UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Forests, floods and channel processes : illuminating links between forest harvesting, the flood regime… Green, Kim Cordelia 2013

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

Item Metadata

Download

Media
24-ubc_2014_spring_green_kim.pdf [ 9.24MB ]
Metadata
JSON: 24-1.0166834.json
JSON-LD: 24-1.0166834-ld.json
RDF/XML (Pretty): 24-1.0166834-rdf.xml
RDF/JSON: 24-1.0166834-rdf.json
Turtle: 24-1.0166834-turtle.txt
N-Triples: 24-1.0166834-rdf-ntriples.txt
Original Record: 24-1.0166834-source.json
Full Text
24-1.0166834-fulltext.txt
Citation
24-1.0166834.ris

Full Text

Forests, floods and channel processes: illuminating links between  forest harvesting, the flood regime and channel response in  snowmelt headwater streams   by Kim Cordelia Green B. Sc., University of British Columbia, 1986 M. Sc., University of Calgary, 1990  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy in The Faculty of Graduate and Postdoctoral Studies (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2013 ? Kim Cordelia Green, 2013 ii Abstract A meta-analysis of four snowmelt catchments with moderate harvest levels (30% to 40%) utilizing a frequency-based approach demonstrates, despite a century?s-worth of studies to the contrary, how harvesting increases the magnitude and frequency of all floods on record including the largest floods (R.I. = 1:50 yrs.) and how such effects increase unchecked with increasing return period as a consequence of changes to both the mean and standard deviation of the flood frequency distribution. Additionally, meta-analysis outcomes reveal up to three-fold increases in the number and duration of peak flows with return periods ranging from 0.6*Q1.5 to Q10, which includes floods capable of mobilizing bedload and altering the form of gravel-bed streams. A frequency-based meta-analysis provides new insights concerning the physical processes governing the relation between forests and floods in snowmelt environments that were previously unrecognized using traditional chronological-pairing methods. The dominant process responsible for flood regime changes following harvesting is the increase in basin-average snowmelt rates that are amplified or mitigated by physical characteristics such as aspect distribution, elevation range, slope gradient and amount of alpine area.  The outcomes of a high-resolution, nested-monitoring investigation of hydrologic and geomorphic controls on bedload mobility indicate that flow regime changes from harvesting in snowmelt streams can alter rates of bedload transport, a first-order determinant of channel form in fluvial systems. Regression analysis shows that annual sediment yield is controlled by the number of peaks-over-threshold discharge. During peak events, repeated destabilization of channel armor and re-mobilization of sediment temporarily stored behind LWD generates bedload transport across the entire snowmelt season. However, the potential for channel response iii to changes in the flow regime depends on characteristics of channel morphology. Study results indicate that patterns of bedload entrainment and mobility are influenced by flow resistance associated with channel form, grain size and LWD while the value of the critical dimensionless shear stress varies with channel gradient, relative and absolute grain effects and flow resistance. Differences in bed texture, gradient and channel form contributes to variations in rates and characteristics of bed mobility, hence differences in potential for alteration due to changes in the flow regime.  iv Preface All components of the research presented herein represent original work for which I have been the lead investigator.  Portions of Chapter 2 have been published: Green, K.C. and Y. Alila. 2012. A paradigm shift in understanding and quantifying the effects of forest harvesting on floods in snow environments. Water Resources Research VOL. 48, W10503, doi:10.1029/2012WR012449, and Green, K.C. Brardinoni, F. and Y. Alila. 2012. Channel morphology and bed-load yield in fluvial, formerly-glaciated headwater streams of the Columbia Mountains, Canada. Geomorphology 188, 96-109. Chapter 3 has been published: Green, K.C. and Y. Alila. 2012 (cited above). In addition the outcomes of this investigation have been summarized and published in an American Geophysical Union press release: Treacy, S., 2012. Deforestation in snowy regions causes more floods. American Geophysical Union, Release No. 12-43 (download from; http://www.agu.org/news/press/pr_archives/2012/2012-43.shtml). This publication has generated two comments by European researchers: Bathurst, J.C., (in press). Comment on: Green, K. C., and Y. Alila (2012), A paradigm shift in understanding and quantifying  the effects of forest harvesting on floods in snow environments, Water Resources Research, 48, W10503, doi:10.1029/2012WR012449. Submitted to Water Resources Research. And, Birkinshaw, S.J., (in press). Comment on ?A paradigm shift in understanding and quantifying the effects of forest harvesting on floods in snow environments? by Kim C. Green and Younes Alila. Submitted to Water Resources Research. v Dr. Younes Alila assisted in study design for the analyses presented in Chapter 3 and provided input on the discussion and conclusions. This chapter has been modified from the original manuscript to fit within the overall dissertation.  Chapter 5 has been published: Green, K.C. Brardinoni, F. and Y. Alila. 2012 (cited above). Dr. Francesco Brardinoni assisted with GIS analysis used in the process domain analysis and with interpretation of study results and discussion. Dr. Younes Alila contributed to setup of the experimental sites. This chapter has been modified from the original manuscript to fit within the overall dissertation. Chapter 6 has been submitted for publication.  Dr. Francesco Brardinoni assisted in interpretation of results and discussion. Dr. Younes Alila assisted in experimental set-up. This chapter has been modified from the original manuscript to fit within the overall dissertation.  A portion of Chapter 7 (Chapter 7.3.1) has been published in Green, K.C. and Y. Alila, 2012 (cited above).  vi Table of contents Abstract ..................................................................................................................................... ii Preface ...................................................................................................................................... iv Table of contents ....................................................................................................................... vi List of tables ...............................................................................................................................x List of figures ............................................................................................................................ xi List of symbols and abbreviations ............................................................................................xiv Acknowledgements ................................................................................................................ xvii Dedication ................................................................................................................................xix 1 Introduction .........................................................................................................................1 1.1 Effects of forest harvesting on floods in snowmelt-dominated streams................5 1.1.1 The problem with traditional research methods .............................................8 1.2 The flood regime and headwater streams .......................................................... 11 1.2.1 Channel maintaining flows ......................................................................... 12 1.2.2 Channel forming floods .............................................................................. 13 1.3 Effects of flood regime changes on headwater streams ..................................... 14 1.4 Thesis objectives and structure ......................................................................... 17 2 Study sites and experimental design .................................................................................. 19 2.1 Forest harvesting effects on the flood regime .................................................... 19 2.1.1 Fool Creek .................................................................................................. 22 2.1.2 Camp Creek ................................................................................................ 23 2.1.3 Redfish Creek ............................................................................................. 24 2.1.4 240 Creek ................................................................................................... 25 2.1.5 Simulation of harvesting scenarios by DHSVM .......................................... 26 vii 2.2 Forested snowmelt headwater stream morphodynamics .................................... 27 2.2.1 Geology and physiography ......................................................................... 27 2.2.2 Monitoring site characteristics .................................................................... 32 3 Forest harvesting effects on frequency and magnitude of snowmelt floods ........................ 38 3.1 Introduction ...................................................................................................... 38 3.2 Methods ........................................................................................................... 41 3.2.1 Overview .................................................................................................... 41 3.2.2 Adjusting for non-stationarity due to forest regeneration ............................. 42 3.2.3 Estimation of expected posttreatment discharges ........................................ 44 3.2.4 Flow frequency curve analysis .................................................................... 45 3.2.5 Statistical vs. physical significance ............................................................. 46 3.3 Results ............................................................................................................. 48 3.3.1 Meta-analysis and the frequency pairing framework ................................... 48 3.3.2 Chronological pairing and the missing dimension of frequency................... 51 3.3.3 Physical processes investigation ................................................................. 56 3.4 Discussion ........................................................................................................ 59 4 Harvesting effects on flood duration and number of peaks over threshold discharge .......... 65 4.1 Introduction ...................................................................................................... 65 4.2 Methods ........................................................................................................... 67 4.2.1 Duration over threshold .............................................................................. 69 4.2.2 Peaks over threshold ................................................................................... 70 4.3 Results ............................................................................................................. 73 4.3.1 Flood duration ............................................................................................ 73 4.3.2 Number of peaks-over-threshold ................................................................. 74 4.4 Discussion ........................................................................................................ 75 5 Channel morphology and bedload yield dynamics forested snowmelt streams. .................. 79 viii 5.1 Introduction ...................................................................................................... 79 5.2 Data collection and methods ............................................................................. 81 5.3 Results ............................................................................................................. 89 5.3.1 Basin structure, channel morphology, and geometry ................................... 89 5.3.2 Daily bedload yield ..................................................................................... 94 5.3.3 Variability of bedload yield across channel morphologies ........................... 98 5.3.4 Morphological and hydrological controls on annual sediment yield .......... 102 5.4 Discussion ...................................................................................................... 104 6 Patterns of bedload entrainment and transport in forested snowmelt streams .................... 113 6.1 Introduction .................................................................................................... 113 6.2 Methods ......................................................................................................... 117 6.2.1 Discharge gauging .................................................................................... 117 6.2.2 Bedload monitoring .................................................................................. 118 6.2.3 Channel bed texture .................................................................................. 120 6.2.4 Initiation of motion ................................................................................... 121 6.2.5 Bedload transport...................................................................................... 124 6.3 Results ........................................................................................................... 125 6.3.1 Characteristics of bed mobility ................................................................. 125 6.3.2 Initiation of bedload transport and relative bed mobility ........................... 127 6.3.3 Relative bedload mobility with increasing discharge ................................. 133 6.3.4 Bedload rating curves ............................................................................... 136 6.4 Discussion ...................................................................................................... 141 6.4.1 Bedload entrainment ................................................................................. 141 6.4.2 Bedload transport...................................................................................... 145 7 Conceptual model of channel response to harvesting in forested snowmelt streams ......... 151 ix 7.1 Introduction .................................................................................................... 151 7.2 Physical process understanding ...................................................................... 152 7.3 Methods ......................................................................................................... 154 7.4 Conceptual model development ...................................................................... 156 7.4.1 Potential for flow regime changes ............................................................. 156 7.4.2 Potential for channel response................................................................... 158 7.5 Results and discussion .................................................................................... 161 7.5.1 Application of model outputs to forest management.................................. 167 8 Summary and conclusions ............................................................................................... 168 8.1 Forest harvesting effects on the flood regime .................................................. 168 8.2 Hydrodynamics of forested snowmelt streams. ............................................... 171 8.3 Potential for forested snowmelt channel response ........................................... 173 8.4 Future studies ................................................................................................. 174 Bibliography ........................................................................................................................... 176 Appendices ............................................................................................................................. 202 Appendix 1: Measurement errors and uncertainties .................................................. 202 Appendix 2: Bedload transport formulae .................................................................. 204 Appendix 3 Bedload transport data and bulk channel bed data for sample sites ........ 209 Appendix 4 Channel survey data for Cotton and Elk by reach .................................. 215 Appendix 5 Sources of data for meta-analysis (Chapters 3 and 4) ............................ 216   x List of tables Table 1.1. Summary of paired watershed studies investigating the influence of forest harvesting on streamflow in snow-dominated regions of western North America .........................................6 Table 2.1. Meta-analysis catchment characteristics .................................................................... 20 Table 2.2. Monitoring site sampling details ............................................................................... 32 Table 2.3 Physical and hydrological characteristics of bedload monitoring sites. ....................... 33 Table 3.1. Linear regression analysis of meteorological variables with discharge. ..................... 57 Table 4.1. Expected (control) discharge magnitude for selected flood quantiles from GEV distribution (m3/s). .................................................................................................................... 68 Table 4.2. Differences in average number of peaks over threshold for pre- and post-treatment periods. Statistically significant differences highlighted in bold type. ........................................ 74 Table 5.1 Correlation matrix ................................................................................................... 103 Table 5.2 Stepwise multiple regression analysis results ........................................................... 104 Table 6.1. Coefficients (a) and exponents (b) determined for the power-law equation defining particle entrainment. ................................................................................................................ 131 Table 6.2. Bedload rating curves (C = complete data set, R = rising stage data). ...................... 139 Table 6.3. Bedload transport as a function of excess shear stress (C = complete data set, R = rising stage data) ..................................................................................................................... 141 Table 6.4 Hydraulic and geomorphic parameters at monitoring sites. ...................................... 142 Table 7.1 Example strategic planning rational by response category ........................................ 167    xi List of figures Figure 1.1 From Buffington et al., 2003 modified from Parker, 1990. Vector illustrates the direction of channel response for increases in discharge without increases in sediment supply. . 16 Figure 2.1. Location and topography of study areas for Chapters 3 and 4. ................................. 19 Figure 2.2.(A) Location of Cotton Creek study area with reach breaks (red) and monitoring locations (white/blue) and (B) distribution of surficial material. ................................................ 29 Figure 2.3. Distribution of (a) elevation, (b) slope gradient, and (c) aspect for Upper Cotton and Elk. ........................................................................................................................................... 31 Figure 2.4. Example of channel types present in Cotton (A-D) and Elk (E-H) Creeks. A) colluvial, B) forced step pool, C) riffle pool, D) boulder cascade, E) colluvial, F) forced step pool, G) boulder cascade, H) step-pool. ..................................................................................... 35 Figure 2.5. Plan-view and longitudinal channel sketches for 15 to 20 meter reaches up stream from bedload monitoring sites C2 to C4 on Cotton (A to C) and E2 to E4 on Elk (D to F) ......... 36 Figure 2.6. Cummulative grain size distribution of channel bed surface and sub-surface at study sites ........................................................................................................................................... 37 Figure 3.1. Flow duration curve analysis for pre- and posttreatment daily peak flows at (a) Camp Creek (19 years), (b) Fool Creek (48 years), (c) 240 Creek (95 years), and (d) Redfish Creek (99 years).  The point at which the two CDFs intersect, marked by the vertical arrow, increases with record length suggesting a no clear upper threshold to the effects of forest harvesting on floods in snowmelt-dominated hydroclimate regimes. .............................................................................. 49 Figure 3.2. The chronologically paired analysis of pre- and posttreatment daily peak flows at (a) Camp Creek (19 years), (b) Fool Creek (48 years), (c) 240 Creek (95 years), and (d) Redfish Creek (99 years). Pretreatment R2 values for 240 and Redfish Creeks are equal to 1 because the control and treatment catchments are the same for these modeled watersheds. ........................... 52 Figure 3.3. The misleading trend of decreased treatment effects for increased control catchment flood magnitude, apparent in all four data sets when CP-based analysis methods are applied (panel designation as in Figure 3.2). .......................................................................................... 53 Figure 3.4. Illustration of the changes in return period for chronologically paired floods for (a,b) Fool Creek and (c,d) Redfish Creek. In both cases moderate pre-treatment floods have been elevated in magnitude and rank to be amongst the largest floods in the harvested catchment while large pre-treatment floods drop in rank becoming more frequent in the treatment watershed. .... 54 Figure 3.5. Flow duration curve analysis of catchment average 3-day snowmelt preceding the annual maximum peak flow for: (a) 40% clearcut and control scenarios in 240 Creek and (b) 33% clearcut and control scenarios for Redfish Creek. .............................................................. 58 xii Figure 4.1. Sample daily hydrograph showing peaks over threshold. ......................................... 71 Figure 4.2. Cummulative percent exceedence of daily mean discharge for the period of record. 73 Figure 5.1. Bankfull was identified using physical indicators including vegetation changes and marks on the banks associated with scour. ................................................................................. 82 Figure 5.2. Correlation between Dmax and Wolman pebble count D84 grain size. ...................... 83 Figure 5.3. (A) Sediment basket at site E2 (0.45m x 0.3m x 0.15m) and (B) channel spanning weir at site C3 (2m x 0.7m x 0.6m). .......................................................................................... 85 Figure 5.4. Example break point analysis for sediment yield at sites (a) E4 and (b) C4. ............. 88 Figure 5.5. (a) Longitudinal profiles with landform and channel morphology and (b) Channel gradient versus drainage area with landform morphology. ......................................................... 90 Figure 5.6 Channel-reach morphology plotted by drainage area versus (a) slope; (b) Dmax; (c) shear stress; (d) Shields stress; (e) specific stream power; (f) total stream power; and (g) number of LWD pieces per unit channel length. Open and closed symbols represent reaches in Elk and Cotton Creek respectively. ........................................................................................................ 92 Figure 5.7 Downstream hydraulic geometry across the study reaches. Black symbols indicate reaches with field-measured discharges. .................................................................................... 94 Figure 5.8 Water discharge hydrograph with daily specific yield and associated Shields stress at monitoring sites C4 and E4. Dashed lines indicate Shields stress at bankfull. Note scaling factors for specific bed load yield in the secondary y-axes. ................................................................... 96 Figure 5.9 Box-plot showing the size distribution of transported mass during the rising and falling limb in Upper Cotton Creek and Elk Creek. ................................................................... 97 Figure 5.10 Daily bed load yield versus daily peak discharge (Qd) expressed as a ratio of the threshold discharge (Qt) for initiation of Phase 2 transport during rising stage (solid circles) and falling stage (open squares) periods. (a) C2 in 2006; (b) C2 in 2007; (c) E4 in 2006; (d) E4 in 2007; (e) C4 in 2006; and (f) C4 in 2007. .................................................................................. 99 Figure 5.11 Grain-size distribution of bed loads associated with rising and falling stages at sites: (a) C2; (b) E4; and (c) C4. ....................................................................................................... 101 Figure 5.12 Suspended sediment yield as a function of drainage area for the Columbia Mountains physiographic region. Filled symbols are yields published by Church et al. (1999); open symbols are yields estimated at the eight monitoring sites in Cotton Creek. Bars indicate 50% error around estimated values (Jordan, 2006) ................................................................................... 103 Figure 6.1 Grain-size distributions of channel bed surface, subsurface, and bedload material at the monitoring stations. Qbf  = bankfull flow; Qmax = largest flood; < Qth = lower than threshold discharge; and > Qth = higher than threshold discharge............................................ 127 xiii Figure 6.2. Determination of dimensionless reference shear by grain size class for each site according to Wilcock and Southard, 1988. The subjectivity associated with visual extrapolation creates greater uncertainty in the selected values of ? *i. A sensitivity analysis conducted at site C4 indicates that for grain sizes with greatest scatter in the ?*i - W*i relations (i.e., 1mm to 5.7mm) this uncertainty can reach up to 25%. ......................................................................... 129 Figure 6.3 ?*i as a function of Di/D50s. for study sites. Power-law equations are listed in Table 6.1. .......................................................................................................................................... 130 Figure 6.4. Relative mobility (left-hand panels) and scaled fractional transport (right-hand panels) of selected bedload samples with increasing discharge relative to the bed subsurface. Data points signify geometric mean diameter of grain size class. Dashed and solid lines in right-hand column differentiate between domains of over-passing (A), equal mobility (B) and partial transport (C). ........................................................................................................................... 135 Figure 6.5. Bedload transport (Qb) as a function of water discharge (Q) in: (a) Cotton Creek; and (b) Elk Creek. Bedload transport rate as a function of excess shear stress (?o-?c) in: (c) Cotton Creek, and (d) Elk Creek. ?c corresponds to ?r 16-20 depending on site. Solid symbols designate rising stage and open symbols are from falling stage periods. Best-fit power law equations given in Table 6.2. ............................................................................................................................ 138 Figure 6.6 Variation in dimensionless critical shear stress (?*c50s) for the six monitoring sites with (a) gradient, (b) bed surface texture and (c) total flow resistance. Lines presented on the figures are intended to highlight general trends. Dashed lines (e) are from Lamb et al. (2008) and represent different ratios of shear stress borne by morphological structures to total shear stress. ............................................................................................................................................... 143 Figure 6.7 Exponents of the six rating curves generally define a trend of increasing exponent ?b? with increasing D90s indicating that this exponent is a function of the size of the surface (armoring) grains. ................................................................................................................... 147 Figure 6.8 Predicted versus observed transport rates at steep forced-alluvial and gentler alluvial sites. See Appendix 2 for details regarding selected transport formulae. .................................. 150 Figure 7.1 Risk matrix for channel response. ........................................................................... 155 Figure 7.2 Conceptual model of channel response in forested snowmelt watersheds. At least three attributes from a single category are required for the assignment of response potential.  .. 162 Figure 7.3 A comparison of shade patterns for May 10th for Redfish Creek and 240 Creek. Compared to 240 Creek, Redfish has steeper gradient slopes and two dominant aspects (E/W) which causes shading and results in less direct solar radiation during the early May snowmelt period. ..................................................................................................................................... 164 Figure 7.4 Examples of colluvial (A-C), semi-alluvial (D-F) and alluvial (G-I) channels in B.C.?s forested snowmelt watersheds. ................................................................................................ 166   xiv List of symbols and abbreviations API  Air photo interpretation ASY Annual sediment yield (kg/km2) BC Boulder cascade CDF  Cummulative Density Function CP Chronological pairing of peak flows CV  Coefficient of variation of annual maximum series Cv Colluvial  D50  Diameter of median mobile particle (m) D50s  Diameter of median mobile bed surface particle (m) D50ss Diameter of median mobile bed subsurface particle (m) D84  Diameter of 84th percentile mobile bed particle (m) D90 Diameter of the 90th percentile mobile bed particle (m) Db Mean diameter of immobile boulders (m) dbf Bankfull depth (m) DHG Downstream hydraulic geometry DHSVM Distributed hydrology, soils and vegetation model Di  Grain diameter Dmax Largest mobile particle diameter Dr Duration of discharge ESL East Saint Louis fi  Relative proportion of a given grain size ?i? within the subsurface bed material fadd Additional resistance due to channel form fo Grain resistance FP Frequency pairing of peak flows FSP Forced step pool ft Total flow resistance g  Gravitational acceleration (9.81 m/s2) GEV Generalized extreme value function GIS Geographic information system xv ?  von Karman?s constant (0.4),  Ks  Roughness parameter (m),  LWD  Large woody debris n  Manning?s coefficient (s m-?) nPk Number of peaks over threshold p Probability of occurrence PDF  Probability density function pi Relative proportion of a given grain size ?i? within the mobile bedload POT Peaks over threshold Q Discharge (m3/s) ?Q Change in peak discharge (m3/s) q* Dimensionless discharge Q2  Discharge equaled or exceeded once in 2 years on average (m3/s) Q20  Discharge equaled or exceeded once in 20 years on average (m3/s) qb  Bedload transport rate (g/m/s) qb* Dimensionless bedload transport Qbf  Bankfull discharge (m3/s)  Qd Daily peak discharge (m3/s) Qi  Initial discharge for sediment mobility (m3/s) Qp  Peak discharge (m3/s) QT Discharge corresponding to the ?T? return period (m3/s) Qth Threshold discharge for bed mobility Rbf Hydraulic radius at bankfull(m) RP Riffle pool S  Channel gradient (m/m), S.D.  Standard deviation of the probability density function of peak flows SP Step pool SSY Specific sediment yield (kg/km2day) SWE Snow water equivalent (mm) u*  Shear velocity (m/s, equal to [?/?]0.5) Wbf Bankfull width (m) xvi Wi*  Dimensionless bedload transport rate WSC  Water Survey of Canada ????? Mean, ranked expected discharge corrected for loss of variance from regression  ???? Ranked expected peak discharge from regression ???? Expected peak discharge estimated through simple linear regression ?c  Critical Shields number (Shields parameter) ?w  Downstream length of macro-roughness elements (m) ?x  Downstream spacing of macro-roughness elements (m) ?  Density of water (1000 kg/m?) ?s Density of sediment (2650 kg/m3) ?*bf  Bankfull dimensionless shear stress. ?*ci  Dimensionless critical shear stress for the ?i? grain size  ?*ri  Dimensionless reference shear stress for the ?i? grain size ?c  Critical shear stress for the initiation of bed mobility (kg/ms2) ?i  Boundary shear stress at the onset of bed mobility (kg/ms2) ?o  Total boundary shear stress (kg/ms2) ?= ?*i / ?*ri Ratio of dimensionless shear for grain size ?i? and reference dimensionless shear    stress for ?i?. ? Specific stream power (Watts/m2) zo  Height above the bed where velocity is zero (m)   xvii Acknowledgements This has been a long but thoroughly engaging journey filled with many challenges. Above all I am grateful for the enduring patience and support of my supervisor Dr. Younes Alila who has proved to be a dedicated guide and mentor through all components of my research project as well as providing numerous days of hard labor in freezing cold water to help with the installation of sediment traps. I could not have made it to the end of this journey without him. I am also very grateful to my supervisor Dr. Francesco Brardinoni who, despite the challenges of thousands of kilometers and a nine-hour time difference, provided much appreciated guidance and assistance in my investigation of hydrodynamics of forested snowmelt streams. I also thank my ?support staff? including Markus Schnorbus, Georg Jost and Charles Luo who assisted with various questions regarding hydrometric data analysis. My research on hydrodynamics of forested snowmelt streams was made possible by the assistance of David Gluns (B.C. Ministry of Forests) and Drs. .Markus Weiler and Georg Jost, in the installation of sediment monitoring sites. Leslie Harker who helped me collect and sieve tons of bedload sediment deserves special recognition for her commitment to the job. Fellow grad students Pascal Szeftel and Russell Smith made evenings in Cranbrook actually enjoyable. Dr. Dan Moore?s logistical support allowed my field program in Cotton Creek to run smoothly. Dr. Ellen Wohl (Colorado State), and Dr. John Buffington (University of Idaho, USFS) provided valuable suggestions and helpful answers to questions that popped up during my studies. Dr. Brett Eaton (UBC) provided much appreciated guidance in surveying methods at monitoring sites. Kelly Elder of the USFS provided access to hydrometric data sets used in this investigation. I thank Dr. Robert Millar for being a willing and supportive member of my supervisory committee and for his contributions to my understanding of flow resistance in gravel bed streams. Dr. Abdel Azim Zumrawi (UBC, xviii Forest Resource Management) is also thanked for helping with the stepwise regression analysis. A special thanks goes to Brian Dureski, R.F.T., Ken Streloff, R.F.T., Joe Gnucci, R.P.F., Eirik Pihgin, R.P.F., Doug Thorburn, R.P.F. and Bruce Pope, R.P.F. for their ongoing support and for asking the questions that needed answering. Most importantly, and from the bottom of my heart, I thank my husband Will Halleran for his unfailing support and enthusiasm throughout this hydrological journey. This work was funded in part by Tembec Industries, Kalesnikoff Lumber Co. Ltd., Apex Geoscience Consultants Ltd, Forest Investment Account Forest Science Program project funding Y051294 (Y. Alila) and Y081214 (D. Moore) and the National Science and Engineering Research Council of Canada Discovery Grant RGPIN 194388?11 to Y. Alila.   xix  Dedication    To my family Will, Claire and Laurel,  and my mum Joyce            1 1 Introduction Most of British Columbia?s forested headwater streams are situated on Crown Land that is currently held under license for forest development. In addition to containing valuable forest resources, forested headwater streams provide important aquatic habitat as well as key physical and biological functions (Gomi et al., 2002; Richardson et al., 2005b). Headwater streams occupy a substantial portion of the land base, accounting for 70% to 80% of the total channel network of major river systems (Sidle et al. 2000, Meyer and Wallace 2001; Benda et al., 2005). A high degree of hydrogeomorphic coupling between hillsides and first and second order headwater streams facilitates the transfer of surface runoff, sediment and organic debris between terrestrial and fluvial domains (Church, 2002). This high level of connectivity between hillsides and stream channels in forested headwater catchments makes them particularly vulnerable to development-related disturbance. Additionally, downstream linkages between headwaters and higher-order reaches creates the potential for downstream cumulative impacts to water quality and aquatic ecosystems from development related disturbance in headwater streams (Reid, 1993; Gomi et al. 2002; MacDonald and Coe, 2007) In British Columbia forest practices regulations state that forest licensees must prevent cumulative hydrological effects of primary forest activities within high value fisheries and consumptive-use watersheds from resulting in material adverse impacts on the quantity of water or the timing of the flow of the water (Chapter 8.2(2) BC Forest Planning and Practices Regulation (2004) of the BC Forest and Range Practices Act). A prerequisite to preventing cumulative hydrological effects is the knowledge of appropriate levels of  2 forest harvesting to minimize the risk of impacting hydrogeomorphic processes governing water quality and channel stability in headwater streams.  The morphology of headwater streams reflect the underlying geology, glacial history and hydro-climate regime of a region (Montgomery, 1999; Brardinoni and Hassan, 2007). Forested headwater streams represent a distinct subset of headwater streams where channel form and processes are influenced by the addition of large woody debris from adjacent riparian stands (Hassan et al., 2005). Much of the past research on logging effects in British Columbia?s forested headwater streams has focussed on steep, rainfall- or rain-on-snow dominated systems where logging-related changes in the frequency of shallow mass wasting events imposes both immediate and long-term changes to fluvial form and processes (Hogan, 1989; Brardinoni et al., 2003; Gomi and Sidle, 2003). Although important progress has been made in understanding the effects of forest development on the hydro-geomorphology of humid, steep headwater channels (Gomi and Sidle, 2003; Richardson et al., 2005a; Hassan et al., 2005), many of British Columbia?s headwater streams are not necessarily synonymous with steep, shallow mass-wasting dominated systems.  Large portions of B.C. exhibit hydro-climatic and topographic boundary conditions that do not favor a sediment transport regime dominated by rapid shallow failures. Over 30% of the Canadian Cordillera, including the Interior and Yukon Plateaus and portions of the Columbia, Skeena, Omineca and Rocky Mountains (Holland, 1964) are situated in hydro-climate regions characterized by snowmelt-dominated, continental climates that experience a naturally low frequency of shallow mass wasting (Jordan, 2002). Impacts associated with forest harvesting in headwater streams that drain these regions are primarily due to changes in the flood regime (i.e. changes in the frequency, magnitude and duration of  3 floods) that may not be as immediately apparent as increases in mass wasting frequently documented in humid environments (e.g. Guthrie, 2002; Dhakal and Sidle, 2003) but are as equally damaging to aquatic ecosystems over the long-term (Naiman et al., 2000).  To date there have been very few studies undertaken in British Columbia to investigate the hydrogeomorphic effects of forest harvesting in snowmelt-dominated headwater streams not dominated by frequent mass wasting. While several studies have examined the influence of forest harvesting on the annual peak flows of snowmelt streams either through the traditional paired-watershed approach (Cheng, 1989; MacDonald et al., 2003; Moore and Scott; 2005) or through the use of physically distributed models (Whitaker et al., 2003; Schnorbus and Alila, 2004; Schnorbus et al., 2004) only one study has attempted to link changes in channel form and sediment yield with level of harvest (Beaudry and Gottesfeld, 2001). Catchment-level studies investigated using traditional paired-watershed methods in snow environments of western North America (Bates and Henry, 1928; Troendle and King, 1985; Cheng 1989) document changes in the timing of peak flows but have been unable to establish correlations between level of harvest and the change in the magnitude of peak flows following forest harvesting (Troendle and King, 1985; 1987; King, 1989; Troendle and Olsen, 1994; Troendle et al 2001; Cheng 1989). Conflicting study outcomes ranging from decreases to large increases in peak flows following harvesting have led many to conclude that harvest level is poorly correlated with peak flow response (Austin, 1999; MacDonald and Stednick, 2003). The apparent lack of correlation with flood response has led to limited guidance from government agencies to B.C.?s forest licensees regarding appropriate levels of harvest in high-value watersheds (Scherer and Pike, 2003; Winkler et al., 2009) forcing land managers to rely on the ?best- 4 guess? of forest hydrology professionals. However, recent revelations concerning the use of incorrect analytical techniques leading to decades of incorrect and conflicting outcomes concerning the influence of harvesting on peak flows (Alila et al., 2009) presents an urgent need to re-evaluate past study outcomes using appropriate analytical methods.  The goal of the research comprising this dissertation is to provide B.C.?s forest managers operating in snow environments with answers to the question of appropriate thresholds of harvest to minimize impacts to water quality and channel morphology in headwater streams. Research hypotheses directing this investigation are stated as follows:  1. Timber harvesting in forested snowmelt watersheds has the potential to alter the flood regime including floods larger than the mean. 2. The morphology of formerly glaciated, forested snowmelt headwater streams is adjusted to contemporary conditions of peak discharge and sediment supply.  3. Dynamics (i.e. rates and patterns) of bedload mobility in forested snowmelt headwater streams reflect differences in channel. The following sections contain a review of literature relating to the hydrologic response of snowmelt streams to forest harvesting as well as the influence of the flood regime on processes controlling channel form and material transfer in headwater streams. Information and research gaps identified in the literature review provide the framework for the main research questions of the dissertation presented in outline form at end of this chapter.    5 1.1 Effects of forest harvesting on floods in snowmelt-dominated streams The first experimental watershed in North America, Wagon Wheel Gap, Colorado, was established in 1910 specifically to address the lack of quantitative data on the influence of forests on the frequency and magnitude of floods (Hoyt and Troxell, 1932; Dobbs, 1969). Since the early 1900?s at least half a dozen subsequent studies have been undertaken in snowmelt-dominated watersheds in western North America to investigate the influence of forest harvesting on streamflow metrics including maximum annual peak flows (a.k.a. maximum annual flood peak or flood flow) (Bates and Henry, 1928; Van Haveren, 1988; Troendle and King, 1985, 1987; King, 1989; Burton, 1997; Cheng, 1989; Troendle et al. 2001; Moore and Scott, 2005; Moore and Wondzell, 2005; Table 1.1). Most studies listed in Table 1.1 were conducted using the traditional experimental method of  paired watersheds in which the annual peak flows of physically similar, adjacent watersheds are correlated through linear regression during a pre-treatment calibration period and then one watershed is subject to a treatment of forest clearing. Treatment effects are measured using the Before-After-Impact-Control (BACI) method in which the post-treatment observed discharges in the treatment watershed are compared against the un-harvested ?pre-treatment? condition through the use of the pre-treatment linear regression. The predicted ?pre-treatment? peak discharge is compared to the chronologically paired ?observed? post-treatment peak discharge and statistical tests of analysis of variance (ANOVA) or analysis of co-variance (ANCOVA) are applied to determine if the differences are statistically significant.    6 Table 1.1. Summary of paired watershed studies investigating the influence of forest harvesting on streamflow in snow-dominated regions of western North America  Study or Catchment Location Size Treatment Method of Analysis Peak Flow Response Citation Wagon Wheel Gap- Watershed B  Colorado  81 ha  100 % Clearcut  CPa/Change in mean Qpeak Elevated and advanced Bates and Henry, 1928 Avg. pre- and post treatment flow duration curves/ ANCOVA Increased by average of 50%. Large floods not affected Van Haveren, 1988 Deadhorse Creek ? North Fork sub basin Colorado FEF 40 ha 36% cut in small circular openings CP/ANCOVA Increased by average of 50% Troendle and King, 1987     Avg. pre- and post treatment flow duration curves/ ANCOVA Flows increased. Largest not affected Troendle and Olsen, 1994 Fool Creek Colorado FEF 289 ha 40% cut in strips of 1 to 7 tree heights wide CP/ANCOVA Increased by average of 23% Troendle and King, 1985 Fool Creek Colorado FEF 289 ha 40% cut in strips of 1 to 7 tree heights wide CP/ANCOVA Increased by average of 23%. Large floods not affected Moore and Wondzell, 2005 Horse Creek basins 12, 14, 16. 18 Idaho 22 ? 86 ha  25 to 36% Clearcut in patches CP/ANCOVA Average increases from 34% to 87% King,1989 Brownie Creek Utah 2145 Clear cut of 25% of catchment all in upper 1/3 CP/ANCOVA Increased by average of 66% Burton, 1997 Coon Creek Wyoming 1673 ha Clear cut of 24% of catchment Annual flow duration curves /ANCOVA No significant increase Troendle et al., 2001 Camp Creek BC 3390 ha 27% clear cut of catchment area CP/ANOVA 21% increase  Cheng, 1989 CP/ANCOVA No significant increase for larger floods Moore and Scott, 2005 Redfish Creek BC 2600 ha Various Scenarios incld 100% clear cut Percent change in magnitude (%?QT) of flood quantile through comparison of CDFs Significant increase for all floods for >16%  harvest at upper elevations Schnorbus and Alila, 2004 Fool Creek Colorado FEF 289 ha 40% cut in strips of 1 to 7 tree heights wide FP ? direct comparison of CDFs Increases in frequency and magnitdude of floods up to 30yr R.P.  Aila et al., 2009 a  CP = Chronological pairing Treatments in these previous studies ranged from removal of 24% to 100% of the forest cover in the treatment catchment (Table 1.1); however, treatment effects varied considerably between studies from no significant increase in peak flows to increases of up to 87% above predicted flows leading researchers to suggest that no clear correlation exists  7 between percent forest cover removed and peak flow increase (Austin, 1999; MacDonald and Stednick, 2003). In addition, a few studies reported that during years with exceptionally large floods the observed peak flow in the treatment watershed showed no change or actually decreased compared to the chronologically paired pre-treatment control peak flow (VanHaveren, 1988; Troendle and Olsen, 1994; Austin, 1999; Moore and Scott, 2005; Moore and Wondzell, 2005) leading to the perception that forest harvesting has minimal effect on the largest floods (Calder et al. 2007; NCR, 2008).  Recently, in their investigation of forests effects on floods, Alila et al. (2009) revealed that the traditional experimental design used over the past century, that measures treatment effect through chronological flood event pairing (CP) together with the standard BACI analysis (including those studies listed above), is flawed because it does not account for changes in frequency of peak flows following harvesting. In fact, despite the original call by the American Society of Civil Engineers for studies to investigate the effect of forest removal on flood frequency (Hoyt and Troxell, 1932), very few forest hydrology studies undertaken over the past century have invoked the aspect frequency in the investigation of harvesting effects on floods (Lieberman and Hoover, 1951; Berris and Harr, 1987). Interestingly, while the forest hydrology community appeared to be limited to the use of CP-based methods, the investigation of the effects of changing land cover on flood frequency was relatively common-practice in other science and engineering disciplines (e.g., Howe et al., 1966; Knox, 1977; Booth, 1990).  Alila et al. (2009) illustrated using two long term paired watershed data sets from contrasting hydroclimate regimes how investigations of flood response conducted using CP-based analyses actually mask the effects of forest harvesting practices on the larger floods.  8 By investigating forest harvesting effects on floods outside of the analytics of frequency distributions, as the traditional studies listed in Table 1.1 have done, the reported changes in magnitude for events larger and smaller than the mean flood are deceptive (Alila et al. 2009). Even if the outcomes of CP-based studies were correct it would be for the wrong reason (Alila et al., 2010). The results of Alila et al., (2009) re-analysis of the Fool Creek data from Colorado revealed surprising increases in peak flows across a wide range of flood magnitudes (Table 1.1).  1.1.1 The problem with traditional research methods Alila et al. (2009) revealed that it is, in part, the incorrect definition (or indexing) of a ?large? flood, based solely on the ranking of the peak flows in the control catchment, which contributes to misleading study outcomes in CP-based analyses. In CP-based studies, when the magnitude of the largest control catchment floods do not display relative increases in the treatment catchment it is mistakenly concluded that harvesting has no effects on large floods. Alila et al. (2009) illustrated how this line of reasoning is flawed because some small and medium control catchment floods may be amplified enough to become some of the largest flood events on record in the treatment catchment, consequently increasing the frequency, and by association the magnitude, of post-harvest large floods. This adjustment in the ranking (frequency) of flood events caused by harvesting changes what should be designated as a ?large? flood in the treatment watershed. Most of the studies listed in Table 1.1 investigated harvesting effects on floods larger than the mean flood without recognizing this change in frequency, which could not be done because harvesting effects were measured as the difference between the chronologically paired control and treatment catchment flood magnitudes (e.g. Troendle and Olsen, 1994; Moore and Scott, 2005).   9 Floods, as with most hydrological and meteorological variables, are randomly occurring events so that their prediction is probabilistic (or stochastic) rather than deterministic (Yevjevich, 1972); that is, while it is not possible to predict exactly when a given flood of some magnitude will occur, we can predict the likelihood that it will occur in a specific period of time according to its frequency distribution. As a stochastic process, floods are described in part by two inextricably linked attributes: magnitude and frequency. Investigating changes in magnitude without controlling for frequency, as conducted in CP-based studies, leads to an ?apples to oranges? type of comparison. A frequency distribution is the only framework that allows for the investigation of one attribute while controlling the other and is the only correct method of addressing the research question: What is the change in magnitude (frequency) for an event of a specific frequency (magnitude) of interest (Alila et al., 2010)? Another critical flaw of traditional CP-based paired watershed investigations that seek to measure treatment effect by comparing chronologically paired peak flows in control and treatment catchments is that it is assumed that peak flow response in the treatment catchment can be predicted based on peak flows in the control catchment. This premise of a strong deterministic association between the peak flow responses is true for physically identical basins subject to identical meteorological inputs and identical runoff processes and is often used by engineers to fill-in missing peak flow data in two neighbouring gauged catchments (e.g. Dalrymple, 1960). Such premise, however, is no longer valid during the posttreatment period because following harvesting, the hydro-meteorological processes largely responsible for the stochastic nature of the peak flow (Qp) response (Yevjevich, 1972) differ between treatment and control catchments. In snowmelt-dominated regions the  10 difference in pre- and posttreatment runoff generating processes is commonly observed as peak flows that are synchronized during the pre-treatment period but separated in time by days or weeks during the post-treatment period (Van Haveren, 1988; Troendle and Olsen, 1994; Troendle & Stednick, 1999; Troendle et al., 2001; King, 1987; Moore and Scott, 2005; Moore and Wondzell, 2005; Troendle and King, 1985).  The independence of hydro-meteorological processes in treatment and control catchments following logging is often observed as a decrease in the statistical strength of the posttreatment Qp response regression relation. This unexplained variability caused by forest harvesting has often been viewed as a nuisance and suppressed by logarithmic transformations of Qp but only to satisfy the basic assumptions of normality and homogeneity of variance in traditional peak flow investigations by ANOVA and ANCOVA analyses (e.g. Jones and Grant, 1996; Thomas and Megahan, 1998; Jones, 2000).  Alila et al., (2009) stress that the key failing of traditional CP-based research methods is that the regression fit does not account for the simultaneous change in magnitude and frequency, nor does it preserve the all-important relation between these two attributes. Focusing on the flood response to harvesting in terms of a collection of annual differences in flood magnitude as done in CP via a regression equation has detracted from the original and most critical research question, namely: How has harvesting changed the frequency distribution of floods? Alila et al. (2009) also point out that the ability to observe how the full frequency distribution has changed is important because of the inverse and highly non-linear relation between magnitude and frequency that makes small changes in the mean and variability translate into surprisingly large changes in the upper tail of the frequency distribution (i.e. the frequency and magnitude of larger events).  11 1.2 The flood regime and headwater streams  Physical and biological sustainability of headwater streams depends on the natural variability of hydrologic and geomorphic processes that control sediment and nutrient flux as well as the temporal and spatial complexity of aquatic ecosystems (Poff et al., 1997; Meyer and Wallace, 2001; Gomi et al., 2001; Richardson et al., 2005b). In headwater streams where mass wasting is infrequent or absent, disturbance events leading to the episodic influx of sediment and nutrients are closely linked to a flood regime comprising a range of flood magnitudes and durations reflective of regional hydro-climate and physiographic controls (Resh et al., 1988; Poff et al., 1997). High predictability of flow variability in snowmelt-dominated streams make British Columbia?s interior fluvial systems particularly high-value aquatic ecosystems (Poff and Ward, 1989). Annual spring snowmelt floods constitute the main flood event of interior hydro-climate regions (Church, 1988). In most inland headwater streams flows begin to rise in early April as snowmelt is transferred to from hillsides to stream networks and remain elevated throughout May and June receding back to low flows during late June or early July. Snowmelt hydrographs of interior headwater streams typically display several independent peak flows associated with isolated periods of exceptional warm spring temperatures. The largest peak flow of the spring snowmelt period corresponds to the annual maximum flood peak. The long-term time series of annual maximum peak flows is characterized by a range of flood magnitudes that have a probability of occurrence described by a cumulative frequency distribution (a.k.a. flood frequency curve).   12 1.2.1 Channel maintaining flows Alluvial rivers, flowing through contemporary alluvial deposits display geometries adjusted to the long-term frequency distribution of floods (Leopold and Maddock, 1953; Andrews and Nankerviz, 1995; Dodov and Foufoula-Georgiou, 2005). The average annual flood, that occurs approximately once every 1.5 to 3 years (avg. 2 yr return period), is referred to as the ?effective? discharge that mobilizes the majority of bedload sediment and maintains channel morphology over the long-term (Wolman and Miller, 1960; Wolman and Gerson, 1978; Andrews, 1980; Whiting et al., 1999; Emmett and Wolman, 2001). In self-formed alluvial rivers effective discharge corresponds to the bankfull flood which just fills the active channel (Andrew, 1980). Similar agreement between effective discharge for bed load transport and bankfull flood with return periods ranging from 1.5 to 3 years has also been shown to apply to smaller, alluvial headwater streams despite the large variability in gradient, substrate texture and the complex nature of bed mobility (Whiting et. al., 1999; Ryan and Emmett, 2001; Torrizo and Pitlick, 2003; Ryan et al., 2005).  In forested headwater streams, channel-spanning pieces or jams of large woody debris (LWD) incorporated into the channel from the adjacent riparian vegetation create steps in the longitudinal channel profile that retain alluvium (Montgomery et al., 1996). Alluvial channel morphologies associated with the LWD steps include forced step-pool and forced riffle-pool channel types that are typically super-imposed on steeper gradient plane-bed, cascade or even bedrock morphologies (Montgomery and Buffington, 1997; Massong and Montgomery, 2000). In British Columbia, forested headwater streams are the standard rather than the exception so forced alluvial morphologies constitute an important and prevalent channel type within headwater systems. LWD steps and jams have been shown to  13 impart controls on flow hydraulics and bed mobility (Bilby, 1981; Heede, 1985; Curran and Wohl, 2003; Comiti et al., 2009; David et al., 2010), however very little research has been undertaken to determine if the morphology (i.e. hydraulic geometry, gradient, bed texture) of these forced alluvial streams is calibrated to the average annual or bankfull flood (Gomi and Sidle, 2003; Hassan et al. 2005; Brayshaw, 2011).  1.2.2 Channel forming floods Channel maintaining floods that occupy the bankfull channel are responsible for the majority of sediment transport over the long-term (Wolman and Miller, 1960; Schmidt and Potyondy, 2004), whereas, less frequent channel forming floods that exceed the bankfull geometry are the primary disturbance event responsible for eroding channel boundaries, altering channel morphology and mobilizing large volumes of sediment and nutrients to downstream reaches (Ashworth and Ferguson, 1989; Warburton, 1992; Lenzi et al., 1999; Church, 2006). In steep terrain debris flows impart episodic disturbance to headwater channels; however, in lower relief uplands, typical of much of the interior region of BC, episodic disturbance to stream channels is closely linked to flood events (e.g. Comiti and Mao, 2012). Catchment scale acts as a first order control on the frequency of channel forming floods (Gupta and Waymire, 1998; Church, 2002; Dodov and Foufoula-Georgiou, 2005). In large alluvial rivers, frequent (R.P.?< 5 yr), small to moderate magnitude floods are capable of eroding banks, occupying floodplains and imparting channel change whereas, in small headwater streams, floods capable of mobilizing large keystone and woody debris structures and altering channel form have a low frequency of occurrence (R.P. > 25 yrs to +200 yrs) (Wolman and Gerson, 1978; Church, 2002; Grant et al. 1990, Lenzi et al., 1999; 2006). The capacity of a flood to effect channel change depends on both the magnitude and  14 the duration of a flood (Costa and O?Connor, 1995, Huckleberry, 1994). Large magnitude short duration floods have been observed to impart little channel change in alluvial gravel-bed systems (Costa and O?Connor, 1995). The effect of changing land-cover on the magnitude and duration of floods has been investigated in rain-dominated regions of North America and Britain. Increases in both the magnitude and duration of floods have been documented following urbanization and forest land to pasture conversion in rainfall-dominated hydroclimate regimes (Konrad et al; 2005; Archer et al., 2010). Such response is less well documented in snowmelt hydroclimate regimes (Troendle and Olsen, 1994; Troendle et al., 2001; Gaeuman et al. 2005). 1.3 Effects of flood regime changes on headwater streams In alluvial rivers, changes in the flood regime have the potential to cause long-term changes to sediment transport dynamics, sediment yield and channel morphology (Knox, 1977; Rumsby and Macklin, 1994; Gordon and Meetenmeyer, 2006; Kiss and Blanka, 2012). Changes in channel cross-sectional area of low-land alluvial streams have been linked to increases in both the frequency and magnitude of floods following the conversion of forests to pastures or urban areas (Booth, 1990; Fitzpatrick et al., 1999). In upland headwater streams, fining of surface textures has been documented as a result of decreases in the magnitude of floods associated with flood water extraction (Parker et al., 2003; Gordon and Meetenmeyer, 2006). Two studies investigating the influence of forest removal-related changes in flood response on sediment yield in snowmelt dominated headwater streams determined that increases in sediment yield can be linked to increases in the magnitude and duration of sediment transporting discharges following harvesting (Troendle and Olsen, 1994; Troendle et al. 2001).   15 Predicting the potential for headwater channel response to changes in the flood regime or sediment supply has received some attention in recent decades (Montgomery and Buffington, 1997, Buffington et al., 2003; Grant et al., 2008). The capacity of a channel to be altered by changing conditions of discharge or sediment supply depends on the degree to which alluvial processes shape channel form and function. Headwater streams containing coarse angular colluvium from mass wasting inputs are unlikely to respond to flood regime changes because streamflow alone is not competent to mobilize channel bed material. Empirically-based competence equations describing the ability of the streamflow to mobilize the available bed material (Gilbert, 1914, Baker and Ritter, 1975; Costa, 1983) are often expressed as: C=aDib          (1-1) Where competence (C) represents the critical average velocity, shear stress or stream power necessary to mobilize a grain of size Di and the coefficient (a) and exponent (b) vary by stream channel. Alternatively, channel competence has been expressed as the balance between hydraulic driving forces (i.e. shear stress, ?) and resisting forces or grain inertia ((?s-?)gD) where ?s and ? are the densities of sediment and fluid respectively and g is acceleration due to gravity; ?? = ???(?? ? ?)??         (1-2) In this equation (Shields equation) ?c is the critical shear stress required to mobilize a grain of diameter D and ??? is a dimensionless value referred to as the dimensionless critical shear stress or Shields parameter (also commonly denoted ?c).   16 As an approach to predicting channel response to changes in hydraulic or sediment loading in alluvial headwater streams Buffington et al., (2003) determined that state diagrams originally developed to represent relationships between driving forces, sediment mobility and channel characteristics in gravel-bed rivers (Parker, 1990), predict reasonably well alluvial channel response in terms of adjustments in channel gradient, bed texture and bedload transport rate.  Figure 1.1 From Buffington et al., 2003 modified from Parker, 1990. Vector illustrates the direction of channel response for increases in discharge without increases in sediment supply. In Buffington et al.?s (2003) modified state diagram, different channel morphologies (states) occupy distinct sub-parallel regions in dimensionless equilibrium bedload transport rate (qb*) versus dimensionless discharge (q*) space (Figure 1.1). Figure 1-1 indicates that, in equilibrium channels, increases in discharge with constant sediment supply (i.e. transport capacity Qc in excess of sediment supply Qs) will lead to channel degradation (i.e. increased relative submergence h*= h/D84s) and decreases in gradient (S). In terms of morphological adjustments the state diagram indicates that with increasing discharge a cascade reach is  17 likely to adjust into a step-pool reach. The development of an increasingly structured stepped-channel with increasing discharge has been observed in flume studies (Whittaker and Jaeggi, 1982). However, the model also predicts that additional increases in discharge would cause further degradation and channel gradient decreases causing the step-pool channel to transition to a plane bed and eventually to a riffle pool channel. Such adjustments are unlikely in headwater systems because they require that the channel be fully alluvial and able to adjust bed gradient. Typically the channel gradient of most headwater systems, particularly those in previously glaciated regions, is nowadays still imposed by past glacial erosion and paraglacial deposition (Brardinoni and Hassan, 2006; 2007).  An additional failing of the state-diagram predictive approach for headwater channels is that it assumes a constant dimensionless shear stress (??) for bedload mobility across morphologies whereas recent studies have determined that ?? increases with increasing channel gradient (Lamb et al., 2008). For this reason the application of a regime-type response model to predict morphological adjustments to changes in discharge are unlikely to apply to alluvial and semi-alluvial headwater streams which display a large range of glacially-imposed channel gradients. 1.4 Thesis objectives and structure  The research presented in this thesis was undertaken with the objective of an improved understanding of forest harvesting effects in formerly glaciated, snowmelt headwater streams characteristic of BC?s inter-montane physiographic regions where mass wasting events are infrequent. To this end the following research questions form the basis of the dissertation;   18 ? How does forest harvesting affect the flood regime of snowmelt-dominated headwater streams?  ? What processes dictate channel form and sediment transport dynamics in fluvial headwater channels? ? How are sediment yield and channel morphology in forested snowmelt-dominated headwater steams influenced by changes to the flood regime?  This dissertation consists of seven additional chapters. Chapter 2 provides details regarding study design for the investigation of (1) flood regime response to forest harvesting and (2) fluvial headwater channel morphodynamics. Chapter 3 present the findings of forest harvesting influences on flood magnitude, and frequency. The effect of harvesting on flood duration and number of peaks over threshold is presented in Chapter 4. Chapter 5 presents findings with respect to hydrological and physiographic controls on channel morphometrics, hydraulic geometry and sediment yield in fluvial, formerly glaciated headwater streams. Chapter 6 presents the findings of a detailed investigation of sediment transport dynamics of fluvial headwater streams. The potential for channel response in headwater channels associated with altered flow regimes due to forest harvesting is explored in Chapter 7. Specifically, Chapter 7 links the investigation of forest harvesting effects on snowmelt flood regime detailed in Chapters 3 and 4 with the potential for changes in channel form, sediment yield and bedload transport from Chapters 5 and 6. Chapter 8 provides a synthesis of study findings and conclusions regarding channel response to forest harvesting in formerly glaciated, fluvial headwater streams, addresses study limitations and provides direction for future research.   19 2 Study sites and experimental design 2.1 Forest harvesting effects on the flood regime   Figure 2.1. Location and topography of study areas for Chapters 3 and 4. The investigation of forest harvesting effects on the flood regime utilizes a meta-analysis approach that employs the simultaneous examination of daily and annual time series of discharges from physically different catchments to expose similarities and/or differences in catchment response in terms of the frequency and magnitude of annual peak flows, the duration of individual flood peaks and the number of peaks over a geomorphically  20 relevant threshold discharge. These similarities or differences can be used to reveal the influence of physical basin characteristics on flood response (Jones, 2005; McDonnell et al., 2007). The four study sites selected for the meta-analysis are typical of interior continental snow environments where flood hydrology is dominated by annual snow melt and where rainfall plays a much less important role either during the freshet or in the remainder of the year.  Table 2.1. Meta-analysis catchment characteristics Basin Characteristics Modeled Catchments Observed Catchments  Redfish(100U) 240 (40T) Fool Camp Yrs of post treatment data 99 96 48 19 Size (km2) 25 5 3 37 Avg. slope (%) 50 24 23 20 Elevation Range (m) 700 to 2300 1600 to 2000 2896 to 3505 1900 to 1050 % Alpine (+ open sub-alpine) 40 0 23 0 Aspect Distribution* E=W/S E/W/S N/NW/NE S/SE/E/W Stand composition* Sp/Bf/Lp/Cw Lp Sp/Bf/Lp Lp/Df/Sp/Bf Crown Closure (%) 50-80 44(avg.) >50% 55-70 Harvest area (%) 33 40 40 37 Elevation of harvesting(m) 1520 to 1880 1750 to 1950 2950 to 3300 1200 to 1700 Aspect dist. of harvesting* E=W/S E/W/S NE/NW/N SE/S/E/W Average date of Qp June 17 May 12 June 15 May 20 * listed in order of abundance Sp = spruce, Bf = balsam fir, Lp = lodgepole pine, Cw =western Cedar, Df= Douglas fir  The four snowmelt-only headwater catchments in our meta-analysis include two large (25 and 37 km2) and two small (<5 km2) basins (Figure 2.1). For contrast, both large and small basin pairs include one basin that contains a component of alpine area (Fool and Redfish Creeks) and one fully forested basin (Camp and 240 Creeks). Two of the hydrometric data sets are from long-term paired watersheds (Camp and Fool Creeks), and two are simulated data sets (240 and Redfish Creeks). The use of simulated flows is appealing because of the length of years of simulated discharges (95 and 99 yrs, respectively) for the modeled catchments compared to 19 and 48 years of posttreatment  21 daily hydrometric data for Camp and Fool Creeks and because simulated discharges represent static land cover conditions without forest regeneration that can obscure treatment effects on larger floods. The use of simulated discharges allows us to estimate changes in frequency and magnitude of flood events during the most critical period after logging before any substantial recovery has occurred. Simulated discharge for 240 and Redfish Creeks was made available by Y. Alila, (UBC, Forestry).Hydrometric data for Fool and East St. Louis Creeks was downloaded from a website of archived USFS hydrometric data. Data for Camp and Greata Creeks was downloaded from Environment Canada archived hydrometric data website.  The paired watersheds have been subject to moderate levels of harvest accounting for 38 and 40% of Camp and Fool Creek watersheds, respectively (Table 2.1). In both cases harvesting is situated at the mid-elevations of the watershed and conventional ground skidding methods were used to transport the logged timber to haul roads. In Fool Creek most of the roads used to access cut blocks were deactivated following harvesting to re-establish natural drainage patterns. An extensive network of roads and old skid trails is apparent on air photographs of Camp Creek, however, it is not known if any of these roads or trails were deactivated to restore natural drainage patterns. Harvesting scenarios selected for the modeled catchments were chosen to most closely resemble the level of harvest and location of openings in the observed paired watersheds (Table 2.1). Modeled catchments do not include the effect of roads on hydrological response.  22 2.1.1 Fool Creek  Fool Creek is a 289 ha, north flowing treatment drainage in a paired watershed experiment at the Fraser Experimental Forest (FEF), located about 100 km northwest of Denver, Colorado. The watershed is characterized by moderate gradient slopes with predominantly northern aspects, ranging in elevation from 2896 to 3505 m (Table 2.1). Mean annual precipitation at the site is around 595 mm, 60?80% of which occurs as snow. The geology of the watershed is metamorphic, consisting of schist and gneiss derived from granite, subjected in the past to extensive glaciation. Soils are gravelly except for deep alluvial soils adjacent to stream courses. Vegetation in the watershed consists of a dense mature stand (aged 250?350 years) of lodgepole pine (Pinus contorta), Engelmann spruce (Picea engelmannii), and subalpine fir (Abies lasiocarpa). The upper 25% of the watershed consists of alpine terrain and open krumholtz forest (Troendle and Kaufmann, 1987).  Calibration of the Fool Creek watershed began in 1943 and ended in 1954. The contiguous East St. Louis watershed (803 ha) was used as a control. Logging began in 1954 and was completed in 1956. The harvesting pattern consisted of alternating cut and leave strips of varying width (one, two, three, and six chains, where a chain equals 20.12 m) running normal to contours between 2950 and 3300 m, with 40% of the watershed harvested (50% of the timbered area) and the forest left to regenerate naturally. Spur roads (14.2 km) built along contours were decommissioned after logging; culverts were removed on alternate roads and all roads were grass seeded. The main haul road (5.3 km), however, is still open and subject to regular maintenance (Alexander and Watkins, 1977). For more details, the reader is referred to Goodell (1958), Alexander and Watkins (1977), and Troendle and King (1985).  23 2.1.2 Camp Creek Camp Creek (37 km2) is located along the western side of the Okanagan Valley approximately 20 km west of Peachland, British Columbia (BC). Camp Creek together with Greata Creek (41 km2) compose an opportunistic treatment?control catchment pair with long term streamflow gauging by Environment Canada. Mean annual precipitation measured 12 km north of Camp Creek at the Brenda Mine climate station (1500 m asl) is approximately 600 mm, 60% of which falls as snow between the months of November and March. Both catchments display dominantly south slope aspects and similar average slope gradients of 20% (Table 2.1). Elevation in Camp Creek ranges from 1900 to 1050 m at the hydrometric gauging site. Both catchments are forested to the headwaters and are underlain by coarse textured bedrock and glacial deposits. Forest cover varies from Lodgepole Pine (Pinus contorta Dougl) leading stands with lesser Douglas fir (Pseudotsuga menziesii) at lower elevations to mixed Douglas fir and Lodgepole pine stands at intermediate elevations and spruce (Picea Engelmannii Parry) and sub-alpine fir (Abies lasiocarpa) at higher elevations.  Concurrent daily discharge gauging of Greata and Camp Creeks began in 1971. Logging commenced in Camp in 1976 in response to a Mountain Pine Beetle (MPB) outbreak and by the end of 1977 over 29% of the 37 km2 watershed had been harvested. A number of blocks accounting for an additional 8% of the watershed were harvested between 1977 and 1990. By 1991 a total of 37% of the Camp Creek watershed was in a clear cut state. Most of the harvesting occurred in stands consisting of Lodgepole pine with lesser amounts of spruce and Douglas fir at elevations between 1300 and 1700 m. Digital forest inventory information available from the B.C. Ministry of Forests (http://geobc.gov.bc.ca) indicates that the blocks harvested in 1977 contain regenerating juvenile pine stands  24 averaging 22 years in age and ranging in height from 4.4 to 7 m (5.5 m avg.). Crown closure in these regenerating stands averages 35%. In contrast, the un-harvested mature forest in Camp Creek consists of greater than 100-year-old Lodgepole pine (120 year avg.), balsam fir and Douglas fir stands averaging 25 m in height and 55 to 70% crown closure (Table 2). A small amount of logging, less than approximately 9% of the 41 km2 area has occurred over the past 30 years at the western headwaters of Greata Creek control catchment. The logging in Greata Creek is thought to have occurred in part before 1980 with the remainder occurring sometime before 1990 (Moore and Scott, 2005). Although it may affect study outcomes, this small amount of logging in Greata Creek is ignored. If anything, it would cause an underestimation of the predicted effects of forest harvesting on peak flows (Moore and Scott, 2005). It is not known if the numerous roads and skid trails constructed to access the cut blocks in Camp Creek have been deactivated to restore natural drainage patterns.  2.1.3 Redfish Creek  Redfish Creek is a 26 km2 catchment located in the Selkirk Mountains of BC, approximately 20 km northeast of Nelson. Redfish Creek ranges in elevation from 700 to 2300 m. Mean annual precipitation is estimated at between 1400 and 1800 mm with precipitation occurring throughout the year but falling as snow from October to May. Basin slopes are moderately steep, with a median gradient of 50% and primarily east and west aspects (Table 2.1). Bedrock underlying Redfish Creek is dominantly coarse crystalline granodiorite. Soils are derived from rapidly drained sandy gravelly glacial till and colluvium. Slopes below approximately 2000 m are densely forested with a mixed coniferous stand including western redcedar (Thuja plicata), western hemlock (Tsuga heterophylla), Douglas fir (Pseudotsuga menziesii), Lodgepole pine (Pinus contorta), spruce  25 (Picea glauca x engelmannii), and subalpine fir (Abies lasiocarpa). Above an approximate elevation of 2000 m the forest stands transition into subalpine parkland, which is a sparsely vegetated subzone that occupies approximately 40% of basin area and does not hold any operable forest. The Distributed Hydrology, Soils, Vegetation Model (DHSVM) has been used by Schnorbus and Alila (2004) to generate 99 years of simulated flows for a fully forested control scenario plus 9 other harvest scenarios in Redfish Creek. Only two of the 10 (Control and 100U) simulated scenarios of Schnorbus and Alila (2004) were used in this study. For the Control scenario the watershed is fully forested which required trees to be re-established over approximately 250 ha (10% of the watershed) on lower and mid-elevation slopes. For the 100U scenario, all merchantable forest is removed from the middle and upper forested slopes between the elevations of 1520 and 1880 m, which accounts for roughly 33% of the total watershed. For a detailed description of forest cover distribution and harvest scenarios the reader is referred to Schnorbus and Alila (2004).  2.1.4 240 Creek  240 Creek is a 5 km2 fully forested catchment located in the Okanagan Highlands roughly 25 km northeast of Penticton, BC. 240 Creek is one of three heavily instrumented catchments in the Upper Penticton Creek Experimental Watershed (UPC) that is maintained by the BC Ministry of Forests. Mean annual precipitation is 750 mm, of which about half falls as snow from November to April. Elevation ranges from 1600 to 2000 m (Table 2). Slopes are of moderate gradient and have predominantly east and west aspects (Table 2.1). The catchment is underlain by coarse crystalline granodiorite and metamorphic orthogneiss  26 overlain by veneers of sandy soil derived from glacial till and colluvium. 240 Creek has a relatively open forest canopy containing predominantly mature Lodgepole pine (Pinus contorta) with small amounts of Engelmann spruce (Picea Engelmannii) and sub-alpine fir (Abies lasiocarpa) (Thyer et al., 2004).  DHSVM has been used by Schnorbus and Alila (2013) to generate 95 years of simulated flows for a fully forested control scenario plus 11 other different harvest scenarios in 240 Creek. Only two of the 12 (Control and 40T) simulated scenarios of Schnorbus and Alila (2013) were used in this study. In the 40T scenario, 40% of the watershed is removed uniformly in a band that extends across the mid and upper slopes between the elevations of 1750 and 1950 m. For the Control scenario forest cover exists over the full catchment area. For a detailed description of forest cover distribution and harvest scenarios the reader is referred to Schnorbus and Alila (2013). 2.1.5 Simulation of harvesting scenarios by DHSVM DHSVM is a spatially distributed model that approximates catchment-scale precipitation inputs and outputs through multi-parameter algorithms that account for changes in the energy and water balance associated with forest cover removal (Wigmosta et al., 1994). As with all physically distributed models, DHSVM may introduce uncertainties in the estimation of large peak flows due to the low number of extreme events represented during the calibration period. However, in both Redfish and 240 Creeks rigorous testing and calibration of DHSVM included reproducing multiple years of observed flows at the basin outlets as well as numerous internal catchment processes and meteorological variables. At Redfish, DHSVM was initially calibrated and tested by Whitaker et al. (2003). Model  27 calibration incorporated an extensive network of field based meteorological and hydrological parameters. A multi-pass approach to calibration, in which potentially correlated parameters were excluded from the same calibration exercise, reduced the potential for equifinality in model output. Whitaker et al. (2003) found that the calibrated model provided good approximation of measured hydro-meteorological parameters including streamflow, and snow accumulation and melt at multiple sites throughout the catchment. Details regarding the calibration and parameterization of the DHSVM model for Redfish Creek are presented in Whitaker et al. (2003). At UPC (240 & 241 Creeks), DHSVM was also extensively calibrated and evaluated by Thyer et al. (2004) who found that the model successfully simulated streamflows as well as other spatially distributed hydro-meteorological parameters including forest and clearcut SWE, tree transpiration and clearcut snowmelt rates. Further testing and validation of the UPC DHSVM model output was undertaken by Kura? et al. (2011) who determined that the model realistically simulated the spatiotemporal variability of road and stream network flows, and subsurface responses in the watershed. The DHSVM applications at Redfish and 240 Creeks are believed to be reliable enough tools for contributing to the ongoing debate on the effects of forest harvesting on the peak flow regimes of snow-dominated watersheds (Schnorbus and Alila, 2004; Kura? et al., 2012) 2.2 Forested snowmelt headwater stream morphodynamics  2.2.1 Geology and physiography A detailed investigation of channel morphometry and bedload sediment mobility was undertaken on Cotton Creek (22.1 km2, Figure 2.2), a forested, mountain basin of the Purcell  28 Range, Columbia Mountains physiographic region, British Columbia (Holland, 1964) between spring 2005 and fall of 2009.  Elevation in Cotton Creek ranges from 900 m a.s.l. in proximity of Moyie Lake, to 2018 m along the south-eastern divide of the basin. Channel gradient in the basin, ranging between 3 and 40 percent, does not support debris-flow initiation. Our observations are consistent with findings from prior work conducted in neighboring Selkirk and Monashee Ranges of the Columbia Mountains, according to which landslide and debris-flow initiation is rare on slopes less than 47 percent (Jordan, 2002).  The drainage network has a dendritic pattern consisting of two main tributaries: Elk Creek (5.7 km2) and Upper Cotton Creek (10.3 km2). The former, in comparison to Upper Cotton Creek, is characterized by lower elevation (Figure 2.3a), and lower slope (Figure 2.3b). With respect to aspect, Elk exhibits a comparatively greater proportion of terrain facing south (i.e., S, SE, and SW; Figure 2.3c) at the expense of north-facing topography (i.e., N and NE). Dominantly south-facing slopes imply higher incoming solar radiation, hence faster snowmelt runoff. As a result, a substantial portion of Elk Creek basin is snow free at least two weeks earlier than Upper Cotton Creek (Jost et al., 2007). The area is underlain by middle Proterozoic meta-sedimentary rocks of the Purcell Group. Silty-sandy, compact, matrix-supported basal till is the dominant surficial material in Cotton Creek and blankets much of the catchment from the lower elevations up to the ridge crests (Figure 2.2b). More loosely consolidated, gravelly ablation till is present as veneers and blankets on the northern slopes of the watershed above about 1500m. Sandy, gravelly  29 glaciofluvial deposits with relict meltwater channels occur in association with the ablation till.   Figure 2.2.(A) Location of Cotton Creek study area with reach breaks (red) and monitoring locations (white/blue) and (B) distribution of surficial material.  30 At the peak of the last glacial period Cotton Creek was overridden by the Cordilleran Ice Sheet that extended south-westward into the Moyie Valley from the Rocky Mountain Trench (Clague et al., 1980). Relict glacial topography suggests that shallow valley-side glaciers descended westward and northward coalescing in Cotton Creek valley before flowing into the main Moyie Valley glacier. Surficial material mapping (Figure 2.2b) indicates that tributaries of both Upper Cotton and Elk contained pro-glacial deposits (eg. glaciofluvial deposits, Figure 2.2b) suggesting that these drainages were ice-free while the surface of the Cordilleran Ice Sheet occupying the Rocky Mountain Trench was still above the highest ridge tops (Ryder, 1981; Jackson and Clague, 1991; Ryder et al., 1991). Glaciofluvial and ablation deposits have been reworked and deposited as paraglacial fluvial deposits (Figure 2.2b) along the length of Upper Cotton and Cotton Creek (below Elk confluence). A small kame terrace (identified as glaciofluvial, Figure 2.2b) along the lower reaches of Cotton Creek likely marks the elevation of the ice surface in the Moyie Valley directly to the west that persisted while pro-glacial meltwater streams were flowing out of the tributary valleys.  Today the region has a continental climate. Precipitation (600 mm annually) falls mainly as snow between October and March. An average maximum snow accumulation of 400mm (snow water equivalent), measured on April 1st since 1971, is reported at the Moyie Mountain snow pillow (1840 m a.s.l; BC Ministry of Environment) located 7 km south of Cotton Creek. Stream flows in the southern Purcell Mountains rise rapidly in mid-April to mid-May in response to solar-radiation driven snowmelt and remain elevated above base flows into mid-July. Typically there are between two to five independent discharge peaks  31 lasting from 2 to 7 days during the spring snowmelt freshet in response to periods of warm, sunny weather.  Figure 2.3. Distribution of (a) elevation, (b) slope gradient, and (c) aspect for Upper Cotton and Elk. Vegetation within the watershed consists of 60 to 70 year old naturally regenerated mixed conifer stands). Starting from the mid 1990's, timber harvesting has affected  32 respectively 34% and 20% of lower-to-mid slopes in Elk Creek and Upper Cotton Creek. Riparian buffers along the drainage network have protected the channel banks and adjacent flood plain from logging disturbance, so that anthropogenic effects on sediment and wood delivery to streams are minimal. 2.2.2 Monitoring site characteristics Bedload and water discharge were monitored through three to four freshets (2005-2008) at respectively eight and nine stations arranged in a nested set up (Figure 2.2a). The bedload monitoring stations enclose sub-basins ranging from 2.2 to 10 km2 (Table 2.2) and include sites E1 to E4 in Elk Creek and Sites C1 to C4 in Cotton Creek. Sites C1 and E1 (Figure 2.2) were accessed less frequently and therefore monitored only for bedload yield (Table 2.2).  Table 2.2. Monitoring site sampling details Site Drainage area  (km2) Channel type Monitoring frequency a Trap typeb Ch. width monitored  (%) Monitoring Period   SY Q X-sec E1 2.2 Colluvial A C A 1  100 2005-08 E2 3.5 Forced SP B C B 2 45 2006-08 E3 5.0 B-cascade B B B 2 45 2006-08 E4 5.7 Step pool A, B C B 3  100 2005-08 C1 3.0 Colluvial A C A 1  100 2006-08 C2 4.0 Forced SP B B B 2 33 2006-08 C3 8.9 Forced SP B B B 2 28 2006-08 C4 10.3 Riffle pool A, B C B 3  100 2005-08 C5 18.5 Riffle pool na C B na na 2006-08 a A ? Once or twice during the freshet, B ? Hourly during the main peak flow period to daily during  intermediate flows and up to weekly during low flows. C ? Continuous b 1 ? Channel spanning weir, 2 ? Removable pit trap, 3 ? Channel spanning pit trap The eight bedload monitoring sites include fully alluvial, semi-alluvial and colluvial channel types (Table 2.3, Halwas and Church, 2002). The physical and hydrological  33 characteristics of the eight bedload monitoring sites is summarized in Table 2.3. Of the two streams, Elk Creek displays substantially coarser bed textures and much lower peak discharges compared to Cotton Creek (Table 2.3) Table 2.3 Physical and hydrological characteristics of bedload monitoring sites. Site Channel type Surface a grain size S Rbf b ?obf ?*c50s Qbf Qmaxc Qid SSYe ?if Max ?o/?ig Breakpointh   mm m/m m (Pa)  m3/s m3/s %Qbf kg/ day  Pa  Qth R2 Qbf   D20 D50 D90        km2   (m3/s)  % E1 Colluvial 3 10 37 0.128 0.120 194.3 1.2 0.15 na na 1.00 na na 0.09 Est. 60 E2 Forced SP 34 65 115 0.048 0.183 89.3 0.089 0.28 0.29 27 0.90 84.2 1.4 0.17 0.49 61 E3 B-Cscd (FS) 42 82 128 0.068 0.204 136.3 0.092 0.3 0.63 38 1.65 112.8 1.4 0.23 0.5 77 E4 Step Pool 22 53 190 0.047 0.238 110 0.102 0.37 0.63 31 1.20 61.9 1.7 0.28 0.61 76 C1 Colluvial 16 30 60 0.130 0.131 204.2 0.17 0.31 na na 0.19 na na 0.25 Est. 80 C2 Fcd SP (BC) 4 11 30 0.077 0.198 213.8 0.702 0.52 0.86 48 0.52 140.5 1.6 0.45 0.33 80 C3 Fcd SP (BC) 10 31 80 0.101 0.261 257.8 0.498 0.82 1.33 35 0.86 210.4 1.6 0.55 0.22 67 C4 Riffle Pool 13 34 60 0.02 0.253 36.1 0.036 0.86 1.37 24 0.88 13.1 3.0 0.5 0.59 58 a From bulk sample analysis; b Hydraulic radius at bankfull flow (R = A/P, where A is channel cross-sectional area and P is  the relevant wetted perimeter); c  Maximum peak flow during monitoring period; d Discharge at the onset of bedload transport; e Specific bedload yield for the three years of monitoring (2006-2008) in kg/day/km2 f Initial shear for mobilization of bedload g Ratio of maximum shear recorded during monitoring to initial shear h Threshold discharge defined through stepwise regression analysis delineates the   break between phase-1 and phase 2 transport. Bold R2 indicate stepwise regression p-value  <0.05. Alluvial sites including step-pools (E4) and riffle-pools (C4) (Montgomery and Buffington, 1997) characterize the lowest monitoring sites on Elk and Cotton Creek respectively. The presence of large immobile clasts (i.e. immobile during average annual floods) and abundant LWD at the intermediate four sites (E3, E2, C3, C2) create semi-alluvial channels that display woody-debris forced step-pool (FSP) and boulder-cascade (BC) morphologies. The three forced step-pool (FSP) sites (E2, C3, C2) display substantial  34 variability with respect to channel gradient and bed texture. In particular, the forced step-pool site on Elk Creek (E2), which has a coarser median grain size (D50s) and gentler channel gradient (S) than the two sites on Cotton (Table 2.3) displays a plane-bed morphology in intervening channel segments (i.e. between forced step-pool units). Whereas an abundance of immobile boulders at the steeper gradient Cotton Creek sites (C2, C3) contributes to boulder cascade morphologies within intervening channel segments. A hybrid morphology also exists at the boulder-cascade (BC) site on Elk (E3) where alluvial boulder-cascade segments occur together with immobile glacial lag boulders and woody-debris forced step units. The upper two sites (C1 and E1) are classified as colluvial, however, they are distinct from the colluvial analogues typical of steepland areas (e.g., Benda and Dunne, 1997; Jakob et al., 2005) in that they display limited indicators of rapid mass wasting disturbance. Instead they are characterized by moderate slopes (i.e. < 40%) where sediment supply is dominated by soil creep, local bank instabilities, and tree throw, with no evidence of debris-flow activity. Photographic examples of channel types at monitoring sites on Elk and Cotton Creeks are provided in Figure 2.4. Plan-view sketches and longitudinal profiles created from channel surveys and photo-mosaics for 15 to 20 bankfull widths upstream from the three lower sampling sites on both streams are illustrated in Figure 2.5  35  Figure 2.4. Example of channel types present in Cotton (A-D) and Elk (E-H) Creeks. A) colluvial, B) forced step pool, C) riffle pool, D) boulder cascade, E) colluvial, F) forced step pool, G) boulder cascade, H) step-pool.  Channel bed texture was assessed at the eight monitoring sites using bulk sample analysis (Figure 2.6). Bulk samples of surface and subsurface bed material ranging from 20 kg to 90 kg (dry weight) were collected at representative locations in the vicinity of each monitoring site (Church et al., 1987; Kondolf et al., 2003)  36  Figure 2.5. Plan-view and longitudinal channel sketches for 15 to 20 meter reaches up-stream from bedload monitoring sites C2 to C4 on Cotton (A to C) and E2 to E4 on Elk (D to F)  37  Figure 2.6. Cumulative grain size distribution of channel bed surface (s) and sub-surface (ss) at study sites    38 3 Forest harvesting effects on frequency and magnitude of snowmelt floods 3.1 Introduction Land cover changes that alter the frequency and magnitude of floods capable of modifying alluvial channels or occupying floodplains represent a high risk negative impacts to humans dependent on surface water supplies, engineered structures and the aquatic ecosystem. A response by Calder et al. (2007) published in Nature summarizing the current state of knowledge in the study of forest hydrology reports that ?Now forest hydrologists generally agree that, although forests mitigate floods at the local scale and for small to medium-sized flood events, there is no evidence of significant benefit at larger scales and for larger events.? The lack of an influence of forest harvesting on large floods at all spatial scales is based on the outcomes of nearly a century of paired watershed studies. Consequently it has become and established precept in forest hydrology forming the basis for land management policy (CIFOR and FAO, 2005; NRC, 2008; Bathurst et al., 2009) and being taught to students in forest hydrology textbooks (e.g. Jeffrey, 1970; Lee, 1981; Brooks et al., 2003; Calder, 2005; Chang, 2006). In the early 1900?s the American Society of Civil Engineers Special Committee on Floods and Flood Prevention rejected the opinion of many of its members that forests reduce the frequency and severity of large floods in its final report commissioned by the US Government on the practical benefits of reforestation because of the lack of quantitative data (Hoyt and Troxell, 1932; Dobbs, 1969). In response to an obvious need for scientific studies the first experimental watershed in North America, Wagon Wheel Gap, Colorado, was  39 established in 1910 specifically to address the lack of quantitative data on the influence of forests on the frequency and magnitude of floods. Since the early 1900?s at nearly a dozen subsequent studies have been undertaken in snowmelt-dominated watersheds in western North America to investigate the influence of forest harvesting on streamflow metrics including maximum annual peak flows (a.k.a. maximum annual flood peak or flood flow) (Bates and Henry, 1928; Van Haveren, 1988; Troendle and King, 1985, 1987; King, 1989; Burton, 1997; Cheng, 1989; Troendle et al. 2001; Schnorbus and Alila, 2004; Moore and Scott, 2005; Moore and Wondzell, 2005; Alila et al., 2009, Table 1.1).  Despite the original call to investigate the influence of forest removal on flood frequency and magnitude, the research questions of most of these past studies overlooked the dimension of flood frequency and focussed only on quantifying a change in magnitude between pre-harvest and post-harvest floods paired by equal meteorology or storm input (i.e. chronological pairing or CP). In a landmark study, Alila et al., (2009) revealed that the results of many of these past studies were misleading because they ignored the dimension of flood frequency. Changes in flood response, regardless of whether the cause is land cover or climate change must be investigated within the context of a frequency distribution that reveals changes in magnitude of floods with equal frequency (i.e. frequency pairing or FP), consistent with the methods employed by climatologists to evaluate the effects of changing climate on weather extremes (Wigley, 1985; Katz, 1993; Wigley, 2009). In the four years since the publication of Alila et al. (2009) numerous studies and literature syntheses have been published that report or re-report the results of CP-based investigations on the influence of forest and land cover changes on flood magnitudes (Buttle, 2011; Sibert and McDonnell, 2010; Bathurst et al., 2011a,b; Birkinshaw et al., 2010; Jones and Perkins, 2010;  40 Schleppi, 2011; Troendle et al., 2010; Z?gre et al., 2010; Zhao et al., 2010; Dung et al., 2012). The reluctance of the forest hydrology community to abandon CP-based analysis is surprising considering years of inconsistencies between study results and the repeated call by researchers to adopt new analytical approaches to provide a more consistent and uniform understanding of watershed scale hydrological response to forest harvesting (e.g. DeWalle, 2003; Jones, 2005; Moore and Wondzell, 2005; McDonnell et al., 2007; Alila et al., 2009; 2010). This study is a response to the recommendation by Jones (2005) and Lewis et al. (2010) to adopt a meta-analysis approach to investigate watershed hydrological response to forest harvesting. This meta-analysis is an intersite comparison using four previously-published flow data sets from snowmelt-dominated headwater catchments of the western North America Cordillera with which the influence of forest harvesting on both flood magnitude and frequency of annual maximum daily peak flows (i.e. flood events with return periods larger than one year) is investigated. This intersite comparison, investigated within the framework of a frequency distribution is used to provide new insights on catchment-scale flood response to forest harvesting not revealed previously in studies using CP-based analyses. The outcomes of FP and CP analyses are contrasted to further reveal how the results and interpretations of traditional CP-based analyses have misled the forest hydrology community in the understanding of the physics of the forests and floods relation in snowmelt-dominated watersheds. This meta-analysis uses both moderate-length (19 to 48 years) observed peak flows (Qp) from paired catchment studies as well as long-term (95 to 99 years) simulated peak flows from modeled catchment studies. Descriptions of the study catchments used in this meta-analysis are presented in Chapter 2. The use of both observed  41 and simulated flows allows uncertainties associated with posttreatment sample size and non-stationarity due to changing land cover associated with forest regeneration to be addressed, both of which may affect the estimated change in frequency and magnitude of the floods (Lewis et al., 2010; Alila et al., 2010). 3.2 Methods 3.2.1 Overview At each of the four study sites (Figure 2.1) treatment effects on the magnitude of peak flows are assessed using both frequency pairing and chronological pairing frameworks. In the frequency pairing framework treatment effects on the magnitude of peak flows are assessed by examining the difference between control and treatment catchment peak flows of the same historic probability of occurrence. In the chronological pairing framework, on the other hand, treatment effects on the magnitude of peak flows are assessed by examining the difference between treatment and control catchment peak flows generated by the same snowmelt freshet every year. At Camp and Fool, the effect of forest harvesting on peak flow regimes was assessed by comparing the observed peak flow sample following harvest (posttreatment sample) with a sample of peak flows expected to occur during the same period in the absence of harvesting (expected posttreatment sample). The expected peak flows are therefore not observed and must be modeled in some fashion. At each of these two sites, expected posttreatment peak flows in both CP and FP frameworks are predicted using the pretreatment calibration regression established from each study site (treatment peak flow regressed on paired control peak flow). Confidence intervals for the chronologically paired assessment (not shown on Figure 3.2) were derived directly from the predictive uncertainty of the calibration equation. Confidence intervals for the frequency-paired assessment are a  42 combination of this predictive uncertainty and quantile sampling uncertainty, both estimated via Monte Carlo simulation. At Fool Creek, the observed posttreatment peak flows are adjusted to remove the effects of forest regrowth by including a recovery trend with time to a regression fit applied to chronologically paired observed and expected posttreatment discharges, however such adjustment was found unnecessary at Camp Creek. This adjustment is conducted for two reasons: (i) justify the basic assumption of stationarity in frequency analysis; and (ii) allow the effects of harvesting during the most critical period prior to any substantial forest regrowth to be evaluated, as well as the effects of other longer-lasting forest land-use changes such as deforestation. At each of Redfish and 240 Creeks, the DHSVM model was used to simulate two time series of peak flows with and without forest cover generated by the same long-term proxy climate data, with no forest regrowth. Therefore, direct comparison of the two time series in FP and CP frameworks was conducted without need to adjust for recovery.  3.2.2 Adjusting for non-stationarity due to forest regeneration A multiple regression analysis that included the variable ?time since harvest? (T since Hv) indicated that no time trend is present (F= 1.71, p=0.2) in the posttreatment annual daily peak flow data set of chronologically paired observed and expected posttreatment discharges (Camp Qp and Greata Qp).  ???? ?? = 1.13 + 1.00 ? ???????? ? 0.01 ? ? ????? ??     (3-1) The variable ?T since Hv? is also not a predictor of ?Treatment effect? measured as the difference between Camp Creek observed and predicted (control) maximum daily discharge (? = 1.92, ? = 0.18).   43 Treatment effect ?? = ???? ? ?????? = 0.66 ? 0.01 ? ? ????? ??  (3-2) Hence, the 19 year posttreatment annual maximum daily peak flow data set from Camp Creek has not been adjusted to remove the effects of forest regrowth. A time trend is present in the Fool Creek data set but only when the entire 48 year data set is included in the analysis (FC = Fool Creek, ESLC = East Saint Louis Creek, (F=7.39, p<0.001)):  ?? ?? = 37.2 + 0.35 ? ?????? ? 1.12 ? ? ????? ??      (3-3) Further investigation determined that a time trend is not present when the 48-year data set is subdivided into three contiguous sub-sets of 30 years each; however, the trend is statistically significant in a data set consisting of the first 20 years and last 10 years of data. This finding suggests that forest regeneration only started to influence stand level processes in the last decade of recorded data. To remove the time trend the correction factor was applied to the data starting with the last 10 years and moving back in time until the ?T since Hv? variable was no longer a significant predictor of discharge in the regression analysis. Using this method of adjusting for recovery only the last 12 years of the Fool Creek data needed to be adjusted for the effects forest regeneration. Limiting the time adjustment to the last 12 years addresses a previous concern that adjusting for recovery in the entire posttreatment data set may contribute to additional increases in the estimated treatment effect between the pretreatment and posttreatment frequency distributions (Lewis et al., 2010).  44 3.2.3 Estimation of expected posttreatment discharges Analysis of the Camp (treatment) ? Greata (control) and Fool (treatment) ? East St. Louis (control) paired watershed data sets requires the development of a pretreatment regression model from which an expected posttreatment data set is derived during the posttreatment period. Typically the regression relation is developed through simple linear regression derived by relating chronologically paired annual peak flow data.  ???? = ?? + ?1??         (3-4) Where Xi is the peak flow of the control watershed and iY?  is the expected untreated peak flow for the treated watershed.  Due to the relatively short pretreatment calibration period for Camp and Greata Creeks the pretreatment regression is derived using the method of multiple, chronologically paired peaks (analogous to the method of peaks over threshold or PoT). This method extends the pretreatment data set from 6 to 16 paired peak flow events from which the pretreatment regression relation is defined, and is appropriate because snowmelt-dominated hydrographs of the semi-arid BC Okanagan region typically display multiple (3 or more) independent peak events (separated in time by at least 5 days, USWRC, 1976) during most freshets. Improving the regression relation for the pretreatment calibration period between control and treatment catchments by including additional paired peaks from a single freshet can be used to extend the pretreatment data set where peak flows are driven by a single meteorologic process (e.g. snowmelt; Waylen and Woo, 1982). Depending on the development of meteorology through the snowmelt season any one of the independent multiple peaks during the freshet could be the annual maximum peak. Regardless of whether  45 a given peak is the maximum annual peak or not it provides additional chronologically matched regression points that further define the relationship between the two catchments. Independence of the 16 pretreatment peak flow events at Camp and Greata Creeks was further confirmed through the nonparametric Spearman rank order serial correlation test which indicates that both time series can be considered a set of independent observations (?=0.05). 3.2.4 Flow frequency curve analysis The comparison of pre- and posttreatment flow frequency curves, also known as empirical cumulative distribution functions (CDFs), enables the assessment of the change in magnitude for a given probability (or return period) flood, or conversely, the change in probability for a given magnitude flood. An empirical approach was used without fitting a frequency distribution model to the data to avoid introducing other sources of uncertainty in our estimation of the effects of harvesting on the peak flow regime. However, for presentation purposes and without loss of generality the CDFs are plotted in probability space defined using the Generalized Extreme Value function. The process of assigning a probability (p) to a given flood (Y) from a time series of peak flows involves ranking the floods in descending order of magnitude from 1 to n such that Y(1) is the largest value and Y(m) is the mth largest value in the sample of n values, where ? ? ? ? ? ?nm YYYY ??? )(21 ? .   An estimate of the probability, p, for ranked event Y(m) is obtained using the cumulative distribution function Fy as:  46 ? ?)(mY YFp ?           (3-5) An estimate of the exceedence probability, 1?p, for ranked event Y(m) is obtained by: ? ? 2.0 40.01 )( ???? nmYF mY         (3-6) where the right-hand side of Equation (3-6) is the approximately quantile-unbiased Cunnane plotting position (Stedinger et al., 1993).  From Equation (3-6) the discharge event of rank m, Ym, is an empirical estimate of the pth quantile yp, and identical estimates exist for the paired expected sample (????) derived from the regression relation (3-4) with the control catchment. The predicted discharge (????) is corrected for loss of variance from the regression model in Equation (3-4) by re-introducing a random error, sampled from a t-distribution with n-2 degrees of freedom, to the expected discharge (Alila et al., 2009). This is done through a Monte Carlo simulation that adds the random error to the expected discharges (??? + ?), ranks the corrected expected data set (???? = ???? + ??) and repeats this for 10,000 iterations to provide an estimate of the mean, corrected, ranked expected discharge (?????). 3.2.5 Statistical vs. physical significance   For a sample of flood flows the uncertainties associated with the estimation of probability increases with event size so that the probability of the largest flood event in a sample of any size cannot be determined accurately by an empirical plotting position equation. Large errors in plotting position for the largest floods can cause a premature convergence (or divergence) of the pretreatment/control and posttreatment/observed CDFs.  47 An apparent negative change in the magnitude of the largest few events, where the upper tail of the observed CDF dips below the upper tail of the control CDF, could be real but could also be an artefact of a mismatch in event return periods and/or uncertainties in estimated expected discharges by regression models, which are often based on a short sample of peak flows (e.g. Camp and Fool Creeks).  The lack of statistical power has always been a hindrance to detecting changes in larger flood events. Lewis et al. (2010) suggest that a prudent course of action when faced with non-significant results is (1) to note the apparent direction of change, regardless of statistical significance, and (2) to conduct metastudies to investigate whether analogous changes have repeatedly been measured but declared insignificant in the absence of sufficient statistical power. In the meta-analysis, therefore, trends in the direction of the peak flow response to harvesting are looked for, irrespective of its statistical significance. By investigating several catchments concurrently it can be determined if the treatment effects on the few largest floods display similar trends regardless of posttreatment sample length. In this way it is possible to identify physically meaningful treatment effects even if they are not statistically significant. A meta-analysis approach does not require the reporting of study results in terms of null hypothesis statistical tests of significance (e.g. Alila et al., 2010, p. 4). The upper confidence intervals (95%) on the expected CDFs from the paired catchment data sets and on the control CDFs from the two DHSVM modeled catchments are included for general information only.  Confidence intervals about the expected CDFs are approximated as normally distributed errors about the estimated mean corrected, ranked discharge that are a combination of the predictive uncertainty of the calibration equation and quantile sampling  48 uncertainty (Alila et al., 2009). Confidence intervals about the control CDFs from the modeled data sets are only a function of the quantile sampling uncertainty. To facilitate sampling of empirical quantile values at the sample extremes for the observed hydrometric data sets the Generalized Extreme Value (GEV) frequency distribution is fit to the (?????) series (Alila et al., 2009). For a comprehensive explanation of the generation of the control empirical CDFs and the associated confidence limits the reader is referred to Alila et al. (2009). 3.3 Results 3.3.1 Meta-analysis and the frequency pairing framework Moderate levels of harvesting changed both the mean and the variability (as measured by the standard deviation) of posttreatment peak flows compared to the pretreatment (control) sample in the four catchments (Figure 3.1). In Camp Creek (Figure 3.1a), a 37% clear-cut harvest results in a 35% increase in the mean and almost no change (1% increase) in the standard deviation of posttreatment compared to the pretreatment peak flows. Harvesting of 40% of the watershed area in 240 Creek (Figure 3.1c) has resulted in a 15% increase in the mean and a 19% increase in the standard deviation of the posttreatment sample of peak flows compared to the pretreatment sample. Redfish Creek (Figure 3.1d), with a harvest level of 33%, displays the smallest increase in the mean (11%) but one of the largest increase in standard deviation (18%) of post treatment peak flows compared to the pretreatment sample. For all three catchments, these changes in mean and standard deviation of posttreatment peak flows have the effect of shifting the posttreatment probability density function (PDF) to the right towards the larger magnitude floods combined with a widening of the PDF so that the probability of occurrence of the larger floods increases (see insets   49  Figure 3.1. Flow duration curve analysis for pre- and posttreatment daily peak flows at (a) Camp Creek (19 years), (b) Fool Creek (48 years), (c) 240 Creek (95 years), and (d) Redfish Creek (99 years).  The point at which the two CDFs intersect, marked by the vertical arrow, increases with record length suggesting a no clear upper threshold to the effects of forest harvesting on floods in snowmelt-dominated hydroclimate regimes.  Figure 3.1). Relative to the pretreatment PDF, Camp Creek displays the largest rightward shift in the upper tail of the PDF followed by 240 and Redfish Creeks. Perhaps one of the most interesting results of this meta-analysis is the different treatment response observed in  50 Fool Creek (Figure 3.1b) where a 40% level of harvest has resulted in a 23% increase in the mean but a 12% decrease in the standard deviation of posttreatment peak flows relative to the pretreatment sample. An increase in the mean coupled with a decrease in the standard deviation of posttreatment peak flows is observed as a larger rightward-shift of the lower tail of the PDF compared to the upper tail. Interestingly, the upward shift in the mean and the negative, zero, or positive change in variability of posttreatment peak flows at the four study sites results in an upward shift of all posttreatment peak flows, save the largest event, irrespective of sample size. In all four catchments even small increases in the magnitude of larger floods translate to substantial increases in flood frequency and, remarkably, in one of the four study sites the largest increases in frequency occurred for the largest flood events. The 40% level of cut in 240 Creek has resulted in nearly a four-fold increase in the frequency of the 50-yr flood while the 20-yr flood has doubled in frequency. Similarly, the 33% level of cut in Redfish Creek resulted in a doubling of the frequency of the 10, 20, and 50-yr pretreatment floods. For the observed data sets of Camp and Fool Creeks the 5, 10, and 20-yr floods had also all doubled in frequency. These findings represent a new insight in the relation between forests and floods; in all cases, increases in frequency of all floods are observed, including the largest on record, regardless of whether the CDFs are diverging, as in the case of 240 and Redfish Creeks, running nearly parallel as in the case of Camp, or converging as in the case of Fool. Considered collectively the four paired frequency analyses suggest there is no clear upper limit to the influence of forest harvesting on flood response in snowmelt-dominated regimes. As posttreatment sample size increases the apparent ?no-effect? threshold signposted by the intersection of the pre- and posttreatment CDFs in Figure 3.1(a ? d) shifts  51 towards the right from the 20-yr for Camp to the 50-yr for Fool and to beyond the 100-yr flood for 240 and Redfish Creeks. Furthermore, the divergence of the pre- and posttreatment CDFs in two of the four study catchments (240 and Redfish) implies that treatment effects on the magnitude of peak flows in snowmelt dominated catchments in absolute terms actually increases with increasing return period. The exception being Fool Creek, where the standard deviation of the posttreatment peak flows decreases causing such effects on the magnitude of peak flows to decrease with increasing return period. It is important to note that the decreasing treatment effects expressed in relative terms as suggested by the rapidly decreasing percent increase for the flood quantiles (Q2, Q10, Q20 and Q50) given in Figure 2(a?d) is deceptive and must not be interpreted to mean that forest harvesting is not substantially affecting the magnitude and frequency of larger floods. While the relative change in flood magnitude may be decreasing (where relative change is measured as (QTPosttreatment ? QTPretreatment)/QTPretreatment) with increasing return period, the absolute change (QTPosttreatment ? QTPretreatment) is increasing in the case of 240 and Redfish Creeks, remaining nearly unchanged for Camp or, in the case of Fool, is decreasing albeit with a slow rate.  3.3.2 Chronological pairing and the missing dimension of frequency By comparing Figure 3.1(a ? d) with Figure 3.2(a ? d) the flaw is revealed in an analysis that measures treatment effects solely as changes in flood magnitude without controlling for changes in flood frequency. In all 4 snowmelt-dominated catchments, the CP-based analysis suggests that the largest floods are either not affected much or are reduced relative to the control catchment (Figure 3.2a ? d). In the CP framework, the deception that the largest floods are reduced or not affected much occurs because the largest  52 floods in the treatment catchment are not occurring at the same time as the largest floods in the control catchment, and there lies the incorrect definition (or indexing) of a ?large? flood event. An increase in variability is also observed around the posttreatment regression in the CP-based analysis, which is best illustrated by the simulated flows in Figure 3(c and d). In these two graphs, the control and treatment catchments are the same so the pretreatment regression is a perfect line with R2 equal to 1. The increased variability about the posttreatment regression line, observed as a decrease in the R2, is strictly an artifact of the wrong type of pairing and is caused by year to year variability between pre- and posttreatment peak flows. In contrast, the change in variability observed in the posttreatment PDF in Figure 3.1(a?d, insets), as measured by the change in the standard deviation, reflects the influence of forest harvesting on the variability of the full sample of posttreatment peak flows and can substantially affect the magnitude and frequency of larger floods.  Figure 3.2. The chronologically paired analysis of pre- and posttreatment daily peak flows at (a) Camp Creek (19 years), (b) Fool Creek (48 years), (c) 240 Creek (95 years), and (d) Redfish Creek (99 years). Pretreatment R2 values for 240 and Redfish Creeks are equal to 1 because the control and treatment catchments are the same for these modeled watersheds.     53 As stated in Chapter 1.1.1, the incorrect estimate of the change in flood magnitude that results from pairing flood events of different frequencies is one of the major failings of CP-based analyses. In all four catchments pairing flood events chronologically rather than by equal frequency produces the repeatedly cited CP-based outcomes that (1) treatment effects rapidly diminish with increasing control basin flood magnitude (e.g. decreasing trend of percent change in Qp with increasing control catchment flood magnitude, Figure 3.3), and (2) that forest harvesting can cause peak flow magnitude to increase, decrease or remain the same (e.g. the positive, zero or negative percent change in Qp, Figure 3.3) (e.g. Thomas and Megahan, 1998; Brath and Montanari, 2000; Calder, 2005; Moore and Scott, 2005; Moore and Wondzell, 2005; Brath et al., 2006; NRC, 2008). However, both outcomes are deceptive because of the missing element of flood frequency. Although FP analysis may also reveal   Figure 3.3. The misleading trend of decreased treatment effects for increased control catchment flood magnitude, apparent in all four data sets when CP-based analysis methods are applied (panel designation as in Figure 3.2).  54 decreasing relative treatment effect trends with increasing flood magnitude as well as positive, zero or negative relative changes in QT, these are not the same relative changes observed in the CP analysis. By ignoring flood frequency, the CP-based analysis does not reveal the correct change in flood magnitude of small, medium, and large events, or, how the frequency of small, medium, and large events have changed relative to the control catchment. An investigation of the changes in the frequency of chronologically paired floods reveals consistent changes in flood return period in the treatment catchment relative to the control catchment following harvesting for both modeled and observed data sets (Figure 3.4). Fool and Redfish Creeks, the two basins selected to illustrate this trend, both reveal that   Figure 3.4. Illustration of the changes in return period for chronologically paired floods for (a,b) Fool Creek and (c,d) Redfish Creek. In both cases moderate pre-treatment floods have been elevated in magnitude and rank to be amongst the largest floods in the harvested catchment while large pre-treatment floods drop in rank becoming more frequent in the treatment watershed.  55 posttreatment large magnitude floods have increased in frequency (decrease in return period, Figure 3.4a,c) and moderate floods have decreased in frequency (increased in return period, Figure 3.4a,c).  The corresponding change in the rank order of floods (floods are ranked from 1 to ?n? with 1 being the largest and ?n? the smallest) associated with the change in return period for large and moderate flood years is illustrated in the adjacent CDFs for both the observed (Fool) and modeled (Redfish) data sets (Figure 3.4b,d). In a cumulative distribution function the flood of rank ?n? is the smallest event located at the extreme left and the flood of rank 1 is the largest event located at the extreme right. One of the largest decreases in flood rank (i.e. rank moving up towards 1) for Fool occurred in 1958. In the accompanying CDF curve it can be seen that while the 1958 flood had a historical return period of approximately 3 years in the control catchment, the flood in the treatment catchment for the same year was nearly doubled in magnitude and had a corresponding return period of approximately 10 years. Conversely, the 1971 flood which was one of the largest floods in the control catchment for the posttreatment period with a return period of approximately 20 years decreased slightly in magnitude in the treatment catchment but the posttreatment frequency increased to a 7-yr return period. This same pattern is also illustrated in the simulated data for Redfish for the flow years 2037 and 2060.  The pattern in the change of rank order of floods between treatment and control catchments observed in both the simulated and observed data sets suggests that there has been some consistent change in the physical processes generating peak flows in the treatment catchments relative to the control catchments. More specifically, it appears that, at least in some years, the seasonal meteorology that is producing medium events in the control  56 watershed is now contributing to the generation of much larger events in the treatment watershed and, consequently, a decrease in the rank (increase in return period) of these floods in the treatment watershed. However, this is not always the case as some larger floods in the control watershed are not changed substantially in magnitude and remain large in the treatment watershed. These unchanged, and at times even slightly reduced, large events end up dropping in rank as they compete with the amplified flood events for the rank positions of larger floods in the treatment watershed. 3.3.3 Physical processes investigation A regression analysis is undertaken using both observed and simulated hydrometric and meteorologic data to explore possible factors contributing to peak flows. Meteorological data including snow water equivalent (SWE) for Camp Creek is from the Environment Canada climate station (#1126077) located 7 km north at an elevation of 1520 m. Increased snow accumulation and increased rates of snow melt are most often implicated as processes contributing to peak flow increases following harvesting (Troendle and Leaf, 1980; Troendle, 1987; Schmidt and Troendle, 1992). For the four catchments peak flows and peak SWE are positively related in all four treatment watersheds (i.e. during the posttreatment period) in this study. However, low R2 values indicate that peak SWE only accounts for 38 to 50% of the variability for the observed catchments and less than 30% of the variability in the modeled catchments (Table 3.1).  Regressions of simulated catchment-average daily snowmelt from the DSHVM studies indicate that peak discharge is more strongly determined by catchment-average 3-day total snowmelt preceding peak flows (R2=0.90 for 240 and 0.58 for Redfish (p<0.001)),  57 Table 3.1. Linear regression analysis of meteorological variables with discharge. Regression R2 P value Fool (Qp) = 6.9442(Max SWE) + 32.501 R? = 0.511, p < 0.001 Camp (Qp) = 0.0056(Max SWE) - 0.2481 R? = 0.381 p < 0.001 240 (Qp) = 0.0019(Max SWE) + 0.6933 R? = 0.076 p < 0.01 Redfish (Qp)= 0.0072(Max SWE) + 1.251 R? = 0.269 p < 0.001 Redfish (Qp) = 0.0705(3 day melt) + 1.3132 R? = 0.582 p < 0.001 240 (Qp) = 0.0184(3 day melt) - 0.0022 R? = 0.902 p < 0.001 Fool Creek (Qp) = 4.5179(3 day avg T) + 206.9 R? = 0.014 p>0.05 Camp Creek (Qp) = 0.107(3 day avg. T) + 0.5125 R? = 0.475 p<0.001 Camp(Daily Q) = 0.115 (7 day avg. T) ? 0.0094 (1981) R? = 0.812 p<0.001 Fool (Daily Q) = 16.67 (7 day avg. T) ? 14.31 (1976) R? = 0.813 p<0.001 Units: Qp (m3/s) except Fool Creek (l/s), Max SWE (mm), 3-day melt (mm), Avg T (oC),  although daily snowmelt data were not available for Camp or Fool Creek. For these catchments observed daily air temperature is used as a proxy for snowmelt (Hock, 2003). Annual peak discharge is found to be moderately correlated with (R2=0.48, p<0.001) 3-day average temperature preceding peak flows in Camp but in Fool neither 24-hr, 3-day or 7-day average temperature preceding peak flows were found to be significant predictors of the magnitude of the annual peak flow (3-day average temperature vs. Qp, R2 = 0.01, p>0.05). However, the daily discharge (Daily Q), for the two week period preceding and including the peak discharge, is strongly determined by the preceding average 7-day air temperature in both Camp and Fool Creeks (R2=0.67 to 0.94 for period of record, p< 0.05) suggesting that the increases in daily net radiation in the days preceding peak flows is the primary factor in catchment scale flood response rather than the measure of total energy (i.e. maximum daily temperature or average 3 day temperature) at the time of peak flow. To help understand how changes in shortwave radiation related to forest harvesting affect flood response in watersheds with different physical characteristics a frequency analysis is used to investigate changes in catchment-average total 3-day snowmelt preceding  58 peak flow between forested and harvested scenarios following 40 and 33% harvest in 240 and Redfish Creeks, respectively (Figure 3.5). These two watersheds represent   Figure 3.5. Flow duration curve analysis of catchment average 3-day snowmelt preceding the annual maximum peak flow for: (a) 40% clearcut and control scenarios in 240 Creek and (b) 33% clearcut and control scenarios for Redfish Creek. physiographic opposites; 240 is a small, fully forested, moderate gradient watershed whereas Redfish is a much larger, steeper basin containing 40% alpine area. The results for both 240 and Redfish reveal similar increases in the mean of the total 3-day snowmelt (increases of 7% in 240 vs. 8% in Redfish) with moderate levels of harvest but larger increases in the variability of 3-day melt in 240 (23%) compared to Redfish (17%). As snowmelt is the dominant process in generating peak flows it is expected that changes in mean and variability of basin average snowmelt preceding peak flows will transmit to changes in the  59 mean and variability of peak flow response at the outlet of the watershed but that this response could be amplified or mitigated by the location of harvested openings and routing of surface and subsurface runoff. 3.4 Discussion  A frequency based meta-analysis reveals that moderate levels of forest harvesting (33 to 40%) has affected the entire flood frequency distribution including the largest floods on record in all four study catchments, irrespective of sample size. In two of the four study catchments (240 and Redfish) harvesting has resulted in increases in both the mean and the standard deviation of the posttreatment series of flood peaks compared to the pretreatment or control series. In the third catchment (Camp) harvesting has increased only the mean with almost no change in the standard deviation while in the fourth catchment (Fool) harvesting has increased the mean but decreased the variability of posttreatment floods.  Investigating changes to the mean and variability of posttreatment flood peaks represents a paradigm shift with respect to the way treatment effects are physically explained and quantified (Alila et al., 2009) in the study of forest hydrology. Past investigations using CP-based analysis, with associated ANOVA or ANCOVA methods, attempted to quantify changes in the mean of the posttreatment sample while controlling for changes in variability, most often through log transformation of flood magnitudes or categorical parsing of data sets (e.g. Jones and Grant, 1996; Thomas and Megahan, 1998; Jones, 2000). However, this traditional approach fails to recognize the importance of changes in variability in quantifying changes in extreme events (Katz and Brown, 1992). By investigating treatment effects in terms of changes in both the mean and variability of the  60 posttreatment frequency distribution it can be revealed how, in snowmelt-dominated regions, harvesting has influenced floods across a much wider range of magnitude including the largest floods on record.  The increase in the mean and variability of posttreatment peak flows at Camp, 240, and Redfish Creeks and the upward shift of all posttreatment peak flows, save the largest event, translate into a divergence of the empirical posttreatment CDFs with increasing return period relative to the pretreatment CDFs. The sudden disappearance of the vertical difference between the pre- and posttreatment CDFs at the largest flood for these three catchments cannot be touted with any physical meaning without considering the error in the plotting position for the largest one or two floods (e.g. Alila et al., 2009). In fact, the dip in the last point that gives the impression of no effect on the largest flood is inconsistent with the physical outcome of an increase in the mean and standard deviation which is, that the posttreatment curve is shifted upward and becomes steeper than the pretreatment curve and therefore the two curves should diverge (or run parallel in case of a no change in standard deviation), at least within the range of observed or simulated peak flows. The intersection between pre- and posttreatment CDFs, imposed by the dipping of the largest flood event, shifts to a higher return period from about 20-yr at Camp to beyond 100-yr at 240 and Redfish Creeks. The meta-analysis therefore clearly illustrates that this dipping of the largest flood event is not real and is an artifact of sample size.  Within a frequency distribution framework, consistent changes in flood frequency are observed in the four study catchments regardless of the changes in the variability of the posttreatment peak flows (i.e. whether the CDFs are converging, diverging, or running parallel). The upward shift in the posttreatment CDF associated with the increase in the  61 mean of the distribution causes the frequency of all floods in all four catchments to increase despite the deceptive rapid decrease in the relative increase in the magnitude of the 2, 10, 20, and 50-yr return period floods (Figure 2). In general, it is observed that events up to approximately the 20-yr flood double in frequency (e.g. 10-yr becomes a 5-yr and 20-yr becomes a 10-yr flood) while larger floods tend to increase in frequency by 2 to 4 times (e.g. 50-yr may become a 30-yr or 13-yr flood). The larger increase in frequency for the larger magnitude floods is due to the highly non-linear relation between flood frequency and magnitude which creates large increases in frequency for floods in the upper tail of the distribution even if the increase in magnitude is below statistically significant levels. The effect of larger increases in frequency for larger floods appears to be particularly evident in these snowmelt-dominated catchments where the slopes of the CDFs are relatively gentle so that small changes in magnitude translate into surprisingly large changes in return periods. It is therefore illustrated for the first time how the occurrence of more frequent events of same magnitude, or higher magnitude events of same frequency, are amplified in watersheds with milder sloping flood frequency curves as a consequence of forest harvesting. In a century of CP-based literature that rarely invoked the dimension of frequency only Berris and Harr (1987; p. 141) hinted to such a scientifically and practically profound generalizing percept that the peak flow regimes of watersheds with milder slope flood frequency curves should be more sensitive to forest harvesting. In all four catchments forest harvest appears to have increased flood magnitude over a wide range of return periods including the 5, 10, and 20-yr events in Camp and Fool Creeks, and up to the 50-yr events in 240 and Redfish Creeks. Most intriguing perhaps is that the outcome of the meta-analysis suggests there is no clear upper threshold to the effects  62 of forest harvesting on floods in snowmelt-dominated hydroclimate regimes. As posttreatment sample length increases between the four catchments the threshold return period beyond which forest harvesting has no effect shifts towards the right. The right-ward shift of the no-effect threshold with increasing posttreatment sample length was also demonstrated by Alila et al. (2009) when they extended the posttreatment sample length for Fool Creek from 29 years to 48 years. Kura? et al. (2012) came to the same conclusion of a no clear upper threshold to the effects of forest harvesting on floods in snow environments using 100 years of model simulated post-harvest peak flows. These findings are inconsistent with the claim reported and re-reported in the literature from snowmelt-dominated regimes, that forests have no or minimal effects on larger return period floods (e.g. MacDonald and Stednick, 2003; Moore and Wondzell, 2005; NRC, 2008; Bathurst et al., 2011a). A frequency analysis of basin average total 3-day melt preceding peak flows using simulated snowmelt output for 240 and Redfish Creeks indicates that in both catchments moderate levels of forest removal increases the mean and variability of basin average total 3-day snowmelt preceding peak flows but despite physical differences between the two catchments only the change in variability displays any substantial difference (standard deviation shows a 23% increase in 240 versus 17% increase in Redfish). This result indicates that differences in basin physiography (size, aspect, slope gradient, etc.) has a greater influence on the behavior of extreme melt events which influence the standard deviation of the sample than on the mean melt event. In both 240 and Redfish Creeks, similar increases in the mean of the total 3-day melt (7% and 8%, respectively) reflects similar average increases in snowmelt with similar levels of forest removal. In both  63 catchments increases in total 3-day melt is maintained over the full range of meteorological conditions included in the model. The results of the physical process investigation of melt rates does not support the assertion that larger floods are not affected by harvesting due to an ?overwhelming? of the interception capacity of the forest canopy and storage capacity of the forest soils that has been used to explain the relative decrease in treatment effects with increasing flood magnitude (MacDonald and Stednick, 2003; Brooks et al., 2003; Lee, 1981; Jeffrey, 1970; Chang, 2006; NRC, 2008). To the contrary, regardless of the relative role of forests on evapotranspiration, the findings presented here suggest that the mitigating effects of the forest canopy on snowmelt, the primary process controlling peak flows, are maintained over the full posttreatment sample, which is consistent with multi-year stand level investigations that have found melt rates to be consistently higher in cut blocks relative to the forest regardless of the seasonal variability in meteorology during the study period (e.g. Golding and Swanson, 1978; Kattelmann, 1991; Toews and Gluns, 1986; Teti, 2004; Winkler et al., 2005; Jost et al., 2007).  The collective outcomes of the meta-analysis and physical process investigation suggest that physical basin characteristics contribute differently to changes in the mean and the variability of the frequency distribution. Smaller differences in changes in the mean of the frequency distribution between the four study catchments (see insets. Figure 3.1) suggests that similar levels of harvest are contributing to consistent increases in net radiation related to the change from longwave to shortwave-dominated melt. Comparitively larger changes in peak flow variability between the four study catchments appears to be a measure of the effectiveness and synchronization of snowmelt delivery to the stream channel. The  64 influence of basin characteristics with respect to synchronization of snowmelt runoff and delivery of runoff to stream networks consistent with the conjectures presented above has recently been confirmed using detailed DHSVM model output for the upper 50% and bottom 50% harvest scenarios in 240 Creek [Schnorbus and Alila, 2012]. The results of their physical process investigation using simulated model outputs indicate that harvesting in the upper 50% increases peak flow response through the synchronization of melt between upper and lower elevation slopes. However, in the case of 240 Creek the timing of delivery of meltwater from upper elevation slopes is influenced more by low drainage density in the upper half of the drainage than by elevation range [Schnorbus and Alila, 2012].    65 4 Harvesting effects on flood duration and number of peaks over threshold discharge 4.1 Introduction A seminal study by Costa and O?Connor (1995) revealed that the capacity of a flood to cause channel change depends on both its magnitude and duration. This assertion is based on observations that some very large floods of short duration have resulted in minimal long-term changes in channel morphology (Kochel, 1988; Costa and O?Connor, 1995; Huckleberry, 1994). The geomorphic effectiveness of a flood is measured by its capacity to transport sediment and alter channel form (Wolman and Miller, 1960; Wolman and Gerson, 1978). Costa and O?Connor (1995) determined that floods of medium to long duration with medium to large total energy expenditure and large peak stream power per unit area have greater effectiveness than floods of short duration but large peak stream power. In the case of the short duration flood, Costa and O?Connor (1995) observed that erosion of the channel banks and floodplain was minimal because the flood did not last long enough to overcome flow resistance associated with vegetation and debris leading them to conclude that the effectiveness of a flood depends on flood attributes as well as the resistance thresholds of the channel boundary material. In their investigation of bedload yield in the gravel-bed headwater stream of East St. Louis Creek, Adenloff and Wohl, (1994) reached similar conclusions regarding the importance of discharge attributes including flood duration. Their findings, which were consistent with the findings of Troendle (1992), indicated that bedload yield in a steep, headwater stream was dependent on flood duration as well as magnitude because of the high resistance associated with a coarse armor layer and woody debris structures on the channel bed.  66 Forest harvesting effects on flood duration are poorly understood. A study of urbanization impacts revealed that conversion from a forested to urban landscape resulted in increases in both the magnitude and duration of geomorphically effective floods (Booth, 1990). Additionally several studies have shown that changes in forest cover influences the cumulative duration of daily and instantaneous discharge (Troendle and Olsen, 1994; Troendle et al., 2001; Archer, 2007).  Changes in the number of flood peaks over threshold discharge is another geomorphically important attribute of the flood regime that has received little attention in the investigation of forest harvesting effects on floods. In gravel-bed channels, patterns of bedload transport are complex due to a surface armor layer that inhibits bedload mobility (Parker et al., 1982; Wilcock and McArdell, 1993; Church et al., 1998; Emmett and Ryan, 2002; Frey and Church; 2002; Mueller, 2005). In such streams bedload yield depends, in part, on the number of peak events exceeding the threshold conditions for armor-layer disruption or mobilization. Several studies have reported that bed mobility is limited during the earliest peak events but increases with subsequent peaks over threshold discharge as bedload accessibility increases with armor layer break-up or as bedload is released from behind large woody debris jams (Reid et al., 1985; Sidle, 1988; Bunte, 1996; Gomi and Sidle, 2001).  Changes in the duration of individual flood peaks or the number of flood peaks over threshold discharge has the potential to alter the dynamics of bedload mobility in headwater gravel-bed streams to a similar extent as changes in the magnitude of floods. In all cases the resulting changes in the flood regime can lead to changes in long-term sediment yield and channel morphology reflective of the dynamic equilibrium between conditions of sediment  67 supply and transport capacity within a watershed (Montgomery and Buffington, 1997). In snow environments alterations to the flood hydrograph due to forest harvesting are likely to persist for decades or longer until sufficient forest regeneration restores the snowmelt hydrograph. The limited focus of past forest hydrology studies that have attempted to quantify harvesting effects based on a single attribute of the flood regime (i.e. flood magnitude) highlights a long-enduring disconnect between the studies of forest hydrology and fluvial geomorphology. 4.2 Methods A meta-analysis approach is used to investigate the effects of moderate levels of forest removal on the duration of flows above threshold discharge and the number of flood peaks over threshold discharge (POT) for initiation of bedload transport. The four snowmelt dominated catchments included in the meta-analysis are the same as those described in Chapter 2.  Harvest levels in the treatment catchment range from 33 to 40 percent (Table 2.1). A summary of the physical characteristics of the catchments included in the meta-analysis is also included in Table 2.1 (Chapter 2.1). In both the flood duration and the POT investigation the average daily discharge time series was used for the analysis as this is the only discharge data available for the two observed paired catchment studies (Fool - East St Louis Creeks and Camp ? Greata Creeks).  Five discharge thresholds (flood quantiles) including the 0.6*Q1.5, Q1.5, Q2, Q5, and Q10 were selected for investigation in both the flood duration and POT analyses. Discharge magnitudes associated with the flood quantiles were estimated using the Generalized Extreme Value distribution (GEV). For the two paired watershed studies (i.e. Camp/Greata  68 and Fool/E.St.Louis) quantile flood magnitudes were estimated using the GEV flood frequency analyses of the expected annual flood series for the treatment watersheds (corrected for loss of variance as described in Chapter 3.2.3). The expected GEV distributions for both Camp and Fool Creeks were compared against the shorter pre-treatment GEV distributions to ensure general consistency in estimated flood magnitudes? particularly for the mean flood (Q2) and bankfull flood (Q1.5) and were found to be within 5% of each other.  Table 4.1. Expected (control) discharge magnitude for selected flood quantiles from GEV distribution (m3/s). Quantile Camp Greata Fool E. St. Louis Redfish 240 0.6*Q1.5 0.54 0.17 0.13 0.40 3.56 0.54 Q1.5 0.9 0.28 0.22 0.68 5.93 0.9 Q2 1.4 0.44 0.25 0.75 6.36 1.0 Q5 1.6 1.0 0.30 0.90 7.47 1.3 Q10 2.1 1.2 0.40 1,00 8.2 1.4 The thresholds of 0.6*Q1.5, Q 1.5, Q2, Q5 and Q10 were selected for investigation because they represent geomorphically significant discharges. In many gravel-bed streams including Fool Creek and E.St. Louis Creek (Ryan et al., 2005), the bankfull flood (Qbf) corresponds closely to the Q1.5. The bankfull flood is associated with high levels of bedload mobility (Ryan et al., 2005) and is typically referred to as the channel maintaining flood (Andrew, 1980, Emmett and Wolman, 2001), however, bedload mobility often initiates at discharges of approximately 60% to 80% of the bankfull flood (i.e. 0.6*Q1.5) (Ryan et al., 2005).  The Q2, which corresponds approximately to the average annual flood, as well as the Q5 and Q10 flood quantiles are included in the analysis to provide insight regarding how the duration and number of POT of floods with magnitudes approaching the channel forming flood are affected by forest harvesting.  The relatively short record length for the observed  69 hydrometric data sets (i.e. Camp/Greata and Fool/E.St. Louis) limits investigation of larger flood quantiles.   4.2.1 Duration over threshold Changes in the duration of flows above threshold discharge is investigated using cumulative flow duration analysis. Total seasonal flow for discharges equal to or greater than 25% of Q1.5 for the snowmelt period of April to July (240 and Camp Creeks) and May to August (Redfish and Fool Creeks) is used as an expression for total snowmelt-related flow volume. Limiting the duration analysis to the days during the snowmelt period where flows are greater or equal to 0.25*Q1.5 captures both the initiation of elevated snowmelt discharges as well as discharges corresponding to mobilization of substantial volumes of bedload in headwater streams (Ryan et al., 2005). Cumulative flow duration analysis compares the total duration in days during which flows equal or exceed the selected flood discharges between the expected (i.e. unharvested) and observed daily discharge time series. For Fool/East St. Louis and Camp/Greata watershed pairs the expected daily mean discharge was estimated based on a regression model developed using the pre-treatment daily discharge data. Fool Creek daily Q(l/s)= 0.3134* Greata daily Q - 21.009 (R? = 0.89, S.E. = 27.4)      (4-1) Camp Creek daily Q(m3/s)= 1.1848*Greata daily Q + 0.1004 (R? = 0.77, S.E. = 0.26) (4-2) The use of regression models to generate the expected daily flow time series introduces considerable error into the analysis particularly for the largest discharges that are not well represented in the pre-treatment data set. The meta-analysis approach that combines results from observed and simulated catchments to identify similar trends in catchment  70 response improves the strength of study outcomes. Cumulative flow duration curves are constructed for the modeled catchments (240 and Redfish Creeks) using 50 years of simulated daily discharge (also limited to snowmelt period for Q ? 0.25*Q1.5) for the unharvested and harvested scenarios (i.e. 40T for 240 Cr. And 100U for Redfish, see Chapter 3.2).  Total number of days that flows equal or exceed a specified discharge is expressed as the cumulative percent exceedence for the total post-treatment time series. Daily discharge from the expected (control) and observed (harvested) time series are ranked in order from largest to smallest. The rank (m) is applied to the total daily series (total number of days = n) such that 1 corresponds to the largest discharge and n corresponds to the smallest discharge. The percent of time flows of a given magnitude are equal or exceed for a given discharge is calculated as;  % exceedence = (?? *100)        (4-3) For Camp Creek the full 30 year posttreatment hydrometric data set (1978-2009) is used in the flow duration analysis because it is unlikely that the relatively small amount of additional harvesting (8%) that occurred between 1977 and 1991 would have a substantial effect on the daily discharge time series.   4.2.2 Peaks over threshold For the number of peaks over threshold (#POT) analysis forest harvesting effects were investigated for the two paired watersheds using the traditional Before-After-Control-Impact (BACI) method of analysis. For the two modeled catchments (Redfish and 240  71 Creeks) the treatment effect was measured directly as the difference in #POT for the control and treatment scenarios.   Figure 4.1. Sample daily hydrograph showing peaks over threshold. The identification of a flood peak is somewhat subjective. For a peak to be counted it must display a rapid increase followed by a decrease in discharge. The required drop in discharge was arbitrarily set at a minimum of ten percent of the Q1.5 to distinguish one peak from another. On the example figure (Figure 4.1) the Redfish Control hydrograph (blue line) displays five peaks equal or exceeding 0.6*Q1.5, three peaks equal or exceeding Q1.5, two peaks equal or exceeding the Q2 threshold and zero peaks equal or exceeding the Q5 and Q10 thresholds. Peaks over threshold were counted manually using graphs of daily discharge. An alternative approach, which involved the use of Matlab to identify peaks on the basis of peak height, was also used to generate POT data for the four study catchments. The Matlab method produced results generally consistent with the manual counting however the limited ability to filter data resulted in the inclusion of too many smaller flood peaks. Due to the  72 time consuming method of analysis investigation of the simulated daily hydrometric data for Redfish and 240 Creeks was limited to 50 years. In the paired catchments treatment effects (TE) were measured as the control (c) to treatment (t) catchment differences in the #POT for each flood quantile (QT) between pre- and posttreatment periods;  Pre-treatment period     Posttreatment period TE(QT) = (#POT(QT) c - #POT(QT) t)  versus   (#POT(QT) c - #POT(QT) t)  (4-4) A t-test was used to determine whether differences in #POT exist between the pre- and posttreatment periods for each flood quantile investigated. A test for consistency in variances found that variances could be considered similar for pre- and post-treatment datasets only for the 0.6*Q1.5 and the Q1.5 flood quantiles (this finding also applied to the simulated flows). For flood quantiles larger that the Q1.5 the t-test assuming unequal variances was used to test whether the differences in the pre-treatment and post-treatment mean differences were similar. The lengths of the pre-treatment periods varied from 10 years for Fool/ESL to 7 years for Camp/Greata. The 7 year pre-treatment period in Camp is not ideal and limits the robustness of the statistical test however, in the meta-analysis framework the objective is to identify similarities or differences in treatment response through repetition of testing rather than focussing on statistical significance. For this reason the results of the statistical testing in the analysis of #POT are included in Table 4.2 but trends in outcomes are considered in the discussion of results.  73 4.3 Results 4.3.1 Flood duration  Figure 4.2. Cummulative percent exceedence of daily mean discharge for the period of record. In all four catchments the duration of discharge equal to or exceeding specified flood quantiles increased following harvesting compared to the control or expected condition (Figure 4.2). Of the four catchments, Camp Creek and Fool Creek display the largest increases in flood duration for a given quantile (i.e. up to 288% and 305% increases respectively) while Redfish and 240 Creeks display smaller increases (increases of 87% and 187% respectively). In all catchments there is a general increase in treatment effect from the 0.6*Q1.5  threshold up to the Q2 or Q5 thresholds and then a comparative drop in treatment effect for the Q10 threshold. In 240, Fool and Camp Creeks the duration of discharges equal or exceeding the Q5 has more than doubled (i.e. >100% increase) compared to the control or  74 pre-treatment condition. In Redfish Creek the Q2 and Q5 floods display consistent increases in duration of over 80%. The decrease in the Q10 treatment effect is less apparent for Fool Creek that for the remaining three catchments. The relatively small increases in duration of flows equal or exceeding the Q10 reported for Camp Creek (17%) and Redfish Creek (40%) translate into 1.2 and 7 additional days respectively of flows equal or exceeding the Q10 threshold over the 50 year post-treatment period. 4.3.2 Number of peaks-over-threshold Table 4.2. Differences in average number of peaks over threshold for pre- and post-treatment periods. Statistically significant differences highlighted in bold type.  Camp Fool Redfish 240 QT ?1 Pre (7 yr) ?2 Post (33 yr) %? # POT P3 ?1 Pre (10 yr) ?2 Post (48 yr) %? in # POT P3 x?4  % ?# POT P3 x?4  %? # POT P3 0.6Q1.5 -1.3 -0.61 53 0.1 1.3 0.35 73 0.02 1.02 2 0.35 1.2 18 0.014 Q1.5 0.14 -0.38 367 0.02 0.6 -0.19 131 0.03 1.4 36 0.007 1.2 21 0.049 Q2 1.29 0.30 76 0.03 0.6 0.08 86 0.05 1.6 59 0.001 1.3 34 0.022 Q5 0.57 -0.15 127 0.05 0.5 -0.25 150 0.002 2.3 135 0.000 2.8 179 0.000 Q10 0.43 -0.03 107 0.09 0.3 0.02 93 0.05 2.2 123 0.005 3.2 222 0.000 1 Difference in mean #POT (Control ? treatment) pre-treatment period 2 Difference in mean #POT (Control ? Treatment) post-treatment period 3 (T?t) 1-tail 4 Change presented as a multiple of the control #POT  In Camp and Fool Creek the change in #POT is measured as the mean difference between control and treatment catchment #POT for the pre-treatment period and post-treatment periods (Table 4.2). A decrease in mean #POT from pre-treatment to post-treatment periods indicates a relative increase in the #POT?s in the treatment catchment during the post-treatment period. In Camp Creek the small pre-treatment sample (i.e. seven years) limits the reliability of these results, particularly for the largest quantile, however when considered together with the three other catchments the results are generally consistent.  In Redfish and 240 Creeks the difference in #POT is measured directly from  75 control and treatment simulated daily time series. Increases in #POT are observed for all flood quantiles in all treatment catchments (Table 4.2). In two of the four catchments (Fool and Redfish) the greatest increase in #POT?s is observed for the Q5 quantile. 240 Creek displays an increase in #POT with increasing flood quanitle across the full range of threshold floods resulting in over three times more POT at the Q10 quantile for the harvested condition compared to the unharvested control. Camp Creek displays the largest relative increase in #POT for the Q1.5 quantile.  4.4 Discussion An investigation of changes in the duration and #POT of flood events contributes towards a comprehensive understanding of the influence of harvesting on the snowmelt hydrograph. Analysis results indicate that harvesting causes the duration and #POT of all snowmelt peak flows to increase and that these increases are generally greater for larger flood quantiles. Such a response implies that, following harvesting, the flood hydrograph in all four catchments has become more responsive during the snowmelt period in terms of an increase in the number of peak flow events. In addition harvesting has resulted in increases in the total volume (i.e. duration) of flood peaks so that discharge remains elevated above specific thresholds for longer periods of time. The meta-analysis results suggest that physical basin characteristics including percentage of alpine area, slope aspect, and gradient, elevation and watershed size all play a role in catchment scale response to harvesting-related increases in flood duration and # POT. However, the individual effects of these characteristics are not straightforward. Harvesting- 76 related increases appear to decrease with increasing alpine area and watershed size. Harvesting effects also appear to decrease with increasing aspect and elevation distribution. Redfish Creek, the largest of the four catchments containing alpine area, displays the smallest increases in the duration (+4%) and #POT (+2%) for floods less than bankfull (Q1.5) compared to the other three catchments. In contrast, Fool Creek the small alpine catchment displays relatively large increases in the duration (+69%) and #POT (+73%) for floods less than bankfull. The contrast between Redfish and Fool Creeks for the smaller flood quantiles suggests that snowmelt from lower elevation harvested areas on east-west aspect slopes in the larger, steeper Redfish Creek contributes less to peak streamflow volume and responsiveness for these smallest flood peaks than in smaller, gentler gradient, Fool Creek. Further comparison of Redfish and Fool Creek reveals that both alpine catchments display a relative decrease in harvesting effect for the largest flood quantile (Q10), an indication that, in both cases, runoff from upper alpine elevations may contribute to a greater degree in the generation of the largest flood peaks. These interpretations regarding the influence of different elevations on peak flow generation are consistent with study outcomes documenting that prior to harvesting peak flows in Fool Creek were generated by snowmelt from alpine areas but following harvesting, melt from lower elevation slopes contributed more often to peak flow generation (Troendle and Leaf, 1981).  In the two non-alpine catchments much larger increases in duration and #POT (Figure 4.2 and Table 4.2) are observed in Camp compared to the smaller 240 Creek for the smallest flood quanitles (?Q2) which likely relates to the distribution of slope aspects and harvesting. In Camp Creek harvesting is concentrated on southern aspect slopes that are quick to respond to transient periods of warm temperatures throughout the snowmelt period  77 (Ellis et al., 2011). Whereas, in 240 Creek, slope aspects and harvesting are more evenly distributed (Table 2.1) so that snowmelt from openings is modulated by indirect solar radiation and possibly shading for part of the day. Less distinction exists for changes in duration and #POT of larger flood quantiles (e.g. Q5, Figure 4.2 and Table 4.2) in the two non-alpine catchments likely because these larger floods relate to multiple days of sunny, warm temperatures contributing to a single peak flow response generated by a wider range of slope aspects and elevations in both catchments.  A comparison of study outcomes for the two smallest catchments reveals substantially larger changes in #POT and duration for floods less than the Q5 (i.e. small to average floods) in Fool compared to 240 Creek (Figure 4.2 and Table 4.2). These differences are less straightforward to interpret given similarities in catchment size, slope gradient and harvesting distribution (Table 2.1). The main physical differences between the two catchments relates to elevation and percent alpine area. Compared to Fool Creek, the outlet of 240 Creek is situated at nearly 1500 meters lower elevation (Table 2.1) and the catchment is fully forested. Differences in hydrograph response between the two catchments suggests that during average years smaller differences in melt dynamics occur between forests and openings in 240 Creek where melt is occurring earlier in the year (i.e. early May, Table 2.1) compared to Fool Creek where snowmelt occurs later in the season (i.e. later June). This interpretation agrees with findings of detailed study of variability in snowmelt dynamics over different elevations and aspects in a snowmelt dominated forested catchment in BC (Jost et al., 2007). Larger increases in #POT for floods exceeding the Q5 in 240 Creek (Table 4.2) again supports the interpretation suggested above, that snowmelt from alpine  78 areas in Fool Creek (and Redfish Creek) are, to some degree, moderating harvesting related increases during the largest floods.     79 5 Channel morphology and bedload yield dynamics of forested snowmelt streams.  5.1 Introduction Headwater streams are a distinct class of channels with characteristic morphologies, processes, and dynamics (Gomi et al., 2002; Hassan et al., 2005). They occupy the outermost, portion of the drainage network and facilitate the transfer of sediment and organic material from hillslopes to high-order streams (Church, 2002). In steep terrain coarse sediment and wood inputs governed by mass wasting exert a primary control on channel morphology (Grant et al., 1990; Benda and Dunne, 1997) and bedload transport (Gomi and Sidle, 2003).  Headwater streams situated in moderate gradient upland regions represent a unique subset of small streams in which fluvial rather than colluvial processes prevail (Buffington et al., 2003).   The spatial distribution of channel-reach morphology in mountain streams has been explained as the expression of the interaction between sediment supply and transport capacity (Montgomery and Buffington, 1997, 1998). The downstream progression of channel types along a concave-up longitudinal profile reflects a general shift from supply-limited (colluvial reaches, boulder-cascades, and step-pools) to transport-limited (riffle-pools) conditions with increasing opportunities for in-channel storage (Church, 1992). The incorporation of large woody debris from forested areas into stream channels can influence channel morphology and sediment mobility by increasing the capacity for sediment storage in steep gradient reaches (Montgomery and Buffington, 1997, 1998).  80 In steep headwater streams the movement of bedload material exhibits high spatial and temporal variability in relation to conditions of peak flows and sediment supply (Bunte, 1996; Ryan et al., 2005). In steep streams seasonal bedload exhaustion patterns (clockwise hysteresis) associated with the maximum annual flood have been associated with depletion of annual mass wasting-supplied sediment (Gomi and Sidle, 2003). Short-term bedload transport variations in mountain streams are usually associated with the break-up and subsequent re-establishment of the bed armor layer (Gomi and Sidle, 2003; Hassan et al., 2005). Although important progress has been made in understanding the hydro-geomorphology of humid, steep headwater channels, in mountain environments headwater streams are not necessarily synonymous with mass-wasting dominated systems. In fact, there exist many mountainous regions where hydro-climatic (e.g., continental) and topographic boundary conditions (e.g., moderate gradient or plateau-like topography) do not favour a landscape configuration dominated by rapid shallow failures. For example, in British Columbia fluvially-dominated headwater streams drain over 30% of the Canadian Cordillera, including the Interior Plateau and portions of the Columbia Mountains and Rocky Mountains (Holland, 1964).  This typology of headwater systems, which has been largely neglected, present distinct morphodynamics. An understanding of their hydro-geomorphic functioning is fundamental for addressing the management of forest and aquatic resources. This study examines fluvially-dominated headwater streams considering both morphological and sediment transport aspects and presents a comprehensive account of headwater streams in which mass wasting disturbance is not prevalent.   81 The objectives of this investigation are to (1) examine the spatial variability of channel-reach morphology and cross-sectional channel variates at the watershed scale, (2) evaluate bedload sediment dynamics in relation to hydrologic (i.e.discharge) variability, (3) elucidate linkages between channel morphology and bedload transport, and (4) investigate the effects of topographic and sedimentary glacial imprints on channel morphodynamics. These objectives are pursued through an experimental set up that relies on a nested network of hydrologic stations and in-channel sediment traps in Cotton Creek, a forested, fluvially-dominated watershed of the southern Columbia Mountains, Canada. 5.2 Data collection and methods Data collection combines aerial photo interpretation (API), GIS analysis, field surveys, and the installation of water and sediment monitoring stations. API served for identifying major landscape structures and sediment sources. GIS analyses were instrumental to extract longitudinal profiles (Figure 5.5a), calculate drainage area, hence to build area-slope plots (Figure 5.5b) from a 25-m digital elevation model (Montgomery and Foufoula-Georgiou, 1993). Mapping (Halleran, 2004) of the surficial materials (Figure 2.2b) was conducted via fieldwork and API following the British Columbia Terrain Classification System (Howes and Kenk, 1997). Specifically, 50% of the terrain polygons delineated through API were field-checked to confirm distribution and characteristics of surficial materials. Intensive field surveys along the study channel network (10.7 km) entailed delineation of 23 channel reaches according to the degree of hillslope-channel coupling and dominant channel bed morphology, as well as measurement of the quantitative  82 morphometric variables that control and best summarize channel morphodynamics (Church, 1992, 2006; Montgomery and Buffington, 1998).  Field-based channel survey data include local slope (S), bankfull width (w), bankfull (thalweg) depth (d), largest actively mobile particle diameter (Dmax), and number and height of functioning large woody debris (LWD) pieces. Channel slope was measured with a theodolite and stadia rod. Bankfull width and the intermediate axis of the five largest stones (here referred as Dmax) were measured with a metric tape; depth measurements at bankfull were taken with a metric rod along the thalweg. Bankfull level was measured as the transition between vegetation types indicative of partially saturated (e.g. grasses, horsetail, and moss) and fully saturated (e.g., moss only) soil moisture conditions, as well as by physical marks on the vegetated channel banks indicative of repeated scour by flood flows (e.g., Wohl and Wilcox, 2005) (Figure 5.1).    Figure 5.1. Bankfull was identified using physical indicators including vegetation changes and marks on the banks associated with scour.  Depending on channel accessibility (i.e., obstructions associated with LWD pieces), field measurements were taken at a length scale equal to 2-10 times the local channel width,  83 and then averaged (weighted by distance) across reaches. Channel reach lengths spanned between 50 and 100 channel widths. Dmax is meant to represent a proxy for first-order bed roughness, insensitive to exceptionally large boulders. Given the foregoing specifications, surface Dmax for one reach is estimated from 30 to 60 stones collected within the active channel bed. To assess Dmax reliability Wolman pebble counts were conducted in 13 reaches and it was determined that Dmax corresponds, on average, to D88 of the active channel bed surface with a relatively narrow range of variability (i.e., D83- D95) (Figure 5.2).  Figure 5.2. Correlation between Dmax and Wolman pebble count D84 grain size. Sediment yield was monitored using a nested network of channel spanning traps and removable pit traps (Figure 2.2a, Figure 5.3) over 3 to 4 snowmelt freshets at 8 locations along the channel network (Table 2.2). In, 2005, the initial year of instrumentation, the sediment monitoring network included one headwater weir on Elk Creek (E1) and two channel spanning pit traps in the lowest reaches of Upper Cotton and Elk Creeks (E4 and C4). During this start-up year sediment monitoring was limited to total yield (Table 2.2). In 2006 and 2007 the monitoring network was completed and included an additional headwater  84 weir (C1) as well as 4 removable rectangular pit traps (E2, E3, C2, C3). All sites were monitored hourly to daily during the largest flows in 2006 and 2007 except for the two headwater weirs that had limited access during the snowmelt period (Table 2.2). To increase the representativeness of the study period, hence have a more robust basis for assessing hydrologic controls on annual sediment yield (see Chapter 5.3.3), monitoring of total yield was extended to 2008. The efficiency of the removable pit traps in comparison with other sampling devices (e.g., Helley-Smith sampler) is not known. Sterling and Church (2002) established that 29 cm-wide cylindrical pit traps used in Harris Creek reliably collected fine sand (0.25 mm) for discharges less than 4 m3/s. Peak discharges at the pit trap sites ranged from 0.3 m3/s to 1 m3/s and sediment captured in the traps at the highest flows included all available bed material sizes, from small boulders to sediment finer than 2 mm, which is the grain size cut-off. During the largest discharge events pit traps were emptied frequently to avoid these being filled to capacity, hence compromising the estimate of bedload yield. Given the relatively low peak discharges and sediment yields in the study streams and the ability of the traps to capture the full range of sediment sizes these traps are believed to be a reasonably reliable tool of monitoring rates of bedload transport.   85 a  b  Figure 5.3. (A) Sediment basket at site E2 (0.45m x 0.3m x 0.15m) and (B) channel spanning weir at site C3 (2m x 0.7m x 0.6m). Odyssey electronic water level recorders at six locations including five of the eight sediment monitoring sites provided a continuous record of water level during the study period (Table 6). At these sites discharge was calibrated with water level annually through salt dilution (Day, 1976; Merz and Doppmann, 2006) gauging over the full range of flows (Szeftel, 2011). Salt dilution gauging is a simple, accurate method for estimating discharge in turbulent mountain streams with discharges less than 5m2/s. This method is based on the principle that a volume of water moving past a channel section over a period of time can be estimated based on the concentration of a known mass of salt that is diluted in the flow. Concurrent measurements of discharge and channel cross-sectional area were collected along with sediment load at the lower six monitoring sites through three freshets. Discharge was also measured by salt dilution gauging at a supplementary location (C5 in Figure 2.2a)  86 downstream from the sediment monitoring sites. Estimated uncertainties for collected field data including channel survey data and discharge measurements can be found in Appendix 1.  Field-based measurements of discharge and channel cross-sectional area at the monitoring sites allowed (i) developing an ad hoc area-discharge relation for the study area, and (ii) calculating cross-sectional water velocity over the full range of flows, hence yielding an estimation of total stream power (?) and specific stream power (?), which are a good reference for those power indices derived elsewhere in the ungaged study reaches from Manning's back-calculated discharges.  Manning`s n values were calculated using direct measurements of bankfull discharge, water surface slope at bankfull stage, and bankfull channel cross-sectional area. Calculated Manning`s n values were then applied to other reaches with similar morphologies and, together with measured bankfull width and depth, were used to estimate bankfull discharge (Qbf) at ungauged locations (Figure 2.2a). Water surface slope was also used to calculate at-a-station boundary shear stress (?o). The area-discharge relation is useful in that it shows how appropriate it is to use drainage area as a proxy for water discharge (e.g., in downstream hydraulic geometry relations). In Cotton Creek, the scaling exponent denotes a virtually isometric relation (i.e., Q = 0.07A1.01; R2 = 0.89), ruling out the existence of non-linear trends. The intercept, a measure of the stream bankfull flow for a drainage area of 1km2, shows a relatively low value compared to humid mountain environments (e.g., Emmett, 1972; Wohl and Wilcox, 2005) but is comparable to values reported for semi-arid, formerly glaciated watersheds in  87 both Idaho and Colorado (e.g., Whiting et al. 1999; Mueller and Pitlick, 2005; Muskatirovic, 2008). The combined analysis of discharge and sediment yield at the monitoring stations has been used to: (i) examine the temporal variability of bedload yield in relation to water discharge (Figure 5.9); (ii) quantify threshold discharge (Qt) for bed mobility (Figure 5.4); and (iii) analyse bedload response across channel morphologies (Figure 5.10). Threshold discharge was identified via stepwise regression (Ryan and Porth, 2007). Specifically, it was defined as the intersection (or breakpoint) of the linear relations describing low levels of bed mobility, also termed Phase 1 transport, and high levels of mobility or Phase 2 transport (Jackson and Beshta, 1982; Ryan et al., 2005; Figure 5.4). The breakpoint is selected iteratively so that combined R2 and the relevant p-value between the two regression trends are respectively maximized and minimized (Table 2.3). At sites C1 and E1, since these were not monitored frequently, Qt is estimated as largest peak discharge that did not initiate substantial bedload movement (i.e., Gomi and Sidle, 2003).   88  Figure 5.4. Example break point analysis for sediment yield at sites (a) E4 and (b) C4. Finally, to gain further insights on the mechanisms modulating fluvial sediment transport in the study streams, the effects of channel morphology on sediment yield (Chapter 5.3.3) are examined in three of the intensively monitored sites: C2 (forced step-pool), E4 (step-pool), and C4 (riffle-pool). The combination of sites (i.e., including their upstream 20-m channel stretches) containing distinct sequences of channel units, in conjunction with the range of flood magnitudes recorded in the 2006-07 period represents a suitable set up for assessing the variability of channel response to floods.  To this purpose, daily sediment yield (in kg day-1) is plotted against daily peak discharge (Qd) expressed as a ratio of the threshold discharge (Qt) for initiation of Phase 2 transport (Figure 5.8) (Jackson and Beschta, 1982; Warburton, 1992).  89 5.3 Results 5.3.1 Basin structure, channel morphology, and geometry Although the entire physiographic area of the southern Columbia Mountains has been repeatedly overridden by the Cordilleran Ice Sheet, morphological evidence indicates that glacial erosion was chiefly focused on main valley glaciers (e.g., Moyie Lake; Figure 2.2a) flowing along the major structural lineaments (Clague, 1989). The lack of well-defined glacial cirques across the southern Purcell ranges of Columbia Mountain (e.g. Yahk and Moyie Ranges) suggests that ice-flow in tributary hanging valley glaciers was relatively stationary. Cotton Creek provides a good example of the foregoing landscape history. Its topography is subdued, with a complex longitudinal profile that exhibits concave-up, straight, and convex portions (Figure 5.5a). This pattern translates into a fragmented area-slope transect (Figure 5.5b). Morphologically, the profile can be classified into three main components including an upper valley wall (or upland area), a hanging valley, and a valley step that terminates at Moyie Lake. Since the glacially-imposed slopes are gentle to moderate, Cotton Creek lacks topographic boundary conditions that can support mass wasting activity (i.e., debris slides and debris flows) (Jordan, 2002). It follows that inflections in the area-slope relation do not imply process replacement or transition (e.g., Brardinoni and Hassan, 2006). In particular, headwater tributary confluences are not synonymous with debris flow fans so that mechanisms of hillslope to channel sediment transfer are limited to localized rock-topple and bank undercutting distributed along the length of the stream network.  90  Figure 5.5. (a) Longitudinal profiles with landform and channel morphology and (b) Channel gradient versus drainage area with landform morphology. In view of the above glacial palimpsest, process domains include headmost colluvial reaches (C, Figure 5.5a) (sensu Montgomery and Buffington, 1998) that have a disturbance regime distinct from the colluvial analogues typical of steepland areas (e.g., Benda and Dunne, 1997; Jakob et al., 2005). They are characterized by moderate slopes (0.03-0.41 m/m; Figure 5.5b), so that sediment supply is dominated by soil creep, local bank instabilities, and tree throw, with no evidence of debris-flow activity. The spatial distribution of fluvial channel morphologies is controlled by local slope, wood abundance and unlimited availability of glacially-derived sediments (Figure 2.2b).  91 Bank undercutting and tree throw (Swanson et al., 1982, 1998) being the dominant mechanisms of sediment supply in fluvial reaches, promote high wood delivery to streams so that plane-bed morphology is highly localized (no discrete reaches) and forced step-pools prevail (9 out of 18 fluvial reaches). Steeper reaches along lower Elk Creek and along the valley step present step-pool and boulder-cascade (SP-BC) morphologies, while riffle-pools (RP) are prevalent in the gentler reaches of the hanging valley floor (Figure 5.5a and Figure 5.6a). Surface Dmax exhibits generalized downstream coarsening both along Elk Creek (open symbols) and Cotton Creek (closed symbols; Figure 5.6b). This systematic increase in grain size contrasts with reports of D50 from mountain streams of the Washington Cascades (Brummer and Montgomery, 2003) and D84 from New Zealand (Wohl and Wilcox, 2005) which show downstream coarsening on debris flow-dominated reaches and downstream fining along fluvially-dominated ones. Heller et al. (2001) attribute the lack of downstream fining in distal reaches of the Hoh River, Washington, to the continuous resupply of coarse glacial material at cutbanks and tributary junctions. The glacial sedimentary imprint in Cotton Creek (Figure 2.2b) supports this hypothesis.  92  Figure 5.6 Channel-reach morphology plotted by drainage area versus (a) slope; (b) Dmax; (c) shear stress; (d) Shields stress; (e) specific stream power; (f) total stream power; and (g) number of LWD pieces per unit channel length. Open and closed symbols represent reaches in Elk and Cotton Creek respectively. Shear stress (?) and specific stream power (?) across the study reaches do not show any systematic variation with drainage area (hence with discharge). The former variate plots consistently between 85 and 330 N/m2 (Figure 5.6c), the latter spans between 120 and 320  93 Watts/m2 (Figure 5.6e). In contrast, total stream power displays a well-defined downstream increase across channel types (Figure 5.6f). This finding contrasts with previous reports (Wohl and Wilcox, 2005) that have shown progressive downstream increase of total power along debris-flow dominated reaches and monotonic decline along fluvially-dominated counterparts. Given that both total power and Dmax increase with contributing area, it follows that Dmax increases as a function of total stream power (Dmax = 2*10-3 ?0.61; R2 = 0.55). A similar relation between D50 and total stream power has been reported for larger drainage basins (20-380 km2) of central and northern Idaho (Whiting et al., 1999). Shields stress (Figure 5.6d), even though characterized by high scatter, due to the downstream increase in Dmax, declines progressively downstream. In particular, the area-Shields stress relation yields a good discrimination between colluvial and alluvial reaches. The number of LWD pieces per unit channel length (LWD density) shows a complex pattern (Figure 5.6g). Values plot between 0.14 and 0.37 along the confined colluvial and fluvial reaches (valley width range: 5m-15m) located upstream of the Elk-Upper Cotton confluence. Downstream, in riffle-pool-bar morphologies of the unconfined hanging valley floor (valley width range: 30m-100m) greater total stream power and valley-side storage opportunity impart a further increase of LWD density, which peaks to 1.06. Finally, when Cotton Creek enters the relatively steep and entrenched step-pool/boulder-cascade reaches of the valley step, drastically reduced wood storage conditions and a further increase in transport capacity (i.e., total stream power) impart an abrupt decline in LWD density (0.04-0.11).  Downstream hydraulic geometry relations (DHG, Figure 5.8) in Cotton Creek are statistically robust (note high R2 values) and show scaling exponents that are in good  94 agreement with the relation originally shown by Leopold and Maddock (1953). This finding suggests that lowland DHG relations can apply to mountain streams, with abundant LWD structures and glacially-imposed slopes, as long as mass wasting disturbance is not active.  Figure 5.7 Downstream hydraulic geometry across the study reaches. Black symbols indicate reaches with field-measured discharges. Inter-basin variability between Elk Creek and Upper Cotton Creek is minimal in terms of channel slope, total stream power, and LWD density (Figure 5.5a and Figure 5.7). However, differences are evident with respect to Dmax and correlated variables, such as Shields stress (?/Dmax) and relative roughness (Dmax/bankfull depth). In particular, along fluvial reaches Elk Creek displays consistently higher surface Dmax and relative roughness than Upper Cotton Creek, while Shields stress plots systematically lower.  5.3.2 Daily bedload yield Bedload mobilization typically occurs with increasing stream flow early in the spring snowmelt period and continues episodically throughout the spring peak flow period, terminating during the final recession flows in mid-summer (Figure 5.8). Flow duration analysis indicates that discharges capable of mobilizing bedload occurred between 0% to  95 25% of the time, depending on the station, between the beginning of March and the end of June. No bedload movement was observed during the seasonal low-flow period between the months of July to March. Examination of the data at the stations with higher temporal resolution (i.e., E4 and C4) in 2006 and 2007 shows (i) an intuitive direct dependence of daily yield on water discharge (Q); and (ii) high variability in temporal patterns of Q in the two seasons monitored (Figure 5.8). Both the 2006 and 2007 freshets included multiple sediment mobilizing peak flows. In particular, the structure of the 2007 hydrograph is regarded as typical of a spring freshet in the southern Columbia Mountains (i.e., Whitaker et al., 2002; Jordan, 2006). Maximum Q in 2006 was the largest recorded in the four study years and is estimated to be a 10-year flood based on the closest (50 km) long-term gauging site that displays similar physiographic characteristics and meteorology (WSC station #08NH016). Bankfull level was also reached in 2007 at 7 of the 9 monitoring sites, in this case maximum Q was estimated as a 2-year flood.  96  Figure 5.8 Water discharge hydrograph with daily specific yield and associated Shields stress at monitoring sites C4 and E4. Dashed lines indicate Shields stress at bankfull. Note scaling factors for specific bed load yield in the secondary y-axes.  97 Higher daily yields tend to mimic the temporal pattern of Q peaks. However, daily yield response is complex: largest yields are not necessarily associated with the seasonal maximum discharge and vice versa Q peaks do not necessarily generate sediment mobilization. For example, in both study streams the second freshet peak of 2006 produces the greatest daily yield even though this event was not the largest of the hydrological season (Figure 5.8).  Analysis of bedload yield associated with rising and falling limbs of individual Q peaks documents that rising limbs (including the peak) mobilize far more sediment than the corresponding falling limbs (Figure 5.9). Differences are especially large for Elk Creek, in which the median mobilized mass during the rising component of a flood event is about one order of magnitude greater than the falling one (i.e., 100 kg vs. 10 kg). In terms of inter-basin comparison, while the size variability of sediment pulses in the two study streams largely overlap during the falling limb, in Elk Creek rising limb-driven sediment transport is about three times higher than in Upper Cotton Creek (Fig. 10).  Figure 5.9 Box-plot showing the size distribution of transported mass during the rising and falling limb in Upper Cotton Creek and Elk Creek.  98 The mechanism responsible for higher bedload transport during the rising stages of the hydrograph becomes clear when considering the Shields stresses associated with mobilization of bed material. At both E4 and C4 sudden increases of bedload are observed once Shields stress (based on the surface D50) approaches bankfull (?*bf). As discharge starts to recede there is a corresponding drop in bedload yield even when Shields stress remains above bankfull (Figure 5.8). The fact that ?*bf values in C4 and E4 (i.e., respectively 0.05 and 0.11, Figure 5.8) lie within the range associated with the destabilization of the amour layer in similar snowmelt-dominated streams of North America (i.e., Mueller et al., 2005; ?*bf (0.05-0.128), H/D50 <5), suggests that armoring must play a prominent role in terms of sediment supply. The dynamics associated with the supply and subsequent exhaustion of bedload material are examined in greater detail in the next Chapter. 5.3.3 Variability of bedload yield across channel morphologies The presentation of bedload sediment dynamics in forced step-pools (C2), free-formed step-pools (E4), and riffle-pools (C4), proceeds from 2007, the average flood year, then continues with the assessment of the more severe hydrological conditions of 2006 (i.e., 10-year flood). In 2007 distinct yield-discharge relations are observed in the three channel types. In riffle pools (C4) rising and falling stages plot approximately along the same discharge-yield relation (Figure 5.10a) indicating adequate sediment supply (no exhaustion) throughout average floods. Free-formed step pools (E4) seem to behave somewhat differently, in that modest clockwise hysteresis are noted (i.e., sediment yield in the rising stage is higher than that associated with the falling stage) for QdQt-1 < 1.5 (Figure 5.10b). In snowmelt gravel-bed streams clockwise hysteresis in bedload yield at the seasonal scale has  99 been attributed to depletion of the available sediment mobilized during the rising limb of the hydrograph (Nanson, 1974; Moog and Whiting, 1998).   Figure 5.10 Daily bed load yield versus daily peak discharge (Qd) expressed as a ratio of the threshold discharge (Qt) for initiation of Phase 2 transport during rising stage (solid circles) and falling stage (open squares) periods. (a) C2 in 2006; (b) C2 in 2007; (c) E4 in 2006; (d) E4 in 2007; (e) C4 in 2006; and (f) C4 in 2007. Lastly, at the forced-alluvial site (C2) delayed mobilization of sediment is observed during the rising stage that results in a counter-clockwise hysteresis (Figure 5.10c). These bedload dynamics reveal substantial channel-bed stability conditions so that very little sediment can move until break-up of the surface armor layer occurs at flows approaching  100 bankfull. In particular, higher transport rates across the falling limb period are associated with the mobilization of fine gravels, previously sheltered by armoring, which remain mobile after the bed surface has re-stabilized (Reid et al., 1985; Lisle, 1986). In 2006, while the behaviour of the riffle-pool channel (C4) (Figure 5.10d) mimics substantially that observed in 2007, the other two sites exhibit a clockwise hysteresis pattern. Namely, the hysteresis loop in the forced-alluvial site (C2) (Figure 5.10f) has a larger amplitude compared to the purely alluvial one (E4) (Figure 5.10e), indicating more pronounced conditions of sediment exhaustion during the falling stage.  Comparisons of average rising limb and falling limb grain-size distributions in C2, C4, and E4 provide further insights while corroborating foregoing interpretations on the mechanisms controlling bedload yield variability (Figure 5.11). In the riffle-pool site (C4) the grain-size distributions associated with the rising and falling limbs are nearly identical in both years (Figure 5.11c). These patterns, showing that rising and falling stages lie approximately on the same discharge-yield relation, point to sediment dynamics typical of transport-limited conditions. By contrast, in the coarse-grained step-pool site (E4) rising stages mobilize consistently coarser bed-load fractions than the falling stages (Figure 5.11b). This outcome is attributed to the effect of exhaustion of sediment supply in conjunction with the reduced mobility of the coarsest fraction as the armor layer re-forms for QdQt-1 < 1.5.  Sediment dynamics at the forced-alluvial site (C2) in 2006 and 2007 appear to be the most diverse. This behaviour is attributed to the presence of LWD. In 2007, loads transported during falling limbs are finer than the rising counterparts, a pattern consistent with increased mobilization of fines following the armor layer break-up. In 2006, where the  101 largest difference between rising and falling limb yields was observed (Figure 5.11a), coarser textured falling limb loads indicate a depletion of fines following the large magnitude flood event (open squares, Figure 5.11a). According to morphological changes observed in the field, the 10-year flood event was capable of mobilizing material stored behind LWD; material that was not accessible to stream flow in 2007.  Figure 5.11 Grain-size distribution of bed loads associated with rising and falling stages at sites: (a) C2; (b) E4; and (c) C4.  102 5.3.4 Morphological and hydrological controls on annual sediment yield In order to decipher postglacial sedimentary responses in headwater alluvial systems and assess relevant hydrological controls, sediment yield is examined at much coarser temporal (annual) and spatial (regional) scales.  In the two study basins specific sediment yield (SSY; expressed in kg km-2day-1) tends to increase with drainage area (Table 2.3). During the study period, Elk Creek displays consistently higher SSY than Upper Cotton Creek. Previous monitoring in headwater tributaries of Gold Creek, a basin that shares its western divides with Upper Cotton Creek, indicate that bedload and suspended load exhibit a 1:6 ratio (Jordan, 2006). By applying this ratio to the bedload yields of the study sites we obtain a first-order estimate of suspended yield (Figure 5.12; given the intrinsic spatial variability associated with suspended sediment yield, 50% error bars are included). These estimates, which plot along the area-yield scaling relation defined by larger basins (50-30,000 km2) of the Columbia physiographic region (Church et al., 1999), indicate a sedimentary pattern of downstream degradation, ultimately the result of contemporary fluvial reworking of inherited glacial and glaciofluvial deposits (Church and Slaymaker, 1989). Within the context of formerly glaciated landscapes, an evaluation of the hydrological effects on downstream patterns of sediment yield has received limited investigation. In the next Chapter the extent that hydrologic forcing can control annual bedload yield (ASY; kg/km2) is explored. To this purpose, a multiple step-wise regression is conducted. Hydrological variables include: (1) magnitude of the maximum annual peak flow measured as the ratio of annual peak discharge to threshold discharge for bed mobility (QpQt-1) (Table 2.3); (2) duration of flows over threshold discharge (Dr); (3) number of  103 peaks over threshold discharge (nPk); and (4) percent exceedence of flows above threshold discharge (Ex).   Figure 5.12 Suspended sediment yield as a function of drainage area for the Columbia Mountains physiographic region. Filled symbols are yields published by Church et al. (1999); open symbols are yields estimated at the eight monitoring sites in Cotton Creek. Bars indicate 50% error around estimated values (Jordan, 2006) Simple correlation analysis (Table 5.1) shows that annual sediment yield is chiefly correlated with the number of peaks over threshold. We also note that virtual complete correlation exists between 'Duration' and '% exceedence', and that these two variables  Table 5.1 Correlation matrix   % Ex Dr  nPk Qp/Qt ASY % Ex 1     Dr  0.9918 1    nPk 0.5977 0.6406 1   Qp/Qt 0.7485 0.7476 0.4592 1  ASY 0.4329 0.4597 0.8019 0.3904 1 correlate well with QpQt-1. ?Duration? was excluded from stepwise regression analysis due to non-normally distributed residuals, even after a number of transformation attempts. The results of the stepwise regression indicate that nPk alone is the strongest predictor of ASY (Model 1, Table 5.2). Model strength does not improve after including QpQt-1 (Model 2), unless one removes the two colluvial sites (E1 and C1) from the regression (Model 3).   104 Table 5.2 Stepwise multiple regression analysis results Model  Regression equation n R2 (adj) F Value P value (adj) 1 ASY = -5.6 + (14.6*nPk)  27 0.63 F=45.02 p<0.001 2 ASY = -7.4 + (1.4*Qp/Qt) + (14.4*nPk) 27 0.61 F=21.67 P<0.001 3 ASY = -16.2 + (9.4*Qp/Qt) + (14.2*nPk) 20 0.65 F=18.53 p<0.001 It appears that only one other study (Ouimet and Deither, 2002) has recognized the influence of nPk in explaining annual variability in sediment yield. The study investigated controls on sediment flux in Birch Brook Creek, a headwater basin in northwestern Massachusetts, whose annual hydrograph is characterized by numerous rain-on-snow and rain-only peak events that occur throughout the year. Using simple linear regression, the authors test the strength of bivariate relations between annual bedload yield and meteorological and discharge variables (i.e., annual precipitation, duration of flows, and number of peaks over bankfull flow). Study results indicated that number of peaks over bankfull discharge and duration of flows in excess of bankfull discharge respectively explained 70% and 82% of the variability of annual bedload yield however, duration of discharge was estimated as a function of peak discharge magnitude rather than measured directly. The variable ?number of days in excess of Qbf ?, a more direct estimate of duration of flows above threshold discharge, explained only 35% of the variability in annual yield. 5.4 Discussion This study characterizes the linkages between landscape structure, channel-reach morphology, and bedload sediment dynamics in two forested headwater streams of the southern Columbia Mountains, Canada. In so doing, it complements prior studies conducted in formerly glaciated larger fluvial systems of Idaho (e.g., Whiting et al., 1999). The combination of bedrock geology, tectonic activity, but chiefly Pleistocene glaciations imposes local slope and valley width, which in turn modulates stream channel  105 morphodynamics including the spatial distribution of morphological variates and temporal and spatial characteristics of sediment yield. Morphological and sedimentological evidence indicates that during the Last Glacial Maximum erosion associated with advancing glaciers primarily occurred along the major valleys, while intervening upland areas experienced a gradual build-up of ice as valley glaciers coalesced (Ryder, 1981). As a result, this region of the Canadian Cordillera is characterized by a subdued, plateau-like topography, in places interrupted by major, entrenched valleys (i.e., relict glacial troughs). Even though the glacial palimpsest generates stepped longitudinal profiles (e.g., Cotton Creek), inflections in the area-slope space do not impart geomorphic process transitions (Figure 5.5) and the landscape is fluvially dominated, with mass-wasting activity relegated to the vicinity of a few highest peaks. In contrast with the steep, rugged terrain of coastal British Columbia (Brardinoni and Hassan, 2006; 2007), colluvial channels are confined to the headmost tips of the drainage network and do not support debris-flow activity, so that the significance of this geomorphic process at the local and regional scales is negligible. The number of LWD pieces per unit channel length increases downstream with increasing stream power as long as the channel is unconfined and provides room for wood storage on adjacent riparian areas (i.e., hanging valley floor of Cotton Creek). Where step-pool and boulder-cascade morphologies are entrenched in high-power reaches (i.e., valley step), wood is less abundant likely as a result of being broken up and transported downstream or remains inactive (overhanging), so that the number of LWD pieces per unit channel length drops dramatically (Figure 5.7g).  106 The spatial distribution of channel-reach morphology is controlled by local slope, in conjunction with wood abundance. Local slope, imposed by relict glacial landforms, produces the classic boulder-cascade/step-pool/plane bed/riffle-pool sequence (Montgomery and Buffington, 1997), which is reset by the presence of hanging valleys (Brardinoni and Hassan, 2007). This sequence is further altered by the abundance of active LWD pieces, in turn modulated by channel confinement and total stream power. LWD abundance dictates: (i) which channel types are dominant (i.e., forced step-pools) and which are excluded (i.e., plane bed reaches; Figure 5.6a); and (ii) the spatial distribution of forced morphologies so that for the reasons detailed above, forced step-pools do not appear downstream of the hanging valley. Previous studies conducted in steep mountain streams have shown that the area-slope inflection marking the transition between colluvial (debris-flow) channels and downstream fluvial analogues imparts inflections in downstream trends of stream power indices and sediment caliber (Brummer and Montgomery, 2003; Wohl and Wilcox, 2005). In these study creeks, the composite area-slope trend does not affect the downstream variation of total stream power and Dmax, and both variables exhibit monotonic increasing trends with drainage area. According to the foregoing comparison, area-slope inflections would be effective in modulating stream power and maximum sediment caliber only when the threshold of debris-flow initiation/transport (i.e., slope > 40 percent) is crossed, a threshold supported by extensive field-based data (Fannin and Rollerson, 1993; Jordan, 2002). Given that both total power and Dmax increase with contributing area, it follows that Dmax increases as a function of total stream power (Dmax = 2*10-3 ?0.61; R2 = 0.55).   107 DHG relations in the study area are well developed and are insensitive to area-slope inflections (cf. Figure 5.5b and Fig. 8). In particular, discharge-width and -depth scaling relations in Cotton Creek are statistically stronger (0.75 < R2 < 0.92) than those reported for basins of comparable size, characterized by debris-flow dominated headwaters (Brummer and Montgomery, 2003; Wohl and Wilcox, 2005; Brardinoni and Hassan, 2007). Interestingly, average ?/Dmax (i.e., 3,600 kg/s3) plots well below the 10,000 kg/s3 threshold condition proposed by Wohl (2004) for attaining well developed DHG relations in mountain channels. This outcome indicates that such energy-based limit to DHG does not hold for Cotton Creek and possibly for similar fluvially-dominated headwater channels of this physiographic region. DHG scaling exponents roughly conform to those originally reported by Leopold and Maddock (1953), as well to more recent ones obtained for larger fluvial systems located within the Columbia River basin, Idaho (Whiting et al., 1999). In this regard, the present work extends the validity of fluvial DHG relations to smaller and steeper fluvial systems, characterized by more complex topography and coarser beds.  In Cotton Creek while the spatial distribution of free-formed reach morphologies is chiefly controlled by local slope, downstream patterns of the coarse grain-size fraction, bankfull width, bankfull depth, and stream power are all insensitive to systematic changes of glacially imposed slope: an indication that these variables are adjusted to the contemporary snowmelt-driven water and sediment transport regimes.  The spring (March-June) snowmelt hydrograph of the Columbia Mountains typically displays a number of distinct peaks that are triggered by periods of warm weather and, to a  108 minor extent, by rain-on-snow events (McCabe et al., 2007). In the study sites, discharge peak events on average are capable of triggering bedload transport for about 12% (11 days) of the spring snowmelt period, with average specific yields ranging between 0.2 and 2 kg/km2 per day. Examination of sediment transport at finer resolution (shorter temporal scales) shows that peak flows of a given magnitude are not necessarily associated with bedload transport and higher yields are observed on rising limb of peak flows compared to those associated with falling limb counterparts. Differences in bedload mobility between rising and falling flood discharges have been reported elsewhere (Wohl, 2010). In steep mountain channels subject to landslide sediment inputs and debris-flow activity, the decrease in bedload yield following large floods is typically interpreted as the result of in-channel sediment depletion/exhaustion (supply-limited conditions) (e.g., Nanson, 1974; Alvera and Garcia-Riuz, 2000; Gomi and Sidle, 2003). In forested streams where sediment is supplied from channel banks or upstream reaches, lower yields associated with falling hydrographs have been related to the break-up and subsequent re-establishment of a channel-bed armor layer (e.g., Sidle, 1988; Kuhnle, 1992). Finally, in low-gradient alluvial systems, depletion or seasonal exhaustion of sediment availability has also been invoked as the cause of decreased bed mobility following peak flows (Dietrich et al. 1989; Moog and Whiting, 1998).  Comparison of relevant Shields stress values with published data from snowmelt-dominated streams in western North America (Mueller et al., 2005), provides quantitative confirmation that in alluvial reaches of Elk and Upper Cotton Creek (E4 and C4) the discrepancy between sediment yields during rising and falling stages of the hydrograph is  109 modulated by the break-up and subsequent re-development of the surface armor layer (Figure 5.9).  A closer look at sediment dynamics in three representative sites characterized respectively by forced step-pools, and free-formed step-pools and riffle-pools allows detailing causal linkages between local morphology and channel response (i.e., bedload transport) to imposed discharges. Examination of daily bedload yield expressed as a function QdQt-1 (daily peak discharge normalized by threshold discharge) is instructive (Figure 5.10). The disparity in hydrological forcing during the two years of monitoring:  (i) had virtually no effect in the riffle-pool site (C4), with rising and falling bedload yields displaying comparable magnitudes. This pattern is characteristic of transport-limited systems.  (ii) imparted in free-formed step-pools (E4) a clockwise hysteresis loop (indicative of sediment exhaustion dynamics), whose amplitude was proportional to the recurrence interval of the peak flood.  (iii) induced in forced step-pools (C2) a hysteresis reversal, from a distinct counter-clockwise pattern in 2007 to a wide clockwise loop in 2006. In the latter case, the 10-year flood event has made available material stored behind and above LWD, which is typically inaccessible in average flood years (supply-limited conditions).  The foregoing channel responses agree with the conceptual framework that considers channel-reach morphology in mountain streams a qualitative expression of the competition between water/sediment supply and transport capacity (Montgomery and Buffington, 1997,  110 1998). In this perspective, riffle-pools are transport-limited systems whereas step-pools and forced step-pools are increasingly limited by sediment supply. Annual bedload specific yield in Elk and Upper Cotton Creek increases as a positive power function of drainage area. Similar trends, which have been reported for suspended yield across larger fluvial systems (~102-104 km2) of British Columbia (e.g., Church and Slaymaker, 1989; Church et al., 1999), are diagnostic of ongoing degradation. This sedimentary pattern has been explained as the result of contemporary fluvial reworking (i.e., mainly bank erosion) of paraglacial sediment surplus (i.e., valley fills) emplaced about 10,000 years ago, during the last continental deglaciation. This interpretation is supported by the spatial distribution of glacial (ablation till), sandy gravelly glaciofluvial (kame terraces) and sandy gravelly fluvial deposits though which the study streams are actively eroding (Figure 2.2b). Based on bed- and suspended-load monitoring efforts conducted in headwater streams adjacent to Cotton Creek (Jordan, 2006), a first-order approximation of the annual suspended load in Elk and Upper Cotton Creek is obtained. Derived total yields plot along the same scaling relation described by larger basins of the Columbia Mountains region (Figure 5.12; Church et al., 1999). Since tree throw and bank erosion are the main mechanisms of sediment supply from till-mantled terrain to both Elk and Upper Cotton Creeks, these outcomes provide further empirical support to that original explanation (i.e., Church and Slaymaker, 1989) and extend it to fluvial systems as small as 2 km2. Stepwise regression analysis, conducted to examine the temporal variability of annual sediment yield in relation to water discharge, suggests that bedload yield is chiefly  111 related to the number of peaks over threshold (nPk, Table 5.2) rather than the magnitude of peak discharge (QpQt-1).  The number of peaks over threshold in Elk (nPk = 42) and Upper Cotton Creek  (nPk = 23) also explains the higher bedload yield recorded in the former system (Table 1) across the 2006-08 study period. In this regard, a crucial question needs be adressed: what controls the inter-basin variability of nPk? Recent studies investigating snowmelt dynamics in mountain environments indicate that, at the basin scale, differences in topographic conditions can combine to generate different snowmelt responses (Ellis et al. 2010; Pomeroy et al. 2011).  Therefore, a first-order explanation of nPk variability can be pursued via analysis of basic topographic variables such as elevation (i.e., hyspometric curve), slope, and aspect (Figure 2.3). In regards to elevation, previous work have shown that: (i) peak flows in mountain basins are chiefly controlled by snowmelt occurring at the H20-H60 elevation band (that is, 20-60% basin upper portion); and (ii) snowmelt from the headmost areas of a watershed does not contribute substantially to hydrograph peaks, due to poor surface flow connectivity to the main channel network (Gluns, 2001; Whitaker et al., 2002). Given that Elk and Upper Cotton Creek have virtually identical hypsometry in the H20-H60 elevation band (Figure 2.3a) it is concluded that elevation likely does not affect nPk inter-basin variability. The same conclusion applies to slope gradient, as Elk and Upper Cotton share similar slope distributions (Figure 2.3b). Slope aspect is the topographic factor that better explains the complex hydrologic response observed in the two study tributaries. Natural shading on north-facing slopes, prevalent in Upper Cotton Creek (Figure 2.3c), would result in relatively slow snowmelt  112 driven by longwave radiation (Jost et al., 2007; Ellis et al., 2010), hence dampening the effects of any external meteorological perturbation on snowmelt rates. In contrast, terrain with prevalent south-facing components (i.e., Elk Creek) by experiencing faster snowmelt, dominated by direct shortwave radiation, would be more sensitive to meteorological fluctuations, generating peaks over bedload threshold more frequently. This interpretation is supported by a recent detailed empirical investigation of aspect and slope effects on runoff dynamics in Cotton Creek (Smith, 2011). The importance of hydrograph variability on sediment yield agrees with previous reports showing poor correlation between peakflow magnitude and annual yield (e.g., Warburton, 1990; Troendle and Olsen, 1994; Ryan and Dixon, 2008). Temporal and spatial variability of bedload yield in such studies has been interpreted as the result of differences in bedrock geology and/or sediment availability. Although these factors must be taken into consideration, data from Cotton Creek indicate that, in average runoff years, the temporal and spatial variability of bedload yield mainly derives from differences in hydrograph complexity. Specifically, while annual variability at a site arises from seasonal differences in meteorologic conditions during the snowmelt period, spatial variability derives from differences in basin morphometry that affect the hydrologic response to seasonal temperature fluctuations.     113 6 Patterns of bedload entrainment and transport in forested snowmelt streams 6.1 Introduction Steep headwater streams constitute the majority of the mountain drainage network, and as such are prominent geomorphic players in the steepland sediment cascade (Gomi et al., 2002; Benda et al., 2005; Church, 2010). Forested streams are particularly complex owing to the stochastic nature of water, sediment, and wood inputs, and the temporary character of in-channel sediment stores (Benda et al., 1998; Gomi et al., 2001; 2002; Faustini and Jones, 2003; Hassan et al., 2005). Knowledge of critical thresholds for entrainment of bedload as well as governing relations between discharge and bedload flux is necessary for modelling channel response potential in problems of sustainable watershed management, aquatic ecology and hydraulic engineering.  In simplified laboratory conditions, critical shear stress for particle entrainment (?c) on a bed of uniform grains is a function of the grain diameter (D) and a constant known as the dimensionless critical shear stress (Shields parameter, ?c ?0.06) (Shields, 1936). However, application of this constant to streams with heterogeneous bed surfaces typically results in over-prediction of bed mobility due to the hiding effects associated with coarse grains (Einstein, 1950; Fenton and Abbott, 1977).  Entrainment is further complicated by the occurrence of a surface armor that inhibits some or all of the mobile bed until critical thresholds of shear stress are exceeded (Parker et al. 1982; Parker and Klingemann, 1982; Andrews, 1983; 2000; Church et al., 1998; Parker and Toro-Escobar, 2002; Parker, 2008; Frey and Church, 2009). In this research context, a  114 large number of field and flume-based studies have endeavoured to establish the degree of size dependence in entrainment thresholds. While some studies have found near equal-threshold entrainment across all mobile grain sizes (e.g., Andrews, 1983; 1994; Wilcock and Southard, 1988; Ferguson et al., 1989) others have determined that at least some degree of size independence is observed (e.g. Komar, 1987; Ashworth and Ferguson, 1989; Kuhnle, 1993; Wilcock and McArdell, 1993; 1997; Whitaker and Potts, 2007). Differences in entrainment dynamics reported in past studies have been partly explained by recent studies showing that the Shields parameter (?c) is not constant but varies with relative and absolute grain size effects (Shvidchenko et al., 2001; Mueller et al., 2005; Bunte et al., 2010) relative roughness (Wilberg and Smith, 1987; Buffington and Montgomery, 1997) and increasing channel gradient (Shvidchenko et al., 2001; Mueller et al., 2005; Lamb et al., 2008; Bunte et al., 2010).  Following entrainment the mobility of a given grain size relative to the available bed material depends on governing conditions of streamflow until a second threshold is exceeded and the grain size becomes fully mobile (Wilcock and McArdell, 1993; 1997; Church and Hassan, 2002). In flumes and low gradient alluvial streams full mobility occurs once the shear stress reaches approximately twice the critical shear for entrainment of a given grain size (Wilcock and McArdell, 1993; 1997; Church and Hassan, 2002). Such a well-defined mobility threshold does not appear to apply to steep mountain streams where high flow resistance associated with step-pool structures limits the occurrence of equal mobility transport for all but the smallest grain sizes (<10mm) (Gomi and Sidle, 2003).  115 In steep streams, channel gradient, grain protrusion (i.e., grain resistance associated with hiding or armoring) and macro-roughness elements (i.e., morphological or form resistance due to large woody debris (LWD), cobble-boulder steps and cascades) function cumulatively to increase flow resistance which in turn decreases the flow energy available to transport bed material (Mueller et al., 2005; Wilcox et al., 2006; Lamb et al., 2008; Mao et. Al. 2008; Zimmerman, 2010; David et al., 2010; Ghilardi and Schleiss, 2011; Nitsche et al., 2011, 2012; Yager et al., 2007,2012). Consequently many bedload transport equations based on fully alluvial systems with low flow resistance over-predict bedload transport rates by several orders of magnitude when applied to steep mountain streams (e.g. Rickenmann, 1997; D?Agostino and Lenzi, 1999; Yager et al., 2007, 2012; Mueller et al., 2008). Recent findings of the influence of bed roughness and channel gradient on bedload entrainment and transport relations are primarily based on flume studies that have been scaled to approximate natural systems (e.g. Zimmerman; 2010; Yager et al., 2007; 2012). Few field-based studies have been conducted that can validate relevant findings (e.g. Bunte et al., 2010; Nitsche et al., 2011). Despite recent advances in bedload surrogate methods (e.g., Rickenmann et al., 2012), measuring bedload intensity in a steep mountain channel at high flows remains problematic due to both logistical and instrumental difficulties. As a consequence, field-based studies involving direct bedload sampling during flood events are scarce (e.g., Reid et al., 1985; Reid and Laronne, 1995; Bunte, 1996; Ryan and Emmett, 2002 Whitaker and Potts, 2007).  Snowmelt-dominated headwater streams of the Columbia Mountains, which are characterized by low unit discharges and seasonal flood events, provide a unique  116 opportunity to collect high resolution (hourly to daily) bedload samples to investigate factors influencing bedload entrainment and transport in natural fluvial systems. In this study we are interested in establishing whether differences in grain size distributions and macro-roughness characteristics across sites manifest as intrinsic differences in bedload entrainment and transport dynamics. In light of the most recent advances cited above we hypothesize that: (1) that entrainment dynamics should vary across morphologies in response to differences in channel gradient, flow resistance and grain size characteristics; (2) Patterns of relative mobility of bedload will vary with differences in form resistance ? morphologies displaying low form resistance should display equal mobility for a wider range of bedload grain sizes; (3) Dynamics of bedload transport should display differences in response to variations in flow resistance across morphologies.  We pursue these objectives by means of a nested monitoring network that includes seven gauging stations and six sediment traps on forested fluvial systems with mobile bedload fractions ranging from 1mm to 128mm. This experimental set up, differs from prior research campaigns (e.g., Lenzi et al., 1999; Rickenmann, 2001; Gomi and Sidle, 2003; Thompson and Croke, 2008; Mao and Lenzi, 2007) in that it was specifically designed to account for spatial heterogeneity both in terms of channel bed texture, which varies between streams, and morphology (Figure 2.6 and Table 2.3).  We begin by presenting patterns of bedload entrainment and relative bed mobility in relation to grain and form resistance to flow. We use grain size distributions and scaled fractional transport analyses to investigate patterns of bedload mobility with increasing discharge. Bedload rating curves provide a method to compare and contrast bedload  117 mobility versus discharge relations between the six sites. Finally, we examine bedload rating curves and evaluate the suitability of selected transport formulae for predicting the observed bedload rates in these forested headwater systems. 6.2 Methods A table documenting estimated uncertainties associated with field measurements is provided in Appendix 1. 6.2.1 Discharge gauging Water level was recorded continuously at six naturally constricted cross-sections using Odyssey capacitance water level probes on a 15 minutes time step (Figure 2.2, sites C1,C4, C5, E1, E2 and E4). Three of the continuous water level gauging sites correspond to bedload monitoring sites (E2, E4, and C4). For these three sites between seventeen and twenty-five discharge measurements made using salt dilution gauging (e.g., Day, 1976; Merz and Doppmann, 2006) were conducted for a wide range of flows. This allowed developing rating curves relating water level to discharge for every stream gauge. At the remaining three bedload sites (C2, C3, and E3) discharge was measured throughout the study period (minimum of fifteen times) via salt dilution gauging. Regression relations were developed using chronologically paired discharge values (n=15) for gauged and ungauged sites over the full range of flows to build a continuous record of discharge for the ungauged sites. Regression correlations (R2) between gauged and ungauged sites range from 0.95 (C3) to 0.88 (C2).  Bankfull flow was estimated at each monitoring site by salt dilution during the period when stream flow filled the channel to a level defined by vegetation and physical  118 profile breaks and highlighted earlier in the year with blue tree paint. Average flow velocity was estimated as discharge (Q, m3/s) divided by flow cross-sectional area (A, m2). The latter variable was measured frequently throughout the monitoring period (i.e., over the full range of flows) at fixed sites located 0.5 m upstream of the bedload monitoring stations. Specifically, flow depth was measured using a stadia rod (rounded to the nearest 0.5 cm) at 10 cm intervals along the cross section, hence flow cross-sectional area was calculated incrementally between intervals and summed over the width of the cross-section. 6.2.2 Bedload monitoring Bedload samples were collected at six rectangular pit traps (Figure 2.2, Table 2.2). Specifically, at sites C4 and E4 fixed, channel-spanning, geotextile lined pit traps were installed. At sites C2, C3, E2 and E3 removable rectangular baskets lined with geotextile were set into the channel bed and levelled with the bed surface. Depending on site location, the removable pit traps cover between 28 and 45 % of the active channel width (Table 2.2). A base basket securely fixed into the channel substrate allowed the sampling basket to be quickly removed and replaced. Fixed traps were emptied by way of a heavy-duty, fine (1mm) mesh sieve. All removable pit traps were placed in rectilinear channel segments, within planar glides located in the central portion of the channel, away from channel banks.  The efficiency of the removable pit traps in comparison with other sampling devices (e.g., Helley-Smith or Bunte samplers, Potyondy et al., 2010) has not been quantified. However, considering that: (i) 29 cm-wide cylindrical pit traps were found to reliably collect (compared to a Helley-Smith sampler) bedload material > 1 mm for discharges < 10 m3/s in Harris Creek (Interior British Columbia) (Sterling and Church, 2002); (ii) that peak  119 discharges at bedload sites ranged from 0.3 m3/s to 1 m3/s; and (iii) that sediment captured at the highest flows included sizes from small boulders (> 256mm) to sediment < 2 mm (the cut-off size); the uncertainty associated with removable traps appears to set within the error envelope of the published field-based bedload literature.  At sites equipped with removable (non-channel-spanning) traps, instead of plainly extending the measured bedload flux to the unmonitored portion of the channel cross section, this figure is scaled by a factor of 0.8. This adjustment is made on the basis of direct field observations (e.g., within the channel-spanning traps we noted significantly smaller amounts of bedload material deposited in proximity of the banks) and on the basis of prior empirical studies. In this context, Emmett (1980) shows that devices which sample only portions of the channel width typically over-predict bedload flux by 1.3 times.  To avoid the effects of short-term fluctuations in bedload transport rate (e.g., Klingeman and Emmett, 1982; Bunte, 1996) and to increase the probability of sampling larger grains, sample collection intervals ranged from hourly time steps, during high flows, to several days, at low flow regimes. During the largest hydrological events pit traps were emptied frequently to avoid infilling to capacity, hence data loss. In about 80% of the cases discharge variability during the sample collection period ranged from 0 to 15%. Higher discharge variability (i.e., up to 40%) is associated with a number of samples spanning periods of rapid fluctuation within 24 hr that marked the onset of higher sampling frequency periods.  High discharge variability during periods of rising stage leads to an under-estimation of transport rates because the rate of transport is averaged over the full sampling period  120 rather than attributed to the period of higher discharge. A sensitivity analysis conducted on 12 samples included in this study indicates that averaging across periods of high discharge variability reduces transport rates by, at most, a factor of 4. Due to similar sampling intervals adopted for each study stream, the bias towards lower-than-actual sampling rates for high variability periods is consistent across sites on a given stream.  It has been long acknowledged that in steep mountain streams bedload mobility is episodic yet strongly correlated to peak discharge (e.g., Bunte, 1996; Rickenmann, 1997; D?Agostino and Lenzi, 1999), because this study is concerned with the dynamics of bedload entrainment and mobility, the maximum discharge recorded during the sampling period is associated with the bedload sample (e.g., Church and Hassan, 2002; Gomi and Sidle, 2003).  Samples collected in the removable pit traps ranged from 100 g to over 29 kg (Table 2.2). Bedload samples collected in the channel spanning pit traps ranged from 120 g to 132 kg (Table 2.2). Typically, the largest clast accounted for less than 1% of the total sample weight. During the largest discharge event of the study period, keystones comprising stone-lines were mobilized at sites E4 and C3. In this case, the largest clast at site C3 accounted for about 10% of the total sample weight. To determine grain size distribution bedload samples were sieved on site through 128, 64, 32, 16 and 8 mm meshes; each sieved sample was then transferred to a field lab, dried, weighed and sieved further through 4 mm and 2 mm meshes.  6.2.3 Channel bed texture At each bedload site surface and subsurface channel bed texture was assessed using Wolman pebble counts and bulk sample analysis. Wolman pebble counts were conducted by  121 sampling incrementally (one foot-step increments) along diagonal transects in a zig-zag manner across the width of the channel upstream of the bedload traps until a minimum of 100 particles were sampled (Kondolf et al., 2003). In so doing, the estimation of bed surface texture at ? phi grain size divisions includes both mobile and immobile grains on the channel bed. Data from the Wolman pebble count is used to estimate the diameter of the largest immobile stones in the channel bed for the parameterization of bedload transport formula (Chapter 3.4). Bulk samples of bed surface and subsurface were limited to mobile material only (i.e exclude moss-covered and/or large angular clasts distinct in appearance from mobile material) and samples ranged from 40 kg to 80 kg (dry weight). Bulk samples were collected at the sites excavated during installation of the traps (Wilcock, 2001). Subsurface samples were collected by removing the surface layer to a depth equal to the largest mobile stone (? D84). At the uppermost sites (C2 and E2) additional sample sites were necessary to ensure that the largest grain accounted for no more than 5 to 10% of the total mass of the sample (Church et al., 1987). Bulk samples were dried, split when possible, and sieved to 1 phi divisions. Bulk sample data is used for: (1) the estimation of reference shear stress by grain size (Chapter 6.3.2); (2) assessing the variability of bedload entrainment as a function of discharge (Chapter 6.3.3); and (3) fractional and scaled fractional transport analysis (Chapter 6.3.3).  6.2.4  Initiation of motion To define thresholds of bedload entrainment in relation to grain size the flow competence method (Andrews, 1983; Ashworth and Ferguson, 1989; Mao and Lenzi, 2007; Mao et al., 2008), and the reference shear stress approach (Wilcock and Southard, 1988; Church and Hassan, 2002; Mueller et al., 2005) are applied. The former method typically  122 expresses average dimensionless critical shear stress (? ?*c) as a power function of the largest grain size class collected in the bedload sample (Di) scaled to the median grain diameter of the channel surface (D50s) (Andrews, 1983). Critical dimensionless shear stress (?*ci) is equal to: ???? =  ???(????)??         (6-1) where g is the acceleration due to gravity, ?s and ? are the densities of sediment (?s = 2650 kg/m3 for quartz-rich sediment) and water respectively. The shear acting on the grain (?i) is approximated by the boundary shear stress: ?o = pgRS          (6-2) where R is the hydraulic radius of the flow and S is the water surface slope. The former is calculated using measurements of flow cross-sectional area (A) (discussed previously) while he latter variable is measured using an engineer's level at discharges around bankfull flow along 8 to 10 channel-width distances upstream of each monitoring station. Parameters of the power law relation are estimated via least-squares regression:  ????? = ?(???50?)?         (6-3) In the reference shear stress approach (Wilcock and Southard, 1988) the threshold for entrainment is estimated as the dimensionless shear stress (?*ri) that produces a small dimensionless reference transport rate (W*i). Specifically,  ??*c is estimated by extrapolating the trend of ?*i, associated to each grain size class Di, as a function W*i to a small reference  123 dimensionless transport rate of W*=0.002. The dimensionless transport rate (W*i) is defined as (Parker et al., 1982; Wilcock and Southard, 1988): ??? =(????1)?????????(??)3          (6-4) where qb*pi/fi  is the fractional transport rate of grain size pi relative to the fraction of ?Di? in the bed subsurface (fi), and ??is the shear velocity of water (?o/?)1/2.  Similarly to the flow competence method, least squares regression is used to define parameters of the power law: ???? = ?(???50?)?         (6-5) A number of studies have established that in mountain stream channels dimensionless Shields stress for grain mobility varies with channel gradient and relative roughness (Buffington and Montgomery 1999; Shvidchenko and Pender, 2000; Mueller et al., 2005; Lamb et al., 2008; Bunte et al., 2010). The presence of LWD has been shown to increase critical shear stress for bedload entrainment up to an order of magnitude due roughness-induced turbulence within the velocity profile (Lamb et al., 2008). The nested sampling design applied here offers the opportunity to investigate how channel bed gradient, bed roughness and flow resistance can influence bedload mobility thresholds across sites (i.e., morphologies). Specifically, calculated values of ?*c50s (critical dimensionless shear for entrainment of the surface median grain size) are plotted using boundary shear (6-2) against channel gradient (S), bed texture (D50s:D90s) total flow resistance (ft) (Figure 6.6) at bankfull flow.   124 Total flow resistance (ft) is calculated as the sum of a base-level resistance analogous to grain resistance (fo) and an additional morphological resistance component (fadd) associated with macro-roughness elements including large woody debris and immobile boulders (e.g. Millar and Quick, 1994; Millar, 1999). Details regarding the calculation of flow resistance are provided in Appendix 2. 6.2.5 Bedload transport To investigate bedload particle mobility in relation to grain size (i.e., equal mobility vs. selective mobility) textural differences are compared between the channel bed surface and subsurface (i.e., bulk samples) and the bedload samples by means of cumulative grain size distributions. These are presented for the bedload samples (i.e., mean) collected at about bankfull discharge, as well as for the samples (i.e., mean) collected at flows below and above threshold discharge (Qth). Threshold discharge (Qth) is defined as the break point or intersection between low (phase-1) and high bedload transport rates (phase-2) (Jackson and Beshta, 1982; Ryan et al., 2005). Threshold discharge for each site is estimated via stepwise regression iteratively so that the combined R2 and p-values of the two regression trends are respectively maximized and minimized (Ryan and Porth, 2007). Threshold discharges at the monitoring sites typically correspond to values ranging between 60 to 80 % of bankfull flow (Table 2.3). Fractional transport analysis (Wilcock and McArdell, 1993, 1997) provides a means to investigate changes in the relative mobility of bedload grain sizes with increasing discharge. Scaled fractional transport is computed as qb*(pi/fi),where qb is the bedload transport rate (g/ms), pi is the proportion of each size fraction "i" present in the transported  125 material, and fi is the proportion of each size fraction in the bulk sample of sediment from the channel bed subsurface (Parker et al., 1982; Wilcock and Southard, 1988). The outcomes of this analysis are compared across sites to investigate the effects of morphological and textural differences on relative mobility.  A first order investigation of morphologic controls on bedload transport rate is made through the comparison of rating curves across morphologies and between streams. Rating curves express bedload discharge (qb) as a log-linear function of water discharge (Q) normalized by bankfull discharge (Qbf): qb = a(Q/Qbf)c          (6-6) where a and c are respectively the site specific constant and scaling exponent.  To appraise the extent to which bedload empirical data from Cotton and Elk Creek can be used as a basis for estimating bedload transport in unmonitored headwaters of the Columbia Mountain region observed bedload transport are compared with that predicted by means of empirically-derived transport formulae developed in mountain gravel-bed streams. In particular, formulae by Parker et al. (1982), Rickenmann (2001) (i.e., as modified by Nitsche et al. (2011)), and Yager et al. (2012) are tested (see Appendix 2 for more details on the selected bedload transport formulae). 6.3 Results 6.3.1 Characteristics of bed mobility In Cotton and Elk Creek channel bed mobilization occurs during peak flows associated with rapid snowmelt, typically between March and July. During the three years of  126 site E4 in Elk Creek, and between 31 and 65 days per year at site C4 in Cotton Creek. This study relies on the collection of 218 bedload samples.  The spring freshets of 2006, 2007 and 2008 included multiple sediment mobilizing peak flows (Chapter 5.3.2). The largest recorded discharge (i.e, 1.3 m3/s) was recorded at station C5 in 2006 (Figure 2.2a) and is estimated to be a 10-year flood event, based on the closest (50-km away) long-term gauging site located within the same physiographic and climatic sub-region (WSC 08NH016). Bankfull level was also attained at all monitoring sites in 2007 and 2008. The return period of the 2007 and 2008 peak flow events was estimated to be respectively 2 and 5 years.  In both channels the size distribution of mobile bed material is unimodal. Inspection of cumulative grain-size distributions (Figure 6.1) indicates that mobile sediment is finer than the channel bed surface at all sites, a well-documented characteristic of armored gravel-bed channels (e.g., Adenlof and Wohl, 1994; Church and Hassan, 2002; Muskatirovic, 2008). The median grain size of bedload samples at bankfull flows corresponds respectively to the D28 (C2), D7 (C3), and D6 (C4) of the channel bed surface in Cotton Creek, and to the D2 (E2 and E3), and D5 (E4) in Elk Creek (Figure 6.1). During the largest flood (Qmax) the D50 of the mobile sediment ranged from D5 (C3) to D60 (C2) of the channel bed surface in Cotton Creek, and from D4 (E3) to D20 (E2) in Elk Creek (Figure 6.1). At five sites the coarsest bed material is mobilized by the largest recorded discharge, with E3 (boulder-cascade morphology) being the exception. Here, the coarsest bedload is associated with discharges approaching bankfull (Figure 6.1e).  127  Figure 6.1 Grain-size distributions of channel bed surface, subsurface, and bedload material at the monitoring stations. Qbf  = bankfull flow; Qmax = largest flood; < Qth = lower than threshold discharge; and > Qth = higher than threshold discharge 6.3.2 Initiation of bedload transport and relative bed mobility Initiation of bedload transport in gravel-bed streams has been customarily described as either equal threshold entrainment, in which all grain sizes are mobilized over a narrow range of shear stress (e.g., Parker et al., 1982; Andrews, 1983), or size selective entrainment, in which progressively larger grains are mobilized with increasing shear stress (e.g., Ashworth and Ferguson, 1989; Komar and Shih, 1992; Whitaker and Potts, 2007). Exponents of the power-law equation (i.e. Equation 6-5) between relative particle size  128 (Di/D50) and dimensionless critical shear stress (?*ic) reflect the degree of equal-threshold entrainment while the coefficient (i.e., a) of the same power-law relation defines the dimensionless shear stress (?*) for the median surface grain size (D50). Using the reference shear method findings indicate that at E2 (forced step-pool), E3 (cascade), E4 (step-pool), C2 (forced step-pool) and C3 (forced step-pool) the ?*i -W*i relation across grain-size classes follows the subsurface-based bedload transport model proposed by Parker et al., (1982), according to which: Wi*= 0.0025 G(?),          (6-7) where ?= ?*i / ?*ri and G(?) is a three-part function dependent on shear stress. ?(?) = {5474 (1 ?0.853?)4.5                                 ? > 1.59???[14.2(? ? 1) ? 9.28(? ? 1)2]          1 ?  ? ? 1.59              ?14.2                                                       ? < 1  (6-8) At these five sites, where ?*ri is estimated as the intersection between the G(?) function and the line of reference dimensionless transport rate (W* = 0.002, Figure 6.2), conditions of near equal-threshold entrainment (Table 6.1,  -0.97 < b < -1.0) are observed over the full range of mobile bedload fractions (i.e., a single power-law relation fits the entire domain) (Figure 6.3a-d).   129  Figure 6.2. Determination of dimensionless reference shear by grain size class for each site according to Wilcock and Southard, 1988. The subjectivity associated with visual extrapolation creates greater uncertainty in the selected values of ? *i. A sensitivity analysis conducted at site C4 indicates that for grain sizes with greatest scatter in the ?*i - W*i relations (i.e., 1mm to 5.7mm) this uncertainty can reach up to 25%. At the remaining gentler site (C4, riffle-pool morphology), the ?*i -W*i relation departs substantially from Equation (6-8). Here reference dimensionless shear is estimated by visually extrapolating the data trend to the reference dimensionless transport rate (W*= 0.002). To aid in delineation of this trend, data are stratified into rising (black symbols) and falling limb (gray symbols) series (Figure 6.2c). During visual estimation greater weight is  130 placed on rising limb data points close to the reference shear value than on falling limb ones. The latter generally display higher reference shear stress for a given transport rate, consistent with a pattern of clockwise hysteresis observed in the study streams (Chapter 5.3.2).  Figure 6.3 ?*i as a function of Di/D50s. for study sites. Power-law equations are listed in Error! Reference source not found. According to the reference shear analysis the riffle-pool site displays a composite behaviour in that entrainment is described by an obvious double power-law relation (Figure 6.3a). Specifically, for Di/D50s < 1 the scaling exponent denotes dynamics of equal threshold entrainment (b = -0.97), which develop into size-selective ones (b = -0.41) for larger calibers.  Similar patterns of entrainment are observed by applying the flow competence approach, although here the range of grain-sizes that can be investigated is smaller (Figure  131 6.3 b,d). Exponents (b) between the maximum mobile grain (Dmax) and dimensionless shear (?*ci) range from -0.84 to -0.94 at the step-pool/forced step-pool and cascade sites (E2, E3, E4, C2 and C3, Table 6.1) to -0.54 at the riffle-pool site (C4). Exponents determined using either approach are well within the envelope of values reported by seminal studies from western North American mountain step-pool and riffle pool streams (i.e. 0.87 ? 0.99, Parker et al., 1982; Andrews, 1994; Andrews and Nankerviz, 1995) but differ from those reported for step-pool systems elsewhere which document a tendency to size-selective entrainment with exponents ranging between -0.69 and -0.74 (Marion and Weirich, 2003; Lenzi et al., 2006; Mao et al., 2008).  Table 6.1. Coefficients (a) and exponents (b) determined for the power-law equation defining particle entrainment. Site Method ???? = ?(???50?)? n R2 p 3 E2 (FSP(PB)) Ref. Shear y = 0.092x-0.992 7 0.99 <0.001  Flow Comp. y = 0.102x-0.936 36 0.99 <0.001 E3 (BC) Ref. Shear y = 0.092x-0.976 7 0.99 <0.001  Flow comp. y = 0.109x-0.914 33 0.96 <0.001 E4 (SP) Ref. Shear y = 0.098x-0.969 8 0.99 <0.001  Flow Comp. y = 0.103x-0.869 52 0.97 <0.001 C2 (FSP(BC)) Ref. Shear y = 0.960x-0.974 6 0.99 <0.001  Flow Comp. y = 0.869x-0.838 22 0.93 <0.001 C3 (FSP(BC)) Ref. Shear y = 0.543x-0.951 6 0.99 <0.001  Flow Comp. y = 0.527x-0.897 27 0.97 <0.001 C4 (RP) Ref. Shear y = 0.035x-0.795 7 0.97 <0.001   y = 0.023x-0.967 (1) 5 0.99 <0.001   y = 0.032x-0.410 (2) 3 - -  Flow Comp. y = 0.046x-0.543 33 0.53 <0.001 1. Di/D50<1,  2. Di/D50>1 3. P-value of regression (of log transformed power law)  As for site C4, a similar composite behavior of entrainment has been previously documented in large, low gradient, alluvial gravel-bed streams in Wyoming and Idaho.  132 Andrews (1983) reports that ?all particle sizes except the very largest, were entrained at nearly the same discharge? but notes size selective dynamics for particle sizes > 4.2 times the median grain size (D50s = 53 to 75mm). The exponent derived via the flow competence approach in C4 (i.e., -0.54) agrees with prior work conducted in a larger (83km2) riffle-pool streams (b = -0.59; S = 1%) in Montana (Whitikar and Potts, 2007).  The coefficients of the power-law equations (corresponding to ?*50s), which exhibit a variability of ?10% in relation to the methodology used, in Cotton Creek display greater than a 20 fold difference in magnitude across sites (Table 6.1). In particular, the coefficients of the two forced-alluvial sites (C2 and C3) stand out for being substantially higher (i.e., 0.543 ? 0.960) than what is typically reported for gravel-bed mountain streams (i.e., 0.041 to 0.088) (Parker et al., 1982; Andrews, 1983; Ashworth and Ferguson, 1989; Andrews and Nankerviz, 1995; Buffington and Montgomery, 1997). By contrast, the coefficients recorded in Elk Creek display minimal variability (i.e., 0.092 ? 0.109) regardless of morphology and up to two percent differences in channel gradient (i.e. 4.7% to 6.8%, Table 2.3). The values of dimensionless shear recorded on Elk Creek fall within the range of those observed in steeper (i.e. 2 to 3%) coarse-grained plane-bed and forced riffle channels (Bunte et al.,2010) but are less than those reported from step-pool streams in Northeastern Italy and Chile (i.e., from 0.189 to 0.288, Lenzi et al., (2006); Mao et al., (2008)). The high coefficients (i.e., ?c*) recorded at the semi-alluvial sites C2 and C3 imply dramatically reduced mobility of the bed material compared to the other study sites. In Chapter 6.4.2, the variability of the coefficients is investigated in terms of the effects of relative grain size, channel gradient, and form roughness associated with forced-step morphologies (Mueller et al., 2005; Wilcox and Wohl, 2006; Canovaro et al., 2007; Lamb et al., 2008; Bunte et al., 2010).   133 6.3.3 Relative bedload mobility with increasing discharge Phase-1 and phase-2 transport are associated with respectively low and high levels of channel bed mobility (Jackson and Beshta, 1982); they can be constrained reasonably well via step-wise regression in the discharge-bedload transport domain (Ryan et al., 2005). The onset of phase-1 transport, defined by the mobilization of sand and fine gravel, occurs at 24 to 48 % and 27 to 38 % of bankfull flow for sites in Cotton Creek and Elk Creek respectively (Table 2.3). The transition (i.e., threshold discharge, Qt) between phase-1 and phase-2 transport occurs at flows between 58% (i.e., C4) and 80% (i.e., C2) of bankfull flow in Cotton Creek, and between 61% (i.e., E2) and 77% (i.e., E3) in Elk Creek (Table 2.3).  In agreement with prior work conducted in mountain channels (e.g., Ryan et al., 2005), Qt separates domains of relatively finer (i.e., phase 1) and coarser (i.e., phase 2) bedload transport (Figure 6.1). Specifically, in Cotton Creek mobilization of material up to coarse gravel (? 32mm) is recorded during phase-1 transport, and up to small cobbles (? 128mm) during phase-2 transport. In E2 and E3, phase 1 and phase 2 are capable of transporting material respectively as large as very coarse gravel (? 64mm) and cobbles (?128mm). Finally, in E4 the coarsest bedload samples are recorded, that is, phase-1 transport includes small cobbles (?128mm), and phase-2 transport includes boulders (>256mm) (Figure 6.1f). Although the foregoing outcomes indicate that higher transport rates are associated with coarser bedload, phase-2 transport does not necessarily imply full (i.e., equal) mobility of the channel bed relative to the available bed material (i.e., Wilcock and McArdell, 1993; 1997).   134 Distinctive patterns of relative mobility are observed between sites that appear to relate chiefly to channel bed texture and gradient, rather than to differing assemblages of morphological units. At steep gradient, semi-alluvial sites C3 (FSP) and E3 (BC) (Table 2.3, Figure 6.4 b,e), the relative proportion of grain sizes in the bedload does not change with increasing discharge (i.e., discharge independent) and data remain compressed in terms of pi/fi. This pattern is similar to that reported for steep forested headwater streams in coastal Alaska (Gomi and Sidle, 2003). In contrast, fully alluvial sites C4 (RP) and E4 (SP) display decreasing relative mobility of the finest grains (< 22 mm) in response to increasing relative mobility of the coarser grains (>22 mm) with increasing discharge (Q > 90% Qbf, Figure 6.4 c, f), creating a vertical spread in pi/fi along the y-axis. A similar mobility pattern was documented in Harris Creek, a larger (220 km2) snowmelt alluvial stream in British Columbia (Church and Hassan, 2002). Despite general similarities in channel morphology (i.e. FSP), the two remaining semi-alluvial sites, E2 and C2, display different mobility patterns: the former behaves similarly to the fully alluvial sites in that it documents increasing relative mobility of coarser grains with increasing discharge, whereas the latter exhibits a lesser discharge-dependent mobility, somewhat consistent with observations made at C3 and E3. Scaled fractional transport plots, in which the bedload to bed material grain-size fraction (pi/fi) is scaled by the transport rate (qb), provide a more insightful picture of the relative mobility of bedload in relation to discharge magnitude. At discharges below bankfull, for all sites, the majority of gravel-to-cobble bedload (i.e.,Di >16mm) falls in the zone of partial mobility (zone C in Figure 6.4, panels g to l) (i.e., only a fraction of the  135  Figure 6.4. Relative mobility (left-hand panels) and scaled fractional transport (right-hand panels) of selected bedload samples with increasing discharge relative to the bed subsurface. Data points signify geometric mean diameter of grain size class. Dashed and solid lines in right-hand column differentiate between domains of over-passing (A), equal mobility (B) and partial transport (C).  136 available bed material within a given size class is mobile). As discharge increases the degree to which bedload becomes equally-mobile (i.e., size-independent transport, Church and Hassan, (2002), zone B in Figure 6.4) varies between sites. At the finest textured site (C2) equal-mobility occurs with discharges approaching bankfull (Qbf) for all available mobile calibers (Di ? 64 mm) (zone B in Figure 6.4g). At coarser textured sites (C3 and C4), the condition of equal-mobility is limited to medium gravel (i.e., Di ? 16 mm and 32 mm respectively) for discharges exceeding bankfull (zone B in Figure 6.4h and i). In Elk Creek the highest discharges (Q>Qbf) impart equal-mobility transport to medium gravel and finer material (Di ?16mm) at semi-alluvial sites E2 and E3 (Figure 6.4j and k), and up to small boulders (Di ? 256 mm, Figure 6.4l) at the fully alluvial site (E4).  The condition of over-passing (e.g., Hassan and Church, 2002; Gomi and Sidle, 2003), in which the finest grain sizes are under-represented in the bedload relative to the bed material, is limited to the fully alluvial sites (E4 and C4) but only (zone A in Figure 6.4i and l) for fine gravel and smaller bedload (Di ? 8 mm and 4 mm) during the highest stages. The increase in relative abundance of particles ? 64 mm, which produces the upward trend in the right-tail of all fractional transport plots at Q > 90% Qbf (Figure 6.4, panels g-l), is interpreted as the under-representation of coarse grains in the subsurface bulk samples due to the random distribution of the largest clasts (e.g., Brayshaw et al., 1983) and to the difficulties of capturing such variability in headwater systems (Church et al., 1987).  6.3.4 Bedload rating curves The nested monitoring network allows us to investigate bedload transport variability in forced step-pool, boulder-cascade, step-pool, and riffle-pool morphologies. Of interest is  137 determining to what extent channel morphology can influence bedload transport relations within and between stream channels. To this purpose water discharge (Q) normalized by bankfull (Qbf) is plotted as a power-law function (estimated via a linear fit to log transformed data) of bedload transport rate (qb, g/m/s) (Figure 7a and 7b, Table 6.2). In consideration of the hysteresis effects noted earlier (Chapter 5.3.2), analysed data are stratified by samples collected during either the rising or falling stage of the hydrograph. This stratification, although it does not alter the relative site ranking of the exponents in the power-law equations, in some cases improves the goodness of fit of the relations, and provides further insights on the variability of transport rates for rising and falling conditions. For example, it is apparent that bedload transport rates during falling stages at site E4 display substantial variability at low discharges (Figure 6.5a). Similar observations apply to sites C2 and C4 (Figure 6.5b). This variability associated with the falling stages relates to hysteresis effects due to the depletion or liberation of sediment (depending on site and hydrograph characteristics) during and following peak discharge (Green et al., 2012). The exponents of the bedload rating curves vary between 1.56 and 5.34 (Table 6.2), and as such plot within the range reported for larger coarse-bedded snowmelt-dominated streams in Idaho and Wyoming (Whiting et al., 1999; Emmett and Wolman, 2001; King et al., 2004).    138  Figure 6.5. Bedload transport (Qb) as a function of water discharge (Q) in: (a) Cotton Creek; and (b) Elk Creek. Bedload transport rate as a function of excess shear stress (?o-?c) in: (c) Cotton Creek, and (d) Elk Creek. ?c corresponds to ?r 16-20 depending on site. Solid symbols designate rising stage and open symbols are from falling stage periods. Best-fit power law equations given in Table 6.2.  Differences are apparent between Elk Creek and Cotton Creek, with the former showing larger exponents (4.18 < b < 5.34) than the latter (1.56 < b < 3.02), clearly indicating that the location effect (Elk vs. Cotton) overrides the morphological effect at the monitoring sites. The lowest and highest exponents are associated respectively with the fully alluvial riffle-pool site (C4) and the coarse-textured step-pool and boulder cascade sites (E3 and E4) (Table 6.2).  In agreement with sediment yield-based findings (Green et al., 2012), the former site displays bedload transport across a much wider range of flows (transport-limited behaviour) than the latter ones, which show extreme transport rates beyond a well-defined threshold (i.e., bankfull).   139 Table 6.2. Bedload rating curves (C = complete data set, R = rising stage data). Site Power law equation n (R2) p E2  C  qb = 0.14(Q/Qbf)4.18 36 0.59 <0.001 R qb = 0.16 (Q/Qbf)3.64 20 0.77 <0.001 E3  C qb = 0.35(Q/Qbf)5.34 33 0.65 <0.001 R qb = 0.49(Q/Qbf)5.63 14 0.71 <0.001 E4  C qb= 0.12(Q/Qbf)5.34 52 0.55 <0.001 R qb = 0.28(Q/Qbf)5.98 22 0.63 <0.001 C2  C qb = 0.04 (Q/Qbf)3.02 22 0.63 <0.001 R qb = 0.07(Q/Qbf)3.35 11 0.82 <0.001 C3  C qb = 0.13 (Q/Qbf)2.03 27 0.72 <0.001 R qb = 0.17(Q/Qbf)2.16 13 0.71 <0.001 C4  C qb = 0.06(Q/Qbf)1.56 33 0.39 <0.001 R qb = 0.10(Q/Qbf)1.47 18 0.55 <0.001 To assess whether these differences are significant an analysis of covariance (ANCOVA) is conducted on the log transformed linear regressions (i.e., Log (qb) = Log (a) +b*Log (Q/Qbf). ANCOVA shows that the slopes (i.e., exponents) at C2 (forced step-pool) and C4 (riffle pool) are statistically different at the 95% confidence level (p<0.05), whereas no difference is apparent between the exponents of the three sites in Elk Creek (p > 0.2).  A closer inspection of the three rating curves in Cotton Creek (Figure 6.5a) suggests that at the riffle-pool site (C4) bedload transport can start at somewhat lower flows than in the semi-alluvial sites (C2 and C3). Notwithstanding similar morphological conditions, these two sites display high variability, that is, the rating curve in C2 trends more closely to C4 than to C3. Elk Creek exhibits less inter-site variability and the rating curves overlap roughly along the same relation. However, among the three sites, E4 (step-pool) stands out for displaying virtually "transport-unlimited" conditions at bankfull flows (note vertical increase of transport rate in Figure 6.5b).  140 Plots of transport rate as a function of excess shear stress, stratified by hydrograph stage, appear to provide greater morphological discrimination among rating curves (Figure 6.5c and d). For example, in Cotton Creek a more pronounced distinction exists between the riffle-pool site (C4) and the semi-alluvial counterparts (Figure 6.5c) with the latter sites display higher transport rates at higher excess shear, producing comparatively larger exponents and lower intercepts (Table 6.3). In the first case, low variability in transport rates are observed over the range of excess shear while the second case highlights a typical behavior of bed mobility in steep gravel bed channels; high flow resistance initially inhibits sediment transport at low excess shear but mobility increases rapidly beyond a critical threshold, as sediment becomes available from beneath the armor layer or macro-roughness elements (Parker and Klingemann, 1982). Similarly to what is observed in Cotton Creek, in Elk Creek the excess-shear representation provides greater inter-site morphological discrimination in that: (i) it highlights the non-linear character of the rating curves in E3 (boulder-cascade) and E4 (step-pool); and (ii) it draws a linear and well-constrained rating curve for rising stages in E2 (forced step-pools). As a result of the above combined observations in Elk and Cotton Creek, a high degree of variability is noted between the forced step-pool morphologies (i.e., E2 vs. C2 and C3).     141 Table 6.3. Bedload transport as a function of excess shear stress (C = complete data set, R = rising stage data) Site Data Power law equation qb = a(?o-?c)b n R2 p E2 C qb = 0.0003(?o-?c)1.87 31 0.47 <0.001  R qb = 0.0016(?o-?c)1.45 20 0.43 <0.001 E3 C qb = 0.001(?o-?c)1.46 33 0.54 <0.001  R qb = 0.002(?o-?c)1.41 15 0.60 <0.001 E4 C qb = 0.0003(?o-?c)1.68 37 0.55 <0.001  R qb = 5.4E-05(?o-?c)2.41 16 0.53 <0.001 C2 C qb = 0.0006(?o-?c)1.31 13 0.46 0.01  R qb = 0.002(?o-?c)1.09 7 0.28 >0.05 C3 C qb = 0.0014(?o-?c)1.04 28 0.46 <0.001  R qb = 0.0002(?o-?c)1.63 13 0.73 <0.001 C4 C qb = 0.006(?o-?c)0.73 30 0.49 <0.001  R qb = 0.010(?o-?c)0.73 18 0.58 <0.001  6.4 Discussion 6.4.1 Bedload entrainment  Analysis conducted using both reference shear and flow competence approaches indicates that channel morphology plays an important part in bedload entrainment dynamics. This is best exemplified by the distinctive behaviour of the riffle-pool site (C4) (Figure 6.3). A comparison of the hydraulic characteristics of the study sites (Table 6.4) reveals that the occurrence of equal-threshold entrainment across the entire range of mobile calibers, except at C4, is associated with high flow resistance (i.e., ft >1) related to high channel gradient (S) and the greater density of macro-roughness elements (i.e., #LWD/m, Am/At, Table 6.4). Consequently, at C4 the pattern of equal threshold entrainment of gravel and finer bedload (Di ? D50s) followed by size-selective entrainment of coarser material appears to relate solely to the resistance afforded by the armoring layer (i.e. grain or base-level resistance, fo).    142 Table 6.4 Hydraulic and geomorphic parameters at monitoring sites.   E2 (FSP) E3  (BC) E4  (SP) C2 (FSP) C3 (FSP) C4  (RP) Vm/sa 0.68 0.8 0.76 0.99 1.01 1.1 dbf 0.23 0.27 0.35 0.31 0.35 0.38 Wbf 1.3 1.7 1.6 1.7 2.3 2.5 D50s:D50ssb 2 2.2 1.7 1.2 3.1 2.6 D50s:D90s 0.57 0.64 0.28 0.37 0.39 0.57 LWDd (#/m) 0.2 0.15 0 0.27 0.25 0.1 LWD ht(m) 0.21 0.68 0 0.58 0.45 0.24 foe 0.104 0.114 0.08 0.042 0.06 0.063 fadd 1.66 7.28 4.34 5.42 4.80 0.68 ft 1.76 7.39 4.42 5.46 4.86 0.74 nf 0.1 0.11 0.11 0.18 0.13 0.05 d/D84m 2.13 2.31 2.36 10 6 5.77 Avg. Dbg 0.16 0.27 0.2 0.26 0.19 0.2 d/Db 1.44 1 1.75 1.19 1.84 1.9 Q>Qbfh (hrs) 120 112 161 166 129 124 a. Average cross-sectional velocity at bankfull flow; b. Armoring ratio d. Large woody debris density for 15-20 metre reach above sampling site. d. Total resistance at bankfull flow; e. Grain resistance; f. Manning?s n at bankfull g. Average diameter of immobile boulders. h. Duration in hours of flows above bankfull in 2006. This explanation is further supported by comparing entrainment dynamics at sites C4 (riffle pool) and C3 (forced step pool). If on one hand these two sites display nearly identical bed texture (Figure 2.6), which imparts similar base-level resistance (fo, Table 6.4), on the other, C4 exhibits substantially lower total resistance (ft, Table 6.4) facilitating the onset of size-selective entrainment once base-level (i.e. grain) resistance is overcome (Shields, 1936; Fenton and Abbott, 1977; Church, 1978; Komar, 1987).  143 A comparison of ?*c50s across the six study sites indicates that in these natural stream channels dimensionless shear varies strongly with channel gradient and grain effects. Simple regression analysis yields the strongest correlations between dimensionless shear (?*c50s), channel gradient (S) and grain resistance (fo) (Figure 6.6a and b), while sheltering effects (D50s:D90s) and flow resistance (ft) (Figure 6.6c and d) draw much weaker correlations.   Figure 6.6 Variation in dimensionless critical shear stress (?*c50s) for the six monitoring sites with (a) gradient, (b) bed surface texture and (c) total flow resistance. Lines presented on the figures are intended to highlight general trends. Dashed lines (e) are from Lamb et al. (2008) and represent different ratios of shear stress borne by morphological structures to total shear stress.  144 Sites C2 and C3, displaying the highest channel gradients and finest mobile bed textures of the six sites (Table 2.3), record dimensionless shear values (?*c50s = 0.53 and 0.87 respectively) that are over an order of magnitude higher than at the lowest gradient riffle pool site (C4, ?*c50s = 0.035). Similar exceptionally high dimensionless shear values (?c*= 0.67 at bankfull for maximum mobile grain size) recently reported by Bunte et al. (2010) are attributed to the combined effects of limited sediment supply, steep-gradients and coarse-textured cobble step-pool channels. In this study, the effect of grain size on dimensionless shear stress becomes apparent when sites E3 and C2 are compared. These two sites display similar gradients and forced morphologies (Table 2.3) and consequently similar flow resistance (Table 6.4), however much finer mobile bed textures at site C2 correspond to a dimensionless shear stress value that is 8 times higher than that at E3. This finding, in part, contradicts the conclusions of Bunte et al. (2010) and suggests that, where forced morphologies occur, such as in steeper gradient forested mountain streams, higher entrainment thresholds are associated with fine (i.e. gravel-sized) rather than coarse (i.e. cobble-sized) bed material.  The three sites on Elk Creek display a much smaller range of dimensionless shear stress values (?*c50s = 0.092 to 0.098) for a range of channel gradients (S=0.047 to 0.068) compared to finer-textured Cotton Creek. This observation suggests that in coarser-textured streams (i.e. cobble streams) dimensionless shear stress is less sensitive to small variations in channel gradient (i.e. at least up to 2%) compared to finer-textured (i.e. gravel) streams.  The complex nature of bedload entrainment in these wood-rich natural channels is apparent when dimensionless shear versus channel slope for the six sites are plotted together with Lamb et al.?s (2008) theoretical curves approximating of the effects of increasing  145 relative roughness and grain emergence with increasing slope gradient for given conditions of morphologic drag (Figure 6.6e). This figure reveals that ?*c50s for the two finer textured LWD-forced sites on Cotton Creek are much greater that what is predicted for gravel streams with similar gradients and high morphological drag (i.e. 90%), while the three Elk Creek sites all plot along the 60% ?m:?T line despite the relatively wide range of form resistance values (fadd = 1.66 to 7.28, Table 6.4) and different morphologies.  The outcomes of this study, which is based on a limited number of data points, suggests that in forested mountain streams dimensionless shear shows different trends with increasing channel gradient depending on the absolute grain size distribution of the mobile bed material. Streams with finer-textured bed material (i.e. gravel streams) exhibit exponential increases in dimensionless shear as channel gradient increases from riffle pool (i.e. S ? 0.02) to forced step pool morphologies (S ? 10%). However, in coarser-textured cobble streams dimensionless shear stress appears to be generally insensitive to small changes in channel gradient (i.e. ?S < 2%). 6.4.2 Bedload transport Patterns of fractional transport observed in this study differ from those observed in flumes and larger low gradient alluvial systems (Wilcock and McArdell, 1993; 1997; Church and Hassan, 2002) but are generally consistent with what has been reported for steep mountain streams elsewhere (Gomi and Sidle, 2003; Mao and Lenzi, 2007). The results of this investigation indicate that for floods below the average annual stage (bankfull), the effects of form resistance impart first-order controls on patterns of fractional transport. At semi-alluvial sites with high form resistance (E3 and C3) near-vertical boundaries between  146 zones of partial and equal-mobility at Di ? 16 mm indicate that this threshold was never attained even for discharges greater than bankfull. In contrast, fully alluvial sites exhibiting lower form resistance develop equal-mobility transport across a broader range of grain sizes (e.g. E4 and C4) with increasing discharge. The development of equal-mobility transport for all mobile grain size (Di ? 64mm) at C2(FSP), a site exhibiting high flow resistance, presents an exception to the observations noted above. In this case the fine texture of the mobile bed material and discharge characteristics likely play a role in the dynamics of bed mobility.  The limited development of equal-mobility transport during average flood events (i.e. 1.5 to 2 year return period) in steep streams exhibiting characteristics of high flow resistance has been reported previously. High form resistance was found to inhibit the development of equal mobility for grains larger than 10 mm during floods with return periods of up to 1.5 years in steep, rainfall-driven headwater streams in Coastal Alaska (Gomi and Sidle, 2003). The limited development of equal-mobility for grains larger than 45 mm has also been documented in a steep cobble-boulder snowmelt-rainfall stream in Italy for floods with return periods of 2 years or less (Mao and Lenzi, 2007). In addition to the primary controls on mobility associated with flow resistance, downstream variability in duration of floods above bankfull (e.g. Lenzi et al., 1999; Powell et al.,2001) likely contributes to the occurrence of equal mobility for all available grain sizes at sites C2 and E4 (Table 6.4). During the 2006 freshet, which corresponded to the highest recorded discharges during the study period (?1:10 year flood), site E4 experienced floods above bankfull 36% longer than site E3 and 28% longer than E2 while site C2 recorded floods above bankfull for 29% and 34% longer than sites C3 and C4 respectively. The  147 overpassing of sand at sites C4 and E4 also occurred during the 2006 flood event while discharge exceeded bankfull.  When considered together exponents of the six rating curves generally define a trend of increasing exponent ?b? with increasing D90s (Figure 6.7) consistent with the finding that rating curve exponents are primarily a function of the size of grains armoring the bed surface (Emmett and Wolman, 2001, Bathurst, 2007). The influence of form resistance (i.e. morphology effects) on the exponent of the rating curve becomes apparent when the two streams are considered separately, (i.e. grain effects are held constant). A comparison of the three Cotton Creek rating curves reveals a delay in the onset of bed mobility at the two forced-step pool sites (C2 and C3, Figure 6.5) compared to the riffle pool site (C4). Similarities in grain resistance across sites (fo, Table 6.4) suggests that this behavior is due to higher form resistance (i.e., fadd) at the two forced step-pool sites (C2 and C3). Elk Creek sites also display a positive correlation between rating curve exponents and form resistance however, in this case, larger exponents at the boulder cascade (E3) and step pool (E4) sites correspond to higher transport rates at higher discharges rather than delayed entrainment.   Figure 6.7 Exponents of the six rating curves generally define a trend of increasing exponent ?b? with increasing D90s indicating that this exponent is a function of the size of the surface (armoring) grains.   148 6.4.2.1 Comparison with empirical transport formulae Modelling potential impacts to forested headwater streams from changing land cover requires the ability to predict sediment transport rates. For example, where aquatic values such a drinking water or spawning habitat exist in a watershed with high timber values it is necessary to be able to predict how sediment mobility that could affect water quality or channel stability might change with increasing peak flows. Accurate prediction of bedload transport in steep gravel-bed streams has been the focus of many investigations over the last three decades (e.g., Bagnold, 1980; Smart, 1984; Bathurst et al., 1987; Parker, 1990; D?Agostino and Lenzi 1999; Rickenmann, 2001; Chiari et al., 2010). Transport formulae derived on lowland rivers typically overpredict sediment flux in steep gravel-bed streams by orders of magnitude, an error that has been chiefly attributed to transport efficiency losses due to higher bed roughness (Bathurst et al., 1987; Rickenmann, 2001; Yager et al., 2007). Recent studies show that the performance of these predictive formulae in mountain streams can be significantly improved by incorporating a factor that accounts for flow resistance associated with immobile roughness elements (Yager et al., 2007; Chiari and Rickenmann, 2011; Ghilardi and Schleiss, 2011) and including a factor that accounts for the reduced area of mobile bed (Yager et al., 2007). To evaluate the extent to which measured transport rates are predictable, empirical data are plotted for a selection of morphologies against those obtained by applying equations originally developed in steep (S > 0.01m/m) coarse-textured channels. Selected equations include those by Parker et al. (1982), Rickenmann (2001), and Yager et al. (2012) (Table A2.1), however, the first two are modified from their original form to account for (1) reduced fraction of the channel bed available for transport due to immobile elements  149 (boulders and LWD) and (2) losses in shear stress due to macro-roughness (see Appendix 2). Among the tested formulae, that by Parker et al. (1982) behaves comparatively worse in that overpredicts bedload transport across morphologies by one to three orders of magnitude (Figure 6.8). Rickenmann's (2001) modified equation replicates transport rates at low excess discharges for both alluvial sites (C4 and E4) but underpredicts transport rates at the two semi-alluvial sites (C3 and E3) by up to two orders of magnitude. Of the three tested formulae, the equation by Yager et al. (2012) provides the closest approximations to the observed rates, however, its application is limited to conditions of high excess shear stress. In particular the Yager et al., (2012) equation predicts bedload transport reasonably well in the coarse textured alluvial step-pool (E4) and boulder cascade (E3) sites; although it underpredicts bedload flux through the finer textured forced step pools (C3) and over predicts flux for the riffle-pool site (C4, Figure 6.8).  Discrepancies between predicted and observed transport rates at the two Cotton Creek sites could reflect parameterization difficulties. Of the three formulae tested, Yager et al.?s (2012) is the most heavily parameterized, requiring seven measured input parameters (Appendix 2, Table A 2). A sensitivity analysis indicates small deviations in input variables (e.g.+/- 50mm for the protrusion value pu, (Table A 2) can cause up to an order of magnitude difference in the estimated transport rate at low excess shear stresses, hence limiting the application of this formulae in natural stream channels with variable topography.  Overall, the poor predictability of bedoad transport rates attests to the complex nature of bedload mobility in steep forested snowmelt streams. In particular, forested channels exhibiting median surface grain textures in the range of gravel to small cobbles  150 (e.g. Cotton Creek sites) exhibit transport rates that differ substantially from those predicted using equations developed for steep alpine streams. In order to achieve reasonable estimates of bedload transport rates in these forested mountain headwater streams additional efforts at modification and testing of existing transport equations is warranted. Of the three equations tested the modified Rickenmann equation is the most simplistic to apply and has the greatest potential for modification.  Figure 6.8 Predicted versus observed transport rates at steep forced-alluvial and gentler alluvial sites. See Appendix 2 for details regarding selected transport formulae.    151 7 Conceptual model of channel response to harvesting in forested snowmelt streams 7.1 Introduction The question of how much harvesting can be undertaken in a watershed without negatively impacting the stream channel is a question frequently asked by BC?s forestry managers, and professional foresters and represents the primary motivation for the research comprising this dissertation. To date forest scientists have had limited success in tackling this question directly. Development of the equivalent clearcut area (ECA) concept by Idaho State forest scientists in the 1970?s represents an early attempt to model the potential for downstream cumulative effects from forest harvesting (King, 1989; USDA-FS, 1974). Unfortunately little progress has been made on this topic in the past several decades and many of British Columbia?s forest managers are struggling to find available wood supplies given, among other restrictions, blanket ECA limitations. Grant et al. (2008) are the most recent to tackle the question of appropriate harvest levels to minimize channel impacts, however their ?state of science? approach that relies on the outcomes of CP-based investigations leads to the incorrect conclusion that peak flow increases diminish with increasing return period.  This purpose of this chapter is to provide BC?s forest managers with an improved understanding of the potential for hydrologic and geomorphic changes in snowmelt watershed associated with forest harvesting. This is achieved by combining a new physical understanding of the influence of harvesting on the flow regime revealed through frequency-based analysis with new insights regarding hydrodynamics of bedload mobility in forested  152 snowmelt streams to produce a strategic-level conceptual model of the potential for channel response to forest harvesting. Specifically, the influence of basin characteristics on hydrological response are considered together with findings of the physical controls on bed mobility associated with different morphologies in forested headwater streams in order to predict the response potential for different combinations of watershed and channel morphologies.  7.2 Physical process understanding Channels are linked to their watersheds through hydrogeomorphic processes governing water and sediment delivery to the stream network (Montgomery and Buffington, 1998; Gomi et al., 2002; Church, 2002). British Columbia?s interior watersheds display hydrogeomorphic processes characteristic of semi-arid snowmelt environments. Hillslope runoff during the spring snowmelt freshet is the primary contributor to flood generation (Church, 1988, Smith, 2011). Annual variability in seasonal meteorology contributing to snow accumulation and melt dynamics produces a range of flood magnitudes and durations that characterizes the regional flood regime. Detailed stand level and watershed scale studies have determined that slope aspect, elevation, gradient and forest cover are the primary factors influencing snow accumulation, melt and hillslope runoff (Winkler et al., 2005; Jost et al., 2007; NRC, 2008; Ellis et al., 2011; Smith; 2011). Removal of forest cover alters snowmelt rates (Winkler et al., 2005; Jost et al., 2007 and Ellis et al., 2011), soil water runoff (NRC, 2008; Smith, 2011) and streamflow volume and timing (Troendle and Olsen, 1994; Troendle et al., 2001) causing increases in the frequency, magnitude and duration of flood events (Alila et al., 2009, Schnorbus and Alila, 2004; 2013; this study, Chapters 3 and 4). Watershed characteristics of scale, slope gradient, elevation and aspect can either  153 exacerbate or limit changes in the flood regime associated with forest harvesting (Chapters 3 and 4).   Processes marking the downstream transfer of sediment from headwater reaches to higher-order channels are inherently linked to the flow regime and vary with watershed scale, underlying geology, channel gradient, confinement, and vegetation (Schumm, 1977; Montgomery and Buffington, 1998; Church, 2002; Montgomery, 2003; Wohl, 2008). The investigation of two forested snowmelt streams of the Columbia Mountains (Chapter 5) documents that sediment yield is strongly correlated with the number of peak flow events above a threshold discharge (peaks-over-threshold, POT), as well as flood magnitude and duration (Table 5.1).  In fully alluvial systems channel morphology reflects the relation between sediment supply (i.e. rate of sediment transfer from upstream) and transport capacity (i.e. discharge) (Montgomery and Buffington, 1997) and changes in either have the potential to cause adjustments in channel form to account for new governing conditions (Buffington et al., 2003). A substantial number of studies have documented changes in channel form resulting from increased flood frequency and magnitude associated with forest clearing for croplands or urbanization of a watershed (Hammer, 1972; Graf, 1975; Knox, 1977; Booth, 1990; Fitzpatrick et al., 1999). In formerly glaciated forested headwater streams typical of BC?s interior snowmelt regions, channel form may be adjusted to long term flow regimes (Figure 5.7) but the presence of glacial lag deposits and vegetation along channel banks limits the extent of boundary adjustments that can occur in response changes in sediment supply or discharge (Eaton and Millar, 2004). In these semi-alluvial reaches (Halwas and Church, 2002) channel adjustment is most likely to occur through aggradation in transport limited  154 reaches (i.e. where low gradients limit the rate of bed mobility) or degradation and incision in supply-limited ones (Booth, 1990; Buffington et al., 2003).  7.3 Methods To model the potential for channel response (assuming impacts to riparian stands and channel banks are limited through riparian buffers and therefore wood supply to forested channels is unaffected) it is necessary to consider: (1) the potential for changes in the flow regime due to forest removal, and (2) the potential for changes in bedload mobility due to changes in the flow regime. A relatively simplistic approach considers three levels of potential response; ?High?, ?Moderate? and ?Low?, similar to the qualitative assignment of risk in the assessment of hazards on fans associated with forest harvesting developed by Wilford et al. (2009). In this approach a ?High? potential for changes to the flow regime would apply to catchments with physical characteristics that act to maximize increases in peak flow magnitude and frequency following moderate levels of harvesting (i.e. ~ 30%) while a ?High? potential for changes in bed mobility applies to channel types that display both high relative mobility (i.e. Wilcock and McArdell, 1993; 1997) and rates of bedload flux at bankfull discharges. Conversely, a ?Low? potential for changes to the flow regime applies to catchments with physical characteristics that limit changes to flood frequency and magnitude. While a ?Low? likelihood of bedload mobility changes apply to channels containing material that the contemporary flow regime is not competent to mobilize. The designation of ?Moderate? potential applies to catchments and channels that display characteristics intermediate to the two end members.  155 Presentation of this simplified response model within a matrix framework results in the delineation of five categories of catchment response potential (Figure 7.1) to harvesting. Category 1 (green) describes catchments with a low potential of flow regime changes and low potential of bed mobility/channel form changes (i.e. Low ? Low), Category 2 (yellow) corresponds to Low-Moderate (or the reverse) response potential combinations. Category 3 (orange) identifies High-Low and Moderate-Moderate response potential combinations. Category 4 (bright red) corresponds to High-Moderate combinations. The highest risk catchments for impacts to aquatic values (i.e. water quality and quantity, channel stability) due to harvesting are those corresponding to Category 5 (dark red) that display both a high potential of flow regime changes and a high potential of bed mobility changes.  Response Potential Potential for flow regime changes Low Mod. High Potential for changes in  bed mobility  Low  (1) (2)  (3) Mod. (2) (3) (4) High (3) (4)  (5) Figure 7.1 Risk matrix for channel response. Identifying the physical attributes of headwater catchments that contribute to response potential is fundamental to successful discrimination between ?robust? catchments that are less sensitive to harvesting and those at greatest risk of developing long term impacts to aquatic values associated with harvesting related flood regime changes. The  156 results of the outcomes of Chapters 3 and 4 are combined with results from numerous stand level investigations on the influence of physical characteristics including aspect, elevation, and slope gradient on hydrological response to assist in developing an understanding of watershed-scale flood response. Likewise, studies of bedload entrainment and mobility in mountain streams are combined with the outcomes of Chapters 5 and 6 to provide the basis for assessment of channel response potential. Combining these two components enables the development of a strategic-level conceptual model to identify catchments with the greatest potential for experiencing harvesting related impacts to aquatic values in headwater snowmelt streams.  7.4 Conceptual model development 7.4.1 Potential for flow regime changes Increased snow accumulation and increased rates of snow melt are most often implicated as processes contributing to peak flow increases following harvesting (Troendle and Kauffman, 1987; Troendle, 1987; Scherer and Pike, 2003; NRC, 2008). However, a stronger correlation between snowmelt rates and peak flow magnitude compared to snow accumulation and peak flow magnitude (Table 3.1) leads this investigation to focus on the former for an explanation of catchment influences. Detailed stand level studies have determined that increases in melt rates following forest clearing are a function of the conversion from longwave-generated melt under the forest canopy to shortwave-generated melt in the cleared openings (Adams et al., 1998). The magnitude of these changes in melt rate vary as a function of slope gradient, aspect, and elevation (Jost et al., 2007; Ellis et al., 2010; Varhola et al., 2010). Melt rate increases associated with forest removal are found to be substantially lower on north aspect slopes than on west and south aspect slopes due to  157 natural shading that limits potential increases in shortwave radiation (Jost et al., 2007; Ellis, et al., 2010) and are also comparatively lower in gentle gradient terrain due to a greater influence of shade-related long wave radiation on snowmelt (Ellis et al., 2010). Larger increases in melt rates are documented at higher elevations where melt occurs later in the spring season when longer daylight hours occur (Jost et al., 2007).  The collective outcomes of the above stand-level energy balance studies when applied to assess the potential for catchment-scale flow response lead to the hypothesis that larger harvesting related changes in the flow regime are likely to occur in catchments in which the conversion from longwave to shortwave-dominated snowmelt is maximized (i.e. largest net radiation increases) and then efficiently delivered as runoff to the stream network. Specifically, basin characteristics of aspect distribution, elevation distribution and slope gradient should all influence catchment scale flood response to harvesting such that larger increases in basin-average melt rates should occur in catchments where the aspect distribution and elevation range are minimal, and slope gradient maximizes absorpion of incoming shortwave radiation (Oke, 1987). Physical characteristics that naturally mitigate the impact of forest removal changes on snowmelt runoff rates at the catchment scale should include the presence of large alpine areas that contribute the majority of seasonal catchment outflow, predominantly north-aspect slopes that naturally shade the snowpack or, conversely, a wide distribution of elevations and aspects that allow for desyncrhonziation of snowmelt.  Outcomes of the meta-analyses undertaken in Chapters 3 and 4 generally support this hypothesis. These results indicate that Redfish Creek which exhibits steep gradient slopes (avg. S = 50%) and a large alpine area (40%) displays comparitively smaller increases in  158 frequency, magnitude and duration of floods following moderate levels of harvest. Conversely, Camp Creek and 240 Creeks, the two fully forested catchments with moderate gradient slopes (avg. S = 20 % and 24% respectively) and, in the case of Camp Creek, a larger proportion of south aspect slopes, display larger increases in the frequency, magnitude and duration of flooding following similar levels of harvest (see Figure 3.1, Figure 4.2 and Table 4.2). The influence of basin characteristics with respect to synchronization of snowmelt runoff and delivery of runoff to stream networks consistent with the conjectures presented above has recently been confirmed through a detailed investigation of slope runoff dynamics in headwater basins of Elk Creek (Smith, 2011) as well as through detailed DHSVM modelling of different harvest scenarios in 240 Creek (Schnorbus and Alila, 2013). The results of both physical process investigations indicate that harvest openings in small forested catchments increases peak flow response through the synchronization of melt over a range of elevations. Smith?s (2011) findings also confirm that these effects are greatest on south aspect slopes.  7.4.2 Potential for channel response Differences in the physical characteristics of channels lead to disparities in entrainment and mobility dynamics in steep headwater streams. Channel gradient, bed texture and characteristics of flow resistance have been identified as important factors governing bedload entrainment (Schvidchenko et al., 2001; Mueller et al., 2005; Lamb et al., 2008) while bedload mobility has been found to vary with the influence of flow resistance (Church et al., 1998; Yager et al., 2007; 2012; Ghilardi and Schleiss, 2011). In forested mountain channels woody debris jams (LWD) are important macro-roughness elements influencing bedload entrainment and transport through the additional resistance imparted by  159 channel spanning jams (Sidle, 1988; Gomi and Sidle, 2002; Curran and Wohl, 2003; Mao et. al. 2008).  High channel gradients reduce mobility due to the low relative submergence and high turbulence. As channel gradient increases it becomes increasingly difficult to mobilize bed material (Bathurst, et al., 1983; Mueller et al., 2005; Lamb et al., 2008). Outcomes of Chapter 6, showing a positive correlation between dimensionless shear stress for grain mobilization and channel gradient, support this finding (Figure 6.6).  In gravel-cobble bed streams a well-developed surface armor layer develops due to the downward sifting of finer grains (Frey and Church, 2009). The armor layer inhibits bedload mobility by sheltering finer material from the shear stress exerted by the flowing water (Einstein, 1950; Fenton and Abbott, 1977; Parker et al., 1982; Parker and Klingeman, 1982; Mueller et al., 2005). The armor layer has the effect of delaying bed mobility until discharge thresholds necessary for disrupting the armor layer are achieved (Emmett, 2010). Results of Chapter 6 support this finding. In particular, they indicate that reaches with the finer bed textures and lower sheltering effects (i.e D50s:D90s) display greater bed mobility at lower relative discharges and excess shear stress compared to coarser textured channels (Figure 6.5).  The high density of macro-roughness including LWD jams and lag boulders has been shown to decrease bedload mobility through decreases in flow velocity associated with high form resistance. Study outcomes of Chapters 5 and 6 as well as several others (e.g., Sidle, 1998; Adenlof and Wohl, 1989; Gomi and Sidle, 2003; Wilcox and Wohl, 2006) show that the presence of abundant woody debris functions to reduce bed mobility relative to  160 reaches without woody debris. Outcomes of Chapter 6 indicate that reaches with high densities of woody debris and lag boulders (and consequently, high flow resistance, Table 6.4) achieve only partial mobility of the available bed material even at discharges exceeding bankfull (Figure 6.4). Those reaches displaying low flow resistance and a low percentage of macro-roughness elements achieve full mobility over a much wider range of grain sizes (e.g. site E4). Although the conditions of low flow resistance and low macro-roughness also occurs at the riffle pool reach (C4), low channel gradients (S=2%) through this reach result in an upper limit to the available excess shear stress thereby inhibiting full mobility of the larger grain sizes even for discharges exceeding bankfull (i.e. RP ? 10 yr). Relative roughness related to protruding grains functions to reduce bed mobility through disruptions in flow velocity near the channel bed (Lamb et al., 2008). Several flume-based studies have established that entrainment is delayed and bed mobility reduced in channels displaying large relative roughness (D50s:dbf) or conversely, low relative submergence (d/D84, the inverse parameter) (Ashida and Bayazit, 1973; Bayazit, 1978; Bathurst et al., 1983, Graf, 1991; Buffington and Montgomery, 1997; Shvidchenko and Pender, 2000). These effects of relative roughness are greatest for relatively shallow flows over emergent grains at high channel gradients. A comparison of the relative mobility of larger grains at the different monitoring sites on Cotton Creek supports this finding (Figure 6.4). Specifically, site C2, which displays the finest grain size distribution and the largest relative submergence during bankfull flows (Table 6.4Table 6.4) of the six sites displays an abrupt transition to full mobility of the channel bed for flows approaching bankfull.  Based on the factors discussed above it follows that channels least likely to experience increased bed mobility associated with changing flow regimes are the colluvial and semi- 161 alluvial morphologies that display high gradients, coarse bed surfaces with large sheltering effects and abundant macro-roughness elements creating high flow resistance. Semi-alluvial and alluvial channels with lower gradients, small sheltering effects and low densities of macro-roughness elements are most likely to respond to changes in peak flow magnitude, frequency and duration. Of the alluvial channel types, riffle pool channels with gradients of less than 2% are likely to be somewhat more resilient to increases in the frequency of large floods (i.e. RP > 10 yr) compared to steeper gradient plane bed, step pool and cascade channels due to the slope-related upper limit on excess shear stress that inhibits mobility of larger grain sizes. The channel response potential outlined here, which considers the combined effects of relative and absolute grain size, varying shields stress with gradient and field-based observations of relative mobility from this study and other recent studies of bedload mobility, differs from the classic response potential model proposed by Montgomery and Buffington (1997). In their model, which relates response potential to the balance between transport capacity (Qc) and sediment supply (Qs), steeper gradient step pool and cascade channels with high transport capacity and coarse bed textures have a lower potential for alteration from changing discharge than the lower gradient, finer textured riffle pool systems with lower transport capacity (Montgomery and Buffington, 1997).  7.5 Results and discussion The information necessary to apply the model is collected through GIS analysis and synoptic channel surveys conducted in reaches upstream from the fan or outlet of the watershed under investigation. Catchment attributes considered in the assessment of flood response potential include aspect distribution, percent of alpine area, average slope gradient, basin size and elevation range. With the exception of basin size, thresholds applied to each attribute for the  162 assessment of flood response potential relate to the outcomes of the meta-analysis investigation (Chapters 3 and 4) and the findings of stand-level studies summarized in Section 7.4.1. Thresholds assigned to channel attributes consider the outcomes of Chapters 5 and 6 in association with the literature synopsis provided in section 7.4.2. The proposed attribute thresholds will require further testing prior to application for forest management purposes. In addition it is likely that the thresholds will vary across physiographic regions and will need to be adjusted to some degree to account for differences in hydrogeomorphic processes between regions. At least three attributes from a given response potential category (i.e. Low, Moderate or High) are necessary for classification of flood and channel response potential.   Figure 7.2 Conceptual model of channel response in forested snowmelt watersheds. At least three attributes from a single category are required for the assignment of response potential.  163 The classification scheme presented in Figure 7.2, indicates that small, fully forested, single-aspect catchments with moderate slope gradients would be assigned a high potential for flood response while, large (>100 km2) watersheds containing steep, alpine areas would be assigned a low flood response potential. A basin area of 100 km2 is selected as the threshold between moderate and low flood response potential because this is the scale that generally corresponds to the transition from steeper mountain headwaters (S ? 1%) to low gradient (S? 1%) fully alluvial streams in interior snowmelt regions (King et al., 2004). A channel gradient of 30% is selected as the threshold between moderate and low channel response potential because this gradient generally corresponds to the division between colluvial, debris flow dominated and semi-alluvial streams (Halwas and Church, 2002).  A comparison of time series of Google Earth images showing hillslope shading provides an informal test of the effects of aspect distribution and slope gradient during the snow melt period. Figure 7.3 clearly illustrates the difference in shade patters between steep gradient Redfish Creek (S = 50%) and moderate gradient 240 Creek (S = 23%), two of the meta-analysis catchments investigated in Chapters 3 and 4. Shade patterns over an 11 hour period for May 10th reveal shading on steeper gradient east and west aspect slopes in Redfish while the majority of 240 Creek is exposed to sun.   164  Figure 7.3 A comparison of shade patterns for May 10th for Redfish Creek and 240 Creek. Compared to 240 Creek, Redfish has steeper gradient slopes and two dominant aspects (E/W) which causes shading and results in less direct solar radiation during the early May snowmelt period.  On the basis of catchment characteristics listed in Table 2.1 Redfish Creek would be considered a ?Moderate? potential for flood response due to its size, slope gradient and aspect distribution while 240 Creek would classify as a ?High? potential based on slope gradient, lack of alpine area, and basin size. These classifications generally agree with the results of the meta-analysis investigations in Chapters 3 and 4 and summarized above which revealed that both Redfish and 240 Creek experience changes in the flood regime associated with moderate levels of harvesting (i.e. 30 to 40%) but of the two catchments 240 Creek displays larger increases in the frequency, magnitude and duration of floods larger than bankfull (i.e. > 2 yr return period) compared to Redfish Creek.   165 Channels displaying characteristics of steep gradients, coarse bed textures and abundant macro-roughness would have a low potential for channel response according to the classification criteria of Figure 7.2, while those displaying low gradients, finer bed textures and minimal macro-roughness would be classified as having a high potential for response (Figure 7.2). B.C.?s interior snowmelt channels can be broadly grouped according to colluvial, semi-alluvial and alluvial morphologies (Halwas and Church, 2002) (Figure 7.4). In this context, colluvial channels display steep gradients (S ? 30%), coarse bed textures (D50s > 100mm) and a high proportion of macro-roughness elements (Am:At < 0.5) (e.g. Figure 7.4A-C), . Semi-alluvial channels include forced-step, forced-riffle-pool and cascade-type morphologies with gradients between 2 % and 30 % that contain abundant woody debris jams and lag boulders (Am:At <0.8) imparting high macro-roughness (e.g. Figure 7.4D-F). Bed textures in semi-alluvial channels vary locally but contain substantial finer textured mobile bed material usually stored upstream from woody debris jams (D50s < 100mm) (e.g. Figure 7.4D and F). Alluvial mountain channels display a range of morphologies (e.g. Montgomery and Buffington, 1997) however the primary distinction from semi-alluvial channels is their low gradients (S ? < 5%) and the low proportion of immobile macro-roughness elements (Am:At > 0.8) (e.g. Figure 7.4G-I). Bed textures in alluvial streams are typically finer than semi-alluvial counterparts (D50s < 60 mm) due to the absence of immobile macro-roughness elements.   166  Figure 7.4 Examples of colluvial (A-C), semi-alluvial (D-F) and alluvial (G-I) channels in B.C.?s forested snowmelt watersheds. In the case of Cotton and Elk Creeks the response potential model predicts that of the six monitoring sites, (i.e. C3 to C4 and E2 to E4) site E4 on Elk Creek would be classified as having a high potential for response due to its gradient, grain size distribution and the lack of macro-roughness elements, while the remaining five sites would be considered as having a moderate response potential. Of the six sites only two sites are fully alluvial (E4 and C4) while the remainder are classified as semi-alluvial and display a high density of macro- 167 roughness elements. Site C4 represents and exception to the generalization that alluvial sites fall into the high response classification. In this case low channel gradients (i.e. transport limited) reduce bed mobility through the riffle pool reach.  7.5.1 Application of model outputs to forest management Applying the outputs from the conceptual model outlined in Figure 7.2 to the risk matrix framework (Figure 7.1) results in the delineation of five categories of response potential (Table 7.1). Such categorization presents a strategic-level planning tool that can be used by forest planners to direct development and management activities. For example, management strategies for watersheds falling into categories 1 and 2 that present a low potential for impacts to aquatic values may include relaxing harvest level limitations established by existing regulations. Conversely, watersheds identified as category 5 require very careful planning and strict limitations on development levels in order to avoid long term impacts such as degraded water quality or aquatic habitat in these highly sensitive watersheds. Management strategies in watersheds meeting the criteria of categories 3 to 5 may include seeking guidance from forest hydrology specialists to maximize development opportunities while minimizing potential risks to aquatic values.  Table 7.1 Example strategic planning rational by response category Risk Category Possible Management Strategy 1 Level of harvest is not a restriction in this watershed 2 Moderate levels of harvest (~30-40 %) possible in these watersheds 3 Low to moderate levels of harvest (~ 25%) possible. Seek guidance from hydrologist if higher development levels are proposed. 4 Limit harvest levels to 20% and seek input from hydrologist to ensure proposed development limits risks to aquatic resources 5 Limit harvest levels to 15% and seek input from hydrologist to assist with identifying areas where development should be avoided.  168 8 Summary and conclusions This project was prompted by a knowledge gap in the current body of literature regarding the influence of forest harvesting on streams in snowmelt environments. To this end the study has addressed two questions: (1) how forest harvesting affects the flood regime of snowmelt catchments, and (2) how the flood regime affects bedload mobility in forested snowmelt headwater streams. The use of frequency pairing methods in the investigation of forests influence on floods has resulted in substantially different outcomes than those reported in past studies and provides clear evidence that moderate levels of forest harvesting have the potential to increase the frequency, magnitude and duration of snowmelt peak flows for a wide range of return periods including floods with return periods as high as the 50-year event. Even more importantly, the use of a frequency-based approach within a meta-analysis framework provides new insights of the influence of physical basin characteristics on harvesting related flood response. By combining this new understanding of flood response to harvesting with new insights regarding the influence of the flood regime and channel morphology on sediment mobility (component 2) this investigation shows that in some watersheds a moderate level of forest harvesting (i.e. 30%) has the potential to increase the frequency, magnitude and duration of floods with magnitudes ranging from geomorphically-effective to channel-forming. The summary and conclusions of key components of this research project are presented in the following sections. 8.1  Forest harvesting effects on the flood regime The meta-analysis of four snowmelt-dominated catchments utilizing a frequency-pairing approach has revealed that harvesting has changed the mean and the variability of  169 the probability distribution of annual floods hence changing the magnitude and frequency of the maximum annual flood series. Harvesting related increases in the magnitude of annual peak flows are observed over a wide range of event sizes, including larger floods of approximately 10, 20, and 50-yr return periods. In addition to observing increases in flood magnitude it is also observed that forest harvesting results in two- to four-fold increase in the frequency of these large floods. In all four treatment watersheds pre-harvest small and moderate annual floods become larger and decrease in frequency while the largest annual floods increase in frequency following harvesting regardless of whether they show statistically significant increases in magnitude. In two of our four study sites, the effects of harvesting on annual maximum peak flows increase with increasing return period with no apparent no-effect threshold. Further investigation of the same four snowmelt catchments has revealed up to three-fold increases in the number and 300% increase in duration of inter-annual peak flow events corresponding to 0.6*Q1.5 to the Q10  flood quantiles.  The outcome of the meta-analysis should be acknowledged as a real effect of forest harvesting irrespective of statistical significance because it represents a repetitive, physically explainable pattern in magnitude and direction among the four study sites [Lewis et al., 2010]. These estimated effects of forest harvesting on peak flows are representative of the most critical period after logging before any substantial recovery has occurred. However, the fact that no time trend was observed in the 19-year posttreatment data set for Camp Creek and only the last 12 years of the 48-year data set in Fool Creek required adjustment to correct for a very small time trend suggests that hydrological recovery occurs slowly in snowmelt-dominated hydroclimate regimes. This is consistent with previous analyses at  170 Fool Creek which speculated that full hydrologic recovery of the flow regime may take as long as 80 years after logging [Troendle and King, 1985]. Harvest induced changes in the frequency and magnitude of peak flows can be explained by applying an understanding of the influence of forest removal on changes in the energy balance at the stand and watershed scales. The dominant physical process associated with increases in peak flows is the increase in net radiation associated with the conversion from longwave-dominated snowmelt beneath the forest canopy to shortwave- dominated snowmelt in harvested areas. At the stand level, physical factors including slope aspect, slope gradient and elevation combine to influence the increase in net radiation following forest removal. At the catchment level basin characteristics including aspect distribution, elevation range, slope gradient, and the amount of alpine area can all influence the changes in the frequency, magnitude and duration of peak flows following harvesting. The outcomes of this investigation indicate that fully forested, south-aspect dominant, moderate gradient catchments display the largest increases in the mean and variability of the annual flow regime as well as the largest increases in the number and duration of peaks over threshold up to the Q5 flood quantile. In contrast, steep-gradient, alpine-dominated catchments display the smallest increases in the mean and variability of the annual flood regime and the smallest increase in the number and duration of peaks over threshold. The increase in variability of the posttreatment frequency distribution and the number of peaks over threshold appears to reflect the efficiency (synchronization) of delivery of increased snowmelt to the stream channel.  Study outcome, that the largest floods experience the largest increase in frequency, has implications with respect to the lifespan of engineered structures, the safety of human  171 settlement as well as the sustainability of fluvial ecosystems. The potential for increased frequency of damaging floods due to upstream forest harvesting was conjectured by engineers in the early 1900?s (Hoyt and Troxell, 1932). The results of this meta-analysis, undertaken a century later, support their original concerns that harvesting in upstream headwater forests can increase the occurrence of damaging floods. In addition to the concerns for impacts to humans and infrastructure associated with the largest flood events, the role of flood frequency in maintaining fluvial and riparian ecosystems has garnered substantial attention in recent decades (Resh et al., 1988; Naiman et al., 2005). While flood disturbance is important for maintaining habitat diversity and sediment and nutrient flux between terrestrial and aquatic environments (Naiman et al. 2005) the form and function of a fluvial system is adjusted to a flood regime consisting of infrequent channel-forming floods (i.e. ? Q10) and more frequent channel maintaining bank-full floods (i.e. (Q1.5 to Q2) (Wolman and Miller, 1960; Andrews, 1984; Schmidt and Potyondy, 2004). Increases in the frequency of occurrence and duration of these smaller floods has the potential to destabilize fluvial stream channels, create chronic problems for water distribution systems and degrade critical aquatic habitat beyond the natural range of (Schmidt and Potyondy, 2004). 8.2 Hydrodynamics of forested snowmelt streams. Formerly glaciated, forested streams of the Columbia Mountains that exhibit a subdued topography inherited from Pleistocene glaciations do not support debris-flow activity. Stream channels investigated in this study display alluvial and semi-alluvial morphologies mainly controlled by glacially-imposed local slope, wood abundance, and the virtually unlimited stock of glacigenic sediment. Slope and sediment availability chiefly control the distribution of free-formed alluvial reaches, whose downstream progression is  172 reset by the presence of glacial hanging valleys. Wood abundance, modulated by total stream power in conjunction with the degree of channel confinement, imparts an overall dominance of semi-alluvial channel types. Observed DHG relations conform to those characteristic of fully-fluvial lowland systems; an indication that these streams have adjusted their geometric configuration and energy expenditure to a post-glacial equilibrium state. Bedload entrainment exhibits two distinct patterns in accordance with morphological variability. In alluvial reaches where both base-level and morphological resistance is low (i.e. riffle pool) equal-threshold entrainment of grains up to the median bed surface diameter is followed by size selective entrainment of coarser material. In reaches where base-level and/or morphological resistance is high (ft > 1, e.g. step-pool and semi alluvial morphologies) equal-threshold entrainment patterns are observed for all mobile grain sizes. For all channel types discharges between 60% of bankfull and bankfull facilitate equal mobility transport for gravel and finer bedload while coarser loads are transported in partial-mobility fashion. For the largest floods where discharges equal or exceed bankfull the over-passing of finer fractions (i.e. sand) and equal mobility of coarser ones (i.e. cobbles and boulders) occur for channel types exhibiting one or more conditions of low total resistance, high relative submergence or high flood duration. Morphologic controls on critical dimensionless shear stress for entrainment of the median surface grain size (?*c50s) pertain to channel gradient and bed surface texture and flow resistance. Morphologies exhibiting low channel gradients and finer textured bed surfaces exhibit the lowest critical dimensionless shear values consistent with those documented for larger, alluvial river systems. Steeper, semi-alluvial gravel-bed channels document dimensionless critical shear values over an order of magnitude larger than those  173 reported for lower-gradient gravel bed channels. In coarse-textured channels (i.e. D50s > 50mm, D90s > 100mm), critical dimensionless shear stress appears to be less sensitive to changes in channel gradient.  The variability of the snowmelt hydrograph directly modulates annual rates of bed load transport through the frequency of peaks over threshold discharge (nPk). A coherent pattern of high yields during rising limbs and lower yields during the corresponding falling ones across the entire freshet is consistent with the notion of unlimited sediment supply, released through destabilization of the channel bed armor layer and LWD structures. Comparative morphometric analysis of Elk and Upper Cotton Creek indicates that a dominant southerly slope aspect in the former basin, imparts a faster response to meteorological fluctuations (e.g., changes of solar energy inputs), which translates into higher frequencies of peak flow events, and consistently higher annual bed load yields.  8.3 Potential for forested snowmelt channel response The combined results from Chapters 3 to 6 provide insights regarding the potential channel responses that are likely to occur in forested snowmelt streams associated with increases in discharge magnitude and frequency. First order controls on sediment yield in these forested headwaters include the frequency of peak flows over the threshold discharge for bed entrainment, as well as flood magnitude and duration. Moderate levels of harvesting in forested snowmelt streams increases the frequency and magnitude of large flood events and increases the number of peak flow events and duration of discharges capable of causing long-term changes to bedload transport and channel form. For a given level of harvest the extent of flow regime changes varies with the physical basin characteristics of catchment  174 size, aspect, slope gradient, elevation and percent of alpine area. In addition, differences in channel gradient, flow resistance and grain size characteristics between morphologies influences the capacity for changes in hydrogeomorphic processes governing bedload mobility. The greatest potential for flow regime changes exist in small, single aspect-dominated, fully forested, moderate gradient watersheds while the greatest potential for channel changes occur in fully alluvial, low gradient, uniform-textured channels. Concurrent assessment of the potential for flow regime changes and the potential for alteration to processes of bedload mobility within a risk-matrix framework provides a strategic-level management tool to identify appropriate levels of harvesting in watersheds to minimize the potential for long-term impacts to downstream infrastructure and aquatic values.  8.4 Future studies By overlooking the dimension of frequency at the outset of the investigation of forest influence on floods nearly a century ago the forest science community took the wrong research path which resulted in dozens of studies that were guided by irrelevant research hypotheses and therefore produced misleading outcomes. To move forwards towards a more comprehensive understanding of the connection between forests and floods we must re-trace our path and re-analyze existing hydrometric data using FP-based methods that enables simultaneous investigation of changes in both the magnitude and frequency of floods. Similar observational studies of existing data could further corroborate the hypothesized physical explanations of the environmental controls on the forests and floods relation but could also lead to interesting new hypotheses that should be tested using the derived flood frequency approach [e.g. Eagleson, 1972]. To reduce the risk of possible long-term impacts to aquatic values from forest harvesting in snowmelt headwater streams and establish  175 sustainable harvest levels, the influence of basin characteristics and hydrological recovery must both be re-investigated within a frequency distribution framework.  Forest harvesting in snowmelt watersheds can increase the magnitude, duration and frequency of flooding, which, in turn, provides first order controls on sediment yield, bedload mobility and channel stability. A long-enduring segregation between forest hydrology and fluvial geomorphology disciplines has hindered the development of a comprehensive understanding of the potential for impacts to channel condition and aquatic values associated with forest harvesting. Future work integrating studies of bedload and channel morphodynamics with hydrologic monitoring networks within physically distributed hydrologic and sediment routing models (e.g., Dolten et al., 2006; Flynn and Van Liew, 2011) has the potential to improve our understanding, hence ability to predict, the transient response of forested snowmelt alluvial and semi-alluvial headwater streams to land-management activities. The conceptual model of response potential developed on the basis of study outcomes contained in this research project represents a first attempt at combining an understanding of hydrology effects with channel response potential associated with forest harvesting. This model requires further testing and calibration across a wider physiographic and hydro-climatic land base in order to become a useful strategic-level management tool for the BC forestry industry.    176 Bibliography Adams, R.S., Spittlehouse, D.L. and Winkler R.D., 1998. The snowmelt energy balance of a clearcut, forest and juvenile stand, Proceedings 23rd Conference on Agricultural and Forest Meteorology, 2-5 November 1998, Albuquerque, N.M.,. American Meteorological Society, Boston, MA Adenlof, K.A, and Wohl, E.E., 1994. Controls on Bedload Movement in a Subalpine Stream of the Colorado Rocky Mountains, U.S.A. Arctic and Alpine Research 26: 77-85. Alexander, P.R., and Watkins R.K., 1977. The Frazer experimental forest, Colorado, General Technical Report 40, Rocky Mountain. Forest and Range Exp. Stn., USDA Forest Service, Fort Collins, Colo. Alila, Y.,  Hudson, R., P. K. Kura?, P.K., Schnorbus, M., and Rasouli K., 2010., Reply to comment by Jack Lewis et al. on ?Forests and floods: A new paradigm sheds light on age-old controversies,? Water Resources Research, 46, W05802,doi:10.1029/2009WR009028. Alila, Y., Kura?, P.K., Schnorbus, M., and Hudson R., 2009. Forests and floods: A new paradigm sheds light on age-old controversies. Water Resources Research, 45, W08416, doi:10.1029/2008WR007207 Alvera, B., and Garcia-Ruiz, J.M., 2000. Variability of sediment yield from a high mountain catchment, central Spanish Pyrenees. Arctic, Antarctic, and Alpine Research 32, 478-484. Anderson, H.W., and Hobba R.L., 1959. Forests and floods in the Northwestern United States, Publ. 48, pp. 30?39, International. Association of Hydrological Sciences, Gentbrugge, Belgium.  Andrews, E.D., 1980. Effective and Bankfull Discharges of Streams in the Yampa River Basin, Colorado and Wyoming. Journal of Hydrology 46: 311-330. Andrews, E.D., 1983. Entrainment of gravel from naturally sorted riverbed material. Geological Society of America Bulletin 94:1225-1231. Andrews, E.D., 1994. Marginal bedload transport in a gravel-bed stream, Sagehen Creek, California, Water Resources Research 30: 2241-2250. Andrews, E.D., 2000. Bed material transport in the Virgin River, Utah. Water Resources Research 36: 585-596. DOI: 10.1029/1999WR900257 Andrews, E.D., and Nankerviz J.M., 1995. Effective discharge and the design of channel maintenance flows for gravel-bed rivers. In: J.E. Costa, A.J. Miller, K.W. Potter, P.R. Wilcock (Editors), Natural and Anthropogenic Influences in Fluvial  177 Geomorphology: The Wolman Volume,.Geophysical Monograph 89, American Geophysical Union: 151-164. Archer, D.R., Climent-Soler, D., and Holman, I.P., 2010. Changes in discharge rise and fall rates applied to impact assessment of catchment land use. Hydrology Research 41, 13?26. Ashida, K., and Bayazit, M., 1973. Initiation of motion and roughness of flows in steep channels, International Association of Hydraulic Research, Proceedings of the 15th Congress, Istanbul, Turkey, 1, 475? 484 Ashworth, P.J., and Ferguson, R.I., 1989. Size-selective entrainment of bed load in gravel-bed streams. Water Resources Research 25: 627?634. Austin, S.A., 1999. Streamflow response to forest management: a meta-analysis using published data and flow duration curves. MSc thesis. Colorado State University, Fort Collins, Colorado Bagnold, R.A., 1980. An empirical correlation of bedload transport rates in flumes and natural rivers. Proceedings of the Royal Society of London A372: 453-473 Baker V.R. and Ritter, D.F., 1975. Competence of Rivers to transport coarse bed load material. Geological Society of America Bulletin 86: 975?978. Bates, C.G., and Henry A.J., 1928. Forest and streamflow experiment at Wagon Wheel Gap, Colorado, U.S. Weather Bureau Monthly Weather Review, Supplement 30, 79p. Bathurst, J.C., Graf, W.H. and Cao, H.H., 1983. Initiation of sediment transport in steep channels with coarse bed material. In B.M. Sumer and A. Mueller (Editors), Mechanics of Sediment Transport, pp. 207?213, Rotterdam. Bathurst, J., Graf W., and Cao H., 1987. Bed load discharge equations for steep mountain rivers. In: C.R. Thorne J.C. Bathurst and R.D Hey (Editors) Sediment Transport in Gravel-Bed Rivers,. Wiley: New York; 453?477. Bathurst, J.C., 2007. Effect of coarse surface layer on bed-load transport. Journal of Hydraulic Engineering 133: 1192?1205.  Bathurst, J.C., Amezaga, J., Cisneros, F., Gavi?o Novillo, M., Iroum?, A., Lenzi, M.A., Aguirre, J., Miranda, M., and A. Urciuolo, 2009. Forest and Floods in Latin America: Science, Management, Policy and the Epic Force Project. Downloaded on April 23, 2012 from; http://research.ncl.ac.uk/epicforce/assets/EPIC%20FORCE%20Year%203%20final%20activity%20report.pdf.  Bathurst, J.C., Birkinshaw, S.J. Cisneros, F., Fallas, J., Iroum?, A., Iturraspe, R., Novillo, M.G., Urciuolo, A., Alvarado, A., Coello, C., Huber, A., Miranda, M., Ramirez, M., and Sarandon R., 2011b. Forest impact on floods due to extreme rainfall and  178 snowmelt in four Latin American environments 2: Model analysis, Journal of Hydrology, 400, 292?304, doi:10.1016/j.jhydrol.2010.09.001. Bathurst, J.C., Iroum?, A., Cisneros, F., Fallas, J., Iturraspe, R., Novillo, M.G., Urciuolo, A., de Bievre, B., Borges, V.G., Coello, C., Cisneros, P., Gayoso, J., Miranda, M., and Ramirez M., (2011a), Forest impact on floods due to extreme rainfall and snowmelt in four Latin American environments 1: Field data analysis, J. Hydrol., 400, 281?291, doi:10.1016/j.jhydrol.2010.11.044. Bayazit, M., 1978. Scour of bed material in very rough channels. Journal of the Hydraulics Division, 104(HY9), 1345? 1349. Beaudry, P.G., and A. Gottesfeld. 2001. Effects of forest-harvest rates on stream channel changes in the central interior of British Columbia. In: D.A.A. Toews and S. Chatwin, (Editors), Watershed Assessment in the Southern Interior of British Columbia. Workshop proceedings, 9-10 March 2000, Penticton, British Columbia, Canada.. British Columbia Ministry of Forests, Research Program, Victoria, Working Paper 57. Pages 151-173. Benda, L. and Dunne, T., 1997. Stochastic forcing of sediment routing and storage in channel networks. Water Resources Research 33, 2865?2880. Benda, L, Hassan, M.A., Church, M., and May, C.L., 2005. Geomorphology of Steepland Headwaters: The transition From Hillslopes to Channels. Journal of the American Water Resources Association 41(4):835-851. Benda, L., Miller D., Dunne T., Agee J., and Reeves G., 1998. Dynamic landscape systems. In R. Naiman  and R. Bilby (Editors.), River Ecology and Management: Lessons from the Pacific Coastal Ecoregion, Springer-Verlag, Reiskirchen, Germany; 261-288. Berris, S.N., and Harr R.D., 1987. Comparative snow accumulation and melt during rainfall in forested and clear-cut plots in the Western Cascades of Oregon, Water Resources Research., 23 (1), 135?142. Beschta, R.L., 1978. Long-term patterns of sediment production following road construction and logging in the Oregon Coast Range, Water Resources Research, 14, 1011 ? 1016, doi:10.1029/WR014i006p01011. Beschta, R.L., Pyles, M.R., Skaugset, A.E., and Surfleet, C.G., 2000. Peakflow response to forest practices in the western Cascades of Oregon, USA, Journal of Hydrology, 233, 102?120. Bilby, R.E., 1981. Role of Organic Debris Dams in Regulating the Export of Dissolved and Particulate Matter from a Forested Watershed. Ecology 62:1234?1243. http://dx.doi.org/10.2307/1937288  179 Birkinshaw, S.J., Bathurst, J.C., Iroum?, A.and Palacios H., 2010. The effect of forest cover on peak flow and sediment discharge: An integrated field and modelling study in central-southern Chile, Hydrol. Processes, 25(8), 1284?1297, doi: 10.1002/hyp.7900. Booth, D., 1990. Stream-channel incision following drainage-basin urbanization. Water Resources Bulletin. 26(3): 407-417. Brardinoni, F., O. Slaymaker, and M. A. Hassan (2003), Landslide inventory in a rugged forested watershed: A comparison between remotely sensed and field survey data, Geomorphology, 54, 179?196. Brardinoni, F., and Hassan M.A., 2006. Glacial erosion, evolution of river long profiles, and the organization of process domains in mountain drainage basins of coastal British Columbia. Journal of  Geophysical Research 111, F01013. doi:10.1029/2005JF000358. Brardinoni, F., and Hassan, M.A., 2007. Glacially-induced organization of channel-reach morphology in mountain streams. Journal of Geophysical Research 112, F03013. doi:10.1029/2006JF000741. Brath, A., and Montanari, A., 2000. The effects of the spatial variability of soil infiltration capacity in distributed flood modelling, Hydrological Processes, 14, 2779? 2794, doi:10.1002/1099?1085(20001030)14:15<2779::AID-HYP91>3.0.CO;2?X. Brath, A., Montanari, A., and G. Moretti, G., 2006. Assessing the effect of flood frequency of land use change via hydrological simulation (with uncertainty), Journal of Hydrology, 324, 141 ? 153, doi:10.1016/j.jhydrol. 2005.10.001. Brayshaw, D.D., 2012. Bankfull and effective discharge in small mountain streams of British Columbia. Unpublished PhD dissertation, University of British Columbia, Vancouver B.C., 173p. Brayshaw, A.C., Frostick, L.E., and Reid I. 1983. The hydrodynamics of particle clusters and sediment entrainment in coarse alluvial channels. Sedimentology 30: 137-143 British Columbia Forest and Range Practices Act: Forest Planning and Practices Regulation (2004). Downloaded Feb, 2012, from: http://www.bclaws.ca/EPLibraries/bclaws_new/document/ID/freeside/14_2004 Brooks, K.N., Ffolliott, P.F. Gregersen, H.M., and DeBano, L.F., 2003., Hydrology and the Management of Watersheds, Blackwell, Ames, Iowa. Brummer, C.J., and Montgomery, D.R., 2003. Downstream coarsening in headwater channels. Water Resources Research 39, 1294. doi:10.1029/2003WR001981.  180 Buffington J.M and Montgomery, D.R., 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resources Research 33(8), 1993?2029. Buffington J.M and Montgomery, D.R., 1999. Effects of hydraulic roughness on surface textures of gravel-bed rivers. Water Resources Research 35(11), 3507-3521. Buffington, J.M., Woodsmith, R.D. Booth, D.B. and Montgomery D.R.. 2003. Fluvial processes in Puget Sound Rivers and the Pacific Northwest. In: D.R. Montgomery, S. Bolton, D.B. Booth and L. Wall (Editors). Restoration of Puget Sound rivers. Univ. Wash. Press, Seattle, Wash., pp.46?78. Bunnell, F.L., Kremsater, L.L., and Houde I., 2011. Mountain pine beetle: A synthesis of the ecological consequences of large-scale disturbances on sustainable forest management, with emphasis on biodiversity, Information Report BC?X-426, Natural Resources Canada, Canadian Forest Service, Pacific Forestry Centre, Victoria, BC.  Bunte, K. 1996. Analyses of the temporal variation of coarse bedload transport and its grain size distribution (Squaw Creek, Montana, USA). General Technical Report RM-GTR-288, USDA Forest Service, Rocky Mountain Forest and Range Experimental Station: Fort Collins, CO.  Bunte, K., Abt, S.R., Swingle,K.W., and Potyondy,J.P. 2010. Bankfull mobile particle size and its predictions from a Shields-type approach. Proceedings of the 4th Federal Interagency Hydrologic Modeling Conference and the 9th Federal Interagency Sedimentation Conference, Las Vegas, NV, 27 June-I July, 2010. Session: sediment transport. CD-ROM ISBN 978-0-9779007-3-2. Burton, T.A., 1997. Effects of basin-scale timber harvest on water yield and peak streamflow, Journal of the American Water Resources Association, 33(6), 1187?1194. Buttle, J.M., 2011. The effects of forest harvesting on forest hydrology and biogeochemistry. In: D. Levia, D. Carlyle-Moses, and T. Tanaka (Editors), Forest Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions, pp. 659?674, Ecological Studies Series, No. 216, Springer-Verlag, Heidelberg, Germany. Calder, I.R., 2005. Blue Revolution: Integrated Land and Water Resource Management, Earthscan, London.  Calder, I.R., Smyle, J., and Aylward B., 2007. Debate over flood-proofing effects of planting forests, Nature, 450, 945, doi:10.1038/450945b  Canovaro, F, Paris, E., and Solari, L., 2007. Effects of macro-scale bed roughness geometry on flow resistance. Water Resources Research 43: W10414, DOI:10.1029/2006WR005727.  181 Chang, M., 2006, Forest and streamflow. In: Forest Hydrology: An Introduction to Water and Forests, chap. 10, pp. 191? 236, CRC Press, Boca Raton, Fla. Cheng, J.D., 1989. Streamflow changes after clear-cut logging of a pine beetle-infested watershed in southern British Columbia, Canada, Water Resources Research 25, 449?456, doi:10.1029/WR025i003p00449 Chiari, M., and Rickenmann D., 2011. Back-calculation of bedload transport in steep channels with a numerical model. Earth Surface Processes and Landforms 36: 805?815. DOI. 10.1002/esp.2108 Chiari, M., Friedl, K., and Rickenmann D., 2010. A one dimensional bedload transport model for steep slopes. Journal of Hydraulic Research 48: 152?160. Church, M., 2002. Geomorphic thresholds in riverine landscapes. Freshwater Biology 47: 541?557. Church, M., 1978. Palaeohydrological reconstructions from a Holocene valley fill. In: A.D. Miall (Editor) Fluvial Sedimentology, Memoirs of the Canadian Society of Petroleum Geolologists 5: 743?772. Church, M., 1988. Floods in Cold Climates. In V.R. Baker, R.C. Kochel and P.C. Patton, (Editors), Flood Geomorphology. New York, Wiley-Interscience: 205-229 Church, M., 1992. Channel morphology and topology. In: P. Calow and G.E. Petts, (Editors), The River Handbook, Blackwell, Oxford, UK, pp. 126-143. Church, M., 2006. Bed material transport and the morphology of alluvial river channels. Annual Review of Earth and Planetary Sciences 34, 325-354. Church, M., 2010. Mountains and montane channels. In: T. Burt and R. Allison (Editors), Sediment Cascades, Wiley-Blackwell: Chichester; 17-53. Church, M., and Hassan, M.A., 2002. Sediment mobility in Harris Creek. Water Resources Research 38: 1237. DOI: 10.1029/2001WR000753. Church, M., and Hassan, M.A., and Wolcott, J.F., 1998. Stabilizing self-organized structures in gravel-bed streams: field and experimental observations. Water Resources Research 34: 3169?3179. Church, M., and Slaymaker, O., 1989. Disequilibrium of Holocene sediment yield in glaciated British Columbia. Nature 337, 452?454. Church, M., Ham, D., Hassan, M., and Slaymaker, O., 1999. Fluvial clastic sediment yields in Canada: scaled analysis. Canadian Journal of Earth Sciences 36, 1267?1280.  182 Church, M., McLean, D.G., and Wolcott, J.F., 1987. River Bed Gravels: Sampling and Analysis. In: C.R. Thorne, J.C. Bathurst and R.D. Hey (Editors) Sediment Transport in Gravel-bed Rivers,. Wiley & Sons Ltd.: Chichester; 43-88. CIFOR and FAO, 2005. Forest and floods: drowning in fiction or thriving on facts? From http://www.cgiar.org/insightdev/upload/291/145_BCIFOR0501.pdf Clague, J.J, Armstrong, J.E., and Mathews, W.H., 1980. Advance of the late Wisconsin Cordilleran Ice Sheet in southern British Columbia since 22,000 Yr B.P. Quaternary Research 13, 322-326. Clague, J.J. (compiler), 1989. Quaternary geology of the Canadian Cordillera, p. 15-96. In: R.J. Fulton, (Editor), Quaternary geology of Canada and Greenland. Geological Survey of Canada, Geology of Canada 1, 839 pp. (also Geological Society of America, The Geology of North America, K-1). Comiti, F., Cadol, D., and Wohl, E., 2009. Flow regimes, bed morphology, and flow resistance in self-formed step-pool channels. Water Resources Research 45: W04424. DOI: 10.1029/2008WR007259 Costa, J.E., 1983. Paleohydraulic reconstruction of flash-flood peaks from boulder deposits in the Colorado Front Range. Geological Society of America Bulletin 94: 986-1004. Costa, J.E., and O'Connor, J.E., 1995. Geomorphically Effective Floods. In: J.E. Costa, A.J. Miller, K.W. Potter and P.R. Wilcock (Editors), Natural and Anthropogenic Influences in Fluvial Geomorphology, American Geophysical Union, Washington, D. C.. doi: 10.1029/GM089p0045 Curran, J.H., and Wohl, E.E., 2003. Large woody debris and flow resistance in step-pool channels, Cascade Range, Washington. Geomorphology 51: 141?157. D?Agostino, V., and Lenzi, M.A., 1999. Bedload transport in the instrumented catchment of the Rio Cordon, Part II: Analysis of the bedload rate. Catena 36 : 191-204. Dalrymple, T., 1960. Flood-frequency analysis, USGS Water Supply Paper, 1543-A. David, G.C.L., Wohl, E., Yochum, S.E., and Bledsoe, B.P., 2010. Controls on spatial variations in flow resistance along steep mountain streams. Water Resources Research 46 : W03513, DOI:10.1029/2009WR008134. Day, T.J., 1976. On the precision of salt dilution gauging. Journal of Hydrology 31, 293-306. DeWalle, D.R., 2003. Forest hydrology revisited, Hydrological Processes, 17, 1255? 1256, doi:10.1002/hyp.5115 Dhakal, A.S., and Sidle, R.C., 2003. Long?term modelling of landslides for different forest management practices. Earth Surface Processes and Landforms 28(8): 853-868.  183 Dietrich, W.E., Kirchner, J.W., Ikeda, H., and Iseya F., 1989. Sediment supply and the development of the coarse surface layer in gravel-bedded rivers. Nature 340: 215?217.  Dodds, G.B., 1969. The stream-flow controversy: A conservation turning point, The Journal of American History, 56, 59?69.  Dodov, B., and Foufoula-Georgiou, E., 2005. Fluvial processes and streamflow variability: Interplay in the scale-frequency continuum and implications for scaling, Water Resources Research, 41, W05005, doi:10.1029/2004WR003408 Doten, C.O., Bowling, L.C., Lanini, J.S., Maurer, E.P., and Lettenmaier, D.P., 2006. A spatially distributed model for the dynamic prediction of sediment erosion and transport in mountainous forested watersheds. Water Resources Research 42: W04417, doi:10.1029/2004WR003829. Eagleson, P.S., 1972. Dynamics of flood frequency, Water Resources. Research, 8(4), 878?898, doi:10.1029/WR008i004p00878.  Eaton, B.C., Millar, R.G., 2004. Optimal alluvial channel width under a bank stability constraint. Geomorphology 62: 35?45. Einstein, H.A., 1950. The bed-load function for sediment transportation in open channel flows, Technical Bulletin No. 1026. United States Department of Agriculture: Washington, DC Ellis, C.R., Pomeroy, J.W., Essery, R.L.H., and Link, T.E., 2010. Effects of needle-leaf forest cover on radiation and snowmelt dynamics in the Canadian Rocky Mountains,  Canadian Journal of Forest Research, 41, 608?620, doi:10.1139/X10-227. Emmett, W.W., 1972. The hydraulic geometry of some Alaskan streams south of the Yukon River. U.S. Geological Survey Open-File Report 72-108, 108 pp. Emmett WW. 1980. A field calibration of the sediment-trapping characteristics of the Helley-Smith bedload sampler: Professional Paper 1139, United States Geological Survey. Emmett WW. 2010. Observations of Bedload Behavior in Rivers and Their Implications for Indirect Methods of Bedload Measurement. United States Geological Survey Scientific Investigations Report 2010-5091.Emmett, W.W., and Wolman, M.G., 2001. Effective discharge and gravel-bed rivers. Earth Surface Processes and Landforms 26: 1369 ?1380. Fannin, R.J., and Rollerson, T., 1993. Debris flows: some physical characteristics and behaviour. Canadian Geotechnical Journal 30, 71-82.  184 Faustini, J.M., and Jones, J.A.,2003. Influence of large woody debris on channel morphology and dynamics in steep, boulder-rich mountain streams, western Cascades, Oregon. Geomorphology 51: 187?205 Fenton, J.D., and Abbott, J.E., 1977. Initial movement of grains on a stream bed: The effect of relative protrusion. Proceedings of the Royal Society of London Ser. A. 352: 523? 537. Ferguson, R.I., Prestegaard, K..L, and Ashworth, P.J., 1989. Influence of sand on hydraulics and gravel transport in a braided gravel bed river. Water Resources Research 25: 635?643 Fernandez Luque, R., and van Beek, R., 1976. Erosion and transport of bed?load sediment. Journal of Hydraulic Research 14: 127-144. Fitzpatrick, F.A., Knox, J.C., and Whitman, H.E., 1999 Effects of historical land-cover changes on flooding and sedimentation, North Fish Creek, Wisconsin. USGS Water-Resources Investigations Report 99?4083: 12p. Flynn, K.F., and Van Liew, M.W., 2011. Evaluation of SWAT for sediment prediction in a mountainous snowmelt-dominated catchment. Transactions of the American Society of Agricultural and Biological Engineers 54: 113-122. Frey, P., and Church, M.. 2009. How river beds move. Science 325: 1509?1510. Fritze, H., Stewart, I.T., and Pebesma, E., 2011. Shifts in Western North American Snowmelt Runoff Regimes for the Recent Warm Decades. Journal of Hydrometeorology 12, 989?1006. doi: 10.1175/2011JHM1360.1 Gaeuman, D., Schmidt, J.C, and Wilcock, P.R., 2005. Complex channel responses to changes in stream flow and sediment supply on the lower Duchesne River, Utah. Geomorphology 64: 185?206 Ghilardi, T., and Schleiss, A.J., 2011. Influence of immobile boulders on bedload transport in a steep flume. 34th IAHR World Congress - Balance and Uncertainty, 33rd Hydrology & Water Resources Symposium, 10th Hydraulics Conference, 26 June - 1 July 2011, Brisbane, Australia. Gilbert, G.K., 1914. The transportation of debris by running water. USGS Professional Paper 86. US Government Printing Office, Washington D.C.  Gluns, D.R., 2001. Snowline pattern during the melt season: Evaluation of the H60 concept. In: D.A.A. Toews, and S. Chatwin, (Editors), Watershed Assessments in the Southern Interior of British Columbia: Workshop Proceedings, March 9-10, 2000, Penticton, British Columbia. British Columbia Ministry of Forests, Research Branch, Working Paper 57, Victoria, B.C., pp. 68-80.  185 Golding, D. L., and Swanson, R. H., (1978), Snow accumulation and melt in small forest openings in Alberta, Canadian Journal of Forest Research 8, 380?388. Gomi, T., and Sidle, R.C., 2003. Bed load transport in managed steep-gradient headwater streams of southeastern Alaska. Water Resources Research 39: 1336. DOI.10.1029/2003WR002440. Gomi, T., Sidle, R.C., and Richardson, J.S., 2002. Understanding processes and downstream linkages of headwater systems. Bioscience 52: 905?916.  Gomi, T., Sidle, R.C., Bryant, M.D., and Woodsmith, R.D., 2001. The characteristics of woody debris and sediment distribution in headwater streams, southeastern Alaska. Canadian Journal of Forest Research 31: 1386?1399. Goodell, B.C. 1958. A preliminary report on the first year?s effects of timber harvesting on water yield from a Colorado watershed. Paper 36, USDA Forest Service, Rocky Mountain Forest and Range Experimental Station, Fort Collins, Colorado Gordon E., and Meentemeyer, R.K., 2006. Effects of dam operation and land use on stream channel morphology and riparian vegetation. Geomorphology 82(3-4): 412-429. Graf, W.L. 1975. The impact of suburbanization on fluvial geomorphology. Water Resources Research. 11:690-692. Graf, W.H., 1991. Flow resistance over a gravel bed: Its consequences on initial sediment movement. In, A. Armanini, and G. Di Silvio (Editors), Fluvial Hydraulics in Mountain Regions pp. 17?32, Springer-Verlag, Berlin Grant, G.E., Lewis, S.L., Swanson, F.J., Cissel, J.H., and McDonnell J.J., 2008. Effects of forest practices on peak flows and consequent channel response: A state-of-science report for western Oregon and Washington, General Technical Report PNW-GTR-760, USDA Forest Service, Pacific Northwest Research Station, Portland, Oregon. Grant, G.E., Swanson, F.J., and Wolman, M.G., 1990. Pattern and origin of stepped-bed morphology in high gradient streams, Western Cascades, Oregon. Geological Society of America Bulletin 102: 340?352. Green, K.C., and Alila, Y. 2012. A paradigm shift in understanding and quantifying the effects of forest harvesting on floods in snow environments. Water Resources Research, 48: W10503, doi:10.1029/2012WR012449 Green, K.C., Brardinoni, F., and Alila, Y., 2012. Channel Morphology and bed-load yield in fluvial, formerly-glaciated headwater streams of the Columbia Mountains, Canada. Geomorphology. 188: DOI:10.1016/j.geomorph.2012.05.004 Gupta, V. K. and Waymire, E., 1998. Spatial variability and scale invariance in hydrologic regionalization. In: G. Sposito (Editor), Scale Dependence and Scale Invariance in Hydrology, Cambridge University Press, pp. 88-135.   186 Guthrie, R.H., 2002. The effects of logging on frequency and distribution of landslides in three watersheds on Vancouver Island, British Columbia.Geomorphology 43(3?4):275?294. Halleran, W., 2004. Terrain map of Cotton Creek watershed. Prepared for Tembec Industries, Cranbrook., B.C. Halwas, K.L., and Church, M., 2002. Channel units in small, high gradient streams on Vancouver Island, British Columbia. Geomorphology 43: 243? 256. Hammer, T.R., 1972. Stream channel enlargement due to urbanization. Water Resources Research 8: 1530-1540. Hassan, M.A., Church, M., Lisle, T.E., Brardinoni, F., Benda, L., and Grant, G.E., 2005. Sediment Transport and Channel Morhpology of Small, Forested Streams. Journal of the American Water Resources Association 41, 853-876. Hassan, M.A., Smith, B.J., Hogan, D.L., Luzi, D.S., Zimmermann, A.E., and Eaton, B.C., 2007. Sediment storage and transport in coarse bed streams: Scale considerations. In: H. Habersack, H. Piegay, and M. Rinaldi (Editors), Gravel-bed Rivers VI: From Process Understanding to River Restoration. Elsevier B.V; 473-496. Heede, 1985; Channel Adjustments to the Removal of Log Steps: an Experiment in a Mountain Stream. Environmental Management 9(5): 427-432.  Heller, P.L., Beland, P.E., Humphrey, N.F., Konrad, S.K., Lynds, R.M., McMillan, M.E., Valentine, K.E., Widman, Y.A., and Furbish, D.J., 2001. Paradox of downstream fining and weathering-rind formation in the lower Hoh River, Olympic Peninsula, Washington. Geology 29, 971-974. Hess, S., 1984. Timber harvesting and flooding, Journal of Soil Water Conservation, 39, 115? 117. Hewlett, J.D., 1982. Forests and floods in the light of recent investigations. In Proceedings of the Canadian Hydrology Symposium, pp. 543?560, National Research Council of Canada, Ottawa, Ontario.  Hicks, N.S., Smith, J.A., Miller, A.J., and Nelson, P.A., 2005. Catastrophic flooding from an orographic thunderstorm in the central Appalachians. Water Resources. Research 41: W12428, doi:10.1029/2005WR004129. Hock, R. 2003. Temperature index melt modelling in mountain areas. Journal of Hydrology, 282, 104?115. Hogan, D.L., 1989. Channel response to mass wasting in the Queen Charlotte Islands, British Columbia: Temporal and spatial changes in stream morphology. In: Proceedings Watersheds ?89: Conference on the Stewardship of Soil, Air and Water  187 Resources, Juneau, Alaska, March 21?23, 1989. USDA Forest Service, Alaska Region, R10-MB-77, pp. 125?144.  Holland, S.S., 1964. Landforms of British Columbia: A Physiographic Outline. British Columbia, Department of Mines and Petroleum Resources, Bulletin 48, Victoria, B.C., 138 pp. Howe, G.M., Slaymaker, H.O., and Harding, D., 1966. Flood hazard in Mid-Wales. Nature, 212(5062): 584-585. Howes, D.E., and Kenk, E., 1997. Terrain Classification System for British Columbia (Version 2) 1997: a system for the classification of surficial materials, landforms and geological processes of British Columbia. In: D.E. Howes, and E. Kenk, (Editors), British Columbia Ministry of Environment, Lands, and Parks, Resource Inventory Branch, Manual 10, Victoria, B.C. 102 pp. H?y, T., 1993. Geology of the Purcell Supergroup in the Fernie west-half map area, southeastern British Columbia. British Columbia Ministry of Energy Mines and Petroleum Resources Bulletin 84. Hoyt, W.G., and Troxell, H.C., 1932. Forests and streamflow. Proceedings of the American Society of Civil Engineers, 58, Aug., pp. 1037?1066 Huckleberry, G., 1994. Contrasting channel response to floods on the middle Gila River, Arizona. Geology 22: 1083-1086. Jackson, L.E., and Clague, J.J., 1991. The Cordilleran Ice Sheet: One Hundred and Fifty Years of Exploration and Discovery. G?ographie physique et Quaternaire 45, 269-280. Jackson, W.L., and Beshta, R.L, 1982. A model of two-phase bed load transport in an Oregon Coast Range stream. Earth Surface Processes and Landforms 7, 515-527.  Jakob, M., Bovis, M., and Oden, M., 2005. The significance of channel recharge rates for estimating debris-flow magnitude and frequency. Earth Surface Processes and Landforms 30, 755-766. doi 10.1002/esp.1188 Jeffrey, W.W., 1970. Hydrology of land use. In: D.M.Gray, (Editor), Handbook on the Principles of Hydrology, pp. 13.1?13.57, Water Inf. Cent., Port Washington, N.Y. Jones, J.A. 2000. Hydrologic processes and peak discharge response to forest removal, regrowth, and roads in 10 small experimental basins, western Cascades, Oregon. Water Resources Research 36(9): 2621-2642. Jones, J.A., 2005. Intersite comparisons of rainfall-runoff processes. In: M. Anderson, (Editor) Encyclopedia of Hydrological Sciences, pp 1839?1854, John Wiley, Chichester, 2005.  188 Jones, J.A., and Grant, G.E., (1996), Peak flow responses to clear-cutting and roads in small and large basins, western Cascades, Oregon, Water Resources Research, 32, 959? 974, doi:10.1029/95WR03493. Jones, J.A., and Perkins, R.M., 2010. Extreme flood sensitivity to snow and forest harvest, western Cascades, Oregon, United States. Water Resources Research 46: W12512, doi:10.1029/2009WR008632. Jones, J.A., Achterman, G.L., Augustine, L.A., Creed, I.F., Ffolliott, P.F., MacDonald, L., and Wemple, B.C., 2009. Hydrologic effects of a changing forested landscape?challenges for the hydrological sciences. Hydrological Processes 23: 2699?2704, doi: 10.1002/hyp.7404. Jordan, P., 2002. Landslide frequencies and terrain attributes in Arrow and Kootenay Lake forest districts. In: P. Jordan, and J. Orban, (Editors), Terrain Stability and Forest Management in the Interior of British Columbia: Workshop Proceedings, Nelson, B.C., May 23?25, 2001. British Columbia Ministry of Forests, Forest Sciences Program, Victoria, B.C. Tech. Rep. 003, pp. 80-102. Jordan, P., 2006. The use of sediment budget concepts to assess the impact on watersheds of forestry operations in southern interior of British Columbia. Geomorphology 79, 27-44. doi:10.1016/j.geomorph.2005.09.019  Jost, G., Weiler, M., Gluns, D.R., and Alila, Y., 2007. The influence of forest and topography on snow accumulation and melt at the watershed-scale. Journal of Hydrology 347, 101-115. doi:10.1016/j.jhydrol.2007.09.006 Kattelmann, R., 1991. Peak flows from snowmelt runoff in the Sierra Nevada, USA. In: H. Bergmann, H. Land, W. Frey, D. Issler, and B. Salm (Editors), Snow Hydrology and Forests in High Alpine Areas, pp. 203?211, IAHS Pub. No. 205.  Katz, R.W., and Brown, B.G., 1992. Extreme events in changing climate: Variability is more important than averages. Climate Change, 21, 289? 302, doi:10.1007/BF00139728. Katz, R.W.,1993. Towards a statistical paradigm for climate change, Climate Research, 2, 167? 175, doi:10.3354/cr002167. King, J.G., 1989. Streamflow responses to road building and harvesting: A comparison with the equivalent cut area procedure. U.S. Research Paper NT-401, USDA Forest Service, Intermountain Research Station,Ogden, Utah. King, J.G., Emmett, W.W., Whiting, P.J., Kenworthy, R.P., and Barry, J.J., 2004. Sediment Transport Data and Related Information for Selected Coarse-Bed Streams and Rivers in Idaho. General Technical Report RMRS-GTR 131. USDA Forest Service, Rocky Mountain Research Station. Boise, ID.   189 Kiss, T., and Blanka, V., 2012. River channel response to climate- and human-induced hydrological changes: Case study on the meandering Hern?d River, Hungary. Geomorphology, 175?176: 115?125. http://dx.doi.org/10.1016/j.geomorph.2012.07.003 Klingemann, P.C., and Emmett, W.W., 1982. Gravel bedload transport processes. In: R.D. Hey, J.C. Bathurst, and C.R. Thorne, (Editors), Gravel-bed Rivers, Wiley: Chichester; 141?169. Knox, J.C., 1977. Human impacts on Wisconsin stream channels. Annals of the Association of American Geographers 67(3): 323-342. doi:10.1111/j.1467-8306.1977.tb01145.x Komar, P.D., 1987. Selective gravel entrainment and the empirical evaluation of flow competence. Sedimentology 34: 1165?1176. Komar, P.D., and Shih, S.M., 1992. Equal mobility versus changing bedload grain sizes in gravel-bed streams. In: P. Billi, R.D. Hey and C.R. Thorne, (Editors), Dynamics of Gravel-Bed Rivers, John Wiley and Sons: Chichester; 73? 93. Kondolf, G.M., and Pi?gay, H. 2003. Tools in Fluvial Geomorphology. John Wiley & Sons, London. Kondolf, M., Lisle, T., and Wolman, G., 2003. Bed sediment measurement. In: M. Kondolf, H. Piegay, (Editors), Tools in Fluvial Geomorphology. Wiley, Chichester, UK, pp. 347-395. Konrad, C.P., Booth, D.B., and Burges, S.J., 2005. Effects of urban development in the Puget Lowland, Washington, on interannual streamflow patterns: Consequences for channel form and streambed disturbance. Water Resources Research 41: W07009, doi:10.1029/2005WR004097 Kuhnle, R.A., 1992. Fluvial transport of sand and gravel mixtures with bimodal size distributions. Sedimentary Geology 85, 17-24.  Kura?, P.K., Alila, Y., and Weiler, M., 2012. Forest harvesting effects on the magnitude and frequency of peak flows can increase with return period. Water Resources Research 48(1), doi:10.1029/2011WR010705, in print. Kura?, P.K., Alila, Y., Weiler, M., Spittlehouse, D., and Winkler R., 2011. Internal catchment process simulation in a snow-dominated basin: Performance evaluation with spatiotemporally variable runoff generation and groundwater dynamics. Hydrological Processes, 25, doi:10.1002/hyp.8037. Lamb, M.P., Dietrich, W.E., and Venditti, J.G., 2008. Is the critical Shields stress for incipient sediment motion dependent on channel-bed slope? Journal of Geophysical Research 113: F02008. DOI: 10.1029/2007JF000831 Lee, R., 1980. Forest Hydrology, Columbia University Press, New York.  190 Lenzi, M.A., D'Agostino, V., and Billi, P., 1999. Bedload transport in the instrumented catchment of the Rio Cordon, Part I: analysis of bedload records, conditions and threshold of bedload entrainment. Catena 36: 171?190.  Lenzi, M.A., Mao, L., and Comiti, F., 2006. When does bedload transport begin in steep boulder-bed streams? Hydrological Processes 20: 3517?3533. DOI: 10.1002/hyp.6168 Leopold, L., and Maddock, T., 1953. The Hydraulic Geometry of Stream Channels and Some Physiographic Implications. U.S. Geological Survey Professional Paper 252, 57 pp. Lewis, J., Reid, L.M., and Thomas, R.B., 2010. Comment on ?Forests and floods: A new paradigm sheds light on age-old controversies? by Younes Alila et al., Water Resources Research, 46, W05801, doi:10.1029/2009WR008766. Lieberman, J.A., and Hoover, M.D., 1951. Stream-Flow Frequency Changes on Coweeta Experimental Watersheds, Transactions of the American geophysical Union.Vol.32. No. 1, 73?76.Lisle, T.E., 1986. Stabilization of a Gravel Channel by Large Streamside Obstructions and Bedrock Bends, Jacoby Creek, Northwestern California. Geological Society of America Bulletin 97, 999-1011. MacDonald, J.S., Beaudry, P.G., MacIsaac, E.A., and Herunter. H.E., 2003. The effects of forest harvesting and best management practices on streamflow and suspended sediment concentrations during snowmelt in headwater streams in sub-boreal forests of British Columbia, Canada. Canadian Journal of Forest Research. 33: 1397-1407. MacDonald, L.H., and Coe, D., 2007. Influence of Headwater Streams on Downstream Reaches in Forested Areas. Forest Science 53(2): 148-168. MacDonald, L.H., and Stednick, J., 2003. Forests and water: A state of-the-art review for Colorado, Completion Rep., 196, 65 pp., Colorado Water Resources Research Institute, Fort Collins. Mao, L., and Lenzi, M.A., 2007. Sediment mobility and bedload transport conditions in an alpine stream. Hydrological Processes 21: 1882-1991. DOI: 10.1002/hyp.6372. Mao, L., Uyttendaelle, G.P., Iroum?, A., and Lenzi, M.A., 2008. Field based analysis of sediment entrainment in two high gradient streams located in Alpine and Andine environments. Geomorphology 93: 368?383 Marion, D.A., and Weirich, F., 2003. Equal-mobility bed load transport in a small, step-pool channel in the Ouachita Mountains. Geomorphology 55: 139-154. Massong, T.M., and Montgomery, D.R., 2000. Influence of sediment supply, lithology, and wood debris on the distribution of bedrock and alluvial channels. Geological Society of America Bulletin 112(5): 591?599.  191 May, C.L., and Gresswell, R.E., 2003. Processes and rates of sediment and wood accumulation in headwater streams the Oregon Coast Range, USA. Earth Surface Processes and Landforms 28: 409-424. McCabe, G.J., Clark, M.P., and Hay, L.E., 2007. Rain-on-snow Events in the Western United States. Bulletin of the American Meteorological Society 88, 319-328. McDonnell, J.J., Sivapalan, M., Vach?, K., Dunn, S., Grant, G., Haggerty, R., Hinz, C., Hooper, R., Kirchner, J., Roderick, M.L., Selker, J., and Weiler, M., 2007. Moving beyond heterogeneity and process complexity: A new vision for watershed hydrology. Water Resources Research, 43, W07301, doi:10.1029/2006WR005467. Mearns, L.O., Katz, R.W., and Schneider, S.H., 1984. Extreme high temperature events: Changes in their probabilities with changes in mean temperature, Journal of Climate and Applied Meteorology 23, 1601 ? 1613, doi:10.1175/1520-0450(1984)02320051601:EHTECI>2.0.CO;2.1984. Merz, J., and Doppmann, G.. 2006. Measuring Mountain Stream Discharge Using the Salt Dilution Method ? A Practical Guide. International Centre for Integrated Mountain Development Kathmandu. Kathmandu 15pp. Meyer, J.L., and Wallace, J.B., 2001. Lost linkages and lotic ecology: Rediscovering small streams. In: M.C. Press, N.J. Huntly, and S. Levin, (Editors), Ecology: Achievement and Challenge, pp. 295?317 Oxford (United Kingdom): Blackwell Scientific.  Meyer-Peter, R., and M?ller, R., 1948. Formulas for bedload transport. In: Proceedings 2nd Meeting International Association of Hydraulic Research, pp. 39-64, Stockholm. Millar, R.G., 1999. Grain and form resistance in gravel bed rivers. Journal of Hydraulic Research 37: 303-312 Millar, R.G., and Quick, MC., 1993. Effect of bank stability on geometry of gravel rivers.  Journal of Hydraulic Engineering, 119: 1343-1363. Montgomery, D.R., 1999. Process Domains and the River Continuum. Journal of the American Water Resources Association 35(2): 397-410. Montgomery, D.R., 2003.Wood in rivers: interactions with channel morphology and processes. Geomorphology. 51(1-3), 1-5. Montgomery, D.R., and Buffington, J.M., 1997. Channel Reach Morphology in Mountain Drainage Basins. Geological Society of America Bulletin 109, 596-611. Montgomery, D.R., and Buffington, J.M., 1998. Channel Processes, Classification, and Response. In: R.J. Naiman, and R.E. Bilby (Editors), River Ecology and Management: Lesson From the Pacific Coastal Ecoregion. Springer-Verlag, New York, pp. 13-42.  192 Montgomery, D. R., Abbe, T. B., Buffington, J. M., Peterson, N. P., Schmidt, K. M., and Stock, J. D., 1996. Distribution of bedrock and alluvial channels in forested mountain drainage basins. Nature 381: 587?589.  Montgomery, D.R., and Foufoula-Georgiou, E., 1993. Channel network source representation using digital elevation models. Water Resources Research 29, 3925?3934.  Moog, D.B., and Whiting, P.J., 1998. Annual Hysteresis in Bedload Rating Curves. Water Resources Research 34, 2393-2399. Moore, R.D., 2005. Introduction to salt dilution gauging for streamflow measurement Part III: Slug injection using salt in solution. Streamline Watershed Management Bulletin 8(2):1-6 Moore, R.D., and Scott, D.F., 2005. Camp Creek revisited: Streamflow changes following salvage harvesting in a medium-sized, snowmelt-dominated catchment, Canadian Water Resources Journal, 30, 331?344. Moore, R.D., and Wondzell, S.M., 2005. Physical hydrology and the effects of forest harvesting in the Pacific Northwest: A review. Journal of the American Water Resources Association, 41, 763?784. Mueller, E., Batalla, R., Garcia, C., and Bronstert, A. 2008. Modeling Bed-Load Rates from Fine Grain-Size Patches during Small Floods in a Gravel-Bed River. Journal of Hydraulic Engineering, 134(10), 1430?1439. doi: 10.1061/(ASCE)0733-9429(2008)134:10(1430) Mueller, E.R., and Pitlick, J., 2005. Morphologically based model of bed load transport capacity in a headwater stream. Journal of Geophysical Research-Earth Surface 110: F02016. DOI: 10.1029/2003JF000117 Mueller, E.R., Pitlick, J., and Nelson, J.M., 2005. Variation in the reference Shields stress for bed load transport in gravel-bed streams and rivers. Water Resources Research 41: W04006. DOI: 10.1029/2004WR003692 Muskatirovic, J., 2008, Analysis of bedload transport characteristics of Idaho streams and rivers. Earth Surface Processes and Landforms 33: 1757?1768. DOI: 10.1002/esp.1646   Naiman R.J., Bilby, R.E., and Bisson, P.A., 2000. Riparian Ecology and Management in the Pacific Coastal Rain Forest. BioScience 50: 995?1011. Naiman, R.J., Bechtold, J.S., Drake, D.C., Latterell, J.J., O?Keefe, T.C., and Balian, E.V., 2005. Origins, patterns, and importance of heterogeneity in riparian systems. In: G. Lovett, C.G. Jones, M.G. Turner and K.C. Weathers (Editors) Ecosystem Function in Heterogeneous Landscapes, pp. 279?309, Springer-Verlag, New York.  193 Nanson, G.C., 1974. Bedload and Suspended-Load Transport in a Small, Steep Mountain Stream. American Journal of Science 274, 471-486. National Council for Air and Stream Improvement, Inc. (NCASI), 2009. Effects of forest management on water resources in Canada: A research review. Technical Bulletin No. 969. Research Triangle Park, N.C.: National Council for Air and Stream Improvement, Inc. National Research Council, 2008. Hydrologic Effects of a Changing Forest Landscape. National Academies Press, Washington, DC. Nitsche, M., Rickenmann, D., Kirchner, J.W., Turowski, J.M., and Badoux, A., 2012. Macroroughness and variations in reach-averaged flow resistance in steep mountain streams. Water Resources Research 48: doi: 10.1029/2012WR012091 Nitsche, M., Rickenmann, D., Turowski, J.M., Badoux, A., and Kirchner, J.W., 2011. Evaluation of bedload transport predictions using flow resistance equations to account for macro-roughness in steep mountain streams. Water Resources Research 47: W08513. DOI:10.1029/2011WR010645. Oke, T.R., 1987. Boundary Layer Climates, 2nd edition. Methuen. New York. Ouimet, W., and Dethier, D., 2002. Modeling Sediment Flux from Birch Brook, an Undisturbed Catchment in Northwestern Massachusetts. Northeastern Geology and Environmental Sciences 24, 176-184. Parker, G., 1990. Surface-based bedload transport relation for gravel rivers. Journal of Hydraulic Research 28: 417?436.  Parker, G., Toro-Escobar, C.M., Ramey, M., and Beck, S., 2003; Effect of floodwater extraction on mountain stream morphology. Journal of Hydraulic Engineering 129(11): 885-895. DOI:10.1061/(ASCE)0733-9429(2003)129:11(885) Parker, G., and Klingeman, P.C., 1982. On why gravel-bed streams are paved. Water Resources Research 18: 1409?1423. Parker, G., Klingeman, P.C., and McLean, D.G., 1982. Bedload size and distribution in paved gravel-bed stream. Journal of the Hydraulics Division, American Society of Civil Engineers 108: 544?571. Parker, G., Toro-Escobar, C.M., 2002. Equal mobility of gravel in streams: the remains of the day. Water Resources Research 38: 1264?1271. DOI: 10.1029/2001WR000669 this should be cited if it is not. Platt, J.R., 1964, Strong Inference. Science, New Series, 146(3642), 347?353.  194 Poff, N.L., Allan, J.D., Bain, M.B., Karr, J.R., Prestgaard, K.L., Richter, B.D., Sparks, R.E., and Stromberg, J.C., 1997. The natural flow regime: a paradime for river conservation and restoration. Bioscience 47: 769-784. Pomeroy, J., Fang, X., Ellis, C., and Guan, M., 2011. Sensitivity of Snowmelt Hydrology on Mountain Slopes to Forest Cover Disturbance. University of Saskatchewan Centre for Hydrology, Report No. 10, Saskatoon, Sask., 105 pp. Potyondy, J.P., Bunte, K., Abt, S.R., and Swingle, K.W., 2010. Bedload movement in mountain channels: Insights gained from the use of portable bedload traps. 2nd Joint Federal Interagency Conference, Las Vegas, NV, June 27 - July 1, 2010 Powell, D.M., Reid, I., and Laronne, J.B., 2001. Evolution of bed load grain size distribution with increasing flow strength and the effect of flow duration on the caliber of bed load sediment yield in ephemeral gravel bed rivers. Water Resources Research 37: 1463?1474. Reid, I., Frostick, L.E., and Layman, J.T., 1985. The incidence and nature of bedload transport during flood flows in coarse-grained alluvial channels. Earth Surface Processes and Landforms 10, 33-44. Reid, L.M., 1993, Research and cumulative watershed effects, Berkeley, California, Pacific Southwest Research Station: General Technical Report PSW-GTR-141, USDA, Forest Service, 118 p. Resh, V.H., Brown, A.V., Covich, A.P., Gurtz, M.E., Li, H.W., Minshall, G.W., Reice, S.R., Sheldon, A.L., Wallace, J.B., and Wissmar R.C., 1988. The role of disturbance in stream ecology. Journal of North American Benthological Society 7, 433?455. Richardson, J.S., Bilby, R.E. and Bondar, C.A., 2005a. Organic matter dynamics in small streams of the Pacific Northwest Journal of the American Water Resources Association 41:921-934.  Richardson, J.S., Naiman, R.J. Swanson F.J. and Hibbs, D.E., 2005b. Riparian communities associated with Pacific Northwest headwater streams: assemblages, processes, and uniqueness. Journal of the American Water Resources Association 41:935-947. Rickenmann, D., 1997. Sediment Transport in Swiss Torrents. Earth Surface Processes and Landforms 22, 937?951 Rickenmann, D., 2001. Comparison of bed load transport in torrents and gravel-bed streams. Water Resources Research 37: 3295-3305.  Rumsby, B.T., and Macklin, M.G., 1994.  Channel and floodplain response to recent abrupt climate change: The Tyne basin, Northern England. Earth Surface Processes and Landforms 19(6)  195 Ryan, S.E., and Dixon, M.K., 2008. Spatial and temporal variability in stream sediment loads using examples from the Gros Ventre Range, Wyoming, USA. In: H. Habersack, H. Piegay, and M. Rinaldi, (Editors), Gravel-Bed Rivers VI: From Process Understanding to River Restoration. Elsevier, Amsterdam, pp. 387-407. Ryan, S.E., and Emmett, W.W., 2002. The nature of flow and sediment movement in Little Granite Creek near Bondurant, WY. General Technical Report RMRS-GTR-90. USDA Forest Service, Rocky Mountain Research Station: Ogden, UT. Ryan, S.E., and Porth, L.S., 2007. A tutorial on the piecewise regression approach applied to bedload transport data. General Technical Report RMRS-GTR-189. USDA Forest Service, Rocky Mountain Research Station: Fort Collins, CO., 41 pp.  Ryan, S.E., Porth, L.S., and Troendle, C.A., 2005. Coarse sediment transport in mountain streams in Colorado and Wyoming, USA. Earth Surface Processes and Landforms 30, 269-288. doi: 10.1002/esp.1128 Ryder, J.M., 1981. Biophysical Resources of the East Kootenay area. Generalized terrain, British Columbia. British Columbia Ministry of Environment. 250,000.82G, J. Ryder, J.M., Fulton, R.J., and Clague, J.J., 1991. The Cordilleran Ice Sheet and the Glacial Geomorphology of Southern and Central British Columbia. G?ographie Physique et Quaternaire 45, 365-377. Scherer, R., and Pike, G.P. 2003. Effects of forest management activities on streamflow in the Okanagan Basin, British Columbia: Outcomes of a literature review and workshop, FORREX Ser. 9, For. Res. Ext. Partnership (FORREX), Kamloops, B. C., Canada. Schleppi, P., 2011. Forested water catchments in a changing environment. In: M. Bredemeier, S. Cohen, D. Godbold, E. Lode, V. Pichler, and P. Schleppi, (Editors), Forest Management and the Water Cycle: An Ecosystem-Based Approach, , Chap. 5, pp. 89?110, Ecological Studies 212. 531pp. Schmidt, L.J., and Potyondy, J.P., 2004. Quantifying Channel Maintenance Instream Flows: An Approach for Gravel-Bed Streams in the Western United States, General Technical Report RMRS-GTR-128, USDA Forest Service, Rocky Mountain Research Station, Fort Collins, CO. 33 p Schmidt, R.A., and Troendle, C.A., 1992. Sublimation of intercepted snow as a global source of water vapor. In: Proceedings of the 60th Western Snow Conference, 60, pp. 1?9, Jackson, WY. Schnorbus, M., and Alila, Y., 2004. Forest harvesting impacts on the peak flow regime in the Columbia Mountains of southeastern British Columbia: An investigation using long-term numerical modeling, Water Resources Research, 40, W05205, doi:10.1029/2003WR002918.  196 Schnorbus, M., and Alila, Y., 2013, Peak flow changes following forest harvesting in a snow-dominated basin: A modeling exploratory analysis of the effects of harvest area, elevation and channel connectivity. Water Resources Research, 49. DOI: 10.1029/2012WR011901. Schnorbus, M.A., Winkler R.D, and Alila, Y., 2004. Modelling forest harvesting effects on maximum daily peak flow at Upper Penticton Creek. B.C. Ministry of Forests Research Branch, Victoria, B.C. Exten. Note 67. <http://www.for.gov.bc.ca/hfd/pubs/Docs/En/En67.htm> Schultz, C., 2012. Model investigation overthrows assumptions of watershed research, EOS Transactions, AGU, 93(14), 150, doi:10.1029/2012EO140026 Schumm, S.A., 1977. The Fluvial System. New York, John Wiley &Sons, 338 p Shields, A., 1936. Anwendung der Aehnlichkeitsmechanik der Turbulenzforschung auf die Geschiebebewegung. Mitteilungen der Pressische Versochsanstalt for Wasserbau und Schiffbau 26: 5?24 (English translation by Ott WP, van Uchelen JC. 1950. USDA Soil Conservation Service Cooperative Laboratories, California Institute of Technology: Pasadena). Shvidchenko, A.B., and Pender, G., 2000. Flume study of the effect of relative depth on the incipient motion of coarse uniform sediments. Water Resources Research 36: 619?628. DOI: 10.1029/1999WR900312 Shvidchenko, A.B., Pender, G., and Hoey, T.B., 2001. Critical shear stress for incipient motion of sand/gravel streambeds. Water Resources Research 37(8): 2273?2283. Sibert, J., and McDonnell J.J. 2010. Land-cover impacts on streamflow: a change-detection modelling approach that incorporates parameter uncertainty. Hydrological Sciences Journal 55(3): 316-332. doi:10.1080/02626661003683264. Sidle, R.C., Tsuboyama, Y., Noguchi, S., Hosoda, I., Fujieda, M., Shimizu, T., 2000. Stormflow generation in steep headwaters: A linked hydro-geomorphic paradigm. Hydrological Processes 14: 369?385. Sidle, R.C., 1988. Bed Load Transport Regime of a Small Forest Stream. Water Resources Research 24, 207-218. Smart, G.M., 1984. Sediment transport formula for steep channels. Journal of Hydraulic Engineering 110: 267-276. Smith, R, 2011. Space-time dynamics of runoff generation in a snowmelt-dominated montane catchment. Unpublishe PhD thesis, University of British Columbia, 170p. Stedinger, J.R., Vogel, R.M., and Foufoula-Georgiou, E., 1993. Frequency analysis of extreme events. In D.R. Maidment (Editor), Handbook of Hydrology, Chapter 18, pp. 18.1?18.66, McGraw-Hill, New York.  197 Sterling, S.M., and Church. M., 2002. Sediment trapping characteristics of a pit trap and the Helley-Smith sampler in a cobble gravel bed river. Water Resources Research, 38(6), DOI: 10.1029/2000WR000052 Swanson, F.J., Fredricksen, R.L., and McCorison, F.M., 1982. Material transfer in a western Oregon forested watershed. Analysis of Coniferous Forest Ecosystems in the Western United States, US/IBP Synthesis Series, Huchinson Ross Publ., Stroudsburg, Penn, pp. 233-266. Swanson, F.J., Johnson, S.L., Gregory, S.V., and Acker, S.A., 1998. Flooding: Natural and Managed. BioScience 48, 681-689. Szeftel, P., 2011. Stream-catchment connectivity and streamflow dynamics in a montane landscape. Unpublished Phd dissertation. University of British Columbia, Vancouver. 183pp. Teti, P., 2005. Effects of small logged openings on snow ablation during a high snow year. In: Proceedings of the 72nd Western Snow Conference, pp. 45?52, April 19?22, Vancouver, Canada.  Thomas, R.B., and Megahan, W.F., 1998. Peak flow responses to clearcutting and roads in small and large basins, western Cascades, Oregon: A second opinion. Water Resources Research 34: 3393 ? 3403, doi:10.1029/98WR02500. Thompson, C.J., and Croke, J., 2008. Channel flow competence and sediment transport in upland streams in southeast Australia. Earth Surface Processes and Landforms 33: 329-352. DOI: 10.1002/esp.1558.  Thyer, M., Beckers, J., Spittlehouse, D., Alila, Y., and Winkler R.D., 2004. Diagnosing a distributed hydrologic model for two high-elevation forested catchments based on detailed stand- and basin-scale data, Water Resources Research 40, W01103, doi: 10.1029/2003WR002414.  Toews, D.A.A., and Gluns, D.R., 1986. Snow accumulation and ablation on adjacent forested and clearcut sites in southeastern British Columbia. In: Proceedings of the 54th Annual Western Snow Conference, pp. 101?111, April 15?17, 1985, Phoenix, Ariz Troendle, C.A., 1987. Effect of clearcutting on streamflow generating processes from a subalpine forest slope. In: Forest Hydrology and watershed Management ? Hydrologie Forestiere et AmBnagement des Bassins Hydrologiques, Proceedings of the Vancouver Symposium, IAHS-AISH Pub1.no.167, August, 1987. Troendle, C.A., and Kaufmann, M.R., 1987. Influence of forests on the hydrology of sub alpine forests. In: Management of Sub Alpine Forests: Building on Fifty Years of Research, USDA Forest Service, Rocky Mountain Forest and Range Experiment Station, General Technical Report RM-149, pp. 68?76, Fort Collins, Co.,   198 Troendle, C.A., and King, R.M., 1985. The effect of timber harvest on the Fool Creek watershed, 30 years later, Water Resources Research 21: 1915? 1922, doi:10.1029/WR021i012p01915.  Troendle, C.A., and King, R.M., 1987. The effect of partial and clear cutting on streamflow at Deadhorse Creek, Colorado, Journal of Hydrology., 90, 145?157. Troendle, C.A., and Leaf, C.F., 1980. Hydrology, Chapter III. In: An Approach to Water Resources Evaluation of Non-Point Silvicultural Sources (A procedural handbook), EPA 60018-80-012, pp. 111-1?111-173., Environmental Research Lab, Athens, GA. Troendle, C.A., and Leaf, C.F., 1981. Effects of timber harvest in the snow zone on volume and timing of water yield. In: Interior West Watershed Management, pp. 231-243, Washington State University, Pullman.  Troendle, C.A., and Olsen, W.K., 1994. Potential effects of timber harvest and water management on steamflow dynamics and sediment transport. In: Sustainable Ecological Systems: Implementing an Ecological Approach to Land Management, USDA Forest Service General Technical Report RMRS-GTR-247, pp. 34?41, Flagstaff, Ariz.  Troendle, C.A., and Stednick, J.D., 1999. Discussion on ??Effects of basin scale timber harvest on water yield and peak streamflow?? by Timothy A. Burton, Journal of the American Water Resources Association 35, 177? 181, doi:10.1111/j.1752-1688.1999.tb05462.x. Troendle, C.A., MacDonald, L., Luce, C.H., and Larsen, I.J., 2010. Fuel management and water yield. In: W.J. Elliot, I. S. Miller, and L. Audin (Editors), Cumulative Watershed Effects of Fuel Management in the Western United States, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-231. pp. 126?148. Fort Collins, CO.  Troendle, C.A., Wilcox, M.S., Bevenger, G.S., and Porth, L.S., 2001. The Coon Creek water yield augmentation project: Implementation of timber harvesting technology to increase streamflow. Forest Ecology and Management, 143: 179? 187. doi:10.1016/S0378-1127(00)00516-8 U.S. Water Resources Council, 1976. Guidelines for determining flood flow frequency, Bulletin 17, Washington, D.C. United States Department of Agriculture, Forest Service (USDA-FS), 1974. Forest hydrology part II ? hydrological effects of vegetation manipulation. Missoula, MT. U.S. Department of Agriculture, Forest Service, 229p. United States Department of Agriculture, Forest Service (USDA-FS), 2010. Final Environmental Impact Statement: Fernow Experimental Forest, Forest Service Northern Research Station,  199 (http://nrs.fs.fed.us/ef/locations/wv/fernow/EIS/resources/Final_EIS_Fernow_Exp_Forest_04Oct.pdf) (Accessed January 1, 2012). Van Haveren, B.P., 1988. A re-evaluation of the Wagon Wheel Gap Forest Watershed Experiment. Forest Science 34: 208?214. Varhola, A., Coops, N.C., Weiler, M., and Moore, R.D., 2010. Forest canopy effects on snow accumulation and ablation: An integrative review of empirical results. Journal of Hydrology 392: 219?233, doi:10.1016/j.jhydrol.2010.08.009. Warburton J., 1992. Observations of bed load transport and channel bed changes in a proglacial mountain stream. Arctic and Alpine Research 24: 195?203. Warburton, J., 1990. Comparison of bed load yield estimates for a glacial meltwater stream. In: Proceedings of two Lausanne Symposia, August 1990. Hydrology in Mountainous Regions. I ? Hydrological Measurements; the Water Cycle. IAHS Publication 193, pp. 315-323.  Waylen, P., and Woo, M., 1982. Prediction of annual floods generated by mixed processes. Water Resources Research 18(4): 1283?1286, doi:10.1029/WR018i004p01283. Whitaker, A., Alila, Y., and Beckers, J., 2002. Evaluating peak flow sensitivity to clear-cutting in different elevation bands of a snowmelt-dominated mountainous catchment. Water Resources Research 38, 1172. doi:10.1029/2001WR000514 Whitaker, A., Alila, Y., Beckers, J., and Toews, D., 2003. Application of the distributed hydrology soil vegetation model to Redfish Creek British Columbia: Model evaluation using internal catchment data. Hydrological Processes 17, 199? 224. Whitaker, A.C., and Potts, D.F., 2007. Analysis of flow competence in an alluvial gravel-bed stream, Dupuyer Creek, Montana. Water Resources Research 43: W07433. DOI:10.1029/2006WR005289. Whiting, P., and Dietrich, W., 1990 Boundary Shear Stress and Roughness Over Mobile Alluvial Beds. Journal of Hydraulic Engineering 116(12): 1495?1511. Whiting, P.J., Stamm, J.F., Moog, D.B., and Orndorff, R.L., 1999. Sediment-transporting flows in headwater streams. Geological Society of America Bulletin 111: 450?466. Whittaker, J.G., Hickman, W.E., and Croad, R.N., 1988. Riverbed Stabilisation with Placed Blocks, Central Laboratories, Works and Development Corporation, Lower Hutt, NZ. Whittaker, J.G., and. Jaeggi, M. N. R, 1982. Origin of Step-Pool Systems in Mountain Streams. Journal of the Hydraulics Division, 108(6): 758-773. Wiberg, P.L., and Smith, J.D., 1987. Calculations of the critical shear stress for motion of uniform and heterogeneous sediments. Water Resources Research 23: 1471?1480.  200 Wigley, T.M.L., 1985. Impact of extreme events. Nature, 316, 106? 107. Wigley, T.M.L., 2009. The effect of changing climate on the frequency of absolute extreme events. Climate Change, 97, 67?76, doi:10.1007/s10584-009-9654-7. Wigmosta, M.S., Vail, L.W., and Lettenmaier, D.P., 1994. A distributed hydrology-vegetation model for complex terrain. Water Resources Research 30(6), 1665?1679. Wilcock, P.R., 2001. Toward a practical method for estimating sediment transport rates in gravel-bed rivers. Earth Surface Processes and Landforms 26: 1395-1408, DOI: 10.1002/esp.301 Wilcock, P.R., and McArdell, B.W., 1993. Surface-based fractional transport rates: mobilization thresholds and partial transport of a sand-gravel sediment. Water Resources Research 29: 1297? 1312. Wilcock, P.R., and McArdell, B.W., 1997. Partial transport of a sand/gravel sediment. Water Resources Research 33: 235? 245. Wilcock, P.R., and Southard, J.B., 1988. Experimental Study of Incipient Motion in Mixed-Size Sediment. Water Resources Research 24: 1137-1151. Wilcox, A.C., and Wohl, E.E., 2006. Flow resistance dynamics in step-pool channels: 1. Large woody debris and controls on total resistance. Water Resources Research 42: W05418. DOI:10.1029/2005WR004277 Wilcox, A.C., Nelson, J.M., and Wohl, E.E., 2006. Flow resistance dynamics in step-pool channels: 2. Partitioning between grain, spill and woody debris resistance. Water Resources Research 42: W05419. DOI:10.1029/2005WR004278 Wilford, D.J., Sakals, M.E., Grainger, W.W., Millard, T.H., and Giles, T.R., 2009. Managing forested watersheds for hydrogeomorphic risks on fans, B.C. Min. of For. Range, For, Sci. Prog. Victoria, B.C. Land Manage. Handb. 61. Winkler, R.D., Moore, R.D., Redding, T.E., Spittlehouse, D.L., Smerdon, B.D., and Carlyle-Moses, D.E., 2009. The Effects of Forest Disturbance on Hydrologic Processes and Watershed Response. In R.G. Pike (Editor), Compendium of Forest Hydrology and Geomorphology in British Columbia, Chapter 7, pp. 179?212. B.C. Ministry of Forest and Range Research Branch, Victoria, B.C., and Land Management Handbook 66, FORREX Forum for Research and Extension in Natural Resources Society, Kamloops, B.C.  Winkler, R.D., Rex, J.F., Teti, P., Maloney, D.A., and Redding, T., 2008. Mountain pine beetle, forest practices, and watershed management. B.C. Extension Note 88, B.C. Ministry of Forests and Range Research Branch, Victoria.  201 Winkler, R.D., Spittlehouse, D.L., and Golding, D.L., 2005. Measured differences in snow accumulation and melt among clearcut, juvenile, and mature forests in southern British Columbia. Hydrological Processes 19, 51?62.  Wohl, E., 2004. Limits of downstream hydraulic geometry. Geology 32, 897-900.  Wohl, E., 2010. Mountain Rivers Revisited. American Geophysical Union Press. Water Resources Monograph 19, 573 pp. Wohl, E.E., and Wilcox, A., 2005. Channel geometry of mountain streams in New Zealand. Journal of Hydrology 300, 252?266. Wolman, M.G., and Gerson, R., 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surface Processes 3: 189-208.  Wolman, M.G., and Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes, Journal of Geology 68: 54?74.  Yager, E.M., Dietrich, W.E., Kirchner, J.W., and McArdell, B.W., 2012. Prediction of sediment transport in step-pool channels. Water Resources Research 48: W01541. DOI: 10.1029/2011WR010829 Yager, E.M., Kirchner, J.W., and Dietrich, W.E., 2007. Calculating bed load transport in steep boulder bed channels. Water Resources Research 43: W07418. DOI.10.1029/2006WR005432. Yevjevich, V., 1972. Stochastic processes in hydrology, Water Research Publications, Colorado. Z?gre, N., Skaugset, A.E., Som, N.A., McDonnell, J.J., and Ganio, L.M., 2010, In lieu of the paired catchment approach: Hydrologic model change detection at the catchment scale. Water Resources Research 46: W11544, doi:10.1029/2009WR008601. Zhao, F., Zhang, L., Xu, Z., and Scott, D.F., 2010. Evaluation of methods for estimating the effects of vegetation change and climate variability on streamflow Water Resources Research 46: W03505, doi:10.1029/2009WR007702. Zimmermann, A., 2010. Flow resistance in steep streams: an experimental study. Water Resources Research 46: W09536, doi:10.1029/2009WR007913.    202 Appendices Appendix 1: Measurement errors and uncertainties Table A 1 Errors and uncertainties for field data Parameter Measuring  method Uncertainty Water discharge   Continuously monitored sites Odyssey capacitance water level probe ? 15min time-step  0.8mm  Salt dilution gauging < 5% Intermittently monitored sites Salt dilution gauging Discharge time-series based on regression with continuously monitored sites Chronologically paired discharge values (n=15) between gauged and ungauged sites over the full range of flows was used to build a continuous record of discharge for ungauged sites. Regression correlations (R2) between gauged and ungauged sites range from 0.95 (C3) to 0.88 (C2) and standard errors of the estimate range from 0.019 to 0.025 (m3/s) respectively Flow velocity   All sites Average flow velocity was estimated as discharge (Q, m3/s) divided by flow cross-sectional area (A, m2) Flow depth for channel x-section was measured using a stadia rod (rounded to the nearest 0.5 cm) at 10 cm intervals along the cross section. An uncertainty of up to 1 cm in the stadia rod reading amounts to a maximum error of 15 % for regression-based flow velocity estimates and a maximum error of 10 % for the velocity estimates at the continuously gauged sites.  Bedload   Channel-spanning pit traps Traps were emptied by hand-held scoop, hourly to every several days depending on discharge  Unknown ? Potyondy et al. (2010) indicate that channel spanning pit traps display the highest level of accuracy compared to other sampling devices. Removable pit traps Baskets were removed and emptied hourly to every several days as above Unknown - The uncertainty associated with our removable traps appears to set within the error envelope of the published field-based bedload literature Channel Survey   Grain size Gravelometer and 1phi seives < 5% Bankfull width/depth Stadia rod and metric tape < 5%  203 Parameter Measuring  method Uncertainty Gradient Engineers level < 2% LWD Step spacing Metric tape < 2% LWD Step height Stadia rod < 2% Estimate of mobile channel area  Arial photo mosaics and measured data from detailed surveys < 20%     204 Appendix 2: Bedload transport formulae A2.1: Parker et al., 1982 (modified) In Parker et al.?s, (1982) bedload transport formula transport rate (qb) is calculated as follows; qb=[g1/2(DSr)3/2G(?50ss)??? ?s??] *(????)       (A1a)   5474(1-0.853?50ss)4.5     ?50ss>1.59 G(?50ss)=  exp[14.2 (?50ss-1)-9.28(?50ss-1)2]  1??50ss ?1.59   ?50ss14.2       ?50ss ?1    (A1b)  ?50ss=?*D50ss/?*r(D50ss)         (A1c)  Where, D is depth (m), Sr is reduced slope (see below), W*r is 0.0025 (dimensionless transport rate), ?s is sediment density (2650 kg/m3), and Rs is the submerged specific gravity of sediment (?s/? ? 1). G(?50ss) is a three-part function revised by Parker (1990) dependent on the excess Shields stress of the median subsurface grain size (?50ss).  Following the approach adopted by Nitsche et al. (2011) we modify Parker et al.?s (1982) equation to include a slope reduction factor that accounts for loss of flow energy due to resistance from immobile clasts and large woody debris (LWD). The slope reduction factor (Sr) is calculated using the ratio of grain resistance (fo) to total resistance (ft) (Nitsche et al., 2011). Total resistance is calculated as the sum of the grain resistance and an additional resistance associated with form roughness (ft=fo+fadd). Grain resistance is a function of the bed grain-size distribution and is calculated using Millar and Quick?s (1993) variation of the Keulegan equation in which D50 replaces the roughness parameter ks: fo = [2.03log (12.2R/D50)]2        (A2)  205 where resistance is calculated as a function of the bankfull hydraulic radius (R). Additional flow resistance due to woody debris and immobile clasts is estimated using Whittaker et al.?s (1988) resistance parameter developed to account for flow resistance associated with blocks (Db) used for river stabilization: fadd = [8/(2.5*ln(12R/kb))2]        (A3) where kb is the mean roughness height associated with the immobile macro-roughness elements (i.e. LWD and boulders). kb = ??? (17.8 ? 0.47???)        (A4) and ?, is the areal concentration of macro-roughness elements per unit area of channel: ? =????           (A5) where Ai is the area of immobile elements (including wood steps and immobile boulders) within a given unit area At of the channel bed. This is the same factor (e.g. At-Am = Ai) we use to account for the reduction in the available bed material that is included in the transport formulae of Yager et al,. (2007).  Reduction in mobile bed area = ????        (A6) Where Am is the channel area occupied by mobile bed sediment and At is the total bed area. The fraction of mobile bed area is estimated for a 10 meter reach upstream from the sampling site using channel survey information and photo mosaics of the channel.      206 A2.2 Rickenmann (2001) modified Formulation of Rickenmann?s (2001) bedload transport equation with adjustment factors to account for reduced energy slope and reduced mobile bed area due to macro-roughness elements is as follows;  qb = 3.1(D90sD30s 0.2)(q-qc)Sr2(2.65-1) -1.5 *(????)       (A7) D90s and D30s are estimated from cumulative grain size distributions of the bed surface at the sampling site (Figure 3). Unit discharge q (m2/s) and qc is critical unit discharge for sediment transport (m2/s), which is identified as the unit discharge corresponding to a small measurable bedload transport rate (qc = Qi/Wbf; see Tables 1 and 6 for these values). Sr is the reduced slope calculated using the ratio of grain to total resistance (Sr = fo/ft) as outlined in Appendix A1. The steep gradient exponent is set equal to 2. The factor to account for reduced availability of mobile bed load (???? ) is also discussed in Appendix 2.1.  A2.3 Yager et al., 2012 The equation used here is the equation presented in Yager et al., (2007) based on the Fernandez Luque & Van Beek (1976) equation but adjusted to account for the reduction of bed shear stress due to the presence of immobile steps as outlined in Yager et al., (2012).  q*mb=(5.7*(??? ? ????(d50))1.5)*( ????)        (A8) where q*mb is the dimensionless sediment transport rate, ?*m and ?*cm are the dimensionless shear stress and critical shear stress acting on the mobile bed. The factor accounting for reduced availability of bedload (Am/At) is discussed in Appendix A2.1.  To parameterize Yager et al.'s (2012) bedload transport equation we use data from detailed longitudinal channel surveys conducted upstream of each bedload sampling site (Figure 2)  207 along stretches equal to 15 to 20 bankfull widths. These surveys were conducted at a resolution capable of capturing the location and spacing of macro-roughness elements (e.g., cascades and LWD steps) and the area of bed occupied by mobile material. Data from these surveys are used to estimate, ?x and ?w, the downstream spacing of and downstream lengths of macro-roughness elements. Protrusion (pu) and sediment depth (zmu) were estimated from close-range vertical photographs, using as base reference grain diameters measured via Wolman pebble counts (section 3.3). We estimate the uncertainty around these parameters to be as high as 50%. A sensitivity analysis conducted on the six study sites shows that imposing a 50% change to pu produces high variability to predicted values of qb* i.e., from as much as 25% at bankfull flows to more than 10-fold difference at low flows.  Calculation of the total dimensionless shear stress acting on the mobile bed (??? ) requires calculation of a ?virutal? reach average flow depth ?ha?. ??? =  ?????          (A9) The virtual flow depth is calculated according to the equation presented in Yager et al., (2012); ? =  ?2??????3??????+??(?????)        (A10) Where q is unit discharge (m2/s), g is gravity, S is reach average water slope, w is channel width (i.e. width at bankfull). Aif is the area of the bed occupied by immobile elements, Ci and Cm are drag coefficients of the immobile and mobile bed material and ?x and ?w, the downstream spacing of and downstream lengths of macro-roughness elements. Equation A10 is rearranged to solve for ha for the range of unit discharges associates with the  208 sediment transport. The drag coefficient of the immobile grains (Ci) is calculated according to; ?? = 157 ? (????)?1.6 Where Pu is the protrusion of the immobile macro-element on the upstream side of the step. Finally, critical dimensionless shear stress for the median surface bed material used in Yager et al.'s (2012) equation is set equal to ?*cm50s and calculated as: ???? =  ???(????)?50?         (A11) Where; ?? = ???  ?22          (A12) Cm is the drag coefficient of the mobile bed material (set to the constant 0.44), p is the density of water and U is the reach average flow velocity (q/ha). Parameters used in the Yager et al., (2012) equation are listed in Table A3.1 Table A 2 Parameters use in Yager et al., 2012 bedload transport equation.   E2 (FSP(PB)) E3  (BC) E4  (SP) C2 (FSP(BC)) C3 (FSP(BC)) C4  (RP) Avg. Dba 0.16 0.27 0.2 0.26 0.19 0.2 ?x (m)b 5 4 4 2.8 3.3 8 ?w (m)c 4 2 0.4 1.8 2 0.5 Am/Atd 0.8 0.73 0.9 0.56 0.57 0.88 pu (m)e 0.13 0.1 0.1 0.14 0.08 0.05 zmu(m)f 0.032 0.171 0.097 0.12 0.11 0.15 AIFg 0.142 0.121 0.126 0.191 0.137 0.077 Cmh 0.44 0.44 0.44 0.44 0.44 0.44 ?*cm50i 0.036 0.054 0.045 0.048 0.034 0.021 a. Average diameter of immobile boulders b. Downstream spacing of macro-roughness elements c. Downstream length of macro-roughness elements d. Proportion of mobile area to total area upstream from sampling site e. Immobile grain protrusion on upstream side. f. Average bed perpendicular height of the mobile bed above the base of the immobile grain. g. Bed perpendicular area of immobile grains h. Drag coefficient due to mobile sediment i. Dimensionless critical shear stress for mobilization of the D50s adjusted according to Yager et al., (2007).   209 Appendix 3 Bedload transport data and bulk channel bed data for sample sites   210   211     212     213    214   215 Appendix 4 Channel survey data for Cotton and Elk by reach     216 Appendix 5 Sources of data for meta-analysis (Chapters 3 and 4) Data Set Source Fool/ ESL Creek daily average discharge USFS Download website: http://www.fs.fed.us/rm/data_archive/dataaccess/FEF_ESLC_LFCRK_dailyflow.shtml Camp/Greata daily average discharge Environment Canada archived hydrometric database website: http://www.wsc.ec.gc.ca/applications/H2O/index-eng.cfm Redfish Creek and 240 Creek DHSVM Daily average discharge In-house data set: Contact Dr. Younes Alila, Forest Resources Management, Forestry, UBC    

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0166834/manifest

Comment

Related Items