UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Some physical and biological factors influencing the fate of fine clastic particles in flowing water Salant, Nira Liat 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2009_fall_salant_nira.pdf [ 2.39MB ]
Metadata
JSON: 24-1.0067301.json
JSON-LD: 24-1.0067301-ld.json
RDF/XML (Pretty): 24-1.0067301-rdf.xml
RDF/JSON: 24-1.0067301-rdf.json
Turtle: 24-1.0067301-turtle.txt
N-Triples: 24-1.0067301-rdf-ntriples.txt
Original Record: 24-1.0067301-source.json
Full Text
24-1.0067301-fulltext.txt
Citation
24-1.0067301.ris

Full Text

SOME PHYSICAL AND BIOLOGICAL FACTORS INFLUENCING THE FATE OF FINE CLASTIC PARTICLES IN FLOWING WATER by NIRA LIAT SALANT  B.A., Dartmouth College, 2003 M.Sc., Dartmouth College, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES (Geography)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2009  © Nira Liat Salant, 2009  ABSTRACT An experimental flume study was conducted to assess the influence of several physical and biological factors on the movement and deposition of fine particles (< 125 µm) in flowing water. Mechanisms of particle movement were elucidated from measurements of flow hydraulics, particle concentrations, surface deposition, and subsurface infiltration for varying flow rates, bed sand fractions, particle densities, initial concentrations, and periphyton structures. Results showed that low flows slowed total deposition, an unexpected result given the lower near-bed Reynolds stresses and velocities of this condition. Similarly, a bed with a high sand fraction also slowed total deposition despite having lower near-bed Reynolds stresses. A higher amount of surface deposition to the high sand bed was offset by limited subsurface deposition, likely due to the clogging of pore spaces by fine sand and reduced advective transport. Particle density also significantly altered deposition rate but had no effect on particle infiltration or flow hydraulics. Along a gradient of low to high initial concentrations, deposition rate and infiltration increased, due to greater particle availability and an increase in particle interactions. A comparison of theoretical and measured concentration profiles showed that for fine particles the Rouse equation, using a depth-integrated particle size, performed as well as or better than more complex models. All models under-predicted concentrations of low-density plastic particles, over-predicted at low concentrations, and performed better with a high sand bed. Periphyton had a significant effect on hydraulics and deposition for a range of structures, densities and spatial scales. High density, closed periphyton patches compacted under high flows resulted in higher velocities and lower near-bed Reynolds stresses by constricting the flow depth and smoothing the bed surface. Lower density patches increased bed roughness, reducing near-bed velocities and transferring turbulent shear upward. Mucilaginous diatoms at low to moderate biomasses increased deposition rate and surface deposition by reducing near-bed Reynolds stress and enhancing particle adhesion. However, at high biomasses, diatom assemblages clogged interstitial spaces and reduced the amount of subsurface deposition thus slowing total deposition. In contrast, deposition occurred more slowly for most growth stages of filamentous algae, possibly due to partial clogging of the bed and a lack of surface adhesion. However, later algal growth stages increased Reynolds stress and advective transport, in turn increasing the amount of subsurface deposition and thus total deposition rate.  ii  TABLE OF CONTENTS ABSTRACT ................................................................................................................................................... ii TABLE OF CONTENTS............................................................................................................................... iii LIST OF TABLES .......................................................................................................................................... v LIST OF FIGURES ...................................................................................................................................... vii LIST OF SYMBOLS AND ABBREVIATIONS .......................................................................................... xi ACKNOWLEDGEMENTS ......................................................................................................................... xiv CO-AUTHORSHIP STATEMENT .............................................................................................................. xv CHAPTER 1 : INTRODUCTION .................................................................................................................. 1 1.1 STATEMENT OF THE PROBLEM ............................................................................................................... 1 1.2 CONCEPTUAL HYPOTHESES ................................................................................................................... 2 1.3 LITERATURE REVIEW ............................................................................................................................ 4 1.3.1 FINE PARTICLE TRANSPORT AND DEPOSITION IN STREAMS ............................................................ 4 1.3.2 MODELS OF PARTICLE DISTRIBUTION AND DEPOSITION ................................................................. 6 1.4 OPERATIONAL HYPOTHESES AND APPROACH ...................................................................................... 11 1.5 REFERENCES ....................................................................................................................................... 14 CHAPTER 2 : PHYSICAL FACTORS INFLUENCING THE DISTRIBUTION, DEPOSITION AND INFILTRATION OF FINE PARTICLES TO A STREAMBED ................................................................. 18 2.1 INTRODUCTION ................................................................................................................................... 18 2.2 METHODS............................................................................................................................................ 19 2.2.1 FLUME DESCRIPTION ................................................................................................................... 19 2.2.2 EXPERIMENTAL DESIGN ............................................................................................................... 20 2.2.3 MEASUREMENTS ......................................................................................................................... 23 2.3 ANALYSIS ........................................................................................................................................... 27 2.3.1 VELOCITY AND TURBULENCE PARAMETERS ................................................................................ 28 2.3.2 PARTICLE DEPOSITION RATES ...................................................................................................... 29 2.4 RESULTS AND DISCUSSION .................................................................................................................. 32 2.4.1 REFERENCE CONDITION ............................................................................................................... 32 2.4.2 FLOW RATE ................................................................................................................................. 43 2.4.3 BED SAND FRACTION ................................................................................................................... 47 2.4.4 PARTICLE DENSITY ...................................................................................................................... 49 2.4.5 INITIAL PARTICLE CONCENTRATION ............................................................................................ 51 2.4.6 COMPARISONS AMONG PHYSICAL CONDITIONS............................................................................ 53 2.5 CONCLUSIONS ..................................................................................................................................... 55 2.6 REFERENCES ....................................................................................................................................... 57 CHAPTER 3 : MODELS FOR THE VERTICAL DISTRIBUTION OF FINE SUSPENDED PARTICLES ...................................................................................................................................................................... 59 3.1 DESCRIPTION AND APPLICATION OF MODELS ...................................................................................... 59 3.1.1 ROUSE EQUATION ........................................................................................................................ 61 3.1.2 MODIFIED ROUSE EQUATION ....................................................................................................... 62 3.1.3 MODEL OF CAO ET AL. (1995) ..................................................................................................... 63 3.1.2 LOCAL EXCHANGE MODEL ......................................................................................................... 64 3.2 RESULTS ............................................................................................................................................. 69 3.2.1 MEASUREMENT ERROR AND EXPERIMENTAL VARIABILITY .......................................................... 69 3.2.2 MODEL PERFORMANCE: ALL FORMULATIONS ............................................................................. 71 3.2.3 EFFECT OF PARTICLE SIZE ............................................................................................................ 77 3.2.4 EFFECT OF SHEAR VELOCITY PARAMETER ................................................................................... 79 3.3 CONCLUSIONS ..................................................................................................................................... 83 3.4 REFERENCES ....................................................................................................................................... 84 CHAPTER 4 : EFFECTS OF PERIPHYTON PATCHES ON HYDRAULICS OF GRAVEL-BED FLOW ...................................................................................................................................................................... 85 4.1 INTRODUCTION ................................................................................................................................... 85 4.1.1 CONCEPTUAL MODEL OF PERIPHYTON-FLOW DYNAMICS ............................................................. 87 4.2 METHODS............................................................................................................................................ 90  iii  4.3 ANALYSIS ........................................................................................................................................... 96 4.3.1 VELOCITY PROFILES AND PARAMETERS....................................................................................... 96 4.3.2 TURBULENCE PROFILES AND PARAMETERS.................................................................................. 97 4.4 RESULTS ............................................................................................................................................. 98 4.4.1 HYDRAULIC CONDITIONS AND PERIPHYTON CHARACTERISTICS................................................. 102 4.4.2 VELOCITY PROFILES .................................................................................................................. 103 4.4.3 TURBULENCE PROFILES AND PARAMETERS................................................................................ 106 4.5 DISCUSSION ...................................................................................................................................... 108 4.5.1 PERIPHYTON EFFECTS ON VELOCITY PROFILES .......................................................................... 109 4.5.2 PERIPHYTON EFFECTS ON TURBULENCE..................................................................................... 110 4.5.3 APPLICATION OF PERIPHYTON-FLOW MODELS ........................................................................... 112 4.5.5 COMPARISONS WITH LARGE-SCALE ANALOGUES ....................................................................... 115 4.6 CONCLUSIONS ................................................................................................................................... 116 4.7 REFERENCES ..................................................................................................................................... 118 CHAPTER 5 : ‘STICKY BUSINESS’: THE INFLUENCE OF STREAMBED PERIPHYTON ON PARTICLE DEPOSITION AND INFILTRATION ..................................................................................................................................... 120 5.1 INTRODUCTION ................................................................................................................................. 120 5.2 METHODS.......................................................................................................................................... 122 5.2.1 PERIPHYTON CULTIVATION ....................................................................................................... 124 5.2.2 MEASUREMENT SCHEDULE AND EXPERIMENTAL DESIGN .......................................................... 124 5.2.3 SURFACE SAMPLES .................................................................................................................... 125 5.2.4 STRATIFIED TRAY SAMPLES ....................................................................................................... 125 5.3 ANALYSIS ......................................................................................................................................... 125 5.4 RESULTS AND DISCUSSION ................................................................................................................ 126 5.4.1 REFERENCE EXPERIMENTS......................................................................................................... 126 5.4.2 PERIPHYTON GROWTH AND DEVELOPMENT ............................................................................... 131 5.4.3 DIATOM EFFECTS ON HYDRAULICS AND PARTICLE DEPOSITION ................................................. 136 5.4.4 ALGAE EFFECTS ON HYDRAULICS AND PARTICLE DEPOSITION ................................................... 148 5.4.5 RELATION BETWEEN PERIPHYTON BIOMASS AND INORGANIC MASS .......................................... 156 5.4.6 COMPARISONS BETWEEN DIATOMS AND ALGAE ........................................................................ 161 5.5 CONCLUSIONS ................................................................................................................................... 162 5.6 REFERENCES ..................................................................................................................................... 164 CHAPTER 6: CONCLUSIONS ................................................................................................................. 166 APPENDIX................................................................................................................................................. 170  iv  LIST OF TABLES Table 2.1: Experiment code names and descriptions. 20%SHF is the reference condition. ......................... 21 Table 2.2: Velocity and turbulence statistics for different physical conditions. ........................................... 33 Table 2.3: Regression coefficients and associated parameters for streamwise velocity regressed against ln(z) for six physical conditions. .................................................................................................................... 33 Table 2.4: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h. ........................................................................................................................................ 39 Table 2.5: Model and depositional parameters from 0-1h and 1-8h for four different particle size classes (in µm) from the three replicate reference condition experiments (20%SHF1, 2, and 3). .......................... 41 Table 3.1: Statistics for the distribution of median particle sizes at 4h and 8h for all experimental conditions............................................................................................................................................... 66 Table 3.2: Replicated concentration and particle size measurements. .......................................................... 67 Table 3.3: Description of modeling approaches for the vertical distribution of suspended particles............ 68 Table 3.4: Statistics for the difference in concentration (C) and median particle size (D50) between two replicated measurements for each experimental condition; ‘n’ is the number of replicated measurements. ....................................................................................................................................... 69 Table 3.5: Regression parameters for measured concentrations normalized by the near-bed measurement regressed against normalized height above the bed from 4h and 8h for all experimental conditions. ... 70 Table 3.6: Regression parameters for normalized measured median particle size profiles from 4h and 8h for all experimental replicates. .................................................................................................................... 71 Table 3.7: Percent RMSE values of suspended particle profiles for seven different models and six experimental conditions. ........................................................................................................................ 74 Table 3.8: Percent bias estimates of suspended particle profiles for seven different models and six experimental conditions. ........................................................................................................................ 75 Table 3.9: Percent RMSE and bias values of concentration profiles predicted by the Rouse-Mean equation using three different estimates of shear velocity (u*) for six experimental conditions. ......................... 82 Table 4.1: Hydraulic and sedimentological parameters for the flume, bed material, and periphyton clusters at high and low flow (HF and LF). ........................................................................................................ 91 Table 4.2: Mat characteristics, velocity and turbulence parameters, coefficients of determination and ANCOVA p-values from velocity-height regressions for each experiment at a) HF and b) LF. ........... 99 Table 4.3: Periphyton structure, velocity, and turbulence statistics for periphyton-covered and nonperiphyton clusters. .............................................................................................................................. 101  v  Table 4.4: Significance values (p-values) for non-parametric Kruskal-Wallis (KW) and Wilcoxon tests for several mat characteristics and hydraulic parameters compared between stone clusters with one of two periphyton mat types – Closed (C) or Open (O) – or without periphyton (N) for two flow rates. ..... 101 Table 5.1: Regression coefficients for velocity as a function of ln(z) and hydraulic parameters from reference experiments with and without trays. .................................................................................... 127 Table 5.2: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for reference experiments with (‘Trays’) and without trays (‘No trays)....................... 128 Table 5.3: Regression coefficients for velocity as a function of ln(z) and hydraulic parameters from reference and periphyton conditions. ................................................................................................... 137 Table 5.4: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for periphyton and reference experiments. ................................................................... 139 Table 5.5: Velocity regression coefficients (SE) and mean hydraulic parameters from reference and algae experiments. ......................................................................................................................................... 149 Table 5.6: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for algae and reference experiments. ............................................................................ 151 Table 5.7: Information about field studies used to determine the relation between AM and AFDM of surface periphyton samples. ................................................................................................................. 158 Table 5.8: Regression coefficients for the relation between periphyton biomass (AFDM) and inorganic content (AM) of surface samples from field studies with a range of locations, stream types, hydraulic conditions, periphyton types, substrates, and sediment regimes. ......................................................... 159 Table A.1: Velocity regression coefficients (SE) and mean hydraulic parameters from each bed location, experimental run, and experimental condition of the physical experiments. ....................................... 170 Table A.2: Velocity regression coefficients (SE) and mean hydraulic parameters from each bed location, experimental run, and experimental condition of the reference and periphyton experiments. ............ 171 Table A.3: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for four different particle size classes from periphyton and reference experiments. .... 172  vi  LIST OF FIGURES Figure 1.1: Schematic illustrating all possible combinations of experimental factors, representing a complete experimental design. .............................................................................................................. 12 Figure 2.1: Schematic illustrating all possible combinations of experimental factors. ................................. 18 Figure 2.2: Schematic of flow and sediment re-circulating flume, University of British Columbia, Canada. ............................................................................................................................................................... 20 Figure 2.3: Bulk particle size distributions for artificial mixtures used in experiments (20% and 80% sand). Solid lines indicate the 50th percentile grain size for each distribution. ................................................. 21 Figure 2.4: Relation between a) expected volume concentrations (VC) based on a known particle dose and water volume and LISST-measured concentrations; and b) LISST-measured volume concentrations and optical transmission of the LISST laser beam (range from 0-1; 1 = 100% transmission)............... 26 Figure 2.5: Log-normalized velocity profiles from the third replicate reference experiment at the a) 2.5 m, b) 3 m, and c) 3.5 m locations of the streambed at three measurement times, as well as at d) the 0 h measurement time at all three locations. ................................................................................................ 35 Figure 2.6: Reynolds stress (τRe) profiles from the third replicate reference experiment at the a) 2.5 m, b) 3 m, and c) 3.5 m locations of the streambed at three measurement times, as well as at the d) 0 h measurement time at all three locations. ................................................................................................ 36 Figure 2.7: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for six physical conditions. .............................................. 38 Figure 2.8: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for the three replicate reference condition experiments from 0-8h; the shaded region is the 0-1h period. ........................................................................................................... 40 Figure 2.9: Percentage of particles from the surface to the flume bottom for the three replicate reference experiments (20%SHF1, 2, and 3). ........................................................................................................ 42 Figure 2.10: Streamwise velocity (u) as a function of log-normalized height above the bed (ln(z)) for six physical conditions. ............................................................................................................................... 44 Figure 2.11: Average decay rates (k) of the 0-1h and 1-8h time periods and the average near-bed (τRe0) and maximum Reynolds stress (τReM) for each experimental condition. ...................................................... 45 Figure 2.12: Percentage of particles <125 µm in each of four bed layers taken from the surface to the flume bottom for the reference condition and a) a low flow rate, b) an 80% sand bed, c) a plastic particle suspension, and d) low and high initial concentration.. ......................................................................... 46 Figure 2.13: Log-normalized velocity (ux) profiles at the a) 3 m and b) 3.5 m locations at four different times during the first 80% sand replicate experiment (80%SHF1). ....................................................... 48 Figure 2.14: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for the first replicate reference condition experiment (20%SHF1) and the first replicate plastic particle experiment (20%PHF1) from 0-8h; the shaded region is the 0-1h period. ............................................................................................................................................................... 50  vii  Figure 2.15: Measured Reynolds stress (τRe) profiles with corresponding loess model fits (solid and dashed lines) from three streambed locations of the high concentration experiment (20%SHFHC) and the three replicate experiments of the reference condition (20%SHF1, 2, and 3) . .............................................. 52 Figure 2.16: Percentage of particles <125 µm in each of four bed layers taken from the surface to the flume bottom for all physical conditions. ........................................................................................................ 55 Figure 3.1: Theoretical particle concentration profile as it varies with Rouse number (s) ........................... 62 Figure 3.2: Normalized measured concentration profiles from 4h and 8h for all experimental conditions. . 70 Figure 3.3: Normalized measured median particle sizes from 4h and 8h for all experimental replicates plotted against normalized height above the bed. .................................................................................. 71 Figure 3.4: Normalized theoretical profiles of suspended particle concentrations for seven model formulations and six experimental conditions at 4h. ............................................................................. 72 Figure 3.5: Normalized theoretical profiles of suspended particle concentrations for seven model formulations and six experimental conditions at 8h. ............................................................................. 73 Figure 3.6: Normalized concentration profiles from the Rouse and LEM equations using the median particle size of size classes 2-4 µm, 4-16 µm, 16-63 µm, and 63-122 µm for four experimental conditions at 4h. ..................................................................................................................................... 78 Figure 3.7: Normalized concentration profiles from the Rouse-Mean equation using three different estimates of shear velocity based on three estimates of shear stress (τRe0, τReM,, τU.) for six experimental conditions at 4h. ..................................................................................................................................... 80 Figure 3.8: Normalized concentration profiles from the Rouse-Mean equation using three different estimates of shear velocity based on three estimates of shear stress (τRe0, τReM,, τU.) for six experimental conditions at 8h. ..................................................................................................................................... 81 Figure 4.1: A conceptual model of periphyton-flow dynamics. One-way arrows indicate an influential effect from one factor to another. .......................................................................................................... 88 Figure 4.2: Schematic representation of the influence of flow rate and a) filament density or b) filament height on local flow structure. ............................................................................................................... 89 Figure 4.3: Measurement set-up for periphyton patches. .............................................................................. 94 Figure 4.4: Relation between the height of the periphyton mat and the lowest height at which the ADV could take a measurement without interference from the bed. .............................................................. 98 Figure 4.5: Near-bed streamwise velocity (u0) and near-bed Reynolds stress (τRe0) as a function of areal density (measured as AFDM per surface area) (a, c) and mat height (b, d) of periphyton clusters at two flow rates (HF and LF). ....................................................................................................................... 102 Figure 4.6: Streamwise velocity (u) profiles from representative closed (D) and open (B) mats for two flow rates...................................................................................................................................................... 104  viii  Figure 4.7: Streamwise velocity (u) profiles at two flow rates from representative experiments K, L, P, and Q for which logarithmic regression lines are not significantly different at one flow rate, but the nearbed region of the profile is slower for the periphyton run and the regression is significantly different at the other flow rate. ............................................................................................................................... 105 2  2  Figure 4.8: Normalized vertical distributions of velocity fluctuations in the three dimensions (■ <u’ >/Ux , 2  2  2  2  ▲ <v’ >/Ux , ● <w’ >/Ux ) from closed and open mats D and B, respectively, at HF and LF. ......... 107 2  Figure 4.9: Normalized vertical distributions of Reynolds stress (♦ τ /ρU ) from closed and open mats D Re  x  and B, respectively, at HF and LF. ...................................................................................................... 108 Figure 5.1: Schematic illustrating all possible combinations of experimental factors ................................ 123 Figure 5.2: Streamwise velocity (u) as a function of log-normalized height above the bed (ln(z)) for reference experiments with and without trays. .................................................................................... 127 Figure 5.3: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for reference experiments with and without trays. ......... 129 Figure 5.4: Percentage of fine particles (<125 µm) in bulk bed layers from three sections of the flume beds of reference experiments. ..................................................................................................................... 131 Figure 5.5: Periphyton biomass (AFDM) from surface samples at growth periods of 0-24 weeks for diatom and algal assemblages. ......................................................................................................................... 133 Figure 5.6: Photographs of periphyton growth stages on a coarse particle substrate (a), showing the changes in structure that occur as periphyton develop and the differences between diatom- and algal-dominated assemblages. Low-profile diatoms at b) four and c) 24 weeks of growth; filamentous green algae at d) 4, e) 12, and f) 16 weeks of growth. .................................................................................................... 135 Figure 5.7: Velocity as a function of ln(z) for periphyton and reference condition experiments. .............. 138 Figure 5.8: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for diatom, algae and reference experiments. ................ 140 Figure 5.9: Decay rates (k) for the 0-1h and 1-8h time periods, near-bed (τRe0) and maximum (τReM) Reynolds stress, and AFDM densities for a) diatom experiments and b) algae experiments compared to the reference condition. ....................................................................................................................... 142 Figure 5.10: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for periphyton and reference experiments from 0-8h. .......................... 144 Figure 5.11: Decay rates (k) of four particle size classes (~2-4, 4-16, 16-63, and 63-122 µm) and the total suspension (black bars) from a) 0-1h and b) 1-8h for diatom experiments and the reference condition. ............................................................................................................................................................. 145 Figure 5.12: Ash-mass of diatom surface samples plotted against diatom AFDM and growth stage. ........ 146 Figure 5.13: Percentage of fine particles (<125 µm) in bulk bed layers of three sections of the flume bed from diatom experiments compared to the reference condition. .......................................................... 147  ix  Figure 5.14: Regression lines of velocity profiles from algae and reference condition experiments.......... 149 Figure 5.15: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for algae experiments. .................................................... 152 Figure 5.16: Decay rates (k) for the 0-1h and 1-8h time periods, near-bed (τRe0) and maximum (τReM) Reynolds stress, and AFDM densities for algae experiments compared to the reference condition.... 152 Figure 5.17: Decay rates (k) of four particle size classes (~2-4, 4-16, 16-63, and 63-122 µm) and the reference condition (black bars) from 0-1h and 1-8h for algae experiments and the reference condition. ............................................................................................................................................................. 153 Figure 5.18: Ash-mass of algae surface samples plotted against diatom AFDM (a) and growth stage (b). 154 Figure 5.19: Percentage of fine particles (<125 µm) in bulk bed layers of three sections of the flume bed from algae experiments compared to the reference condition. ............................................................ 155 Figure 5.20: Relation between periphyton biomass (AFDM) and inorganic content (AM) of surface samples from field studies with a range of locations, stream types, hydraulic conditions, periphyton types, substrates, and sediment regimes (Table 5.7), compared to experimental data and field samples of green algae collected from Hope Slough, British Columbia................................................................ 157  x  LIST OF SYMBOLS AND ABBREVIATIONS AFDM  ash-free dry mass; characterizes the areal density of organic matter in periphyton samples, g m-2  AM  ash mass; characterizes the areal density of inorganic matter in periphyton samples; a measure of particle deposition, g m-2  AFDM/h  volumetric density of periphyton samples, mg cm-3  BE  bed elevation above flume bottom of pebble substrate, cm  BEeff  bed elevation detected by the ADV probe; the difference between the measured water surface elevation (WSE) and the height detected by the ADV probe when positioned at the water surface; equal to the height of the measurement farthest from the bed plus the 5 cm sampling volume of the ADV, cm.  BEp  bed elevation at the top of the periphyton mat, cm  β  depth-averaged ratio of sediment to momentum diffusivity, assumed equal to 1  δ  boundary layer thickness, cm  C  total volume concentration measured by the LISST, µl/l or ppm  c*  normalized steady-state concentration  Ca  near-bed volume concentration measured by the LISST, µl/l or ppm  C0  initial total volume concentration determined from the intercept of regression lines for to exponential models fit to the decline in water column concentration over time, µl/l or ppm  Cz  volume concentration at height z measured by the LISST, µl/l or ppm  DM  median particle size measured by the LISST, µm  DR  bed grain ratio; the ratio of bed grain size to suspended particle size (Fries and Towbridge, 2003)  Ed  enhancement factor; ratio of measured depositional velocity to still-water settling velocity (Fries and Towbridge, 2003)  D  cluster height, cm  Di  the ith percentile of the particle size distribution, mm  h  height or thickness of periphyton mats, cm  h/D  plant-to-roughness height ratio  H  flow depth, cm  H/h  relative submergence of periphyton mats  z0  roughness length, cm  k  decay rate from exponential model fit to the decrease in concentration over time, s-1  κ  von Karman’s constant (0.40)  K  dispersion rate (the sum of the components of molecular diffusion and turbulence, which are proportional to kinematic viscosity (M) and eddy viscosity (l2 (z) du/dz), respectively)  xi  l ( z ) = κz 1 − z / h )  l(z)  Prandtl’s mixing length, assumed  R*  grain Reynolds number; a measure of nearbed roughness; describes the nature of nearbed flows; R* = z0u*/v, using a roughness length z0 = 0.25D84 = 0.008 m, shear velocity u* = τ 0 / ρ m/s, and fluid viscosity, v, of 1.0 x 10-6 m2 s-1 for water at 20 ºC (Dade et al., 1991)  Q  discharge, m3 s-1  L  length of the wetted surface where the boundary layer takes place, cm  ρ  density of water, (assumed equal to 1000 kg m-3)  ŝ  Rouse number, s =  τRe  Reynolds shear stress; τ Re = − ρ < u ' w' > ; also indicated as ‘Re’; τRe0 is the value  ^  v fall  βκu*  calculated from the near-bed velocity measurement, Pa u*  shear velocity, cm s-1  u0  nearbed streamwise velocity, cm s-1  Ux  depth-integrated streamwise velocity, cm s-1  umax  maximum streamwise velocity, cm s-1  u’  streamwise velocity fluctuation, cm s-1  <u’w’>  covariance of the streamwise and vertical velocity fluctuations, cm2 s-2  < u' 2 >  root-mean-square of the streamwise velocity fluctuation, cm s-1 lateral velocity fluctuation, cm s-1  v’ < v' 2 >  root-mean-square of the lateral velocity fluctuation, cm s-1  vdep  particle deposition velocity, equivalent to ws, cm h-1 or cm s-1  vfall  particle fall velocity, equivalent to still-water settling velocity, ws, cm h-1 or cm s-1  w’  vertical velocity fluctuation, equivalent to uz, cm s-1  < w' 2 > wd  root-mean-square of the vertical velocity fluctuation, cm s-1 particle deposition velocity, equivalent to vdep, calculated from decay rate (k) and flow depth (H) as wd = Hk , cm h-1 or cm s-1  WSE  water surface elevation above flume bottom, cm  ws  still-water settling velocity, equivalent to particle fall velocity, vfall, calculated as  ws =  RgD 2 (Ferguson and Church, 2004), where R is the particle’s C1v + (0.75C2 RgD 3 ) 0.5  submerged specific gravity (  ρ p − ρw ; where ρp and ρw are particle and water density, ρw  respectively), g is gravitational acceleration, D is mean particle diameter, and v is the  xii  kinematic viscosity of water (1.0 x10-6 kg m-1 s-2 for water at 20 ºC); C1 and C2 are constants with values of 18 and 1.0 for typical natural sands , cm h-1 or cm s-1 z  height above bed of individual velocity or concentration measurement, cm  za  reference height za (normally taken at the top of the bed, in this case the lowest measurement point is used), cm  xiii  ACKNOWLEDGEMENTS Funding for this research came from a Natural Sciences and Engineering Research Council of Canada (NSERC) research grant (M. Hassan), an Isaak Walton Killam Predoctoral Scholarship (N. Salant), and a Geological Society of American Student Research Grant (N. Salant). A NSERC Research Tools and Instruments (RTI) grant (M. Church) provided funds for the LISST-100X Particle Size Analyser, a major piece of equipment used in this study. Many individuals generously gave their time and expertise to help with research, analysis, and manuscript preparation. I would like to thank Violeta Martin in the Department of Civil Engineering at the University of British Columbia for her help with experimental design, ADV use, flume operation, and data analysis, as well as her ever-present support. Thanks also: to Craig Layne in the Department of Biology at Dartmouth College for his help identifying periphyton assemblages; to Andre Zimmermann, Josh Caulkins, Tony Lagemaat, Melissa Ewan, Joao Sarmiento, Ilana Klinghoffer, and Tim Argast for advice and physical assistance with flume construction, sediment preparation, periphyton collection, sieving, and lab upkeep; to Dean Blinn and Walter Dodds for advice on periphyton cultivation and analysis; to Chuck Pottsmith and Doug Keir at Sequoia Scientific for advice and assistance regarding LISST operation; to Harald Schremp, Bill Leung, Doug Smith and all staff members of the U.B.C. Department of Civil Engineering Workshop for their advice and assistance with the construction and maintenance of lab equipment; and to Vladimir Nikora, Aaron Packman and the members of my thesis committee (M. Hassan, M. Church, D. Moore, and J. Richardson) for thoroughly reviewing several drafts of this manuscript or portions thereof and providing many helpful revisions.  xiv  CO-AUTHORSHIP STATEMENT One section and one chapter of this thesis were co-authored. Section 1.2 is a shortened version of a published literature review, invited for contribution to the book Coastal Watershed Management (2008) and co-written with Marwan Hassan. For Chapter 5, Marwan Hassan helped with sample collection, data interpretation and writing. I designed the experiments, conducted all laboratory tests, processed and analysed the data, wrote the first complete draft of the manuscript, and created all figures and tables.  xv  CHAPTER 1: INTRODUCTION 1.1 STATEMENT OF THE PROBLEM  Fine particulate matter is an important component of many physical and biological processes in streams. Particles deposited on the streambed surface or within the interstices of the bed interact with benthic organisms and bed topography, whereas suspended particles impact free-living organisms, water quality, and water chemistry. Particle dynamics at the streambed and in the water column may differ for inorganic and organic material due to the effects of particle density, shape, size, surface charge, and nutritional quality. For example, the deposition and infiltration of fine inorganic sediment has been repeatedly shown to degrade benthic habitat for fish and other organisms (Hynes, 1970 ; Waters, 1995). In contrast, fine organic particles are a significant source of carbon for benthic organisms (Webster et al., 1987). In the water column, high concentrations of both inorganic and organic particles can harm the feeding habits of free-swimming consumers such as filter-feeding invertebrates and fish (Anderson et al., 1996; Runde and Hellenthal, 2000) and degrade water quality (Reiser, 1998). For example, fine organic particles can play a large role in the movement and deposition of hydrophobic contaminants (Karichkhoff, 1984). However, at moderate concentrations, organic particles transported in suspension provide energy and nutrients to free-swimming organisms; similarly, exports of fine organic particles from forested headwater reaches can provide a significant energy subsidy to downstream biotic communities (Karlsson et al., 2005). Although the impacts of fine particle deposition have been well-documented, few studies have investigated the factors controlling deposition rates or the degree of particle infiltration. In fact, fine particle movement is influenced by a variety of physical and biological processes and mechanisms. Physical mechanisms include advective forces (Dade et al., 1991) and upward flow velocities from eddies generated by an irregular boundary (Minshall et al., 2000; Wanner and Pusch, 2000; Paul and Hall, 2002; Georgian et al., 2003), direct trapping by physical structures, particle flocculation, and the downward force of gravity. Biological mechanisms include adhesion to biofilm (Lock, 1981; Battin et al., 2003), invertebrate manipulation (Wallace et al., 1991), and removal by filter-feeders (Mccullough et al., 1979; Georgian and Thorp, 1992; Monaghan et al., 2001). Which factors dominate may also be influenced by particle composition; it has been suggested that rapidly depositing particles with low settling velocities may be propelled by factors other than gravity (Thomas et al. 2001), such as advective transport, filtration at the sediment-water interface (Hoyal et al., 1997) or biofilm adhesion. In contrast, particles that deposit more slowly than their settling velocities may be influenced by upward turbulent mixing and entrainment factors (e.g. Paul and Hall, 2002). As a consequence of our limited understanding and the complexity of these processes, current models for predicting the deposition and water column distribution of particles are constrained by assumptions and conditions unrealistic for most natural systems. In particular, no available model provides an explicit representation of channel and streambed morphology, benthic ecology, or particle composition. Qualitative theories of particle deposition from turbulent water describe the instantaneous rate of change in concentration as  1  solely a function of the particle still-water settling velocity, particle concentration, and the depth of the mixed layer (e.g.Einstein, 1968) thus ignoring conditions of the flow or the streambed. Traditional hydrodynamic models of vertical particle distribution (e.g. Rouse equation, advection-dispersion model) are limited because they do not consider the small-scale mechanics of particle motion, particularly the effect of particle-water interactions on turbulent flows and particle movement. The Local Exchange Model (LEM) (McNair et al. 1997) improves upon traditional models by describing the movement of an individual particle due to turbulence as a stochastic-diffusion process that includes a depth-dependent measure of vertical dispersion. Nevertheless, the LEM is still limited because it does not consider the effect of particle concentration, bed topography, or biological condition on particle movement or turbulent flows. No theoretical model exists that fully encapsulates the numerous factors influencing particle exchange. A conceptual model of particle movement that includes physical and biological mechanisms is needed as a framework on which to base predictions and field sampling schemes. As a starting point to developing this model, we require more accurate measurements of the deposition rate, water column distribution and infiltration of fine particles under a range of hydraulic and surface conditions. A flume-based, experimental approach allows for controlled conditions and detailed measurements at scales relevant to the processes being investigated  1.2 CONCEPTUAL HYPOTHESES  In this study, ‘fine particles’ are defined as those <125 µm, comprising clay, silt, very fine sand and fine organic particles, which typically make up the wash load of gravel-bed streams (Wentworth, 1922; Church, 2006). Five factors and their relation to fine particle movement and flow hydraulics are considered: 1) the fraction of sand-sized material (<2 mm) in the bed, 2) flow rate, 3) particle density, 4) initial particle concentration, and 5) periphyton1 structure and density on the streambed surface. Five conceptual hypotheses each related to a single factor are tested in this study: Hypothesis 1: Bed sand fraction will influence the vertical distribution, deposition, and infiltration of fine particles by its effect on hydraulic conditions, particle trapping, and interstitial pore space. For high sand fractions, a smoother hydrodynamic surface may produce roughly logarithmic velocity profiles; corresponding concentration profiles are expected to be well-predicted by theoretical models, which typically assume a logarithmic velocity profile over a hydrodynamically smooth bed. In contrast, velocity profiles over a coarse bed are expected to be more linear than logarithmic (Bridge, 2003) thus concentrations high in the water column may be larger than predicted by theoretical models. As a result, measured  1  ‘Periphyton’ is used in this text interchangeably with ‘biofilm’, as in the aquatic literature, although these terms are different in theory. Biofilm is defined generally as a structured community of microorganisms contained in a self-developed matrix, adhering to either a living or inert surface. Periphyton refers more specifically to microscopic plants attached to submerged surfaces in most aquatic ecosystems; thus periphyton does not technically include bacteria, protists, or associated detritus and excretions. Nevertheless, this term is frequently used in reference to a broad microbial assemblage attached to aquatic surfaces, including algae, bacteria, heterotrophs, detritus, and excretions.  2  concentration profiles over a coarse bed are expected to be more uniform than predicted profiles. For high sand conditions, turbulence intensity in the near-bed region should be reduced relative to coarser, higher roughness substrates thus surface deposition may be enhanced. However, fewer protruding particles may reduce the amount of particle trapping thus slowing the rate at which particles are lost from the water column. Furthermore, large amounts of sand will clog interstitial spaces, reducing the amount and depth of additional fine particle infiltration. In contrast, surface deposition and infiltration into a coarse bed with a low sand fraction may be enhanced by particle trapping and large pore spaces, but reduced by an increase in turbulence and upward fluid motion. Overall, total deposition (including surface deposition and subsurface infiltration) to a high sand bed is expected to be slower than to a coarse bed. Hypothesis 2: Flow rate will affect the vertical distribution, deposition, and infiltration of fine particles through its influence on local velocities and turbulent forces. Low flow rates will have slower velocities and lower shear stresses, allowing for the deposition of coarser particles and resulting in a general fining of particle sizes with height above the bed. Surface deposition will be enhanced by the reduction in turbulent forces and thus upward fluid motion, but the depth and amount of particle infiltration into the bed will be less due to reduced vertical turbulent transport. Conditions should reverse for high flow rates, which will more uniformly distribute particles of different sizes and increase particle mixing, trapping, and advection. Hypothesis 3: Particle density will affect the vertical distribution, deposition, and infiltration of fine particles by influencing the rate of downward movement of particles under the force of gravity. Low-density particles on the streambed surface will be more readily entrained and transported higher in the water column than high-density particles thus concentration profiles of low-density particles will be more uniform than profiles of high-density particles under all bed and flow conditions. High-density particles will deposit in greater amounts at the surface and move to greater subsurface depths than will low-density particles of equal size and shape. Hypothesis 4: Initial particle concentration will influence the deposition and infiltration of fine particles by changing bed-particle interactions, particle-particle interactions, the number of particles available to deposit, and turbulent forces. At high concentrations, particle-bed interactions will be more frequent and depositing coarse particles may sweep finer particles towards the bed thus deposition from the water column will be faster than at low or moderate concentrations. Furthermore, turbulent energy of the flow will be reduced by the transport of large numbers of particles, in turn lower shear stresses and enhancing surface deposition. High water column concentrations will ultimately result in greater numbers of particles entering the bed, increasing infiltration amounts, although the rate of particle infiltration may not be linearly related to the increase in particle numbers. Hypothesis 5: Surface periphyton will influence the deposition and infiltration of fine particles via surface adhesion, physical trapping, bed clogging, and changes to near-bed hydraulics, but the direction and magnitude of effects will depend on periphyton structure and density.  3  Highly developed or filamentous assemblages can increase friction and near-bed turbulence, potentially reducing the amount of surface deposition. However, filamentous structures can also directly capture more particles than lower profile forms, increasing deposition rates. Diatoms produce a ‘sticky’ exopolysaccharide (EPS) matrix that has been shown to enhance particle deposition (Battin et al., 2003). High densities of periphyton could also smooth the boundary, reducing surface roughness and turbulence intensities and enhancing deposition, similar to the case of the high sand bed. However, also similar to the high sand condition, dense, cohesive assemblages may clog pore spaces, in turn reducing particle deposition and infiltration. Thus the rate at which particles are lost from the water column will depend on the balance of periphyton density and structure. 1.3 LITERATURE REVIEW2  Design and analysis of this study was based on an extensive literature review of fine particle dynamics in streams that addressed four main components of fine particle dynamics: sources and supply mechanisms; instream transport and deposition; biological impacts; and spatial and temporal scales of study and variability. The following sections present a review of two topics considered most relevant to the present study: 1) instream transport and deposition and 2) models of particle distribution and deposition.  1.3.1 FINE PARTICLE TRANSPORT AND DEPOSITION IN STREAMS  Upon entering the fluvial system, fine particles may be transported downstream or deposited on the surface of the streambed. Once deposited, these particles may be retained, accumulating or infiltrating into the bed, consumed, or entrained back into the water column for further transport. A number of physical and biological factors determine the fate of fine particles, including suspended particle concentrations (Muste et al., 2005), particle size and composition (Kranck, 1979; Droppo et al., 1997), bed composition (Minshall et al., 2000), channel morphology (Diplas and Parker, 1985; Maridet et al., 1995), benthic ecology (Georgian and Thorp, 1992), and discharge (Broekhuizen and Quinn, 1998; Lenzi and Marchi, 2000). For decades, both physical and ecological researchers have quantified and modeled particle transport and deposition, greatly enhancing our understanding of particle movement and storage. A high degree of variability and uncertainty among results, however, arises from the inherent complexity of factors that govern particle dynamics. Particle entrainment occurs when turbulent forces mobilise particles from the streambed. In natural streams, fine particle concentrations are generally related to discharge and often exhibit hydrograph hysteresis, a term for cases when concentrations at a given flow on the rising limb are different than at the corresponding flow on the falling limb (Bilby and Likens, 1979; Langlois et al., 2005; Nistor and Church, 2005). Patterns of hysteresis in 2  An version of this section has been published. Salant, N.L. and Hassan, M.A. (2008) Fine particles in streams: physical, ecological, and human connections. In Coastal Watershed Management, eds. A. Fares and A. I. El-Kadi, WIT Press Royal: Southampton, U.K, p 124-181. (Invited contribution)  4  the relation between suspended load and water discharge are related to types and locations of active sources (Lenzi and Marchi, 2000; Nistor and Church, 2005), which can be a function of bed and bank composition. Mixed beds are characterized by a range of particle sizes that can affect particle entrainment by controlling both the supply of particles and the degree of grain-grain interactions. Grain-grain interactions may influence the mobilisation and suspension of particles. How much grain-grain interactions will affect the mobilisation of fine particles, such as fine sand or silt, is still an open question. Small sand grains may infiltrate the pore spaces of the bed surface, moving only when the entire bed is entrained (Diplas and Parker, 1985). Thus the mobilisation and entrainment of fine particles will be limited when their abundance is low and discharge is less than needed to move the coarser fraction. In this case, the mean bed grain size will be larger than the grain size of suspended particles. Alternatively, selective entrainment of fine particles at low flows when their abundance is high may transport these particles higher in the water column such that grain size varies with depth. However, a substrate dominated by very fine particles (silt and clay) can experience compaction and cohesion, increasing the shear stress required to entrain particles and decreasing erosion and suspension rates (Krone, 1976; Fukuda and Lick, 1980). For these reasons, bed composition – the relative amounts of fine and coarse material – may be an important factor governing the entrainment and distribution of fine particles. Once entrained, particles are transported downstream before depositing again. Most simply, the mode, rate, and distance of particle transport is a function of particle size, particle density, and flow velocity. Generally, fine particles such as silts, clays, or fine particulate organic matter are carried in suspension; therefore, under normal flow conditions they are unlikely to frequently interact with the bed or have long residence times within the stream. In contrast, larger particles roll and saltate along the bed as bed load (Knighton, 1998) and have longer residence times. The boundary between these two transport modes is transitional; depending on flow magnitude, medium and coarse sand (0.25 – 2 mm) may move either short distances in suspension or roll along the bed as bed load. Numerous physical models and measurement techniques have been developed to quantify suspended and bed load transport (Einstein and Chien, 1952; Emmett, 1980; Wilcock, 2001; Hassan and Ergenzinger, 2002). Information on the long history of this field is summarized in a comprehensive review by Garcia (2008). Deposition, infiltration and retention of fine particles can occur by several physical mechanisms, including passive percolation, kinematic sieving, passive settling, burial, and vertical hydraulic transfer. On an armoured bed, small sand grains may infiltrate the pore spaces of the bed surface, moving again only when the entire bed is entrained (Diplas and Parker, 1985). Passive percolation into the hyporheic zone sediments varies spatially and temporally, depending on geomorphic and hydraulic conditions of the streambed, leading to the formation of patches of differing exchange capacities and transient storage characteristics (Pusch et al., 1998). Kinematic sieving occurs when the movement of grains within the bed creates openings for the downward movement of smaller particles. Passive settling occurs in regions of low flow, such as pools, eddies or backwaters, created by structural elements within the channel (Mulholland, 1981; Hedin, 1990; Jones and Smock, 1991; Cushing et al., 1993). High sedimentation and low re-suspension rates produce zones of accumulation. Burial by shifting sediments occurs in lowland, sand-bedded rivers when migrating bedforms bury organic particles deposited in  5  the lee of the structure (Savant et al., 1987; Meyer, 1988; Rutherford, 1994). Vertical hydraulic transfer occurs in well-sorted, open frame sediments, where local differences in water pressure produce vertical hydraulic gradients of up- and downwelling water, mixing materials between the surface and interstitial water. Several studies have documented vertical exchange in riffle-pool sequence (Valett et al., 1990; Hendricks, 1993; Hendricks and White, 1995; Jones et al., 1995), where even single boulders can increase vertical interactions (Williams, 1993) and the concentrations of fine particulate organic matter (FPOM) in the hyporheic zones of riffles (Pusch, 1996; Pusch, 1997). Bed roughness elements, whether inorganic or organic, can produce vertical fluxes of water and materials between the stream and bed sediments (Savant et al., 1987; Rutherford, 1994; Huettel et al., 1996), with ensuing effects on nutrient and carbon dynamics.  1.3.2 MODELS OF PARTICLE DISTRIBUTION AND DEPOSITION  In the physical sciences, decades of research have led to the development of theoretical equations for the vertical distribution of suspended particles in water, based primarily on mathematical representations of turbulent flow, eddy diffusion, and particle movement. Early studies in uniformly turbulent tanks known as “turbulence jars” (Hurst, 1929; Rouse, 1939; Dobbins, 1944) confirmed the ability of these equations to predict the vertical distribution of particles under controlled conditions, using several empirically-derived constants and basic assumptions. Recognizing that some calibration may be necessary to define this system, the constants and assumptions required may not be valid under natural conditions. Several such equations exist (Rouse, 1937; Halbronn, 1949; Einstein and Chien, 1952; Bagnold, 1954; Hunt, 1954; Tanaka and Fugimoto, 1958), each attempting to improve those developed before, but all containing the same fundamental model, basic assumptions, and empirical constants. Rouse (1937) developed the original form from a differential equation of suspended sediment that assumes two-dimensional, steady, uniform flow and particles with uniform size, shape, and density. Steady flow implies that the average concentration of particles remains constant, such that the flux of particles in all three dimensions is balanced; in other words, for a given volume, sediment fluxes in and out are equal. Thus particle distribution in the vertical direction represents a balance between the downward settling of particles due to gravity and the upward movement of particles due to turbulent diffusion (O'Brien, 1933):  − Cws = ε p  dC dz  In this formulation, C is the time-averaged concentration, ws is the particle still-water settling velocity, εp is the vertical coefficient for particle turbulent diffusion, and z is height above the bed. Integration of this equation yields an equation for the average concentration of particles at a given height; the equation can be further modified to incorporate particles of different sizes. These derivations are presented in Chapter 4. Similar to the Rouse equation, the advection-dispersion model (Bencala and Walters, 1983; Valett et al., 1996; Webster and Ehrman, 1996) describes solute or particle flux in three spatial dimensions as a balance between the transport driven by particle fall velocity and flow velocity (advection) and that driven by molecular diffusion and turbulence (dispersion). In its basic form, the model is a continuity equation,  6  ∂c + ∇ ⋅ (cu − K∇c) = 0 ∂t where c is particle concentration, t is time, ∇c is the concentration gradient vector, u is the advective velocity vector, and K is the matrix of the longitudinal, transverse, and vertical dispersion coefficients. In this framework, the concentration of an assemblage of particles is modeled as a deterministic diffusion process. Integration of the advection-dispersion model over the flow depth produces an estimate of the average concentration of suspended particles. Once solved for a single dimension (typically the longitudinal, or streamwise direction), the advection-dispersion model can be expanded to include additional terms for the effects of stream processes such as groundwater and tributary inputs, transient storage, or the removal of solutes from the water column via adsorption or plant uptake (i.e. immobilization). Numerous ecological studies have used the advection-dispersion model for the modeling of FPOM transport and deposition in natural streams; in particular, several studies have focused on the relation between particle dynamics and the extent of transient storage (e.g. Paul and Hall, 2002). Several authors have proposed a qualitative theory of particle deposition from turbulent water that describes the instantaneous rate of change in concentration,  dC , as solely a function of the particle still-water dt  settling velocity (ws)3, the particle concentration, and the depth of the mixed layer (usually taken as the depth of flow, H, assuming complete vertical mixing) (e.g. Einstein, 1968). An equation for particle concentration can then be derived from this general model (Smith, 1982; McCave, 1984): −(  C (t ) = C 0 e  ws )t H  As noted above, in coarse-bedded streams with turbulent flow, particle deposition can be influenced by other factors, such as grain interactions and turbulent advection. Thus the actual deposition velocity (wd) may differ from ws. Deposition velocities larger than settling velocities are considered ‘enhanced’ deposition rates, which can be represented by an enhancement factor E d =  wd (Fries and Towbridge, 2003). Values of Ed > 1 ws  indicate the influence of factors that increase the rate of deposition relative to settling (e.g. trapping, advection) whereas values of Ed < 1 indicate mechanisms that slow the settling of particles (e.g. entrainment, upward directed turbulent velocity component). None of these models considers the small-scale mechanics of suspended particles, assuming instead that particles and water act as a single phase or mixture. As a result, predictions based on these assumptions may be of limited validity because interactions between water and particles can influence the dynamics of turbulent  3  In the ecological literature, the rate at which particles are lost from the water column, or depositional velocity, is symbolized as vdep. Notation for still-water settling velocity, or particle fall velocity, is vfall. However, the hydraulic and sedimentological literature uses the w-v-u convention, signifying movement in the vertical, transverse, and streamwise directions, respectively. For consistency, in this text I will use the w-v-u convention throughout. Thus wd and ws are used for depositional and still-water settling velocity, respectively, regardless of the original terminology used by the papers I review.  7  flows and streamwise velocities. Several studies have demonstrated that because of these interactions the assumptions necessary to these models are questionable (Squires and Eaton, 1990; Elghobashi, 1994; Kaftori et al., 1998; Rouson and Eaton, 2001; Muste et al., 2005). Image-based techniques that separately quantify particle and water movement (Muste et al., 2005) provide evidence that most of the key assumptions in the models do not hold. In essence, these models do not take into account particle-particle interactions, particle-water interactions, the depth-dependent nature of model parameters, or the influence of streambed morphology. For example, in sediment-laden flows, interactions between sediment and fluid particles can alter velocity and concentration profiles. Numerous single- and two-phase models have been proposed for the prediction of velocity and concentration profiles in sediment-laden flows. In the two-phase framework, interactions and differences between particle and fluid movement are considered; mechanisms of particle-fluid interactions are incorporated through modification of the advection-dispersion framework. For example, Cao et al. (1995) derived a diffusion equation for particle concentration that accounts for the downward mass flux that arises from the inequality between water and particle density. Cao et al. (1995) show that the mixture’s time-averaged vertical velocity, wd, is related to concentration, C, by the following equation: wd = −  ρ p − ρw ρw  ws C  where ρp and ρw are the particle and fluid density, respectively. Other models attempt to incorporate the effects of drift flux and density stratification; these effects are described in more detail in Chapter 4. Results from ecological field studies demonstrate the limited application of the advection-dispersion framework. In theory, the measured wd should be approximately equal to ws, but most field studies show no consistent relation between these two parameters (Cushing et al., 1993; Hall et al., 1996; Minshall et al., 2000; Thomas et al., 2001). The ratio of wd to ws in these studies ranges from 0.04 to 1690, depending on particle properties and stream characteristics, but generally increases with decreasing particle size and density. Fluorescently labelled bacteria have the highest ratio (1690) (Hall et al., 1996) while natural FPOM and corn pollen are much lower (0.04-0.56). For cases in which wd is << ws (Minshall et al., 2000; Wanner and Pusch, 2000; Paul and Hall, 2002), it is proposed that turbulent mixing and re-suspension factors overwhelm gravitational settling in controlling particle movement. This is represented in theoretical models that predict wd < ws when local shear stresses exceed a critical threshold for re-suspension (Einstein and Krone, 1961). The ratio of gravitational velocity to turbulent mixing velocity (due to bed shear) is expressed by the Rouse number (ŝ). There is little effect of gravity at values of ŝ < 0.1 but gravitational factors increasingly dominate as ŝ approaches 1 (McNair, 2000; McNair and Newbold, 2001). Georgian (2003) calculated ŝ << 0.01 and low wd/ws for both FPOM and pollen in the field and ŝ approaching 0.1 for FPOM in a flume, demonstrating that shear forces were more important than gravity in the field and less important in a flume. A large wd may occur in situations where energy at the streambed is dissipated in slow flowing pools or stagnant zones, therefore decreasing bed shear stress and turbulent resuspension. Minshall et al. (2000) reported wd ~ ws in small streams with high bed complexity, while Georgian et al. (2003) reported that wd in a smooth-bedded flume is less than wd of the same particles in the field, both suggesting that bed complexity and its effect on turbulence may strongly influence the deposition of particles.  8  Particles with very low settling velocities and larger depositional velocities (i.e. bacteria) may also be deposited via advective transport into interstitial spaces, hyporheic entrainment, or adhesion to the substrate. Traditional models explain the discrepancy between wd and ws by the hydrodynamic/gravitational mechanisms described above, but so far, measurements of bed roughness and shear stress have not revealed a consistent or significant relation with wd (Minshall et al., 2000). Furthermore, conflicting and limited results have not fully elucidated the mechanistic role of transient storage zones in FPOM deposition. For example, Minshall et al. (2000) reported that wd was positively correlated with the relative size of transient storage zones and a coefficient of transient storage exchange, but not related to the advective exchange of water into these zones. Paul and Hall (2002) reported no relation between wd and any measure of transient storage or exchange. Newbold et al. (2005) measured brief retention of FPOM at a rate similar to that of water retention in transient storage zones, indicating simple advective transport to and from these zones without deposition. Based on earlier findings that hyporheic exchange increases particle removal from the water column (Packman and Mackay, 2003), Newbold et al. (2005) proposed that most transient storage may not occur in hyporheic zones, but rather in deep lateral areas where turbulence is high enough to keep particles suspended. Differences between studies may be due to the size and composition of particles, the relative amount of in-channel versus hyporheic zone storage, or biological properties of the streambed. Newbold et al. (2005) also calculated that average residence times of deposited particles were longer than turbulent fluctuations, suggesting the influence of biological retention processes. Evidence for the role of biological or bio-physical mechanisms in particle retention is limited, but suggestive. For example, periphyton adhesion is proposed as one explanation for the discrepancy between wd and ws (Lock, 1981; Battin et al., 2003), while invertebrate manipulation (Wallace et al., 1991) or removal by filter-feeders may account for a measurable proportion of total deposition (Mccullough et al., 1979; Georgian and Thorp, 1992; Monaghan et al., 2001). Particle composition may also explain much of the observed discrepancy, an explanation that has received increasing attention in the past decade. As noted above, most suspended particles are composites of organic and inorganic components. Composite particles may enter the stream from the watershed as aggregates, retaining their structure during transport, or may form in-channel via physical, biological, or chemical flocculation processes (Petticrew, 1996; Droppo et al., 1998; Woodward et al., 2002). Cohesive properties of organic matter tend to enhance flocculation (Kranck, 1979; Droppo et al., 1997); however some evidence exists for electrochemical flocculation in glacial meltwaters that lack organic matter (Woodward et al., 2002). Particle size and hydrodynamic properties of composite particles differ considerably from their mineral components. Thus, predictions of deposition rates assuming single grain settling are likely to be inaccurate when composite particles predominate (Ongley et al., 1992; Petticrew, 1996; Droppo et al., 1998). McNair et al. (1997) propose an alternative approach to modeling suspended particle dynamics that considers the behaviour of an individual particle, rather than an assemblage of particles. Like the advectiondispersion model, the focus and intended application of the McNair et al. (1997) approach is FPOM. The authors argue that theoretical models developed for suspended inorganic particle dynamics are inadequate for ecological applications because of their focus on inorganic particles with high fall velocities, passive  9  gravitational settling, and physical forces, as well as an interest in large-scale processes. In contrast, ecological studies emphasize the biological significance of particles that often vary in shape, composition and density and may be deposited and mobilized via behavioural means. McNair et al. (1997) describe the process of fine particle transport as including four key components: the attachment problem, the entrainment problem, the hitting-time problem, and the hitting-distance problem. The attachment and entrainment problems address how a particle at the bed-water interface becomes fixed to the substrate and its residence time before re-suspension into the water column. The hitting-time/distance problems consider the temporal and spatial dimensions of longitudinal transport, namely, how long a particle remains in the water column and the distance it travels. Time spent in the water column may be relevant to consumers, while travel distance is important to determining the rate of downstream dispersal. Fine particles in turbulent water move along irregular trajectories, buffeted up and down by fluid eddies thus particle vertical movement and elevation may be considered a stochastic process (Denny and Shibata, 1989; McNair et al., 1997). McNair et al. (1997) provided a discrete representation of this stochastic process by considering the motion of a neutrally buoyant, non-motile particle as occurring in two ways: particles can be propelled by molecular collisions or may be incidentally carried by the turbulent transport of water. Because of the complex, non-linear structure of turbulent fluid motion, a simplified approximation, known generally as a stochastic-diffusion process, is used to model turbulent transport. All stochastic-diffusion processes are defined by forward and backward Kolmogorov equations, which include the infinitesimal mean and variance functions. Once specified, these functions convert the abstract stochastic-diffusion process into a meaningful description of the process of interest, i.e. the dynamics of particle motion. Assuming steady-state suspended sediment concentrations, zero lateral velocity, constant longitudinal velocity, and a laterally and longitudinally homogenous vertical concentration profile, McNair et al. (1997) derive an expression for the steady-state vertical distribution of particles:  d  dc *  uz c * −K =0  dz  dz  where c* is the normalized steady-state concentration, K is the turbulent dispersion coefficient (the sum of the components of molecular diffusion and turbulence, which are proportional to kinematic viscosity (M) and eddy viscosity (l2 (z) du/dz), respectively), uz is the vertical component of the advective velocity, and z is elevation above the bed. For more detailed explanation of the background and equations of the Local Exchange Model, see McNair et al. (McNair et al., 1997; McNair, 2000) and McNair and Newbold (2001). In this study, four models will be tested for their ability to describe the vertical distribution of fine particles in the water column. All four models can be applied to particles of any size, but were designed for application to particles larger than those tested in this study (i.e. sand-sized particles 0.125-2.0 mm). For this reason, this assessment does not provide a critical test of these models, but it does test the applicability of these theories to finer particles and the effect of varying physical conditions on model performance. The four models tested are: 1) the Rouse equation, 2) a modified Rouse equation that incorporates particle size distribution, 3) the Cao et al. (1995) model for sediment-laden flows, and 4) the Local Exchange Model. Particles used in this study are non-  10  motile and of a mineral composition thus the Rouse equation may perform as well, if not better, than the Local Exchange Model, which is specifically designed for application to organic particles and motile organisms. Similarly, the concentrations tested in this study are relatively dilute compared to typical sediment-laden flows thus the Cao et al. (1995) model may underperform.  1.4 OPERATIONAL HYPOTHESES AND APPROACH  Review of the literature demonstrates that a range of physical and biological factors influence the movement of fine particles in streams. In this study, I focus on five factors: flow rate, bed composition, particle density, initial concentration, and periphyton structure and density. My study aims to elucidate the direct and indirect mechanisms by which these factors alter the deposition rates, infiltration amounts, and water column distributions of fine particles. Operational hypotheses address the conceptual hypotheses presented in section 1.1: 1.  Varying the fraction of sand (< 2 mm) in the streambed at a single flow rate will alter local hydraulic conditions and the water column distribution, deposition, and infiltration of high-density particles (cf. Hypothesis 1).  2.  Changing the average flow rate will alter local hydraulic conditions and the water column distribution, deposition, and infiltration of high-density particles to a coarse bed (cf. Hypothesis 2).  3.  Varying the particle density of a suspension at a single flow rate will alter hydraulic conditions and the water column distribution, deposition, and infiltration of particles to a coarse bed (cf. Hypothesis 3).  4.  Varying the initial particle concentration at a single flow rate will alter local hydraulic conditions and the water column distribution, deposition, and infiltration of high-density particles to a coarse bed (cf. Hypothesis 4).  5.  Periphyton structure and density at the patch scale will influence local flow hydraulics of coarsegrained clusters (cf. Hypothesis 5).  6.  Periphyton structure, density, and areal coverage on a coarse bed will influence local hydraulic conditions, particle deposition, and particle infiltration (cf. Hypothesis 5).  As presented, this approach tests each factor independently, comparing a change in each of the other factors to the reference condition in an ‘incomplete’ experimental design (Figure 1.1). The reference condition is defined by a coarse pebble-gravel-sand bed (~20% sand), a relatively high flow rate (~8.5 L/s), high density particles (silica), a moderate initial particle concentration (~100 ppm), and a bed surface without periphyton growth – comparable to a narrow, shallow, low-gradient, coarse-bedded stream with limited sediment supply. Given the available time and resources, the number of factors being tested, and the importance of replication, this approach was the most feasible. Although an incomplete design does not account for the fact that some factors may interact, there is little evidence to suggest that the interaction of factors will be significant in this study. Only two possible effects merit mention: 1) a high flow rate can alter bed composition by mobilising fine particles and coarsening the bed surface and 2) high concentrations of particles can alter hydraulic conditions. I  11  attempted to avoid the former effect by choosing a reference flow rate with shear stresses insufficient to mobilise sand-sized particles (0.25-2 mm); however, I did experience complications related to this effect that are discussed below. For the latter, although suspended particles have been shown to enhance or attenuate turbulence intensities (Rashidi et al., 1990; Rogers and Eaton, 1991; Yarin and Hetsroni, 1994; Muste et al., 2005) and modify vertical velocities (Muste et al., 2005), these are considered ‘local’ hydraulic parameters; high concentrations should not affect the average flow rate.  Figure 1.1: Schematic illustrating all possible combinations of experimental factors, representing a complete experimental design; shaded boxes are the combinations tested in this study. Flow Rate  Par  ticl eD ens  it y  H  L  it y ticl eD ens  Flow Rate  L  H Initial M Concentration  H  L  L  a) 20% Sand Bed  Par  H  b) 80% Sand Bed  H Initial M Concentration  H  L  L  c) Diatoms at 4, 8, 12, and 24 weeks growth  d) Algae at 4, 8, 10, 16, and 20 weeks growth  *Notes: Factors include particle density (high or low), flow rate (high or low), initial concentration (high, medium, or low) and bed composition or periphyton coverage (20% sand, 80% sand, algae at one of 5 growth stages, or diatoms at one of 4 growth stages). Box with diagonal lines is the reference condition. Operational hypotheses 1 through 4 were addressed using data from a series of flume experiments that test how changing the flow rate, bed composition, particle density, and initial concentration from the reference condition influences the deposition and infiltration of fine particles. Results from these ‘physical’ experiments are described in Chapter 2. In the original experimental design, I planned to conduct and compare experiments with a bed of 20% sand to those with 80% sand, keeping all other conditions constant and replicating each three  12  times. However, difficulties controlling bed movement in the 80% sand experiments forced me to abandon this approach; because of changes to bed configuration and non-uniform flow conditions, results from the 80% sand experiments are not conclusive. Nevertheless, suggestive results are presented and discussed in Chapter 2. In Chapter 3 I then apply the data from the physical experiments to assess and compare the performance of theoretical models of vertical particle distribution. To address operational hypothesis 5, I conducted experiments on the same 20% sand bed and at the same flow rates as those used in the physical study, but no suspended particles were present and only hydraulic conditions were measured. I tested how the presence of filamentous periphyton patches on the sedimentary bed altered local flow conditions in relation to patch structure and how the patch-hydraulic relation varied with bulk flow rate; results are presented in Chapter 4. Finally, a series of experiments identical in design and using the same reference condition as the physical experiments expanded upon the periphyton patch study by incorporating particle suspension, a wider range of periphyton densities (small patches to uniform coverage), and additional periphyton structures (several growth stages of a mucilaginous diatom assemblage). Results and analysis from these experiments are presented in Chapter 5.  13  1.5 REFERENCES Anderson P. G., Taylor B. R., and Balch G. C. (1996) Quantifying the effects of sediment release on fish and their habitats. In Canadian Manuscript Report of Fisheries and Aquatic Sciences No. 2346. Bagnold R. A. (1954) Experiments on a gravity free dispersion of large solid spheres in a Newtonian fluid under shear. In Proceedings, Royal Society of London, Series A, Vol. 225, pp. 49-70. Battin T. J., Kaplan L. A., Newbold J. D., and Hansen C. M. E. (2003) Contributions of microbial biofilms to ecosystem processes in stream mesocosms. Nature 426(6965), 439-442. Bencala K. E. and Walters R. A. (1983) Simulation of solute transport in a mountain pool-and-riffle stream: a transient storage model. Water Resources Research 19, 718-724. Bilby R. E. and Likens G. E. (1979) Effect of hydrologic fluctuations on the transport of fine particulate organic-carbon in a small stream. Limnology and Oceanography 24(1), 69-75. Bridge J. S. (2003) Rivers and floodplains. Blackwell. Broekhuizen N. and Quinn J. M. (1998) Influences of stream size and catchment land-use on fine particulate organic matter retention in streams. New Zealand Journal of Marine and Freshwater Research 32(4), 581-590. Cao Z., Wei L., and Xie J. (1995) Sediment-laden flow in open channels from twophase flow viewpoint. Journal of Hydraulic Engineering 121(10), 725-735. Church M. (2006) Bed material transport and the morphology of alluvial river channels. Annual Review of Earth and Planetary Sciences 34, 325-354. Cushing C. E., Minshall G. W., and Newbold J. D. (1993) Transport dynamics of fine particulate organic-matter in two Idaho streams. Limnology and Oceanography 38(6), 1101-1115. Dade W. B., Nowell A. R. M., Jumars P.A. (1991) Mass arrival mechanisms and clay deposition at the seafloor, In Microstructure of finegrained sediment: From mud to shales (eds. R. H. Bennett, W. R. Bryant, and M. H. Hulbert), p. 161-165, Springer-Verlag. Denny M. W. and Shibata M. F. (1989) Consequences of surf-zone turbulence for settlement and external fertilization. American Naturalist 134, 859-889. Diplas P. and Parker G. (1985) Pollution of gravel spawning grounds due to fine sediment. In Proj. Rep. 240. St. Anthony Falls Laboratory, Univ. of Minn. Dobbins W. E. (1944) Effect of turbulence on sedimentation. Transactions - ASCE 109(2218), 629-678. Droppo I. G., Jeffries D., Jaskot C., and Backus S. (1998) The prevalence of freshwater flocculation in cold regions: A case study from the Mackenzie River Delta, Northwest Territories, Canada. Arctic 51(2), 155-164. Droppo I. G., Leppard G. G., Flannigan D. T., and Liss S. N. (1997) The freshwater floc: A functional relationship of water and organic and inorganic floc constituents affecting suspended sediment properties. Water Air and Soil Pollution 99(1-4), 43-53. Einstein H. A. (1968) Deposition of suspended particles in a gravel bed. Journal of the Hydraulics Division ASCE 94, 1197-1205. Einstein H. A. and Chien N. (1952) Second approximation to the solution of the suspended load theory. In Series No. 47. Institute of Engineering Research. Einstein H. A. and Krone R. B. (1961) Experiments to determine modes of cohesive sediment transport in salt water. Journal of Geophysical Research 67, 1450-1461. Elghobashi S. (1994) On predicting particle-laden turbulent flows. Applied Scientific Research 52(4), 309-329. Emmett W. W. (1980) A field calibration of the sediment-trapping characteristics of the Helley-Smith bedload sampler. In U.S. Geological Survey Professional Paper, Vol. 1139. Fukuda M. K. and Lick W. (1980) The entrainment of cohesive sediments in frshwater. Journal of Geophysical Research 85, 8244-8464. Garcia M. H. (2008) Sedimentation engineering: processes, measurements, modeling, and practice. American Society of Civil Engineers. 1132 p. Georgian T., Newbold J. D., Thomas S. A., Monaghan M. T., Minshall G. W., and Cushing C. E. (2003) Comparison of corn pollen and natural fine particulate matter transport in streams: can pollen be used as a seston surrogate? Journal of the North American Benthological Society 22(1), 2-16. Georgian T. and Thorp J. H. (1992) Effects of microhabitat selection on feeding rates of net-spinning caddisfly larvae. Ecology 73, 229-240.  14  Halbronn G. (1949) Remarque sur la theorie de l'Austausch applique au transport des materiaux en suspension. International Association of Hydraulic Research, 3rd Meeting, 1-6. Hall R. O., Peredney C. L., and Meyer J. L. (1996) The effect of invertebrate consumption on bacterial transport in a mountain stream. Limnology and Oceanography 41, 1180-1187. Hassan M. A., Church M., Lisle T. E., Brardinoni F., Benda L., and Grant G. E. (2005) Sediment transport and channel morphology of small, forested streams. Journal Of The American Water Resources Association 41(4), 853-876. Hassan M. A. and Ergenzinger P. (2002) Use of tracers in fluvial geomorphology. In Tools in Fluvial geomorphology (ed. G. M. Kondolf and H. Piégay), pp. 397-423. John Wiley and Sons. Hedin L. O. (1990) Factors controlling sediment community respiration in woodland stream ecosystems. Oikos 57(1), 94-105. Hendricks S. P. (1993) Microbial ecology of the hyporheic zone - a perspective integrating hydrology and biology. Journal of the North American Benthological Society 12(1), 70-78. Hendricks S. P. and White D. S. (1995) Seasonal biogeochemical patterns in surface-water, subsurface hyporheic, and riparian ground-water in a temperate stream ecosystem. Archiv Fur Hydrobiologie 134(4), 459-490. Hoyal D.C.J.D., Bursik M. I., Atkinson J.F., and Depinto J. V. (1997) Filtration enhances suspended sediment deposition from surface water to granular permeable beds. Water Air and Soil Pollution 99, pp. 157171. Huettel M., Ziebis W., and Forster S. (1996) Flow-induced uptake of particulate matter in permeable sediments. Limnology and Oceanography 41(2), 309-322. Hunt J. N. (1954) The turbulent transport of suspended sediment in open channels. In Proceedings, Royal Society of London, Vol. 224, pp. 322-335. Hurst H. E. (1929) The suspension of sand in water. In Proceedings, Royal Society of London, Vol. 24, pp. 196201. Hynes H. B. N. (1970 ) The ecology of running waters. University of Toronto Press. Jones J. B., Fisher S. G., and Grimm N. B. (1995) Vertical hydrologic exchange and ecosystem metabolism in a Sonoran Desert stream. Ecology 76(3), 942-952. Jones J. B. and Smock L. A. (1991) Transport and retention of particulate organic-matter in two low-gradient headwater streams. Journal of the North American Benthological Society 10(2), 115-126. Kaftori D., Hetsroni G., and Banerjee S. (1998) The effect of particles on wall turbulence. International Journal of Multiphase Flow 24(3), 359-386. Karichkhoff R. F. (1984) Organic pollutant sorption in aquatic systems. Journal of Hydraulic Engineering (ASCE) 110, 707-735. Karlsson O. M., Richardson J. S., and Kiffney P. A. (2005) Modelling organic matter dynamics in headwater streams of south-western British Columbia, Canada. Ecological Modelling 183(4), 463-476. Knighton D. (1998) Fluvial Forms and Processes: A New Perspective. Arnold. Kranck K. (1979) Dynamics and distribution of suspended particulate matter in the St. Lawrence Estuary. Naturaliste Canadien 106(1), 163-173. Krone R. B. (1976) Engineering interest in the benthic boundary layer. In The Benthic Boundary Layer (ed. I. N. McCave), pp. 143-156. Plenum. Langlois J. L., Johnson D. W., and Mehuys G. R. (2005) Suspended sediment dynamics associated with snowmelt runoff in a small mountain stream of Lake Tahoe (Nevada). Hydrological Processes 19(18), 3569-3580. Lenzi M. A. and Marchi L. (2000) Suspended sediment load during floods in a small stream of the Dolomites (northeastern Italy). Catena 39(4), 267-282. Lock M. A. (1981) River epilithon - a light and energy transducer. In Perspectives in running water ecology (ed. M. A. Lock and D. D. Williams), pp. 4-40. Plenum. Maridet L., Wasson J. G., Philippe M., and Amoros C. (1995) Benthic organic matter dynamics in three streams - riparian vegetation or bed morphology control. Archiv Fur Hydrobiologie 132(4), 415-425. McCave I. N. (1984) Erosion, transport and deposition of fine-grained marine sediments. In Fine-grained sediments: deepwater processes and facies. (ed. D. A. Stow and D. J. W. Piper), pp. 35-70. Blackwell Scientific Publications. McCullough D. A., Minshall G. W., and Cushing C. E. (1979) Bioenergenics of lotic filter-feeding insects Simulium spp. (Diptera) and Hydropsyche occidentalis (Trichoptera) and their function in controlling organic transport in streams. Ecology 60, 585-596.  15  McNair J. N. (2000) Turbulent transport of suspended particles and dispersing benthic organisms: The hittingtime distribution for the local exchange model. Journal of Theoretical Biology 202(3), 231-246. McNair J. N. and Newbold J. D. (2001) Turbulent transport of suspended particles and dispersing benthic organisms: the hitting-distance problem for the local exchange model. Journal of Theoretical Biology 209(3), 351-369. McNair J. N., Newbold J. D., and Hart D. D. (1997) Turbulent transport of suspended particles and dispersing benthic organisms: How long to hit bottom? Journal of Theoretical Biology 188(1), 29-52. Meyer J. L. (1988) Benthic bacterial biomass and production in a blackwater river. Verhandlungen der Internationalen Vereinigung fur theoretische und angewandte Limnologie 23, 643-651. Minshall G. W., Thomas S. A., Newbold J. D., Monaghan M. T., and Cushing C. E. (2000) Physical factors influencing fine organic particle transport and deposition in streams. Journal of the North American Benthological Society 19(1), 1-16. Monaghan M. T., Thomas S. A., Minshall G. W., Newbold J. D., and Cushing C. E. (2001) The influence of filter-feeding benthic macroinvertebrates on the transport and deposition of particulate organic matter and diatoms in two streams. Limnology and Oceanography 46(5), 1091-1099. Mulholland P. J. (1981) Organic-carbon flow in a swamp-stream ecosystem. Ecological Monographs 51(3), 307-322. Muste M., Yu K., Fujita I., and Ettema R. (2005) Two-phase versus mixed-flow perspective on suspended sediment transport in turbulent channel flows. Water Resources Research 41, 1-22. Newbold J. D., Thomas S. A., Minshall G. W., Cushing C. E., and Georgian T. (2005) Deposition, benthic residence, and resuspension of fine organic particles in a mountain stream. Limnology and Oceanography 50(5), 1571-1580. Nistor C. J. and Church M. (2005) Suspended sediment transport regime in a debris-flow gully on Vancouver Island, British Columbia. Hydrological Processes 19(4), 861-885. O'Brien M. P. (1933) Review of the theory of turbulent flow and its relation to sediment transportation. Transactions of AGU (April 27-29), 487-491. Ongley E. D., Krishnappan B. G., Droppo I. G., Rao S. S., and Maguire R. J. (1992) Cohesive Sediment Transport - Emerging Issues For Toxic-Chemical Management. Hydrobiologia 235, 177-187. Packman A. I. and MacKay J. S. (2003) Interplay of stream-subsurface exchange, clay particle deposition, and streambed evolution. Water Resources Research 39(4). Paul M. J. and Hall R. O. (2002) Particle transport and transient storage along a stream-size gradient in the Hubbard Brook Experimental Forest. Journal of the North American Benthological Society 21(2), 195205. Petticrew E. L. (1996) Sediment aggregation and transport in northern interior British Columbia streams. In Erosion and Sediment Yield: Global and Regional Perspectives. IAHS Publication No. 236. Pusch M. (1996) The metabolism of organic matter in the hyporheic zone of a mountain stream, and its spatial distribution. Hydrobiologia 323(2), 107-118. Pusch M. (1997) Community repiration in the hyporheic zone of a riffle-pool sequence. In Groundwater/Surface Water Ecotones: Biological and Hydrological Interactions and Management Options (ed. J. Gibert, J. Mathieu, and F. Fournier), pp. 51-56. Cambridge University Press. Pusch M., Fiebig D., Brettar I., Eisenmann H., Ellis B. K., Kaplan L. A., Lock M. A., Naegeli M. W., and Traunspurger W. (1998) The role of micro-organisms in the ecological connectivity of running waters. Freshwater Biology 40(3), 453-495. Rashidi M. G., Hetsroni G., and S. B. (1990) Particle turbulence interaction in a boundary laye. International Journal of Multiphase Flow 16(6), 935–949. Reiser D. W. (1998) Sediment in gravel bed rivers; ecological and biological considerations Gravel-bed rivers in the environment 4, 199-228. Rogers C. B. and Eaton J. K. (1991) The effect of small particles on fluid turbulence in a flat-plate, turbulent boundary-layer in air. Physics of Fluids A-Fluid Dynamics 3(5), 928-937. Rouse H. (1937) Modern conceptions of the mechanics of fluid turbulence. Transactions - ASCE 24(1965), 463543. Rouse H. (1939) An analysis of sediment transportation in the light of fluid turbulence. In Soil Conservation Service Report No SCS-TP-25. United States Department of Agriculture. Rouson D. W. I. and Eaton J. K. (2001) On the preferential concentration of solid particles in turbulent channel flow. Journal of Fluid Mechanics 428, 149-169.  16  Runde J. M. and Hellenthal R. A. (2000) Effects of suspended particles on net-tending behaviors for Hydropsyche sparna (Trichoptera : Hydropsychidae) and related species. Annals of the Entomological Society of America 93(3), 678-683. Rutherford J. C. (1994) River Mixing. J. Wiley and Sons. Savant S. A., Reible D. D., and Thibodeaux L. J. (1987) Convective-transport within stable river sediments. Water Resources Research 23(9), 1763-1768. Smith I. R. (1982) A simple theory of algal deposition Freshwater Biology 12, 445-449. Squires K. D. and Eaton J. K. (1990) Particle response and turbulence modification in isotropic turbulence. Physics of Fluids A-Fluid Dynamics 2(7), 1191-1203. Tanaka S. and Fugimoto S. (1958) On the distribution of suspended sediment in expermental flume flow. In Memoirs of the Faculty of Engineering, Vol. 5. Kobe University. Thomas S. A., Newbold J. D., Monaghan M. T., Minshall G. W., Georgian T., and Cushing C. E. (2001) The influence of particle size on seston deposition in streams. Limnology and Oceanography 46(6), 14151424. Valett H. M., Fisher S. G., and Stanley E. H. (1990) Physical and chemical characteristics of the hyporheic zone of a Sonoran Desert stream. Journal of the North American Benthological Society 9(3), 201-215. Valett H. M., Morrice J. A., and Dahm C. N. (1996) Parent lithology, groundwater-surface water exchange and NO{NE}3- retention in headwater streams. Limnology and Oceanography 41, 333-345. Wallace J. B., Cuffney T. F., Webster J. R., Lugthart G. J., Chung K., and Goldowitz B. S. (1991) Export of fine organic particles from headwater streams - effects of season, extreme discharges, and invertebrate manipulation. Limnology and Oceanography 36(4), 670-682. Wanner S. C. and Pusch M. (2000) Use of fluorescently labelled Lycopodium spores as a tracer for suspended particles in a lowland river. Journal of the North American Benthological Society 19(4), 648-658. Waters T. F. (1995) Sediment in streams: sources, biological effects and control. In American Fisheries Society Monograph, pp. 251. Webster J. R., Benfield E. F., Golladay S. W., Hill B. H., Hornick L. E., Kazmierczak R. F., and Perry W. B. (1987) Experimental studies of physical factors affecting seston transport in streams. Limnology and Oceanography 32, 848-863. Webster J. R. and Ehrman T. P. (1996) Solute dynamics. In Methods in stream ecology (ed. F. R. Hauer and G. A. Lamberti). Academic Press. Wentworth C. K. (1922) A scale of grade and class terms for clastic sediments. Journal of Geology 30, 377392. Williams D. D. (1993) Nutrient and flow vector dynamics at the hyporheic groundwater interface and their effects on the interstitial fauna. Hydrobiologia 251(1-3), 185-198. Woodward J. C., Porter P. R., Lowe A. T., Walling D. E., and Evans A. J. (2002) Composite suspended sediment particles and flocculation in glacial meltwaters: preliminary evidence from Alpine and Himalayan basins. Hydrological Processes 16(9), 1735-1744. Yarin L. P. and Hetsroni G. (1994) Turbulence intensity in dilute 2-phase flows .1. Effect of particle-size distribution on the turbulence of the carrier fluid. International Journal of Multiphase Flow 20(1), 115.  17  CHAPTER 2: PHYSICAL FACTORS INFLUENCING THE DISTRIBUTION, DEPOSITION AND INFILTRATION OF FINE PARTICLES TO A STREAMBED4 2.1 INTRODUCTION  Fine particle deposition and infiltration into the streambed are influenced by physical characteristics of the flow and the solid bed. Hydraulic properties of flowing water (e.g. velocity, turbulence intensity) determine the size and quantity of particles in suspension, while hydraulic properties are in turn a function of suspended load and bed composition. Suspended particles can influence deposition through direct interactions with the fluid or other particles, indirect hydraulic effects, and inherent particle characteristics such as size and density. Bed properties such as roughness and pore space are related to the size distribution of bed particles and influence particle deposition through direct trapping mechanisms and indirect effects on flow. Bed composition can also influence the concentration and distribution of suspended particles by controlling in-channel supply. In this chapter, I investigate the influence of several dominant physical factors on the rate of fine particle deposition and infiltration into the streambed. I establish a reference condition, including a standard flow rate, bulk sand fraction, particle density, and suspended particle concentration, and then assess how changes to each of these factors influence the deposition rate and infiltration of fine particles in relation to near-bed hydraulics. Flow rate, particle density, and suspended particle concentration are conditions of the flowing water, while bed sand fraction influences characteristics of the streambed, such as roughness and porosity. Figure 2.1 illustrates all the possible combinations of these factors, the reference condition, and the combinations tested in this study; each cube represents a different bed condition and each axis a different quantity: flow rate, particle density, or initial concentration. Figure 2.1: Schematic illustrating all possible combinations of experimental factors. Flow Rate  Par t icl eD ens  it y  H  L  H  Flow Rate  L  H Initial Concentration M  H  L  L  a) 20% Sand Bed  b) 80% Sand Bed  *Notes: Factors include particle density (high or low), flow rate (high or low), initial concentration (high, medium, or low) and bed composition (20% sand or 80% sand). Shaded boxes are the combinations tested in this study; box with diagonal lines is the reference condition. 4  A version of this chapter will be submitted for publication. Salant, N. L. and Hassan, M.A. Physical factors influencing the deposition and infiltration of fine particles to a streambed.  18  Flow rate alters mean velocity and turbulence, in turn regulating the capacity and competence of the flow. Relative to the reference condition, lower flow rates are expected to have slower velocities and lower shear stresses, in turn reducing particle transport and enhancing deposition rates. An increase in bed sand fraction is expected to smooth hydrodynamic conditions and thus enhance surface deposition of fine particles, but may limit particle infiltration by reducing pore space. Lower density particles have slower settling velocities, which should translate into slower rates of deposition. In many natural streams, suspended particles may be fully or partially composed of low-density organic matter, as the product of wood or leaf disintegration, senesced organisms, or flocculated organic and inorganic particles. Although organic and flocculated particles have additional characteristics (e.g. irregular shapes; positively-charged surfaces) that influence their movement, these experiments are designed to isolate the effect of low density. Particle concentration may influence deposition by altering hydraulic conditions or particle interactions. For more than 60 years, researchers have studied the effects of suspended sediment on flow resistance (see reviews in Arora et al., 1984; Khullar et al., 2007), demonstrating that in the presence of suspended sediment, flow resistance can either decrease due to turbulence damping or increase due to energy lost in the process of entrainment and increased viscous resistance; the balance of these mechanisms determines the net effect. Mechanisms of density stratification, lift force, particle-fluid interactions and particle-particle interactions (Vanoni and Nomicos, 1960; Smith and McLean, 1977; Czernuszenko, 1998; Righetti and Romano, 2004; Wang and Fu, 2004) have been invoked to explain why high concentrations of particles can alter turbulence characteristics, mean velocities, and velocity profiles (see review in Righetti and Romano, 2004); the magnitude and direction of effects depend on sediment loading conditions (i.e. particle concentration and size). In some cases, no turbulence modulation occurs (e.g. Muste and Patel, 1997). Particle interactions at high concentrations of different particle sizes may directly influence deposition velocity, as the presence of larger particles can affect the settling velocities of smaller particles (Garcia-Aragon et al., 2002, 2004). Experiments were conducted in a laboratory flume to allow for carefully controlled conditions and highresolution measurements. In order to reduce confounding effects and facilitate comparisons, I tested each factor in isolation. In Chapter 3, I apply the measurements from these experiments to theoretical models of particle distribution to assess model performance for the particle sizes used in this study and the influence of each factor on the vertical profile of these particles.  2.2 METHODS  2.2.1 FLUME DESCRIPTION  All experiments were conducted in a small (0.15 m wide x 7.75 m long x 0.44 m deep) recirculating flume located in the Hydraulics Laboratory of the Department of Civil Engineering at the University of British Columbia, Canada (Figure 2.2). Water was supplied from a holding tank with two separate chambers for the recirculation of water and sediment. For these experiments, water and sediment were transported only by the  19  sediment recirculating pump supplied from the smaller sediment recirculating chamber (0.76 m x 0.64 m x 1.5 m); a plywood wall sealed with silicone prevented mixing of particles and water between the two chambers. While the flume was running, turbulence generated by the water flowing into the sediment recirculating chamber kept all fine material in suspension; therefore the loss of particles due to settling was minimal. Furthermore, in order to keep the water and particles mixed and prevent particle settling in the supply tank, a sump pump in the centre of the tank and four small bilge pumps in each corner of the tank were operated continuously to agitate water and particles in the tank. Water level in the tank was maintained at ~10 cm below the maximum height when the pump was not running for a total water volume of ~685 L. Water flow was controlled by the pump setting and by opening or closing a gate at the end of the flume. Flume slope remained at 0.1% for all experiments.  Figure 2.2: Schematic of flow and sediment re-circulating flume, University of British Columbia, Canada. End gate  0.0  1.0  2.0  3.0  3.5  4.0  5.0  Sediment recirculating pump  6.0  7.0  Slope adjustment  Sediment Water recirculating recirculating chamber chamber  2.2.2 EXPERIMENTAL DESIGN  The reference condition for these experiments had a bed sand fraction of 20%, an initial dose of silica particles, a relatively high flow rate, and a reference initial particle concentration. Subsequent experiments tested variations in a single factor from the reference condition. Experiments are identified using the following codes: ‘20%’ or ‘80%’ indicates the percentage of sand (< 2 mm) in a bed otherwise composed of coarse material; ‘S’ indicates a dose of high-density ground silica, while ‘P’ indicates a dose of low-density plastic particles; ‘HF’ and ‘LF’ indicate comparatively high and low flow rates (average discharge ~8.5 L/s and 4.2 L/s), respectively, although both would be considered low for natural systems; and ‘HC’ and ‘LC’ indicate high and low initial concentrations, respectively, relative to the reference condition (Table 2.1). Thus the reference condition is identified as ‘20%SHF’. Due to limited time and resources, only three conditions were replicated three times (plastic, low flow, and reference); the high concentration and low concentration experiments were only conducted once and the 80% sand bed experiments were conducted twice.  20  Table 2.1: Experiment code names and descriptions. 20%SHF is the reference condition. Code Bed sediment Flow rate 20%SHF* 20% bulk sand fraction High flow (~8.5 L/s) 20%SLF 20% bulk sand fraction Low flow (~4.2 L/s) 20%PHF 20% bulk sand fraction High flow (~8.5 L/s) 20%SHFHC 20% bulk sand fraction High flow (~8.5 L/s) 20%SHFLC 20% bulk sand fraction High flow (~8.5 L/s) 80%SHF 80% bulk sand fraction High flow (~8.5 L/s) *Notes: Asterisk indicates reference condition.  Particle type Silica Silica Plastic Silica Silica Silica  Dose conc. 0.31 kg/l 0.31 kg/l 0.15 kg/l 0.52 kg/l 0.10 kg/l 0.31 kg/l  Rep. 3 3 3 1 1 2  Reference bed composition was a pebble-gravel-sand mixture, 20% of which was composed of sand-sized particles (0.25-2 mm); the full particle size distribution for this mixture is presented in Figure 2.3. Particles were sieved into standard 0.25-0.5, 1-2, 8-11, 16-22, and 30-45 mm size fractions. The volume of each size fraction was determined as the appropriate percentage of total bed volume, calculated from the flume width, length, and desired bed depth of 10 cm (0.15 m wide x 7.75 m long x 0.10 m deep = 0.116 m3); some material was lost during sieving and mixing so that final bed depths ranged between 7-9 cm. Bulk mass for each size fraction was estimated from the volume using a bulk density of 1400 kg/m3. The sediment was mixed together by hand and placed in batches on the flume floor, then graded by hand to create a relatively level surface down the length of the flume. After each replicate experiment, the bed was removed, sieved to remove all fine particles (<0.125 mm), and replaced in the flume so that each experiment had a slightly different bed configuration but the same bulk particle size distribution. Figure 2.3: Bulk particle size distributions for artificial mixtures used in experiments (20% and 80% sand). 100 Cumulative frequency (%)  80% sand 20% sand  80 60 40 20 0  *  0.1  1 10 Median particle size (mm)  100  *Notes: Solid lines indicate the 50th percentile grain size for each distribution. Reference flow rate was ~8.5 L/s, identified as ‘high flow (HF)’. At this discharge, none of the bed particles in the 20% sand mixture was mobile thus the bed remained static throughout the experiment. A static bed condition was desired to eliminate complications with bedform development and temporally changing bed conditions. A relatively constant flow rate was maintained for the entire experiment, for a total of nine hours; a clamp meter operating throughout each experiment measured a range of 4.18-4.26 L/s for low flow and 8.29-  21  8.68 L/s for high flow. Water level, bed elevation, and longitudinal velocity measurements indicated that flow was uniform between the 2 and 5 m points of the flume; all measurements were restricted to within this uniform zone. Water surface and bed elevation measurements were repeated periodically throughout each experiment to ensure uniform flow and to determine the average flow depth. For the reference particles, I used SIL-CO-SIL® 75 ground silica (at least 99.4% SiO2) particles from U.S. Silica Company (http://www.u-s-silica.com/scs.htm), with a manufacturer-specified particle size of 75 µm and finer. However, an independent analysis indicated a distribution of 4.5 to 125 µm and a median particle size of ~20 µm. Ground silica particles are inherently inert, white, low moisture, and do not dissolve or clump in moving water. Density is equal to that of pure silica (specific gravity of 2.65); particles examined under a microscope appeared sub-angular in shape. After one hour of flow to condition the bed and allow for a set of measurements without a particle suspension, a high-concentration slurry was homogenised in a 1 L bucket, poured into the sediment recirculating chamber, and mixed vigorously by hand to evenly distribute it prior to recirculation. Particle mass for the slurry was chosen to produce an initial average concentration that was within the range of the instrument detection limits (Sequoia Scientific LISST-100X Particle Size Analyser, described in detail below) but also a measurable proportion of the bed material. A silica particle mass equal to only 1% of the bed mass equates to ~3.1 kg. Preliminary tests showed that this mass produces concentrations that far exceed the LISST upper concentration limit, estimated at ~1000 µl/l. Instead, I used a particle mass of 0.31 kg for the silica particles, amounting to ~0.1% of the bed volume and producing a reference initial concentration of ~160 ppm. Flow rate was varied from the reference condition by reducing the bulk discharge to ~4.2 L/s and narrowing the end gate slightly to maintain uniform flow. Bed composition was varied from the reference condition by increasing the fraction of sand to 80% (Figure 2.3) in order to test a bed composition dominated by sand versus a bed dominated by gravel. Coarse particles were the same as those from the 20% sand bed, but the sand fraction contained a wider range of sizes; half of the 80% fraction was composed of a coarse sand (median particle size 1.0 mm) and half was composed of a fine sand (median particle size 0.5 mm). Before the first experiment and between the first and second experiments, this mixture was sieved in the same manner as the 20% sand bed to remove all particles < 0.125 mm and ensure that any particles present in the bed after the experiment were due to deposition and infiltration from the water column. Particle density was varied from the reference condition by replacing the silica particles with a low-density plastic particle in the initial particle dose. For the low-density particles, ‘Clear-Cut PlastiGrit’ (-200 mesh) fine plastic particles (Composition Materials Co.; http://www.compomat.com) were used. These particles, designed for vehicle paint removal, have a manufacturer-specified specific gravity of 1.3 and particle size 74 µm and finer. However, an independent grain size analysis (James Buttle, personal communication) measured a distribution of 5 to180 µm and a mean of ~90 µm. Like silica particles, plastic particles examined under a microscope appear sub-angular in shape. Plastic particles do not clump or flocculate and are not chemically charged thus the only characteristic they share with natural organic particles and flocs is a density less than pure inorganic matter. I used a plastic particle mass with a volume concentration equivalent to the silica particles  22  (~0.15 kg), created a homogenous slurry, and dosed the sediment recirculating chamber in the same manner as for the reference experiments. Initial particle concentration was varied from the reference condition by increasing or decreasing the mass of silica particles used in the particle dose. For the high concentration (HC) experiment, the reference silica mass was increased by a factor of five-thirds to 0.52 kg; for the low concentration (LC) experiment, the dose was decreased by a factor of one-third to 0.10 kg.  2.2.3 MEASUREMENTS  Measurement schedule and instrument operation for each nine hour experiment are described in the following paragraphs; all experiments were conducted in an identical manner.  ADV velocity profiles  Soon after the flow was started, velocity profiles were measured at the centreline of the flume at the 2.5, 3, and 3.5 m locations, within the uniform section of the flume. Profiles were then repeated at the same locations one hour later, immediately following the particle dose, and twice more during the experiment, four and eight hours after the dose. Velocity measurement times are thus represented as ‘-1h’, ‘0h’, ‘4h’, and ‘8h’, in reference to the hour relative to the time of the particle dose. Measurements were made before and after the dose in order to determine whether the addition of particles to the flow influenced hydraulic conditions. Velocity measurements were made with a Sontek 16 MHz-Micro Acoustic Doppler Velocimeter (ADV) at a sampling rate of 50Hz and a velocity range setting of 100 cm/s. The instrument body was supported by a stand that was separate from the flume and the probe was held in a small brace attached to the flume walls. Velocity profiles were compiled from point measurements taken vertically in the water column. In preliminary tests, vertical velocity profiles were measured at several locations along the length of the flume and across the width of the flume to assess the degree of spatial variability in flow velocity. Although some lateral variation was observed due to wall drag effects, little change in velocity was observed longitudinally within the 2 to 4 m test section; thus measurements were restricted to the centreline of the flume. Each vertical profile consisted of a minimum of eight measurements throughout the flow depth, spaced with increasing resolution near the bed. Although the exact spacing differed between experiments depending on flow depth, profiles typically had the lowest 2-3 measurements at 0.5 cm intervals, the middle 2-4 measurements at 1 cm intervals, and the uppermost 1-2 measurements at 2 cm intervals. Because the ADV has a 5 cm sampling volume that extends below the tip of the probe, no measurements were taken within the top 5 cm of flow. The deepest measurement was taken as close to the bed as feasible without allowing the ADV sampling volume to include the bed (0.3-0.6 cm). All velocity measurements were time averages of at least 60 seconds in order to reduce error around the mean (Sontek, 1997) but were increased to 90 seconds in the case of high standard errors or low signal correlations (Martin et al., 2002 ).  23  Because of the small width-to-depth ratio of the flume (aspect ratio < 5), wall-generated turbulence likely generated secondary currents that influenced velocity measurements taken at the centre of the flume, as evidenced by relatively high lateral velocity fluctuations. However, these secondary currents were present during all experiments, i.e. for all experimental conditions. I compared mean velocity and turbulence statistics between experimental conditions and for all I assumed that the degree of wall-generated turbulence was the same. I am only concerned with the relative difference in hydraulic parameters due to physical condition, with wall effects presumed to be the same for all scenarios. I acknowledge that the nature of these effects may differ for large river systems with wide channels and limited wall effects.  LISST suspended particle concentrations  Following the particle dose, water column particle concentrations were measured continuously at a single near-bed depth at 3 m for the first hour of the experiment. Preliminary tests showed that during this initial period particle concentrations decreased rapidly due to deposition on the streambed. Thus continuous measurements were used to assess initial rates of particle deposition. After the first hour, rates of deposition slowed so that temporal effects were less significant. Thus beginning at 1h and continuing hourly until 8h vertical concentration profiles were measured to assess particle distribution in the water column. At the start of each profile measurement, water temperature was also measured with a thermometer inserted into the flow in the centre of the flume. Like velocity profiles, concentration profiles contained at least eight point measurements throughout the flow depth, spaced with increasing resolution towards the bed. Although the exact spacing differed between experiments depending on flow depth, profiles typically had the lowest 2-3 measurements at 0.5 cm intervals, the middle 2-4 measurements at 1 cm intervals, and the uppermost 1-2 measurements at 2 cm intervals. Measured concentration profiles are analysed and discussed in Chapter 3. Concentration measurements were made by pumping water from the flume through a Sequoia Scientific LISST-100X Particle Size Analyser (type C; size range 2.5 – 500 µm) equipped with a flow-through chamber. A peristaltic pump moved water and sediment from the flume through a plastic intake nozzle and rubber tubing mounted onto the flume wall. A ruler attached to the mount indicated the height of the nozzle above the bed; all near-bed measurements were taken as close to the bed as permitted by the radius of the plastic nozzle (0.3 cm). A digital flow-meter measured discharge through the tubing and the pump rate was adjusted so that flow through the tubing approximated the flume velocity at that height. Water and sediment entered the tubing from upstream and was returned via an identical outtake nozzle on the downstream side so as not to influence subsequent measurements. Raw data were immediately offloaded from the instrument and analysed using the LISST data processing program version 4.65 (Sequoia Scientific, Inc.) which reports values of total volume concentration (C), mean particle size (D50), and particle size distribution (PSD) with the concentration for each of 32 logarithmically-spaced size classes. Sixty measurements are averaged over a ~15 s sampling period for each concentration/PSD measurement.  24  The LISST measures particle concentration by laser diffraction; a laser beam moving through the water is scattered by particles and sensed by a multi-ring detector behind a receiving lens. Scattering intensity is recorded over a range of small angles and is then converted to particle size distribution by a mathematical inversion. Details of the laser diffraction technique and principles of operation of the LISST (Laser In-Situ Scattering and Transmissometry) can be found in other papers (Agrawal and Pottsmith, 2000; 2004) and on the Sequoia Scientific website: http:www.sequoiasci.com. At and above a maximum particle concentration, the laser light is completely blocked by the particles and cannot reach the optical sensor. At high concentrations below this upper limit, scattered light may be re-scattered by nearby particles (‘multiple scattering’) adding error to the concentration estimate. Multiple scattering occurs when the optical transmission through the water drops below ~30%; the error is most severe below 10% of the optical transmission (Chuck Pottsmith, personal communication). Because small particles have more surface area per unit volume than larger particles, multiple scattering is likely to occur at lower concentrations for small particles than for larger particles. Dose amounts for the reference condition produced an expected maximum concentration of ~160 µl/l. At this concentration, the optical transmission of the LISST was ~50-70%, depending on particle type, which is well above the level when multiple scattering occurs. Thus LISST measurements are expected to be relatively accurate. In the high concentration experiment, starting concentrations were still within the range of the LISST detection limits, although optical transmission was temporarily reduced to <30%, potentially introducing error into our measurements due to multiple scattering. However, this effect was short-lived, lasting only 8 minutes, and not very severe; optical transmission was never lower than 24% and increased above 30% once concentrations dropped below ~250 µl/l.  LISST Calibration  In order to determine the accuracy of the LISST-measured concentrations for the materials and particle sizes used in this study, I performed controlled tests that compared LISST readings to known suspended particle concentrations of the silica and plastic particles. From these tests, I developed calibration relations between LISST concentrations and actual concentrations for each particle type. In addition, I recorded the optical transmission for a range of particle concentrations of each particle type, determining the concentration level at which transmission dropped below 30% and 10%. From this I was able to set a limit on the concentrations that could be accurately measured with the LISST. To conduct the test, I filled a 50 litre plastic tub with distilled water. At the bottom of the tub I installed four small bilge pumps positioned to recirculate water and prevent particle settling. I used the same peristaltic pump, set at the same rate, as well as the plastic intake-outtake nozzles and rubber tubing from the flume experiments to move water from the centre of the tub through the LISST flow-through chamber and back. An initial background measurement determined scattering due to residual particles in the water or marks on the optical windows. For each subsequent measurement, I added an incremental dose of fine particles, identical to those used in the experiments. Each dose was mixed thoroughly by hand, allowed to recirculate, and pumped  25  through the entire system before the measurement was taken. Doses were added in increments of 0.1 g up to 1 g, 1 g from 1-10 g, and 10 g thereafter. Dosing ended when the optical transmission dropped to zero and no concentration was detected. Mass concentrations (g/l) were converted to volume concentrations (µl/l) by dividing by particle density. Dose amounts result in expected concentrations of ~1-500 µl/l. Silica concentrations were well-measured by the LISST; the ratio of LISST-measured to expected concentrations was virtually 1:1 (Figure 2.4a; slope = 1.0001, r2 = 0.97). Optical transmission decreased exponentially with the increase in concentration, dropping below 30% at ~262 µl/l and below 10% at ~506 µl/l (Figure 2.4b). LISST-measured concentrations for plastic, however, were greatly biased but extremely precise; the ratio of LISST-measured to expected concentrations was 2.42:1 (r2 = 0.99). Optical transmission through plastic particles also decreased exponentially with concentration, reaching 10% at 1250 µl/l and 30% at 639 µl/l; these corresponded to actual concentrations of 264 and 516 µl/l, respectively, very similar to the levels for silica particles. Overestimation of plastic concentrations cannot be explained by multiple scattering due to an abundance of fine particles in the plastic doses; size distributions for the plastic and silica particles are very similar. Instead the difference may be due to the mathematical inversion used by the LISST software program for converting measured particle size distributions to particle concentration that assumes higher particle density. Given the robustness of the relation between measured and expected plastic concentrations, I simply used the 2.4:1 calibration to convert all measured plastic concentrations to actual values for comparison with silica measurements. Put another way, measured LISST concentrations were converted to actual concentrations by the following equation: C = 0.4094 Cmeas, where C is the actual concentration and Cmeas is the measured concentration of plastic particles detected by the LISST.  Figure 2.4: Relation between a) expected volume concentrations (VC) based on a known particle dose and water volume and LISST-measured concentrations; and b) LISST-measured volume concentrations and optical transmission of the LISST laser beam (range from 0-1; 1 = 100% transmission). 1.1  800  Plastic y = 2.42x – 11.0 r2 = 0.99  0.9 0.8  600 Silica Plastic  500  Silica y = 1.00x – 7.10 2 r = 0.97  400 300  Optical transmission  Measured VC  700  b)  1  Silica Plastic Expon. (Plastic)  0.7 0.6 0.5 0.4 0.3 0.2  200  0.1 0  0 0  100  200  300  400  500  Expected VC  0  500 1000 M easured VC (ul/l)  1500  *Notes: Solid horizontal line indicates 30% optical transmission.  26  Stratified subsurface samples  At the end of the nine hour experiment, the flow was stopped and the flume was allowed to drain. Three 0.5 metre-long sections within the test zone were divided into stratified bed samples in order to determine the amount and depth of particle infiltration into the bed. Four different layers of the section were removed by hand: ‘armour,’ ‘subsurface,’ ‘middle’, and ‘base.’ Layers were approximately 0-2, 2-4, 4-6, and 6-7 cm from the bed surface. Each layer was oven-dried and sieved for particle size analysis; size classes included a fine particle class <125 µm. Because each layer had slightly different masses, the proportion of layer mass in the fine particle class < 125 µm was used for analysis, rather than the raw mass in order to correct for the fact that differences in sample mass may bias the comparison of fine particle mass (i.e. two samples from the same depth and condition will have different fine particle masses if one sample is simply larger than the other). Given that plastic particles are approximately half as dense as silica particles, the same number of particles weigh half as much. Thus plastic sample masses <125 µm were corrected for comparison with silica by multiplying by the ratio of silica to plastic density (2.65:1.3), giving the equivalent mass for the same number of silica particles.  2.3 ANALYSIS  Data from all experiments were analysed using the same calculations and statistical procedures in order to determine the effects of a change in each physical factor from the reference condition. For all metrics, a repeated measures model (Maindonald and Braun, 2003) was fit to the data, treating experimental condition as a fixed factor and experimental run within each condition as a random factor. A repeated measured model was used because measurements were repeated at different bed locations (velocity profiles, hydraulic parameters, and subsurface samples), heights (velocity profiles), and times (near-bed concentration measurements). Model error is based on the assumption that the variance is constant among conditions, which allowed me to test whether there were significant differences among replicated and non-replicated conditions. Among conditions, error was due to between-run variability, but for factors that varied within runs (e.g. location) the error was due to within-run variability. Because variance was assumed constant for each condition, I could apply the variance observed in the replicated experiments (e.g. reference condition) to the non-replicated experiments (e.g. high concentration condition), correcting for the smaller sample size (e.g. n =1 instead of n = 3). As a result, the model estimate and standard error for each condition are not absolute values, but represent the difference between the factor levels. Sources of between-run variability included a slightly varying bulk discharge (< 5% change) due to inconsistent pump output, different bed configurations due to the removal, remixing, and replacement of the sediment between each experiment, and natural short-term fluctuations in turbulence. Given the noise introduced by these factors, as well as the exploratory nature of this study, a p-value of 0.1 was adopted to indicate a significant effect. Using a less conservative p-value than the conventional 0.05 or 0.01 biases the analysis toward a Type I error (false positive) in order to limit the risk of ignoring an effect that might be  27  significant in future study (Type II error). From an exploratory standpoint, I deemed the risk of falsely identifying an effect less serious than failing to observe a difference when there actually was one. For this reason, a p-value < 0.1 is considered significant throughout this study. A repeated measures model was used to compare velocity profiles, hydraulic parameters, particle concentrations, and subsurface samples among experimental conditions. Additional analyses relevant to the velocity and concentration measurements are described in the following sections.  2.3.1 VELOCITY AND TURBULENCE PARAMETERS  Velocities and three-dimensional velocity fluctuations measured by the ADV were used to calculate a suite of local velocity and turbulence parameters. Local, rather than channel-averaged estimates, were used because turbulence anisotropy across the narrow flume channel (aspect ratio < 5) likely generated secondary currents that influenced velocity measurements taken at the centre of the flume. Parameters computed from at-a-point measurements reflect local hydraulic conditions of the bed and eliminate the requirement of laterally uniform flow needed for the calculation of channel-averaged parameters. Velocity profiles from all hours at each location (2.5, 3, and 3.5 m) were assessed to confirm that there were no temporal trends in velocity during the experiment. No systematic difference was observed between pre- and post-dose profiles, but I chose to use only the profiles taken after the particle dose (i.e. 0h, 4h, and 8h) in order to eliminate any possible bias introduced by the absence of suspended particles prior to the dose. All post-dose profiles were combined into a single profile for each location. Combined profiles from the three locations were then compared to verify flow uniformity and a lack of longitudinal trends. Variability in local and upper-flow parameters among locations was assessed for evidence of spatially varying bed conditions within each run. Hydraulic parameters were computed as follows. First, combined velocity profiles from all post-dose hours at each of the three locations were fitted with regression lines of velocity as a function of log-normalized height above the bed; this is considered the appropriate regression given that most of the error is associated with the velocity value. Residual plots were then checked to ensure linearity and a lack of trends in the residuals. The regression equation for each location was then used to predict a depth-integrated streamwise velocity (Ux), maximum streamwise velocity (umax) and near-bed horizontal velocity (u0). Maximum velocity was taken as the value of the regression extrapolated to the water surface (when height, z is equal to flow depth, H); near-bed velocity was determined from the regression at z = 0.05H, which was the average height of the near-bed measurement for all experiments. Depth-integrated velocity was determined from integration of the region from the near-bed height to the water surface. Velocity parameters determined for each experimental condition from the velocity regressions were tested for significant differences among conditions using the repeated measures model described above. Each hydraulic parameter was the response variable regressed on experimental condition as a fixed categorical variable and experimental run as a random variable. Because measurements were repeated within each run at the three locations, values from each location were considered repeated measures in the model as an estimate of within-run variability. Replicated runs provided the within-condition  28  (or between-run) variability. In addition, I used the repeated measures model to test for significant differences among the log-normalized velocity profiles of the different experimental conditions. Velocity was treated as the response variable, log height (ln(z)) as the covariate (continuous variable), experimental condition (flow level, bed condition, particle density, or initial concentration) as a fixed categorical variable, and run as a random variable; using profiles from the three locations within each run provided an estimate of within-run variability due to location and height. For all tests, a p-value < 0.1 was considered significant. Three-dimensional velocity measurements were used to calculate local estimates of shear stress. Data output from each ADV measurement relevant to this parameter included the covariance of the streamwise and vertical velocity fluctuations (u’ and w’, respectively), equal to <u’w’>, where ‘< >’ indicates the average, as well as the root-mean-square of the velocity fluctuations in all three directions (equal to < u ' 2 > , < v' 2 > , and < w' 2 > , where v’ is the lateral component). Local estimates of shear stress were calculated from the Reynolds stress method (Biron et al., 2004). Reynolds stress at each depth was calculated as  τ R = − ρ < u ' w' > where ρ is the density of water (assumed to be 1000 kg/m3). Like velocity profiles, Reynolds stress profiles were plotted for determination of turbulence parameters for each location within each experimental run. Reynolds profiles typically demonstrated a shift in slope mid-way through the water column, increasing from the bed to a maximum part-way up the flow depth and decreasing towards the water surface. Thus a nonparametric lowess model was fit to Reynolds stress as a function of height above the bed, using a span of 0.75 and a Gaussian (least-squares) fit; a span of 0.75 was found to provide a smooth fit without removing the trends in the data. Near-bed Reynolds shear stress (τRe0) was predicted from the regression as the value of the regression when z = 0.05H. Maximum Reynolds stress (τReM) was determined by visual inspection of the regressed profiles as the value at the height where the shift in slope occurs. This height is often considered the top of the roughness layer thus I used the maximum shear stress as representative of turbulence conditions in the region outside the inner boundary layer. Like velocity parameters, predicted nearbed and maximum Reynolds stresses were statistically compared among physical conditions using the same repeated measured model described above; a p-value < 0.1 was considered significant. Velocity and Reynolds stress parameters computed for each location and experimental condition are provided in the Appendix, Table A.1.  2.3.2 PARTICLE DEPOSITION RATES  Near-bed volume concentrations (C) measured by the LISST were plotted against time to determine concentration decay rates, interpreted as the rate of particle deposition. Concentration as a function of t, C(t), approximately fit the exponential decay function, C (t ) = C0 e − kt , where C0 is the initial concentration at time t = 0 and the coefficient k is the decay rate per unit of time. For a plot of ln(C/C0) against time, the slope is equal to  29  the decay rate, k. Functions were fit from time t = 0 at peak C (such that C0 = peak concentration at time t = 0; C/C0= 1 when t = 0) to the end of the measurement period. After the dose of fine particles to the sediment recirculating chamber, the slug of particles moved as a wave through the flume system, becoming completely mixed with the full water volume after one or two cycles through the system. As the wave entered the LISST intake at the centre of the flume, the measured concentration increased, peaked, and decreased as the tail of the wave passed by. Depending on the degree of mixing, subsequent measured concentrations either peaked again at the next cycling of the wave or decreased continuously until the end of the experiment. Although every effort was made to fully mix the particles in the recirculating chamber prior to recirculation, each experiment had a different level of mixing and thus exhibited either two peaks (in the case of incomplete mixing) or only one peak (in the case of complete mixing). For those experiments with two peaks, defining the peak concentration and thus the start time (t = 0) for fitting the exponential function was somewhat difficult. Using the initial peak (from the first cycle of the wave) could artificially elevate the peak value because the particles were not fully mixed with the total water volume; subsequent decreases in concentration would thus be not solely a result of deposition, but also due to subsequent mixing. In contrast, although at the secondary peak (the maximum value prior to continuous decrease in concentration) I assumed that particles and water were fully mixed, deposition since the initial peak decreased the peak value from what it would be had complete mixing occurred prior to recirculation. Given that neither the initial nor the secondary peak represents the true maximum concentration, I elected to define the peak concentration in an alternative manner that incorporates the information provided by both peaks. Least-squares regression lines were first fit to the raw data (ln(C) against time in seconds), including both peaks, with time t = 0 at the time of the initial peak. Regression lines using the raw data provide a model fit in-between the functions defined by the two different peaks, weighted by the magnitude of each peak. Thus the initial peak value C0, defined by the regression line when t = 0, can be considered a best estimate of the true peak concentration if complete mixing had occurred prior to recirculation. For consistency, C0 values for experiments with only one peak were determined in the same way. Because of the imperfect fit of the exponential model, in all cases peak values determined this way were lower than any of the actual measured peak concentrations. Concentrations normalized by the model-determined C0 were log-transformed (ln(C/C0)) and fit to measurement time (in seconds) using a linear least-squares regression. Residual plots were checked for linearity. Like velocity profiles, a repeated measures model was performed on normalized and log-transformed concentrations as a function of time, treating ln(C/C0) as the response variable, time (seconds) as the continuous covariate, and run as a random variable. Concentrations were measured at only one location, thus the only measure of within-run variability was the error associated with time. A value of p < 0.1 corresponding to the interaction between condition and time indicated a significant effect of a condition on particle decay rate (slope of the regression). Given that all regressions were normalized to have an intercept of one, intercepts were not significantly different among conditions. However, a repeated measures model fit to non-normalized but logtransformed data was also used to determine whether intercepts, and thus model-predicted initial concentrations, were significantly different between conditions. Concentrations of four particle size classes within the  30  suspension measured by the LISST (2.72-4.46 µm, 4.46-16.78 µm, 16.78-63.11 µm, and 63.11-122.39 µm) were analyzed in the same manner to calculate and compare decay rates for particles of different sizes. Several authors have proposed a qualitative theory of particle deposition from turbulent water that describes the instantaneous rate of change in concentration,  dC , as solely a function of the particle still-water dt  settling velocity, ws, the particle concentration, and the depth of the mixed layer (usually taken as the depth of flow, H, assuming complete vertical mixing) (e.g. Einstein, 1968). An equation for particle concentration and deposition rate can then be derived from this general model (Smith, 1982; McCave, 1984): −(  C (t ) = C 0 e  ws )t H  In gravel-bedded streams with turbulent flow, particle deposition can be influenced by other factors, such as grain interactions and turbulent advection. Thus the actual deposition velocity, wd, can differ from the stillwater settling velocity. By replacing ws with wd in this equation, I was able to use values of k from the measured concentration decay functions and the average flow depths from measured water surface and bed elevations to estimate the particle deposition velocity, wd = Hk. For comparison with these depositional velocities, I then computed ws for the suspension using the equation of Ferguson and Church (2004):  ws =  ρ − ρw RgD 2 , where R is the particle’s submerged specific gravity ( p ; where ρp and ρw 3 0.5 C1v + (0.75C2 RgD ) ρw  are particle and water density, respectively), g is gravitational acceleration, D is mean particle diameter, and v is the kinematic viscosity of water; C1 and C2 are constants with values of 18 and 1.0 for typical natural sands. I assumed particle densities equal to 2650 kg/m3 for silica and 1300 kg/m3 for plastic particles. For computation of ws, I used the measured median particle size (D50) at the initial concentration (C0) from the deposition models. Over the course of the experiment, D50 of the suspension decreased as coarse particles preferentially deposited, reducing the ws of the particles in suspension. Thus I also calculated the settling velocity of four size classes within the distribution detected by the LISST (2.72-4.46 µm, 4.46-16.78 µm, 16.7863.11 µm, and 63.11-122.39 µm) for comparison with the actual deposition velocity of each class. Because the distribution within these particle size classes is difficult to determine, I used the mean particle size of each class for computation of settling velocity. In addition, water temperature increased over the course of the experiment due to the heat generated from the flume operation. Temperatures over the course of the experiment ranged from 15 to 25 ºC, corresponding to a decrease in viscosity from 1.1 to 0.9 x 10-6 m2 s-1. For the range of particle sizes used in this study (2-122 µm), this decrease in viscosity results in a 13-17% increase in settling velocity, depending on the particle size used (small particles experience a greater percent increase in settling velocity as viscosity decreases). For the purpose of comparison with wd, which integrates the conditions over the course of the entire experiment (including the change in temperature), I use a median temperature and viscosity (20 ºC and 1.0 x10-6 m2 s-1) for the computation of ws. Thus the error in the calculated settling velocities was 10% or less.  31  Deposition and settling velocities were used to calculate enhancement factors (Ed) for each experiment, time period, and particle size class, using the equation E d =  wd (Fries and Trowbridge, 2003). Comparisons ws  were then made between enhancement factors of different experimental factors and particle sizes to determine whether depositional or entrainment factors have a more dominant effect.  2.4 RESULTS AND DISCUSSION  Following a discussion of results from the reference condition experiments, each of the subsequent sections discusses how a change in a given physical condition influenced local hydraulic conditions, particle deposition, and particle infiltration.  2.4.1 REFERENCE CONDITION  Velocity and turbulence parameters  Despite scatter in some velocity profiles, only three of the logarithmic regressions of streamwise velocity against depth for all locations, runs, and conditions were not significant (p > 0.1; Appendix, Table A.1). Residual plots demonstrate linearity for all regressions except two, which show a non-linear trend. Because the repeated measures model relies on linearity, the inclusion of these two non-linear profiles degraded the model fit. However I deemed this effect insignificant given the small number of non-linear cases. Both of these profiles were visibly uniform, with near-bed velocities nearly as fast as maximum velocities and roughly linear non-transformed profiles. As predicted, these linear velocity profiles occurred under high roughness conditions (20%SHF and 20%SLF) although the effect was not as apparent in other high-roughness profiles. Mean velocity and Reynolds stress parameters for the reference condition and each of the other experimental conditions as estimated by the repeated measures model are provided in Table 2.2. Because variance was assumed to be the same for all conditions, reported standard errors are determined from the variance associated with the replicated conditions and the number of runs for each condition. Regression parameters for the velocity profiles from all physical conditions are presented in Table 2.3.  32  Table 2.2: Velocity and turbulence statistics for different physical conditions. Condition n Ux (cm s-1) u0 (cm s-1) umax (cm s-1) τRe0 (Pa) τReM(Pa) 20%SHF 3 42.44(3.63) 21.97(3.94) 62.92(4.66) 1.12(0.39) 2.39(0.33) 20%SLF 3 21.25(3.63)*** 8.16(3.94)** 34.34(4.66)*** 0.17(0.39) 0.75(0.33)*** 20%PHF 3 41.49(3.63) 24.93(3.94) 58.06(4.66) 1.22(0.39) 2.05(0.33) 20%SHFHC 1 31.40(6.82) 9.11(6.82) 53.68(8.07) 0.66(0.67) 2.75(0.57) 20%SHFLC 1 40.56(6.82) 24.09(6.82) 57.03(8.07) 1.36(0.67) 2.04(0.57) 80%SHF 2 33.29(4.45) 24.64(4.82) 41.95(5.75)** 0.28(0.48) 0.57(0.40)*** DF 7 7 7 7 7 F 4.96 2.89 4.96 1.28 5.16 p <0.05 <0.1 <0.05 0.37 <0.05 *Notes: See Table 2.1 for description of experimental code names; ‘20%SHF’ is the reference condition. Values are parameter estimates (SE) of conditions determined from a repeated measures model. n’ is the number of experimental replicates. For each hydraulic parameter, the model is described by degrees of freedom (DF), F-statistic (F), and significance value (p) associated with experimental condition. u0, Ux, and umax are the near-bed, depth-integrated, and maximum streamwise velocity, respectively. Near-bed and maximum values of the Reynolds shear stress (τRe0 and τReM,) were determined from fluctuations of the three velocity components. See text for equations. Significance codes next to each parameter indicate the p-value of the comparison with the reference condition (20%SHF): ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. Table 2.3: Regression coefficients and associated parameters for streamwise velocity regressed against ln(z) for six physical conditions. Condition 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF DF F p  n 3 3 3 1 1 2  Slope 10.54(1.43) 4.68(1.43)** 9.20(1.43) 14.08(2.48) 10.45(2.48) 5.54(1.75)*** 705 3.51 <0.01  Intercept 30.34(3.08) 15.38(3.08)** 28.27(3.08) 20.46(5.34) 32.53(5.34) 26.33(3.78) 7 3.32 <0.1  *Notes: Values are model estimates (SE) determined from a repeated measures model. Model parameters include the degrees of freedom (DF), F-statistic (F), and significance value (p) associated with experimental condition and the condition-ln(z) interaction. Significance codes next to each parameter indicate the p-value of the comparison with the reference condition (20%SHF): ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. ‘n’ is the number of experimental replicates. See Table 2.1 for experimental code names and descriptions. Velocity profiles from the three bed locations over time for the third replicate reference experiment (Figures 2.5a, b, and c) demonstrate that there was no consistent temporal trend in velocity as concentration decreased; neither of the other two replicate experiments exhibited a temporal trend in velocity profiles. Profiles from a single measurement time during the experiment (0h) for the three locations (Figure 2.5d) show the spatial variability in velocity due to differences in local bed configurations; all reference condition replicates had a similar degree of variability among locations. Spatially-varying bed conditions within all experimental runs arose from the non-uniform sediment mixture that was randomly placed; although the bulk particle size distribution was the same across the length of the flume, small-scale topographic variations such as protruding stones or clusters altered local hydraulic conditions. Furthermore, the removal, remixing, and replacement of the  33  sediment reconfigured bed conditions for each experiment; thus the comparisons between experiments include a degree of local-scale variability that is independent of changes to ‘bulk’ physical conditions (e.g. bed composition, flow rate, particle density). Thus differences in hydraulics and deposition among physical conditions may be difficult to detect if there is a large degree of within-condition or within-experiment variability arising from these bed effects. Near-bed hydraulic conditions are dominated by local effects while upper flow regions integrate upstream and local conditions. Evidence for local-scale effects is seen in the higher variability of near-bed hydraulic metrics within experimental runs relative to metrics representing the upper region of flow. Profiles measured at different bed locations (2.5, 3, and 3.5 m) were compared to assess small-scale spatial variability within runs. Coefficients of variation (CV) among the three locations are higher for near-bed velocities and Reynolds stresses than maximum velocities recorded at the top of the profile. For near-bed velocity, CV ranges from 0.062.2 (mean = 0.82); for near-bed Reynolds stress, CV ranges from 0.2-26 (mean = 3.7); and for maximum velocity, CV ranges from 0.02-0.76 (mean = 0.19). All physical experiments exhibit higher variability in the near-bed velocity and near-bed Reynolds stress than the maximum velocity, indicating that local-scale bed effects lessen with height above the bed. I acknowledge that these local-scale bed effects introduce a degree of error to the measurements that could confound comparisons among conditions. However, this variability is inherent to natural streams with heterogeneous substrates; thus the effects of changing bulk conditions such as flow rate or particle concentration will vary depending on existing bed configurations. By measuring at more than one location and replicating experiments within conditions, I attempted to incorporate a range of possible configurations and detect effects larger than this variability. I recognize, however, that the error may be too large to identify statistically significant differences. For this reason, my analysis will highlight apparent trends in the data that may not be statistically significant but are relevant to my hypotheses.  34  Figure 2.5: Log-normalized velocity profiles from the third replicate reference experiment at the a) 2.5 m, b) 3 m, and c) 3.5 m locations of the streambed at three measurement times, as well as at d) the 0 h measurement time at all three locations. 2  a) 2.5 m  b) 3 m  ln(z ) (z in cm)  1  0  -1  0h  0h  4h  4h  8h  8h  2.5 m  3m  -2  2  c) 3.5 m  d) 0 h  ln(z ) (z in cm)  1  0 0h  -1  4h  3.5 m  8h  2.5 m  3.5 m  3m  -2 -10  10  30 u (cm/s)  50  70  -10  10  30 u (cm/s)  50  70  *Notes: Solid black lines are the linear least-squares regression lines fit to all measurement points at each location, treating velocity (u) as a function of log-normalized height (ln(z) where z is height above the bed). Reynolds stress profiles from the same replicate experiment, locations, and times (Figure 2.6) also show no consistent temporal change in turbulence and the large spatial variability between locations. Also notable at two of the locations (3 and 3.5 m) is the distinct shift in slope partway up the profile and the decrease in Reynolds stress towards the bed (Figures 2.6 b and c) characteristic of shear stress profiles above rough beds (e.g. Nikora and Goring, 2002). The height of maximum shear stress is often considered the top of the inner boundary or roughness layer (Nikora et al., 2002). In this experiment, only the 2.5 m location does not show this characteristic shape, perhaps because the flow was simply too shallow in this location. Since the ADV cannot measure within the top 5 cm of flow, the full velocity profile cannot be measured thus velocity measurements were likely restricted to the inner boundary layer. Alternatively, shallow flow may have simply limited the formation of an ambient boundary layer, above the point of maximum shear stress , thus Reynolds stresses may continue to increase with height above the bed.  35  Figure 2.6: Reynolds stress (τRe) profiles from the third replicate reference experiment at the a) 2.5 m, b) 3 m, and c) 3.5 m locations of the streambed at three measurement times, as well as at the d) 0 h measurement time at all three locations. 5  a) 2.5 m  0h  b) 3 m  0h  4h  4h  4  8h  z (cm)  8h  3m  2.5 m  3 2 1 0  5  0h  c) 3.5 m  3.5 m d) 0 h  4h 4  2.5 m  8h  3m  z (cm )  3.5 m 3 2 1 0 -2  -1  0  1  τRe (Pa)  2  3  4  5  -2  -1  0  1  2  3  4  5  τRe (Pa)  *Notes: Solid black lines are the lowess smooth fit lines fit to all measurement points at each location, treating Reynolds stress (τRe) as a function of height (z). Particle deposition rates  Measured particle concentrations decreased continuously from 0 to 8h, showing a roughly exponential decline (Figure 2.7). Log-transformed concentration does not show a perfectly linear relation with time for the 0-1h period thus residual plots show a distinct trend. Non-linearity is apparent in plots of log-transformed and normalized concentration against time for the 0-1h periods (Figure 2.7a). Additional terms in the 0-1h model did not significantly improve model fit thus an exponential model was considered an appropriate approximation of the relation between concentration and time. Furthermore, from 1-8h, transformed data fit a more linear trend (Figure 2.7b). Because of the rapid decrease in deposition rate from the early (0-1h) to the late (1-8h) part of the experiment, regression lines for the full eight hour period did not fit as well as lines for the two time periods fit individually. Also noticeable from the plots is that between-run differences are much more pronounced in the 0-  36  1h period, as seen by the separate trends in measurement points in the replicated experiments, particularly the 80% sand and plastic conditions (Figure 2.7a). Given the differences in fit between the two time periods, I conducted separated analyses for the 0-1h and 1-8h periods to provide a more accurate representation of deposition rate. Model parameters from the repeated measures model, including decay rate (k) (regression slope) and estimates of initial concentration (C0), as well as deposition parameters (wd, ws and Ed) of both time periods for the reference condition and all other experiments are provided in Table 2.4. Median particle size (D50) used in the calculation of ws is the measured particle size of the suspension when the concentration equals C0. Also provided is the p-value for statistical comparisons of k between the reference condition and each of the other experiments; these differences will be discussed in subsequent sections.  37  Figure 2.7: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for six physical conditions.  38  Table 2.4: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h. Condition n k (x 10-5 s-1) C0 (µl/l) D50 (µm) wd (cm h-1) ws (cm h-1) Ed 0-1h 20%SHF 3 24.1(0.28) 158.7(1.2) 14.0(0.6) 8.91(0.10) 100.8 0.09 20%SLF 3 20.8(0.28)**** 165.2(1.2) 14.0(1.0) 5.91(0.80) 101.5 0.06 20%PHF 3 5.87(0.28)**** 109.8(1.2) 36.3(1.1) 2.16(0.10) 324.7 0.01 20%SHFHC 1 27.4(0.48)**** 278.7(2.1) 15.0(4.0) 8.87(0.16) 56.8 0.16 20%SHFLC 1 20.7(0.48)**** 58.4(2.1)** 11.6 7.45(0.17) 34.1 0.22 80%SHF 2 16.1(0.34)**** 108.6(1.5) 11.5 7.46(0.16) 37.4 0.20 DF 2803 F 577.07 P <0.001 1-8h 20%SHF 3 3.67(0.71) 78.8(1.2) 6.8(0.2) 1.35(0.3) 24.2 0.06 20%SLF 3 4.33(0.71) 90.85(1.2) 8.5(1.4) 1.23(0.2) 39.4 0.03 20%PHF 3 5.93(0.71)** 86.28(1.2) 20.3(1.7) 2.18(0.3) 105.0 0.02 20%SHFHC 1 10.6(1.23)*** 261.9(1.9)*** 12.4(0.7) 3.46(0.4) 38.9 0.09 20%SHFLC 1 4.13(1.23) 32.85(1.9)** 13.4 1.48(0.4) 45.4 0.03 80%SHF 2 4.77(0.87) 91.95(1.4) 9.15 2.21(0.4) 21.3 0.10 DF 84 F 6.13 P <0.001 *Notes: Decay rate (k) was determined from the slope of the regression line (log-transformed concentration vs. time) fit to each condition with a repeated measures model. ‘n’ is the number of experimental replicates. Values in parentheses are standard errors of the model estimates (k, C0) or the means of experimental replicates, when available. Initial concentration C0 was determined from the intercept of the model fit to the raw data and was used to normalize measured concentrations (C/C0). D50 is the median particle size measured by the LISST at the start of each experiment, averaged for the experimental replicates. wd is the depositional velocity, computed from the decay rate and the average depth of the experimental replicates; ws is settling velocity of the median particle size; Ed is the enhancement factor. Regressions fit through normalized concentrations vs. time were compared to the reference condition (20%SHF) with a repeated measures model. Significance codes next to each parameter indicate the p-value of the comparison with the reference condition: ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. See Table 2.1 for experiment code names and descriptions. Also provided are the degrees of freedom (DF), F-statistic (F), and significance value (p) associated with the k estimate for the whole model. Plots of the decline in the near-bed concentration for the four different particle size classes for the three replicate reference condition experiments from 0-8h are shown in Figure 2.8. For each particle size class, I determined the decay rate (k) by fitting a linear model to the decline in log-normalized concentration from 0-1 and 1-8h. In this case, I did not use a repeated measures model because I was fitting separate models to each experimental replicate and not comparing between conditions or replicates. Like the total suspension, models fit to the two time periods separately had higher r2 values. Model parameters for each of the particle size classes, including k, depositional velocity (wd), initial concentration (C0), and coefficient of determination (r2) for the three replicate reference experiments are provided in Table 2.5; I have also calculated the settling velocity (ws) and enhancement factor (Ed) based on the mean particle size of each size class.  39  Figure 2.8: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for the three replicate reference condition experiments from 0-8h; the shaded region is the 0-1h period. 300  c) 20%SHF3  b) 20%SHF2  a) 20%SHF1 T otal  250  2-4 microns 4-16 microns  200 C (ppm)  16-63 microns  150  63-122 microns  100  50  0 0  10000 20000 Seconds  30000  0  10000 20000 Seconds  30000  0  10000 20000 Seconds  30000  *Notes: Size classes, from the uppermost line moving downward on the plot at the y-axis, are: total concentration (all classes), ~2-4 µm, ~4-16 µm, ~16-63 µm, and ~63-122 µm. Values from the three replicates are similar in magnitude and show the same trends. All particle sizes at both time periods have enhancement factors less than 1 (Ed < 1 or ws > wd), indicating that the actual deposition velocity of particles was slower than expected under the force of gravity. High turbulence intensity and upwarddirected velocities due to high bed roughness must have acted to reduce the actual rate of deposition. As expected, larger particle sizes have higher decay rates (k) and faster depositional velocities (wd) during both time periods, due to higher settling velocities (ws). However, the increase in wd is not proportional to the increase in ws; particles in the 63-122 µm size class have a settling velocity more than 700 times faster than particles in the 2-4 µm class, but depositional velocities less than 100 times faster. Thus the smallest particles have higher enhancement factors than larger particles (though still less than 1), indicating that depositional and settling velocities are closer in value and suggesting that these particles were either 1) less affected by upwarddirected flow at the bed surface or 2) more influenced by processes promoting deposition, infiltration ,and retention than large particles. I expect the latter is more likely, because there is no mechanistic explanation for why small particles would be resuspended from the bed surface less easily than large particles. Rather, it is logical to expect that the advection and passive percolation of fine particles in the bed would occur more readily, as they would fit into smaller pore spaces. Once retained below the surface, these particles would be entrained less frequently, increasing the rate at which they are lost from the water column.  40  Table 2.5: Model and depositional parameters from 0-1h and 1-8h for four different particle size classes (in µm) from the three replicate reference condition experiments (20%SHF1, 2, and 3). Range D50 k (10-5 s-1) C0 (µl/l) r2 wd(cm h-1) ws(cm h-1) Ed Total 14.0 25.47(0.62) 148.7 0.89 9.48 100.5 0.09 2-4 3.6 9.09 (0.29) 53.57 0.83 3.38 7.20 0.47 4-16 10.6 20.39(0.30) 53.62 0.96 7.59 57.6 0.13 16-63 40.0 95.74(1.61) 48.62 0.94 35.62 781.2 0.05 63-122 92.8 899.9(41.7) 1.40 0.93 334.76 3664.8 0.09 20%SHF2 Total 13.0 24.97(0.76) 152.9 0.84 8.93 86.8 0.10 2-4 3.6 9.22(0.29) 57.97 0.82 3.29 7.20 0.46 4-16 10.6 20.76 (0.69) 55.26 0.81 7.42 57.6 0.13 16-63 40.0 110.0(2.81) 50.20 0.88 39.32 781.2 0.05 63-122 92.8 1197.8(29.9) 0.97 0.98 428.18 3664.8 0.12 20%SHF3 Total 15.0 22.15(0.45) 176.6 0.92 8.44 115.2 0.07 2-4 3.6 8.24(0.20) 61.56 0.88 3.14 7.20 0.44 4-16 10.6 15.43(0.19) 59.38 0.97 5.88 57.6 0.10 16-63 40.0 71.20(0.82) 63.24 0.97 27.13 781.2 0.03 63-122 92.8 601.4(10.5) 2.74 0.98 229.16 3664.8 0.06 C0 (µl/l) r2 wd(cm h-1) ws(cm h-1) Ed 1-8h Range D50 k (10-4 s-1) 0.94 20%SHF1 Total 7.0 3.80(0.37) 76.02 1.41 25.3 0.06 2-4 3.6 2.17(0.13) 44.12 0.98 0.81 7.20 0.11 4-16 10.6 8.51(0.77) 34.74 0.95 3.17 57.6 0.11 16-63 40.0 17.9(6.08) 1.83 0.52 6.67 781.2 0.12 63-122 92.8 NA NA NA NA 3664.8 NA 20%SHF2 Total 6.5 3.88(0.36) 76.86 0.95 1.39 21.9 0.06 2-4 3.6 2.26(0.14) 46.62 0.97 0.81 7.20 0.05 4-16 10.6 10.0(0.62) 36.31 0.97 3.60 57.6 0.06 16-63 40.0 16.59(4.32) 1.77 0.70 5.93 781.2 0.04 63-122 92.8 NA NA NA NA 3664.8 NA 20%SHF3 Total 7.0 3.28(0.38) 85.20 0.62 1.25 25.3 0.05 2-4 3.6 2.25(0.19) 50.35 0.95 0.86 7.20 0.01 4-16 10.6 6.54(0.63) 37.64 0.94 2.49 57.6 0.01 16-63 40.0 9.45(2.20) 10.4 0.71 3.60 781.2 0.01 63-122 92.8 NA NA NA NA 3664.8 NA *Notes: Values in parentheses are the standard error of the regression. Model parameters include decay rate (k), initial concentration (C0), depositional velocity (wd), and coefficient of determination (r2). Initial concentrations (C0) are the predicted concentration at time t = 0 from exponential models fit to the concentrations measured during each time period. Theoretical particle sizes were used for computation of the settling velocity (ws) and enhancement factor (Ed) for the four particle size classes. For the total distribution, ws and Ed were calculated from the particle size measured for the initial concentration (D50). NA indicates that no particles in that size class were measured in the suspension. 0-1h 20%SHF1  As noted previously, decay rates for the total suspension decrease from the early to the late part of the experiment, due to the rapid loss of coarse particles from the water column. Due to the shift in the proportion of particles in each size class, the decay rate, depositional velocity, and enhancement factor of the total suspension are closer in value to the 4-16 µm size class in the early part of the experiment and the 2-4 µm size class in the late part of the experiment. In other words, the dominance of the 2-4 µm size class in suspension resulted in a slower decay rate for the total suspension from 1-8h.  41  Stratified subsurface samples FIne particle infiltration measured in stratified subsurface samples for the three replicate reference experiments are plotted in Figure 2.9; for illustration purposes only percentages were averaged for the three sections although these three locations are simply repeated measures within the runs and not true replicates. Error bars are the standard deviation of the three bed locations, presented solely to demonstrate the degree of within-run variability. Figure 2.9: Percentage of particles from the surface to the flume bottom for the three replicate reference experiments (20%SHF1, 2, and 3).  Percent fine particles (<125 µm) 0.00  0.05  0.1  0.15  0.20  0.25  0 20%SHF1 -1  Depth (cm)  -2  20%SHF2 20%SHF3  -3 -4 -5 -6 -7 -8  *Notes: Depths representing the middle of each layer have been offset vertically so that error bars can be distinguished; true depths are the center of each layer. Dotted lines separate layers. Error bars are the standard deviation of three streambed locations within each run. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer. Samples show some variability within runs, especially in the subsurface and base layers (3 and 7 cm depths), but generally the percentage of particles increases with depth, reflecting the percolation of particles through the highly porous bed over time. Particle mass in the armour layer is low, perhaps because particles were not well-retained at the surface and downward particle movement was relatively rapid. In these experiments, the 20% sand bed remained static thus kinematic sieving would not have occurred and all particle infiltration would have been due to passive percolation or advective transport as flow moved through the bed.  42  2.4.2 FLOW RATE  As hypothesized, reducing the bulk flow rate from ~8.5 to ~4.2 L/s (high flow to low flow) altered local hydraulic conditions, particle deposition rates, and infiltration amounts for the 20% sand bed, high-density silica particles, and moderate initial concentration of the reference condition. In comparison with the reference condition, depth-integrated velocities, near-bed velocities and maximum streamwise velocities were significantly slower in the low flow experiments (Table 2.2; p < 0.05), as predicted. Furthermore, maximum Reynolds stress, higher in the water column, was significantly less for the low flow condition than the reference condition (Table 2.2; p < 0.05). However, near-bed Reynolds stresses were not significantly different between the low flow and reference condition, indicating that near-bed turbulence conditions were more affected by local bed conditions than bulk flow rate. Local roughness effects influence the intensity of turbulent fluctuations near the bed and thus may override bulk flow conditions, resulting in similar near-bed Reynolds stresses for high and low discharges. A comparison of velocity profiles for the low flow and reference case show a similar result (Table 2.3; Figure 2.10a and b); as predicted, velocities for the lower discharge were slower throughout the water column, resulting in a significantly lower slope and intercept of the velocity regression (p < 0.05). Plots in Figure 2.10 present measurement points from all locations, heights, and runs for a given condition, as well as the regression line fit to the data with the repeated measures model. All conditions are presented on the same plot to facilitate comparisons.  43  Figure 2.10: Streamwise velocity (u) as a function of log-normalized height above the bed (ln(z)) for six physical conditions.  u (cm/s)  60  ln(z) (z in cm) 0.5 1.0 1.5 2.0 -1.0 -0.5 0 0.5 1.0 1.5 2.0 -1.0 -0.5 0  a) 20%SHF  b) 20%SLF  0.5 1.0 1.5 2.0  c) 20%PHF  60  40  40  20  20  0  0  60  e) 20%SHFLC  d) 20%SHFHC  f) 80%SHF  60  40  40  20  20  0  0 -1.0 -0.5 0  0.5 1.0 1.5 2.0 -1.0 -0.5 0  0.5 1.0 1.5 2.0 -1.0 -0.5 0  u (cm/s)  -1.0 -0.5 0  0.5 1.0 1.5 2.0  ln(z) (z in cm)  *Notes: See Table 2.1 for description of experimental code names; ‘20%SHF’ is the reference condition.  According to the repeated measures test, decay rate (k) for the low flow condition was significantly less than the reference condition for the early 0-1h period, but not for the late 1-8h period (Table 2.4; p > 0.1). Thus it appears that the reduction in flow rate from ~8.5 to 4.2 L/s actually reduced the rate of particle deposition in the early part of the experiment, counter to my predictions. Later in the experiment (from 1-8h), however, decay rate of the low flow condition was slightly higher, but the difference was not significant. Despite slower velocities, it seems that the reduction in maximum Reynolds stress slowed deposition at low flow early in the experiment when coarse particle (>16 µm) deposition was dominant. Coarse particle deposition was likely enhanced in the reference condition by high vertical velocity fluctuations. Later in the experiment, when fine particle deposition dominated, similar bed structures and near-bed Reynolds stresses in the reference and low flow conditions may have resulted in similar rates of deposition. A plot comparing decay rates for the 0-1h and 1-8h periods from the reference condition, low flow condition, and all other experimental conditions also shows the relation with the average near-bed and maximum Reynolds stress (Figure 2.11). Regressions fit to the exponential decline in concentration over time for each condition are plotted in Figure 2.7a and b for the 0-1h and 1-8h periods, respectively.  44  Figure 2.11: Average decay rates (k) of the 0-1h and 1-8h time periods and the average near-bed (τRe0) and maximum Reynolds stress (τReM) for each experimental condition. 3.5  0-1h  1-8h  τRe0  4  τReM  Experiment codes:  3.0 2.5 2  2.0 1.5  1  1.0  Reynolds stress (Pa)  Decay rate (k; x 10-4 s-1)  3  20%SLF = Low flow rate 20%SLC = Low initial concentration 20%SHF = Reference condition 20%SHC = High initial concentration 20%PHF = Plastic particle dose 80%SHF = 80% sand fraction  0 0.5 0.0  -1 20%SHF  20%SLF  20%PHF  20%SHFHC 20%SHFLC  80%SHF  *Notes: Values and standard errors (error bars) were estimated from a repeated measures model assuming equal variance among conditions. Decay rates for individual size classes differed, providing insight into the processes of deposition. From 01h, all particle sizes except the smallest (2-4 µm) deposited at a significantly lower rate at low flow than in the reference condition. From 1-8h, particles < 16 µm deposited at statistically similar rates in the low flow and reference conditions, but particles > 16 µm deposited at significantly lower rates in the low flow condition. Slower deposition of large sizes at low flow suggests that these particles may require strong advective forces for large amounts of infiltration, and hence net deposition, to occur. As noted above, significantly reduced maximum Reynolds stresses at low flow may have limited the deposition of large particles, explaining why deposition was slower at low flow early in the experiment only. Slower deposition at low flow is reflected in the base layer of subsurface samples; according to the repeated measures model, the percentage of fine particles in the base layer is significantly less for the low flow condition than for the reference condition (p < 0.1). As seen in Figure 2.12a, however, fine particle mass is the same for low flow and the reference condition in all other layers of the bed.  45  Figure 2.12: Percentage of particles <125 µm in each of four bed layers taken from the surface to the flume bottom for the reference condition and a) a low flow rate, b) an 80% sand bed, c) a plastic particle suspension, and d) low and high initial concentration.  Percent fine particles < 125 µm -0.1  0.1  0.3  0.5  0.7  -0.3  -0.1  0.1  0.3  0.5  0.7  0 -1 -2 -3  20%SHF (Reference) 20%SLF 20%PHF 80%SHF 20%SHFHC 20%SHFLC  -4 -5 -6  Depth (cm)  -7 a)  b)  c)  d)  -8 0 -1 -2 -3 -4 -5 -6 -7 -8  *Notes: Values and standard errors (error bars) were estimated from a repeated measures model assuming equal variance among conditions. Depths representing the middle of each layer have been offset vertically so that error bars can be distinguished; true depths are those aligned with the reference condition (20%SHF). Dotted lines separate layers. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer.  46  2.4.3 BED SAND FRACTION  Observations and experimental complications  Because of a high sand fraction, the beds of the two 80% sand experiments were mobile, even under low flow rates. In the first experiment (80%SHF1), despite attempts to gradually saturate the bed and incrementally increase the flow rate, the initial front of water entering the flume produced a sharp increase in shear stress that mobilised the bed surface. At low discharge, flow was shallow and fast, rapidly moving the bed as a thin uniform layer (~ 0.5 cm thick); scour was visible around stable and static coarse grains. As I increased the flow rate, the flow deepened and the rapid uniform bed movement subsided. However, three migrating dunes began forming at the head of the flume, each ~0.25 m long and ~0.5 cm high at their front. Observation of the water surface profile clearly indicated that flow was non-uniform under the flow conditions imposed in the 20% sand experiments, likely because of the hydrodynamically smoother bed. Movement of the dunes was most rapid at the head of the flume (~0.08 cm/s) where flow was shallowest and gradually slowed as the dunes migrated into deeper water at the downstream end of the flume. During the first thirty minutes of flow, prior to the particle dose or any concentration measurements, the dunes migrated to the far end of the test section (~3.75 m). At the time of the particle dose (one hour of flow, ‘0h’ of the experiment), the dunes had migrated and virtually stalled at the 5 m location. During the period of continuous near-bed measurements (0-1h), bed movement through the test section was still visible. However, after two hours of flow, at the end of the first hour of concentration measurements, the test section appeared virtually immobile, the bed had degraded by 0.5-3 cm, and coarse grains were clearly visible on the surface. Thus all concentration profile measurements taken from 1-8h were over a stable and coarser-grained bed surface than the first hour of continuous near-bed measurements. After four hours of flow, the migrating dunes reached the end of the flume and stopped moving. I estimated the volume of sediment moved over the course of the experiment to be ~5000 cm3. Due to the scour and subsequent coarsening of the bed surface, I observed systematic changes in velocity profiles between those taken before and after the bed stabilized. Velocities measured at 3 and 3.5 m during bed movement were faster throughout the flow depth prior to bed stabilization and decreased steadily with each subsequent velocity measurement (Figure 2.13); no difference in velocity profiles was observed at the 2.5 m location. This observation suggests that the gradual exposure of coarse grains at the surface increased roughness, dissipating flow energy and slowing streamwise velocities.  47  Figure 2.13: Log-normalized velocity (u) profiles at the a) 3 m and b) 3.5 m locations at four different times during the first 80% sand replicate experiment (80%SHF1). 6  b) 3.5m  a) 3 m 5 ln(z) (z in cm)  -1h 4 3  -1h  0h  0h  4h  4h  8h  8h  2 1 0 0  10  20  30 u (cm/s)  40  50  60  0  10  20  30 u (cm/s)  40  50  60  *Notes: ‘-1h’ is the hour prior to the particle dose; subsequent measurements occurred at the time of the particle dose (‘0h’) and at 4 and 8 hours after the dose. In the second 80% sand experiment, I attempted to prevent the formation of bed forms and sediment transport. To do this, I gradually saturated the bed and filled the flume with water at the lowest possible discharge, closing the end gate completely. Although the initial shallow wave of flow mobilized some sand along the surface, movement stopped when the flow depth reached ~3 cm. Once the flow was as deep as possible (~30 cm), I slowly opened the gate and increased the flow rate while monitoring the bed for movement, attempting to prevent any drastic changes in shear stress. Doing so, I was able to prevent significant bed movement while establishing flow conditions. Unfortunately, as in the first experiment, flow at the head of the flume was much shallower than at the downstream end and a small, slowly migrating dune formed in the upper reaches. Movement was at first relatively rapid (~0.03 cm/s) but after two hours of flow, dropped to only ~0.005 cm/s; by this point the dune was still upstream of the test section (at ~2 m). Furthermore, by the end of the eight hour experiment, the dune had only progressed to the 2.8 m location. As a result, almost all velocity measurements and all concentration measurements were taken above a stable or minimally moving bed. Only the 8h ADV measurements at the 2.5 m location were taken over the migrating dune. As a result, I observed only a slight difference in the velocity profile near the bed. Furthermore, I did not observe the scour, degradation, and coarse grain exposure that occurred in the first experiment thus there are no systematic or significant differences in velocity profiles over time at any of the three locations. Although I recognize that the issues of bed movement and non-uniform flow potentially confound any conclusions I can draw from the data, the results from the concentration measurements and bed samples are intriguing and merit discussion.  48  Comparison with reference condition  I predicted that an increase in the bed sand fraction would reduce near-bed turbulence intensities and enhance surface deposition, but reduce particle trapping, infiltration and deposition rate. Results generally support these predictions. Both 80% experiments were considered replicates in the repeated measures model, despite the differences described above, to determine whether the effect of the change in sand fraction was detectable despite the effects of bed movement and non-uniform flow. From 0-1h, the decay rate (k) of the 80% sand condition was significantly slower than the reference condition, as predicted (Table 2.4; p < 0.001). From 1-8h, however, decay rates of the 80% sand and reference conditions were not significantly different. Slower deposition in the early part of the 80% sand experiment can be explained by either a change in hydraulics or physical trapping effects from the dominantly gravel-bedded reference condition. As predicted, near-bed velocity was higher and Reynolds stresses were lower for the 80% sand bed than for the reference condition, but only maximum Reynolds stresses were significantly lower (p < 0.001). For near-bed metrics, the variability within each experimental condition due to bed effects may have been too high to detect a significant difference. Slope of the 80% sand velocity regression was also significantly less than the reference velocity regression (Figure 2.10f; Table 2.3), due to a more uniform velocity profile. Faster near-bed velocities and lower Reynolds stresses for the 80% sand condition indicate smoother hydrodynamic conditions, a result of fewer large surface grains and lower roughness. A reduction in near-bed Reynolds stress could enhance surface deposition, but advective forces that move water and particles into the bed should also be less, decreasing the depth and degree of particle infiltration. Stratified bed samples provide some evidence for decreased infiltration and increased surface deposition (Figure 2.12b), supporting my predictions. Deposition of particles <125 µm to the 80% bed was significantly higher than the reference condition in the two uppermost armour and subsurface layers (p < 0.05; Figure 2.12b). Below the subsurface layer, infiltration into the 80% sand bed decreased considerably and was not significantly different from the reference condition. A reduction in fine particle infiltration may have been due to reduced advective forces or the clogging of pore spaces by fine sand particles.  2.4.4 PARTICLE DENSITY  As predicted, replacing the high-density silica particles of the reference condition with low-density plastic particles had a large effect on particle decay rate, but no discernible effect on local flow hydraulics. None of the mean hydraulic parameters (Table 2.2) was significantly different between the plastic and the reference condition. Furthermore, the regressions of the velocity profiles were not significantly different between the plastic and the reference condition. Contrary to my predictions, however, the depth and amount of infiltration did not differ between plastic and silica particles (Figure 2.12c).  49  Decay rates for the plastic condition during both time periods were significantly different from the reference condition. For the early part of the experiment (0-1h), the k value for the total distribution of plastic particles was significantly lower than for the reference condition (Table 2.4; p < 0.001), as predicted. In fact, the plastic decay rate was approximately one-quarter the decay rate of reference experiments (Figure 2.11; Table 2.4). Given that hydraulic and bed conditions were the same between the plastic and reference conditions, slower plastic decay rates were undoubtedly due to the lower settling velocities of the low-density plastic particles. All particle sizes showed the same trend as the total suspension, depositing at a rate significantly lower than silica particles of the same size. However, this trend showed an unexpected reversal in the later part of the experiment (1-8h; Figure 2.11). From 1-8h, deposition of plastic particles was actually significantly faster than the reference condition (Table 2.4; p < 0.05). Whereas decay rates for the silica particles decreased rapidly after the first hour of flow, plastic decay rates remained virtually constant throughout the experiment, explaining the reversal. Silica particles deposited more slowly late in the experiment because of the rapid loss of coarse particles from the water column. In contrast, large-sized plastic particles deposited much more slowly, remaining in suspension for longer and maintaining a fairly constant total decay rate (Figure 2.14). In fact, only plastic particles 63-122 µm were still in suspension after one hour of flow; silica particles in this size class completely dropped out of suspension within one hour. Meanwhile, plastic particles sized 2-63 µm actually deposited slower than silica particles in the 1-8h period, but the presence of particles in the 63-122 µm size class resulted in a faster total decay rate for the plastic suspension.  Figure 2.14: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for the first replicate reference condition experiment (20%SHF1) and the first replicate plastic particle experiment (20%PHF1) from 0-8h; the shaded region is the 0-1h period.  300  a) 20%SHF1  b) 20%PHF1  Total  250  2-4 microns 4-16 microns  C (ppm)  200  16-63 microns 150  63-122 microns  100  50  0 0  10000 20000 Seconds  30000  0  10000  20000  30000  Seconds  *Notes: Size classes, from the uppermost line moving downward on the plot at the y-axis, are: total concentration (all classes), ~2-4 µm, ~4-16 µm, ~16-63 µm, and ~63-122 µm.  50  Differences in decay rate did not correspond with different infiltration amounts in the plastic particle and reference conditions, however (Figure 2.12c). Despite slower decay rates in the early part of the experiment, the plastic stratified samples did not have significantly fewer fine particles (<125 µm) in any of the bed layers than the reference samples; the percentage of fine particles in each layer was virtually the same for the plastic and reference condition. Because the bed samples reflect a cumulative amount of particle deposition and infiltration (i.e. they were collected eight hours after the particle dose), the slowing of particle deposition (and hence infiltration) in the late part of the reference experiments may have offset the rapid deposition and infiltration that occurred early in the experiment, so that final infiltration amounts were similar between plastic and reference conditions.  2.4.5 INITIAL PARTICLE CONCENTRATION I predicted that an increase in particle concentration would alter turbulence characteristics through its effects on flow resistance, but the direction of effect (damping or intensifying) was not clear. A change in concentration was also predicted to influence deposition and infiltration by modifying grain-grain interactions and particle availability. Results are suggestive, but not conclusive. None of the mean hydraulic parameters was significantly different between the high or low concentration and reference conditions, nor were the velocity regressions significantly different, but both the near-bed velocity and Reynolds stress in the high concentration experiment were notably lower than in both the low concentration and reference experiments (Table 2.2). As noted previously, these near-bed measurements exhibit a high degree of variability due to local bed effects, possibly explaining why the means are not significantly different among conditions. Nevertheless, the trends suggest that high concentrations of suspended particles near the bed may have slightly damped turbulence intensities, as has been previously observed in both laboratory and field settings (e.g. Vanoni, 1946; Einstein and Chien, 1954; Einstein and Chien, 1955; Nordin, 1964; Jordan, 1965; Scott and Stephens, 1966; Muste and Patel, 1997). Vertical density stratification induced by the concentration has been proposed as one mechanism of turbulence damping, in which a buoyancy force created by the density gradient inhibits the turbulent mixing of fluid and sediment of different densities (e.g. Vanoni, 1975). Studies have also shown that large particles in transport can increase fluid turbulence via eddy shedding (Best et al., 1997; Garcia-Aragon et al., 2002) thus the settlement of these large particles decreases the turbulence of the flow. Others have suggested that the observed velocity lag between particles and fluid (a.k.a. ‘particle slip’), due to the resistance to particle motion and the drag force exerted by neighbouring particles, causes turbulence dissipation (Barenblatt, 1953; Czernuszenko, 1998). However, more recent research has shown that while turbulence attenuation in the presence of particles occurs in the outer region of flow, turbulence is enhanced near the boundary (z/h < 0.2-0.3). Relative particle and fluid velocities are also different: particles move faster than the fluid near the boundary and slower in the outer region, due to the fact that particles in the outer region are lifted up by low-momentum fluid ejections (bursts) and particles in the inner region are carried downward by injections (sweeps) that retain their momentum and cause them to be faster than the near-bed fluid. Momentum exchange between the fluid and solid (particle) phases can result in the loss of turbulent energy in the outer region. Sand-sized particles have  51  also been observed to attenuate turbulence in the outer region of flow, where the scale of the grains is smaller than the eddy scale, and enhance turbulence near the bed where particles are larger relative to the eddy scale. In the present study, the near-bed Reynolds stress was typically measured at 0.3-0.6 cm, or z/H~0.05, which is well within the inner region. Near-bed Reynolds stress of the high concentration experiment was decreased, but the maximum Reynolds stress, in the outer region, was not; thus this does not correspond to the velocity lag/momentum exchange mechanism. Density stratification is also unlikely, because this mechanism results in turbulence reduction throughout the water column and an increased velocity gradient in the velocity profile, neither of which were observed (Figures 2.9 and 2.15). Potential effects of density stratification on both the velocity and concentration profiles will be discussed further in Chapter 3. Possibly, energy may have been lost in the transport of particles, which were in higher concentrations near the bed, reducing turbulent fluctuations (Guo and Julien, 2001) and resulting in a low near-bed velocity and low near-bed Reynolds stress.  Figure 2.15: Measured Reynolds stress (τRe) profiles with corresponding lowess model fits (solid and dashed lines) from three streambed locations of the high concentration experiment (20%SHFHC) and the three replicate experiments of the reference condition (20%SHF1, 2, and 3).  5  a) 20%SHFHC  2.5 m 3m 3.5 m  2.5 m 3m 3.5 m  b) 20%SHF1  z (cm)  4 3 2 1 0  5  d) 20%SHF3  c) 20%SHF2  z (cm)  4 3 2 1 0 0  1  2 3 4 Reynolds stress (Pa)  5  6  0  1  2 3 4 Reynolds stress (Pa)  5  6  As predicted, low initial concentrations resulted in significantly slower deposition rates than the reference condition, but only for the 0-1h period; deposition rates were not significantly different during the 1-8h period. High concentration resulted in significantly faster deposition during both time periods, however (p < 0.001).  52  Thus results clearly show that the rate of particle deposition is directly and positively related to initial concentration. Unlike the plastic, low flow, and low concentration experiments, the high concentration experiment showed the same pattern in decay rate for the 0-1h and 1-8h periods. Decay rate of the high concentration experiment was significantly higher than the reference condition during both the early and late parts of the experiment. Higher concentrations persisting later into the experiment may explain the continued rapid deposition of the high concentration experiment. Increased rates of particle deposition for the high concentration experiment cannot be sufficiently explained by a decrease in the near-bed Reynolds stress, because a similarly low near-bed Reynolds stress did not increase decay rate for the low flow condition. Instead, particles may have been directly accelerating the deposition of neighbouring particles through particle contact and collision. At high concentrations, depositing large particles may have swept smaller particles towards the bed, increasing total deposition rates. High concentrations may have also increased the frequency of collisions between suspended particles and the bed, increasing trapping and deposition rate. As predicted, stratified bed samples show a distinct trend in infiltration with suspended particle concentration (Figure 2.12d). Although fine particle mass in the armour layer was relatively low for the concentration and reference experiments, from the subsurface to the base layer the high concentration experiment had the highest and the low concentration had the lowest percentage of particles <125 µm. The repeated measures comparison, however, did not detect significant differences between any of the low concentration or high concentration layers with the reference condition layers, possibly due to the high error of the non-replicated conditions. Nevertheless, the trend is clear: suspended particle concentration and particle infiltration are directly related. Combined with the trend in deposition rate, these results show that as suspended particle concentration increased, so did the rate of deposition and the cumulative amount of infiltration. At high concentrations, the frequency of particle interactions and bed trapping was likely high, as was the total number of particles available to infiltrate.  2.4.6 COMPARISONS AMONG PHYSICAL CONDITIONS  Certain experimental conditions were compared with a Student’s t-test using t-values computed by the repeated measures model. The number of pairwise comparisons was limited, however, in order to reduce the risk of a Type I error due to multiple comparisons. Three general comparisons were made that were deemed relevant to the study: 1) deposition rates of low concentration versus low flow; 2) deposition rates of low flow versus plastic; and 3) subsurface samples of 80% sand versus low concentration. Hydraulic parameters for all experiments are provided in Table 2.2; decay rates in Table 2.4 and Figure 2.11; and stratified samples in Figures 2.12 and 2.16. Firstly, I suspected that a change in concentration may potentially have similar effects on hydraulics and particle deposition as a change in flow rate. From 0-1h, the decay rate (k) for the low flow condition was statistically the same as the low concentration experiment (Table 2.4; Figure 2.11) and both were significantly  53  lower than the reference experiment. Slower deposition at low concentrations was likely due to the reduced frequency of particle-particle and particle bed interactions, while slow deposition at low flow is likely due to reduced vertical advective transport of large particles. A lower maximum Reynolds stress at low flow may have limited the degree of eddy penetration and advection into the bed thus slowing total deposition. Any effect of the reduction in maximum Reynolds stress appears to have been counteracted by an increase in particle concentration, however; despite Reynolds stresses similar to the low flow and low concentration conditions, the high concentration experiment had the highest decay rate overall. Secondly, there appears to be an interaction between particle composition, particle size and flow rate. From 0-1h, the full suspension of plastic particles had a notably lower decay rate than the suspension of silica particles under all experimental conditions. Plastic particles 63-122 µm, however, deposited significantly faster than silica particles of the same size at the low flow rate. Thus it appears that a low flow rate slowed the deposition of large particles more than did a low particle density, perhaps because particle size has a more dominant effect on settling velocity than density as particle diameter increases. As seen in the equation for settling velocity, a particle with a low density and a large diameter could have a higher settling velocity than a particle with a high density and small diameter. At low flow, passive settling may have been more important to particle deposition because of the reduction in advective forces identified above. Finally, low concentration conditions reduced particle infiltration more than clogging by sand particles in the 80% sand condition. Deposition of particles <125 µm to the 80% bed was notably greater than the other conditions in the armour and subsurface layers (Figure 2.16). Below the subsurface layer, infiltration into the 80% sand bed decreased considerably, but was still significantly greater than the low concentration condition at the base layer (p < 0.05). Thus the 80% sand condition reduced the depth of fine particle infiltration, but not to the same degree as low particle concentrations.  54  Figure 2.16: Percentage of particles <125 µm in each of four bed layers taken from the surface to the flume bottom for all physical conditions.  Percent fine particles < 125 µm -0.2 0  -0.1  0.0  0.1  0.2  0.3  0.4  0.5  0.6  -1  -2  Depth (cm)  -3  -4  -5  -6  -7  -8  20%SHF (Reference) 20%SLF 20%PHF 80%SHF 20%SHFHC 20%SHFLC  *Notes: Values and standard errors (error bars) were estimated from a repeated measures model assuming equal variance among conditions. Depths representing the middle of each layer have been offset vertically so that error bars can be distinguished; true depths are with the center of each layer. Dotted lines separate layers. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer. 2.5 CONCLUSIONS  A series of flume experiments was used to analyse the effects of the following four physical factors on local flow hydraulics, particle deposition rate, and the amount of particle infiltration: 1) flow rate, 2) bed sand fraction, 3) particle density, and 4) particle concentration. Experiments tested the influence of varying a single factor from a reference condition consisting of a 20% sand bed, high flow rate, and a moderate concentration of high-density particles. Although results are variable and complex, some general conclusions can be made. A reduction in flow lowered near-bed, average and maximum velocities and upper-flow maximum Reynolds stresses, slowed particle deposition rates, and reduced infiltration at depth, but did not significantly  55  alter near-bed Reynolds stresses. Meanwhile, a change in bed composition had a significant effect on deposition rate, infiltration, and local hydraulics. A high bed sand fraction resulted in faster near-bed velocities and significantly lower Reynolds shear stresses than the reference condition, indicating smoother hydrodynamics conditions due to the reduction in surface roughness. Although lower Reynolds stresses might be expected to enhance surface deposition, particle deposition to the 80% sand bed was slower than to the 20% sand bed early in the experiment. Instead it appears that the rate at which particles were lost from the water column to the 80% bed was influenced by the degree and depth of infiltration. Infiltration into the surface and subsurface layers of the 80% bed was significantly greater than into the 20% sand bed, but below the subsurface layer, infiltration into the 80% sand bed decreased considerably, although infiltration at depth was still greater than low flow and low concentration conditions. Infiltration into the 80% bed may have been hindered by sand-clogged pore spaces or a reduction in downward particle transport by turbulent advection. Among experiments on a rough, 20% sand bed, particle density and initial concentration had the greatest effects on particle deposition rates, although the magnitude and direction of effects differed over time and among particle size classes. Early in the experiment, the total suspension of low-density plastic particles and those sized 2-63 µm deposited significantly slower than high-density silica particles; only plastic particles 63122 µm deposited slightly faster than silica particles at low flow, indicating that the importance of particle size and passive settling increased relative to particle density as particle size increased and flow rate decreased. Late in the experiment, however, plastic particles deposited faster than all silica particles for all conditions except the high concentration experiment, due to the persistence of large particles in suspension in both the plastic and high concentration experiments. Despite its effects on deposition rate, particle density had no effect on cumulative particle infiltration or local flow hydraulics. In contrast, particle concentration influenced local flow hydraulics, particle deposition rate, and infiltration amounts. At high concentrations, the full suspension of particles deposited faster than at all other conditions throughout the experiment. Furthermore, infiltration amounts increased from low to high concentrations, following the trend in deposition rate. High concentrations also had lower near-bed velocities and Reynolds stresses, perhaps because of the turbulent energy lost in the transport of particles. Low upward turbulent stresses may be one mechanism explaining the increased deposition and infiltration of high particle concentration. However, these are more likely explained by a higher frequency of particle-particle and particle-bed interactions and greater quantities of particles available to infiltrate, because lower Reynolds stresses at low flow rates did not increase deposition rates or infiltration amounts relative to higher flows.  56  2.6 REFERENCES Agrawal Y. C. and Pottsmith H. C. (2000) Instruments for particle size and settling velocity observations in sediment transport. Marine Geology 168(1-4), 89-114. Agrawal Y. C. and Pottsmith H. C. (2004) LISST laser diffraction sensors advance sediment monitoring - Laser system also measures water's inherent optical properties and volume scattering function. Sea Technology 45(8), 33. Arora A. K., Ranga Raju K. G., and Garde R. J. (1984) Resistance to flow and velocity distribution in rigid boundary channels carrying sediment-laden flow. Water Resources Research 22(6), 943-951. Barenblatt G. I. (1953) On the motion of suspended particles in a turbulent flow. Applied Mathematics and Mechanics 17(3), 261-274. Best J., Bennet S., Bridge J., and Leeder M. (1997) Turbulence modulation and particle velocities over flat sand beds at low transport rates. Journal of Hydraulic Engineering 123(12), 1118–1129. Biron P. M., Robson C., Lapointe M. F., and Gaskin S. J. (2004) Comparing different methods of bed shear stress estimates in simple and complex flow fields. Earth Surface Processes and Landforms 29(11), 1403-1415. Czernuszenko W. (1998) The drift velocity concept for sediment-laden flows. Journal of Hydraulic Engineering 124(10), 1026-1033. Einstein H. A. (1968) Deposition of suspended particles in a gravel bed. Journal of the Hydraulics Division ASCE 94, 1197-1205. Einstein H. A. and Chien N. (1954) Second approximation to the solution of the suspended load theory. MRD Sediment Series No. 3 Einstein H. A. and Chien N. (1955) Effects of heavy sediment concentration near the bed on velocity and sediment distribution. MRD Sediment Series No. 8. Ferguson R. I. and Church M. (2004) A simple universal equation for grain settling velocity. Journal Of Sedimentary Research 74(6), 933-937. Fries J. S. and Trowbridge J. H. (2003) Flume observations of enhanced fine-particle deposition to permeable sediment beds. Limnology And Oceanography 48, 802-812. Garcia-Aragon J., Salgado-Salazar S., and Funes-Gallanzi M. (2002) Experimental observations of particlefluid and particle-particle interactions in two-phase flows using PTV. Proceedings of the American Society of Civil Engineers (ASCE). Garcia-Aragon J., Salgado-Salazar S., Funes-Gallanzi M., and Sahagun L. (2004) Experimental observations of particle-particle interactions in settling velocity of polydispersed sediment mixtures. Ingenieria Hidraulica En Mexico 19(4), 57-64. Guo J. and Julien P. Y. (2001) Turbulent velocity profiles in sediment-laden flows. Journal of Hydraulic Research 39(1), 11-23. Jordan P. R. (1965) Fluvial sediment of the Mississippi River at St. Louis, Missouri U.S. Geological Survey Water-Supply Paper 1802. Khullar N. K., Kothyari U. C., and Ranga Raju K. G. (2007) Effect of wash load on resistance. Journal of Hydraulic Research 45(4), 497-504. Maindonald J and Braun J. (2003) Data Analysis and Graphics Using R: An Example-based Approach. Cambridge University Press. Martin V., Fisher T. S. R., Millar R. G., and Quick M. C. (2002 ) ADV data analysis for turbulent flows: low correlation problem. Proceedings 2002 EWRI/ASCE/IAHR Joint Int. Conf. on Hydraulic Meas. and Experimental Methods (CD-ROM). McCave I. N. (1984) Erosion, transport and deposition of fine-grained marine sediments. In Fine-grained sediments: deep water processes and facies. (ed. D. A. Stow and D. J. W. Piper), pp. 35-70. Blackwell Scientific Publications. Muste M. and Patel V. C. (1997) Velocity profiles for particles and liquid in open-channel flow with suspended sediment. Journal of Hydrologic Engineering (ASCE) 123, 742-751. Nikora V. I. and Goring D. G. (2002) Fluctuations of suspended sediment concentration and turbulent sediment fluxes in an open-channel flow. Journal of Hydraulic Engineering-Asce 128(2), 214-224. Nikora V. I., Goring D. G., and Biggs B. J. F. (2002) Some observations of the effects of micro-organisms growing on the bed of an open channel on the turbulence properties. Journal of Fluid Mechanics 450, 317-341.  57  Nordin C. F. (1964) Aspects of flow resistance and sediment transport Rio Grande near Bernallilo, New Mexico. U.S. Geological Survey Water-Supply Paper 1498-H. Righetti M. and Romano G. P. (2004) Particle–fluid interactions in a plane near-wall turbulent flow. Journal of Fluid Mechanics 505(93-121). Scott L. H. and Stephens H. D. (1966) Special sediment investigations Mississippi River at St. Louis, Missouri, 1961–1963. U.S. Geological Survey Water-Supply Paper 1819-J. Smith I. R. (1982) A simple theory of algal deposition. Freshwater Biology 12, 445-449. Smith J. D. and McLean S. R. (1977) Spatially averaged flow over a wavy surface. Journal of Geophysical Research 82(2), 1735-1746. Sontek. (1997) Technical Notes, Pulse coherent Doppler processing and the ADV correlation coefficient, pp. 5. Vanoni V. A. (1946) Transportation of suspended sediment by water. Transactions of ASCE 111, 67-133. Vanoni, V.A. and Nomicos, G.N. (1960) Resistance properties of sediment-laden streams. Transactions of ASCE 125, 1140-1175. Vanoni V. A. (1975) Sedimentation engineering. ASCE. Wang G. Q. and Fu X. D. (2004) Mechanisms of particle vertical diffusion in sediment-laden flows. Chinese Science Bulletin 49(10), 1086-1090.  58  CHAPTER 3: MODELS FOR THE VERTICAL DISTRIBUTION OF FINE SUSPENDED PARTICLES5 3.1 DESCRIPTION AND APPLICATION OF MODELS  Most previous attempts to model fine particle dynamics have been based on mathematical expressions for the vertical distribution of suspended particles. In the framework originally formulated by Rouse (1937) and expanded upon by later researchers, particle flux is described as a balance between downward movement due to gravity (particle settling) and turbulent motion, and upward movement due to diffusion and turbulence (dispersion). In this, the ‘advection-dispersion’ framework, particles and fluid are assumed to act as a single phase. In its complete form, the advection-dispersion model describes particle movement in three directions; the classical Rouse equation is simply the solution in the vertical dimension, assuming steady and uniform flows and a mean vertical velocity of zero. Although the advection-dispersion framework is mathematically robust, predictions based on this model require a set of assumptions and empirical calibrations. For example, particle settling velocity can be modified in the presence of other particles due to particle collisions and drag effects; thus to account for these effects, an empirical relation has been developed to determine the settling velocity of particles in a mixture as a function of particle concentration and the individual particle settling velocity. Similarly, a coefficient for the turbulent diffusion of particles must be related to the turbulent diffusion of clear water by accounting for the effects of particles on fluid turbulent structure and the difference in diffusion between a fluid and sediment particle. However, in most cases, the coefficient for particle turbulent diffusion is assumed equal to the coefficient for fluid turbulent diffusion (both assumed to be constants equal to one). Thus although empirical assumptions have been employed to account for effects not represented in the theoretical framework, the empirical coefficients are not well-defined. In dilute flows, particle effects are not likely to greatly alter the distribution of suspended particles, but significant deviations in both velocity and concentration profiles have been observed in sediment-laden flows. Typically, as mean concentrations increase, velocity and concentration profiles depart from log-law and Rouse profiles, exhibiting steeper velocity and concentration gradients. Numerous single- and two-phase models have been proposed for the prediction of velocity and concentration profiles in sediment-laden flows. In the twophase framework, interactions and differences between particle and fluid movement are considered; mechanisms of particle effects on flow are incorporated through modification of the advection-dispersion framework. For example, Cao et al. (1995) derived a diffusion equation for particle concentration that accounts for the downward mass flux of the fluid-solid mixture that arises from the inequality between water and particle density. At a single point in the flow, the net vertical velocity of the fluid-solid mixture is not zero because of  5  A version of this chapter will be submitted for publication. Salant, N.L. and Hassan, M.A. An assessment of models for the vertical distribution of fine suspended particles.  59  higher density solid particles exchange with lower density fluid particles, creating a local downward mass flux. Globally, this downward flux is balanced by an upward turbulent diffusion flux to achieve steady-state equilibrium. For a single profile, vertical velocity of the suspension is thus a function of particle concentration. Czernuszenko (1998) introduced an additional mechanism, called ‘drift flux’, into the advection-dispersion equation to model the effect of particles on turbulence structure; drift flux accounts for the fact that in dense concentrations particles migrate towards regions of low concentration. A third mechanism incorporated into several models of sediment-laden flow is density stratification, in which a buoyancy force created by the density gradient of suspended particles inhibits the turbulent mixing of sediment and fluid. Several formulations for this mechanism exist. Smith and McLean (1977) were the first to use the analogy of a thermally stratified atmospheric boundary layer to describe density stratification in sediment-laden flows. A similar approach was adopted by later researchers, developing slightly modified models with new data and different empirical constants (Gelfenbaum and Smith, 1986; McLean, 1991; McLean, 1992). More recently, Wright and Parker (2004) used additional experimental data to refine the density stratification model. In their formulation, the strength of stratification is indicated by the flux Richardson number, a dimensionless grouping that represents the ratio of energy lost working against density stratification to the energy produced by shear. Stream gradient is included in the denominator of the flux Richardson number, such that rivers with lower gradients may potentially have much higher Richardson numbers and stronger stratification effects. Stochastic models offer an alternative approach to the single- and two-phase frameworks, treating the motion of individual particles as a stochastic process. A stochastic approach has been adopted by some researchers in the ecological sciences (e.g. Denny and Shibata, 1989; McNair et al., 1997) because biological mechanisms of particle deposition cannot be represented in the traditional frameworks of bulk particle and fluid flow. Net deposition and entrainment have less relevance to organic particles, particularly living organisms carried as plankton, which may be removed from the water column by consumption or self-propulsion. An advantage of the stochastic framework is that it can be used to model the vertical and longitudinal movements of an individual particle, namely the time and distance between particle entrainment, deposition, re-entrainment, and final removal from the water column (‘hitting-time’ and ‘hitting-distance’), as well as the mechanisms and determinants of particle entrainment. However, stochastic models are considered unsuitable for simulating particle-particle or particle-fluid interactions (Ni, 2002). In this study, I applied four particle flux models to the concentration profiles measured using the methods and experimental design described in Chapter 2. My objective is to test the ability of these models to predict the distribution of fine particles under varying physical conditions. Fine particles, as defined in this study, are < 125 µm and transported as washload with minimal interaction with bed. Because these fine particles are readily suspended, concentration gradients are expected to be low; homogenous profiles are typical of wash load material but are unusual for suspended sands. Given this important distinction between particle size classes, this analysis should not be considered a critical test of these models, which were constructed for application to larger sand-sized material. However, fine particle concentration is a major determinant of water quality, pollutant transport, and suspended sediment load; thus an ability to model the concentration profile of this  60  material is important and merits evaluation. The four models assessed in this chapter are: 1) the classical equation developed by Rouse (1937); 2) a modified version of the Rouse equation (Glenn and Grant, 1987); 3) the two-phase model of Cao et al. (1995); and 4) the recently developed stochastic Local Exchange Model (McNair et al., 1997; McNair, 2000; McNair and Newbold, 2001). Although many attempts have been made to improve the original Rouse equation, it is still used in many applications and deserves assessment here. One of the major limitations of the traditional Rouse equation, however, is the use of a single particle size for all points in the profile; in reality, there is a distribution of particle sizes at each height. Thus, I chose to test a modified Rouse equation that incorporates the full particle size distribution at each height. The model of Cao et al. (1995) accounts for the effect of particle concentration on the vertical velocity of the sediment-fluid mixture, which arises due to lift effects and particle-particle interactions that occur as particles settle. Given the relatively low concentrations tested in this study (< 300 ppm), the Cao et al. (1995) model was the only one of the four that is designed for sediment-laden flow; particle effects caused by high concentrations are not expected to be a major factor in this study. Finally, I chose the Local Exchange Model as a representative stochastic model in order to test its suitability to describe clastic particle suspension in turbulent flows. For all four, I compared theoretical particle distributions to measured particle profiles to assess model performance for fine particles under varying physical conditions.  3.1.1 ROUSE EQUATION  Derived from the three-dimensional advection-dispersion equation, the Rouse equation represents the vertical transport of suspended particles as a balance between the settling of suspended particles due to gravity and upward turbulent diffusion  − Cws = ε p  dC dz  (1)  where C is the time-averaged concentration, ws is the particle still-water settling velocity, εp is the vertical coefficient for particle turbulent diffusion, and z is height above the bed. Conventionally, εp is related to the coefficient of turbulent viscosity vt by εp= β·vt, and β is assumed equal to one thus yielding εp≡ vt. For uniform and unidirectional flow, the vertical distribution of vt is related to friction velocity (u*) von Karman’s constant (κ = 0.40), z, and flow depth (H) by the theoretically-based parabolic relation vt = κu* y ( h − z )h . Substitution and integration of (1) and the specification that Cz = Ca at z = za. yields an equation for the average concentration of particles at a height z, (Cz) as:   z H − za  C z = Ca  ⋅   za H − z  ws and s = βκu *  −s (2)  (3)  61  where Ca is the sediment reference concentration at reference height za (normally taken at the top of the ‘near-bed layer’, also considered the base of the ‘suspension layer’; conventionally, the lowest measurement point is used). The shape of the profile is determined by the suspension index or Rouse number (s). Concentrations at each elevation are typically plotted non-dimensionally in terms of the reference concentration and flow depth; the shape of the profile varies with values of s. As s approaches 0 (due to a small settling velocity or high shear stress), the value in brackets approximates 1; thus elevation differences have less effect in the equation and the profile becomes more uniform (Figure 3.1).  Figure 3.1: Theoretical particle concentration profile as it varies with Rouse number (s) 1.0  s =0.001  0.9 0.8 0.7  z/H  s =0.01  s =0.1  0.6  s =0.25  0.5 s =0.5  0.4 s =1 0.3 0.2 0.1 0.0 0.0  s =2  0.2  0.3 C/Ca  0.4  0.5  0.6  Non-dimensional profiles allow for comparison between flow levels and bed conditions that may have different absolute concentrations.  3.1.2 MODIFIED ROUSE EQUATION  A modified Rouse equation that uses the full particle size distribution at each height (Glenn and Grant, 1987) takes into account the proportion of each particle size in suspension. In its original form, the equation determines the Rouse concentration profile using nine size classes, but this can be modified to include all 32 size classes detected by the LISST. In this equation, the average concentration of particles at height z (Cz) is expressed as 32  C z = Ca  ∑ P (i)( zz ) a  i =1  a  − A( i )  (4)  62  where A(i ) =  w s (i ) is the Rouse number corresponding to the ith particle size class with a settling velocity βκu*  ws(i) and Pa(i) is the proportion of the reference concentration (Ca) made up of that class.  3.1.3 MODEL OF CAO ET AL. (1995)  Cao et al. (1995) developed a model for the concentration and velocity profiles of both low- and highconcentration sediment flows in a two-phase framework. Continuity equations for fluid and sediment were used to develop a transport equation for suspended sediment, from which concentration profiles could be derived. Although based in physical theory and mechanistic explanations, the model does rely on some empiricism for mechanisms that are too complex to represent mathematically. Like the Rouse equation, the Cao et al. (1995) model is based on the theory that the vertical distribution of particles depends upon the balance between upward turbulent diffusion and downward movement due to advection and particle settling. In the Rouse equation, the net vertical velocity of the water-sediment mixture is assumed to be zero, but derivation of the continuity equation for the mixture shows that this only holds true for very low concentrations. Mass inequality between sediment and fluid particles when they exchange position in the water column produces a net downward flux of the mixture. Cao et al. (1995) show that the mixture’s timeaveraged vertical velocity, wd, is related to concentration, C, by the following equation: wd = −  ρ p − ρw ρw  ws C  (5)  where ρp and ρw are the particle and fluid density, assumed equal to 2650 (inorganic) or 1300 (plastic) and 1000 kg/m3, respectively. Average vertical velocity of the mixture is thus downward for particles denser than water and increases with concentration. Balancing this downward flux plus the particle settling velocity with the upward turbulent diffusion flux yields the following equation:  εp  ρ p − ρw dC = − wsC − wsC 2 ρw dz  (6)  If wd= 0, replacement in this equation yields the basic form of the Rouse equation; the relation between wdand C suggests, however, that this may only be applicable to very low concentration conditions (i.e. if C→ 0). From this equation, a general solution is obtained for the mean suspended sediment concentration, C, normalized by the near-bed concentration, Ca:  η( z) Cz = Ca 1 + [1 − η ( z )]( ρ s − ρ f ) ⋅ Ca / ρ f  (7)  63  where η(z) is defined differently for theoretical and empirical computations of vt. In the Rouse equation, above, the vertical distribution of vt is defined by a theoretically based parabolic distribution. In the Cao et al. (1995) model, empirical formulations for vt are used instead to account for the fact that particle settling velocity in clear, still water differs from settling velocity in a mixture. Given the complexity of the mechanisms that create this difference, including effects of lift and particle-particle interactions, Cao et al. (1995) do not attempt a theoretical representation. Empirical distributions of vt are found to be parabolic in the lower half of the water column (z < 0.5H) and linear in the upper half (z > 0.5H), producing the following formulations:  s   H − z ⋅ z a  , z ≤ 0.5H  z H − za  η(z) =    za   ⋅ exp[ −4 s( z / H − 0.50)],     H − z a   (8) z ≥ 0.5H  (9)  An alternative formulation for the full water column is expressed as a power function: z  η ( z ) =  a   3s   z   (10)  As concentration approaches zero, the equation for the mean concentration of particles reduces to C = η (z ) , producing the Rouse formula (8), a Rouse-exponent profile (9), or a power function (10) profile Ca  depending on which of the three formulations for η(z) is used. Cao et al. (1995) do not specify which of the three closures they use for validation of the model. I tested two of the formulations: (8) and (10), identified as ‘Cao et al. (1995)-Rouse’ and ‘Cao et al. (1995)-Power’, respectively. I selected (8) because of its similarity to the Rouse equation for low concentrations in order to determine whether the modifications of the Cao et al (1995) model improve the Rouse equation predictions for dilute suspensions. I tested (10) because it employs a mathematical formulation distinctly different from the Rouse equation. Cao et al. (1995) also provide a separate but similar derivation for determining the mean concentration of the near-bed layer, which is then used as the reference concentration Ca for integration into the model for the suspension layer. The authors acknowledge, however, that bed-load transport in the near-bed layer may introduce a velocity lag between the particle and water phases, invalidating two necessary assumptions: 1) particle inertia is negligible and 2) there is a difference in particle and fluid velocity only in the vertical direction. Given the uncertainty and the complexity of determining the near-bed concentration theoretically, I simply use the measured near-bed concentration for the reference concentration in the model computation.  3.1.2 LOCAL EXCHANGE MODEL  The Local Exchange Model (LEM), proposed by McNair et al. (1997), offers an alternative approach to modeling suspended particle dynamics that considers the behaviour of each individual particle, rather than an  64  assemblage of particles. The focus and intended application of the McNair et al. (1997) approach is determining the behaviour of fine organic particles. The LEM represents the vertical movement and elevation of a particle by considering that the motion of a neutrally buoyant, non-motile particle occurs in two ways: particles are propelled by molecular collisions or are incidentally carried by the turbulent transport of water. Assuming steady-state suspended sediment concentrations, zero lateral velocity, constant longitudinal velocity, and a laterally and longitudinally homogenous vertical concentration profile, McNair et al. (1997) derive an expression for the steady-state vertical distribution of particles:  d  dc *  uz c * −K =0  dz  dz   (11)  where c* is the normalized steady-state concentration, K is the dispersion rate (the sum of the components of molecular diffusion and turbulence, which are proportional to kinematic viscosity (M) and eddy viscosity (l2(z)du/dz), respectively), uz is the vertical component of the advective velocity, and z is elevation above the bed. Assuming particles are non-motile and uz is simply the particle settling velocity, ws (and thus negative), this equation is easily modified to yield an equation almost identical to the Rouse equation for the concentration of particles at height z (Cz). Using the form of Prandtl’s mixing length (l(z)) recommended by the authors ( l ( z ) = κz 1 − z / H ) ) and assuming no component due to molecular diffusion (Brownian motion), K can be reduced to (1 − z  H ) (κzu*)  . Replacing these values in the above equation and integrating over the profile yields an  equation for Cz.   z H − za  C z = Ca  ⋅   za H − z   − w s /[(1− z / H )(κu*)] (12)  Thus, like the other models, the LEM can be used to predict vertical profiles for comparison with measured profiles. Although similar to the Rouse equation, this formulation includes a depth-dependent parameter in the exponent.  3.1.4 MODEL APPLICATION AND ASSESSMENT  Models are designed to estimate the true mean suspended particle profile. In order to assess model performance, the precision and the accuracy of predicted profiles should be quantified relative to this mean condition. However, determining the true mean condition from measurements is impossible; as with any samples from a true distribution, measured profiles reflect inherent natural variability and measurement error. Thus the first step of model assessment was to estimate the mean profile from replicate measured profiles and assess the error around this mean. To do this, I plotted the measured concentrations normalized by the near-bed measurement of each profile against normalized height for a given time period (4h or 8h) from all replicates for a given experimental condition, fitting all points with a linear least-squares regression. Although the general  65  form of the theoretical models predicts a non-linear relation between concentration and height, the measured profiles were highly uniform and therefore best fit by a linear regression. For the non-replicated high and low concentration experiments, only one measurement profile was available for determining the best fit line. Normalized rather than raw concentration values were used because starting concentrations varied among replicates due to unintended variation in mass of the initial dose; normalizing corrected for these differences. Particle size varied among replicates for the same reasons as concentration thus I also determined the mean particle size profile for each experimental condition. I used the same method as the concentration profiles, plotting measured median particle size normalized by the near-bed particle size of each profile against normalized height for a given time period from all replicates for a given experimental condition and fitting measurement points with a linear least-squares regression. Before fitting regression lines, however, I also examined profiles and removed clear outliers – extremely large concentrations that I deemed outside the range of natural variability. I based my outlier criterion on the median particle size of each concentration measurement. Large median particle sizes indicate the presence of debris or air bubbles interpreted by the LISST as large particles. I determined the mean and standard deviation of the measured median particle size distribution for each time period and experimental condition. I defined outlier measurements as those with measured particle sizes falling outside one standard deviation from the mean and used the remaining measurement points to determine the best-fit profile. Doing so eliminated a total of 13 out of 170 measurement points obtained from 26 profiles, including five measurements from the 4h period and eight from the 8h period; the maximum number of eliminated points for a single profile was two. Final statistics for the median particle size analysis are provided in Table 3.1.  Table 3.1: Statistics for the distribution of median particle sizes at 4h and 8h for all experimental conditions. 4h Measured median particle size (µm) 8h Measured median particle size (µm) Condition n Mean Mean+1SD Max n Mean Mean+1SD Max 20%SHF 21 12.25 34.84 108.80 21 12.07 33.76 98.00 20%SLF 21 6.30 11.57 29.03 21 4.58 5.43 7.70 20%PHF 21 11.66 12.89 13.16 21 7.56 7.85 7.94 20%SHFHC 9 5.17 6.17 7.32 7 6.97 15.90 27.21 20%SHFLC 7 5.43 8.53 11.47 7 2.92 2.95 2.96 80%SHF 17 6.33 6.88 7.78 15 6.24 8.13 12.30 *Notes: Measurements with particle sizes greater than one standard deviation from the mean (Mean+1SD) were considered outliers. See Table 2.1 for experiment codes and descriptions. Because of the gradual decrease in average concentration over time, sequential profile measurements were slightly biased by temporal changes in concentration (i.e. sequential measurement points in the same profile may have had slightly lower concentrations even if there was no vertical difference between points). Thus I used measured profiles from late in the experiment (8h) when the deposition rate was very slow and temporal differences were negligible. I used profiles from the middle of the experiment (4h) for comparison. Secondly, I determined what proportion of the error from the mean condition was due to measurement error by using replicated measurements. Although time constraints prohibited me from replicating every  66  measurement or extending my sampling period longer than 15 seconds, I did replicate some individual measurements for all experimental conditions at certain time periods and heights. Details of these replicated measurements are provided in Table 3.2, including experiment, time period, height above the bed, volume concentration, and median particle size. For each set of replicated measurements, I calculated a percent error between the first and the second replicate for both concentration and median particle size. I then plotted the distribution of this percent error to determine the mean, maximum, and standard deviation of the measurement error. I considered the mean percent error as an estimate of the expected measurement error due to complications such as instrument error, air bubbles, or debris.  Table 3.2: Replicated concentration and particle size measurements.  Experiment 20%SHF1 20%SHF2  20%SLF1  20%SLF2  20%PHF2 20%PHF3 20%SHFHC  20%SHFLC 80%SHF1  Time (h) 1 4 3 5 7 2 3 4 4 7 1 2 3 5 7 1 4 1 3 3 3 4 4 7 3 4 6  Height (cm) 9 9 5.3 1.3 1.3 3.5 5.5 3.5 5.5 5.5 5 5 3 5 6 7.5 1.5 7 7 2.5 7.5 1 7.5 4 9 9 9  Rep 1 C (ppm) 65.0 42.5 42.4 34.6 28.9 63.3 49.4 43.0 40.6 30.5 60.1 49.8 40.9 30.9 25.6 96.5 28.1 107.1 54.9 65.4 59.9 60.8 62.1 10.2 54.1 39.5 35.9  D50 (µm) 5.9 9.0 4.1 104.1 20.6 6.1 5.3 5.1 4.6 3.9 5.5 5.0 4.6 3.8 4.1 18.9 12.9 22.6 12.7 4.8 4.4 4.7 7.3 2.9 6.9 5.1 14.1  Rep 2 C (ppm) 62.8 40.5 44.1 39.0 32.4 62.6 51.5 41.9 50.8 27.5 61.6 50.4 38.3 31.8 25.6 96.9 28.3 106.5 56.4 61.9 63.8 60.9 86.5 10.2 53.1 44.0 38.6  D50 (µm) 5.7 4.6 5.1 25.9 4.3 6.1 5.6 5.2 6.6 26.0 5.9 6.2 4.1 5.5 6.8 18.9 12.6 22.0 12.6 4.4 5.1 4.6 6.1 2.9 6.8 6.5 23.2  Percent error C (ppm) D50 (µm) 3.4 3.5 4.6 49.5 4.0 25.1 12.9 75.1 11.8 79.1 1.0 0.7 4.3 5.9 2.4 1.8 25.4 43.5 9.9 563.4 2.5 6.5 1.1 23.9 6.5 10.9 2.6 41.8 0.2 67.6 0.4 0.6 0.9 2.5 0.5 2.5 2.7 0.7 5.3 8.2 6.5 16.0 0.3 2.2 39.4 16.7 0.5 0.1 1.8 1.7 11.5 27.5 7.8 65.1  *Notes: ‘Time’ is the time during the experiment at which the measurement profile was started; ‘height’ is the measurement height above the bed; ‘C’ is volume concentration; and ‘D50’ is median particle size. Percent error is calculated as (Rep 2 – Rep 1)/Rep 1 × 100. A summary of the modeling approaches used is provided in Table 3.3. For all models, I used the near-bed u* computed from the near-bed Reynolds stress estimate and the deepest point (za/H ~ 0.02) of the mean profile for Ca. For Rouse and modified Rouse equation profiles, I assumed a constant β = 1. For both the original Rouse equation and the LEM, profiles were constructed using two different particle sizes in the computation of ws, which was calculated from the Ferguson and Church (2004) equation presented in Chapter 2. For ‘Point’ profiles, I used the median particle size at each point of the mean particle size profile. For ‘Mean’ profiles, I  67  used the depth-integrated particle size from the profile, such as might be derived from a single, depth-integrated water sample. In addition, in order to determine how the choice of particle size affects the shape of the profile, I constructed profiles using the median particle size of size classes 2-4 µm, 4-16 µm, 16-63 µm, and 63-122 µm. Size-specific profiles were constructed for comparison with the mean profile to determine both the influence of particle size on the shape of the profile and the particle size profile which best approximates the mean condition. For the modified Rouse equation, profiles were constructed using the LISST-detected particle size distribution at each height. For the Cao et al. (1995) model formulations (‘Rouse’ and ‘Power’) I used the depth-integrated particle size from the profile in the calculation of ws and different values of particle density for the silica and plastic particles. Finally, I tested the effect of the choice of shear velocity parameter on the shape and accuracy of the theoretical profiles using three different estimates of u*, calculated from: 1) the near-bed Reynolds stress value; 2) the maximum Reynolds stress value; and 3) from a method that uses the vertically averaged mean velocity and a roughness value estimated from the D84 of the bed, known as the ‘quadratic stress law’ (Sime et al. 2007).  Table 3.3: Description of modeling approaches for the vertical distribution of suspended particles. Model  Description  Rouse-Point  Rouse equation; mean particle size at each point in profile; 4h and 8h  Rouse-Mean  Rouse equation; depth-integrated particle size; 4h and 8h  LEM-Point  LEM; mean particle size at each point in profile; 4h and 8h  LEM-Mean  LEM; depth-integrated particle size; 4h and 8h  Modified Rouse  Modified Rouse; LISST-measured particle size distribution at each height; 4h and 8h  Cao et al. (1995) - Rouse  Cao et al. (1995) Rouse-like formulation; depth-integrated particle size; 4h and 8h  Cao et al. (1995) - Power  Cao et al. (1995) power function formulation; depth-integrated particle size; 4h and 8h  *Notes: All models used the near-bed u* computed from the near-bed Reynolds stress estimate and the deepest measurement of the mean profile (za/H ~ 0.02) for Ca. Model performance was assessed by comparing predicted profiles to the mean profile determined from measured profiles. Model precision was assessed using a root-mean-square error (RMSE), which measures the variance of the prediction, in which I took the square root of the squared, summed and averaged differences between the predicted concentration and the corresponding concentration at each height of the mean profile. I computed the RMSE for each profile and averaged for the hour (4h or 8h) and physical condition. Model accuracy was assessed using a simple estimate of average bias, the sum of difference between the predicted concentration and the mean profile concentration for all points in a profile and divided by the number of points in the profile. As for RMSE, I averaged the bias estimates from the replicate profiles corresponding to a given hour and physical condition. Positive bias values indicate models that generally over-predict concentrations; negative bias values indicate under-prediction. I used concentrations normalized by the near-bed measurement for the computation of RMSE and bias, then multiplied by one-hundred to determine percent RMSE and percent bias. I acknowledge that application and assessment of these models to the 80% sand condition is limited  68  somewhat by the bed movement problems described in Chapter 2, but the 80% sand analyses are included for completeness.  3.2 RESULTS  3.2.1 MEASUREMENT ERROR AND EXPERIMENTAL VARIABILITY  Using the median particle size criterion (Table 3.1), only five outliers were removed from the 4h measurements; eight outliers were deleted from the 8h measurements. Outliers were likely more common at 8h because debris may have built up in the sampling tubes over the course of the experiment, resulting in more frequent spikes in concentration. I observed the accumulation of debris in sections of the rubber tubing and flushed this material out before each experiment by pumping water at a rapid flow rate through the tubes. Although I attempted to remove debris in this manner during the experiment, the interval between measurements was not long enough to adequately remove all the accumulation. Replicated measurements from all experimental conditions and times (n = 26) had a mean percent error for concentration measurements of 6.3% (SD = 8.6%) and a range of 0.18-39.4% (Table 3.4). Replicated measurement error was greater for median particle size measurements, with a mean of 42.3% (SD = 107.2%) and a range of 0.10-563.3%.  Table 3.4: Statistics for the difference in concentration (C) and median particle size (D50) between two replicated measurements for each experimental condition; ‘n’ is the number of replicated measurements.  Condition 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF  n 5 10 4 4 1 2  Mean 7.35 5.59 1.14 12.87 0.48 9.55  C (ppm) Min Max 3.42 12.90 0.18 25.35 0.41 2.71 0.31 39.38 0.48 0.48 7.57 11.53  D50 (µm) Mean 46.44 76.60 1.58 10.77 0.10 46.33  Min 3.45 0.71 0.64 2.16 0.10 27.52  Max 79.14 563.39 2.50 16.69 0.10 65.14  Measured concentrations from 4h and 8h for all experimental replicates regressed on normalized height are shown in Figure 3.2; see Table 3.5 for regression parameters. Most of the linear regressions of concentration against height were non-significant fits (Table 3.5) because concentration throughout the water column was roughly uniform and did not vary much with depth. Very low and in some cases negative r-squared values demonstrate that the model has low predictive power. Specifically, negative r-squared values mean that the linear least-squares model is actually a worse fit than a horizontal line (vertical line as plotted in Figure 3.2); in other words, concentration remains constant with the increase in height. Of the significant regressions at 4h, the reference (4h), low concentration (8h), and 80% sand condition (4h and 8h) show a decline in concentration with height, while the plastic concentration increases with height (4h). Standard errors of the slope estimates are generally high due to the variability in individual measurements and among replicates. In order to determine whether theoretical profiles fall outside this range of variability, ninety-five percent confidence intervals were calculated for each regression and plotted around the mean profile. Upper and lower confidence limits (2.5 and  69  97.5%, respectively) associated with the independent variable (z/H) are used to determine a range of concentration values for each height; if predicted concentrations fall outside this range, the model and mean values are considered significantly different. Table 3.5: Regression parameters for measured concentrations normalized by the near-bed measurement regressed against normalized height above the bed from 4h and 8h for all experimental conditions. 4h 20SHF 20SLF 20PHF 20SHFHC 20SHFLC 80SHF 8h 20SHF 20SLF 20PHF 20SHFHC 20SHFLC 80SHF  Slope -0.085(0.04) -0.076(0.07) 0.104(0.04) -0.040(0.09) -0.063(0.08) -0.194(0.04)  Intercept 0.98(0.02) 1.01(0.03) 1.00(0.02) 1.01(0.04) 1.05(0.04) 1.02(0.01)  DF 16 18 19 5 4 10  r2 0.16 0.02 0.23 -0.16 -0.07 0.69  F 4.26 1.31 6.81 0.19 0.66 26.06  P <0.1 0.27 <0.05 0.68 0.46 <0.001  -0.050(0.05) -0.012(0.03) 0.059(0.20) -0.010(0.04) -0.104(0.01) -0.124(0.02)  1.00(0.02) 1.00(0.01) 0.92(0.08) 1.00(0.02) 1.01(0.01) 1.00(0.01)  19 17 18 4 5 12  0.0001 -0.05 -0.05 -0.23 0.92 0.73  1.00 0.13 0.08 0.06 70.21 36.29  0.33 0.72 0.78 0.81 <0.001 <0.001  *Notes: Values in parentheses are standard errors; DF = degrees of freedom; r2 = coefficient of determination; F = F-statistic; p = significance value of the regression. Figure 3.2: Normalized measured concentration profiles from 4h and 8h for all experimental conditions. a) 4h  20SHF 20PHF 20SHFLC  1.0  b) 8h  20SLF 20SHFHC 80SHF  0.9 0.8 0.7  z/H  0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.8  0.9  1.0 C/C a  1.1  1.2  0.8  0.9  1.0 C/C a  1.1  1.2  *Notes: Regressions were fit for normalized concentration as a function of normalized height, but profiles are plotted with normalized height on the y-axis by convention. Normalized median particle sizes regressed against normalized height for each experimental condition (Figure 3.3; regression parameters in Table 3.6) show the high variability in particle size for all conditions. Like concentration regressions, model fits for the most of the normalized measured particle size profiles were nonsignificant and r-squared values low, because particle size also varies little with height (Table 3.6). Among the  70  significant regressions, particle sizes of the reference (8h), low concentration (4h and 8h), and 80% sand (4h) conditions all decreased with height. Table 3.6: Regression parameters for normalized measured median particle size profiles from 4h and 8h for all experimental replicates. 4h 20SHF 20SLF 20PHF 20SHFHC 20SHFLC 80SHF 8h 20SHF 20SLF 20PHF 20SHFHC 20SHFLC 80SHF  Slope -0.27(0.18) -0.20(0.17) -0.005(0.03) 0.003(0.03) -0.10(0.01) -0.12(0.02)  Intercept 1.08(0.08) 1.10(0.07) 0.99(0.01) 1.00(0.02) 1.01(0.01) 1.00(0.01)  DF 14 18 13 4 5 12  r2 0.07 0.02 -0.07 -0.23 0.92 0.73  F 2.09 1.39 0.02 0.06 70.21 36.29  0.17 0.25 0.88 0.81 <0.001 <0.001  -0.17(0.08) 0.09(0.12) 0.05(0.07) 0.04(0.11) -0.03(0.01) -0.08(0.08)  1.04(0.03) 0.98(0.05) 1.04(0.06) 1.03(0.01) 1.00(0.03) 0.96(0.03)  18 19 18 4 5 11  0.20 0.002 -0.03 -0.21 0.86 0.002  4.69 1.02 0.50 0.12 32.27 1.02  <0.05 0.33 0.49 0.74 <0.01 0.33  P  *Notes: Values in parentheses are standard errors; DF = degrees of freedom; r2 = coefficient of determination; F = F-statistic; p = significance value of the regression. Figure 3.3: Normalized measured median particle sizes from 4h and 8h for all experimental replicates plotted against normalized height above the bed. 1.0  20SHF 20PHF 20SHFLC  a) 4h  0.9  20SLF 20SHFHC 80SHF  a) 8h  0.8 0.7  z/H  0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0  0.5  1.0 D 50 /D 50a  1.5  2.0  0.0  0.5  D 50 /D 50a  1.0  1.5  *Notes: Regressions were fit for normalized particle size as a function of normalized height, but profiles are plotted with normalized height on the y-axis by convention. 3.2.2 MODEL PERFORMANCE: ALL FORMULATIONS  Figure 3.4 and Figure 3.5 show the normalized predicted and mean profiles for each experimental condition at 4h and 8h, respectively, as well as the 95% percent confidence intervals of the mean profiles.  71  Figure 3.4: Normalized theoretical profiles of suspended particle concentrations for seven model formulations and six experimental conditions at 4h. 1.0 a) 20%SHF  b) 20%SLF  0.8  z/H  0.6  0.4 Mean Fit Modified Rouse Rouse-Point Rouse-Mean LEM-Point LEM-Mean Cao-Power Cao-Rouse 2.50%  0.2  0.0  1.0  c) 20%PHF  d) 20%SHFHC  0.8  z/H  0.6  0.4  0.2  0.0  1.0  f) 80%SHF  e) 20%SHFLC 0.8  z/H  0.6 0.4 0.2 0.0 0.0  0.5  C/C a  1.0  1.5  0.0  0.5  C/C a  1.0  1.5  *Notes: Experiments are: a) reference (20%SHF), b) low flow (20%SLF), c) plastic (20%PHF), d) high concentration (20%SHFHC), e) low concentration (20%SHFLC), and f) 80% sand (80%SHF). Red lines indicate the upper and lower limits of the 95% confidence interval. ‘Mean Fit’ is the mean profile of measured profiles for each condition. Concentrations (C) are normalized by the near-bed concentration of the mean profile (Ca); height above the bed (z) is normalized by total flow depth (H).  72  Figure 3.5: Normalized theoretical profiles of suspended particle concentrations for seven model formulations and six experimental conditions at 8h. 1.0 a) 20%SHF  b) 20%SLF  0.8  z/H  0.6 Mean Fit Modified Rouse Rouse-Point Rouse-Mean LEM-Point LEM-Mean Cao-Power Cao-Rouse 2.5  0.4 0.2 0.0  1.0 c) 20%PHF  d) 20%SHFHC  0.8  z/H  0.6  0.4  0.2  0.0  1.0  f) 80%SHF  e) 20%SHFLC  0.8  z/H  0.6 0.4 0.2 0.0 0.0  0.5  1.0 C/C a  1.5  0.0  0.5  C/C a  1.0  1.5  *Notes: Experiments are: a) reference (20%SHF), b) low flow (20%SLF), c) plastic (20%PHF), d) high concentration (20%SHFHC), e) low concentration (20%SHFLC), and f) 80% sand (80%SHF). Red lines indicate the upper and lower limits of the 95% confidence interval. ‘Mean Fit’ is the mean profile of measured profiles for each condition. Concentrations (C) are normalized by the near-bed concentration of the mean profile (Ca); height above the bed (z) is normalized by total flow depth (H).  73  Normalized mean profiles from both time periods (Figures 3.4 and 3.5) show the virtually uniform distribution of fine particle concentration throughout the water column for most physical conditions. As indicated by generally non-significant regressions of normalized concentration and height (Table 3.6), most mean profiles are better described by a vertical line than a least-squares linear regression. Only four profiles are well-described by a linear regression: low concentration at 4h and 8h, 80% sand at 4h, and the reference condition at 8h. Furthermore, visual examination of the profiles (Figures 3.4 and 3.5) shows that the predicted concentrations generally fall outside the confidence interval of the mean profiles, indicating that a linear model is a better predictor of wash material distribution than more complex non-linear formulations, at least for the conditions tested in this study. Four additional features of the profiles can be assessed in more detail using the RMSE and bias values presented in Tables 3.7 and 3.8: 1) the two Cao et al. (1995) models have high negative bias for all conditions except the low concentration condition at 8h, although the Cao-Rouse model is slightly less biased than the Cao-Power model for all conditions; 2) all models were negatively biased for the plastic model, with deviations from the mean profile increasing in the upper water column; 3) predictions of all LEM and Rouse formulations vary less than the error around the mean profile for all except the plastic condition (i.e. use of point particle size measurements or the full particle size distributions did not improve model performance over use of a depthintegrated particle size); and 4) in several cases, the LEM profiles showed a pronounced decrease in the upper portion of the water column, particularly for the low flow, high concentration and plastic conditions. Table 3.7: Percent RMSE values of suspended particle profiles for seven different models and six experimental conditions. Rouse Mean Modified 4.90 6.10 4.53 8.17 6.42 3.33 0.60 0.30 1.44 0.74 3.46 2.96 3.58 3.60  LEM Point Mean 4.47 3.87 2.98 4.12 13.15 8.00 3.23 2.39 1.05 1.09 1.29 1.98 4.97 3.89  Cao et al. (1995) Rouse Power 22.98 41.03 46.78 66.65 64.08 81.60 45.08 61.20 5.03 9.75 39.46 58.41 37.24 53.11  Average 4h 20%SHF 11.06 20%SLF 16.62 20%PHF 24.13 20%SHFHC 12.85 20%SHFLC 2.73 80%SHF 13.08 Average 13.61 8h 20%SHF 3 2.48 2.00 2.79 1.87 1.62 17.93 35.52 7.73 20%SLF 3 6.69 6.65 6.89 5.14 5.94 33.44 53.86 14.85 20%PHF 3 5.18 4.08 5.89 7.12 4.27 34.28 48.29 13.11 20%SHFHC 1 0.46 0.29 0.20 5.19 0.42 29.01 42.42 8.73 20%SHFLC 1 3.04 3.03 3.19 2.34 2.30 1.27 5.25 2.94 80%SHF 2 0.27 0.55 2.66 3.60 1.35 35.35 48.13 10.33 Average 13 3.02 3.21 3.60 4.33 2.91 25.21 38.91 9.82 *Notes: ‘n’ = number of experimental replicates. All models used the average near-bed u* computed from n 3 3 3 1 1 2 13  Point 5.25 5.18 8.63 0.39 1.84 3.11 4.07  replicated near-bed Reynolds stress estimates and the deepest measurement of the mean profile (za/H ~ 0.02) for Ca. See Table 2.1 for experiment codes and descriptions. See Table 3.3 for model descriptions.  74  Table 3.8: Percent bias estimates of suspended particle profiles for seven different models and six experimental conditions.  4h 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF 8h 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF  n 3 3 3 1 1 2  Point 3.86 3.50 -7.29 0.14 1.38 1.98  Rouse Mean Modified 3.39 4.64 4.16 4.96 -5.55 -0.22 0.35 -1.00 1.16 -0.31 1.97 0.90  3 3 3 1 1 2  1.94 5.55 -4.12 -0.36 2.64 -0.26  1.22 5.39 -3.15 0.21 2.27 0.25  -1.79 6.14 -5.00 -0.18 2.75 -13.81  LEM Point Mean 3.39 2.85 2.25 3.02 -10.29 -6.81 -1.68 -1.10 0.85 0.04 0.62 0.83 1.55 4.54 -5.56 -2.60 2.15 -2.08  0.99 4.94 -3.21 -0.10 1.83 -0.76  Cao et al. (1995) Rouse Power -22.50 -37.42 -51.10 -72.74 -60.99 -79.03 -40.23 -56.09 -3.87 -8.07 -36.21 -55.20 -17.02 -36.97 -31.08 -24.22 -1.15 -35.31  -34.30 -60.88 -45.35 -37.61 -5.06 -50.04  *Notes: ‘n’ = number of experimental replicates. All models used the average near-bed u* computed from replicated near-bed Reynolds stress estimates and the deepest measurement of the mean profile (za/H ~ 0.02) for Ca. See Table 2.1 for experiment codes and descriptions. See Table 3.3 for model descriptions. Firstly, as seen clearly in the plotted profiles (Figures 3.4 and 3.5) both the Rouse and power function of the Cao et al. (1995) model were highly negatively biased for almost all conditions at both time periods. In contrast, the LEM and Rouse models were generally positively biased or within the confidence interval of the mean profile, for all except the plastic condition. Interestingly, the Cao et al. (1995) models, particularly the Cao-Rouse model, were much more precise and more accurate for the low concentration condition. At 4h only Cao-Power formulation falls outside the confidence interval of the mean profile; all other formulations fall within the error of the measurements. Effects of particle concentration are included in the Cao et al. (1995) model through a density term in the denominator, which is modified by the near-bed concentration and the closure (i.e. Rouse or power function) selected (Eq. 7). For the Cao-Rouse formulation, the closure is identical to the Rouse equation (Eq. 8) thus it is the value of the denominator that skews model predictions. High Rouse numbers or high concentrations will increase the denominator, reducing the predicted concentration. A high Rouse number may be due either to a high settling velocity or a low shear velocity. In the case of plastic particles, density was low, but median particle size was high due to the gradual loss of large grain sizes thus settling velocity and Rouse numbers were higher in the denominator, reducing the Cao-Rouse predicted concentration far below the mean profile and the Rouse equation predictions. For the low concentration condition, particle sizes were small and near-bed concentration was low, producing a lower value in the denominator and higher predicted concentrations, hence the closer correspondence with the mean profile. All concentrations tested in this study were far below the ‘sediment-laden’ conditions for which the Cao et al. (1995) model is intended, so that the measured profiles did not show the steep concentration gradient that the model predicted. In my analysis, neither velocity nor concentration profiles exhibited much evidence of particle-fluid interactions or density stratification effects. As shown in Chapter 2, although the high concentration experiment did have a lower average near-bed Reynolds stress than moderate or low concentration experiments, the high concentration velocity profiles did not have a steeper velocity gradient than  75  those of the reference condition. Given the comparatively low concentrations and small particle sizes of the experiments in the present study, it is perhaps not surprising that particle effects were not significant. Particle and density stratification effects are found to be strongest in sediment-laden flows with concentrations reaching ~10,000 ppm (Coleman, 1986), two orders of magnitude larger than my ‘high concentration’ experiment. Furthermore, density stratification is not expected to occur for particles transported only as washload, i.e. those that do not interact with the bed, because particles of this size typically distribute uniformly in the vertical. My measured profiles exhibit virtually constant or slightly varying concentrations, making a stratification effect highly unlikely. Secondly, all models at both time periods consistently under-predicted concentrations of plastic particles throughout the water column. Plastic profiles may have been consistently under-predicted by all formulations because the models are not designed for particles with densities only slightly larger than water. Plastic particles remained in higher concentrations than predicted because they were easily buoyed by upward turbulent stresses generated by the coarse bed; as shown in Chapter 2, plastic particles deposited significantly slower than silica early in the experiment, with a wd of ~2 cm/h compared to ~8 cm/h for silica particles at high flow, translating into higher suspended sediment concentrations. Thirdly, all the Rouse and LEM formulations, whether based on depth-integrated particle sizes, point measurements, or full particle size distributions, predicted similar concentrations relative to the error around the mean profile. Furthermore, both the Rouse-Mean and LEM-Mean formulations were relatively precise, with average RMSE values <5% for both time periods. Bias values were also lower than the point-based versions of these models for all conditions during both time periods. Thus detailed particle size measurements may not be necessary to produce relatively accurate profiles, nor may they be suitable. Models are designed to use a depthintegrated, rather than a depth-varying, particle size; point measurements may be biased by local fluctuations in particle size due to small-scale hydraulic conditions and may not be representative of the full vertical distribution of particles. Finally, the observed decrease in concentration in the upper portion the LEM-Mean profiles occurred because the exponent in the equation ( w s / (1 − z H ) (κzu*) ) was relatively larger than points lower in the water column. Due to the change with z and because the same particle size was used for the entire profile, all profiles showed a slight decrease with depth. This decrease became more pronounced however, for a large settling velocity, small u*, or large z/H ratio (due to a shallow flow depth). In the case of low flow, the lower exponent was likely due to a small u* (~0.005 cm/s), such that particles were not suspended in high concentrations in the upper region of flow. Shear velocity estimated from the near-bed Reynolds stress was also relatively low in the high concentration experiment (~0.011 cm/s). In the case of plastic, ws was larger, despite a lower density, due to a measured median particle size that was more than twice that of the silica particles, a function of the slower loss of coarse particles from the water column, and the dominant effect of particle size over density in the settling velocity equation.  76  3.2.3 EFFECT OF PARTICLE SIZE  Particle size has a visible effect on the shape of the theoretical profile, regardless of model used. Plots of normalized theoretical profiles for the Rouse and LEM equations using the median particle size of size classes 2-4 µm, 4-16 µm, 16-63 µm, and 63-122 µm compared to mean profiles from the reference, low flow and plastic, and 80% sand conditions show the effects of particle size and physical condition (Figure 3.6). Profiles become less uniform as particle size increases. Low flow and the 80% sand condition have lower concentrations of all particle sizes, with coarse particles dropping out completely in the upper portion of the flow, due to lower shear velocity. Plastic concentrations of all particle sizes are slightly larger than silica concentrations due to lower particle density and the reduction in ws, but the effect is not as great as the shifts due to particle size. LEM profiles are less uniform than Rouse profiles; LEM concentrations are lower in the upper portions of flow and the difference increases with particle size, due to the depth-dependency of the LEM exponent (i.e., the 1-z/H component). For the reference (20%SHF) condition, the mean profile corresponds most closely with the 4-16 µm profile of the LEM equation, with only a slight deviation in the upper water column; Rouse profiles for the 2-4 and 416 µm classes are higher than the upper confidence limit of the mean profile. Actual median particle size of the mean profile ranged from 4-6 µm for the reference condition thus the Rouse equation slightly over-predicts these particle sizes under this condition. For the low flow condition, the 2-4 µm profiles of both the Rouse equation and the LEM fall within the mean profile confidence interval, but the 4-16 µm fall slightly below the lower confidence limit. Similar to the reference condition, the low flow mean profile had an actual median particle size of 4-5 µm. In contrast, the mean plastic profile had a larger median particle size of 11-12 µm but all Rouse and LEM profiles fall below the confidence interval of the plastic mean profile. Mean profile for the 80% sand condition had an actual mean particle size of 6-7 µm and corresponds best with the 4-16 µm profiles of the Rouse equation and the LEM; thus the models predict the 80% sand condition more precisely than the other conditions.  77  Figure 3.6: Normalized concentration profiles from the Rouse and LEM equations using the median particle size of size classes 2-4 µm, 4-16 µm, 16-63 µm, and 63-122 µm for four experimental conditions at 4h. 1.0  b) LEM - 20%SHF  a) Rouse - 20%SHF  0.8  z/H  0.6 0.4  Mean Fit 2-4 um 4-16 um 16-63 um 63-122 um 2.5  0.2 0.0  1.0  c) Rouse - 20%SLF  d) LEM - 20%SLF  z/H  0.8 0.6 0.4 0.2 0.0  1.0  e) Rouse - 20%PHF  f) LEM - 20%PHF  0.8  z/H  0.6 0.4 0.2 0.0 1.0  g) Rouse - 80%SHF  h) LEM - 80%SHF  0.8  z/H  0.6 0.4 0.2 0.0 0.00  0.20  0.40  0.60 C/C a  0.80  1.00  1.20  0.00  0.20  0.40  0.60 C/C a  0.80  1.00  1.20  *Notes: ‘Mean Fit’ is the mean profile of measured profiles for each condition. Concentrations (C) are normalized by the near-bed concentration of the mean profile (Ca); height above the bed (z) is normalized by total flow depth (H). Red lines indicate the upper and lower limits of the 95% confidence interval. ‘Mean Fit’ is the mean profile of measured profiles for each condition.  78  3.2.4 EFFECT OF SHEAR VELOCITY PARAMETER  In the above analysis, the near-bed Reynolds stress (τRe0) estimate was used to calculate u* for constructing theoretical profiles. However, use of a local, near-bed estimate of shear stress is potentially problematic because turbulent fluctuations and instrument error can be high near a rough bed. Thus Reynolds stress calculated from a near-bed velocity measurement may not be representative of the full water column. As discussed in Chapter 2, Reynolds stress profiles reach a maximum part-way up the water column (z/H ~ 0.15-0.25) and then decrease towards the water surface. Considering the height of maximum Reynolds stress (τReM) as the top of the roughness layer, this parameter can be considered representative of turbulence conditions above the inner boundary layer. An alternative method for determining shear stress, known as the ‘quadratic stress law,’ does not use local measurements, instead incorporating the vertically averaged mean velocity and a roughness height estimated from bed D84; in a large river, this method has been shown to provide more repeatable estimates and better precision than estimates based on single near-bed measurements (Sime et al., 2007). According to the quadratic stress law, shear stress τU is calculated as τU = ρCdU 2 where C d = κ 2 / ln 2 ( H / ez 0 ) is the drag coefficient, z0 = 0.1D84, and U is the depth-integrated streamwise velocity. Particle size distributions of the 20% and 80% sand beds provide D84 values of ~30 and 5 mm, respectively. I used these three different estimates of shear stress to calculate u* for each experiment and then applied the three u* estimates to the Rouse-Mean equation. I selected the Rouse-Mean equation for this analysis because of its consistently good performance and simple formulation. Comparisons between the three u* profiles and the mean profiles at 4h and 8h for the reference, plastic, high concentration, and 80% sand conditions are provided in Figure 3.7 and Figure 3.8. Methods are identified as Rouse-τRe0, Rouse-τReM, and Rouse-τU. Percent RMSE and bias values and average shear velocities for each method and experimental condition are provided in Table 3.7.  79  Figure 3.7: Normalized concentration profiles from the Rouse-Mean equation using three different estimates of shear velocity based on three estimates of shear stress (τRe0, τReM,, τU.) for six experimental conditions at 4h. 1.0  b) 20%SLF  a) 20%SHF  Mean Fit Rouse-τRe0 Rouse-τReM Rouse-τU  0.8  z/H  0.6 0.4 0.2 0.0 1.0  d) 20%SHFHC  c) 20%P HF  0.8  z/H  0.6 0.4 0.2 0.0 1.0 e) 20%SHFLC  f) 80%SHF  0.8  z/H  0.6 0.4 0.2 0.0 0.5  1.0 C/C a  1.5  0.5  1.0 C/C a  1.5  *Notes: ‘Mean Fit’ is the mean profile of measured profiles for each condition. Concentrations (C) are normalized by the near-bed concentration of the mean profile (Ca); height above the bed (z) is normalized by total flow depth (H). Red lines indicate the upper and lower limits of the 95% confidence interval. ‘Mean Fit’ is the mean profile of measured profiles for each condition.  80  Figure 3.8: Normalized concentration profiles from the Rouse-Mean equation using three different estimates of shear velocity based on three estimates of shear stress (τRe0, τReM,, τU.) for six experimental conditions at 8h. 1 .0  b) 20%SLF  a) 20 %SHF  Mean Fit Rouse-τRe0 Rouse-τReM Rouse-τU  0 .8  z/H  0 .6 0 .4 0 .2 0 .0 1.0  d) 2 0%SHFHC  c) 20%P HF  0.8  z/H  0.6 0.4 0.2 0.0 1 .0 f) 80%SHF  e) 2 0%SHFLC 0 .8  z/H  0 .6 0 .4 0 .2 0 .0 0.5  1.0 C/C a  1.5 0.5  1.0 C/C a  1.5  *Notes: ‘Mean Fit’ is the mean profile of measured profiles for each condition. Concentrations (C) are normalized by the near-bed concentration of the mean profile (Ca); height above the bed (z) is normalized by total flow depth (H). Red lines indicate the upper and lower limits of the 95% confidence interval. ‘Mean Fit’ is the mean profile of measured profiles for each condition.  81  Table 3.9: Percent RMSE and bias values of concentration profiles predicted by the Rouse-Mean equation using three different estimates of shear velocity (u*) for six experimental conditions.  4h 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF Average  Rouse- τRe0 4.92 4.65 6.68 0.60 1.44 3.46 3.66  RMSE Rouse-τReM 5.52 5.86 5.67 1.26 1.82 3.56 4.03  Rouse-τU 5.44 5.34 5.82 1.22 1.79 3.56 3.92  Rouse- τRe0 23.70 24.94 -38.82 2.43 6.97 11.80 3.84  Bias Rouse-τReM 28.85 31.94 -32.30 6.37 8.25 14.62 8.62  Rouse-τU 28.38 28.74 -33.27 6.16 8.11 14.58 7.63  Shear velocity (cm/s) u*Re0 u*ReM U*U 0.037 0.065 0.054 0.017 0.043 0.025 0.035 0.063 0.055 0.017 0.062 0.055 0.023 0.062 0.058 0.014 0.038 0.037 0.024 0.056 0.047  8h 20%SHF 20%SLF 20%PHF 20%SHFHC 20%SHFLC 80%SHF Average  Rouse- τRe0 2.00 6.65 4.08 0.29 3.03 0.55 3.21  Rouse-τReM 2.71 7.85 4.37 0.29 3.03 0.59 3.65  Rouse-τU 2.64 7.21 4.48 0.26 3.00 0.48 3.52  Rouse- τRe0 8.57 35.57 -22.05 1.44 15.87 0.63 7.88  Rouse-τReM 15.20 45.74 -24.49 1.44 15.87 0.91 10.75  Rouse-τU 14.69 41.17 -25.43 1.29 15.72 1.72 9.49  u*Re0 0.037 0.017 0.035 0.017 0.023 0.014 0.024  u*ReM 0.065 0.043 0.063 0.062 0.062 0.038 0.056  u*U 0.054 0.025 0.055 0.055 0.058 0.037 0.047  *Notes: See Table 2.1 for experiment codes and descriptions. See text for shear velocity calculations and the model formulation. For all experimental conditions, the τU estimate of u* falls in-between the near-bed and maximum Reynolds estimates (Table 3.9), demonstrating that this method effectively integrates the high and low turbulent stresses in the water column. Use of the τU estimate does not significantly improve the precision or accuracy of the Rouse-Mean equation overall, however. As clearly seen in the profiles at 4h and 8h (Figures 3.7 and 3.8), the range of concentrations predicted by the three shear velocity methods is less than or equal to the variability around the mean profile. For some conditions, the three estimates fall within the confidence interval of the mean profile, indicating that the model predictions are statistically similar to the least-squares regression. For other conditions, the model estimates fall outside the error bounds, but the differences between theoretical profiles are negligible relative to the differences with the mean profile. Differences among theoretical profiles are more pronounced in the cases when near-bed Reynolds stress is low (low flow, high concentration, and 80% sand conditions) or settling velocity is high (plastic particle experiments). In these cases, the Rouse-τRe0 profile is less uniform than both the Rouse-τReM, and Rouse-τU profiles, with lower concentrations in the upper region of flow, due to the lower capacity of the shear velocity or the higher settling velocity applied to the model. This decrease with height above the bed more closely reflects measured profiles for all except the plastic condition. In the case of plastic, the Rouse-τReM, and Rouse-τU profiles are relatively uniform and thus slightly closer to the mean concentrations (which increase with height) than the Rouse-τRe0 concentrations (which decrease with height). RMSE and bias values (Table 3.9) reflect the slight differences in performance among methods and the negative bias in the plastic predictions.  82  3.3 CONCLUSIONS  Theoretical models for the vertical distribution of suspended sediment were assessed for their ability to predict concentrations of fine wash load material under varying physical conditions by comparing modelpredicted concentration profiles with mean profiles derived from measured concentration measurements. Model precision and accuracy were assessed using visual examination of concentration profiles and RMSE and bias estimates. Three forms of the classical Rouse equation were tested. Two forms of a model designed for sediment-laden flow (Cao et al., 1995), including a power-function and Rouse-like formulation, were applied to comparatively dilute concentrations. The Local Exchange Model (LEM), a stochastic model designed for particles with motility and low densities (McNair et al., 1997), was modified for application to non-motile clastic particles, producing a Rouse-like equation with a depth-varying exponent. Results demonstrate that the fine particles tested in this study generally distribute uniformly in the water column; thus most measured profiles are better described by a vertical line than a least-squares linear regression or any of the non-linear models. Among models, both Cao et al. (1995) models had RMSE and negative bias values up to ten times the other models for virtually every condition, due to the inclusion of a density-concentration term in the denominator of the equation. However, the low concentration condition was relatively well-predicted by the Cao et al. (1995) models, with the Cao-Rouse formulation performing best among models for low concentrations. Low concentrations were better-predicted by the Cao et al. (1995) models because the densityconcentration term in the denominator was reduced, limiting its effect on the model prediction. Conversely, the plastic condition was the most poorly predicted by the Cao et al. (1995) models because the large mean particle size of the plastic suspension increased the denominator, skewing concentrations to lower values than measured. All models were negatively biased for the plastic model, with deviations from the mean profile increasing in the upper water column. Predictions using a depth-integrated particle size in both the Rouse equation and the LEM were generally closer to measured concentrations than those using point-based particle sizes. Furthermore, the inclusion of the full particle size distribution did not generally improve model performance. However, the variation among LEM and Rouse models was generally less than or equal to the error of the measured profiles. LEM concentrations decreased relative to measured and Rouse profiles in the upper portions of flow due to the depth-dependency of the LEM exponent. Particle size had a large effect on profile shape for both the Rouse equation and the LEM; model uniformity decreased with the increase in particle size, with concentrations of particles > 63 µm dropping to zero high in the water column if shear velocities were low. Finally, use of the quadratic stress law to estimate shear velocity did not significantly improve or degrade model performance relative to the use of localized Reynolds stress estimates. Given that the quadratic stress law is a more stable and integrative measure, this method may be a suitable replacement for localized estimates in model applications. Furthermore, near-bed estimates were somewhat problematic in instances of damped near-bed turbulence or large particle settling velocities.  83  3.4 REFERENCES Cao Z., Wei L., and Xie J. (1995) Sediment-laden flow in open channels from two phase flow viewpoint. Journal of Hydraulic Engineering 121(10), 725-735. Czernuszenko W. (1998) The drift velocity concept for sediment-laden flows. Journal of Hydraulic Engineering 124(10), 1026-1033. Denny M. W. and Shibata M. F. (1989) Consequences of surf-zone turbulence for settlement and external fertilization. American Naturalist 134, 859-889. Elghobashi S. (1994) On predicting particle-laden turbulent flows. Applied Scientific Research 52(4), 309-329. Gelfenbaum G. and Smith J. D. (1986) Experimental evaluation of a generalized suspended-sediment transport theory. In Shelf Sands and Sandstones (ed. R. J. Knight and J. R. McLean), pp. 133-144. Canadian Society of Petroleum Geologists, Memoir II. Glenn S. M. and Grant W. D. (1987) A suspended sediment stratification correction for combined wave and current flows. Journal of Geophysical Research 92(C8), 8244-8264. McLean S. R. (1991) Depth-integrated suspended-load calculations. Journal of Hydraulic Engineering 117(11), 1440–1458. McLean S. R. (1992) On the calculation of suspended load for noncohesive sediments. Journal of Geophysical Research (Oceans) 97(C4), 5759–5770. McNair J. N. (2000) Turbulent transport of suspended particles and dispersing benthic organisms: The hittingtime distribution for the local exchange model. Journal of Theoretical Biology 202(3), 231-246. McNair J. N. and Newbold J. D. (2001) Turbulent transport of suspended particles and dispersing benthic organisms: the hitting-distance problem for the local exchange model. Journal of Theoretical Biology 209(3), 351-369. McNair J. N., Newbold J. D., and Hart D. D. (1997) Turbulent transport of suspended particles and dispersing benthic organisms: How long to hit bottom? Journal of Theoretical Biology 188(1), 29-52. Ni J. (2002) Particle suspension in sediment-laden flow. Progress in Natural Science 12(7), 481-492. Rouse H. (1937) Modern conceptions of the mechanics of fluid turbulence Transactions of the American Society of Civil Engineering 102 463-554. Sime L. C., Ferguson R. I., and Church M. (2007) Estimating shear stress from moving boat acoustic Doppler velocity measurements in a large gravel bed river. Water Resources Research 43, W03418. Smith J. D. and McLean S. R. (1977) Spatially averaged flow over a wavy surface. Journal of Geophysical Research 82(2), 1735-1746. Wright S. and Parker G. (2004) Flow resistance and suspended load in sand-bed rivers: simplified stratification model. Journal of Hydraulic Engineering 130(8), 796-805.  84  CHAPTER 4: EFFECTS OF PERIPHYTON PATCHES ON HYDRAULICS OF GRAVEL-BED FLOW6 4.1 INTRODUCTION  Coarse substrates create micro-environments with unique structural and hydraulic conditions that support a diverse and abundant group of organisms. In particular, benthic organisms are greatly affected by near-bed flow conditions; flow is generally believed to be the dominant forcing factor shaping all stream processes and patterns (Hart and Finelli, 1999). Organisms are adapted to both withstand and benefit from the turbulent flows generated by streambed structures. For example, studies have shown that filter-feeders will alter their distribution or feeding appendages to maximize feeding efficiency under a given set of flow conditions (Morin et al., 1988; Shimeta and Jumars, 1991; Malmqvist and Sackmann, 1996). Even non-motile benthic algae are influenced by flow; higher velocities increase the rate of nutrient exchange, promoting algal growth (Denny, 1973; Vogel, 1994). Furthermore, sediment deposition and scour are also controlled by flow, in turn influencing streambed morphology and physical heterogeneity. Thus organisms that influence their local flow environment can have a large impact on both habitat condition and organism behaviour. Submerged aquatic vegetation can alter flow and sedimentation patterns in stream channels. Numerous studies have documented the effects of large-scale plant forms such as grasses and macrophytes on channel roughness, velocity gradients, and turbulence intensities (e.g. Kouwen and Unny, 1973). However, large plant forms and extensive plant canopies are uncommon in small gravel-bedded streams, where the dominant submerged vegetation is periphyton. Periphyton is a complex matrix of photosynthetic and heterotrophic microbes attached to the streambed surface. Many forms are pervasive in streams and rivers throughout the world, yet only a handful of studies have investigated the hydrodynamic impact of periphyton (Dodds and Biggs, 2002; Nikora et al., 2002a; Labiod et al., 2007). Previous studies of periphyton-flow interactions have generally evaluated surface effects above the matrix and in the overlying flow, focusing primarily on periphyton contributions to roughness. While most studies documented an increase in surface shear stress, shear velocity, roughness length, and resistance as periphyton biomass increased (Reiter, 1989a; Reiter, 1989b; Nikora et al., 1997; Nikora et al., 2002a; Labiod et al., 2007), others suggested that periphyton mats may have ‘smoothed’ the boundary layer, reducing turbulent shear, an effect that increased with mat thickness (Godillot et al., 2001). A conceptual model of periphyton–flow interactions presented by Nikora et al. (1998) evaluated the relationship between periphyton- and bed-derived roughness as a function of the drag and buoyancy forces, which varied with periphyton density, filament height/diameter, and near-bed flow velocity.  6  A version of this chapter has been submitted for publication. Salant, N. and Hassan, M. Effects of periphyton patches on hydraulics of gravel-bed flow.  85  However, Nikora et al. (1998) acknowledged that this simple model neglected potentially relevant factors, such as the scale of periphyton filaments relative to other roughness elements and the effects of flow penetration into the mat. Although they hypothesized that penetration would intensify at higher velocities and increase surface roughness, they did not speculate on flow conditions within the mat. Only two known studies have investigated effects below the periphyton surface (Dodds and Biggs, 2002; Nikora et al., 2002a); both indicated distinct reductions in mean velocities and turbulence intensities, the degree of which depended on periphyton density and structure (Dodds and Biggs, 2002). Furthermore, available evidence suggests that periphyton effects on surface roughness are variable, depending on periphyton structure, density, and roughness scale (Nikora et al., 2002a). Indeed, many studies of vegetation-flow dynamics have demonstrated the importance of plant characteristics, such as height, density, degree of compaction, or flexibility (Dawson and Charlton, 1988). In addition, whereas most previous studies of periphyton-flow dynamics used a channel bed fully covered with periphyton of uniform density and structure, natural periphyton assemblages are often patchily distributed and vary in form and size. Furthermore, these and other previous studies of coarse-bed hydraulics often used channel-averaged, spatially integrated hydraulic measures, rather than at-a-point, local-scale parameters. Thus results and theories derived from full-bed measurements may not be applicable to single locations on heterogeneous streambeds with dispersed periphyton growth. Our primary objective was to investigate how filamentous periphyton patches associated with pebble clusters alter local-scale flow hydraulics, focusing on the change in the vertical velocity and Reynolds stress distribution at a single location near the boundary. Our approach differed from previous periphyton and cluster studies in several important ways. Firstly, we concentrated our efforts on replication at a single point (i.e. the centre of a periphyton-pebble cluster), rather than measure over a broad spatial scale. Although our measurements did not, therefore, provide information about streamwise or lateral flow patterns around a single cluster, they allowed us to analyze the effects of a range of cluster configurations, periphyton densities, and mat thicknesses. In contrast, most previous studies used a single set of flow measurements taken before and after growing a bed-covering periphyton assemblage (Labiod et al., 2007; Nikora et al., 2002a). Secondly, for each replicate, we compared the effects of periphyton mats directly to conditions imposed by the coarse substrates that host them. Both these aspects helped to accommodate the inherent complexity of bed roughness elements and the variability in hydraulic effects that result. We were thus able to assess whether the scale of periphyton impact was significant relative to the scale of turbulent eddies and velocity fluctuations generated by other surface roughness elements. Thirdly, we restricted our analysis to local estimates of hydraulic conditions, not channel-averaged conditions. We recognize that above clusters, the upper region of flow is influenced by both local and upstream conditions, but the boundary region is dominated by local effects. Thus we analyzed only near-bed, at-a-point parameters, including the near-bed streamwise velocity, near-bed Reynolds stress, and the Reynolds stress at the top of the inner boundary layer. By using an ADV we were able to measure at-a-point, time-integrated, three-dimensional velocities at locations close to the bed, permitting the direct calculation of local Reynolds stresses. Local estimates of turbulence removed the need for assuming logarithmic or exponential velocity functions, which may be inapplicable to high roughness, heterogeneous beds and shallow  86  flows (Katul et al., 2002). Finally, although we used only filamentous periphyton mats (most prior studies used a mixed or diatom-dominated assemblage), we characterized each mat in terms of density and height so that we could more closely decipher the relation between hydraulic effect and periphyton structure.  4.1.1 CONCEPTUAL MODEL OF PERIPHYTON-FLOW DYNAMICS  Using previous periphyton-flow studies, macrophyte analogues, and basic understandings of periphyton structure and hydrodynamic principles, we can develop a generic model of the relationship between periphyton and flow in terms of its governing factors (Figure 4.1). Our conceptual model demonstrates the two-way interaction between hydraulic conditions and periphyton properties: flow conditions imposed on basic, unchangeable periphyton properties produce a given mat morphology; in turn, the structure of the periphyton mat influences local flow distribution. Thus the emergent properties of this model are the microbial mat morphology and the local flow structure; the independent variables, or governing factors, are the imposed flow and system geometry (e.g. flow hydraulics, bed material characteristics, bed morphology, and channel geometry) and basic properties of the microbial community, related to the species and growth stage. An issue of scale arises here that is important to clarify. Whereas bulk flow and bed morphology operate at the channel scale, the effect of periphyton mat morphology is manifest at the local (particle or cluster) scale and thus is reflected only in local estimates of hydraulic condition (at-a-point vertical velocity and turbulence profiles). In our study, channel and bed morphology remain constant; only bulk discharge varies from low to high flow (~4.2 to ~8.5 L/s). We can predict the influence of this change in flow rate on mat structure: compaction of the flexible periphyton mats by high flows should decrease mat height. Filament flexibility, density, and height interact with discharge to produce a resulting mat morphology (illustrated schematically in Figure 4.2). We assume that the material properties (e.g. flexibility, skin friction) of the filaments are constant for the single species of algae used in this study, but expect this to vary for other microbial types. Interaction between these microbial properties and the imposed flow produce a mat morphology with a set of properties (e.g. porosity, permeability, height) that in turn influences local flow structure. We quantify the effect on local hydraulic conditions using at-a-point, near-boundary velocity and turbulence parameters. Based on other studies of flowvegetation dynamics, we can hypothesize how the interaction between mat structure and flow rate will influence local hydraulic parameters (Figure 4.2).  87  Figure 4.1: A conceptual model of periphyton-flow dynamics. One-way arrows indicate an influential effect from one factor to another.  Independent Variables  Flow and System Geometry Flow hydraulics Bed material characteristics (Size Distribution, Clusters, Surface Roughness) Bed morphology (Riffles, pools) Channel geometry (Depth, Width, Slope)  Flow Structure  Basic Properties of Microbial Community Filament flexibility Filament density Friction properties  Microbial Mat Morphology (a.k.a. ‘Mat Structure’)  Velocity distribution Turbulence distribution  Height Porosity Permeability  Emergent Properties Figure 4.2a illustrates how the relationship between bulk flow and filament density should influence mat height and hydraulic condition. Flows imposed on low-density, and therefore high-porosity, high-permeability mats will be able to move through, as well as over, the filaments. Resistance caused by the filaments will thus lower velocities within the mat. Velocities will be lowest for the low bulk flow rate. Turbulence intensity will be higher in and above the mat, however, due to high momentum exchange with the overlying flow and roughness generated by the filaments. As density increases (and permeability decreases), flow will be diverted over the mat, causing constriction of the flow depth and a consequent acceleration of flow velocity in the region above the mat. A high filament density will block the flow of water through the mat, forming what has been called a ‘closed’ or ‘shielded’ structure (cf. Sand-Jensen and Pederson, 1999). In this case, flow into the mat is limited in all directions (vertically, laterally, and horizontally) thus turbulence intensity will also be reduced. As flows increase, both low- and high-density mats will become more compacted and streamlined. Near-bed velocities will increase for all densities due to the increase in bulk discharge, but closed mats will again have the highest velocities due to flow constriction, although the effect may be reduced slightly as mats decrease in  88  height. Turbulence intensity will be lowest for high-density, closed mats at high flows because streamlined filaments should be hydrodynamically smoother than upright filaments and compaction will further reduce flow exchange with the mat. Figure 4.2: Schematic representation of the influence of flow rate and a) filament density or b) filament height on local flow structure.  a) Density Low  b) Filament height High  Short  Long  High  LF  Mat height  Turbulence intensity  Flow rate  Tall  HF  Short  Low  *Notes: Gray ovals represent stone clusters; thin black lines represent mat filaments. Arrows above each stone represent velocity vectors directly above and within the mat; the thickness of each arrow represents the relative magnitude of the velocity. In cases of high density or highly compacted mats, velocities within the mats are considered negligible thus no arrow is drawn. Figure 4.2b illustrates how the relationship between bulk flow and filament height influences mat height and hydraulic conditions, given a constant filament density. In this case, we consider a relatively low filament density, or an ‘open’ mat. At low flows, both short and long filaments remain relatively upright (although we acknowledge that the upper parts of tall filaments may experience slightly higher velocities and may bend somewhat). Momentum exchange with the mats should be similar for both, although the depth of flow penetration will decline with filament height, as has been shown for submerged grass canopies (Nepf and Vivoni, 2000). Thus turbulence intensity above the mats should be high, but may be reduced at the base of the tall filaments. Because we are considering a low-density, open mat, velocities will be low for short and tall mats, as flow moving through the mat experiences resistance. In the case of a high-density, closed mat, however, velocities will be higher for the long filaments, which will cause greater constriction of the upper flow  89  layer than the short filaments. As flows increase, compaction of the filaments will decrease the height of the mat for both the short and long filaments. Compaction will also limit momentum exchange and smooth the mat surface, resulting in lower turbulence intensity. Higher velocities will result from less resistance from the filaments and higher bulk flow for both short and long filaments. Long filaments may have slightly higher velocities above the mat than short filaments because the additional biomass may create more of an obstruction, slightly constricting the flow. We have provided a generic model of periphyton-flow dynamics that describes how the imposed hydraulic/morphological conditions and basic microbial properties influence the emergent periphyton mat morphology and resulting local flow structure (Figure 4.1). We then used this model to qualitatively describe the relationship between flow rate, filament density, filament height, mat height, and local hydraulic parameters (Figure 4.2). Similar descriptions could be made for varying periphyton structures, including different microbial species. Our experiments tested this model for patches of a filamentous green alga at a range of filament densities, filament heights, mat heights, and cluster configurations at two bulk flow rates.  4.2 METHODS  Our procedure was designed to provide data that explicitly address our primary objective – to determine how the presence of periphyton alters near-bed hydraulics at a single location on a coarse bed. Our overall experimental procedure for a single experiment was as follows: 1) in a coarse-bedded flume, we constructed a surface cluster of three pebbles covered with filamentous periphyton; 2) we measured a vertical threedimensional velocity profile above the centre of the cluster at two flow rates; 3) after stopping the flow, we recorded the location of the cluster, scrubbed periphyton from the pebbles, and reconstructed periphyton-free cluster in the same configuration; and 4) we then repeated the vertical profile measurements at the same location and flow rates. Twenty different clusters were tested in this manner, covering a range of configurations and periphyton mat characteristics. Direct comparisons between the vertical profiles and near-bed hydraulic parameters for clusters with periphyton and the identical cluster without provide evidence of periphyton effects on local-scale flow conditions. Details of the procedure are described below. Experiments were conducted in a small (0.15 m wide x 7.75 m long x 0.44 m deep) recirculating laboratory flume at the University of British Columbia (Figure 2.1), with a bed covered by a pebble-gravel-sand substrate ~7-9 cm deep (Table 4.1). The experiments were run at two flow rates: ‘low’ (LF = ~4.2 L/s) and ‘high’ (HF = ~8.5 L/s). Water level, bed elevation, and longitudinal velocity measurements indicated that both of the selected flows were uniform within the 2-5 m section of the flume. All measurements were restricted to the centreline at the 3 m location, in order to minimize sidewall effects and be within this uniform zone.  90  Table 4.1: Hydraulic and sedimentological parameters for the flume, bed material, and periphyton clusters at high and low flow (HF and LF). Flume and cluster hydraulic conditions HF LF Discharge (L/s) ~8.5 ~4.2 Flow depth H (cm) (mean; range) 10.6; 8.9-12.1 7.2; 5.3-9.4 Flume width (cm) 15.0 15.0 Average aspect ratio 1.4 2.1 Average friction slope 0.002 0.001 1.004 1.004 Kinematic viscosity ν (m2/s) x 10-6 50.4 36.9 Average cross-sectional velocity Uxs (cm/s) 22,039.4 13,530.8 Average cross-sectional Reynolds number Re = UxR/ν 0.77 0.61 Average cross-sectional Froude number Fr = Ux/(gR)1/2 Average local depth-averaged velocity (Ux) (cm/s) (mean; range) 35.0; 33.5-36.2 13.5; 11.4-15.0 Bed properties Bulk Surface Cluster D16 (mm) 1.4 5.6 32.2 D50 (mm) 12.8 16.5 46.4 30.3 33.2 64.4 D84 (mm) 4.68 2.43 1.41 Geometric standard deviation S = (D84 /D16)1/2 Cluster height D (cm) (range) --2.5-3.9 Periphyton mat height h (cm) (range) --0.33-2.7 *Notes: Average values are for all trials, with and without periphyton, combined. ‘Local’ values are for measurements directly above the periphyton clusters. ‘Bulk’ bed properties include the coarse particles (> 2 mm) and sand-sized particles (< 2 mm) not present on the armored surface layer; ‘surface’ bed properties are for the bed material without the 20% sand fraction; ‘cluster’ includes only the particles comprising the periphyton clusters. D16, D50, and D84 are the 16th, 50th, and 84th percentile of the particle size distribution, respectively. We acknowledge that natural periphyton patches develop while adjusting to a complex set of flow and geomorphic cluster configurations that are difficult to reproduce in the laboratory. Thus an artificial construction of pebble clusters is unlikely to recreate the exact structure and density of periphyton assemblages found in the field. To address this problem, we attempted in our experiments to measure a wide range of possible conditions and configurations that would provide us with sufficient data for understanding general trends in flow-patch interactions. Pebbles covered by visibly filamentous periphyton assemblages were collected from Hope Slough, a tributary of Fraser River in south-western British Columbia. With a median particle size of ~46 mm, pebbles were at the upper limit of or slightly larger than the coarse fraction of the flume substrate (32-45 mm) and were rounded or rod-like in shape. Periphyton-covered pebbles were stored in the lab in a small holding tank containing river water and were used within three days of collection. We chose to collect natural periphyton assemblages, rather than cultivate them artificially because this approach allowed us to study naturally formed periphyton patches associated with pebble clusters and a large number of replicates over a wide and natural range of conditions. Similar approaches have been used in related studies and fields; for example, Dodds and Biggs (2002) transplanted field-grown periphyton mats for a study in a flume, while Nikora et al. (1998) measured turbulence dynamics of moss-covered stones placed on a flume bed. An advantage of this approach is that it can be replicated easily, allowing for the analysis of many different periphyton and bed structures.  91  For each experiment, three periphyton-covered pebbles were selected that had visibly similar biomass densities and arranged in a small cluster on the surface of the bed. Existing surface pebbles were removed prior to placement so that the cluster did not protrude above the bed surface. Clusters were approximately triangular in shape, depending on the shape of each stone. Although each cluster varied in size and shape, all were restricted to within a 13 cm width and most were ~10 cm wide, leaving ~2.5 cm of non-periphyton covered substrate on either side (Figure 4.3a). Given the narrowness of our flume, some caution must be taken against extending our results to wide flumes or natural channels. Dense mats in our flume redirected flow above and around the periphyton patch, but the narrow periphyton-free bed areas on either side of the patch (~1/6 flume width) may have partially restricted the flow – creating a ‘blockage’ effect. Flow redirected above the mat was also constricted and accelerated as a result. In wider flumes, where flow can move more freely around the patch, acceleration of flow above the patch should be weaker. However, the effect of geometrical blockage may not be significant. For instance, Cooper et al. (2007) showed that blockage effects of macrophytes in flumes were minor for blockage ratios (plant-to-flume area ratios) of ~10%, especially at higher velocities when plant bending and streamlining was high (30-70 cm/s). In this study, maximum blockage ratio was ~ 9% (assuming a maximum mat width and height of 13 and 3 cm, respectively and a flume cross-sectional area of 150 cm3) which is considered a ‘low’ ratio by Cooper et al. (2007; defined as < 26%). Flow rate was raised first to LF and then HF; three sequential velocity profiles were measured at each flow rate at the centre of each cluster. Each set of three profiles at each flow rate constitutes an ‘experiment’, identified by flow rate (HF or LF) and a letter A through Q identifying each cluster (Table 4.2); the three profiles of each experiment were combined into a single profile for analysis. Within each cluster there was a degree of internal patchiness related to the non-uniform distribution of filaments over a single stone or between stones. In order to quantify this spatial variability, we performed two sets of measurements on some of the larger clusters, one in the downstream section and one at the centre of the cluster. Three of the clusters – D, L, and K – were large enough to permit this; corresponding downstream measurements were identified as D2, L2, and K2. When bed elevation and ADV measurements were taken at different positions on the mat both the mat height and degree of flow penetration varied with location, indicating that periphyton mat structure differed between up- and downstream positions of the cluster. Thus each location was treated as an independent experiment. No downstream trends were observed between these paired sets of measurements on the same cluster. Velocity profiles were measured at each flow rate with a Sontek 16 MHz-Micro Acoustic Doppler Velocimeter (ADV) with a sampling rate of 50Hz and a velocity range setting of 100 cm/s. Each of the three profiles was constructed from a minimum of eight point measurements throughout the flow depth, spaced with increasing resolution near the bed; measurements were taken at ~0.1-0.25 cm intervals below 0.15 of the flow depth (H), at ~0.5 cm intervals from 0.15-0.3H and at 1 cm intervals above 0.3H. Because the ADV has a 5 cm sampling volume that extends below the tip of the probe, no measurements were taken within the top 5 cm of flow (Figure 4.3). The shallowest measurement was always taken with the ADV transmitter at the water surface (Figure 4.3a), while the deepest measurement was taken as close to the bed as feasible (Figure 4.3b) without  92  allowing the ADV sampling volume to include the bed (0.3-0.6 cm). All ADV measurements were time averages of at least 60 seconds in order to reduce error around the mean (Sontek, 1997) but were increased to 90 seconds in the case of high standard errors or low signal correlations (Martin et al., 2002). In addition to velocity profiles, for each experiment and flow rate the elevations of the water surface (WSE) and the top of periphyton mat (BEp) were measured with a digital Fowler Economy IP54 Water-Resistant Caliper mounted at the top of flume walls. A metal pointer was attached to the head of the caliper, positioned vertically downward from the top of the flume wall towards the bed. By opening the caliper until the pointer touched the top of the water surface or the periphyton mat, we could measure the distance from the top of the flume wall to both these levels. We then subtracted each distance from the distance between the flume wall top and the flume base, measured with the caliper before the sedimentary bed was put in place. Measurement uncertainty of the digital caliper is ±0.01 mm. After velocity profiles and elevation measurements were completed with the periphyton mat in place, we then stopped the flow and recorded the location and orientation of the three pebbles by inserting several wooden toothpicks into the bed around the perimeter of each stone. Toothpicks were inserted carefully so as not to disrupt the bed, but densely enough to clearly designate the location and position of each stone. Furthermore, care was taken while removing and handling the pebbles to retain the orientation of the stones in all three dimensions. Once removed, pebbles were scrubbed clean with a toothbrush, rinsed with deionized water, and returned to their exact location by placing them into the divot outlined by the toothpicks. Because of this attention to detail, we are confident that the removal and replacement of the stones did not alter their positions significantly. We acknowledge that some slight shift in location or orientation may have occurred, but we believe this to be negligible relative to the variability in hydraulic measurements. Rinse solution from the scrubbed stones was saved for AFDM analysis (described below). Flow was then restarted and velocity profiles were re-measured in the exact same location at both flow rates. Bed (BE) and water surface elevations were also re-measured using the digital caliper in the same manner as described for the clusters with periphyton. Flow depth (H) was determined by subtracting the BE of the pebble bed without periphyton from the average WSE measured with the caliper both with and without periphyton (Figure 4.3). WSE did not change significantly between these two conditions; mean variance between paired WSE measurements was <5%. Periphyton thickness (h) was calculated as the difference between the top of the periphyton mat (BEp) and the bed (BE), both measured with the caliper. Flow depth and periphyton thickness were then used to calculate the relative submergence – the ratio of flow depth (H) to plant height (h) – of the periphyton mat. Relative submergence indicates the degree of flow constriction of the overlying flow and thus the extent to which velocity may be increased (Nepf and Vivoni, 2000). A decrease in periphyton thickness between LF and HF was used to assess whether compaction of the mat increased with flow rate. Periphyton height (h) and the height of the pebble clusters (D, estimated by averaging the short axes of the cluster stones) were used to calculate the plant-to-roughness height ratio, h/D (Nikora et al., 2002a).  93  Figure 4.3: Measurement set-up for periphyton patches. Schematic of a) typical cluster shape and orientation from above; and (b-e) the side view of a mat, ADV probe location, and characteristics of the measurement set-up for b) an upright mat under low flows with the probe at the water surface (shallowest measurement), c) a compacted mat under high flows with the probe at its deepest measurement, and measurements used to calculate the effective bed elevation (BEeff) and thus characterize a mat as either d) open or e) closed, including the ADV sample volume, the height detected by the ADV, the elevation and height of the periphyton mat (BEp and h), the cluster elevation (BE), the water surface elevation (WSE) and the flow depth H ( ).  a)  ~2.5 cm ~10 cm  Flow  ~2.5 cm Flume wall  ADV Probe b)  WSE  c)  ADV Sample Volume (5 cm)  H H  ADV Detected Near-bed Height (~0.3-0.6 cm )  BE + h + H = WSE BE + h = BE p  h  BEp  h  BEp  BE Flume Base d)  WSE  e)  ADV Sample Volume (5 cm)  ADV Sample Volume (5 cm)  ADV Detected Height  BEp  ADV Detected Height  BEeff  BEp  BEeff BEp > BEeff = Open  BEp = BEeff = Closed  WSE – ADV Sample Volume – ADV Detected Height = BE  eff  94  In addition to mat height, relative submergence, and plant-to-roughness height ratio, we used two other parameters to characterize the periphyton mats: degree of flow penetration and biomass density. We used the ability of the ADV to detect a signal between periphyton filaments within the mat as a measure of flow penetration. Highly compacted, densely packed periphyton filaments interfered with the ADV sampling volume, preventing measurements within the mat (Figure 4.3e). However, if the assemblage of filaments was sparse enough, the ADV sampling volume could penetrate between filaments and measure flow between them (Figure 4.3d), although the plant material comprising the filaments is itself not acoustically transparent. A similar approach has been used with an ADV in an artificial grass meadow, where a swath was cut from the stand to provide room for the ADV to penetrate and measure velocity between grass blades (Nepf and Vivoni, 2000). In our study, low-density, non-compacted mats provided adequate space for the ADV signal to penetrate within the mat, between the filaments, so that velocity measurements could be taken. Thus we considered the occurrence of ADV interference as an indicator of filament density and compaction, separating our test subjects into ‘closed’ and ‘open’ mats (Sand-Jensen and Pederson, 1999). A closed mat effectively raised the elevation of the bed (‘effective bed elevation’, BEeff), the surface of the mat becoming physically analogous to the sedimentary bed, whereas the BEeff of an open mat was the top of the stones at the base of the mat. In order to calculate the BEeff detected by the ADV, we placed the probe at the water surface and took a measurement. With the probe at the surface, the ADV measured the transmitter height above the bed – the ‘bed’ being either the stones at the base of an open mat or the top of a closed mat. By subtracting the measured height above the bed from the total distance between the flume base and water surface (WSE), we obtained the effective bed elevation (BEeff) (Figures 4.3d and 4.3e). Effective bed elevation (BEeff) was compared to actual periphyton bed elevation measured with the caliper (BEp) to determine whether the mat was open or closed. If the elevation of the periphyton mat was greater than the effective bed elevation (BEp > BEeff), the ADV was able to penetrate and the mat was considered open (Figure 4.3d); the opposite was true for closed mats (Figure 4.3e). In other words, if the ADV measured a bed elevation that was beneath the top of the periphyton mat (BEeff < BEp), we assumed the mat was sparse enough for the signal to penetrate between filaments and we considered the mat open. In order to measure biomass density, rinse solutions from the scrubbed stones were homogenized, diluted to a known volume, and subsampled. A subsample was filtered through a pre-weighed glass-fiber filter, ovendried overnight at 65ºC, allowed to cool, weighed, ashed at 450ºC for four hours, and re-weighed to determine ash-mass (AM) and ash-free dry mass (AFDM). Ash-free dry mass was used to determine periphyton biomass (living and non-living), which was divided by the total stone surface area to calculate the areal biomass density (g/m2). Surface area of the three stones was determined by the method of Graham et al. (1988). All twenty experiments, with and without periphyton at HF and LF, covered a range of hydraulic conditions. A summary of hydraulic and sedimentological parameters is provided in Table 4.1. Because of the small width-to-depth ratio of the flume (aspect ratio < 5), turbulence anisotropy across the channel likely generated secondary currents that would be reflected in velocity measurements taken at the centre of the flume. High values of the transverse velocity component relative to the streamwise and vertical components provide  95  evidence for this effect. However, these secondary currents are assumed to have been present during all experiments, with and without periphyton. Thus, we compared mean velocity and turbulence statistics directly between clusters with periphyton and without and at identical large-scale conditions that included these secondary currents.  4.3 ANALYSIS  We computed several hydraulic and periphyton mat properties that characterized each cluster (Table 4.2). Hydraulic parameters include near-bed streamwise velocity, near-bed Reynolds stress, streamwise velocity profiles, and Reynolds stress profiles. Periphyton properties include mat type, mat height, relative submergence, plant-to-roughness height ratio, and biomass density. We then assessed periphyton effects in three ways. First, we compared turbulence and velocity parameters between mat types. Each parameter was averaged over all experiments corresponding to a specific periphyton type (Open, Closed, or None) and flow rate (HF, LF) and comparisons made between means using non-parametric Kruskal-Wallis and Wilcoxon rank sum tests. A nonparametric test was used because of the small sample sizes and unbalanced design. Second, we analyzed the relationship between periphyton mat properties and hydraulic parameters for all clusters and each flow rate. Third, we compared velocity and Reynolds stress profiles for single clusters with and without periphyton.  4.3.1 VELOCITY PROFILES AND PARAMETERS  Data output from each ADV measurement includes the covariance of the streamwise and vertical velocity fluctuations (u’ and w’, respectively), equal to <u’w’>, where ‘< >’ indicates the average; the root-mean-square of the velocity fluctuations in all three directions (equal to < u' 2 > , < v' 2 > , and < w' 2 > , where v’ is the lateral component); and mean velocities in all three directions (u, v, and w). For each combined velocity profile (for each experiment), we calculated the depth-integrated streamwise velocity (Ux), maximum streamwise velocity (umax) and near-bed streamwise velocity (u0), taken as the value of the measurement closest to the bed (0.3-0.6 cm). We acknowledge that the depth-integrated and maximum streamwise velocities calculated from our profiles are not representative of the full water column because the ADV cannot measure in the top 5 cm of flow; these measures are, however, representative of the near-bed and lower region of flow, the intended focus of this study. Each velocity parameter was averaged for all experiments in one of six categories based on the combination of flow rate (HF or LF), periphyton presence (Periphyton or None), and for the periphyton categories, mat type (Open or Closed). For each experiment, flow, and condition, we fit regression lines to u plotted as a function of height above the bed (z). For each experiment, an analysis of covariance (ANCOVA) was performed to determine whether periphyton presence had a significant effect on the velocity regression. Because ANCOVA relies on the assumption of linear effects, the data were first transformed to give a linear relation. Residual plots were then checked to ensure there were no trends in the residuals. In all cases, a logarithmic transformation of height resulted in suitably linear regressions. Velocity was treated as the response  96  variable, periphyton presence as a categorical independent variable, and ln(z) as the covariate. Given the high variability inherent to heterogeneous substrates and periphyton structures, as well as the exploratory nature of this analysis, we used a less conservative p-value of 0.1 to detect whether periphyton had a significant effect on the velocity regression.  4.3.2 TURBULENCE PROFILES AND PARAMETERS Vertical distributions of velocity fluctuations (<u’2>, <v’2>, <w’2>) were compared between periphyton and non-periphyton experiments. All distributions were normalized by total flow depth (H) and the depth-integrated velocity (Ux); Ux is considered the most relevant velocity scale because shear velocity (u*) is difficult to accurately determine in rough flows (Strom and Papanicolaou, 2007). Velocity fluctuations at each location in the vertical profile were also used to calculate local estimates of Reynolds stress at each depth from the following equation (Biron et al., 2004):  τ R = − ρ < u ' w' > where ρ is the density of water (assumed to be 1000 kg/m3). We plotted the distributions of Reynolds stress and velocity fluctuations for comparison between periphyton and non-periphyton conditions. We consider the Reynolds stresses closest to the bed and at the top of the inner boundary layer reflective of the locally induced effect of the periphyton mat. Thus we use the nearbed value of Reynolds shear stress (τRe0) and the maximum Reynolds shear stress (τReM, assumed to be the top of the inner boundary layer) for statistical comparisons between mat types; we use τRe0 in our regressions with mat height and density. Typically, the ADV experienced bed interference below ~0.3-0.6 cm thus the near-bed measurement was always somewhere within this range. Although this cannot be confirmed from our measurements, we suspect that for high density mats ‘near-bed’ refers to a point above the mat rather than within the mat, but above the solid substrate. Thus in some trials, the height of the near-bed measurement was higher for the non-periphyton condition, indicating that the mat generated less interference in the near-bed flow layer. For mats less than 0.3-0.6 cm tall, the ADV cannot reach within the mat but can measure at the mat surface if interference is low (i.e. if only the solid substrate generates interference in the near-bed layer). In fact, we found that for the periphyton condition, the near-bed height increased with mat height (Figure 4.4), indicating that very tall mats (up to 3 cm) generated more interference (up to 0.6 cm). Non-periphyton clusters (with ‘mat heights’ of zero), however, had near-bed heights spanning the range of 0.3-0.6 cm. In other words, although there is a weak relation between mat height and ADV detected near-bed height, the total variance is greater than the variance in near-bed height over non-periphyton clusters.  97  Figure 4.4: Relation between the height of the periphyton mat and the lowest height at which the ADV could take a measurement without interference from the bed.  Near-bed height (cm)  0.7 0.6 0.5 0.4 0.3 0.2 Periphyton  0.1  None 0 0  0.5  1  1.5  2  2.5  3  Mat height (cm)  *Notes: Non-periphyton clusters (‘None’; open circles) are considered to have ‘mat heights’ of zero. We consider the near-bed Reynolds shear stress as representative of the Reynolds bed shear stress  4.4 RESULTS  Mat characteristics, velocity parameters and shear stress estimates, coefficients of determination and ANCOVA p-values from velocity-height regressions for each experiment and flow rate are given in Table 4.2. Averages for selected characteristics and parameters for each category of periphyton mat structure (Closed, Open, or None) and flow rate are given in Table 4.3. Results from Kruskal-Wallis and Wilcoxon tests for each mat type and flow rate are in Table 4.4.  98  Type  Open Open Open Closed Open Closed Open Closed Closed Closed Open Open Open Closed Open Open Open Closed Open Open  Experiment  A B C D D2 E F G H I J K K2 L L2 M N O P Q  a) HF  AFDM (mg/cm2) 1.01 2.04 1.92 3.16 3.16 2.23 2.17 2.06 1.63 1.05 1.19 1.50 1.50 1.37 1.37 1.88 3.74 3.15 1.11 0.97 0.34 0.12 0.24 0.81 0.98 0.26 0.36 0.49 0.32 0.42 0.26 0.29 0.51 0.14 0.60 0.11 0.65 0.31 0.32 0.18  h/D  h (cm) 1.2 0.4 0.6 2.2 2.7 0.7 1.3 1.5 1.2 1.3 0.9 0.9 1.5 0.4 1.9 0.4 1.9 0.9 1.2 0.7 8.3 24.9 15.9 5.4 4.3 15.3 8.5 7.9 9.0 9.0 12.1 13.7 7.9 22.3 5.7 26.0 5.9 12.2 8.6 16.0  H/h  AFDM/h (mg/cm3) 0.86 5.09 3.20 1.41 1.16 3.24 1.70 1.34 1.31 7.93 1.25 1.77 0.99 3.19 7.30 4.85 1.99 3.66 9.22 1.47  Periphyton U0 umax (cm/s) (cm/s) 7.86 47.57 4.23 25.35 -0.07 46.34 48.71 22.19 22.81 47.39 11.35 45.83 33.27 48.54 2.87 44.57 28.87 46.75 8.52 54.58 2.36 54.15 17.2 54.65 7.61 54.94 17.85 50.84 0.44 51.33 2.98 50.68 6.41 50.85 20.04 49.5 10.45 51.98 14.06 50.14 Ux (cm/s) 31.43 35.85 28.82 38.97 37.15 31.35 41.23 28.76 41.23 37.44 36.8 40.36 37.83 38.83 29.84 33.55 31.18 37.11 37.71 38.34 τRe0 (N/m2) -0.26 0.07 0.004 -0.11 -0.09 0.06 0.62 0.11 0.69 -0.13 0.01 0.40 0.06 0.15 0.04 0.09 0.35 1.00 -0.02 0.58 0.87 0.93 0.83 0.95 0.91 0.94 0.94 0.98 0.93 0.98 0.95 0.99 0.97 0.94 0.87 0.98 0.97 0.99 0.97 0.98  r  2  None u0 (cm/s) 30.2 15.2 -0.66 46.4 12.4 19.6 16.3 7.70 0.14 22.8 9.52 25.5 9.91 31.1 1.45 16.05 9.82 8.02 25.29 24.55 umax (cm/s) 45.0 30.2 43.3 10.4 45.7 47.0 46.4 45.3 46.5 54.4 52.4 52.3 52.6 49.2 50.5 50.2 50.4 48.5 49.8 49.5  Ux (cm/s) 39.47 34.31 27.73 25.62 33.89 30.56 32.12 28.24 27.82 38.96 36.56 40.09 37.79 39.04 24.73 34.43 30.97 30.53 38.23 38.49 τRe0 (N/m2) 0.62 0.38 0.59 -0.17 -0.11 -2.79 0.49 0.44 0.84 0.16 1.8 0.74 -26 1.16 0.31 1.7 -0.83 1.65 1.26 1.19 0.95 0.98 0.95 0.93 0.92 0.98 0.96 0.98 0.92 0.95 0.98 0.97 0.97 0.92 0.79 0.98 0.98 0.95 0.94 0.96  r2  <0.001 0.070 <0.01 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.05 <0.05 0.972 0.963 <0.001 <0.001 <0.05 0.084 <0.001 0.159 0.760  p-value  Table 4.2: Mat characteristics, velocity and turbulence parameters, coefficients of determination and ANCOVA p-values from velocity-height regressions for each experiment at a) HF and b) LF.  99  Open Open Open Open Closed Closed Open Closed Open Open Open Closed Open Closed Open Open Open Closed Open Open  A B C D D2 E F G H I J K K2 L L2 M N O P Q  AFDM (mg/cm2) 1.01 2.04 1.92 3.16 3.16 2.23 2.17 2.06 1.63 1.05 1.19 1.50 1.50 1.37 1.37 1.88 3.74 3.15 1.11 0.97 0.36 0.17 0.24 0.86 0.94 0.32 0.50 0.61 0.52 0.44 0.41 0.25 0.34 0.11 0.60 0.12 0.63 0.52 0.31 0.30  h/D  h (cm) 1.2 0.5 0.6 2.4 2.6 0.8 1.8 1.9 2 1.4 1.5 0.7 1 0.3 1.9 0.4 1.8 1.5 1.2 1.1 5.5 15.0 11.6 3.5 3.3 9.5 4.1 5.0 4.1 5.8 5.2 10.9 7.9 21.1 4.3 16.5 4.3 5.8 6.2 7.2  H/h  AFDM/h (mg/cm3) 0.83 3.77 3.20 1.32 1.22 2.66 1.22 1.08 0.80 0.75 0.79 2.03 1.48 4.14 7.30 4.48 2.08 2.17 0.95 0.91  Periphyton u0 umax (cm/s) (cm/s) 3.26 19.01 1.84 14.66 0.01 16.72 24.31 11.91 19.72 26.92 0.4 24.62 25.32 25.79 0.38 21.4 11.49 24.27 10.73 28.19 0.14 11.34 4.23 27.02 0.75 24.3 4.12 23.86 0.02 21.98 1.28 25.9 1.03 21.39 10.29 29.34 0.04 24.87 3.19 23.43 Ux (cm/s) 5.14 16.23 4.00 12.76 21.54 10.94 24.85 10.86 17.22 17.03 2.8 14.57 11.41 12.44 8.71 9.69 8.33 19.35 8.82 13.09 τRe0 (N/m2) -0.11 -0.02 0.0002 -0.10 1.16 -0.005 0.36 0.02 -0.30 -0.19 0.01 0.02 -0.01 0.07 0.005 0.009 0.01 0.25 0.0005 0.04 0.57 0.77 NA 0.79 0.57 0.88 0.092 0.91 0.67 0.90 0.55 0.97 0.92 0.96 0.96 0.93 0.92 0.75 0.85 0.95  r  2  None u0 (cm/s) 13.03 7.98 -3.98 21.71 3.8 17.3 7.06 6.06 -4.39 10.72 5.00 15.29 14.11 15.31 -0.54 7.36 14.17 4.44 13.42 10.41 umax (cm/s) 27.9 10.9 21.2 9.95 20.8 22.6 20.8 29.3 24.1 27.8 23.9 28.8 24.1 27.2 20.3 24.8 24.4 19.9 25.9 28.3 Ux (cm/s) 20.83 15 5.68 15.37 15.19 12.46 14.39 14.79 7.73 18.37 11.18 20.68 11.88 20.74 6.46 14.16 7.76 10.61 18.01 20.13  τRe0 (N/m2) 0.50 0.18 0.91 0.16 -0.09 0.31 0.34 0.42 1.04 -0.20 0.70 0.30 -5.32 0.38 0.01 0.98 -2.76 4.32 0.63 0.49 0.88 0.84 NA 0.77 0.92 0.94 0.95 0.90 0.67 0.90 0.69 0.93 0.81 0.88 0.97 0.95 0.08 0.80 0.89 0.97  r2  <0.001 <0.01 <0.001 <0.001 <0.001 <0.001 <0.001 0.147 <0.001 0.130 <0.001 0.058 0.112 0.710 0.900 <0.001 0.265 <0.001 <0.001 0.069  p-value  100  *Notes: See text for definition of mat type. AFDM = ash-free dry mass; h/D = plant-to-roughness height ratio; h = measured mat height; H/h = relative submergence of mat = flow depth/mat height; AFDM/h = volumetric density of periphyton. u0, Ux, and umax are the nearbed, depth-integrated, and maximum streamwise velocity, respectively. Near-bed values of the Reynolds shear stress (τRe0) were determined from 3-D velocity fluctuations of the near-bed measurement, assumed representative of the Reynolds bed shear stress. See text for equations. Significance (p-) values are for the difference in slope between velocity profiles over periphyton-covered clusters and profiles over the same clusters without periphyton, indicating whether or not periphyton had a significant effect on the velocity profile.  Type  Experiment  b) LF  Table 4.3: Periphyton structure, velocity, and turbulence statistics for periphyton-covered and nonperiphyton clusters.  n  AFDM (mg/cm2)  h (cm)  AFDM/h (mg/cm3)  u0 (cm/s)  Ux (cm/s)  umax (cm/s)  τRe0 (N/m2)  HF C  7  2.09(0.3)  1.19(0.2)  2.13(0.4)  19.75(5.8)  36.24(1.7)  44.89(4.0)  0.25(0.2)  O  13  1.81(0.2)  1.19(0.2)  1.99(0.4)  9.97(2.7)  32.63(2.9)  48.76(2.1)  0.14(0.1)  N  20  NA  NA  NA  16.56(2.7)  31.76(2.0)  45.98(2.2)  0.83(1.4)  C  6  2.25(0.3)  1.31(0.3)  2.22(0.4)  6.52(2.8)  14.95(1.7)  25.53(1.1)  0.25(0.2)  O  14  1.83(0.2)  1.35(0.2)  1.73(0.3)  6.17(2.4)  10.67(1.8)  21.14(1.5)  -0.01(0.04)  N  20  NA  NA  NA  8.91(1.6)  13.32(1.3)  23.15(1.2)  0.16(0.4)  LF  *Notes: Values are averages (SE) of all experiments (n = number of experiments) in one of six categories based on flow rate (HF = High, LF = Low) and type of periphyton mat (C = Closed, O = Open, or N = None). AFDM and AFDM/h are the areal and volumetric density of periphyton mats, respectively; h is mat height. u0, Ux, and umax are the near-bed, depth-integrated, and maximum streamwise velocity; τRe0 is the near-bed value of the Reynolds shear stress. See text for definition of mat types, flow rates, and equations. Table 4.4: Significance values (p-values) for non-parametric Kruskal-Wallis (KW) and Wilcoxon tests for several mat characteristics and hydraulic parameters compared between stone clusters with one of two periphyton mat types – Closed (C) or Open (O) – or without periphyton (N) for two flow rates. Parameter KW High flow (HF) AFDM <0.001 H <0.001 AFDM/h <0.001 0.128 u0 0.223 umax Ux 0.393 0.143 τRe0  C v. N  O v. N  C v. O  <0.001 <0.001 NA 0.685 0.808 0.240 0.268  <0.001 <0.001 NA <0.1 0.141 0.372 <0.1  0.383 0.968 0.699 <0.1 0.157 0.606 0.606  KW C v. N Low flow (LF) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.215 0.324 0.212 0.388 0.275 0.882 <0.001 <0.1  O v. N  C v. O  <0.001 <0.001 <0.001 0.112 0.323 0.169 <0.001  0.160 0.901 0.179 0.547 <0.1 0.207 0.386  Differences between closed mats, open mats, and non-periphyton clusters are suggestive, but not conclusive. Areal densities of periphyton ranged from ~10-38 g/m2 and mat heights were between 0.4-2.7 cm. On average, closed mats had higher AFDM than open mats (Table 4.3), but this difference is not significant (Table 4.4) thus we cannot generalize a relation between biomass density and flow penetration. Both closed and open mats were taller at low flow (LF) than at high flow (HF), indicating less compaction at LF; at HF, both types of mat were equally compacted. At HF there were slightly more closed mats (n = 7) than at LF (n = 6), suggesting that higher flows may have increased mat compaction and interference, reducing flow penetration for one experiment, but the difference in number is too small to make any firm conclusions about this effect. Mat submergence ranged from 3.3-26.1 (mean = 9.5), spanning the range of low to deep submergence (Nepf and Vivoni, 2000; Raupach et al., 1996). All periphyton mats were shorter than the roughness height of the stone clusters (h < D), with h/D ranging from 0.11-0.94 (mean = 0.40).  101  4.4.1 HYDRAULIC CONDITIONS AND PERIPHYTON CHARACTERISTICS  Plots of near-bed velocity (u0) and near-bed Reynolds stress (τRe0) against periphyton density and height (Figure 4.5) at both HF and LF show highly scattered relations.  Figure 4.5: Near-bed streamwise velocity (u0) and near-bed Reynolds stress (τRe0) as a function of areal density (measured as AFDM per surface area) (a, c) and mat height (b, d) of periphyton clusters at two flow rates (HF and LF).  Near-bed velocity (cm/s)  60  HF HF LF LF  a)  b)  HF NP M ean  50 40 30 20 10  Near-bed Reynolds stress (N/m 2)  0 -10 5  c)  d)  4 3 2 1 0 0  1  2  3 2  Areal density (g/m )  4 0  1  2 3 Mat height (cm)  4  *Notes: Solid lines indicate the mean values for non-periphyton (zero density clusters). See text for details. Some measurements suggest a possible increase in near-bed velocity with areal density and height (e.g. HF velocities vs. Height, Figure 4.5b), but overall the measurements do not show a clear trend. For instance, several mid-density (~2 mg/m2) clusters and two high density clusters (~3-4 mg/m2) have comparatively slow near-bed velocities. Mats with heights of 0.25-2 cm span a range of velocities (nearzero up to 35 cm/s); only the tallest mat has a notably faster u0 (at HF and LF). At both HF and LF, most periphyton clusters have lower near-bed velocities than the average for non-periphyton clusters although some of the mid- to high-density mats are clearly higher. Near-bed Reynolds stress shows no clear trend with density or height. However, the comparison with non-periphyton clusters is notable; all periphyton  102  clusters have lower-than-average near-bed Reynolds stresses at LF. In addition, mats with heights < 1 cm have near-bed Reynolds stresses lower than all non-periphyton clusters at LF. At HF, only four periphyton clusters have higher-than-average near-bed Reynolds stress; these mats span the range of densities, but fall within the 0.5-1.5 cm height.  4.4.2 VELOCITY PROFILES  Other than experiment C at LF, all regressions of u versus ln(z) are significant fits (p < 0.001) with rsquared values > 0.5 for all experiments and > 0.8 for most (Table 4.2). Thirty out of the forty LF and HF experiments are significantly different between the periphyton and non-periphyton run (Table 4.2; p < 0.1). Vertical distributions of u are plotted against the normalized height above the bed for representative closed and open mats (experiments D and B, respectively) at each of the two flow rates (Figure 4.6). No clear pattern emerges for the ten non-significant experiments; they encompass the full range of densities and heights, are both open and closed, and occur at both flow rates. However, several clusters have velocity profiles that are significantly different at one of the two flow levels but not at the other flow level (e.g. experiments K, L, P, and Q). Upon examining these individual cases, we see that although the dominant trends of the regressions are not significantly different the near-bed region of the profile is slower for the periphyton run. Furthermore, these experiments show a significant effect of periphyton at one flow but not at the other thus suggesting that bulk flow rate influences the magnitude of periphyton effect. Four examples are shown in Figure 4.7, comparing the non-significant and significant run for each experiment.  103  Figure 4.6: Streamwise velocity (u) profiles from representative closed (D) and open (B) mats for two flow rates (HF ~8.5 L/s, LF ~4.2 L/s). 0.6  a) Experiment DHF  b) Experiment DLF  0.5 Periphyton None  z/H  0.4 0.3 0.2 0.1 0.0 0.6  d) Experiment BLF  c) Experiment BHF  0.5  z/H  0.4 0.3 0.2 0.1 0.0 0  10  20  30 -1  u (cm s )  40  50 0  10  20  30  40  50  -1  u (cm s )  *Notes: 'Periphyton' is the three-cobble cluster with a mat of filamentous green algae; 'None' is the identical cluster without any algae.  104  z/H  Figure 4.7: Streamwise velocity (u) profiles at two flow rates (HF ~8.5 L/s, LF ~4.2 L/s) from representative experiments K, L, P, and Q for which logarithmic regression lines are not significantly different at one flow rate, but the near-bed region of the profile is slower for the periphyton run and the regression is significantly different at the other flow rate. 0.6 a) Experiment KHF (Not Significant) 0.5 Periphyton 0.4 None 0.3 0.2  b) Experiment KLF (Significant)  0.1 0.0  z/H  0.6 0.5  c) Experiment LHF (Significant)  d) Experiment LLF (Not Significant)  e) Experiment PHF (Not Significant)  f) Experiment PLF (Significant)  0.4 0.3 0.2 0.1 0.0 0.6  z/H  0.5 0.4 0.3 0.2 0.1 0.0 0.6  g) Experiment QHF (Not Significant) h) Experiment QLF (Significant)  0.5 z/H  0.4 0.3 0.2 0.1 0.0 0  20  40 u (cm s-1)  60 0  10  20  30  u (cm s-1)  *Notes: 'Periphyton' is the three-cobble cluster with a mat of filamentous green algae; 'None' is the identical cluster without any algae. Plots of these paired significant-non-significant experiments show the consistent reduction in near-bed velocity for the periphyton runs, even those that are not significant, illustrating the pattern observed in mean hydraulic parameters (Table 4.3). On average, near-bed velocity (u0) was significantly slower for  105  open mats than for closed mats or non-periphyton clusters at LF (p < 0.1), but not at HF. As seen in the plots, the reduction in u0 is more pronounced at LF. At HF, maximum velocity (umax) was significantly faster for open mats than for closed mats, but the differences in umax between periphyton and nonperiphyton mats are not significant. Low near-bed velocities and high maximum velocities created a steep velocity gradient above open mats. Although the reported effects are not all significant, these results lend support to the hypothesis that flow compression above periphyton mats increases streamwise velocities, while flow resistance near the surface of open mats decreases velocities (Figure 4.2a). Most likely, the differences in velocity parameters among groups are not always significant because of the large variation in bed configurations and local hydraulic conditions of individual clusters.  4.4.3 TURBULENCE PROFILES AND PARAMETERS Normalized vertical distributions of velocity fluctuations in the three dimensions (<u’2>/ Ux 2, <v’2>/Ux2, <w’2>/Ux2) for representative experiments D and B at HF and LF (Figure 4.8), show the high magnitude of the lateral and streamwise components and relatively lower vertical fluctuations. High lateral stresses may result from secondary currents generated by coarse substrates and the flume wall. However, for all three directions, both experiments, and both flows, velocity fluctuations in most portions of the flow were lower in the experiment with periphyton than without periphyton. For open experiment B, turbulence reduction is most pronounced in the lower portion of the flow (z/H < 0.2), whereas for closed experiment D turbulence is damped throughout the flow depth. Plots of normalized Reynolds stress distributions (τRe0/ρUx2) for the two experiments at HF and LF (Figure 4.9) demonstrate the same pattern of lower stresses in the presence of periphyton. Also notable is the distinct shift in slope at ~0.15-0.25 of the flow depth and the decrease in shear towards the bed characteristic of shear profiles above rough beds (Biron et al., 1998; Voulgaris and Trowbridge, 1998; Martin et al., 2002; Nikora and Goring, 2002).  106  2  2  Figure 4.8: Normalized vertical distributions of velocity fluctuations in the three dimensions (■ <u’ >/Ux , 2  2  2  2  ▲ <v’ >/Ux , ● <w’ >/Ux ) from closed and open mats D and B, respectively, at HF and LF. 0.6  a) Experiment DHF  0.5 z/H  0.4 0.3 0.2 0.1 0.6 0.5  b) Experiment BHF  z/H  0.4 0.3 0.2 0.1 -0.01 0.5  0.01  0.03  0.05  0.07  0.09  0.11  0.13  0.15  c) Experiment DLF  z/H  0.4 0.3 0.2 0.1 0.0 0.5  d) Experiment BLF  z/H  0.4 0.3 Periphyton ■ <u’2>/Ux2 ▲ <v’2>/Ux2 ● <w’2>/Ux2  0.2 0.1 0.0 0.0  0.1  0.2  0.3 2  2 x  2  None □ <u’2>/Ux2 ∆ <v’2>/Ux2 ○ <w’2>/Ux2 0.4  2  2  <u’ >/U , <v’ >/Ux , <w’ >/Ux  0.5  0.6  2  *Notes: U = depth-integrated streamwise velocity in m/s. Please note the difference in scale between HF x  and LF plots.  107  2  Figure 4.9: Normalized vertical distributions of Reynolds stress (♦ τ /ρU ) from closed and open mats D Re  x  and B, respectively, at HF and LF.  z/H  0.6 0.5  a) Experiment DHF  0.2  0.5 0.4 0.3 0.2  0.1  0.1  0.4 0.3  0.6 0.5 z/H  0.6  b) Experiment BHF  0.3 0.2  0.6 0.5 0.4 0.3 0.2  0.1  0.1  0.4  -0.01  0.0  0.01  0.02  0.03  0.04  -0.05  c) Experiment DLF  Periphyton None  d) Experiment BLF  0.0  τRe/ρUx2  0.05  0.1  2 x  τRe/ρU  *Notes: Ux = depth-integrated streamwise velocity in m/s; ρ = density of water (assumed equal to 1000 kg/m3). Please note the difference in scale between HF and LF plots. Near-bed values of Reynolds stress (τRe0) for closed mats are the same at HF and LF, but are increased relative to the non-periphyton run only at HF (Table 4.3). At HF τRe0 is significantly higher for closed mats than open mats and non-periphyton clusters. Open mats have significantly lower near-bed Reynolds stresses than non-periphyton clusters at LF and both closed mats and non-periphyton clusters at HF (Tables 4.3 and 4.4).  4.5 DISCUSSION  Most previous periphyton-flow studies used extensive periphyton growths covering the entire bed surface (Nikora et al., 1998; Nikora et al., 2002a; Labiod et al., 2007). In contrast, we focus on individual, small-scale clusters and periphyton mats. Our scenario is more realistic for small, forested streams, where periphyton growth is patchy and heterogeneous, reflecting the distribution of flow fields, geomorphic features, light, and nutrients. Patch hydrodynamics, however, are highly transient and form from a combination of upstream roughness and local periphyton effects. Thus our observations describe a unique scenario that is not necessarily comparable to long, uniform beds of periphyton growth. For this reason, we use local velocity and turbulence parameters to characterize the effects of periphyton on hydraulic conditions, rather than spatially or depth-averaged hydraulic measures. We focus on the near-boundary  108  region of a single bed location, where the vertical velocity and turbulence profiles are more strongly influenced by local effects (i.e. the presence or absence of a periphyton mat) than upstream conditions.  4.5.1 PERIPHYTON EFFECTS ON VELOCITY PROFILES  As expected, filamentous periphyton mats altered near-bed flow hydraulics above pebble clusters. However, the effects were variable and depended on the structure of the periphyton assemblage, a function of both density and flow rate. Thirty out of the forty velocity profiles were significantly altered by the presence of periphyton. On average, however, not all hydraulic parameters were significantly different among periphyton types and non-periphyton clusters, likely because of the high variability in local hydraulic conditions between clusters. Such hydraulic variability and streambed heterogeneity is characteristic of coarse-bedded streams, as has been well-documented in the sedimentological literature for pebble clusters and other bed structures (Belanger and Roy, 1998; Strom and Papanicolaou, 2007). Differences between cluster configurations introduced a degree of variability that likely influenced the magnitude of periphyton effect, particularly for measurements in the near-bed region. Although the effects were not always significant, some general trends were observed that support our predictions; we acknowledge that these results are only suggestive, not conclusive. At high and low flows, closed mats increased depth-integrated velocities relative to non-periphyton surfaces. Closed mats raised the effective bed height (BEeff = BEp) and the resulting flow constriction in turn increased velocities throughout the flow depth. Similar results have been found above mixed diatom-algae assemblages (Nikora et al., 2002a), densely packed macrophyte canopies (Sand-Jensen and Pedersen, 1999) and some eelgrass beds (Gambi et al., 1990). In contrast, open mats decreased near-bed, depth-integrated, and maximum velocities at low and therefore, shallow flows. However at higher, and deeper, flows, open mats actually increased depth-integrated and maximum velocities – velocity reduction was restricted to the near-bed region. Velocity reduction within the mats of periphyton (Nikora et al., 2002a) and canopies of other vegetation types (Sand-Jensen and Pedersen, 1999) has been previously observed. Depth of flow relative to the height of the mat may explain the discrepancy between LF and HF. Velocity was reduced above open mats because of the increase in resistance and momentum exchange with the mat filaments. However, the vertical extent of velocity reduction scales with the height of the mat and reaches the same height at both flow rates. Above the local effect of the mat, determined by mat height, upstream (non-local) effects dominated. At high flows, the column of water above the mat was deeper than at low flows thus velocities in the upper section of flow were not reduced by the momentum exchange below. Similarly, near-bed velocities above closed mats were much higher at high flows, but this effect did not extend into the upper portions of the water column where velocities were slightly reduced or the same. At low flows, some nearbed velocities above closed mats were reduced while others were increased, likely due to the fact that lower flows and less compaction may have allowed for limited flow exchange in some cases. Given this variability, not all velocity parameters were significantly different between groups.  109  4.5.2 PERIPHYTON EFFECTS ON TURBULENCE  Although most velocity profiles above periphyton and non-periphyton clusters were relatively welldescribed by a logarithmic function of depth, turbulence stress profiles demonstrate the two-layer nature of the flow, including a near-bed ‘roughness sublayer’ and an overlying ‘logarithmic layer’ (Raupach et al., 1991, Nakagawa et al. 1991, Nikora et al. 2001). Nikora et al. (2002a) identified the boundary between these two regions as the height of peak shear stress, which occurred at ~0.15-0.25H for all experiments (with and without periphyton), similar to the values previously reported for a periphyton-covered bed (Nikora et al. 2002a) and stone clusters on a coarse-bedded stream (Strom and Papanicolaou, 2007). Our results can be viewed in the framework of a patch-created internal boundary layer (IBL), arising from the transition from a coarse bed without periphyton to a periphyton patch. According to this model, flow upstream over the sedimentary bed contains an ‘ambient’ boundary layer (ABL) comprised of both a roughness sublayer and an overlying logarithmic layer. Upon encountering the patch, an IBL develops at the base of the ABL, increasing in thickness downstream of the patch until it replaces the original ABL with a new roughness sublayer and logarithmic layer associated with the patch roughness elements. Such a phenomenon has been described for a patch of estuarine bivalves (Nikora et al., 2002b), in which the patchderived IBL was a region of lower streamwise velocities but higher turbulent energies relative to the upstream ABL. In our study, the coarse bed without periphyton supported a fully developed ABL; placement of the periphyton patch created an isolated bed area with a different roughness that generated an IBL beneath the existing ABL. Our measurements were at a single vertical at the centre of the patch thus they represent a single section within the growing IBL. We can estimate the thickness of this boundary layer ( δ ) with a simple empirical equation (U.S.A.C.E, 1965) in which thickness is a function of roughness length (z0) and the length of the wetted surface where the boundary layer takes place (L):  δ L  = 0.08(  L −0.223 ) z0  We can use the length of each cluster for L, estimated by doubling the mean intermediate axis of each pebble within the cluster (mean = 9.0 cm), and use mat height (h) as the roughness length. Applying these values to the equation, we compute boundary layer thicknesses ranging from 0.35-0.56 cm at the downstream edge of the mat, corresponding to ~0.05H, for both HF and LF. Predicted boundary layers are therefore thinner than observed at the centre of the mat. However, applying an estimate of bed roughness length (0.25D84 of cluster height) (Strom and Papanicolaou, 2007) to the equation, boundary layer thickness for all clusters is also estimated to be ~0.5 cm or 0.05H. Thus the equation underestimates δ for clusters with and without periphyton. An alternative method incorporates the Reynolds number of the flow (Re) and the length over which flow occurs (L) (Aerospace, Mechanical & Mechatronic Engineering, 2005):  110  δ L  =  0.385 Re  0.02  If we use mat height (h) as the characteristic length-scale in the computation of Re and cluster length as the length of flow, predicted boundary layer thicknesses range from 0.53-1.15 cm (HF mean = 0.68; LF mean = 0.81) or ~0.05-0.12H, which are less than but closer to those observed. Using just particle roughness height in the calculation of Re produces slightly smaller thicknesses. Some studies have suggested that boundary layer thickness can be defined as the height above the bed at which the local velocity approaches the free stream mean velocity (Strom and Papanicolaou, 2007). For our clusters with and without periphyton, this height was ~0.15-0.25H, which is greater than predicted by these equations. Nevertheless, these equations demonstrate that the length of the cluster was sufficient for a new boundary layer to develop, from ~0.35 – 1.15 cm in height, as flow moved over the periphyton mat. Filamentous periphyton mats had a distinct effect on the magnitude of turbulent stresses throughout the water column. At HF, both closed and open mats reduced the near-bed value of Reynolds stress relative to the same clusters without periphyton mats. Closed mats may have resulted in reduced turbulence at HF because highly compacted, flexible filaments created a hydrodynamically smooth boundary with few roughness elements, streamlined flow, and limited eddy development, similar to the findings of Godillot et al. (2001). However, this finding runs somewhat counter to previous studies of high density periphyton mats, which found that as periphyton assemblages developed and heightened, shear velocity and shear stress increased due to the increase in surface area and resistance (Reiter, 1989a; Reiter, 1989b; Labiod et al., 2007). Discrepancies in effect may be partly due to the structural and hydraulic differences between our periphyton patches and the fully-covered periphyton bed. Because of the three-dimensional and transient nature of patch flow dynamics, some assumptions and models relevant to uniform plant coverage may be inapplicable in this case. Furthermore, small-scale variability in the morphologies of different periphyton species may influence the effect of mat growth and development on turbulent properties of the flow. In our study, it is possible that the closed periphyton mats had virtually impenetrable surfaces with no velocity gradient and low hydraulic roughness – thus turbulence was also low. Negligible vertical turbulence penetration into the periphyton mat (‘interfacial sublayer’) was also observed by Nikora et al. (2002a). Furthermore, Labiod et al. (2007) observed ‘phases of smoothing’ during periphyton development that might have damped turbulence intensities. Variation among our closed mats was similar: in some cases, turbulence is reduced by smoothing, while in others a degree of momentum exchange occurred, reducing near-bed velocities and increasing near-bed turbulence (e.g. experiments HFO, LFD2). Increased turbulence was more prevalent at LF, lending support to the idea that when mats were less compacted, some closed mats may have become slightly opened to the flow and turbulent motion increased. This may also explain why on average, at LF the near-bed value of the Reynolds stress was significantly higher for closed mats than for non-periphyton clusters. Although the effect of compaction at high flows cannot be explicitly tested by our measurements, our results suggest it is a factor that should be considered when assessing the hydrodynamic impact of vegetation. A model of periphyton-flow interactions developed by  111  Nikora et al. (1998) proposes that high flows and larger eddies should increase vertical flow penetration, but our results suggest this must be balanced with potential compaction effects. For highly flexible and dense filamentous assemblages, compaction may play a dominant role. Future study should consider how the interplay between eddy size and compaction at high flows affects the degree of penetration and momentum exchange. Open mats also reduced the near-bed value of Reynolds stress at HF and LF. However, it is important to point out that whereas ‘near-bed’ refers to the mat surface in the case of the closed mat, it refers to the base of the open mat within the algal filaments. Measurements higher in the flow, above the open mat, are more comparable to near-bed measurements of the closed mat. Vertical distributions of velocity fluctuations and Reynolds stress demonstrate the high turbulent stresses in the mid-portion of the water column, peaking at the roughness layer height. Low near-bed stresses and high mid-flow stresses indicate that the flow resistance generated by open mats transferred peak shear stress higher in the water column. Other studies have suggested that turbulence reduction might occur when periphyton mats completely submerge bed roughness elements; i.e. when h >> D (Nikora et al. 2002a). However, none of the mats had heights greater than cluster heights, nor did the clusters protrude above the surrounding bed grains; thus we found no clear relation between h/D and turbulent effects. In fact, some of the mats with the lowest submergence, both open and closed (e.g. experiments B, L, and M), significantly reduced turbulence and near-bed shear stress relative to the non-periphyton experiment. Thus our results demonstrate that turbulence reduction can occur even when h << D. A reduction in turbulence observed within the open periphyton mats agrees with the lower turbulence intensities and vertical turbulent fluxes beneath the periphyton surface measured by Nikora et al. (2002a). Nikora et al. (2002a) explained the reduced turbulence as a consequence of weakened eddy penetration, because the periphyton ‘forelocks’ blocked vertical flow movement at the bed surface, suggesting that even our open mats experience an attenuation in flow penetration relative to the clean substrate.  4.5.3 APPLICATION OF PERIPHYTON-FLOW MODELS  First, we compare our measurements to our conceptual model of periphyton-flow dynamics; we then discuss application of our results to previously developed quantitative periphyton-flow models. Our model qualitatively represents the interactions between bulk flow rate, mat properties (density and height), and flow structure (velocity and turbulence distributions). Our results illustrate complexities, additional factors, and small-scale variations in the periphyton-flow relation that fall outside the model framework, although some general predictions hold true. We predicted that an increase in mat density would block and divert flow over the mat, constrict the flow depth, and increase velocities throughout the water column. Our results show that on average, highdensity, closed mats did in fact increase depth-integrated velocities relative to non-periphyton surfaces, although the extent of the effect varied with flow depth and flow rate. At high flows, compacted high-  112  density mats had higher near-bed velocities, but the same mats at low flows often showed a reduction in near-bed velocity, due to less compaction and more resistance generated by the filaments. Furthermore, even at high flows, several mid-density mats and two high density mats had comparatively slow near-bed velocities. Meanwhile, not all low-density mats had slower velocities. At shallow flows, open mats decreased velocities throughout the flow depth, but at deeper flows, open mats increased depth-integrated and maximum velocities so that velocity reduction was restricted to the near-bed region. Thus the relation between mat density and velocity depended on location in the water column, degree of compaction, and flow depth. Our model also predicted that near-bed turbulence intensity would decrease with the increase in mat density due to a reduction in vertical momentum exchange, but near-bed Reynolds stress showed no trend with density. At low flows, near-bed Reynolds stress was lower for all periphyton mats than the average for non-periphyton clusters, either due to weakened eddy penetration or hydrodynamic smoothing by periphyton filaments. At high flow, four periphyton mats had higher-than-average near-bed Reynolds stress, perhaps due to roughness generated by periphyton filaments. However, these high-Reynolds stress mats spanned the range of densities; thus mat density is not a clear predictor of near-bed turbulence intensity. Mat height also showed a mixed relation with velocity and turbulence. Our model predicted that flow velocities moving through a low-density mat would not increase with mat height, but velocities over a highdensity mat would increase with mat height as flow became increasingly constricted. Despite spanning a range of relative submergence values, near-bed velocities of our mats showed no trend with mat height; only the tallest mat had a notably faster near-bed velocity, suggesting a possible increase in flow compression. Given the variability in near-bed velocity among closed mats (described above), it is perhaps not surprising that there was no near-bed velocity-height relation for these high-density mats. Our model cannot precisely predict the degree of resistance generated at the mat surface, which determines near-bed velocities and the height of hydraulic effect. Our model also predicted that the depth of flow penetration into open mats would decline with filament height; thus turbulence intensity would be reduced at the base of tall mats. In fact, near-bed Reynolds stress showed no clear trend with height, but was lower for all periphyton mats than the non-periphyton average at low flow. At high flow, only four periphyton mats had near-bed Reynolds stresses higher than the nonperiphyton average; these mats spanned the range of densities, but fell within the 0.5-1.5 cm height. Thus periphyton presence appears to have reduced near-bed turbulence intensity in most cases, increasing Reynolds stress only for a few cases in the middle of the range of heights. Reynolds stress reduction was likely due to hydrodynamic smoothing over short, compacted mats and limited vertical exchange into tall mats; thus turbulence may have been increased for mats falling between these extremes. Finally, our model predicted that as flow rates increased, compaction of periphyton filaments would decrease mat height and reduce momentum exchange. As expected, all mats were taller at low flow than at high flow.  113  Our results illustrate the complex interactions and small-scale variations in the periphyton-flow relation, factors that are difficult to account for even in a qualitative model. Nevertheless, our measurements have some pattern and structure that can be placed into the context of other, more quantitative periphyton-flow models. For example, vertical distributions of velocity fluctuations and Reynolds stress (Figures 4.8 and 4.9) show locations of the near-bed roughness and outer logarithmic layers for clusters with and without periphyton. Similarly, Nikora et al. (2002a) also revealed a marked difference between flow properties in the near-bed region and those in the upper logarithmic layer for flows over a coarse bed both with and without periphyton. Nikora et al. (2002a) used τxz/ρ = -<u’w’>= (τ0/ ρ) (1-z/H) as representative of the vertical transfer of momentum, assuming two-dimensional flow (Monin and Yaglom, 1971; Nezu and Nakagawa, 1993), essentially equivalent to our Reynolds stress (- ρ<u’w’>). However, instead of using the near-bed value, as we have done, they plotted -<u’w’> = f(1-z/H) and estimated τ0/ ρ = u*2 from the slope of the regression line. Their approach relies on fitting a regression to the shear stress profile in the logarithmic layer; thus estimates achieved this way are in close agreement with values of u* obtained using a log-law fit to the velocity profile. A similar approach has been used to determine near-bed shear stresses by extrapolation of the Reynolds stress profile to the bed (i.e. τ0 when z = 0) (Martin, 2003), using the profile in the upper logarithmic layer. However, if a shift in slope occurs partway up the water column, due to the transition from inner roughness to outer logarithmic layer, the estimate of near-bed shear stress obtained this way is much higher than shear stress values measured close to the bed. Thus this approach is best applied to flows dominated by a logarithmic layer. In shallow, high roughness flows, the inner roughness layer may extend into most of the water column, limiting the extent of the logarithmic layer and invalidating the use of the log-law (Katul et al., 2002). In addition, periphyton patches and stone clusters like those in our study set up transient, dynamic, and three-dimensional flow fields that further complicate the use of these methods. Flow in the near-bed region arises from local conditions above the mat, whereas the outer layer forms from upstream roughness conditions thus the full profile represents a combination of average and local bed effects. For these reasons, our analysis focuses on examining the distribution of turbulent stresses and near-bed hydraulic measurements in order to assess the influence of periphyton on flow structure in the near-bed region. In natural streams, the presence of periphyton patches and associated flow fields could similarly invalidate estimates of near-bed turbulent stresses from the loglaw formulation or extrapolated upper flow profiles. Similarly, application of the periphyton-flow model of Nikora et al. (1998) to our periphyton patches is questionable. In their model, a characteristic number (P), representing the nature of the periphyton-flow interaction, is related to roughness length (z0), from which the effect of periphyton on roughness can be assessed in relation to the type of interaction. Low values of P indicated a weak interaction, where the periphyton filament is completely upright and z0 is directly related to periphyton height. At high values of P, the flow completely compresses the filament and roughness was directly related to the filament diameter. Between these two extremes, P changes with the angle of the filament relative to the bed, and thus with the balance of drag and buoyancy forces; roughness thus varies with both filament height and P. As bed shear  114  stress (and thus flow level) increases, the periphyton contribution to roughness decreases as filaments became more streamlined. However, determination of z0 requires application of the log-law to velocity measurements, which cannot be done for our case of the periphyton patch. Conceptually, the model predicted slightly different scenarios than we observed. Although this model supports our observation that highly compacted periphyton (high P) reduced roughness effects and turbulence intensities, it does not explain the reduction in near-bed turbulence at low flows and less compacted mats (low P). Both our open and closed mats exhibited this reduction at LF thus turbulence was reduced both at the surface of (closed) and within (open) the mats. Possibly, the localized shift in the height of peak turbulence may explain this discrepancy; we observed that the periphyton mat, like an elevated cluster, shifted the location of peak turbulent stress higher in the water column. Thus near-bed turbulent stresses were lower above the periphyton mat, but upper flow stresses may have been higher, as the model predicts. Application of this model to periphyton patches would require a modification that accounted for the two-layered, threedimensional flow-field characteristic of a heterogeneous streambed.  4.5.5 COMPARISONS WITH LARGE-SCALE ANALOGUES  Our findings agree relatively well with previous studies of periphyton-flow interactions, demonstrating the contrasting changes in roughness, turbulence intensity, and mean velocity that can occur depending on location relative to and structure of the periphyton mat. However, discrepancies between our results and results of others may be due to the differences between patch and full-bed dynamics. Similarly, comparisons with submerged macrophytes may also be limited by scale-dependent and structural differences between these two plant types. Here we outline the dominant ways in which they differ. Firstly, the relative submergence depth ratio of the plant (H/h) should be considered. Large-scale aquatic vegetation is typically tall relative to the depth of the flow, producing depth-limited flows and low depth ratios. In shallow coastal waters, for example, depth ratios of sea grass beds are typically 1-3 (Orth and Moore, 1988), encompassing a transition zone between submerged and emergent (H/h <1) canopies. Deeply submerged canopies are considered those with H/h >10 and behave like unconfined canopies, as has been described by models of terrestrial canopy air flow (see (Raupach et al., 1996), for review). In this study, the relative submergence of periphyton mats ranges from 3.3-26.1, spanning the range of low to deep submergence, but still deeper than typical sea grass beds. Secondly, as noted previously, large aquatic canopies are generally long and through-flow is uniform such that measured shear stresses are assumed to reflect only vertical turbulent fluxes. In contrast, flow around and within periphyton-cobble clusters varies strongly downstream so that shear stresses are associated with both longitudinal and vertical turbulent fluxes. Thirdly, comparing density measurements between periphyton and larger-scale vegetation is potentially complicated by structural differences. For example, the study by Sand-Jensen and Pederson (1999) reports volumetric plant densities of ~5 mg/cm3 for the shielded canopies and ~4 mg/cm3 for the  115  open canopy, which are approximately twice the average of volumetric densities (AFDM/h) measured for closed and open periphyton mats. Thus open macrophyte canopies are considerably denser than closed periphyton mats, most likely because of the macrophyte form and the vertical distribution of plant biomass. Whereas periphyton filaments are uniform in shape and thickness and are compressed easily by the flow, the open canopy is a so-called ‘meadow-former’ and has its biomass concentrated deep in the canopy. In contrast, grass leaves and periphyton are similar in shape and flexibility, but densities differ. Artificial grass canopies used in the experimental studies of Nepf and Vivoni (2000) and Ghisalberti and Nepf (2002, 2006) were modeled to mimic packing densities of natural seagrass meadows, which are typically reported as grass shoots per area in the range of 100-1200 shoots cm -1 (e.g. Gambi et al., 1990). Biomass estimates are difficult to find for hydraulic studies, but ecological studies report areal biomass densities up to ~50 mg/cm2 and grass heights of 25-100 cm (Keller and Harris, 1966), producing volumetric densities of 0.5-2 mg/cm3, at the low end of the periphyton range. Thus even high density grass beds are low in comparison to periphyton mats. Given the above-outlined differences, it is clear that although large-scale analogues may help direct future periphyton research, comparisons must be carefully scrutinized when making inferences about periphyton-flow dynamics. As our knowledge of periphyton dynamics increases, such a comparison should not be necessary.  4.6 CONCLUSIONS  Velocity and turbulence profiles over filamentous periphyton patches were used to evaluate a conceptual model of periphyton-flow dynamics in coarse-bedded, open-channel flows. Profiles measured over identical clusters with and without periphyton demonstrate how the presence of periphyton can significantly alter local hydraulic conditions, namely the velocities and Reynolds stresses within the inner boundary layer. However, the nature of effect depended on periphyton properties such as density, compaction and height. High density, highly-compacted, closed mats raised the effective bed height, becoming physically analogous to inorganic surface elements. In these cases, flow constriction accelerated velocities throughout the flow depth, but hydrodynamic smoothing decreased near-bed turbulence intensities. Low-density, open mats increased surface roughness, reducing near-bed velocities and transferring turbulent stress higher in the water column. For all profiles, regardless of flow level or periphyton presence, vertical distributions of turbulent stresses included two layers, an inner roughness and outer logarithmic layer. More information is needed on a range of periphyton structures and densities to assess and understand hydraulic differences between periphyton forms. Our results indicate that like large-scale macrophytes, periphyton assemblages may alter flow conditions of the near-bed region, which could influence sediment transport, streambed morphology, and habitat quality. For example, patchily distributed periphyton mats could create zones of stagnant flow and sediment accumulation overlain by high velocities in the water column, altering living conditions for both  116  benthic and free-swimming organisms. High density algal blooms, already known to harm water quality and oxygen levels, could cause additional problems by promoting streambed clogging and sedimentation. Researchers are beginning to recognize the strong feedbacks that exist between the physical and ecological components of the aquatic system (Cardinale et al., 2002; Thomson et al., 2004). This study advances our knowledge of the ways living organisms can influence their physical environment. The next chapter further investigates the effects of periphyton, extending from hydraulic effects to direct particle trapping, adhesion, and infiltration below the bed surface.  117  4.7 REFERENCES Aerospace, Mechanical & Mechatronic Engineering (2005), Laminar and Turbulent Boundary Layers, Aerodynamics for Students, University of Sydney, 7 February, 2008 <http://wwwmdp.eng.cam.ac.uk/library/enginfo/aerothermal_dvd_only/aero/fprops/introvisc/node8.html> Belanger T. B. and Roy A. G. (1998) Effects of a pebble cluster on the turbulent structure of a depthlimited flow in a gravel-bed river. Geomorphology 25, 249–267. Biron P. M., Lane S. N., Roy A. G., Bradbrook K. F., and Richards K. S. (1998) Sensitivity of bed shear stress estimated from vertical velocity profiles: The problem of sampling resolution. Earth Surface Processes and Landforms 23(2), 133-139. Biron P. M., Robson C., Lapointe M. F., and Gaskin S. J. (2004) Comparing different methods of bed shear stress estimates in simple and complex flow fields. Earth Surface Processes and Landforms 29(11), 1403-1415. Cardinale B. J., Palmer M. A., and Collins S. L. (2002) Species diversity enhances ecosystem functioning through interspecific facilitation. Nature 415(6870), 426-429. Dawson F. H. and Charlton F. G. (1988) Bibliography on the hydraulic resistance or roughness of vegetated watercourses. In Freshwater Biological Association, Occasional Publications Vol. No. 25. . River Laboratory. Denny M. W. (1973) Air and Water: the Biology and Physics of Life’s Media. Princeton University Press. Dodds W. K. and Biggs B. J. F. (2002) Water velocity attenuation by stream periphyton and macrophytes in relation to growth form and architecture. Journal of the North American Benthological Society 21(1), 2-15. Gambi M. C., Nowell A. R. M., and Jumars P. A. (1990) Flume observations on flow dynamics in Zostera marina (eelgrass) beds. Marine Ecology-Progress Series 61 159-169. Godillot R., Caussade B., Ameziane T., and Capblancq J. (2001) Interplay between turbulence and periphyton in rough open-channel flow. Journal of Hydraulic Research 39(3), 227-239. Graham A. A., McCaughan D. J., and McKee F. S. (1988) Measurement of surface-area of stones. Hydrobiologia 157(1), 85-87. Hart D. D. and Finelli C. M. (1999) Physical-biological coupling in streams: The pervasive effects of flow on benthic organisms. Annual Review of Ecology and Systematics 30, 363-395. Katul G., Wiberg P., Albertson J., Hornberger G. (2002) A mixing layer theory for flow resistance in shallow streams. Water Resources Research 38(11), 1250, doi:10.1029/2001WR000817. Keller M. and Harris S. W. (1966) The growth of eelgrass in relation to tidal depth. The Journal of Wildlife Management 30(2), 280-285. Kouwen N. and Unny T. E. (1973) Flexible roughness in open channels. Journal of the Hydraulics Division ASCE 99(5 ), 713–728. Labiod C., Godillot R., and Caussade B. (2007) The relationship between stream periphyton dynamics and near-bed turbulence in rough open-channel flow. Ecological Modelling 209(2-4), 78-96. Malmqvist B. and Sackmann G. (1996) Changing risk of predation for a filter-feeding insect along a current velocity gradient. Oecologia 108(3), 450-458. Martin V. (2003) Hydraulic roughness of armoured gravel beds: the role of grain protrusion, University of British Columbia. Martin V., Fisher T. S. R., Millar R. G., and Quick M. C. (2002) ADV data analysis for turbulent flows: low correlation problem. EWRI/ASCE/IAHR Joint International Conference on Hydraulic Measurements and Experimental Methods. Monin A. S. and Yaglom A. M. (1971) Statistical Fluid Mechanics: Mechanics of Turbulence. MIT Press. Morin A., Back C., Chalifour A., Boisvert J., and Peters R. H. (1988) Empirical models predicting ingestion rates of black fly larvae. Canadian Journal of Fisheries and Aquatic Sciences 45(10), 1711-1719. Nezu I. and Nakagawa H. (1993) Turbulence in Open Channel Flow. IAHR. Nikora V. I. and Goring D. G. (2002) Fluctuations of suspended sediment concentration and turbulent sediment fluxes in an open-channel flow. Journal of Hydraulic Engineering-Asce 128(2), 214224. Nikora V. I., Goring D. G., and Biggs B. J. F. (1997) On stream periphyton-turbulence interactions. New Zealand Journal of Marine and Freshwater Research 31(4), 435-448.  118  Nikora V. I., Goring D. G., and Biggs B. J. F. (1998) A simple model of stream periphyton-flow interactions. Oikos 81(3), 607-611. Nikora V. I., Goring D. G., and Biggs B. J. F. (2002a) Some observations of the effects of micro-organisms growing on the bed of an open channel on the turbulence properties. Journal of Fluid Mechanics 450, 317-341. Nikora V.I., Green M.O., Thrush S.F., Hume T.M., Goring D. (2002b) Structure of the internal boundary layer over a patch of pinnid bivalves (Atrina zelandica) in an estuary. Journal of Marine Research, 60, 121-150. Orth R. J. and Moore K. A. (1988) Distribution of Zostera-Marina L and Ruppia-Maritima L Sensu Lato along depth gradients in the Lower Chesapeake Bay, USA. Aquatic Botany 32(3), 291-305. Raupach M. R., Finnigan J. J., and Brunet Y. (1996) Coherent eddies and turbulence in vegetation canopies: The mixing-layer analogy. Boundary-Layer Meteorology 78(3-4), 351-382. Reiter M. A. (1989a) Development of benthic algal assemblages subjected to differing near-substrate hydrodynamic regimes. Canadian Journal of Fisheries and Aquatic Sciences 46(8), 1375-1382. Reiter M. A. (1989b) The effect of a developing algal assemblage on the hydrodynamics near substrates of different sizes. Archiv Fur Hydrobiologie 115(2), 221-244. Sand-Jensen K. and Pedersen O. (1999) Velocity gradients and turbulence around macrophyte stands in streams. Freshwater Biology 42, 315-328. Shimeta J. and Jumars P. A. (1991) Physical-Mechanisms and Rates of Particle Capture by SuspensionFeeders. Oceanography and Marine Biology 29, 191-257. Sontek. (1997) ADV technical documentation. Sontek. Strom K. and Papanicolaou A. (2007) ADV measurements around a cluster microform in a shallow mountain stream. Journal of Hydraulic Engineering-ASCE 133, 1379-1389. Thomson J. R., Clark B. D., Fingerut J. T., and Hart D. D. (2004) Local modification of benthic flow environments by suspension-feeding stream insects. Oecologia 140(3), 533-542. Vogel S. (1994) Life in Moving Fluids. Princeton University Press. Voulgaris G. and Trowbridge J. H. (1998) Evaluation of the acoustic Doppler velocimeter (ADV) for turbulence measurements. Journal of Atmospheric and Oceanic Technology 15(1), 272-289.  119  CHAPTER 5: ‘STICKY BUSINESS’: THE INFLUENCE OF STREAMBED PERIPHYTON ON PARTICLE 7 DEPOSITION AND INFILTRATION 5.1 INTRODUCTION  Fine particle deposition can greatly affect the chemical, physical, and biological characteristics of a streambed. For example, excess sediment deposition has been repeatedly shown to degrade benthic habitat for fish and other organisms (Hynes, 1970; Waters, 1995). Particle infiltration can clog interstitial spaces, reducing oxygen and nutrient exchange in the ecologically vital hyporheic zone (Brunke, 1999). Although the impacts of fine particle deposition have been well-documented, much less is known about the factors controlling particle deposition rates or the degree of particle infiltration. Several field studies have shown that measured rates of particle deposition often differ from still-water particle settling velocities calculated from particle size and density (e.g. Cushing et al., 1993), but explanations for this discrepancy remain largely speculative. For example, although adhesion of particles to surface periphyton has been proposed as a mechanism for enhancing particle deposition (Lock, 1981), few have investigated this phenomenon (Battin et al., 2003). In addition, although periphyton has been shown to significantly alter near-bed and interstitial flow velocities (Dodds and Biggs, 2002), the effect of these changes on particle deposition has not been explored. Periphyton is pervasive in rivers and streams throughout the world thus it may play an important, and heretofore underestimated, role in the deposition of fine particles to the streambed. Numerous studies have shown how large-scale plant forms such as grasses and macrophytes alter flow and sedimentation patterns in stream channels (e.g. Kouwen and Unny, 1973). In small streams at the local scale, periphyton may have similar effects. In this study, I investigate the interactions between periphyton composition (i.e. density and structure) and particle deposition using an experimental design that incorporates both ecological and physical variables of the stream system. Periphyton may influence particle deposition and infiltration indirectly via changes to flow hydraulics or directly by particle trapping and adhesion. Periphyton structure, density, and areal coverage are likely to determine the nature and magnitude of effect. For example, some forms of photosynthetically-active periphyton produce a ‘sticky’ exopolysaccharide (EPS) matrix that has been shown to enhance particle deposition (Battin et al., 2003). However, periphyton growth may also fill the spaces behind coarse substrates that typically capture and retain fine particles. Thus direct particle trapping by periphyton could be the sum of these two effects; high densities of sticky EPS assemblages are expected to capture more particles than non-EPS assemblages at equivalent densities, but high biomass and surface retention could reduce subsurface infiltration. In addition, recent studies have documented periphyton-induced changes to interstitial flow hydraulics (Nikora 2002, Dodds and Biggs, 2002) that may in turn influence particle deposition and infiltration. For 7  A version of this chapter has been submitted for publication. Salant, N.L., ‘Sticky business’: the influence of streambed periphyton on particle deposition and infiltration.  120  example, Nikora et al. (2002) measured a reduction in turbulence intensity and vertical turbulent fluxes into the periphyton ‘interfacial sublayer’ (between the crests of roughness elements), a consequence of weakened eddy penetration due to the presence of periphyton ‘forelocks’ at the bed surface. Similarly, Dodds and Biggs (2002) measured high rates of velocity attenuation with depth into densely packed, cohesive assemblages dominated by diatoms. Velocity and turbulence reduction within periphyton mats is analogous to the stagnant flow and high sedimentation generated by dense macrophyte canopies (SandJensen and Pedersen, 1999; Vermaat et al., 2000; Pluntke and Kozerski, 2003; Schulz et al., 2003). In contrast, vertical flow penetration and advective transport into open-weave filamentous algae is higher (Dodds and Biggs 2002), much like the turbulent conditions within low-density macrophytes (Sand-Jensen and Pedersen, 1999) and wave-dominated seagrass beds (Koch, 1993), which may increase particle transport into the subsurface. Above periphyton mats, velocities can accelerate due to flow constriction or the hydrodynamic smoothing of the surface (Godillot et al., 2001). Near-bed turbulence intensities and shear velocities can also increase due to roughness effects of the periphyton (Nikora et al., 2002; Labiod et al., 2007). Both effects are expected to promote particle suspension and transport, reducing deposition rates. In this study, I test how the presence of periphyton on a gravel-pebble substrate alters the rate and quantity of fine particle deposition and infiltration. My study measures both the change in water column concentration over time and the fine particle content of surface and subsurface samples for varying periphyton densities and structures. Specifically, I test two distinct forms of periphyton: a low-profile, EPSproducing assemblage dominated by diatoms and a high-profile, filamentous green algae that does not produce EPS. My specific hypotheses regarding the effects of these two forms on surface deposition, infiltration, and near-bed hydraulics are as follows: H1: High densities of low-profile forms will cover surface roughness elements, reducing near-bed shear stresses and increasing surface deposition. H2: Particles will adhere directly to EPS-producing assemblages, so that the amount of surface deposition will increase with periphyton density until the surface saturates with particles and the deposition-density relation plateaus. H3: High densities of EPS-producing assemblages will fill subsurface pore spaces, reducing particle infiltration. H4: For high profile, non-EPS-producing periphyton, surface particle deposition will not increase with periphyton density. H5: Growth of high-profile, filamentous assemblages will increase surface roughness and near-bed shear stresses, limiting surface deposition but increasing vertical flow penetration, subsurface particle transport, and thus particle infiltration  121  5.2 METHODS  Several densities of two types of surface periphyton were used in the experiments: a low-profile, mucilaginous form dominated by diatoms (Achnanthes minutissima), bacteria, and associated EPS excretions and a high-profile form dominated by filamentous green algae (Rhizoclonium riparium). I consider these two types as representative end-members of the range of complex architectures that exist in natural periphyton assemblages. Five reference experiments containing beds without periphyton were also conducted for comparison with the periphyton experiments. Reference condition was the same as described in Chapter 2, consisting of a 20% sand bed, a relatively high (~8.5 L/s) flow rate, and a moderate concentration of fine silica particles (< 125 µm). Periphyton experiments used the same substrate, flow rate, and particle dose as the reference experiments (Figure 5.1), with one small modification. Three metal trays filled with the 20% sand substrate were used for periphyton cultivation (described below) and installed in the test section of the flume thus the flume bed of the periphyton experiments was not identical to the reference bed without trays. Trays were constructed from aluminum gutters that were cut into 0.5 m sections and closed on either end with gutter end caps that were held into place with screws and sealed at the joints with silicon sealant. Trays were ~13 cm wide, 9 cm deep, and 0.5 m long; once filled with sediment, their surface was level with the remainder of the streambed. In order to test whether the presence of the trays alone (i.e. regardless of periphyton coverage) significantly altered the conditions of the bed or the flow, I installed trays without periphyton into two of the reference experiments (Reference A and Reference B) and compared results among all five references (with and without trays). Once I established that trays did not have a significant effect, I treated all reference experiments as representative of the reference condition for comparison with periphyton experiments.  122  Figure 5.1: Schematic illustrating all the possible combinations of experimental factors. Flow Rate  Par  ticl eD ens  it y  H  H  L  Flow Rate  L  H Initial M Concentration  H  L  L  a) 20% Sand Bed  b) 20% Sand Bed plus Diatoms at 4, 8, 12, and 24 Weeks Growth c) 20% Sand Bed plus Algae at 4, 8, 10, 16, and 20 Weeks Growth  *Notes: Factors include particle density (high or low), flow rate (high or low), initial concentration (high, medium, or low) and bed composition (20% sand with or without several periphyton growth stages). Shaded boxes are the combinations tested in a) Chapter 2 and b/c) this chapter; box with diagonal lines is the reference condition for both this chapter and Chapter 2. Tests were conducted in a small (0.15 m wide x 7.75 m long x 0.44 m deep) recirculating laboratory flume at the University of British Columbia, containing a pebble-gravel-sand substrate ~7-9 cm deep. See Chapter 2 for a more detailed flume description. Substrate was identical to the 20% sand bed described in Chapter 2. Experiments were conducted for a total of nine hours at a single flow rate (average discharge ~8.5 L/s), identical to the ‘HF’ flow rate described in Chapters 2 and 4. Water level, bed elevation, and longitudinal velocity measurements indicated that flow was uniform within the 2-5 m section of the flume; all measurements were restricted to the centreline at the 2.5, 3, and 3.5 m locations, in order to minimize sidewall effects and be within this uniform zone. Water surface and bed elevation measurements were repeated periodically throughout each experiment to ensure uniform flow and to determine the average flow depth. Because of the small width-to-depth ratio of the flume (aspect ratio < 5), wall-generated turbulence likely generated secondary currents that influenced velocity measurements taken at the centre of the flume. However, these secondary currents were most likely present during all experiments, without periphyton and for all periphyton structures and densities. I compared mean velocity and turbulence statistics between experiments, but for all I assumed that the degree of wall-generated turbulence was the same. I am concerned only with the relative difference in hydraulic parameters due to periphyton presence and form, with wall effects assumed to be the same for all scenarios.  123  5.2.1 PERIPHYTON CULTIVATION  Periphyton assemblages were cultivated in two different artificial stream systems for each assemblage type. Cultivation systems each consisted of a long narrow flume connected via rubber tubing to a large reservoir filled with continuously recirculating streamwater. Water was replenished approximately once per month to provide adequate nutrients. Photosynthetically active radiation (PAR) was provided 24h per day by a 1000-watt metal halide lamp hanging directly above the flumes. Twelve 0.5 m-long shallow aluminum trays were filled with the same substrate as the flume bed and placed in the cultivation flumes along with several inocula hosting the two different periphyton types collected from two streams near Vancouver, British Columbia: East Creek (diatoms) and Hope Slough (algae). Diatom and bacterial assemblages were allowed to grow for 4, 8, 12, and 24 weeks; three trays were removed after each growth interval, transported to the experiment flume, and fitted into the 2.25-3.75 m section of the flume immediately prior to the experiment. Green algal assemblages were tested after 4, 8, 13, and 16, and 20 weeks (logistical complications prevented testing at exactly the same intervals as the diatom experiments). Extra trays were used for taxonomic analysis: surface stones were scrubbed with a toothbrush and rinsed with deionised water; organisms within the rinse solution were identified under a microscope. Throughout the cultivation period, photographs were taken of the assemblages in order to document changes in growth and structural development. A camera held on a stable stand was positioned directly above the trays; photographs were taken at the same locations every two days.  5.2.2 MEASUREMENT SCHEDULE AND EXPERIMENTAL DESIGN  Experiments are identified by periphyton type and growth period: ‘D’ = diatoms and bacteria, ‘A’ = green algae and number = weeks of growth. The measurement protocol during each experiment was identical to the procedures described in Chapter 2, including ADV velocity profiles, particle dose, and LISST suspended particle concentrations, with the following exceptions: 1) only a silica dose at the reference concentration (0.31 kg) was used; 2) in addition to stratified samples, surface deposition samples were also collected at the end of each experiment (described in Section 5.2.3); and 3) the procedure for the stratified sampling was modified slightly to adjust for the trays and the presence of organic matter (described in Section 5.2.4). Each experiment began immediately after periphyton trays were installed (experiment hour -1; i.e. 1 hour before the particle dose). Three trays filled with substrate but without periphyton were tested in the same manner for two of the reference experiments (Reference A and Reference B). Data from all five reference experiments, including the two with trays and the tray-less three from Chapter 2 (20%SHF1, 2, and 3) were first compared and then combined into a single reference condition.  124  5.2.3 SURFACE SAMPLES  At the end of the experiment, the flow was stopped and the flume was allowed to drain. Three randomly selected pebble-sized stones from the surface layer of each tray were removed, scrubbed with a toothbrush and rinsed with deionised water. Rinse solutions from the scrubbed stones were homogenised, diluted to a known volume, subsampled, and filtered through a pre-weighed glass fibre filter. Filters were oven-dried overnight at 65ºC, allowed to cool in open air, weighed, ashed at 450ºC for four hours, and reweighed to determine dry-mass (DM), ash-mass (AM) and ash-free dry mass (AFDM). Surface area of the three stones was determined by the method of Graham et al. (1988) and used to calculate the areal density (g/m2) of each value. AFDM was used to determine the density of living and non-living periphyton and AM used to determine the amount of inorganic particle (silica) deposition.  5.2.4 STRATIFIED TRAY SAMPLES  After surface samples were removed, the remainder of each cultivation tray was divided into a stratified bed sample. Four different layers of the tray were removed by hand: ‘armour,’ ‘subsurface,’ ‘middle’, and ‘base.’ Armour was what remained of the surface layer (~0-2 cm from the surface), to which the scrubbed surface stones were later added; the subsurface layer was just below the armour (~2-4 cm); the middle layer extended 2 cm below the subsurface layer (~4-6 cm); and the base layer extended to the flume bottom (~6-7 cm). Each layer was dried and sieved for particle size analysis; particle size classes included a fine particle class <125 µm. Because each layer had slightly different masses, the proportion of layer mass in the fine particle class <125 µm was used for analysis, rather than the raw mass in order to correct for the fact that differences in sample mass may bias the comparison of fine particle mass (i.e. two samples from the same depth and condition will have different fine particle masses if one sample is simply larger than the other). Samples with periphyton growth were not ashed to remove organic matter because it was impractical to do with the large sample sizes. Instead, fine particle proportions of the top two surface layers (A and SS) were corrected by multiplying by the fraction assumed to be inorganic, determined from the inorganic fraction (by mass) of the surface samples (AM/DM).  5.3 ANALYSIS  Data analysis was performed in the same manner as for the experiments described in Chapter 2, including the calculation of velocity and turbulence parameters from regressions fit to the measured data and the derivation of particle deposition rates for the total suspension and four particle size classes from 01h and 1-8h. Velocity and Reynolds stress parameters computed for each location and experimental condition are provided in the Appendix, Table A.2. As in Chapter 2, a repeated measures model was used to compare hydraulic parameters, velocity profiles, deposition rates, and subsurface samples between  125  periphyton experiments and the reference condition; AFDM and AM of surface samples were also compared in this way. Due to time and resource limitations, periphyton experiments were not replicated (i.e. each combination of composition and growth stage was tested only once). As a result, the variance used in the model was computed from the replicated reference experiments. As noted above, reference experiments included two containing aluminum trays filled with a periphyton-free substrate as well as the three non-tray reference experiments from Chapter 2. Before these two groups of reference experiments were tested as replicates, however, a repeated measures model was fit to the reference experiments to determine if the presence of trays alone, without periphyton, significantly altered any of the measured conditions. Results of these comparisons are provided in subsequent sections. Given the exploratory nature of this study and the high variability inherent to the study design, for all statistical comparisons a p-value < 0.1 was considered significant.  5.4 RESULTS AND DISCUSSION  5.4.1 REFERENCE EXPERIMENTS  Data from the reference experiments with and without trays were compared to determine whether the presence of the trays alone, without periphyton coverage, influenced hydraulic and depositional metrics. Comparisons were made using the same repeated measures model used to compare reference and periphyton conditions, treating experimental run as a random variable in the model and measurements at different locations, heights, and times as repeated measures within each run. Non-tray experiments were replicated three times, while experiments with trays were replicated twice; variance was assumed equal for both conditions. None of the hydraulic parameters was significantly different between the two reference condition groups; velocity regressions also had statistically similar slopes and intercepts (Table 5.1; Figure 5.2). Thus there is no evidence for a significant or systematic effect of trays on the hydraulic characteristics of the bed.  126  Figure 5.2: Streamwise velocity (u) as a function of log-normalized height above the bed (ln(z)) for reference experiments with and without trays. -1.0 -0.5 0 0.5 1.0 1.5 2.0  u (cm/s)  60  With trays  Without trays  40 20  0 -1.0  -0.5  0  0.5  1.0  1.5  2.0 ln(z) (z in cm)  *Notes: Measurement points are from three streambed locations and three time periods within each experimental run. Experiments with trays were replicated twice; experiments without trays were replicated three times. Regression lines were fit with a repeated measures model. Regression parameters are provided in Table 5.1. Table 5.1: Regression coefficients for velocity as a function of ln(z) and hydraulic parameters from reference experiments with (‘Trays’) and without trays (‘No trays’). Trays No trays DF F p  N 2 3  Ux (cm/s) 31.3(7.0) 42.4(5.7) 3 1.50 0.31  u0 (cm/s) 19.1(7.1) 21.9(5.8) 3 0.10 0.77  umax (cm/s) 43.6(8.5) 62.9(6.9) 3 3.10 0.18  τRe0 (Pa) 0.62(0.7) 1.12(0.6) 3 0.30 0.62  τReM(Pa) 2.13(0.7) 2.39(0.6) 3 0.08 0.80  Slope 12.3(1.6) 9.98(1.3) 374 1.01 0.32  Intercept 19.9(3.8) 30.6(3.1) 3 4.76 0.12  *Notes: Values are parameter estimates (SE) of conditions determined from a repeated measures model. n’ is the number of experimental replicates. For each hydraulic parameter, the model is described by degrees of freedom (DF), F-statistic (F), and significance value (p) associated with experimental condition. u0, Ux, and umax are the near-bed, depth-integrated, and maximum streamwise velocity, respectively. Near-bed and maximum values of the Reynolds shear stress (τRe0 and τReM,) were determined from fluctuations of the three velocity components. See text for description of trays and experimental design. In all experiments, measured particle concentrations decreased continuously from 0 to 8 h. Over time, the loss of coarse particles from the water column changed the particle size distribution of the suspension, in turn slowing the rate of decay. As in Chapter 2, I separated my analysis into two time periods: rapid deposition from 0-1h and the slower decay from 1-8h in order to improve the model fit and provide a better estimate of deposition rate. Concentration data over time was fit with a repeated measures model for the two time periods and two reference condition groups (with and without trays). Model estimates, including decay rate (k) (regression slope) and initial concentration (C0), as well as deposition parameters (flow depth, wd, ws and Ed) for individual reference experiments (fit with a simple linear model) and for both reference conditions groups (fit with the repeated measures model) are provided in Table 5.2. Median  127  particle size (D50) used in the calculation of ws was the measured particle size of the suspension when the concentration equalled the model-determined C0, i.e. at the start of the analysis period. Table 5.2: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for reference experiments with (‘Trays’) and without trays (‘No trays).  0-1h Reference A Reference B 20%SHF1 20%SHF2 20%SHF3 Trays No trays No 20%SHF3 DF/F/p DF/F/p-No 20%SHF3 1-8h Reference A Reference B 20%SHF1 20%SHF2 20%SHF3 Trays No Trays No 20%SHF3 DF/F/p DF/F/p-No 20%SHF3  Trays Trays No trays No trays No trays  1061 835 Trays Trays No trays No trays No trays  33 26  D50 (µm)  k (10-4 s-1)  C0 (µl/l)  r2  H (cm)  wd (cm h-1)  ws (cm h-1)  Ed  13.8 16.5 14.0 13.0 15.0 15.2 14.0 13.5 3.08 0.04  2.50(0.06) 2.57(0.07) 2.55(0.06) 2.50(0.08) 2.22(0.05) 2.51(0.03) 2.41(0.03) 2.51(0.04) 0.08 0.84  91.38 108.20 148.71 152.93 176.62 131.84 158.82 150.72  0.89 0.88 0.89 0.84 0.92 ----  12.5 9.3 10.3 9.9 10.6 10.9 10.3 10.1  11.21 8.58 9.47 8.94 8.44 9.85 8.91 9.13  98.25 139.13 100.51 86.78 115.22 118.7 100.8 93.65  0.114 0.062 0.094 0.103 0.073 0.083 0.088 0.100  6.6 7.7 7.0 6.5 7.0 7.2 6.8 6.8 1.77 1.48  0.39(0.02) 0.28(0.05) 0.38(0.37) 0.39(0.36) 0.33(0.38) 0.33(0.02) 0.38(0.02) 0.38(0.03) 0.19 0.24  62.30 71.88 76.02 76.86 85.20 66.89 82.32 76.47  0.97 0.71 0.94 0.94 0.92 ----  12.5 9.3 10.3 9.9 10.6 10.9 10.3 10.1  1.74 0.94 1.41 1.39 1.25 1.30 1.41 1.38  22.33 30.24 25.34 21.86 25.34 26.3 24.2 23.6  0.078 0.031 0.056 0.063 0.049 0.049 0.058 0.059  *Notes: Values (SE) are model estimates. Decay rates (k) for individual experiments were determined from the slope of a linear least-squares regression of log-transformed concentration vs. time. Values for grouped experiments (with and without trays) are estimates from a repeated measures model. ‘No 20%SHF3’ is from the repeated measures test with the data from the 20%SHF3 experiment removed from the ‘No tray’ group. Also provided are the degrees of freedom (DF), F-statistic (F), and significance value (p) associated with the k value (concentration-time interaction). Initial concentration C0 was determined from the intercept of the model fit to the raw data from each time period and was used to normalize measured concentrations (C/C0). D50 is the median particle size measured by the LISST; r2 the coefficient of determination of the regression; H is flow depth. See text for descriptions and calculations of depositional velocity (wd), particle settling velocity (ws), and enhancement factor (Ed). See text for description of trays and experimental design. Results suggest that decay rate was not influenced by the presence of trays, although the test did detect a significantly different decay rate between reference condition groups in the 0-1h period (p < 0.1). Deposition was slower for the group without trays because of a low decay rate in the third replicate nontray experiment (20%SHF3; Table 5.2); the other two non-tray experiments have decay rates very similar to the experiments with trays. Decay rate was not significantly different between groups in the 1-8h period. When the 20%SHF3 experiment was removed from the comparison, decay rates were not significantly different between reference groups (p = 0.84; Table 5.2), demonstrating that the presence of the trays did  128  not have a systematically significant effect on particle deposition. Regression plots of concentration over time for the two groups also show the similarity in decay rate, even with the 20%SHF3 data included (Figure 5.3). Slower deposition in 20%SHF3 may have been due to other unintended factors, such as differences in bed configuration or dose amount. By including all reference experiments in the comparison with periphyton experiments, I could determine whether the effect of periphyton was significant despite these sources of variability.  Figure 5.3: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for reference experiments with and without trays. a) 0-1h  Without trays  With trays  0.5 0.0  ln(C/C 0)  -0.5 0 0.0  1000  2000  b) 1-8h  3000  0  1000  2000  Without trays  3000 With trays  -0.5 -1.0  5000  15000  20000  5000  15000  20000  Time (Seconds) In all experiments, both with and without periphyton, and for both time periods, measured depositional velocities were less than the particle settling velocity (wd << ws or Ed << 1). Thus particles deposited slower than expected under the force of gravity, due to the effects of entrainment and suspension factors. Trends in enhancement factor (Ed) roughly parallel trends in decay rate and depositional velocity, with some variation due to differences in particle size. Higher Ed values indicate faster rates of deposition relative to settling velocity (Table 5.2); however, none of experimental conditions ‘enhance’ particle deposition. Particle size used in the computation of ws strongly influenced the value of Ed thus experiments with very different starting particle sizes but similar decay rates had different values of wd and Ed. Given that large particles (>16 µm) deposited very rapidly from the water column, the median particle size of the water column could have changed in a relatively short period of time thus differences in my estimates of ‘initial’ median particle size between experiments may have been artificially large. Similarly, slight differences in depth  129  can also skew values of wd and Ed, potentially confounding the analysis. Given these concerns, I considered decay rates (k) to be the most accurate measure of deposition for reference and periphyton experiments. Like velocity parameters and deposition rates, surface samples from reference experiments were not affected by the presence of trays. Ash-mass and AFDM of surface samples were not significantly different between experiments with and without trays (p = 0.81 and 0.55, respectively). Ash-mass averaged 3.96±0.70 g/m2 for experiments with trays and 3.74±0.60 g/m2 for experiments without; ash-free dry mass was on average 0.34±0.06 g/m2 for experiments with trays and 0.29±0.05g/m2 for experiments without. Subsurface samples were slightly affected by the presence of trays, however. A plot of fine particle percentages from the subsurface layers of all five reference experiments (Figure 5.2) shows the correspondence between experiments in most of the bed layers; all reference experiments exhibit a steady increase in fine particle mass with depth. Only one measurement point deviates from the others; the base layer of Reference A has a visibly higher percentage of fine particles than the other experiments. As a result, the base layer of the experiments with trays had a significantly higher percentage of fine particles than the experiments without trays (p < 0.001). Pooling of water in the bottom of the trays in both Reference A and Reference B may have caused an increase in fine particle mass in the base layers relative to the non-tray reference experiments. Thus the presence of trays may slightly alter infiltration measurements, even if hydraulic parameters, decay rates, and surface samples are not affected. Analysis of subsurface samples from periphyton experiments must consider this potentially biasing factor.  130  Figure 5.4: Percentage of fine particles (<125 µm) in bulk bed layers from three sections of the flume beds of reference experiments.  0 -1 -2  0.00  Percent fine particles (<125 µm) 0.01 0.1  1.0  20%SHF1 20%SHF2 20%SHF3 Reference A Reference B  Depth (cm)  -3 -4 -5 -6 -7 -8 *Notes: Depths of each experiment have been offset vertically so that symbols can be distinguished; true depths are those aligned with the 20%SHF1 values. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer. Note that the x-axis is on a logarithmic scale to better illustrate differences between experiments. Despite the potential effect of the metal trays on particle infiltration, none of the other metrics used in this study was influenced by the presence of trays. For this analysis, I considered the five reference experiments with and without trays replicates. All reference experiments were assigned the same categorical variable in the repeated measures model for comparison with non-replicated periphyton experiments, assuming equal variance for reference and periphyton conditions. Although a lack of replication of periphyton conditions reduces the power of the statistical tests, replicated reference experiments reflect the range of bed configurations that occur without the influence of periphyton. By using a high number of reference experiments, I could be more certain that any significant differences I observed between periphyton and reference conditions were likely due to the presence of periphyton and not natural variation in flow or sediment structure.  5.4.2 PERIPHYTON GROWTH AND DEVELOPMENT  Biomass measurements were taken from the surface samples removed after each experiment, not before. Scour was observed during the experiments that removed portions of the periphyton coverage thus biomass measurements are underestimates of periphyton growth. Trays were installed under low flow conditions and flow was increased incrementally to avoid abrupt shifts in shear stress and minimize scour;  131  nevertheless, small patches of periphyton growth were removed from both diatom and algal assemblages. Generally, scour was most severe for high profile, less cohesive assemblages, such as the highly filamentous A13. Scour occurred following the increase from low to high flow, but ceased when flow reached equilibrium conditions. Thus the biomass measured at the end of the experiment was equal to the biomass present throughout the experiment. At four weeks, both diatoms and algae had a biomass density of ~3 g/m2, but at eight weeks, algal mat density was ~15 g/m2, while diatom mats were only ~5 g/m2. At eight weeks, the AFDM of algae was higher than all diatom trays – even those that grew for 24 weeks (Figure 5.5). Due to a relatively low profile form, diatoms were subjected to less scour than algae thus higher algal biomasses must be due to higher growth rates. Algal growth rate to 16 weeks was ~ 1.5 g/m2/week, twice that of the diatoms to 24 weeks (~0.81 g/m2/week). Difference in growth between algal and diatom assemblages cannot be attributed to differences in light or flow level, because these were identical for the two growing systems. Nutrient level is a possible factor because the two cultivation flumes were supplied with different water sources in order to provide the same nutrient levels they would experience naturally: algae were grown in water from Hope Slough; diatoms were grown in water from East Creek. Furthermore, diatoms require silica for growth, which is potentially more limiting than other nutrients. As expected, all diatom and all algae samples had significantly higher biomass than the reference samples (p < 0.01). For diatoms alone, AFDM density was similar at D4 and D8, but increased at D12 and again at D24 (Figure 5.5). Thus diatoms exhibited a gradual but steady increase in biomass over time. Among algae, biomass and growth stage did not directly correspond. Peak biomass occurred at A16 (Figure 5.5), while biomasses of A20 and A8 were both higher than A13, despite light supply problems and shorter growth periods, respectively, likely because of the visibly significant scour that occurred at the start of the A13 experiment. Given the complexity of the algae biomass-growth stage relationship, my analysis considered both AFDM density and growth stage as potential factors influencing hydraulics and particle deposition.  132  Figure 5.5: Periphyton biomass (AFDM) from surface samples at growth periods of 0-24 weeks for diatom and algal assemblages.  30 Diatoms Algae  AFDM (g/m2)  25 20 15 10 5 0 0  4  8  12 13 Growth stage (weeks)  16  20  24  *Notes: Surface samples from reference experiments without periphyton were averaged for the ‘0 week’ growth period (white column). Error bars are standard error determined from a repeated measures model. Despite a longer growing period, A20 had a lower biomass than A16 because of an incorrectly positioned light supply during the early stages of growth. I also observed distinct structural changes of the periphyton assemblages. Diatom assemblages in my study were dominated by a common species of diatoms, associated exopolysaccharide excretions, and some forms of bacteria. Visible only under a high-powered microscope, the diatoms were extremely small and densely packed. After four weeks growth (D4), diatom coverage of surface stones was sparse, characterized as a non-descript, light brown, low-profile film (Figure 5.6b). By D8, the brown film uniformly covered the bed surface, but no three-dimensional structures were visible. However, between D8 and D12, I witnessed the development of very distinct structures on the diatom assemblages that were absent from the earlier growth stages; numerous brown tufts, stout and cylindrical in shape and ~0.5 cm tall were patchily distributed over the surface stones. By 24 weeks (D24), these projections had disappeared; the assemblage instead appeared ‘rippled’ or ‘streaked’, covered with evenly-spaced, wavy ridges with lower-profile coverage in-between (Figure 5.6c). At four weeks (A4), algae had grown into small patches on the stone surfaces, but did not cover the entire bed (Figure 5.6d) thus pore spaces between particles were still visible at the surface. Later growth stages, however, were characterized by thick, webbed layers that covered most of the bed surface, between and over substrate particles (Figure 5.6e and f), blocking the spaces between particles at the surface. By thirteen weeks (A13), the algal mat was over a centimetre thick in some locations, with long (> 10 cm) flexible filaments that became fully streamlined and flattened against the rocks at high flow. Much of the A13 mat, however, was removed by scour at the start of the experiment. At A16 and A20, I also observed small, densely-packed nodules or ‘bumps’ spread out over the algal surface (Figure 5.6f). I attribute these  133  differences in structure to the developmental stages and growth forms of diatom- and algal-dominated assemblages.  134  Figure 5.6: Photographs of periphyton growth stages on a coarse particle substrate (a), showing the changes in structure that occur as periphyton develop and the differences between diatom- and algaldominated assemblages. Low-profile diatoms at b) four and c) 24 weeks of growth; filamentous green algae at d) 4, e) 12, and f) 16 weeks of growth.  a)  b)  c)  d)  Diatomaceous ripples  e)  f) Algal bumps  Algal webs  135  5.4.3 DIATOM EFFECTS ON HYDRAULICS AND PARTICLE DEPOSITION  Velocity and turbulence parameters  Velocity regression coefficients and hydraulic parameters for the reference condition and all periphyton experiments, as estimated by the repeated measures model, are summarized in Table 5.3. Nearbed and maximum Reynolds stresses were lower than the reference condition for all diatom experiments (Table 5.3). Although the effect of diatoms on Reynolds stress was not significant, it was consistent; as predicted, the presence of diatoms appears to have damped both near-bed and upper flow turbulence intensities (Hypothesis 1). Furthermore, some of the diatom velocity profiles were significantly different from the reference condition (Table 5.3). Early stage diatoms (D4 and D8) had regression slopes lower than the reference condition (significant for the D8 experiment; p <0.05) as well as low maximum Reynolds stresses, while late-stage diatoms (D12 and D24) had steeper regression slopes (significant for the D12 experiment; p <0.05) and higher maximum Reynolds stresses. Regression lines and velocity measurements from the diatom and reference conditions illustrate these differences (Figure 5.7). Profile slopes are the inverse of the regression lines; D12 and D24 have shallower velocity profiles than the reference condition throughout the profile, indicating a greater difference between near-bed and upper flow velocities and therefore stronger shear stresses. In contrast, velocities of D8 are similar to the reference condition in the upper region of flow but remain above ~20 cm/s near the bed (~0.04-0.05H); reference condition velocities at the same heights range between 2-10 cm/s. Because of these high near-bed velocities, the D8 profile has a steeper, more uniform profile. High near-bed velocities in this case may have been due to a reduction in flow resistance over the smooth eight-week diatom surfaces. Hydrodynamic smoothing by periphyton has been observed in a small number of previous studies, but seems to be fairly uncommon. Most studies of periphyton hydraulics report an increase in shear velocity and roughness length with periphyton growth, due to the resistance induced by periphyton structures (Reiter, 1989a; Reiter, 1989b; Nikora et al., 1997; Nikora et al., 2002; Labiod et al., 2007). However, others suggest that thick periphyton mats may create a more uniform and hydrodynamically smooth boundary layer, reducing turbulent shear (Godillot et al., 2001). Furthermore, as periphyton develops, hydraulic conditions can switch between intermittent periods of enhanced roughness and periods of smoothing (Reiter, 1989a; Reiter, 1989b; Nikora et al., 1997; Nikora et al., 2002; Labiod et al., 2007). A reduction in Reynolds stresses for all diatom experiments relative to the reference condition may have been due to coverage of roughness elements by the initially uniform diatom mat. However, the growth of structures like those visible in D24 could have caused the subsequent increase in near-bed and maximum Reynolds stress due to roughness induced by these forms.  136  Table 5.3: Regression coefficients for velocity as a function of ln(z) and hydraulic parameters from reference and periphyton conditions.  Reference D4 D8 D12 D24 A4 A8 A13 A16 A20 DF F P  n 5 1 1 1 1 1 1 1 1 1  Ux (cm/s) 37.99(4.3) 50.06(9.6) 54.60(9.6) 50.64(9.6) 60.53(9.6) 70.22(9.6) 55.69(9.6) 13.38(9.6)* 59.12(9.6) 75.89(9.6) 4 1.63 0.34  u0 (cm/s) 20.80(3.7) 17.76(8.9) 26.69(8.9) 12.69(8.9) 25.29(8.9) 41.39(8.9) 20.86(8.9) -1.55(8.9)* 20.36(8.9) 40.13(8.9) 4 1.99 0.26  umax (cm/s) 55.19(6.1) 47.97(13.5) 48.13(13.5) 54.20(13.5) 61.37(13.5) 64.65(13.5) 56.12(13.5) 28.29(13.5) 63.47(13.5) 77.26(13.5) 4 0.92 0.58  τRe0 (Pa) 0.92(0.4) 0.24(0.9) 0.57(0.9) 0.13(0.9) 0.52(0.9) 0.47(0.9) 0.42(0.9) 0.02(0.9) 1.82(0.9) 0.95(0.9) 4 1.27 0.44  τReM (Pa) 2.29(0.4) 0.66(0.9) 0.65(0.9) 1.91(0.9) 1.95(0.9) 0.88(0.9) 1.51(0.9) 0.03(0.9)* 3.02(0.9) 1.36(0.9) 4 0.37 0.90  Slope 11.13(0.9) 9.93(1.9) 6.74(1.9)** 13.97(1.9) 16.98(1.9)** 3.49(1.9)** 11.24(1.9) 4.80(1.9) 8.00(1.9) 5.35(1.9)** 1040 3.26 <0.01  Intercept 26.33(3.3) 23.44(7.3) 29.77(7.3) 20.37(7.3) 30.49(7.3) 46.67(7.3)* 28.30(7.3) 4.19(7.3)* 28.41(7.3) 44.67(7.3)* 4 2.65 0.18  *Notes: Values are parameter estimates (SE) of conditions determined from a repeated measures model. ‘n’ is the number of experimental replicates.For each hydraulic parameter, the model is described by degrees of freedom (DF), F-statistic (F), and significance value (p) associated with experimental condition. u0, Ux, and umax are the near-bed, depth-integrated, and maximum streamwise velocity, respectively. Near-bed and maximum values of the Reynolds shear stress (τRe0 and τReM,) were determined from fluctuations of the three velocity components. See text for equations. Significance codes next to each parameter indicate the pvalue of the comparison with the reference condition: ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1.  137  Figure 5.7: Velocity as a function of ln(z) for periphyton and reference condition experiments. -1 60  D4  D8  D12  D24  A4  A8  A13  A16  Reference  A20  0  1  2  40 20 0 60 40 20 0 u (cm/s)  60 40 20 0 60 40 20 0 60 40 20 0  -1  0  1  2 ln(z) (z in cm)  *Notes: ‘D’ (‘Diatom’) refers to a periphyton assemblage comprised of diatoms, bacteria, and associated excretions; ‘A’ (‘Algae’) refers to a periphyton assemblage dominated by filamentous green algae; experiment codes include the number of weeks of periphyton growth. Particle deposition rates  Diatom and algae experiments showed the same roughly exponential decline in concentration observed in the reference experiments, although the rate of decline varied. Particle decay rates for the 0-1h and 1-8h period for all experiments are provided in Table 5.4, along with the other model and deposition parameters. Measured concentrations and regression lines for the periphyton and reference experiments are plotted in Figure 5.8. .  138  Table 5.4: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for periphyton and reference experiments.  0-1h Reference D4 D8 D12 D24 A4 A8 A13 A16 A20 DF F p 1-8h Reference D4 D8 D12 D24 A4 A8 A13 A16 A20 DF F P  D50 (µm)  k (10-4 s-1)  C0 (µl/l)  H (cm)  wd(cm h-1)  ws(cm h-1)  Ed  14.5 14.1 16.4 16.4 12.9 13.6 13.4 14.5 12.6 16.2  2.45(0.03) 2.53(0.07) 3.00(0.07) *** 2.63(0.07) ** 2.66(0.07) *** 2.57(0.07) 2.22(0.07)*** 1.84(0.07)**** 2.09(0.07)**** 2.64(0.07)** 2955 21.32 <0.0001  147.4(1.1) 132.3(1.2) 129.7(1.2) 131.5(1.2) 151.1(1.2) 159.7(1.2) 132.3(1.2) 84.0(1.2)** 133.0(1.2) 159.6(1.2) 4 1.62 0.34  10.5 11.7 13.0 11.4 9.7 9.7 10.9 9.6 9.6 10.2  9.26 10.66 14.04 10.79 9.29 8.97 8.71 6.36 7.22 9.69  107.28 101.57 137.65 137.21 86.04 94.9 92.58 107.75 81.24 134.23  0.086 0.105 0.102 0.079 0.108 0.095 0.094 0.059 0.089 0.072  7.0 5.4 6.2 8.0 7.0 7.5 6.0 13.0 6.9 7.8  0.36(0.02) 0.41(0.05) 0.63(0.05)*** 0.43(0.05) 0.38(0.05) 0.52(0.05)*** 0.16(0.05)**** 0.42(0.05) 0.36(0.05) 0.41(0.05) 86 5.72 <0.0001  75.76(1.1) 68.72(1.2) 51.54(1.2) 64.69(1.2) 73.63(1.2) 84.10(1.2) 70.96(1.2) 39.36(1.2)** 69.49(1.2) 76.23(1.2) 4 1.76 0.31  10.5 11.7 13.0 11.4 9.7 9.7 10.9 9.6 9.6 10.2  1.36 1.73 2.95 1.76 1.33 1.82 0.63 1.45 1.24 1.51  24.93 15.32 20.02 33.31 25.26 29.07 18.63 86.78 24.26 31.35  0.055 0.113 0.147 0.053 0.053 0.062 0.034 0.017 0.051 0.048  *Notes: Decay rate (k) and initial concentration (C0 ) are slope and intercept estimates (SE) from a repeated measures model. D50 is the median particle size measured by the LISST at the start of the measurement period. Significance codes next to k and C0 indicate the p-value of the comparison with the reference condition: ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. Also provided are the degrees of freedom (DF), F-statistic (F), and significance value (p) associated with the k and C0 estimates. H is flow depth. See text for descriptions and calculations of depositional velocity (wd), particle settling velocity (ws), and enhancement factor (Ed). Each experiment corresponds to a different growth period (number in weeks).  139  Figure 5.8: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for diatom, algae and reference experiments.  a) 0-1h 0.5  0  2000  0  Reference  D4  0  2000  2000  D8  D12  D24  A13  A16  A20  0  ln(C/C0)  -0.5 -1.0 0.5  A4  A8  0 -0.5 -1.0 0  2000  5000 15000 25000  2000  0 5000 15000 25000  5000 15000 25000  b) 1-8h 0  Reference  D4  D8  D12  A13  A16  D24  -0.5  ln(C/C 0)  -1.0 -1.5 0  A4  A8  A20  -0.5 -1.0 -1.5 5000 15000 25000  5000 15000 25000 Time (Seconds)  Trends in decay rate among diatom experiments and the reference condition are identical for the 0-1h and 1-8h time periods. All diatom experiments have a higher rate of decay than the reference condition; D8, D12, and D24 decay significantly faster from 0-1h, but only D8 decays significantly faster in the later time period as well. Given the difference in Reynolds stresses and velocity profiles between these three experiments and the reference condition identified above, there appears to be some correspondence between hydraulic conditions and decay rate of diatom experiments. Decay rates from diatom experiments  140  and the reference condition for the two time periods are plotted with near-bed and maximum Reynolds stresses in Figure 5.9. Also shown are the corresponding diatom biomasses measured as the AFDM of surface samples. Results from algae experiments are included for comparison but will be discussed in later sections. As seen in Figure 5.9, diatom growth resulted in steady biomass accumulation, but decay rate peaked at the low profile, moderate biomass D8 and decreased with the subsequent increase in biomass and structural development. At eight weeks of diatom growth (D8), both near-bed and maximum Reynolds stresses were low relative to the reference conditions and to the other growth stages, indicating that the biomass and structure of this growth stage resulted in hydrodynamic smoothing of the bed. As noted previously, D8 also had a more uniform velocity profile relative to the reference condition and the later diatom growth stages due to high velocities in the near-bed region. D8 also had relatively low near-bed and maximum Reynolds stresses. In contrast, velocity profiles from D12 and D24 had much shallower gradients than the reference condition, with faster velocities in the upper region of flow and higher Reynolds stresses due to the structural roughening that occurred as the diatom assemblages grew surface nodules and ridges. Decay rate and maximum Reynolds stress from the high-density and full-coverage D8, D12, and D24 experiments are negatively correlated (r2 = 0.99), but the low-biomass D4 had both a low decay rate and low Reynolds stress. Given the negative relation between Reynolds stress and decay rate at diatom biomasses > 5 g/m2, it seems probable that the low Reynolds stresses of D8 promoted the deposition of particles to the surface by decreasing upward entrainment forces and allowing particles to settle (Hypothesis 1). Thus particle deposition, as measured from the decrease in suspended concentrations, was likely related to the hydraulic conditions associated with diatom surface structures.  141  Figure 5.9: Decay rates (k) for the 0-1h and 1-8h time periods, near-bed (τRe0) and maximum (τReM) Reynolds stress, and AFDM densities for a) diatom experiments and b) algae experiments compared to the reference condition. 4  a) Diatoms  0-1h 1-8h τ τ Re0 ReM  4  b) Algae  3  2  1  2  0  Reynolds stress (Pa)  Decay rate (k; x 10 -4 s -1  3  1  AFDM (g/m2)  -1  -2  0 30 20 10 Reference  D4  D8  D12  D24  A4  A8  A13  A16  A20  *Notes: Decay rates and Reynolds stress values and standard errors (error bars) are estimates from a repeated measures model fit to the decline in log-transformed concentration versus time. ‘D’ (‘Diatom’) refers to a periphyton assemblage comprised of diatoms, bacteria, and associated excretions; ‘A’ (‘Algae’) refers to a periphyton assemblage dominated by filamentous green algae; experiment codes include the number of weeks of periphyton growth. Decay rates varied among particle size classes within the total suspension due to the effects of particle size on settling velocity; not surprisingly, decay rates increased with particle size. However, trends with periphyton biomass and growth stage differed slightly for the different particle size classes: some classes followed the trends in the total concentration; others showed a different pattern (Appendix, Table A.3), indicating that particle size influences which factors control deposition. Plots of concentration over time for four different particle size classes (~2-4, 4-16, 16-63, and 63-122 µm) from periphyton and reference experiments (Figure 5.10; Chapter 2, Figure 2.8) illustrate the differences in decay rates. Large particles (63-125 µm) deposited rapidly, dropping completely out of the water column in less than 30 minutes. Plots also show how the decline in concentration of the total suspension was similar to the 16-63 µm class in the early part of the experiment, but reflected the decline of the finer fractions (2-4 and 4-16 µm) later in the experiment due to the shift in proportions of the different size classes. Among diatoms, the rapid decay of the D8 suspension is clearly visible in comparison to the other diatom experiments and the reference experiments (Figure 5.8; Figure 5.10). However, not all particle sizes deposited fastest in D8 throughout the experiment. Total deposition is a sum of the deposition of individual  142  size classes within the suspension. In a given experiment, the relatively faster deposition of one particle size can be counter-balanced by the slower decay of another, such that the overall deposition rate does not significantly change. Decay rates, model parameters, and statistical comparisons between diatom and reference experiments are provided in the Appendix, Table A.3; decay rates for each size class and time period are plotted in Figure 5.11. From 0-1h, very fine particles (<16 µm) reflected the trend of the total suspension, with high decay rates at D8 (Figure 5.11a). Very fine particles also deposited rapidly at D24. In contrast, larger particles (>16 µm) deposited fastest at D4, but because of the slow deposition of the fine fractions total deposition was only slightly faster than the reference. From 1-8h, all particle sizes (<63 µm) reflected the trend in the total suspension; decay rate clearly peaked at D8 (Figure 5.11b). Differences between the trends of the individual particle size classes and the total suspension in the early part of the experiment indicate that the effects of depositional factors can vary with particle size. Furthermore, enhancement factors during both time periods are largest for the smallest particle size class and generally decrease with particle size (Table A.3, Appendix), although all are less than one (i.e. deposition is not enhanced for any size class). Most likely, differences in the trends of different particle sizes were due to the fact that the mechanisms influencing particle deposition differ among particle sizes and experiments. Large particle deposition should be promoted by high bed porosity and low Reynolds stresses, which would allow these particles to settle and percolate into the bed. In contrast, very fine particles should remain in suspension unless shear stresses drop very low or particles are trapped and retained by bed grains or adhesive surfaces; even small pore spaces will allow these particles to infiltrate the bed. Thus the comparatively low maximum Reynolds stress combined with patchy diatom coverage of the D4 experiment allowed large particles to settle and percolate rapidly. At D8, Reynolds stresses dropped lower, promoting the deposition of both large and small particles; meanwhile, diatom coverage increased, reducing the degree of large particle infiltration, but increasing fine particle adhesion. For most of the experiment (1-8h), D24 had the lowest rate of fine particle decay and the lowest total decay among diatom experiments, possibly due to the high biomass but striated coverage of this diatom assemblage; high biomass would have clogged interstitial pore spaces, obstructing infiltration, but striated coverage would have reduced diatom surface area and surface adhesion. Evidence from surface and subsurface samples allowed me to further explore these hypotheses and variable effects.  143  Figure 5.10: Decline in the near-bed suspended concentration (C) of the total suspension and four size classes within the suspension for periphyton and reference experiments from 0-8h.  *Notes: The shaded region is the 0-1h period. Concentration is normalized by the initial concentration (C0) of the exponential model fit to the measured concentrations; because the model C0 was less than the actual initial concentration, the plotted values of the total start at a number greater than one. Irregular spikes in the concentrations are due to instrument error.  144  Figure 5.11: Decay rates (k) of four particle size classes (~2-4, 4-16, 16-63, and 63-122 µm) and the total suspension (black bars) from a) 0-1h and b) 1-8h for diatom experiments and the reference condition. a) 0-1h  -4 -1  Decay rate, k (x 1 0 s )  4  2-4 µm  0.8  4-16 µm  4  2-4 µm  3  3  0.6  3  2  2  0.4  2  1  1  0.2  1  0 15  0 80  0.0  0  Decay rate, k (x 1 0 s )  -4 -1  b)1-8h 4  16-63 µm  25  63-122 µm  16-63 µm  4-16 µm  Ref D4 D8 D12 D24  20  60  10 15 40 10 5 20  0  5  0 Ref  D4  D8  D12 D24  0 Ref  D4  D8  D12 D24  Ref  D4  D8 D12 D24  *Notes: Values and standard errors (error bars) are estimates from a repeated measures model. Note the difference in scale among particle size classes. No data plotted indicates there were not enough measurements to fit the repeated measures model. ‘Ref’ values are estimates from the data of five replicate reference experiments; ‘D’ (‘Diatom’) refers to a periphyton assemblage comprised of diatoms, bacteria, and associated excretions; experiment codes include the number of weeks of periphyton growth. Surface and subsurface samples  Surface samples do not mimic the trend in particle decay rates of the total suspension. Ash-masses (AM) from diatom surface samples plotted against diatom biomass (AFDM) and growth stage show a general increase up to 12 weeks of growth or ~9 g/m2 AFDM (Figure 5.11a and b), subsequently decreasing at 24 weeks. Thus D8, at eight weeks of growth, did not have the highest surface deposition, despite having the highest particle decay rate throughout the experiment. Meanwhile, D12 had the highest surface deposition, but decay rates throughout the experiment slower than or similar to the other diatom experiments. Surface AM is lowest for the reference condition, which also has the slowest decay rate; in fact, all diatom experiments had significantly higher AM amounts than the reference condition (p < 0.001). As predicted (Hypothesis 2), adhesive properties of the exopolysaccharides (EPS)-producing diatom assemblages appear to increase particle retention at the surface. Battin et al. (2003) found that the accumulation of microbial EPS was a strong predictor of particle deposition rate. Complex periphyton surfaces enhanced particle retention by over 100%. Microbial adhesion of particles occurs on a wide range of substrate types; from marine mudflats (Westall and Rincé, 1994) to tooth enamel (Busscher and Van Der Mei, 1997). Adhesive interactions are driven by physico-  145  chemical forces, including Van der Waals forces, hydrogen bonding, and Brownian motion. Microbial assemblages on stream substrates are likely to interact with particles in similar ways; particles may adhere to mucilaginous excretions because of chemical bonds and electrostatic attractions.  Figure 5.12: Ash-mass of diatom surface samples plotted against diatom AFDM (a) and growth stage (b). 35 a)  b)  30  AM (g/m2)  25 20 15 10 5 0 0  5  10 AFDM (g/m2)  15  0  10 20 Growth period (weeks)  30  *Notes: Error bars are standard errors determined from a repeated measures model. Additionally, some of the AM of the diatom assemblages may have been due to the silica contained in diatoms; silica content typically ranges between ~30-70% diatom dry-weight (Reynolds, 1984). However, my diatom assemblages also hosted bacteria and mucilaginous exudates – the relative proportion of which likely varied with growth stage (Battin et al, 2003) – which would not have contributed to the AM. Furthermore, AM did not increase directly with AFDM nor vary much among experiments despite larger changes in AFDM (Figure 5.5). Thus very little of the AM could have been due to silica in the periphyton biomass. A complex AM-AFDM relation indicates that diatom biomass and surface deposition are not directly related by adhesive effects; surface deposition also depends on the effects of structure and hydraulics discussed above. In order to understand why the moderate-biomass, low-profile D8 has the fastest decay rate without the highest amount of surface deposition, one must consider that particles are lost from the water column to both the surface and the subsurface of the bed. Thus total deposition measured as the decline in suspended concentration is the sum of surface and subsurface deposition. Stratified subsurface samples from diatom experiments and the combined reference condition demonstrate a distinct trend with diatom biomass and growth stage (Figure 5.12) that explains the discrepancy between surface samples and decay rates. Reference values in the lower two layers (middle and base) were lower than all diatom experiments except the highest biomass experiment, D24. Base layers of D4 and D8 were significantly higher than the base of the reference condition (p < 0.1) but base layers of D12 and D24 were not significantly different from the reference, indicating that the increase in diatom biomass reduced the depth and amount of particle  146  infiltration. Similarly, fine particle mass of the armour layer was significantly higher for the low-biomass D4 than for the reference condition (p < 0.1). An identical trend is observed in all layers; fine particle mass was highest in all bed layers for D4 and consistently decreased as diatom biomass and growth stage increased (D4>D8>D12>D24). However, as noted previously, pooling of water in the bottom of the trays may positively bias the base layer values of the periphyton experiments.  Figure 5.13: Percentage of fine particles (<125 µm) in bulk bed layers of three sections of the flume bed from diatom experiments compared to the reference condition.  0 -1  Percent fine particles (<125 µm) 0.01 0.1  0.001  1.0  Reference D4 D8  -2  Depth (cm)  -3  D12 D24  -4 -5 -6 -7 -8  *Notes: Depths representing the middle of each layer have been offset vertically so that error bars can be distinguished; true depths are those aligned with the reference condition. Dotted lines separate layers. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer; note that values are plotted on a logarithmic scale. Error bars are standard error estimates from a repeated measures model ‘D’ (‘Diatom’) refers to a periphyton assemblage comprised of diatoms, bacteria, and associated excretions; experiment codes include the weeks of periphyton growth. Overall, surface and subsurface samples indicate that, as predicted (Hypothesis 3), the increase in diatom biomass limited fine particle infiltration, but low diatom densities increased surface deposition and infiltration relative to periphyton-free beds. Near-bed Reynolds stress reduction and particle adhesion to mucilaginous surfaces likely retained more particles in the bed surface ‘dead-zone’ where entrainment was limited and gravity-driven percolation carried them into the bed. Without diatoms, it appears that particles were too quickly swept back into the water column for significant amounts of percolation to occur. As diatom biomass increased, particles were probably physically blocked from entering interstitial spaces thus infiltration decreased. Furthermore, densely-packed diatoms have been shown to reduce velocities and vertical turbulent fluxes in the bed (Dodds and Biggs, 2002; Nikora et al., 2002) which may have also limited particle infiltration.  147  Sum of effects  Combined information about diatom structure and biomass, hydraulics, surface deposition, and subsurface infiltration provides a more detailed picture of the mechanisms influencing particle deposition to diatom assemblages. High variability within and among experiments due to spatially-varying substrates and periphyton structures, as well as the lack of replication of periphyton conditions, may have reduced the significance of some diatom effects; as a result, even non-significant trends are important to consider. At low diatom biomasses (D4), surface adhesion was relatively low, but particle infiltration was high due to abundant pore space; as predicted, reduced Reynolds stresses relative to reference conditions likely also facilitated the settling of particles onto and into the substrate (Hypothesis 1). As expected, an increase in diatom biomass (D8 and D12) initially increased surface adhesion and deposition (Hypothesis 2), but infiltration was limited by the clogging of pore spaces by diatom assemblages and exudates (Hypothesis 3). At later growth stages (D24), however, surface adhesion declined due to increased structural development and Reynolds stresses that reduced particle settling; subsurface deposition continued to decline with the increase in biomass, a finding that is counter to my predictions. Thus the fastest deposition occurred at eight weeks of growth (D8) when diatom biomass was sufficient to facilitate surface adhesion, but still sufficiently moderate to allow for subsurface infiltration; furthermore, low-profile structures and low Reynolds stresses promoted particle settling. Despite having the highest surface deposition, D12 also had less subsurface deposition than D8, hence slower deposition rates. Both surface and subsurface deposition were reduced at D24 relative to the other diatom experiments because of high Reynolds stresses and high clogging effects, but both decay rate and surface deposition were still higher than the reference condition because of the adhesive properties of the D24 surface.  5.4.4 ALGAE EFFECTS ON HYDRAULICS AND PARTICLE DEPOSITION  Velocity and turbulence parameters  For hydraulic parameters, only A13 was significantly different from the reference condition, with a significantly lower depth-integrated velocity, near-bed velocity, and maximum Reynolds stress (p < 0.1) (Table 5.5). None of the other algae experiments had significantly different hydraulic parameters relative to the reference condition. However, velocity regressions were significantly different from the reference condition for A4, A13 and A20. All three had lower regression slopes, indicating a more uniform velocity profile, but only A13 also had a lower intercept, indicating that the profile was also shifted to slower velocities throughout the water column (Figure 5.14).  148  Table 5.5: Velocity regression coefficients (SE) and mean hydraulic parameters from reference and algae experiments.  Reference A4 A8 A13 A16 A20 DF F P  n 5 1 1 1 1 1  Ux (cm/s) 37.99(4.3) 70.22(9.6) 55.69(9.6) 13.38(9.6)* 59.12(9.6) 75.89(9.6) 4 1.63 0.34  u0 (cm/s) 20.80(3.7) 41.39(8.9) 20.86(8.9) -1.55(8.9)* 20.36(8.9)) 40.13(8.9)) 4 1.99 0.26  umax (cm/s) 55.19(6.1) 64.65(13.5) 56.12(13.5) 28.29(13.5) 63.47(13.5) 77.26(13.5) 4 0.92 0.58  τRe0 (Pa) 0.92(0.4) 0.47(0.9) 0.42(0.9) 0.02(0.9) 1.82(0.9) 0.95(0.9) 4 1.27 0.44  τReM (Pa) 2.29(0.4) 0.88(0.9) 1.51(0.9) 0.03(0.9)* 3.02(0.9) 1.36(0.9) 4 0.37 0.90  Slope 11.13(0.9) 3.49(1.9)** 11.24(1.9) 4.80(1.9)** 8.00(1.9) 5.35(1.9)** 1040 3.26 <0.01  Intercept 26.33(3.3) 46.67(7.3)* 28.30(7.3) 4.19(7.3)* 28.41(7.3) 44.67(7.3)* 4 2.65 0.18  *Notes: Values are parameter estimates (SE) of conditions determined from a repeated measures model. ‘n’ is the number of experimental replicates. For each hydraulic parameter, the model is described by degrees of freedom (DF), F-statistic (F), and significance value (p) associated with experimental condition. u0, Ux, and umax are the near-bed, depth-integrated, and maximum streamwise velocity, respectively. Near-bed and maximum values of the Reynolds shear stress (τRe0 and τReM,) were determined from fluctuations of the three velocity components. See text for equations. Significance codes next to each parameter indicate the pvalue of the comparison with the reference condition: ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. Figure 5.14: Regression lines of velocity profiles from algae and reference condition experiments. 60  A4  A8  A13  A16  Reference  A20  40 20  u (cm/s)  0 60 40 20 0 60 40 20 0 -1  0  1  2 ln(z) (z in cm)  Like diatoms, all algae experiments except A16 had lower near-bed and maximum Reynolds stresses than the reference condition, but none had significantly lower near-bed Reynolds stresses and only A4 and A13 had significantly lower maximum Reynolds stresses. Thus counter to my predictions, the presence of algae in most cases damps turbulence intensities relative to non-periphyton substrates; the effect is most pronounced for the low-to-moderate biomass, uniformly structured assemblages from four to 13 weeks of growth. However, as predicted, the development of high-profile, visible structures at A16 and A20 (i.e.  149  small, densely-packed nodules and bumps) seems to have increased bed roughness and Reynolds stresses to values greater than or similar to those of the non-periphyton bed (Hypothesis 5). A decrease in slope coupled with an increase in intercept, as occurred in A4, A16 and A20, indicates a simple rotation of the velocity profile, due to higher near-bed velocities and lower upper flow velocities. In contrast, decreases in both the slope and intercept indicate a displacement of the entire profile to lower velocities throughout the water column (e.g. A13). High velocities throughout the flow depth in A16 and A20 are likely due to the high biomass of these mats. As noted in Chapter 4, high-density algal mats raise the effective bed height, constricting the flow and increasing velocities throughout the flow depth. Velocities over A4 may be high because of the lack of rough surface structures and relatively uniform bed that also resulted in low Reynolds stresses. In contrast, the profile of A13 remained below 20 cm/s despite a moderate biomass and shallow flow depth, as seen in the limited height of the profile (Figure 5.13). Possibly, the assemblages of A13 may have been structurally similar to the ‘open’ mats discussed in Chapter 4; given the tall height of the mat, the ADV may have penetrated its upper layers, where drag induced by the filaments impeded and slowed the flow. Particle deposition rates  Decay rates of the total suspension varied among algae and reference experiments (Table 5.6; Figure 5.15). Like diatoms, variations in particle decay corresponded with changes in hydraulic conditions. Unlike diatoms, however, trends in total decay differed slightly from the early to the late parts of the experiment Table 5.6; Figure 5.16). Although decay rate did not show a clear correspondence with biomass, growth stage, or Reynolds stress, three observations are suggestive: 1) significantly faster deposition in A20 early in the experiment corresponded with moderate biomass, high maximum Reynolds stress, and a moderate near-bed Reynolds stress; 2) significantly slower deposition throughout the A8 experiments corresponded with low maximum and near-bed Reynolds stress and a moderate biomass; and 3) deposition was faster than the reference experiment in the low-biomass A4. To determine whether these relations can be mechanistically explained, I examined the trends in decay for the individual particle size classes (Figure 5.17).  150  Table 5.6: Exponential model and deposition parameters for the decrease in concentration over time from 0-1h and 1-8h for algae and reference experiments.  0-1h Reference A4 A8 A13 A16 A20 DF F p 1-8h Reference A4 A8 A13 A16 A20 DF F P  D50 (µm)  k (10-4 s-1)  C0 (µl/l)  H (cm)  wd(cm h-1)  ws(cm h-1)  Ed  14.5 13.6 13.4 14.5 12.6 16.2  2.45(0.03) 2.57(0.07) 2.22(0.07)*** 1.84(0.07)**** 2.09(0.07)**** 2.64(0.07)** 2955 21.32 <0.0001  147.4(1.1) 159.7(2.4) 132.3(2.4) 84.0(2.4)** 133.0(2.4) 159.6(2.4) 4 1.62 0.34  10.5 9.7 10.9 9.6 9.6 10.2  9.26 8.97 8.71 6.36 7.22 9.69  107.28 94.9 92.58 107.75 81.24 134.23  0.086 0.095 0.094 0.059 0.089 0.072  7.0 7.5 6.0 13.0 6.9 7.8  0.36(0.02) 0.52(0.05)*** 0.16(0.05)**** 0.42(0.05) 0.36(0.05) 0.41(0.05) 86 5.72 <0.0001  75.76(1.1) 84.10(2.5) 70.96(2.5) 39.36(2.5)** 69.49(2.5) 76.23(2.5) 4 1.76 0.31  10.5 9.7 10.9 9.6 9.6 10.2  1.36 1.82 0.63 1.45 1.24 1.51  24.93 29.07 18.63 86.78 24.26 31.35  0.055 0.062 0.034 0.017 0.051 0.048  *Notes: Decay rate (k) and initial concentration (C0 ) are slope and intercept estimates (SE) from a repeated measures model. D50 is the median particle size measured by the LISST at the start of the measurement period. Significance codes next to k and C0 indicate the p-value of the comparison with the reference condition: ‘****’ = p < 0.001; ‘***’ = p < 0.01, ‘**’ = p < 0.05, ‘*’ = p < 0.1. Also provided are the degrees of freedom (DF), F-statistic (F), and significance value (p) associated with the k estimate for the whole model. H is flow depth. See text for descriptions and calculations of depositional velocity (wd), particle settling velocity (ws), and enhancement factor (Ed). Each experiment corresponds to a different growth period (number in weeks).  151  Figure 5.15: Log-transformed concentration normalized by the initial concentration (ln(C/C0) as a function of time (in seconds) for a) 0-1h and b) 1-8h for algae experiments. 0.5  A4  a) 01h  A8  A13  A16  A20  0 -0.5  ln(C/C 0)  -1.0 0 0  2000 b) 1-8h  0  2000 A4  2000  A8  0 A13  2000  0  2000  A16  A20  -0.5 -1.0 -1.5 5000 15000 25000  5000 15000 25000  Time (Seconds)  Figure 5.16: Decay rates (k) for the 0-1h and 1-8h time periods, near-bed (τRe0) and maximum (τReM) Reynolds stress, and AFDM densities for algae experiments compared to the reference condition. 4  3  3 2  1  2  0  Reynolds stress (Pa)  Decay rate (k; x 10-4 s-1  4  0-1h 1-8h τRe0 τReM  1  AFDM (g/m2)  -1  -2  0 30 20 10 Reference  A4  A8  A13  A16  A20  *Notes: Decay rates and Reynolds stress values are estimates (SE) from a repeated measures model fit to the decline in log-transformed concentration versus time. ‘A’ (‘Algae’) refers to a periphyton assemblage dominated by filamentous green algae; experiment codes include the number of weeks of periphyton growth.  152  Figure 5.17: Decay rates (k) of four particle size classes (~2-4, 4-16, 16-63, and 63-122 µm) and the reference condition (black bars) from 0-1h and 1-8h for algae experiments and the reference condition. a) 0-1h 3  2-4 µm  4-16 µm  0.6  b) 1-8h 2-4 µm  2  4-16 µm  -4  -1  Decay rate, k (x 10 s )  3  2  2  1  1  1  0 25  0.2  0  16-63 µm  200  63-122 µm  0.0 3  0 Ref  16-63 µm  A4  A8 A13 A16 A20  -4  -1  Decay rate, k (x 10 s )  0.4  20 150 2  15 100  10  1 50  5 0  0 Ref A4  A8 A13 A16 A20  0 Ref A4  A8 A13 A16 A20  Ref  A4  A8 A13 A16 A20  *Notes: Values and standard errors (error bars) are estimates from a repeated measures model. Note the difference in scale among particle size classes. No data plotted indicates there were not enough measurements to fit the repeated measures model. ‘Ref’ values are estimates from the data of five replicate reference experiments; ‘A’ (‘Algae’) refers to a periphyton assemblage dominated by filamentous green algae; experiment codes include the number of weeks of periphyton growth. Trends in deposition differed among particle sizes in both time periods. Some of these differences may help explain the three relations outlined above. First, total deposition was fastest in A20 early in the experiment despite low-to-moderate deposition rates for large particles (> 63 µm) because the rate of fine particle deposition was high relative to the other experiments (Figure 5.17a). Meanwhile, although fine particle deposition was also rapid at A13, very low rates of deposition for particles > 63 µm slowed total deposition. Algae assemblages of these two experiments differed in biomass, structure, and resulting hydraulic conditions; A20 had a bumpy surface, a moderate biomass, and high maximum Reynolds stresses, while A13 lacked structures, was at a lower biomass, and had a low maximum Reynolds stress. However, both A13 and A20 had a relatively low near-bed Reynolds stress. Possibly, the high maximum Reynolds stress of A20 increased large particle deposition by increasing eddy penetration and particle transport into the bed (Hypothesis 5). High advective transport and vertical flow penetration into openweave filamentous algae has been documented previously (Dodds and Biggs 2002). Meanwhile, fine particle deposition was likely promoted in both A13 and A20 by low near-bed Reynolds stresses that allowed small particles to settle and infiltrate. Similarly, A4 had relatively low maximum and near-bed Reynolds stresses that would have promoted the deposition of very fine particles and increased total decay rate, particularly late in the experiment when very fine particles dominated the suspension (Figure 5.17b). In contrast, A8 had the slowest deposition late in the experiment, likely a consequence of Reynolds stresses lower than A20 (limiting advection) but a biomass higher than A13 (clogging pore spaces). Thus algal  153  biomass appears to have a generally negative effect on deposition in cases where Reynolds stresses are not high enough to increase the transport of particles into the bed.  Surface and subsurface samples  Surface samples further support the prediction that an increase in algal biomass did not increase deposition (Hypothesis 4). Although all algal experiments had higher surface ash-mass (AM) than the reference condition, suggesting that surface algae did promote surface deposition in some way, AM shows no clear relation with biomass (Figure 5.18a). However, AM did decrease gradually from A8 through A20 (Figure 5.18b) indicating that algal growth stage and structure influenced surface deposition. Structural development appears to have decreased surface deposition; thus the high decay rates measured in the lategrowth stage, moderate-biomass A20 were likely due to higher infiltration into the bed and retention below the surface. An increase in deposition with growth stage is not detectable in surface samples.  Figure 5.18: Ash-mass of algae surface samples plotted against diatom AFDM (a) and growth stage (b). 25 a)  b)  AM (g/m2)  20  15  10  5  0 0  10 20 AFDM (g/m2)  30  0  5  10 15 Growth period (weeks)  20  25  *Notes: Error bars are standard errors determined from a repeated measures model Subsurface samples show generally lower amounts of infiltration for algae experiments relative to the reference condition (Figure 5.19), in correspondence with generally slower deposition. Whereas low biomass algae experiments (A4 and A8) had percentages of fine particles higher than the reference condition in the armour layer, all algal experiments have less particle deposition in the deeper layers. For example, while the armour layer percentage of A8 was significantly greater than the reference (p < 0.1), the subsurface, middle, and base layer percentages of A8 were all significantly less than the reference (p < 0.05). Interestingly, A4, A13, and A20 all had higher base layer percentages than other layers and other algal experiments. Low biomass and high porosity of A4 likely allowed for deep infiltration similar to the  154  porous reference bed; A13 also had a moderately low biomass with a similar effect. As noted above, deep infiltration of A20 was likely promoted by high maximum Reynolds stresses, eddy penetration and particle transport into the bed (Hypothesis 5). Figure 5.19: Percentage of fine particles (<125 µm) in bulk bed layers of three sections of the flume bed from algae experiments compared to the reference condition. -0.1 0  -1  -2  0.0  Percent fine particles (<125 µm) 0.1 0.2  0.3  0.4 Reference A4 A8 A13 A16 A20  Depth (cm)  -3  -4  -5  -6  -7  -8  *Notes: Depths representing the middle of each layer have been offset vertically so that error bars can be distinguished; true depths are the middle of each layer. Dotted lines separate layers. Percentages were calculated from the mass of particles <125 µm divided by the total mass of each bed layer. Error bars are the standard error of the estimate from a repeated measures model. ‘A’ (‘Algae’) refers to a periphyton assemblage of a filamentous green algae; experiment codes include the weeks of periphyton growth. Sum of effects  Randomly varying bed configurations, a variable biomass-growth stage relation, and changing hydraulics due to algal development produce an inconsistent effect of algae on particle deposition. As a result no clear pattern of effects is discernible from the data. Nevertheless, some general observations are suggestive. Slow deposition at A8 reflected lower levels of infiltration, likely due to the lower Reynolds stresses and higher biomass of this experiment that resulted in less vertical transport of particles into the bed. In contrast, the low Reynolds stresses of A4 and A13 were counteracted by the effects of lower  155  biomass, patchy algal coverage and high bed porosity allowing deep infiltration and faster total deposition. Decay rate of A16 was low in the early part of the experiment, but similar to other experiments later on, balancing the influence of high maximum Reynolds stresses (enhancing vertical transport) and high biomass (clogging) effects. Low subsurface particle amounts in A16 suggest that the high biomass of this experiment reduced infiltration, as was observed in the high biomass diatom experiments. Total decay rate of A20 early in the experiment was higher than the reference condition, probably due to higher armour layer (but not surface) deposition and similar particle amounts with depth. Relatively high amounts of infiltration but low levels of surface deposition in both A20 and the reference condition suggest that under these high-roughness surface conditions, increased turbulence intensity, high porosity and frequent trapping by surface structures increased particle infiltration below the surface, while retention on the stone surfaces was limited. In sum, the relation between particle deposition and algae growth was influenced by interacting effects of turbulent transport, bed porosity, and surface trapping.  5.4.5 RELATION BETWEEN PERIPHYTON BIOMASS AND INORGANIC MASS  As seen in the pattern of decay rates among experiments, the relation between particle deposition and periphyton biomass was complex and variable depending on periphyton type (diatoms or algae) and structure (low profile or highly structured). For diatoms, decay rates did not simply increase with biomass, as might be expected if only adhesion influenced the number of particles removed from suspension. Instead, decay rate reflected the combined effects of adhesion, hydraulic conditions, and bed clogging imposed by the diatom assemblages at different growth stages, densities, and structures. Inorganic content of surface samples did not differ much between diatom experiments, despite significant changes in biomass, suggesting either that a) other factors affecting surface deposition (e.g. hydraulics) counteracted the increase in biomass or b) deposition was high enough to saturate the diatom surface. Furthermore, all diatom surface samples had significantly higher inorganic contents than all non-periphyton surfaces; thus any observed decrease in decay rate (e.g. experiment D12) must be attributed to a reduction in subsurface deposition. Among algae experiments, there was no relation between biomass and inorganic content of surface samples. Furthermore, the experiment with the lowest surface deposition actually had the fastest decay early in the experiment (A20), indicating that low surface deposition must have been outweighed by higher levels of subsurface deposition. Decay rates also show that high biomass was associated with lower total deposition, likely by limiting percolation and subsurface retention. Meanwhile low biomass and late algal growth stage structures (producing high shear stresses) increased decay rates. My results thus demonstrate that surface deposition and total decay rate (i.e. the rate of decline in suspended concentration) may not correspond due to the additional contribution of subsurface infiltration and retention to particle removal from the water column. Furthermore, my results show that while adhesion to diatoms may increase surface  156  deposition to a certain degree, this effect may be either lessened or enhanced by hydraulic conditions generated by the diatoms themselves. The amount of surface deposition should be a function of both trapping by periphyton surfaces and near-bed hydraulics. Trapping includes direct interference by high-profile structures or adhesion to mucilaginous surfaces; thus the amount of particle trapping depends on the extent of surface structures, the stickiness of periphyton, and the surface area available to which particles may bind, in turn a function of periphyton composition, density, and structure. Hydraulic conditions can also be altered by periphyton structure and composition. As expected, my results show that the mucilaginous composition of diatom assemblages did increase surface deposition, but the relation with diatom biomass was counteracted by hydraulic effects. Non-mucilaginous algae also increased surface deposition relative to non-periphyton surfaces, but showed no relation between inorganic content and biomass due to the increase in shear stresses. My dataset is relatively limited, however, showing the effects of only two periphyton compositions at a small number of densities and structures. In order to better assess the relation between periphyton biomass and inorganic surface deposition, I compiled data from several previous studies that measured AM and AFDM of surface periphyton samples for a range of locations, stream types, hydraulic conditions, periphyton types, substrates, and sediment regimes (Table 5.7). I compared these results to my experimental data and to the field samples of green algae I collected from Hope Slough for the study presented in Chapter 4 (Figure 5.20).  Figure 5.20: Relation between periphyton biomass (AFDM) and inorganic content (AM) of surface samples from field studies with a range of locations, stream types, hydraulic conditions, periphyton types, substrates, and sediment regimes (Table 5.7), compared to experimental data and field samples of green algae collected from Hope Slough, British Columbia. 10000.00 1000.00  2  AM (g/m )  100.00 10.00  Graham (1990) Yamada and Nakamura (2002) Kiffney et al. (2003) Jowett and Biggs (1997) van Dijk (1993) Collier (2002) Runck (2007)  1.00  Hope River Flume - Diatoms Flume - Algae  0.10 0.01 0.1  1  2  AFDM (g/m )  10  100  *Notes: Values are plotted on a log-log scale because of the wide range of values.  157  Table 5.7: Information about field studies used to determine the relation between AM and AFDM of surface periphyton samples. Authors Collier, 2002  Graham, 1990  Site and Substrate Description Flow-regulated stream, reach width ~40 m Affected by volcanic eruption and dam flushing Prior to flushing: Small to large cobble substrate After flushing: increased sand to 22%, gravel to 40% Flow-regulated braided river, riffle-run reach High suspended glacial silt  Jowett et al., 1997  Kiffney et al. 2003  Runck, 2007  vanDijk et al., 1993  Tributary (reference) with low SSC Two New Zealand rivers One with high SSC/ Q; one with low SSC/Q Artificial substrates, weekly accumulation Forested streams, logged, pool-riffle morphology  Flow Conditions Controlled below dam  Data Provided  Periphyton Type  AFDM (g/m2)  Unknown  Storms >100 m3/s  AFDM/DM  Baseflow ~0.6 m3/s  Chl a (g/m2)  Downstream <20.5 m3/s  ISIS  Steady Q = 4.7 m3/s  AFDM, OM%  Diatoms (dominant)  Mean velocity = 0.8 m/s  Silt (g/m2)  Green algae  Chl a (g/m2)  Cyanobacteria 2  High (60-90 cm/s)  AFDM (g/m )  Several genera  Low (20-50 cm/s)  Chl a (g/m2)  Unknown form  Silt (#/m2) Unknown  AFDM (mg/m2)  Artificial substrate (tiles)  AM (mg/m2)  Buffer widths 0, 10, 30 m and clear-cut  Chl a (mg/m2)  Industrially-contaminated  High  Nutrient enriched  Low  AFDM (mg/m2) AM (mg/m2) Chl a (mg/m2)  Epiphytic lake periphyton  Chl a (mg/m2)  Glass slide and macrophyte substrates  Chl a/AFDM AM/DM  Yamada and Nakamura, 2002  Three streams, different channel condition:  Steady, low flows  Channel works (decreased slope, velocity) Bank stabilization None  Diatoms (low light) Filamentous (high light)  Unknown  Green algae (Scenesmus sp.) Diatoms (Diatoma sp.) (dominant) Nitzschia sp. (dominant)  AFDM (g/m2)  Filamentous  DM (g/m2)  Ulothrix spp.  2  AM (g/m ) Chl a (g/m2)  Cladophera spp  Although there is a large degree of scatter, AM is positively related to AFDM (Figure 5.20). An early paper on this subject (Graham, 1990) highlighted the positive relation between periphyton biomass and silt deposition in a flow-regulated, braided stream with high loads of suspended glacial silt. Periphyton assemblages in his study were dominated by diatoms mixed with cyanobacteria and green algae. Graham (1990) found that the silt mass of samples increased at three times the organic weight of periphyton and most of the periphyton dry mass was inorganic (65-90% inorganic).  158  Other studies measured similarly high inorganic percentages (up to 90%) and a relation between AM and AFDM (Figure 5.20). Most of the field data, with the exceptions of Hope Slough, Yamada and Nakamura (2002), and Runck (2007), closely fit a power relation between AM and AFDM with an exponent of ~1-1.5), indicating that AM often increases out of proportion to AFDM (Table 5.8). However, AM values are shifted higher for one data set; AM values measured by Jowett and Biggs (1997) for a given AFDM are up to 100 times higher than other studies. In this study, silt content and periphyton accumulation on artificial substrates were sampled weekly in two New Zealand rivers with contrasting flow and sediment regimes: one with high and the other with low discharge and suspended sediment concentrations (SSC). Periphyton biomass was determined to be a better predictor of silt density (measured as the number of individual particles per area) than either local hydraulics or SSC. Although high SSC provided a higher supply of particles available for deposition, Jowett and Biggs (1997) suggested that silt accumulation itself decreased SSC by removing particles from the water column thus there was not a clear relation between these factors. Neither periphyton composition nor structure was specified in this study thus it is impossible to assess whether these factors might explain the high AM values. In general, periphyton composition of field studies is uncertain; most studies report a mixed assemblage of diatoms, filamentous green algae, and cyanobacteria or simply do not identify the form of periphyton under study. Thus it is difficult to determine how important periphyton structure is to particle deposition relative to other factors such as SSC or flow level.  Table 5.8: Regression coefficients for the relation between periphyton biomass (AFDM) and inorganic content (AM) of surface samples from field studies with a range of locations, stream types, hydraulic conditions, periphyton types, substrates, and sediment regimes. Authors/Site Collier, 2002 Graham, 1990 Jowett et al., 1997 Kiffney et al. 2003 vanDijk et al., 1993 Yamada and Nakamura, 2002 Hope Slough  n 28 5 55 36 168 12 20  Regression equation AM = 7.99 × AFDM 0.91 AM = 2.25 × AFDM 1.04 AM = 12.37 × AFDM 1.34 AM = 1.54 × AFDM 1.53 AM = 2.51 × AFDM 1.34  r2 0.70 0.58 0.75 0.84 0.86  AM = 13.7 × AFDM 0.66  0.10  0.53  0.29  AM = 3.32 × AFDM  *Notes: ‘n’ is the number of AFDM/AM samples used to fit the regression. Data from Runck, 2007 contained only two measurement points thus the coefficients for this data set are not reported. My experimental data show a distinctly different trend than most the field data, as well as a difference in the AM-AFDM relation between diatoms and algae. Instead of increasing, the AM of my experimental samples remain virtually constant with AFDM. For a given density, however, the AM of diatoms was higher than for algae, such that the percent inorganic matter was higher for diatoms (61-91%) than algae (32-61%). As noted before, some of this AM may have been due to the silica content of diatoms, although the contribution is likely to be small. My experimental data differ from the field data in two main ways: my experiments were run for a fixed time (8 hours) and with a constant sediment load, while field samples  159  reflect a long period of flow with sediment renewal of varying frequency and amount. For these reasons, the two sets of data may not be directly comparable. Field samples from Hope Slough did have similar levels of AM to the experimental algae samples, suggesting that periphyton composition also had effect on particle deposition. However like my experiments, Hope Slough has a very low suspended sediment load thus the relative importance of periphyton composition and sediment cannot be conclusively determined from these data. Only one of the field studies I reviewed specified a filamentous form of algae (Yamada and Nakamura, 2002), but samples from this study had higher inorganic content than my experimental algae, similar to other studies at the same density. Another study sampled both diatom and algae-dominated assemblages (Kiffney et al., 2003), but the AM-AFDM relation is not distinctly different between types. In this study, small forested streams were subjected to varying buffer widths and logging intensities, from which Kiffney et al. (2003) documented how changes in light availability related to periphyton composition, biomass, and inorganic mass. For wide buffer widths, low light levels corresponded with diatom-dominated assemblages, low biomass, and low inorganic mass. As buffer width decreased, light levels, periphyton biomass and periphyton inorganic mass increased. What cannot be determined from these results, however, is whether the increase in biomass caused the increase in inorganic mass, or whether both were responding to the reduction in buffer width, which might alter other factors such as sediment load or flow rate. Thus, given the many potentially interacting factors that can influence surface sediment deposition in the field, it is arguably not surprising that the effects of periphyton structure and composition cannot be deciphered from the field data. Likewise, several studies emphasize the importance of feedbacks and competing effects between periphyton biomass and inorganic content. For instance, an increase in sediment load might be expected to increase deposition, but high suspended sediment concentrations can reduce light availability and photosynthesis, in turn reducing periphyton growth and its effect on deposition. Furthermore, scour by sand-sized particles during high flows can also reduce periphyton biomass (Francoeur and Biggs, 2006). One response to the potential negative effects of sediment on periphyton growth is to use the ‘autotrophic index’ (AI), or the ratio of AFDM to chlorophyll a, which estimates the proportion of non-living periphyton biomass, for comparisons with AM (Yamada and Nakamura, 2002), but this was not necessary in my experimental study because my growing periphyton were not subjected to high sediment loads. Conversely, sediment deposition to existing periphyton can reduce nutritional quality, leading to limited grazing intensity and thus higher standing stocks. In my study, however, I cultivated periphyton without suspended sediment or invertebrate grazers and increased SSC for only a short period on a static level of periphyton growth thus isolating the one-way effect of periphyton on deposition. In addition, my sediment was <125 µm and velocities were relatively low (<50 cm/s) thus sediment-induced scour of periphyton was unlikely, even at high suspended concentrations.  160  5.4.6 COMPARISONS BETWEEN DIATOMS AND ALGAE  Four major differences were observed between diatom and algal assemblages. Firstly, whereas nearbed Reynolds stresses of both types steadily increased with periphyton growth stage, Reynolds stresses directly above green algal assemblages were almost twice as high as above diatom and bacterial assemblages (Table 5.3), with the exception of the smooth-surface, moderate-biomass A13. Maximum Reynolds stresses, mid-flow depth, were also higher above green algae, though not to the same degree. I speculate that changes in shear stress with growth were related to changes in structure that increased bed roughness; green algae may have had more irregular surfaces and larger roughness elements, such as the bumps and webbed filaments I observed. Secondly, general trends in particle decay rates, as compared to the reference condition, differed between diatom and algal assemblages. Overall, most particle sizes deposited faster in the diatom experiments than in the reference condition, particularly D8, which had a moderate biomass (~5 g/m2) and relatively low shear stresses (0.50-0.9 Pa). In contrast, particles generally deposited slower in algae experiments than in the reference condition, with the exception of the low-biomass A4 and the moderatebiomass, highly structured A20. However, it is important to note that relative measures of biomass and shear stress also differ between algae and diatoms. After eight weeks of growth, algae growth stages had significantly higher densities than all diatom stages; algal biomass accumulation was nearly twice that of diatoms. Thirdly, inorganic surface deposition was higher for diatoms and algae relative to non-periphyton surfaces, but deposition to all diatom growth stages was higher than to algae despite generally lower biomasses. Deposition and retention of particles at the bed surface was likely enhanced by the mucilaginous polysaccharide exudates of diatoms and bacteria, while this form of green algae does not produce these substances. Finally, the relation among biomass, shear stress, and infiltration also differed for the two periphyton types. For diatoms, an increase in biomass resulted in a steady decrease in the amount of particle infiltration to all subsurface layers, despite the increase in near-bed and maximum Reynolds stresses. In contrast, particle infiltration into algae beds increased with growth stage and structural development as Reynolds stresses increased. Unlike diatoms, however, an increase in algal biomass only slightly reduced the degree and depth of infiltration, despite much higher densities than diatoms. Compared to the reference condition, however, algae had similar or lower amounts of particle infiltration, while the low biomass, low Reynolds stress diatoms (D4 and D8) had higher amounts of infiltration. Possibly, the high biomass of the later-stage (> 4 week) algae experiments limited infiltration somewhat (hence lower-than-reference infiltration levels at A8 and A13) but this effect was offset by high shear stresses and downward advection generated by very late-stage and highly structured algae (e.g. A20). Diatoms did not show an increase in infiltration as Reynolds stresses increased because advective forces were still much lower than late-stage algae, even at  161  24 weeks of growth. Furthermore, the structure of diatom assemblages may have been less porous and permeable than algae because exopolysaccharides form a cohesive seal over the surface. Similar effects have been observed in the sediments of tidal flats and estuaries, where a matrix of diatoms and exopolysaccharides covering a sediment surface can trap particles and increase bed cohesion (Staats et al., 2001). Infiltration into diatom beds was thus highest when biomass was low.  5.5 CONCLUSIONS  Data from a series of flume experiments were used to investigate the influence of streambed periphyton on particle deposition. I measured the effects of two contrasting forms of periphyton at several densities and growth stages on near-bed hydraulics, particle loss from the water column, surface deposition, and subsurface infiltration. Relative to a non-periphyton reference condition, periphyton assemblages altered the rate and quantity of particle deposition via several mechanisms, including shear stress modification, surface adhesion, and bed clogging. Effects were highly variable, however, being dependent on periphyton structure and biomass, particle size, and whether deposition is to the surface or subsurface of the bed. In general, diatoms increased deposition rates relative to reference conditions by reducing shear stresses and enhancing surfac