The Planform Mobility of River Channel Confluences: Insights from Analysis of Remotely Sensed Imagery

River channel confluences are widely acknowledged as important geomorphological nodes that control the downstream routing of water and sediment, and which are locations for the preservation of thick fluvial deposits overlying a basal scour. Despite their importance, there has been little study of the stratigraphic characteristics of river junctions, or the role of confluence morphodynamics in influencing stratigraphic character and preservation potential. As a result, although it is known that confluences can migrate through time, models of confluence geomorphology and sedimentology are usually presented from the perspective that the confluence remains at a fixed location. This is problematic for a number of reasons, not least of which is the continuing debate over whether it is possible to discriminate between scour drivers of confluence mobility are broadly the same as those that drive channel change more generally. Thus in the GBM basin, a high sediment supply, large variability in monsoonal driven discharge and easily erodible bank materials result in a catchment where over 80 % of large confluences are mobile over this 40-year window; conversely this figure is less than 40 % for the Amazon basin. These results highlight that: i) the potential areal extent of confluence scours is much greater than previously assumed, with the location of some confluences on the Jamuna (Brahmaputra) River migrating over a distance of 20 times the tributary channel width; ii) extensive migration in the confluence location is more common than currently assumed, and iii) confluence mobility is often tied to the lithological and hydrological characteristics of the drainage basins that determine sediment yield.

drivers of confluence mobility are broadly the same as those that drive channel change more generally. Thus in the GBM basin, a high sediment supply, large variability in monsoonal driven discharge and easily erodible bank materials result in a catchment where over 80 % of large confluences are mobile over this 40-year window; conversely this figure is less than 40 % for the Amazon basin. These results highlight that: i) the potential areal extent of confluence scours is much greater than previously assumed, with the location of some confluences on the Jamuna (Brahmaputra) River migrating over a distance of 20 times the tributary channel width; ii) extensive migration in the confluence location is more common than currently assumed, and iii) confluence mobility is often tied to the lithological and hydrological characteristics of the drainage basins that determine sediment yield.

Introduction
River confluences are important nodal points in alluvial networks, often representing abrupt downstream changes in discharge, grain size and channel geometry, which in turn may exert a significant control on channel morphology, migration and avulsion (Mosley, 1976;Richards, 1980;Ashmore, 1991;Bridge, 1993;Ashmore and Gardner, 2008;Best and Rhoads, 2008). The morphology of the confluence zone also has many ramifications for understanding and managing aspects of river behaviour, such as the fact that the dynamic morphological adjustments at these sites may make managing land use and infrastructure difficult (Ettema, 2008). Meanwhile, the morphological and geochemical heterogeneity often present at confluence sites has led ecologists to conclude that they are 'hotspots' of high biodiversity (e.g. Benda et al., 2004), and/or may form sites of appreciable biological change (e.g. Rice et al., 2008). Even at confluences that possess a relatively stable planform location, the hydraulic processes at junctions are still highly complex, which makes understanding of pollutant pathways, for example, problematic (Biron and Lane, 2008). In the present paper, we focus on exploring the planform morphodynamics of large confluences and linking this to the subsurface sedimentology River confluences have the potential to create some of the points of deepest incision into underlying sediments (Mosley, 1976;Best, 1988;Bristow et al., 1993;Salter, 1993;Siegenthaler and Huggenberger, 1993;Best and Ashworth, 1997;Miall and Jones, 2003;Ullah et al., 2015) and hence their subsequent fill has been argued to possess the highest preservation potential of fluvial channels (Huber and Huggenberger, 2015). Since the depth of junction scour and mobility of the confluence are determined by flow processes in the confluence hydrodynamic zone (Best and Rhoads, 2008), it can be argued that differing junction dynamics may produce a range of characteristic confluence zone sedimentology from sandy bar development to mud-filled scours. Furthermore, understanding the planform mobility of confluences, and thus the potential spatial extent of basal scour surfaces, particularly in large rivers, is key to interpreting alluvial stratigraphy and discriminating between autocyclic and allocyclic scour surfaces (Best and Ashworth, 1997;Fielding, 2008), reconstructing palaeohydraulics and channel sedimentary architecture (Bristow et al., 1993;Siegenthaler and Huggenberger, 1993;Miall and Jones, 2003), as well as identifying potential sites for hydrocarbon exploration (Ardies et al., 2002).
Despite the fact that the sedimentary fill of confluences may be preferentially preserved and that their large scale may lead to confusion in discriminating between autocyclic and allocyclic scour, to date there has been no comprehensive analysis of confluence mobility to resolve questions concerning the extent and ubiquity of migrating confluence locations. For example, Holbrook and Bhattacharya (2012) question whether confluences can migrate sufficiently to produce a scour large enough to resemble that of an incised valley, and hence be mistaken for a product of allocylic-driven erosion. However, some case studies, such as the confluence of the Ganges and Jamuna rivers, Bangladesh, show junction migration over distances of several kilometres in a year (Best and Ashworth, 1997). In addition, the course of the Jamuna River has also been shown to avulse on centennial to millennial timescales Pickering et al., 2014;Reitz et al., 2015), thus changing the location of its confluence with the Ganges River by hundreds of kilometres. High-angle confluences in meandering rivers have also been demonstrated to adjust their confluence planform over decadal timescales (Riley, 2013). Ettema (2008)   as well as changes in confluence morphology (Graf, 1980;Petts, 1984;Allen et al., 1989;Grant et al., 2003;Gilvear, 2004;Petts and Gurnell, 2005;Phillips et al., 2005) There is a broader theoretical basis for assuming confluence location and morphology may change substantially over time. Mosley (1976) showed that confluence morphology (Figure 1) is dynamic and responds and adjusts to upstream boundary conditions of flow and sediment supply in each tributary, and thus confluences may be expected to adjust to three broad factors. Firstly, upstream boundary conditions of discharge, or momentum, ratio between the tributaries, where momentum ratio exerts a control on scour morphology (Mosley, 1976;Best, 1986;Best, 1988;Best and Rhoads, 2008) and tributary bar morphology (Best, 1988;Biron et al., 1993;Rhoads, 1996;Biron et al., 2002;Boyer et al., 2006;Best and Rhoads, 2008). There is also some evidence that inter-event fluctuations in momentum ratio can lead to changes in bar morphology (Boyer et al., 2006), and where tributaries drain different lithological or climatic areas there could be annual or seasonal variations in momentum flux. Secondly, junction angle controls both scour morphology (Mosley, 1976;Best, 1988;Sambrook Smith et al., 2005) and tributary mouth bar morphology (Best, 1988). Where the channels upstream of the confluence are meandering, the junction angle could thus change over time in response to bend migration and channel cut-off.
Finally, formation of a mid-channel bar in the postconfluence channel (Mosley, 1976;Best, 1988),can occur through convergence of sediment transport pathways (Best, 1988;Best and Rhoads, 2008) and declining flow velocities and turbulence intensities downstream of the zone of maximum flow acceleration (Best, 1987;Best, 1988;Sukhodolov and Rhoads, 2001;Rhoads and Sukhodolov, 2004).
Such bar formation can promote bank erosion and channel widening (Mosley, 1976) below are used to help identify these key controls (section 4) from which an overall classification is derived (section 5). The rationale for focusing on large rivers is briefly outlined below.   Mosley, 1976;Best, 1988;Roy and Bergeron, 1990;Best and Roy, 1991;Biron et al., 1993;Kenworthy and Rhoads, 1995;Rhoads and Kenworthy, 1995;Rhoads, 1996;Rhoads and Kenworthy, 1998;De Serres et al., 1999;Rhoads and Sukhodolov, 2001;Biron et al., 2002;Boyer et al., 2006;Leite Ribeiro et al., 2012), and it is only with recent advances in technology that the direct field investigation of large river confluences has been possible (e.g. McLelland et al., 1999;Ashworth et al., 2000;Richardson and Thorne, 2001;Parsons et al., 2005;Parsons et al., 2007;Lane et al., 2008;Parsons et al., 2008;Sambrook Smith et al., 2009). There is therefore a need to critically examine, describe and quantify the decadal morphodynamics of large river junctions in order to better understand the extent to which river confluences are mobile, how mobility is expressed and the rates of change. With recent advances in remote sensing, the planform characteristics and decadal evolution of large rivers can be described in greater detail (Ashworth and Lewin, 2012;Trigg et al., 2012;Lewin and Ashworth, 2014a), and the temporal morphodynamics of large rivers can be quantified (e.g. Mount et al., 2013).  conversely, it is assumed that bare sediment has been disturbed by channel processes within a time frame that is not greater than that required for the establishment of vegetation. Therefore, "within braidplain" migration is defined as the reworking of exposed sedimentary material assumed to be within the active braidplain, whilst "braidplain migration" is Where confluence angle (see Figure 1) is reported, this was measured using the approach of Hackney

Styles of confluence evolution
This section presents data on 14 large confluences (summarised in Table 1) (Table 1).

Bar Migration in Tributary Channels
The   Past research (Mosley, 1976;Best, 1988)  that is driven by island and bar migration. Bank erosion along these rivers is relatively low due to the presence of permafrost, with lateral channel migration rates of 2-4 m yr -1 (Are, 1983;, whereas downstream island migration is an order of magnitude greater , with rates up to 40 m yr -1 Costard et al., 2014). The junction between these two rivers is occupied by many braid bars and thus the confluence zone consists of multiple smaller junctions rather than one single confluence.
In this case, it is likely that a series of smaller, mobile, confluence scours may yield a more complex pattern of intersecting scour surfaces and scour fills linked to the migration of these smaller junctions.

Tributary Channel Migration
In a multi-channel river, the migration, bifurcation or avulsion of tributary channels within a braid belt will cause corresponding migration and/or avulsion of the confluence location, and thus drive channel mobility at a greater spatial scale relative to active river width than that mediated by bar dynamics within the confluence zone. The width of the active channel belt of a multi-channel tributary therefore sets the potential migration length of the confluence location. An outstanding example of a confluence driven by channel migration is that described by Best and Ashworth (1997)     bend (labelled 1 in Figure 6), whilst the bends in the     Australia, is now also fixed ( Figure 13), but this imagery indicates that confluence mobility can

Controls on confluence evolution
The preceding examples illustrate that confluences can adjust their planform position over a range of relative spatio-temporal scales and that such changes can occur in a broad range of river planform types. Some inferences concerning the processes that may be driving the style and rate of change observed at these confluences are now discussed briefly, focusing on the role of discharge, sediment supply, tectonics, climate, bank material and human influence.
In broad terms, it would appear that the same drivers of channel planform change are also responsible for controlling confluence evolution.
Thus it might be expected that confluences in areas with high rates of sediment supply, high water discharges and easily erodible banks would be highly mobile, due to bar migration driving changes in channel orientation and location, thus resulting in alluvium (Rittenour et al., 2007) and have a high suspended sediment load that contributes to the formation of abundant islands and bars that can become stabilised by vegetation (Knox, 2008). The rates of channel migration in the Mississippi River were quantified by Hudson and Kesel (2000) who showed an average meander bend migration rate for the 825 km section of the lower Mississippi containing the Arkansas confluence to be 38.4 myr -1 .
However, for the four measurement points closest to the confluence, there is an average meander bend migration rate of around 60 m yr -1 (Hudson and Kesel, 2000). The Arkansas River provides a large input of medium sand to the main river and the shallower slope of the Mississippi River in the vicinity of the confluence, as compared to up-and down-stream (Schumm et al., 1994), promotes deposition of this sediment input. The high sediment load in both the Arkansas and Mississippi rivers, coupled with the easily erodible floodplain, and possible paucity of clay plugs restricting migration in this region (Hudson and Kesel, 2000) contributes to rapid bank erosion in the Arkansas River, with rapid migration of the meander bends yielding rapid changes in confluence location.
In contrast, where there is significant geological control, confluences may be essentially static over decadal timescales, as illustrated by the confluence of the Congo and Kasai Rivers (Figure 12). At this location, the confluence remains fixed due to the inability of either channel to migrate laterally in the presence of bedrock control. Changes in climate may also lead to a change in confluence dynamics, as is likely in the case of the Murray-Darling rivers ( Figure 13). River discharges in this region were much higher than at present during the last glacial maximum (LGM) through to the early Holocene (Page et al., 1996;Nanson et al., 2008;Fitzsimmons et al., 2013), with channel size and lateral migration decreasing since the LGM (Nanson et al., 2008;Fitzsimmons et al., 2013). The average annual flood and long-term mean annual discharge have also been reduced substantially over the later part of the 20 th century by human intervention through water diversions, and the construction of dams (Maheshwari et al., 1995) and over 3600 weirs (Arthington and Pusey, 2003). As a result, the present day Murray-Darling River has a remarkably low annual discharge for its catchment area (Maheshwari et al., 1995;Arthington and Pusey, 2003), resulting in a confluence with no detectable movement over decadal timescales. Within-channel engineering works have also had a direct impact on the movement of the Padma-Meghna and Yangtze river confluences described herein, by introducing an artificial hardpoints that prevent the migration of these junctions.

A new classification of planform confluence behaviour
A new classification of confluence morphodynamics over management timescales is proposed herein        (Syvitski et al., 2005), that is reasoned to reduce the rate of morphological change at the junctions, and may have contributed to fixing the planform morphology. The preponderance of fixed confluences in the Meghna basin could be due to its low sediment yields compared to the Ganges-Brahmaputra, with the Meghna contributing ~12% of the GBM water discharge but just ~2% of its sediment load (Milliman and Farnsworth, 2013). Although the Meghna River drains the tectonically active uplands of the Shillong Massif crustal block and the Tertiary mud-and sand-stones of the Indo-Burman foldbelt (Mukherjee et al., 2009), most sediment yield is extracted within the subsiding Sylhet Basin upstream of the confluence (Goodbred et al., 2003) In contrast to the GBM basin, in the Amazon Basin Overall, the main channel of the lower Amazon system has low sinuosity, and is entrenched and confined to its valley over a scale of hundreds of kilometres (Mertes et al., 1996;Mertes and Dunne, 2008). Here the combination of intracratonic deformation and structural highs results in a channel system that is relatively immobile (Mertes and Dunne, 2008), with structural features such as the Purus and Garupá arches (Figure 16 geological inset) promoting entrenchment of the river and restricting channel movement (Mertes et al., 1996).

Confluence Type GBM Amazon
Thus, as the morphodynamics of junctions are inextricably linked in scale and process to the morphodynamics of their confluent channels, the junctions of the lower Amazon are also immobile.
It has been argued that deep confluences have a high preservation potential in the rock record (e.g. Huber and Huggenberger, 2015),  (Mertes et al., 1996;Mertes and Dunne, 2008;Constantine et al., 2014) and thus also fixed confluences over longer timescales.

Sedimentological implications of confluence mobility
Identifying the type and scale of erosional surfaces in the sedimentary record is important for reconstructing palaeoenvironments and palaeoenvironmental change (Bristow et al., 1993;Miall and Jones, 2003

Examples of the former include the Mississippi and
Arkansas river confluence, which moved ~ 5km (or 4 channel widths), and the Ganghara and Sarda River confluence which moved ~ 23km (or more than 11 channel widths) due to upstream channel avulsions.
These examples were typically complete within 10 years, with abandoned channels appearing to infill rapidly. Other larger-scale channel avulsions, such as that of the Brahmaputra in the late 18 th century (Best et al., 2007), may also relocate the locations of major river confluences by large distances, in this case by approximately 125km.

Conclusions
The planform morphodynamics of river confluences have received little attention in the literature, potentially leading to a perception that such junctions tend to be fixed nodal points within a channel network. The case studies presented herein demonstrate that, far from being fixed, confluences in large rivers can display a range of adjustments in response to external forcing. These adjustments range in scale from within-channel change, to bar deposition and erosion within the confluence zone, to channels migrating within a defined belt via meandering or braiding, to highly mobile confluences that migrate an order of magnitude greater than the channel width.
Initial basin-wide analysis of the patterns of confluence mobility for the Amazon and Ganges-Brahmaputra-Meghna rivers, suggests that confluent channels with high sediment loads have a higher probability of being mobile, in contrast to confluent channels with low sediment loads (such as in cratonic settings) that are more likely to be fixed.
Where tributary channels have a braided planform, confluence mobility is likely to be high and driven by changes in the position of dominant flow within the braid belt(s). In meandering channels with high sediment loads, the confluence location will be strongly dependent on meander neck cut-off in the tributary channel(s). Where the tributaries have any combination of very low sediment loads, low discharge variability or banks with high resistance to erosion, confluences will likely be fixed in their positon or migrate far more slowly.
The present results suggest several implications for the interpretation of scour surfaces in the stratigraphic record and reconstructions of past environmental change. Mobile confluences may generate scour over an area much wider than that of the channel width at the junction, thus generating significantly larger, and more complex, erosional surfaces than suggested in previous models (Bristow et al., 1993). The The present study highlights the need for further research into the scour and fill of large river confluences, in order to further refine the diagnostic criteria (Best and Ashworth, 1997) that may differentiate such scours Best, J.L., Roy, A.G., 1991. Mixing-layer distortion at the confluence of channels of different depth. Nature, 350, 411-413. Biron, P., Roy, A.G., Best, J.L., Boyer, C.J., 1993. Bed morphology and sedimentology at the confluence of unequal depth channels. Geomorphology,8(2), 115-129. Biron, P.M., . Modelling hydraulics and sediment transport at river confluences. In: S.P. Rice, A.G.