Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Linking sediment supply, channel morphology, and aquatic habitat in forested, gravel-bed streams : spatial… Reid, David A. 2019

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

Item Metadata


24-ubc_2019_september_reid_david.pdf [ 17.76MB ]
JSON: 24-1.0379797.json
JSON-LD: 24-1.0379797-ld.json
RDF/XML (Pretty): 24-1.0379797-rdf.xml
RDF/JSON: 24-1.0379797-rdf.json
Turtle: 24-1.0379797-turtle.txt
N-Triples: 24-1.0379797-rdf-ntriples.txt
Original Record: 24-1.0379797-source.json
Full Text

Full Text

Linking sediment supply, channel morphology, and aquatic habitatin forested, gravel-bed streams: spatial and temporalconsiderationsbyDavid A. ReidBA, University of British Columbia, 2011MSc, University of British Columbia, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Geography)The University of British Columbia(Vancouver)July 2019c© David A. Reid, 2019The following individuals certify that they have read, and recommend to the Faculty of Grad-uate and Postdoctoral Studies for acceptance, the thesis entitled:Linking sediment supply, channel morphology, and aquatic habitat in forested,gravel-bed streams: spatial and temporal considerationssubmitted by David A. Reid in partial fulfillment of the requirements for the degree of Doctorof Philosophy in Geography.Examining Committee:Marwan Hassan, GeographySupervisorMichael Church, GeographySupervisory Committee MemberJohn Richardson, Department of Forest and Conservation SciencesSupervisory Committee MemberOlav Slaymaker, GeographyUniversity ExaminerJohn Innes, Faculty of ForestryUniversity ExaminerEllen Wohl, Warner College of Natural Resources, Colorado State UniversityExternal ExamineriiAbstractGravel-bed streams in mountain environments are complex systems which respond unpre-dictably to episodic inputs of sediment from hillslopes. Stream channel response to episodicsediment supply has implications for channel morphology, stability, and aquatic habitat. Whileinsight into the behaviour of these systems can be gained from numerical models and physi-cal experiments, a paucity of long-term, comprehensive, and multi-scale field data has limitedresearchers’ ability to describe channel response to episodic sediment supply, and to situatethis response in a broader landscape context. This thesis aims to overcome these limitationsby examining long-term channel dynamics in response to episodic sediment supply and vari-able wood loading and then linking this variability to modeled aquatic habitat for juvenilesalmonids. A 45 year field dataset from Carnation Creek, a small forested, gravel-bed streamlocated in a deglaciated catchment on coastal British Columbia, is used for analysis.Results indicate that temporal patterns of sediment storage are governed by the sedimentconditions and local erosional and depositional processes, while spatial patterns are associ-ated with sediment travel distance statistics. A conceptual model is proposed which presentschannel response to episodic supply as a function of channel position downstream relative tocolluvial input. In-stream wood is also found to influence channel morphology: logjams impactsediment throughput and lead to locations of elevated sediment storage, which decay exponen-tially over a 10-20 year period. A wood budget model indicates that wood loads will take upto 200 years to recover following riparian timber harvesting, with long-lasting implications forchannel morphology.Modeled habitat for juvenile salmonids, simulated using Carnation Creek topographic andwood data, is found to vary through time by up to a factor of ten as a function of variablechannel morphology and wood abundance. These results are found to support a conceptualmodel which links the contribution to habitat variability from channel morphology to a wa-tershed’s sediment supply regime. Collectively, this work has improved our understanding ofhow episodic sediment supply and wood impact channel morphology, sediment storage, andaquatic habitat in a forested, gravel-bed stream over the long term.iiiLay SummaryDebris flows often travel downs hillslopes into streams in mountain landscapes. The sedimentcontributed from these events impacts channel form and must therefore influence habitat con-ditions for fish. However, few long-term field studies have examined this connection. Theobjective of this thesis is to examine how streams respond to sediment input from debris flowsover the long-term, how log jams in rivers affect this response, and how fish habitat is im-pacted. A 45-year dataset from a stream in coastal British Columbia is used to examine theseinter-related stream responses. Findings indicate that debris flow sediment drives changes inchannel form at locations near to the flow, but is less important further downstream. Woodin the channel disrupts the downstream movement of sediment, trapping it for decades. Thecombination of changes through time in wood and sediment abundance leads to changes in thequantity of habitat available for salmon.ivPrefaceThis thesis was completed under the supervision of Marwan Hassan, who provided feedbackand guidance on data analysis and editing for all chapters of this work. I am the lead thesisauthor, and undertook the conceptualization, data analysis, and writing of all chapters.Much of the data for this thesis was made available from Dr. Peter Tschaplinski and RobinPike at the B.C. Ministry of Environment and Climate Change. I collected all channel mor-phology and in-stream wood data used in this thesis from 2011 to 2017, with the exceptionof 2013. This thesis contains one published manuscript, one manuscript submitted for pub-lication, and one additional research chapter. A version of Chapter 3 is published in EarthSurface Processes and Landforms with co-authors Marwan Hassan, who supervised the work,and Stephen Bird and Daniel Hogan, who were involved in data collection at the study siteprior to the start of my PhD. I undertook all data preparation, analysis and writing, in additionto conceptualization.A version of Chapter 5 has been submitted for publication with co-authors Marwan Has-san, who supervised the work, and Stephen Bird, Robin Pike, and Peter Tschaplinski, whocollected historical data in the study watershed, and provided editorial feedback on the submit-ted manuscript version. I undertook all data preparation, analysis and writing, in addition toconceptualization.Reid, D.A., Hassan, M.A., Bird, S.A, and Hogan, D.H. (2019). Spatial and temporal patternsof sediment storage over 45 years in Carnation Creek, BC, a previously glaciated mountaincatchment. Earth Surface Processes and Landforms, of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2 Motivation and study objectives . . . . . . . . . . . . . . . . . . . . . . . . . 42 Carnation Creek study watershed . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1 Carnation Creek: An interdisciplinary long-term monitoring project . . . . . . 82.2 Watershed setting and biophysical characteristics . . . . . . . . . . . . . . . . 82.3 Catchment history: A summary of key events . . . . . . . . . . . . . . . . . . 153 Spatial and temporal patterns of sediment storage . . . . . . . . . . . . . . . . . 183.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21vi3.3.1 Topographic data collection and preparation . . . . . . . . . . . . . . . 213.3.2 Assessment of sediment supply and transport . . . . . . . . . . . . . . 253.3.3 Calculation of sediment stored in bars . . . . . . . . . . . . . . . . . . 253.3.4 Autocorrelation and spectrum analysis . . . . . . . . . . . . . . . . . . 263.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.4.1 Hydrological conditions . . . . . . . . . . . . . . . . . . . . . . . . . 273.4.2 Sediment supply and transport conditions . . . . . . . . . . . . . . . . 273.4.3 Organization of stored sediment and wood . . . . . . . . . . . . . . . . 303.4.4 Spatial and temporal patterns of sediment storage . . . . . . . . . . . . 333.4.5 Storage and sediment transfer rate . . . . . . . . . . . . . . . . . . . . 353.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.5.1 Storage and sediment transfer . . . . . . . . . . . . . . . . . . . . . . 403.5.2 Landscape organization and storage . . . . . . . . . . . . . . . . . . . 413.5.3 Conceptual response of storage to supply . . . . . . . . . . . . . . . . 423.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444 Long-term characteristics, morphological impacts, and simulated budgets ofin-stream large wood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464.3 Methods and data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.3.1 Wood characterization and volumetric estimates . . . . . . . . . . . . . 474.3.2 Wood budget model . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.4.1 Large wood characteristics . . . . . . . . . . . . . . . . . . . . . . . . 554.4.2 Spatial and temporal patterns of large wood . . . . . . . . . . . . . . . 584.4.3 Wood-morphology interactions . . . . . . . . . . . . . . . . . . . . . 614.4.4 Wood budget model and large wood response timescales . . . . . . . . 644.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.5.1 Large wood characteristics and patterns . . . . . . . . . . . . . . . . . 684.5.2 Morphological influence of large wood . . . . . . . . . . . . . . . . . 724.5.3 Timber harvesting influence on LW storage and channel morphology . 744.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 765 Testing a process-based linkage between catchment structure and aquatic habitat 775.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785.2.1 Conceptual model and study objectives . . . . . . . . . . . . . . . . . 79vii5.3 Methods and data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.3.1 Field data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.3.2 Generation of channel bed surfaces . . . . . . . . . . . . . . . . . . . 825.3.3 Hydrodynamic modeling . . . . . . . . . . . . . . . . . . . . . . . . . 825.3.4 Variable selection and definition . . . . . . . . . . . . . . . . . . . . . 845.3.5 Analysis of model output . . . . . . . . . . . . . . . . . . . . . . . . . 845.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865.4.1 Variability in channel morphology . . . . . . . . . . . . . . . . . . . . 865.4.2 Variability of modeled habitat . . . . . . . . . . . . . . . . . . . . . . 885.4.3 Habitat by flow level . . . . . . . . . . . . . . . . . . . . . . . . . . . 905.4.4 Role of hydrological relative to morphological variability . . . . . . . . 915.4.5 Scale considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 985.5.1 Evaluating the landscape-habitat linkage . . . . . . . . . . . . . . . . . 985.5.2 Spatial and temporal variability of modeled habitat . . . . . . . . . . . 995.5.3 Habitat by flow level . . . . . . . . . . . . . . . . . . . . . . . . . . . 1005.5.4 Importance of morphological variability for habitat change . . . . . . . 1005.5.5 Study limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1026 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046.1 Summary of contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1056.2 Broad research implications . . . . . . . . . . . . . . . . . . . . . . . . . . . 1086.3 Future research opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . 111Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113A Autocorrelation and spectrum analysis methodological details . . . . . . . . . . 128B Wood budget model term time series . . . . . . . . . . . . . . . . . . . . . . . . . 134C Nays2DH settings and performance evaluation . . . . . . . . . . . . . . . . . . . 139viiiList of TablesTable 2.1 Characteristics and channel geometry of the eight intensive study areas . . . 13Table 3.1 Field data and data collection methods . . . . . . . . . . . . . . . . . . . . 23Table 3.2 Characteristics of logjams and stored sediment along the lower 3.0 km ofCarnation Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Table 3.3 Channel geometry, sediment storage, and wood characteristics for studyareas of Carnation Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Table 4.1 Model parameters varied for simulations . . . . . . . . . . . . . . . . . . . 54Table 4.2 Characteristics of logjams located along the lower 3.0 km of Carnation Creek 55Table 4.3 Wood and morphology characteristics for Carnation Creek study areas . . . 56Table 4.4 Distribution of wood piece orientation angles relative to cross-channel di-rection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Table 4.5 Model results summary for all simulation configurations . . . . . . . . . . . 67Table 5.1 Flow levels (% of mean annual discharge) scaled by contributing area . . . . 83Table 5.2 Variance explained by streamflow relative to channel form and wood loading 95Table B.1 Reach characteristics of bank erosion and fluvial wood transport for woodbudget simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138Table B.2 Summary input and output wood budget terms for pre-harvest (baseline)conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138ixList of FiguresFigure 2.1 Carnation Creek watershed and channel profile . . . . . . . . . . . . . . . 10Figure 2.2 Carnation Creek climate averages . . . . . . . . . . . . . . . . . . . . . . 11Figure 2.3 Example debris flows entering Carnation Creek main stem . . . . . . . . . 12Figure 2.4 Maps of intensive study areas SA-2 to SA-9 . . . . . . . . . . . . . . . . . 14Figure 2.5 Photos of SA-2 to SA-5 . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Figure 2.6 Photos of SA-6 to SA-9 . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Figure 2.7 Timeline of notable events . . . . . . . . . . . . . . . . . . . . . . . . . . 17Figure 3.1 Grain size distributions and particle travel distances . . . . . . . . . . . . . 24Figure 3.2 Competent flow hours, peak flows, and sediment delivery from hillslopes . 28Figure 3.3 Bank erosion and stabilization . . . . . . . . . . . . . . . . . . . . . . . . 30Figure 3.4 Main-stem sediment storage in bars . . . . . . . . . . . . . . . . . . . . . 32Figure 3.5 Sediment storage and wood volume in study areas SA-2 to SA-9 . . . . . . 33Figure 3.6 Corellellograms and periodograms for main stem and study area sedimentstorage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 3.7 Aggradation-degradation cycles of sediment storage . . . . . . . . . . . . 37Figure 3.8 Duration of small-scale aggradation-degradation cycles . . . . . . . . . . . 38Figure 3.9 Cumulative departures and scaled values of study area sediment storage . . 38Figure 3.10 Conceptual model of storage variability as a function of position relativeto hillslope-channel coupling . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 4.1 Example maps of wood pieces . . . . . . . . . . . . . . . . . . . . . . . . 49Figure 4.2 Wood piece size distribution and piece persistence . . . . . . . . . . . . . 56Figure 4.3 Wood piece orientation relative to thalweg . . . . . . . . . . . . . . . . . . 57Figure 4.4 Logjams located along the lower 3.0 km of Carnation Creek . . . . . . . . 59Figure 4.5 Wood piece and volume time series in the eight study areas . . . . . . . . . 60Figure 4.6 Topographic complexity relative to wood piece abundance . . . . . . . . . 62Figure 4.7 Logjam location and sediment storage along channel main stem . . . . . . 63Figure 4.8 Logjam evolution in SA-5 and SA-8 . . . . . . . . . . . . . . . . . . . . . 65xFigure 4.9 Decay of sediment stored behind logjams . . . . . . . . . . . . . . . . . . 66Figure 4.10 Wood budget model terms for Scenario 1 . . . . . . . . . . . . . . . . . . 68Figure 4.11 Wood budget simulation results for Scenario 1 . . . . . . . . . . . . . . . 69Figure 4.12 Wood budget simulation results for Scenario 2 . . . . . . . . . . . . . . . 70Figure 4.13 Wood budget simulation results for Scenario 3 . . . . . . . . . . . . . . . 71Figure 5.1 Conceptual model contrasting habitat variability in channels with steadyand episodic supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81Figure 5.2 Example Nays2DH flow model output . . . . . . . . . . . . . . . . . . . . 85Figure 5.3 Channel width, bed elevation, and topographic roughness through time . . 87Figure 5.4 Variability in channel morphology as a function of position relative tohillslope-coupled regions . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Figure 5.5 Time series of modeled habitat . . . . . . . . . . . . . . . . . . . . . . . . 89Figure 5.6 Comparison of habitat between study areas at two flow levels . . . . . . . . 91Figure 5.7 Habitat availability relative to discharge . . . . . . . . . . . . . . . . . . . 92Figure 5.8 Summer low flow characteristics . . . . . . . . . . . . . . . . . . . . . . . 93Figure 5.9 Cumulative seasonal habitat . . . . . . . . . . . . . . . . . . . . . . . . . 94Figure 5.10 Variability in channel morphology relative to flow influence on habitat . . . 96Figure 5.11 Variance in habitat relative to record length . . . . . . . . . . . . . . . . . 97Figure A.1 Channel width and bed elevation through time in the eight study areas . . . 130Figure A.2 Aggradation-degradation cycles in six of eight study areas . . . . . . . . . 131Figure A.3 Spectral density for bar width along the lower 3.0 km of channel . . . . . . 132Figure A.4 Spectral density for bar height along lower 3.0 km of channel . . . . . . . 133Figure B.1 Image of mobile wood pieces . . . . . . . . . . . . . . . . . . . . . . . . 134Figure B.2 Wood budget term time series for Scenario 1 . . . . . . . . . . . . . . . . 135Figure B.3 Wood budget term time series for Scenario 2 . . . . . . . . . . . . . . . . 136Figure B.4 Wood budget term time series for Scenario 3 . . . . . . . . . . . . . . . . 137Figure C.1 Nays2DH model output evaluation . . . . . . . . . . . . . . . . . . . . . . 142Figure C.2 Calculation of cumulative seasonal habitat . . . . . . . . . . . . . . . . . . 143xiList of SymbolsSymbol Definitionθ Slope angle of a surface [deg]A2D The two-dimensional area of a topographic surface [m2]A3D The three-dimensional area of a topographic surface [m2]Ad Upstream contributing catchment area [km2]Ap Modeled pool area [m2]Ahv Modeled high-velocity wetted area [m2]Apwc Modeled pool area with wood cover [m2]as Fall angle of a stem [deg]B Annual rate of bank erosion [m/yr]Cp Cumulative modeled pool area [m2]Chv Cumulative modeled high-velocity wetted area [m2]Cpwc Cumulative modeled pool area with wood cover [m2]D50 Median grain size of bed sediment [m]D In-situ decay of in-stream wood [m3/yr]Db Bankfull channel depth [m]E(Vb) Expected volume of a tree entering a stream from a single fallevent [m3]I Wood input to a channel from all sources [m3yr−1]Ibe Wood input from bank erosion [m3yr−1]IbeS Bank erosion wood input of standing, live timber [m3yr−1]IbeCWD Bank erosion wood input of coarse woody debris [m3yr−1]Im Wood input from riparian mortality [m3yr−1]It Wood input due to fluvial transport [m3yr−1]k Decay constant for in-stream woodkcwd Decay constant for coarse woody debrisxiiL Overbank losses or loss of wood from a channel due to migration[m3yr−1]Lw Length of wood pieces [m]O Combined wood output from a channel reach [m3yr−1]Oca Wood output due to channel migration and abandonment[m3yr−1]Od Wood output due to in-situ decay [m3yr−1]Ot Wood output due to fluvial transport [m3yr−1]Pθ Probability of piece movement based on orientation anglePf Probability of tree fall over a specified time intervalPmove Probability of a wood piece movement based on ratio of piecelength to channel widthPmove Probability of wood piece movement over a given time intervalPn Percentile flow level, for nth percentilePspan Proportion of the channel spanned by a gravel barPvar Proportion of variance explained by channel morphologyQ Water discharge [m3s−1]Qi Wood transported into a channel reach [m3yr−1]Qo Wood transported out of a channel reach [m3yr−1]S Storage, used in the context of both sediment and wood [m3]Sb Sediment stored in bars per unit length of channel [m3m−1]Sw Wood travel distance over a given time interval [m]t Time [s] or [yr]Tr Topographic roughness or complexity [m2/m2]Vc Volume of wood stored in a stream channel [m3/m]VCWD Density of coarse woody debris in the riparian zone [m3/ha]Vo Volume of wood at time t = 0 for budget model [m3/m]Vs Density of standing, live timber in riparian zone [m3/ha]Wb Bankfull channel width [m]x Downstream or segment distance [m]Z Elevation [m]Zb Elevation of bar tops [m]Z j Distance from a given tree to the riverbank [m]Zr Relative bed elevation [m]Zt Elevation of the channel thalweg [m]xiiiList of AcronymsAcronym DefinitionACF Autocorrelation FunctionADV Acoustic Doppler VelocimeterBC British ColumbiaBCVRI British Columbia Vegetation Resource InventoryBE Bank erosionCIP Cubic Interpolated Pseudo-particleCWD Coarse Woody DebrisDEM Digital Elevation ModelDS Downstream, usually in relation to another featureFFT Fast Fourier TransformLP Longitudinal Profile, usually referred to in relation to survey ap-proachLW Large Wood, also referred to as in-stream woodMAD Mean Annual DischargeNays2DH Two dimensional hydrodynamic modelRiver2D Two dimensional hydrodynamic modelSA Study Area, in reference to Carnation Creek channel segmentsSE Standard error of the meanSI Site Index, used for wood regeneration simulationsSD Standard DeviationTIPSY Table Interpolation Program for Stand Yields. Model to simulateforest regenerationUS Upstream, usually in reference to another locationUAV Unmanned Aerial VehiclexivAcknowledgmentsI would first like to thank my very patient family. My parents, Bruce and Hilary, have alwaysprovided endless love, encouragement, and support, and also inspired in me a keen interest inthe natural world. My brothers Tim and Iain were always up for adventures which were often awelcome distraction and a great chance to explore our home province of B.C. My grandmotherMerle and aunt Ros also deserve a special mention for their support and constant interest inthe work I have been doing. To the rest of my family members, who I probably saw much lessthan I should have over the past several years, we have some catching up to do!To Leonora King, my partner in research, adventures, and life, I’m not sure I could havedone this work without your love and support. The many words of wisdom you provided madethe journey much easier, and you have always helped me challenge my assumptions and keptwhat matters in perspective.I thank my friends made at UBC and elsewhere. My long-time Vancouver friends Eric,Alex, and Graham were always willing to talk and take part in a diverse range of adventuresranging from mountain biking to sea kayaking. Dave, Julie, Joel, Charles and Tobias werealways game for a fishing, cycling, or camping trip, not to mention a good debate. Marc andKiely exposed me to many new ideas and helped shape the way I see how science is done, oftenwhile out searching for wildlife. Raphael and Lucy, conversations with you about the trials andtribulations of research were very cathartic. You have all kept me on my toes!Marwan, thank you for your supervision, patience, trust, and general life guidance overthe past years. You have taught me a lot about rivers, but at least as much about how totake a compassionate, outward-focused approach to science and beyond. John and Mike, thesupervision you’ve provided and the courses you’ve taught over the years left a big impression.To all the members of Marwan’s lab group past and present, thank you for your feedback andgreat ideas. Similarly, I couldn’t have asked for better office mates in room 109.The work undertaken in this thesis would not have been possible without the help of manyindividuals. On top of spending hour after hour helping me sort out Carnation Creek data, SteveBird taught me many of the field and data analysis skills that I have, and was very patient in theprocess. Dan Hogan was always willing to meet and talk about the colourful history of Car-xvnation Creek, especially if a Mexican restaurant was involved. Peter Tschaplinski and RobinPike entrusted me with the Carnation Creek dataset, and answered many questions related toit; I hope this thesis does the project justice. Dan Moore was always willing to talk about thedetails of data analysis, R packages, and craft beer. Leo and Tobias also provided significanteditorial support during the writing of this thesis. Finally, Carina Helm, Ryan Buchanan, SteveVoller, and Andrew Westerhof all helped with field data collection and logistics, for which Iam very grateful.I also must thank the UBC Geography Department staff for their logistical support over theyears. Sandy, you are the guardian angel to so many in the department! Danny and Suzannealways helped to keep me on track. Vincent provided technical support and was always happyto talk about cycling. Eric assisted with the aesthetics of many of the figures in Chapters 2, 3,and 5. Sumi, conversations with you always brightened my day!xviChapter 1IntroductionLow-order gravel-bed streams in forested, mountainous environments constitute a large frac-tion of remaining unmodified freshwater aquatic habitat. These channels often receive themajority of the coarse sediment supply to the system (Dietrich and Dunne, 1978; Roberts andChurch, 1986; Hassan et al., 2005a), possess a high degree of spatial and temporal heteor-geneity driven by sediment supply and wood loading (Gurnell et al., 2002; Montgomery et al.,2003; Hassan et al., 2005b, 2019), and thus provide valuable aquatic habitat to many organisms(Bjornn and Reiser, 1991; Pess et al., 2002; Lapointe, 2012; Richardson, 2019). Gravel-bedstreams are often defined by processes which are discontinuous in time and space. Sedimentand wood supply may be largely episodic (Church, 2002; Andreoli et al., 2007; Hassan et al.,2016) and delivery to the channel may be dictated by the spatial organisation of hillslope-channel coupling (Whiting and Bradley, 1993), which is in part a function of landscape history(Brardinoni and Hassan, 2007). In spite of their geomorphic and ecological significance inthe broader river network, the complexity, variety, and variability of geomorphic processes inforested, gravel-bed streams often renders predictions of channel response to episodic sedimentsupply and changes in wood abundance unsuccessful (Wohl, 2013). A common limitation ofprevious work is a lack of comprehensive field data where linkages between sediment sup-ply and channel processes can be evaluated over a time scale suitable for capturing irregular,episodic events. As a consequence, many aspects of channel response to sediment supply andwood load variability remain insufficiently understood.This thesis aims to improve our understanding of three inter-related and fundamental com-ponents of forested, gravel-bed streams in mountain environments: (1) how the sediment sup-ply regime, dictated by the spatial organization of hillslope-channel coupling, impacts channelmorphology and sediment storage; (2) how in-stream large wood mediates sediment through-put and impacts morphology; and (3) how supply, morphology, and wood collectively impactthe abundance and spatial distribution of aquatic habitat. This thesis focuses on these topics in-1dividually and also strives to explicitly link them through a framework of sediment connectivityand glacially imposed landscape organization.The following section (Section 1.1) provides a brief overview of the relationship betweenlandscape organization and channel form with a focus on the impact of episodic process. InSection 1.2, an argument is then made regarding the value of field data for addressing theseinterconnected problems, the primary research questions are identified, and the thesis structureoutlined.1.1 BackgroundThe morphology and stability of gravel-bed streams is largely a function of the sediment supplyconditions, including the magnitude, frequency, and texture of colluvial material delivered tothe channel (Church, 2002; Muller and Hassan, 2018). As the dominant sediment supplyregime in steepland channels is often regulated by colluvial sediment input (Slaymaker, 1993;Brummer and Montgomery, 2003), the controls on the location of this input and the inputcharacteristics are of great importance to channel form. Streams which experience this typeof hillslope sediment input are generally described as ”coupled” to their hillslopes (Nakamuraand Swanson, 1993; Whiting and Bradley, 1993), while channels with adjacent floodplains orlow-gradient regions are ”decoupled”. The typical view of mountain channels is that hillslope-channel coupling is mainly limited to headwaters (Schumm, 1973; Brummer and Montgomery,2003). However, large-scale processes altering the landscape surface (e.g. glaciation) caninfluence watershed topography through the formation of erosional and depositional featuressuch as sediment terraces and hanging valleys (Church and Ryder, 1972; Ballantyne, 2002;Tunnicliffe et al., 2012). The net effect of this glacial reconfiguration is that colluvial input isnot limited to headwaters and may occur much farther downstream (Brardinoni and Hassan,2007; Hassan et al., 2019). In this way, the landscape history influences the connectivity ofsediment from hillslope to channel (Fryirs, 2013; Wohl et al., 2019).Along sections of channel which are coupled to hillslopes, sediment (and wood, in manycases) will often be delivered through events which occur episodically such as debris flows andlandslides (Wohl, 2013). These irregular sediment supply events lead to a high degree of localvariability in sediment transport rate and storage (Church, 2002; Hassan et al., 2005a, 2008;Pryor et al., 2011). Following input to the channel, colluvial material may be transported andreworked through fluvial processes (Sutherland et al., 2002; Church, 2006), or remain in placeas stored sediment (Brummer and Montgomery, 2006). The role of episodic supply in affectingsediment storage is thought to be important in mountain gravel bed streams as this storage canact as a source of future supply (Lisle and Church, 2002). At a given location along a channelof this type, the supply state relative to the channel’s capacity for fluvial transport can lead toa characterisation of this channel segment as either (a) supply limited, or (b) transport limited2(Montgomery, 1997). A channel segment can fluctuate from one state to another through time.”Supply limited” refers to a state where the flow conditions are able to mobilize sediment of acalibre typical of the supply, yet insufficient supply is available for transport. Conversely, anexcess of supply relative to fluvial transport capacity will thus render the channel ”transportlimited”. A supply limited channel will likely degrade while a transport limited system willaggrade. While this characterisation is conceptually simple, it is often complicated by thepresence of sediment storage in the channel (Lisle and Church, 2002; Pryor et al., 2011) andalso modulation of sediment movement by wood (Montgomery et al., 2003; Wohl and Scott,2017). Sediment may cease to be supplied from external sources (i.e. hillslopes), but materialstored in the channel can act as a continuing source of supply, thus complicating the simpletransport vs. supply limited binary.Given that sediment storage and transport rate are likely to vary through time as a functionof supply conditions, specific aspects of channel morphology (such as the size and locationof bars and pools) will be affected by temporal variability of supply. Rapid and substantialdownstream responses in channel morphology are often observed following episodic sedimentinput from natural (Hoffman and Gabet, 2007) and anthropogenic (Major et al., 2016) sources.Widespread aggradation followed by degradation has been observed as a pulse of colluvialsediment moves downstream (Sutherland et al., 2002). Similarly, pool infilling and bar build-ing have both been observed as a function of episodic supply increases (Hoffman and Gabet,2007). Numerous experimental and modeling studies also highlight the influence of supplyvariability on channel gradient and grain size (Elgueta-Astaburuaga et al., 2018; Muller andHassan, 2018).A complicating factor in channel response to episodic sediment supply is the presence ofin-stream large wood (LW). In channels of intermediate size and smaller in forested regions,wood added to the channel from colluvial and riparian sources strongly modifies sedimentthroughput (Hogan, 1989; Haschenburger and Rice, 2004; Eaton et al., 2012; Davidson andEaton, 2013; Wohl and Scott, 2017), forming spatially discrete storage areas (Hogan et al.,1998; Wohl and Scott, 2017) and altering channel morphology (Montgomery et al., 2003).Large and immobile wood pieces and accumulations intercept sediment moving downstream(Haschenburger and Rice, 2004). This interception is achieved through impacts to channelhydraulics whereby reduced velocities upstream of LW force the deposition of sediment, whilea lack of sediment throughput results in scour pool areas downstream of logjams (Montgomeryet al., 2003; Davidson and Eaton, 2013; Hafs et al., 2014). Large channel-spanning jams inconfined systems are capable of storing many times the annual bed sediment load (Hogan,1986; Andreoli et al., 2007; Ryan et al., 2014), and this material can then become stabilizedwith vegetation and incorporated as part of the floodplain (Hogan et al., 1998). The nearlyrandom nature of wood additions in upland streams (Davidson and Eaton, 2015) combined3with mixed but limited transportability of wood pieces (Benda and Bigelow, 2014) leads tohighly variable magnitudes and locations of wood storage, and therefore significant variabilityof impact on sediment throughput and channel morphology. The spatial and temporal influenceof LW on sediment storage and channel morphology is therefore a function of wood abundance(Wohl and Scott, 2017), mobility (Montgomery et al., 2003; Davidson and Eaton, 2013; Dixonand Sear, 2014), source (May and Gresswell, 2003; Rigon et al., 2012), and how easily achannel can migrate around a jam (Hogan, 1986; Wohl and Scott, 2017).Associating channel conditions with aquatic habitat quantity and quality, especially forsalmonids, has historically been a common motivation for research in upland gravel-bed streams(Montgomery et al., 2003). Habitat conditions for salmonids synthesize numerous processesat the watershed to local scale (Bjornn and Reiser, 1991; Montgomery et al., 2003; Wipfliand Richardson, 2015). Channel morphology and cover provided by wood are both key com-ponents which form the physical template within which salmonids exist (Cederholm et al.,1997; Beecher et al., 2002; Hafs et al., 2014). Suitable hydraulic conditions are essential forallowing organisms access to resources such as low velocity pool areas (Bjornn and Reiser,1991; Beecher et al., 2002) which are sufficiently deep to allow for passage, foraging, andoptimal cover (Nickelson and Reisenbichler, 1977; Bjornn and Reiser, 1991; Beecher et al.,2002). Structural cover elements in stream channels, such as LW, also serve an important rolein sheltering juveniles from predation and reducing water velocities, and in providing accessto food (Bjornn and Reiser, 1991; Hafs et al., 2014). As a consequence, LW quantities areoften positively associated with juvenile salmonid abundance (Cederholm et al., 1997; Benkeand Wallace, 2003; Pess et al., 2012). As channel morphology and wood abundance influencehabitat, variability through time in these factors must also result in changes to the quantity,quality, and location of available habitat for salmonids.This brief introduction describes the scales and process connections present in forestedupland streams. In summary, sediment connectivity, impacted and modified by wood, servesto regulate the way a channel will behave through time with direct implications for aquatichabitat. This framework provides a means to associate large scale structure, such as that of alandscape influenced by glaciation, to small scale process and form, such as bar building andpool filling at the channel unit scale.1.2 Motivation and study objectivesSediment delivery, channel morphology, in-stream wood, and the association amongst thesevariables and aquatic habitat have been the subject of a great deal of research over the past fivedecades or more (Ruiz-Villanueva et al., 2016). Evaluation of these river system componentshas traditionally involved the investigation of separate components in isolation. However, it hasbeen known for some time that quantifying and relating major processes in systems where these4components occur irregularly is challenging when short term, localized data are used (Dietrichand Dunne, 1978). The response of channel structure to episodic sediment supply has primarilybeen investigated through the use of experimental and numerical methods and field data of alimited spatial and temporal scale, as the unpredictable nature of sediment input requires fieldstudies of long duration or of opportunity (e.g. Hoffman and Gabet, 2007; Pryor et al., 2011).Similarly, temporal aspects of interactions between LW and channel morphology are oftenmodeled numerically or are evaluated through physical experiments, again due to difficultieswith long term data collection. These components of physical aquatic habitat are similarlyrarely evaluated over the long-term and linked over time to other river processes, as a result ofthe limitations above (Fabris et al., 2017). There is a need to synthesize separate componentsof upland gravel-bed streams to gain a better perspective on how sediment connectivity andprocess linkages are realized at local scales in terms of channel morphology and aquatic habitat.In summary, many open questions remain in regards to temporal patterns of habitat, wood-morphology interactions and sediment supply, and how these are inter-related through timeand in the broader context of catchment structure.Syntheses of sediment supply, morphology, and habitat elements are challenging from anexperimental and numerical standpoint as the detailed physics or other process controls arenot always well understood. Comprehensive field data, however, can capture this complexity,but with a trade-off in ease of attributing cause to effect. However, If many variables can becollected concurrently, then it may be possible to link process to form across scales. In thecase of fluvial systems, this type of data would include sediment supply (source and quantity),a record of channel response and/or transport rate, and a record of wood quantity and inputsources. Collection of this type of data, particularly over the long term, allows for concurrentevaluation of processes which are otherwise challenging to connect using numerical or physicalmodels.The primary objective of this thesis is to collectively evaluate the above components offorested, gravel-bed streams over temporal and spatial extents that are sufficient to captureepisodic processes driving change in the system. The thesis framework is structured aroundthree broad questions:1. What duration and extent of measurement are necessary to capture the fundamental vari-ability of a forested, gravel bed-stream?2. What is the nature of channel dynamics observed over a multi-decadal timescale?3. How are watershed structure, sediment supply, channel morphology, wood abundance,and aquatic habitat linked through time and across space in relation to these dynamics?A detailed 45-year record of sediment delivery, channel morphology, and wood abundance5in a coastal gravel bed stream is used to meet the primary thesis objective. This dataset, col-lected in Carnation Creek, B.C., captures highly variable supply conditions stemming fromcolluvial sediment input. Using this dataset, several specific research questions are addressed:1. How is stored channel bed sediment distributed spatially and temporally, and what ex-plains these distributions?2. How are changes in sediment storage related to sediment transfer rate over time inepisodic systems?3. What role does in-stream large wood play in modulating sediment storage and channelmorphology through time?4. Does a loss of wood supply from the riparian zone affect wood storage, and in turnchannel morphology?5. How does variability in sediment supply, channel morphology, and wood abundanceinfluence habitat conditions for salmonids?6. How does historical watershed glaciation influence watershed structure and resultingspatial and temporal organization of sediment delivery, channel morphology, and aquatichabitat?To address these research questions and objectives, the body of this thesis is divided intoone chapter describing the study watershed, and three research chapters, through which a cen-tral theme of process linkages, cascades, and scale dependencies flows. The Carnation Creekstudy area and history are described in Chapter 2, which provides a detailed overview of thephysical watershed setting and the natural and anthropogenic history of the catchment. Inaddition, data collection sites within the watershed are described.Chapter 3 aims to address research questions 1, 2, and 6 by examining the spatial and tem-poral organization of sediment storage at the channel unit and main stem scale. Additionally,sediment transfer and storage state are linked to the timing and location of colluvial sedimentsupply. A conceptual model is proposed which describes channel response to supply at differ-ent positions along a network. This chapter addresses the need for a better understanding ofsediment storage patterns and dynamics in upland gravel bed streams over the long term.Chapter 4 examines the spatial and temporal dynamics of large wood at the same twospatial scales as Chapter 3, linking LW, and particularly logjam evolution to sediment storagepatterns and other aspects of channel morphology. A simulated wood budget is developedto forecast LW abundance in relation to history of riparian harvesting. Using this approach,this Chapter addresses questions 3 and 4, and, to the author’s knowledge, presents the longestsemi-continuous record of wood piece data available.6In Chapter 5, the findings of the previous chapters are synthesized into a conceptual modelwhich proposes that temporal variability in habitat is a function of changes in the sedimentsupply regime of a system. This Chapter aims to address questions 5 and 6. To assess habi-tat variability and the conceptual model, flow depths and velocities are simulated with a 2Dhydrodynamic model, using annual topographic survey data from the Carnation Creek studyareas as input.7Chapter 2Carnation Creek study watershed2.1 Carnation Creek: An interdisciplinary long-term monitoringprojectData collected in Carnation Creek, B.C., are central to this thesis. Beginning in 1970, a mon-itoring program was initiated in the watershed with the objective of evaluating the influenceof forest practices on watershed process and Pacific Salmonid populations (Hartman et al.,1996; Tschaplinski and Pike, 2017). The project was initially a collaboration amongst severalgovernmental, industry, and academic organizations, including the federal Department of Fish-eries and Oceans (the project lead), the B.C. Ministry of Forests, and MacMillan Bloedel Ltd.,who were contracted to undertake harvesting and road building in the watershed (Hartman andScrivener, 1990). A scientific panel composed of up to 20 cross-disciplinary experts was in-volved in the project design and general oversight (Hetherington, 1987). As of 2019, the B.C.provincial government provides primary project management, mainly from the Ministry of En-vironment and Climate Change, though other partners remain involved. A partner addition ofnote is the Huu-Ay-Aht First Nation, whose traditional territory encompasses the watershed.The particular catchment was selected as it was one of few small Vancouver Island wa-tersheds which remained unharvested by the mid 1960s, and yet had reasonable access. Datacollected in the watershed have included fish populations and behaviour, streamflow, aquaticand riparian invertebrates, forest hydrology and groundwater, meteorology, channel morphol-ogy, slope stability and landsliding, sediment transport, and forest regeneration (Hogan et al.,1998). Well over 200 publications have resulted from the project to date.2.2 Watershed setting and biophysical characteristicsThe 11.2 km2 basin (Figure 2.1a) is located on the southeastern edge of Barkley Sound, ap-proximately 20 km northeast of the village of Bamfield, on the west coast of Vancouver Island,8British Columbia. The catchment is oriented roughly along a northeast - southwest axis, anddrains directly into the ocean where the creek has formed a substantial alluvial fan. The cli-mate of the region (Figure 2.2) is temperate and humid, with annual precipitation totals rangingfrom close to 290 cm/yr at sea level to over 500 cm/yr at higher elevations (Tschaplinski andPike, 2017). As a result, Carnation Creek falls within the Coastal Western Hemlock (CWH)biogeoclimatic zone, typical of the west coast of North America from the Olympic Peninsula toHaida Gwaii, B.C. (Krajina, 1969). Hillslopes are predominantly covered with western hem-lock (Tsuga heterophylla) and Amabilis fir (Abies amabilis). A greater diversity of species isfound in the valley bottom and in the riparian zone including western redcedar (Thuja plicata),Sitka spruce (Picea sitchensis), and red alder (Alnus rubra) in addition to Hemlock and Fir.The primary bedrock geology of the watershed consists of the Vancouver Island Intrusion,a large Jurassic-age batholith composed of granodiorite (Hartman et al., 1996). As with therest of coastal B.C., the catchment was glaciated several times, most recently during the Fraserglaciation, which ended approximately 13 000 years ago (Church, 2010). As a result, till coversbedrock to an average (but variable) depth of approximately 0.7 m, though the valley bottomcontains some fine sediment of marine origin. The glacial legacy of the region has shapedmany features of the watershed’s topography: catchment relief is close to 900 m, and steepgradients exceeding 50% are found throughout the basin (Hartman and Scrivener, 1990).As with many glaciated basins in this region (e.g. Brardinoni and Hassan, 2007; Hassanet al., 2019), the catchment profile displays a stepped structure (Figure 2.1b), with an alter-nating sequence of steep and shallow gradients. The specific origin of this structure may be aproduct of lateral scour within the flat lowland area along the lower 3.0 km of channel, and abedrock base to steep channel section upstream indicates that colluvial material isn’t respon-sible for the overall profile shape. Upstream of this step, it appears that lateral ice flow acrossa breached divide may have occurred, resulting in the comparatively wide valley section here.Tectonic activity is unlikely to be responsible for catchment structure, as activity in the regionis low. The channel therefore contains multiple depositional zones along its length, and varyingdegrees of hillslope-channel coupling.Slope failures have been common within the catchment, and are a major source of sedimentto the stream channel (Zimmermann et al., 2004). Examples of slope failures reaching thechannel in coupled areas are shown in Figure 2.3a and b. The channel is coupled to hillslopesprimarily in two locations: (1) a confined channel section between 2.9 and 4.5 km from theriver mouth (see Figures 2.1 and 2.3c) hearafter referred to as the ”canyon reach”, and (2) inthe channel headwaters. Most slope failures connected to the channel network occurred in theearly 1980s and late 1990s.The hydrological regime of Carnation Creek is rain dominated, and most floods occurduring storms in the autumn and winter months. Discharge can vary by more than two orders9B WeirC TributaryFigure 2.1: (a) Carnation Creek location (see inset) and watershed. Note locations ofintensive study areas, which ascend in order from downstream to upstream (sitesSA-2 to SA-9). B Weir is the primary hydrometric gauging station in the watershed.(b) Profile of channel main stem, with locations of channel-coupled landslides anddebris flows marked.of magnitude over a 48 hour period (Hartman and Scrivener, 1990), and bankfull dischargeranges from ∼ 25 - 35 m3s-1 (Haschenburger and Wilcock, 2003). Summers are relatively dry,with only 15-25% of total annual precipitation falling between the months of May and October.Mean annual discharge is 0.81 m3s−1 and summer flow conditions are usually well below 0.5m3s−1. During very dry summers sections of the channel may lose surface flow.The primary channel morphology is riffle-pool, although the channel in the canyon reachis mainly step pool. Gradients are typically below 1% in surveyed areas, but exceed 8% inthe canyon (see Figure 2.1b). Active channel width averages approximately 15 m, and fewislands are present. Numerous logjams and individual pieces are found along the channel10PrecipitationTemp. rangeAvg. tempFigure 2.2: Precipitation and temperature characteristics for Carnation Creek, collectedat low elevation. Data are from Environment Canada station ID: 1031413.and both act as important controls on channel morphology in the system (Hartman et al.,1996; Haschenburger and Rice, 2004). The threshold of motion for bedload sediment in thechannel downstream of the canyon reach has been estimated at between 4-7 m3s−1 from re-lationships between discharge and tracer particle mobility (Haschenburger and Rice, 2004;Haschenburger, 2011), and based on direct sampling and analysis of magnetic tracer stonemobility, bed material is fully mobile above approximately 15 m3s−1 (Hassan et al., 2008).Bed texture ranges from small gravel (median size 20-40 mm) near the river mouth to cobblesand boulders in the canyon reach (see Figure 2.3) and headwaters. Large bars spanning overhalf the channel’s width are common in areas up and downstream of the canyon, and mobilesediment is generally abundant.Aside from sediment added to the channel from colluvial sources, other sediment sourcesinclude tributary input and bank erosion. Overall, sediment is supplied in relative abundanceto the channel, but the system likely alternates between supply and transport-limited statesthrough time (Hassan et al., 2008), particularly in upstream locations.The creek contains several anadromous fish species, including chum salmon (Oncorhynchusketa), coho, (Oncorhynchus kisutch), steelhead trout (Oncorhynchus mykiss), and sea-run cut-throat trout (Oncorhynchus clarkii) (Tschaplinski and Pike, 2017). The anadromous speciesare limited to the channel downstream of approximately 3.1 km from the river mouth, at whichpoint a waterfall impedes upstream migration. Upstream of this point, resident cutthroat troutare present. Chum salmon generally spawn in the lower 500 m of the stream channel, whilecoho make use of most of the lower 3.0 km of the creek. Off-channel rearing habitat is lo-11Figure 2.3: Overview of sediment supplied to canyon reach, with examples of debrisflows and landslides shown in (a) and (b), highlighted with red lines. A ground-level photograph from within the canyon reach is shown in (c); note the coarse lagdeposits.12Table 2.1: Characteristics and channel geometry of the eight intensive study areasStudyareaYearssurveyedaAdb(km2)Length(m)Wb(m)Dbc(m)Slope(%)dSA-2 44 10.1 82 13.8 0.99 0.73SA-3 44 9.3 68 19.6 0.97 0.36SA-4 42 8.8 62 21.2 0.42 1.35SA-5 42 8.1 75 13.7 0.61 0.91SA-6 44 7.8 70 14.9 0.74 0.53SA-7 42 7.6 51 19.6 0.77 0.86SA-8 44 7.4 58 16.9 0.87 0.95SA-9 41 3.8 148 11.0 0.59 1.90a As of 2015; surveys are ongoing. No data were collected in 2010.b Contributing area, measured to approximate center of study areac Bankfull depth.b Calculated from thalweg elevations.cally abundant. A structure built near the river mouth (hereafter referred to as the ”fish fence”)traps incoming and outgoing fish which are then captured and measured prior to release on theopposite side of the fence.Locations of geomorphic researchStream channel morphology in the catchment has been studied at two spatial scales: (1) thelongitudinal profile, channel width, and basic topographic data have been surveyed over threekilometers of channel from the river mouth to the canyon reach (Figure 2.1b) four times be-tween 1991 and 2017, and (2) detailed channel bed topography has been captured annually ineight study areas (see Figure 2.4) since 1971, referred to throughout as SA-2 to SA-9. A 9thstudy area was established in the river estuary in 1971 (SA-1), but was abandoned in the late1980s and is not used for this study. Seven of these study areas (SA-2 to SA-8) are spaced 300-500 m apart along the lower 3 km of channel, overlapping with the four profile surveys. Theeighth study area (SA-9) is located upstream of the canyon reach (see Figure 2.1). Photographsof the study areas are found in Figures 2.5 and 2.6, and study area characteristics are located inTable 2.1. While the study areas, which were originally established to encompass fish samplinglocations, are relatively short (5-10 bankfull width equivalents), the high spatial and temporalresolution of their surveys provide invaluable information on localized morphological featuresand sediment storage dynamics through time.13Figure 2.4: Carnation creek watershed, with expanded main channel, study area loca-tions, and corresponding images shown. Study area maps for this figure are gener-ated from 2015 data. The channel length coloured in red is shown on the watershedinset for scale.14Figure 2.5: Photos of study areas (a) SA-2, (b) SA-3, (c) SA-4, and (d) SA-5 lookingdownstream. All photos are from 2015 except (c) which was taken in 2013. Notethe abundance of sediment of a mobile size, especially in (a) and (b). Photos fromIain Reid and Steve Bird with permission.2.3 Catchment history: A summary of key eventsOver the 48 years since the project onset, several events of note have occurred, including timberharvesting, major floods, and debris flows (Figure 2.7). As part of the watershed study design,the catchment was partly logged from the late 1970s to the early 1980s. The riparian distur-bance history of the lower watershed can be broadly divided into two zones: one region whereriparian vegetation was left undisturbed along channel banks in buffer zones of widths rangingfrom 1 to 70 m (river mouth to approximately 1300 m upstream), and another region wherenear-complete and complete removal of material (1300-3100 m) was undertaken (Hartman andScrivener, 1990; Tschaplinski and Pike, 2017). In a 900 m stretch from 1300 m to 2200 m,merchantable timber within the channel was extracted. Post-harvest regeneration included amix of species but was predominantly western redcedar and Amabalis fir, with some Douglasfir and western hemlock. Red alder was also common as an early-stage successional species inthe riparian zone. Between 1975 and 1981, approximately 41% of the total catchment area washarvested. Further details of the basin’s logging history and riparian disturbance can be foundin Hartman and Scrivener (1990), Hartman et al. (1996), and Tschaplinski and Pike (2017).15Figure 2.6: Photos of study reaches (a) SA-6, (b) SA-7, (c) SA-8, and (d) SA-9 lookingdownstream. All photos are from 2015 except (d) which was taken in 2013. Noteabundance of mobile sediment in (c). Photos from Iain Reid and Steve Bird withpermission.In the ten years after the bulk of harvesting took place, the two highest recorded flowsoccurred. The second highest, close to 50 m3s−1, occurred in 1982. The highest flow was in1984, at 64 m3s−1, and also resulted in numerous debris flows which delivered sediment andlogging slash to the canyon reach. While several additional large storms occurred in the early1990s, no major debris flows or input from hillslopes occurred until 2007, when a minor debrisflow entered C tributary (2.1) but did not reach the main channel. The most recent major stormevent occurred in January 2015, and resulted in notable channel change, wood transport, anddamage to the fish fence below SA-2.Collection of data pertaining to channel morphology and wood was underway in all studyareas by 1973, and is still collected annually to date. No data were collected for one year in2010, and no data were collected in SA-9 in 1990. The first long-profile survey was conductedin 1991, and recently, a full survey conducted with an unmanned-aerial-vehicle (UAV) was un-dertaken (Figure 2.7). Environmental and biological data collection for the project is expectedto continue for the foreseeable future.16Begin present investigationFigure 2.7: Timeline of major research and other notable events in Carnation Creek overthe project history. Purple shaded area corresponds to the pre-harvest baseline datacollection period, while the yellow shaded area corresponds to the period over whichharvesting took place in the riparian zone.17Chapter 3Spatial and temporal patterns ofsediment storage3.1 SummaryThe movement of sediment through mountain river networks remains difficult to predict, asprocesses beyond streamflow, particle size, and channel gradient are responsible for the en-trainment and transport of bedload sediment. In deglaciated catchments, additional complexityarises from glacial impacts on landscape organization. Research to date indicates that thequantity of sediment stored in the channel is an important component of sediment transportin systems which alternate between supply and transport limited states, but limited long-termfield data exist which can capture storage-transfer dynamics over a timescale encompassingepisodic supply typical of mountain streams. The 45 year dataset of annual and decadal-scaledata on sediment storage, channel morphology, and wood loading in Carnation Creek is usedto investigate the spatial and temporal organization of sediment storage in the channel.Sediment is supplied episodically to the channel, including additions from debris flows inthe early 1980s just upstream of the studied channel region. Analyzing the spatial and temporalorganization of sediment storage along 3.0 km of channel main stem reveals a characteristicstorage wavelength similar to the annual bedload particle travel distance. Over time, two scalesof variation in storage are observed: small scale fluctuation of 3-10 years corresponding tolocal erosional and depositional processes, and larger scale response over 25-35 years relatedto supply of sediment from hillslopes. Complex relationships between storage and sedimenttransfer (i.e. annual change in storage) are identified, with decadal scale hysteresis present instorage-transfer relations in sites influenced by hillslope-sediment and logjams. A conceptualmodel is proposed linking landscape organization to temporal variability in storage and tostorage-export cycles. Collectively, results in this study reaffirm the importance of storage to18sediment transport and channel morphology, and highlight the complexity of storage-transportinteractions.3.2 IntroductionBedload sediment transport in gravel bed rivers is a function of streamflow, sediment supplyand caliber, bed surface structures, and catchment history. Practically, sediment transport rateis calculated using a hydrological record empirically tied to limited particle size and channelgeometry data. These transport capacity-based calculations produce results that differ from ob-served transport rates often by an order of magnitude or more (Gomez and Church, 1989), withdiscrepancies attributed to non-uniform sediment (Andrews, 1983), bed surface armoring andother bedforms (Buffington and Montgomery, 1997; Church et al., 1998; Hassan et al., 2008),and limited sediment of a caliber that can be transported (Lisle and Church, 2002). Field andexperimental laboratory work suggests that sediment stored in stream channels may have sig-nificant bearing on sediment transport rates in systems which are subject to episodic sedimentsupply (Lisle and Church, 2002; Hassan et al., 2008). Systems with variable supply can switchfrom supply to transport limited states over annual to decadal timescales independently of flowconditions (Pryor et al., 2011). However, our understanding of the connection between sed-iment stored in the stream channel and sediment transport remains limited, as few field dataexist that span timescales sufficient to capture these storage state changes.The importance of sediment storage on transport dynamics has been suspected for manydecades (e.g. Schumm, 1973; Graf , 1987). However, the direct influence of storage on trans-port rates and transport capacity appears complex, with both linear and non-linear relationshipsobserved (Lisle and Church, 2002; Lisle and Smith, 2003; Smith, 2004; Lisle, 2008). Field andexperimental studies illustrate the importance of storage on the export rate of sediment froma catchment, but consistent relationships remain elusive. By tracking sediment storage in theHigashi-Gouchi River of southern Japan, for example, Maita (1991) observed an exponentialdecrease in storage over four years. In contrast, a more linear rate of storage loss was observedin two tributaries of the Waipoa River, New Zealand, over a 10 year period (Marutani et al.,1999). A reanalysis of experimental work demonstrated roughly linear (but scattered) storagedecreases over time, which also correspond to a linear relation between storage and sedimentoutput rate (Lisle and Church, 2002). Observations of storage change in Cuneo Creek, Califor-nia, indicated exponential losses through time, while experiments simulating the creek showedapproximately linear relations between sediment transfer rate and storage volume at severaldifferent storage levels (Pryor et al., 2011). Flume experiments by Elgueta-Astaburuaga et al.(2018) also suggest that sediment stored in the channel leads to long-term persistence in trans-port rates, implying a positive transport-storage relationship.A key aspect of the complexity of transport-storage connections in mountain channels is19the spatial organization of stored sediment. This can be dictated by the specific location ofmass movement or steep tributary input (e.g. Owczarek, 2008), which can be influenced by acatchment’s glacial history (Hassan et al., 2019), or by other channel features, such as logjams,should sediment be remobilized and deposited farther downstream (Sutherland et al., 2002). Ifsediment is very coarse or the deposit armors, the limited mobility may cause the sediment toremain close to the point of entry to the channel (Reid and Dunne, 2003; Brummer and Mont-gomery, 2006). In other cases, sediment from hillslopes may move through channels rapidly(Sutherland et al., 2002; Hoffman and Gabet, 2007), resulting in sediment deposits in bars orbehind logjams distant from the input location (Hogan et al., 1998). In channels with riparianforests or other wood sources, logjams and other wood pieces store sediment by forcing deposi-tion of material through a reduction of water velocities upstream of wood obstructions (Hogan,1986; Nakamura and Swanson, 1993; Webb and Erskine, 2003; Davidson and Eaton, 2013;Jackson and Wohl, 2015). In the absence of logjams, sediment delivered to a confined, steepchannel reach will travel rapidly downstream and deposit in lower gradient regions (Bountryet al., 2013), or zones of decreasing transport capacity (Gartner et al., 2015). The spacing andlocation of depositional features may also be related to the particle step lengths: sediment haslong been suggested to travel from bar to bar (Neill C., 1967; Pyrce and Ashmore, 2003a,b),and for instance in steeper bedrock channels, sediment is found to have a step length similar tothe spacing of alluvial sediment patches (Hodge and Sklar, 2011).In catchments with episodic sediment delivery, storage will vary through time at a givenlocation. The timescale of sediment storage variability is governed by competition betweenrates of deposition and erosion: the size, texture, and location of a storage unit can allow it topersist only briefly (< one year) or for many decades (Dietrich et al., 1982; Kelsey et al., 1987;Madej and Ozaki, 1996). When sediment transport is in relative equilibrium with supply, stor-age at a broad scale is unlikely to vary substantially through time; if a system becomes eitherstarved or oversupplied with sediment, however, then a negative or positive change in storage,respectively, occurs (Major et al., 2016). If added sediment is assumed to move through achannel as a dispersive wave (Sutherland et al., 2002), then storage will change through timeat a given location as this wave passes, with the rate and magnitude of change related to theproximity and magnitude of the sediment addition. The texture of stored sediment also hasbearing on how the sediment body changes through time. Sediment deposits with a wide grainsize distribution may armour upon degradation, providing some resistance to further erosion(Brummer and Montgomery, 2006).The persistence of sediment storage units generated by features such as logjams is con-trolled in part by the integrity of the structure, and in part by a channel’s ability to avulse(Montgomery et al., 2003). Sediment deposited behind logjams can persist for many years ifit is shielded from erosion and then stabilized with vegetation growth (O’Connor et al., 2003;20Collins et al., 2012). This material can subsequently be re-introduced to the active channel asthe jam structure degrades with time (Haschenburger and Rice, 2004).Several researchers have investigated sediment supply dynamics, storage, and transport,often through the lens of connectivity (e.g. Beylich and Laute, 2015; Gran and Czuba, 2017).While a few articles have synthesized these elements (e.g. Hogan, 1989; Wohl and Goode,2008; Jackson and Wohl, 2015; Wohl and Scott, 2017), studies are often flume-based or usecomparatively short field datasets. Without a data record in geomorphic systems which cap-tures fundamental dynamics of episodic supply, difficulties remain in connecting spatial andtemporal patterns of sediment storage to sediment transport characteristics. In this Chapter,an attempt is made to define and characterize these scales of sediment storage in mountainstreams and also the implications for predicting bedload sediment transport. Using the 45year Carnation Creek dataset which contains detailed information on sediment supply, channelform, wood loading and sediment storage at two spatial scales, the aim of this study is to meetthree objectives: (1) to quantify and describe the spatial and temporal distributions of sedimentstorage; (2) to evaluate observed patterns in the context of sediment transfer rates and hillslopesediment supply; and (3) to assess these findings in terms of general channel adjustment andresponse in previously glaciated mountain streams. Sediment storage units are hypothesizedto be organized across space in a non-random way, while temporal variability hypothesizedto be episodic and tied to supply. Additionally, a conceptual model is proposed which linkslandscape organization to temporal variability in storage and to storage-transfer cycles. Thisresearch aims to improve our understanding of sediment transport and storage dynamics at atimescale which more closely matches that of infrequent but dominant episodic supply eventsin deglaciated mountain catchments.3.3 Methods3.3.1 Topographic data collection and preparationTopographic data in the study areas and along the channel profile (see Figure 2.4) have beencollected using several methods, details of which are found in Table 3.1. Channel surfaces inthe study areas were initially mapped with a theodolite along established 1-3 m interval cross-sections until 1987, followed by mapping with an automatic level (1987-2008) and then witha total station (2009-2015). Survey precision in the study areas is high, and closing traverseerrors typically range from 0.01-0.04 m in the x and y coordinates, and 0.005 - 0.03 m in thez (elevation) coordinate. These errors are less than or similar to the median grain size (D50) inall study areas (Figure 3.1a). Uncertainty in surveyed channel area was assessed by comparingtwo consecutive years of data between which no major floods occurred (2011-2012). Acrossall sites, area differences amounted to less than 2%. The point density of the study area surveys21generally ranges between 0.5 and 1.5 points per m2, with the highest resolution associated withtotal station data. The effect of point density and other survey quality elements on topographicanalysis has been discussed extensively by Lane (1998). These data are of moderate resolution;while grain-scale features are not captured, larger scale bars, lobes of sediment, and othertopographic features are well described at this point density. Important channel attributes weresurveyed, such as bank tops and bottoms, thalweg position, and margins of the wetted channel.Following inspection of each year of data for correct attribute labeling and erroneous points,digital elevation models of the annual topographic surveys were generated from interpolatedsurvey data. Interpolations were done to 10 cm resolution using the R programming language(RCoreTeam, 2017).22Table 3.1: Field data and data collection methodsSurvey type DateRangeInterval # surveys Method Variables collected NotesTopographyStudy area topog-raphy1971-2015annual 44 Theodolite (1971 -1986), Automatic level(1987-2008), Totalstation (2008-2015)Elevation, Easting, Nor-thing, water depth, bankedges, thalweg, water’sedge.Surveyed point densityranges from 0.5-1.5 ptsper m2Long-profile (LP)survey1991-20177 yrs 4 Automatic level Distance; thalweg, bar,and bank elevation;channel morphology;water depth; channelwidth; bar spanPositions referencedagainst survey bench-marks along channelWoodSA wood 1971-2015annual 44 Survey with hand tape(mapping) and directmeasurement (dimen-sions)Position, orientation,and approx dimensions(1971-2000), length,diameter of each piece(2001-2015)Volumetric wood infor-mation obtained frommaps 1971-2000LP wood 1991-20177 ys 4 Categorical methods,survey with hand tape.Categorical piece dimen-sions and orientations;logjam span, age, size,sediment storage, anddecay classSee Hogan (1987) fordetails230.05 0.50 5.00 50.00 500.00020406080100Particle size (mm)Percent finer thanMobileSurfaceTravel distance per year (m)Years0 100 200 300 400 500 60002468(a) (b)Figure 3.1: (a) Particle size distributions for mobile sediment and bed surface sediment.”Mobile sediment” was sampled in 2006 from sediment repeatedly excavated froma pool near the river mouth, and is also representative of the subsurface material.Surface particle size is derived from Wolman pebble counts, of which 18 were con-ducted along the lower 2.3 km of stream channel in 2017. The plotted line is thesummed distribution of all counts. (b) Distribution of annual travel distance formobile sediment. Values are based on virtual velocities reported in Haschenburger(2011) applied to the historical flow records for the period.To supplement the study area surveys, four longitudinal profile surveys were conductedbetween 1991 and 2017 over a distance of 3 km (see Figure 2.1). As part of these surveys, thal-weg, bank, and bar top elevations, and fraction of the channel spanned by bars were recordedevery channel width, while bankfull width (wb) was measured every two channel widths in1991 and 1999, and every five channel widths in 2009 and 2017 with a fiberglass tape to thenearest 0.1 m. Surveys were conducted with a Nikon automatic level for elevations and a sur-veyor’s hip chain for distance, where measurements were taken to the nearest 0.1 m. Elevationsand distances were checked against known benchmarks along the channel during surveys to aidin inter-annual comparison of results. Distributed elevation errors for the level surveys rangefrom 0.5 to 1 mm per meter of channel surveyed. For analysis, width values were linearlyinterpolated to match the data resolution of the other profile variables (i.e. 1 wb).As part of both the longitudinal profile and study area surveys, characteristics and quantitiesof in-stream wood were captured. In the study areas, wood piece locations, orientations, anddimensions were mapped annually to 2001, after which pieces were both mapped and measuredwith a fiber tape. Wood data are not available from 1990-2001 in some study areas. For thelongitudinal profile surveys, wood pieces and logjams were classified based on dimensions andorientation criteria outlined in Hogan (1987) and Hassan et al. (2016) for each 1 Wb interval.243.3.2 Assessment of sediment supply and transportA detailed inventory of mass movement events exists for the catchment. Dhakal and Alila(2003) used a combination of air photo analysis (13 sets of imagery spanning from 1963 to2001) and field mapping (including channel walks to assess connectivity) to characterize over90 slope failures occurring to 1996. This inventory was updated subsequently from air photosin 2004 (see Zimmermann et al., 2004), 2009, and again in 2015. Field observations alongthe lower 3.5 km of stream channel were used to verify some air photo observations made in2015. For each slope failure identified, area, approximate depth, and connectivity to the streamnetwork was assessed.Sediment supplied from bank erosion was estimated in the study areas by examining ac-tive channel boundaries, defined from presence of erosion, moss or other vegetation lines fromsuccessive years. Active channel area in year 2 that was inactive in year 1 qualifies as ero-sion, while inactive area in year 2 that was active in year 1 qualifies as stabilization. Valueswere averaged over study area length to obtain an annual erosion and stabilization length, anddepth determined from subtraction of mean bank elevation from mean bed elevation for thatyear. Bank erosion calculations are not possible for the longitudinal profile data as exact wbmeasurement positions may vary from survey to survey as channel position changes over time.Limited data exist which directly describe sediment transport in Carnation Creek over theperiod of study. However, annual and bi-annual excavations of sediment in a ∼ 500 m3 pool atthe river mouth for maintenance of a fish passage structure (the fish fence) provide a minimumestimate of bedload transport. Similar minimum estimates can be obtained from subtraction ofsuccessive digital elevation models, coupled with a particle virtual velocity: magnetic tracerparticle studies in the catchment with more than 10 recoveries over 30 years (see Haschen-burger, 2011) have also been conducted and provide data on travel distances, burial depths,and virtual particle velocities.3.3.3 Calculation of sediment stored in barsStored sediment was quantified along the lower 3 km of channel using the longitudinal pro-file data, and in the eight study areas using the local, higher detail survey data. Followingmethodology in Hogan (1987), bar storage per unit length of channel (Sb) was calculated as:Sb = (Zb−Zt)pspanwb (3.1)where Zb is bar top elevation, Zt thalweg elevation, pspan proportion of the channel widththat the bar spans, and wb is bank-full channel width. This approach assumes flat bar tops, andmay therefore overestimate storage in some cases, as true bar shape is variable and not known ateach measurement location. While a small floodplain is present adjacent to the channel, much25of the sediment is composed of glacial, rather than alluvial material. However, an absence ofdetailed floodplain survey data renders floodplain storage difficult to estimate. Similarly, a lackof data pertaining to depth of alluvial material above bedrock precluded an investigation of thisstorage component.A digital elevation model (DEM) subtraction approach was used to calculate sedimentstorage for the eight study areas. For each site, a baseline surface of uniform elevation wasselected against which all years of data could be compared. For each study area, the baselinesurface elevation was set to be below the minimum from any year of data, and storage is thuscalculated relative to this baseline. This baseline surface was subtracted from each DEM, andboth the DEM and the base level layer were then clipped by an outline of the active channel foreach year, allowing both changing bed elevations and changing channel areas to be consideredin the calculation of storage.3.3.4 Autocorrelation and spectrum analysisScale information pertaining to temporal and spatial data can be extracted from autocorrelationfunctions and from the results of spectrum analysis. Both approaches have previously beenused in a geomorphic context to aid in, for example, examining the role of channel bed his-tory in streambed evolution (Elgueta-Astaburuaga et al., 2018), or characterising subglacialtopography (Siegert et al., 2005). All analyses here were completed using the R programminglanguage, and details of the method are shown in Appendix A.Analysis of autocorrelation in data provides information on how values are related throughthe series in time or space (Chatfield, 2003). Prior to running any analysis on the data series,Ljung-Box tests (Ljung and Box, 1978) were conducted to assess whether the data differ sig-nificantly from random noise. Results from this test indicate that all channel profile (spatial)and study area (temporal) data show some degree of autocorrelation (p< 0.05) and the data aretherefore not white noise. As continuous series are needed for the autocorrelation and spectralanalysis, the data gap in storage for the 2010 year was filled by taking the mean of 2009 and2011 values.Spectral analysis methods are helpful for checking for the presence of periodicity in thedata. In essence, this approach serves to break data series into sine and cosine functions ofdifferent frequencies using Fourier decomposition, and then assess the combined fit of thesefunctions through inspection of their power spectra (Chatfield, 2003). The method is somewhatanalogous to multiple linear regression, in that the goal is to find which functions and associatedfrequencies explain the most variance in the dataset. For this analysis, the ”Spectrum” functionof the base R package is used. This function uses the Fast Fourier Transform (FFT) algorithmto perform calculations (RCoreTeam, 2017). Following recommendations for applying themethod, the spatial and temporal data series were detrended and scaled by their means prior to26analysis.3.4 Results3.4.1 Hydrological conditionsLong-term precipitation and streamflow data collected near the river mouth (Figure 2.1a) showlarge-scale variation year to year over the period of study. Annual flow conditions and precipi-tation are shown in Figure 3.2a and 3.2b. From 1972-2015, precipitation ranged from 1766 to4008 mm, and averaged 2917 mm/yr near the river mouth, with notably wet years in 1997 and1999 (4008 and 3849 mm, respectively), and dry years in 1978 and 1985 (2152 and 1766 mm,respectively). However, there is little apparent long term trend in total precipitation (Figure 3.2b). A notable rainfall event occurred on January 3 1984, when 190 mm of rain fell in 24 hours.This particular storm event triggered many slope failures in Carnation Creek and elsewhere onVancouver Island (for additional detail see Hartman and Scrivener, 1990).Streamflow conditions varied greatly year to year over the period of study. Instantaneouspeak discharge (Figure 3.2a) ranged from 11 to 64 m3s−1, with the highest flow occurringin 1984, and a period of generally elevated peak flows present from the late 1970s to mid1990s. From the late 1990s on, all instantaneous peak flows were below 40 m3s−1. Competentflow time, defined here as total time during which flows can entrain sediment (flows above6.7 m3s−1, see Haschenburger, 2011), ranges from less than 20 hours to nearly 300 hoursannually, with the greatest duration of flows in 1991 and 1997. Cumulative departures of hoursof competent flow (Figure 3.2a) shows inflection points near 1985 and again near 2005, wherecompetent flow conditions trend above, and below average respectively. Flow is capable ofentraining sediment for an average of 147 hours per year, while peak instantaneous annualflows averaged 28.7 m3s− Sediment supply and transport conditionsSediment is delivered to the channel from hillslopes in relative abundance (Figure 3.2c), withover 40 individual channel - coupled slope failures identified as having occurred during, orslightly before the period of record (Dhakal and Alila, 2003). Two regions coupled to hillslopesreceive the vast majority of sediment: first, the highly coupled canyon reach, and second, thechannel near the valley headwall upstream of 6.0 km (see Figure 2.1b). Of the field measuredslides for the 1996 inventory, the average length, width, and depth were, respectively, 48 m, 12m, and 1.2 m. Sediment additions from individual slides ranged in magnitude from small (<100 m3) to relatively large (> 2000 m3), but on average are less than 500 m3 and uncertaintyin volume is likely in the order of +/- 10% (see Hassan et al., 2016). Major sediment input27/Figure 3.2: (a) Total hours per year when flow is capable of entraining sediment (basedon values reported in Haschenburger, 2011), and instantaneous peak annual flows.Note that flow records for 1971 and 1972 are incomplete, and therefore omittedfrom the calculation. Similarly, no peak flow values were available for 1990, 2000,and 2015. Cumulative departures in competent hours, defined as the cumulativesum of value differences from a series mean, are plotted as the red dashed line. (b)Annual precipitation as measured near the river mouth, with cumulative departuresin precipitation values plotted. (c) Volume of sediment delivered through time fromhillslopes, distinguishing between total contributions (grey) and delivery to the con-fined canyon reach (red), 500 m above SA-8.28occurred in 1970, followed by a period of relative inactivity until the early 1980s, at whichpoint several large inputs occurred between 1981 and 1984. Over this four year period, morethan half of all sediment delivered to the channel was supplied to the canyon reach. Followingthis period of activity, little material entered the main stem until a major event in 1996, whenover 6000 m3 of sediment was delivered to the headwaters of the catchment. The succeeding20 year period experienced comparatively little sediment input.Net bank erosion rates (difference between area eroded and area stabilized) averaged overthe period of record are generally low in all study areas and similar to the measurement un-certainty, ranging from 0.09 m/yr of erosion in SA-3 to net stabilization of 0.06 m/yr in SA-7(Figure 3.3). These values also translate to small changes in net volumes eroded, with SA-3averaging 6.2 m3/yr of erosion, and SA-5 3.2 m3/yr of stabilized sediments. While net erosionand deposition is small, some study areas, such as SA-5, SA-7 and SA-8 experienced periodsof greater instability in the 1980s and 1990s, where annual erosion and stabilization rates ex-ceeded 1-2 m/yr. Cumulative net erosion (Figure 3.3) shows little trend in SA-2 to SA-4 andSA-9, with consistent net erosion in SA-8 during the 1980s followed by net stabilization to2015. SA-5, SA-6 and SA-7 all show little cumulative net change until the late 1990s, at whichpoint the channels all narrow for several consecutive years, indicating net stabilization.Six excavations over ten years (2006-2016) in the pool near the river mouth imply anannual minimum bedload transport rate ranging from 128 m3/yr to 642 m3/yr (unpublisheddata, Tschaplinski, 2016). DEM subtractions in some study areas corroborate these values, butseveral active years in SA-4 and SA-8, for example, indicate potentially higher local transportrates (> 1000 m3yr−1, see Figure 3.5). In contrast to stable bed systems with low supply, theratio of surface to subsurface median grain size in Carnation Creek is low (1.5) and most sizeclasses of bed sediments are often fully mobile at flows > 15 m3s−1 with the largest sedimentsize fractions mobilizing most years (Hassan et al., 2008).The spatial organization of sediment storage is thought to be related to particle travel dis-tances (e.g. Pyrce and Ashmore, 2003a,b). Information on travel distance can be estimatedin Carnation Creek from virtual velocities reported by Haschenburger (2011). Values of 2.15m/hr for all competent flows > 6.7 m3s−1 are reported from long-term tracer particle studies inthe catchment. Applying this value to the flow series, a distribution of annual travel distancesfor the period of record can be generated (Figure 3.1b). Mean annual travel distance is 301m, ranging from 40 to 625 m/yr. As these values are derived from sediment tracer particles,however, these may overestimate true transport estimates as much of the buried bed sedimentis not accessible to the streamflow.29Erosion/stabiliza�on (m)Figure 3.3: Erosion, stabilization, net change, and cumulative net change in the eightstudy areas. Two subsequent years of data are used for calculations: erosion (redline) is calculated as channel areas active in year 2 which were inactive in year 1;stabilization (blue line) defined as areas inactive in year 2 which were active in year1. Values are then divided by study area length to provide an erosion or stabilizationlength. Net change (black line) is the sum of erosion and stabilization.3.4.3 Organization of stored sediment and woodSediment in Carnation Creek is not stored uniformly along the lower 3 km of channel, orthrough time in the eight study areas. Figure 3.4 displays sediment stored in bars along thechannel from the 1991, 1999, 2009, and 2017 longitudinal profile surveys, while Table 3.2describes channel-averaged geometry and wood loading for each survey. Several pronouncedspikes in Sb are visible in all years of data. In 1991, the largest accumulations (near 50 m3m−1)are found upstream of 1300 m, with comparatively uniform bar storage downstream. Twolarge spikes in storage persist near 1300 and 2700 m for 1999, but a large spike near 1500 m isreduced in magnitude. Regions downstream of 1500 m have also organized into more clearly30Table 3.2: Characteristics of logjams and stored sediment along the lower 3.0 km of Car-nation CreekSurveyyearWb (m) Db (m) Storedsediment(m3m−1)# log-jamsMean jamvolume (m3)Jamvolume(m3m−1)1991 16.8 1.37 7.39 56 55.1 1.011999 17.6 0.99 6.49 44 83.8 1.312009 16.0 1.14 5.51 34 37.1 0.492017 13.4 1.08 5.27 44 19.6 0.33defined zones of high storage. For the 2009 survey, a major new accumulation has formed near2900 m in addition to numerous smaller accumulations downstream of 1500 m. The large spikein storage near 2500 m in 1999 has reduced in magnitude for 2009, as is the spike near 1300m. For 2017, most major accumulations have reduced in size or disappeared, with only threepeaks approaching or exceeding 20 m3m−1 of storage. Wood loads from the profile surveysare shown in Table 3.2. Overall, the number of logjams has fluctuated but wood volume hasdecreased from 1999 to 2017, and spacing increased. Scaled to channel length, wood volumehas decreased by 50% since 1991.Sediment and wood storage also varies substantially through time in the study areas. Stor-age varied through adjustment of either active channel area, or depth of alluvial sediment (seeFigure A.1), and these variables generally adjusted in tandem. Time series of wood volumeand stored sediment in the study areas are shown in Figure 3.5, and time-average values inTable 3.3. Sites can be grouped based on magnitude of variability in storage: SA-2, SA-3, andSA-9 show relatively low magnitude variation, while sites SA-4 to SA-8 show higher magni-tude variability. Overall, the greatest temporal variability in storage occurs in sites SA-4 andSA-8, while the lowest variability is in SA-2, SA-3, and SA-9. Sites SA-4 through SA-8 showpronounced peaks in storage between the mid 1980s and the early 2000s, with positive anoma-lies ranging from more than 2000 m3 in SA-4 and SA-8 to 900 m3 in SA-6 during this time.However, peaks generally occur earlier farther upstream, with maximum storage in SA-8 in1988 but occurring 11 years later for SA-4.In regards to wood loading, SA-3 and SA-9 show sharp increases in 1980 and 2012, respec-tively, which are a result of the addition of a few very large wood pieces from the riparian zone.Relatively large logjams formed in SA-5 and SA-8, in 1990 and 1982, respectively. In thesetwo sites, stored sediment and wood follow similar patterns, while SA-3, SA-4 and SA-9 showpatterns in storage independent from wood. SA-6 and SA-7 show persistently low wood load-ing, in spite of variable sediment storage. This is likely a product of the direct wood removalfrom this section of channel in the late 1970s, and continued lack of riparian wood supply.31Figure 3.4: Sediment stored in bars during the four longitudinal survey periods of 1991,1999, 2009, and 2017. Approximate study area locations are noted as verticaldashed lines.Table 3.3: Channel geometry, sediment storage, and wood characteristics for study areasof Carnation CreekStudyareaLength(m)Wb(m)Db(m)StoredSediment(m3m−2)WoodpiecesaWoodvol.b(m3m−2)SA-2 82 13.8 0.99 2.35 35 0.03SA-3 68 19.6 0.97 4.00 40 0.07SA-4 62 21.2 0.42 4.06 44 0.04SA-5 75 13.7 0.61 3.60 47 0.04SA-6 70 14.9 0.74 3.21 24 0.01SA-7 51 19.6 0.77 3.78 23 0.01SA-8 58 16.9 0.87 3.73 39 0.05SA-9 148 11.0 0.59 3.96 95 0.08aexcludes wood in logjams; average values over period of studybIncludes wood in logjams; average over period of study32Figure 3.5: Sediment storage and wood volume plotted through time for SA-2 throughSA-9. Storage has been normalized to the mean for each study area. Uncertaintyin wood volume is based on the standard error of the mean for wood pieces in sizeclasses (see Hassan et al., 2005a), and on estimated error for volumetric estimatesof wood in logjams.3.4.4 Spatial and temporal patterns of sediment storageAutocorrelation and spectrum analysis reveal regular patterns in the storage data. Figure 3.6aand 3.6b illustrates autocorrelation coefficients plotted against lag (timestep relative to time= 0), which provides information on length and time scales over which values of storage arecorrelated. For sediment stored along the channel main stem (Figure 3.6a), similar patternsemerge in all surveys, with rapidly declining positive correlation to between 80 and 120 meters.In all years, a significant negative correlation exists at distances of between 150 and 300 m,implying a cyclical nature to the data.33Figure 3.6: Corellellograms for (a) sediment stored along the channel profile, and (b) inthe study areas. Red dashed lines correspond to 95% confidence intervals, outside ofwhich correlations are considered significant. Periodograms for (c) sediment storedalong the main channel, and (d) sediment stored in study areas. Shaded areas in(c) correspond to persistent peaks. Values have been scaled by the series maximafor plotting, and were also smoothed for ease of interpretation by sampling andinterpolating the raw output. Data in (d) were not smoothed due to relatively smallsample size.34Autocorrelation patterns are also observed in the temporal sediment storage data from thestudy areas (Figure 3.6b). Patterns are broadly similar in five of the eight sites (SA-4 to SA-8), showing positive correlation in the order of seven or eight years. In these study areas,significant negative correlation emerges after 10-20 years, with the most pronounced negativecorrelation in SA-8. Sites SA-2, SA-3, and SA-9 have different patterns, with autocorrelationvalues of less than five years, and no significant negative correlation.Figure 3.6c shows smoothed periodograms for sediment storage along the channel main-stem, while Figure 3.6d illustrates periodograms for the time series storage data in the studyareas. The lowest possible wavelength, or frequency (Nyquist frequency, see Chatfield, 2003)detectable given the data resolution is approximately 1/30 m, while the highest is 1/3000 m.At least two prominent spikes in spectral density appear in Figure 3.6c. The obvious dominantpeak in all years is near a frequency of 0.004 m−1, which corresponds to a wavelength of 250m, while a smaller peak is apparent at a frequency of 0.12 m−1 (wavelength of 80 m). As afraction of the total variance, the lower frequency spectral density is much higher, indicating itsgreater relative importance in explaining spatial organization of storage. For 1991 and 2009,additional peaks are found near frequency 0.009 m−1, and 0.007 m−1, respectively.For the temporal study area storage data (Figure 3.6d), spectral density results can begrouped broadly into two categories. Study areas SA-3 and SA-9 show spikes in spectraldensity at frequencies of 0.07 and 0.12 hz, (14 and eight years) respectively, while SA-2 andSA-4 show some evidence of maxima near 0.05 hz (20 years). All other sites reach spectraldensity maxima at a wavelength equivalent to the length of the data record, indicating either alack of detectable periodicity to the series, or that the series is not long enough to capture anyperiodicity. The similar separation of patterns by study area in both Figures 3.6b and d wouldsuggest that record length is more likely to explain the lack of detectable periodicity in Figures3.6d.3.4.5 Storage and sediment transfer rateTemporal storage dynamics and sediment transfer rate (Pryor et al., 2011) can be further ex-amined by plotting the quantity of sediment stored in a study area against change in storagefrom the previous year (see Hassan et al., 2008). Figure 3.7 displays ∆S (sediment transfer)vs. S (total storage) in each of the study areas, while Figure 3.8 shows the distribution of cycleperiods for the entire dataset. A positive value in Figure 3.7 indicates a net increase in storageof sediment year to year, while a negative value represents a decrease. Additional detail ofcycles can be found in Figure A.2. With Carnation Creek data, cycle duration was defined asthe time interval over which the channel state has switched from aggradation to degradation,and then back to aggradation (see Figure 3.7b for a visual description of the definition). Over-all, complex relations between ∆S and S are found in all sites. As with the results previously35described, SA-2, SA-3, and SA-9 display different behaviour than SA-4 to SA-8, with smallannual changes in storage, and aggradation-degradation cycles of similar durations and mag-nitudes. In comparison, study areas SA-4 and SA-8 have larger cycles and greater change instorage over short periods of time, with large-scale clockwise hysteresis evident. While smallercycles are present in these sites, they are superimposed on larger cycles of duration > 15 years.The most prominent cycle is found in SA-8, which lasts nearly the full period of record. SitesSA-5 to SA-7 have some cycles of intermediate duration but most cycles are comparativelyshort, between five and ten years. When all cycles are combined (Figure 3.8), most cycles arefound to be short, and median cycle duration is approximately of six years.The larger scale aggradation/degradation cycles can be further explored by assessing thecumulative departures in storage for each study area. Figure 3.9a displays cumulative depar-tures for all study areas overlaid, while 3.9b shows collapsed and smoothed storage. Positiveslopes in Figure 3.9a indicate a period of consistently above-average storage, while negativeslope indicate consistently below-average storage. By tracking the broad trajectories of storagedepartures, larger-scale patterns of aggradation followed by degradation are apparent in studyareas SA-4 through SA-8, but not in SA-2, SA-3, or SA-9, consistent with other observations.Periods of consistently elevated storage range from the early 1980s to the late 1990s in SA-6 to SA-8, and from 1990 to 1999 in SA-5. The period of positive anomalies in storage islater in SA-4, occurring from 1990 to 2005. The magnitude of peak departures also decreasesdownstream, with the exception of SA-5.3.5 DiscussionSpatial and temporal variability in sediment storageThe lower three kilometers of Carnation Creek are in the process of adjustment to a sedimentadditive disturbance. The higher magnitude variation in storage, patterns in cumulative depar-tures, and similar broad autocorrelation patterns in SA-4 to SA-8 indicate a response in storageto a major pulse of sediment moving through the channel, supplied from debris flows in thecanyon reach in the early to mid 1980s. Collectively, these debris flows delivered over 5000m3 of material, equivalent to between 10 and 40 times the annual load as measured at the rivermouth.The timing of the SA-8 storage increase corresponds well with a response time (3-5 yrs)calculated from reported virtual velocities from Haschenburger (2011). Additionally, the pro-gressively delayed peaks in storage downstream (to SA-4, see Figure 3.9b) provides evidenceof downstream propagation of the hillslope-supplied sediment. Considering the timing of peakstorage in the downstream direction, the added sediment is displaced at a rate of 122 m/yr, abouthalf the average annual travel distance (Figure 3.1b). Similar patterns have been observed by36Figure 3.7: Stored sediment vs. annual change in storage (sediment transfer) for studyareas SA-2 to SA-9. (a) Expanded panels of SA-7 and SA-8 illustrate three and twoexample cycles respectively, with arrows indicating direction of hysteresis. Notethat axis values for the enlarged SA-7 plot are different from other subplots. Thecolour of datapoints on all plots corresponds to date. (b) plot of sediment transferthrough time in SA-8 as example of how cycles are defined and identified. Colouredlines correspond to cycles shown in expanded SA-8 plot in (a). (c) plots of storagevs. sediment transfer for study areas SA-2 to SA-6, and SA-9. Expanded versionsof all subplots in (c) can be found in Figure A.237Duration (yrs)Count5 10 15 20051015Figure 3.8: Distribution of durations of small-scale aggradation/degradation cycles forsediment stored in all study areas combined. Thirty-nine individual cycles wereidentified.Figure 3.9: (a) Cumulative departures of storage in the study areas, (b) storage in sectionsscaled to study area length. Dashed lines correspond to study areas with logjams.Cumulative departures are calculated as the difference between a given value and theseries mean, summed through time. Values in panel (b) have been smoothed witha Loess function of span 0.3 for plotting purposes, and can be seen individually inFigure 3.5.38others, such as a downstream propagating wave by Madej and Ozaki (1996) in Cuneo Creek,California, and a more dispersive wave in the Navaro River, California, by Sutherland et al.(2002).The sediment pulse observed in Carnation Creek is not apparent in sites SA-2 and SA-3,while SA-9 is upstream of the pulse and is thus not affected. SA-3 shows only weak evidence ofa storage peak in 2005-2008, while SA-2 and SA-9 show no evidence of this pattern, both withminor storage maxima in 1982. In SA-2 and SA-3, the current rate of pulse movement indicatesthat material may have either fully dispersed or may not yet have reached the sites. While SA-9is located downstream of several debris flows delivering sediment in the early 1980s and mid1990s, no clear response in storage is apparent. SA-9 may be far enough (as with SA-2 and SA-8) from the sediment source that pulses may have fully dispersed, or much of the sediment hasyet to reach the depositional area where SA-9 is located. These results collectively suggest thatthe timescale of adjustment to a major pulse of sediment can be quite long (> 30 years in thissystem) in close proximity to the input location, but will vary substantially over short distances.The specific timescale of adjustment will depend on the magnitude of the sediment deliveryevent, the grain size distribution of the delivered material, the frequency of slope failures, andthe transport capacity of the system (e.g. Muller and Hassan, 2018).Temporal patterns in sediment storage are complex but vary consistently in a downstreamdirection. Spectral analysis methods only detected meaningful frequencies for SA-3 and SA-9,two sites with storage unaffected by the debris flow sediment pulse. Given the tendency formost study areas to have the largest periodigram value at the lowest frequency, it is likely thatthe series in other sites were not long enough to capture the adjustments in storage. However,in SA-3 and SA-9, the dominant periodicity at 14 and eight years, respectively, falls within therange of shorter-term aggradation/degradation cycles (Figures 3.7 and 3.8). These shorter termcycles are present in all study areas, effectively superimposed on top of longer-period signals(Figure 3.7). However, variance is dominated by low frequency periodicity imposed by thelarge scale increase and decrease in storage in SA-4 to SA-8, thereby muting their influencein the periodogram. The direct explanation for these shorter term aggradation/degradationcycles is likely tied to changes in internal supply related to logjam degradation and erosion ofsediment wedges, or changes to active channel area. Bank erosion and stabilization patterns(Figure 3.3) agree with these values, with fluctuations in the order of 5-10 years in most studyareas.Our results provide strong evidence of non-random spatial organization of sediment stor-age, with at least two persistent wavelengths of 250-330 and 70-100 m. The greater wavelengthcorresponds closely to that of the major bars in the system (see Figures A.3 and A.4), where thechannel is wider than average, and bar elevations are higher. A plausible explanation for theparticular wavelength is related to the average annual particle travel distances (Figure 3.1b),39which have a mean of close to 300 m. While travel distance here is calculated from flow data,background virtual velocity relationships (Haschenburger, 2011) implicitly consider the com-bined effect of other variables such as particle size distribution and channel morphology. Whilethe major periodogram peaks appear closely related to particle travel distances, the smallerpeak at a wavelength of approximately 80 m is likely related to the spacing of geomorphicallyeffective logjams in the system (Table 4.2).Wood loading in Carnation Creek may also explain some of the spatial and temporal vari-ability in storage. Including non-jam wood pieces, total wood loading in Carnation rangesfrom 368 to 880 m3ha−1 between 1991 and 2017, with the highest value in 1999 and lowest in2017. In spite of the direct removals and lack of robust riparian supply, these values effectivelyencompass the upper half of those reported in Wohl and Scott (2017) (see their Table 1) andare even above some other values found in the Pacific Northwest (e.g. Hogan, 1986; Nakamuraand Swanson, 1993). In other forested mountain streams (e.g. Hogan et al., 1998; Gomi et al.,2004; Wohl and Scott, 2017), logjams are associated with zones of elevated sediment storage.While logjams do not appear to be the dominant factor explaining the spatial organization ofsediment storage in Carnation Creek, the clear association between storage and logjams in SA-5 and SA-8 lends strength to the argument that logjams are at least locally significant agents ofsediment storage in forested systems, increasing sediment trapping efficiency along channels(see Eaton et al., 2012) and delaying release of material. Wood loads have generally reducedthrough time since the early 1990s (see table 3.3 and Figure 3.5), a product of the ripariandisturbance and direct removals of in-stream wood in the late 1970s between SA-6 and SA-8. This reduction in wood loads likely corresponds to a reduced capacity to affect sedimentstorage, implying that undisturbed systems could see a greater wood-storage association. Theincreased trapping efficiency provided by historically large, young logjams could also havereduced sediment mobility and led to lower annual travel distances, and potentially differentspatial organizations of sediment storage.3.5.1 Storage and sediment transferEvaluating changes in sediment storage through time confirms that movement of materialthrough the system is not simply a function of flow conditions, but also one of variable sedi-ment supply. While ∆S is not an equivalent proxy for transport rate, it is reasonable to expect acorrelation between the magnitude of geomorphic change and a measure of a stream’s capac-ity to do work. In Carnation Creek, some of the greatest scour and fill events occurred duringbelow-average flow years in several study areas (see Figures 3.2 and 3.5), while some high flowyears resulted in little change in storage. Although many of the largest year-over-year storagechanges in Carnation Creek do occur when storage is highest, the overall relation is complex.Our results do not show a simple connection between storage quantity and the rate of stor-40age loss even in degrading study areas (see lower half of all panels in Figure 3.7). Previouswork (e.g. Lisle and Church, 2002; Hassan et al., 2008) has suggested that sediment trans-fer and transport rate may be either linearly or non linearly, but positively correlated to thequantity of sediment stored in a system. The comparatively simple relation between sedimenttransfer and storage in Lisle and Church (2002) and Pryor et al. (2011) may therefore not befully representative of complex natural systems with glacial histories and wood loading. Flumeexperiments are not able to capture many of the elements of natural channels, and the relativelyshort and infrequent field data from Pryor et al. (2011) may miss or mask smaller-scale vari-ability in the system. While the large-scale patterns shown in cumulative departures of storage(Figure 3.9a) provide compelling evidence of the role of external supply on regulating stor-age, explanations for the smaller scale complexity of the storage/transfer relation is likely tiedto local conditions. These conditions include the degree of channel bed armoring, presenceand type of bed surface structures, and antecedent sediment texture, in addition to the particlesize distribution of supply (Brummer and Montgomery, 2006; Hassan et al., 2008). These fea-tures, in combination with wood loading, may render a deposit more difficult to erode. Woodmay play a large role in regulating the storage and transfer of sediment; In SA-5 and SA-8,the logjams that formed over the period of study trapped volumes of sediment > 1000 m3.The trapping efficiency and stabilizing effect of the logjams may have temporarily disruptedthe sediment pulse, resulting in comparatively small but temporary year-to-year changes instorage. Though the remaining study areas did not have such dynamic wood loads, smallerchanges in wood volume or configuration could have resulted in a temporary stabilizing effecton smaller sediment deposits.3.5.2 Landscape organization and storageCarnation Creek’s glacial history has direct bearing on the complexity of relationships betweensediment export and storage, and on sediment storage and redistribution through the channelnetwork. The repeating sequence of coupling-decoupling observed along the channel is aptto partly dictate the spatial organization of sediment storage: steep, confined reaches maycontain storage in colluvial fans but large bars are unlikely given lateral confinement. Materialadded to these reaches from hillslopes is thus likely to rapidly travel downstream to lowergradient reaches, where it is reworked by fluvial processes. Therefore, the effect of supply onstorage and other channel response will not necessarily be observed in close proximity to theexternal supply location. In Carnation Creek, the similar responses in SA-9 relative to SA-2and SA-3, both far from direct hillslope input, may illustrate this effect. Based on the resultsof sediment storage analysis, it appears that sediment propagates with at least some dispersivecharacteristics (see Sutherland et al., 2002), with greater magnitude of change in storage nearerto the hillslope sediment supply. This implies that the glacially imposed spatial organization41of hillslope-channel coupling can indirectly, but critically influence decadal-scale adjustmentin storage at different points in the channel.Similar connections between landscape history, hillslope-channel coupling, supply dynam-ics and transport conditions have been made for several catchments in regions outside the Pa-cific Northwest. For example, Beylich and Laute (2015) provide sediment budgets for twodeglaciated catchments in Norway, illustrating the role of profile shape and valley form onsediment delivery and storage, and finding transport rate to be more closely tied to supply thanhydrological conditions. Similarly, Heckmann and Schwanghart (2013) evaluate hillslope-channel connectivity in the central Alps, highlighting that while hillslope material is a dominantinput in the system, longitudinal sediment connectivity of the channel main stem is disruptedby the glacial legacy. In an alpine catchment with an actively retreating glacier, Lane et al.(2017) also link bedload transport rates to sediment supplied from hillslopes destabilized bythe retreating ice.While the studies listed above focus on catchments more recently deglaciated, CarnationCreek is experiencing late-stage paraglacial conditions (Church and Ryder, 1972; Ballantyne,2002), a phenomenon which is globally widespread (e.g. Mercier, 2008) and affects the sedi-mentological connectivity of the landscape (Wohl et al., 2019). Deep glacial drift material overmuch of south-west British Columbia is largely stabilized with vegetation, but disturbancesfrom major storms or anthropogenic activity result in re-activation of this material in the formof episodic supply events, leading to a type of paraglacial adjustment such as that outlinedby Ballantyne (2002). While Carnation Creek findings represent conditions in a single catch-ment, glacially-affected landscapes elsewhere are likely to produce broadly similar patterns ofhillslope sediment delivery and in-stream storage, should glacial drift material be episodicallysupplied to channels.3.5.3 Conceptual response of storage to supplyWe propose a conceptual framework to describe and contrast the adjustment of sediment stor-age and storage-transfer patterns in reaches coupled and decoupled to hillslopes. The effectof variable hillslope-channel coupling is illustrated in Figure 3.10, with two theoretical catch-ments and associated profiles: (a) a catchment with channel-hillslope coupling limited to head-waters (Figure 3.10a), and (b) a deglaciated catchment with a hanging valley and multiplecoupling zones along the channel length (Figure 3.10b). This second catchment contains char-acteristics found widely throughout deglaciated regions (e.g. Beylich and Laute, 2015; Hassanet al., 2019).Along channel reaches experiencing direct, or nearly direct contributions of hillslope sed-iment, high-magnitude variation in storage is likely as a pulse of material moves through thesystem. Past research suggests that at least some dispersive characteristics are usually present42StorageStorageDistance DistanceCoupling  Coupling CouplingDecoupling Decoupling Decoupling(a) Upland coupling (b) Variable downstream coupling: stepped long profileLogjam LogjamZ, dS/dtΔSZ, dS/dtΔSZ dS/dtZdS/dtSediment storage-transfer cycleFigure 3.10: Conceptual model of adjustment in storage and storage-export cycles atdifferent positions in (a) unglaciated catchments with simple concave-up profilesand (b) deglaciated catchment with stepped longitudinal profile. In (a), hillslope-channel coupling (shown as grey bands) and associated sediment delivery is lim-ited to headwaters but can occur again downstream in (b), reflected as grey bandsin both panels. The state of channel coupling and episodic nature of sediment de-livery affects the timescale and magnitude of storage adjustment, shown as the reddotted line. This in turn is tied to the magnitude of aggradation/degradation cyclesin storage, with larger-scale cycles found in coupled regions, and smaller cycles indecoupled regions dominated by local depositional and erosive processes. Smallcycles can be superimposed upon larger ones, and are also influenced by the pres-ence of logjams, which are likely to increase storage a pulse of sediment moving downstream in a mountain channel (Sutherland et al., 2002;Pryor et al., 2011). In this case, the magnitude of storage variation with time will also de-crease in a downstream direction, away from the sediment source. The specific timescale andmagnitude of adjustment will depend on sediment texture and factors influencing sediment mo-bility, such as logjams. A greater abundance of wood, or other mobility-reducing factors willgive a sediment pulse more dispersive, rather than translating tendencies as material is trappedmore readily along the channel.Storage-transfer patterns will differ between coupled and decoupled reaches in tandemwith storage. In coupled reaches, both large and small scale cycles are likely: larger scalepatterns correspond to the movement of a sediment pulse through the channel, and smallercycles correspond to local variation related to bar migration and bank erosion. In decoupled43reaches, these large-scale patterns are muted or absent, with cycles limited to a smaller scale intime and magnitude. In the case where a channel is heavily affected by wood loading, such aswith a channel-spanning logjam, sediment moving downstream is likely to be more efficientlytrapped and retained, released only when the jam breaks up or if the channel avulses. Thesedynamics result in comparatively elevated and persistent storage in such reaches, and amplifiedstorage-transfer cycles.The studied channel region of Carnation Creek falls mainly between the two categoriesoutlined here, and therefore illustrates intermediate responses and variability in storage. How-ever, SA-2, SA-3 and SA-9 all show signs of patterns present in decoupled reaches, while SA-8shows a logjam-modified response in a section of channel immediately downstream of a cou-pled reach. Remaining study areas show declining coupling effect, with local scale processesincreasingly dominant in a downstream direction. Collectively, data in the seven study areasdownstream of the canyon reach indicate that the conceptual model proposed above can beuseful for broadly capturing dominant scales of storage variability in complex catchments withhillslope-channel coupling.3.6 ConclusionsIn this study, spatial and temporal scales of sediment storage in the active stream channel areexamined in the context of wood loading and external sediment supply over a 45-year period inCarnation Creek, a small, forested mountain stream in coastal British Columbia. Collectively,our findings highlight the importance and complexity of storage for regulating supply, and alsoassociations between storage conditions and landscape organization imposed by glaciation.Analysis of spatial and temporal sediment storage data using autocorrelation and spectralanalysis reveals characteristic length and time scales. Storage is correlated through time inmost study areas, but this correlation timescale varies based on position relative to sedimentsources. High-storage areas are found to display periodic behaviour over 3.0 km of surveyedchannel, with dominant wavelengths of approximately 250-300 m, but this behavior is notas apparent in the temporal data. Average annual sediment travel distances are similar to thislength. Logjams and other wood structures may serve to store and regulate sediment locally butdo not correspond to the annual travel distance of sediment in the catchment. Temporal vari-ability in the study areas, manifest as aggradation/degradation cycles, displays similar patternsto wood loads in some study areas, but may also be a product of hydrological conditions.We propose a conceptual model which links landscape organization, variability in stor-age response and storage-export cycles in mountain channels. Findings from this study rein-force our understanding of how supply conditions and in-stream wood dynamics serve to shapemountain streams over a multi-decade timescale.44Chapter 4Long-term characteristics,morphological impacts, and simulatedbudgets of in-stream large wood4.1 SummaryIn-stream large wood (LW) is a prominent feature of forested aquatic ecosystems worldwide,yet many questions remain regarding temporal patterns of LW and LW influence on channelmorphology, particularly in systems affected by riparian timber harvesting. In this chapter,spatial and temporal patterns of LW are examined over the long term and used to provideinformation on piece persistence, mobility, and interactions with channel morphology. A woodbudget model is developed to assess the long-term impacts of riparian harvesting on woodstorage in streams, including timing of minimum storage and recovery timescales.The 45-year record of wood piece characteristics in Carnation Creek is used in this chapter.Wood pieces and logjams are distributed unevenly along the channel, a distribution which isshaped in part by the riparian logging history of the catchment. Similarly, patterns of woodand sediment storage through time reflect this history, and provide detailed information onlogjam-channel morphology evolution in the system. Analysis of the formation and the decayof two jams reveals an avulsion style of channel adjustment and major sediment accumulationupstream which persists for up to 20 years in the active channel, and likely longer in the flood-plain. Topographic complexity of the channel bed is found to be positively related to woodpiece abundance. Overall, wood is an important geomorphic agent in the channel over time.Results from the wood budget model indicate that impacts from timber harvesting are pro-jected to continue for at least 150 years after harvesting as riparian forests recover. Losses ofwood from harvesting also correspond to a reduction in sediment storage along the channel45over a 26 year period. Minimum modeled wood loads are projected to occur 50 to 80 yearspost-logging, depending on the influence of wood transport and proximity to harvesting. Col-lectively, these findings indicate that wood is acting as an important geomorphic agent throughtime, and that riparian logging will lead to century-scale changes to wood loads with implica-tions for channel morphology and aquatic habitat.4.2 IntroductionLarge wood (LW) is a key structural and functional component of forested streams worldwide,altering physical channel form and serving as a critical component of river ecosystems (Ceder-holm et al., 1989; Abbe and Montgomery, 1996; Gregory et al., 2003; Gurnell, 2013). LWpieces and logjams influence channel morphology, geometry and sediment throughput (Hogan,1989; Montgomery et al., 2003; Haschenburger and Rice, 2004; Wohl and Scott, 2017) throughchanges to flow hydraulics at the channel-unit (Wohl et al., 1997; Buffington et al., 2002) tothe reach scale and above (Buffington and Montgomery, 1999). By modulating channel mor-phodynamics and providing cover and shelter from high flow velocities, LW adds diversity toaquatic habitat for a variety of organisms and is thus of high ecological significance (Bjornnand Reiser, 1991; Abbe and Montgomery, 1996; Gurnell, 2013; Wipfli and Richardson, 2015).The value of wood in river systems is now well recognized, but prior to the latter third ofthe 20th century, LW was often viewed with indifference or through a negative lens in westernNorth America (Richardson et al., 2012; Wohl et al., 2019). As a result, little consideration waspaid to impacts on channel form and aquatic habitat stemming from loss of wood supply dueto harvesting on hillslopes and in the riparian zone. Collectively, these hillslope and riparianwood sources often constitute the vast majority of wood input to stream channels (Lienkaemperand Swanson, 1987; Murphy and Kosksi, 1989; Benda et al., 2002; Rigon et al., 2012; Bendaand Bigelow, 2014; Hassan et al., 2016). Several studies have documented harvesting-relatedlosses of wood with associated impacts to channel morphology and sediment storage, thoughresponses are sometimes complex: reductions in wood loads were often (Bilby and Ward, 1991;McHenry et al., 1998; Nowakowski and Wohl, 2008; Benda and Bigelow, 2014) but not always(Carlson et al., 1990) observed, while reductions in pool areas and sediment retention werenoted (Bilby and Ward, 1991). While a few studies have projected wood-load response (Me-leason et al., 2003) and recovery times (Murphy and Kosksi, 1989; Bragg, 2000; Stout et al.,2018) from riparian wood loss, model results are rarely compared with field data. Given thatwood is supplied to rivers from both proximal (Gregory et al., 2003) and distal (upstream andhillslope) sources (Hassan et al., 2016), the long-term impacts of landscape-scale reductionsto wood supply are of key importance to river system dynamic and therefore warrant a morethorough examination.Previous field studies have examined many aspects of LW-channel form interactions, but46more often with a focus on spatial (e.g. Hogan et al., 1998; Montgomery et al., 2003; Hassanet al., 2005b; Andreoli et al., 2007; Jackson and Wohl, 2015) rather than temporal characteris-tics (but see McHenry et al., 1998; Wohl and Goode, 2008; Iroume et al., 2014). A commonlimitation with these studies is that field datasets usually span a short timeframe relative to im-portant but often episodic river processes and disturbance events, such as fires, debris flows, ortimber harvesting. In particular, long-term concurrent data of logjam and channel morphologyevolution are rare (Eaton et al., 2012; Ruiz-Villanueva et al., 2016), and few continuous dataare available to examine wood load and channel response to logging-related reductions in woodsupply (Nowakowski and Wohl, 2008). Longer-term data on the spatial and temporal variabil-ity of in-stream wood are therefore needed in order to capture and characterise the evolutionof processes linking riparian disturbance history to LW characteristics, and in turn to channelmorphology. In summary, the length of field data records has been an outstanding limitation inprevious LW research (Ruiz-Villanueva et al., 2016).The primary objectives of this chapter are therefore to (i) describe the characteristics ofin-stream wood with particular focus to spatial and temporal patterns in wood storage over thelong term; (ii) examine these characteristics and patterns as they pertain to LW functionality asa morphological agent, especially regarding sediment storage by logjams; and (iii) use a woodbudget model to evaluate LW storage changes, the time to minimum wood load, and wood loadrecovery times as a result of riparian timber harvesting and forest regeneration. These objec-tives are synthesized by testing two related hypotheses: (1) that there is a positive relationshipbetween wood abundance and channel morphology in terms of (a) sediment storage and (b)the overall topographic complexity of the channel bed surface; and (2) that changes in woodabundance from timber harvesting will be reflected in both complexity and storage.This chapter aims to capitalize on the long-term wood piece dataset available from Carna-tion Creek in order to address these objectives and test the proposed hypotheses. This datasetis uniquely valuable for investigating long-term, post-harvest dynamics of LW and channelmorphology in tandem, given the complementary spatial and temporal resolution.4.3 Methods and data4.3.1 Wood characterization and volumetric estimatesInformation on in-stream wood was gathered in the study areas during topographic surveys(1971-2015) and along the lower 3 km of channel during the four long profile surveys (seeFigure 2.4). Within each 15 m profile survey interval, individual wood pieces were assignedinto size classes based on length and diameter, and identified as either part of a jam or asindividual non-jam pieces. Details of the classification and measurement strategy used forjams and wood pieces can be seen in Hogan (1986) and Hassan et al. (2016). Volumes of47wood were then calculated from the median characteristic size of pieces corresponding to eachclass.In the study areas, wood piece positions were mapped and dimensions sketched to scaleannually until 1998 (see Figure 4.1), but pieces were not measured directly. Beginning in 2001,individual wood piece lengths and diameters were measured each year with a fiber surveytape to the nearest centimeter, allowing for quantification of relatively precise wood volumes.Between 1989 and 1998, wood maps were not available for some study areas, resulting in arecord gap length of 2 to 11 years, depending on the site. To obtain volumetric wood estimatesfrom the pre-2001 study area maps, each map was georeferenced using visible benchmarks(see Figure 4.1). Wood pieces on each map were then classed based on length using a modifiedversion of the relative size criteria outlined in Hassan et al. (2005b), whereby classes aredefined in relation to the ratio of piece length (L) to bankfull channel width (Wb). Here, threelength classes are used, defined as L/Wb < 0.3 (Class 1), 0.3 < L/Wb < 1.0 (Class 2), andL/Wb < 1.5 (Class 3). To convert the piece count to volume and provide an error estimate, afull inventory of pieces measured from survey year 2007 (length and end diameters, n = 419)was compiled and segmented on the same class criteria as outlined above. For each piece sizeclass, a median characteristic volume was calculated and applied to all identified wood pieces,thus providing a wood volume estimate for each year, in each study area. The year 2007 wasselected for estimation of size class characteristics due to high survey data quality, when a newpiece tagging approach was employed.In two study areas (SA-5 and SA-8) large logjams formed in the late and early 1980s,respectively (see Figure 4.1 b for SA-8 example). As individual wood pieces in jams werenot always possible to identify, jam outlines were digitized from maps to obtain a measure ofeach jam’s two-dimensional footprint. To obtain wood volume from this footprint, ground-level and overhead photographs and field notes were used to obtain estimates of jam height.As logjams are not solid objects, a porosity factor must be applied to estimate void space.Relatively few published data are available which describe logjam porosity; however, Liverset al. (2015) noted values ranging from 0.28 to 0.54 in jams located in several North Americanlocations, while Dixon (2016) noted a range of values from 0.16 to 0.63 in a watershed insouthern England. A porosity value of 0.4 was selected based on these published values andinspection of the imagery described above. It is worth noting, however, that porosity is likelyto change through time if the jam becomes compressed, or wood decays within the jam.The orientation of wood pieces is thought to be of importance regarding their stabilityand probability of transport (Eaton et al., 2012). Piece orientations were estimated from the2007 survey year database. Using the wood piece maps generated for this year, linear thalwegsegments were drawn along the center of the channel. A line parallel to the long axis of eachwood piece was then drawn to intersect the thalweg segment, and the angle relative to the48(a)(b)Figure 4.1: Example maps derived from annual survey data for (a) SA-3, and (b) SA-8for the 1982 survey year. Note the presence of the logjam in (b) during early stagesof formation. Following georectification using visible survey benchmarks, woodpiece counts and jam volume estimates were derived from these maps and ground-level imagery. Note imperial units of length and flow rate on figure: all values wereconverted to SI units.49thalweg estimated as falling into one of eight bins of 22.5◦. For example, a wood piece nearlyparallel to the thalweg would fall into the bin of 67.5◦ to 90◦, or 90◦ to 112.5◦. Orientationsdetermined for each piece were then matched to the other wood piece attributes using theunique piece IDs. While this approach was effective and reliable in most study areas, the largenumber of pieces in SA-9 rendered labels difficult to associate with corresponding pieces onthe maps; therefore, SA-9 data were excluded from the orientation estimates.4.3.2 Wood budget modelTo assess the magnitude of wood loss and the response timescales to catchment and riparianharvesting, a reach-scale wood budget model was developed. The budget spans a period of300 years, from 1900 to 2200. Similar budgets have been used in other studies to evaluatethe relative contributions of wood sources, wood transport, and change in loading over time(Martin and Benda, 2001; Benda and Sias, 2003; King et al., 2013; Hassan et al., 2016). Here,a budget model similar to that used by Hassan et al. (2016) is applied. At the reach scale, ageneralized wood budget can be represented as:∆S =[I∆x−L∆x+(It −Ot)−Od]∆t (4.1)where ∆S is the change in wood storage, I∆x input over reach length x, L∆x loss of wooddue to overbank losses or channel migration over length x, It wood transported into the reach,Ot wood transported out of the reach, Od in-situ decay, and ∆t the time period of interest. Fora given reach, total inputs (I) are characterized as:I = Ibe+ Im+ It (4.2)where Ibe is input from bank erosion and Im input from mortality within the riparian zone.In mountain catchments, wood delivered from hillslopes through debris flows and landslidescan be a dominant source (e.g. Nakamura and Swanson, 1993; May and Gresswell, 2003; Rigonet al., 2012). While input from hillslopes likely occurs in some sections of the watershed, directinput does not occur along the main stem section of interest, which is buffered from hillslopesby a narrow floodplain. The assumption is thus made that colluvial wood input is a negligiblebudget term in this region of the channel. Similarly, other potential wood sources deemedinsignificant include input from disturbances such as fire or insects, anthropogenic wood input,and exhumation and burial of wood pieces. A full description of common budget terms is givenin Benda et al. 2002. Wood outputs from a reach (O) are described here as:O = Oca+Ot +Od (4.3)50where Oca is loss due to channel migration and abandonment and Od losses from wooddecay.For the two reaches of Carnation Creek, Ibe is calculated as the sum of two components:Ibe = IbeS+ IbeCWD (4.4)where IbeS is input from standing, live timber, and IbeCWD input from coarse wood debris(CWD) on the channel banks. IbeS is calculated as:IbeS = ∆xBVs (4.5)and IbeCWD similarly as:IbeCWD = ∆xBVCWD (4.6)Where ∆x is reach length, B bank erosion rate (m/yr), and Vs and VCWD the volume perhectare of standing timber and CWD, respectively. Annual erosion rate B is derived fromdelineation of active channel areas which are then compared year to year to identify erodedarea (see Chapter 3 for more details). As material eroded in a particular year may have beendeposited only a few years before, a 10 year window was used for the calculation of erosionrate, whereby any material eroded in one year that was deposited in the previous ten yearswas assumed to contain no woody vegetation with stems of diameter sufficient to meet theminimum LW criteria of 0.1 m (e.g. Hogan et al 1998). Published growth curves for Redalder, one of the early-stage successional species in the region, were used to define the 10 yearwindow (Nigh and Courtin, 1998).Information on live tree stem density and volume for pre-harvest conditions was obtainedfrom the BC Vegetation Resource Inventory database (BCVRI), which provides a synthesisof many forest cover attributes at the sub-catchment scale or below (BCVRI, 2011). As theBCVRI database was not compiled until long after harvesting took place in the lower sectionof Carnation Creek (2005 vs. late 1970s), average values from nearby unlogged sections wereused as a proxy for pre-harvest stand characteristics.To obtain values of Vs and VCWD for post-harvest conditions, the Table Interpolation Pro-gram for Stand Yields (TIPSY) model was used. This model, first developed by the BritishColumbia Ministry of Forests in 1991, can be applied to forecast stand growth for timber har-vesting purposes. This model requires inputs describing site characteristics (average slope,biogeoclimatic zone, and site productivity index, for example), in addition to stand and re-planting characteristics (planted stem density, cutblock treatment, species distribution). Fulldetails of the original TIPSY model can be found in Mitchell et al. (1992). Most TIPSY inputvariables can be obtained from information found in the BCVRI database. Tree species compo-51sition, post-harvest stand treatment, and some site characteristics were obtained from Hartmanand Scrivener (1990) and Tschaplinski and Pike (2017). TIPSY model output consists of timeseries of stand volume estimates (m3/ha) subdivided into stem size classes. Similar output forVCWD is also generated.Input from mortality (Im) will result if a living tree dies and then falls into the streamchannel. Factors that dictate the quantity of LW contributed from mortality include the lengthof the stem relative to channel width, the position of the tree relative to the stream bank, theprobability of the tree falling towards the stream, and the probability of a tree dying over agiven time interval. Collectively, Im can be formulated after Van Sickle and Gregory (1990)and Hassan et al. (2016) as:Im =∑Vs∆Z j∆xPf E(Vb) (4.7)where Z j is the distance from a tree to the streambank, and Pf is the probability of treefall over the interval ti to ti+1. E(Vb), the expected volume of a tree entering a stream from asingle fall event, is calculated as:E(Vb) =∫ 180−asasVb f (a)da (4.8)where f (a) is the probability of a tree falling relative to the banks at angle a. Equation4.8 is integrated over the limits of 180−as to as representing the possible arc of a tree fallingtowards a stream channel.Previous research (e.g. Benda and Bigelow, 2014; Hassan et al., 2016) indicates that smallchannels usually experience limited wood transport. To incorporate this term into the budgetmodel, the method following Eaton et al. (2012) is applied. This method uses an estimate fortransport based on the L/Wb ratio and piece orientation, using assumed probability distribu-tions. For a given timestep, wood transported into or out of a reach (It and Ot , respectively)can both be calculated as:It = SwPmoveVc (4.9)where Sw is the travel distance over a given time interval, Pmove the probability of piecemovement, and Vc the volume of wood per unit channel area. Using empirical relations devel-oped from data in Mack Creek, Oregon (Gurnell et al., 2002), Sw in channel width equivalentsis calculated as:Sw = 10 · e−3.8(Lw/Wb) (4.10)In Carnation Creek, a size distribution of wood pieces from an unlogged section of channel52(SA-9) was generated from the 2007 survey year and used to estimate a mean piece length.Reach-average channel width was calculated from the 1991 longitudinal profile survey data(see Chapter 3). Pmove, the probability of piece movement, is calculated asPmove = Pl ·Pθ (4.11)Where Pl is the probability of movement based on the Lw/Wb ratio, and Pθ the probabilityof movement based on the mean piece orientation. Values of Pl were determined by applyingCarnation Creek data to probability distributions proposed by Eaton et al. (2012). Similarly,mean piece orientation was determined from the 2007 wood map data and this orientationapplied to the proposed distribution outlined in Eaton et al. (2012). Tnet , net wood transport, isthe difference between It and Ot in a given reach.An exponential decay function has been found to represent in-situ mass loss of wood piecesin natural settings reasonably well (Murphy and Kosksi, 1989; Hyatt and Naiman, 2001; Bendaand Bigelow, 2014; Ruiz-Villanueva et al., 2016). As a budget output term, wood decay (Od)over a given time interval is calculated as:Od =∫ TVo(t) · (1− e−kt)∆t (4.12)where Vo is the initial wood volume, and k the decay constant, calculated as 1/Aw, whereAw corresponds to the weighted mean age of wood in the system. Given that detailed woodage data are not available for Carnation Creek, values calculated for two streams on HaidaGwaii (Hassan et al., 2016) are used. As climatic conditions, tree species, and channel sizesare similar to Carnation Creek, these values are a reasonable approximation. Finally, lossesfrom a reach due to channel migration and abandonment (Oca) are calculated as:Oca = ∆xBVc (4.13)As channel width is assumed to remain constant and wood evenly distributed within thechannel, bank eroded by length B must be compensated for by abandonment also of length B.Simulation settings and evaluationTo implement the wood budget model, a Monte-Carlo approach was used to perform manysimulations with varying input parameters. Where possible, information on the distributions ofinput variables was used to define the range of values within which random samples were taken.A summary of best estimates, distributions, and standard deviations/ranges for input values isshown in Table 4.1, and un-varied component parameters of transport and bank erosion inTable B.1. Estimates of variability were derived from the BCVRI database, regional forest53characteristics literature (e.g. Wells and Trofymow, 1997), and aspects of the Carnation Creekwood dataset. The modeling approach incorporates several assumptions: (a) that the woodpiece size distribution remains constant through time; (b) that average channel width does notvary through time; (c) that wood loads and characteristics upstream of 3000 m are unaffectedby harvesting (relevant for transport calculation), and (d) that no large one-time input of slashor riparian blowdown occurred at the time of logging.Table 4.1: Model parameters varied for simulationsParameter Distribution Best estimate Range/SDa Units nb SourceVs Normal 880 +/- 162 m3/ha 8 BCVRIVcwd Normal 636 +/- 122 m3/ha 4 Wells and Trofy-mow, (1997)SIc Uniform 24 +/- 4 m - BCVRIkcwdd Uniform 0.027 +/- 0.002 yr−1 - TIPSYk Normal 0.021 +/- 0.0024 yr−1 7 Hassan et al.(2016)Lw Lognormal 0.76 +/- 0.32 log(m) 140 Carnation fielddataseta Where SD is standard deviationb Sample size from which statistics were derivedc Site Index, used for TIPSY Vs and Vcwd simulationsd Used for TIPSY Vcwd simulations of post-harvest CWD decaySimulations were run in the R programming language (RCoreTeam, 2017). The lower 3.0km of Carnation Creek was split into two reaches, each simulated separately but concurrently.The downstream reach (0-1300 m), which contains SA-2, SA-3 and SA-4, historically did notexperience intensive riparian harvesting, as variable-width buffer strips were retained. How-ever, it is important to note that these buffers were quite narrow in places, often < 10 m. Theupstream reach (1300-3000 m), which contains study areas SA-5 to SA-8, experienced inten-sive harvesting in the riparian zone, with near-complete removal of woody vegetation. In orderto estimate the sensitivity of wood volume to different logging configurations and to evaluatethe impact of configuration on recovery timescales at different locations along the main stem,three logging configurations are used: (a) Scenario 1, the historical scenario, with riparian har-vesting in the upstream reach but not downstream; (b) Scenario 2, where the downstream reachis harvested but upstream remains unharvested; and (c) Scenario 3, where both reaches are har-vested concurrently. While logging took place between 1976 and 1981 at different locationsin the catchment, the year 1978 is used as the dividing point between pre- and post-harvestconditions.Carnation Creek field data are used as a measure of model performance. Data from the fourprofile surveys and the study areas are used to assess performance over the full wood length54record (45 years). To generalize the study area wood data to larger areas of channel, wooddensities (volume per meter of channel) were averaged for all study areas located within thedownstream and the upstream reaches. In this way, average values from SA-2 to SA-4 wereused downstream, and SA-5 to SA-8 upstream.4.4 Results4.4.1 Large wood characteristicsCarnation Creek possesses wood storage values of up to 1.5 m3m−1, approximately 80% ofwhich is found in jams (see Table 4.2). When scaled by area, wood storage translates to ap-proximately 700 m3/ha, though this value fluctuates through time and varies along the channel(see Table 4.3). The distribution of piece sizes relative to channel width is shown in Figure 4.2,and average count by size class in each study area in Table 4.3. In general, most wood piecesare small relative to channel dimensions, with a median piece length of 5.7 m (0.36 L/Wb),diameter of 0.23 m, and volume of 0.24 m3. Within the study areas, the proportion of pieceswhere L/Wb > 1 is usually below 20% except for SA-9, where approximately 25% of all woodpieces are > 1 L/Wb. In SA-6 and SA-7, only one or two pieces of this size were found after1980.Table 4.2: Characteristics of logjams located along the lower 3.0 km of Carnation CreekSurveyyear# jams Mean jamvolume(m3)Jamvolume(m3m−1)Jam spac-ing (m)Non-jamwood(m3m−1)Totalwood(m3m−1)Storedsedimenta(m3m−1)1991 56 55.1 1.01 48.7 0.10 1.11 7.391999 49 83.8 1.31 48.9 0.13 1.44 6.492009 38 37.1 0.49 80.1 0.32 0.81 5.512017 52 19.6 0.33 57.1 0.17 0.50 5.27a Sediment stored in bars along the channel profileA summary of orientation information is shown in Figure 4.3 and Table 4.4. Pieces aremore likely to be oriented parallel to the streamflow direction (in the angle range of 67.5◦ to112.5◦) than perpendicular to flow (angles less than 22.5◦ or greater than 157.5◦). Aside fromthe angles close to parallel, there is little difference in proportion between the other orientationranges. Orientation angle distributions do not vary substantially between piece size class (Table4.4), though the sample size for the largest pieces is small (n = 5).Given the detailed nature of the Carnation wood dataset, it is possible to examine severalaspects of wood piece persistence which also provide insight into piece mobility (an exampleof mobile wood pieces in Carnation Creek is shown in Figure B.1). The distribution of relative55Table 4.3: Wood and morphology characteristics for Carnation Creek study areasWood pieces (mean count)a Wood volumeb MorphologyStudy area Class1Class2Class3Mean (m3) m3m−1 Storedsedimentcm3m−2Topographycomplexityd(m2/m2)SA-2 13.6 16.1 4.9 38.8 0.47 2.35 1.033SA-3 21.3 15.4 3.3 96.9 1.43 4.00 1.055SA-4 23.1 16.6 4.3 56.9 0.92 4.06 1.034SA-5 19.9 24.6 2.6 38.9 0.52 3.60 1.047SA-6 14.9 8.7 0.9 10.8 0.15 3.21 1.038SA-7 13.6 8.5 0.5 13.8 0.27 3.78 1.029SA-8 23.7 13.5 1.4 50.9 0.88 3.73 1.046SA-9 22.2 49.8 23.4 135.8 0.92 3.96 1.049a Excludes wood in logjamsb Includes wood in logjamsc Calculated relative to baseline surface (minimum observed elevation minus 1 m), see Chapter 3d See Figure 4.6 for calculationClass 1 Class 2 Class 3Class 2 Class 3Class 1Figure 4.2: (a) Probability density estimate of piece length relative to channel width forall pieces identified in 2007, and all 2007 pieces re-identified nine years later, in2016. (b) weak positive relationship between L/Wb and proportion of wood remain-ing nine years late using bins of 0.2 L/Wb. Vertical lines correspond to size classbreaks identified by Hassan et al. (2005b)56Figure 4.3: Wood piece orientations as identified from 2007 survey data. (a) illustrationof wood pieces (a, b, and c) relative to thalweg, and corresponding orientation angleplotted below. (b) Distribution of orientation angles for all wood pieces identifiedin study areas SA-2 to SA-8 during 2007, and again in 2016. SA-9 data were ex-cluded due to challenges identifying orientation of individual pieces, and linkingorientations to piece IDs.Table 4.4: Distribution of wood piece orientation angles relative to cross-channel direc-tionPiece orientation (degrees from perpendicular)Size class 0-22.5 22.5 -4545-67.567.5 -9090-112.5112.5 -135135-157.5157.5 -180SumClass 1 0.10 0.13 0.13 0.17 0.19 0.08 0.09 0.10 218Class 2 0.08 0.14 0.02 0.24 0.16 0.18 0.12 0.08 51Class 3 - - - - 0.20 0.60 - 0.20 5Total 0.09 0.13 0.11 0.18 0.18 0.11 0.09 0.10 274Persistinga 0.11 0.18 0.10 0.16 0.11 0.12 0.11 0.11 109a Pieces re-identified from 2016 survey data57piece size and orientation for wood persisting from 2007 to 2016 is shown in Figures 4.2 and4.3. In Figure 4.2a, the distribution is shifted rightward for 2016, with a loss of small pieces< 0.3 L/Wb but a gain in proportion of larger pieces. The largest shift is for pieces falling inthe range of 0.5-1.0 L/Wb. When the proportion of pieces remaining is plotted against binnedrelative piece size, a weak positive relationship emerges (Figure 4.3b). The distribution is moreeven for the persistent pieces relative to the full 2007 inventory, but with slightly elevated valuesfor the 67.5-90 degree and the 22.5-45 degree interval. Persistent pieces near to flow-parallelcompose 27% of the total, relative to 36% for the full 2007 record.4.4.2 Spatial and temporal patterns of large woodWood, and in particular logjams, show an uneven spatial distribution along the lower 3.0 km ofCarnation Creek (Figure 4.4 and Table 4.2). The total number of logjams identified ranges from56 in 1991 to 34 in 2009 (Table 4.2), and for all surveys, a greater number of jams are founddownstream of SA-4 than upstream. However, the jams upstream of SA-4 are comparativelylarge in the 1991 and 1999 surveys, with four or more channel-spanning jams exceeding 100m3 in volume. One very large jam is located upstream of SA-8, with a volume of over 800m3. Jams downstream of SA-4 are smaller but more numerous, with few jams > 50 m3 involume. Average jam spacing ranges from 55 to 83 m, with the greatest spacing in 2009 andsmallest in 1991. Spacing is greater above SA-4, with several intervals exceeding 200 m, whiledownstream intervals rarely exceed 100 m.Wood piece and volume time series in the study areas are shown in Figures 4.5a-h andi-p, respectively, and averages summarized in Table 4.3. The proportion of pieces in each sizeclass varies through time in all study areas, but in sites SA-2 to SA-8 the proportion of Class 3pieces generally decreases, while Class 1 increases. Two sites, however, are exceptions here:SA-3 and SA-9 do not show clear trends in loss of Class 3 wood, though some increase in theproportion of small pieces still occurs. SA-8 shows a high degree of variability, with a majorspike in proportion of small pieces in the mid 1980s, concurrent with jam formation. Patternsof wood volume (Figure 4.5i-p) also differ greatly between study areas: SA-2, SA-6, and SA-7show long-term declines, while volume in SA-3 increases, primarily a function of the additionof several very large stems in 1977 and again in 2012. Wood volume in SA-4 increased to1975 and then decreased to a minimum in 2005. Increases in the mid and early-1980s in SA-5and SA-8, respectively, are a result of the formation and growth of the previously describedjams. In SA-9, wood levels increase and then appear to fluctuate at a timescale of 10-15 years,though the gap in this data series is long.The spatial pattern of jams shown in Figure 4.4 also varies with time: the large logjamsfound upstream of SA-4 present in the 1991 and 1999 surveys reduce in volume dramaticallyfor the 2009 survey, and many have disappeared. The remaining large jam in the upstream58● ● ●● ●●● ●●●●●●● ●●●●●● ●●● ●●●●●●●● ● ● ●● ●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●● ●●●● ●● ● ● ● ●●●●●●●● ● ●●●● ● ●●●●●●●●●●●●●●●●●● ● ● ●●●●●●● ● ●●● ● ● ●● ●●● ● ● ●●●●●●●●●● ● ●●●●● ●●● ●● ●●●● ● ●●●● ● ● ●●● ●●●●●●●●● ●●●●●(a) 1991(b) 1999(c) 2009(d) 2017●●●Logjam volume (m3)50020050● 200 500 mNSA-2SA-3SA-4SA-5SA-6SA-8SA-7Figure 4.4: Overhead map of the lower 3.0 km of Carnation Creek, with logjam locationsand sizes mapped for (a) 1991, (b) 1999, (c) 2009 and (d) 2017 longitudinal profilesurveys. Study area locations are indicated as grey bands. Flow direction is fromtop right to bottom left of the Figure.59Figure 4.5: Proportion of wood pieces falling into relative size classes 1-3 (panels a-h):Class 1: L/Wb < 0.3; Class 2: L/Wb 0.3-1.0; Class 3; L/Wb >1.0. Wood piecesin jams are excluded from these Figures. Panels i-p show wood volume, derivedfrom the size classes described above, through time in the different study areas.Uncertainty and volume for wood pieces and jams is calculated separately: non-jamwood volume is based on standard error of the mean piece size in each class, whilejam uncertainty is based on published ranges for jam porosity (see Dixon, 2016) andvolume. Data gaps ranging from 2 to 11 years are a product of missing piece maps.60region, found between SA-7 and SA-8, has reduced in volume by more than half. Downstream,additional jams appear in the 2009 survey relative to 1999, and several also grow in size. Whilejams are lost in the 2009 survey, several new small ones have formed by 2017 (see Figure 4.4d and Table 4.2). Over the entire 3.0 km of channel, total wood storage in jams increases by25% between 1991 and 1999, then falls by half for 2009, and by another 30% for 2017.4.4.3 Wood-morphology interactionsRelationships between wood piece abundance and topographic complexity (the ratio of 3D bedarea to 2D bed area) are shown in Figure 4.6, divided into piece size classes. For all classes,a significant (p < 0.05 in all cases), positive relationship between the degree of topographiccomplexity and wood abundance is found. While variance explained by wood piece abundanceis low, it is higher (8%) when all classes are grouped (Figure 4.6d) than for individual sizeclasses.Bed sediment storage and logjams located along the lower 3.0 km of channel are shown inFigure 4.7 (see Chapter 3 for details of sediment storage calculations). While volume alone isnot always a good indicator of a jam’s capacity to store sediment (Hogan et al., 1998), severalmajor spikes in storage correspond closely to jams along the channel. Though findings inChapter 3 suggest that jams are not the dominant sediment storage mechanism in CarnationCreek, they generally occur in association with zones of elevated storage (Figure 4.7), withthe largest spikes in storage immediately upstream of some of the large jams. However, thereare also several locations (e.g. 1991 near 650 m) where relatively large jams do not coincidewith substantial sediment storage upstream. For 1999 and 2009, regions along the channelwith closely spaced jams (500-1500 m) contain many spikes in storage of lower magnitude,and zones are present where few jams exist and also little material is stored (e.g. 1500 to 2000m for 1999, and 1800 to 2300 in 2009). A few spikes in close proximity to jams (e.g upstreamof 2500 m 1991-2009) are relatively persistent through time. While few locations of elevatedstorage are present in 2017, those remaining are in close proximity to jams (e.g. near 2350 m).The mechanisms of sediment storage behind jams and reduction of storage following lossof jam functionality can be examined in greater detail in Figure 4.8. Prior to the formationof the SA-5 jam (Figure 4.8a), a tributary entered the channel from the left bank near thedownstream end of the reach. Several years after jam formation (Figure 4.8c), the main SA-5channel was blocked, the jam is effectively closed and impermeable to sediment throughput(Hogan, 1989), and major channel aggradation has occurred upstream. An avulsion into theleft-bank tributary channel has occurred in two locations. By 2008 (Figure 4.8e), the mainchannel abandoned its previous course downstream of the jam, and much of the depositedsediment upstream was scoured or is now stored outside of the active channel. The channelin SA-8 responded in a similar way to SA-5, with relative impermeability forcing upstream61●●●●●●●●●●●●●●●●●y = 0.011 x + 1.035R2 = 0.021.0251.0501.0750.0 0.2 0.4 0.6Topographic roughness (m2/m2)●●●●●●●●●●●●●●●●●●y = 0.025 x + 1.035R2 = 0.051.0251.0501.0750.0 0.2 0.4●●●●●●●●●●y = 0.047 x + 1.035R2 = 0.031.0251.0501.0750.0 0.1 0.2 0.3Wood abundance (Pieces/m)●●●●●●●●●●●●●●●y = 0.013 x + 1.035R2 = 0.081.0251.0501.0750.0 0.5 1.0 1.5● Binned dataRaw data(a) (b)(c) (d)Figure 4.6: Plots of topographic complexity vs. in-stream wood abundance (pieces permeter) for (a) Class 1 pieces, (b) Class 2 pieces, (c) Class 3 pieces, and (d) all woodgrouped. Topographic complexity is defined as the ratio of 3D bed surface area to2D area, calculated as A3D = A2Dcosθ , where A2D is the two dimensional area of a rasterpixel, and θ is the slope angle of a given point on a 2D surface. Data for all studyareas are plotted in each panel, and blue circles represent binned data. Fit linescorrespond to relationships for the raw data, and in all panels are significant at p =0.05, though variance explained is low.62Figure 4.7: Plots of logjam location and volume (black bars), and sediment storage (bluelines) along the lower 3.0 km of Carnation Creek for (a) 1991, (b) 1999, (c) 2009,and (d) 2017. Sediment storage is calculated from methods discussed in Reid et al.2019.63aggradation, and as a consequence channel avulsion had occurred by 1991 (Figure 4.8c andd, respectively). By 2008 (Figure 4.8f) the channel occupied a position left of the pre-jamcourse, and left much of the stored sediment outside the active channel area. For both jams,the primary mechanism of wood removal from the active channel was channel abandonment,rather than transport or decay. Large wood in these study areas, therefore, had a major influenceon morphology and sediment storage over the period of record.Topographic survey data in SA-5 and SA-8 can be used to examine rates and timescalesof sediment wedge loss following avulsions around the two logjams. Figure 4.9 illustrates theloss of storage through time upstream of both jams. In SA-5 and SA-8, scaled sediment storagedeclined to low values after 10 and 17 years, respectively. When data points from both jamsare combined, the best fit is one of negative exponential decay.4.4.4 Wood budget model and large wood response timescalesWood budget terms are shown in Figure 4.10 and Table B.2 for pre-harvest conditions andthose at t = minimum storage. Prior to harvesting, wood input of live stems from bank erosionis dominant (4.10a and b), followed by bank erosion additions of CWD. Mortality is low inboth reaches, constituting less than 1% of input. Decay is the largest source of wood loss,followed by net transport and then channel abandonment. When reach results are combined(Figure 4.10c) losses through channel abandonment are slightly greater than net transport. Att = minimum storage, inputs and outputs are much lower but in similar proportions for theupstream reach, as storage is low and riparian zone inputs are still recovering. Downstream(4.10b), inputs and outputs are generally similar relative to pre-harvest conditions, but channelabandonment losses are lower (as they depend on storage), and net transport is more negativeas input from upstream is greatly reduced.Results of the wood budget simulations are shown in Table 4.5 and Appendix Figures 4.11to 4.13, and time series of budget terms in Figures B.2 to B.4. Overall, the field and modeldata show similar trends and magnitudes over the 45-year period of overlap. While the 95thpercentile range of model outcomes is relatively wide, the model predicted within the range ofobserved field values, but somewhat under-predicted pre-harvest conditions, particularly in theupstream reach (Figure 4.11). Relative to the wood volumes from the 1991 and 1999 surveys,the model under-predicts storage in both reaches but is very close to the 2009 and 2017 fieldvalues. In the upstream reach, pre-harvest SA data fall below the model predictions, but thenmatch well during the 1980s and early 1990s. For the downstream reach, the study area dataare highly variable but generally fall within the model prediction range. Similarly, the modeldata fall within the profile survey data point ranges, but these also show substantial variability.When the two reaches are grouped, most of the study area data fall close to the model meanprediction line, with the exception of pre-harvest data and data from 2000 to 2010.64Figure 4.8: Panels illustrating evolution of logjams and channel morphology in SA-5(panels a, c, and e), and SA-8 (panels b, d, and f). Flow in all panels is fromright to left. Relative bed elevation in the maps is shaded from low (black) to high(white) while wood pieces are shown as brown polygons.653Figure 4.9: Ratio of sediment stored by jams in SA-5 and SA-8 at time t (Vt) relative tomaximum storage (Vo). Storage is calculated by isolating bed areas upstream of jamlocations, and subtracting surfaces in these areas from the pre-jam surface.Results of the budget simulations for Scenario 1 (observed logging history) are shown inFigure 4.11. Pre-harvest wood loads, similar for all simulations, are in the order of 1 m3m−1.Following harvesting, storage drops rapidly in the upstream reach (Figure 4.11a), falling tohalf of pre-harvest levels in approximately 30 years, and reaching a minimum between 50 and75 years post-harvest (Table 4.5). Recovery to pre-harvest wood storage takes just under 200years. Downstream in Figure 4.11b, the rate of decline is driven solely by the reduction intransport from upstream, and is thus delayed and lower than the decline observed in Figure4.11a. Minimum wood storage in the downstream reach occurs later than upstream, approxi-mately 85 years post-harvest. In spite of this delayed minimum, the recovery time is similar tothe upstream reach. When results are grouped (Figure 4.11c), minimum storage occurs approx-imately 70 years following harvesting, and LW storage recovers nearly 200 years post-harvest.Minimum storage values are less than half of pre-logging levels in the upstream reach, 80% ofpre-harvest in the downstream reach, and roughly 55% of pre-harvest levels when values aregrouped.Simulation results for Scenario 2 (downstream logging), are shown in Figure 4.12. While66Table 4.5: Model results summary for all simulation configurationsStorage (m3m−1)a Timescale (yrs)Scenario Reach Pre-logPre-log+/-Minb Min+/-Minc Min+/-Rec.d Rec.+/-e1 US 1.00 0.77-1.280.35 0.27-0.4463 51-74 197 152->225DS 1.01 0.76-1.310.79 0.60-1.0186 73-98 196 -All 1.01 0.76-1.290.55 0.42-0.6969 56-81 197 150->2252 US 1.00 0.77-1.28- - - - - -DS 1.01 0.76-1.310.54 0.41-0.7056 47-67 159 134->225All 1.01 0.76-1.290.80 0.61-0.9956 47-67 159 -3 US 1.00 0.77-1.280.35 0.27-0.4463 51-74 197 152->225DS 1.01 0.76-1.310.32 0.26-0.4567 54-80 170 139->225All 1.01 0.76-1.290.38 0.29-0.4863 51-77 172 140->225a Volumes scaled by reach lengthb Minimum wood storage valuec Time to minimum storaged Recovery time to pre-harvest conditionse > 225 years if lower 95% interval outside of pre-harvest intervalsimulation results are simply a continuation of historical conditions (Figure 4.12a), Figure4.12b displays a sharp reduction in storage similar to (a) in Figure 4.11. Minimum storagehere is approximately 0.55 m3m−1, a smaller reduction than for the upstream reach as thetransport influx remains relatively large. The time of this minimum is approximately 55 yearspost harvest, a shorter delay than for Scenario 1. Similarly, the time to recovery is shorter,with pre-harvest wood loads reached after approximately 160 years. Given the lack of loggingresponse in the upstream reach, the timescale of minimum wood loads and recovery are similarwhen the two reaches are grouped (Figure 4.12c) as for the downstream reach. The magnitudeof the wood loss is overall lower, a reduction from 1.01 to 0.80 m3m−1.Simulation results for Scenario 3 are shown in Figure 4.13. Overall, results for the in-dividual reaches are similar to those for the previous simulations, but when the reaches arecombined (Figure 4.13c) a greater reduction in wood storage is apparent, to 0.54 m3m−1. Thetime to minimum is 60 years post harvest, while recovery time is approximately 180 years.67(live) (live) (live)(CWD) (CWD) (CWD)Figure 4.10: Budget model input and output terms for Scenario 1, which aims to matchhistorical conditions. Results are shown for (a) the upstream, logged reach, (b)the downstream, unlogged reach, and (c) the reaches grouped. BE (live) is bankerosion input of live stems, CA channel abandonment, BE (CWD) coarse woodydebris (from bank erosion), D in-situ decay, M mortality, and TR net fluvial trans-port of wood. Bars with only one shade have equal values between pre-harvest andminimum-load conditions.4.5 Discussion4.5.1 Large wood characteristics and patternsWood data from Carnation Creek reveal complex patterns of storage and mobility which varyalong the stream channel. Overall, LW storage is within the upper range of values reportedfor other systems (Hogan, 1986; Nakamura and Swanson, 1993; Andreoli et al., 2007; Jacksonand Wohl, 2015), but varies considerably through time and across space.As with other small streams (e.g. Benda and Bigelow, 2014), wood mobility appears rel-atively low, and small pieces either decay more rapidly, or are preferentially transported orlost from the channel. Figure 4.2 provides some evidence of greater mobility of small pieces,and the inverse of Figure 4.2b could be considered a mobility function. The loss of a clearbias towards flow-parallel orientation seen in Figure 4.3 also suggests that in spite of having asmaller cross-sectional area exposed to streamflow, flow-parallel pieces are also more likely tobe mobile or otherwise lost from the channel, though the flow-parallel orientation may in factbe a product of previous wood transport. Wood piece mobility in previous studies is usuallylinked to relative piece size, shape, orientation, and flow magnitude (Braudrick and Grant,2001; Manners et al., 2007): in considering relative size, most pieces not re-located after 9years in Carnation Creek are small relative to both channel width and depth, confirming previ-ous findings (e.g. Lienkaemper and Swanson, 1987; Hassan et al., 2005b; Seo and Nakamura,68Time of wood minimum Figure 4.11: Budget model output for Scenario 1 (historical logging conditions). Resultsare shown for (a) the upstream, logged reach, (b) the downstream, unlogged reach,and (c) the reaches grouped. Study area (SA) wood data are scaled by combinedlength of areas falling within a reach. Note that in (b), bars with only one shadehave equal values between unlogged and t = minimum conditions. Errors in profile(LP) wood survey volumes correspond to a misclassification by +/- 1 size categoryfor all pieces (see Hogan, 1986)69●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ● 2000 2050 2100 2150 2200Wood storage (m3 m−1)(a)●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 2000 2050 2100 2150 2200Wood storage (m3 m−1)(b)●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 2000 2050 2100 2150 2200DateWood storage (m3 m−1)(c)95% Prediction intervalMinimum rangePre-harvest 95% prediction interval●● LP wood vol.SA wood vol.Prediction medianHarvestTime of wood minimum Figure 4.12: Budget model output for Scenario 2 (upstream reach is not logged, whilethe downstream reach is logged). Results are shown for (a) upstream reach, (b)downstream reach, and (c) reaches combined. Study area (SA) wood data arescaled by combined length of areas falling within a reach. Errors in profile (LP)wood survey volumes correspond to a misclassification by +/- 1 size category forall pieces (see Hogan, 1986)70Time of wood minimum Figure 4.13: Budget model output for Scenario 3 (both reaches are logged). Results areshown for (a) upstream reach, (b) downstream reach, and (c) reaches combined.Study area (SA) wood data are scaled by combined length of areas falling within areach. Errors in profile (LP) wood survey volumes correspond to a misclassifica-tion by +/- 1 size category for all pieces (see Hogan, 1986)712009). It is worth noting that the loss of pieces through time may be a product of wood lossesbut could also be a function of the measurement strategy. Single-tag regimes such as thoseused in Carnation Creek are simple and easy to implement, yet piece breakage, decay, or burialcan lead to difficulties in piece re-identification.Spatial and temporal patterns of LW storage in Carnation Creek appear to reflect the historyof riparian management. In SA-5 and SA-8, the dominant temporal patterns in LW volumeare controlled by the major logjams, but explanations for patterns in other study areas aremore complex. SA-2 and SA-3, for instance, have similar riparian histories but while SA-2 shows a persistent decline in volume, SA-3 experiences step-increases through time. Thetwo SA-3 steps (in the late 1970s and again in 2012) are a product of additions of very largestems from the riparian zone, trees of a size which SA-2 generally lacks, possibly due to athin riparian buffer. The increase and then decline of wood in SA-4 is likely attributable tolosses of large pieces either through decay or channel abandonment, as pieces of this size areunlikely to be fluvially transported (Gurnell et al., 2002; Eaton et al., 2012). Wood storageand piece size distribution patterns in SA-6 and SA-7 most clearly illustrate the influence ofhistorical timber harvesting practices: abrupt drops in storage and presence of large pieces(Figure 4.5) corresponds closely to riparian harvest dates in this area, which included some in-stream salvage (see Chapter 2). Patterns of storage in SA-9 are more difficult to explain, as noriparian harvesting occurred here and wood transport is likely to be low given the comparativelynarrower and shallower channel. Localized windthrow or mass mortality could explain therapid additions of large pieces in the early to mid 1980s.In relation to the wood survey data in the downstream part of the channel (see Figure4.11b), the explanation for the higher storage in 1999 relative to 1991 may be related to mo-bility of small wood pieces from upstream as jams decay without resupply. The very largedrop in wood storage from 1999 to 2009 appears to be at the expense of large jams in theupstream half of the channel (Figures 4.4 and 4.7), a configuration which is similar in 2017.Collectively, the harvesting-driven changes in the spatial organization of wood appear to beextending downstream with time.4.5.2 Morphological influence of large woodAnalysis of wood variables in relation to channel morphology indicates that both sedimentstorage and topographic complexity are positively related to the presence and abundance ofLW, lending clear support to the first hypothesis. While the association between topographiccomplexity and wood abundance is relatively weak, results indicate that individual jams canstore a large quantity of sediment, up to several times the average annual bed load (see Chapter3), and storage can persist for over a decade. Data presented here have therefore provided ahigh-resolution view into concurrent logjam formation, sediment storage, and degradation over72the full logjam lifecycle.The mechanism of channel adjustment to the jams in SA-5 and SA-8 is aggradation fol-lowed by avulsion. The consequence of this adjustment style is that much of the jam-storedsediment remains in the floodplain as the channel migrates away from the sediment wedge.Therefore, rather than eroding previously deposited sediment, jams of this type have the ef-fect of driving erosion of old floodplain deposits through lateral migration. In systems whichare laterally confined, sediment stored by jams will remain in place until the jam itself breaksapart, a timescale which is likely to vary substantially as a function of the specific jam charac-teristics (see Eaton et al., 2012), particularly the state of decay and structural integrity whichare a function of time and wood species (Hogan, 1989).The lifecycle of the logjam - sediment storage unit can be split into three phases: (1)the formation phase, when the jam and sediment wedge itself are created. The timescale ofthis phase will depend on the jam formation mechanism, and may range from a single stormevent (Hogan et al., 1998; Lancaster et al., 2003; May and Gresswell, 2003) to much longer(Keller and Swanson, 1979; Dixon and Sear, 2014). Additionally, the bedload transport rateand jam both integrity and permeability will likely influence the timescale and magnitude ofthe wedge formation; (2) A period of high storage when the jam is at maximum integrity andimpermeability to sediment, and a large sediment body has formed upstream. The timescaleof persistence here will be a function of mechanism of channel adjustment to the jam. Achannel which can avulse will potentially result in a very short or non-existent second phase,while a jam and sediment wedge in a laterally confined channel may persist for many years,in some cases even converting bedrock channels to alluvial ones (Massong and Montgomery,2000); and (3) the degradation and wedge erosion phase, which will again be a function of thechannel adjustment style. In the case of avulsion, in-channel storage may take up to 20 years toreturn to pre-jam levels (Figure 4.9), but could occur rapidly should the entire jam fail at once.Previous research provides evidence of major sediment storage and impacts to channelmorphology from logjam formation in channels similar to Carnation Creek. The impact oflarge jams on channel avulsion and floodplain dynamics has been well documented in (Brum-mer et al., 2006), and (Collins et al., 2012) where jams are found to be a key formative andmaintenance process of floodplains. Similarly, Wohl (2011) provided evidence of logjamscausing single-thread channels to transition towards multi-thread. More generally, Wohl andScott (2017) proposed several hypotheses related to sediment storage from wood, and alsofound a significant positive relationship between volume of wood and quantity of sedimentstorage. While other studies have examined wood characteristics in relation to several otherchannel geometry and morphology variables (e.g. pool spacing and channel width), the influ-ence of wood metrics on the diversity of bed elevations induced by complex wood-modulatedhydraulics do not appear to have been evaluated. While the topographic influence of wood in73Carnation Creek is broadly similar to other studies, the timescale after which jams cease tobe effective geomorphic agents is much shorter than suggested in Eaton et al. (2012), wheremost simulated jams are reported to fail after 20 years or more. However, their study did notconsider avulsion as an adjustment option. While the influence of logjams on sediment storageis clear, little evidence of a timescale and pattern of erosion for jam-forced sediment deposits isfound in the literature. Therefore, this work provides a basis to evaluate loss of material fromsediment wedges in cases where channels can adjust through avulsion.4.5.3 Timber harvesting influence on LW storage and channel morphologyResults from the wood budget modeling in tandem with patterns in field data suggest a recov-ery time in the order of two centuries is needed to reach pre-harvest wood levels, and that aminimum in LW abundance will only be reached after more than 50 years. Additionally, aresponse in sediment storage associated with wood loss at the profile scale is apparent. Thesefindings present a first look at recovery times using a wood budget modeling approach, andhave clear implications for many regions which have experienced widespread historical timberharvesting.The hypothesis that morphological channel change will stem from harvesting-related woodlosses can be evaluated by examining data in Table 4.2 and 4.3. With the exception of increasedwood abundance in 1999, the general decline in wood along the channel main stem also corre-sponds to a reduction in stored sediment. However, morphological differences between studyareas with different logging histories are less apparent, and while wood loads clearly droppedin some of the most affected sites (SA-6 and SA-7, see Figure 4.5), this has not translated tonotable differences in sediment storage or topographic complexity between sites as a functionof logging history (Table 4.3). As the diversity of channel bed morphology arising from jamscan be localized (Abbe and Montgomery, 1996), it is possible that the relatively short studyareas are unable to capture the logging effect noted at broader scales.The wood budget model performs reasonably well, and it is likely that differences betweensimulation results and model output (e.g. Figure 4.11a) are at least in part a product of ex-trapolating localized wood data from study areas to a larger scale. It is worth noting that nostudy area in Reach 2 (the upstream reach) contained a logjam at the time of study area es-tablishment, yet large jams compose the majority of the LW load along the channel during thefirst two profile surveys. Once jams have formed in SA-5 and SA-8, the model and field datamatch closely (Figure 4.11a and c). However, the overall model performance is similar to thatof Hassan et al. (2016).The spatial organization of harvesting in the watershed has a moderate impact on recoverytimescales (Table 4.5). However, wood transport introduces a lag effect in storage downstream,delaying the wood minimum timing, and most likely recovery time. When transport from74upstream is intact, the time to minimum is shorter, recovery time faster, and absolute valueshigher than when upstream transport is affected. As expected, the magnitude of wood loss ishighest when harvesting has occurred along a greater length of channel. Wood volume alone isnot necessarily a good proxy for the influence that wood can have on channel morphology andaquatic habitat (King et al., 2013): the large, long-lasting jams in Carnation Creek are typicallycomposed of at least some very large wood pieces with a diameter greater than 0.5 m. Basedon TIPSY model output, trees of this size do not begin appearing in the channel for over 60years post-harvest, and therefore some elements of wood-channel interaction may be impactedbeyond the minimum suggested by total LW volume alone.Results of the TIPSY regeneration model were produced from best estimates of inputs forCarnation Creek. However, in regions with different climate, species composition, replantingstyles and disturbance regimes, the recovery timescales are likely to vary. Recovery times areapt to be sensitive to episodic inputs: if a large quantity of wood is delivered from hillslopes(e.g. Andreoli et al., 2007; Ruiz-Villanueva et al., 2016), then very rapid increases in woodloading are possible, reducing recovery times. Similarly, different historical riparian harvestingapproaches can have a large impact on temporary in-stream loads, should excess slash be leftin the channel, or should a riparian buffer blow down (Nakamoto, 1998).While few studies have examined timescales and storage reductions of wood over the longterm, results presented here can be evaluated against projections from Murphy and Kosksi(1989); Bragg (2000); Stout et al. (2018). Murphy and Kosksi (1989) used a simple approachof comparing losses of in-stream wood from a depletion function to inputs from regeneratingriparian zone trees in order to evaluate the effectiveness of a 30 m buffer strip. Ultimately, theyfound that minimum wood loads were reached almost 100 years after harvesting, with recov-ery to pre-harvest levels not having occurred even 250 years post harvest. While their time tominimum and recovery times are somewhat delayed over those projected for this chapter, theyprovide further support for century-scale impacts to stream channels from riparian harvesting.Both Bragg (2000) and Stout et al. (2018) use similar Monte-Carlo approaches to that appliedhere to characterise recovery in a Pacific Northwest and Australian stream, respectively. Bothstudies find that recovery takes two centuries or longer, with Bragg (2000) also suggesting atime to minimum of approximately 75 years. (Davidson and Eaton, 2015) examined simulatedimpacts to habitat from a loss of riparian trees from both fire and timber harvesting. Whiletimescale of recovery was not considered, wood additions from fire, and reductions from har-vesting translated to positive and negative changes to habitat availability, respectively.Collectively, this work provides further evidence of the long-term impacts of timber har-vesting on in-stream wood loading and channel morphology, with clear implications for aquatichabitat. As many small streams experienced intensive riparian timber harvesting until the lastthird of the 20th century, many are now likely approaching, or have just passed their minimum75wood loading. As a consequence, stream channels will likely remain in a degraded but im-proving state for many years to come. Riparian buffer strips have been used since the 1960sin western North America (Richardson et al., 2012), and appear to assist in preserving at leastpart of the natural wood supply regime to the system.4.6 ConclusionsWood loading and channel morphology in forested streams which have undergone widespreadtimber harvesting remains poorly studied over the long term. This chapter has examined long-term spatial and temporal patterns of wood storage, relationships to channel morphology, andimpacts from timber harvesting in the riparian zone. To evaluate longer-term impacts fromlogging, a wood budget model was developed and applied to determine post-harvest wood loadcharacteristics and timescales of recovery in the system. A unique long-term dataset of woodpiece locations and volume was used for analysis.Carnation Creek historically contained a high degree of wood loading, and most woodwas stored in logjams, some of which were very large. However, a reduction in wood supplythrough time has led to declines in storage along much of the channel. Presently, the spatialdistribution of wood is uneven, and few large jams remain in the system. Overall, sedimentstorage in the channel decreased as LW decreased, suggesting an extension of the impacts fromharvesting to channel morphology, and in turn aquatic habitat.While wood abundance has lessened through time, LW in the channel has had a significantimpact on channel morphology, particularly in relation to topographic complexity of the bed,and in local sediment storage. The evolution of two logjams captured in detail reveals anavulsion channel adjustment style, and a timescale of storage response in the order of 10-20years following the avulsion. Sediment upstream of the jams displayed a negative exponentialdecay through time.The simulated wood budget matched observed conditions reasonably well, and results sug-gest that LW volume recovery times are in the order of 150 to more than 200 years. Woodstorage reaches a minimum more than 50 years post-harvest, as in-stream material decays, istransported downstream, or is lost through channel abandonment.Overall, this work provides novel temporal information on wood loads in a forested, gravelbed stream over a timescale which captures key wood-channel morphology interactions. Giventhe widespread nature of riparian timber harvesting which has occurred, these findings are rel-evant to a large geographic area, and indicate that logging impacts may extend to morphology(and in turn habitat) for over a century. Continued monitoring of wood abundance, morphol-ogy, and habitat will be important to track the recovery of these systems.76Chapter 5Testing a process-based linkagebetween catchment structure andaquatic habitat5.1 SummaryStream channel morphology forms the template upon which hydraulic aspects of aquatic habi-tat are imposed, yet spatial and temporal variability in habitat created by changing morphologyis not well understood. This chapter presents a conceptual model linking catchment organiza-tion and sediment delivery to variability in channel form and aquatic habitat. To evaluate thismodel, spatial and temporal variation of three habitat variables are quantified using a 2D hy-drodynamic modeling approach. The 45-year record of topographic data from Carnation Creekis used as input for the flow modeling. Using the Nays2DH modeling platform, water depthsand velocities were simulated in eight channel segments and for seven flow levels ranging from3% to 400% of mean annual discharge over the period of record. From the model output, timeseries of pool area, high-velocity flow area, and pool area with wood cover were produced.Results indicate that habitat availability changes through time as a result of changing chan-nel morphology and wood loads, but patterns between study areas vary as a function of theirdominant morphology. These changes also influence habitat availability at a given flow level,leading to non-stationary habitat-discharge rating curves. When habitat areas are predictedfrom these curves applied to daily flow series spanning annual May 1st to Sept. 30th dry sea-sons, over 50% of the variance in seasonal cumulative habitat can be explained by changes tochannel morphology and wood loading within individual study areas. This value is found to berelated to the temporal morphological variability of a study area, which in turn is related to thesegment position relative to zones of colluvial input. Collectively, these results lend support77to the conceptual model and allow for the evaluation of variability in habitat on the basis ofvariability in sediment supply.5.2 IntroductionPacific salmonids are of significant economic, ecological, and cultural importance in the Pa-cific Northwest of North America (Gislason et al., 2017). Historical salmonid populationshave been observed to vary over decadal timescales (e.g. Tschaplinski and Pike, 2017), butsome species are thought to be in decline (Irvine and Fukuwaka, 2011) partly as a result ofhabitat loss in river systems (Gregory and Bisson, 1996; Tschaplinski and Pike, 2017). Man-aging watersheds for salmonids requires balancing competing demands for water and otherresources in fish-bearing streams (Bradford and Heinonen, 2008). However, to successfullymanage streams for present and future conditions, it is necessary to understand linkages be-tween habitat and physical channel structure, and how channel-altering processes determinehabitat availability, with potential consequences for fish abundance and distribution throughtime.Freshwater habitat is important to Pacific salmonid species as some spend one or moreyears rearing in river systems as juveniles (Meehan and Bjornn, 1991). Over this period, juve-niles have specific physical habitat requirements. Metabolic efficiency and growth rate dependon stream temperature and availability of food sources (Becker and Genoway, 1979; Murphyand Meehan, 1991), in addition to dissolved oxygen and turbidity levels (Bjornn and Reiser,1991). Suitable hydraulic conditions are essential for allowing juveniles access to resources,such as low velocity pool areas (Bjornn and Reiser, 1991; Beecher et al., 2002) which are deepenough to allow for passage, foraging and optimal cover (Nickelson and Reisenbichler, 1977;Bjornn and Reiser, 1991; Beecher et al., 2002). Structural cover elements in stream chan-nels, such as in-stream large wood (LW), also serve an important role in sheltering juvenilesfrom predation and in reducing water velocities (Bjornn and Reiser, 1991; Hafs et al., 2014).Additionally, LW quantities are often positively associated with juvenile salmonid abundance(Cederholm et al., 1997; Benke and Wallace, 2003; Pess et al., 2012).For a given river discharge, hydraulic conditions forming habitat for juvenile salmonids areshaped by a channel’s topographic structure and LW (Abbe and Montgomery, 1996; MacVicarand Roy, 2007; Caamano et al., 2012). Previous research highlights the connection betweenchannel morphology and topography and larger-scale sediment supply dynamics, such as thoseimposed by channel-hillslope coupling (Hoffman and Gabet, 2007; Hassan et al., 2019). Fora given valley gradient, a range of channel morphologies is possible as a function of sedimentsupply and calibre (Church, 2006; Hassan et al., 2008). Several flume-based and dam removalstudies have noted that variations in sediment supply lead to changes in channel bed relief, suchas bar building and erosion (Lisle et al., 1993; Dietrich et al., 2005; Venditti et al., 2012; Major78et al., 2016) and pool filling (Hoffman and Gabet, 2007; Major et al., 2016). Changes in sedi-ment loads from debris flows are common in mountain landscapes, and have also been shownto influence bar and pool characteristics downstream (Madej and Ozaki, 1996; Hoffman andGabet, 2007; Pryor et al., 2011). Collectively, this work suggests a clear impact of sedimentsupply on morphology and topographic relief, which will ultimately influence the hydraulicconditions forming suitable habitat.Managing an aquatic system for salmonids is complex and requires consideration of manyvariables including linkages between physical channel structure and stream channel hydraulics,and how these are likely to vary over time and space, and at different flow levels. Temporal vari-ability in channel bed form is well documented (e.g. Beschta, 1983; Pryor et al., 2011; Buffing-ton, 2012), but this variability is not often explicitly considered when modeling aquatic habitatfor management purposes, nor is it clear what effect morphology has on habitat-regulating hy-draulics at different flow levels through time. Due to a limited number of long-term datasets,habitat modeling studies typically focus on short timescales or spatial aspects of habitat (Har-rison et al., 2011; Cienciala and Hassan, 2013; Hafs et al., 2014; Carnie et al., 2016), orexamine conditions over longer timescales but assume static morphology (Fabris et al., 2017).However, to better understand how changing morphology influences aquatic habitat throughtime and at different flow levels, it is necessary to consider both spatial and temporal aspectsof aquatic habitat variability at a scale which captures key controlling processes.While the presumption of unchanging morphology is reasonable for short-term habitatmodeling projections or in systems where stream channel stability is relatively assured, it is notvalid in catchments with highly variable sediment supply, especially when considering multi-year or decadal timescales. In many regions, such as northwestern North America, historicalglaciation, deep soils, and steep hillslopes result in large and episodic delivery of sediment andwood to stream channels (Roberts and Church, 1986; Imaizumi and Sidle, 2007; Hassan et al.,2019), which commonly leads to spatially and temporally changeable channel morphology(Hoffman and Gabet, 2007). These dynamic conditions are fundamental for shaping streamchannels, yet few studies have described how hydraulic aspects of aquatic habitat for juvenilesalmonids vary over time as a result of these processes.5.2.1 Conceptual model and study objectivesTo better understand connections between spatiotemporal variability in habitat and processesthat regulate channel morphology, a conceptual model is proposed here. This model describesthe connection between sediment supply regime and temporal patterns in channel form, whichin turn influence hydraulics for aquatic habitat (Figure 5.1). This conceptual model contrastschannel sections representing (1) a reach with quasi-steady sediment supply (Figure 5.1a), and(2) a reach with episodic supply (Figure 5.1b). In a catchment which experiences minimal79variability in sediment supply and wood load (Figure 5.1a), channel form is likely to remainrelatively constant over the multi-decade scale, and therefore the dominant source of year-to-year variability in aquatic habitat will be driven by hydrological variability. However, incatchments with temporally variable sediment supply and wood loading (Figure 5.1b), chan-nels will experience substantial changes in morphology at annual to multi-decade timescales,which in tandem with hydrology, will further drive variability in hydraulic conditions formingaquatic habitat through time. This model, explored further in this paper, allows the contribu-tion of dynamic morphology to habitat variability to be evaluated on the basis of a catchment’ssediment supply regime.In order to evaluate this conceptual model and examine temporal patterns of aquatic habitat,a 2D hydrodynamic modeling approach based on a long-term empirical dataset is used to meetfour objectives: (1) to examine the temporal variability in modeled habitat induced by variabil-ity in channel topography and wood loads; (2) to examine the relationship between availabilityof modeled habitat and spring/summer streamflow at several flow levels of significance; (3) toassess the significance of changes to channel form relative to hydrology in explaining observedhabitat variability; and (4), to examine the effects of temporal and spatial scale on variabilityin habitat. The hypothesis in this chapter is that channel sections which experience greatertemporal variations in morphology will also see greater variations in area of suitable hydraulicconditions for salmonids.To meet these objectives, the 45-year record of topographic channel and wood data (1971-2015) in the eight study areas of Carnation Creek is used as input for a 2D hydrodynamicmodel, which simulates flow depths and velocities over this period of record. The focus of thiswork is on the annual low-flow period of May 1st to September 30th when coho salmon (On-corhyncus kisutch) juveniles are residing in the channel. Three habitat variables are examined:pool area, wetted area with high flow velocities, and pool area with wood cover. Collectively,this paper aims to improve our understanding of the role that channel form plays in providingsuitable salmonid habitat over multiple decades.5.3 Methods and data5.3.1 Field dataDetailed data on channel morphology and in-stream wood have been collected in all eightstudy areas since 1973, and since 1971 in SA-2, SA-3, SA-6, and SA-8. Throughout theperiod of record, several channel attributes were surveyed and recorded, such as bank topsand bottoms, thalweg position, and margins of the wetted channel. Data have been collectedpredominantly during low flow conditions between May and mid-October. Topographic datawere not collected for any study area in 2010, or in SA-9 for 1990. Data pertaining to in-stream80Figure 5.1: Conceptual diagram contrasting (a) a stable channel with low or quasi-steadysupply, and (b) an unstable channel with high/episodic supply. (c) key process link-ages, shown as line plots beneath each block diagram, illustrate the relative contri-butions of water discharge, sediment supply, and complexity of channel morphologyto shaping temporal patterns of habitat. The final panel at the bottom shows howthese individual components combine to produce habitat in the contrasting channelstudy areas (a) and (b).wood include piece orientations and dimensions. These data were collected concurrently withthe study area topographic surveys, though wood data are not available for several years in thelate 1990s. A detailed description of Carnation Creek topographic and wood data collectionand preparation can be found in Chapter 3.Hydrometric data have been collected at several locations in the watershed for nearly theduration of the experiment. While five stations are currently active, for this study, flow data81collected at a main-channel weir between SA-2 and SA-3 (B weir, Figure 2.4, Water Survey ofCanada station ID: WSC-08HB048) are used to estimate several low flow parameters necessaryfor the hydrodynamic modeling.5.3.2 Generation of channel bed surfacesDigital elevation models (DEMs) of the study areas were generated from the topographic sur-vey data. The spatial resolution of these data ranges from 0.5 to 1.5 points per m2, with densitylocally adjusted to ensure that primary topographic features such as bars, pools, or lobes ofsediment were captured. Uncertainty in survey coordinates is low, with calculated closing tra-verse errors ranging from 0.01-0.04 m in the x and y coordinates, and 0.005 - 0.03 m in thez (elevation) coordinate. In context, these errors are less than the median grain size (D50) inall study areas. From the topographic survey data, DEMs of 10 cm resolution were generatedusing the ”Raster” package within the R programming language (RCoreTeam, 2017), resultingin a total of 340 unique topographic surfaces.To characterize variability of study area morphology over time, width (Wb), relative bedelevation (Zr), and topographic roughness (Tr) were calculated for each year of data for eachstudy area. Zr is calculated as the minimum mean bed elevation over the period of recordfor each study area subtracted from the mean annual elevation, and is essentially a measurethickness of sediment. Topographic roughness is defined here as the ratio of the 3D (i.e. slopearea) to 2D channel bed area, where 3D area (A3D) is calculated as:A3D =A2Dcosθ(5.1)where A2D is the two dimensional area of a raster pixel, and θ is the slope angle of a givenpoint on a 2D surface. To avoid the influence of steep channel banks on the determination ofTr, the calculation was performed only on bed areas. All values of Tr are equal to or greaterthan 1, where a value of 1 indicates a perfectly flat surface, and greater values indicate morecomplex surfaces.5.3.3 Hydrodynamic modelingTo assess the influence of variable channel morphology on aquatic habitat, depths and veloci-ties in the study areas were simulated with the Nays2DH hydrodynamic model (Nelson et al.,2016). As with other 2D hydrodynamic models (e.g. River 2D, see Steffler and Blackburn,2002), Nays2DH makes use of depth-averaged Navier Stokes shallow water equations andorthogonal curvilinear grid systems for calculations. Momentum advection is calculated us-ing the cubic interpolated pseudo-particle (CIP) method. Input boundary conditions include achannel bed surface and downstream water surface elevation for a given discharge. In addition,82Table 5.1: Flow levels scaled by contributing area relative to weir near river mouth (inliters per second).StudyAreaAda(km2)400%MAD100%MAD40%MAD20%MAD10%MAD5%MAD3%MADSA-2 10.1 3310.1 827.5 331.0 165.5 82.8 41.4 24.8SA-3 9.3 3078.3 769.6 307.8 153.9 77.0 38.5 23.1SA-4 8.8 2906.2 726.5 290.6 145.3 72.7 36.3 21.8SA-5 8.1 2677.7 669.4 267.8 133.9 66.9 33.5 20.1SA-6 7.8 2594.9 648.7 259.5 129.7 64.9 32.4 19.5SA-7 7.6 2522.1 630.5 252.2 126.1 63.1 31.5 18.9SA-8 7.4 2439.4 609.8 243.9 122.0 61.0 30.5 18.3SA-9 3.8 1260.8 315.2 126.1 63.0 31.5 15.8 9.5cContributing area upstream of approximate study area midpointa roughness value (Manning’s n) characterizing the channel bed must be supplied. All simu-lations were run through the International Rivers Cooperative Interface (IRIC). Further detailsof Nays2DH and IRIC can be found in Nelson et al. (2016). Detailed information regardingsimulation conditions is found in Appendix C.Simulations were run for all years of survey data in each of the eight study areas, and forseven flow levels ranging from 3% to 400% MAD (see Table 5.1), resulting in nearly 2500simulations. These flow levels encompass nearly the full range of flow conditions experiencedover the seasonal May 1st to September 30th dry period at Carnation Creek, and are typical offlows experienced by many streams in coastal British Columbia. Additionally, several of theselected flow levels (e.g. 10% MAD) are used for management and conservation purposes todefine thresholds below which ecological functionality is considered impacted (Bradford andHeinonen, 2008). To extrapolate discharge values collected near the river mouth to dischargein the study areas, streamflow was scaled by catchment contributing area upstream of eachstudy area, and this approach confirmed by checking values from three additional hydrometricstations upstream. These flows are shown in Table 5.1. Manning’s n values of 0.04 were usedfor all simulations, model runtimes ranged from 800 to 1000 seconds, and simulation solvertimesteps were 0.01s, the default value for the model. Output data were saved in 10 secondsteps. For each simulation, downstream water surface elevation was determined from empiricalrating curves generated from historical data.Given the large number of simulations run with varying input conditions, it was not possibleto individually calibrate simulations. Instead, best estimates of initial input values were used,and a detailed assessment of model performance through comparison with field depth andvelocity measurements was undertaken. Point measurements were taken during a range of flowconditions in the summer and fall of 2017 in three study areas (SA-3, SA-6, and SA-7) with aSonTek FlowTracker 1 acoustic Doppler velocimeter (ADV). In addition, study-area average83depth and width values were evaluated against wetted channel boundaries delineated duringtopographic surveys. Finally, the influence of reach length on model output was considered,because some of the study areas are comparatively short. Details of this performance evaluationare found in Appendix C and Figure C.1.Overall, the model performed well when evaluated against field measured depths and ve-locities. Prediction intervals (95%) for mean depth was ± 0.12 m, velocity ± 0.23 m/s, andarea-averaged comparisons resulted in good agreement. Although depth is predicted well,the model has a tendency to slightly overpredict low velocities and underpredict higher ones.While the model performed well when evaluated against field data, authors are aware that it isnot possible to evaluate the model under all conditions for which it was applied in this study.However, given the objective of parsing broad temporal and spatial trends from the data, thesedata are considered suitable for this purpose.5.3.4 Variable selection and definitionThree flow-based variables were used to evaluate the influence of changing channel morphol-ogy on aquatic habitat conditions: (i) pool area (Ap), (ii) wetted area with high water velocities(Ahv), and (iii) pool area with wood cover (Apwc). Ap is defined as all wetted area of depth> 0.1 m, with velocities < 0.6 m/s. Ahv is defined as all wetted area of velocity > 0.6 m/sregardless of depth. Apwc is defined as Ap overlain by in-stream wood pieces. These variableswere selected to reflect the best compromise between known habitat preferences for juvenilecoho salmonids occupying the channel during the summer low-flow period, and collectively,these definitions encompass a relatively broad range of conditions both optimal and marginal(Bjornn and Reiser, 1991). While the focus of these three variables is on coho juveniles, simi-lar depth/velocity combinations could be used to illustrate the role of channel form in shapinghabitat for other species.5.3.5 Analysis of model outputFor each simulation, raw data output was converted to depth and velocity raster surfacesthrough Inverse Distance Weighting interpolation using the R programming language (RCoreTeam,2017). Example depth and velocity output for two flow levels in study area SA-2 is shown inFigure 5.2. Once these rasters were generated, areas were selected which fell within the cri-teria of a habitat variable of interest. For Apwc, wetted areas falling within the criteria for Apwere cropped by shapefiles delineating wood pieces. Summary information was then calcu-lated from the cropped area, resulting in an area value for each variable, flow level, and year.All analysis was done with the R programming language (RCoreTeam, 2017), using packages”Raster” (Hjimans, 2019) and ”RGDAL” (Bivand, 2019).The quantity of habitat available at each flow level was evaluated by comparing area for84Depth (m)Velocity (m/s)(a)(c)(b)(d)Figure 5.2: Example model output of depth (a and b) and velocity (c and d) for SA-2 fortwo modeled flows in 2015: 5% MAD (a and c), and 100% MAD (b and d). Flowis from right to left. Wood pieces are shown as brown polygons.each variable (scaled to active channel area) against discharge in each study area, resultingin discharge-habitat relationships for each study area and year. These curves were linear, loglinear, or in the case of Apwc, usually a 2nd degree polynomial function.From these curves and the historical flow record, the total cumulative quantity of habitatavailable over each May 1st - to Sept 30th period, for each variable (denoted as Cp, Chv, andCpwc for pool area, high velocity area, and pool area with wood cover, respectively) was cal-culated. This enables the evaluation of the relative significance of hydrology and morphologyfor varying habitat. The historical daily flow record for each season was used to predict dailyhabitat areas from the empirical habitat-discharge relationships determined above, and the cu-mulative sum of each daily habitat area taken over the 152-day dry season period for each year(see Figure C.2 for a visual description of the approach). To avoid problems with predictingoutside of the range of values used to fit the habitat-discharge relationships, flows > 400%MAD in the annual low-flow series were replaced with the 400% MAD value. This affected30 daily flow values over the 45 year series, or about 0.4% of the record.To evaluate the effect of channel morphology relative to discharge for explaining habitat85variability, a second set of seasonal cumulative habitat values were calculated using the meanrelationship between habitat area and flow over the entire period of record. Therefore, for eachyear, two habitat areas were predicted: (a) cumulative area from the annually varying flow-habitat relationship, and (b) cumulative area predicted from the mean, unvarying habitat-flowrelationship. The resulting difference between values therefore reflects the difference in theshape of the habitat-flow relationship through time, which is driven by channel morphologyand, in the case of Apwc, abundance and location of wood pieces. This approach allows for theintegration of both morphological and hydrological variability in the assessment of availablehabitat through time, similar to the approach used by Fabris et al. (2017), and is analogous tothe calculation of ”growing degree days”.5.4 Results5.4.1 Variability in channel morphologyThe eight study areas display variable patterns of bankfull width (Wb), relative bed elevation(Zr), and topographic roughness (Tr) over the period of record (Figure 5.3). All study areasdisplay a relatively high degree of geomorphic variability, but they can be broadly groupedinto sites displaying small-scale change (SA-2, SA-3, and SA-9), and more dynamic sites (SA-4 to SA-8). All sites exhibit values of Tr between 1.02 and 1.1.In the relatively stable study areas, Wb and Zr show only small-scale fluctuations, with theexception of bed elevations in SA-9, which drop abruptly near 1990. Topographic roughnessis also fairly constant, but with higher values in SA-3 near 1980, and elevated values in SA-9after 2008: the high values in SA-9 may correspond to a shift in survey methods over the studyperiod. Cross-sectional surveys in this study area did not capture complex topography as wellas in other sites due to the presence of three channel bends, therefore the switch to total stationsurveys offered a greater improvement in survey quality. The dynamic study areas show Wband Zr varying by up to a factor of two, and also exhibit variability in Tr. In particular, highvariability is present in SA-5 and SA-8.Variability in Wb, Zr, and Tr is also evaluated against distance from major sediment sourceareas, specifically the downstream limit of the confined canyon reach (see Figure 2.4) for studyareas SA-2 to SA-8, and the downstream limit of confined headwater reaches for SA-9. Stan-dard deviation of Wb and Zr (scaled to study area means), and Tr are shown in Figure 5.4. Asignificant (p = 0.036) trend of declining variability with distance is evident for Wb, but nosignificant trend is apparent for Zr and Tr, though they both appear to decline slightly down-stream.All study areas contain some degree of wood loading, but SA-5 and SA-8 also containedlogjams over a portion of the study period. In the case of SA-5 a logjam formed in 1991,86SA 2 SA 3SA 4 SA 5SA 6 SA 7SA 8ZrWbTrSA 911.041.0811.041.0811.041.0811.041.08210210210210YearWb(m/m) and Zr(m)T r(m2 /m2 )1970 1980 1990 2000 20101970 1980 1990 2000 2010Figure 5.3: Plots of bed elevation relative to mean bank height, scaled width, and topo-graphic roughness for the eight study areas. Width (Wb) is scaled by the mean ofeach study area, while relative bed elevation (Zr) is not scaled. Topographic rough-ness (Tr) is calculated as the ratio of 3D/2D surface areas of each study area, foreach year.87ZrWbTr10. from C–H coupling (m)σZrand scaled WbσTr500 250015001000R2 = 0.54–0.092(log(dist))+0.72Figure 5.4: Plots of the standard deviation of channel width, relative bed elevation, andtopographic roughness relative to the position of a study area from major sedimentsource areas. The dashed line indicates a statistically significant relationship (p< 0.05), with a negative slope. A negative but non-significant slope is found fortopographic roughness.resulting in accumulation of sediment upstream, and leading to a major avulsion in 1994. InSA-8, a channel-spanning logjam formed after debris flows entered the channel a short distanceupstream, and resulted in major widening and aggradation until the channel cut around thelogjam in 1989. In both study areas, the temporal evolution of Wb, Zr, and Tr partly reflects theinfluence of logjams.5.4.2 Variability of modeled habitatModeled pool area (Ap), high water velocity area (Ahv), and pool area with wood cover (Apwc)vary over time in all study areas. Example time series of a geomorphically stable site (SA-2),an unstable site (SA-8), and all study areas grouped at all flow levels are shown in Figure 5.5.Similar patterns of temporal variability between flow levels are observed for Ap in SA-2,with peaks in habitat area near 1980, 1990, and 2005 for the highest three flows, and a long-term trend of decline at all levels. For the same variable in SA-8, similar patterns are only seenin the higher five flow levels, with 3% and 5% MAD flows showing relatively muted changeover time. The year-to-year variability is overall much greater in SA-8 than SA-2, with theexception of the lowest flow levels. When results from all sites are combined, the pattern ofAp through time is again similar between flow levels, and variability over time is less than inthe individual study areas. There is a slight decrease in habitat in the late 1980s through to the8800.20.40.600.20.40.635102040100400% MADSA 2 SA 8 SA 2 - SA 9YearApwc(m2 /m2 )Ahv(m2 /m2 )Ap(m2 /m2 )00.0250.0500.0751970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010Figure 5.5: Time series of the three modeled habitat variables with areas scaled to ac-tive channel area for SA-2, SA-8, and all study areas combined. Plots have beensmoothed with a loess function of span 1.7 for clarity.1990s. Overall, between 10% and 50% of the active channel area consists of Ap.High-velocity wetted area (Ahv) is found in large quantities only at flow levels above 40%MAD, but at 400% MAD can occupy close to 50% of the channel area. In SA-2, a spike inAhv is seen in the mid-1980s at 400% MAD, and from this point there is a slight upward trendthrough time. At lower flow levels, Ahv remains relatively constant after 1985. Contrary to themid-1980s spike in SA-2, in SA-8 there is a major decrease near 1980 followed by an increasein the mid-1990’s for 400% MAD, and similar variability through time at lower flow levels asfor SA-2. When all sites are grouped, only small areas of Ahv are present below 40% MAD.While a decrease followed by increase in Ahv is evident at 400% MAD, little long-term changeis observed at lower flow levels.Overall, pool area with wood cover (Apwc) typically represents 2% - 7% of all active chan-nel area, and displays the greatest variability through time. In SA-2, a pronounced decrease is89observed through time at all flow levels. In SA-8, a major increase in Apwc occurs from 1973to 1984, followed by a consistent decrease by 1994, after which Apwc stays comparatively low,but increases slightly. For both SA-2 and SA-8, gains in Apwc provided at increasing flow levelsfalls through time. When all study areas are combined, Apwc declined at a similar rate at mostflow levels until the early 2000s, when a temporary increase is apparent.Notable differences in scaled habitat are found between study areas. A direct comparisonbetween all study areas for two example flow levels, 10% and 100% MAD, is shown in Figure5.6. At 10% MAD, SA-2 and SA-8 exhibit more Ap than other study areas, especially SA-9 and SA-7. The relationship between study areas changes at higher flows levels, where at100% MAD, SA-2 and SA-3 posses the greatest Ap, while SA-5, SA-6 and SA-9 have similarquantities. SA-7 continues to contain relatively little Ap at the higher flow level, and SA-4 isbelow-average.High-velocity area Ahv is essentially absent in all sites at 10% MAD, with the exception ofa few years in SA-5. At 100% MAD, SA-2 again has the largest Ahv area relative to the activechannel area (but is similar to SA-5 to SA-7), while SA-3, SA-8, and SA-9 contain the least.The remaining sites are relatively similar, even though SA-6 has several years of elevated Ahv.Pool area with wood cover (Apwc) shows the greatest variability between study areas. SA-2and SA-3 show the elevated habitat at 10% MAD, while SA-5 and SA-6 show very little Apwc.At 100% MAD, SA-3 and SA-9 have the greatest Apwc, followed by SA-2. At both flow levels,there is a general decrease in Apwc in the upstream direction to SA-8. Aside from SA-3 andSA-9 which increase by nearly 50% and more than 100%, respectively, most sites gain lessthan 30% Apwc at 100% MAD relative to 10% MAD.5.4.3 Habitat by flow levelExamining the area of habitat available at the seven modeled flow levels reveals distinct rela-tionships for the three variables of interest. The connection between flow level and modeledhabitat is shown in Figure 5.7, again for study areas SA-2, SA-8, and all sites grouped. Ex-amining flow level relative to Ap, both SA-2 and SA-8 show increases in habitat area to 100%MAD, after which Ap is reduced. Across all sites, the relationship is similar as for SA-2 andSA-8, but with a slighter decrease in Ap between 100% MAD and 400% MAD.Area of Ahv by flow level shows a different relationship: below 40% MAD there is relativelylittle Ahv across all study areas individually and combined (Figure 5.7). Above 40% MAD, Ahvincreases rapidly with increasing discharge consistently across study areas. By 400% MADover 30% of the channel contains Ahv on average.The relationship between Apwc and flow level is the most variable, with wide interquartileranges at all flow levels relative to Ap and Ahv. In SA-2, there is a gradual increase in Apwc to100% MAD, followed by a decrease to 400% MAD, the flow level at which Apwc is lowest. In900.100.202 3 4 5 6 7 8 9 2-9 2 3 4 5 6 7 8 9 2-92 3 4 5 6 7 8 9 2-9 2 3 4 5 6 7 8 9 2-92 3 4 5 6 7 8 9 2-9 2 3 4 5 6 7 8 9 2- MAD 100% MADStudy area (SA)(a) (b)(c) (d)(e) (f)Apwc(m2 /m2 )Ahv(m2 /m2 )Ap(m2 /m2 )Figure 5.6: Comparison of habitat area scaled to active channel area between all studyareas for 10% and 100% MAD flows. Subplots (a) and (b) show pool area (Ap), (c)and (d) area of high velocity flow (Ahv), while (e) and (f) pool area with wood cover(Apwc).SA-8, a broadly similar pattern exists, but with a median relationship of much less Apwc withflow than for SA-2, and which is skewed toward the lower boxplot range. Across all sites, apattern fairly similar to that of Ap emerges, but with greater variability in habitat for a givenflow level.5.4.4 Role of hydrological relative to morphological variabilityBoth flow conditions and morphological variability appear to influence habitat availability overthe period of record. Flow conditions vary substantially year to year, with cumulative discharge91000.20.40.600.0250.0500.0750.1000.20.40.6SA 2 SA 83 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 4003 5 10 20 40 100 400Flow level (% MAD)SA 2 - 9Apwc(m2 /m2 )Ahv(m2 /m2 )Ap(m2 /m2 )(a)(d)(g)(b)(e)(h)(c)(f)(i)Figure 5.7: Boxplots displaying scaled habitat area relative to each modeled flow level.Results for SA-2, SA-8, and all reaches grouped are shown.over the May 1st to September 30th period varying by more than a factor of five from thedriest (1998) to the wettest (1997) year. Annual May 1st to September 30th flow conditionssegmented into (a) time by flow level and (b) cumulative discharge by flow level are shown inFigure 5.8. By time, the majority of daily flow values are at levels below 10% MAD, howevera majority of outflow is contributed from infrequent flows above 40% MAD. Typically, thisresults from a few wet days over the dry season that are influencing the total seasonal waterexport. For example, 1997, the wettest year on record during the dry season, experienced fewerthan 10 days above 100% MAD, but these contributed more than four times the water of allflows over the same period in 1998.Cumulative seasonal habitat for all study areas combined was segmented into three sea-sonal sub-periods: May 1st to June 15th, June 16th to July 31st, and Aug 1st to Sept. 15th(Figure 5.9). Different temporal patterns in annual cumulative habitat are apparent betweenthe variables. Cumulative pool area (Cp, Figure 5.9a) shows a period of elevated and con-sistent values from 1973 to 1981, followed by a period of greater variability and a return to92Figure 5.8: (a) Annual days during which each flow level occurs; (b) total dry seasondischarge, segmented into relative water contributions by flow level.higher values in 2007, followed by a decline. Sightly less than half of the cumulative habitatoccurs before June 15th. Approximately 70% of variability in Cp can be explained by seasonaldischarge, and 30% by channel morphology changing through time.Cumulative high velocity area (Chv, Figure 5.9b) is more variable through time than Cp(Figure 5.9a). As Ahv is largely absent at flows less than 40% MAD, the comparatively wetterMay 1st to June 15th period dominates the annual values of Chv most years. Slightly morevariability (75%) is explained by the flow record than for Cp, and variance is low during low-flow years.The habitat-flow relationships for Apwc are a function of both morphology and wood load-ing, and thus add additional element of complexity to the temporal response. The pattern ofcumulative pool area with wood cover (Cpwc, Figure 5.9c) more closely resembles Cp thanChv. However, a pronounced decline in Cpwc is apparent from the mid-1970s to mid-1990s,after which point Cpwc remains relatively stable to 2009, when it declines further. Slightly lessthan 50% of Cpwc occurs prior to June 15th. Hydrological variability explains only 49% of thevariance in Cpwc, notably less than Cp and Chv.The relative contributions of discharge and channel form in the individual study areas isshown in Table 5.2. Study areas show more variability individually than when combined: for93hvpwcpwcpwchvhvFigure 5.9: Annual cumulative habitat series for (a) pool area, (b) high-velocity wettedarea, and (c) pool areas with wood cover, segmented into three seasonal sub-periodsof May 1st to June 15, June 16 to July 31, and Aug. 1st to Sept. 15th. Valuesare for all sites grouped, with scaled values added. Scatterplots to the right of eachseries illustrate the difference between cumulative habitat predicted with the meanhabitat-discharge relationship (x axis) and values predicted with the annually vary-ing habitat-flow relationships (y axis).94Table 5.2: Variance explained by streamflow relative to channel form and wood loadingVariable SA-2 SA-3 SA-4 SA-5 SA-6 SA-7 SA-8 SA-9Pool area 0.53 0.62 0.41 0.51 0.23 0.51 0.39 0.61High velocity area 0.74 0.60 0.61 0.43 0.78 0.69 0.53 0.83Pool area with structural cover 0.20 0.03 0.14 0.16 0.27 0.30 0.25 0.21Cp, SA-3 and SA-9 have the greatest variability explained by the flow record, and SA-6 andSA-8 the least. For Chv SA-9 is again the most flow-driven, though SA-2 and SA-6 also havea high degree of variability explained by discharge. Cpwc variance explained by flow is low tovery low across all sites, but notably so in SA-3 to SA-5.The influence of channel form and morphological variability on cumulative seasonal habi-tat area can be further examined by evaluating standard deviation of the three morphologyvariables against variability explained by channel morphology and wood (Figure 5.10). Weakpositive relationships are observed between morphology - driven variability in Cp and Wb, andpositive relationships between Chv and all three geomorphic variables are apparent. A signifi-cant positive relationship between the geomorphic contribution to Chv variance and topographicroughness was found (p = 0.0047), linking channel form variability to variability in aquatichabitat. No statistically significant relationship was found between Cpwc and any of the threemorphological variables.5.4.5 Scale considerationsUsing these results, it is possible to evaluate whether temporal variability in habitat in studyareas translate to larger scale changes in the system. It is clear that the magnitude of variabilityis reduced when results of all study areas are grouped in relation to variability in the individualstudy areas (see Figures 5.5, 5.6, and 5.7). This effect is most pronounced when examiningApwc, the spread of which drops dramatically when grouped variability is examined. Variancein cumulative seasonal habitat explained by discharge (Table 5.2) is also less for individualreaches than for reaches grouped, implying that as the spatial scale of interest increases, therelative role of hydrology in explaining habitat variability also increases.It is also possible to estimate the record length needed to capture key elements of observedhabitat variability through examination of changing standard error of the mean (SE) relativeto record length. Standard error of Ap plotted against record length (Figure 5.11a) shows theclearest inflection point after five to ten years, with a sharp drop in both median, 75th percentileand 90th percentile SE. Standard error in Ahv (Figure 5.11b) shows a more complex pattern,dominated by large-scale variability in the 1990s in some sites. There is a decrease in me-dian and 75th percentile SE, with median reaching stable values after five years but increasing95Pvar morphology0.75 10.500.2500246σ W  b00.20.40.6σZr0.010.0080.0060.0120.014σ Tr(a)(b)(c)ApAhvApwcFigure 5.10: Relationships between standard deviation of (a) channel width, (b) relativebed elevation, and (c) topographic roughness and variance in cumulative seasonalhabitat explained by channel morphology and wood. Fit lines with R2 valuesshown are significant, while other fit lines show positive but non-signficant trends(p = 0.05.)9600.10.20.3σ Ap00.20.40.6σ Ahv00 10 20 30 400.20.40.6Record length (yrs)σ ApwcMedianP10 and P90P25 to P75 range(a)(b)(c)Figure 5.11: Standard error of the mean as record length increases for (a) pool habitat,(b) high velocity wetted area, and (c) pool habitat with structural cover time series.All flow levels are shown for all study areas, scaled by their respective means priorto plotting. For each record length value, there are 56 values of standard error fromwhich the percentile ranges are determined.again after 20, but 75th percentile values take approximately 20 years to reach a minimum,after which they increase. Standard error in median and 75th percentile Apwc series (Figure5.11c) gradually decreases to a minimum near 20 years, but the 90th percentile series reachesa minimum at n = record length. Collectively, these results suggest that a very long record isneeded to capture the dominant temporal variability of habitat through time.975.5 Discussion5.5.1 Evaluating the landscape-habitat linkageResults presented above provide support for the catchment-habitat linkages outlined in Figure5.1, with morphological variability decreasing with increasing distance from sediment supply,and aquatic habitat partly regulated by this variability.In Carnation Creek, study areas SA-2, SA-3, and SA-9 are most likely to represent Figure5.1a, while other study areas, closer to sediment sources, are more likely to reflect linkages in5.1b. Patterns of variability in channel morphology clearly follow this organization in termsof width variations, but this trend is weaker for Tr and is absent for Zr. Since Tr incorporatesfeatures occurring at a smaller spatial scale than the mean Wb and Zr values, wood pieces, bankroughness or bedrock outcrops can have a strong influence on Tr. While wood pieces (and inparticular logjams) are known to strongly influence Wb (Montgomery et al., 2003), Zr appearsto be particularly affected by logjams in study area SA-5 and SA-8. However, changes to Zrmay also be a product of processes acting beyond the study areas, such as degradation of amajor sediment wedge upstream.Overall, the positive relationships shown in Figure 5.10 also support the theory of morphology-driven variability in habitat. However, Apwc does not show a clear trend with distance fromsediment sources, in spite of debris flows often contributing a large fraction of wood loads tomountain streams (Hassan et al., 2016). This may be a product of the trapping of wood piecesin confined channel reaches upstream of the study areas, and overall low wood mobility inmountain streams (Benda and Bigelow, 2014).Previous authors (see review in Lapointe, 2012) have identified general relationships be-tween ecological processes and catchment process in riverine systems. For example, Frissellet al. (1986) states how landslides might impact channels at the reach-scale through alteringsediment storage units, while Montgomery (1999) attempts to connect geomorphic process do-mains to disruption of the River Continuum Concept (Vannote et al., 1980). Thompson and Lee(2000) and Pess et al. (2002) empirically link abundance of juvenile salmonids to landscape at-tributes, including channel-hillslope coupling. Collectively, work here reinforces these modelsand findings, provides direct evidence of a process-based link between sediment supply dy-namics and depth and velocity attributes important for fish, and can serve as a useful means toevaluate habitat conditions partly on the basis of how sediment supply is likely to vary throughtime in a catchment.985.5.2 Spatial and temporal variability of modeled habitatBy examining modeled habitat at constant flow levels ranging from very low (3% MAD)to moderate (400% MAD), notable morphology-driven variability through time and betweenstudy areas is observed. The range of variability, sometimes up to a factor of three for Apwc, issignificant for planning and design of restoration projects, allocating streamflow in managedsystems, and potentially for understanding variations in populations of juvenile salmonids.In Carnation Creek, variation in channel morphology is largely driven by sediment sup-ply conditions to the stream channel, and aggradation-degradation cycles have been previouslyidentified in the catchment (see Chapter 3 and Hassan et al., 2008). For Ap and Ahv, twopatterns of variability are observed which contrast study sites impacted by hillslope-derivedsediment from those less affected by colluvial input. Sites SA-2 and SA-8 represent the ex-treme examples, as SA-8 is not far downstream from major sediment sources, while SA-2is located more than 3 km from any notable hillslope-channel coupling. While the effect ofsupply in SA-8 is modified and probably exacerbated by the concurrent formation of a majorlogjam, aggradation and channel widening was observed in all reaches from SA-4 to SA-8 (seeFigure 5.3). During the period of maximum aggradation and widening (late 1990s), Ap andAhv are both at a minimum, indicating low habitat availability relative to active channel area.This response of widening with loss of pools is similar to that observed following colluvialsediment input in other systems from both natural sediment additions (e.g. Madej and Ozaki,1996; Hoffman and Gabet, 2007), and those resulting from non-natural sources (Major et al.,2017).Apwc shows the most spatial and temporal variability, diversity which is driven in part bymorphology but also by wood loading in the system. The conditions supplying and removingwood to Carnation Creek are complex, as the riparian zone between 1.3 and 3.1 km from theriver mouth was logged in the 1970s, effectively reducing the natural wood supply to much ofthe stream channel. Downstream of the 1.3 km point, a variable width buffer was retained nextto the channel, but in places this is sufficiently narrow to restrict input, such as near SA-2. Thepersistent decline in pool habitat with wood cover, even when all reaches are averaged, is thuslikely a result of a reduction of wood supply to the system from the riparian zone.There has been little work which quantitatively assesses the temporal variability of aquatichabitat from empirical data. Temporal patterns in ”Relative Available Habitat” (a velocity-based criteria) in Fabris et al. (2017) show maximum annual variation of less than 50%, whichis entirely driven by streamflow given their assumption of a static channel bed. Carnation Creekis a channel with dynamic supply conditions at the multi-decadal timescale, and therefore pos-sesses greater variability in channel structure through time. Previous work examining historicalcoho salmon and other salmonid populations in the context of habitat variables described herehas not always been able to discern a clear association between flow-driven variability, habitat,99and population dynamics (Soulsby et al., 2012); it is possible that morphology-driven variabil-ity can help to partly explain observed patterns in fish populations.5.5.3 Habitat by flow levelResults of this study illustrate clear relationships between flow level and modeled habitatthrough time, with associated tradeoffs between low-velocity pool area, and higher velocityareas less suitable for juvenile salmonids (see Figure 5.7). While there are differences betweenstudy areas in the parameters describing flow-level - habitat area relationships, the shape ofthe relationships are generally consistent between sites, with maximum habitat area achievedbetween 40% and 100% MAD for Ap and Apwc, but increasing area for Ahv above these levels.Principles of channel geometry broadly dictate relationships between depth, velocity, andwetted width (Leopold and Maddock, 1953), but smaller-scale topographic features are clearlyimportant for influencing hydraulics. At 400% MAD, between 50% and 90% of the activechannel area is typically wetted, but a tradeoff between Ap and Ahv occurs at 30% to 50% ofthe channel and 40% MAD, the point at which Ahv rapidly increases with increasing discharge.These findings indicate that as flows approach a level where much of the channel is wetted,additional flows mainly increase velocity and push thresholds outside of the criteria for poolhabitat typically preferred by salmonids.The flow-habitat relationships (Figure 5.7) for Apwc show substantial variability in somesites and years, especially SA-8. While year-to-year variability is large, there is comparativelylittle difference between flow levels for Apwc, with only minor gains in habitat between 3% and100% MAD in comparison to Ap. The comparatively flat relationship between Apwc and flowlevel implies that wood is often located in regions providing good pool habitat at lower flows,but at higher flows, velocity increases and habitat falls outside the Ap criteria.Assumed relationships between habitat quantity and flow level are often used to manageand protect aquatic habitats while providing opportunities for water allocations for other uses.In many regions, threshold-based approaches to water management simplify the conservationof water for aquatic ecosystems. For instance, flow levels from 10% to 100% MAD maybe used depending on the habitat conditions and aquatic species of interest (Bradford andHeinonen, 2008). The above results suggest that the habitat-flow relationships from whichthese values are implicitly based may shift through time, with potential implications for waterallocation decisions.5.5.4 Importance of morphological variability for habitat changeBy applying discharge-habitat relationships to historical flow records, these results indicate thatdynamic channel morphology and wood loading are responsible for substantial variability intime series of cumulative habitat. There is a relationship with a positive slope between change100in channel morphology and the contribution of variation in habitat that can be attributed to dy-namic channel form. Given that habitat is often considered primarily a function of streamflow,this finding is important for understanding how hydrology and channel form interact to cre-ate aspects of aquatic habitat in fluvial systems. Additionally, these findings lead to a greaterability to assess the impact of disturbances related to sediment input on habitat through time.This study indicates that the strongest relationships between habitat variability and geo-morphic change is with topographic roughness (Tr), rather than bankfull width (Wb) or relativebed elevation (Zr). Tr is able to capture many aspects of channel form which ultimately lead todiverse habitats; systems which possess low Tr are unlikely to have a great diversity of bed ele-vations, and most often lack substantial pool habitat. This is the case in SA-8 during the periodof maximum aggradation in the 1980s, where a low-relief bed has formed, providing little poolhabitat area. These findings are also supported by the observed variability over time (Figure5.5), which clearly demonstrates the effect of changing channel form on habitat availability.Hydrological variability is least important when considering pool area with wood cover, asvariability in wood abundance adds a third element of complexity to the system. Additionally,the area that Apwc generally constitutes in all study areas is small (< 5-10% of the channel),and therefore comparatively small changes in wood loading or velocity can strongly affectthe habitat available at a given flow. Hydrological variability explains most variation for Ahv;given that more of the wetted channel is apt to be composed of higher-velocity area as dischargeincreases, it is reasonable to assume that the effect of channel morphology on habitat variabilitywill diminish as discharge increases.The spatial scale of interest will influence the specific contribution that channel form hason habitat quantity over time. In Carnation Creek, individual study areas display greater overallvariability in comparison to their grouped sum. If the geomorphic variability is driven by localprocesses, such as bank erosion, logjam formation, or bar formation, the net variability ofthese features, when averaged over a larger area, will decrease. However, if change is drivenby larger scale processes, such as major colluvial sediment input, then variability through timemay be more widespread.Given knowledge of geomorphic variability in the system, it is apparent that a data recordspanning even four decades is not sufficient to capture key aspects of variability in habitatthrough time (Figure 5.11). In a system dominated by episodic and unpredictable change, along data record is needed to characterize the range of possible habitat conditions in a system,depending on the frequency of sediment supply events. However, in a catchment which is ge-omorphically stable and where hydrological variability dominates, then a much shorter recordwould likely suffice. The authors acknowledge that in many cases, no geomorphic or base-line habitat data are available for other studies, but general knowledge of the sediment supplyregime can provide some indication of the likely morphological variability of a catchment.1015.5.5 Study limitationsThe approach for this study of linking long-term channel morphology data with hydrodynamicmodeling provides valuable insight into connections between channel form and habitat throughtime. However, it is important to acknowledge limitations of this approach introduced by thesimplified representation of channel bed surface, and the habitat variables selected.Topographic surveys do not represent all elements of the system which are critical to fish,and using models with these input data will miss the hydrodynamic effect of these elements.Grain-scale features, for instance, are found to be of significance for providing refuge to juve-nile salmonids during higher flows (Swales et al., 1986).A related source of additional uncertainty is tied to how wood is represented in the channel;the assumption is made that wood pieces have no effect on the flow field at the modeled flowlevels, as there are not sufficient data through the period of record to determine if they areresting on the channel bed or are suspended, even partly, above it. Though field observationsduring summer low flow conditions confirm that wood has minimal influence on the flow field,at higher modeled flows (100% and 400% MAD), some wood pieces likely influence channelhydraulics, adding spatial variability to velocity fields within study areas, and reducing averagevelocities. Similarly, pieces resting directly on the bed may provide different cover to fish thanthose suspended above the bed.The selected habitat variables were chosen to capture a range of conditions experiencedby fish in the system, but consider only hydraulic and wood-cover aspects of habitat. Manyother variables have bearing on habitat conditions, such as stream temperature, presence ofoverhanging banks, water quality, and predation (Bjornn and Reiser, 1991). Consideration ofthese variables through time will also be important for drawing conclusions related to temporalvariability of aquatic habitat conditions.5.6 ConclusionsAquatic habitat for salmonids is under increasing pressure, yet little is known about the spatialand temporal variability of habitat driven by variable channel morphology. Here, this vari-ability has been assessed using a hydrodynamic modeling approach coupled with 45 years ofchannel morphology and in-stream wood data.Overall, pool area (Ap), high-velocity area (Ahv), and pool area with wood cover (Apwc)vary substantially through time, but in different ways depending on location of these habitatswithin the watershed. Each of these variables has a distinct relationship with flow level, with atrade-off apparent between low-velocity pool area and area of high velocity flow as dischargeincreases. Variability in channel morphology and wood abundance is found to explain between25% and 50% of the variation in habitat variables through time. The magnitude of variability102in habitat differs between channel study areas, and it appears that a long record is needed tocapture the key aspects of this variability, particularly if sediment is supplied eposodically.Collectively, these findings lend support to a conceptual linkage between catchment-drivensediment supply dynamics, and temporal variability in aquatic habitat. Channel morphology isthus a key driver of habitat suitability over time in dynamic channel environments.Understanding the drivers of channel geomorphic variability in a given system will be ofkey importance in assessing the likelihood of evolving channel morphology being a significantfactor for varying habitat. If a system is geomorphically stable, then hydrology is going tocontribute more to the availability of habitat within the system. Similarly, if one is interested inhabitat dynamics at a scale larger than the channel reach, geomorphic variability will be mostsignificant if related to a widespread event or process. At smaller scales, local processes anddynamics are likely to play a larger role.Further work is needed to refine the connection between sediment and water supply, andaquatic habitat, in particular by examining habitat variability in a highly stable system, andalso through examination of fish population data.103Chapter 6Concluding remarksThis thesis has focused on the relationships amongst sediment supply dynamics, wood abun-dance, channel morphology and aquatic habitat in forested, gravel-bed streams. This work hasbeen motivated by a need for greater understanding of how streams with episodic sedimentsupply behave as a function of the wood and sediment supply conditions, which are driven bythe landscape history and organization. Long-term field data records are necessary to capturetemporal variability in these watersheds, while a large spatial data extent is important to trackchanges along the channel. Field investigations of this scope are therefore important for char-acterising the complexity of upland gravel-bed streams, a characterisation which is a precon-dition for the development of realistic models and experiments related to channel adjustmentfrom sediment supply and wood variability.The overarching objective of this work has been to link a cascade of geomorphic processesfrom the watershed scale to the channel unit scale. These linkages connect catchment organi-zation to sediment supply dynamics, which in turn regulate channel morphology and aquatichabitat. To this end, findings in this thesis have addressed questions related to the spatial andtemporal patterns of: (a) sediment storage and channel morphology as they relate to supply;(b) in-stream large wood and wood-morphology interaction; and (c) modeled aquatic habitatfor juvenile salmonids, all over a timescale of multiple decades. A long-term field dataset col-lected in Carnation Creek, B.C. is used to address these questions. This dataset is unique inthat it contains annual channel morphology, hydrological and in-stream wood data spanning45 years, in addition to a record of colluvial sediment input.The following section (Section 6.1) presents a summary of primary research contributionsfrom this thesis. A brief discussion of some broad implications of the work relevant for fielddata collection and study design is then described in Section 6.2. Finally, a selection of possibleopportunities for future research is outlined in Section 6.3.1046.1 Summary of contributionsSediment is stored in stream channels unevenly in time and space. The spatialorganization of stored sediment is dictated by travel distance statistics, while temporalvariability is a function of proximity to episodic sediment supply and local processes.Chapter 3 presents contributions related to the spatial organization of sediment storage in Car-nation Creek, a forested, gravel-bed stream. Locations of high storage correspond to majorbars along the channel, the locations of which are associated with the median annual traveldistance of bedload sediment in the channel. Logjams appear to have a minor influence on thespatial organization of storage at large spatial scales, but their local influence on storage canbe substantial. In channel sections near the river mouth and upstream of a hillslope-coupledsection of channel, low-amplitude variability is present in sediment storage. However, tempo-ral variability in storage increases with proximity to hillslope-coupled channel regions. Thesefindings improve the state of knowledge regarding the controls on sediment storage location,abundance, and dynamics in gravel bed streams, and the possible scales and magnitudes ofchange expected as a function of channel position in a deglaciated watershed.The relation between sediment storage and export is complex and non-linear, and glacialhistory is an important high-level control on these relationships.Previous research indicates that sediment transport in rivers is not simply a function of flowconditions, but also of the availability of mobile sediment. Analysis of Carnation Creek datain Chapter 3 supports this finding, with complex patterns observed between sediment storageand transfer (i.e. export).Cycles of aggradation and degradation are apparent through time along Carnation Creek.In downstream, transport-capacity limited sections of channel, small-scale cycles are observedat the channel-unit scale, on the order of a several (5-10) years. These cycles are primarilya product of local bank erosion and bar deposition. However, in locations closer to colluvialsediment sources where fluctuations between transport- and capacity-limited states are morepronounced, the duration and magnitude of dominant aggradation and degradation cycles area function of episodic supply events. Large logjams appear capable of greatly modifying andeven forcing these cycles.Given that the cycle characteristics appear related to the position of a channel segment rel-ative to colluvial sediment sources, glacial history and resulting watershed organization appearto play an overarching role in dictating spatial and temporal storage patterns. A comparison ofaggradation and degradation cycles near the river mouth (SA-2 and SA-3) to those in a decou-pled upland section (SA-9) illustrates the effect of glacial history in Carnation Creek data.Results presented in Chapter 3 lead to a conceptual model which describes the connection105between watershed structure, sediment supply, and the temporal variability of sediment storage.This model states that temporal variability in storage is highest closest to locations of colluvialsediment input, and contrasts glaciated and non-glaciated channels.In-stream large wood, and particularly logjams, influence channel morphology at thechannel-unit to reach scale. The topographic complexity of a channel bed is related towood piece abundance, and logjams lead to major accumulation of sediment upstream,followed by channel avulsion around the jams and exponential loss of stored sediment.Research and findings in Chapter 3 indicate that logjams have impacts to channel morphologyat a range of spatial scales. Chapter 4 provides a novel characterization of spatial and tem-poral patterns of wood abundance, and, in particular, statistical relationships between woodabundance and the topographic complexity of the channel bed surface.A primary contribution of this chapter is the characterization of how sediment stored behinda logjam is eroded or otherwise removed from the active channel. A negative exponentialfunction fits the observed loss of sediment storage from behind two jams in the main channel.However, a large quantity of material stored behind jams remains in the floodplain as a productof channel avulsion. Previous studies have assumed particular decay distributions to model lossof stored sediment from jams with little field evidence, as no clear characterisation is availablein the literature.In addition to morphological impacts, the detailed Carnation Creek wood dataset enabledexamination of several characteristics relevant to wood mobility which have not been welldescribed by past research. In particular, this work contributes a better understanding of therelationship between wood piece size and persistence. Small wood pieces < 0.3 L/Wb arefound to be preferentially lost from a given channel segment over time, as are pieces orientedparallel to the primary direction of streamflow.Spatial and temporal patterns of wood abundance are a function of riparian history. Insmall channels where wood transport is relatively limited, loss of riparian supply willresult in wood load recovery times at the century scale and minimum loading at least 50years post-harvest in temperate rainforest ecosystems.The quantity of wood found in a river channel will vary through time depending on the domi-nant input sources, but in cases where riparian zone logging has taken place, a long period ofreduced wood storage is likely following timber harvesting. Chapter 4 contributes estimatedtimescales of recovery to such a disturbance. These timescales have broad geographic implica-tions given the widespread extent of riparian zone harvesting in much of the Pacific Northwest(and elsewhere) over the first two-thirds of the 20th century.The detailed data available in British Columbia characterising forest conditions coupledwith the Carnation Creek record enable a wood budget simulation approach. The analysis pre-106sented in this chapter synthesizes many data sources and uses this budget approach to estimaterecovery timescales. Modeled results and field data indicate similar reductions in wood load,collectively suggesting that the spatial organization of harvesting has relatively little impacton overall recovery time. However, the magnitude of in-stream wood loss will be larger if agreater extent of riparian zone is harvested.Chapter 4 also contributes some evidence that the channel morphology of Carnation Creekis responding to harvesting-driven wood losses. The loss of LW through time as a product ofharvesting also corresponds to widespread loss of sediment storage over the past 26 years.Modeled aquatic habitat as a function of flow depth, velocities, and structural coverfrom wood varies through time by up to a factor of ten. Temporal variability in channelmorphology and wood abundance can be equally important as hydrology in varyinghabitat through time.Channel morphology and streamflow both control critical aspects of aquatic habitat. However,variability in habitat through time stemming from changes to channel morphology is poorlyunderstood, and has important implications for the management of aquatic ecosystems. Resultsin Chapter 5 indicate that coupling a hydrodynamic model with long-term topographic channeldata is an effective method of characterising variability in modeled aquatic habitat as a functionof changes to bed topography and wood loading. Results from this chapter indicate that at low-flow levels, habitat availability can vary through time by more than a factor of two, and whenstructural cover from wood is considered as a habitat component, can vary by a factor of ten.A key contribution from Chapter 5 is evidence that morphological and LW variability trans-lates to major change in aquatic habitat availability through time. Structural components ofhabitat are found to explain as much variation through time as natural year-to-year variabilityfrom the hydrological regime. This chapter also examines issues of spatial and temporal scalein this matter: findings indicate that the importance of morphological relative to hydrologicalvariability decreases as spatial scale increases, and that at least ten years of data are necessaryto capture the bulk of temporal variability in habitat.Landscape structure and proximity to colluvial sediment sources dictate the variabilityin habitat through time and across space.A key contribution of Chapter 5 is evidence of a connection between sediment supply-drivenchannel form change and temporal variability in aquatic habitat. The conceptual model pre-sented in Chapter 5 proposes that the contribution of morphological variability to habitat vari-ability will depend on the sediment supply regime of the system and proximity to the sedimentsource. As illustrated in Chapter 3, landscape history will then have a broad impact in regulat-ing habitat variability through time.107This conceptual model is evaluated with field data, and is found to be supported in Carna-tion Creek: greater morphological variability is noted near to hillslope-coupled channel sec-tions, and change through time in habitat is related to variability in channel morphology.Examination of the relationship between discharge and habitat availability given changesin channel morphology provides novel and important contributions to the field of watershedmanagement at the channel scale. Managers can use these findings to help make decisions onwatershed activities (such as water withdrawals or forest harvesting) on the basis of a channel’ssediment supply regime.Long-term datasets are needed to capture important but infrequent formative riverprocesses.Characterising linkages between watershed-scale processes, channel change, and aquatic habi-tat is challenging, and previous studies generally focus on subsets of the complex processlinkages occurring in these environments. Numerical and physical models are also limited intheir capacity to capture the detailed and often unexpected interactions between landscape his-tory, sediment supply, channel adjustment, wood dynamics, and aspects of physical habitat. Anotable research contribution presented in Chapters 3, 4 and 5 is a demonstration of the valueof major long-term, interdisciplinary field studies which provide a holistic understanding ofwatershed and ecological processes.While such projects are challenging to undertake and maintain, the value of long-term datacollection is very high, especially in systems where many of the key formative processes oc-cur infrequently or episodically. The Carnation Creek project is particularly useful as majorepisodic sediment input did in fact occur, meaning that the adjustment of many channel vari-ables to this unpredictable event could be monitored concurrently. However, as Chapters 4 and5 indicate, the duration of some process responses suggest that even a remarkably long 45-yearrecord will miss key events and processes.6.2 Broad research implicationsThe results and discussions presented in Chapters 3 to 5 provide insight into a range of sedimentsupply-driven watershed processes which shape channel form and aquatic habitat. There are,however, additional implications of this work spanning multiple chapters which are relevant forgeneral watershed science, and highlight the benefits and drawbacks of long-term field studies.Several implications of this type are discussed below.108Duration and scale of field measurementsField data collection has historically been an expensive and labour-intensive undertaking, andselecting an appropriate project scope remains a challenge for many researchers. It is thereforeimportant to design a study which captures the key aspects of the system or processes of interestas efficiently as possible without missing important spatial and temporal information.Results from this thesis provide some insight into the relationship between record length,spatial extent, and quantity of information gained from various scales of data collection. Atradeoff is usually present between the spatial and temporal scale of the data to be collected.This tradeoff is clearly illustrated in results from Chapter 4, when the wood data collected overthe 3.0 km of channel and study area wood data records are compared. A great deal of temporalvariability is found in the study area data, driven by localized events. However, it is impossibleto know if these local occurrences are meaningful at a larger scale. The profile wood datahelps to address this question, but at a lower temporal resolution. Similarly, results presentedin Chapter 5 indicate that necessary record length will depend on the particular variable ofinterest.Information on the dominant process regimes in a system can also be used to guide thetimescale of data collection. For example, an upland river with episodic sediment supply willnecessitate a longer data record than a lowland system where temporal variability in supplyis comparatively muted. Based on data from Carnation Creek, records of different lengthscould have captured patterns of sediment storage change through time equally well in differentparts of the channel: the full record was essential for characterising sediment storage patternsin SA-6 to SA-8, but SA-2 and SA-3 and SA-9 had much of the variability described afterapproximately ten years.Suggested improvements to the design of a long-term study of watershed processesAn overarching theme of this work is that interdisciplinary, multi-organizational field studiesproduce information amounting to more than the sum of their component parts. Concurrentcollection of sediment source, channel morphology, and hydrological data allows for a greaterability to attribute cause to effect (for example, episodic sediment supply to variability in sed-iment storage and habitat). However, the Carnation Creek study was not designed with theevaluation of widespread geomorphic process as a priority.Several changes could be made to better capture key aspects of sediment supply, mobility,channel change and landscape structure. First, fewer study areas are needed than currentlypresent, but studied segments need to span a greater length of channel. Reach length should beselected in proportion to the measurement frequency: if surveys are to be conducted annually,then a study reach should be equivalent to at least the length of the median annual particle traveldistance. While this would increase the total channel area surveyed (even if the total number of109study areas decreases), improving survey methods, particularly from unmanned aerial vehicles(UAVs) could aid in more efficient data collection. Similarly, longer reaches are more likelyto capture erosion and deposition patterns in better context and eliminate the over-influence ofhighly localized processes.Second, an additional study area could be located upstream of SA-9, in a zone closer toheadwater colluvial sources. This would allow for better evaluation of the influence of water-shed structure on channel processes, replicate the study area configuration found downstreamof the canyon reach, and help clarify the role of glacial history in the watershed. Similarly,a systematic inventory of sediment sources coupled to the main channel would be a valuableasset. This inventory should include data on the grain size distribution of material delivered tothe channel, if possible.Third, detailed temporal sediment texture data of the channel bed surface and subsurfacehas been a limitation of the work conducted in this study. Annual quantification of basic grainsize parameters such as the median (D50) would be an asset. Collectively, better informationon spatial and temporal patterns of sediment texture are useful for characterising sedimentmobility, supply state, and assessing the suitability of a channel section for habitat.Data collection and database management over the long termThrough working with the database from Carnation Creek, several important lessons can becommunicated regarding the planning, implementation, and ongoing maintenance of a long-term, multidisciplinary field research project. First, consistency of data collection is critical.There is greater value in clear and consistent collection of variables through time than in ahighly detailed collection regime which is not sustainable. If a project is intended to persist fora length of time greater than that spanned by a single researcher’s career, then sustainability ofthe data collection regime must be carefully considered.The theme of consistency must also translate to the management of project databases.Carefully thought-out data assimilation, quality control, and documentation is of critical im-portance. Without good metadata, consistent labeling, and a clear database structure, finding,analyzing, and interpreting data can be very challenging. Similarly, specialized data formats,such as project files associated with specific software, may become obsolete and potentiallyunusable. Data and notation should be saved as basic text files whenever possible to max-imize long-term accessibility. Ease of data sharing and communication is also important,and database organization should remove any barriers to dissemination. Researchers in theseprojects must share their data with one another and the broader research community wheneverpossible. A centralized project database would serve well in this regard, which strives to beaccessible to others involved in the work.Finally, very few long-term datasets within old-growth, natural river systems are avail-110able. Perhaps, as these watersheds become increasingly rare, the need to characterise pro-cesses within them is also reduced. However, parsing natural processes from those introducedby human landscape disturbance is often difficult, and baseline data in systems less affected byanthropogenic disturbance continues to be in short supply.6.3 Future research opportunitiesDevelop and test an improved sediment routing model: Numerical models used to track theinput, transport and storage of bed load sediment in natural river systems are improving ascomputational power increases. However, several limitations remain in regards to the repre-sentation and behaviour of sediment stored in channel networks, and in the role of wood as atransport modulator.Work described in Chapters 3 and 4 provides new information on temporal patterns ofsediment storage in the context of sediment supply, and in the way that sediment storage andchannel adjustment occur in response to the formation of major logjams. The clear exponentialdecline in storage through time from behind jams in Chapter 4 can be coupled with stochasticjam formation strategies to mimic natural sediment-wood interactions in small streams. It ispossible that a degree of stochastic behaviour can also be applied to rates of transport as afunction of storage.While not used in this thesis, a detailed dataset in Carnation Creek stemming from mag-netic tracer particles could be used to better characterise sediment transport parameters for usein a routing model, particularly grain size-dependent step length and mobility properties.Evaluation of climate change impacts to aquatic habitat for juvenile salmonids: Changes toprecipitation and temperature regimes as a function of climate change will have widespreadimplications for the availability of aquatic habitat. However, natural variability in channelmorphology as a function of (a) sediment supply and (b) LW dynamics can potentially providea buffer to these impacts and insight into how they can be mitigated.Using a similar approach to that discussed in Chapter 5, it is possible to estimate whatmagnitude of flow reductions are needed to realize a significant shift in habitat given historicalvariability in channel morphology. Habitat-discharge relationships for all years of data can beapplied to percentile flow series, and significant differences between percentiles evaluated onthe basis of morphological variability. This approach can also be used to estimate the impactto habitat availability from reductions in streamflow from other causes.Extending habitat-morphology linkages to include fish population data: Results in Chapter 5demonstrate that the cumulative availability of habitat for juvenile salmonids varies year to yearas a function of both streamflow and channel morphology. However, other variables, such as111water temperature and quality and availability of food sources are not considered in this workyet are important for supporting juvenile salmonids. It would be useful to evaluate the natureof findings in Chapter 5 against fish population data and extend analysis to include additionalvariables. For example, total wetted area may be less important than water temperature as alimitation for fish abundance in many systems, and therefore population data would show littlecorrelation to an area-based habitat metric.Testing novel survey methods to characterise channel morphology and wood: Traditional meth-ods of collecting geomorphic data are generally expensive and time consuming. Recent devel-opments in surveying technology, principally in the form of unmanned aerial vehicles (UAVs),are proving increasingly capable tools for geomorphic data collection in rivers and streams.While some recent work has proven the utility of UAVs as data collection tools in a range ofriver environments, the focus has mainly been on channel bed topography in open-canopy sys-tems where the stream bed is not obscured. Carnation Creek presents an ideal site for testingthese tools in a smaller, densely forested system, as other overlapping datasets are availableagainst which UAV results can be evaluated. In particular, information on channel bed particlesize and channel topography can be gained, and sediment sources in the watershed can alsobe more easily identified. Perhaps most importantly, analysis of UAV-based imagery couldserve as a more rapid and reliable means of quantifying and tracking LW characteristics insmall streams. Rapid repeat surveys can be undertaken to track wood pieces and other changesimpossible to evaluate with traditional methods.112BibliographyAbbe, T., and D. R. Montgomery (1996), Large Woody Debris Jams, Channel Hydraulics andHabitat Formation in Large Rivers, Regulated Rivers: Research and Management, pp.201–221. → pages 46, 74, 78Andreoli, A., F. Comiti, and M. Lenzi (2007), Characteristics, distribution and geomorphicrole of large woody debris in a mountain stream of the Chilean Andes, Earth SurfaceProcesses and Landforms, 32, 1675–1692. → pages 1, 3, 47, 68, 75Andrews, E. D. (1983), Entrainment of gravel from naturally sorted riverbed material.,Geological Society of America Bulletin, 94(10), 1225–1231,doi:10.1130/0016-7606(1983)94〈1225:EOGFNS〉2.0.CO. → page 19Ballantyne, C. K. (2002), A general model of paraglacial landscape response, The Holocene,3, 371–376, doi:10.1191/0959683602hl553fa. → pages 2, 42BCENVCCS (2009), Manual of British Columbia Hydrometric Standards, Tech. rep., BCMinistry of Environment, Science and Information Branch, Victoria, B.C. → page 140BCVRI (2011), British Columbia Vegetation Resource Inventory, British Columbia Ministryof Forests, doi: → page 51Becker, C., and R. Genoway (1979), Evaluation of the critical thermal maximum fordetermining thermal tolerance of freshwater fish, Environmental Biology of Fishes, 12,245–256. → page 78Beecher, B., H. A. Caldwell, and S. B. DeMond (2002), Evaluation of Depth and VelocityPreferences of Juvenile Coho Salmon in Washington Streams, North American Journal ofFisheries Management, 22(July), 785–795, doi:10.1577/1548-8675(2002)0222.0.CO;2. →pages 4, 78Benda, L., and P. Bigelow (2014), On the patterns and processes of wood in northernCalifornian streams, Geomorphology, 209, 79–97, doi:10.1016/j.geomorph.2013.11.028.→ pages 4, 46, 52, 53, 68, 98Benda, L., and J. Sias (2003), A quantitative framework for evaluating the wood budget,Forest Ecology and Management2, 172, 1–16. → page 50Benda, L., P. Bigelow, and T. Worsley (2002), Recruitment of wood to streams in old-growthand second-growth redwood forests, northern California, U.S.A., Canadian Journal ofForest Research, 32, 1460–1477. → page 46113Benke, A., and J. Wallace (2003), Influence of wood on invertebrate communities in streamsand rivers, American Fisheries Society Symposium, 37, 149–177. → pages 4, 78Beschta, R. (1983), Long-term changes in channel widths of the Kowai River, Torlesse Range,New Zealand, Journal of Hydrology (New Zealand), 22(2), 112–122. → page 79Beylich, A. A., and K. Laute (2015), Geomorphology Sediment sources , spatiotemporalvariability and rates of fl uvial bedload transport in glacier-connected steep mountainvalleys in western Norway ( Erdalen and Bødalen drainage basins ), Geomorphology, 228,552–567, doi:10.1016/j.geomorph.2014.10.018. → pages 21, 42Bilby, R. E., and J. W. Ward (1991), Characteristics and Function of Large Woody Debris inStreams Draining Old-Growth , Clear-Cut , and Second-Growth Forests in SouthwesternWashington, Canadian Journal of Fisheries and Aquatic Science, 48, 2499–2509. → page46Bivand, R. (2019), Package ”rgdal”,doi: → page 84Bjornn, T., and W. Reiser (1991), Habitat Requirements of Salmonids in Streams, AmericanFisheries Society Special Publication, 19, 83–138, → pages 1, 4, 46, 78, 84, 102Bountry, J., L. Yong, and T. Randle (2013), Sediment impacts from the Savage Rapids Damremoval, Rogue River, Oregon, in The challenges of dam removal and river restoration,edited by J. De Graff and J. Evans, p. 12, → page 20Bradford, M. J., and J. S. Heinonen (2008), Low Flows, Instream Flow Needs and FishEcology in Small Streams, Canadian Water Resources Journal, 33(2), 165–180,doi:10.4296/cwrj3302165. → pages 78, 83, 100Bragg, D. C. (2000), Simulating catastrophic and individualistic large woody debrisrecruitment for a small riparian system, Ecology, 81(5), 1383–1394,doi:10.1890/0012-9658(2000)081[1383:SCAILW]2.0.CO;2. → pages 46, 75Brardinoni, F., and M. A. Hassan (2007), Glacially induced organization of channel-reachmorphology in mountain streams, Journal of Geophysical Research, 112, 18. → pages1, 2, 9Braudrick, C., and G. Grant (2001), Transport and deposition of large woody debris instreams: a flume experiment, Geomorphology, 41, 263–283. → page 68Brummer, C., and D. Montgomery (2006), Influence of coarse lag formation on the mechanicsof sediment pulse dispersion in a mountain stream, Squire Creek, North Cascades,Washington, Water Resources Research, 42, doi:→ pages 2, 20, 41Brummer, C. J., and D. R. Montgomery (2003), Downstream coarsening in headwaterchannels, Water Resources Research, 39(10), doi:10.1029/2003WR001981. → page 2114Brummer, C. J., T. B. Abbe, J. R. Sampson, and D. R. Montgomery (2006), Influence ofvertical channel change associated with wood accumulations on delineating channelmigration zones, Washington, USA, Geomorphology, 80(3-4), 295–309,doi:10.1016/j.geomorph.2006.03.002. → page 73Buffington, J. (2012), Changes in Channel Morphology over Human Timescales, in GravelBed Rivers: Processes, Tools, Environments, edited by A. R. Michael Church, PascaleBiron, pp. 433–463, Wiley, doi: → page 79Buffington, J. M., and D. R. Montgomery (1999), Effects of hydraulic roughness on surfacetextures of gravel-bed rivers, Water Resources Research, 35(11), 3507–3521. → page 46Buffington, J. M., T. E. Lisle, R. D. Woodsmith, and S. Hilton (2002), Controls on the sizeand occurrence of pools in coarse-grained forest rivers, River Research and Applications,18(6), 507–531, doi:10.1002/rra.693. → page 46Buffington, M., and D. Montgomery (1997), A systematic analysis of eight decades ofincipient motion studies, with special reference to gravel bed rivers, Water ResourcesResearch, 33, 1993–2029, doi: → page 19Caamano, D., P. Goodwin, and J. Buffington (2012), Flow structure through pool-rifflesequences and a conceptual model for their suitability in gravel-bed rivers, River Researchand Applications, 28, 377–389, doi:DOI:10.1002/rra.1463. → page 78Carlson, J., C. Andrus, and H. Froelich (1990), Woody debris, channel features, andmacroinvertebrates of streams with logged and undisturbed riparian timber in northeasternOregon, Canadian Journal of Fisheries and Aquatic Sciences1, 47, 1103–1111,doi:10.1139/f90-127. → page 46Carnie, R., D. Tonina, J. A. McKean, and D. Isaak (2016), Habitat connectivity as a metric foraquatic microhabitat quality: application to Chinook salmon spawning habitat,Ecohydrology, 9(6), 982–994, doi:10.1002/eco.1696. → page 79Cederholm, C., D. Houston, D. Cole, and W. Scarlett (1989), Fate of coho salmon(Oncorhynchus kiutch) carcasses in spawning streams, Canadian Journal of Fisheries andAquatic Science, 46, 1347–1355. → page 46Cederholm, C. J., R. E. Bilby, P. A. Bisson, T. W. Bumstead, B. R. Fransen, W. J. Scarlett, andJ. W. Ward (1997), Response of Juvenile Coho Salmon and Steelhead to Placement ofLarge Woody Debris in a Coastal Washington Stream, North American Journal of FisheriesManagement, 17(4), 947–963,doi:10.1577/1548-8675(1997)017〈0947:ROJCSA〉2.3.CO;2. → pages 4, 78Chatfield, C. (2003), The analysis of time series: an introduction, 6th ed., 352 pp., Chapmanand Hall/CRC. → pages 26, 35, 128Church, M. (2002), Geomorphic thresholds in riverine landscapes, Freshwater Biology, 47(4),541–557, doi:10.1046/j.1365-2427.2002.00919.x. → pages 1, 2115Church, M. (2006), Bed Material Transport and the Morphology of Alluvial River Channels,Annual Review of Earth and Planetary Sciences, 34(1), 325–354,doi:10.1146/ → pages 2, 78Church, M. (2010), Gravel-Bed Rivers, in Sediment cascades: an integrated approach,chap. 9, pp. 241 –269, Wiley-Blackwell, Chichester. → page 9Church, M., and J. Ryder (1972), Paraglacial Sedimentation: A Consideration of FluvialProcess Conditioned by Glaciation, GSA Bulletin, 83, 3059–3072. → pages 2, 42Church, M., M. A. Hassan, and J. F. Wolcott (1998), Stabilizing self-organized structures ingravel-bed stream channels: Field and experimental observations, Water ResourcesResearch, 34(11), 3169, doi:10.1029/98WR00484. → page 19Cienciala, P., and M. Hassan (2013), Linking spatial patterns of bed surface texture, bedmobility, and channel hydraulics in a mountain stream to potential spawning spawningsubstrate for small resident trout, Geomorphology, 197, 96–107,doi:10.1016/j.geomorph.2013.04.041. → page 79Collins, B. D., D. R. Montgomery, K. L. Fetherston, and T. B. Abbe (2012), The floodplainlarge-wood cycle hypothesis: A mechanism for the physical and biotic structuring oftemperate forested alluvial valleys in the North Pacific coastal ecoregion, Geomorphology,139-140, 460–470, doi:10.1016/j.geomorph.2011.11.011. → pages 21, 73Davidson, S., and B. Eaton (2015), Simulating riparian disturbance: Reach scale impacts onaquatic habitat in gravel bed streams, Water Resources Research, 51, 7590–7607. → pages3, 75Davidson, S. L., and B. C. Eaton (2013), Modeling channel morphodynamic response tovariations in large wood: Implications for stream rehabilitation in degraded watersheds,Geomorphology, 202, 59–73, doi:10.1016/j.geomorph.2012.10.005. → pages 3, 4, 20Dhakal, A., and Y. Alila (2003), Assessment of impacts of forestry practices on shallowlandslides in the Carnation Creek watershed (Poster). → pages 25, 27Dietrich, W., T. Dunne, N. Humphrey, and L. Reid (1982), Construction of sediment budgetsfor drainage basins, in Sediment budgets and routing in forested drainage basins:Proceedings of the symposium, p. 19, US Forest Service, Corvallis, Oregon. → page 20Dietrich, W., P. Nelson, E. Yager, J. Venditti, and M. Lamb (2005), Sediment patches,sedimen tsupply, and channel morphology, in 4th conference on River, Coastal, andEstuarine Morphodynamics, edited by G. Parker and M. Garcia, pp. 79–90, Rotterdam,Germany. → page 78Dietrich, W. E., and T. Dunne (1978), Sediment Budget for a Small Catchment inMountainous Terrain, Z. Geomorph. N. F, 191(206), 190–205. → pages 1, 5Dixon, S. (2016), A dimensionless statistical analysis of logjam form and process,Ecohydrology, 9, 1117–1129, doi:10.1002/eco.1710. → pages 48, 60116Dixon, S., and D. Sear (2014), The influence of geomorphology on large wood dynamics in alow gradient headwater stream, Water Resources Research, pp. 9194–9210,doi:10.1002/2014WR015947.Received. → pages 4, 73Eaton, B., M. Hassan, and S. Davidson (2012), Modeling wood dynamics, jam formation, andsediment storage in a gravel bed stream, Journal of Geophysical Research, 117, 18,doi: → pages 3, 40, 47, 48, 52, 53, 72, 73, 74Elgueta-Astaburuaga, M., M. Hassan, and G. Clarke (2018), The effect of episodic sedimentsupply on bedload variability and sediment mobility, Water Resources Research, 54, 17,doi: → pages 3, 19, 26Fabris, L., I. A. Malcolm, W. B. Buddendorf, K. J. Millidine, D. Tetzlaff, and C. Soulsby(2017), Hydraulic modelling of the spatial and temporal variability in Atlantic salmon parrhabitat availability in an upland stream, Science of the Total Environment, 601-602,1046–1059, doi:10.1016/j.scitotenv.2017.05.112. → pages 5, 79, 86, 99Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley (1986), A hierarchical frameworkfor stream habitat classification: Viewing streams in a watershed context, EnvironmentalManagement, 10(2), 199–214, doi:10.1007/BF01867358. → page 98Fryirs, K. (2013), (Dis)Connectivity in catchment sediment cascades: a fresh look at thesediment delivery problem, Earth Surface Processes and Landforms, 38, 30–46,doi:doi:10.1002/esp.3242. → page 2Gartner, J., W. Dade, C. Renshaw, F. Magilligan, and E. Buraas (2015), Gradients in streampower influence lateral and downstream sediment flux in floods, Geology, 43(11), 983–986,doi: → page 20Gislason, G., E. Lam, K. Gunnar, and M. Guettabi (2017), Economic Impacts of PacificSalmon Fisheries, Tech. Rep. July, GSGislason and Associates Ltd. → page 78Gomez, B., and M. Church (1989), An assessment of bed load sediment transport formulaefor gravel bed rivers, Water Resources Research, 25(6),doi: → page 19Gomi, T., R. C. Sidle, and D. N. Swanston (2004), Hydrogeomorphic linkages of sedimenttransport in headwater streams, Maybeso Experimental Forest, southeast Alaska,Hydrological Processes, 18(4), 667–683, doi:10.1002/hyp.1366. → page 40Graf, W. (1987), Late Holocene sediment storage in Canyons of the Colorado Plateau, GSABulletin, 99(2), 261–271,doi:〈261:LHSSIC〉2.0.CO;2. → page 19Gran, K. B., and J. A. Czuba (2017), Geomorphology Sediment pulse evolution and the roleof network structure, Geomorphology, 277, 17–30, doi:10.1016/j.geomorph.2015.12.015.→ page 21117Gregory, S., and P. Bisson (1996), Degradation and loss of anadromous salmonid habitat inthe Pacific Northwest, in Pacific Salmon and their Ecosystems, edited by D. J. Stouder,P. A. Bisson, and R. J. Naiman, pp. 277–314, Springer, Boston, MA, → page 78Gregory, S., K. Boyer, and A. Gurnell (2003), The ecology and management of wood in worldrivers, American Fisheries Society Symposium, 37. → page 46Gurnell, A. (2013), Wood in fluvial systems, in Treatise on Geomorphology, edited byJ. Schroder and E. Wohl, chap. 9.11, pp. 163–188, Academic Press, San Diego. → page 46Gurnell, A. M., H. Pie´gay, F. Swanson, and S. Gregory (2002), Large wood and fluvialprocesses, Freshwater Biology, 47, 601–619. → pages 1, 52, 72Hafs, A. W., L. R. Harrison, R. M. Utz, and T. Dunne (2014), Quantifying the role of woodydebris in providing bioenergetically favorable habitat for juvenile salmon, EcologicalModelling, 285, 30–38, doi:10.1016/j.ecolmodel.2014.04.015. → pages 3, 4, 78, 79Harrison, L. R., C. J. Legleiter, M. A. Wydzga, and T. Dunne (2011), Channel dynamics andhabitat development in a meandering, gravel bed river, Water Resources Research, 47(4),1–21, doi:10.1029/2009WR008926. → page 79Hartman, G., and J. Scrivener (1990), Impacts of forestry practices on a coastal streamecosystem, Carnation Creek, British Columbia, Tech. rep. → pages 8, 9, 10, 15, 27, 52Hartman, G., J. Scrivener, and M. Miles (1996), Impacts of logging in Carnation Creek, ahigh energy coastal stream in British Columbia, and their implications for restoring fishhabitat, Canadian Journal of Fisheries and Aquatic Science, 53, 237–251,doi: → pages 8, 9, 11, 15Haschenburger, J. K. (2011), Vertical mixing of gravel over a long flood series, Earth SurfaceProcesses and Landforms, 36, 1044–1058, doi: → pages11, 25, 29, 36, 40Haschenburger, J. K., and S. P. Rice (2004), Changes in woody debris and bed materialtexture in a gravel-bed channel, Geomorphology, 60(3-4), 241–267,doi:10.1016/j.geomorph.2003.08.003. → pages 3, 11, 21, 46Haschenburger, J. K., and P. Wilcock (2003), Partial transport in a natural gravel bed channel,Water Resources Research, 39(1), 1–9, doi:10.1029/2002WR001532. → page 10Hassan, M., S. Bird, D. Reid, and D. Hogan (2016), Simulated wood budgets in two mountainstreams, Geomorphology, 259, 119–133,doi: → pages1, 24, 27, 46, 47, 50, 52, 53, 74, 98Hassan, M., S. Bird, D. Reid, C. Ferrer-Boix, D. Hogan, F. Brardinoni, and S. Chartrand(2019), Variable hillslope-channel coupling and channel characteristics of forestedmountain streams in glaciated landscapes, Earth Surface Processes and Landforms, 44,736–751, doi: → pages 1, 2, 9, 20, 42, 78, 79118Hassan, M. A., M. Church, T. E. Lisle, F. Brardinoni, L. Benda, and G. E. Grant (2005a),Sediment transport and channel morphology of small, forested streams, Journal of theAmerican Water Resources Association, 97331, 853–876. → pages 1, 2, 33Hassan, M. A., D. L. Hogan, S. A. Bird, C. L. May, T. Gomi, and D. Campbell (2005b),Spatial and temporal dynamics of wood in headwater streams of the Pacific Northwest,JAWRA Journal of the American Water Resources Association, 2(41), 899–919. → pages1, 47, 48, 56, 68Hassan, M. A., B. J. Smith, D. L. Hogan, D. S. Luzi, A. E. Zimmermann, and B. C. Eaton(2008), Sediment storage and transport in coarse bed streams: scale considerations, inGravel Bed Rivers 7, vol. 11, pp. 473–496, doi:10.1016/S0928-2025(07)11137-8. → pages2, 11, 19, 29, 35, 41, 78, 99Heckmann, T., and W. Schwanghart (2013), Geomorphic coupling and sediment connectivityin an alpine catchment — Exploring sediment cascades using graph theory,Geomorphology, 182, 89–103, doi:10.1016/j.geomorph.2012.10.033. → page 42Hetherington, E. (1987), Carnation Creek, Canada: a review of a west-coast fish/forestrywatershed impact project, in Forest Hydrology and Watershed Management : Proceedingsof the Vancouver Symposium, p. 8, Vancouver, B.C. → page 8Hjimans, R. (2019), Package ”raster”,doi: → page 84Hodge, R., and L. Sklar (2011), Bedload transport in bedrock rivers: the role of sedimentcover in grain entrainment, translation, and deposition, Journal of Geophysical Research:Earth Surface, 116, 19, doi: → page 20Hoffman, D. F., and E. J. Gabet (2007), Effects of sediment pulses on channel morphology ina gravel-bed river, Geological Society of America Bulletin, (1), 116–125,doi:10.1130/B25982.1. → pages 3, 5, 20, 78, 79, 99Hogan, D. (1987), The influence of large organic debris on channel recovery in the QueenCharlotte Islands, British Columbia, in Erosion and Sedimentation in the Pacific Rim, p.165. → pages 24, 25Hogan, D. (1989), Channel response to mass wasting in the Queen Charlotte Islands: temporaland spatial changes in stream morphology, in Proceedings of Watersheds ’89: a conferenceon the stewardship of soil, air and water resources, pp. 125–142. → pages 3, 21, 46, 61, 73Hogan, D., S. Bird, and M. A. Hassan (1998), Spatial and Temporal Evolution of SmallCoastal Gravel Bed Streams, in Gravel Bed Rivers in the Environment, pp. 365–392,Highlands Ranch, Colorado. → pages 3, 8, 20, 40, 47, 61, 73Hogan, D. L. (1986), Channel morphology of unlogged, logged, and debris torrented streamsin the Queen Charlotte Islands. British Columbia Ministry of Forests and Lands, Landmanagement Report 49, Tech. rep., Victoria, B.C. → pages 3, 4, 20, 40, 47, 68, 69, 70, 71119Hyatt, T. L. H., and R. Naiman (2001), The residence time of Large Woody Debris in theQueets River, Washington, USA, Ecological applications : a publication of the EcologicalSociety of America, 11(1), 191–202. → page 53Imaizumi, F., and R. C. Sidle (2007), Linkage of sediment supply and transport processes inMiyagawa Dam catchment , Japan, Journal of Geophysical Research, 112(August), 1–17,doi:10.1029/2006JF000495. → page 79Iroume, A., L. Mao, H. Ulloa, C. Ruz, and A. Andreoli (2014), Large wood volume andlongitudinal distribution in channel segments draining catchments with different land use,Open Journal of Modern Hydrology, 4(57-66). → page 47Irvine, J. R., and M. A. Fukuwaka (2011), Pacific salmon abundance trends and climatechange, ICES Journal of Marine Science, 68(6), 1122–1130, doi:10.1093/icesjms/fsq199.→ page 78Jackson, K., and E. Wohl (2015), Instream wood loads in montane forest streams of theColorado Front Ranges, Geomorphology, 234, 161–170,doi: → pages 20, 21, 47, 68Keller, E., and F. J. Swanson (1979), Effects of large organic material on channel form andfluvial process, Earth Surface Processes and Landforms, 4, 361–380. → page 73Kelsey, H., R. Lamberton, and M. Madej (1987), Stochastic model for the long-term transportof stored sediment in a river channel, Water Resources Research, 23(9), 1738–1750,doi: → page 20King, L., M. a. Hassan, X. Wei, L. Burge, and X. Chen (2013), Wood dynamics in uplandstreams under different disturbance regimes, Earth Surface Processes and Landforms,38(11), 1197–1209, doi:10.1002/esp.3356. → pages 50, 75Krajina, V. (1969), Ecology of forest trees in British Columbia, Ecology of Western NorthAmerica, 2, 1–146. → page 9Lancaster, S., S. Hayes, and G. Grant (2003), Effects of wood on debris flow runout in smallmountain watersheds, Water Resources Research, 39, 1168. → page 73Lane, S. (1998), The use of digital terrain modelling in the understanding of dynamic riverchannel systems, in Landform monitoring, modelling, and analysis, edited by S. Lane,K. Richards, and J. Chandler, pp. 311–342, Wiley. → page 22Lane, S. N., M. Bakker, C. Gabbud, N. Micheletti, and J.-n. Saugy (2017), GeomorphologySediment export , transient landscape response and catchment-scale connectivity followingrapid climate warming and Alpine glacier recession, Geomorphology, 277, 210–227,doi:10.1016/j.geomorph.2016.02.015. → page 42Lapointe, M. (2012), Ecological Aspects of Gravel-bed Rivers, in Gravel Bed Rivers:Processes, Tools, Environments, edited by M. Church, P. Biron, and A. Roy, chap. 17, p. 25,doi:10.1002/9781119952497. → pages 1, 98120Leopold, L., and T. Maddock (1953), The hydraulic geometry of stream channels and somephysiographic implications, Tech. rep., USGS, Washington D.C., doi:10.3133/pp252. →page 100Lienkaemper, G., and F. Swanson (1987), Dynamics of large woody debris in streams in oldgrowth Douglas Fir forests, Canadian Journal of Forest Research, 17, 150–156. → pages46, 68Lisle, T. (2008), The evolution of sediment waves influenced by varying capacity inheterogeneous rivers, in Gravel Bed Rivers 6, Volume 11, edited by H. Habersack,H. Piegay, and M. Rinaldi, p. 836, Elsevier Science. → page 19Lisle, T., and B. Smith (2003), Dynamic transport capacity in gravel-bed river systems, inProceedings of the international workshop for ”source to sink” sedimentary dynamics incatchment scale, 16-20 June, pp. 187–206. → page 19Lisle, T., F. Iseya, and H. Ikeda (1993), Response of a channel with alternate bars to adecrease in supply of mixed size bed load: a flume experiment, Water Resources Research,29, 3623–3629, doi:10.1029/93WR01673. → page 78Lisle, T. E., and M. Church (2002), Sediment transport-storage relations for degrading, gravelbed channels, Water Resources Research, 38(11), 1–14, doi:10.1029/2001WR001086. →pages 2, 3, 19, 41Livers, B., K. Lininger, and E. Wohl (2015), Porosity problems, in Proceedings of the Wood inRivers Conference, Padova, Itality. → page 48Ljung, G., and G. Box (1978), On a measure of lack of fit in time series models, Biometrika,65, 297–303. → page 26MacVicar, B., and A. Roy (2007), Hydrodynamics of a forced riffle pool in a gravel bed river1: mean velocity and turbulence intensity, Water Resources Research, 43, 1240,doi: → page 78Madej, M., and V. Ozaki (1996), Channel response to sediment wave propagation andmovement, Redwood Creek, California, USA, Earth Surface Processes and Landforms,(21), 911–927,doi:〈911::AID-ESP621〉3.0.CO;2-1.→ pages 20, 39, 79, 99Maita, H. (1991), Sediment dynamics of a high gradient stream in the Oi River Basin ofJapan, Tech. rep., USDA Forest Service General Technical Report. → page 19Major, J., D. Bertin, T. Pierson, A. Amigo, A. Iroume, H. Ulloa, and J. Castro (2016),Extraordinary sediment delivery and rapid geomorphic response following the 2008-2009eruption of Chaiten Volcano, Chile, Water Resources Research, 52. → pages 3, 20, 78, 79Major, J., A. East, J. O’Connor, G. Grand, A. Wilcox, C. Magirl, C. Collins, and D. Tullos(2017), Geomorphic responses to U.S. dam removals: a two decade perspective, in GravelBed Rivers 8, doi: → page 99121Manners, R., M. Doyle, and M. Small (2007), Structure and hydraulics of natural woodydebris jams, Water Resources Research2, 43, 17, doi:doi:10.1029/2006WR004910. → page68Martin, D., and L. Benda (2001), Patterns of in-stream wood recruitment and transport at thewatershed scale, Transactions of the American Fisheries Society, 130, 940–958. → page 50Marutani, M., M. Kasai, L. Reid, and N. Trustum (1999), Influence of storm-related sedimentstorage on the sediment delivery from tributary catchments in the upper Waipaoa River,New Zealand, Earth Surface Processes and Landforms, 24(10), 881–896,doi:〈881::AID-ESP17〉3.0.CO;2-I.→ page 19Massart, D., B. Vandeginste, L. Buydens, S. De Jong, P. Lewi, and S.-V. J. (1998), Chapter 8:Straight line regression and calibration, in Data handling in science and technology,chap. 8, pp. 171–230, Elsevier ScienceDirect. → page 140Massong, T., and D. Montgomery (2000), Influence of sediment supply, lithology, and wooddebris on the distribution of bedrock and alluvial channels, GSA Bulletin, 112(5), 591–599.→ page 73May, L., and R. Gresswell (2003), Processes and Rates of Sediment and Wood Accumulationin Headwater Streams of the Oregon Coast Range, USA, Earth Surface Processes andLandforms, 28, 409–424. → pages 4, 50, 73McHenry, M. L., E. Shott, R. H. Conrad, and G. B. Grette (1998), Changes in the quantity andcharacteristics of large woody debris in streams of the Olympic Peninsula, Washington,U.S.A. (1982-1993), Canadian Journal of Fisheries and Aquatic Sciences, 55(6),1395–1407, doi:10.1139/f98-013. → pages 46, 47Meehan, W., and T. Bjornn (1991), Salmonid distributions and life histories, in Influences offorest and rageland management on salmonid fishes and their habitats, edited byW. Meehan, pp. 47–82, Bethesda, Maryland. → page 78Meleason, M., S. Gregory, and J. Bolte (2003), Implications of Riparian ManagementStrategies on Wood in Streams of the Pacific Northwest, Ecological Society of America,13(5), 1212–1221. → page 46Mercier, D. (2008), Paraglacial and paraperiglacial landsystems : concepts , temporal scalesand spatial distribution, Geomorphologie: relief, processus, environnement, 4, 223–234. →page 42Mitchell, K., S. Grout, R. Macdonald, and C. Watmough (1992), User’s guide for TIPSY,Tech. rep., Victoria, B.C. → page 51Montgomery, D., B. Collins, J. M. Buffington, and T. Abbe (2003), Geomorphic Effects ofWood in Rivers, American Fisheries Society Symposium, pp. 1–28. → pages1, 3, 4, 20, 46, 47, 98122Montgomery, D. R. (1997), Channel reach morphology in mountain drainage basins, GSABulletin, 109(5), 596–611, doi:10.1130/0016-7606(1997)109〈0596:CRMIMD〉2.3.CO;2.→ page 3Montgomery, D. R. (1999), Process Domains and the River Continuum, Journal of theAmerican Water Resources Association, 35(2), 397–410,doi:10.1111/j.1752-1688.1999.tb03598.x. → page 98Muller, T., and M. Hassan (2018), Fluvial response to changes in the magnitude andfrequency of sediment supply in a 1D model, Earth Surface Dynamics, In Press,doi: → pages 2, 3, 39Murphy, M., and K. Kosksi (1989), Input and depletion of woody debris in Alaska streamsand implications for streamside management, North American Journal of FisheriesManagement1, 9, 427–436. → pages 46, 53, 75Murphy, M., and W. Meehan (1991), Stream ecosystems, American Fisheries Society -Special Publication, 19, 17–46. → page 78Nakamoto, R. (1998), Effects of timber harvest on aquatic invertebrates and habitat in NorthFork Caspar Creek, in Proceedings of the conference on coastal watersheds, pp. 87–96. →page 75Nakamura, F., and F. J. Swanson (1993), Effects of coarse woody debris on morphology andsediment storage of a mountain system in Western Washington, Earth Surface Processesand Landforms, 18, 43–61, doi: → pages2, 20, 40, 50, 68Neill C. (1967), Mean velocity criterion for scour of coarse uniform bed material,Proceedings of the 12th Congress of the International Association of Hydraulic Research,3, 46–54. → page 20Nelson, J. M., Y. Shimizu, T. Abe, K. Asahi, M. Gamou, T. Inoue, T. Iwasaki, T. Kakinuma,S. Kawamura, I. Kimura, T. Kyuka, R. R. McDonald, M. Nabi, M. Nakatsugawa, F. R.Simo˜es, H. Takebayashi, and Y. Watanabe (2016), The international river interfacecooperative: Public domain flow and morphodynamics software for education andapplications, Advances in Water Resources, 93, 62–74,doi:10.1016/j.advwatres.2015.09.017. → pages 82, 83Nickelson, T., and R. Reisenbichler (1977), Streamflow requirements of salmonids: OregonDepartment of Fish and Wildlife, Annual Progress Report, Tech. rep., Portland. → pages4, 78Nigh, G., and P. Courtin (1998), Height models for Red Alder (Alnus rubra Bong.) in BritishColumbia, New Forests, 16, 59–70, doi: → page51Nowakowski, A., and E. Wohl (2008), Influences on wood load in mountain streams of theBighorn National Forest, Wyoming, USA, Environmental Management2, 42, 557–571. →pages 46, 47123O’Connor, J. E., M. A. Jones, and T. L. Haluska (2003), Flood plain and channel dynamics ofthe Quinault and Queets Rivers, Washington, USA, Geomorphology, 51, 31–59,doi: → page 20Owczarek, P. (2008), Hillslope deposits in gravel bed rivers and their effects on the evolutionof alluvial channel forms: A caste study from the Sudetes and Carpathian Mountains,Geomorphology, 98(1-2), 111–125, doi:→ page 20Pess, G., M. Liermann, M. McHenry, R. Peters, and T. Bennett (2012), Juvenile salmonresponse to the placement of engineered log jams (ELJs) in the Elwha River, WashingtonState, USA, River Research and Applications, 28, 872–881, doi:10.1002/rra.1481. →pages 4, 78Pess, G. R., D. R. Montgomery, E. A. Steel, R. E. Bilby, B. E. Feist, and H. M. Greenberg(2002), Landscape characteristics, land use, and coho salmon (Oncorhynchus kisutch)abundance, Snohomish River, Wash., U.S.A., Canadian Journal of Fisheries and AquaticSciences, 59(4), 613–623, doi:10.1139/f02-035. → pages 1, 98Pryor, B. S., T. Lisle, D. S. Montoya, and S. Hilton (2011), Transport and storage of bedmaterial in a gravel-bed channel during episodes of aggradation and degradation: A fieldand flume study, Earth Surface Processes and Landforms, 36(15), 2028–2041,doi:10.1002/esp.2224. → pages 2, 3, 5, 19, 35, 41, 43, 79Pyrce, R., and P. Ashmore (2003a), The relation between particle path length distributions andchannel morphology in gavel-bed streams, Geomorphology, 56, 167–187,doi: → pages 20, 29Pyrce, R., and P. Ashmore (2003b), Particle path length distributions in meandering gravelbed streams: results from physical models, Earth Surface Processes and Landforms,28(951-966), doi: → pages 20, 29RCoreTeam (2017), A language and environment for statistical computing,doi: → pages 22, 26, 54, 82, 84, 128Reid, L., and T. Dunne (2003), Sediment budgets as an organizing framework in fluvialgeomorphology, in Tools in Fluvial Geomorphology, edited by G. Kondolf and H. Piegay,p. 696, Wiley, Chichester. → page 20Richardson, J. (2019), Biological Diversity in Headwater Streams, Water, 11(2), 366,doi:10.3390/w11020366. → page 1Richardson, J. S., R. J. Naiman, and P. A. Bisson (2012), How did fixed-width buffers becomestandard practice for protecting freshwaters and their riparian areas from forest harvestpractices?, Freshwater Science, 31(1), 232–238, doi:10.1899/11-031.1. → pages 46, 76Rigon, E., F. Comiti, and M. Lenzi (2012), Large wood storage in streams of the easternItalian Alps and the relevance of hillslope processes, Water Resources Research, 48, 1–8.→ pages 4, 46, 50124Roberts, R., and M. Church (1986), The sediment budget in severely disturbed watersheds,Queen Charlotte Range, British Columbia, Canadian Journal of Forest Research, 16,1092–1106, doi: → pages 1, 79Ruiz-Villanueva, V., H. Piegay, A. M. Gurnell, R. Marston, and M. Stoffel (2016), Recentadvances quantifying the large wood dynamics in river basins: New methods and remainingchallenges, Reviews of Geophysics, pp. 611–652, doi:10.1002/2015RG000500. → pages4, 47, 53, 75Ryan, S. E., E. Bishop, and J. Daniels (2014), Influence of large wood on channelmorphology and sediment storage in headwater mountain streams, Fraser ExperimentalForest, Colorado, Geomorphology, 217, 73–88. → page 3Schumm, S. (1973), Geomorphic thresholds and complex response of drainage systems, inFluvial Geomorphology, Publications of Geomorphology, edited by M. Morisawa, pp.299–310, State University of New York, Binghampton. → pages 2, 19Seo, J., and F. Nakamura (2009), Scale-dependent controls upon the fluvial export of largewood from river catchments, Earth Surface Processes and Landforms2, 34, 786–800,doi: → page 68Siegert, M., J. Taylor, and A. Payne (2005), Spectral roughess of subglacial topography andimplications for former ice-sheet dynamics in East Antarctica, Global and PlanetaryChange, 45, 249–263, doi: → page 26Slaymaker, O. (1993), The sediment budget of the Lillooet River, British Columbia, PhysicalGeography, 14, 304–320, doi: → page 2Smith, B. (2004), Relations between bed material transport and storage during aggradationand degradation in a gravel bed channel (unpublished master’s thesis). → page 19Soulsby, C., J. Grant, C. Gibbins, and I. Malcolm (2012), Spatial and temporal variability inAtlantic Salmon spawning habitat in braided river channels: a preliminary assessment,Aquatic Sciences, 74(3), 571–586, doi:10.1007/s00027-012-0249-4. → page 100Steffler, P., and J. Blackburn (2002), Two Dimentional Depth Average Model of Riverhydrodynamics and fish habitat, Tech. rep., University of Alberta. → page 82Stout, J. C., I. D. Rutherfurd, J. Grove, A. J. Webb, A. Kitchingman, Z. Tonkin, and J. Lyon(2018), Passive Recovery of Wood Loads in Rivers, Water Resources Research, 54(11),8828–8846, doi:10.1029/2017WR021071. → pages 46, 75Sutherland, D., M. H. Ball, S. J. Hilton, and T. E. Lisle (2002), Evolution of aLandslide-Induced Sediment Wave in the Navarro River , California, Geological Society ofAmerica Bulletin, 7606, 13, doi:10.1130/0016-7606(2002)114〈1036. → pages2, 3, 20, 39, 41, 43Swales, S., R. Lauzier, and C. Levings (1986), Winter habitat preferences of juvenilesalmonids in two interior rivers in British Columbia, Canadian Journal of Zoology, 64,1506:1514. → page 102125Thompson, W. L., and D. C. Lee (2000), Modeling relationships between landscape-levelattributes and snorkel counts of chinook salmon and steelhead parr in Idaho, CanadianJournal of Fisheries and Aquatic Sciences, 57(9), 1834–1842, doi:10.1139/f00-135. →page 98Tschaplinski, P. (2016), Unpublished sediment excavation data. → page 29Tschaplinski, P. J., and R. G. Pike (2017), Carnation Creek watershed experiment — long -term responses of coho salmon populations to historic forest practices, Ecohydrology,10(August 2016), 12, doi:10.1002/eco.1812. → pages 8, 9, 11, 15, 52, 78Tunnicliffe, J., M. Church, J. J. Clague, and J. K. Feathers (2012), Postglacial sedimentbudget of Chilliwack Valley, British Columbia, Earth Surface Processes and Landforms,37(12), 1243–1262, doi:10.1002/esp.3229. → page 2Van Sickle, J., and S. Gregory (1990), Modeling inputs of large woody debris to streams fromfalling trees, Canadian Journal of Forest Research, 20(10), 1593–1601. → page 52Vannote, R. L., G. Minshall, K. Cummins, J. Sedell, and C. Cushing (1980), The RiverContinuum Concept, Canadian Journal of Fisheries and Aquatic Sciences, 37, 130–137. →page 98Venditti, J. G., P. A. Nelson, J. T. Minear, J. Wooster, and W. E. Dietrich (2012), Alternate barresponse to sediment supply termination, Journal of Geophysical Research: Earth Surface,117(2), 1–18, doi:10.1029/2011JF002254. → page 78Webb, A., and W. Erskine (2003), Distribution, recruitment, and geomorphic significance oflarge woody debris in an alluvial forest stream: Tonghi Creek, southeastern Australia,Geomorphology, 51, 109–126. → page 20Wells, R., and J. Trofymow (1997), Coarse woody debris in chronosequences of forests onsouthern Vancouver Island, Tech. rep., Center for Applied Conservation Biology. → page54Whiting, P., and J. Bradley (1993), A process-based classification system for headwaterstreams, Earth Surface Processes and Landforms, 18(7), 603–612,doi: → pages 1, 2Wipfli, M. S., and J. S. Richardson (2015), Riparian management and the conservation ofstream ecosystems and fishes, in Conservation of Freshwater Fishes, pp. 270–291,doi:10.1017/cbo9781139627085.010. → pages 4, 46Wohl, E. (2011), Threshold-induced complex behavior of wood in mountain streams,Geology, 39(6), 587–590, doi:10.1130/G32105.1. → page 73Wohl, E. (2013), Mountain rivers revisited, Vol. 19, 565 pp., Wiley-Blackwell,doi:10.1029/WM019. → pages 1, 2Wohl, E., and J. R. Goode (2008), Wood dynamics in headwater streams of the ColoradoRocky Mountains, Water Resources Research, 44(9), 1–14, doi:10.1029/2007WR006522.→ pages 21, 47126Wohl, E., and D. Scott (2017), Wood and sediment storage and dynamics in river corridors,Earth Surface Processes and Landforms, 42, 5–23, doi:→ pages 3, 4, 21, 40, 46, 73Wohl, E., S. Madsen, and L. MacDonald (1997), Characteristics of log and clast bed-steps instep-pool streams of northwestern Montana, Geomorphology, 20, 1–10. → page 46Wohl, E., N. Kramer, V. Ruiz-Villanueva, D. N. Scott, F. Comiti, A. M. Gurnell, H. Piegay,K. B. Lininger, K. L. Jaeger, D. M. Walters, and K. D. Fausch (2019), The natural woodregime in rivers, BioScience, In Press, 1–15, doi:10.1093/biosci/biz010. → pages 2, 42, 46Zimmermann, A., S. Tsang, S. Bird, M. Hassan, and D. Hogan (2004), Stream operationalchannel monitoring in real time, Tech. rep., Victoria, B.C. → pages 9, 25127Appendix AAutocorrelation and spectrum analysismethodological detailsAutocorrelation and spectrum analysis methodsAnalysis of autocorrelation in data provides information on how values are related through theseries in time or space. Sample autocorrelation coefficients (r) are dimensionless quantitiesthat provide a descriptive measure of how observations are correlated at different positionsrelative to the first observation at t or x = 0 (Chatfield, 2003). The autocorrelation coefficient(r) at lag k is calculated as:rk = ck/c0 (A.1)where ck is the autocovariance coefficient at lag k and c0 at lag zero. ck is calculated as:ck =1NN−k∑t=1(xt − x¯)(xt+k− x¯) (A.2)where N is the number of observations, t observation time, and x¯ observation mean. Valuesfor rk will range from 1 to -1, with negative values indicating an inverse correlation. Usingthis method, autocorrelation is commonly assessed by plotting rk against k. This method wasapplied to Carnation Creek profile data (spatial) and study reach storage data (temporal) usingthe ”acf” function in the base R package (RCoreTeam, 2017).While assessing autocorrelation provides information on scales over which phenomenaare related, periodicity in the data series can be explored using spectral analysis methods. Inessence, spectral analysis approaches serve to break data series into sine and cosine functionsof different frequencies using Fourier decomposition, then assess the combined fit of thesefunctions through inspection of their power spectra (Chatfield, 2003). The method is somewhat128analogous to multiple linear regression, in that the objective is to find which functions andassociated frequencies best fit the data. Given n-1 predictor variables represented as cos(2pitn ),sin(2pitn ) and data series xt of length n, a Fourier series can be formed, represented as:xt = ao+n−1∑k=1[akcos(2pikt/n)+bksin(2pikt/n)] (A.3)where ak and bk are the regression coefficients for frequencies k, which tell us how well theindividual sine and cosine functions fit the data. To interpret the contribution of each functionto explaining variability in the series, the periodogram value Pk is calculated asPk = ak +bk (A.4)for all frequencies in the series. The periodogram is then a plot of Pk against k, and peaks inPk can be identified as frequencies which fit the data well, while the periodogram integral equalsthe series variance. Minimum and maximum frequencies used in the analysis are, respectively,a function of the sampling interval (Nyquist Frequency) and the record length. Further detailsof the theory behind approach can be found in Chatfield (2003).For analysis presented in this thesis, the ”Spectrum” function of the base R package isused. This function uses the Fast Fourier Transform (FFT) algorithm to perform calculations.Following recommendations for applying the method, the spatial and temporal data series weredetrended and scaled by their means prior to analysis. The resulting periodograms were alsosmoothed for ease of interpretation by sampling and interpolating the raw output.Figures129SA-2 SA-3SA-4 SA-5SA-6 SA-7SA-8 SA-9Figure A.1: Width and mean bed elevation plotted through time for channel segmentsSA-2 through SA-9. Both width and depth have been normalized to the mean valuefor each study reach.130SA-3SA-2SA-5SA-4SA-9SA-6Figure A.2: Storage vs. change in storage for six of eight study areas, in expanded form.Two aggradation/degradation cycles are marked on each sub-panel in blue and redlines. Dates correspond to cycle beginnings and endings.1310.000 0.005 0.010 0.015 0.020 0.025 0.0300. density1991199920092017Figure A.3: Normalized spectral density of bar width along the channel main stem for thefour long-profile survey periods. Data have been smoothed with a loess function ofspan 0.01.1320.000 0.005 0.010 0.015 0.020 0.025 0.0300. density1991199920092017Figure A.4: Normalized spectral density of bar heights for the four long profile surveys.Data have been smoothed with a loess function of span 0.01.133Appendix BWood budget model term time seriesFigure B.1: Image of fluvially transported wood pieces caught on the fish fence aftera major storm in January 2015. A substantial quantity of wood was also notedtravelling over the fence during the period of high flow. Photo courtesy of SteveVoller.134YearFigure B.2: Simulated wood budget model terms for Scenario 1. (a) Terms for the up-stream reach, Reach 2. (b) Terms for the downstream reach, Reach 1. (c) Termsfor both reaches grouped. BE (live) is live wood input from bank erosion; Mort. ismortality input; Trans. is net transport; CA is losses from channel abandonment,and BE (CWD) is bank erosion input from coarse woody debris.135YearFigure B.3: Simulated wood budget model terms for Scenario 2. (a) Terms for the up-stream reach, Reach 2. (b) Terms for the downstream reach, Reach 1. (c) Termsfor both reaches grouped. BE (live) is live wood input from bank erosion; Mort. ismortality input; Trans. is net transport; CA is losses from channel abandonment,and BE (CWD) is bank erosion input from coarse woody debris.136YearFigure B.4: Simulated wood budget model terms for Scenario 3. (a) Terms for the up-stream reach, Reach 2. (b) Terms for the downstream reach, Reach 1. (c) Termsfor both reaches grouped. BE (live) is live wood input from bank erosion; Mort. ismortality input; Trans. is net transport; CA is losses from channel abandonment,and BE (CWD) is bank erosion input from coarse woody debris.137Table B.1: Reach characteristics of bank erosion and fluvial wood transport for woodbudget simulationsReach BE (m/yr) Length(m)Wba Lw/Wbb Pmovec Pθ d Step length(m)eReach 1 0.23 1300 16.2 0.35 0.78 0.87 42.5Reach 2 0.28 1700 15.4 0.37 0.75 0.87 37.7Grouped 0.26 3000 15.8 0.36 0.77 0.87 39.8a Wb is 10.4 m for channel upstream of Reach 2b 0.55 for wood entering from upstream of Reach 2c 0.48 for wood entering from upstream of Reach 2d Calculated from Carnation Creek orientation data applied to Eaton et al. 2012 functionse 12.9 m for wood entering from upstream of Reach 2Table B.2: Summary input and output wood budget terms for pre-harvest (baseline) con-ditionsReach IbeCWDa(m3/m)IbeSb(m3/m)Imc(m3/m)Tnet d(m3/m)Ocae(m3/m)D f(m3/m)Reach 1 0.0146 0.0200 0.0004 -0.0037 -0.0100 -0.0146Reach 2 0.0178 0.0244 0.0004 -0.0116 -0.0101 -0.0179Total 0.0164 0.0224 0.0004 -0.0081 -0.0101 -0.0164a Bank erosion input of coarse woody debrisb Bank erosion input of standing timberc Input from mortalityd Net wood transporte Output from channel abandonmentf Output from in-situ decay138Appendix CNays2DH settings and performanceevaluationNays2DH simulation settingsA total of nearly 2500 individual simulations were run through the Nays2DH modeling plat-form. Calculation grids were prepared for study areas individually, and then used for all simula-tions for a given study area. These grids were prepared to encompass lateral channel instability,with a wide buffer given to active channel boundaries. Calculation grids of roughly 1 m2 wereused, similar to the resolution of the original survey data. Simulation runtimes were determinedthrough experimentation, with output checked to determine the time needed for achievementof stable output. For most sites, this was in the order of 800-1000 seconds of model run time,depending on the specific site and flow. Model timestep was 0.01s. Manning’s n values of0.04, typical for high-roughness gravel bed streams, were used for all simulations.In order to run a simulation, it is necessary to have information on the downstream watersurface elevation for a given discharge in the reach of interest. For most study sites, it waspossible to fit a relationship between observed discharge and downstream surface elevation bycomparing discharge at the date and time of the topographic surveys to the surveyed watersurface. This relationship could then be used to predict downstream water surface elevationfor a given discharge for model input. However, in sites which displayed a high degree ofgeomorphic variability (e.g. SA-5 and SA-8), it was not possible to obtain a robust discharge-water surface elevation relation. In these cases, mean downstream water surface elevationfrom survey data was used for all simulated flow levels. This assumption is evaluated within abroader assessment of model performance discussed below.139Assessment of model performanceGiven the large number of simulations run with varying input conditions, it was not possible toindividually calibrate simulations. Therefore, authors opted to use best estimates of initial inputvalues, and perform a detailed assessment of model performance through comparison withfield data. The success of Nays2DH in simulating depth and velocity conditions in CarnationCreek study areas was determined by evaluating model output against field measured depthsand velocities, and study area-average data determined from the historical survey record. Flowconditions and channel bed DEMs generated shortly prior to field measurements were used asmodel inputs, and the model was run with the same parameters used for all other simulationsfor each study area.Field data was obtained in study areas SA-2, SA-3, and SA-7 during flows ranging from0.012 to 1.3 m3s−1 during July and October of 2017. Over 60 water velocity and depth mea-surements were captured with a SonTek FlowTracker 1 acoustic Doppler velocimeter and ac-cording to the standards described in BCENVCCS (2009). All velocities were observed for 60seconds and were measured at 0.6 x depth (from the water surface) when water depth was <1 m, and at both 0.2 and 0.8 of depth when the depth was ≥ 1 m with the results averaged.Measurement cross-sections were oriented perpendicular to the flow using pre-existing surveybenchmarks located within each study area, or were established with a total station and refer-enced to the survey hubs. Thus, each velocity measurement was geo-referenced to the samecoordinate system as the respective study area DEM, enabling direct comparison of modeloutput to survey results.The prediction errors (Sχ s) of the relationship between modeled and observed values arederived for each model observation according to equation 8.25 of Massart et al. (1998):Sχ s =Seb1√1m+1n+(y¯s− y¯)b21∑(xi− x¯)2(C.1)where Se is the residual error in the regression model, b1 is the regression slope, m is thenumber of replicates (m = 1 in this case), n is the sample size, x is the field-measured value,and y is the response. Note that (Sχ s) is the standard deviation error and can be converted to aconfidence interval (CI) by multiplying by t0.025;n−2 (for the 95% CI).The relationships between modeled and observed depth and velocity are given in FigureC.1. Although the prediction errors vary with both the modeled and measured values (seeEquation C.1), overall errors are of ± 0.13 m for depth (figure C.1a) and ± 0.23 m/s forvelocity (figure C.1b). Residuals about the regression line show some stratification by studyarea given the presence of relatively large scatter in the velocity data for SA-3. Additionally,a subset of depth data collected for SA-3 at low discharge (flows of 0.012 m3/s) appear toplot as a near-straight line with modeled values < 0.02 m. In both cases, the relatively coarse140resolution of the Nays2DH prediction grid (1 m grid cells) forces the model to average theresults over a relatively large area of bed material, which results in a tendency for higher errorswhen evaluated against point measurement field data. Field observations of flow characteristicsat Carnation Creek suggest a wide range of both depths and velocities is possible within a 1m2 area of bed, especially in close proximity to large roughness elements (e.g. logs, isolatedcobbles, etc.).The effect of relatively large roughness elements becomes magnified when considering lowflows, and in SA-3 this effect may be responsible for local back-watering observed in the fieldbut not predicted by the model. The slope and intercept of the depth calibration curve was0.90 and -0.026 m, respectively, approximating a 1:1 relation with a slight systematic under-estimation of depths throughout the distribution. Conversely, the slope and intercept of thevelocity calibration curve was 0.53 and 0.14 m/s. This suggests that the model over-estimatesvelocities < 0.2 m/s and then under-estimates higher predicted velocities, although the dataseems somewhat separated into two groups.In addition to point measurements, average modeled wetted widths (figure C.1c), depths(figure C.1d), and velocities (figure C.1e) are compared to averages extracted from field surveydata. To generate depth and velocity values from field data where only a water surface wasdefined, a raster of this surface was generated from survey points and the DEM of the bedbeneath subtracted from it, providing a resulting depth raster, similar to that determined fromthe flow modeling. From this depth raster, a mean value was extracted. Wetted width wasdetermined by dividing the field-defined wetted channel area by the study area length. Velocitywas calculated as V = Q/W*d, as discharge at the time of survey is known. As with the modelsimulations, discharge was scaled using the contributing area relation developed for each studyarea.Results of this evaluation are shown in figure C.1 c-e. These findings generally agree withour point measurements, with a good relationship for depth but under predictions of velocityabove 0.5 m. Width is fairly well predicted but the model overpredicts at higher flows, whichalso explains the velocity discrepancy.To evaluate the potential effect of reach length on model output, reaches SA-2, SA-6, andSA-8 were artificially extended by duplicating the reach DEM and correcting for elevationgiven known gradient in the study areas. These study areas were selected as the downstreamand upstream channel widths are relatively similar. Three years of data were selected from SA-2 and SA-6, and two years from SA-8, and two flow levels run for each year. Models were runand areas calculated from the downstream half of the reach as per methods described above.Wetted areas were then compared between the extended study area output, and that from theregular reaches.Results of this evaluation are shown in figure C.1f. Overall, very little difference is ob-141●●● ●● ●●●●●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●● ●●0.0 0.2 0.4 0.6 0.8 1.0 depth (m)Modeled depth (m) (a)0.896 x+ −0.026 ●●●●●●●●●●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●● ●●● ●●0.0 0.5 1.0 velocity (m/s)Modeled velocity (m/s) (b)0.531 x+ 0.137●● ●●●●●● ●●●●●●●●●● ●●●●●● ●●●●●0.0 0.1 0.2 0.3 0.4−averaged field depth (m)SA−averaged modeled depth (m)(c)0.797 x+ 0.039 ●●●●●●●●●●● ●●●●● ● ●●●●●●0.0 0.2 0.4 0.6 0.8−average field velocity (m/s)SA average modeled velocity (m/s)(d)0.462 x+ 0.112●●●●●●●●●●●●●●●●●●●●●●●● ●●●0 5 10 15051015SA average field Wb (m)SA average modeled Wb (m)(e)1.08 x+ −0.134●● ●●● ●●●●●●● ●●●●● ●0 200 400 600 800 100002006001000Wetted area (long)Wetted area (short)(f)1:1Best fitFigure C.1: Comparison of model output to field data for (a) depth, and (b) velocitypoint data collected in summer and fall of 2017; (c), (d) and (e), depth, velocityand width values calculated from reach-averaged data collected during historicaltopographic surveys, and (f) a comparison of model output between extended andregular reaches.served between the extended-reach simulations and the regular study area simulations, in theorder of 1-3% in total area. There is some indication that the minor differences are a product ofhow well the extended reach can be created: if reach boundaries do not match well, particularlyin the Z direction, then differences could arise simply as a result of the mismatch, unrelated tothe actual influence of boundary conditions.142Calculation of cumulative seasonal habitat150 200 2500.01.02.0Discharge (m3)150 200 25050150250350Day of yearHabitat area (m2)0500015000Cumulative habitat area (m2)Habitat area (m2 )Discharge (m3)0.5 1.5 2.51000750500250Figure C.2: Example of how seasonal cumulative habitat is calculated. A habitat-discharge relationship is applied to a seasonal daily flow record, and daily valuesare summed over the season.143


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items