputra is significantly higher in comparison to these notable dynamic river reaches. In fact, high bank erosion of 100 m/year (Mosselman, 2006; Mount et al., 2013) along with local scale erosion up to 1–2 km/year is even observed in these river reaches (Sarker et al., 2014). The spatio-temporal variability in the bank erosion process has further developed a dynamic braided belt width in the Brahmaputra River. The old maps of the 1800s in the Bangladesh region suggest the Brahmaputra may have followed a meandering channel pattern (Thorne et al., 1993). In the later period, with the increase of both fluvial and anthropogenic disturbances, the river has significantly altered the process–form relationships. The meandering planform is eventually replaced by a considerably wider braided channel pattern accompanied by a significant rate of bank line retreat. As suggested by Lawler (1995), one or combination of the processes such as sub-aerial erosion, fluvial entrainment and mass failure caused by weakened bank materials is majorly responsible for river bank erosion. It has been observed that, in the Brahmaputra river, the water level falls rapidly during the recession limb and the pressure against the wall decreases, which resulted in the backwater movement from the bank and causes subaqueous failure by lateral flowage of sand and silt to the river (Karmaker & Dutta, 2011). Furthermore, the river experiences frequent channel migration, which enables the migrating channel intersects the bank line at some locations. Eventually, that uplifts the water level in the bank, saturates the sand and silt, and in the subsequent falling limb, saturated sediments are liquified by forming well-defined shear planes. These processes contribute to the braided belt dynamics of the Brahmaputra River. The mean width of the Brahmaputra River in Assam valley is about 5949 m during the year 1912–1928 (Sarma & Acharjee, 2018). The mean width has increased up to 7455 m due to intense bed aggradation after the great Assam earthquake of 1950 (Goswami, 1985; Takagi et al., 2007). The braided belt width has rapidly increased up to 1980–90s and gradually stabilized in the next two decades (Takagi et al., 2007) (reproduced from Karmaker et al., 2017 and integrating recent satellite imageries). In the 2000s, the mean braided belt width of Brahmaputra River is observed close to 9012 m, with distinct nodal locations and multichannel segments. Figure 4a shows spatio-temporal variability of braided belt width for the middle stretch of the Brahmaputra (approximately 240 km). The percentage change in braided belt width variation is higher in the 1970s as compared to the 1990s, 2000s and 2010s. This further indicates the reach is approaching a dynamic stable state after the 1970s. However, in recent years, the braided belt width variability is relatively higher, resulting in a dynamic, unstable (or transient) state. The downstream stretch from Guwahati is morphologically more dynamic in terms of braided belt width variability, which is more than 80 percentages (in the early 1990s) and 30 to 60 percentages (in the late 2010s), respectively.

The temporal dynamics in braided belt width (channel instability) can be attributed to numerous hydrological (floods and sediment wave), topographical (nodal points and local bed slope) and geological factors (bank erodibility, earthquakes and landslides). The Brahmaputra River has experienced a sequence of geomorphological stages (1950 to 2020) with an interval of 5 to 10 years (Takagi et al., 2007). Furthermore, during the monsoon season, this river generates a considerable quantity of energy in the system due to a sequence of large flood waves. The morphological parameters mainly affected by this huge incoming energy are thalweg shifting (sinuosity), bank erosion, deposition, alteration in bed slope, channel widening and so on. According to previous studies, flood stream power arguably regulates the morphodynamics of the braided rivers (Akhtar et al., 2011; Bizzi & Lerner, 2012; Magilligan et al., 2015). However, in case of Brahmaputra River, the stream power has no direct influence on the braided belt variation (Karmaker et al., 2017). The fluvial energy is dissipated within the system by forming a series of braided loops, resulting in the generation of new secondary/tertiary channels as well as the blockage of the existing defined water ways (Chembolu & Dutta, 2016). This process leads to the increase in planform disorderness of the system. Additionally, the energy associated with the rising peak flood has a significant erosion potential, which can be capable of reshaping the shallower channel to deeper, washing out smaller unstable (mobile) bars and triggering acute bank erosion. Part of the fluvial energy is also stored/absorbed within the system eventually before it initiates a new planform change (Charlton, 2007). A conceptualization of the channel instability in the Brahmaputra River basin integrating the studies of Goswami (1985), Takagi et al. (2007), Karmaker et al. (2017) and Chembolu and Dutta (2018) is shown in Figure 4b. After the 1950s, the extreme earthquake and series of landslides stimulated high sediment loads completely dominated the morphological processes. This resulted in an instability of the braided belt width and initiation of (de)stabilizing waves. The propagation of these waves was unclear in the succeeding decades, and the river was in a quasi-dynamic equilibrium state. But from the 1980s onward, the river was morphologically active and showed high fluctuations in the braided belt width with complex landform distributions. The sandbar disorderness study has established a stabilizing state in the first decade of the 2000s. However, after the 2010s, the river has again increased the channel instability in terms of bar disorderness as observed from the recent satellite imageries and field observations. This knowledge indicates a higher magnitude of energy is expended during the braided belt adjustments, which cannot be overlooked and needs to be addressed in process-based understanding.

3.2 Channel bar dynamics and formation of large-scale confluence–bifurcation

Bars are the key geomorphic unit and the fundamental building block of a braided river system. Several braiding formation processes are formulated, entailing the bar development as a single unit or assemblages. In most cases, braiding is related to the conversion of the alternate bar to a multiple row bar or mid-channel bar via cut-offs formation. In sandy braided rivers, the lobate bars are known as ‘unit bars’ (Ashmore, 1991a; Smith, 1974), which have individual downstream restrictions with avalanche faces (Bridge, 1993). However, with the progression of time, these unit bars are gradually deposited, sculpted, or dissected and developed into braid or compound bar in the braided rivers (Bridge, 2003; Bridge & Lunt, 2006). Depending upon the mobility of bars, forced bar (less migration rate) and free bar (high migration rate) are taxonomized for braided river systems (Bridge, 2003).

The growth and dynamics of confluence–bifurcation are also key features of braided channel networks (Ashmore, 2013; Leopold & Wolman, 1957; Tubino & Bertoldi, 2007). A number of studies have focused on understanding the formation and evolution of channel bifurcations through laboratory experiments (Ashmore, 1991b; Bertoldi & Tubino, 2007; Federici & Paola, 2003), field observations (Ferguson et al., 1992; Richardson & Thorne, 2001; Zolezzi et al., 2006) and numerical modelling (Bolla Pittaluga et al., 2003; Jagers, 2003). Ashmore (1991a) has reported four dominant mechanisms responsible for bifurcation commencement in braided rivers, such as (i) chute cut-off, (ii) initiation of the central bar, (iii) transverse bar alteration and (iv) multiple bar braiding. Bifurcations may also be symmetric or asymmetric depending upon the division of flow sediment an individual branch. In particular, the sand-bed rivers (relative to gravel-bed channels) are more prone to symmetrical bifurcations due to the increased Shield's stress (Best et al., 2007). It should also be noted that the flow–sediment partitioning of the active channel relies on the bifurcation dynamics (Bolla Pittaluga et al., 2001; Bristow & Best, 1993), and this activity may switch between the individual branch depending upon the upstream bar migration rate. The confluence node points serve as an interface zone for upstream lateral erosion and downstream deposition. Similar to the bifurcation nodes, confluences may also have symmetric and unsymmetric formations. To understand the dynamic confluence morphology, precise field observations of the flow structure near the zone, shear stress distribution and sediment discharge pattern are necessary (Ashmore et al., 1992; Ashworth et al., 1992; Davoren & Mosley, 1986; Ferguson et al., 1992). Numerous factors also influence the morphology of confluences, including total and relative discharge at confluent anabranches, bedload yield to the confluence, boundary shear stress, particle size distribution, angle of confluence and several anabranches (Ashmore, 1993; Mosley, 1976). In addition, the confluence angle and discharge of the anabranches control the size and depth of scouring (Ashmore, 1993, 2013). In the case of high angle confluences and equivalent anabranch discharge, scour depth is more expensive. This scour depth is marginally less for considerable variation in anabranch discharges, irrespective of confluence angle (Ashmore & Gardner, 2008; Ashmore & Parker, 1983; Best, 1986; Mosley, 1976).

In the Brahmaputra River, the formation of bars along with confluence–bifurcation nodes can be attributed to a number of dominant hydrogeomorphological factors. Due to its complicated flow–sediment discharge characteristics, the Brahmaputra is one of the few braided river systems in the world that exhibits dynamic morphological behaviour (Dubey et al., 2021; Goswami, 1985; Karmaker & Dutta, 2011, 2013; Pradhan, Dutta, & Bharti, 2021). A very limited number of studies have been conducted to understand the sandbar dynamics (growth/adjustment/propagation) in this river system. The formation of braided patterns in the river is mainly determined by the formation/growth of sand bars. In addition, the sandbar dynamics are linked to the series of erosion–deposition processes in the river system. During high-flooding episodes, the incoming energy to the river system eventually regulates the underlying sandbar dynamics. During this period, morphological processes such as bar displacement and river channel fragmentation become prominent, which disperse a significant amount of energy. Furthermore, it governs the grain size distribution of the sediment material and its conveyance to a bar head, which influences the size and stability criteria of a bar. Depending upon the hydraulic regime and reach parameters, the river bed configuration (bed roughness) also provides varied resistance to the flow, which also dissipates a substantial amount of energy (Dubey et al., 2020; Pradhan, Chembolu, Bharti, & Dutta, 2021). According to Dubey et al. (2020), the channel has reduced roughness during the monsoon season compared to the non-monsoon season, which results in a steady river morphology in the course of monsoon period. During high floods, the flow energy increases and results in the erosion of mobile bars, which then gradually falls after the flood recedes, and irregular deposition of sediment material is observed (Chembolu & Dutta, 2018). At first, the river may encounter up to nine flood events of varying durations and sediment transport in a year (Karmaker & Dutta, 2010) and develop geomorphic units with multiple inundation surfaces. In addition, the substantial reduction of bed elevation at the India–China border (a drop of around 4800 m within a length of about 1700 km) and change in average bed slope from 2.8 to 0.1 m/km in the Assam valley have produced a large braided channel of high energy dissipation potential. The nodal points also facilitate the channel to scour at the local scale and eventually deposit the excess sediment load across the widened braided belt sections. Table 1 shows the development of channel bars (based on mobility) and probable bifurcation–confluence statistics along two braided reaches (Tezpur to Guwahati and Guwahati to Jogighopa) of the Brahmaputra. The morphometric analysis is performed focusing on the sandbar dynamics. The informations in Table 1 were obtained by analysing the 20 years (2000 to 2019) of satellite data using Google Earth Engine (GEE) platform. The data extracted from GEE were then processed in ArcGIS environment to determine statistics related to sandbar area and bifurcation/confluence configurations. The modified normalized difference water index (MNDWI) (Han-Qiu, 2005; Xu, 2006, 2007) was used with a threshold value (MNDWI < 0) to identify the sand bars. Further, the sand bars were categorized into seven different classes based on their surface area to understand the influence of bar dynamics on the morphological behaviour. Baki and Gan (2012) reported, the large bars are generally referred to as forced bars with low mobility (or immobile) rates in terms of shape, size and location, whereas small bars are referred to as free bars with high mobility rates that further play a key role in morphological changes. In the case of the Brahmaputra River, huge immobile bars form the large islands, whereas smaller mobile bars appear to have transient characteristics. It is interesting to note that bar surface area varies from 0.0006 to more than 100 km², and clusters of mobile, moderate–low mobile and immobile bars have formed as instream geomorphic units. Most of the bars can be classified as mid-channel longitudinal and diagonal bars, and occasionally side-channel and lateral bars are developed close to the cohesive high banks. Furthermore, more than 800 channel bars (temporal average of 20 years) have produced a complex braided planform of varying confluence–bifurcation geometries. The mobile bars have mostly asymmetric bifurcation (or confluence) configurations, whereas the large bars (surface area > 1 km²) are symmetric in nature. Similarly, the configuration angle is less (close to 30–45°) for mobile bars, which gradually increases for static mid-channel islands.

Figure 5 illustrates in-channel geomorphic units of varying surface water occurrence (SWO) in the Brahmaputra. The active low-flow channel flows through numerous confluences and bifurcations and develops transient morphological features across the variable braided belt (Figure 5a). The low-flow channels have SWO varying between 80% and100%, signalling permanent water availability. However, the thalweg is not defined at few locations, which provides notable lateral freedom to the channel and attacks the fixed bars. The fixed bars are mostly emerged (SWO below 20%) and periodically subjected to sculpting during the high flows. The chutes are developed on the surface resulting in short-circuiting of flood channels. The bar assemblages are highly dynamic in nature resulting in a planform disorderness at spatio-temporal scales. For example, in the pre-monsoon season, the Brahmaputra is dominated by medium to large-sized sand bars. However, in the post-monsoon season, the same bars are dissected to smaller mobile and transient bars (SWO of 45% to 60%) and increases the entropy of the fluvial system. These smaller sized bars are mostly submerged during floods and subjected to migration in the highly sediment-charged environment. Figure 5b signals the presence of probable channel-in-channel physiography of the Brahmaputra River. The low-flow channel (mega-channel) has maximum width up to 4–5 km and carries a significant portion of the flow–sediment discharge. In addition, these zones can be regarded as the active braided belt segment with a combination of dominant confluences and bifurcations. The fixed bars and floodplains can be considered as inactive braided belt zone, which are mostly active during high flows and extreme events. Therefore, a comprehensive evaluation of the bar dynamics along the active morphological reach is necessary and can aid in quantifying the energy utilization in braided physiographic settings.

3.3 Process–morphodynamics–vegetation interactions at landform scale

Instream vegetation is a vital component of fluvial systems and has a significant impact on river rejuvenation schemes by improving the physical features and ecological conditions (Bennett & Simon, 2004; Gurbisz et al., 2016, 2017; Gurnell, 2014; Kondolf et al., 2013; Lera et al., 2019; Nandi et al., 2021; Nepf & Ghisalberti, 2008; Vargas-Luna et al., 2018; Wilson et al., 2003). Similarly, riparian vegetation has a significant impact on river morphodynamics since it interacts with the flow–sediment discharge at different spatio-temporal scales, triggering a series of intricate feedbacks and linkages (Camporeale et al., 2013, 2019; Gurnell, 2014; Gurnell & Petts, 2002; Henriques et al., 2021). In addition, aquatic vegetation considerably adjusts the flow and turbulent structure and affect the riverine habitat (Bornette & Puijalon, 2011; Boruah et al., 2008), water quality (Dosskey et al., 2010), pollutant-nutrient distributions (Perucca et al., 2009; Shucksmith et al., 2010) and sediment transport process (Baptist, 2003; Bertoldi, Gurnell, et al., 2009; Gonzalez et al., 2019; Jordanova & James, 2003; Kothyari et al., 2009; López & García, 1998). Vegetation is also related to the morphodynamics by governing the braiding intensity (Belletti et al., 2015; Gran & Paola, 2001; Kui et al., 2017; Tal & Paola, 2007), promoting bar accretion and bank stabilization (Bertoldi et al., 2011; Gurnell, 2014). A number of laboratory-scale studies (Coulthard, 2005; Gran & Paola, 2001; Tal & Paola, 2010) have been performed to understand the vegetation interactions with braided river morphodynamics integrating non-natural rigid (Bennett et al., 2002; King et al., 2012; Liu et al., 2008) and flexible cylindrical stem (Chembolu et al., 2019; Chen et al., 2011; Velasco et al., 2008) submerged (or emergent) vegetation.

In the braided rivers, instream/riparian vegetations and aquatic plants represent the fluvial environment, climatic conditions and local geological factors. The vegetation can either actively or passively contribute to the morphological activity of rivers (Camporeale et al., 2013, 2019). The passive role represents the impact of vegetation on roughness, hydraulic resistance and erosion/deposition characteristics which have alike morphological and mechanical properties. On the other hand, the active role refers to the biotic process linked with plant community colonization and life cycle, which regulates the morphodynamics of the river. Furthermore, certain species have the ability to colonize exposed or flooded alluvial sediments and create pioneer landforms (Bertoldi & Gurnell, 2020). Thus, individual species or patches in a braided fluvial system can operate as an ecosystem concocts in a variety of environmental conditions. Through a spatio-temporal series of interaction between the vegetation and geomorphic processes, the associated energy in the braided reach can be dissipated by various means. The growing vegetation root networks has been demonstrated to stabilize sediment deposits by diverting the flow and strengthen the soil (Bywater-Reyes et al., 2018; Polvi et al., 2014), which further reduce the flow velocity (Vargas-Luna et al., 2015) and increase the deposition of sediment minerals (Cotton et al., 2006; Meier et al., 2013; Nepf, 2012). This also leads to the formation of pioneer planforms (bars/islands) which affects the local physical conditions and accelerates eventual colonization of different species (Gurnell, 2014). Additionally, the pioneer islands actively contribute to the hydrogeomorphological processes, giving rise to the establishment of a larger island with a broader space during the lean seasons and sagregated during the extreme flood events. In this way, vegetation can operate as a ‘ecosystem engineer’, aiding its development and establishment (Gurnell, 2014; Gurnell & Petts, 2006), which is the primary stairway towards the accretion of river banks, the formation of river islands, and, as a result, changes in river characteristics and planforms (Bertoldi et al., 2014; Wintenberger et al., 2015). There are limited studies available for understanding the interactions between active channels and instream/flood plain green corridors in the braided rivers. Further, for a highly braided river like the Brahmaputra, these bio-morphological processes become extremely complex and unpredictable due to the variable growth cycle of vegetation species and their controlling factors. The Brahmaputra has varied instream or floodplain vegetated landforms that affect the bio-morphological processes (Figure 6). During the high flows and extreme events, these vegetated structures offer resistance to the incoming flow and further help in stabilizing the submerged landforms.

To establish a framework for a more comprehensive investigation of island distribution, its character and dynamics, a precise understanding of an island in terms of size, form and structure is required. It is important to carry out this objective as size of island provides significant information on its stability-structure and can be used as a parameter for identifying islands from maps and aerial imageries (Gurnell et al., 2000). The braided rivers can be classified based on the ratio of island area to gravel (sand) area (IBR) as bar-braided (no island present), bar-braided with occasional islands (<0.25), island-braided (in between 0.25 to 0.5) and heavily island-braided (>0.5) (Gurnell et al., 2000). This classification is applied to the Brahmaputra by dividing the entire study reach into eight zones with lengths varying between 19 and 55 km (Figure 7a). The zones are defined by local nodals identified using the high-resolution global surface water maps (Pekel et al., 2016). The vegetation cover information was also mapped for the whole stretch using the normalized differenced vegetation index (NDVI) (Figure 7b). The GEE platform was used to obtain NDVI images for the selected study reach, and the vegetated pixels were identified using a threshold value (NDVI > 0.2) (Bertoldi et al., 2011). In addition, the vegetated area was calculated, and the previously determined sand bar area (Section 3.2) is included in the IBR value computation. Zones 1, 5, and 7 confirm to occasional islands category; Zones 2 and 3 fall under island-braided; and Zones 4, 6, and 8 represent heavily island-braided segments (Figure 7c). Interestingly, zone without vegetation cover on the bar surface is not present in the entire study area. The heavily island-braided zones (4, 6 and 8) (IBR of 0.62, 0.76 and 0.59, respectively) have higher braided belt variability (average of 10 to 11 km). These reaches accommodate larger bars that are stable in nature and provide a favourable condition for vegetation growth and succession. On the other hand, Zones 2 and 3 (IBR of 0.25, 0.76 and 0.40) featuring medium and small-size bars have a low island-to-bar ratio due to less growth of vegetation. However, Zones 1, 5 and 7 (IBR of 0.21, 0.20 and 0.22, respectively) have the lower island-to-bar ratio attributed to the presence of smaller size bars that are mobile in nature and possess the least vegetation cover.

In the Brahmaputra River, the island formation process begins with the deposition of fine sediments around vegetated landforms, and further trapping of sediments provides stability characteristics (Charlton, 2007; Latrubesse, 2015; Orton & Reading, 1993; Pradhan et al., 2020; Rogers & Goodbred, 2014). This concept encompasses varieties of interdependent processes occurring on instream and riparian vegetation, such as seed dispersal, recruitment, growth and mortality. The spatial scale hydromorphological cascade processes and the landforms appear according to vegetation growth-succession, which are further responsible for the bio-morphological stabilization, river rejuvenation and restoration (Figure 8). In the Brahmaputra, a multiscale approach can be considered (reach to catchment or catchment to reach scale) for instream or riparian vegetation characterization and succession. In the first approach, recruiting and establishing pioneer species on bare alluvial bars can be facilitated by water availability, flood disturbance and geomorphic unit recreation at the local scale (i.e., location-specific shear stress, time and frequency of flood) (González et al., 2018; Johnson & Jones, 2000; Karrenberg et al., 2002). In the second approach, scale-dependent dominancy of climatic conditions, geographic formation and progressively flow regime, valley contexts and local lateral flood plains control channel dynamics and geomorphic unit development and maintenance (Fryirs & Brierley, 2012). River corridor vegetation serves as a substantial energy dissipation component and has the ability to operate as a morphogenetic agent by exerting a large influence on the processes of erosion, transportation and deposition of sediment. Thus, it is decisive to acknowledge the process–form–interaction of instream-floodplain vegetation, which contributes to the stabilization of fluvial landforms and control the braided morphology.