Analytical models of river evolution predict meander narrowing and elongation which creates sinuosity-driven hyporheic exchange across the meander neck, by decreasing flow distance and increasing head loss. We used a laboratory river table and close range photogrammetry to map and analyze sinuosity as a driver of head gradients and hyporheic exchange during cutoff. The river valley had relatively high slopes (1.8%) and moderately cohesive sediment (10% talc, 90% sand) to facilitate cutoff, and ratios of horizontal to vertical scaling were distorted to achieve dynamic similitude (Re = 3200). Incipient to cutoff, the head gradient across the neck increased due to a narrowing neck, upstream aggradation, and downstream degradation. Longitudinal and transverse river surface slopes around the meander bend increased as the meander approached cutoff. The steep head gradient across the moderately cohesive meander neck generated seepage erosion and scour that formed a low-sinuosity avulsion. Sediment-rich flow in the avulsed channel aggraded the downstream bed and separated the active channel and oxbow lake. The limitation in geometric and dynamic similitude in the river table limits extrapolation to natural rivers, yet river evolution may involve aggradation and degradation induced channel head loss and turnover hyporheic exchange as well as seepage-induced meander neck erosion. Our submillimeter maps of meander morphology and water stage provide data to parameterize river evolution and hyporheic exchange models, and may inform analysis and mapping of field sites.
River evolution that increases meander sinuosity ultimately causes river avulsion and oxbow lake formation in a process called meander cutoff. Oxbow lakes can form by chute or neck cutoff; chute cutoff occurs between the meander neck and apex, and neck cutoff occurs near the neck creating a larger oxbow lake and adjacent intrameander zone [Stølum, 1996; van Dijk et al., 2012] (Figure 1). The formation of oxbow lakes fundamentally changes local and regional ecosystem structure and processes [Knight and Welch, 2004; Ward, 1998], but less is known about changes to intrameander hyporheic exchange structure and processes during meander evolution through neck cutoff.
Such changes are poorly understood, in part because the neck cutoff process is difficult to anticipate and observe. The intrameander zone in alluvial rivers contains the mixing of river water and groundwater [Kasahara and Hill, 1997], which is termed hyporheic exchange and is an ecologically important process [Boulton et al., 2010]. Research into meander evolution using field observations [Constantine and Dunne, 2008; Erskine et al., 1982; Fares, 2000; Gagliano and Howard, 1984; Gay et al., 1998; Micheli and Larsen, 2011] and laboratory experiments [Braudrick et al., 2009; Schumm and Khan, 1971; Smith, 1998; van Dijk et al., 2012] has not provided detailed observations of changes to intrameander hyporheic exchange structure and processes. Analysis of meander evolution using space for time substitution has not provided a coherent and detailed set of intrameander hyporheic exchange observations [Bartley and Rutherfurd, 2005; Fares, 2000].
Established analytical models of river evolution with increasing sinuosity generate river planimetry with statistical properties equivalent to real-world meanders [Camporeale et al., 2005; Ikeda et al., 1981; Johannesson and Parker, 1989; Sun et al., 1996], where sinuosity is the quotient of river distance and valley distance. These analytical hydro-dynamic models simulate meander migration leading to increasing sinuosity, using equations that assume no groundwater interaction, constant river width, and instantaneous adjustment of flow and bed topography to the evolved planimetry [Zolezzi and Seminara, 2001]. Boano et al.  and Revelli et al.  used river stage and planimetry output from an analytical river evolution model to parameterize a groundwater model and determine how increasing sinuosity affected intrameander hyporheic exchange. They found sinuosity determined head gradients, which together with sediment hydraulic conductivity in their groundwater model, determined rates of intrameander hyporheic exchange [Boano et al., 2006; Revelli et al., 2008]. Head gradient is defined as the quotient of head loss and transect distance (e.g., transect N-N′ in Figure 2), where head is piezometric head and does not include velocity head. Head loss across the intrameander zone is determined by the difference in river stage across the transect bisecting the meander. The analytical river evolution model predicted that longitudinal distance around the meander determined the decrease in river stage due to a decreasing bed elevation, with minor variations in the river stage caused by secondary currents creating superelevation in the meander apex [Revelli et al., 2008]. As the meander evolved and sinuosity increased, the model predicted larger head loss around the meandering channel and shorter transect distances across the meander [Revelli et al., 2008].
This sinuosity-driven increase in head gradient leads to increasing rates of hyporheic exchange across the intrameander zone, which has been described as intensification [Revelli et al., 2008; Han and Endreny, 2012]. We define spatial intensification as an increase in lateral hyporheic exchange rates from the apex toward the neck for a given meander form (e.g., from transect M-M' to N-N' in Figure 2). And we define temporal intensification as an increase in lateral hyporheic exchange rates at a given transect location with evolving meander form before cutoff. Observations of river stage, groundwater level, and morphology during cutoff are desired to examine the role of sinuosity and other controls on intrameander head gradients and hyporheic exchange. Han and Endreny's  river table experiment documented spatial and temporal intensification of hyporheic exchange across the intrameander as sinuosity increased, but several aspects of the experiment were limiting, including: each meander was carved rather than sequentially evolved which prohibited observations of evolutionary transitions in head gradient; river bed elevation and stage were measured with a coarse scaled hook-gauge at only seven cross sections, which introduced uncertainty in estimates of head loss; and the relatively small river table (2.1 m × 0.9 m) prohibited distancing the meander unit from the influence of upstream and downstream reservoir conditions. As a result, high accuracy observations of an unconstrained meander evolution sequence are still needed for detailed analysis of changes to the intrameander river and head gradients. This study will map river bed, river stage, groundwater in observation wells, and document changes to intrameander head gradients and processes driving hyporheic exchange during meander neck cutoff in a laboratory river table.
2. Materials and Methods
2.1. River Table Setup and Experimental Procedure
Laboratory work was conducted in the James M. Hassett Laboratory at SUNY ESF using a mobile bed river table [Paola et al., 2009]. The river table was 3.66 m long, 1.83 m wide, and 0.2 m deep. Water entered the upper reservoir from a flow-stabilizing baffle and drained from the lower reservoir through a 5 cm diameter drainage hole and perforated downstream board into a tank (Figure 2). A recirculating pump and hose connected the tank and the flow-stabilizing baffle in the upstream reservoir. The flow rate was set to 80 mL/s by an Alix Digital Flow Controller. The valley slope and valley average head gradient was 1.8%, which is steep for meandering rivers but within the range of natural meandering pool-riffle rivers [Montgomery and Buffington, 1993, 1997; Schumm, 2005]. The substrate material was F-75 unground silica with a median diameter (D50) of 0.18 mm, D7 of 0.1 mm, D94 of 0.3 mm, with 10% talc powder added to the channel banks create a moderately cohesive substrate with adequate bank strength to avoid braiding and bank failure. The saturated hydraulic conductivity (K) value of the substrate at the bank was tested to be 7E-07 m/s, nearly two orders of magnitude smaller than the K for sand at the meander valley without talc (K = 2E-04 m/s). Substrate was placed in the river table to an average depth of 15 cm, and the experiment allowed for a mobile channel bed and unconstrained channel (Table 1).
Table 1. Initial Physical Model Parameters for River Table Experimentsa
Re and Fr values were computed using a characteristic length of four times the hydraulic radius.
River table width
River table length
River table substrate depth
Average channel depth
Average channel width
Median grain size
Reynolds number, Re
Froude number, Fr
Boundary shear stress
Critical shear stress
Grain Reynolds number
Weber number, We
The experiment used nomenclature of Boano et al.  to describe the meander evolution, with meander age and sinuosity increasing from stage M1 (young) to M2, M3, M4, etc., with M4 considered incipient to meander neck cutoff. The river meanders were initiated in the center of the river table and connected with channels of the same meander planimetry to distance them from the wall and upper inflow reservoir and lower outflow reservoir boundary conditions (Figure 2). The stage M1 meander had a river surface slope of 1.1%, and the M3 meander had a river surface slope of 0.5%, within the slope range given for pool-riffle meandering channels [Montgomery and Buffington, 1993, 1997]. The stage M1 meander had an initial sinuosity of 1.6 and an initial radius of curvature (rc) of 175 mm, and the M3 meander had an initial sinuosity of 3.4 and an initial rc of 123 mm (Figure 2). River discharge was kept at a subcritical flow (Fr < 1), which is representative of single thread meandering rivers. The river Reynolds number (Re) was 3200 based on four times hydraulic radius as the characteristic length, indicating that the channel was in transition toward fully turbulent flow. The dimensions, slope, and discharge placed the channel within the meandering regime, as defined by Parker , yet the relatively high valley slope and moderately cohesive substrate were intended to accelerate meander evolution.
The stage M1 meander did not evolve to M2 during our preliminary experiments, perhaps due to simplifications of keeping a stationary flow inlet, uniform flow regime, and moderately cohesive substrate without vegetation [Braudrick et al., 2009; Schumm and Khan, 1972; Tal and Paola, 2009]. We did not create a stage M2 meander, but do report M1 data to contrast the low-sinuosity channel head gradients to stage M3 and later meanders we observed. The stage M3 meander was carved into the river table and was allowed to evolve through neck cutoff. Our laboratory stage M1 and M3 meanders were scaled to Boano et al.'s  analytical M1 and M3 meanders, at 1:280 to their planimetric pattern and at 1:12.5 to their vertical depth. While critical parameters were scaled into normal meandering rivers (e.g., Reynolds number, Froude number), certain parameters remained distorted (e.g., planimetric scale versus vertical scale of river depth and well diameter) due to the dimensional constraints of laboratory simulations. Physical model distortion, however, is often unavoidable and accepted, and still yields valuable results that may reveal hydraulic and geomorphic drivers of change [Paola et al., 2009; Peakall et al., 1996]. Even in natural rivers, geometric and dynamic similitude between sites is challenging as meanders may range across meters to kilometers [Langbein and Leopold, 1966], introducing significant variation in riverbed relative roughness and open channel hydraulics.
Six observational wells monitored groundwater elevation in the river table valley and intrameander zone, kept some distance away from the eroding meander neck. The wells were 12 cm deep, fully screened, and 3–5 cm in diameter, which was an oversized diameter relative to the planimetric scale, but necessary to provide a shadow-free view of the well water surface for photogrammetric imaging. Observational wells (Figure 3) were compared against narrower test wells to ensure that oversized diameters did not create pool of water that experienced a lagged response times. Wells W1 and W2 were closest of all wells to the river table wall boundary, but in tests of boundary effects, where we incrementally moved well location away from wall, we detected no discernible effect on well response time and head level. No water table transverse slopes, perpendicular to valley slope, were present in the table, and MODFLOW groundwater [Harbaugh, 2005] simulations of the domain confirmed the groundwater at the well locations were not influenced by boundary effects or transverse slopes.
2.2. Data Collection
Photographs of the river table experiment were recorded with a pair of Nikon D5100 Digital Single Lens Reflex cameras, each equipped with a 20 mm prime lens, mounted 1.3 m above the river table surface (Figure 2). The meander stage M1 experiment ran for 6 h under a constant discharge. The meander stage M3 experiment ran for a nonconsecutive 50 h under constant discharge to allow meander evolution to stage M4–M7; the flow was intermittently interrupted for a total of 10 h to allow for river drainage and photographs of the river bed. The drainage of river water was controlled to minimize shear forces and there was no observed modification of the river bed morphology during drainage; river water drained into the sediment and exited the river table as groundwater. For the stage M1 and M3 experiments, initial conditions were established by running the flow for 3 h to reach a steady water table elevation; drainage of the river bed during evolution of stage M3–M7 did not completely drain the water table, and less time was required to reestablish an equilibrium elevation.
Eleven sets of stereo-pair photographs with nearly 100% overlap were taken during the experiments, capturing the river stage and bed surface at distinct meander evolution stages. Elevations of the river stage and groundwater well levels were captured using a novel method that involved covering the water with a fine wax powder. The wax powder was 20 mesh in size with a 0.8 mm maximum diameter and a 0.3 mm average diameter. During the experiment wax powder was uniformly distributed as seeds along the river and floated on the water surface, which created an opaque water surface needed for stereo-pairs. Each stereo-pair set was taken simultaneously using a Nikon ML-L3 remote control to avoid seed distortion caused by flowing water, and six sets were selected for data analysis, which were at meander stages M1, M3, incipient to cutoff (M4), immediately after cutoff (M5), 5 h after cutoff (M6), and 10 h after cutoff (M7). Each camera sensor was 16.2 Mega pixels, providing 0.3 mm resolution of images captured from 1.3 m above the ground for a 1.5 m2 area, which revealed submillimeter spatial and temporal changes. Triangular camera configuration steps were used to calibrate the cameras, and calibrated camera interior parameters were maintained by hard fastening the focus setting of the prime lenses with adhesive tape and switching off the camera autofocus function [Matthews, 2008; Han and Endreny, 2014]. The image planimetric coordinates were determined based on the distortion corrected images in 3DM CalibCam software. The vertical coordinates of the control points were measured using ultrasonic sensors (SICK UM30-2) with an accuracy of up to 0.3 mm.
2.3. Data Processing
Digital elevation models (DEMs) of the river table were generated using the 3DM CalibCam and 3DM Analyst software (ADAM Technology, Belmont, WA, Australia), and then exported into ArcGIS (ESRI, Redlands, CA, USA) for post processing. DEMs of the river bed, river stage, and groundwater in observation wells were extracted for each meander evolution sequence (Figure 3). DEM noise was removed by averaging elevation values in a 3 × 3 window, and comparisons of overlapping image areas determined DEM accuracy of 0.6 mm, equivalent to a 1% measurement error across the river table given its 1.8% valley average head gradient and 3.66 m length. The largest discrepancies in DEM values were in shaded regions, near steep river banks and the edge of observation wells. These areas were not used to collect head loss data.
Distance around the meander was quantified along a normalized curvilinear longitudinal coordinate s*, with points N and N' corresponding to the meander neck and points M and M' referring to the centroid (Figure 3). Distance around the meander, s* was divided into three parts of equal distance, [Han and Endreny, 2012; Revelli et al., 2008], and these chords shifted during meander evolution. Distance across the meander was quantified using straight-line transects, denoted with an overbar, i.e., . Head loss across a transect was computed as the difference in upstream and downstream river stage elevations, taken adjacent to the intrameander transect, i.e., at location N and N′ for transect . River stage adjacent to the intrameander zone was considered at equilibrium with groundwater levels inside the channel's right bank.
3.1. Head Gradient Changes
Head gradients across the meander neck monotonically increased along the evolutionary sequence until cutoff, due to increasing head loss and narrowing meander neck. The gradient started at 7.9% in meander stage M3, increased to 47.5% in stage M4 (incipient to cutoff) and to 52.6% in stage M5, at cutoff (Figure 4). The transect distance across neck transect decreased by 73%, from 140 mm at stage M3 to 38 mm at cutoff. Head loss across the meander neck increased by 150% during cutoff, starting at 8 mm in stage M3, and increasing to 20 mm at cutoff. The increase in head loss across the intrameander zone during cutoff (M3–M5) was primarily due to rising stage at neck location N, as backwater formed upstream of 4 mm of aggraded material, and lowering stage at neck location N′ due to 3 mm of channel degradation; upper and lower reservoir stage remained constant. Lateral migration at the meander apex was minimal between stages M3 and M5, and curvilinear distance only increased by 4% from 2084 to 2172 mm. With a 0.65% average river surface slope along curvilinear path , meander elongation accounted for 0.57 mm of the 12 mm increase in head loss; less than a 5% contribution to the head loss between meander stages M3 and M5. The cutoff process scoured the intrameander sediment at the neck, which decreased the upstream head and increased downstream base level, decreasing the head gradient across the neck. The breaching of the intrameander zone occurred when distance across transect was 38 mm (M5, Figure 4). Note that transect locations shift between meander stages to maintain equal curvilinear distances, . After cutoff, the length of the avulsion channel across the transect increased to 360 mm by stage M7. The avulsion caused head loss to decrease by 80%, from 20 mm at cutoff to 4 mm at the later evolution stage of M7 (Figure 4).
Changes in head gradient in the meander core transect were significantly smaller than changes in the neck transect , despite their proximity. Head gradients at the meander core transect increased from 1.0% at M3 to 1.4% at M4, and then decreased to 1.3% at M5, and to 0.2% at M6 and M7 (Figure 5), lower than the 1.8% valley head gradients. The increases in head gradients were driven by reductions in transect length and increases in head loss, which increased from 5 mm at stage M3 to 7 mm at stage M4. The increase in head loss was generated by meander elongation and changes in water surface gradients due to bed aggradation and degradation near transect . Meander elongation between stages M3 and M4, seen by development of compound meander planimetry (Figure 3), caused the curvilinear path to increase by 10%, from 690 to 757 mm. With a 0.8% average river surface slope, elongation accounted for a 28% of the increase in head loss, creating 0.55 mm of the 2 mm increased head loss. During cutoff, intrameander zone bank erosion caused transect length to decrease by 8% from 510 to 471 mm. After cutoff, head loss across decreased by 83%, from 6 mm at stage M5 to 1 mm at stage M7, while transect length increased by 5% to 496 mm. This relatively large reduction in head gradient at the meander core was attributed to changes in head loss as the oxbow lake formed and caused upstream stage to decrease and downstream stage to increase.
Changes in head gradient across valley transect were similar to changes at transect , and small compared with head gradient changes observed in transect . Head loss was monitored upstream of the meander neck, across transect , which was near the river valley wells W1 and W2 (see Figure 3). Of all transects monitored, head loss at was largest, which was explained by the transect connecting more distant river sections than were connected by or . Head gradient across transect was 2.7% at meander stage M3 and increased to 5.2% at stage M5, which was much smaller than the 52.6% gradient across the neck transect at M5. Head gradients in the valley groundwater, measured using head loss between wells W1 and W2, denoted by in Figure 4, were smaller than the gradients in intrameander transects and the transect . At evolution stage M3, the valley groundwater head gradient was 1.4%, increased to 2.5% at M4, and then decreased to 1.8% during cutoff at M5. Peak head gradient occurred at evolution stage M4 for both transect and the valley groundwater transect , while peak head gradient occurred at M5 for transects and , during cutoff. At evolution stage M6, after cutoff, both transect and the valley groundwater transect had head gradients reduced to ∼1%, while head gradient across transect was 6.6% at stage M6, reducing to 1.1% at M7. The difference in lagged timing and damped magnitude of head gradient values between transect and transects and , despite their physical proximity, indicates the localized impact of aggradation and degradation on head loss (Figure 4).
Longitudinal and transverse river surface slopes also experienced episodes of steepening and flattening during the meander evolution sequence, responding to the cutoff dynamics (Figure 6). The evolution from meander stage M3 to M4 increased the longitudinal slope in the curvilinear path around the meander bend (Figure 6), but meander neck cutoff at stage M5 flattened the longitudinal slope and initiated oxbow lake formation. The cutoff sequence also changed transverse river surface slopes in the meander apex. During evolution stage M3, the transverse slope had a classic superelevation pattern, with water surfaces higher along the outer bank due to curvature and river velocity generating secondary currents. At evolution stage M4, the upstream water depth increased and the transverse superelevation slope increased. With formation of the oxbow lake, during meander neck cutoff, the transverse slopes near the apex flattened (Figure 6).
Spatial variation of head in the intrameander zone was captured in groundwater equipotential surface maps, created for each evolutionary time step (Figure 7). These equipotential surfaces were computed using inverse distance weighted interpolation between river stage at the right bank, bounding the intrameander, and well water; river stage along the right bank was presumed to equal the adjacent intrameander groundwater. The equipotential surfaces capture the steeper head gradient in the neck region, as compared to the meander core, at meander stage M4 and M5, which presumably facilitates greater hyporheic exchange flows through the meander neck. After cutoff, however, there was a relatively uniform equipotential surface (401–405 mm) within the meander zone bounded by the oxbow lake, along . The change in equipotential surfaces between meander stage M3 and M4 at the meander neck transect illustrates the temporal intensification of head gradient, suggesting increasing hyporheic exchange. At cutoff, head gradients and hyporheic exchange no longer intensify, and begin to decrease, indicating the cyclical nature of these processes. As head gradients between the oxbow lake and intrameander groundwater flattened to less than 0.5% after cutoff (transect , Figure 5), head gradient upstream and downstream of the transect must increase to maintain the 1.8% average valley head gradient. We did not monitor upstream or downstream of the oxbow lake, but these larger head gradients could be localized at the channel bank, and would cross from the upstream valley to oxbow lake at curvilinear section , and correspondingly from the oxbow lake to the downstream valley at section . Head gradients between the valley, oxbow lake, and intrameander zone would continue to mix these waters.
3.2. Geomorphologic Changes
River bed elevation changes during the evolution sequence illustrate how meander cutoff involves lateral and longitudinal adjustment, avulsion, and sediment transport. At the meander apex prior to cutoff (M4, Figure 8), a weak pattern of lateral migration was identified by river bed scour erosion along the left bank (looking downstream) and deposition along the right bank near the pointbar. More proximate to the meander neck transect , elevation data reveal 4 mm of bed aggradation upstream of the neck and 3 mm of degradation downstream of the neck (M4, Figure 8), which increased longitudinal river surface slopes. Incipient to cutoff seepage-induced bank erosion at the meander neck was observed, which is a process that can occur with an adequately large ratio of Darcy flux, q, and critical seepage flow rate, qc [Fox et al., 2007]. In our experiment, bank erosion by river scour removed the talc modified substrate, exposing 100% sand substrate and increasing the hydraulic conductivity of the intrameander hyporheic exchange. Under these conditions q across the neck transect was estimated at 3.3E-04 m/s, which was adequately larger than the qc, estimated by Midgley et al.  at 8.5E-06 m/s, to induce seepage erosion. After the seepage-induced erosion narrowed the channel banks, a cascade then formed and scoured sand along transect . This lowered bed elevations near N by entraining sediment that had aggraded during the evolution of meander stage M3 and M4. This sand was transported across the neck and it aggraded the downstream zone near N′, raising bed elevations where stage M3 and M4 evolution had caused degradation (see Figures 4 and 8). Bed elevation data show how the evolution from meander stage M6 to M7 involved channel widening across the neck, causing a retreat of floodplain in the intrameander zone and in the valley zone bounding transect (compare floodplain extent between M6 and M7 in Figure 8).
Maps of river bed elevation and water stage revealed the heterogeneous array of pools, defined as deep water above localized scour, along longitudinal and transverse bed profiles (Figure 9). Longitudinal bed slope around the meander bend, along curvilinear path , trended with the longitudinal river surface slope prior to cutoff. Local bed slopes were highly variable, and changed during evolution. At the early meander stage of M1, there were four small and relatively deep scour pools located throughout the meander bend (Figure 9), but in the recarved M3 channel the meander bend had a single dominant pool along curvilinear path (Figure 9). During cutoff evolution, new pools formed by localized scour and the hydraulically forced morphology departed from the prototypical morphology of a single meander bend pool positioned opposite the apex pointbar [Montgomery and Buffington, 1997].
Sediment transport during the meander evolution sequence created an oxbow lake by depositing sediment and disconnecting flow between the active channel and curvilinear path The oxbow lake occupied the old channel denoted by the curvilinear path and the avulsed channel contained the active channel, with the thalweg denoted by an arrow in meander stage M6 and M7 (Figure 9). During the experiment, we observed a berms form along the upper and lower terminus of the oxbow lake (see orange color left of in M7 of Figure 9). The berm along the upstream section of the oxbow lake, near location N, was composed of sediment that had eroded from an upstream pointbar. The berm along the downstream section of the oxbow lake, near location N′, was composed of sediment that had eroded from the pointbar along curvilinear path and the sediment-rich flow after cutoff (M6 in Figure 9). The oxbow lake stage was lower than berm height at evolution stage M7. There was negligible flow observed crossing these sediment berms, yet the downstream berm had not fully closed.
4.1. Meander Evolution and Hyporheic Exchange
Head gradients across the intrameander zone drive lateral hyporheic exchange, and these gradients change with meander evolution. Hyporheic exchange in turn regulates the concentration and residence time of subsurface biogeochemical materials important to microbial and invertebrate communities [Boulton et al., 2010; Stanford and Ward, 1988]. When and where meander evolution creates locally steep head gradients, increased hyporheic exchange may lead to the development of hot spots and hot moments in surface water-groundwater exchange in natural river systems [Vidon et al., 2010; Revelli et al., 2008]. By contrast, gentle head gradients create lower rates of hyporheic exchange across a wider intrameander region with longer residence times that enable different nutrient transformation mechanisms [Boano et al., 2010; Lautz and Fanelli, 2008; Cardenas, 2008]. Our meander evolution experiment captured seepage erosion of the river bank at the meander neck, which was evidence of high rates of groundwater flux into the river water, a temporal intensification of hyporheic exchange across the meander neck. The precutoff upstream bed aggradation and downstream bed degradation were evidence of turnover hyporheic exchange, a mixing of river water and groundwater due to the moving bed material trapping and releasing interstitial fluid [Elliot and Brooks, 1997]. Elliot and Brooks  observed turnover exchange in a flume where dunes on the channel bed migrated downstream, a channel evolution phenomenon different than meander cutoff but also comprised bed aggradation and degradation processes.
Our river table experiment confirmed analytical model predictions [Boano et al., 2006; Revelli et al., 2008] that the cyclical nature of meander evolution increased head gradients and intensified hyporheic exchange, with head gradients and hyporheic exchange diminishing after cutoff. The analytical models attributed increasing head gradients to sinuosity-driven processes of meander elongation, which increases head loss across the intrameander zone, and narrowing of the neck, which decreases distance across the intrameander zone. In three river evolution sequences simulated by Revelli et al. , evolution from stage M3 to M4 led to meander elongation of 22–27%, and this increase in curvilinear distance led to the model predicted increase in head loss of 5–30%. Our experiments found head loss across the meander neck was driven by local bed aggradation and degradation near the neck, with less than 5% of head loss generated by the meander elongation simulated by the analytical models. Upstream aggradation and downstream degradation in river beds are known to increase river slope [Schumm, 2005], and these erosion/deposition processes contributed to a 7 mm change in bed elevation across the neck transect. This bed elevation change is larger than the head loss change across the neck, indicating that the river depth is regulated during the evolution. The smaller meander elongation ratio and larger bed elevation change indicated that the upstream aggradation and downstream degradation were the primary drivers of increasing head loss in our evolution sequence. The aggradation/degradation-based head loss led to head gradients at the neck increasing at a greater rate and to a larger magnitude than in the nearby intrameander core and river valley. Our laboratory results provide an alternative explanation to meander elongation as the cause for changes in head gradients during meander evolution. The localized aggradation/degradation, as compared with evenly distributed meander elongation, may create important differences in the spatial and temporal pattern of hyporheic exchange across the intrameander zone.
4.2. Groundwater Processes in River Evolution
Groundwater is not explicitly modeled in analytical river evolution models [Camporeale, 2005; Zolezzi and Seminara, 2001], yet these models perform remarkably well in replicating statistical patterns of river planimetry by simulating erosion, lateral and longitudinal adjustment, and avulsion [Camporeale, 2005]. Model performance suggests groundwater simulation is not needed to represent meander planimetry. From a process-based perspective, surface water, not groundwater, has greater hydraulic power and shear force to perform lateral and longitudinal adjustment, avulsion, and sediment transport [Schumm, 2005]. However, groundwater is known to do work on soil and directly contribute to bank erosion through seepage; Midgley et al.  report seepage-induced erosion as a significant component of bank erosion, with seepage and erosion rates proportional to head gradients in the near-streambank region. Seepage imposes a viscous force on the soil, leading to subsurface sediment entrainment, erosion, and increased permeability [Baldock and Holmes, 1998; Fox et al., 2007; Rorabaugh, 1964]. Seepage piping has been studied in earthen dams where head gradients across the dam can compromise dam stability and lead to catastrophic flows of soil and water; one illustration of this was the 1994 dam failure on the Chagrin River, in Ohio, US that released 38,000 m3 of impounded water and sediment [Evans et al., 2000]. Seepage may have been active in the St. Anthony Falls Laboratory Outdoor Streamlab, where intrameander hyporheic exchange was associated with subsurface sediment transport and increasing permeability across the intrameander zone [Nowinski et al., 2011]. Over time, such a positive-feedback process contributes to increasing hydraulic conductivity and rates of seepage, and perhaps erosion.
The steep valley gradients and moderately cohesive sediment in our experiment allowed groundwater seepage to directly contribute to breaching the meander neck during cutoff. This occurred when transect head gradients were near or exceeding 50%, reaching the threshold for the rapid seepage erosion considered quicksand flux [Baldock and Holmes, 1998]. While seepage erosion of river banks has been observed in the field [Midgley et al., 2013], our moderate cohesive sand-talc substrate is more vulnerable to seepage erosion than higher cohesion natural river banks. Even with more cohesive soils, however, erodibility increases exponentially with increasing seepage forces [Al-Madhhachi et al., 2011], and the seepage forces are proportional to head gradient [Lobkovsky et al., 2004; Midgley et al., 2013], which increases during cutoff. In meander cutoff conditions with large head gradients and moderately cohesive soils, groundwater seepage erosion may be an important simulation process for analytical river evolution models. Seepage-induced piping may indirectly affect evolution by modifying river bank pore-water pressure and reducing the head loss across the neck via macropore conveyance of water [Rinaldi et al., 2004; Sophocleous, 2002] through the intrameander aquifer and into the downstream river channel. Our river table instrumentation was not designed to track groundwater flow or pore-water pressures, and monitoring in follow-on experiments should utilize more precise groundwater instrumentation to examine these indirect controls on meander evolution. These observations of the interplay of groundwater seepage, erosion, and meander evolution, while enhanced by the short comings of the river table (e.g., distorted scale and similitude), should nonetheless motivate subsequent experiments on the role of seepage erosion in river evolution.
4.3. Limitations of the Experiment
We achieved a meander neck cutoff evolution sequence for a single thread channel, but our experiment had limitations that constrain inferences about natural rivers. The experiment used a talc-enriched river sand substrate, and this still allowed areas of excess bank erosion and widening of the river channel, and channel widening departs from analytical model assumptions of a uniform and fixed channel width. Within our widened channel, the thalweg of the cutoff channel was well established (Figure 9), and maps of water depth identified a dominant flow path and not a braided river reach. Natural single thread meandering rivers can avoid widening with vegetation enhanced banks. Given seepage-induced and shear-induced bank erosion is known to occur in the field, some widening should be accepted in laboratory experiments representing the banks without vegetation. However, future experiments might manage overwidening by addressing upstream sediment supply, local hydraulics causing scour along the banks, lack of bank vegetation and tensile resistance, and use of a uniform flow rate and a single flow inlet location [Tal and Paola, 2009; van Dijk et al., 2012]. In an experiment achieving a chute cutoff in a single thread channel, the prior study alternated the flow inlet location and modified the substrate size distribution to reduce overwidening [van Dijk et al., 2012]. These and other innovations are encouraged for experiments controlling against river widening.
There are no uniform or perfect transformations to convert the time scale of laboratory river evolution to natural river evolution time scales. Laboratory rivers may transition between meander evolution stages in hours to days, while natural rivers have geological time scales of years to millennia, and experience a different set of upstream and local forcing agents such as climate and human modification [Schumm, 2005]. The temporal scaling between the laboratory experiment and natural system is inexact due to the simplicity of the experiment compared with the complex deterministic and stochastic processes involved in natural river evolution. Methods have been developed, however, to scale time between the laboratory and natural rivers. Yalin  analytically investigated time scaling in physical models, and these were summarized by Peakall et al. . Using dimensional analysis, where λT is the time scale of sediment transport and λL is the scaling of length, research has shown a power relationship of 1.5 between spatial and temporal scales, [Yalin, 1963; Yalin, 1971]. Applying this power relationship to our laboratory evolution at its 1:280 planimetric length scale for a 50 h simulation time, the M3–M7 cutoff experiment represented a natural prototype time of 22 years. This laboratory time scale is ∼5 times faster than the time scale in the analytical model simulation, as meander stage M3 took 100 yr to evolve to stage M4 [Boano et al., 2006]. The short time scale achieved in our experiment may represent the accelerated erosion for a river with homogeneous and moderately cohesive sediment and no plants. Rather than use our experiment to predict the time for cutoff in natural rivers, we suggest time scaling remain limited to first-order approximation, as the scaling laws do not consider important factors such as stream power and river substrate material strength.
Dynamic similitude between our laboratory experiment and natural rivers can also lead to differences in rates and processes of meander evolution. Scaling of laboratory experiments to natural rivers can use the Weber number (We, Table 1) to consider the ratio of fluid inertia to surface tension, and how tension influences erosion. The river table had a We value of 0.7, which suggests surface tension may be significant relative to fluid motion. While this We value is smaller than most natural river conditions, similarly small We values have been reported in other small river experiments; there is no consensus on when the We value indicates sediment transport is nonrepresentative to field conditions [Peakall and Warburton, 1996]. Theoretically, it is unavoidable to use smaller We in small river experiments since sediment size and water dynamic viscosity cannot be completely scaled. Analytical river evolution models control for the We number, but the effects of We number on seepage erosion cannot be tested in these models as they do not simulate groundwater. Such simplifications and distorted scaling may help explain why analytical and physical models have different time scales for cutoff, and different processes driving head gradient evolution. Even with differences between analytical and physical model simulation of head gradient evolution, both model types show meander evolution causes temporal and spatial intensification of intrameander hyporheic exchange prior to cutoff.
This laboratory research experiment successfully achieved a meander neck cutoff and captured detailed maps of channel morphology, water stage, and head gradients, with submillimeter vertical accuracy and horizontal resolution. The head gradients increased across the intrameander zone due to erosion decreasing the planimetric distance across the neck and aggradation and degradation increasing the head loss across the neck; meander elongation accounted for less than 5% of increase in head loss. Incipient to cutoff, when the head gradient was steepest, rapid hyporheic exchange and seepage erosion were triggered, which accelerated the breaching of the meander neck. Upstream bed aggradation and downstream degradation both increased intrameander hyporheic exchange rates and generated turnover hyporheic exchange. The large head gradients observed near the meander neck incipient to cutoff were much smaller in nearby valley groundwater transects, illustrating the local natural head loss was induced by bed aggradation and degradation rather than meander elongation. The laboratory experiment's moderately cohesive soil and distorted scaling limits our ability to directly translate results to natural rivers, yet the experiment provides new insights on how the intrameander structure and processes evolve during cutoff, underlining the importance of understanding groundwater when considering meander evolution. Further, the river stage and bed elevation maps of the cutoff sequence provide important temporal and spatial details of meander evolution that should inform river science and engineering. This paper also demonstrated how the novel close range photogrammetry method can generate river surface water topography and bathymetry in a laboratory meandering river, revealing the utility of close range photogrammetry in river surface water topography mapping.
This research was supported in part by NSF EAR 09–11547 and a SUNY ESF Faculty Seed grant. The ideas for this work were advanced via conversation with S. Schumm, N. Matthews, G. Zolezzi, F. Boano, P. Ashmore, and P. Gao. Technical assistance was provided by EmRiver staff, H. Chu, A. Greenhill, H. Tieu, P. Szemkow, A. Li, and T. Zhou. We acknowledge B. Reschke for his thoughtful construction of the river table. We are grateful for the editorial guidance and scientific insights provided by the Associate Editor, A. Sawyer, E. Hester, and an anonymous reviewer.