Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modelling channel morphodynamics : the effects of large wood and bed grain size distribution MacKenzie, Lucy 2014

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2014_september_mackenzie_lucy.pdf [ 27.47MB ]
JSON: 24-1.0166010.json
JSON-LD: 24-1.0166010-ld.json
RDF/XML (Pretty): 24-1.0166010-rdf.xml
RDF/JSON: 24-1.0166010-rdf.json
Turtle: 24-1.0166010-turtle.txt
N-Triples: 24-1.0166010-rdf-ntriples.txt
Original Record: 24-1.0166010-source.json
Full Text

Full Text

Modelling Channel Morphodynamics: The Effects of Large Woodand Bed Grain Size DistributionbyLucy MacKenzieB.Sc., University of British Columbia, 2012A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinTHE FACULTY OF ARTS(Geography)The University Of British Columbia(Vancouver)August 2014© Lucy MacKenzie, 2014AbstractWithin this thesis the results of a set of four stream table experiments are presented in order toexamine the role that bed texture adjustments play in the development of a equilibrium channelform in the presence of large wood. Experiments were conducted using two physical modelsof Fishtrap Creek, an intermediate sized stream in the interior of British Columbia. While bothflumes were Froude-scaled models with fixed banks and mobile beds, Model 1 contained a singlegrain size representative of the D50 of the prototype stream while Model 2 contained a scaled grainsize distribution (GSD) of Fishtrap Creek. Two treatments of wood load were run in each model:a moderate wood load (a scale equivalent of 160 m3/m2) and a high wood load (a scale equivalentof 220 m3/m2). Channel morphology was captured at five-hour intervals in order to create DEMsof the evolving bed surface. The results of this study show that bed grain size composition plays adominant role in shaping channel morphology, even in the presence of large wood. The addition oflarge wood increased sediment storage which resulted in an increase in reach-averaged bed slope,the magnitude of which was proportional to the wood load added. Large wood also caused newareas of scour and deposition to be imposed onto the channel morphology that had been establishedprior to the addition of wood, causing an overall decrease in pool spacing and median pool area.The presence of a grain size distribution constrained the range of depth values in the flume asit allowed the bed to self-stabilize by limiting scour depth through the process of armouring.Regardless of the presence of large wood, maximum depths were approximately twice as deep inthe single grain size flume and pools were deeper relative to their area. These results highlightthe necessity of considering the full grain size distribution when modelling channel response tochanges in the governing variables that influence channel morphology.iiPrefaceThis dissertation is based on a set of four experiments that were designed by Brett Eaton. Ex-periment 1 was run by Siri Hermanski as part of a work study position in 2011. Experiment 2was run by Lucy MacKenzie as part of a work study in 2012. Experiments 3 and 4 were run bySarah Davidson as part of her M Sc Thesis in 2009 and 2010. All of the analysis and writingpresented in this dissertation was done by Lucy MacKenzie and none of the text is taken directlyfrom previously published or collaborative articles, including Davidson’s M Sc thesis.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 How Rivers Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 How Have We Studied Rivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Objectives and Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . 91.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1 Issues of Scale in Physical Modelling . . . . . . . . . . . . . . . . . . . . . . . 112.1.1 Types of Scale Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.1.2 Modeling Large Wood in Stream Channels . . . . . . . . . . . . . . . . 152.2 Prototype Stream: Fishtrap Creek . . . . . . . . . . . . . . . . . . . . . . . . . 152.3 Model Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.4 Experimental Design and Data Collection . . . . . . . . . . . . . . . . . . . . . 243 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.1 Bed Elevation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27iv3.2 Water Surface Elevation and Water Depth . . . . . . . . . . . . . . . . . . . . . 303.3 Pool Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.4 Wood Location Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.1 Experiment 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.2 Experiment 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.3 Experiment 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.4 Experiment 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.5 Synthesis of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535.1 Equilibrium Without Wood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535.2 The Addition of Large Wood . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555.3 Equilibrium with Wood Present . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.4 The Role of a GSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626.1 Grain Size Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626.2 The Presence of Large Wood . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636.3 Wood Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646.4 Applications of this Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646.5 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66A LiDAR Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77A.1 LiDAR: Remotely Sensed Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 77A.1.1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78A.2 LiDAR Visualization Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 79A.3 Proposed Method of Visualization . . . . . . . . . . . . . . . . . . . . . . . . . 81A.3.1 Comparison of Visualization Methods . . . . . . . . . . . . . . . . . . . 88B Digital Elevation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92C DEMs of Difference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105vList of TablesTable 2.1 Model and Prototype Parameters . . . . . . . . . . . . . . . . . . . . . . . . 23Table 2.2 Model Wood Piece Characteristics . . . . . . . . . . . . . . . . . . . . . . . 23Table 2.3 Summary of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Table 2.4 Facies Grain Size Categories . . . . . . . . . . . . . . . . . . . . . . . . . . 26Table 3.1 DEM Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Table 4.1 Comparison of Sediment Storage and Bed Slope . . . . . . . . . . . . . . . . 37Table 4.2 Summary of Steady State Morphology . . . . . . . . . . . . . . . . . . . . . 38Table 4.3 Within-Pool Bed Surface Texture . . . . . . . . . . . . . . . . . . . . . . . . 45Table 5.1 Summary of the Effects of Wood Addition, Wood Load, and GSD . . . . . . . 60Table A.1 Summary of PCA Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 84viList of FiguresFigure 2.1 LiDAR Image of Fishtrap Creek Watershed and Surroundings . . . . . . . . 17Figure 2.2 Location of Fishtrap Creek and the Study Reach . . . . . . . . . . . . . . . 18Figure 2.3 Study Reach at Fishtrap Creek in 2008 . . . . . . . . . . . . . . . . . . . . . 19Figure 2.4 The Approximate Area of Fishtrap Creek Represented by the Two Models . . 20Figure 2.5 Images of the Two Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Figure 2.6 Grain Size Distributions of Prototype and Models . . . . . . . . . . . . . . . 22Figure 2.7 Examples of Bed Surface Texture from the Two Models . . . . . . . . . . . 22Figure 2.8 Model Wood Piece Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 2.9 Laser Mount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Figure 3.1 Depth Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 3.1 Depth distributions continued . . . . . . . . . . . . . . . . . . . . . . . . . 32Figure 3.2 Pool Areas for Various Depth Criteria . . . . . . . . . . . . . . . . . . . . . 33Figure 3.3 Mean Pool Areas for Various Pool Depth Criteria . . . . . . . . . . . . . . . 35Figure 4.1 Example DEMs from Experiment 1 . . . . . . . . . . . . . . . . . . . . . . 39Figure 4.2 Longitudinal Profiles Prior to and Following Wood Addition . . . . . . . . . 40Figure 4.3 Example DEMs from Experiment 2 . . . . . . . . . . . . . . . . . . . . . . 42Figure 4.4 Example DEMs from Experiment 3 . . . . . . . . . . . . . . . . . . . . . . 44Figure 4.5 Example DEMs from Experiment 4 . . . . . . . . . . . . . . . . . . . . . . 46Figure 4.6 DoDs of Equilibrium Morphologies . . . . . . . . . . . . . . . . . . . . . . 48Figure 4.7 Cumulative Sediment Storage Since Wood Addition . . . . . . . . . . . . . 49Figure 4.8 Pool Spacing and Area Through Time . . . . . . . . . . . . . . . . . . . . . 50Figure 4.9 Pool Morphologies Relative to Wood Load . . . . . . . . . . . . . . . . . . 51Figure 4.10 Pool Morphologies Relative to Bed Textures . . . . . . . . . . . . . . . . . . 52Figure 5.1 Wood Movement Through Time . . . . . . . . . . . . . . . . . . . . . . . . 55viiFigure A.1 Components of a PCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Figure A.2 Example of LiDAR Visualization Method . . . . . . . . . . . . . . . . . . . 83Figure A.3 Various Methods of PCA Visualization . . . . . . . . . . . . . . . . . . . . 85Figure A.4 Flow Chart of LiDAR Visualization Method . . . . . . . . . . . . . . . . . . 87Figure A.5 Visualization Methods Compared for Location 1 . . . . . . . . . . . . . . . 89Figure A.6 Visualization Methods Compared for Location 2 . . . . . . . . . . . . . . . 90Figure A.7 Visualization Methods Compared for Location 3 . . . . . . . . . . . . . . . 91Figure B.1 DEM Legends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Figure B.2 Digital Elevation Models for Experiment 1 . . . . . . . . . . . . . . . . . . 96Figure B.3 Digital Elevation Models for Experiment 2 . . . . . . . . . . . . . . . . . . 100Figure B.4 Digital Elevation Models for Experiment 3 . . . . . . . . . . . . . . . . . . 102Figure B.5 Digital Elevation Models for Experiment 4 . . . . . . . . . . . . . . . . . . 104Figure C.1 DoD Legend . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105Figure C.2 DEMs of Difference for Experiment 1 . . . . . . . . . . . . . . . . . . . . . 109Figure C.3 DEMs of Difference for Experiment 2 . . . . . . . . . . . . . . . . . . . . . 113Figure C.4 DEMs of Difference for Experiment 3 . . . . . . . . . . . . . . . . . . . . . 115Figure C.5 DEMs of Difference for Experiment 4 . . . . . . . . . . . . . . . . . . . . . 117viiiAcknowledgmentsI would first of all like to thank my two supervisors, Dr Brett Eaton and Dr Michele Koppes, for thehelp and guidance they have provided throughout my academic career thus far. I am not sure thatI would have pursued my master’s degree without the encouragement and support of Dr Koppes.She has been an inspiring mentor throughout this process and I am thankful for her advice andsupport. I am also grateful to Dr Eaton who helped shape my background in geomorphology aswell as fostered the development of my own ideas about the way things work. His knowledge andenthusiasm have motivated me to do my very best.I would like to thank my office mates, Lawrence Bird, Matt Chernos, and Alexis Moyer, whohave provided valuable help as well as valuable distractions throughout the process of my degree.Likewise, many, many thanks to Sarah Davidson who has been a true friend as well as a (future)collaborator. I would also like to thank the rest of my lab group, in particular Lea Zhecheva, ArielKettle, and Siri Hermanski, for their help in the lab, the field, and in our many brainstormingsessions.Finally, I want to express how lucky I am to have the very best friends and family. AlthoughI have encountered a number of tough and stressful challenges within this process I have alwaysfelt supported and well-loved; there is no way I could have succeeded without them.ixChapter 1Introduction1.1 How Rivers WorkFrom small mountaineous channels to huge rivers that cut through plateaus and plains, all streamsact to shape the landscape. It is generally accepted that, regardless of size, the range of possibleforms that channels may exhibit is constrained by the timing and magnitude of flows in the channel(Q), the material that is transported and deposited along its length (Qb), and local boundary con-ditions such as bank strength and bed sedimentology (e.g. Mackin, 1948; Lane, 1957; Millar andQuick, 1993; Montgomery and Buffington, 1998; Eaton and Church, 2004). The relations betweenthese three governing conditions have been described using both empirically-derived equations aswell as physically-based theories. Predicting how a channel reacts to changes in one or severalof these conditions has long been a source of interest, as changes to environmental drivers withinfluvial systems is inevitable.Although moderate flows can rework both bank and bed material, it is the peak annual dis-charge, or bankfull flow, that is the most geomorphically effective in shaping channel form (Wol-man and Miller, 1960). The magnitude of discharge that a channel experiences is a function of thesize of the contributing watershed area (Dunne and Leopold, 1978) and the hydrologic connectiv-ity within the watershed (Jencso et al., 2009). The timing and duration of the peak flows are drivenby the general climate of the area and the presence of glaciers or other anthropogenic diversions(e.g. dams or agricultural demands) in upstream areas (De´ry et al., 2009).Material found within a channel reach may either be transported in from upstream or erodedfrom banks and the surrounding area. The flux of sediment coming into the system, Qb(t), in-fluences the relative magnitude of erosion and deposition that occurs within the channel (Mont-gomery and Buffington, 1998). However, because the ability of the channel to move sediment1through the system ultimately depends on Q(t), the tendency for a channel to aggrade or degradeover a period of time is dictated by the relative ratio of sediment transport capacity to sedimentsupply (Montgomery and Buffington, 1997).The size distribution of particles, Qb(D), that are transported in from upstream reaches or iseroded from channel banks also influence channel morphology. The size of particles relative tothe discharge influences the amount of sediment transport that can occur (Buffington and Mont-gomery, 1999b) while the distribution of sizes present allows for stabilizing processes such asarmouring and the development of other features, such as stone cells, to form (Church et al.,1998).The range of boundary conditions that account for a huge amount of variability in channelmorphologies worldwide. Consequently, it is important to consider the geological, fluvial, and,if applicable, glacial legacies of an area when examining channel morphologies. These processesplay a part in forming the longitudinal valley slope, Sv, which constrains the range of channelslopes that are possible and thereby influences the sediment transport ability of the channel (Eatonand Church, 2004). Bank strength, which is dictated by the cohesiveness of bank sediment andthe presence of vegetation, also influences channel form but must be considered relative to bedstrength as well as in relation the the width to depth ratio of the channel (Eaton and Millar, 2004;Millar, 2005; Eaton and Giles, 2009).Outside of the governing conditions, although partially linked to bank stability, in-stream largewood (LW), or wood pieces larger than 0.1 m in diameter and 1 m in length (Hassan et al., 2005),has also been found to influence channel morphology at a channel-unit, channel-reach, and water-shed scale (Abbe and Montgomery, 2003; Thompson, 1995; Brummer et al., 2006). Input fromadjacent forest areas or transported in from upstream reaches (Webb and Erskine, 2003; Boc-chiola, 2011), LW plays a significant role in modifying reach-scale channel hydraulics and sedi-ment transport by increasing flow resistance, and thereby decreasing flow velocity (Buffington andMontgomery, 1999a; Manga and Kirchner, 2000; Wilcox et al., 2006; Davidson and Eaton, 2012).Trapping over 150 % of the annual bed load and sediment yield (e.g.Megahan, 1982; Andreoliet al., 2007), in-stream LW accounts for the storage of more sediment than the amount that isstored in riffles or by other channel obstructions (Megahan, 1982; Thompson, 1995). The volumeof sediment stored within a stream channel correlates to the amount of LW present in a streamchannel: streams with higher wood loads store more sediment than those with little or no wood(Gomi et al., 2001; Tecle et al., 2001; May and Gresswell, 2003; Davidson and Eaton, 2012). Inmountaineous terrain, where wood loads are high and streams are of intermediate size, log jams actas one of the dominant storage reservoirs for sediment at a watershed scale (May and Gresswell,2003).2LW affects where within the stream channel sediment erosion and deposition will occur byaltering flow patterns within a stream channel (e.g. Thompson, 1995; Abbe and Montgomery,1996; Buffington et al., 2002; Montgomery et al., 2003). Erosion tends to occur in areas of vorticesand flow compression whereas deposition occurs primarily in areas of flow divergence. The areasof backwater often caused by valley dams are key locations of deposition in stream channels as thechannel-spanning jam imposes a new, higher base level which results in reduced flow velocitiesand decreased sediment transport (Lisle, 1986; Thompson, 1995; Lancaster et al., 2001; Wohl,2011).A fluvial system in equilibrium, or steady state, is defined as one in which the functionalrelationship between the input and output variable(s) is invariant for some period of time (Howard,1982; Renwick, 1992). Streams in equilibrium often display an optimal channel form, one thatis ‘in regime’ with the given hydraulic regime, which is dictated by Q(t) and Qb(t,D). It mustbe noted however, that this is a dynamic equilibrium, meaning that although the overall channeldimensions remain relatively fixed (e.g. width-to-depth ratio), the location of bars, banks, pools,and meanders are not necessarily stable through time (Millar, 2005).The time it takes for a system to respond to a change in one or several of the governingconditions is characterized by its response, or relaxation, time which is a function of the size ofthe channel network (or stream length), the hydraulic regime, and to a lesser extent, the magnitudeof change itself (Howard, 1982, 1988). Systems tend to be insensitive to regime changes that cycleon time scales shorter than the response time (e.g. annual fluctuations in discharge and sedimentyield) (Howard, 1982). Likewise, if input variables cycle at a lower frequency than the responsetime, the system has time to adjust while remaining in equilibrium (Howard, 1982).Rapid changes in trends, step changes, pulse inputs, or an intermediate frequency input cyclescan cause the morphology of a channel to alter (Howard, 1982). Such changes can be causedby processes and events such as mass movements from valley sides, upstream mining or forestrypractices, dam construction or removal, forest fires, and watershed urbanization (Montgomeryand Buffington, 1998). Due to the prevalence of such forcings in watersheds worldwide, themorphology of many real world streams are constantly adjusting to changes in their governingconditions.Despite this, the concept of a steady state channel configuration is pervasive within the fieldof fluvial gemorphology and underlies much of our qualitative understanding of morphodynamicswithin fluvial systems (Lane, 1955). This is driven by the fact that studies have to assume morpho-logical equilibrium of a channel in order to isolate the role that single variables play in dictatingchannel morphology.Regime theory has been developed with the goal of better understanding how channel mor-3phology responds to environmental change (Eaton and Church, 2004). This approach applies theconcept of governing conditions to stream channels in order to predict an optimized channel ge-ometry for any given set of conditions. This is accomplished by utilizing relations that incorporatebed material transport, flow resistance, and bank stability (Kirkby, 1977; Chang, 1979; Daviesand Sutherland, 1983; Millar and Quick, 1993). Using these, regime models attempt to find singlesolutions that represent an equilibrium stream geometry configuration for any given set of inputs.While empirical relations, derived from regression analysis of observed channel geometries, havebeen used to solve for the hydraulic geometry of simple channels, such as canals, rational regimemodels utilize physically-based equations to guide our understanding of channel response.A rational regime model proposed by Eaton et al. (2004) predicts that channel stability, andtherefore an equilibrium form, is achieved by adjustments made at three scales: grain scale, bed-form scale, and reach scale. Adjustments occur at grain and bedform scales through changes tothe surface texture and structure (e.g. Ashmore, 1991a; Church et al., 1998; Wilcox et al., 2006).In channels where banks are erodable, changes to flow resistance occur at a reach scale throughadjustments to channel gradient achieved through changes to channel sinuosity (e.g. Eaton andChurch, 2004).Understanding how different processes act to shape channel morphology is important in achanging world as humans play a large role in modifying governing processes that act to shapefluvial systems at a range of scales. Development and land use change influence watersheds byincreasing runoff in urban areas, by rerouting water for agricultural uses and by limiting sedimentsupply in areas downstream of dams, among other things (Stromberg et al., 2004). Overlaid ontop of that are anthropogenic changes to the climate system which influence many geomorphicprocesses. Evidence for non-linear changes in cycles of stream discharge, in large part due tochanges in precipitation patterns resulting from increased melt of permafrost and glaciers (e.g.Clarke, 2007; Shook and Pomeroy, 2012), have been observed on several continents. By alteringtemporal patterns of stream discharge, Q(t), this “loss of hydrologic stationarity” in conjunctionwith increased pressure on our freshwater systems will continue to play a role in altering themorphology of many fluvial systems (Sandford, 2012).1.2 How Have We Studied RiversOur current state of knowledge regarding fluvial systems has come from three principle methodsof amassing data: field studies, numerical modelling, and physical modelling, or experimentation.All three methods have their own inherent strengths and weaknesses that must be considered whenconducting analysis on the results. In many circumstances, the best understanding of physical4processes is acquired when all three are used in conjunction with one another.Field studies are based on observations in which quantitative measurements and qualitativeclassifications are made on real-life systems. Case studies are a type of field-based method ofdata collection, conducted at one location, in which a single system is studied in depth to betterunderstand the unique context that gave rise to its current state but also in order to get at theunderlying processes that drive the system. Field studies may also be undertaken in the formof a survey in which many field sites are examined and the results are compared and contrastedbetween sites in order to understand the relationship between two or more variables. It is from theresults of these multi-site field studies that many empirical relations are derived.Field studies are very important in shaping our understanding of how the driving forces affectthe actual morphology of stream channels, and for guiding the development of conceptual modelsregarding the functioning of fluvial systems. Additionally, measurements of real stream channelsare commonly used to validate the results of numerical and physical models.Numerical models use empirical and physical relations to model the functioning of systemsover a long period of time or for a range of different scenarios. These types of studies rely onpre-existing concepts of how systems operate, which generally have been formed through fieldobservations. Studies that employ numerical modelling are best used to assess patterns that existwithin the real world and to construct hypotheses that may be tested using field studies.While it is very appealing to take the results of numerical models at face value, it is impor-tant to recognize that their results are dictated by the choices made by the user as the conceptualmodel on which it is based as well as which the equations used in the model, whether physical orempirical, ultimately govern the results of the model. Furthermore, the high sensitivity of numer-ical models to their initial inputs demonstrates the necessity of having reliable measurements ofexisting systems derived from field work.Many rational regime models are numerical models that use optimality criterion to obtain thesolution for the hydraulic geometry of a channel for a given set of conditions (Millar and Quick,1993; Eaton et al., 2004). The type of criterion used depends both on the extremal hypothesisis invoked as well as the processes under study. Extremal hypotheses attempt to describe whichchannel variable(s) is optimized by the fluvial system in order achieve an equilibrium geome-try. Examples of such hypotheses include minimizing stream power (Chang, 1979), maximiz-ing sediment transport efficiency (Kirkby, 1977), maximizing sediment transport capacity (Whiteand Paris, 1982; Millar and Quick, 1993; Millar, 2005), maximizing friction factor (Davies andSutherland, 1983) and maximizing resistance to flow (Eaton et al., 2004). In many ways all ofthese hypotheses are addressing the same fundamental processes and comparison of the results ofdifferent hypotheses for similar conditions have been shown to yield equivalent results (White and5Paris, 1982). While numerical modelling affords the ability to compare the results derived fromthe different hypotheses, it does not allow us to determine how river systems actually adjust tochange.Physical modelling is useful in investigating how well hypotheses explain or predict how flu-vial systems come into equilibrium. Unlike in field studies, where the results are purely obser-vations of the existing conditions, physical modelling is an experimental method and allows forsingle or multiple conditions to be varied. Additionally, because physical modelling allows for adecoupling of process from any specific environmental conditions, this type of model is useful forstudying relationships and processes that are applicable to entire populations of phenomena ratherthan single case studies (Hooke, 1968).Physical modeling has several advantages over data collection in the field. First of all, workingin a laboratory or controlled environment reduces the hazards and technical difficulties associatedwith collecting field data. Many geomorphic and hydrologic processes of interest in fluvial en-vironments occur predominantly during periods of high flow when the ability to collect data iscompromised due to logistical and safety constraints (Braudrick and Grant, 2001). For example,high rates of discharge during high flows make stream crossing hazardous, while the increasedability of transport of sediment and debris poses threats to in-stream instruments. Physical ex-periments permit high flow conditions to be mimicked, allowing for hydraulic and geomorphicvariables to be measured, while reducing the risk associated with measurement.Secondly, physical modeling of stream channels allows us to address the role that specificvariables have on complex processes as they allow us to alter a single variable while holding therest of the system steady. Consequently, it is possible to determine the how changes to a specificvariable may alter channel conditions. This type of control also allows us to deconstruct processesthat are too complex to fully understand based on field data alone (Wallerstein et al., 2001).Lastly, the scaling down of models from field prototypes in the field allows us to decrease notonly the spatial dimensions of a stream channel, but also the temporal dimension as well, meaningthat it becomes possible to study in the time frame of hours to days the effects of processes that,in real life, operate on time scales ranging from years to decades. This is important as the lengthof time series required to study certain fluvial processes are not readily available and inherentlyrequire long periods of time to acquire.Stream tables, or flumes, are scaled-down physical models of stream channels. Two generaltypes of models exist: prototype models and generic models. Prototype models are scaled versionsof a specific stream channel and are used to examine how changes to certain variables affect thesystem as a whole. “Generic” models (which have also been called “analogue” or “similarity ofprocess” models (Hooke, 1968)) are not models of any specific prototype stream, but instead obey6gross scaling relationships found in fluvial systems and are used to study general fluvial processes.Chapter 2 goes further into detail regarding the methods employed when creating physical models.Flume experiments began to appear in the scientific literature around the late 1800s and early1900s. In many of these first studies, the observations made from the models were qualitativelyrather than quantitatively related back to the original system (Gilbert and Murphy, 1914). Someof the first modelling techniques that employed scaling equations through dimensional analysiswere first described by Buckingham (1915) and later refined throughout the rest of the 1900s byboth fluvial geomorphologists and hydraulics engineers alike (e.g. Bruun, 1966; Hooke, 1968;Hebertson, 1969; Schumm et al., 1987; Postma et al., 2008). Key findings from these early studiesconfirmed that if gross scaling relationships were met it could be assumed that the same morpho-dynamic processes operated in both the model and the prototype (Bruun, 1966; Hooke, 1968).Physical models have been used to examine a range of processes in fluvial geomorphologyincluding: sediment transport within channels (e.g. Ashmore, 1988; Young and Davies, 1991),the development of channel pattern (e.g. Schumm and Khan, 1972; Whiting and Dietrich, 1993;Termini, 2009), how channel patterns are sustained through time (e.g. Ashmore, 1991b; Braudricket al., 2009), how channel patterns change (e.g. Schumm and Khan, 1972; Davies and Lee, 1988),and local scour patterns due to roughness elements within the channel (e.g. Cherry and Beschta,1989; Kuhnle et al., 2002; Dey and Raikar, 2007; Pagliara and Carnacina, 2010; Bocchiola, 2011).When designing a flume experiment to examine hypotheses within regime theory it is neces-sary to consider the conditions that govern channel morphology: water discharge, sediment fluxand size distribution, and local environmental factors (e.g. valley slope, bank cohesion, geologicalhistory). Manipulating and isolating these many different components within a physical modelallow the user to focus in on the relationships between different variables in the system.Experiments conducted in a generic physical model of a laterally active gravel stream bed byEaton and Church (2004) examined the relationship between channel geometry and three indepen-dent variables, discharge (Q), sediment discharge (Qb), and valley slope (Sv), in order to investigatethe hypothesis proposed by Eaton et al. (2004) that adjustments to governing conditions occurs atthree different scales. When Q and Qb were varied and changes to channel morphology were pos-sible at all scales,Eaton and Church (2004) found that the primary channel adjustments occurredat a reach scale through the modification of channel slope. Changes in grain and bedform scale,such as surface armouring and changes in channel cross section geometry, were found to onlycontribute in a minor way to the overall changes in channel morphology.While Eaton and Church (2004) found that adjustments to channel slope occurred throughchanges in channel sinuosity (i.e. channel length), a similar set of experiments conducted byMadej et al. (2009), found that although channel slope adjusted in response to changes in Qb,7these adjustments occurred through changes in bed elevation driven by aggredation or degradation.Additionally, Madej et al. (2009) observed changes in bed armouring and channel geometry underhigh sediment feed and zero sediment feed conditions. This response was attributed to the fact thatalthough the streams in both experiments were able to erode through the substrate, the flume bedused by Madej et al. (2009) was narrower than that used by Eaton and Church (2004), meaningthat channel migration was limited by the flume walls. These results indicate that if a channel islimited at one scale, in this case at a reach scale, in its ability to respond to changes in governingconditions, this must be compensated for through changes to flow resistance at other scales.In order to further examine the ability of the system to respond to changes in governing con-ditions at a bedform and grain scale, Eaton and Church (2009) conducted a set of experiments inwhich Qb was varied in a flume with non-erodible banks. Like what was seen by Madej et al.(2009) in moderate sediment feed conditions, the results from these experiments showed that bedstate adjustments were able to compensate for nearly a twofold range of Qb without significantchanges to channel gradient. However, at high and sediment feed rates, the system was foundto experience net aggredation in the upper part of the study reach and was unable to establish amorphologic equilibrium form with respect to the governing conditions.The results from experiments mentioned above contribute to our understanding of how streamchannels adjust to changes in governing conditions. In unconstrained systems with erodible banks,stability is achieved by the channel adjusting its sinuosity to accommodate the imposed conditions(Eaton and Church, 2004). In systems where the channel is constrained or the banks are non-erodible, bed form adjustments can accommodate moderate changes in Q and Qb, (Madej et al.,2009; Eaton and Church, 2009).Given its influence on channel morphology and its prevalence in streams worldwide, largewood in stream channels is another important variable to study in relation to regime theory. Inorder to investigate the relationship between large wood and reach-scale channel morphology, aset of experiments were conducted on a non-erodible bank flume under the conditions of constantSv, Qb and Q (Davidson, 2011).Davidson (2011) found the addition of LW resulted in decreased a reach-averaged flow ve-locity. This decrease in sediment transport lead to the aggradation of material in the flume, themagnitude of which corresponded to the wood load present in the channel. In the presence ofhigh wood loading, this storage of sediment resulted in a higher reach-averaged bed gradient atequilibrium. Additionally, Davidson (2011) found that the higher spatial variability of local flowpatterns around LW pieces increased the heterogeneity of the bed surface as evidenced throughthe development of a greater number of bed facies patches with wood present.Altogether, the results of Davidson (2011) suggest that stream channels adapt to increases in8flow resistance caused by the addition of LW by adjusting both at a reach-scale, through changesto reach-averaged gradient, and at a bed form scale, through changes to bed facies. These findingscontradict those of Eaton and Church (2009) which predict that, in a non-erodible bank flume, bedstate alone should adjust to changes to governing variables and if changes to sediment supply aretoo large for the channel to adjust to by altering the bed state alone the system will be unable toachieve equilibrium.1.3 Objectives and Research QuestionsStudies that utilize physical models of stream channel reveal that, in highly simplified systems, thestability of the stream boundary dictates whether the channel will accommodate changes in gov-erning conditions either through adjustments at a reach scale in systems where banks are erodible(Eaton and Church, 2004) or through changes in bed state where bank erosion is constrained(Eaton and Church, 2009). Introducing additional roughness elements into a non-erodible flumein the form of modelled large wood pieces, increases the complexity of the response of the sys-tem to the newly imposed conditions as both reach-scale and bed-scale adjustments are observed(Davidson, 2011).This study aims to further examine the morphologic response of stream channels to the addi-tion of wood by i) examining the changes in bed morphology that occur following the addition oflarge wood and ii) assessing the role that the bed grain size distribution has on channel morpho-dynamics. The following questions will be addressed in relation to these themes:1. How does bed texture influence the progression of the channel into equilibrium in the ab-sence of large wood?2. What are the main differences in equilibrium channel morphology in the absence of largewood when only a single grain size is used compared to when a wide range of grain sizesare present?3. What impacts does the addition of large wood have on channel morphodynamics?4. In a non-erodible bank, mobile bed flume, how does the system adjust to accommodate thepresence of large wood?5. In the presence of large wood, what are the main differences in equilibrium channel mor-phology with and without a full grain size distribution?6. How do different wood loads influence the resulting equilibrium channel morphology?91.4 Thesis OrganizationChapter 2 discusses the experimental design of the flume experiments. Issues of scaling of boththe prototype models and large wood pieces are introduced and addressed. The prototype stream,Fishtrap Creek, is described and previous work on the field site is summarized. Both the designof the models themselves and the design of the experiments are outlined and the methods of datacollection are explained.Chapter 3 presents the methods used in the analysis of the data. This includes the creation ofdigital elevation models of the bed surface elevation and depth maps from the bed elevation dataand water surface elevations. It also presents a method for identifying pool areas from depth maps.Chapter 4 presents the results from the flume experiments. Channel morphology is qualita-tively assessed based on the digital elevation models of the bed surface. The temporal patternsof erosion and deposition are considered by examining the spatial correlation between subsequentruns. The relative amount and location of sediment storage that occurred in each experiment ispresented. Pool morphology is compared between different wood loads and with and without afull grain size distribution.Chapter 5 discusses the results of this study in relation to the questions outlined in the ob-jectives. This chapter is organized into sections that examine the channel morphodynamics 1)associated with the development of an equilibrium morphology without wood present, 2) thoserelated with the addition of wood, and 3) those linked to the flume coming into equilibrium withwood present.Chapter 6 summarizes the results of this study and proposes areas of further study related tothe main themes that are presented.Appendix A presents a short study that outlines a LiDAR visualization method developedin order minimize the biasing associated with the hillshading of DEMs using a single angle ofillumination.Appendix B supplies the digital elevation models of the flume bed at the culmination of eachrun of each experiment.Appendix C supplies DEMs of difference showing the change in elevation between each sub-sequent run for all experiments.10Chapter 2Experimental Design2.1 Issues of Scale in Physical ModellingWhen utilizing flumes, especially prototype models, it is necessary to employ scaling techniquesin order to assure that the types of processes and results observed during the experimental runscan be scaled up and applied to the real streams which they are attempting to model. Scalingtechniques aim to properly scale key variables are commonly known as ’pi’ (Π) terms (Peakallet al., 1996). The Π terms differ slightly on whether the model is attempting to simulate sedimenttransport or not.For fixed-bed models with non erodible channel boundaries and no sediment transport (i.e. noloose substrate), the Π terms are: the flow Reynolds number (Re, Eqn. 2.1), the Froude number(Fr, Eqn. 2.2), the relative roughness (Eqn. 2.3), and the channel bed slope (Eqn. 2.4). Fixed-bed models have been used to examine processes that occur in stream channels but are separatefrom sediment movement, such as wood transport and movement (Braudrick and Grant, 2000;Bocchiola, 2011). The controlling variables that make up these equations for fixed-bed openchannel flow model are (Yalin, 1971; Peakall et al., 1996):• properties of the fluid: the dynamic viscosity (µ) and density (ρ)• boundary conditions of the channel: hydraulic radius (R) and surface roughness (ks)• bed slope (S)• average downstream velocity (U) and• gravitational constant (g)11Π1 =ρRUµ = Re (2.1)Π2 =U√gR= Fr (2.2)Π3 =ksR(2.3)Π4 = S (2.4)Mobile-bed studies are more common in practice as they are used to examine key aspectsof channel morphodynamics such as sediment transport and/or channel morphology (e.g. Madejet al., 2009; Malverti et al., 2008; Wallerstein et al., 2001; Whiting and Dietrich, 1993; Schummand Khan, 1972). When a mobile bed model is used, the system is modelled as a two-phaseflow comprised of the sediment particles and the overlying fluid. Consequently, in addition tothe Re number and the Fr number, other Π terms, namely the relative roughness of the sediment(Eqn. 2.5), the relative density of the sediment (Eqn. 2.6), the grain Reynolds number (Eqn.2.7), and an expression of the Shields relationship (Eqn. 2.8), must be considered. In mobile-bedexperiments, two additional parameters, the sediment density (µs) and the characteristic grain sizeof the sediment (D) are employed in addition to µ,ρ,R,ks,S,U , and g.Π1 =RD(2.5)Π2 =ρsρ (2.6)Π3 =ρU∗Dµ = Re∗ (2.7)Π4 =ρU2∗γsD(2.8)In mobile bed studies it is necessary mimic the sediment size distribution of the prototypestream in order to both assure a rough boundary layer and to correctly model sediment transport.The sediment size distribution of a flume is associated with flow and hydraulic constraints (as itdictates Re∗), bedload transport and deposition, and particle settling. It is also related to bankcohesion in flumes with unconstrained channels. Consequently it is necessary to scale the distri-12bution as accurately as possible. For most prototype–model relationships, this is done by scalingdown grain size distribution of the prototype (Madej et al., 2009). It is, however, difficult to main-tain accurate sediment scaling in cases where the lower end of the sediment size distribution scalesdown to less than 300 µm as past this point the Re∗ drops lower than 70, meaning that the flowover these particles is no longer fully rough and the dimensionless shear stress of entrainmentchanges rapidly. This is mitigated against by using sediment only greater than 300 µm.The cohesion of sediment, and therefore bank strength, can also be changed during scalingdown of sediment as clay minerals have different intermolecular forces that sand or gravel. Theincrease in cohesion of sediment due to the addition of clay has been shown to limit channelerosion in flumes (Schumm and Khan, 1972). This issue is especially relevant for microscale riverexperiments as it is impossible to truncate the grain-size distribution at silt-sized particles, as isoften done in other studies. In response to this issue, studies have shown that it is possible touse alternative materials, such as inert silica flour (Garcia, 1993), Urea type II plastic (Gaines andMaynord, 2001) or small glass beads (Malverti et al., 2008), to replace the clays without affectingthe outcome of the experiment drastically.Models can also be divided into experiments that use unconstrained (mobile banks) and thosethat use constrained channels (fixed banks). Constrained banks are often used to study the affectsof a specific channel pattern on sediment transport and deposition processes, for example, thepattern of channel bars in a sinusoidal channel (Whiting and Dietrich, 1993; Eaton and Church,2009). Flumes with unconstrained, erodible banks have been used to study how changes in channelpattern are related to changes in sediment discharge or other key variables (e.g. Schumm andKhan, 1972; Wallerstein et al., 2001; Eaton and Church, 2004; Malverti et al., 2008; Braudricket al., 2009; Madej et al., 2009).Studies that have used unconstrained channels have also experimented with using techniquesto model the presence of vegetation in riparian areas (e.g. Tal and Paola, 2007). This may bedone in order to model the elevated bank strength associated with near-channel vegetation and/orassess the that bank stability has on channel migration. Vegetation has been modeled using alfalfasprouts, toothpicks, and other seeded plants (Tal et al., 2004; Pollen and Simon, 2006; Tal andPaola, 2007; Braudrick et al., 2009). There are still issues that arise when modeling vegetation asthe change in bank and soil cohesion that is associated with the roots of riparian vegetation cannotbe scaled (Pollen and Simon, 2006).2.1.1 Types of Scale ModelsIn an ideal situation, all Π terms would maintain a 1:1 ratio between the model and the prototype.However, in most situations, it is not possible to accomplish this, meaning that there are one or13more scaling exceptions to be made. It is generally the research objective of a study that dictateswhich relationships may be relaxed and which must be preserved.Froude-scale modeling is a commonly used method of scaling in which the flow conditionsdictated by the Froude number (subcritical vs. supercritical flow) are preserved while those dic-tated by the flow Reynolds number (laminar vs. turbulent flow) are relaxed enough that a smallerexperimental model can be created but that the resulting model still experiences fully turbulentflow (i.e. the flow Reynolds number must remain greater than 2000). This compromise is neces-sitated by the fact that, without changing the fluid viscosity between the model and the prototype,it is impossible to satisfy both the Froude and the Reynolds criteria for a scaled-down model.In Froude-scale models, there is the assumption that the other channel parameters (i.e. relativeroughness, channel slope), along with the Froude number, are correctly scaled between the modelin the prototype. Studies may choose to distort scaling relationships for parameters other than theRe (and Re∗) number in order to build smaller models or to study larger prototypes (Madej et al.,2009; Peakall et al., 1996). An example of this are distorted Froude-scale models which increasethe horizontal scale relative to the vertical scale while maintaining similarity of Froude number(Peakall et al., 1996).While distorted Froude-scale models still attempt to maintain a high Re number (i.e. turbulentflow), ”microscale rivers” operate with laminar flows, meaning that the Re number falls below2000 (Malverti et al., 2008). The main advantage to these laminar flow models is that time scalescan be greatly reduced, allowing many experiments to be run in a short period of time. In addition,these ”microscale rivers” do not require large laboratories to be run in as are quite small in size,with flow depths of only millimeters and lengths and widths on the order of tens of centimeters tometers. There are, however, several issues that arise with this type of scaling technique: (1) thesemodels are believed to have unrealistic friction coefficients, (2) suspended sediment transport isnot possible due to the absence of turbulence, (3) surface tension at these scales is believed toeffect the physics of channelized flow (Malverti et al., 2008). Despite the limitations that theseissues seem to impose on microscale models, several studies have suggested that it is still possibleto apply the findings from microscale studies to large scale fluvial systems (e.g. Lajeunesse et al.,2009; Malverti et al., 2008; Davies et al., 2003).Generic models, which have also been called ”analogue” or ”similarity of process” models(Hooke, 1968), are used to examine the underlying processes of large-scale phenomena. As thename suggests, these models are not scaled from a specific prototype, instead they obey gross scal-ing relationships and model some processes present in natural systems. Generic models have beencommonly used to examine fluvial responses to sea-level change (e.g. Van Heijst and Postma,2002; Koss et al., 1994) and long-term landscape evolution (e.g. Postma et al., 2008; Van Hei-14jst et al., 2002). The main downfall of generic models are that, unlike Froude-scaled and evendistorted-scale models, their results cannot be applied directly to any specific field example. Theresults from generic models are instead used to describe key trends and processes that underlie thefluvial systems that the model is simulating.2.1.2 Modeling Large Wood in Stream ChannelsWhen modeling LW in flume experiments, the physical characteristics of the wood pieces mustbe designed to replicate the properties of LW that makes them influential in channel morphody-namics. The characteristics that are relevant include: wood density (Bocchiola, 2011; Braudrickand Grant, 2001; Wallerstein et al., 2001), piece length (Wallerstein et al., 2001; Braudrick andGrant, 2000; Braudrick et al., 1997), “trunk” shape (Davidson, 2011; Braudrick and Grant, 2001;Wallerstein et al., 2001; Braudrick et al., 1997), and presence of branches and/or rootwads on thepieces (Davidson, 2011; Braudrick et al., 1997; Braudrick and Grant, 2001). Like bed sediment,LW pieces may be mobile (Davidson, 2011; Braudrick and Grant, 2001, 2000; Braudrick et al.,1997) or fixed (Bocchiola, 2011; Wallerstein et al., 2001) within the flume.In studies that use mobile LW pieces, the quantity and the placement of pieces will influencechannel dynamics, since individual pieces behave differently than log jams made up of severalwood pieces (Bocchiola, 2011; Davidson, 2011). Two types of wood loading have been describedby Bocchiola (2011): distributed wood loading and lumped wood loading. Distributed wood load-ing, in which individual pieces enter the stream at random intervals, tends to mimic the addition ofwood due to processes such as tree senescence or bank erosion. Lumped wood loading, in whichseveral LW pieces enter the channel at a single point, emulates the addition of wood due to a singleevent (e.g. forest fires, landslides and debris flows).2.2 Prototype Stream: Fishtrap CreekFishtrap Creek, the prototype stream for the flume models used in this study, is located in theinterior region of British Columbia near the town of Barriere and approximately 50 km north ofKamloops. The Fishtrap Creek watershed drains an area of about 158 km2 (Eaton et al., 2010a).Elevations within the area range from about 300 m a.s.l. to 1600 m a.s.l. The terrain is madeup of incised stream channels, steep valley slopes, as well as gently sloping plateaus. Althougha large portion of the surficial deposits present in the area are associated with the last glaciationwhich culminated around 11 500 years BP (Ryder et al., 1991), post-glacial fluvial activity, andmass movements have reworked many of these deposits resulting in a complex assortment ofdeposit types and landforms. This glacial and post-glacial legacy is indicated by the landforms,15such as drumlins and landslides, visible on the LiDAR image shown in Figure 2.1. The novelmethod of LiDAR data visualization developed during this thesis research project is described indepth in Appendix I. The relationship between the glacial legacy and the current fluvial processesis notable in Figure 2.1 as Fishtrap Creek and the prototype study area is seen to run through aglacial meltwater channel that extends out to the North Thompson river in the lower right-handcorner of the map.In August 2003 the McClure forest fire, a high intensity crown fire, burned through 62 % (98km2) of the watershed, killing almost all of the vegetation on the Fishtrap Creek floodplain up tothe edges of the stream channel (Eaton et al., 2010a). Since then, several long term studies havebeen conducted in a study reach located approximately 5 km up from the mouth of the creek,documenting the changes that have ensued since the fire.The study reach, established in 2004 and shown in Figure 2.2, is located just upstream of aWater Survey of Canada stream gauge (station no. 08LB024) that has been monitoring streamflowalmost continuously since 1971. Initially the study reach spanned a length of approximately 130m and encompassed 11 cross-sections (1-11) (Figure 2.3), however in 2006 the study reach wasexpanded on either side in order to better accommodate a gravel tracer displacement study (Eatonet al., 2010c). Cross-sections A-H were added on the downstream end and cross-sections 12-19were added to the upstream end, resulting in a total study reach of 440 m (along the thalweg).As reported in Eaton et al. (2010a), the study reach itself is relatively steep, with a gradientvarying from 1.5-2.0%. With a D50 of 55 mm, the surface of the bed is quite coarse while theunderlying bed material has a finer D50 of 35 mm. The average bankfull width of the channel is10-12 m. According to the WSC records, the mean annual peak flow, which is on average 7.5m3/s, occurs during the spring snow melt sometime between April and June.There have been several studies conducted within the study reach at Fishtrap Creek (see Pet-ticrew et al., 2006; Phillips, 2007; Andrews, 2010; Leach and Moore, 2010). These studies foundthat Fishtrap Creek had a relatively unusual response to the forest fire, as there was no significantincrease in peak flows or suspended sediment loads in the years immediately following the forestfire (Petticrew et al., 2006; Eaton et al., 2010c) in contrast to an increase in suspended sedimentthat is typically observed post-fire (Keller et al., 1997). Instead, the primary changes have oc-curred in the channel morphology: there has been a shift from featureless plane-bed morphologyto that of a riffle-pool morphology coinciding with an increase in local bed material transport rates(Eaton et al., 2010a,c). Additionally, extensive bank erosion caused some areas of the channel todouble in width (Eaton et al., 2010a,c). This morphological change has largely been attributed tochanges in the bank strength and stability resulting from a decrease in cohesion as burnt riparianvegetation decays (Eaton, 2008; Eaton and Giles, 2009; Eaton et al., 2010c).160 2 41 KmNStudy ReachFigure 2.1: LiDAR image of Fishtrap Creek watershed and the surrounding area createdfrom LiDAR data collected in 2007 and visualized using the methods described inAppendix I.17Figure 2.2: Location of Fishtrap Creek and the study reach (Figure taken from Eaton et al.,2010b)2.3 Model DesignTwo fixed-bank, mobile bed, Froude-scaled models of Fishtrap Creek were used to collect thedata for this study. The prototype dimensions and parameters used to scale the two models weremodeled after actual measurements made of Fishtrap Creek (Figure 2.4) in 2008 and 2009 byChristie Andrews (see Andrews, 2010). Model 1 represents a large portion of the lower half ofFishtrap Creek and encompasses both a relatively straight reach as well as two meander bends.Model 2 is comparatively straighter overall and represents a shorter section of the stream channel.The experiments presented for Model 1 were conducted in 2011 and 2012. The apparatus usedfor Model 1 measured 9 m long and 2 m wide, although the channel itself was 8.5 m long and 0.3m wide. The experiments conducted on Model 2 were done in 2010 and 2011(Davidson, 2011;Davidson and Eaton, 2012). The Model 2 apparatus measured 5 m by 0.85 m. The channel itselfspanned 4.5 m in length and was on average 0.35 m wide. The width of the channel was varied forModel 2 but not for Model 1, in order to better isolate the effects of LW in Model 1 as changes tochannel width result in slightly variable sediment transport rates (due to variable discharge) alongthe length of the channel.18Figure 2.3: Study reach at Fishtrap Creek in 2008 (Figure taken from Eaton et al., 2010a)Model 1 was run with a single grain size representative of the median subsurface grain size(D50) of Fishtrap Creek. Model 2 was run with a modeled subsurface grain size distribution (GSD)measured in Fishtrap Creek in 2006 (Figure 2.6) The low end of the distribution is truncated inorder to avoid the issues of entrainment noted above. The difference in the resulting bed textureof the two models is exemplified in Figure 2.7.The width, depth, and bed material grain size (Wf ,d f , and D f ) of both models were scaledusing a characteristic length scale (Lr) of 1:30. Discharge (Q f ) and time (Tf ) were scaled in bothmodels according to the following equations:19Figure 2.4: The approximate area of Fishtrap Creek represented by the two modelsTf = L0.5r (2.9)Q f = L2.5r (2.10)Given the difference in composition of the bed material in the two models manifests in thelack of ability for a surface armour to develop in Model 1, it was necessary for the two modelsto have different slopes in order to maintain a similar Qb : Q ratio. Maintaining similar sediment20(a) Model 1(b) Model 2Figure 2.5: The two models used to conduct the experiments for this study21Figure 2.6: Cumulative distribution functions showing the GSD of Model 2 and the grainsize used in Model 1.(a) Bed texture from Model 1 (b) Bed texture from Model 2Figure 2.7: Examples of surface bed texture from the two models shown in order to comparethe difference in bed texture between the model with a GSD and the one with a singlegrain size.22Table 2.1: Model and Prototype ParametersParameter Prototype ModelsWidth 10-12 mm 0.34 mmD50 35 mm 1.14 mmPeak Discharge 7.5 m3/s 1.6 l/sTime 27 hours 5 hoursTable 2.2: Comparison between the length classes and frequencies of LW in the prototypestream and those used in the experiments.Prototype Prototype Model Model Frequency (%)Piece Length (m) Frequency (%) Piece Length (m) Exp. 1 Exp. 2 Exp. 3 Exp. 42-4 27 0.1 32 33 33 344-8 36 0.2 35 33 38 378-16 35 0.4 32 33 29 2816-32 2 – – – – –concentrations between the two models is important when comparing rates of geomorphic change.The slope of the full GSD model (Model 2) was given a slope that approximated that of FishtrapCreek (0.018 m/m or 1.8%) while it was necessary to lower the slope of the single grain size model(Model 1) to 0.008 m/m or 0.8%.Six different forms of wood pieces (Figure 2.8) were used to represent the range of size classesand wood piece characteristics found in Fishtrap Creek by Eaton et al. (2010a). Characteristics ofwood pieces used in the experiments are based on a set of pilot experiments described in (Davidsonand Eaton, 2012) which found that pieces made of out of maple with square cross sections weremuch better at simulating the dynamics of wood pieces in the real stream channel than pieces madeout of lighter wood with circular cross sections. Rootwads and branches were added onto manyof the pieces as these were found to increase the stability of the pieces and promote interactionsbetween the pieces within the channel.The three size classes of wood pieces modeled in the two models were scaled geometricallybased on length classes and frequencies observed in Fishtrap Creek (Table 2.2). The largest lengthclass observed in the field was omitted from the experiments due to its infrequency within theprototype stream (makes up only 2% of LW surveyed) and because pieces in the 8-16 m size class(the second largest size class observed in the prototype) were considered large enough to functionas key members of log jams (Davidson, 2011).23Figure 2.8: Six different model wood piece forms used in the experiments2.4 Experimental Design and Data CollectionTable 2.3: Summary of the treatments and the run times of all four experiments.Exp. Treatment S f WLm WLp Tt Tw NrBed Texture Wood Load [m/m] [m3/m2] [m3/m2] [hr] [hr] -1 Single GS Moderate 5.4 x10−4 0.008 1.6 x10−2 90 35 182 Single GS High 7.4 x10−4 0.008 2.2 x10−2 90 35 183 Full GSD Moderate 5.4 x10−4 0.018 1.6 x10−2 45 15 104 Full GSD High 7.4 x10−4 0.018 2.2 x10−2 55 10 11Sv = flume slope, WLm = Scaled wood load (in the model), WLp = prototype wood load, Tt =total length of the experiment, Tw = time of wood addition, Nr = number of runsTwo experimental treatments were run on each model: moderate wood load and high woodload (Table 2.3). The moderate wood load treatment represents pre-fire wood load conditions atFishtrap Creek while the high wood load approximates the in-stream wood load five years afterthe fire as documented by Eaton et al. (2010a). For this study, the two experiments conducted onModel 1 will be referred to as Experiment 1 (moderate wood load) and Experiment 2 (high woodload) while those conducted on Model 2 will be referred to as Experiment 3 (moderate wood load)and Experiment 4 (high wood load).Each of the experiments were made up of a series of five-hour runs, each meant to simulateone year of morphologic change at Fishtrap Creek. Given that the model is run at the equivalent ofpeak flow discharge and that peak flows represent the period in which flow is sufficient to mobilizethe majority of the bed material (Buffington et al., 2004), these five hour periods were taken to24represent one year of morphologic change.Throughout all experiments, sediment was input to the flume model at a constant rate of ap-proximately 62 g/min using a rotating feeder. This sediment flux is consistent with values ofupstream sediment yield measured at Fishtrap Creek in the period since the fire (Eaton et al.,2010a). Sediment output from the flume was collected from the outlet of the flume at 15 minuteintervals. These samples were then dried and weighed in order to determine the sediment outputrate. This value of sediment output was also compared to sediment input rate in order to determinewhen the flume reached steady state. Steady state, or sediment transport equilibrium was definedas a period of five hours or more in which sediment input rate roughly approximated sedimentoutput rate.The initial bed conditions for each experiment was an imposed flat bed. In each experimentthe flume was first run without wood until a steady baseline morphology was reached. Once theflume reached an initial steady state, in which sediment output approximated sediment output fora period of 5 hours, the wood pieces were added all at once while the flume was run at a dischargeapproximating low flow (1/3*Qb f or 0.4 l/s). The location and orientation at which the woodpieces were introduced were chosen randomly in order to better simulate the natural recruitmentof wood into a stream. Following their addition, the location (i.e. the cross section) and theorientation of each wood piece was documented at the end of each five hour run.Bed morphology data was collected at the end of every five-hour run using a laser profilingsystem similar to that described by Zimmermann (2009) (Figure 2.9). Following each run theflume was drained of water and images of cross-sections, highlighted by the laser line, were takenat intervals of 38 mm on Model 1 and of 50 mm on Model 2.Bed facies data was collected by Davidson for the two experiments run on Model 2 (Davidson,2011). Facies maps, which distinguish between patches of different surface textures, were createdby manually mapping the boundaries of each different facies present on the bed. A photo was thentaken of a 6” by 8.5” segment of each facies patch in order to later help characterize grain size.Facies patches were classified according to the grain size categories, developed by Davidson,and presented in Table 2.4. This was done by first conducting Wolman samples on twenty of thefacies photos, accomplished by measuring grain sizes in Adobe Illustrator along a super-imposedgrid in order to establish representative facies for each of the grain size categories. The remainingfacies images were classified visually using the measured facies as reference. Error analysis of thevisually classified images revealed a 90% accuracy (Davidson, 2011).25Figure 2.9: Apparatus used to take images of cross-sections in order to create DEMs of thebed surfaceTable 2.4: Facies grain size categories used to determine representative grain sizes for eachof the facies maps created by Sarah Davidson. Table first presented in Davidson (2011).Category Size Range (mm) Characteristic Size (mm)Very Fine 0.35-0.50 0.43Fine 0.50-0.71 0.60Medium 0.71-1.0 0.85Coarse 1.0-1.4 1.2Very Coarse 1.4-2.0 1.726Chapter 3Data Analysis3.1 Bed ElevationData extracted from the images of the cross sections were used to determine the bed elevationsand create DEMs of the channel bed after each run. Images for each run were digitized using asemi-automated code written in Matlab®. The automated part of the code cycled through eachimage and first converted it into black and white in order to highlight the laser line in the imageand background noise (created through this process) was removed using a patch detector. Next,the cross sections present in the image were converted into a ‘skeleton’ version where the x and ycoordinates (point values) were given for all values along the cross section in which the x valuesrepresented the distance across the flume while the y values represented the elevation at each ofthe points across.User input was required to georeference the xy coordinate data from the images to the actualelevation values of the flume, which were collected using a total station. Two control points atthe corners of the channel, one on the left back and one on the right, were selected on each crosssection by the user. If the difference in the relative elevations of control points of the digitizedcross section did not match those taken using a total station of the flume, the cross section wastilted to match the actual difference in elevation. All points along the digitized cross section werethen assigned actual elevation values based on the known elevations of the control points.Digital Elevation Models (DEMs) were created from the georeferenced cross sections by in-terpolation. Although the resolution of the bed elevation points across each cross section are high,the lower resolution in the downstream direction constrains the potential precision of the DEM. Inruns where there was no wood present in the flume, linear 1D interpolation was used to up-sampledata in the across stream direction after which linear 2D interpolation was used to down-sample27data in the down stream direction. The spatial resolution of the resulting DEMs was approximatelyhalf the distance between subsequent cross-sectional images (20 mm in Model 1 and 25 mm inModel 2). This resolution was chosen in order to minimize the error and uncertainty associatedwith interpolation while maintaining a low enough spatial resolution to identify bed forms such aspools.Wood pieces were digitized out manually in runs where wood was present. To do this, a codewas created to run through each of the cross sections individually and the user must determinewhether or not each cross section contains any wood pieces. In the cross sections where woodpieces were present, the user estimated the surface of the bed in the area obscured by the wood.This estimation was facilitated by plotting the same cross section from the previous run and theprevious cross section from the current run on top of the cross section to be digitized. The xycoordinates manually generated from the digitization of these wood containing cross section werethen interpolated to create a DEM using the same methods described above.Uncertainty was introduced during several periods of the digitization process. The largestsource of error during the creation of DEMs is the presence of shadows in the image of the crosssection. Because the image captured of the bed relies on a laser line shot from above, if anything islocated between the laser light and the bed, it will be obscured and the image of the cross sectionwill not capture that portion of the bed. The two causes of shadowing are the edges of the bed andthe wood pieces of the channel.Shadowing by the edges of the bed arises due to the fact that the edges of the bed are rigid andoriented at a 90°to the bed. In areas of the channel where the sides of the bed are not perpendicularto the laser line across the bed (i.e. on meander sections), the bed edges shadow a portion of thebed along the side. In order to mitigate for this, measurements of bed elevation (taken as the depthfrom the top of the bank) were taken along the length of the largest meander in Model 1 whereshadowing was prone to occurring. These bed elevation measurements were then employed whendigitizing the data in order to fill in the areas where the edges shadowed parts of the bed.The uncertainty of the bed elevation DEMs must be considered when conducting analysis onthe data. The uncertainty in the across stream direction is the same as the resolution of the DEM(20 mm in Model 1 and 25 mm in Model 2) as data was upsampled in this direction. Because datawas interpolated in the downstream direction, the uncertainty in this direction is equivalent to thedistance between two subsequent cross sections: 38 mm in Model 1 and 50 mm in Model 2.Additional uncertainty in the elevation values of the DEM is associated with sediment move-ment between runs and user input error, introduced during the digitization process. Uncertaintyassociated with user input was estimated to be approximately 1 mm. Because the grid cells ofthe DEMs are actually averages of several points from the original cross sections, elevation error28Table 3.1: Values used to calculate the standard error (model uncertainty) of the elevationvalues of the bed elevation DEMs and the resulting standard error values associated withthe two flumesModel D90 n SEx¯1 1.4 mm 8 0.5 mm2 2.8 mm 12 0.8 mmassociated with sediment movement was calculated by determining the standard error of thosepoints using the following equation:SEx¯ =D90√n(3.1)where D90 is a representation of the grain size uncertainty and n is the number of originaldata points used to determine the bed elevation for each cell of the DEM. Values used to in thecalculations and the resulting standard error associated with each of the flumes is presented inTable 3.1.Given the calculated values of standard error and the estimated error associated with digitiza-tion, total uncertainty in the elevation values of the bed elevation DEMs was determined to be ±1.5 mm for Model 1 and ± 1.8 mm for Model 2.DEMs of Difference (DoDs) were created to examine changes in sediment storage betweenruns. These were made by differencing between two subsequent DEMs of bed elevation. DoDsfor all experiments are presented in Appendix C. The DoDs were then used to examine the spatialpattern of erosion and deposition in the flume. For DoDs, the elevation uncertainty is equal to thecombined uncertainty of the two DEMs from which it is calculated using the following equation:Uc(x) =√(u1(x))2 +(u2(x))2 + ...+(un(x))2 (3.2)where ui(x) is the combined uncertainty associated with each of the DEMs used to create agiven DoD and Uc(x) is the total uncertainty of the DoD. Using the uncertainty values for the bedelevation DEMs, reported above, the combined error for the DoDs is ± 2.1 mm for Model 1 and± 2.5 mm for Model 2. These uncertainty values are used as the threshold for change detection(i.e. difference in elevation must be ≥Uc(x) to be considered real) when analyzing the DoDs.293.2 Water Surface Elevation and Water DepthWater surface elevation profiles were created from measurements taken for all of the runs on Model2 and for some of the runs for Model 1. Water surface elevations (WSE) were determined by firstmeasuring the distance between the top of the flume and the water surface at regular intervalsalong the flumes (30 cm intervals on Model 1 and 50 cm intervals on Model 2). Next WSE werecalculated at each of those points by subtracting the measured distance from the known flume-topelevation. Once the WSE was found at each of the points along the flume channel, a 1D cubicinterpolation was used to find the water surface elevation for all cross sections located between themeasured points.These WSE elevation profiles were then extrapolated to create DEMs containing the values ofwater surface elevation at each point in the channel. Due to a lack of water surface measurementsin the across stream direction, it was necessary to assume that WSE was the same across thechannel width.For Experiment 1, conducted on Model 1, WSE were only measured for the run in which theflume was in equilibrium without wood (Run 8) and the one in which it was in equilibrium withwood present (Run 18). The water surface elevations for the remaining runs were approximatedby using the WSE of Run 8 for Runs 1 to 9 (all runs without wood) and the WSE from Run 18 forRuns 10 to 17. For Experiment 2, also on Model 1, although the WSE was taken for each run oncewood was added into the flume (Runs 8 to 18), the WSE taken for the flume at equilibrium withoutwood (Run 7) was used for Runs 1 to 7. WSE was measured for all runs for both Experiment 3and 4.DEMs of water depth were created by subtracting the bed surface elevation from the watersurface elevation for each of the runs for all experiments. The uncertainty in depths associatedwith these is determined using Equation 3.2 and therefore have the same uncertainty values as theDoD maps (± 2.1 mm for Model 1 and ± 2.5 mm for Model 2).The distributions of depths differ between the two flumes (Figure 3.1a). The distributionsplotted for the experiments conducted on Model 1 are positively-skewed and have a wide range ofdepths stretching out to approximately 2.5*d¯ (where d¯ is the mean depth). On the other hand, thedepth distributions shown for Model 2 are more normally distributed around the mean depth andonly have maximum depths of approximately 1.75*d¯.It should be noted that although the distributions shown in Figure 3.1a encompass all of theexperimental runs conducted on each flume, the depth distributions from individual runs, bothwith and without wood present, display similar trends as is portrayed by the general population.300 0.5 1 1.5 2 2.5 300.10.2DensityModel 1: Exp. 1  Typical Fishtrap Max. Depth0 0.5 1 1.5 2 2.5 300.10.2DensityModel 1: Exp. 20 0.5 1 1.5 2 2.5 300.10.2DensityModel 2: Exp. 30 0.5 1 1.5 2 2.5 300.10.2Normalized DepthDensityModel 2: Exp. 4(a) Distributions of depths for all experiments shown as a function of mean depth. The dashed black linerepresents the typical maximum pool depth at Fishtrap Creek.Figure 3.1: Depth distributions for each of the experiments. Each distribution is created fromdepth values gathered off of all runs for that given experiment.31−5 −4 −3 −2 −1 0 1 2 3 4 500.10.2Density  Model 1: Exp. 1 1σ Above Mean−5 −4 −3 −2 −1 0 1 2 3 4 500.10.2DensityModel 1: Exp. 2−5 −4 −3 −2 −1 0 1 2 3 4 500.10.2DensityModel 2: Exp. 3−5 −4 −3 −2 −1 0 1 2 3 4 500.10.2Standard Devation, DepthDensityModel 2: Exp. 4(b) Distributions of depths, normalized by standard deviation above the mean. The red line represents thedepth criteria used to identify pool areas.Figure 3.1: Depth distributions continued320 2 4 6 800.511.5  Single GS, No WoodMean (0 σ)1 σ  Above Mean2 σ  Above Mean1 2 3 400.51 Full GSD, No WoodDistance Downstream (m)Distance Across (m)Figure 3.2: Pools areas selected using three different depth criteria shown on an example run from each model.333.3 Pool IdentificationDEMs of water depth were used to identify pool areas. Pools are defined as areas of high flowdepths and low flow velocities relative to the rest of the channel (e.g. Whiting and Dietrich, 1993;Abbe and Montgomery, 1996) and, in practice, are most commonly identified from longitudinalbed elevation data using pre-determined criteria (e.g. Lisle, 1987; Montgomery et al., 1995).The identification of pools from DEMs of water depth allows for a rigorous analysis of thechannel morphology as it can account for the fact that more than one pool may be present alongany cross section of a stream channel and that pools can vary in form such that the longest axisof the pool may be oriented in any direction relative to the flow, a characteristics that may beoverlooked when identifying pools using a longitudinal stream data analysis.For this study it was established that the criteria used to identify pools should: a) select forareas deeper than the mean depth, b) consistently differentiate between riffle tops and adjacentpools, and c) be applicable to different populations of depth data.Standard deviation above the mean was used as criteria with which to select pool area from thedepth data. As seen in Figure 3.1a, the depth distributions are more comparable between modelswhen normalized by standard deviation. Both visual assessments and sensitivity testing was usedto help determine exactly which standard deviation to use as the depth criteria for pools.Visual assessments of the pools delineated on the DEMs are useful as they reveal how changesto the depth criteria affected the location and areas selected as pools. As shown in Figure 3.2, itwas observed that when the mean of the sample was used (i.e. 0 σ ), riffle tops were selected aspart of the pools. On the other hand, when a high standard deviation (i.e. 3 σ ) was used criteria,the resulting pool areas encompassed only the very deepest points of bed depressions.For the sensitivity testing, the depth criteria threshold was varied between the 0 to 3 standarddeviations above the mean (Figure 3.3). A line, representing a pool area of 1/3*W 2b (where Wb isthe bankfull width) is shown as a dashed pink line on Figure 3.3. This pool size is consistent withrelative pool sizes that have been reported in previous studies (e.g. Bilby and Ward, 1991).Mean pool area varied over several orders of magnitude depending on the depth criteria em-ployed. As with the results of the visual assessments, when low percentiles closer to the mean ofthe depth distribution are employed as criteria, the average pool size was found to be too large,whereas the mean pool size was small when the upper percentiles were used. Comparing the re-sults to the representative pool size of 1/3*W 2b , it was seen that mean pool size for all experimentscoincide approximately with this value when depth criteria is around 1σ above the mean.Based on the results of the visual assessments and the sensitivity testing, 1σ above the meanwas chosen as the depth criteria for isolating pool area. This criteria is a simple measure that is340 0.5 1 1.5 2 310−2100(Mean Area)/Wb2Single Grain Size, Mod. Wood Load  0 0.5 1 1.5 2 310−2100(Mean Area)/Wb2Single Grain Size, High Wood Load0 0.5 1 1.5 2 310−2100(Mean Area)/Wb2Modeled GSD, Mod. Wood Load0 0.5 1 1.5 2 310−2100(Mean Area)/Wb2Standard Deviations Above the MeanModeled GSD, High Wood LoadNo Wood RunsWood RunsDepth Criteria1/3*Wb2Figure 3.3: Mean pool area, normalized by a representative unit area of W 2b , is shown for arange of depth criteria for pool selection. The depth criteria value shown at the base ofthe graph is the standard deviation from the mean. The red horizontal line represents thedepth criteria eventually used in the analysis of the data. The dashed red line indicatesa pool size of 1/3*W 2b .35applicable to different depth distributions and is able to consistently isolate pool areas from otherlow lying bed forms within the stream channel.When applying this criteria, it is also necessary to consider the uncertainty of the DEMs. Forthis study it was determined that any areas selected as pools should be over three pixels in size,which is equivalent to areas of approximately 12 cm2 in Model 1 and 19 cm2 in Model 2. Settingthis minimum pool size limit helps to assure that all pools are larger than the uncertainty associatedwith the interpolation used in this study.The facies maps in Davidson (2011) and the pool areas determined from the depth DEMswere used to find the approximate median grain size found within each pool areas. This wasdone by georeferencing the facies maps and the delineated pools to each other in ArcGIS®. Firstthe extents of the facies patches and pool areas were then digitized. Next, each facies patch wasassigned a characteristic grain size according the methods described in Chapter 2. The extents ofthe pools were then used to select parts of the bed found within the pool area. The median grainsize located in each pool was determined using a weighted average calculated using the faciesarea found within the pool and its characteristic grain size. In order to compare the characteristicsurface grain size of the pool to the rest of the bed surface, weighted averages surface grain sizewere calculated using the relative facies area and the characteristic grain size for each patch forthe whole bed.3.4 Wood Location DataThe location and the orientation of each wood piece in the flume was recorded at the end of eachrun. Location was recorded as the cross section at which the upstream-most point of the woodpiece was located and orientation was recorded as the direction of the wood piece relative to thechannel. This data was later used to plot the location of individual wood pieces and jams alongthe stream channel.36Chapter 4ResultsThe following chapter reviews the results of a set of four comparable experiments conducted intwo different fixed-bank, mobile-bed, Froude-scaled prototype flume models of Fishtrap Creek.The main difference between the two models was the grain size distributions of the bed material:Model 1 contained only a single grain size representative of the D50 of the prototype stream whileModel 2 contained a scaled grain size distribution of that found in the prototype stream.Two experiments were conducted on each model, one in which a moderate wood load wasintroduced to the channel at sediment transport equilibrium and another one in which a high woodload was added. Discharge, sediment input, and slope was held constant throughout the length ofall experiments.Table 4.1: Difference in reach-averaged sediment storage (∆SS) and reach-averaged bedslope (∆Sb) between equilibrium with and without wood for all experimentsExp. ∆SS ∆Sb[cm3/m3] [m/m]1 4148 0.0012 5378 0.0033 3433 0.0024 7163 0.0034.1 Experiment 1: Single Grain Size Model, Moderate Wood LoadExperiment 1 was conducted in the single grain size model (Model 1) and was treated to theaddition of a moderate wood load (5.4 x10−4 m3/m2).Figure 4.1 shows the DEMs of the bed surface for four time periods: at the end of the first37Table 4.2: Summary of steady state morphology with and without wood.No Wood WoodExp. Te Sb D50 N f acies Te Sb D50 N f acies[hr] [m/m] [mm] [hr] [m/m] [mm]1 35 0.007 1.14 – 50 0.008 1.14 –2 35 0.007 1.14 – 50 0.010 1.14 –3 15 0.013 1.04 16 30 0.015 1.00 274 10 0.014 1.10 14 50 0.017 1.03 26Te = time to equilibrium, Sb = reach-averaged bed gradient, D50 = reach-averaged me-dian grain size of bed surface, N f acies = number of distinct facies (from the facies mapsof the bed)five hours of run time (i.e. the first run), at equilibrium without wood present, five hours after theaddition of wood, and equilibrium with wood present. Similar figures are presented for Experi-ments 2, 3, and 4 (Figures 4.3, 4.4, and 4.5). The small black lines displayed on the DEMs withwood present represent the approximate locations of LW pieces derived from the LW location datacollected at the end of each run (described in Section 3.4).Pools and riffles were observed to develop from the initial flat-bed surface that was imposedin the flume at the beginning of the experiment during the first hour of run time. By the end of thefirst run (Figure 4.1a), a pool-riffle pattern had developed that would remain relatively stable untilthe addition of LW. The DEMs in Figure 4.1 show that there was very little difference in channelmorphology after five hours of run time (Figure 4.1a) and after 35 hours of run time (Figure4.1b)). The location of pools and bars that developed in the absence of wood corresponded to thealignment of the fixed banks, with pools forming in the areas of high scour and flow convergence,for example on the outside of the large meander bend at approximately 5.5 m downstream.The DEMs of difference created from the DEMs of Experiment 1 (Appendix C) show thatalthough the general morphology of the channel remained stable after the first hour of the experi-ment, differences in elevation between consecutive runs indicate that sediment continued to movedown the length of the flume throughout the first 35 hours of the experiment.38  Elevation (mm)a)00.511360138014001420b)00.51c)  00.51LW Pieced)Distance Across (m)Distance Downstream (m)0 2 4 6 800.51Figure 4.1: DEMs from Experiment 1: a) After the first five hours of run time, b) Equilibriumwith wood (35 hours in), c) Five hours after the addition of LW (40 hours in), and d)Equilibrium with LW present (90 hours in)39Figure 4.2: Longitudinal profiles of mean and minimum bed elevation shown for equilibrium runs without wood present andwith wood present. The water surface elevation (WSE) is shown for the equilibrium run with wood present. The shadedrectangles represent the location of log jams made up of 10 pieces or more.40Without wood present, equilibrium was reached after a total of 35 hours. Scaling this back upto the prototype, this equals to approximately 7 years of morphologic change. Figure 4.2 showsthe longitudinal profiles of minimum and mean bed elevation at equilibrium with and withoutwood present for all experiments. Considering the profiles from Experiment 1 prior to the addi-tion of LW, the profile of mean bed elevation (No LW) is quite uniform and does not show anypronounced pools or bars except for at approximately 5.5 m downstream, which coincides with thelocation of the model’s large meander bend. The equilibrium bed slope without wood present wascalculated to be 0.007 m/m. The pool and riffle morphology is apparent on the longitudinal profileof minimum bed elevation (Figure 4.2) with pools characterized as locations where the minimumbed elevation is lower than the mean elevation and riffles characterized by areas where the meanand minimum elevations are similar.The addition of wood caused local alterations to the background morphology established priorto wood addition. While the location of many of the pools and bars remained constant, oncewood was added, new pools and bars developed in proximity to LW pieces, especially whereseveral wood pieces interacted to form jams (Figure 4.1b and Figure 4.1d). Likewise to areas ofmore intense scour, many of the bars grew in area and in elevation following the addition of LW,indicative of an increase in sediment storage.A new state of equilibrium was achieved with wood present after 50 hours, which is equivalentto approximately a decade of change in the prototype. Changes to the longitudinal profile of themean bed elevation reveal that the distribution of sediment storage was not equal throughout thelength of the flume as most storage occurred in the area upstream of the largest log jam whichoccurred at about 2.5 m downstream (Figure 4.2). The equilibrium bed slope with wood presentwas 0.008 m/m, which is 13% steeper than that of the equilibrium slope without wood. Thepresence of wood resulted in the creation of a more complex and variable bed surface as evidencedby more pools visible in both the mean and minimum bed elevation profiles (Figure 4.2).4.2 Experiment 2: Single Grain Size Model, High Wood LoadExperiment 2 was conducted in the single grain size model (Model 1) and was treated to theaddition of a high wood load (7.4 x10−4 m3/m2). As in Experiment 1, the development of a fairlystable channel morphology occurred within the first hour of run time (Figure 4.1a). Pools andbars developed in similar locations as what was observed in Experiment 1, as areas of erosion anddeposition were driven by flow patterns dictated alignment of the fixed flume banks.The time to equilibrium without wood present (35 hours) and the reach-averaged equilibriumslope (0.007 m/m) were the same for both Experiment 1 and 2. The longitudinal profile of equi-41  Elevation (mm)a)00.5113601380140014201440b)00.51c)  00.51LW Pieced)Distance Across (m)Distance Downstream (m)0 2 4 6 800.51Figure 4.3: DEMs from Experiment 2: a) After the first five hours of run time, b) Equilibriumwith wood (35 hours in), c) Five hours after the addition of LW (40 hours in), and d)Equilibrium with LW present (90 hours in)librium mean bed elevation was also similar between the two experiments (Figure 4.2), howeverExperiment 1 did not exhibit as pronounced a drop near the large meander bend at 5.5 m down-stream.Like what was observed in Experiment 1, the addition of wood caused new pools and bars tobe overlaid onto the pre-existing channel morphology (Figure 4.3c). Two large log jams developedin Experiment 2, one at about 1 m and the other at about 5.5 m downstream. As evidenced by boththe DEMs and the longitudinal profiles of bed elevations (Figure 4.2), new scour pools formedin areas around log jams while new bars developed in areas upstream of log jams. Net sediment42storage occurred upstream of both jams, as seen by a rise in the mean bed elevation, while thelast 3 m of the model (i.e. below the two log jams) experienced areas of net erosion as a result ofupstream sediment trapping by the jams (Figure 4.2).It took the same amount of time (50 hours) for the model to reach equilibrium with woodpresent in Experiment 2 as it did in Experiment 1. The combination of deposition in the upstreampart and degradation in the lower part of the flume resulted in a equilibrium reached-averagedslope of 0.010 m/m with wood present, a 35% increase over the equilibrium slope without wood.This change in slope was greater than what was observed in Experiment 1 following the additionof a lower4.3 Experiment 3: Full Grain Size Distribution Model, ModerateWood LoadExperiment 3 was conducted in the full grain size distribution model (Model 2) and was treated tothe addition of a moderate wood load (5.4 x10−4 m3/m2).In Experiment 3, a persistent channel morphology developed within the first hour (Figure4.4a) that was in large part controlled by the alignment of the channel banks. While the rapiddevelopment of a stable channel morphology is consistent with the experiments conducted onModel 1, the resulting channel morphology differed in that pools tended to form along the centerof the channel while bars formed in the bends along the edges of the flume, resulting in less of apool-riffle morphology than had been seen in Model 1. This difference can be attributed to lack oflarge meanders present in Model 2 relative to Model 1.At 15 hours (prototype equivalent of 3 years), the establishment of equilibrium took less thantwice the amount of time in Experiment 3 than it did for the two experiments conducted on Model1. The equilibrium reach-averaged channel slope was 0.013 m/m which is approximately twicethe slope observed in Experiment 1 and 2. This discrepancy is due to the difference in the imposedslope of the two models (Table 2.3).At equilibrium, the longitudinal profiles of the minimum and mean bed elevation were similarin the upper half of the flume which suggests a plane-bed channel morphology. The profile of theminimum bed elevation deviated from that of the mean bed elevation at about 3 m downstream,suggesting the presence of a large pool. The D50 of the equilibrium bed surface without woodwas calculated to be 1.04 mm based on the facies maps created by Davidson, with pool areas 15%coarser in than the rest of the average bed surface (Table 4.3).The flume experienced changes to bed surface texture following the addition of wood (Table4.2). The number of facies present increased to 27 and median bed surface texture fined by ap-43  Elevation (mm)a)00.5110120130140b)  00.5LW Piecec)Distance Across (m)Distance Downstream (m)0 2 400.5Figure 4.4: DEMs from Experiment 3: a) After the first five hours of run time, b) Equilibriumwith wood (15 hours in), c) Five hours after the addition of LW (20 hours in), and d)Equilibrium with LW present (50 hours in)proximately 5%. Pool areas remained coarser relative to the bed surface with wood present. Asseen by the longitudinal profile of minimum bed elevation, most of the pools formed in the lowerhalf of the flume as there are very few differences between the equilibrium minimum bed profilesin the upper half (Figure 4.2).The reach-averaged slope increased by 14% following the addition of wood, resulting in anequilibrium slope of 0.015 m/m. This was slightly higher than what was observed in Experiment1, which was conducted with the same wood load, but lower than what was seen in Experiment 2,which had a higher wood load. The longitudinal profile of mean bed elevation (Figure 4.2), showsthat most sediment accumulated in the area upstream of the large log that developed at about 2m downstream. However, unlike what was seen in Model 1, the area of accumulation was morerestricted to the area immediately (i.e. one meter) upstream of the jam.44Table 4.3: A comparison between the reach-scale median grain size and the median grainsize found within pool areas. The average median grain size of the bed surface and themedian grain size for pool areas are given for the two experiments conducted in the fullGSD flume for the runs at equilibrium without and with wood present.Experiment 3 Experiment 4No Wood Wood No Wood WoodBed D50 (mm) 1.04 1.00 1.10 1.03Pool D50 (mm) 1.20 1.25 1.28 1.21% Coarser 15 % 25 % 16 % 17 %4.4 Experiment 4: Full Grain Size Distribution Model, High WoodLoadExperiment 4 was conducted in the full grain size distribution model (Model 2) and was treated tothe addition of a high wood load (5.4 x10−4 m3/m2).At 10 hours (or the equivalent of two years in the prototype), the development of an equi-librium channel morphology in Experiment 4 was three times shorter than what was observed inModel 1 (Exp. 1 and 2) and five hours shorter than Experiment 3, which conducted in the samemodel. The longitudinal profile of mean bed elevation was uniform and the profile of the minimumbed elevation suggests that the upper half of the flume was dominated by a plane-bed morphologywhile the lower half contained some pools and riffles. The equilibrium reach-averaged bed slopewithout wood was 0.014 m/m, which is slightly higher than what was observed in Experiment 3.Like in Experiment 3.It took 20 hours longer for equilibrium to establish in the presence of LW in Experiment 4 thanin Experiment 3 (Table 4.2). This is equivalent to a difference of 4 years in the prototype channel.The addition of LW resulted in an 19% increase in the equilibrium reach-averaged bed slope (upto 17.2 %). This increase was higher than what was observed in Experiment 3 but was lower thanwhat was observed in Experiment 2 which was conducted with a similar wood load.Experiment 4 exhibited comparable responses in bed texture to Experiment 3 in response tothe addition of LW. The number of facies present at the bed surface increased from 14 to 26, thereach-averaged bed surface D50 decreased from 1.10 mm to 1.03 mm, and the surface texture ofpool areas remained coarse relative to the averaged bed texture in the presence of LW (Table 4.2and 4.3).45  Elevation (mm)a)00.5100120140b)  00.5LW Piecec)Distance Across (m)Distance Downstream (m)0 2 400.5Figure 4.5: DEMs from Experiment 4: a) After the first five hours of run time, b) Equilibriumwith wood (10 hours in), c) Five hours after the addition of LW (15 hours in), and d)Equilibrium with LW present (55 hours in)4.5 Synthesis of ResultsChannel morphology was visibly influenced by the addition wood in both models. This can easilyseen by contrasting the DEMs of bed surface prior to and following the addition of large woodin all of the experiments. The DEMs of difference shown in Figure 4.6 were created by sub-tracting the equilibrium DEMs (two with and two without wood) of experiments conducted inthe same model from one another. The DoDs in Figure 4.6a show the variability in equilibriumruns conducted on Model 1 (Experiments 1 and 2) while the DoDs given below in Figure 4.6bshow the variability between the equilibrium runs on Model 2 (Experiments 3 and 4). In bothfigures, the upper DoD shows the difference between the equilibrium channel morphologies with-out wood present, meaning that the bed grain size distribution and model is constant between thetwo DEMs. The lower DoDs show the difference between the equilibrium morphologies withLW present, meaning that while the bed grain size distribution and models are constant, the woodloads of the two equilibrium DEMs is different.46For both models the differences in elevation between the equilibrium morphologies withoutwood were relatively small in magnitude, indicating that the general morphology of the channel(i.e. the location of pools and bars) was comparable between the two experiments and that thealignment of the fixed model banks play a important role in dictating channel morphology. TheDEMs of difference show a great deal more variability between the two DEMs of the bed at equi-librium with wood present than was seen prior to wood addition. The differences in equilibriumbed elevation (with wood present) occurred in localized areas and elevational differences betweenthe two equilibrium DEMs were as large as 30 mm. Such areas occurred primarily in proximityto the location of the large log jams that developed during the experiments (see Figure 4.2 forlocations) which suggests that the larger magnitude of differences in elevation are indicative oflocal areas of deposition of material and scour caused by the presence of wood at that location.As seen qualitatively on the DEMs, the addition of wood led to an increase in sediment storagein all experiments (Figure 4.7). Total sediment storage was higher in the high wood load experi-ments (Experiments 2 and 4) than for the moderate wood load experiments (Experiments 1 and 3)(Table 4.2). These results indicate that wood load, rather than bed texture governs reach-averagedsediment storage.The storage rate of sediment suggest a two-phase pattern following the addition of wood. Thefirst phase represents a period of a high rate of sediment accumulation immediately following theaddition of sediment and can be seen for Experiments 1, 2, and 4 (Figure 4.7). In Experiment 3this increase in sediment storage does not begin until about five to ten hours into the experiment.The length of this first phase is influenced by the amount of wood present in the channel assediment storage rates remain higher for longer when there is more wood present in the channel.In the two experiments containing high wood loads (Exp. 2 and 4), the rate of storage remainshigh for about 35 hours, while in the two moderate wood load experiments (Exp. 1 and 3), the rateof storage begins to taper out at around 20 hours (Figure 4.7).The second observed phase of sediment storage occurs when the rate of sediment storage slowsdown and/or drops to zero as the system comes into a new equilibrium with the wood present. Al-though periods of low to moderate accumulation followed by sediment release persist throughoutthis phase, reach-averaged sediment storage remains relatively constant for all experiments duringthis phase (Figure 4.7).Coinciding with the increase in sediment storage, an increase in reach-averaged bed slope wasobserved in all experiments (Table 4.1). The total change in slope was found to be related to woodload, as the increase in reach-averaged bed slope was greater in the experiments run with a higherwood load.Pools were identified for each of the runs by selecting all areas that are deeper than one stan-47  No Wood0 2 4 6 800.51WoodDistance Downstream (m)Distance Across (m)∆ Z (mm)0 2 4 6 800.51−30−1501530(a) The upper DEM of difference (DoD) shows the difference in channel morphology between the flumeruns in equilibrium without wood present from Experiment 1 and Experiment 2, both conducted on Model1. The lower DoD shows the difference between the channel morphologies of the flume once equilibrium isreached with wood present.  No Wood0 2 400.5WoodDistance Downstream (m)Distance Across (m)∆ Z (mm)0 2 400.5−30−1501530(b) The upper DEM of difference (DoD) shows the difference in channel morphology between the flumeruns in equilibrium without wood present from Experiment 3 and Experiment 4, both conducted on Model2. The lower DoD shows the difference between the channel morphologies of the flume once equilibrium isreached with wood present.Figure 4.6: DoDs of equilibrium morphologies with and without wood on the two models.480 10 20 30 40 50 6000.511.52Time since wood addition, hoursSediment Storage, g/cm 2  Exp. 1Exp. 2Exp. 3Exp. 4Figure 4.7: Cumulative amount of sediment stored in the flume following the addition ofwood for each experiment.dard deviation above the mean and greater than at least three cells in area (see Chapter 3). Figure4.8 shows the pool spacing, calculated using the method presented in Montgomery et al. (1995),for each of the runs. Pool spacing is of the same order for all experiments which suggests that thegrain size distribution of the bed does not play a role in constraining the number of pools that formin the presence of LW (Figure 4.8).There was slight decrease in pool spacing, from around 3-4 Wb to 1-2 Wb (where Wb is thebankfull width), following the addition of wood for the experiments conducted in Model 1. Ad-ditionally, the variability in pool spacing is higher prior to the addition of wood, while followingthe addition of wood pool spacing remains relatively constant. There is not enough data prior towood addition to make similar Experiments 3 and 4, although the results post-wood addition arecomparable. These results indicate that it is the presence of wood rather than the wood load thatdictates pool spacing and that bed texture does not play a distinguishable role in pool spacing.The presence of wood in the flume influences the range of pool morphologies present in thestream channel (Figure 4.9). For this study, pool morphology is represented by pool area andmaximum pool depth. Pool area is normalized by W 2b , which represents a unit area of the stream49−30 −20 −10 0 10 20 30 40 50012345a)Time Since Wood Addition (hrs)Pool Spacing (W b/ Pool)  Exp 1 Exp 2 Exp 3 Exp 4Figure 4.8: Pool spacing, calculated using the method presented in Montgomery et al.(1995), shown through time for all four experiments. The points along the verticalgrey line represent the pool spacing in channel equilibrium prior to the addition of LWchannel. The absolute value of maximum pool depth is not given due to the uncertainty associatedwith digitization, instead, maximum pool depth is approximated by a value representing the depthassociated with two standard deviations above the mean depth of the pool (µp +2σ ). This value isthen normalized by mean depth (µtot)for that given run. Data is only shown for Model 1 as thereis not enough “No Wood” data to present from Model 2.Shown only for Experiments 1 and 2 (not enough pre-wood data for Model 2), median poolsize decreased in the presence of wood in both experiments (Figure 4.9). The increased number ofsmall pools exhibited a large range of maximum pool depths than pools of similar size that formedprior to wood addition. Very large pools (i.e. >1∗W 2b ) occurred less frequently and pools greaterthan 1.5 ∗W 2b were not observed. Maximum pool depths remained close to 3 ∗ µtot regardless ofthe presence of wood. The range of pool morphologies measured for the high and low wood loadswere very similar. All together these results indicate that the presence of wood acts to increase thefrequency of smaller, deeper pools while limiting the size of the largest pools regardless of woodload.Bed texture also played a role in constraining the range of pool depths that were observedin the flumes with wood present in the stream channel (Figure 4.10). The median of maximumpool depths in the single GSD model was found to be close to 1.5∗ µtot , while the median in thesingle grain size model was just above 2∗µtot . In the full GSD model, the largest maximum pooldepths were limited to 2 ∗ µtot , with a single outlier at just below 2.5 ∗ µtot , whereas the largestmaximum pool depths were found to be up to 3 ∗ µtot in the single GS model. This difference in50Figure 4.9: The range of pool morphologies found in Model 1 for the given three experimen-tal treatments: no wood, moderate wood load, and high wood load. Each point on theplot represents a pool recorded in the stream channel for the given wood load condition.The dashed lines represent the median pool area (horizontal line) and the median pooldepth (vertical line).51Figure 4.10: Difference in pool morphologies between the two models when wood waspresent in the stream channel. The results for both the moderate and the high woodload experiments are combined. Each point on the plot represents a pool recorded inthe stream channel with wood present on the given model. The dashed lines representthe median pool area (horizontal line) and the median pool depth (vertical line).the range of pool morphologies mirrors the difference in the depth distributions seen in (Figure3.1a), where the range of depth values found in Model 1 (single grain size model) had a largerrange and maximum depth values than those found in Model 2 (full GSD model).Pool areas do not appear to be influenced by bed texture in the same way as maximum pooldepths (Figure 3.1a). Consequently, the relationship between pool area and maximum pool depthis different between the two models. In the full GSD model (Model 2) small pools had a tendencyto be shallow, whereas small pools in the single grain size model (Model 1) varied over a muchwider range of maximum pool depths. Conversely, where in Model 2 large pools were found tohave maximum depths ranging across the entire observed spectrum, the largest pools in Model 1were all also quite deep.52Chapter 5DiscussionThe results of the experiments presented Chapter 4 show that both large wood and the particlesthat make up bed surface texture play a role in dictating equilibrium channel morphology. This isevidenced by the fact that fundamental differences channel morphology between the two modelspersist even in the presence of large wood.5.1 Equilibrium Without WoodWithout wood present in the channel, it is possible to see the role that armouring plays in control-ling the depth of scour in the flume by comparing the range of depth values observed in each ofthe two models (Figures 3.1a and 3.1b). In Model 1, with the single grain size bed material, themaximum depth value is about 2.5 times the mean depth. Conversely, in Model 2, where there is amixture of particle sizes present, maximum depth values are about 1.5 times the mean depth, andare more similar to those found in Fishtrap Creek, the prototype stream, where maximum depthsare approximately 1.75 times the mean depth. This similarity between the prototype and Model 2almost certainly arises due to the presence of armouring of the surface in response to the dischargethat they experience.Without the ability to armour, different processes are needed in order to achieve equilibrium inModel 1. Eaton and Church (2009) proposed that a channel can adjust to imposed conditions byeither a) adjusting its bed state, for example through the formation of an armour surface or otherbed structures, or by b) reach scale adjustments, such as through changes to channel gradient.Considering the development of steady state morphology prior to the addition of LW, it can bereasoned that the greater range in depths that occur in the single grain size model play a role instabilizing the channel morphology. When there are no larger particles present, Buffington andMontgomery (1999a) found that flow resistance can be augmented by increasing the amplitude of53the pools and bars present in the channel, thereby increasing the stability of the channel. In thisstudy, the presence of high amplitude bed forms, as indicated by the wide range of depths foundin Model 1 suggests that such a process plays a role in stabilizing the bed in this study (Figures4.1 and 4.3). By comparing the minimum bed elevations (Figure 4.2) as well as the pool spacing(Figure 4.8) between Models 1 and 2 at equilibrium pre-wood, we see that pools occur much morefrequently in Model 1. This is further indicative of the stabilizing role that the pool-riffle patternplays in the single grain size model.The response time of the two models, characterized by the amount of time it took each of themto come into equilibrium, is highly dependent on the composition of the bed material. It took morethan twice the amount of time, or the equivalent of seven years, for the single grain size modelto come into equilibrium compared to the equivalent of two to three years in the full GSD model(Table 5.1). This most likely has to do with the lack of armouring capabilities in Model 1 as thedevelopment of a armour layer as well as the structures that form from interactions between thelarger particles play an important role in the progression towards a stable morphology (Parker andKlingeman, 1982; Dietrich et al., 1989; Church et al., 1998). This difference in time to equilibriummay also stem from discrepancy in reach length between the two models as stream length has beenlinked to the response time of systems (Howard, 1988). Because the length of Model 1 is almosttwice that of Model 2, it is possible that the difference in time to equilibrium between the twomodels may not actually reflect the differences in bed texture.The equilibrium morphologies that developed in the absence of wood are similar between thetwo experiments conducted on each model (Figure 4.6). This is consistent with the main premiseof regime theory which suggests that channel will optimize its geometry in response to givenany set of imposed governing variables (Eaton and Church, 2004). If the experiments had notbeen conducted in a fixed-bank flume it is probable that the morphology or pattern of the channelmay not have been so similar the imposed channel pattern may have been responsible for thedevelopment of similar equilibrium channel configurations.It has been proposed that the surface grain size can be used to infer the approximate valuesof the sediment supply and the magnitude of discharge (Dietrich et al., 1989; Lisle et al., 1993;Buffington and Montgomery, 1999a,b). This arises from the premise that where Q is high relativeto QB, the surface grain size will be coarser than that of the subsurface, and vice versa. It followsthat comparing the size of particles found at the surface to those within the subsurface should beindicative of the Q:Qb ratio. This logic does not extend itself to the results of Model 1, as thesystem has no ability to modify the surface texture in relation to the imposed governing variables.This shows that the use of such relations to lend an understanding of sediment supply and dischargemay be more difficult in implementing in practice as this relation is highly dependent on a non-54uniform size distribution of material.5.2 The Addition of Large WoodThe mobility of wood pieces simulates how large wood interacts in unregulated forested channelsrather than pre-engineered log jams. The majority of wood movement occurred during the firsttwo runs after the addition of wood (Figure 5.1) as individual wood pieces coalesced to formmore stable log jams. Most wood pieces remained stable following this period of adjustmentalthough individual pieces were observed to move between subsequent jams. Given that the woodmovement slows well before before the models come to equilibrium, it is likely that, while thestability of wood pieces is a prerequisite to the establishment of equilibrium, it is the developmentof bedforms around the stable LW that largely influence the equilibrium channel morphology.5 10 15 20 25 30 35 40 45 50 5500. Since Wood Addition [hr]Mean Travel Distance [m]  Exp. 2Exp. 3Exp. 4Figure 5.1: The average travel distance of each wood piece following the addition of LW.Experiment 1 is not shown due to some errors in the collection of data.The morphologic effects of the addition of large wood are visible at two scales: the local scalearound LW pieces and at a reach scale. This indicates that the development of an equilibriummorphology is dependent on the interplay between these two scales of forcing. Local scale effectsinclude the formation of scour pools around wood pieces (e.g. Figure 4.1) while at the reach scaleeffects are manifested as an increase in reach-average slope (Figure 4.2).Changes to local patterns of erosion and deposition were observed immediately following theaddition of LW and are manifested in changes to the characteristics and frequency of pools. Withwood present, pools spacing decreased and pools were deeper relative to their areal extent (Figures4.9 and 4.8), which is consistent with the characteristics of scour pools around obstructions in otherstudies (e.g. Raudkivi and Ettema, 1985; Abbe and Montgomery, 1996; Kuhnle et al., 2002). It55can be seen from the DEM images that most of the scour pools and bars occur in proximity to LWpieces (e.g. Figure 4.1), and are controlled by altered patterns of flow convergence and divergencearound these roughness elements (Cherry and Beschta, 1989; Keller and Swanson, 1979; Abbeand Montgomery, 1996; Wallerstein et al., 2001).Reach-average sediment transport capacity decreased following the addition of wood as seenin the increase in sediment storage increased following the addition of the LW in almost all exper-iments (Figure 4.7). Sediment transport capacity responds to changes in channel roughness dueto the presence of LW both decreases reach-averaged flow velocity (Davidson and Eaton, 2012)and increases the total shear stress that is required to initiate the movement of material (Assaniand Petit, 1995). This change in sediment transport capacity due to wood addition puts the systemout of equilibrium, without changing any of the governing variables (Q, Qb, SV , or bank strength).In order to regain equilibrium, it is necessary that the system alter its reach-averaged resistanceand/or its bed state resistance (Eaton and Church, 2004).In real stream channels with erodible banks the addition of wood alters bank stability by bothincreasing bank erosion due to a loss of cohesion associated with roots (Keller and Swanson,1979; Zelt and Wohl, 2004) as well as by stabilizing banks by lining channel edges and preventingfurther erosion (Montgomery and Buffington, 1997). At Fishtrap Creek, the 2003 forest fire hasled to a significant loss in bank stability as a result of the fallen and standing dead trees (Eatonet al., 2010c). This loss of cohesion has facilitated the creation of new side channels and led tochannel widening (Eaton et al., 2010a). Consequently, the progression of Fishtrap Creek towardsa new equilibrium form is quite different than what was observed in the model runs.Changes to sediment supply often accompany changes to bank stability as increased rates ofbank erosion provide additional sediment to the system (Wondzell and King, 2003). However, inthis study, the sediment input rate remained constant throughout the length of the experiments.While constant input rate is consistent with response of the prototype stream in the years sincethe fire (Eaton et al., 2010a), it could be expected that if sediment input was increased followingwood addition, more aggradation would have occurred, as the ratio of sediment transport capacityto sediment availability would have been much lower.5.3 Equilibrium with Wood PresentReach-averaged equilibrium bed gradient increased in all experiments in response to the decreasein sediment transport capacity caused by the presence of LW (Figure 4.2). Given that the additionof LW results in a lower sediment transport capacity than without wood present, it would be ex-pected that the system should adjust by increasing transport capacity in order to re-attain sediment56transport equilibrium with the input rate. However, the results from this study are inconsistentwith those of Eaton and Church (2009) which predict that in a fixed-bank flume, the fluvial systemshould adjust to changes in the governing conditions primarily through changes to its bed statewhile maintaining a comparable bed gradient and morphology for a range of conditions. Bed stateadjustments in the form of surface armouring does not appear to contribute substantially to thedevelopment of an equilibrium morphology in this study as equilibrium is still attained even whenthe bed is composed of a single grain size.Differences in equilibrium bed morphology with and without wood present can be attributedto local changes in erosion and deposition induced by the presence of LW. This is supported bythe fact that the location of scour pools and bars coincide with the location of LW (e.g Figure 4.1).The DEMs of difference (Figure 4.6) show that the equilibrium morphologies are highly variablewith wood present, which suggests that LW ultimately governs the location and size of bedforms.The net aggradation of material in response to the addition of wood was not uniform alongthe length of the flume. In all experiments most aggradation occurred in the area upstream oflargest log jam(s) that developed (Figure 4.2). In contrast, the areas downstream of the largestjams experienced degredation as a result of a low ratio of sediment supply to sediment transport.Both the aggradation in the upper half and the degradation in the lower half contributed to theincrease in channel gradient.The locations of sediment storage occur in relation to the location of large log jams (≥ 10pieces) observed in the flume. This occurs due to the fact that as large log jams form throughpiece accretion (Bocchiola et al., 2008), they function as barriers, decreasing sediment transportimmediately upstream of the jam (Assani and Petit, 1995; Keller and Swanson, 1979). This de-position of sediment acts in conjunction with the log jam barrier to further decrease sedimenttransport capacity by further lowering the channel gradient (May and Gresswell, 2003; Webb andErskine, 2003). The long term of effects of this, as the channel approaches a new equilibriummorphology with wood present, is an increase the mean bed elevation in areas upstream of the logjams, while adjusting to downstream sediment starvation by maintaining the pre-wood mean bedelevation and affording local areas of degradation (Figure 4.2).Net sediment storage in the flume was proportional to the amount of wood present, as moresediment was stored for experiments with higher wood loads (Figure 4.7). Following the additionof wood, all experiments exhibited a similar rate of sediment storage during the first 20 hours(although Experiment 3 did experience a short period of net loss). This net accumulation continueduntil which point a threshold was reached and the rate of storage decreased to approximately zero.The time at which this threshold was reached appears to be related to the wood load present in thechannel, with lower wood load experiments reaching the threshold sooner.57This threshold may either represent the reach-averaged storage capacity of the flume whichcan be thought of as the sum of the sediment storage capacities of all the wood barriers presentin the channel. Following wood addition, areas behind these jams start to fill up with sedimentin response to the decrease in local sediment transport. Sediment storage rate levels off once ajam becomes “filled” Megahan (1982). The threshold at which sediment storage rate decreases tozero may be when all or most of the jams within the flume have reached their sediment storagecapacity.Alternatively, the threshold may represent the point at which net aggradation within the flumehas created a new gradient that has allowed the flume to come into sediment transport equilibriumgiven the governing variables. If this is the case, it would be expected that the total amount ofsediment storage should be less for a lower sediment input given the same wood load, discharge,and slope.In reality, this threshold is likely related to both of these processes in that individual log jamsstore the sediment needed to alter the channel gradient in response to wood addition, however it isthe total wood load of the reach that dictates how much sediment must aggregate in order to allowfor a sediment transport equilibrium. Given this scenario, it would be expected that the same logjam would store more or less sediment depending on the amount of other wood pieces present inthe surrounding area.These results suggest that while predicting an exact equilibrium form in the presence of mobilewood pieces is difficult due to the variability in the interactions and spacing between individualwood pieces and log jams, it may be possible to predict reach-averaged responses to differentwood loads. Further experimentation in which channel morphodynamics were explored by com-paring runs with mobile wood pieces and runs in which log jams were engineered in place couldhelp to further our understanding of the role that LW plays in local and reach-averaged channelmorphology.5.4 The Role of a GSDWhen considering the importance of LW pieces compared to particles found on the bed, it wouldbe expected that, given the large discrepancy in the size between these two roughness elements,properties of the grain size distribution should play a only relatively small part in dictating channelmorphology (Manga and Kirchner, 2000; Wilcox and Wohl, 2006). The results of this study showthat, while the presence of LW does increase the complexity of the bed morphology in both mod-els, there is a consistent difference in the distribution of depths measured in two models, regardlessof the presence of LW. Although a study conducted by Gill (1972) found that equilibrium scour58depth is governed by the grain size of bed material in uniform sand beds, it is important to notethat the difference in scour depth observed in this study did not occur due to a difference in mediangrain size, as the bed material of both models have the same median grain size. Consequently itcan proposed that the difference in depth distributions between the two models arises in large partdue to the absence or presence of a range of grain sizes.A bed composed of a non-uniform mixture of sediment sizes can act to limit scour: this self-stabilization process occurs as smaller particles, transported by the imposed flow conditions, aremoved out of the system, while larger particles remain in place and eventually make up the bedsurface (Parker and Klingeman, 1982; Gomez, 1983; Dietrich et al., 1989). Two lines of evidenceshow the role that armouring plays a role in limiting the depth of scour and stabilizing the bed.First of all, in Model 2, the median grain size of pool areas were all over 15% coarser than theaverage median surface grain size of the entire bed. This suggests that pools depths are limitedby the creation of an armour layer, which is consistent with previous studies (e.g. Borah, 1989;Kassem and Chaudhry, 2005). Second, median pool depths from the single grain size flume weredeeper then those found in the full GSD flume. This suggests that deeper scouring is more likelyto occur when the system does not have the ability to armour.Aside from the ability of the full GSD model to form an armour surface there are other pro-cesses or characteristics that may have contributed to the difference in depths between the twomodels. It is possible that the difference in the channel form imposed by the fixed-banks may havecontributed to the disparity between the channel morphologies given that single grain size modelcontained a large meander bend, however it was observed that the distribution of depths for Model1 remained wide even when the data from the meander bend was excluded. Discrepancies in theReynolds number and turbulent structures present in the channel, due to the differences in bed sur-face texture between the two models, may have also contributed to the two alternate morphologies,although this was not measured in any of the experiments.In previous experiments, it has been observed that when the bed material is composed of arange of grain sizes, changes to the governing variables can be compensated for through adjust-ments of the bed texture (Eaton and Church, 2009). These changes, either towards a coarser ora finer bed surface, can help to increase or decrease flow resistance within the reach in order toestablish a new stable form. That there was a shorter response time (i.e. time to equilibrium) in theexperiments conducted in the full grain size distribution model is consistent with these previousfindings, although it is important to consider that part of this discrepancy may arise as a result ofthe differences in the imposed shape and length of the two models.59Table 5.1: Summary of the Effects of Wood Addition, Wood Load, and GSDPresence of Wood Wood Load GSDTime toEquilibriumLonger time toequilibrium withwood presentNo discernible effect Longer time toequilibrium with asingle grain sizeSedimentStorageMore sediment storedin the presence ofwoodMore sediment storedwith higher wood loadNo discernible effectChannel slope Mean BedElevationIncreased mean bedelevation in the upperhalf of the flume withwood presentNo discernible effect No discernible effectMin. BedElevationMore variability in theminimum bedelevation with woodpresentNo discernible effect Less variability in theminimum bedelevation with a fullGSDReach-averagedSlopeHigher reach-averagedchannel slope withwood presentHigher reach-averagedchannel slope with ahigher wood loadNo discernible effectNumber ofFaciesIncreased number offacies with woodpresentNo discernible effect –Surface D50 Higher surface D50without wood presentSlightly higher surfaceD50 with a higherwood load–Distribution ofDepthsNo discernible effect No discernible effect Wider range of depthswith a single grainsize60Table 5.1 ContinuedPresence of Wood Wood Load GSDPool Spacing Decrease in poolspacing following theaddition of woodNo discernible effect No discernible effectPoolMorphologyArea Larger pool areas withwood presentNo discernible effect No discernible effectMax. Depth No discernible effect No discernible effect Deeper pool depthswith a single grainsizeDimensions More small, deeppools and fewer small,shallow pools withwood presentNo discernible effect More small, deeppools and fewer small,shallow pools with asingle grain size61Chapter 6ConclusionsThe morphology of stream channels is dictated by the interactions between discharge, sedimentflux and size distribution, and the characteristics of the surrounding environment. While fluvialsystems can accommodate minor adjustments to these governing variables without significantmorphologic change (Montgomery and Buffington, 1997; Eaton and Church, 2009), sudden orsustained large scale alterations to these conditions forces the morphology of the channel to adjustin order to re-establish a new equilibrium morphology (Howard, 1982). Analytic regime theoryhas been developed in order to help guide our understanding of how fluvial systems adjusts tosuch changes by advancing a physically-based understanding of the processes that drive channelmorphodynamics. The results from this study help to further this knowledge by isolating theeffects that two different roughness elements, bed texture and large wood, have on channel form.The findings of this study show that bed texture plays a large role in shaping channel morphologyregardless of whether it is the dominant roughness element in a stream channel. This highlightsthe necessity of considering the full grain size distribution when modelling channel response tochanges in governing variables.6.1 Grain Size DistributionThe role that bed grain size distribution plays in channel morphology is independent of the effectsof large wood as indicated by the fact that the range of depths recorded in the channel remainedhigher in the single grain size flume than in the full GSD flume regardless of the presence of LWin the channel. Maximum depths as much as 2.5 times the mean depth of the channel were foundusing a uniform sediment size, whereas maximum depths were limited to approximately 1.5 timesthe mean depth in the full GSD model. Furthermore, pools were deeper relative to pool area in thesingle grain size model.62The results of this study suggests that armouring played a role in limiting the range of depthsin the full GSD model. Armouring occurs as finer particles are transported away while largerparticles remain, eventually creating a coarse surface layer in areas of local flow convergence,thereby limiting the local depth of scour (Borah, 1989; Parker and Klingeman, 1982). This processis apparent in the full GSD model as the median surface grain size in pools was found to be higherthan the reach-averaged D50.It was proposed that range of particle sizes also facilitated the development of an equilibriumchannel form, as the response time of the single grain size flume to wood addition was almost twiceas long as it was for the full GSD flume. This discrepancy can also be linked to the availabilityof a range of particles sizes, as previous studies have found that changes to bed texture helpstabilize channel morphology (Parker and Klingeman, 1982; Eaton and Church, 2009). Theseresults suggest that, in addition to the sediment supply and sediment transport capacity (Howard,1982; Montgomery and Buffington, 1997), the grain size distribution of sediment present in astream channel plays a significant role in the response time of the system.6.2 The Presence of Large WoodMorphologic changes occurred at both a reach-scale and bed-form scale in response to the additionof wood into the flume. At a reach scale, the presence of LW increased flow resistance anddecreased the sediment transport capacity within the channel. Individual wood pieces and log jamsinfluenced local channel hydraulics, leading to changes in the pattern of erosion and depositionwithin the flume.Sediment storage increased following LW addition in response to the decrease in sedimenttransport capacity. Sediment storage occurred disproportionally in the areas upstream of the largestlog jam(s) in the flume, with some reaches below the log jams experiencing net degradation. Thisincrease in sediment storage coincided with a increase in bed slope following wood addition. Inthe full grain size model, median grain size decreased following the addition of wood, due to thefact that the wood pieces impounded a greater amount of finer material in the flume.Prior to the addition of wood, the location of pools and bars was dictated by the alignment ofthe non-erodible flume walls. The introduction of LW into the channel caused new areas of scourand deposition to be imposed onto the pre-established background form. In the full GSD flume,the number of distinct facies almost doubled with wood present due to the increased heterogeneityof flow patterns. Local flow convergence around LW pieces induced additional areas of scour, asseen by a decrease in pool spacing from three to four channel widths to about two channel widths.Both pool area and maximum pool depth were found to decrease as a result of this change in63channel hydraulics.6.3 Wood LoadWhile wood load was found to influence the total sediment stored in the flume and the magnitudeof change in reach-averaged bed slope, it was found to have little effect on channel morphologyat a bed-form scale, as characteristics such as pool spacing and pool morphology were similarregardless of wood load. As such, it can be concluded that the system re-equilibrates its mor-phology to an imposed wood load by altering its reach scale resistance as opposed to its bed formresistance.Instead of being driven by the wood load of the channel, pool characteristics were driveninstead by local flow patterns around LW, either as individual pieces or amalgamated as log jams(Abbe and Montgomery, 1996). The results from this study suggest that at a local scale, the totalamount of wood present in the surrounding reach has little effect on the patterns of scour anddeposition that are imposed by individual wood pieces.Wood load was found to dictate the total amount of reach-scale change that must occur inorder for a new equilibrium form to be attained. More sediment storage, and consequently asteeper equilibrium slope was observed under the conditions of a high wood load.6.4 Applications of this StudyAll methods of modelling require the reduction of a complex system. This can mean that compo-nents of the system are omitted or certain relationships or variables are simplified. An exampleof this is the use of a single grain size or a highly distorted grain size distribution to represent therange of grain sizes found in a real channel (e.g. Wallerstein et al., 2001; Malverti et al., 2008).While this application is necessary in order to isolate relationships between individual variablesor, in the case of numerical modelling, to reduce the computation power required, it runs the riskthat processes are not accurately portrayed.The results of this study show that it is important to consider the full range of sediment sizeswhen attempting to model equilibrium channel morphodynamics. The capacity of a stream chan-nel to armour and to limit the depth of scour plays a key role in dictating the response time of asystem to changes in the governing variables, both with and without wood present. The studiesthat disregard or oversimplify a modelled grain size distribution may poorly predict the properrange of depths that occur in a channel, in particular the depth of scour.This study also presented a novel method of identifying pool areas from DEMs of depth.While previous studies used other methods of selecting pool areas (e.g. Montgomery and Buff-64ington, 1998), these methods have a tendency to be somewhat subjective as they often requirethe researcher to qualitatively identify and measure morphologic features such as riffle tops orpool depths. By selecting for areas greater than one standard deviation above the mean depth ofthe reach, this method can be applied to any fluvial system, regardless of the morphology of thechannel, as it uses the unique depth distribution of the channel in question.6.5 Future WorkPhysical modelling provides an excellent opportunity to further our understanding of how equilib-rium conditions are met in a range of scenarios. While this study has broadened our understandingof the types of variables that are influenced by the addition of wood into the channel, the next stepis to further understand the processes that drive the observed outcomes.While this study examined the effects of mobile wood pieces added to the channel, manystream restoration projects employ engineered log jams (ELJ) which are stable, fixed structures.Consequently, understanding how a channel adjusts to the addition of these ELJ as opposed to theaddition of individual mobile wood pieces would prove to be useful when utilizing such methodsin rehabilitation projects. Given that, in the absence of LW, this study showed that equilibriumchannel form was dictated by the pattern of the stable, non-erodible banks, it could be hypothe-sized that, like channel banks, the presence of stable log jams should produce comparable channelmorphologies between experiments if all other variables were held constant.Results from this study suggest that the mobility of LW pieces contributes to the time it takesfor the channel to come into equilibrium as the small adjustments in wood location that occurthroughout the length of the experiment prolong the establishing of a stable channel form. If thelocation of LW pieces were stable throughout the length of the experiment, it is predicted that theresponse time of the system would be shorter.Broadening our understanding of channel response to changes in governing conditions is keyto building better regime models in order to better predict how our fluvial systems respond tochanges in governing conditions. The results from this study exemplify the contribution thatphysical modelling can make towards how we understand and model fluvial systems.65BibliographyAbbe, T. and Montgomery, D. 1996. Large woody debris jams, channel hydraulics and habitatformation in large rivers. Regulated Rivers Research & Management, 12(23):201–221. →pages 3, 34, 55, 56, 64Abbe, T. and Montgomery, D. 2003. Patterns and processes of wood debris accumulation in theQueets River Basin, washington. Geomorphology, 51(1):81–107. → pages 2Andreoli, A., Comiti, F., and Lenzi, M. 2007. Characteristics, distribution and geomorphic roleof large woody debris in a mountain stream of the Chilean Andes. Earth Surface Processesand Landforms, 32(11):1675–1692. → pages 2Andrews, C. 2010. A Stream in Transition: Short Term Morphodynamics of Fishtrap CreekFollowing Wildfire. Master’s thesis, University of British Columbia. → pages 16, 18Ashmore, P. 1991a. Channel morphology and bed load pulses in braided, gravel-bed streams.Geografiska Annaler. Series A. Physical Geography, pages 37–52. → pages 4Ashmore, P. 1991b. How do gravel-bed rivers braid? Canadian Journal of Earth Sciences,28(3):326–341. → pages 7Ashmore, P. E. 1988. Bed load transport in braided gravel-bed stream models. Earth SurfaceProcesses and Landforms, 13(8):677–695. → pages 7Assani, A. and Petit, F. 1995. Log-jam effects on bed-load mobility from experiments conductedin a small gravel-bed forest ditch. Catena, 25(1):117–126. → pages 56, 57Bennett, R., Welham, K., Ford, A., et al. 2012. A comparison of visualization techniques formodels created from airborne laser scanned data. Archaeological Prospection, 19(1):41–48.→ pages 80, 81Bilby, R. and Ward, J. 1991. Characteristics and function of large woody debris in streamsdraining old-growth, clear-cut, and second-growth forests in southwestern washington.Canadian Journal of Fisheries and Aquatic Sciences, 48(12):2499–2508. → pages 34Bocchiola, D. 2011. Hydraulic characteristics and habitat suitability in presence of woody debris:A flume experiment. Advances in Water Resources, 34(10):1304–1319. → pages 2, 7, 11, 1566Bocchiola, D., Rulli, M., and Rosso, R. 2008. A flume experiment on the formation of woodjams in rivers. Water Resources Research, 44(2). → pages 57Borah, D. K. 1989. Scour-depth prediction under armoring conditions. Journal of HydraulicEngineering, 115(10):1421–1425. → pages 59, 63Braudrick, C., Dietrich, W., Leverich, G., and Sklar, L. 2009. Experimental evidence for theconditions necessary to sustain meandering in coarse-bedded rivers. Proceedings of theNational Academy of Sciences, 106(40):16936–16941. → pages 7, 13Braudrick, C. A. and Grant, G. E. 2000. When do logs move in rivers? Water resources research,36(2):571–583. → pages 11, 15Braudrick, C. A. and Grant, G. E. 2001. Transport and deposition of large woody debris instreams: a flume experiment. Geomorphology, 41(4):263–283. → pages 6, 15Braudrick, C. A., Grant, G. E., Ishikawa, Y., and Ikeda, H. 1997. Dynamics of wood transport instreams: a flume experiment. Earth Surface Processes and Landforms, 22(7):669–683. →pages 15Brummer, C., Abbe, T., Sampson, J., and Montgomery, D. 2006. Influence of vertical channelchange associated with wood accumulations on delineating channel migration zones,Washington, USA. Geomorphology, 80(3):295–309. → pages 2Bruun, P. 1966. Model geology: prototype and laboratory streams. Geological Society ofAmerica Bulletin, 77(9):959–974. → pages 7Buckingham, E. 1915. Model experiments and the forms of empirical equations. → pages 7Buffington, J. M., Lisle, T. E., Woodsmith, R. D., and Hilton, S. 2002. Controls on the size andoccurrence of pools in coarse-grained forest rivers. River Research and Applications,18(6):507–531. → pages 3Buffington, J. M. and Montgomery, D. R. 1999a. Effects of hydraulic roughness on surfacetextures of gravel-bed rivers. Water Resources Research, 35(11):3507–3521. → pages 2, 53,54Buffington, J. M. and Montgomery, D. R. 1999b. Effects of sediment supply on surface texturesof gravel-bed rivers. Water Resources Research, 35(11):3523–3530. → pages 2, 54Buffington, J. M., Montgomery, D. R., and Greenberg, H. M. 2004. Basin-scale availability ofsalmonid spawning gravel as influenced by channel type and hydraulic roughness in mountaincatchments. Canadian Journal of Fisheries and Aquatic Sciences, 61(11):2085–2096. →pages 24Chang, H. H. 1979. Minimum stream power and river channel patterns. Journal of Hydrology,41(3):303–327. → pages 4, 567Cherry, J. and Beschta, R. 1989. Coarse woody debris and channel morphology: A flume study.JAWRA Journal of the American Water Resources Association, 25(5):1031–1036. → pages 7,56Church, M., Hassan, M. A., and Wolcott, J. F. 1998. Stabilizing self-organized structures ingravel-bed stream channels: Field and experimental observations. Water Resources Research,34(11):3169–3179. → pages 2, 4, 54Clark, C. D. 1997. Reconstructing the evolutionary dynamics of former ice sheets usingmulti-temporal evidence, remote sensing and gis. Quaternary Science Reviews,16(9):1067–1092. → pages 77Clarke, R. 2007. Hydrological prediction in a non-stationary world. Hydrology and Earth SystemSciences, 11(1):408–414. → pages 4Davidson, S. 2011. Modeling channel morphodynamics associated with large wood in anintermediate-sized stream. Master’s thesis, University of British Columbia. → pages 8, 9, 15,18, 23, 25, 26, 36Davidson, S. L. and Eaton, B. C. 2012. Modeling channel morphodynamic response to variationsin large wood: Implications for stream rehabilitation in degraded watersheds. Geomorphology.→ pages 2, 18, 23, 56Davies, T. and Lee, A. 1988. Physical hydraulic modelling of width reduction and bed levelchange in braided rivers. Journal of hydrology. New Zealand, 27(2):113–127. → pages 7Davies, T., McSaveney, M., and Clarkson, P. 2003. Anthropic aggradation of the Waiho River,Westland, New Zealand: microscale modelling. Earth Surface Processes and Landforms,28(2):209–218. → pages 14Davies, T. R. and Sutherland, A. J. 1983. Extremal hypotheses for river behavior. WaterResources Research, 19(1):141–148. → pages 4, 5De´ry, S. J., Stahl, K., Moore, R., Whitfield, P., Menounos, B., and Burford, J. E. 2009. Detectionof runoff timing changes in pluvial, nival, and glacial rivers of western Canada. WaterResources Research, 45(4). → pages 1Devereux, B., Amable, G., and Crow, P. 2008. Visualisation of LiDAR terrain models forarchaeological feature detection. Antiquity, 82(316):470–479. → pages 80, 81Dey, S. and Raikar, R. V. 2007. Clear-water scour at piers in sand beds with an armor layer ofgravels. Journal of Hydraulic Engineering, 133(6):703–711. → pages 7Dietrich, W. E., Kirchner, J. W., Ikeda, H., and Iseya, F. 1989. Sediment supply and thedevelopment of the coarse surface layer in gravel-bedded rivers. Nature, 340(6230):215–217.→ pages 54, 5968Dunne, T. and Leopold, L. B. 1978. Water in environmental planning. W.H. Freeman andCompany, New York. → pages 1Eaton, B. 2008. Potential Channel Changes Following Wildfire: Applying the UBC RegimeModel to Fishtrap Creek, British Columbia. Streamline Water Management Bulletin,12(1):28–33. → pages 16Eaton, B., Andrews, C., Giles, T., and Phillips, J. 2010a. Wildfire, morphologic change and bedmaterial transport at Fishtrap Creek, British Columbia. Geomorphology, 118(3):409–424. →pages 15, 16, 19, 23, 24, 25, 56Eaton, B. and Church, M. 2004. A graded stream response relation for bed load–dominatedstreams. Journal of geophysical research, 109(F3):F03011. → pages 1, 2, 4, 7, 8, 9, 13, 54, 56Eaton, B. and Church, M. 2009. Channel stability in bed load-dominated streams withnon-erodible banks: Inferences from experiments in a sinous flume. Journal of GeophysicalResearch, 114:F01024, doi:10.1029/2007JF000902. → pages 8, 9, 13, 53, 57, 59, 62, 63Eaton, B., Giles, T., Heise, B., Moore, R. D., Owens, P., and Petticrew, E. 2010b. Fishtrap Creekwatershed project. Streamline Water Management Bulletin, 14:12–13. → pages 18Eaton, B., Moore, R., and Giles, T. 2010c. Forest Fire, bank strength, and channel instability:The ”unusual” response of Fishtrap Creek, British Columbia. Earth Surface Processes andLandforms, 35(10):1167–1183. → pages 16, 56Eaton, B. C., Church, M., and Millar, R. G. 2004. Rational regime model of alluvial channelmorphology and response. Earth Surface Processes and Landforms, 29(4):511–529. → pages4, 5, 7Eaton, B. C. and Giles, T. R. 2009. Assessing the effect of vegetation-related bank strength onchannel morphology and stability in gravel-bed streams using numerical models. EarthSurface Processes and Landforms, 34(5):712–724. → pages 2, 16Eaton, B. C. and Millar, R. G. 2004. Optimal alluvial channel width under a bank stabilityconstraint. Geomorphology, 62(1):35–45. → pages 2Gaines, R. and Maynord, S. 2001. Microscale loose-bed hydraulic models. Journal of HydraulicEngineering, 127(5):335–338. → pages 13Gantenbein, C. 2012. Creating shaded relief for geologic mapping using multiple light sources.In Digital Mapping Techniques ’10 - Workshop Proceedings. → pages 78, 79Garcia, M. 1993. Hydraulic jumps in sediment-driven bottom currents. Journal of HydraulicEngineering, 119(10):1094–1117. → pages 13Gilbert, G. K. and Murphy, E. C. 1914. The transportation of debris by running water,volume 86. US Government Printing Office. → pages 769Gill, M. A. 1972. Erosion of sand beds around spur dikes. Journal of the Hydraulics Division,98(hy9). → pages 58Glenn, N. F., Streutker, D. R., Chadwick, D. J., Thackray, G. D., and Dorsch, S. J. 2006. Analysisof LiDAR-derived topographic information for characterizing and differentiating landslidemorphology and activity. Geomorphology, 73(1):131–148. → pages 77Gomez, B. 1983. Temporal variations in the particle size distribution of surficial bed material:The effect of progressive bed armouring. Geografiska annaler. Series A. Physical geography,65(3-4):183–192. → pages 59Gomi, T., Sidle, R. C., Bryant, M. D., and Woodsmith, R. D. 2001. The characteristics of woodydebris and sediment distribution in headwater streams, southeastern Alaska. Canadian Journalof Forest Research, 31(8):1386–1399. → pages 2Hassan, M. A., Hogan, D. L., Bird, S. A., May, C. L., Gomi, T., and Campbell, D. 2005. Spatialand Temporal Dynamics of Wood in Headwater Streams of the Pacific Northwest. JAWRAJournal of the American Water Resources Association, 41(4):899–919. → pages 2Haugerud, R. A., Harding, D. J., Johnson, S. Y., Harless, J. L., Weaver, C. S., and Sherrod, B. L.2003. High-resolution lidar topography of the Puget Lowland, Washington. GSA Today,13(6):4–10. → pages 77Hebertson, J. G. 1969. Scaling procedures for mobile bed hydraulic models in terms ofsimilitude theory. Journal of Hydraulic Research, 7(3):315–353. → pages 7Hobbs, F. 1995. The rendering of relief images from digital contour data. Cartographic Journal,The, 32(2):111–116. → pages 79, 80Hobbs, K. F. 1999. An investigation of RGB multi-band shading for relief visualisation.International Journal of applied earth observation and geoinformation, 1(3):181–186. →pages 79, 80Hooke, R. 1968. Model geology: prototype and laboratory streams: discussion. GeologicalSociety of America Bulletin, 79(3):391–394. → pages 6, 7, 14Howard, A. 1988. Equilibrium models in geomorphology. Modelling GeomorphologicalSystems, pages 49–72. → pages 3, 54Howard, A. D. 1982. Equilibrium and time scales in geomorphology: Application to sand-bedalluvial streams. Earth Surface Processes and Landforms, 7(4):303–325. → pages 3, 62, 63Jencso, K. G., McGlynn, B. L., Gooseff, M. N., Wondzell, S. M., Bencala, K. E., and Marshall,L. A. 2009. Hydrologic connectivity between landscapes and streams: Transferring reach-andplot-scale understanding to the catchment scale. Water Resources Research, 45(4). → pages 1Jenny, B. and Hurni, L. 2006. Swiss-style colour relief shading modulated by elevationand byexposure to illumination. Cartographic Journal, The, 43(3):198–207. → pages 8070Jones, A. F., Brewer, P. A., Johnstone, E., and Macklin, M. G. 2007. High-resolutioninterpretative geomorphological mapping of river valley environments using airborne LiDARdata. Earth Surface Processes and Landforms, 32(10):1574–1592. → pages 77Kassem, A. and Chaudhry, M. H. 2005. Effect of bed armoring on bed topography of channelbends. Journal of hydraulic engineering, 131(12):1136–1140. → pages 59Keller, E. and Swanson, F. 1979. Effects of large organic material on channel form and fluvialprocesses. Earth surface processes, 4(4):361–380. → pages 56, 57Keller, E. A., Valentine, D. W., and Gibbs, D. R. 1997. Hydrological response of smallwatersheds following the Southern California Painted Cave Fire of June 1990. HydrologicalProcesses, 11(4):401–414. → pages 16Kennelly, P. J. 2008. Terrain maps displaying hill-shading with curvature. Geomorphology,102(3):567–577. → pages 77, 80Kennelly, P. J. and Kimerling, A. J. 2004. Hillshading of terrain using layer tints withaspect-variant luminosity. Cartography and Geographic Information Science, 31(2):67–77. →pages 80Kirkby, M. 1977. Maximum sediment efficiency as a criterion for alluvial channels. Universityof Leeds, School of Geography. → pages 4, 5Koss, J., Ethridge, F., and Schumm, S. 1994. An experimental study of the effects of base-levelchange on fluvial, coastal plain and shelf systems. Journal of Sedimentary Research, 64(2). →pages 14Kuhnle, R. A., Alonso, C. V., and Shields Jr, F. D. 2002. Local scour associated with angled spurdikes. Journal of Hydraulic Engineering, 128(12):1087–1093. → pages 7, 55Lajeunesse, E., Malverti, L., Lancien, P., Armstrong, L., Me´tivier, F., Coleman, S., Smith, C.,Davies, T., Cantelli, A., and Parker, G. 2009. Fluvial and submarine morphodynamics oflaminar and near-laminar flows: A synthesis. Sedimentology, 57(1):1–26. → pages 14Lancaster, S. T., Hayes, S. K., and Grant, G. E. 2001. Modeling sediment and wood storage anddynamics in small mountainous watersheds. Water Science and Application, 4:85–102. →pages 3Lane, E. W. 1955. Design of stable channels. Transactions of the American Society of CivilEngineers, 120(1):1234–1260. → pages 3Lane, E. W. 1957. A study of the shape of channels formed by natural streams flowing in erodiblematerial. US Army Engineer Division, Missouri River. → pages 1Lasaponara, R. and Masini, N. 2011. On the processing of aerial LiDAR data for supportingenhancement, interpretation and mapping of archaeological features. In ComputationalScience and Its Applications-ICCSA 2011, pages 392–406. Springer. → pages 80, 8171Leach, J. and Moore, R. 2010. Above-stream microclimate and stream surface energy exchangesin a wildfire-disturbed riparian zone. Hydrological Processes, 24(17):2369–2381. → pages 16Lidmar-Bergstro¨m, K., Elvhage, C., and Ringberg, B. 1991. Landforms in Ska˚ne, South Sweden.Geografiska Annaler. Series A. Physical Geography, pages 61–91. → pages 77, 79Lisle, T. 1987. Using” residual depths” to monitor pool depths independently of discharge. →pages 34Lisle, T. E. 1986. Stabilization of a gravel channel by large streamside obstructions and bedrockbends, Jacoby Creek, northwestern California. Geological Society of America Bulletin,97(8):999–1011. → pages 3Lisle, T. E., Iseya, F., and Ikeda, H. 1993. Response of a channel with alternate bars to a decreasein supply of mixed-size bed load: A flume experiment. Water Resources Research,29(11):3623–3629. → pages 54Loisios, D., Tzelepis, N., and Nakos, B. 2007. A methodology for creating analyticalhill-shading by combining different lighting directions. In Proceedings of 23rd InternationalCartographic Conference, Moscow. → pages 78, 79, 80, 88Lukas, K. and Weibel, R. 1995. Assessment and improvement of methods for analyticalhillshading. In Proc. 17 th Internat. Cartogr. Conf., Barcelona, volume 2, pages 2231–2240.→ pages 78, 79, 80Mackin, J. H. 1948. Concept of the graded river. Geological Society of America Bulletin,59(5):463–512. → pages 1Madej, M., Sutherland, D., Lisle, T., and Pryor, B. 2009. Channel responses to varying sedimentinput: A flume experiment modeled after Redwood Creek, California. Geomorphology,103(4):507–519. → pages 7, 8, 12, 13, 14Malverti, L., Lajeunesse, E., and Me´tivier, F. 2008. Small is beautiful: upscaling from microscalelaminar to natural turbulent rivers. Journal of Geophysical Research, 113(F4):F04004. →pages 12, 13, 14, 64Manga, M. and Kirchner, J. W. 2000. Stress partitioning in streams by large woody debris. WaterResources Research, 36(8):2373–2379. → pages 2, 58Mark, R. K. 1992. A multidirectional, oblique-weighted, shaded-relief image of the island ofHawaii. US Department of the Interior, US Geological Survey. → pages 79, 80, 81, 84, 88May, C. L. and Gresswell, R. E. 2003. Processes and rates of sediment and wood accumulation inheadwater streams of the Oregon Coast Range, USA. Earth Surface Processes and Landforms,28(4):409–424. → pages 2, 5772Megahan, W. F. 1982. Channel sediment storage behind obstructions in forested drainage basinsdraining the granitic bedrock of the Idaho batholith. Sediment budgets and routing in foresteddrainage basins, pages 114–121. → pages 2, 58Millar, R. G. 2005. Theoretical regime equations for mobile gravel-bed rivers with stable banks.Geomorphology, 64(3):207–220. → pages 2, 3, 5Millar, R. G. and Quick, M. C. 1993. Effect of bank stability on geometry of gravel rivers.Journal of Hydraulic Engineering, 119(12):1343–1363. → pages 1, 4, 5Montgomery, D., Buffington, J., Smith, R., Schmidt, K., and Pess, G. 1995. Pool spacing inforest channels. Water Resources Research, 31(4):1097–1105. → pages 34, 49, 50Montgomery, D., Collins, B., Buffington, J., and T.B., A. 2003. Geomorphic effects of wood inrivers. In American Fisheries Society Symposium. American Fisheries Society. → pages 3Montgomery, D. R. and Buffington, J. M. 1997. Channel-reach morphology in mountain drainagebasins. Geological Society of America Bulletin, 109(5):596–611. → pages 2, 56, 62, 63Montgomery, D. R. and Buffington, J. M. 1998. Channel processes, classification, and response.River ecology and management, 112:1250–1263. → pages 1, 3, 64Notebaert, B., Verstraeten, G., Govers, G., and Poesen, J. 2009. Qualitative and quantitativeapplications of LiDAR imagery in fluvial geomorphology. Earth Surface Processes andLandforms, 34(2):217–231. → pages 77, 79Onorati, G., Ventura, R., Poscolieri, M., Chiarini, V., and Crucilla, U. 1992. The digital elevationmodel of italy for geomorphology and structural geology. Catena, 19(2):147–178. → pages 79Pagliara, S. and Carnacina, I. 2010. Temporal scour evolution at bridge piers: Effect of wooddebris roughness and porosity. Journal of Hydraulic Research, 48(1):3–13. → pages 7Parker, G. and Klingeman, P. C. 1982. On why gravel bed streams are paved. Water ResourcesResearch, 18(5):1409–1423. → pages 54, 59, 63Pawley, S. and Atkinson, N. 2010. Application of airbourne LiDAR imagery for surficial geologymapping in densely vegetated (Boreal Zone) terrain, Lesser Slave Lake, Northern Alberta.Poster. → pages 80, 81Peakall, J., Ashworth, P., and Best, J. 1996. Physical modelling in fluvial geomorphology:principles, applications and unresolved issues. The Scientific Nature of Geomorphology, pages221–253. → pages 11, 14Petticrew, E. L., Owens, P. N., and Giles, T. R. 2006. Wildfire effects on the quantity andcomposition of suspended and gravel-stored sediments. In The Interactions BetweenSediments and Water, pages 283–292. Springer. → pages 1673Phillips, J. 2007. Post-Fire Dynamics of a Gravel Bed Stream: Fishtrap Creek, British Columbia.Master’s thesis, University of British Columbia. → pages 16Podobnikar, T. 2012. Multidirectional visibility index for analytical shading enhancement.Cartographic Journal, The, 49(3):195–207. → pages 80Pohl, C. and Van Genderen, J. 1998. Review article multisensor image fusion in remote sensing:concepts, methods and applications. International journal of remote sensing, 19(5):823–854.→ pages 80Pollen, N. and Simon, A. 2006. Geotechnical implications for the use of alfalfa in experimentalstudies of alluvial-channel morphology and planform. In AGU Fall Meeting Abstracts,volume 1, page 1455. → pages 13Postma, G., Kleinhans, M., Meijer, P., and Eggenhuisen, J. 2008. Sediment transport in analogueflume models compared with real-world sedimentary systems: A new look at scaling evolutionof sedimentary systems in a flume. Sedimentology, 55(6):1541–1557. → pages 7, 14Raudkivi, A. J. and Ettema, R. 1985. Scour at cylindrical bridge piers in armored beds. Journalof Hydraulic Engineering, 111(4):713–731. → pages 55Renwick, W. H. 1992. Equilibrium, disequilibrium, and nonequilibrium landforms in thelandscape. Geomorphology, 5(3):265–276. → pages 3Ryder, J. M., Fulton, R. J., and Clague, J. J. 1991. The Cordilleran Ice Sheet and the glacialgeomorphology of southern and central British Colombia. Ge´ographie physique etQuaternaire, 45(3):365–377. → pages 15Sandford, R. W. 2012. Cold matters: the state and fate of Canada’s fresh water. RockyMountain Books Ltd. → pages 4Schowengerdt, R. A. 2006. Remote sensing: models and methods for image processing.Academic press. → pages 80Schumm, S. and Khan, H. 1972. Experimental study of channel patterns. Geological Society ofAmerica Bulletin, 83(6):1755–1770. → pages 7, 12, 13Schumm, S. A., Mosley, M. P., and Weaver, W. 1987. Experimental fluvial geomorphology. →pages 7Shook, K. and Pomeroy, J. 2012. Changes in the hydrological character of rainfall on theCanadian prairies. Hydrological Processes, 26(12):1752–1766. → pages 4Smith, M. J. and Clark, C. D. 2005. Methods for the visualization of digital elevation models forlandform mapping. Earth Surface Processes and Landforms, 30(7):885–900. → pages 78, 79,8074Smith, M. J. and Wise, S. M. 2007. Problems of bias in mapping linear landforms from satelliteimagery. International Journal of Applied Earth Observation and Geoinformation,9(1):65–78. → pages 78Stromberg, J., Briggs, M., Gourley, C., Scott, M., Shafroth, P., Stevens, L., Baker Jr, M., Ffolliott,P., DeBano, L., and Neary, D. 2004. Human alterations of riparian ecosystems. Riparian Areasof the Southwestern United States: Hydrology, Ecology, and Management, pages 101–126. →pages 4Sˇtular, B., Kokalj, Zˇ., Osˇtir, K., and Nuninger, L. 2012. Visualization of LiDAR-derived reliefmodels for detection of archaeological features. Journal of Archaeological Science,39(11):3354–3360. → pages 80Tal, M., Gran, K., Murray, A. B., Paola, C., and Hicks, D. M. 2004. Riparian vegetation as aprimary control on channel characteristics in multi-thread rivers. Riparian vegetation andfluvial geomorphology, pages 43–58. → pages 13Tal, M. and Paola, C. 2007. Dynamic single-thread channels maintained by the interaction offlow and vegetation. Geology, 35(4):347–350. → pages 13Tecle, A., Bojonell, H. A., and King, J. G. 2001. Modeling sediment storage behind obstructionsin small headwater streams. Journal of the Arizona-Nevada Academy of Science, pages 83–92.→ pages 2Termini, D. 2009. Experimental observations of flow and bed processes in large-amplitudemeandering flume. Journal of Hydraulic Engineering, 135(7):575–587. → pages 7Thompson, D. 1995. The effects of large organic debris on sediment processes and streammorphology in Vermont. Geomorphology, 11(3):235–244. → pages 2, 3Van Heijst, M. and Postma, G. 2002. Fluvial response to sea-level changes: a quantitativeanalogue, experimental approach. Basin Research, 13(3):269–292. → pages 14Van Heijst, M., Postma, G., Meijer, X., Snow, J., and Anderson, J. 2002. Quantitative analogueflume-model study of river–shelf systems: principles and verification exemplified by the LateQuaternary Colorado river–delta evolution. Basin Research, 13(3):243–268. → pages 14Wallerstein, N. P., Alonso, C. V., Bennett, S. J., and Thorne, C. R. 2001. Distorted Froude-scaledflume analysis of large woody debris. Earth Surface Processes and Landforms,26(12):1265–1283. → pages 6, 12, 13, 15, 56, 64Webb, A. A. and Erskine, W. D. 2003. Distribution, recruitment, and geomorphic significance oflarge woody debris in an alluvial forest stream: Tonghi Creek, southeastern Australia.Geomorphology, 51(1):109–126. → pages 2, 57White, W.R., B. R. and Paris, E. 1982. An analytical approach to river regime. Journal of theHydraulics Division, 108(10):1179–1193. → pages 575Whiting, P. and Dietrich, W. 1993. Experimental studies of bed topography and flow patterns inlarge-amplitude meanders: 1. observations. Water Resources Research, 29(11):3605–3614. →pages 7, 12, 13, 34Wilcox, A. C., Nelson, J. M., and Wohl, E. E. 2006. Flow resistance dynamics in step-poolchannels: 2. partitioning between grain, spill, and woody debris resistance. Water ResourcesResearch, 42(5). → pages 2, 4Wilcox, A. C. and Wohl, E. E. 2006. Flow resistance dynamics in step-pool stream channels: 1.large woody debris and controls on total resistance. Water Resources Research, 42(5). →pages 58Wohl, E. 2011. Threshold-induced complex behavior of wood in mountain streams. Geology,39(6):587–590. → pages 3Wolman, M. G. and Miller, J. P. 1960. Magnitude and frequency of forces in geomorphicprocesses. The Journal of Geology, pages 54–74. → pages 1Wondzell, S. M. and King, J. G. 2003. Postfire erosional processes in the Pacific Northwest andRocky Mountain regions. Forest Ecology and Management, 178(1):75–87. → pages 56Yalin, M. S. 1971. Theory of hydraulic models, volume 266. Macmillan London. → pages 11Young, W. and Davies, T. 1991. Bedload transport processes in a braided gravel-bed river model.Earth Surface Processes and Landforms, 16(6):499–511. → pages 7Zelt, R. B. and Wohl, E. E. 2004. Channel and woody debris characteristics in adjacent burnedand unburned watersheds a decade after wildfire, Park County, Wyoming. Geomorphology,57(3):217–233. → pages 56Zimmermann, A. E. 2009. Experimental investigations of step-pool channel formation andstability. PhD thesis, University of British Columbia. → pages 2576Appendix ALiDAR VisualizationA.1 LiDAR: Remotely Sensed DataLiDAR (Light Detection And Ranging) is a type of remote sensing technology that measuresthe distance between the source of a light beam and a remote target surface in order to create adigital elevation model (DEM) where the value of each cell in the resulting raster image representsthe elevation of the terrain at that point. Airborne LiDAR data, collected using a downward-looking LiDAR system mounted on a plane, is of great importance to terrain mapping and othergeomorphological endeavors due to its high spatial resolution and the ability to filter out vegetationduring the processing of the data in order to create a “bare earth” model of a land surface.Many types of geomorphic maps have been created using LiDAR-derived DEM. In glacialapplications, DEMs are commonly used to delineate drumlins, eskers, spillways, and other glacialfeatures (e.g. Lidmar-Bergstro¨m et al., 1991; Haugerud et al., 2003) whose orientations and thatcan be used to reconstruct the dynamics of former ice sheets (Clark, 1997). In fluvial applications,DEMs have been used to conduct watershed assessments and determine flood risk (e.g. Joneset al., 2007; Notebaert et al., 2009). LiDAR data has also proven useful for locating mass-wastingevents in order to monitor landslide activity (e.g. Glenn et al., 2006).Because digital elevation models represent the distributions of elevations for a location theycan be used to derive other types of data about the landscape (Kennelly, 2008). Slope gradient(the rate of change in elevation) and slope aspect (the compass direction of the steepest slope) arefirst order derivatives of a DEMs while profile and planimetric curvature (measured across andalong the direction of maximum slope) represent second order derivatives. While elevation andfirst and second order derivatives may be visualized as rasters, these “unbiased” representationsof the landscape are often unintuitive to interpret. In order to get a real sense of the landscape it is77often necessary to employ other visualization techniques.Originally developed as a complex hand-shading process, hill-shading, or analytical shading(when referring specifically to the computational procedure), is a process in which a diffuse reflec-tion illumination model is applied to a digital elevation (or terrain) model to create a raster image.Compared to a DEM and its first and second order derivatives, these ”illuminated,” hill-shadedlandscapes more closely resemble what a land surface looks like when lit by the sun. This methodis also easy to apply as most GIS programs have a hill-shading function that allows the user toeasily create one from an inputted DEM.Issues arise when a single hill-shaded image (i.e. created from one illumination source) isused in terrain interpretation (Lukas and Weibel, 1995; Loisios et al., 2007; Smith and Wise,2007). Landforms, especially high ridges, running perpendicular to the direction of illuminationexperience overexposure of light on one side and improper or lack of lighting on the other. Linearfeatures, such as drumlins, ridges, and stream channels, that run parallel to the light source, blendinto to the background as a result of both sides being illuminated equally (Lukas and Weibel,1995). Additionally, if the light source is projected from a southern or easterly direction, reliefinversion may occur in which the viewer perceives that concave landforms (e.g. stream channel)protrude from the landscape and convex landforms (e.g. ridges) appear incised (Lukas and Weibel,1995; Gantenbein, 2012).A.1.1 Research ObjectivesThe illumination bias of single hill-shaded images has been accounted for through various vi-sualization methods (see Smith and Clark, 2005 for a review). Because these different methodsrange in complexity and their portrayal of the landscape, their applicability depends on the goalof the study for which they are employed. The first half of this paper will discuss the technique ofhill-shading and will address issues and methods pertaining to it.Given that many field-based geomorphic studies now employ DEMs as a tool from which toplan field sites or to get a sense of the landscape prior to field work, it is necessary to maximizethe clarity of the DEM visualization without requiring that the user have a large understandingof different visualization methods or take too much time processing the data. The second half ofthe paper will apply several visualization methods to a dataset in order to determine a simple yeteffective method of visualizing a LiDAR-derived DEM for use in glacial and fluvial geomorpho-logical studies. Methods considered in this paper must be relatively quick and easy to do and mustnot require extensive knowledge of GIS.78A.2 LiDAR Visualization MethodsWhen creating a hill-shaded image it is necessary to define an azimuth, the altitude and any verticalexaggeration (aka. the z factor). Hill-shaded tools in most GIS applications have default valuesfor each of these variables. The defaults mentioned below pertain to those found in ArcGIS™.Azimuth is the compass direction, in degrees (i.e. between 0° and 360°), of the light source.The default azimuth is 315°, which represents a light source shining from the north-west. Thisdefault azimuth creates a visually appealing landscape (Gantenbein, 2012) but biases against land-forms parallel to that direction, as will be shown later. Because of this, knowing something aboutthe direction of linear features within the landscape prior to choosing the azimuth is valuable whenproducing a single hill-shaded image.The altitude is the angle of the light source up from the horizon (i.e. between 0° and 90°). If ahigh altitude is chosen, the terrain will appear “washed out” while the selection of a low altitudewill cause more of the landscape to be shadowed. The default altitude is 45°, which is high enoughto minimize shadows, but low enough that the surface does not appear expressionless.Although the default choice is for there to be no vertical exaggeration, choosing a value greaterthan one may help to accentuate low lying terrain features. However, like with a low altitude, thismust be balanced out against the shadowing of terrain due to the increased relief.Hill-shaded images may be combined either through raster addition or by layering the two andchanging the opacity of the overlying images to create a composite image that exhibits less biasthan a single hill-shaded image alone. A simple and commonly employed method is to combinetwo orthogonally oriented hill-shaded images in order to enhance the visibility of linear features(e.g. Lidmar-Bergstro¨m et al., 1991; Onorati et al., 1992; Lukas and Weibel, 1995; Notebaertet al., 2009). In theory, the two illumination sources should be parallel and perpendicular to thegeneral trend of linear features within the landscape, however this is not always possible due toeither a lack of prior knowledge of the terrain (e.g. Lidmar-Bergstro¨m et al., 1991) or the presenceof several directional trends within the landscape (e.g. Smith and Clark, 2005).In order to diminish the necessity of defining a single linear trend within the terrain, otherstudies have proposed combining three (Hobbs, 1995; Lukas and Weibel, 1995; Hobbs, 1999;Gantenbein, 2012) or four (Mark, 1992; Loisios et al., 2007) hill-shaded images. In these situ-ations the width and direction of the illumination sector (the range and values of the azimuths),the altitude, and the vertical exaggeration is often predefined within these methods in order tominimize the subjectivity in creating the hill-shaded images.In methods that utilize raster addition, it is necessary to defined the relative weight that eachof the input images go into creating the final raster. Some methods predefine the relative weights79of input raster (Lukas and Weibel, 1995) while others weight them depending on distributionof aspect values within a landscape (Mark, 1992; Loisios et al., 2007). In a study building ona method originally proposed by Mark (1992) in which aspect is used to determine weighting,Loisios et al. (2007) highlight the differences between assigning a single global weight to eachimage added together versus determining the weights of the hill-shaded images on a cell-by-cellbasis. While cell-by-cell weighting resulted in a slightly more detailed image, the procedure doesrequire a more rigorous knowledge of GIS programming.The single largest problem that arises during the addition of hill-shaded images is the creationof a single image with an overall lower contrast than all of the input rasters (Lukas and Weibel,1995; Devereux et al., 2008; Sˇtular et al., 2012). This is due to the fact many terrain features arereplicated in several, if not all of the images, so when they are added to each other the light areason one image will effectively cancel out the darker areas of a one, resulting in an expressionlessimage. This is problem can be mediated by weighting each of the rasters independently rather thanweighing them all equally as well as making sure the illumination directions of inputted imagesare never directly opposite of each other (i.e. 315° and 135°).Hill-shaded images have been visualized in other ways. Colour and shading techniques, suchas swiss-style shading (Jenny and Hurni, 2006), aspect variant luminosity (Kennelly and Kimer-ling, 2004), multidirectional visibility index (Podobnikar, 2012), and RGB layering (Hobbs, 1995,1999), are often used to emphasize the range of aspects present in a landscape. Hill-shaded imagesmay also be combined with other raster representations of the landscape. Because hill-shadingdoes not indicate terrain elevation, adding a coloured DEM behind a hill-shaded image can help toclarify the elevation trends for the viewer (Kennelly and Kimerling, 2004). Other studies suggestlayering hill-shaded images with first or second order derivative maps such as slope or curvature toaid in reducing the illumination bias (Smith and Clark, 2005; Kennelly, 2008). Unfortunately, lay-ering these derivative maps, like adding in many hill-shaded images, tends to reduce the contrastof the resulting image.One method, namely Principal Component Analysis (PCA), has surfaced that has proven use-ful in reducing data redundancy and other problems associated with adding several hill-shadedimages together (Smith and Clark, 2005; Devereux et al., 2008; Pawley and Atkinson, 2010; Las-aponara and Masini, 2011; Bennett et al., 2012). PCA works by taking a set of inter-correlatedvariables and transforming them into a new set of uncorrelated variables, or components (Dev-ereux et al., 2008). A more thorough explanation of the principles and methods behind PCA canbe found in other sources such as Pohl and Van Genderen (1998) or Schowengerdt (2006). Forthe purposes of this paper it suffices to understand that, in this context, a PCA takes the variabilitypresent in all of the inputted hill-shades and creates a set of components, or images, where the80major portion of the variance of the inputs is represented by the first three components (Devereuxet al., 2008). Essentially, the first component represents the areas that are highlighted in all of theinputted hill-shades, the second component shows areas present in all but one of the hill-shades,and so on.Because most GIS software programs have a PCA function, this procedure is relatively simple.There are, however, certain considerations that must be taken into account when conducting aPCA on a dataset. First of all, the number of hill-shaded images that must be inputted has not beenwell defined. Although Devereux et al. (2008) report using 16 hill-shaded images with azimuthsequally spaced between 0° and 360°, it remains unclear as to whether increasing the number ofinputted images affects the results of the PCA. Secondly, like when creating hill-shades for usein a weighted raster addition, the range of illumination sector must be considered, for example,whether it is constrained to a single section (i.e. Mark, 1992) or if hill-shades from all directionsshould be considered.When visualizing the resulting PCA components most studies tend to present each of the com-ponents separately (e.g. Pawley and Atkinson, 2010; Lasaponara and Masini, 2011; Bennett et al.,2012; see Figure A.1b-e) The first three components may also be viewed as single, false colouredcomposite image, although this creates an image that is overwhelming to analyze (Devereux et al.,2008; see Figure A.1a). Despite the many advantages that PCA has over raster addition, littleeffort has been put into examining the range of visualization possibilities that exist for the data.A.3 Proposed Method of VisualizationThe following section describes the steps taken and variables considered to create the visualizationof the study area that fulfills the main goal of this study: to be able to easily discern and interpreta range of fluvial and glacial landforms from a LiDAR-derived DEM using a simple and quickmethod of visualization (Figure A.2). The method described below, which uses a combination ofhill-shading and PCA techniques, was deemed to present the clearest representation of the terrainfor almost all landforms present in the study area. Although, this study was conducted usingArcGIS™, most GIS software programs have similar tools that can be used to replicate theseresults, although certain settings may differ slightly.The first step is to create the hill-shaded images to be used in the PCA. Although Devereuxet al. (2008) use 16 hill-shaded images as input to their PCA, this study found no discernibledifference between using 16 and 4 images as input to the analysis. When experimenting withthe illumination sector of the input hill-shades it was found that using hill-shaded images withazimuths that represent all of the main compass directions (i.e. North, South, East, and West) have810 2 41 Km¯Component 1 Component 2 Component 3Component 1Component 2Component 3Component 4a)b)c)d)e)Figure A.1: Components of a PCA820 2 41 KmNStudy ReachFigure A.2: Visualization of LiDAR-derived DEM of the study area using the method pre-sented in this study.83Table A.1: Percent of variability present in the landscape represented by each of the compo-nents of the PCA.Component Variation1 48.5%2 36.8%3 14.3%4 0.3%better results than if the azimuths of the input images only represent a illumination from a limitedspectrum (i.e. the directions used by Mark, 1992). This is evidenced by the fact that when a PCAis conducted on images with azimuths restricted to the north and north-west the resulting PCAcomponents under-represent south-east facing slopes. This study therefore proposes using fourhill-shades created with the following azimuths: 360°, 270°, 180°, and 90°.Choosing the proper altitude and vertical exaggeration for the hill-shades requires finding theright medium between a washed out landscape (i.e. high altitude/low vertical exaggeration) andone in which there are too many dark shadows (i.e. low altitude/high vertical exaggeration). Aftertesting several different values for the two, it was found that an altitude of 45° and a verticalexaggeration of two allows low-lying features to be discerned without casting large areas of thelandscape into shadow.Once the four hill-shaded images are created, the next step is to conduct a PCA using themas input rasters. The PCA conducted on the hill-shaded images of the study area resulted in a setof components in which the first three account for over 99% of the variation of the landscape (seeTable A.1).The next and most important step in the process is the visualization of the PCA components.Several methods of visualization were experimented with (see Figure A.3) in order to determinewhich one gives the overall best representation of the study area. The largest problem that arisesin visualizing PCA components is finding the right degree of contrast such that there are no areasthat are over or under exposed but there is still enough contrast to discern as many landformsas possible. When the first three components are added together using equal weights (FigureA.3a), the resulting image is slightly over-exposed in flat areas which causes small scale featuresand textures in these areas to become indiscernible. The opposite is true (i.e. east facing slopesare quite under-exposed and lack fine detail) for the case when the PCA components, weightedaccording to the percent of variance that they represent, are added together (Figure A.3b) as wellas when only the first component is visualized (Figure A.3c).Layering of PCA components was experimented with in order to find the medium between the840 1.5 30.75 Kma¯)b)c)d)e)Equally Weighted Addition of PCA ComponentsWeighted Addition of PCA ComponentsComponent 1 of PCA Layering without value inversionLayering with value inversionComponent 2High : 415Low : 0Component 1High : 412Low : 0Component 2High : 415Low : 0Component 1High : 412Low : 0Figure A.3: Several methods of visualizing the PCA results examined for this study. Theimage shown in e) is the method that is found to best represent the study area.85contrast issues innate in PCA component addition and the under-exposure of the single component.Only the first two components were used as using all through three was found to lower the contrastto the extent that many of the fine details were lost. In experimenting with the transparency of theoverlying layer (Component 2) it was found that a transparency level between 60 and 70% lightensthe dark areas of component 1 without lowering the overall contrast of the image too much.Although Figure A.3d does a fairly good job of displaying many details and textures withinthe study area, it can be noted that there exists an illumination bias against north-west trendingobjects, as evidenced by the lack of contrast present for the valley in the mid-right side of FigureA.3d. The default for ArcGIS™is to symbolize the PCA components with high values as a lightcolour and low values as a dark colour (see legend in Figure A.3d), however it was found that ifthis is inverted for Component 1 (see legend in Figure A.3e), many of the linear features withinthe landscape become more pronounced.This simple method, found to create the most detailed and useful visualization of the studyarea (Figure A.2), is summarized in a flowchart shown in Figure A.4.86Digital Elevation Model Azimuth: 90° Altitude: 45 ° Z factor: 2 Azimuth: 180° Altitude: 45 ° Z factor: 2 Azimuth: 270° Altitude: 45 ° Z factor: 2 Azimuth: 360° Altitude: 45 ° Z factor: 2 Hill-Shade Tool Principal  Component  Analysis Tool PCA PCA Component 1 PCA Component 2 PCA Component 2 PCA Component 1 O V E R L Y I N G  high low Transparency: 70% high low Transparency: 0% Displayed as: SYMBOLOGY Final Visualization Figure A.4: Flow chart summarizing the LiDAR visualization method proposed in this study87A.3.1 Comparison of Visualization MethodsAlthough the visualization method presented in this paper allows for the geomorphic interpretationof LiDAR-derived DEMs, there are other simple visualization methods that are commonly usedfor similar applications. In this section the method proposed in this paper (to be known from nowon as the Hill-shade + PCA method) will be compared to three other commonly-used visualizationtechniques:1. Default Hill-shading: A single hill-shaded image created using the ArcGIS™hillshade tooldefault settings (i.e. azimuth: 315°; altitude: 45°, z factor: 1)2. Equal-weighted Orthogonal Addition: Two hill-shaded images with orthogonal azimuthsadded together using equal weights.3. Mark, 1992 adapted by Loisios et al., 2007: Four hill-shaded images with north-westerlyazimuths added together and globally weighted according to the relative proportion of thetotal landscape made up of that aspect.It is important in areas, such as British Columbia where the geomorphology is dictated byboth modern fluvial processes and quaternary glacial deposits, to visualize the terrain in such away that a range of landform types may be identified. Three locations that highlight landformsassociated with different geomorphic processes are shown in Figures A.5 to A.7. Locations 1and 2 are areas dominated by linear glacial features (drumlins) while Location 3 contains fluvialfeatures like terraces, gullies, abandoned stream channels, and bars.Comparing it to the rest of the visualization methods, it is clear that the image created using thedefault hill-shade setting does an overall poor job in highlighting many of the landforms presentacross the study area (Figures A.5a, A.6a, and A.7a). This is especially true for the drumlinfields shown in Figures A.5 and A.6 which happen to be oriented parallel to the default azimuth.Although not all landscapes will have linear features parallel to the angle of illumination, thesefigures show how a single hill-shaded image can easily under-represent the range of landformspresent.The number of landforms, especially linear ones, visible in the landscape is greatly increasedby simply adding an orthogonally illuminated image to the default hill-shaded image (FiguresA.5b, A.6b, and A.7b). This method it does a very good job in visualizing the large-scale featuresof landscape considering its simplicity. The main shortcoming of this method is that the displayedsurface lacks fine textural detail as it appears smoothed compared to the methods shown in c) andd). Given that texture often plays a large role in diagnosing terrain features, certain features maybe overlooked or mistakenly identified given this visualization method.880 0.5 10.25 Km¯Default Hill-shadingHill-shading + PCA a)d)Equal-weighted Orthogonal Additionb)Mark 1992 adapted by Loisios et al 2007c)Figure A.5: Visualization methods compared for location 1.The ability of the weighted raster addition method shown in Figures A.5c, A.6c, and A.7cto represent the landscape is highly inconsistent. This method does a very good job displayingcertain parts of the landscape, for example the area shown in Figure A.5c, while completely under-representing others, like the area displayed in Figure A.6c. In comparison, the hill-shade + PCAmethod defined in this paper consistently highlights both large landforms of all orientations aswell as the finer surface texture of the study area (Figures A.5d, A.6d, and A.7d).890 0.3 0.60.15 Km¯Default Hill-shadingHill-shading + PCA a)d)Equal-weighted Orthogonal Additionb)Mark 1992 adapted by Loisios et al 2007c)Figure A.6: Visualization methods compared for location 2.900 0.75 1.50.375 Km¯Default Hill-shadingHill-shading + PCA a)d)Equal-weighted Orthogonal Additionb)Mark 1992 adapted by Loisios et al 2007c)Figure A.7: Visualization methods compared for location 3.91Appendix BDigital Elevation ModelsDigital elevation models were created using the methods described in Section 1 of Chapter 3. Theexperiments conducted on Model 1 use the legend shown in Figure B.1a while those conductedon Model 2 use the legend shown in Figure B.1b.(a) DEM legend for Model 1 (b) DEM legend for Model 2Figure B.1: Legends used for all of the DEMs presented below. Values show elevation inmillimeters.920 2 4 6 800.51(a) Experiment 1: Hour 50 2 4 6 800.51(b) Experiment 1: Hour 100 2 4 6 800.51(c) Experiment 1: Hour 150 2 4 6 800.51(d) Experiment 1: Hour 200 2 4 6 800.51(e) Experiment 1: Hour 25930 2 4 6 800.51(f) Experiment 1: Hour 300 2 4 6 800.51(g) Experiment 1: Hour 350 2 4 6 800.51(h) Experiment 1: Hour 400 2 4 6 800.51(i) Experiment 1: Hour 450 2 4 6 800.51(j) Experiment 1: Hour 50940 2 4 6 800.51(k) Experiment 1: Hour 550 2 4 6 800.51(l) Experiment 1: Hour 600 2 4 6 800.51(m) Experiment 1: Hour 650 2 4 6 800.51(n) Experiment 1: Hour 700 2 4 6 800.51(o) Experiment 1: Hour 75950 2 4 6 800.51(p) Experiment 1: Hour 800 2 4 6 800.51(q) Experiment 1: Hour 850 2 4 6 800.51(r) Experiment 1: Hour 90Figure B.2: Digital Elevation Models for Experiment 1960 2 4 6 800.51(a) Experiment 2: Hour 50 2 4 6 800.51(b) Experiment 2: Hour 100 2 4 6 800.51(c) Experiment 2: Hour 150 2 4 6 800.51(d) Experiment 2: Hour 200 2 4 6 800.51(e) Experiment 2: Hour 25970 2 4 6 800.51(f) Experiment 2: Hour 300 2 4 6 800.51(g) Experiment 2: Hour 350 2 4 6 800.51(h) Experiment 2: Hour 400 2 4 6 800.51(i) Experiment 2: Hour 450 2 4 6 800.51(j) Experiment 2: Hour 50980 2 4 6 800.51(k) Experiment 2: Hour 550 2 4 6 800.51(l) Experiment 2: Hour 600 2 4 6 800.51(m) Experiment 2: Hour 650 2 4 6 800.51(n) Experiment 2: Hour 700 2 4 6 800.51(o) Experiment 2: Hour 75990 2 4 6 800.51(p) Experiment 2: Hour 800 2 4 6 800.51(q) Experiment 2: Hour 850 2 4 6 800.51(r) Experiment 2: Hour 90Figure B.3: Digital Elevation Models for Experiment 21000 2 400.5(a) Experiment 3: Hour 150 2 400.5(b) Experiment 3: Hour 200 2 400.5(c) Experiment 3: Hour 250 2 400.5(d) Experiment 3: Hour 300 2 400.5(e) Experiment 3: Hour 351010 2 400.5(f) Experiment 3: Hour 400 2 400.5(g) Experiment 3: Hour 450 2 400.5(h) Experiment 3: Hour 50Figure B.4: Digital Elevation Models for Experiment 31020 2 400.5(a) Experiment 4: Hour 100 2 400.5(b) Experiment 4: Hour 150 2 400.5(c) Experiment 4: Hour 200 2 400.5(d) Experiment 4: Hour 250 2 400.5(e) Experiment 4: Hour 301030 2 400.5(f) Experiment 4: Hour 350 2 400.5(g) Experiment 4: Hour 400 2 400.5(h) Experiment 4: Hour 450 2 400.5(i) Experiment 4: Hour 500 2 400.5(j) Experiment 4: Hour 55Figure B.5: Digital Elevation Models for Experiment 4104Appendix CDEMs of DifferenceDEMs of difference (DoDs) are created by subtracting two DEMs from one another. The DoDspresented in this appendix show the difference in elevation between subsequent runs in each ex-periment. All of the figures use the same legend (Figure C.1). The units of the x- and y- axes aremeters and the units in the z- direction (elevation) are in millimeters.Figure C.1: Legend used for all DoDs presented below. Values show the difference in ele-vation between the two DEMs in millimeters.1050 2 4 6 800.51(a) Experiment 1: Hours 10 - 50 2 4 6 800.51(b) Experiment 1: Hours 15 - 100 2 4 6 800.51(c) Experiment 1: Hours 20 - 150 2 4 6 800.51(d) Experiment 1: Hours 25 - 200 2 4 6 800.51(e) Experiment 1: Hours 30 - 251060 2 4 6 800.51(f) Experiment 1: Hours 35 - 300 2 4 6 800.51(g) Experiment 1: Hours 40 - 350 2 4 6 800.51(h) Experiment 1: Hours 45 - 400 2 4 6 800.51(i) Experiment 1: Hours 50 - 450 2 4 6 800.51(j) Experiment 1: Hours 55 - 501070 2 4 6 800.51(k) Experiment 1: Hours 60 - 550 2 4 6 800.51(l) Experiment 1: Hours 65 - 600 2 4 6 800.51(m) Experiment 1: Hours 70 - 650 2 4 6 800.51(n) Experiment 1: Hours 75 - 700 2 4 6 800.51(o) Experiment 1: Hours 80 - 751080 2 4 6 800.51(p) Experiment 1: Hours 85 - 800 2 4 6 800.51(q) Experiment 1: Hours 90 - 85Figure C.2: DEMs of Difference for Experiment 11090 2 4 6 800.51(a) Experiment 2: Hours 10 - 50 2 4 6 800.51(b) Experiment 2: Hours 15 - 100 2 4 6 800.51(c) Experiment 2: Hours 20 - 150 2 4 6 800.51(d) Experiment 2: Hours 25 - 200 2 4 6 800.51(e) Experiment 2: Hours 30 - 251100 2 4 6 800.51(f) Experiment 2: Hours 35 - 300 2 4 6 800.51(g) Experiment 2: Hours 40 - 350 2 4 6 800.51(h) Experiment 2: Hours 45 - 400 2 4 6 800.51(i) Experiment 2: Hours 50 - 450 2 4 6 800.51(j) Experiment 2: Hours 55 - 501110 2 4 6 800.51(k) Experiment 2: Hours 60 - 550 2 4 6 800.51(l) Experiment 2: Hours 65 - 600 2 4 6 800.51(m) Experiment 2: Hours 70 - 650 2 4 6 800.51(n) Experiment 2: Hours 75 - 700 2 4 6 800.51(o) Experiment 2: Hours 80 - 751120 2 4 6 800.51(p) Experiment 2: Hours 85 - 800 2 4 6 800.51(q) Experiment 2: Hours 90 - 85Figure C.3: DEMs of Difference for Experiment 21130 2 400.5(a) Experiment 3: Hours 20 - 150 2 400.5(b) Experiment 3: Hours 25 - 200 2 400.5(c) Experiment 3: Hours 30 - 250 2 400.5(d) Experiment 3: Hours 35 - 300 2 400.5(e) Experiment 3: Hours 40 - 351140 2 400.5(f) Experiment 3: Hours 35 - 300 2 400.5(g) Experiment 3: Hours 40 - 35Figure C.4: DEMs of Difference for Experiment 31150 2 400.5(a) Experiment 4: Hours 15 - 100 2 400.5(b) Experiment 4: Hours 20 - 150 2 400.5(c) Experiment 4: Hours 25 - 200 2 400.5(d) Experiment 4: Hours 30 - 250 2 400.5(e) Experiment 4: Hours 35 - 301160 2 400.5(f) Experiment 4: Hours 40 - 350 2 400.5(g) Experiment 4: Hours 45 - 400 2 400.5(h) Experiment 4: Hours 50 - 450 2 400.5(i) Experiment 4: Hours 55 - 50Figure C.5: DEMs of Difference for Experiment 4117


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