Coarse sediment transport in a bedrock channel with complex bed topography



[1] Independent lithologic and structural controls in fluvial bedrock systems interact with coarse sediment transport processes to play a key role in bedrock incision processes such as abrasion. During a 3 year study on the Ocoee River in the Blue Ridge Province of the southern Appalachians, USA, we used painted tracer clasts to measure coarse sediment transport dynamics and address whether different scales of morphologic variability in bedrock channels lead to coarse sediment transport processes that differ from alluvial channels. At the reach scale, folded metasedimentary units are exposed in the channel bed and appear as linear bedrock ribs that vary in amplitude and orientation to flow. Under similar flow conditions for which size-dependent transport has been observed in alluvial channels (dimensionless Shields stress within 1.5–2.0 times the dimensionless critical shear stress), transport distance was a significant function of grain size where bedrock ribs were longitudinal to flow (Reach 1 and Reach 2). However, transport distance was not size dependent where bedrock ribs were oblique to flow (Reach 3). At different intrareach scales, the variables characterizing the local bedrock topography and sediment architecture were the best predictors of transport distance in all three reaches. Therefore, bed load transport processes may be influenced by bed forms in bedrock streams (bedrock ribs) but at potentially smaller scales than bed forms in alluvial channels. The strong influence of bedrock ribs on coarse sediment transport suggested that coarse sediment transport processes are controlled by different factors in bedrock channels only when bedrock ribs cross the channel at a high angle to the flow.

1. Introduction

[2] In order to constrain parameters required for landscape evolution models [e.g., Howard, 1994; Whipple and Tucker, 2002; Sklar and Dietrich, 2004; Turowski et al., 2007], recent studies have focused on reach scale processes involved in bedrock channel incision [Johnson and Whipple, 2007; Finnegan et al., 2007; Chatanantavet and Parker, 2008; Johnson et al., 2009]. Particular attention has focused on the important role of bed load material in controlling incision via abrasion in bedrock systems [Sklar and Dietrich, 1998, 2001; Hartshorn et al., 2002], illustrating the need to understand the dynamics of coarse sediment transport in bedrock channels. While an extensive literature describes coarse sediment transport processes at the reach scale in alluvial channels [e.g., Einstein, 1950; Beschta, 1987; Ferguson et al., 2002; Wilcock and Crowe, 2003; Parker, 2008], the reach scale controls on coarse sediment transport in bedrock channels are poorly understood, except from a few experimental studies [Johnson and Whipple, 2007, 2010; Finnegan et al., 2007; Chatanantavet and Parker, 2008], and even more limited field examinations [Johnson et al., 2009]. These studies document interactions and feedbacks among sediment supply, localized transport, bedrock erosion rate, and channel morphology. Although it is reasonable to assume that alluvial and bedrock channels have fundamentally different controls on coarse sediment transport, the details of these different controls have not yet been explored and adequately quantified in the field. By examining the controls on coarse sediment transport in natural bedrock channels, we can improve our understanding of the processes that control bedrock incision by abrasion.

[3] In the process of abrasion, saltating bed load acts to scour bedrock surfaces by fine-scale removal of bedrock material. Abrasion is distinguished from other physical erosion processes such as plucking and macroabrasion, which remove cobble- and boulder-sized blocks from the bedrock channel substrate [Hancock et al., 1998; Whipple et al., 2000; Chatanantavet and Parker, 2009]. Bedrock incision by abrasion is conditioned on the interplay between bed load material either providing the tools that promote abrasion and incision or being the cover that inhibits incision, depending on the ratio of sediment supply to transport capacity [Gilbert, 1877; Sklar and Dietrich, 1998, 2001; Turowski et al., 2007]. The incision rate is maximized for intermediate sediment supply relative to the transport capacity, such that a sufficient amount of bed load tools are available for abrasion, and the bed is not fully covered by alluvium. Whereas this threshold concept is simple, there are unconstrained details (e.g., interactions between sediment transport and bedrock bed forms) that arise upon experimental examination [Johnson and Whipple, 2007; Finnegan et al., 2007; Chatanantavet and Parker, 2008].

[4] The saltation-abrasion model [Sklar and Dietrich, 1998, 2004] incorporates the concepts of tools versus cover effects quantitatively by considering bed load supply and transport. This model assumes a planar bed surface, and it does not incorporate channel morphology as a degree of freedom. Natural bedrock channels, however, commonly display high spatial variability in bed topography as a result of sculpting and bedrock characteristics [Whipple, 2004; Richardson and Carling, 2005]. This variability has also been shown to control local erosion rates [Hancock et al., 1998; Johnson and Whipple, 2007] and promote feedbacks between incision and bed morphology [Finnegan et al., 2007; Johnson and Whipple, 2007, 2010]. It is well documented from alluvial systems that variation in bed topography largely influences the local flow field [Furbish, 1993; Nelson et al., 1995; Papanicolaou et al., 2001], the dynamics of sediment transport [Pyrce and Ashmore, 2003; Yager et al., 2007; Thompson, 2007], and spatial variation in the surface texture of bed sediments [Dietrich et al., 1989].

[5] Sediment transport and deposition in bedrock channels are generally different from alluvial systems, in that (1) sediment loads are relatively low compared to transport capacity, (2) highly turbulent flows are capable of transporting up to boulder-sized particles for large distances, and (3) lateral hillslope connection directly supplies coarse sediment through diffusive hillslope processes, landslides, rockfalls, or debris flows [Wohl, 1999; Whipple, 2004]. Bedrock characteristics (i.e., lithology and structure) also have a strong influence on channel geometry [Montgomery and Gran, 2001; Whipple, 2004; Goode and Wohl, 2010], potentially regulating bed load transport.

[6] In bedrock channels, sediment supply, grain size, and bedrock detachment are all interrelated variables that tend to offset variation in one another [Sklar and Dietrich, 2008]. Roughness supplied from sculpted forms in bedrock systems increases form drag, which likely reduces the local shear stress [Shepherd and Schumm, 1974; Wohl and Ikeda, 1997; Wohl, 1998; Wohl et al., 1999; Johnson and Whipple, 2007; Finnegan et al., 2007; Turowski et al., 2007; Chatanantavet and Parker, 2008]. As a result, roughness may locally reduce sediment transport capacity, such that the local bedrock topography controls the spatial distribution of sediment transport. In this study, we examine how spatial heterogeneity at different scales controls the dynamics of coarse sediment transport in a bedrock channel with bed morphology that is influenced by lithologic and structural variation.

[7] In alluvial channels, strong spatial and temporal variability in bed load transport rates are explained by variability in transport over bed forms [Schmidt and Gintz, 1995; Thompson et al., 1996; Cudden and Hoey, 2003], temporal variability in sediment supply [Beschta, 1987], variation in armoring and sorting [Whiting et al., 1988], and burial and vertical mixing [Ferguson and Hoey, 2002; Ferguson et al., 2002]. The spatial and temporal controls on bed load transport in gravel bed streams have been characterized by tracking the movement of individually marked gravels [e.g., Church and Hassan, 1992; Wilcock, 1997; Ferguson et al., 2002; Lenzi, 2004]. Results from these tracer studies indicate that under moderate flow conditions, the structure of the bed sediment (i.e., variability in sediment size, sorting, and packing) is an important control on bed load transport, whereas the flow characteristics are the dominant control on bed load transport at higher flows [Ferguson and Wathen, 1998; Lenzi, 2004]. Size selectivity in bed load transport tends to occur in either of two scenarios: (1) over a range of flows for particles that are unconstrained by interparticle packing and fully exposed to the fluid forces or (2) at moderate flows, when the shear stress is slightly above the threshold of motion for the median grain size, for particles that are constrained by interparticle contacts (i.e., imbricated and incorporated into the armor layer) [Church and Hassan, 1992; Ferguson et al., 2002; Lenzi, 2004].

[8] Here we use painted tracer clasts in order to examine the controls on coarse sediment transport in a bedrock channel with a high degree of spatial variability in bed topography. We frame this study around a fundamental question: Is coarse sediment transport in bedrock channels controlled by the same hydraulic conditions and sediment properties that control coarse sediment transport in gravel bed streams? Specifically, we seek to understand what conditions (flow or sediment architecture), if any, lead to size-dependent transport. We address this question by testing three specific alternative hypotheses. (H1) In all study reaches, transport distance is a significant function of grain size, D. (H2) Differences in coarse sediment transport distance exist between study reaches can be explained by differences in reach scale hydraulics and bedrock topography. This second hypothesis is designed to test the idea that, if transport distance is not size dependent, then observed differences in transport distance can be explained by reach scale variations in bedrock channel bed morphology. (H3) The local scale variability in bedrock channel topography and sediment architecture is the dominant control on transport distance. This hypothesis is based on the assumption that bedrock ribs in the channel examined here exert a strong control on coarse sediment transport. We use the term coarse sediment transport rather than bed load transport because the nature of this study precludes determination of the mode of transport. The focus is on coarse gravel, cobbles, and boulders (32 mm < D < 362 mm).

2. Study Area

[9] The study area is located in the Ocoee River gorge, Tennessee between the Tennessee Valley Authority (TVA) Ocoee No. 3 dam and the 1996 Olympic whitewater course (Figure 1). Here the Ocoee River flows through the Blue Ridge Province of the Southern Appalachians, with a drainage area of approximately 1300 km2. Although the gorge is deeply incised, hillslopes exhibit limited bedrock exposure and are densely vegetated and mantled with thick soils typical of humid temperate landscapes. Homogenous denudation rates (25 ± 5 m/Myr) over 104–105 year time scales [Matmon et al., 2005] characterize this tectonically quiescent region. Bedrock in this region consists of slates and metasandstones contained within the Precambrian Ocoee Supergroup. Specifically, bedrock exposures through the Ocoee River gorge are the Precambrian-age Sandsuck Formation in the western gorge, which is composed of phyllites thinly interbedded with arkosic and calcareous quartzites; the Dean Formation, composed of thinly bedded quartzites and phyllites; and the Hothouse Formation, composed of metagreywacke and mica schist in the eastern gorge [Sutton, 1991]. Through the gorge, the rock units alternate between resistant ledges of metagreywacke and quartzite and softer phyllite sequences. Because the channel meanders across the folded metamorphic units at scales larger than the reach scale (Figure 1), the orientation of flow relative to the dip and lithology of the bedrock varies.

Figure 1.

Digital Orthophoto of the Ocoee River study area [U.S. Geological Survey, 1997]. Reach boundaries are indicated by the solid white lines. The downstream variation of rib orientation can be seen in this areal photo of the dry channel bed, with black dashed lines indicating the rib orientation. Reach boundaries are indicated with white lines.

[10] This lithologic and structural heterogeneity appears in the channel bed as undulating rib-like bedrock forms that we refer to as bedrock ribs. These bed forms are not unique to the Ocoee River [Goode and Wohl, 2010] but have not received much attention in the literature. In their report on sculpted forms in bedrock channels, Richardson and Carling [2005] describe similar structurally influenced features such as concave sculpted joint furrows and bedding plane furrows. Bedrock ribs differ from these features as opposing topographic features. Concave sculpted joint furrows and bedding plane furrows are long, narrow portions of bedrock that protrude above the surrounding bed, whereas bedrock ribs are asymmetrical in cross section, regardless of planform orientation, consistently oriented parallel to the metamorphic foliation in the rock, and occasionally follow dominant joints. The strike of the bedrock ribs varies as the trend of the sinuous channel changes downstream, which suggests that rib orientation is controlled by structural features in the underlying folded metasedimentary units. The occurrence of sculpted forms such as potholes (along the upstream and downstream faces of transverse bedrock ribs and in the troughs between longitudinal bedrock ribs) suggests that abrasion is the dominant mechanism of fluvial incision in this system.

[11] Alluvial material in this system consists of sand- to boulder-sized sediment, which covers roughly half of the channel bed area in three distinct areas: discontinuous patches, the intervening lows between bedrock ribs, and within potholes. In isolated locations of the channel bed, this alluvial material is armored by large cobbles and boulders. The wake zones of bedrock ribs tend to be associated with high concentrations of well sorted, gravel-sized material (Figure 2a). Sediment angularity ranges from very well-rounded particles within potholes to slightly more angular particles across the channel bed. This difference in angularity appears to reflect not only the local hydraulic environment but also the lithologic origin (phyllite produces more angular particles, whereas the metagreywacke corresponds to well-rounded particles) and fluvial transport distance.

Figure 2.

(a) Photograph illustrating the coarse sediment deposition in between bedrock ribs in Reach 3. (b) Tracer clast (circled) downstream of a bedrock rib in a local hydraulic environment classified as shielded. Tape measure for scale. Arrows indicate flow direction. Photos were taken between summer release flows when the channel was dry.

[12] After the closure of the Ocoee No. 3 dam in 1942, the natural-channel flow was diverted through a bypass to the hydropower station downstream, beyond the study area. Since the 1996 Olympics, however, the TVA has guaranteed recreational flows in this section of roughly 45 m3/s for 6 h on both Saturday and Sunday, from the last weekend in May through the last weekend in August. Only exceptional winter stormflows that exceed the hydropower capacity are routed through the natural channel. Winter and spring storms produce flows for which the daily average releases from the dam range from 40 to 60 m3/s, and peak flows can reach 800 m3/s. Infrequent flow releases and flow diversion has led to woody vegetation establishment in many locations of the channel bed. We intentionally avoided these sections in this study.

3. Methods

[13] Three study reaches within the study area were selected according to the dominant lithology and the orientation of the bedrock ribs to flow (Figure 1 and Table 1): Reach 1 (phyllite and longitudinal ribs), Reach 2 (metagraywacke and longitudinal ribs), and Reach 3 (metagreywacke and oblique ribs). A fourth reach, Reach 4 (metagraywacke and transverse ribs), was initially included in this study, but anthropogenic disruption of the experiments forced us to eliminate it [Goode, 2009]. Longitudinal ribs varied 0° ± 10° with respect to the main flow direction, transverse ribs were oriented 180° ± 10° with respect to flow, and oblique ribs occurred at all other angles to flow. The length of each surveyed reach was determined by the persistence of the rib morphology: reaches continued for as long as the bedrock rib orientation and dominant lithology was maintained within a straight, single flow channel.

Table 1. Summary of Differences in Reach Morphology, Lithology, Hydraulic Roughness, and Grain Size
 Reach 1Reach 2Reach 3
Rib orientation to flowlongitudinallongitudinaloblique
Dominant lithologyphyllitemetagreywackemetagreywacke
Mean rib amplitude (m)0.440.570.79
Reach gradient0.00750.00820.0106
Manning's n0.0590.0750.094
D16 (mm)504648
D50 (mm)115100150
D84 (mm)240230252

[14] In each of the three study reaches, we used painted tracer clasts, which we selected from the existing alluvial patches. A total of 300 tracers were randomly selected using a Wolman point count, where we paced the channel in a conceptual grid. While it is argued that this method tends to bias the sample toward the coarse grains [Kellerhals and Bray, 1971], our interest was focused on sediment that was larger than coarse gravel (>32mm), which supported the validity and practicality of this selection approach in this field study. The pattern that we followed for this selection procedure allowed us to span the channel width several times and capture both a representative distribution of particle sizes and a wide range of depositional zones associated with the highly variable bedrock channel bed topography (i.e., grains that were located upstream and downstream of bedrock ribs, in the thalweg, or in pools). If a selected particle was fully incorporated into the armor layer, such that the particle boundaries could not be seen, it was not included in the sample. This occurred for less than five cobble- and boulder-sized grains in each reach.

[15] For each tracer, we recorded the size (measured along the intermediate axis), applied a roughly 5 cm square patch of yellow concrete paint, and wrote the tracer number on the paint patch with a black permanent marker. Care was taken not to disrupt the bed material when clasts were tagged with paint. If the tracer was not incorporated in the armor layer, a paint patch and number were applied to a second side to increase the likelihood of finding the tracer if it flipped over after transport. A laser total station positioning system was used to map the initial tracer locations. During this survey, the area within one grain diameter was visually assessed for its potential influence on the transport of a given tracer (Figure 2b). Five categories defined this local scale: unconstrained (no surrounding particles or bedrock ribs protruded above the tracer), shielded (bedrock ribs protruded above the tracer, and likely interfered with transport), imbricated (surrounding particles constrained the movement), buried (required removing sediment in order to recover the tracer), or embedded (packed into the armor layer). Tracers that were both constrained by surrounding grains and bedrock ribs were assigned imbricated and shielded.

[16] In addition to the local scale, we defined two other spatial scales within each reach: (1) cross-section scale, with tracers linked to the nearest cross section, in the previous survey and (2) zone scale, with tracers linked to geomorphically similar bed regions. The cross-section scale was quantified by the cross-sectional averaged hydraulic metrics, obtained from the modeling results of Goode and Wohl [2010], wherein we applied the U.S. Army Corps of Engineers (USACE) one-dimensional flow model HEC-RAS [U.S. Army Corps of Engineers (USACE), 2002] to iteratively determine the Manning's roughness coefficients by matching the known discharges from the 2006 summer release flows (Q = 45 m3/s) and the surveyed water surface elevations at each cross section. Using these Manning's roughness coefficients, we used HEC-RAS to model each peak discharge between tracer resurveys. Without known water surface elevations at the upstream and downstream reach boundaries for these peak flows, because we were not present in the field area during these flows, we set the downstream boundary condition to the normal flow depth. A sensitivity analysis on the choice of the downstream boundary and response of the calculated hydraulic variables showed no significant difference (P > 0.05) in the hydraulic variables, especially in the upstream cross sections where the tracers existed. This enhanced our confidence in the hydraulic modeling results despite the uncertainty in the downstream boundary conditions. Although supercritical flow likely occurs locally within each reach, model runs were performed assuming subcritical flow and typical step-backwater calculations were performed. The substantial protrusion of the bedrock ribs from the channel bed, which appear to exert a downstream hydraulic control, and the relatively shallow bed slopes, support this assumption of subcritical flow. The following hydraulic variables were obtained from this modeling and used in both cross section and reach-averaged scale analysis (Table 2): streamwise velocity u, total boundary shear stress τ, and unit stream power ω.

Table 2. Hydraulic Variables From HEC-RAS Modeling of Peak Discharges in Each Reach Between Each Recovery
VariableSummer 2006: Mean of Four Resurveys2007 Resurvey2008 Resurvey
Reach 1Reach 2Reach 3Reach 1Reach 2Reach 3Reach 1Reach 2Reach 3
Qpeak (m3/s)474747818181888888
Manning's n0.
Velocity (m/s)
Unit stream power (W/m2)1007288168145146184157160
Total boundary shear stress (N/m2)8775108115123140123129147
Shields parameter τ*500.0470.0460.0440.0620.0760.0580.0660.0800.061
D50 (mm)115100150115100150115100150
D50mobile (mm)606060807080757080

[17] To examine the effect of variable bed morphology on coarse sediment transport within each reach, we divided each reach into four to six zones (∼500 m2) of contiguous and internally consistent bed morphology (e.g., an alluvial patch with limited bedrock rib exposure versus a zone with substantial bedrock exposure and high amplitude bedrock ribs). We visually defined these zones in the field and used the surveyed boundaries to assign each tracer to a given zone after plotting the tracer locations and morphological zones. These zones typically did not span the channel width. Within each morphologic zone, we calculated the standard deviation of the bed elevation Zstdev, from bed topography points that were sampled at a 1 m resolution. We also surveyed transects, orthogonal to the bedrock ribs, to calculate the mean amplitude of the bedrock ribs A within each zone and reach. These values provided a quantitative metric for comparing the mesoscale variability in morphology.

[18] The locations of the recovered tracers were resurveyed after each of four summer recreational flow releases in 2006 (Qpeak = 47 m3/s) and two annual hydrographs with peak flows Qpeak in 2007 and 2008 of 81 and 88 m3/s, respectively (Figure 3). Tracers were recovered by walking the channel in progression of the numbering sequence of the initial installment. If a tracer was “missing” from its initial location, the surrounding channel was visually searched without disrupting the bed. We only searched for buried clasts in the well-sorted patches associated with bedrock ribs, where we suspected vertical mixing. Transport distance was calculated as the linear displacement of a tracer between each resurvey. During each resurvey, we visually categorized the local grain scale for each tracer. The corresponding morphologic zone and nearest cross section were determined from the surveyed coordinates of each tracer after each resurvey.

Figure 3.

Annual hydrograph for Ocoee No. 3 for 3 years of study. Note the systematic fluctuations in summer recreational release flows. Hourly discharge data were obtained from the TVA.

[19] Tracers were considered mobile if the calculated transport distance between surveys was greater than the particle diameter. This accommodated for any error inherent to repeat surveys. We minimized this type of error by consistently reoccupying the setup location of the laser total station, so that the backsight did not exceed 5 mm in distance or elevation. Also, we consistently surveyed the patch of paint on the boulder-sized tracers, which enhanced our confidence that repeat surveys reliably documented the same location on each tracer, and transport distances were not misrepresented by inconsistent survey points.

4. Results

[20] The size distributions of tracers in the three reaches were relatively similar (Figure 4), with median sizes D50 of 115, 100, and 110 mm, respectively. This consistency allowed interreach comparisons of transport dynamics without adjusting for differences in the bed material size distribution. In all three reaches during the recoveries between summer flows in the first year (2006), the recovery rate was nearly 100% (Table 3). There were several cases in the 2007 and 2008 resurveys where the tracer number was indistinguishable. Renumbering and measurement, along with spatial comparison to the previous year's location, allowed for these tracers to be accurately matched to the original number and included in the recovery. Despite these difficulties, recovery rates remained above 60% in all three reaches (61%, 63%, and 70%, respectively).

Figure 4.

Size distribution of painted tracers, which were sampled from the existing bed material in each reach.

Table 3. Summary of Transport Distances and Tracer Sizes Transporteda
 YearQpeak (m3/s)nmobile% RecoveredLmean (m)Lmax (m)LD50 (m)D50mobile (mm)DLmax (mm)
  • a

    Lmean, mean transport distance for all tracers in the reach; Lmax, maximum transport distance of an individual clast; LD50, transport distance of the median grain size; D50mobile, median grain size of mobile tracers only; DLmax, grain size of the farthest transported tracer.

Reach 12006 (1)4881990.38 ±
 2006 (2)47166990.27 ±
 2006 (3)48129990.16 ±
 2006 (4)46110990.22 ±
 2006 (mean)47122990.3 ±
 200781142773.5 ± 0.39272.78050
 20088968614.7 ± 0.71294.27545
 Cumulative 266 4.2 ± 0.40404.77535
Reach 22006 (1)48117990.2 ±
 2006 (2)4786990.27 ±
 2006 (3)4873990.2 ±
 2006 (4)4694990.16 ±
 2006 (mean)4793990.2 ±
 200781112820.87 ±
 20088955612.8 ± 0.81262.570105
 Cumulative 232 2.5 ± 0.26292.58020
Reach 32006 (1)48127990.45 ±
 2006 (2)47105990.35 ±
 2006 (3)48179990.34 ±
 2006 (4)46168990.36 ±
 2006 (mean)47145990.4 ±
 200781130731.2 ± 0.16120.368030
 20088969701.9 ± 0.32142.68030
 Cumulative 227 2.3 ± 0.20271.89030

4.1. Grain Size-Dependent Transport Distance

[21] An examination of the transport distance of mobile tracers after each recovery indicated trends in transport distance and grain size were not consistent for all reaches, all recovery periods, or for all grain size classes (Figure 5). In 2006 and 2007, Reach 1 showed a decrease in transport distance with increasing grain size, only for the finer tail of the size distribution (D < 54 mm, in D < 108 mm, respectively). Reach 2 also showed a similar trend in 2007, only for D < 108 mm. In 2008, this trend occurred in all three reaches for D < 108 mm. Considering the cumulative transport distance over the study period, Reach 2 and Reach 3 showed a decrease in transport distance for D < 152 mm. These grain size threshold values, for which transport distance depends on grain size, compare well to the size of the D50 within each reach (Table 1 and Figure 4).

Figure 5.

Transport distance versus the tracer size. Tracers are binned into 0.5ϕ size diametric classes (ϕ = −log2D (mm)), and transport distances represent the mean for all tracers in that size class, with corresponding standard errors. Each plot shows results from that recovery period. The 2006 plot shows averages for all four resurveys during summer release flows. The cumulative plot represents the total transport over entire study period.

[22] Transport distance was a significant function of grain size only in Reach 1 and Reach 2 in 2007 (P < 0.01) and Reach 1 in 2008 (P = 0.01), based on regression analyses of nonbinned, log-transformed data in each reach. Although this result in these two reaches supported the alternative hypothesis that transport distance is a function of grain size, inconsistency in these relationships among all three reaches resulted in a failure to reject the first null hypothesis that transport distance is not a function of grain size in all reaches. Lack of a significant result in all three reaches led to the second hypothesis, which investigated reach scale controls on transport distance.

4.2. Interreach Variability in Transport Distance

[23] The size distributions of the tracers that were transported during the entire study period were similar for all three reaches (Figure 6). Whereas the median grain size of all the transported tracers, D50mobile, did not vary greatly among the three reaches, the corresponding transport distance for tracers of that size, LD50, varied substantially, with the largest LD50 in Reach 1 (Table 2). In the 2008 resurvey after the largest peak flow (89 m3/s) and the longest duration of flows (Figure 3), LD50 in Reach 1 was roughly 2 times the distance in Reach 2 and Reach 3 (LD50 = 4.21, 2.45, and 2.55 m, for D50mobile = 75, 70, and 80 mm, respectively). Considering the sediment sizes and transport distances integrated over the entire 3 year study period the median grain size mobilized varied slightly between the three reaches (75, 80, and 90 mm, respectively).

Figure 6.

Size distribution of mobile tracers from each reach over the entire study period. Tracers were considered mobile if transport distances were greater than one particle diameter in at least one resurvey period.

[24] Reach-averaged hydraulic data from HEC-RAS modeling of all peak flows are presented in Table 2. The trends in total boundary shear stress in each reach were consistent with the D50mobile in each reach. For example, Reach 3 was associated with the largest D50mobile, as well as the greatest total boundary shear stress. The Shields parameters for the D50 in each reach ranged from 0.04 to 0.05 in the 2006 flows and from 0.06 to 0.08 for the 2007 and 2008 peak flows.

[25] Comparison of the distribution of transport distance after the 2007 and 2008 recoveries indicated that more tracers were transported longer distances in Reach 1 (Figure 7). More tracers were transported shorter distances in Reach 3, whereas Reach 2 yielded an intermediate distribution of transport distances. In all reaches, more tracers were transported longer distances during the 2008 flows, which were longest in duration and had the highest peak discharge in the study period. According to the Tukey HSD multiple comparison tests on the log-transformed mean transport distance between reaches, Reach 1 was consistently associated with significantly greater transport distances than the other two reaches (Figure 8). Transport distances in Reach 2 and Reach 3 were statistically similar after all resurveys. Hypothesis 2 was partly rejected because the mean transport distances were not significantly different in all three reaches: transport distances in Reach 1 were significantly different from the other two reaches, but transport distances were not significantly different in Reach 2 and Reach 3.

Figure 7.

Cumulative distributions of transport distances after the 2007 and 2008 resurveys for all three reaches. Corresponding flows are reported in Table 2.

Figure 8.

Reach scale variation in transport distances. Years indicate the resurvey period. The box plots indicate upper and lower quartiles as box ends, 10th and 90th percentiles as whiskers, and median values as the line within each box. Dots indicate values outside the 10th and 90th percentiles. Transport distances were log-transformed for normality. Different mean transport distances are indicated by contrasting letters above each box (Tukey's HSD following ANOVA on log-transformed data, P < 0.05).

4.3. Intrareach Variability in Transport Distance

[26] Transport distance according to the three spatial scales (cross section, zone, local) was considered for the 2007 and 2008 resurveys only, because these categorical variables from the four 2006 resurveys varied in each of the four resurveys. ANOVA of transport distance by cross-sectional differences was not significant in either the 2007 or 2008 recovery period in any of the reaches (Table 4). Within each reach, transport distances did not differ significantly when the tracers were grouped according to the cross section for which they were nearest to in the proceeding survey. This suggested that differences in cross-section averaged hydraulics cannot account for differences in coarse sediment transport. In Reach 1, despite a nearly twofold difference in unit stream power between two cross sections, mean transport distances of the groups of tracers referenced to those cross sections, were not significantly different between these groups (Figure 9).

Figure 9.

Box plot comparing the transport distance of tracers linked to the nearest cross sections in Reach 1. (a) Transport distance (2007) by cross section (ANOVA, F = 1.77, P = 0.14). (b) Transport distance (2008) by cross section (ANOVA, F = 1.15; P = 0.34), so comparison of means was not appropriate. The box plots indicate upper and lower quartiles as box ends, 10th and 90th percentiles as whiskers, and median values as the line within each box. Dots indicate values outside the 10th and 90th percentiles. Hydraulic variables at each cross section: streamwise velocity u, total boundary shear stress τ, and unit stream power w, were obtained from HEC-RAS modeling of the peak discharge between recoveries [Goode and Wohl, 2010].

Table 4. Summary of ANOVA Analysis of Transport Distance by Intrareach Spatial Scalea
Reach 11.770.143.460.00519.2<0.0001
Reach 20.520.470.760.478.9<0.0001
Reach 31.440.233.790.00636.1<0.0001
  • a

    Bolded values indicate significance (P < 0.05). XS, cross-section scale; ZONE, scale of morphologically similar bed zones; LOCAL, local scale characteristics.

Reach 11.150.341.560.1926.3<0.0001
Reach 20.220.640.930.4019.3<0.0001
Reach 30.480.701.590.216.8<0.0001

[27] Considering the morphologically similar zones, significant differences in mean transport distance were not consistently controlled by zone scale differences in bed topography (i.e., bedrock rib amplitude). In Reach 1, transport distance was significantly different in some of the zones, but not in the zones that we expected to correspond to these differences (Figure 10, 2007). For example, the mean transport distance in the zone where the bedrock rib amplitude was lowest (A = 0.42 m) was not significantly different than the mean transport distance of tracers in the zone where the bedrock rib amplitude was the highest (A = 0.65 m). For the 2007 tracer recovery, the significant ANOVA results indicated zone scale differences in transport distance in Reach 1 and Reach 3 (Table 4). Zone scale differences were not associated with significantly different mean transport distances in Reach 2 for 2007 or any reach for the 2008 recovery. The mean transport distances were not greater in bed regions where the bedrock rib amplitude A and standard deviation of the bed topography Zstdev were relatively low, which indicated that at the zone scale, these metrics were not a significant control on transport distance. Also, in Reach 3, pairwise comparison of all zones, using the Tukey HSD following ANOVA, indicated that the only zones with significantly different mean transport distances had similar mean rib amplitudes (A = 0.65 and 0.69 m, respectively) and standard deviations of the bed elevation (Zstdev = 0.33 and 0.36 m, respectively) within each zone. Within each reach, at the scale of morphologically similar zones, neither bedrock rib amplitude nor topographic variability controlled differences in transport distance.

Figure 10.

Box plot comparing tracer transport distance by zone topographic metrics in Reach 1. (a) Transport distance (2007) by zone (ANOVA, F = 3.46, P < 0.01). Significant pairwise differences in means (P < 0.05) are indicated by contrasting letters above each box (Tukey's HSD following ANOVA on log transformed data). (b) Transport distance (2008) by zone was not significant (ANOVA, F = 1.56; P = 0.19), so comparison of means was not appropriate. The box plots indicate upper and lower quartiles as box ends, 10th and 90th percentiles as whiskers, and median values as the line within each box. Dots indicate values outside the 10th and 90th percentiles. Zstdev is the standard deviation of the bed elevation within the bed zone, and A is the mean amplitude of bedrock ribs within that area. Boxes are in order of increasing A from left to right.

[28] In all three reaches and for both recovery periods, there was a strong relationship between transport distance and the local scale characteristics (unconstrained, shielded, imbricated, buried embedded; Table 4). In Reach 1, particles that were unconstrained by surrounding grains and unobstructed by bedrock ribs were transported the greatest distances (Figure 11). In some cases, particles that were shielded by ribs were transported unexpectedly long distances, which may have been a result of local turbulence fluctuations in the wake zones of ribs. Pairwise comparisons of all local scale categories, using the Tukey HSD following ANOVA, consistently showed that the transport distance of unconstrained tracers was significantly different (P < 0.05) from imbricated, shielded, or buried particles, in all three reaches and for both years. The significant difference in transport distance according to local scale distinctions provided support for hypothesis 3. This hypothesis was more rigorously tested in a multiple regression analysis.

Figure 11.

Box plot comparing tracer transport distance by local hydraulic environment classification in Reach 1. (a) Transport distance (2007) by local classification (ANOVA, F = 19.2; P < 0.01). (b) Transport distance (2008) by local classification (ANOVA, F = 26.3; P < 0.01). Significant pairwise differences in means (P < 0.05) are indicated by contrasting letters above each box (Tukey's HSD following ANOVA on log transformed data). The box plots indicate upper and lower quartiles as box ends, 10th and 90th percentiles as whiskers, and median values as the line within each box. Dots indicate values outside the 10th and 90th percentiles.

[29] Multiple linear regression with model selection, based on the minimum Akaike's Information Criterion (AIC) [Burnham and Anderson, 2002] and adjusted R2, further supported that local scale characteristics explained the largest proportion of the variability in transport distance (supporting hypothesis 3). The parameters tested in these models included: tracer size D (mm), Reach (categorical), total boundary shear stress τ, (at the nearest cross section), mean rib amplitude within the morphologic zone, A, and the local scale (categorical). The best one-parameter model included only local scale classification (P < 0.01 in both years; adjusted R2 = 0.45 and 0.20 for 2007 and 2008, respectively). The best two-parameter model included the reach and local scale (P < 0.01 in both years; adjusted R2 = 0.48 and 0.32 for 2007 and 2008, respectively). Both parameters in these models were also significant at P < 0.01.

[30] Recognizing that the local scale sediment characteristics had the strongest influence on transport distance, we reexamined the relationship between transport distance and grain size by stratifying the data according to local categories (Figure 12). For the unconstrained tracers, transport distance was a significant power function of grain size (L = aDb) in Reach 1 and Reach 2 in 2007 (P < 0.01, R2 = 0.54, and P < 0.01, R2 = 0.53). No other significant relationships were found for nonlinear regression analyses of grain size and transport distance.

Figure 12.

Transport distance versus the tracer size segregated by local scale categories for each reach in 2007 and 2008. Transport distance is significantly (P < 0.05) correlated with grain size only for unconstrained tracers in Reach 1 in both years and Reach 2 in 2007.

5. Discussion

5.1. Controls on Transport Distance

[31] Transport distance was a significant function of grain size in Reach 1 and Reach 2, but not in Reach 3. As one possible explanation for this, bedrock ribs were oriented parallel to flow in both of these reaches, leaving downstreamflow and sediment transport generally unobstructed (Figure 2a). In Reach 3, where bedrock ribs were oblique to flow and of greater amplitude (Table 1), transport distance was not a significant function of grain size after any recovery period (Figures 5 and 12). Because the trend in transport distance and grain size was not similar in Reach 3, where the bedrock rib characteristics were the most different from the other two reaches, it is likely that bedrock ribs influence sediment transport when they are oriented oblique to flow. In Reach 3, the Manning's roughness was substantially larger than in Reach 1 and Reach 2, suggesting that bedrock ribs account for a large proportion of the total roughness. Hence, less of the total boundary shear stress is available for sediment transport in Reach 3. This may also explain why Reach 3 did not show a trend in transport distance with grain size in any of the recovery periods. As a result, where bedrock ribs are larger in amplitude and oriented in a direction that obstructs flow and sediment transport, these bedrock bed forms exert the greatest control on sediment transport.

[32] The central finding from this field study (variable bedrock topography (bedrock ribs) influences sediment transport distance) corresponds well with experimental observations of sediment transport patterns in bedrock channels with spatial variation in bed topography [Johnson and Whipple, 2007, 2010; Finnegan et al., 2007; Chatanantavet and Parker, 2008]. In the field, the additional boundary roughness supplied by bedrock ribs most likely limits the amount of boundary shear stress required to transport coarse sediment. This was apparent from the lower transport distances in Reach 3 (oblique ribs and largest hydraulic roughness, n). Similarly, experimental studies have shown that erosional bed forms act to dissipate energy and reduce the sediment transport capacity [Johnson and Whipple, 2007; Finnegan et al., 2007; Chatanantavet and Parker, 2008].

[33] In a comparison of variables thought to control transport distance at different spatial scales, none of the models selected in the multiple regression analysis explained more than 50% of the variability in transport distance. However, more variability was explained by the local scale bedrock topography and sediment architecture than the relationships of transport distance as a power function of grain size. Also, grain size did not appear as a significant explanatory variable in any of the models, which further indicates that in this bedrock river, bedrock rib characteristics at the reach scale, and variations in morphology and sediment architecture at the local scale, are the most important controls on sediment transport.

5.2. Different Controls on Coarse Sediment Transport in Alluvial and Bedrock Channels

[34] It is well established that bed load transport is an event dependent process and, whether or not transport distance is size dependent, is controlled by the flow conditions and sediment structure [e.g., Parker, 2008]. If size-dependent transport is governed by similar processes in alluvial and bedrock channels, then we would expect transport distance to be size dependent when the Shields stresses τ*50 are 1.5–2.0 times the dimensionless critical shear stress of the D50τ*c50, as demonstrated in gravel bed streams by Ferguson and Wathen [1998], Wilcock and Crowe [2003], and Parker [2008]. When the dimensionless Shields stresses were within this range for the 2007 and 2008 peak flows (τ*50 = 0.06–0.08, compared to for bedrock rivers τ*c50 = 0.03) [Sklar and Dietrich, 2004], transport distance was size dependent in Reach 1 (2007 and 2008) and Reach 2 (2007 only), both of which have longitudinal ribs. Although the dimensionless Shields stress was also within this range in Reach 3 for both years, size-dependent transport did not occur. This suggests that, in bedrock channels with bed forms that do not intersect the flow (longitudinal ribs) ,the processes controlling coarse sediment transport are similar to alluvial channels. However, in bedrock channels with bed forms that intersect and obstruct the flow (oblique ribs), processes that control coarse sediment transport are not similar to alluvial channels. Furthermore, the transport distance of unconstrained clasts in Reach 3 (oblique ribs) was not dependent on tracer size as has been reported in alluvial studies [Church and Hassan, 1992].

[35] Similar to alluvial channels, where variations in channel morphology and sediment structure at different scales are an important control on variations in transport distance [Ferguson et al., 2002; Lisle et al., 2000; Wilcock and Crowe, 2003], the variations in bedrock topography also control bed load transport differences, but at scales potentially smaller than bed form scales of alluvial channels [Schmidt and Gintz, 1995; Thompson et al., 1996; Cudden and Hoey, 2003]. In alluvial channels, long-term storage can occur through deposition in bars, zones of reduced velocity, or clast burial [Ferguson and Hoey, 2002; Ferguson et al., 2002]. The bedrock reaches examined here behave similar to these alluvial systems, but it is the variation in bedrock morphology (bedrock rib geometry) that controls variability in transport distance. Because the unconstrained tracers showed some size dependence in transport distance in Reach 1 and Reach 2, but not in Reach 3, coarse sediment transport may be influenced by factors that are similar to alluvial systems, only in bedrock rivers with relatively homogenous bed topography. On the other hand, if the underlying bedrock lithology and structure create strong heterogeneity in bed forms and channel morphology, then local scale factors are the dominant control on sediment transport. The key difference is that bed forms in bedrock channels can be partly controlled by independent processes (structure and lithology), whereas bed forms in alluvial channels are directly related to the flow and sediment properties. The bedrock ribs examined here exert a strong control on transport patterns at the reach scale. Also, the finest spatial resolution of intrareach variability, which captures local variation in bedrock ribs, was the best predictor of coarse sediment transport distance.

5.3. Temporal Scales and Resolution

[36] Although transport distance was not consistently related to grain size over the study period, it is possible that at longer time scales transport distance varies more closely with grain size. This is because local scale controls that influence transport distance during single flows might become less important when averaged over longer time spans that include multiple transport events.

[37] The D50 of the unrecovered tracers was similar to the D50 of the mobile tracers in the last resurvey. This suggests that unrecovered particles were transported, despite unknown transport distances. It is uncertain whether these particles were transported beyond the reach or moved to locations within the highly variable alluvial deposits and bedrock channel bed topography (i.e., within potholes or between bedrock ribs). Although the recovery rates were relatively large between surveys, there are inherent difficulties in using visually identified tracers. For example, it was necessary to repaint and number most tracers each year because Sun bleaching and algal growth obscured the tracer number.

[38] The shorter transport distances recorded after the four summer 2006 flows (47 m3/s) may reflect the shorter temporal resolution, which facilitates documenting transport distances that may be closer to the hop length of single transport events. However, these flows were also half the magnitude of the yearly peak, after which larger transport distances were measured. It is not clear how the magnitude of individual flow peaks affects the integration of all flow events. Complexity in coarse sediment transport in this system is also demonstrated by some clasts being transported upstream (∼5 m). These few upstream transport events always occurred in sites with substantial bedrock rib exposure, which suggests complex flow patterns in these areas. The turbulent flow structure that is typically associated with bed forms [Nelson et al., 1995; Papanicolaou et al., 2001] is likely to play an important role in determining sediment transport dynamics.

[39] Spatial variations in turbulence structure and thereby sediment flux, which are not describable in terms of local bed shear stress, are important aspects of coarse sediment transport [Nelson et al., 1995; Papanicolaou et al., 2001]. We consistently observed redistribution of gravel located in the lee of bedrock ribs. The gravel-sized tracers in this hydraulic environment showed measurable transport distances in each resurvey, but overall these particles typically remained within the same lee deposit. Tracer particles were commonly buried within these deposits and recovery required sorting through the deposit. This observation is akin to gravel bed rivers [Ferguson et al., 2002], but instead of particles becoming buried as a result of variations in alluvial environment, here, variations in bedrock morphology set up the conditions for burial and vertical mixing.

6. Conclusions

[40] This study demonstrates that lithologic and structural controls on bedrock channel topography (bedrock ribs) exert a major control on the transport dynamics of coarse material over time scales up to 3 year and spatial scales up to 102 m. A comparison of three reaches with different bedrock rib orientations and amplitudes indicated that where bedrock ribs were longitudinal to flow and of lower amplitude, transport was less obstructed, which led to greater transport distances than in reaches where the bedrock ribs were oblique to flow and of larger amplitude. Under similar flow conditions for which size-dependent transport has been observed in alluvial channels (dimensionless Shields stress within 1.5–2.0 times the dimensionless critical shear stress), transport distance was a significant function of grain size where bedrock ribs were longitudinal to flow. Where bedrock ribs were oriented at a high angle to flow however, greater reach scale roughness and lower reach-averaged velocity corresponded to lower transport distances with no relation to tracer size. Because local scale bedrock topography and sediment architecture correlated most strongly with tracer transport distance, channel bed properties at this scale (100–101 m) likely play a dominant role in the transport of coarse sediment. The importance of local scale variables and the lack of size dependence on coarse sediment transport distance where bedrock ribs are oriented oblique to flow may limit the use of reach scale sediment transport formulas, developed for gravel bed streams, in bedrock channels with similar morphologic variability.


[41] This study was funded by the National Science Foundation grant EAR-0507098. We thank William Stubblefield, William Lyons, and Roy Syrmanske for their valuable field assistance. Dmitri Hawk and the Tennessee Valley Authority supplied discharge records for the Ocoee No. 3 dam. The manuscript benefited greatly from insightful and thorough reviews by Phairot Chatanantavet and two anonymous reviewers. Kristen Jaeger and Kyle Nichols provided helpful reviews of an earlier draft.