`10.1029/2019WR027010
`
`Key Points:
`• Initially suspended fine clay
`particles within the water column
`rapidly accumulate within the
`sediment bed due to hyporheic
`exchange
`• Fine clay particle storage occurs
`beneath the mobile layer of the
`sediment bed defined by the extent
`of bedform scour
`• Formation of the fine particle layer
`results in reductions of bedform
`celerity, height, and sediment flux
`while length is unchanged
`
`Supporting Information:
`• Supporting Information S1
`
`Correspondence to:
`J. Dallmann,
`jonathandallmann2020@u.
`northwestern.edu
`
`Citation:
`Dallmann, J., Phillips, C. B.,
`Teitelbaum, Y., Sund, N., Schumer, R.,
`Arnon, S., & Packman, A. I.
`(2020). Impacts of suspended clay
`particle deposition on sand-bed
`morphodynamics. Water Resources
`Research, 56, e2019WR027010. https://
`doi.org/10.1029/2019WR027010
`
`Received 23 DEC 2019
`Accepted 28 APR 2020
`Accepted article online 12 MAY 2020
`
`©2020. American Geophysical Union.
`All Rights Reserved.
`
`DALLMANN ET AL.
`
`Impacts of Suspended Clay Particle Deposition
`on Sand-Bed Morphodynamics
`J. Dallmann1
`, C. B. Phillips2
`, Y. Teitelbaum3, N. Sund4
`and A. I. Packman1
`1Department of Mechanical Engineering, Northwestern University, Evanston, IL, USA, 2Department of Civil and
`Environmental Engineering, Northwestern University, Evanston, IL, USA, 3Zuckerberg Institute for Water Research,
`The J. Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Beersheba, Israel, 4Desert Research
`Institute, Reno, NV, USA
`
`,
`
`, R. Schumer4
`
`, S. Arnon3
`
`Abstract Fine particles (0.1–100 microns) are ubiquitous within the water column. Observations on
`the interactions between suspended fine particles and sediment beds remain limited, reducing our ability
`to understand the interactions and feedbacks between fine particles, morphodynamics, and hyporheic
`flow. We performed laboratory experiments to explore changes in bedform morphodynamics and
`hyporheic flow following the progressive addition of kaolinite clay to the water column above a mobile
`sand bed. We characterized these interactions by taking high-frequency time series measurements of bed
`topography and freestream clay concentration combined with solute injections and bed sediment cores to
`characterize subsurface properties. Deposition of initially suspended clay resulted in a decrease of bedform
`height, celerity, and sediment flux by 14%, 22%, and 29% when 1000 g was accumulated within the bed
`(equal to clay/sand mass ratio of 0.4% in the bed). The hyporheic exchange flux decreased by almost a
`factor of 2 for all clay additions, regardless of the amount of clay eventually deposited in the bed. Post
`experiment sediment cores showed clay accumulation within and below the mobile layer of the bedforms,
`with the peak concentration occurring at the most frequent bedform scour depth. These results
`demonstrate the tight coupling between bed sediment morphodynamics, fine particle (clay) deposition,
`and hyporheic exchange. Suspended and bed load transport rates are diminished by the transfer of
`suspended load to the sediment via hyporheic exchange. This coupling should be considered when
`estimating sediment transport rates.
`
`1. Introduction
`In rivers, fine particles with a diameter of <100 microns consist of particulate organic carbon, minerals such
`as clay, algal and bacterial cells, and other contaminants (Drummond et al., 2014, 2018). Natural sources
`of these fine particles include induced overland flow and erosion, remobilization of fine particles stored in
`the stream bed, bank erosion, landslides, and other mass failures (Belmont et al., 2011; Mueller & Pitlick,
`2013; Owens et al., 2005; Rose et al., 2018; Sekely et al., 2002). Anthropogenic activity, such as mining, agri-
`culture, logging, and urbanization can increase fines within rivers (Karwan et al., 2011; Nelson & Booth,
`2002; Vaughan et al., 2017; Wood & Armitage, 1997; Wolman, 1967). Fine particles represent a significant
`water quality concern (Bilotta & Brazier, 2008) with harmful effects including increased turbidity in pris-
`tine waters (Lloyd et al., 1987), decreasing stream productivity (Ryan, 1991), damage to benthic ecological
`systems (Owens et al., 2005), and hypoxia in coastal systems due to excess nutrients (Ansari, 2005; Paerl
`& Otten, 2013). In addition, the fate of contaminants are linked to the dynamics of fine particles (Foster &
`Charlesworth, 1996; Horowitz, 2009; Zhang et al., 2010).
`The flow of water into and out of the stream bed (hyporheic exchange), fine particle transport, and deposition
`are tightly coupled in river systems (Boano et al., 2014; Harvey et al., 2012; Karwan & Saiers, 2012; Packman
`& Mackay, 2003; Preziosi-Ribero et al., 2020). In the presence of fine particles, hyporheic exchange leads
`to deposition and filtration of fine particles due to advective pumping and turbulent exchange with the
`stream bed (Boano et al., 2014; Packman et al., 2000a, 2000b). The accumulation of fines in the bed via
`filtration, in turn, leads to decreasing hyporheic exchange (Fox et al., 2018; Packman & Mackay, 2003). These
`fines may be stored there for long periods of time spanning multiple flood events (Drummond et al., 2014;
`Harvey et al., 2012). Through their role in setting the storage and release times of fine particles, hyporheic
`exchange and bed sediment transport are important for understanding the long-term fate of contaminants
`
`1 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`and waterborne pathogens (Boano et al., 2014; Drummond et al., 2014). Excessive deposition of fines results
`in the siltation (colmation) of stream beds reducing the transfer of various solutes and particles such as
`organic carbon and regulating heat transfer (Hartwig & Borchardt, 2015). Siltation is expected to impact the
`microbial biomass residing within the upper sediment bed (Merill & Tonjes, 2014) and harm the spawning
`potential of diadromous fish (Chapman, 1988; Greig et al., 2005; Louhi et al., 2011). Reduction in hyporheic
`exchange negatively impacts these communities, leading to increased instream nutrient content (Feris et al.,
`2003, 2004, Li et al., 2017).
`Clay and fine particles are prevalent across many fluvial and marine systems from coarse-grained mountain
`streams to estuarine and shallow marine systems. Coupled fine particle and bed morphodynamic inter-
`actions are expected to occur in sand bedded rivers, estuaries, near coastal environments, and shallow
`marine settings. The presence of stationary sand bedforms, such as dunes and ripples, have been shown to
`greatly increase hyporheic exchange compared to a featureless bed (Elliot & Brooks, 1997; Fox et al., 2018;
`Thibodeaux & Boyle, 1987; Packman & Mackay, 2003). The addition of active bed sediment transport
`remains understudied, though mobile bedforms are known to change hyporheic exchange pathways and
`reduce the rate of nitrogen removal relative to stationary ones (Zheng et al., 2019). In addition, as mobile
`bedforms alter the patterns of fine particle deposition and remobilization (Boano et al., 2014; Packman et al.,
`2001; Phillips et al., 2019), understanding how exchange is impacted by sediment transport is necessary for
`modeling the fate of fine particles in natural systems where mobile bed conditions are common (e.g., floods).
`For stationary bedforms, even relatively small amounts of fine particles can disrupt hyporheic exchange (Fox
`et al., 2018; Packman et al., 2000a, 2000b). However, mobile bedforms disrupt the surface clogging layers
`that develops in stationary beds, leading to no impact on hyporheic exchange with small fine particle addi-
`tions and lower flow rates (Rehg et al., 2005). It remains unclear though how this process will be impacted
`under higher concentrations of fine particles or for sustained background concentrations of fines.
`High concentrations of suspended fine particles have been observed to impact bed morphodynamics due to
`modulation of the stream turbulence. Both river field measurements (Smith & McLean, 1977) and labora-
`tory experiments (Wan, 1984) show that the height to wavelength ratio for sandy bedforms tend to become
`smaller in the presence of large concentrations of suspended clay. Experiments on mixed clay and sand beds
`and premixed clay and sand slurries reveal complex feedback mechanisms between clay concentration, bed-
`form morphodynamics, and flow structure (Baas & Best, 2002, 2008; Best, 2005). In particular, the turbulent
`characteristics of the flow are impacted by higher clay concentrations (conc. > 4 g/L), leading to morphody-
`namic changes (Baas & Best, 2008). This increased turbulence leads to an increase in both bedform height
`and wavelength for increasing freestream clay concentration (Baas et al., 2011). However, in the more cohe-
`sive beds (clay percentage > 13%), winnowing of clay particles produced a segregated bed composed of a
`mobile sand layer above a mixed clay/sand bed (Baas et al., 2013).
`Though the impact of clay on bed morphodynamics under premixed conditions have been well studied, it
`remains unclear how deposition and accumulation of suspended fine particles may impact sand bed mor-
`phodynamics. This mode of interaction is especially relevant in rivers and estuaries, where the introduction
`of clay and other fine particles is episodic in nature, often covarying with higher flows due to runoff gen-
`erated during storms. Further, the ecological implications of the human-induced increases in fine particles
`to rivers can only be inferred without an understanding of how fine particles (clay) introduced to rivers
`from their catchment may impact bed morphodynamics and hyporheic exchange. This study uses four
`experiments to explore the role of clay concentration within the water column on mobile bedforms, clay
`accumulation within the bed, and hyporheic exchange.
`
`2. Methods
`2.1. Experimental Methods
`We performed four experimental runs for a constant freestream velocity consisting of episodic injections of
`kaolinite clay leading to water suspensions of varying concentrations. The experiments differed only in the
`total amount of clay added to the flume and the sequence (i.e., number) of clay injections. The experiments
`were conducted using an 8.5 m long by 0.2 m wide tilting recirculating flume, equipped with a pump (Baldor
`Industrial Motors) that recirculated both water and sediment from the endwell (Figure 1a). Each experiment
`initially consisted of an initially flat clay free mobile sand bed approximately 10 cm thick—250 kg of Flint
`Silica 12 (US Silica, Ottawa, IL) with a D50 of 0.420 mm given by the manufacturer—under a constant flow.
`
`DALLMANN ET AL.
`
`2 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Figure 1. Schematic diagrams of experimental setup and data processing. (a) Schematic of the experimental setup
`(flow is from left to right). The foreground bedform profile and location of the clay layer represent a partial trace from
`sidewall images during an experiment. Experimental measurement devices are located in their approximate locations.
`A porous endplate maintains a minimum sand bed elevation while allowing hyporheic flow to pass. The blue rectangle
`represents the visualization region of the sidewall camera setup shown in panel (b). (b) Sidewall imaging set up for
`bedforms following clay injection during Run 3. The blue rectangle represents the FOV of the cameras. The ADV
`profiler is positioned on a cart directly above the center of the camera FOV. The inset (purple border) shows a close-up
`of a bedform crest and trough and the accumulation of clay below the mobile layer. (c) Smoothed bedform elevation
`data from the ADV profiler showing the extracted bedform heights (peak to trough) identified for a portion of Run 1.
`Note that small bedform ripples (see structure at 44.1 hr) are not treated as individual bedforms.
`
`Mean freestream height (15 cm), mean velocity (0.43 m/s), and shear velocity (u* = 0.026 m/s) were the
`same for all experiments. The u* was determined by fitting a log law velocity profile to a time-averaged
`downstream velocity profile sampled using a Nortek Vectrino Profiler.
`The bed was allowed to run for at least a day until mobile bedforms developed and the size distribu-
`tion reached statistical stationarity. After this developmental period, each experiment consisted of a 4-day
`clay-free period of sand bed load transport (baseline) followed by one or more clay injections. Baselines
`were long enough to ensure that enough bedforms were recorded to accurately determine the clay-free aver-
`age morphodynamic conditions for each run. Relative standard error in measurements of mean bedform
`height dropped below 5% and 0.4% when 75 and 120 bedforms were measured. We performed four experi-
`ments, referred to hereafter as Runs 1–4 (see Table 1 for details). Run 1 consisted of a single clay injection
`of 1,000 g followed by 261.5 hr of bed elevation measurements. Run 2 consisted of an injection of 333 g
`every 4 days totaling 277 hr of observations and three injections. Run 3 represented an initial 700 g injection
`followed by a 300 g clay addition approximately every 1.4 days for the first 300 hr. After 300 hr, the obser-
`vation time between clay injections was increased. Run 4 consisted of a single initial injection of 5,500 g
`followed by 256 hr of observation. The injected clay was kaolinite (Snobrite 75, the Cary Company), with
`a median listed particle diameter of 0.5 μm. For Runs 1–3, the clay was mixed with water matching the
`flume background salinity (350 μS/cm) in beakers with automatic stirrers for 12 hr prior to the injection.
`At the background salinity levels, the clay flocculates and the mean D50 diameter rises to just above 30 μm.
`Due to the large amount of clay added, Run 4 was rapidly mixed over a short period (30 min), leading to
`noticeable amounts of unsuspended clay during the injection. For each injection, the clay was continuously
`poured into the endwell over the measured recirculation time of the flume (40 s). Following each injection,
`
`DALLMANN ET AL.
`
`3 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Table 1
`Information Concerning the Experimental Setup for All Four Runs
`Run 4
`Run 1
`Run 2
`Run 3
`100
`99.5
`98
`100
`Baseline length (hr)
`256
`261.5
`277
`586
`Post baseline length (hr)
`1
`1
`3
`17
`Number of injections
`300a
`5,500
`1,000
`333
`Injection size (g)
`5,500
`1,000
`1,000
`5,500
`Total injected mass (g)
`Note. Each run consists of a baseline without clay after which the first clay
`injection was conducted. The post baseline period occurs after this first injec-
`tion. The number of injections (including the initial injection) and the size
`of each injection are shown.
`aThis run consisted of two closely timed initial injections of 700 and 300 g
`followed by regular injections of 300 g.
`
`the system was allowed to evolve and changes in the clay concentration, hyporheic exchange, and bedform
`morphodynamics were observed.
`Suspended clay concentration was continuously measured at 1-min intervals with a Xylem turbidity meter
`(Runs 1 and 2—WTW Visoturb 700IQ SW, Runs 3 and 4 - WTW Visolid 700IQ SW) positioned just upstream
`of the flume endwell (Figure 1a). Concentration measurements for the first 6.75 hr of Run 1 were taken
`by hand every hour using a syringe and processed via a spectrophotometer (Hach Company, DR/4000).
`Because initial instream clay concentrations exceeded the measurement range of the Visoturb 700IQ SW, a
`calibration curve relating known concentrations of kaolinite to the absorbance of 600 nm light was used to
`determine the concentration of the samples.
`The hyporheic exchange flux (HEF) was measured through salt tracer injections during the baseline and
`following the end of the experiment. To measure HEF, freestream salinity was recorded following dissolved
`NaCl tracer injections, typically 10 hr long, using a salinity meter (SM Star Comm, resolution of 0.01 μS/cm).
`The initial HEF was calculated via regression of the rate of decline in salt concentration with time immedi-
`ately following the NaCl tracer injection, following the methodology of Fox et al. (2014). No measurement
`of HEF was performed for the end of Run 4 due to a leak in the flume, which significantly increased the rate
`of flow into the bed.
`Elevation of the sand bed was recorded at a point with an acoustic Doppler velocimeter (ADV) profiler and
`over a large spatial area with two digital single-lens reflex Nikon D5300 cameras. The ADV was positioned
`in the center of the camera visualization region located 355 cm from the downstream end of the flume (see
`Figure 1a). The ADV profiler recorded depth to the sediment bed (2 Hz) from a fixed elevation. Images
`were taken every minute to visualize bedform propagation and determine bedform length and celerity (see
`section 2.2, Figure 1b). The cameras were affixed to mounting arms attached to a table adjacent to the flume.
`They were faced perpendicular to the bed, providing a combined field of view of 180 cm, centered on the
`profiler and approximately 2.25 times the length of an average bedform. The visualization region was backlit
`to provide sufficient contrast for automated feature extraction.
`At the conclusion of Runs 1–3, 36 cores were taken of the bed following the protocol of Fox et al. (2014) and
`analyzed for bed clay composition to yield depth averaged clay masses for each 0.5-cm depth interval. In
`order to extract a core, the flow was stopped, free stream clay was allowed to settle, and the water level was
`slowly decreased until touching the top of the bed. A syringe was used to remove the settled clay from the
`surface of the bed by suctioning out the water immediately above the bed in the core measurement region.
`For two bedforms, three cores were taken, equally spaced in the cross stream direction at six locations: at
`the downstream trough, at the crest, and at four equally spaced locations between the trough and crest. In
`total, 18 cores were taken for each bedform sampled. Each core consisted of a 35-ml syringe, 11 cm long
`by 2.13 cm wide with the tapered end removed. Each core was carefully inserted into the bed, sealed from
`the bottom, and removed from the flume. Sediments from the cores were extracted in 0.5-cm increments
`and mixed with 50 ml of deionized water to create a clay suspension. The mixture was weighed and the
`concentration was determined via absorbance at 600 nm using a spectrometer (Hach Company, DR/4000).
`
`DALLMANN ET AL.
`
`4 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Table 2
`Experiment and Bedform Statistics for All Four Runs
`Qt
`C
`t
`H
`L
`CB
`HEF
`N
`Run
`Base − 1
`12.1
`146
`0.00541
`0.770
`362
`0
`0.0227
`0.777
`Base − 2
`12.5
`175
`0.00624
`0.904
`379
`0
`0.0235
`0.918
`686a
`Base − 3
`13.5
`169
`0.00603
`0.870
`0
`0.0234
`0.865
`N/Ab
`Base − 4
`156
`0.00681
`1.039
`356
`0
`0.0223
`0.838
`Clay − 1
`4.6
`146
`0.00400
`0.634
`362
`322
`0.0205
`0.730
`Clay − 2
`4.7
`128
`0.00556
`0.835
`379
`239
`0.0214
`0.825
`686a
`Clay − 3
`5.7
`132
`0.00486
`0.763
`609
`0.0209
`0.796
`N/Ab
`Clay − 4
`143
`0.00442
`0.769
`356
`990
`0.0181
`0.805
`Note. The total length of the entire run (baseline and post baseline) is given by t and measured in
`hours. The first four rows represent clay-free baseline data, and the last four represent data taken
`after clay injection. Each measurement period is 4 days long. For the baseline data, the clay mass in
`the bed - CB (g) is 0, while for the clay runs, CB represents the average clay in the bed over the final
`4-day window. H, L, C, and Qt are averages of bedform height (m), length (m), celerity (m/hr), and
`sediment flux (m2/hr) for the initial 96 hr (Base) and final 96 hr (Clay). N is the number of bed-
`forms measured during each 4-day period as recorded by the ADV profiler. HEF is the hyporheic
`exchange flux in cm/day.
`aTime here is the duration of the entire run. Because of a camera data collection failure near the
`end of the run (see Figure S2 in the supporting information), results in this table were gathered
`bDue to a flume leak, it was
`for the contiguous 4-day period preceding the failure (479-575 hr)
`impossible to obtain final hyporheic exchange data for this run.
`
`These concentrations were subsequently converted into a clay percentage by mass for each depth slice within
`the core.
`
`2.2. Data Processing
`The time series of the bed elevation recorded by the ADV profiler was smoothed using a Savitzky-Golay
`filter (Python 2.7 SciPy, window size of 509 data points, 255 s) and processed to remove extraneous noise on
`the elevation signal. A “find peaks” algorithm (Python 2.7 SciPy) was used to identify both the troughs and
`the peaks of the bed elevation. The height of individual bedforms (H) was defined as the vertical distance
`between the bedform peak and downstream (lee side) trough (Figure 1c). Small transient ripples, persisting
`for no more than several minutes with H < 0.5 cm, were removed from the time series prior to calculating
`the final bedform statistics. For each run, Table 2 shows the number of bedforms identified during both the
`baseline period and the last 4 days of data collection.
`Bedform length (L) and celerity (C) were obtained by image analysis of the sidewall camera images. Raw
`images from each camera were thresholded using a simple black/white thresholding procedure (MATLAB
`R2019a) that involved using a manually identified black/white pixel cutoff to determine the sediment water
`interface. The images from both cameras then were stitched together to extract an elevation profile for
`the full 180-cm field of view, allowing for the simultaneous visualization of multiple bedforms. Increased
`freestream clay concentration decreased the light exposure requiring manual calibration of the thresholding
`algorithm following each clay addition. Stitched images were generated, and the black/white pixel cutoff was
`shifted until the sediment water interface was correctly identified. During data processing, sample images
`were saved every 100 min to ensure that the interface was correctly identified and that clay deposition didn't
`alter the camera light exposure. The resultant sidewall bed elevation profile was processed to extract bedform
`height using the same methods developed for the ADV profiler. Bedform lengths were calculated as the dis-
`tance between successive troughs, while celerity was calculated as the slope of a linear regression line fit to
`a trough's downstream location over time. The bedload sediment flux (Qt) was determined as Qt (t) = 𝛽𝜓cH
`(Bagnold, 1941; Martin & Jerolmack, 2013; McElroy & Mohrig, 2009; Simons, 1965) where the bedforms
`were approximated as triangles with a shape factor 𝛽 = 0.5 (Martin & Jerolmack, 2013). The relative pro-
`portion of clay within the bed sediment remained low in all experiments, so the porosity was assumed to
`remain constant for the sand mass flux calculations (𝜓 = 0.48). Average quantities for the baseline and final
`set of bedforms are provided in Table 2.
`
`DALLMANN ET AL.
`
`5 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Table 3
`Statistical Comparison of Bedform Morphologies Pre and Post Clay Injection
`PH (R2
`PL (R2
`PQt (R2
`Pc (R2
`Run
`)
`c )
`H)
`L)
`Qt
`Base − 1
`0.96 (0.00)
`0.75 (0.00)
`0.78 (0.00)
`0.22 (0.01)
`Base − 2
`0.56 (0.00)
`0.40 (0.00)
`0.16 (0.02)
`0.90 (0.00)
`Base − 3
`0.90 (0.00)
`0.31 (0.01)
`< 0.01 (0.15)
`0.03 (0.04)
`Base − 4
`0.70 (0.00)
`0.38 (0.00)
`0.25 (0.01)
`0.51 (0.00)
`Clay − 1
`0.74 (0.00)
`0.09 (0.01)
`< 0.01 (0.08)
`0.01 (0.02)
`Clay − 2
`0.72 (0.00)
`0.59 (0.00)
`0.81 (0.00)
`0.72 (0.00)
`Clay − 3
`0.01 (0.01)
`0.40 (0.00)
`0.02 (0.01)
`< 0.01 (0.03)
`Clay − 4
`0.14 (0.01)
`0.82 (0.00)
`< 0.01 (0.04)
`< 0.01 (0.03)
`Mean − 1
`0.016
`0.390
`< 0.001
`< 0.001
`Mean − 2
`0.045
`0.131
`< 0.001
`0.005
`Mean − 3
`0.011
`0.436
`< 0.001
`0.005
`Mean − 4
`0.010
`0.330
`< 0.001
`< 0.001
`Note. The rows labeled Base and Clay contain the statistical results of linear
`regressions (p values and R2 values) for trends over time within the baseline and
`post-clay injection bedform morphologies (H, L, C and Qt). Rows labeled Mean−
`represent results (p values) of Mann-Whitney U tests between the mean of the
`baseline and final 96 hr of bedforms in the presence of clay for H, L, C, and Qt.
`
`Statistical tests were applied separately to the bedform time series data and between the beginning and
`end of each run to assess potential differences between the pre-clay and post-clay bedform data. First, the
`degree to which the morphodynamic properties H, L, C, and Qt changed over time was determined via
`linear regression. Significance of trends in the baseline and post-clay injection time series was assessed by
`considering the regression p value where a value greater than 𝛼 = 0.05 was taken to indicate that the trend
`was not significantly different from zero over time (see Table 3). Second, a Mann-Whitney U test was used to
`compare whether H, L, C, and Qt from the final period (final 96 hr) of data collection post-clay injection were
`less than the same metrics collected during the clay-free baseline period (initial 96 hr, Table 3). Differences
`were considered significant for an 𝛼 ≤ 0.05. Cumulative distribution functions (CDFs) and kernel density
`estimation (KDE) were used to visualize distributions and potential changes. The CDFs were used to assess
`how the addition of clay impacted the entire population of measured and derived bedform quantities. KDE
`was used to create probability density functions (PDFs) of the trough and bed elevation to assess how these
`quantities impacted clay accumulation. Gaussian kernels were used with Scott's Rule employed to calculate
`the estimator bandwidth.
`
`3. Results
`The results of the four experiments consist of time series of bed elevation, freestream concentration of clay,
`and the conductivity of the water column following salt tracer injections. From these data sources, we derive
`bedform morphodynamic variables, clay accumulation within the bed, and the HEF. The four experimental
`runs were designed to explore the impact of clay accumulation on a population of mobile bedforms and
`in turn how these combined effects impacted hyporheic exchange. A secondary aim of these experiments
`was to determine if the current state of bed morphodynamics depended on the history of clay additions
`(sequence) or only on the current free stream concentration at any given time. We first discuss the results
`of the clay deposition rates and within-bed patterns of accumulation and how these impacted hyporheic
`exchange. Second, we discuss the impact of clay deposition on bedform morphodynamics.
`3.1. Hyporheic Exchange and Clay Deposition
`The four experimental runs are best conceptualized as two sets of paired experiments based on their total
`added clay mass (Figure 2a). Runs 1 and 2 consisted of 1,000 g of clay, while Runs 3 and 4 consisted of 5,500 g
`of clay. Runs 1 and 4 were single injections, while the 1,000 g in Runs 2 and 3 consisted of multiple injections.
`Final mean freestream concentrations over the last 5 hr of measurement are 66.0%, 76.5%, 88.4%, and 81.5%
`of the initial concentration for Runs 1–4, respectively (Figure 2a). The multi-injection runs (2 and 3) final
`
`DALLMANN ET AL.
`
`6 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Figure 2. Patterns of clay deposition. (a) Time series of freestream clay concentrations for Runs 1–4. Jumps in
`freestream concentration reflect clay injections. (b) Accumulation of clay within the bed over time.
`
`freestream concentration exceeds that of their paired single injection run, indicating less deposition overall
`for the multi-injection runs (Figures 2a and 2b). The absence of storage or loss within the recirculating flume
`means that persistent decreases in freestream clay concentration can be taken as deposition within the bed
`(Figure 2b). Comparisons between experiments with the same amount of total mass injected show that runs
`with multiple injections produce less deposition over time by a factor of 1.3 to 1.55 relative to runs with a
`single injection. For Runs 1–3, the rate of deposition is initially rapid (111.0, 33.8, and 104.2 g/hr over the
`first 2 hr) and is roughly proportional to the injection size. Though the rate of clay accumulation within
`the bed decreases over time, we did not observe the emergence of a steady-state concentration within any
`of the runs indicating that deposition was ongoing (Figure 2b). Run 4 shows unexpected clay depositional
`behavior compared to other experiments resulting in an initial increase in deposition on the surface of the
`bedforms followed by decline and eventual stabilization (Figure 2b). The variations in Run 4 may be due
`to incomplete mixing of the clay prior to the injection as described in section 2.1. For all runs short time
`scale (1–2 hr) periodicity within the freestream clay concentration timeseries (Figure 2a) is a mixture of
`short-term bedform deposition and remobilization and sensitivity of the turbidity meter to the changing
`
`Figure 3. Decreasing hyporheic exchange with increasing clay deposition. A hyporheic exchange measurement was
`not made for Run 4 due to a flume leak just before the exchange measurement that pulled excessive clay into the
`subsurface. (a) Normalized freestream salt concentration following the clay-free baseline (lower three curves) and end
`of experiment (upper three curves). (b) Hyporheic exchange flux as a function of clay mass deposited in the bed. The
`inset shows how the HEF is calculated as the slope of the normalized concentration following the injection.
`
`DALLMANN ET AL.
`
`7 of 15
`
`
`
`Water Resources Research
`
`10.1029/2019WR027010
`
`Figure 4. Visualization of clay accumulation for a single experimental bedform from Run 1. The bold blue line
`represents the surface of an experimental bedform at the end of a run traced from the sidewall (see inset for
`photograph of the bedform). The bed profile is not to scale as the downstream distance has been compressed by a factor
`of approximately 9 relative to the vertical dimension. The PDF of elevation to the right of the profile represents the
`complete probability of surface elevation location for the entirety of the run following the clay injections. The deepest
`scour is the lowest recorded elevation and is denoted by the blue dashed line. The red vertical lines and shading
`represent the average clay concentration of three cores in the cross-stream direction normalized by the maximum
`observed concentration for the whole bedform. The zero-concentration point represents the approximate location of the
`core on the bed profile. The inset above the figure shows this bedform from a sidewall view. Flow is from left to right.
`
`distance to the bed surface due to the passage of bedforms. Tests in clear water conditions indicate a 15%
`increase in turbidity over bedform crests compared to troughs.
`The impact of clay deposition on the HEF was assessed through the injection of a conservative salt tracer
`(Figure 3). HEF was computed as the rate of change in freestream tracer concentration immediately after
`the injection (see Figure 3b inset). Initial tracer injections under baseline (i.e., clay free) conditions show
`similar early time exchange rates of between 12.1 and 13.5 cm/day (Figure 3b). Following all clay injections,
`the HEF declined to between 4.6 and 5.7 cm/day, a decline of between 38% and 42% for all experiments. The
`long-term salt concentration in the freestream decreases slightly with clay in the bed (Figure 3a), and the
`overall difference in values of normalized conductivity between experiments remains small (less than 1%
`after 10 hr). Overall, the measured HEF at the conclusion of the experiment is approximately constant
`regardless of the amount of clay accumulated within the bed (Figure 3b).
`Despite a very low Stokes settling velocity (Us = 8.10 ∗ 10−4 m/s) and sufficient shear velocity to keep the
`clay suspended (Rouse number of 0.08), accumulation within the bed begins almost immediately following
`injection and accumulates visually in a layer approximately 2 cm below the active region of sand transport
`(Figure 4). The thickness of the clay accumulation layer is not constant throughout the flume or in a given
`location over time and depends on the history of bed elevation changes at a specific location. Deposition
`from burial within the troughs and hyporheic flow are functions of local bedform size, while scour from bed-
`forms results in remobilization of previously deposited clay. Because of this, individual sediment cores show
`variations in overall clay concentration and the variation in concentration depth profile (Figure 4). How-
`ever, when depth profiles are averaged across all cores, a clear pattern of accumulation emerges (Figure 5).
`In aggregate, the concentration of clay increases over the first 3–4 cm within the bed before reaching a peak
`value that is between 4.2 and 11.6 times the concentr