UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Influence of feed timing on sediment transport and bed evolution during hydrographs Mitchell, Alexander Jonathan 2018

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_2018_november_mitchell_alexander.pdf [ 31.42MB ]
JSON: 24-1.0372075.json
JSON-LD: 24-1.0372075-ld.json
RDF/XML (Pretty): 24-1.0372075-rdf.xml
RDF/JSON: 24-1.0372075-rdf.json
Turtle: 24-1.0372075-turtle.txt
N-Triples: 24-1.0372075-rdf-ntriples.txt
Original Record: 24-1.0372075-source.json
Full Text

Full Text

Influence of Feed Timing on Sediment Transport and BedEvolution During HydrographsbyAlexander Jonathan MitchellB.S. Earth Science, University of California, Santa Cruz, 2015B.A. Computer Science, University of California, Santa Cruz, 2015A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Geography)The University of British Columbia(Vancouver)September 2018c© Alexander Jonathan Mitchell, 2018The following individuals certify that they have read, and recommend to the Fac-ulty of Graduate and Postdoctoral Studies for acceptance, the thesis entitled:Influence of Feed Timing on Sediment Transport and Bed EvolutionDuring Hydrographssubmitted by Alexander Jonathan Mitchell in partial fulfillment of the require-ments for the degree of Master of Science in Geography.Examining Committee:Marwan Hassan, Professor, GeographySupervisorCarles Ferrer-Boix, Universitat Polite`cnica de CatalunyaSupervisory Committee MemberIan McKendry, Professor, GeographyAdditional ExaminerShawn Chartrand, Postdoctoral Research Fellow, GeographyAdditional ExamineriiAbstractSediment transport in gravel-bed streams is an important component for river ecol-ogy, infrastructure design, and hazard assessment. Yet making accurate predictionsof sediment transport remains elusive despite many decades of field, lab, and nu-merical modeling research. Most lab and modeling experiments use either constantdischarge, constant feed, or both are constant. While these works have helped dis-cover many aspects of sediment transport, they do not reflect the variability in dis-charge and sediment supply found in natural settings. We performed several flumeexperiments to challenge an assumption made by Parker, Hassan, and Wilcock(2007) that feed timing during a hydrograph does not matter for sediment transportbehavior and bed evolution. We used mixed grain sizes for both the initial bed andfeed (ranged between 0.5mm and 32mm; D50 = 7.83mm), a symmetrical steppedhydrograph, and five different feed scenarios: no feed, constant feed, rising-limbonly feed, falling-limb only feed, and capacity-scaled feed. All feeding scenarioshad the same total mass fed. We show that the assumption made by Parker et al.is wrong. Feeding on the rising limb strongly controlled sediment transport rates,overall bedload yield of the hydrograph, and bed scouring severity, but not bedloadgrain sizes. Feeding on the falling limb controlled bedload grain sizes, bedloadtransport rates in the early portion of the falling limb, and recovery of the bedtowards a pre-flood morphology (e.g. elevation and slope). Shifting the feed tim-ing towards different parts of the hydrograph highlighted different processes andinfluenced the overall hysteresis trends.iiiLay SummarySediment transport in mountain streams is an important component for river ecol-ogy, infrastructure design, and hazard assessment, yet accurate predictions remainelusive despite many decades of field, lab, and numerical modeling research. Mostlab and modeling experiments use either constant flow, constant sediment sup-ply, or both are constant. While these works have helped discover many aspectsof sediment transport, they do not reflect the variability found in natural settings.We performed several laboratory experiments to test how sediment supply timingduring a flood impacts sediment transport. We show that the supply timing signif-icantly influences sediment transport behavior during floods. High supply early inthe flood increases sediment transported downstream and reduces scouring of thestream bed. High supply towards the end of the flood controls transported grainsizes and bed elevation recovery.ivPrefaceThis thesis is original, unpublished, independent work by the author, Alex Mitchell.Experiments were executed with the help of several lab technicians (David Waine,Kyle Wlodarczyk, and Marcelino Secaira) and some programs used for analysiswere written by others (Shawn Chartrand, Tobias Mu¨eller, and Andre` Zimmer-mann) for their own projects.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Experiment Conditions . . . . . . . . . . . . . . . . . . . . . . . 102.2.1 Flume Settings . . . . . . . . . . . . . . . . . . . . . . . 102.2.2 Sediment . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.3 Hydrograph . . . . . . . . . . . . . . . . . . . . . . . . . 132.3 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 14vi3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.1 The 2B Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.2 Bedload Transport Rates and Yields . . . . . . . . . . . . . . . . 173.3 Bedload and Bed Surface Grain Sizes . . . . . . . . . . . . . . . 233.4 Depth and Slope . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.5 Summary by Feed Scenarios . . . . . . . . . . . . . . . . . . . . 384 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.1 Rising Limb Feed Effects . . . . . . . . . . . . . . . . . . . . . . 414.2 Falling Limb Feed Effects . . . . . . . . . . . . . . . . . . . . . 424.3 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48A Extra Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54viiList of TablesTable 3.1 Bedload Yield per Limb . . . . . . . . . . . . . . . . . . . . . 19viiiList of FiguresFigure 1.1 Hysteresis Diagram . . . . . . . . . . . . . . . . . . . . . . . 3Figure 2.1 Flume Picture . . . . . . . . . . . . . . . . . . . . . . . . . . 9Figure 2.2 Flume Design . . . . . . . . . . . . . . . . . . . . . . . . . . 10Figure 2.3 Feeder Picture . . . . . . . . . . . . . . . . . . . . . . . . . 11Figure 2.4 Feed Grain Size Distribution . . . . . . . . . . . . . . . . . . 12Figure 2.5 Hydrograph and Feed Design . . . . . . . . . . . . . . . . . 14Figure 2.6 Step Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . 15Figure 2.7 Sample Grain Identification Picture . . . . . . . . . . . . . . 16Figure 3.1 Total Bedload Transport . . . . . . . . . . . . . . . . . . . . 20Figure 3.2 Total Bedload Transport Hysteresis . . . . . . . . . . . . . . 21Figure 3.3 Cumulative Mass Balance . . . . . . . . . . . . . . . . . . . 22Figure 3.4 Bedload D50 . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 3.5 Bedload D84 . . . . . . . . . . . . . . . . . . . . . . . . . . 25Figure 3.6 Bedload D50 Hysteresis . . . . . . . . . . . . . . . . . . . . . 26Figure 3.7 Bedload D84 Hysteresis . . . . . . . . . . . . . . . . . . . . . 27Figure 3.8 Bed surface D50 . . . . . . . . . . . . . . . . . . . . . . . . . 29Figure 3.9 Bed surface D84 . . . . . . . . . . . . . . . . . . . . . . . . . 30Figure 3.10 Bed Surface D50 Hysteresis . . . . . . . . . . . . . . . . . . 31Figure 3.11 Bed Surface D84 Hysteresis . . . . . . . . . . . . . . . . . . 32Figure 3.12 Flow Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 3.13 Flow Depth Hysteresis . . . . . . . . . . . . . . . . . . . . . 35Figure 3.14 Water Surface Slope . . . . . . . . . . . . . . . . . . . . . . 36ixFigure 3.15 Water Surface Slope Hysteresis . . . . . . . . . . . . . . . . 37Figure 4.1 Bedload Yield per Limb . . . . . . . . . . . . . . . . . . . . 42Figure A.1 Trap Bedload Yield . . . . . . . . . . . . . . . . . . . . . . . 55Figure A.2 Ratio of Bedload D84 to D50 . . . . . . . . . . . . . . . . . . 56Figure A.3 2m Average Water Surface Slopes . . . . . . . . . . . . . . . 57Figure A.4 8m DEM for Experiment 1A . . . . . . . . . . . . . . . . . . 58Figure A.5 8m DEM for Experiment 1B . . . . . . . . . . . . . . . . . . 59Figure A.6 8m DEM for Experiment 2A . . . . . . . . . . . . . . . . . . 60Figure A.7 8m DEM for Experiment 2B . . . . . . . . . . . . . . . . . . 61Figure A.8 8m DEM for Experiment 3A . . . . . . . . . . . . . . . . . . 62Figure A.9 8m DEM for Experiment 3B . . . . . . . . . . . . . . . . . . 63Figure A.10 8m DEM for Experiment 4A . . . . . . . . . . . . . . . . . . 64Figure A.11 8m DEM for Experiment 5A . . . . . . . . . . . . . . . . . . 65Figure A.12 8m DEM Semivariogram for Experiment 1A . . . . . . . . . 66Figure A.13 8m DEM Semivariogram for Experiment 1B . . . . . . . . . 66Figure A.14 8m DEM Semivariogram for Experiment 2A . . . . . . . . . 67Figure A.15 8m DEM Semivariogram for Experiment 2B . . . . . . . . . 67Figure A.16 8m DEM Semivariogram for Experiment 3A . . . . . . . . . 68Figure A.17 8m DEM Semivariogram for Experiment 3B . . . . . . . . . 68Figure A.18 8m DEM Semivariogram for Experiment 4A . . . . . . . . . 69Figure A.19 8m DEM Semivariogram for Experiment 5A . . . . . . . . . 69xGlossaryHydrograph Water discharge through time during a flood event.Feed Scenario Sediment supply rates during a hydrograph designed to test differ-ent sediment input times. See Section 2.2.2 and Figure 2.5 for details.Experiment One cycle of the hydrograph (8 hours) using a feed scenario.Step A one-hour period of time where the discharge was held constant. (i.e. Onestep in the stepped hydrograph.)Period 20 minutes of continuous operation of the flume (3 periods per step)Mix Distribution Grain size distribution of the sediment used for both the feedand initial bed material. Mix Distribution and feed distribution are usedinterchangeably. See Section 2.2.2 for details.Hysteresis In the case of hydrographs, hysteresis occurs when a variable (e.g.bedload transport rate) responds differently for the same discharges found onthe rising and falling limbs. Clockwise hysteresis occurs when the variable ishigher on the rising limb than the falling limb. Counterclockwise hysteresisoccurs when the variable is higher on the falling limb than the rising limb.See figureBedload Sediment that travels on along the bed surface by saltating (short hops),rolling, or sliding. Sediment is usually medium-sized sand or larger, butdepends on how energetic the flow is.DEM A Digital Elevation Model is a 2D representation of the elevations of a sur-face.xiAcknowledgmentsI thank my supervisor, Marwan Hassan, for his endless support and advice at anytime of the day (or night) at all stages of the research process.I thank my partner in life, Alanna Klassen, for her love and support, keeping meemotionally strong during the tougher times.I thank Marwan’s research group, especially Shawn Chartrand, for providing ex-cellent feedback and putting up with my occasional email that everything is broken.I thank the lab manager, Rick Kelter, and lab technician, David Waine, for helpingme with all aspects of running a flume experiment.I thank the helpers in the flume lab, Kyle Wlodarczyk and Marcelino Secaira, fordoing the bulk of the work I really did not want to do. I would still be sieving todayif it weren’t for you two!I thank the Department of Geography for providing the opportunity and funding tobe a student here.I thank my parents for giving me a solid foundation to grow from and the extrafunding when Vancouver became too expensive.xiiChapter 1IntroductionSediment transport behavior during flood events has important implications foraquatic ecology, channel stability, local hydrology, and the integrity of built struc-tures. While water discharge is the primary driver of sediment transport in streams,several other factors can influence the resulting sediment transport behavior. Mostsediment transport models are based on the hydraulic capacity of the flow, which inturn is based on discharge, and often overestimate bedload transport rates (Gomezand Church, 1989). Bedload transport, it seems, is not a simple function of dis-charge. For example, the same discharge seen on the rising and falling limbs of ahydrograph or two sequential hydrographs, often have different observed bedloadtransport rates (Reid et al., 1985). More recent models recognize the importance ofsediment supply and bed surface characteristics and incorporate elements of bothin some form (e.g. Wilcock and Crowe (2003)). In general, characteristics of thesediment supply (e.g. grain size, volume, or timing), the hydrograph (e.g. shapeand magnitude), and the current state of the channel (e.g. morphology, armor, andgrain structures) can all influence sediment transport behavior.Sediment transport progresses through three stages as discharge increases dur-ing a flood (Ashworth and Ferguson, 1989): I Overpassing; II Partial Mobility, andIII Full Mobility . However, it is important to keep in mind that the bed surfacedoes not react uniformly in space. Intermediate discharges will have a mix of allthree stages across regions of the bed surface at any one time (Haschenburger andWilcock, 2003).1Bedload transport during Stage I is limited to isolated, small grains movingover a static bed. Discharge is relatively low and Shields stresses are below thecritical Shields stress (Church, 2010).Partial mobility occurs when the flow is strong enough to entrain the finer frac-tion of grains from the bed surface but not so strong as to mobilize all grains fromthe surface. Shields stresses for Stage II are normally between one to two times thecritical Shields stress (Church, 2010). As discharge increases through Stage II, theflow can entrain increasingly larger grains from the bed surface and the flow willmove a higher percentage of grains from each size class already in motion (Ash-worth and Ferguson, 1989). Furthermore, increasing discharge also increases thearea of the bed surface that is mobile (Haschenburger and Wilcock, 2003). Movinggrains typically interact with the static bed rather than other moving grains.As discharge increases and Shields stress rises above twice the critical value(Church, 2010), all grain sizes become fully mobilized over the whole bed. Inter-actions between moving grains begin to dominate over grain to static bed interac-tions. Fully mobile beds are often in a state of equal mobility, where the bedloadand the bed subsurface have the same grain size distributions (Parker and Klinge-man, 1982). Stage III transport is rare in gravel bed rivers because there often isnot enough available sediment to sustain transport for very long.In addition to discharge, sediment supply plays a critical role in bedload trans-port. Sediment supply in natural streams can vary significantly in magnitude andtiming relative to the stream hydrograph. Landslides, debris flow, and bank erosionare common ways for sediment to enter the river network from the landscape andrange in frequency and size by several orders of magnitude (Benda and Dunne,1997a,b; Guthrie and Evans, 2004; Simon and Klimetz, 2008; Stark and Hovius,2001). Alternatively, sediment within the stream network that has been temporarilystored in the stream bed can be remobilized if a flood is strong enough (Moog andWhiting, 1998).Sediment supply influences channel morphology (Montgomery and Buffing-ton, 1997), bed surface characteristics (Buffington and Montgomery, 1999; Hassanet al., 2006), and bedload transport (Elgueta-Astaburuaga and Hassan, 2017). Sev-eral studies have attributed the availability of easily mobilized sediment as the pri-mary driver of an observed hysteresis trend (See Figure 1.1 for explanation of hys-2Figure 1.1: Diagrams of basic hysteresis from Juez et al. (2018). Althoughthis digram shows suspended sediment, similar diagrams can be usedfor nearly any variable related to sediment transport (e.g. bedload trans-port rate). Column (i) shows no hysteresis where there is no differencebetween the rising and falling limbs and the peak transport is coinci-dent with the peak discharge. Column (ii) shows clockwise hysteresiswhere the peak transport occurs before peak discharge. Column (iii)shows counterclockwise hysteresis where peak transport occurs afterpeak discharge. These trends can be combined in different orders and atdifferent scales to create more complicated hysteresis trends (e.g. smallclockwise loop on a larger counterclockwise loop).teresis) (Gaeuman, 2010; Hassan et al., 2006; Hsu et al., 2011; Humphries et al.,2012; Mao et al., 2014; Moog and Whiting, 1998; Reid et al., 1985; Yuill andNichols, 2011). If there is a limited amount of easily mobilized sediment relativeto the transport capacity of the flood, then most of the available sediment will betransported early in the flood, leaving little to move later. However, if sedimentis difficult to entrain, such as when protected by a coarse layer of bed armor, thenpeak bedload transport occurs later in the flood after the stabilizing features havebeen destroyed. With unlimited supply of easily mobilized sediment, the hysteresistrend depends on the distance of the reach from the sediment source (Mao et al.,2014).Commonly seen in natural streams, channel bed armoring plays and important3role for sediment transport dynamics, stream ecology, hyporheic exchange, andchannel hydraulics (Wilcock and DeTemple, 2005). There are two types of armor:static armor and mobile armor. Both kinds are defined by a layer of coarse-grainedsediment blanketing a relatively fine-grained subsurface. Static armor tends to becoarser, have a smoother surface, be more imbricated, and have fewer structures(e.g. clusters) than mobile armor, which reflect the processes that created them(Mao et al., 2011). Armor increases the shear stress needed to mobilize the surfacegrains and reduces the amount of time surface grains are subjected to dischargeabove the critical threshold, often shifting peak bedload transport to the falling limbfor subsequent hydrographs (Hsu et al., 2011; Reid and Laronne, 1995). Armor canbe strong enough that it persists through floods (Wilcock and DeTemple, 2005),even if the flow is more than large enough to mobilize the largest grains (Claytonand Pitlick, 2008).Static armor forms when the preferential transport of finer grains during partialmobility, known as winnowing, leaves behind a coarse immobile layer on the bedsurface (Hassan et al., 2006). For the larger grains than remain, vibrating causedby turbulence can encourage grains to settle and interlock, further strengtheningthe bed (Monteith and Pender, 2005; Paphitis and Collins, 2005).Mobile armor is the result of equal mobility created by kinematic sieving athigh discharge. Kinematic sieving is triggered by the dilation of pore spaces in thelarger framework grains as they are jostled around by a strong flow (Frostick et al.,1984). Smaller grains are able to fall through the expanded pores into shelteredareas (Gibson et al., 2009; Parker and Stefan, 1982), even if the smaller grainswere part of the actively moving bedload (Monteith and Pender, 2005). Largergrains become over-represented and smaller grains become underrepresented onthe bed surface. By decreasing the mobility of smaller grains through shelteringand increasing the mobility of larger grains through exposure, the mobility of eachsize class will converge on a common value (Parker and Stefan, 1982). Once themobile armor forms, its grain size distribution becomes somewhat insensitive tohigher discharge (Mao et al., 2011; Powell et al., 2016).Whether armor develops is a function of both hydrograph shape (Hassan et al.,2006) and sediment supply regime (Dietrich et al., 1989), of which the supplyis more important (Hassan et al., 2006). Within a given supply regime, longer4hydrographs provide more time to develop armor whereas flashier hydrographsdo not. However, a high supply regime overwhelms the sorting processes andarmor cannot develop even with a long hydrograph (Hassan et al., 2006). Sedimentcaliber of the supply can impact armor development as well. Introducing sedimentthat is finer than the bed surface will increase the mobility of the larger grainson the bed, potentially damaging the armor (Venditti et al., 2010a,b; Wilcock andCrowe, 2003). Concentrating supplied sediment into pulses can disrupt normalbed behavior. Pulsed sediment supply initially causes a spike in bedload transportand surface fining, but can cause aggradation due to part of the sediment pulsebeing trapped under developing armor and bed structures (Elgueta-Astaburuagaand Hassan, 2017).When given enough time, a flow will reorganize the bed into a more stablestate by consolidation of the bed and creation bed surface structures. Even flowsbelow the entrainment threshold are able to rearrange the most unstable particleson the surface into more stable positions (Monteith and Pender, 2005; Paphitis andCollins, 2005). The bed consolidates as grains are reoriented and fine grains fillthe interstices (Haynes and Pender, 2007; Paphitis and Collins, 2005; Reid et al.,1985). With some transport, larger grains can jam together or entrap grains behindthemselves, leading to grain clusters (Brayshaw et al., 1983). Grain clusters andrelated surface structures can carry a sizable portion of the shear stress and willsignificantly reduce the mobility of the bed (Brayshaw et al., 1983; Hassan andChurch, 2000).ObjectiveMost experiments in sediment transport research use a constant discharge with ei-ther a constant feed or no feed. The minority of experiments which explore theeffects of variable discharge or variable sediment supply will often hold the othercomponent (sediment supply or discharge, respectively) constant. While these sim-plifications have greatly aided in furthering our understanding of sediment trans-port by isolating processes acting in a complex system, both discharge and sedi-ment supply are highly variable in natural streams. Therefore, sediment transportin natural streams may not behave as predicted by existing research.5In a study on bedload transport during repeated hydrographs, Parker, Hassan,and Wilcock (2007) assumed that sediment supply timing did not matter by usinga constant feed rate for their numerical models. In fact, they conclude that thefeed timing does not matter because the surplus or deficit of sediment supply isaccommodated by cyclic aggradation and degredation in a boundary reach at thehead of the stream reach. Would their experiments show the same results if thesame total volume of sediment was fed during different parts of a hydrograph?This thesis challenges the assumption made by Parker et al. with a series offlume experiments to allow better control of the sediment supply. Five differentfeed scenarios were used with a symmetrical hydrograph to highlight differencesin the response to feed timing: no feed, constant feed, rising limb only feed, fallinglimb only feed, and capacity-scaled feed. The total volume of sediment fed, grainssize distributions of the feed and initial bed, and the hydrographs were identical foreach feeding scenario. Our hypotheses are:1 The assumption Parker et al. made is wrong. Sediment supply timing doesmatter for sediment transport and bed evolution during a hydrograph.2 Hysteresis of the bedload transport is controlled by the timing and magnitudeof sediment fed during the hydrograph. Higher feed on the rising limb willlead to clockwise hysteresis whereas higher feed on the falling limb willlead to counterclockwise hysteresis for both bedload transport rates and grainsizes because the feed rate directly controls the amount of sediment availablefor transport.3 Bedload yield depends on the magnitude of sediment fed during the ris-ing limb and not the falling limb. Sediment fed during the rising limb willspend more time above-threshold flow and therefore will have a larger por-tion transported through the flume than sediment fed during the falling limb.4 Bed surface armoring and structures are influenced by sediment fed on thefalling limb and not the rising limb. Increasing discharge on the rising limbinhibits the development of surface features or destroys any that do form,regardless of feed rates. When there is no feed on the falling limb, winnow-ing will coarsen the bed surface and build bed surface structures. However,6feeding on the falling limb will interfere with armor and surface structuredevelopment.7Chapter 2MethodsFrom late July 2017 to early September 2017, I ran a non-scaled flume experimentat the Mountain Channel Hydraulic Experimental Laboratory at the University ofBritish Columbia.2.1 EquipmentThe flume used in this study is 18m long, 1m wide, and 1m deep, though the fulllength was not used (Figure 2.1). The slope can vary from 0% to 18%. Water ispumped to the top of the flume using a peristaltic pump with a discharge meterand a manually-set discharge controller. A conveyor belt with adjustable speedsat the top of the flume supplies sediment. A sediment trap at the end of the flumecollects the transported sediment. Sediment collected in the trap during a flumerun was mixed thoroughly and bifurcated repeatedly until the subsample weighedbetween 2.5kg and 5kg. The subsample was dried in an oven and then sieved downto 0.5mm.Of the variety of sensors available for the flume, this study uses cameras andlasers mounted on a motorized cart and a light table at the end of the flume. Themotorized cart can travel the length of the flume and has a Laserglow green laserpaired with an Allied Vision Technologies Prosilica GC1350 camera with a Kowaf=5mm lens to generate DEMs and a Canon Power Shot G12 camera for takingcolor photographs of the bed surface. At the base of the flume, a diffusive white-8Figure 2.1: Picture of the flume looking upstream.9Figure 2.2: Flume design.light table is paired with an Allied Vision Technologies Prosilica GX2300 camera.The high resolution light table camera operates at 27-32 frames per second and al-lows for real time measurements of grains leaving the flume. Labview code collectsthe raw data from the light table and cart sensors.2.2 Experiment Conditions2.2.1 Flume SettingsOnly the lower 9.25m of the flume, but the full width of 1m, were used for thisexperiment (Figure 2.2). The upper half of the flume was left empty. Water traveledover a rough wooden ramp before reaching the sediment bed. Sediment from thefeeder entered the channel at 10m through a randomizer (Figure 2.3). The first halfmeter had buried bricks and large, relatively immobile rocks on the surface. Thebricks and large rocks prevented a waterfall from forming at the end of the rampas the flow scoured away the regular sediment. The cart could only capture imagesfor 8m of the bed, from roughly 1.25m to 9.25m.The flume slope was set to 0.022m/m. The steeper slope helped prevent back-watering at a constriction entering light table and helped promote quicker bed re-action times.The flume bed was reset before every hydrograph. Sediment was added to theflume if necessary, the bed surface was mixed with shovels, and the new surfacewas leveled with a wooden board in order to to bring the bed surface back to thestarting elevation. The large rocks at the head of the flume were randomly placedand pressed into the surface.10Figure 2.3: Picture of the feeder and randomizer. Note the large blue rocksto break up the flow.112.2.2 SedimentThe experiments used poorly-sorted sand and gravel for both the feed sedimentand bed sediment (Figure 2.4). Sediment was painted different colors based on 12φintervals.Figure 2.4: Grain size distribution of the feed and bulk sediment. Calcu-lated from the sum of five samples gathered throughout the experiments.D50 = 7.83mm and D84 = 16.77mm.12There were five feed scenarios:1. No feed2. Constant3. Only rising limb4. Only falling limb5. Capacity-scaled feedAll feeding scenarios supplied a total of 800kg of sediment during the sevenhour hydrograph. Some feed scenarios were repeated twice (no feed, constant feed,rising feed), but there was not enough time to repeat all scenarios.The experiment proposal called for calculating transport capacity of each dis-charge step and using the sum for the total feed mass. However, the several meth-ods used gave significantly different answers, were much larger than observedtransport rates during practice runs, and would require more sediment than wasavailable in the lab. Therefore, for the scaled feed scenario, estimated transportcapacities were scaled down such that the total feed was 800kg.One source of error for these calculations is that it is difficult to determine thecritical shear stress. The initial beds in these experiments were well mixed andloose. Grains start moving immediately.2.2.3 HydrographThe bed was subjected to seven-hour symmetric stepped hydrographs plus one hourof preconditioning (Figure 2.5). Including the preconditioning step, the dischargesequence is (L/s): 50, 62, 75, 87, 100, 87, 75, 62. Sediment was never fed andmid-step measurements were limited during the preconditioning step. Rising andfalling limb feeds included the full hour of peak discharge.Every 20 minutes during a step, starting at the 10 minute mark, the bed andwater surface profiles are manually measured at 0.5m spacing (Figure 2.6). At the20 and 40 minute marks, the water is temporarily shut off for a 2m photo between4.5m and 6.5m. At the end of the hour, flow is shut off for an 8m photo and laser13scan.Figure 2.5: Symmetric stepped hydrograph (dark blue) used for this experi-ment with variable feed rates. The first hour of each hydrograph wasa preconditioning flow with no sediment feed. Red = No feed; Gold= Constant feed; Green = Rising limb feed; Cyan = Falling limb feed;Magenta = Capacity-scaled feed2.3 Data ProcessingA combination of Labview 2012, Matlab R2017b, and Python 3.6 was used tocollect and process data.Labview code generates bedload transport rates from the light table at a onesecond resolution and DEMs from the green laser at a 2mm/px resolution. How-ever, due to a calibration mistake before the experiment, the raw DEMs had atransverse resolution around 2.5mm/px and had to be interpolated in Matlab togenerate 2mm/px DEMs. Additionally, raw DEMs had many small holes due to14Figure 2.6: Measurement schedule during each hour-long flow step.larger grains blocking the line of sight of the camera. Holes were filled with Matlabwith the mean value of the transverse row of pixels.Semi-automated photo analysis in Matlab identified the grain size distributionsin the bed surface photos based on the grain colors. To generate a distribution, a10 X 10 grid of measurement points were overlain on 320mm X 320mm regions ofa bed surface photo. The code attempted to automatically identify the grain sizeunder each measurement point, but would delegate the task to the user if there wasuncertainty (Figure 2.7). It was difficult to line the DEM and bed photos up whenselecting subsample regions due to the DEM calibration error and to camera per-spective issues from the bed surface elevation varying up to 10cm over the lengthof the flume. Therefore, subsample regions should be close to 320mm X 320mm,but may be slightly off.Python is used to synthesize the output data from the Labview and Matlabprograms. The code can be found on Github 1 2.1https://github.com/alexmitchell/data wrangling Will likely require significant modification foruse in other projects. See readme.txt and contact author for questions.2https://github.com/alexmitchell/helpyr15Figure 2.7: Sample grain identification picture. The grain in the blue circleis identified manually if the code cannot do it automatically, such as inthis image because the measurment point is on the edge of a grain.The Labview and Matlab programs are primarily preexisting code written byShawn Chartrand, Andre` Zimmermann, and Tobias Mu¨eller for their own researchexperiments. I updated these legacy programs to fix a few bugs and to increaseautomation. All of the Python code was written by me for this project.16Chapter 3Results3.1 The 2B ProblemExperiment 2B (Constant feed) is not included in the analysis. This experimentbehaves much differently than its duplicate experiment (2A) or any of the other ex-periments for other feed scenarios. This is almost certainly due to the flow duringthe rising 87L/s and 100L/s steps excavating and moving one of the bricks. Thebricks were intended to prevent the flow from scouring into the flume bed too farand forming a waterfall off the end of a wooden ramp. However, displacement ofthe brick concentrated the flow to the right side of the channel. I chose to continuethe experiment because I did not realize how significant the brick movement wouldbe. Over the next several steps, the bed continued to be scoured on the right sidewhile a large bar formed on the left side (Figure A.7). Clearly, this event supportsthe idea of keystone grains having a significant impact on bedload transport behav-ior and channel morphology. The data from 2B are included in the figures with anX over it for the sake of completeness, but the data will not be used to draw anyconclusions.3.2 Bedload Transport Rates and YieldsBedload transport rates varied considerably throughout the hydrograph (Figure 3.1).Individual steps did not show consistent trends, but transport rates tended to be17highest from the rising 75L/s step through the 100L/s step (hours 2-5).Transport rates measured by the light table that were higher than 800g/s wereremoved from the analysis. Unusually high scatter in the light table data duringsome periods (e.g. 5A between 1h:00m and 1h:20m) could have been the result ofabnormally low camera frame rates and, therefore, individual transport rate valuesmay be unreliable. Despite the possible source of error, the 20-minute sums ofthe unscaled light table transport rates (i.e. kg per 20-minute period) matched thesediment trap data reasonably well (See Figure A.1 for trap data). To improvethe reliability of the long term trends, the trap mass data was used to scale thecorresponding periods of data shown in Figure 3.1.All experiments, except 4A, had a clear clockwise hysteresis of bedload trans-port (Figures 3.2 and 3.3). The experiments with clockwise hysteresis transported2.3 to 7.7 times as much sediment on the rising limb than the falling limb (Ta-ble 3.1). The values in Table 3.1 were based on splitting the 100L/s step in halfand ignoring the 50L/s step. However, it is not clear what would be the best way toassign the peak flow to the limbs given that the hydrograph is stepped. Experiment4A showed a slight counterclockwise hysteresis, with the falling limb transporting1.15 more sediment than the rising limb. However, the rising limb of 4A had un-usually low transport compared to other similar experiments (See 3.5). While therising limb for experiment 5A was always had higher mean transport rates than thefalling limb, the difference in transport rates between the rising and falling limbsduring the 62L/s and 75L/s steps was much larger than during the 87L/s and100L/s steps. The sediment transport rate decreases significantly after the 100L/sstep for all experiments and almost reaches zero g/s by the end of the falling 87L/sstep for most experiments. Experiments 4A and 5A have some transport during thefalling 75L/s step, but also reach near zero transport during the 62L/s step.All experiments showed a net mass export of at least 300kg at some point dur-ing the hydrograph (Figure 3.3). Experiments with feed on the falling limb (2A,4A, and 5A) had the maximum net sediment lost near the end of the peak discharge(between −300kg and −400kg), then the bed aggraded to almost the original bedmass. Only experiment 4A surpassed the original bed mass. Experiments withoutfeed on the falling limb (1A, 1B, 3A, 3B) had the maximum net sediment lost atthe end of the hydrograph. The sediment beds during the two no-feed experiments18Experiment Rising limb Falling limb Hydrograph Rising/fallingtotal (kg) total (kg) total (kg) ratio1A 521 93 614 5.571B 479 82 561 5.862A 707 92 800 7.652B 960 597 1557 1.613A 873 309 1182 2.823B 830 226 1056 3.674A 277 319 596 0.875A 631 274 904 2.30Table 3.1: Comparison of total bedload transport on the rising and fallinglimbs. The 100L/s step is split in half and the 50L/s step is excluded.(See Figure 3.2)degrade until peak discharge then change very little for the remainder of the exper-iment, ending at−633kg and−609kg for experiment 1A and 1B, respectively. Thetwo rising-limb feed scenarios aggrade above the original bed surface during therising 62L/s step, but then rapidly degrade until the end of the falling 87L/s step.Experiments 3A and 3B end at −383kg and −334kg, respectively.Interestingly, the 50L/s preconditioning step (first hour) from each experimentbehaves quite differently despite that they are supposed to have identical condi-tions, have a shear stress lower than critical, and have no interruptions for scans.19Figure 3.1: Total bedload transport rates from the light table data scaled by the sediment trap measurements. Greydots are individual bedload measurements. The blue line is the median and the cyan lines are the 25th and 75thpercentiles based on a 10 minute moving window.20Figure 3.2: Hysteresis plot of the median total bedload transport rates from 3.1. Blue curve is the rising limb andorange curve is the falling limb. Instead of using discharge, as there were only five discharge rates during thestepped hydrographs, time was combined with discharge to imitate a hysteresis plot of a triangular hydrograph(i.e. time was folded around the 4.5 hour mark). The preconditioning step (50L/s) is ignored because it had nomatching step on the falling limb.21Figure 3.3: Cumulative sum of the bedload transport rates and mass blance of the flume. Bedload rates are based onlight table data and was scaled to match the trap data.223.3 Bedload and Bed Surface Grain SizesThe mean bedload D50 typically stayed between 2 and 11 mm and the mean bed-load D84 typically stayed between 4 and 22 mm (Figures 3.4 and 3.5). Bedload Divalues stayed near the feed Di values during the rising limb for all experiments andeither maintained a similar Di value (4A and 5A) or declined well below the feedDi (1A, 1B, 2A, 3A, and 3B) on the falling limb (Figures 3.6 and 3.7).Both the light table and sediment trap samples had similar grain size trendsduring the hydrograph. However, there was a bias in the light table data such thatthe Di values were systematically below the corresponding values from the sievedsediment trap samples. The light table data is scaled proportionally by the averageof the ratios between trap Di and raw light table Di values for all experiments. Thescaling factors for D50 and D84 light table data are 1.48 and 2.03, respectively.Despite the trends between the two data sources matching well after scaling, thereare some notable discrepancies:• The trap D50 and D84 for experiment 2A stays near the corresponding feedDi value until the falling 62L/s step before declining, whereas the than de-clining during the falling 87L/s step.• The trap D50 and D84 stay near the associated feed Di rather than declineduring the falling 62L/s step for experiments 4A and 5A.The trends in the trap data will take precedence for these discrepancies as it is lesslikely to be in error.23Figure 3.4: Scaled bedload D50 as measured by the light table and sediment trap. Grey dots are individual light tablemeasurements. The blue line is the median and the cyan lines are the 25th and 75th percentiles based on a 10minute moving window. The red stars are the D50 values from sieved sediment trap samples.24Figure 3.5: Scaled bedload D84 as measured by the light table and sediment trap. Grey dots are individual light tablemeasurements. The blue line is the median and the cyan lines are the 25th and 75th percentiles based on a 10minute moving window. The red stars are the D84 values from sieved sediment trap samples.25Figure 3.6: Hysteresis plots of median bedload D50. Blue is the rising limb, orange is the falling limb, lines are thelight table D50 and stars are the trap D50. Curves are based on a 10 minute moving window of the scaled lighttable data. The 50L/s step is ignored as it was preconditioning and had no matching step on the falling limb.26Figure 3.7: Hysteresis plots of median bedload D84. Blue is the rising limb, orange is the falling limb, lines are thelight table D84 and stars are the trap D84. Curves are based on a 10 minute moving window of the scaled lighttable data. The 50L/s step is ignored as it was preconditioning and had no matching step on the falling limb.27The bed surface grain size distribution was similar to or slightly coarser thanthe feed for nearly the entire hydrograph for all experiments (Figures 3.8, 3.9,3.10, and 3.11). Trends for the bed surface grain sizes were not as distinct as forthe bedload grain sizes. The bed surface D50 and D84 for experiments 1A, 1B, 2A,and 3A were close to the mix distribution on the rising limb and increased slightlyon the falling limb. Experiment 3B had no trend for D50 but the D84 increaseson the falling limb. Experiment 4A did not have a clear trend in either percentile.Experiment 5A had a slightly increasing D50 on the falling limb and no trend forthe D84.28Figure 3.8: D50 of the bed surface between stations 4.5m and 6.5m. Grey dots are individual D50 measurements from a32x32cm subsection of a bed surface photo. The blue line shows the mean of the D50 values for each period andthe dashed black lines are the armor ratio.29Figure 3.9: D84 of the bed surface between stations 4.5m and 6.5m. Grey dots are individual D84 measurements from a32x32cm subsection of a bed surface photo. The blue line shows the mean of the D84 values for each period andthe dashed black lines are the mix D84.30Figure 3.10: Hysteresis of the mean D50 of the bed surface between stations 4.5m and 6.5m. Blue is the rising limband orange is the falling limb. The 50L/s step is ignored as it was preconditioning and had no matching step onthe falling limb.31Figure 3.11: Hysteresis of the mean D84 of the bed surface between stations 4.5m and 6.5m. Blue is the rising limband orange is the falling limb. The 50L/s step is ignored as it was preconditioning and had no matching step onthe falling limb.323.4 Depth and SlopeAverage flow depths ranged from 6cm to 11cm and roughly followed the hydro-graph discharge (Figures 3.12 and 3.13). All experiments increased in depth duringthe rising limb in a similar way. However, the falling limbs showed two differentbehaviors. Experiments 1A (likely), 1B, 3A, and 3B had flows that are noticeablydeeper on the falling limb than during the same discharge on the rising limb. Ex-periments 2A, 4A, and 5A had falling-limb flow depths that are similar to the risinglimb.Flume-averaged slopes ranged from 0.011 to 0.020 (Figures 3.14 and 3.15).Slopes declined on the rising limb for all experiments, though the feeding scenariostended to be slightly steeper at the peak than the no-feed scenarios. The slopes forexperiments 1A and 1B did not change significantly during the falling limb. Exper-iments 3A and 3B, despite having no feed on the falling limb, had slight increasesin slope. Experiments with feed on the falling limb (2A, 4A, and 5A) increasedin slope during the falling limb with little to no hysteresis. The average slope be-tween 4.5m and 6.5m (Figure A.3) is not used as it shows no consistent trend forany experiment. Instead, the slope values chaotically vary between roughly 0.01and 0.025 for the whole hydrograph. This is likely due to passing bedforms andlow spatial resolution (0.5m) of the measurements.33Figure 3.12: Average water depths between stations 4.5m and 6.5m. Grey dots are individual depth measurements.34Figure 3.13: Hysteresis of the average water depths between stations 4.5m and 6.5m. Blue is the rising limb and orangeis the falling limb.35Figure 3.14: Flume-averaged water surface slopes. Slope is calculated using OLS with no fixed intercept.36Figure 3.15: Hysteresis of flume-averaged water surface slopes. Slopes are calculated using OLS with no fixed inter-cept. Blue is the rising limb and orange is the falling limb.373.5 Summary by Feed ScenariosNo FeedThe no-feed experiments were among the lowest for sediment yields. They hadstrong clockwise hysteresis for bedload transport rates having transported at least5.5 times more sediment (around 400kg) on the rising limb than falling limb. Over-all bedload transport peaked during the rising 75L/s and 87L/s steps, then steadilydeclined to nearly zero during the falling 87L/s step. The bedload grain size distri-bution was similar to or slightly coarser than the original sediment mix distribution(same as feed distribution) during the rising limb. However, the bedload steadilybecame finer on the falling limb, with both the D50 and D84 ending at about halfof their original values on the rising limb. The bed surface grain size distributionshowed a counterclockwise hysteresis by staying similar to the original mix distri-bution on the rising limb then coarsening near peak discharge and on the fallinglimb. Slope had a strong clockwise hysteresis from intense scouring on the risinglimb reducing the slope and no slope recovery on the falling limb. The flow depthshad a weak counterclockwise hysteresis.Constant FeedThe constant feed scenario yielded a moderate amount of sediment and had thestrongest clockwise hysteresis for bedload transport (7.6x more or +600kg on therising limb). The falling limb had a similar total bedload mass to the no-feedscenario, but the rising limb moved 200kg (40%) more than during the no-feedscenario. The bed began aggrading after the 100L/s step and the bedload transportrate leaving the flume was nearly zero after the falling 87L/s step. The bedloadgrain size distribution was similar to or slightly coarser than the mix distributionduring the rising limb and through the falling 87LL/s step. However, the bedloadfined rapidly during the falling 75L/s and 62L/s steps, ending with similar Divalues as the no-feed scenario. The bed surface grain size distribution was similarto or slightly coarser than the feed distribution on the rising limb, but coarsenedsignificantly on the falling limb. Both depth and slope changed significantly overthe course of the hydrograph, but did not show hysteresis.38Rising-Limb FeedThe rising-limb feed scenario transported the most sediment for both the overallhydrograph and for during just the rising limb. The falling limb was among thehigher producing scenarios too despite no feed after the 100L/s step and a nearzero transport rate after the falling 87L/s step. Similar to the constant-feed sce-nario, the bedload grain size distribution was close to or slightly coarser than themix distribution on the rising limb and fined significantly on the falling limb. Thebed surface grain size distribution may have had a weak counterclockwise hys-teresis, but the distributions on the rising limb for both experiments didn’t agreewell. The falling limb on both experiments was more consistently coarser thanthe mix distribution. Depths during the rising-limb feed scenario had the strongestcounterclockwise hysteresis of all experiments, but only had a moderate clockwisehysteresis for slope values.Falling-Limb FeedThe falling-limb feed scenario transported among the least amount of sediment.Surprisingly, the rising limb transported about half the amount of sediment as therising limb during the no-feed scenario despite having the same experimental con-ditions. There are two possible reasons for the discrepancy, both with unknowncauses and unknown magnitudes of the effects:1. The 50L/s step transported almost twice as much sediment (140kg) as anyother 50L/s step. Excessive sediment transport during the preconditioningstep may have taken away from the amount of sediment than could moveduring the rest of the rising limb. Because sediment transported during the50L/s step was not included in the calculations, this could have created un-characteristically low yields.2. A bar was present from the rising 75L/s step through the 100L/s step (Fig-ure A.10). The flow appeared to preferentially scour the left side of the bed,leaving the right side higher. Sediment in the bar would have been trans-ported out of the flume based on the scour patterns of experiments 1A and1B. The bar disappeared as the bed aggraded during the falling 87L/s step39and should not have affected the data later in the falling limb.Based on these two issues, bedload yield between rising 62L/s and peak dis-charge should have been much higher and the yield during the falling 87L/s shouldlikely be slightly lower (had to aggrade over a larger bed surface). Therefore bed-load transport should have a weak clockwise hysteresis rather than the weak coun-terclockwise hysteresis observed.The grain size distributions of both the bedload and bed surface did not show ahysteresis trend. As seen in other scenarios, both distributions stayed near the mixdistribution on the rising limb. However, unlike other scenarios, both distributionsstayed near the mix distribution during the entire falling limb as well, albeit withhigh variability in values. Like the constant feed scenario, flow depths and slopeshad no hysteresis trends.Capacity-Scaled FeedThe capacity-scaled feed scenario yielded a moderately high amount of sediment.The rising limb moved 2.3 times more sediment than the falling limb, primarilyfrom differences at 62L/s and 75L/s. Bedload grain size distributions remainednear the mix distribution for the entire hydrograph. The bed surface grain sizes hada weak counterclockwise hysteresis best seen in the 62L/s and 75L/s steps for theD50, but did not have hysteresis for D84. Depth had no hysteresis and slope hadweak clockwise hysteresis.40Chapter 4DiscussionIn these experiments, changing the timing of the feed produced clear differences inbedload transport, confirming Hypothesis 1. Furthermore, each limb had a differentresponse to feeding.4.1 Rising Limb Feed EffectsSediment fed on the rising limb primarily impacted sediment transport rates andoverall yields. Higher sediment supply increased bedload transport rates (Figures3.2 and 4.1) and reduced net sediment loss by the bed (Figure 3.3). Scouring ofthe bed was reduced as the flow accommodated extra material from the feed. Ifthe feed rate was high enough late in the rising limb or during peak flow, it causedelevated transport rates to lag into the early portions of the falling limb. The rising-limb and capacity-scaled feed scenarios both had high bedload transport during thefalling 87L/s step whereas the constant feed scenario did not (See 3A, 3B, and 5Aversus 2A in Figure 3.2). Grain size distributions for both the bedload and bedsurface were similar to the mix distribution regardless of the feed rate (Figures 3.4and 3.5). Even with low feed rates or no feed, the increasing discharge could scourdeeper into the bed and access new sediment at the mix distribution.41Figure 4.1: Bedload yield for each limb compared with the mass of sedimentfed during that limb. Red is for the rising limb. Blue is for the fallinglimb. Lines are the OLS trend lines.4.2 Falling Limb Feed EffectsSediment feed rates on the falling limb had a greater influence on bedload grainsizes and bed characteristics than on bedload transport rates. Without feed or withlow feed rates early in the falling limb, the bedload grain sizes declined rapidlyafter the peak flow and the bed surface coarsened slightly (Figures 3.4, 3.5, 3.8,and 3.9). Higher feed rates led to longer delays in the bedload grain size distribu-tion deviating from the mix distribution, even if the overall bed load transport ratewas declining. Additionally, higher feed rates early in the falling limb increasedbedload transport rates during the falling 87L/s step and possibly the 75L/s step,but not to the same degree as feeding during the rising limb (Figure 3.2). Higherfeed rates towards the end of the falling limb had little effect on bedload transport.Bedload transport rates in all experiments declined to nearly zero transport by the42falling 62L/s step. Feeding on the falling limb at any rate caused the bed to beginaggrading during the falling 87L/s step and allowed the flow depth and slope torecover values similar to the rising limb (Figures 3.3, 3.12, 3.13, 3.14, and 3.15).It is not clear why the bed surface during the falling limb of the constant feedexperiment became so coarse compared to other experiments. Overall bedloadtransport, slope, and depth behaved similar to other experiments. The bedloadgrain size distribution lingered near the mix distribution longer than experimentswithout falling limb feed but shorter than experiments with higher falling limb feed.A sediment wave can be seen in the DEM for the falling 62L/s step ( betweenstations 4m and 6m on the bottom DEM Figure A.6). However, the wave is notevident in earlier steps and cannot explain why the grain sizes are already muchlarger than other experiments.4.3 ImplicationsHysteresis trends of the bedload transport were influenced by both the magnitudeand timing of the sediment feed but not exactly as expected in Hypothesis 2. Asevidenced by the no-feed scenarios, the experiment design itself created a strongclockwise hysteresis for bedload transport rates and grain sizes. Some feeding onthe rising limb (2A) provided additional sediment for transport when there is aplenty of stream power to mobilize it, leading to a stronger clockwise hysteresis.Too much feed on the rising limb, particularly near the peak (3A, 3B, and 5A),resulted in high transport rates early in the falling limb and reduced the differencesbetween the limbs. Heavy feeding on the falling limb produced counterclockwisehysteresis for the bedload transport rates and no hysteresis for bedload grain sizes.However, assuming a correction is required (see Section 3.5, a weak clockwisehysteresis for the bedload transport rates and no hysteresis for bedload grain sizeswould be expected if this experiment is repeated in the future. To summarize,modest feed rates on the rising limb will strengthen the clockwise hysteresis forbedload transport whereas heavy feed rates on the falling limb will weaken it. Toomuch feed on the rising limb near the peak discharge or too little feed on the fallinglimb will reduce the effect on the hysteresis.Figure Figure 4.1 highlights the differences in bedload yield from rising and43falling limb feed. In particular, higher feed rates on the rising limb correspondedwith higher bedload transport rates whereas feed rates on the falling limb did nothave much effect on bedload yield, confirming Hypothesis 3. However, there aresome issues with comparing sediment yield from each limb of the hydrograph forthe experiment design used, but they don’t impact the conclusions of the plot. Therising-limb (3A and 3B) and falling-limb (4A) feed scenarios both have 100kgof feed on the non-feeding limb due to an issue with the experiment design andanalysis requirements. Sediment was fed for the entire 100L/s step to stay consis-tent with the design of other steps (i.e. feed at the same rate for the whole hour).However, the step was split in half during analysis to keep the limb durations anddischarges balanced. Additionally, lag effects are important for sediment transportduring a hydrograph and this plot makes no attempt to address it. For example, ifrising limb feed actually stopped at the start of the falling limb, bedload transportlag effects would lead to a higher bedload yield than if there were no feeding onthe rising limb despite both scenarios having no feed on the falling limb.Hypothesis 4 is rejected because the armor ratios rarely were below 1 or above1.5 in all experiments, suggesting that the bed surfaces had no armor or wereweakly armored. Furthermore, a weak coarsening trend observed in some of theexperiments did not follow a consistent pattern based on feed timing. The DEM’swere not analyzed for surface structures (DEMs shown in Appendix). However,given that armor did not develop significantly, it is unlikely that sediment structureswere able to form either. Both mobile and static armor require time to develop. Inthese experiments, the hydrographs appear to have been too quick relative to therange of discharges for the bed surface to evolve appreciably. The results indicatethat the experiment design resembles a flash flood setting. However, the litera-ture indicates that a different shape or a lower magnitude hydrograph would helpdevelop these features.Recent work by An et al. (2017), which built upon the model used by Parkeret al. (2007), found that the bed surface became invariant beyond a hydrographboundary layer when the sediment was uniform in size or sediment feed ratesmatched the transport capacity during the hydrograph. Otherwise, sediment wastransported via persistent low-amplitude bedload sheets outside the boundary reach.An et al. attributed the disagreement between the two studies to a higher resolution44model mesh, a focus on bed elevation changes, and the inclusion of bed stratigra-phy into the newer model. While their study showed that constant feed produceddifferent results than capacity feed, they did not test situations where the sedimentwas supplied more on the rising or falling limb (i.e. peak sediment supply did notalign with peak discharge). Their results would suggest that these scenarios wouldproduce bedload sheets due to an imbalance of supply and capacity throughoutthe hydrograph. Given the differences observed in our study of sediment transportbehavior, the feed timing could also impact the development of bedload sheets.However further analysis of the data is needed.In addition to improving flume and numerical model designs, this research canaid gravel augmentation projects downstream of dams. Among the many reper-cussions of a dam on the river system, sediment transport is altered significantly(Ligon et al., 1995). Dams typically reduce the flow regime and reservoirs inter-cept sediment supplied from upstream. The imbalance between stream flow andsediment supply downstream of a dam often leads to bed degradation, bed sur-face coarsening, change in channel morphology, and river entrenchment (Schmidtand Wilcock, 2008; Smith and Mohrig, 2017; Williams and Wolman, 1984), whichcan disrupt fish habitat (Ligon et al., 1995; Pasternack et al., 2008). Efforts havebeen made to mitigate the effects of dams by adding gravel to the river down-stream of the dams, but are often unsuccessful (Wheaton et al., 2004) Much of theresearch on sediment augmentation has focused on the grain size, quantity, andplacement of the supplied sediment (See Gaeuman (2012) for a good overview).Few studies involve sediment supply timing, likely because the easiest methodsfor sediment augmentation place sediment in or near the channel during low flowbetween floods. A few studies (e.g. Gaeuman (2013) and Gaeuman et al. (2017))used gravel injection to add sediment to a river during high-flow flood events. Inparticular, Gaeuman et al. (2017) added sediment sediment in two equal pulses:one during the rising limb and one on the first day of a 2.5 day peak flow. Ourresults suggest that their decision to feed on the rising limb amplified transport ofthe injected sediment while mitigating scour of the bed, similar to our rising-limbfeed scenario. Should a project design call for injected sediment to stay closer tothe injection site (perhaps to reduce the amount of sediment required to dynami-cally build a feature), then injections times should be shifted towards the falling45limb. However, Gaeuman et al. observed significant localized scouring and aggra-dation associated with gravel dunes and large-scale bar development, which ourstudy does not address.46Chapter 5ConclusionIn these experiments, we tested how feed timing during a hydrograph can impactbedload transport to challenge an assumption made by Parker et al. (2007). Weperformed a flume study using a symmetrical stepped hydrograph and five differentfeed scenarios: no feed, constant feed, rising-limb only feed, falling-limb onlyfeed, and capacity-scaled feed. Each scenario with feed produced different bedloadtransport behaviors despite having the same total feed mass.Feeding on the rising limb strongly controlled sediment transport rates, overallbedload yield of the hydrograph, and bed scouring severity, but not bedload or bedsurface grain sizes. Feeding on the falling limb controlled bedload grain sizes,bedload transport rates in the early portion of the falling limb, and recovery of thebed to pre-flood morphology (eg. elevation and slope). Shifting the feed timingtowards one of the limbs accentuated the effects of that limb and influenced theoverall hysteresis trends. While the experiment design defaulted to a clockwisehysteresis for bedload transport, intermediate feed on the rising limb strengthenedthe hysteresis whereas heavy feed near the peak discharge or on the falling limbweakened the clockwise hysteresis. The bed surface did not develop armor orstructures, likely due to not enough time during the falling limb.Given that the sediment supply in natural streams can vary in both magnitudeand timing during a flood hydrograph, these experiments show that it is incorrect toassume that a constant feed will accurately represent bedload transport in naturalsettings.47BibliographyC. An, X. Fu, G. Wang, and G. Parker. Effect of grain sorting on gravel bed riverevolution subject to cycled hydrographs: Bed load sheets and breakdown of thehydrograph boundary layer. Journal of Geophysical Research: Earth Surface,122(8):1513–1533, aug 2017. doi:10.1002/2016jf003994. → page 44P. J. Ashworth and R. I. Ferguson. Size-Selective Entrainment of Load in GravelBed Streams. Water Resources Research, 25(4):627–634, 1989. → pages 1, 2L. Benda and T. Dunne. Stochastic forcing of sediment supply to channelnetworks from landsliding and debris flow. Water Resources Research, 33(12):2849–2863, 1997a. → page 2L. Benda and T. Dunne. Stochastic forcing of sediment routing and storage inchannel networks. Water Resources Research, 33(12):2865–2880, 1997b. →page 2A. C. Brayshaw, L. E. Frostick, and I. Reid. The hydrodynamics of particleclusters and sediment entrainment in coarse alluvial channels. Sedimentology,30:137–143, 1983. → page 5J. M. Buffington and D. R. Montgomery. Effects of sediment supply on surfacetextures of gravel-bed rivers. Water Resources Research, 35(11):3523–3530,1999. → page 2M. Church. Gravel-Bed Rivers. In Timothy Burt and Robert Allison, editor,Sediment Cascades: An Integrated Approach, chapter 9, pages 241–269. JohnWiley & Sons, Ltd., 2010. → page 2J. A. Clayton and J. Pitlick. Persistence of the surface texture of a gravel-bed riverduring a large flood. Earth Surface Processes and Landforms, 33:661–673,2008. doi:10.1002/esp.1567. → page 448W. Dietrich, J. Kirchner, H. Ikeda, and F. Iseya. Sediment supply and thedevelopment of coarse surface layer in gravel-bedded rivers. Nature, 340:215–217, 1989. doi:10.1038/340215a0. → page 4M. A. Elgueta-Astaburuaga and M. Hassan. Experiment on temporal variation ofbed load transport in response to changes in sediment supply in streams. WaterResources Research, 53:763–778, 2017. doi:10.1002/2016WR019460. →pages 2, 5L. E. Frostick, P. M. Luca, and I. Reid. The infiltration of fine matrices intocoarse-grained alluvial sediments and its implications for stratigraphicalinterpretation. J. Geol. Soc. London, 141:955–965, 1984. → page 4D. Gaeuman. Mechanics of bedload rating curve shifts and bedload hysteresis inthe Trinity River, California. In 2nd Joint Federal Interagency Conference onSedimentation and Hydrologic Modeling, Las Vegas, NV, June 2010. AdvisoryCommittee on Water Information. → page 3D. Gaeuman. Mitigating Downstream Effects of Dams. In Gravel-Bed Rivers,pages 182–189. John Wiley & Sons, Ltd, feb 2012.doi:10.1002/9781119952497.ch16. → page 45D. Gaeuman. HIGH-FLOW GRAVEL INJECTION FOR CONSTRUCTINGDESIGNED IN-CHANNEL FEATURES. River Research and Applications, 30(6):685–706, may 2013. doi:10.1002/rra.2662. → page 45D. Gaeuman, R. Stewart, B. Schmandt, and C. Pryor. Geomorphic response togravel augmentation and high-flow dam release in the Trinity River, California.Earth Surface Processes and Landforms, 42(15):2523–2540, jul 2017.doi:10.1002/esp.4191. → pages 45, 46S. Gibson, D. Abraham, R. Heath, and D. Schoellhamer. Vertical gradationalvariability of fines deposited in a gravel framework. Sedimentology, 56:661–676, 2009. doi:10.1111/j.1365-3091.2008.00991.x. → page 4B. Gomez and M. Church. An assessment of bed load sediment transportformulae for gravel bed rivers. Water Resources Research, 25(6):1161–1186,jun 1989. doi:10.1029/wr025i006p01161. → page 1R. H. Guthrie and S. G. Evans. Magnitude and frequency of landslides triggeredby a storm event, Loughborough Inlet, British Columbia. Natural Hazards andEarth System Sciences, 4:475–483, 2004. → page 249J. K. Haschenburger and P. R. Wilcock. Partial transport in a natural gravel bedchannel. Water Resources Research, 39(1), jan 2003.doi:10.1029/2002wr001532. → pages 1, 2M. A. Hassan and M. Church. Experiments on surface structure and partialsediment transport on a gravel bed. Water Resources Research, 36(7):1885–1895, 2000. → page 5M. A. Hassan, R. Egozi, and G. Parker. Experiments on the effect of hydrographcharacteristics on vertical grain sorting in gravel bed rivers. Water ResourcesResearch, 42:W09408, 2006. → pages 2, 3, 4, 5H. Haynes and G. Pender. Stress History Effects on Graded Bed Stability.Journal of Hydraulic Engineering, 133(4):343–349, 2007.doi:10.1061/(ASCE)0733-9429(2007)133:4(343). → page 5L. Hsu, N. J. Finnegan, and E. E. Brodsky. A seismic signature of river bedloadtransport during storm events. Geophysical Research Letters, 38:L13407, 2011.doi:10.1029/2011GL047759. → pages 3, 4R. Humphries, J. G. Venditti, L. S. Sklar, and J. K. Wooster. Experimentalevidence for the effect of hydrographs on sediment pulse dynamics ingravel-bedded rivers. Water Resources Research, 48:W01533, 2012.doi:10.1029/2011WR010419. → page 3C. Juez, M. A. Hassan, and M. J. Franca. The Origin of Fine SedimentDetermines the Observations of Suspended Sediment Fluxes Under UnsteadyFlow Conditions. Water Resources Research, aug 2018.doi:10.1029/2018wr022982. → page 3F. K. Ligon, W. E. Dietrich, and W. J. Trush. Downstream Ecological Effects ofDams. BioScience, 45(3):183–192, mar 1995. doi:10.2307/1312557. → page45L. Mao, J. R. Cooper, and L. E. Frostick. Grain size and topographical differencesbetween static and mobile armour layers. Earth Surface Processes andLandforms, 36(10):1321–1334, mar 2011. doi:10.1002/esp.2156. → page 4L. Mao, A. Dell’Agnese, C. Huinache, D. Penna, M. Engel, and G. Niedrist.Bedload hysteresis in a glacier-fed mountain river. Earth Surface Processesand Landforms, 39:964–976, 2014. doi:10.1002/esp.3563. → page 350H. Monteith and G. Pender. Flume investigations into the influence of shear stresshistory on a graded sediment bed. Water Resources Research, 41:W12401,2005. doi:10.1029/2005WR004297. → pages 4, 5D. R. Montgomery and J. M. Buffington. Channel-reach morphology in mountaindrainage basins. Geological Society of America Bulletin, 109(5):596–611, may1997. doi:10.1130/0016-7606(1997)109〈0596:crmimd〉2.3.co;2. → page 2D. B. Moog and P. J. Whiting. Annual hysteresis in bed load rating curves. WaterResources Research, 34(9):2393–2399, 1998. → pages 2, 3D. Paphitis and M. B. Collins. Sand grain threshold, in relation to bed ’stresshistory’: an experimental study. Sedimentology, 52:827–838, 2005.doi:10.1111/j.1365-3091.2005.00710.x. → pages 4, 5G. Parker and P. C. Klingeman. On why gravel bed streams are paved. WaterResources Research, 18(6):1409–1423, 1982. → page 2G. Parker and S. D. H. Stefan. Model Experiements on Mobile, Paved Gravel BedStreams. Water Resources Research, 18(5):1395–1408, 1982. → page 4G. Parker, M. Hassan, and P. Wilcock. Adjustment of the bed surface sizedistribution of gravel-bed rivers in response to cycled hydrographs. In H.Habersack and H. Piegay and M. Rinaldi, editor, Gravel-Bed Rivers VI: FromProcess Understanding to River Restoration, chapter 10, pages 241–289.Elsevier B. V., 2007. doi:10.1016/S0928-2025(07)11127-5. → pagesiii, 6, 44, 47G. Pasternack, A. A. Fulton, and S. L. Morford. Yuba River analysis aims to aidspring-run chinook salmon habitat rehabilitation. California Agriculture, 64(2):69–77, may 2008. doi:10.3733/ca.v064n02p69. → page 45D. M. Powell, A. Ockelford, S. P. Rice, J. K. Hillier, T. Nguyen, I. Reid, N. J. Tate,and D. Ackerley. Structural properties of mobile armors formed at differentflow strengths in gravel-bed rivers. Journal of Geophysical Research: EarthSurface, 121(8):1494–1515, aug 2016. doi:10.1002/2015jf003794. → page 4I. Reid and J. B. Laronne. Bed Load Sediment Transport in an Ephemeral Streamand a Comparison with Seasonal and Perennial Counterparts. Water ResourcesResearch, 31(3):773–781, mar 1995. doi:10.1029/94wr02233. → page 4I. Reid, L. E. Frostick, and J. T. Layman. The incidence and nature of bedloadtransport during flood flows in coarse-grained alluvial channels. Earth SurfaceProcesses and Landforms, 10:33–44, 1985. → pages 1, 3, 551J. C. Schmidt and P. R. Wilcock. Metrics for assessing the downstream effects ofdams. Water Resources Research, 44(4), apr 2008.doi:10.1029/2006wr005092. → page 45A. Simon and L. Klimetz. Relative magnitudes and sources of sediment inbenchmark watersheds of the Convervation Effects Assessment Project.Journal of Soil and Water Conservation, 63(6):504–522, 2008.doi:10.2489/jswc.63.6.504. → page 2V. B. Smith and D. Mohrig. Geomorphic signature of a dammed Sandy River:The lower Trinity River downstream of Livingston Dam in Texas, USA.Geomorphology, 297:122–136, nov 2017.doi:10.1016/j.geomorph.2017.09.015. → page 45C. P. Stark and N. Hovius. The characterization of landslide size distributions.Geophysical Research Letters, 28(6):1091–1094, 2001. → page 2J. G. Venditti, W. E. Dietrich, P. A. Nelson, M. A. Wydzga, J. Fadde, and L. Sklar.Mobilization of coarse surface layers in gravel-bedded rivers by finer gravelbed load. Water Resources Research, 46:W07506, 2010a.doi:10.1029/2009WR008329. → page 5J. G. Venditti, W. E. Dietrich, P. A. Nelson, M. A. Wydzga, J. Fadde, and L. Sklar.Effect of sediment pulse grain size on sediment transport rates and bed mobilityin gravel bed rivers. Journal of Geophysical Research, 115:F03039, 2010b.doi:10.1029/2009JF01418. → page 5J. M. Wheaton, G. B. Pasternack, and J. E. Merz. Spawning habitat rehabilitation-I. Conceptual approach and methods. International Journal of River BasinManagement, 2(1):3–20, mar 2004. doi:10.1080/15715124.2004.9635218. →page 45P. R. Wilcock and J. C. Crowe. Surface-based Transport Model for Mixed-SizeSediment. Journal of Hydraulic Engineering, 129(2):120–128, 2003.doi:10.1061/(ASCE)0733-9429(2003)129:2(120). → pages 1, 5P. R. Wilcock and B. T. DeTemple. Persistence of armor layers in gravel-bedstreams. Geophysical Research Letters, 32:L08402, 2005.doi:10.1029/2004GL021772. → page 4G. P. Williams and M. G. Wolman. Downstream effects of dams on alluvial rivers.Professional Papers 12826, U.S. Geological Survey, 1984. → page 4552B. T. Yuill and M. H. Nichols. Patterns of grain-size dependent sediment transportin low-ordered, ephemeral channels. Earth Surface Processes and Landforms,36(3):334–346, Feb. 2011. doi:10.1002/esp.2041. → page 353Appendix AExtra Figures54Figure A.1: Total bedload mass in the sediment trap at the end of each period.Periods are 20 minutes long, except for the first one (rising 50L/s),which is one hour long.55Figure A.2: Ratio of D84 to D50 using a 10 minute moving average of the datashown in Figures 3.4 and 3.5.56Figure A.3: Average water surface slopes between 4.5m and 6.5m. Slopeis calculated using OLS with no fixed intercept. The lack of consis-tent trends may be due to passing bedforms and low spatial resolution(0.5m) of the measurements.57Figure A.4: 8m DEM for Experiment 1A. Flow is from right to left. Colorbarshows millimeters above the flume floor.58Figure A.5: 8m DEM for Experiment 1B. Flow is from right to left. Colorbarshows millimeters above the flume floor.59Figure A.6: 8m DEM for Experiment 2A. Flow is from right to left. Colorbarshows millimeters above the flume floor.60Figure A.7: 8m DEM for Experiment 2B. Flow is from right to left. Colorbarshows millimeters above the flume floor.61Figure A.8: 8m DEM for Experiment 3A. Flow is from right to left. Colorbarshows millimeters above the flume floor.62Figure A.9: 8m DEM for Experiment 3B. Flow is from right to left. Colorbarshows millimeters above the flume floor.63Figure A.10: 8m DEM for Experiment 4A. Flow is from right to left. Color-bar shows millimeters above the flume floor.64Figure A.11: 8m DEM for Experiment 5A. Flow is from right to left. Color-bar shows millimeters above the flume floor.65Figure A.12: 8m DEM semivariogram for Experiment 1A. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.Figure A.13: 8m DEM semivariogram for Experiment 1B. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.66Figure A.14: 8m DEM semivariogram for Experiment 2A. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.Figure A.15: 8m DEM semivariogram for Experiment 2B. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.67Figure A.16: 8m DEM semivariogram for Experiment 3A. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.Figure A.17: 8m DEM semivariogram for Experiment 3B. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.68Figure A.18: 8m DEM semivariogram for Experiment 4A. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.Figure A.19: 8m DEM Semivariogram for Experiment 5A. Longitudinal and transverse lags range up to 300px and50px (60cm and 10cm), respectively.69


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