UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Advancements in measuring bed load transport with a magnetic detection system Argast, Timothy 2012

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

Item Metadata

Download

Media
24-ubc_2012_spring_argast_timothy.pdf [ 6.34MB ]
Metadata
JSON: 24-1.0105187.json
JSON-LD: 24-1.0105187-ld.json
RDF/XML (Pretty): 24-1.0105187-rdf.xml
RDF/JSON: 24-1.0105187-rdf.json
Turtle: 24-1.0105187-turtle.txt
N-Triples: 24-1.0105187-rdf-ntriples.txt
Original Record: 24-1.0105187-source.json
Full Text
24-1.0105187-fulltext.txt
Citation
24-1.0105187.ris

Full Text

Advancements in Measuring Bed Load Transport with a Magnetic Detection System by Timothy Argast  B.Sc., The University of British Columbia, 2007  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in The Faculty of Graduate Studies (Geography)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 20, 2012 © Timothy Argast 2012  Abstract A large number of bed load measuring devices have been developed in order to estimate sediment transport in rivers. In spite of that, the geomorphic and engineering communities do not have a reliable method to estimate bed material load. Available techniques suffer from insufficient temporal and spatial resolution to capture the variability inherent in bed load movement. This work involves the design and construction of an in situ magnetic detection device, which shows promise as a method capable of overcoming these limitations. The sensors work by inducing a magnetic dipole in naturally magnetic stones via magnets installed in the bed of the channel. These stones then pass over a coil of wire, inducing a small voltage, which is recorded. The system is installed in a five meter flume and is calibrated using video based particle tracking as well as manual sediment collection and sieving. Initial results indicate the new design performs significantly better than its predecessor.  ii  Table of Contents Abstract  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iii  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  v  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vi  Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ix  Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  x  1  Introduction . . . . . . . . . . . . . . . . . 1.1 Research Context . . . . . . . . . . . . 1.2 Research Problem . . . . . . . . . . . 1.3 Research Objective . . . . . . . . . . . 1.4 Bedload Transport Phases . . . . . . . 1.5 Methods of Bedload Sampling . . . . 1.6 Bedload Surrogate Monitoring Methods 1.7 In situ Magnetic Detection Devices . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  1 1 1 2 2 3 7 9  2  Design and Construction . 2.1 Theoretical Model . . 2.2 Design Improvements 2.3 Construction . . . . . 2.4 Model Results . . . . 2.5 Turntable Experiments 2.6 Flume Experiments . . 2.7 Field Site . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  12 12 17 19 20 27 28 34  3  Analysis and Results . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Turntable Results . . . . . . . . . . . . . . . . . . . . . . . . . .  40 40 40  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  iii  Table of Contents 3.3 3.4 3.5 3.6  Single Grain Size Flume Results Mixed Grain Size Flume Results Comparison of Methods . . . . Field Results . . . . . . . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  46 53 63 70  4  Discussion . . . . . . . . . . . . . . . . . . . 4.1 Progress in Estimating Bedload Transport 4.2 Validity of Assumptions . . . . . . . . . 4.3 Particle Counts . . . . . . . . . . . . . . 4.4 Particle Mass . . . . . . . . . . . . . . . 4.5 Future Work . . . . . . . . . . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  . . . . . .  71 71 71 73 74 75  5  Conclusions  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  81  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  83  Appendices A Sieve Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  88  iv  List of Tables 3.1 3.2 3.3 3.4  A.1 A.2 A.3 A.4 A.5 A.6 A.7 A.8 A.9  Particle Response for varying particle sizes and distance between the stone and the sensor. . . . . . . . . . . . . . . . . . . . . . . Comparison of signal response between the original and redesigned sensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary Statistics for the peak signal strength distribution. . . . Total masses calculated by each method for each of the single grain size runs. The BMD shows inconsistent results, over-predicting in all the size classes except for the 32-44mm class which it significantly under-predicts. . . . . . . . . . . . . . . . . . . . . . . . . February 4, 2010 Mixed grain size flume run sieve results . . . . . February 11, 2010 Mixed grain size flume run sieve results . . . . March 12, 2010 Mixed grain size flume run sieve results . . . . . March 9, 2010 Mixed grain size flume run sieve results . . . . . . November 4, 2010 11-16mm single grain size flume run sieve results November 24, 2010 22-32mm single grain size flume run sieve results December 5, 2010 32-45mm single grain size flume run sieve results December 12, 2010 8-11mm single grain size flume run sieve results December 12, 2010 32-45mm single grain size flume run sieve results  44 44 50  70 89 90 91 92 92 93 93 94 94  v  List of Figures 1.1 1.2 2.1 2.2 2.3 2.4 2.5 2.6  A Helley-Smith sampler being deployed in the Seymour River, British Columbia. . . . . . . . . . . . . . . . . . . . . . . . . . . A Pit trap and recording load cell trap . . . . . . . . . . . . . . .  5 6  2.15 2.16 2.17 2.18  Difference between effective height and bottom rolling height . . The effect of lateral displacement on signal response . . . . . . . Configuration of particle trajectory for the model. . . . . . . . . . A BMD sensor unit . . . . . . . . . . . . . . . . . . . . . . . . . The completed BMD for the flume experiments. . . . . . . . . . . Simulated Voltage and Integrated Signal for a single particle with no displacement. . . . . . . . . . . . . . . . . . . . . . . . . . . Simulated Total Integrated Signal for a single particle with no displacement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simulated Voltage (top) and Integrated Signal (bottom) for a single particle displaced to the right. . . . . . . . . . . . . . . . . . . . . Simulated Total Integrated Signal for a single particle with displacement to the left. . . . . . . . . . . . . . . . . . . . . . . . . Simulated Total Integrated Signal for two particles with no displacement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Turntable experimental setup . . . . . . . . . . . . . . . . . . . . The flume used in experiments . . . . . . . . . . . . . . . . . . . The GSD mounted at the end of the flume . . . . . . . . . . . . . The sediment collection bin at the end of the flume and collected samples awaiting sieving (insert) . . . . . . . . . . . . . . . . . . Single grain size experimental setup . . . . . . . . . . . . . . . . The grain size distribution of the flume bed material . . . . . . . . Map of the field study site . . . . . . . . . . . . . . . . . . . . . The BMD installed in East Creek. . . . . . . . . . . . . . . . . .  3.1 3.2 3.3  Diagram of integral limits used in the analysis . . . . . . . . . . . 41 Turntable results for varied stone sample sizes. . . . . . . . . . . 42 Signal response for varying distances from the sensor to the particle 43  2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14  13 14 16 20 21 22 23 24 25 26 29 30 32 33 35 36 37 39  vi  List of Figures 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12  3.13  3.14 3.15  3.16  3.17  3.18  3.19 3.20 3.21  3.22  Comparison of the old sensor to the new design for two particle size Particle count for the 8-11mm single grain flume run . . . . . . . Particle counts for the 11-16mm single grain size flume . . . . . . Particle counts for the 22-32mm single grain size flume . . . . . . Particle counts for the 32-45mm single grain size flume . . . . . . Summary of the frequency of the peak signal strengths for the various grain sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . Box and whisker plot for the various grain sizes tested in the single grain size experiments . . . . . . . . . . . . . . . . . . . . . . . Calculated mass separated into three minute intervals for the 811mm single grain size flume run. . . . . . . . . . . . . . . . . . Calculated mass separated into three minute intervals for the 1116mm single grain size flume run. Note that BMD yield is plotted on the right ordinate axis. . . . . . . . . . . . . . . . . . . . . . . Calculated mass separated into three minute intervals for the 2232mm single grain size flume run. Note that BMD yield is plotted on the right ordinate axis. . . . . . . . . . . . . . . . . . . . . . . Calculated mass separated into three minute intervals for the 3245mm single grain size flume run. . . . . . . . . . . . . . . . . . Particle count for the first mixed grain size (MX-LF1). A low discharge was used for this initial test. A hand count of the particles was not completed for this run. . . . . . . . . . . . . . . . . . . . Particle count for the second mixed grain size run (MX-LF2). Like the first, this experiment was run with a low discharge and had low transport rates after an initial burst. . . . . . . . . . . . . . . . . . Particle Count for the third mixed grain size run (MX-MF1). Flow was increased for this run; however, hand counts are not available due to a malfunction with the manual sediment collection. . . . . Particle count for the first high flow mixed grain size run (MXMF2). Two increases in discharge were conducted during this run, the first during the seventh interval and the second in the tenth. . . Particle count for the second medium flow mixed grain size run (MX-MF2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Particle count for the second high flow mixed grain size run (MXHF2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Particle count for the final mixed grain size run. (MX-MF3). This run had medium discharge and a single increase in flow during the third interval. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Calculated particle mass for the MX-LF1 run. The manually collected samples from this run were weighed, but not counted. . . .  45 47 47 48 49 50 51 52  52  53 54  55  55  56  57 57 58  58 59 vii  List of Figures 3.23 Particle mass for the second mixed grain size run, MX-LF2. The sieve and GSD show consistent results while the BMD does not. . 3.24 Calculated particle mass for the MX-MF1 run. Sieve data is not available for this run. . . . . . . . . . . . . . . . . . . . . . . . . 3.25 Particle mass for the first high flow run (MX-HF1). There is poor correlation between the three methods. . . . . . . . . . . . . . . . 3.26 Calculated particle mass for the MX-HF2 run. The BMD results do not correlate well with the sieve results. . . . . . . . . . . . . . 3.27 Particle mass for the MX-HF2 run. Manual sediment collection for this run was interrupted due to overflowing of the outfall tank. . . 3.28 Particle mass for the final mixed grain size run (MX-MF3). . . . . 3.29 A Particle count comparison between the GSD and BMD for the single grain experiments. The diagonal line represents a 1:1 ratio. 3.30 A magnified view of the low transport rates from Figure 3.29 . . . 3.31 A Particle count comparison between the GSD and BMD for the single grain experiments. . . . . . . . . . . . . . . . . . . . . . . 3.32 A Particle count comparison between the GSD and the sieved data for the single grain size experiments. . . . . . . . . . . . . . . . . 3.33 Particle count comparison between the BMD and the sieve data. This shows a systematic shift from the 1:1 line and a departure at higher rates indicating the BMD can be calibrated. . . . . . . . . 3.34 Particle count comparison for the low transport intervals between the BMD and the sieve data . . . . . . . . . . . . . . . . . . . . . 3.35 Particle count comparison between the GSD and BMD. . . . . . . 3.36 Particle count comparison for the low transport intervals between the GSD and BMD . . . . . . . . . . . . . . . . . . . . . . . . . 3.37 Particle count comparison between the GSD and the sieve data . . 4.1  4.2  A plot of sediment movement across the channel width through time for the 10-02-11 mixed grain size flume experiment.The graph shows spikes of sediment moving through the center of the channel. This is not a remarkable result from a flume experiment, but demonstrates the BMD’s potential for revealing patterns of sediment transport in natural channels. . . . . . . . . . . . . . . . . . A second plot of sediment movement in the flume shows similar results to Figure 4.4. This data is from the 10-03-05 mixed grain size experiment. . . . . . . . . . . . . . . . . . . . . . . . . . .  60 61 61 62 62 63 64 64 65 66  67 68 68 69 69  76  77  viii  Acknowledgments There are numerous people who have helped and advised me through the process of researching and writing this thesis. I would like to thank Marwan Hassan for his constant support and enthusiam for this project. I would also like to thank Randy Enkin for developing the theoretical background behind the BMD sensors and his technical advice. Sam Robinson and David Reid, my research assistants who worked many hours setting up the flume, helping run experiments and sieving sediment.I would also like to thank Andre Zimmermann for help getting the lighttable up and running and troubleshooting Labview. Last, but not least, I would like to thank my parents for their support and understanding my need to prolong my involvement in reality by a couple more years by deciding to go into graduate school.  ix  Dedication This is dedicated to Tom Bennett. Rock and roll, buddy, rock and roll.  x  Chapter 1  Introduction 1.1  Research Context  Sediment transport in rivers consists of the movement of individual grains along a stream channel, and can be divided into two categories based on the mode of transport. Sediment can be transported either as suspended load, material carried in the water column or bedload, material that rolls, slides or bounces along the bed. Movement and deposition of bedload is responsible for most of the changes to the morphology of gravel-bed streams, which compose the majority of head water streams in British Columbia, and most mountainous drainage basins. This makes understanding bedload movement important for land management practices, and restoration or rehabilitation of streams. Protection and management of aquatic habitat, including spawning gravel used by salmonids, requires an understanding of the bedload transport rates to reflect changes in flow and sediment transport regimes [Gomez, 1991]. Assessing channel stability and design stable channels, specifically determining channel maintenance or channel flushing flows, important for river engineering works, requires an understanding of bedload transport [Gomez, 1991]. Also, on a theoretical level, an accurate measurement of bedload is needed for the calibration of bedload transport formulas, which represents an entirely different approach to estimating bedload movement. Finally, because we do not have a reliable way to estimate sediment transport in streams, accurate data-sets of bedload transport are not available to calibrate these formulas.  1.2  Research Problem  Although considerable effort has been given to quantifying bedload movement, fluvial geomorphologists are still lacking a reliable method of measuring bedload transport. A large number of methods used today rely on taking samples at a single point within the channel as opposed to covering the entire width of the channel. Transport rates can vary considerably across different sections of the channel, so an accurate measurement is required to have adequate spatial resolution to capture these variations. Many bedload measurement techniques are also taken over a short  1  1.3. Research Objective period of time during bedload mobilizing events. The stochastic nature of particle mobilization means that bedload transport rates can vary significantly on short time scales. Therefore, to have an accurate measurement of the bedload moving in a channel, continuous sampling is needed for the duration of the event. It is also important for the method to be able to monitor bedload transport over a wide range of flow magnitudes. The amount of bedload transported in a single event can range over orders of magnitude, making it difficult to capture high transport rates using any of the currently available methods. Large events can overwhelm many methods, so that the resulting sample covers part of the event but not its entire length and only part of the channel. Conversely, it desirable for a method to also detect very low transport rates where particle movement is sporadic and discontinuous but adds up to a significant amount over the period of the entire event.  1.3  Research Objective  This gap represents a significant hurdle to estimating bedload movement and there is a clear need for the development of a method that can deal with the large volumes, and provide spatial and temporal information on bedload. This is what is needed to develop better models and provide frequent, consistent measurement of bedload transport. One such method that is currently in development is known as the Bedload Movement Detector (BMD)[Tunnicliffe et al., 2000, Hassan et al., 2009]. The BMD, which works on the principle of magnetic induction, shows promise as a technique that is capable of measuring transport over a range of scales and providing excellent spatial and temporal resolution. This thesis will further develop this technique by making design improvements suggested by prior research and attempt to calibrate the new design using field and laboratory settings[Rempel, 2005, Hassan et al., 2009].  1.4  Bedload Transport Phases  An understanding of how and when bedload moves is important for determining the design parameters for a new measurement technique. Field observations have shown that sediment mobilization can be divided into three phases[Wilcock and McArdell, 1993, 1997, Wilcock, 1998, Wilcock and Crowe, 2003, Hassan and Church, 2001]. In the first phase, the bed is relatively static and few particles are being mobilized. Fine material (sand and small gravel), originating from upstream or bank erosion, is moved over the bed, but there is no interaction with bed material. This phase of transport is common during low flow periods and in armoured channels. As discharge increases, the bed enters the second phase of transport, 2  1.5. Methods of Bedload Sampling known as partial transport. In this phase there is partial mobilization of individual particles or classes from the local bed surface. Usually the large fractions remain static. As discharge continues to increase, bedload transport enters the third phase of full mobility. As suggested by name in this phase all particles are fully mobile, meaning all particles are mobile in proportion to their presence in the bed surface. Full mobility for all particle sizes is not common in natural streams, and is usually of very short duration as the discharge levels required to attain full mobilization are not frequent. In each phase, the sediment transport rate increases but our ability to quantify it decreases. This means that the largest events that create the most morphological work are not captured, as measurement of bedload is a difficult undertaking, especially in high discharge events. There is no current method of bedload sampling that provides reliable measurements over a range of flow conditions with high spatial and temporal resolution. There is a strong need to develop a method that is able to over come these difficulties to improve our understanding of fluvial processes.  1.5  Methods of Bedload Sampling  The problem of measuring bedload has been addressed in various ways, including the development of several different sampling techniques. The use of sampling methods of bedload measurement all involve the removal of the sediment from the flow for measurement. While this is the most logical and straightforward approach, they all disturb the flow to some degree, which alters the transport rate. It is also important to examine if what they capture is actually representative of the amount and texture of bedload that is being transported. While each method does work in certain situations, they have limitations that prevent them from being used under different conditions or environments.  1.5.1  Box, Basket and Net Samplers  A wide variety of samplers have been developed and been in use since the early 1900s. The simplest designs are merely baskets placed in the channel while some are hand-held or cable mounted devices that consist of an opening to funnel sediment in at one end and a net at the other to catch the moving sediment. Ehrenberger [1931], and Nesper [1937] developed and tested a basket sampler made of fine chain mesh that is anchored directly to the bed. The advantage of these samplers is that they are relatively inexpensive and simple to use. The most common type is the Helley-Smith sampler, which is widely used by the USGS [Helley and Smith, 1971]. The Helley-Smith samplers use a nozzle on their openings to funnel  3  1.5. Methods of Bedload Sampling sediment in, capturing it in a bag (Figure 1.1). They provide a simple and reliable method of bedload sampling that can easily be deployed for large scale sampling programs. The design works well for material less than 4mm but under-samples larger sediment. The original design had an opening of 7.62cm, which has since been increased, and used 0.2mm mesh bags. It is suggested that Helley-Smith samplers can catch 90 to 100 percent of particles 0.5 mm to 32mm in size in stream velocities up to 1.5m/s[Hubbell, 1987]. Sample collection, however, is labour intensive and can be dangerous in high flows. The samples collected also do not reflect the temporal and spatial variability of bedload transport as the openings are small relative to the channel width and are too small to capture sediment over a significant width of the channel. To over come this discrepancy, it is necessary to make multiple traverses of the stream with multiple samples per traverse [Ryan and Troendle, 1999]. Sample duration varies from 30 to 60 seconds [Ryan and Troendle, 1999] to up to four minutes [Andrews, 1994] to attempt to capture random fluctuations in sediment transport. The presence of the sampler itself alters the flow and thus the sediment transport. The sampler may also not sit flush on the bed or disturb the bed causing sediment to be scooped up [Ryan and Troendle, 1999]. In some situations, the peak flow can occur during the night, making sample collection even more challenging [Bunte, 1996, Tunnicliffe et al., 2000]. Also, the small size opening on Helley-Smith samplers biases them against large sediment, and is not representative of what moves in the bed and cannot be used to reliably estimate bedload movement.  1.5.2  Pit Traps  A common alternative to sampling is the use of pit traps, which attempt to address some of the shortcomings of samplers. The simplest pit method was used by Hollingshead [1971]which consisted of digging a trench across the Elbow River in southwestern Alberta, and making consecutive surveys as it was filled during a nival event. More commonly, pit traps consist of a box or bucket that is installed in the bed of the stream [Powell and Ashworth, 1995, Hassan and Church, 2001, Church and Hassan, 2002], or as a trough that spans the entire channel (Figure 1.2). By using multiple buckets it is possible to obtain a coarse sampling of the spatial variability of transport across the channel [Hassan and Church, 2001, Church and Hassan, 2002]. They are flush with the bed to minimize the disruption to the flow as possible and provide a much more representative sample of the bedload. The temporal resolution of the data they provide is often only per event as emptying during events is difficult and can be dangerous at high discharge. If the trap cannot be emptied during an event then simple pit traps cannot provide information on the rate or timing of sediment movement within an individual event. Traps are also 4  1.5. Methods of Bedload Sampling  Figure 1.1: A Helley-Smith sampler being deployed in the Seymour River, British Columbia. subject to over passing, which is sediment that bounces over the trap, as well as over filling in large events (e.g. Hassan and Church, 2001). Installation and maintenance of pit traps can be very difficult and labour intensive. A successful addition to the pit trap has been to place a pressure pillow or load cell underneath the trap to record the weight of sediment as it accumulates [Reid et al., 1980, Sear et al., 2000]. The samples can be sieved to obtain a grain size distribution for the entire event but the grain sizes cannot be associated with transport rates. Recording traps are still, however, subject to the same over filling and over passing problems. Over filling is a significant limitation as the majority of bedload transport often occurs in a limited number of large events. Leopold and Emmett [1977] used a trap on the East Fork River which included a conveyor belt in the bottom of the trap that was capable of handling transport rates up to 9000kg/hr. Their system also included a series of gates that divided the channel into sections in an attempt to increase the spatial resolution across the channel. Due to their substantial setup requirements, they are limited to sites with good access and reliable power supply.  5  1.5. Methods of Bedload Sampling  Figure 1.2: A recording load cell trap similar to the one described by Sear et al. [2000] beside a traditional channel wide trap installed in East Creek, British Columbia.  6  1.6. Bedload Surrogate Monitoring Methods  1.5.3  Vortex Tube Traps  One of the more complex types of bedload measuring devices is the vortex tube. These are based on a design adapted from a method used to eject unwanted sand and silt from irrigation canals [Robinson, 1962]. They work by installing a tube inset into the bed spanning the channel, orientated at an angle to the direction of flow. The movement of water over the tube creates a vortex that pulls sediment out of the flow into the tube. The downstream component of the vortex moves the sediment down the tube to the edge of the channel where it is deposited in a box to be sieved and weighed. This method was pioneered by Klingeman and Milhous [1970] at Oak Creek and their data set remains one of the best bedload records today. Similar devices have been installed at Torlesse Stream, New Zealand by Hayward and Sutherland [1974], at Flynn Creek, Oregon by O’Leary [1981], and at Virginio Creek, Italy by Billi and Tacconi [1987]. Sediment was weighed every ten to twenty minutes by Hayward and Sutherland [1974] and every minute byBilli and Tacconi [1987]. Such data collection, however, is very labour intensive and the system is expensive to install. Also, the system does not record any of the spatial variation across the channel. Despite its sophistication, vortex tube devices are still subject to over passing in high flow [Hayward and Sutherland, 1974].  1.6  Bedload Surrogate Monitoring Methods  Due to the difficulties and disadvantages with the direct methods of bedload sampling, researchers have attempted to develop alternative methods collectively known as bedload-surrogate monitoring methods. A wide variety of surrogate methods have been developed ranging from very simple to quite sophisticated in their design and application. These methods do not disturb the flow or disrupt the bed as direct methods can, but require more complex analysis, and assumptions must be made that do not always hold true. This section will briefly examine the strengths and disadvantages of these methods.  1.6.1  Tagged Sediment  The use of sediment that is marked in some way so it can be identified, or tagged, as it moves downstream is one of the methods that is widely used. The tracers are placed in the bed in a starting zone and are recovered and surveyed following bedload movement events [Crickmore, 1967, Laronne and Carson, 1976]. Various marking methods have been used including paint, magnets, RFIDs and radioactive stones [Butler, 1977, Gintz et al., 1996, Hassan, 1990, Sobocinski et al., 1990]. Their travel distance and burial depth is recorded and used as an indirect measure 7  1.6. Bedload Surrogate Monitoring Methods of sediment transport and depth of the active layer. This method provides detailed information on the movement of individual particles and can be used to estimate the volume of mobile sediment during the event but not transport rate. In order to properly represent the movement of bedload, thousands of stones are required and recovery is very labour intensive.  1.6.2  Morphological Estimates  First used by Popov [1962], this method involves comparison of the changes in bed topography of a channel by repetitive cross-section or topographic surveys. Griffiths [1979] made extensive use of this method on the Waimakariri River in New Zealand. The density of the survey depends on the purpose of the investigation and how the data will be used, but is generally in the range of 3-5 points per square meter [Lane et al., 1994]. As several cycles of scour and fill can occur in a single event, surveys only measure the net bed elevation changes, which may not reflect the actual amount of transport that occurred. By identifying adjacent areas of erosion and deposition, it is possible to get an estimate of the typical transport distance of particles. An estimate of the transport can also be made by using the average erosion and deposition and the mean travel distance [Ashmore and Church, 1999]. Multiple surveys are required, which can become labour intensive for larger reaches.  1.6.3  Sonar Surveys  Sonar surveys are similar to morphological surveys only they use active sonar to monitor the height of the bed. This method is prone to creating noisy data that is difficult to interpret as mobile sediment, turbulent flow and suspended sediment can affect the readings by increasing the amount of back-scatter created from the water column and effectively masking the bedload from the sensor. Dinehart [1989] set up a sonar boom over a 60 meter wide section of the North Fork Toutle River, Washington and was able to observe coarse gravel dunes migrating downstream during a flood. Although it is difficult to collect good quantitative data with this method, it has been shown by Dinehart[1989, 1992] to be useful for making qualitative observations of migrating bed forms.  1.6.4  Acoustic Measurements  Several researchers have developed a method of using hydrophones to record the sound generated by colliding sediment. This method was initially developed by  8  1.7. In situ Magnetic Detection Devices Williams et al. [1989] in a four kilometer wide ocean channel by suspending hydrophones 0.25 meters above the ocean floor. They have shown that there is an inverse relation between the frequency of the impacts and the particle size and the acoustic intensity is directly proportional to the mass of the mobile sediment.Bänziger and Burch [1990] and Rickenmann [1994] used this method on Erlenbach Torrent, Switzerland, calibrating it against overhead video recordings and periodic surveys of retained sediment. Nine hydrophones were installed across a transverse structure spanning the channel that recorded sediment generated impacts due to the collision of particles on a steel plate. The results show a weak relation between impacts and particle count as the data was very noisy and difficult to interpret. Large variations in the impulses from the hydrophones were seen over short time intervals that did not correspond with flow conditions, requiring the data to be averaged on an event-by-event basis for a clear trend to be seen. The infrastructure used in this setup was quite expensive and required access by heavy machinery for construction.  1.6.5  Scour Chains  Another approach to measuring the aggradation and degradation of the bed is the use of scour chains. Scour chains consist of a length of chain attached to an anchor. The chain is installed using a steel tube, driven into the bed and then the tube is removed leaving the chain protruding from the surface. After an event the amount of scour or fill can be determined by the length of chain exposed compared to the initial amount [Colby, 1964, Leopold et al., 1966, Carling, 1987, Hassan, 1990]. While similar to morphologic surveys, the scour chains have the advantage of being able to measure one cycle of scour and fill. Chains are useful for obtaining an estimate of the amount of sediment moved during an event but provide no information on transport rate. Also, this method measures local sediment movement and is difficult to extrapolate to what is occurring at the reach scale. Hundreds of chains would be required to estimate scour and fill at a reach scale. Scour chains can also be very difficult to install in gravel or cobble bed streams. Disturbance to bed-forms or compromising the armour layer, if present, can lead to more scour than would occur naturally.  1.7 1.7.1  In situ Magnetic Detection Devices Early Development  One of the promising methods of measuring bedload is the use of in situ magnetic detection devices. This method was initially developed by Ergenzinger and Custer 9  1.7. In situ Magnetic Detection Devices [1982] at Calabria, Italy and Reid et al. [1984] at Turkey Brook, UK. The method involves the use of magnetic induction to detect the movement of individual particles. A line of magnets is installed in the bed, which induces a magnetic dipole in rocks containing ferrous and nickel minerals. A detector array is also installed in the bed of the creek consisting of a series of induction coils. As stones pass over the coils, the movement of the magnet dipole over the coil induces a voltage spike, which is recorded by a data logger, providing a continuous record of sediment passage. The device at Turkey Brook consisted of a single 2.3m long metal detector installed in a concrete case and was able to detect 100 artificial clasts with implanted ferrite magnets [Reid et al., 1984]. Ergenzinger and Custer [1983] were able to improve the design of their detector at Calabria, Italy so that stones with naturally occurring magnetic minerals could be detected instead. This is only a portion of the total transported bedload, however, as not all stones have naturally magnetic minerals. It was estimated that 40% of the bed material in Squaw Creek was magnetic enough to be detected, a greater percentage than what is caught by any sampler or trap [Ergenzinger and Custer, 1983]. In 1986 an improved sensor was installed in Squaw Creek which consisted of five 1.55 meter units with multiple paired coils allowing for some spatial resolution across the channel of the differences in transport [Custer et al., 1986]. It also had two arrays spaced 15cm apart to allow for an estimate of the velocity of passing sediment. Bunte [1996] used a strip chart system at Squaw Creek, which required manual counting and had a maximum resolution of 200 peaks per hour. Digital logging has improved processing the raw input signal, so many more peaks can be logged per hour and allows for complex post-processing and time series analysis.  1.7.2  Recent Work  The most advanced magnetic detection device has been developed and tested in O’Ne-ell Creek by Tunnicliffe et al. [2000]. This device, called the the Bedload Movement Detector (BMD), consisted of 82 sensors sampling at 100 Hz (see Figure 3 in Tunnicliffe et al., 2000). Each sensor was 10 cm wide and spaced every 15cm providing much greater spatial resolution. Like the sensor at Squaw Creek, the BMD was able to detect the passing of naturally magnetic stones. Due to heterogeneity in the lithology it was estimated that only 30% of the transported material could be detected. Testing and calibrating the sensors used by Tunnicliffe was carried out by Rempel [2005] in laboratory experiments at the University of British Columbia. Rempel designed experiments with both natural and artificial stones to isolate the effects of particle size, velocity and magnetic content on the shape of the recorded signal. He was able to produce empirical relations between the amplitude, width and the 10  1.7. In situ Magnetic Detection Devices integral of the sensor response with particle size, velocity and magnetic content. The problem with the Bedload Movement Detector design also lies in the interpretation of the collected data[Hassan et al., 2009]. Rempel [2005] concluded that improvements to the sensor design could lead to better results. These included creating a more uniform magnetic field, creating a continuous magnet that spanned the entire channel and analyzing signals from adjacent coils to estimate size and location of the particles. There has also not been any attempt to complete any field calibration of the magnetic sensor. By using a pit trap it would be possible to calibrate the signal with bedload measurement. This remains an open research path that is investigated here. The magnetic detection system has several advantages over other methods. It is capable of providing both high resolution spatial data across the stream channel as well as high temporal resolution data per event and throughout the flood season. Once set up, it has the potential to run without supervision, and the only regular maintenance required would be downloading the data. It could provide data from extreme events as it does not over fill like pit traps, though it would still be sensitive to over passing.  11  Chapter 2  Design and Construction The testing of the new sensor occurred in two distinct phases: construction and implementation. A theoretical model of how the sensor behaves was also developed prior to construction. The model was used to simulate the behavior of the sensor and provide a baseline data set of the theoretical results for comparison with the experimental results. Both field and laboratory experiments were planned, but due to complications discussed in further detail in following sections, only the laboratory component was completed successfully.  2.1  Theoretical Model  A theoretical model of the magnetic system has been developed to predict the response that the magnetic sensor system should produce under ideal conditions. The model shows that the time integral of the signal (I) produced by the sensor is only dependent on the stone volume and the trajectory height. I=  NA (χ0V ) B0 4π  2 z30  (2.1)  In Equation 2.1, NA 4π is a constant, dependent on the sensor components, χ0V is the magnetic perturbation of the stone, B0 is the applied magnetic field and z0 is the distance from the stone to the sensor. In application, all of these variables are known except for V , the stone volume and z0 . Calibration of the sensor will be required to see how each of these two variables affects the output of the sensor. The theoretical model also makes several assumptions, a few of which will need to be tested to see if they are valid for the experiments. The assumptions of the model are as follows: 1. The applied magnetic field is uniform and vertically applied (B0 ). 2. The sensing coil is horizontal. 3. The stone’s magnetic susceptibility (χ0 ) is known. The susceptibility distribution depends on the mineralogy of the rocks found in the stream. 12  2.1. Theoretical Model 4. The stone’s magnetic remanence is negligible, thus the model only considers induced magnetism. This assumption must also be tested using samples. 5. The stone is small enough to be modeled as a dipole. 6. The coil has zero size. 7. The stone travels at a fixed height over the center of the coil. The finite size of the stone and coil will lead to a wider signal response than would occur under assumptions five and six. This makes the effective height, the distance from the sensor to the center of the particle, of the trajectory higher than the bottom rolling height of the stone, the distance from the bottom of the stone to the sensor (Figure 2.1). Also, lateral displacement of the stone’s trajectory relative to the center of the coil will decrease the signal strength (Figure 2.2).  Figure 2.1: An illustration of the difference between effective height and bottom rolling height. The theoretical model assumes that the particle has no volume and therefore the effective height is equal to the bottom rolling height. In practice, the stone does have a volume, so the distance from the bottom edge of the stone to the sensor (hr ) is less than the distance from the center of the stone to the sensor (he ). These departures from the model’s assumption require laboratory analysis to determine their effect on the signal strength. It is believed that despite these assumptions, the model will still predict similar results as to what will be produced  13  2.1. Theoretical Model  Figure 2.2: An illustration depicting how the lateral trajectory of the particle will affect signal response. In (A) there is decreased signal strength due to lateral displacement. (B) shows full response when particles pass directly over the sensor.  14  2.1. Theoretical Model in the lab. The model uses Faraday’s Law of Induction which states the electromotive force in a coil is equal to the negative time rate of change of the magnetic flux that traverses the coil (Equation 2.2). ν =−  df d (NABz ) dBz =− = −NA dt dt dt  (2.2)  where • Bz = the vertical component of the stone’s induced field at the position of the coil. • N = number of coil windings • A = cross sectional area of the coil The magnetic moment of the stone is: m=  χ0V B0 µ0  (2.3)  where • χ0 = susceptibility • V = stone’s volume • B0 = the applied field • µ0 = permeability of free space The vertical component of the stone’s magnetic field experienced in the coil is: Bz =  (µ0 m) 3cos2 θ − 1 4π rP 3  (2.4)  where • m = the magnetization of the stone • θ = angle off from vertical between the center of the coil and the center of the dipole • z0 = distance from the center of the coil to the center of the dipole  15  2.1. Theoretical Model  Figure 2.3: Configuration of particle trajectory for the model. Assuming height (h) and particle speed (x = Ut,where U is speed) are constant v(t) = −  NA d3cos2 θ − 1 /dt χ0V B0 4π r3  (2.5)  where •  NA 4π  = constant dependent on sensor components  • χ0V the magnetic perturbation of the stone •  d3cos2 θ −1 /dt r3  is the stone’s trajectory  • B0 is the applied field The measured voltage rises to a peak, quickly crosses the zero and drops to a negative peak, followed by a rise to zero. The area under each signal lobe is independent of the stone’s speed, since the time integral (I) of the signal is the same as the spatial integral: ˆ  t1  df dt = dt  ˆ  t1  d f dx dt = dx dt  ˆ  t1  df dt = dx  ˆ  x1  U  d f dx dx U  ˆ  x1  df dx = f (x1 )− f (x2 ) dx t2 t2 t2 x2 x2 (2.6) where f is the magnetic flux traversing the coil. Since f (−∞) = 0, the integral of the signal from −∞ to the crossing over the coil at x = 0, where θ = 00 = 0 and r=z is: NA (χ0V )B0 (2/(z30 )) (2.7) I= 4π The only unknowns are the stone volume (V ) and the trajectory height (z0 ) assuming χ has been measured. U  U  16  2.2. Design Improvements  2.2  Design Improvements  Before either the laboratory or field experiments could begin, the new version of the sensor had to be designed and constructed. The first generation BMD created by Tunnicliffe contained several inherent weaknesses related to the sensor design. The work completed by Hassan et al. [2009] outlined the following issues: 1. Most of the recorded signals showed a single ideal shape. In some cases, however, individual passing particles yielded double peaked signals. This requires extra attention because such signals hinder the determination of signal width, which is likely to introduce errors into signal width measurements and hence particle size estimates. Double peaks are attributed to the annular shape of the magnet/sensor. 2. The width of the signal changes across the sensor face with a distinct dip near the center of the sensor. Furthermore, the signal amplitude and integral decline as the distance from the sensor centre increases. These deviations are likely related to the shape of the sensor. 3. Because of the shape of the sensor and the steel casing, the area close to the edge of the sensor and between the sensors presents a very weak field. Their experimental work shows that stones as large as 45 mm could pass between the sensors undetected. 4. Stones with particularly high susceptibility are likely to induce an inverted response (trough followed by peak) in adjacent sensors. 5. A large range of particle sizes is described by a relatively narrow range of response, making the ability to resolve different particle sizes difficult. These problems can be addressed by modifying the inductor coil and the magnet used in the sensor. The following two sections will address the changes made to each of these components.  2.2.1  Inductor Coil Improvements  Inductor coils consist of conducting wire wrapped around a core. The core can consist of ferromagnetic or other material of high permeability, or can be hollow, which is referred to as an air core. The inductance of the coil, which is measured in heneries, is the measurement of the conducting wires’ resistance to a change in current. As shown by Faraday’s Law, the higher the inductance of the coil will create a larger voltage spike. The inductance of the coil is calculated using equation 2.8. 17  2.2. Design Improvements  L = µ0 µr  N 2A l  (2.8)  • L= inductance (H) • µ0 = permeability of free space = 4π × 10−7 H/m • µr = relative permeability of core material • N= number of turns • A= cross sectional area of coil (m2 ) • D= length of coil (m) Equation 2.8 shows that the inductance is dependent on the core material and the density of the conducting wire around the core. The original sensor had 3000 turns of 40 gauge magnet wire, with a diameter of 1.3cm and length of 0.5cm, producing an inductance of 108mH. These coils were purchased as a pre-built, off-the-shelf item. Due to the unique application, it was not possible to find an off-the-shelf coil that improved on the first generation, so a custom coil was built. The new coils had 3200 turns of 36 gauge magnet wire, with a diameter of 3.0cm and length of 1.5cm, which produced an inductance of 5.0 heneries. The original coil had a ferrite core, which greatly improves the inductance, however including a core in the new custom coil was too expensive to be practical. The increase in inductance due to the larger surface area is sufficient to provide the needed improvement in inductance.  2.2.2  Magnet Improvements  Many of the problems arising from the previous design were due to the use of torus shaped magnets. This shape was used as it allowed for the coil to be placed in the center of the magnet so that the entire assembly had a flush surface, minimizing the distance between the particle and the coil (z). One of the consequences of using the torus magnet is that the magnetic field it creates is quite variable near its surface, and it has a change in polarity over the hole of the torus. This is believed to be the cause of the double peaked response mentioned in the beginning of this section. Another effect of the torus shape is a significant dip in signal width when the particle passes over the center of the sensor, as opposed to being offset to the side. Despite being the ideal path to induce voltage in the coil, the variability of the magnetic field causes a much weaker response than otherwise expected. The third weakness of the torus shape is that the round shape requires the individual 18  2.3. Construction sensors to be spaced apart along the array. This means that there are significant gaps where particles can pass over, without being detected by the sensor. It was originally thought that large stones passing over these gaps would trigger simultaneous response in adjacent sensors, but testing by Rempel showed that stones as large as 45mm could pass over undetected[Rempel, 2005, Hassan et al., 2009]. For the new design the torus shaped magnets were replaced with ceramic grade 8 anisotropic bar magnets, with a surface field strength (Br ) of 385 mT. They had a coercivity of 2.95 KOe and were 152mm long by 101mm wide by 25mm thick and are polarized through their thickness. The regular magnet shape allows for a much more uniform magnetic field and removes the problems associated with the hole in the center. It is also possible to place them directly end-to-end, forming a continuous bar magnet across the entire channel. This way even if particles pass between the coils, they are still over the magnets, so have a magnet dipole induced in them and create a voltage.  2.3  Construction  The Bedload Movement Detector was designed to be modular to allow for easy maintenance and replacement of individual sensors. The system consists of a main channel in which the sensors are slotted in next to each other. The field BMD used a custom built aluminum beam, constructed for the original BMD. Five millimeter gauge aluminum was used in the construction to ensure the system would be protected from the demands of the field environment but was not so thick that it would diminish the magnetic signal. Aluminum was the preferred material as it had sufficient strength, and was non-magnetic so as not to create background interference. The channel was covered with a 5mm thick aluminum cover which was sealed with industrial strength silicon. At each end of the channel was an aluminum upright used to anchor the system into the banks on the sides of the creek. Each sensor is built on ceramic block magnets 101.6mm long by 76.2mm wide by 25.4mm thick. Each magnet has two 30mm wide induction coils glued to its surface, spaced 30mm from each other and 15mm from the edge (Figure 2.4). This means that when the block magnets are placed end-to-end, all the coils are 30mm apart, except for the coils at the end which are 15mm from the edge of the sensor. The leads from the coils are soldered to a telecommunications cable consisting of twenty-four twisted pairs, which runs to the data acquisition board. Care was taken to shrink wrap and tape each individual connection to minimize the chance of a short circuit developing. The row of sensors are placed inside the aluminum channel that was then installed in the bed of the channel. Care was taken to minimize the distance between the lid of the case and the sensor as the sensitivity of the 19  2.4. Model Results  Figure 2.4: One sensor unit consisting of a bar magnet with two induction coils. These units are placed end to end to form a continuous row of sensors across the channel with no gaps between them. response varies as the inverse cube to the distance. The laboratory BMD was built in a similar manner, using the same modular sensors. As the flume was narrower only six units were needed for a total of twelve induction sensors. These sensors were encased in a plexiglass box and installed at the end of the flume immediately upstream of the GSD light table (Figure 2.5).  2.4  Model Results  A model was programmed in Matlab using the equations and assumptions outlined in Section 2.1 to demonstrate the theoretical response of the system in different scenarios. This proved to be a valuable asset when evaluating flume results by creating a visual output of what is expected from the theoretical signal. The computer model uses three adjacent sensors and a user specified number of particles. The properties of the sensor, the number of stones, and the particle characteristics are all entered as manual inputs. It then moves the particle along a linear trajectory and outputs the position at a user defined time interval. This code is then repeated for the number of particles chosen by the user and allows modification of the properties of each individual particle. The position data is fed into the sensor function which  20  2.4. Model Results  Figure 2.5: The completed BMD for the flume experiments. calculates the voltage and total signal strength at each time step. These results are returned in real-time in a graphical format. After the program has run a number of the particles defined by the user, the total signal generated is summed and returned graphically. This graph includes a breakdown of the signal of each sensor. To test the model several different scenarios were run. All runs had the same coil properties, stone volume and susceptibility. The results shown here illustrate the models ability to handle a single stone as well as multiple particles. The effects of lateral displacement are also shown. There are many more possible variations that could be run with this version of the model, but this set best illustrates the range the model is capable of producing. Other useful variations of the model would likely be a series of summations of these basic scenarios. The most basic configuration is a single particle with no lateral displacement. In this scenario, due to symmetry, the left and right sensors produce identical signals that exactly overlap. The results are shown below in Figures 2.6 and 2.7. To show the effects of lateral displacement, a single stone moving along a trajectory to the right of the sensor was modeled. In this scenario, there is a lack of symmetry so the result of each of the three sensors is unique. The results are shown in Figures 2.8 and 2.9. To show the effect of multiple particles, two stones with the same trajectory where modeled. The total integrated signal is shown in Figure 2.10. The voltage and integrated signal would be exactly the same as shown in Figure 2.6. 21  2.4. Model Results  Figure 2.6: Simulated Voltage and Integrated Signal for a single particle with no displacement.  22  2.4. Model Results  Figure 2.7: Simulated Total Integrated Signal for a single particle with no displacement.  23  2.4. Model Results  Figure 2.8: Simulated Voltage (top) and Integrated Signal (bottom) for a single particle displaced to the right.  24  2.4. Model Results  Figure 2.9: Simulated Total Integrated Signal for a single particle with displacement to the left.  25  2.4. Model Results  Figure 2.10: Simulated Total Integrated Signal for two particles with no displacement.  26  2.5. Turntable Experiments These graphs show exactly what the theoretical model predicts. All three sensors show different results when the particle trajectory is asymmetrical, as expected. These results can be used for comparison to empirical experiments with the bed load sensor, but they are still only considering point masses. The current model only considers point masses which does not account for the shape effect particles could have on the signal. The results in the previous section show exactly the type of signal that was expected. The double peak in the voltage signal behaves exactly as expected with varying distance from the sensors and the integrated signal shows the expected associated peak. The magnitude of the signals is less than what would be expected but this is simply due to the constants used in the signal calculations. These constants could be refined following further experimentation to produce more realistic results. The model requires refinement in two important areas: finite masses and time delays between particles. The first is in the modeling of finite masses. To deal with this, a two dimensional shape could be used with the third dimension integrated to a point in the plane. While this would be an improvement on the current model, it would still not fully capture the variations that would arise from real stones. Programming three dimensional shapes in Matlab quickly becomes problematic and difficult to handle. One way to get around this would be modeling a series of point masses at various positions and summing the total together to get an idea of what a three dimensional shape would produce. The current version of the model is in theory capable of doing this, however, assigning a discrete volume and particle susceptibility is not trivial. Further experiments with real particles would be required to determine what appropriate values should be used. The second issue is that the model considers all particles to be passing the sensor simultaneously in time. The model was designed like this to allow for the simulation of particles of finite volume to be modeled as a series of point masses. It would be helpful, however, to allow for time lags to show how this would affect the signal. This could be accomplished by establishing an absolute time scale and situating particles along this time scale. During the summation in the last step of the program this temporal offset could be accounted for.  2.5  Turntable Experiments  Following the completion of the new sensor design, a method was required to determine the effect of the design modifications and improvements. It was decided that a rotating platter, or turntable, would be used as this method was originally used by Rempel [2005] and isolates the sensor from external environmental vari27  2.6. Flume Experiments ables. The testing rig consists of a rigid styrofoam platter in which the sample stones are placed. The platter is attached to a rotating axle which has a drive motor at one end and is fixed to a support frame at the other (Figure 2.11). Care was taken to ensure that the system was well balanced and there was minimal friction on the platter as vibrations or precession of the platter would cause the frame and sensor to move. Movement of the sensor would add another velocity component to the particle and lead to a nosiy signal response. A dual speed record player was used for the drive motor which allowed for either thirty-three or sixty-six rotations per minute. The test sensor was fixed to an arm extending from the support frame so that it was positioned just above the platter. The height of the platter was also adjustable, so that the relationship between signal strength and distance could be tested. The sensors were connected to a PC with a National Instruments USB 6225 Data Acquisition Board and sampled at 1000 Hz. A series of tests were performed with varying distance from the sensor, various stone sizes, and two different speeds with both the original and new sensor. For the stone, samples constructed by Rempel were used to allow for comparison with results from his original study. The rocks were composed of a mix of iron filings, silica and a bonding agent. The artificial stones were preferred as it allowed control of the magnetic susceptibility of the stones between various sizes by using a fixed ratio of iron to silica, a result that would have been difficult to achieve with natural stones. Stone sizes used to test the new sensor were 11, 16, 20, 25, 32mm across the b-axis. Testing with the turntable was not as exhaustive as was what conducted with the original design by Rempel. The testing conducted here wasdone only to establish the increase in signal return due to the improvements, and not replicate the previous tests.  2.6 2.6.1  Flume Experiments Flume Setup  A series of flume experiments were conducted in the recently constructed Mountain Channel Hydraulics Experimental Laboratory. A flume channel was used in all of the experiments. This channel is five meters long, has a maximum depth of 70 cm and maximum width of 83 cm (Figure 2.12). For the purposes of these experiments, the full size was not required so a false bottom was installed and styrofoam inserts were placed on the edges of the channel to reduce the width, resulting in a channel that was 60 cm wide and 60 cm deep. The flume is capable of variable grades from 0 to 18%, but was kept at a constant five percent for these experiments. Sediment was collected from East Creek in the University of British Columbia Research Forest for use in the mixed grain 28  2.6. Flume Experiments  Figure 2.11: The setup used for the turntable experiments. The sensor is not mounted in this photograph.  29  2.6. Flume Experiments  Figure 2.12: The flume used in the BMD experiments nearing completion of construction. The blue gantry crane could be used to change the slope.  30  2.6. Flume Experiments size experiments. For the single grain size experiments, sediment from the lab was used as it had been previously sorted and painted for other projects. The Bedload Movement Detector was installed at the end of the flume immediately upstream of the GSD. For the GSD to operate, most of the water is drained off through a mesh bottomed section at the end of the flume so that there is minimal depth of water on the light-table. This drain would occasionally cause sediment to become trapped and manual cleaning would be required to keep it moving out of the flume. This would create a delay between when the BMD and GSD registered the particle, which could possibly result in a difference as to which interval it was logged. The BMD was connected to a National Instruments USB-6218 Data Acquisition Board and sampled at 500 Hz. The DAQ board feeds into a central computer running Labview to log the data as well as control the GSD. As the conditions in the channel itself are not important to these experiments and the morphology was not of any concern, no attempt at scaling was undertaken. The only requirement for the testing of the BMD was that sediment is mobilized and transported out of the flume. For each experiment, 10 cm of sediment was placed in the bottom of the channel. The first 80cm of bed was epoxied to the channel bottom to allow for an area of no sediment transport where flow velocity measurements could be made without risk of damage to the instrumentation. For the first two flume experiments, a Vectorino Acoustic Doppler Velocimeter was used but this was replaced by an electric conductivity meter, as the ADV data was very noisy and required considerable post-processing. Additionally, the high temporal resolution and directional data was not needed so the ECM was sufficient.  2.6.2  Grain-size Distribution and Solid Discharge Measurements  The Grain-size distribution and Solid Discharge (GSD) measuring device was developed for the study of step-pool formation in steep channels[Zimmermann et al., 2008]. The system uses a high resolution video camera (1392 x 1024 pixels) situated over a light table to continuously capture images of 2 - 181mm material as it exits the flume (Figure 2.13). Using the GSD was important as it is capable of matching the temporal resolution of the BMD, which allows for continuous comparison of sediment transport rates throughout the duration of the run.  2.6.3  Manual Sediment Collection  In order to obtain a real measurement of the material transported out of the flume, all material was collected to be weighted and sieved. For this purpose a collection system was installed at the outflow of the GSD that allowed sediment exiting the flume to be collected in a bin. These bins were changed every three minutes 31  2.6. Flume Experiments  Figure 2.13: The GSD mounted at the end of the flume. The white plexiglass is back lit allowing the camera mounted above to capture their silhouettes. The BMD is installed immediately upstream. 32  2.6. Flume Experiments  Figure 2.14: The sediment collection bin at the end of the flume and collected samples awaiting sieving (insert) through the course of each experiment in order to provide a coarse temporal resolution to calculate transport rates and isolate when sediment was being mobilized. Each three minute sample was weighed and sieved to obtain a grain size distribution. In addition, a particle count was done for all particles 5.6mm and larger.  2.6.4  Single Grain size Experiments  The first phase of the flume experiments were conducted using a single grain size. For these runs, pre-existing sediment that had already been sieved and painted by size class was used. Four different size classes were used: 8-11mm, 11-16mm, 2232mm and 32-45mm. As there was a limited amount of sediment in each size class available, the length and width of the flume were decreased by installing styrofoam panels along the edges (Figure 2.15). This created an area about half a meter wide and a meter long to be the sediment source for the experiments. In these runs, the flow was gradually ramped up to slowly mobilize particles and then the sensors could be tested with only one or two stones passing over it at a 33  2.7. Field Site time. This proved to be difficult due to the homogeneous nature of the sediment and often resulted in sheets of stones peeling off when the flow was increased and then no further transport as the bed surface stabilized. This was especially challenging with the smaller sediment which would all be washed out of the flume at once. The GSD struggled with these flashes of high transport rate and often misidentified clumps of smaller particles as as single large stone. For the 8-11mm stones, such a large number were flushed out of the flume, it was not practical to count them all. Instead, 1000 stones were counted out and weighed to determine an average weight per stone. This figure was then used to calculate the number of stones in the rest of the high volume bins.  2.6.5  Mixed Grain size Experiments  The second phase of testing involved using mixed grain size bed material. The bed material was collected from the East Creek and field sieved up to 90mm (Figure 2.16). As there was much more sediment available than in the single grain size experiments, a wider and longer channel was used as a sediment source. This gave an area of approximately seventy centimeters wide and two and half meters long. The sediment was thoroughly mixed to ensure there was no armouring of the surface. The fines were not excluded from the flume; as they are beyond the threshold of detection by either the GSD and the BMD, they add a level of background noise that is to be expected in field runs of the equipment. By running the experiments with sand, it allows for the analysis program to be calibrated with data similar to what would be obtained from field events, simplifying future analysis.  2.7 2.7.1  Field Site Site Selection  The ideal field site for Bedload Movement Detector is a small, gravel-bed piedmont stream, with frequent flood events that are of sufficient magnitude to mobilize bed material. In theory, it is possible to install the Bedload Movement Detector in any size channel, however anything larger than a couple meters is logistically complicated and cost prohibitive. The site chosen is located in the University of British Columbia Malcolm Knapp Research Forest on East Creek (Figure 2.17). East Creek is a small (3-5m) wide stream with a pool-riffle morphology. The high flow events are generated by cyclonic rainstorms and usually fall between the months of October and April. The creek is also the site of an on-going research project, so there is an extensive data-set extending back for several years. The site chosen is immediately upstream of an established pit trap, so that the sediment collected 34  2.7. Field Site  Figure 2.15: The setup used for the single grain size experiments. 32-45mm sediment is shown in this photograph. The BMD can be seen immediately downstream from the sediment. 35  2.7. Field Site  Figure 2.16: The grain size distribution of the material placed in the bed of the flume. The sieving and grain size analysis was completed immediately after obtaining the material from East Creek resulting in a large fraction of fine material. This was reduced during the flume runs which essentially washed the material.  36  2.7. Field Site from the trap can be used for calibration purposes. Quick access to the site is important so there can be immediate response following an event.  Figure 2.17: Map of field study site in the Malcolm Knapp Research Forest. The sensor was installed at the end of the first study reach.  2.7.2  Establishing Field Site  The chosen site was not connected to the power grid so an off-the-grid power supply system had to be installed. This consisted of four 6 volt 230 AHr Golf Cart batteries, wired in parallel and series to create a 12 volt system with 460 Ah. The 37  2.7. Field Site batteries were charged by a Honda EM3800SXC gasoline generator via a pair of DLS 45 amp DC battery chargers. The system was controlled by an Auto-Gen Start system made by Magnum Energy, which monitored the battery levels and ran the generator for a set period of time when the charge of the batteries dropped below a critical level. This configuration allows the system to be fully automated as long as the generator is supplied with gasoline. The generator was placed in a separate shelter to the data logging equipment, as far as possible from the sensor arrays in order to reduce any electromagnetic interference generated while charging the batteries. The data logging system installed used a low power 10V DC computer; however, it had USB bus controller conflicts with the National Instruments data acquisition board, so was replaced with a EeePc netbook running Windows 7. Despite requiring the addition of a Samlex 150W, 12V Pure sine wave inverter to charge the netbook, the new configuration required less power, and therefore could run longer between charging cycles. The netbook was connected to a National Instruments USB-6225 16-Bit, 250 kS/s M Series Multifunction Data Acquisition board, which has 80 analog inputs, forty for each array. The netbook was running a custom program created in the National Instruments programming software, Labview, which logged the raw voltage input for later processing. Each channel was sampled at 500 Hz to ensure adequate resolution of each signal spike. The Bedload Movement Detector was placed in a wooden weir that was installed in the bed of the creek in the summer during low flow (Figure 2.18). The weir consisted of a flat surface that extended half a meter upstream of the BMD and provided two troughs to set the sensor arrays in so they would be flush with the bed surface. The decreased roughness over the weir ensured that sediment would move over the sensor rather than deposit on it.  38  2.7. Field Site  Figure 2.18: The BMD installed in East Creek. Two arrays are installed in the gray wooden weir which was used to promote passage of sediment so it would not aggrade on top of the sensors.  39  Chapter 3  Analysis and Results 3.1  Analysis  The analysis was completed using a custom program created with Labview, National Instruments’ graphical development environment. Data is read back in 200 millisecond intervals. In each interval the signals for each sensor are adjusted to a common base level. This is accomplished by calculating the average for each section and then subtracting that value, on a point-by-point basis, from the original signal. This corrects any systematic differences between channels. Once on a common baseline, the signal is run through a filter to remove some of the noise caused by interference from background electronics and electrical wiring in the building. This is accomplished using one of Labview’s built in filtering packages. Following on the work of Tunnicliffe and Rempel, a low-pass Butterworth filter was used, which blocks any signals above a given frequency while signals below the threshold are not affected. A threshold of 55Hz was used as this removes the majority of interference from AC wiring in the building, which operates at 60Hz, while not filtering out much of the desired signal. Following the signal cleaning, each channel is then run through a peak detection algorithm included in the Labview signal analysis package. All peaks over a set threshold are counted, and an integral of the area beneath them is calculated. The limits for the integration are set at half the amplitude of the signal peak, which is located via a recursive algorithm that is run on each peak. This analysis provides both a count of the number of peaks and the integral underneath the peak, which was outlined in the Theoretical Model section. For flume runs, the data is then lumped into three minute intervals so it may be compared directly with the sieve data.  3.2  Turntable Results  To investigate how the new sensor performed, the speed of the particle, the distance from the particle to the sensor and the size of the stone were varied. The velocities tested were limited to the two speeds available on the turntable, 33 or 66 rotations per minute. Due to the high sampling rate of 1000 Hz, the faster velocities were  40  3.2. Turntable Results  Figure 3.1: An illustration of how the integral limits were determined in the signal analysis. A peak of amplitude A that is above the threshold of the background noise was selected by the peak detection algorithm. The integral limits,a and b are determined by a recursive algorithm that determines where the voltage is equal to A/2 on either side of the peak.  41  3.2. Turntable Results  Figure 3.2: Turntable results for varied stone sample sizes showing signal response. Signals are staggered in time for clarity. well within the instruments’ ability to resolve the signal spike. As predicted by the theoretical model, the faster velocities resulted in a stronger spike. To test the effect of stone size 11, 16, 22, and 32mm sample stones were used with the stones set 40mm from the sensor. The signal spikes showed a gradual increase in strength with increasing grain size, with the exception of the 32mm stone which resulted in an order of magnitude increase from the 25mm stone, while the 11mm stone produced a peak that was just discernible above the background noise level. As the turntable tests are as noise free an environment in which the sensors will ever be run, this sets the lower limit of the size of particles the sensor can detect. In field conditions, it is expected that 16mm will be the effective cut-off. These results are summarized in Table 3.1 and plotted in Figure 3.2. To test the response to change in the distance from the sensor to the stone, three test distances were used. First the stones were set so that they were as close as possible to the sensor without touching, a distance estimated to be approximately 2mm. This distance is expected to be representative of stones moving in contact with the bed, as in the practice the lid of the sensor case will be separating the 42  3.2. Turntable Results  Figure 3.3: Signal response for varying distances from the sensor to the particle. Note that the signal response at 100mm is barely discernible from the background noise. stones from the sensor. Second, the stones were set a distance of four centimeters, the nominal distance expected for saltating, or rolling stones moving in transport. Thirdly, the stone was set a distance of ten centimeters from the sensor. In the stream, stones at this height in the water column are essentially in suspension, and not the intended target of this sensor. Nonetheless, this condition was tested to determine how the sensor performed beyond its intended design. For all of these tests a 22mm stone was used and the turntable was set to 33 rotations per minute. As the signal strength varies with the cube of the distance between the stone and sensor, the changes between each level were quite noticeable. Both the two millimeter and forty millimeter test showed a strong signal spike that was easily detected above the background noise level (Figure 3.3). The one hundred millimeter test did not produce a strong enough signal to allow for detection, as was expected as this is beyond the design range of the sensor. The most important test was to compare the new sensor with the old to determine if the design modifications had in fact made a significant improvement in the 43  3.2. Turntable Results  Table 3.1: Particle Response for varying particle sizes and distance between the stone and the sensor.  Table 3.2: Comparison of signal response between the original and redesigned sensor.  signal return (Table 3.2). For this experiment two particle sizes were used: 32mm and 22mm. For each run the sensor was placed 40mm from the particle and the platter was set to spin at 33 rotations per minute (Figure 3.4). The new sensor clearly out-performed the old one with a 0.0632 volt (V) response to the 32mm particle compared to a 0.0280 V response from the old sensor. The 22mm particle had a similar result giving 0.0160 V for the new sensor compared to 0.0079 V for the old sensor. These results showed the new sensor was performing as expected and confirmed that the steps taken to improve the design had a significant effect on the signal response. Tests were also run at both 33 and 66 rotations per minute. The faster speed was only possible with the 22 and 32mm stones as the smaller sizes were thrown from the platter; however, the sensor and data acquisition board had no difficulties with the higher velocities. The resulting signal spikes showed a higher peak value and narrow width, as was expected.  44  3.2. Turntable Results  Figure 3.4: Comparison of the old sensor (solid lines) to the new design (dashed lines) for two particle sizes. The new design performed significantly better.  45  3.3. Single Grain Size Flume Results  3.3 3.3.1  Single Grain Size Flume Results Particle Count  The single grain size experiments produced some varying results. These results are plotted by separating each flume experiment into three minute intervals to allow comparison with the sieve results, as the manual sediment collection trap was swapped every three minutes (Section 2.6.3). The individual sieve results are given in Appendix A. The Bedload Movement Detector (BMD) and the Grain-size distribution and Solid Discharge (GSD) results were summed over the same three minute intervals. Generally, the particle count from the BMD is less than the trap sieve count but greater than the GSD, although there are variations to this trend. The first run was with stones 8 to 11mm in diameter. As 11mm is at detection threshold limit, this size class was not expected to perform particularly well and the count was significantly less than the trapped sieved results (Figure 3.5). Interestingly, the BMD gave a higher result than the light-table, which was also operating at its detection threshold limit. This may be due to the fact that the light-table was counting clumps of particles as larger grain sizes, which were not counted in the final tally. As groups of particles tended to exit the flume together, it is likely that this clumping behavior also accounted for many of the peaks detected by the BMD. As the light-table was further down the flume from the sediment source, these clumps were more likely to have dispersed as they passed over the light-table, and thus not have been detected, leading to a lower count from the GSD. The second single grain size class tested was 11-16mm. This size was within the BMD’s ability to detect and the results reflected that (Figure 3.6). This run featured a large mobilization of sediment in the last interval as flow was increased to the point at which this size class was fully mobile, leading to a nearly complete scour of sediment from the flume. The count for this interval had to be estimated from the total weight using the average weight of a single stone. For the first four intervals, the transport rate was relatively low and both the BMD and GSD show comparable results to the sieve data. In the fifth interval when the flow was increased, the BMD reported fewer stones than either the GSD or the sieve analysis. For the 22 to 32mm run, there was a marked improvement in the BMD count (Figure 3.7). During this run, sediment mobilization was much more discrete, and the waves of gravel that were mobilized in the smaller two class sizes did not occur. This likely increased the number of particles the BMD was able to detect. Noticeably, the light-table performed significantly worse for this particle size. This may be caused by the image processing algorithms of the GSD misidentifying the sizes. The final run consisted of sediment 32-45mm and showed the best correlation 46  3.3. Single Grain Size Flume Results  Figure 3.5: Particle counts separated into three minute intervals for the 8-11mm single grain size flume run. The three minute interval is set by the sieve results of the manual sediment collection. There was significant under-counting by both the BMD and GSD in this experiment.  Figure 3.6: Particle counts separated into three minute intervals for the 11-16mm single grain size flume run. There was mobilization of a large amount of stones at the end of this experiment that is reflected in both the BMD and GSD counts.  47  3.3. Single Grain Size Flume Results  Figure 3.7: Particle counts separated into three minute intervals for the 22-32mm single grain size flume run. The BMD performed well on this run, consistently tracking the changes in sediment transport. between all three methods (Figure3.8). This is likely due to the large size, making them easily detectable and the transport rate was fairly low so there were no waves of particles to complicate the peak detection. The cause for the light-table showing significantly more particles in the final interval than the sieve results is likely due to false positives created by air bubbles. The flow for this run was fairly high to maintain sediment transport and consequently the flow over the GSD was not as laminar as other runs and air bubbles were common. In the final interval, the BMD showed the opposite trend and reported relatively fewer particles, when compared to the sieve data than the previous intervals. To further investigate, the amplitude of all the detected peaks was summarized and grouped into 25mV bins. The signal strength was then plotted against the number of times, or frequency, that it occurred during the flume run (Figure 3.9.) The sensor gave a strong response for all four particle sizes but did not differentiate between them well. The mean signal response between the various grain sizes is not significantly different. This suggests that the sensor was having difficulty determining the actual signal created by a passing particle from the background noise. It is believed that this mean value represents the background noise and is not representative of the response for the given particle size. The frequency increases with decreasing particle size simply as there were more particles in transport during those flume runs. This plot also shows that most  48  3.3. Single Grain Size Flume Results  Figure 3.8: Particle counts separated into three minute intervals for the 32-45mm single grain size flume run. Transport rates were much lower and occurred in discrete pulses leading to a good count by the BMD. of the peak values were around 200mV for all grain sizes tested. Based on the results from the turntable experiments, it was expected that there would be an increase in the average peak signal with increasing grain size. Summary statistics were calculated for this data and are presented in Table 3.3 and plotted in a box and whisker diagram (Figure 3.10). This clearly shows that there is little variation between the mean peak amplitudes for each size class. The maximum for each size class, however, does show a clear trend of increasing peak amplitude with increasing grain size.  3.3.2  Particle Mass  Following the particle count, the mass for each interval was calculated for the BMD, GSD and weighed. Unlike the particle count, there was no clear trend in these results. For the 8-11mm stones, the BMD gave a reasonable result for the first interval followed by large overestimates for the following two intervals and a large underestimate for the last interval (Figure 3.11). In the 11-16mm experiment the BMD consistently highly overestimated the yield for all intervals (Figure 3.12). Oddly, when there was a spike in sediment transport in the fourth interval, the BMD results did not reflect this and instead returned a value in the same range as the previous intervals. For the 22-32mm particles, the BMD again consistently overestimated all in49  3.3. Single Grain Size Flume Results  Figure 3.9: A summary of the frequency of the peak signal strengths for the various grain sizes tested. Most detected signal peaks were around 200mV, regardless of particle size.  Table 3.3: Summary Statistics for the peak signal strength distribution.  50  3.3. Single Grain Size Flume Results  Figure 3.10: A box and whisker plot for the various grain sizes tested in the single grain size experiments. There is not a significant difference between the mean signal strength for each size. The maximum signal strength for each size class does, however, follow the expected trend.  51  3.3. Single Grain Size Flume Results  Figure 3.11: Calculated mass separated into three minute intervals for the 8-11mm single grain size flume run.  Figure 3.12: Calculated mass separated into three minute intervals for the 1116mm single grain size flume run. Note that BMD yield is plotted on the right ordinate axis.  52  3.4. Mixed Grain Size Flume Results  Figure 3.13: Calculated mass separated into three minute intervals for the 2232mm single grain size flume run. Note that BMD yield is plotted on the right ordinate axis. tervals (Figure 3.13). The BMD does seem to follow an increasing trend from the initial interval to the second, although this does not continue through the rest of the experiment, as it reported an increase in the third interval while the sieve results showed a decrease. The last single grain experiment was the 32-45mm size class and shows contrary results to the first three runs (Figure 3.14). In this run the BMD consistently underestimated mass for all three experiments. It also does not show the same trend as either the sieve or GSD results, giving the lowest yield in the last interval when the highest transport was actually observed.  3.4 3.4.1  Mixed Grain Size Flume Results Particle Counts  For each run, the number of particles was counted by the BMD, the GSD and by hand during sieving. The hand count only counted particles down to 5.6mm as labour constraints made counting any more impractical. Additionally, both the BMD and GSD have a significant drop in efficiency around this size due to large amounts of noise in the data so particles of this size are below their detection thresholds. Unlike the single grain size experiments, periods of low intensity transport were much easier to produce and individual grains were easier to detect. 53  3.4. Mixed Grain Size Flume Results  Figure 3.14: Calculated mass separated into three minute intervals for the 3245mm single grain size flume run. Figure 3.15 is the first run (MX-LF1) in the sensor testing. No sieve data is available for this run as there were technical difficulties with the sediment collection bin; however, this run shows the performance of the BMD versus the GSD. Except during the spike in transport, centered around the third interval, the BMD is counting more stones than the GSD. The count trend is similar between the two devices with the exception of the second interval and the last, suggesting that perhaps operations related to starting and stopping the flume run may be affecting the results. The second experimental run has data from the BMD, GSD and sieving is shown in Figure 3.16. This experiment shows good agreement in trend of all three data-sets. After the fourth interval, the BMD begins over-counting the number of stones compared to the sieved data, and shows an anomalous increase in the last interval. For the third experiment (Figure 3.17), like the first, no sieve data was collected, and was only run for fifteen minutes, whereas the other runs are all approximately thirty minutes. As in the first run, sediment accumulation along the sides of the bin jammed the bins in the collection area and prevented us from swapping them every three minutes. This run is shown as the GSD seems to count significantly more stones at high transport rates than the BMD, as can be seen in the fourth interval. This is the opposite of the trend seen in other events and at lower transport rates. Figure 3.18 shows a full run with all three data sets. There is very good agree54  3.4. Mixed Grain Size Flume Results  Figure 3.15: Particle count for the first mixed grain size (MX-LF1). A low discharge was used for this initial test. A hand count of the particles was not completed for this run.  Figure 3.16: Particle count for the second mixed grain size run (MX-LF2). Like the first, this experiment was run with a low discharge and had low transport rates after an initial burst.  55  3.4. Mixed Grain Size Flume Results  Figure 3.17: Particle Count for the third mixed grain size run (MX-MF1). Flow was increased for this run; however, hand counts are not available due to a malfunction with the manual sediment collection. ment in the trends between all data sets with the GSD counting fewer stones than the BMD, both of which under-counted the trapped number of stones in each event. There is a noticeable drop in performance during high transport intervals. A peculiar result from the GSD is show in Figure 3.19. The experiment shows that the trends between all three methods are quite similar; however, there is an initial peak seen in the GSD data which is significantly greater than the other data sets. This may have been caused by air bubbles in the system, which would have been counted as stones by the light table. During one run (Figure 3.20) reduced drainage of the outflow tank resulted in significant flooding of the lab. As a result, no manual sediment collection was complete, so no sieve data is available, although the GSD and the BMD continued to operate normally. The trends in the data sets are quite similar, with the peak transport lining up in each data sets. Similar to the last event, there is an over-prediction at the end of the run for the GSD. This may be caused by a stone becoming stuck as the flow levels drop. The final mixed grain size experiment (Figure 3.21) shows similar results to the run shown in Figure 3.18. The BMD results are considerably lower than either the GSD or sieve results for the high transport intervals indicating that there is some character of these burst of sediment that prevent the BMD from acquiring good particle counts.  56  3.4. Mixed Grain Size Flume Results  Figure 3.18: Particle count for the first high flow mixed grain size run (MX-MF2). Two increases in discharge were conducted during this run, the first during the seventh interval and the second in the tenth.  Figure 3.19: Particle count for the second medium flow mixed grain size run (MXMF2).  57  3.4. Mixed Grain Size Flume Results  Figure 3.20: Particle count for the second high flow mixed grain size run (MXHF2).  Figure 3.21: Particle count for the final mixed grain size run. (MX-MF3). This run had medium discharge and a single increase in flow during the third interval.  58  3.4. Mixed Grain Size Flume Results  Figure 3.22: Calculated particle mass for the MX-LF1 run. The manually collected samples from this run were weighed, but not counted.  3.4.2  Particle Mass  The second variable measured in each data set is the particle mass. Unlike the count, this was done for all grain sizes in the sieve data. It should be noted that some of following graphs plot the BMD yield on the right ordinate axis as the magnitude of the values differs so greatly from the GSD or sieve data that plotting them on the same access is not practical. While this does not allow for comparison of the magnitude of each interval, it does allow for a comparison of the trends in the data. Figure 3.22 shows the mass calculated by each of the three methods for the first run, split into three minute intervals. While a hand particle count was not completed for this run, the mass of each bin was measured, so unlike the count data there is sieve data for the mass. This figure shows little agreement between the three data sets. While both the sieve and GSD show an early peak in transport, they occur on two adjacent intervals. The BMD shows a peak before these two peaks, and a declining trend over the next two intervals. There is also a later peak which is not reflected in either of other two data sets. The second run (Figure 3.23) shows better agreement between the GSD and the sieve data but trends in the BMD are strikingly different. There is some evidence of a small peak in all three data-sets at the third interval; however, the lack of similarity in the rest of the run makes this peak unconvincing. The third run (Figure 3.24) does not have any sieve data. Compared to the GSD, the initial few intervals seem to be following the same trend; however, the 59  3.4. Mixed Grain Size Flume Results  Figure 3.23: Particle mass for the second mixed grain size run, MX-LF2. The sieve and GSD show consistent results while the BMD does not. GSD shows a peak in sediment transport in the fourth interval while the BMD shows a drop. The results from the fourth run (Figure 3.25) are much more promising. The peak at the fourth interval is clearly seen in all three data-sets, as is the peak at the seventh interval. Following the second peak, the BMD trend does not match the other two data-sets as closely, although with the exception of the last point, is still fairly similar. The fifth run (Figure 3.26) also seems to show better agreement between the data sets. The peak over the fourth and fifth intervals is reflected in all data-sets. After this, however, the BMD data diverges from the other two, showing peak at the seventh interval and a sharp drop in the last interval where both the GSD and the sieve data show the opposite. The sixth run (Figure 3.27), which was the experiment without sieve data due to flooding, shows very good agreement between the GSD and the BMD. The peak at the fifth interval and during the last interval is evident in both data sets The final run (Figure 3.28) shows good agreement between all data-sets, with the peaks at the third and sixth interval evident in all. It does however, seem as though the BMD data is not scaling linearly, as the second peak is of a greater magnitude than the first, even though this not reflected in the GSD or sieve data.  60  3.4. Mixed Grain Size Flume Results  Figure 3.24: Calculated particle mass for the MX-MF1 run. Sieve data is not available for this run.  Figure 3.25: Particle mass for the first high flow run (MX-HF1). There is poor correlation between the three methods.  61  3.4. Mixed Grain Size Flume Results  Figure 3.26: Calculated particle mass for the MX-HF2 run. The BMD results do not correlate well with the sieve results.  Figure 3.27: Particle mass for the MX-HF2 run. Manual sediment collection for this run was interrupted due to overflowing of the outfall tank.  62  3.5. Comparison of Methods  Figure 3.28: Particle mass for the final mixed grain size run (MX-MF3).  3.5  Comparison of Methods  In addition to comparing the results over the span of a single run, it is possible to combine all the interval bins into one data set to see how each method of measurement performs against each other. By plotting one method against the other it is possible to identify trends and relations between methods. It is important to remember that while the sieve data is assumed to be correct, there is error associated with this as well, though it is relatively minor. If the measurements for a given interval are exactly equal, they will fall on the 1:1 line. A comparison of the single grain size experiments shows a fairly good relation at lower transport rates but variable results in intervals with higher counts (Figure3.29). There are three intervals from the 8-11mm particle experiment in which the BMD show significantly lower counts than the sieved data. There were also two intervals from the 11-16mm experiment that the BMD over-predicted at high transport rates. Focusing on the intervals with fifty or less counted particles, it is seen that the BMD generally under-predicts the sieve data, which is expected (Figure3.30). There are several bins where there was no sediment transport and the BMD still recorded several false positives as well as one bin where the BMD over-predicted. This interval was also from the 11-6mm experiment. If these intervals are ignored the BMD is seen to be predicting 70% of the sieved results. Comparison between the GSD and BMD counts shows rough agreement at 63  3.5. Comparison of Methods  Figure 3.29: A Particle count comparison between the GSD and BMD for the single grain experiments. The diagonal line represents a 1:1 ratio.  Figure 3.30: A magnified view of the low transport rates from Figure 3.29 64  3.5. Comparison of Methods lower transport rates, although there is scatter on either side of the 1:1 line (Figure 3.31). There is one outlier where the GSD predicted a considerably higher number of particles in the 8-11mm experiment. Although the 11-16mm experiment created outliers when comparing the BMD to the sieve data, there is better agreement between the GSD and BMD as both methods over-predicted, with the exception of one interval where the GSD counted significantly more stones. It would be expected that the experiment with material larger than 16mm (Figures 3.7and 3.8) would provide better results, but this is not seen in Figure 3.31.  Figure 3.31: A Particle count comparison between the GSD and BMD for the single grain experiments. For contrast, the comparison of the GSD and the sieved data is given in Figure 3.32. It also shows better performance at lower transport rates, with fairly little scatter around the 1:1 line but quite variable results for intervals of high transport rate. Comparison of the mixed grain experiments show similar results to the single grain size runs. The first two graphs (Figures 3.33,3.34) compare the sieved data against the results from the BMD. The first shows all transport rates (Figure 3.33) while the second graph (Figure 3.34) plots the same data but is magnified on the lower transport events. There is a clear under-prediction at higher counts, while there is more scatter in the data at lower counts. Although there is more scatter at the lower counts, it is still centered around the 1:1 line. 65  3.5. Comparison of Methods  Figure 3.32: A Particle count comparison between the GSD and the sieved data for the single grain size experiments. The data for the GSD and BMD comparison is not clear as was the case with the single grain size experiments (Figure 3.35). There is considerable scatter at high transport rates, however the data at lower transport rates indicates that the BMD may be recording higher counts. This, however, is a very weak relation (Figure3.36). It would be expected that there would be the opposite trend, as the GSD should be able to detect more particles than the BMD on average. Again, the comparison of the GSD and sieve data is provided for contrast (Figure 3.5). There is a clear trend of under-predicting by the GSD although there are several outliers. To compare the calculated masses between all three methods for the single grain size experiments, the masses of all intervals for each size class were summed (Table3.5).The total mass for the 8-11mm run shows close results between all three methods. There was considerable variance between the individual intervals suggesting that perhaps most of the sediment transport occurred at interval changes and was mistakenly attributed to the wrong interval. Interestingly, both the BMD and GSD reported a greater mass than the sieve results indicating there were a large number of false positives in this size class. While the first single grain size experiment showed promising results, the following three did not. The 11-16mm and 22-32mm experiments showed a massive 66  3.5. Comparison of Methods  Figure 3.33: Particle count comparison between the BMD and the sieve data. This shows a systematic shift from the 1:1 line and a departure at higher rates indicating the BMD can be calibrated.  67  3.5. Comparison of Methods  Figure 3.34: Particle count comparison for the low transport intervals between the BMD and the sieve data  Figure 3.35: Particle count comparison between the GSD and BMD. 68  3.5. Comparison of Methods  Figure 3.36: Particle count comparison for the low transport intervals between the GSD and BMD  Figure 3.37: Particle count comparison between the GSD and the sieve data  69  3.6. Field Results  Table 3.4: Total masses calculated by each method for each of the single grain size runs. The BMD shows inconsistent results, over-predicting in all the size classes except for the 32-44mm class which it significantly under-predicts.  over-prediction by the BMD by approximately five and seventeen times respectively, when compared to the sieve results. This trend did not continue to the 3245mm size class where the BMD predicted only 42% of the sieved material. This is contradictory to what would be expected from the initial turn table experiments as the 8-11mm size class was at or below the detection limit for the BMD. The GSD over predicts the 8-11mm class but predicts approximately 80-94% of the remaining three class sizes.  3.6  Field Results  The site at East Creek suffered from a series of disabling setbacks. The sensor arrays were completed in the summer of 2008 and installed that fall so monitoring could commence in time to capture any rain events that year. While there were a couple of events through the season, the low power computer used to control the BMD and collect the data had a conflict communicating with the data acquisition board. Due to the remote location troubleshooting was a long and arduous process, which finally resulted in the computer being replaced with a netbook. Unfortunately, by the time the issue had been resolved the wet season was over and there were no more bedload mobilizing events that year. The following year the wet season began with a large flood event that scoured out the leading edge of the weir causing the entire weir and sensor to be lifted off the bed and shifted downstream. No useable data was collected from this event. The weir was repaired and reinforced with sandbags and boulders to ensure that it would not move again. Unluckily, the 2009-2010 winter was uncharacteristically dry and there were no further events that mobilized sediment large enough to be detected by the BMD. Weather is an unpredictable variable that simply did not work during the time line of this project.  70  Chapter 4  Discussion 4.1  Progress in Estimating Bedload Transport: A Problem Refined but not Resolved  The timing and location of bedload movement is highly variable and can range over several orders in magnitude. As exhibited in Sections 3.3 and 3.4, it is clear that even in the controlled and simplified environment of a flume, these variations can prove difficult to quantify. These complications only intensify when attempting to make bedload measurements in natural channels. The Bedload Movement Detector is capable of overcoming these obstacles and the new design shows marked improvement over the first generation. The sensor design created by Tunnicliffe et al. [2000] and tested by Hassan et al. [2009] had several problems related to the annular geometry of the sensor. The complex magnetic field it produced caused double peaked signals, signal dips near the center of the sensor and variations in the signal amplitude and integral that were dependent on the particle distance from the center of the sensor. By replacing the magnet with a block magnet with a more uniform magnet field, these problems were eliminated. Furthermore, unlike the ring magnets, the block magnets could be placed end to end forming a continuous strip across the channel to prevent stones from passing between the sensors without being detected. The uniform magnetic field also ensures that high susceptibility particles will induce the same signal response in immediate sensors as they do in adjacent sensors. Further examination of the steps used in the experiments and analysis is considered in the remainder of this chapter. While significant progress was made in the sensor design and implementation, improvements to the signal analysis are required for the BMD to reach its full potential.  4.2  Validity of Assumptions  In order to support the theoretical model behind the calculations used for the simulations (Section 2.4) and analysis (Section 3.1), seven assumptions were made (Section2.1). The majority of these hold true for the purposes of this work, and the 71  4.2. Validity of Assumptions rest are believed to have negligible effects. The first was that the applied magnetic field is uniform and vertical. This is true in the range immediately adjacent to the bar magnets used, which is also where the bedload is being transported. Farther from the surface of the magnet, the magnetic field lines do become curved, but this is too distant from the sensor to have a noticeable effect. The second assumption is that the stones magnetic susceptibility is known. While the exact susceptibility of each stone is not known, the stones sampled by Rempel [2005]allow for an average susceptibility to be calculated. This is likely a major source of error in these experiments but is extremely difficult to determine, so the average will have to suffice for these calculations. The other assumption of the stone’s magnetic properties is that the remanence is negligible. The homogeneous nature of the sediment from East Creek ensures that this is a safe assumption. East Creek is the site of an artificial magnetic tracer study, where every year the tracers are located using a magnetic detector. It has been found during these surveys that the number of stones with high magnetic remanence is minimal. This field data indicates that for the majority of stones transported, it is a safe assumption that the induced magnetism is all that need be considered. Another assumption of the stone is that it is small enough to be modeled as a dipole. For the smaller particles this is certainly a valid assumption as they do not have a significant size relative to the sensor. For coarse gravel sized particles and larger (>32mm), this assumption does not hold as well, as their specific composition may mean that there is considerable variation in magnetic properties through the volume of the particle. This assumption cannot be avoided as accounting for more complex magnetic distributions would require a much more complex model and introduce several new variables that would not be possible to know for individual stones. So while this assumption is not ideal, it is necessary to model the particles as a dipole. The sixth assumption that is not strictly true is that the coil has zero size. Clearly the coil has a physical size, but this assumption is made to maintain the simplicity of the model. Adding the complexity of a finite sized sensor cannot be justified as it would not improve the results significantly due to the low signal to noise ratio. The final assumption is that the stones travel at a fixed height over the center of the coil. As most of the stones are rolling or sliding over the sensor, the fixed height assumption is valid. Moving directly over the center of the coil, however, is more problematic. The spacing of the coils was kept to a minimum to increase the likelihood of particles crossing the array near a sensor, as well as preventing the array from having any holes where a particle could pass without detection.  72  4.3. Particle Counts  4.3  Particle Counts  The data from the flume runs shows fairly consistent trends in terms of particle counts. Generally speaking, the BMD under-counts the number of particles when compared to the sieve or GSD results. This is expected as not all stones will have enough magnetically susceptible material to allow for detection. What is important is that the BMD results show the same relative increases and decreases that the GSD and sieve results show. This can be seen in Figures 3.16, 3.18 and 3.21 which all show good agreement in the trends between all three data sets. A noticeable area of difference between all data-sets is the count values for three minute intervals at the beginning and end of each experiment. In several experiments, where there was good agreement for most of the duration of the run, the beginning and end varied considerably. Error in the GSD counts may be due to air bubbles exiting the system in the initial stages of the run as the GSD cannot differentiate between stones and air bubbles when they are back-lit. This does not explain the high counts from the BMD, as it is not affected by air bubbles. It is possible that this discrepancy may be partly due to variations in the start and cutoff times between the three methods of measurement. The GSD and the BMD were running on different clocks, and while synchronization was attempted in the postprocessing, it is quite likely they were not perfectly synchronous. This also applies to the switching of the sediment collection bins. The timing of the switch was set on the the computer clock that was running the BMD; however, it is not the instantaneous transition that the analysis assumes. During high flows, the switching process can take between twenty and thirty seconds. During low transport rates, the variations in the timing of the cutoff of each bin have a minimal effect, but if they occur during a burst of high sediment transport, which are often fairly short in duration, it could cause the peak to appear to shift to the next interval. In the future, it may be possible to reprogram both the GSD and BMD data logging programs so that their clocks are synchronous while recording data. The switching of the sediment collection bins is an irreducible source of error that cannot be completely eliminated. It could, however, be reduced by not using a set time period as the interval, and instead switch the bins during periods of low transport. This would require careful notation of when the bins were switched relative to the clocks running the other sensors, and would require additional steps in the processing in order to normalize the sediment transport rates in each interval. This additional work could be worthwhile if it were successful in avoiding bin switches during high transport rates. Looking at the sum of each flume experiment eliminates the binning as a possible error source from the analysis. When this is done, the number of inconsistent results is reduced and there are only a handful of outliers remaining (Figure 3.33). 73  4.4. Particle Mass This suggests that a relation does exist between the BMD count and sieve count and that further calibration work will allow the BMD to perform accurate particle counts over a range of transport rates.  4.4  Particle Mass  While the particle count produced by the BMD shows promise, the calculated particle mass for the flume runs indicates that further refinement is required. The particle counts appear to vary in relation to the observed transport rate in the flume, while the mass does not. In the single grain size experiments, the calculated mass does not increase to reflect spikes in sediment transport that occurred during the flume run. A possible cause of the calculation errors may be the algorithm used for determining the integral boundaries. The recursive algorithm used is relatively simple in that it does not account for the overall shape of the signal spike and assumes it has a fairly clean shape, similar to the one shown in Figure 3.1. The actual signals produced can have much more complex shapes, including multiple peaks or periods of elevated signal as waves of gravel are passing over the sensor, which may cause incorrect integration limits to be set. Rolling sediment, which was frequently observed during flume runs, is likely to produce these complex signals. Particle response may need to be characterized by more than simply the peak to determine if the time intercept at half the amplitude is a reasonable choice for the integration limits. One confounding issue is that while the count data appears to be fairly consistent, the integral data appears to be much more random, and often does not reflect the pattern shown by the GSD or sieve data. It is possible that this difference is caused by low frequency, high magnitude noise. It was noticed during the operation of the flume, that movement of personnel on or around the flume during operation, did in fact create a noticeable trace on the BMD output. Unlike the interference caused by the ambient electromagnetic background, such as building lighting, this type of noise is impossible to filter out. As this source was of relatively low frequency, it would not be expected to have a significant effect on the count for each interval, but they would likely have an impact on the integration data. In the future, it could be possible to compare the area of the peaks created by the largest particles to peaks known to be caused by external influences. Another main difference between the mass and count data was that the count data was cut off at 5.6mm while the mass data used all grain sizes. For these smaller particles the signal to noise ratio approaches one as the size decreases. To avoid identification of too many false positives, the threshold for peak identification was set higher 74  4.5. Future Work than would allow for the detection of most of the smaller particles, unless they had an ideal trajectory. As identified by Hassan et al. [2009], a large range of particle sizes is described by a relatively narrow range of response. The changes made to the induction coil and the magnet were meant to address this issue; however the improvements were not as pronounced as hoped. The increased sensitivity of the sensor amplified the response of all particles so the over all range did not increase that much. For example, the ratio of 32mm peak to the 22mm peak for the original sensor is 3.5, while the new design only increased this to 3.9. Increasing this range would be beneficial in determining the mass of each particle in mixed sediment runs. While the Bedload Movement Detector requires further calibration, it is capable of providing high resolution data of bedload transport. As an example two mixed sediment runs have been plotted showing signal response across the channel width versus time (Figures 4.4 and 4.4). These figures clearly show a pattern of sediment moving through the center of the channel, with spikes in sediment transport when sections of the bed were scoured out. Signal response has been lumped into one minute intervals for ease of plotting, but could be plotted at one second intervals. These results do not shed any more light on a flume experiment; however, with no further modification or calibration this system could provide valuable information on the dynamics and timing of sediment movement in a natural channel. With the addition of a water level sensor, this system could reveal when bedload was mobile in relation to a flood hydrograph and through which part of the river the sediment was moving. Examination patterns of bedload transport at these resolutions have not been accomplished to date.  4.5 4.5.1  Future Work Field Site  One of the main problems related to the field was having a computer capable of dealing with the the high rate of data transfer. This can be dealt with by either increasing the computing power or by decreasing the resolution of the data being captured. The original 10W computer lacked a powerful enough USB bus host Controller to handle the data stream coming from the data acquisition board and would quickly crash due to buffer overflows. The use of the netbook appears to have fixed this issue as it contains a more powerful processor and is running Windows 7, which has improved USB transfer protocols. Several event length test scenarios should be run to see if this holds true. While the netbook uses very little power, it needed to be run through the inverter as a power plug of the correct size could not be found. If the computer could be run directly from the DC system with 75  4.5. Future Work  Figure 4.1: A plot of sediment movement across the channel width through time for the 10-02-11 mixed grain size flume experiment.The graph shows spikes of sediment moving through the center of the channel. This is not a remarkable result from a flume experiment, but demonstrates the BMD’s potential for revealing patterns of sediment transport in natural channels.  76  4.5. Future Work  Figure 4.2: A second plot of sediment movement in the flume shows similar results to Figure 4.4. This data is from the 10-03-05 mixed grain size experiment.  77  4.5. Future Work out using the inverter, it would be much more efficient than switching from DC to AC and then back to DC. While it is not possible to create sediment mobilization in the absence of significant precipitation, their rarity makes it important to be prepared well in advance. Another source of potential future problems is the weir that was installed in East Creek. It was not designed to easily allow for adjustment to changes in the bed surface. So if the site is significantly aggraded or degraded in the next few years, considerable work will be required to adjust the sensor to the correct height. Another source of problems with the weir is that the slots containing the arrays are prone to catching sediment. As the particles work their way into the slot, they act to wedge the slot open further leading to more sediment being caught. One way to prevent this is to lay a sheet of plastic over the slots in the weir; however, this decreases the distance between the particle and sensor leading to a weaker response. Work should also be done to reduce the noise from external sources that is being picked up by the data acquisition board. One simple improvement would be to reduce the cable length between the sensors and data acquisition board. Longer lengths of cable are more likely to pick up background electromagnetic interference and degrade the already weak signals from the sensors. A second improvement would be to rearrange the hardware in the instrument shed to distance the data acquisition board from the electrical charging equipment and logging hardware. As these are currently in close proximity, it is likely that they are causing a level of background noise that could easily be eliminated.  4.5.2  Sensor Design  One of the major features on the original sensor that was not replicated on the new design was the use of a low carbon steel yoke. The yoke helped to contain the magnetic field, making it more uniform. It was not possible to find a source in time to include it in the new design, however, the use of the bar magnets should have helped to balance out the lack of yoke. In future versions, the inclusion of the yoke could help further improve the response. Another feature not included in the second version was a ferrite core. Inclusion of a core would increase the inductance, and therefore the response of the system. These two adjustments to the current design would greatly improve the system and could easily be added to the current sensors. Another important improvement would be to encase the coils a resin to protect them from corrosion and reduce the likelihood of short circuits. Some initial testing with epoxy proved problematic as the chemical reaction that occurred while curing was quite exothermic and damaged the sensor. The addition of a coat of protective resin is important for any field study that will be conducted for multiple years. 78  4.5. Future Work The difficulty of removing the sensor from the channel and performing any field maintenance requires that the sensor array be as robust as possible. Care should be taken to ensure that the casing will not effect the magnetic properties of the sensor. As in the field installation, minimizing the cable length between the sensor and the data acquisition board will help decrease signal degradation and background noise levels.  4.5.3  Case Design  The response of the system varies inversely to the cube of the distance from the particle to the sensor. This means that minimizing the thickness of the sensor case can have a significant effect on the response. The second generation BMD reused the original case as it was convenient, of the correct size and tough enough to withstand being left in a creek bed for several years. The original case is quite sturdy and over-built, making getting the sensor close to the lid . The lid was built with a non-standardized drill pattern, so building a new lid for the case would require considerable custom work. In the future a new lid could be created of thinner material with out risking damage to the sensor. It could also be designed to allow for the sensors to be attached directly to the lid. In the current design the sensor are recessed below in an aluminum trough, creating a gap of about 10mm. As the sensor response decreases with the square of the distance, any changes to the design that will bring the sensor closer to the sediment will be a significant improvement.  4.5.4  Signal Analysis  The primary issue currently preventing further development of the Bedload Movement Detector is the limitations of the signal analysis. While the sensor showed good results from the turntable experiments, the signal response generated by real sediment transport in the flume was considerably more complex and varied. In addition, the background noise in the flume runs was considerably higher making peak detection difficult. This is evident from the single grain size experiments which show that the mean value of the detected peaks does not change significantly with grain size (Figures 3.9 and 3.10). This indicates that the majority of peaks detected are actually false positives and the threshold requires further adjustment. The maximum value for each grain size is clearly seen to be increasing so the sensor is responding to the change in grain size (Figure 3.10). While the maximum is not an appropriate value to use as a threshold, it may be used as a scaling factor to help calibrate the analysis. There is too much variability in the signal to increase the threshold too much as it would result too many particles being missed. 79  4.5. Future Work It is assumed by the integration algorithm that the signal peak follows a relatively simple shape similar to the one shown in Figure 3.1. In practice, the shape is often more complex and may not return to the baseline at all. When clusters of particles move over the sensor it results in multiple peaks which the current analysis program only treats as a single peak. In order to handle this complexity, more variables are need to classify the shape of these signals. One such variable could be determining the time between adjacent peaks. Several consecutive peaks can then be flagged as a cluster of stones and a separate algorithm can be used to determine the integration limits. Peak width at half amplitude is another variable that can be used in signal classification. While this will be related to how quickly the particle is passing over the sensor, it could also be used to determine if the signal is returning to baseline voltage or not. Situations were this doesn’t happen could be flagged to be integrated in separate manner than normal peaks. A rolling particle is an example of a scenario that may produce such a signal response. Once the issues with peak identification are resolved, a valuable addition to the analysis would be to compare the signal response between adjacent coils. The current analysis cannot differentiate between a large particle with an average amount of ferrous material and a small particle with a large amount of ferrous material. Looking for concurrent signal spikes between adjacent sensor could be used as a secondary factor when determining particle size as a large particle would be expected to trigger multiple sensors. A small particle on the other hand, would only be expected to trigger a response in the closest sensor as the induced voltage decreases at 1/r2 . With an improved signal response, it should be possible to increase the thresholds the noise filters to clean the signal without removing real peaks.  80  Chapter 5  Conclusions The modifications made to the sensor design significantly improved the sensitivity and increased the signal strength of the BMD. A side by side comparison of the new design to the previous design by Tunnicliffe showed a marked improvement. Several tests on the turntable apparatus showed the new sensor was performing well and gave a strong response over a range of particle sizes and distances. The use of a large block magnet rather than an annular shaped magnet reduced the number of double peaked signals and provides a more even response across the width of the sensor face. The block shape also allowed the sensor to be placed end to end, effectively creating a single continuous bar magnet eliminating the possibility of stones passing between sensors undetected. The uniform magnetic field of the block magnets also prevents s adjacent sensors from recording an inverted response from stones with high susceptibility. The range of signal responses for varying particle sizes was slightly increased; however, resolving different particle sizes still remains a challenge. In place of field data, two series of flume experiments were conducted. The first series of single grain size flume experiments displayed promising results for the particle count showing a systematic variation from the sieve results. With further calibration the BMD should be able to provide accurate particle counts. This will allow for recording and examination of temporal and spatial patterns of bedload transport in natural streams. Figures 4.4 and 4.4 demonstrate the ability of the BMD to record and visualize these patterns. It would also allow for comparison of relative transport rates, although determining absolute amounts will require further improvements to the BMD. The second series of flume experiments consisted of mixed grain sizes. As in the single grain size experiments, the particle count returned promising results but the particle mass calculations were inconsistent. While it is the ultimate goal of the BMD to provide an accurate measurement of bedload transport, further development of the analysis program is required before this can be achieved. Currently, the calculated mass from the BMD does not follow the trends of either the sieve or GSD data, and therefore require more work before calibration is possible. Improvements are required to the analysis program’s ability to classify and properly integrate the range of signal peaks created by real sediment movement. 81  Chapter 5. Conclusions Two arrays of forty sensors were installed in East Creek but due to an unfavorable field season, it remains untested. Examination of the temporal patterns of sediment transport in relation to flood hydrographs is possible with the current installation. The array of 40 sensor across the channel sampling at 500Hz and would create a high resolution data set of sediment transport patterns unlike any other currently available. The full potential of the Bedload Movement Detector is still limited by our ability to correctly analyze the data it collects. Improvement on this limitation can be made on two fronts: fine tuning the design of the sensor and development of the analysis program. From the design side, creating a magnetic yoke for the sensors and adding a ferrite core would help improve the signal strength, making peaks stand out more from the background noise. Also, decreasing the cable length from the sensor to the data acquisition board is a simple fix that will improve results. Improvements to the analysis program will yield the greatest improvement to the Bedload Movement Detector, but will require considerable work to accomplish. A methodology needs to be developed to allow the analysis software to identify and classify different signal peak shapes that are created by clusters of sediment or rolling particles. There is a wide range possible signals, so this task is not trivial and will require an extensive amount of programming to correctly identify peaks created by stones and determine meaningful integration limits. However, if successfully completed, the BMD would provide bedload transport studies an unparalleled methodology capable of overcoming the difficulties that are currently limiting development of this field.  82  Bibliography E.D. Andrews. Marginal bed load transport in a gravel bed stream, sagehen creek, california. Water Resource Research, 30(7):2224–2250, 1994. P.E. Ashmore and M. Church. Gravel-bed Rivers in the Environment, chapter Sediment Transport and river morphology: a paradigm for study, pages 115–140. Water Resource Publications, Colorado, USA, 1999. R. Bänziger and H. Burch. Acoustic sensors (hydrophones) as indicators for bed load transport in a mountain torrent. In Proc. Int. Symp. on Hydrology in Mountainous Regions, I Hydrological Measurements; the Water Cycle, pages 207– 214, 1990. P. Billi and P. Tacconi. International Geomorphology 1986, chapter Bed load transport processes monitored at Virginio Creek measuring station, Italy, pages 549– 559. John Wiley and Sons, New York, 1987. K. Bunte. Analyses of the temporal variation of coarse bedload transport and its grain size distribution. General Technical Report RM-GTR-288, USDA Forest Service, Fort Collins, USA, 1996. R. Butler. Movement of cobbles in gravel-bed stream during a flood season. Geological Society of America Bulletin, 88:1072–1074, 1977. P.A Carling. River channels: Environment and Process, chapter Bed stability in gravel streams with reference to stream regulation and ecology., pages 321–347. Basil Blackwell Ltd, Oxford, U.K., 1987. M. Church and M.A. Hassan. Mobility of bed material in harris creek. Water Resources Research, 38(11):1237, 2002. B.R. Colby. Discharge of sands and mean-velocity relationships in sand-bed streams. U.S. Geological Survey Professional Paper, 462-A:47, 1964. M.J. Crickmore. Measurement of sand transport in rivers with special reference to tracer methods. Sedimentology, 8:175–228, 1967. 83  Bibliography S.G. Custer, K. Bunte, R. Spieker, and P. Ergenzinger. Timing and location of coarse bedload transport: Squaw creek, montana. Transactions-American Geophysical Union, 67:943–945, 1986. R.L. Dinehart. Dune migration in a steep, coarse-bedded stream. Water Resources Research, 25(5):911–923, 1989. R.L. Dinehart. Gravel-bed deposition and erosion by bedform migration observed ultrasonically during storm flow, north fork toutle river, washington. Journal of Hydrology(Amsterdam), 136(1):51–71, 1992. R. Ehrenberger. Direkte geschiebemessung an der donau bei wien und deren bisherige ergebnisse [direct bedload measurements at the danube near vienna and results to date]. Die Wasserwirtschaft, 34:1–9, 1931. P. Ergenzinger and S. Custer. First experiences measuring coarse material bedload transport with a magnetic device. Euromech: Mechanics of Sediment Transport, 156:223–227, 1982. P. Ergenzinger and SG Custer. Determination of bedload transport using naturally magnetic tracers: first experiences at squaw creek, gallatin county, montana. Water Resources Research, 19(1):187–193, 1983. D. Gintz, M.A. Hassan, and K.H. Schmidt. Frequency and magnitude of bedload transport in a mountain river. Earth Surface Processes and Landforms, 21(5): Page 95, 1996. B. Gomez. Bedload transport. Earth-Science Reviews, 31(2):89–132, 1991. G.A. Griffiths. Recent sedimentation history of the waimakariri river, new zealand. Journal of Hydrology, New Zealand, 18(1):6–28, 1979. M.A. Hassan. Scour, fill, and burial depth of coarse material in gravel bed streams. Earth Surface Processes and Landforms, 15(4):341–356, 1990. M.A. Hassan and M. Church. Sensitivity of bed load transport in harris creek: Seasonal and spatial variation over a cobble-gravel bar. Water Resources Research, 37(3):813–825, 2001. M.A. Hassan, M. Church, J. Rempel, and R.J. Enkin. Promise, performance and current limitations of a magnetic bedload movement detector. Earth Surface Processes and Landforms, 34(7):1022–1032, 2009. JA Hayward and AJ Sutherland. The torlesse stream vortex-tube sediment trap. Journal of Hydrology, New Zealand, 13(1):41–53, 1974. 84  Bibliography EJ Helley and W. Smith. Development and calibration of a pressure-difference bedload sampler. Open-file report, US Geological Survey, December 1971. A.B. Hollingshead. Sediment transport measurements in gravel river. Journal of the Hydraulics Division, 97(11):1817–1834, 1971. D.W. Hubbell. Sediment Transport in Gravel-Bed Rivers, chapter Bedload sampling and analysis, pages 89–120. Wiley, 1987. PC Klingeman and RT Milhous. Oak creek vortex bedload sampler. In 17th Annual Pacific Northwest Regional Meeting, American Geophysical Union, Tacoma, Washington, volume 52, page 17, Univ. of Puget Sound, Tacoma, Washington, October 1970. SN Lane, KS Richards, and JH Chandler. Developments in monitoring and modelling small-scale river bed topography. Earth Surface Processes and Landforms, 19(4):349–368, 1994. JB Laronne and MA Carson. Interrelationships between bed morphology and bedmaterial transport for a small, gravel-bed channel. Sedimentology, 23(1):67–85, 1976. L.B. Leopold and W.W. Emmett. 1976 bedload measurements, east fork river, wyoming. Proceedings of the National Academy of Sciences, 74(7):2644–2648, 1977. L.B. Leopold, W.W. Emmett, and R.M. Myrick. Channel and hillslope processes in a semiarid area, new mexico. US Geological Survey Proffesional Paper, 352: 193–253, 1966. F. Nesper. Ergebnisse der messungen uèber die geschiebe±und schlammfuèhrung des rheins an der brugger rheinbruècke [results of bedload and silt movement observations on the rhine at the brugg bridge]. Schweizerische Bauzeitung, 110: 143–148, 1937. S.J. O’Leary. Bedload transport in an oregon coast range stream. Water Resources Bulletin, 17(5):886–894, 1981. I.V. Popov. Application of morphological analysis to the evaluation of the general channel deformations of the river ob. Soviet Hydrology, 3:267–324, 1962. D.M. Powell and P.J. Ashworth. Spatial pattern of flow competence and bed load transport in a divided gravel bed river. Water Resources Research, 31(3):741– 752, 1995. 85  Bibliography I. Reid, JT Layman, and LE Frostick. The continuous measurement of bedload discharge. Journal of Hydraulic Research, 18(3):243–249, 1980. I. Reid, A.C. Brayshaw, and L.E. Frostick. An electromagnetic device for automatic detection of bedload motion and its field applications. Sedimentology, 31 (2):269–276, 1984. J Rempel. Calibrating and measuring bed load transport with a magnetic detection system. Master’s thesis, University of British Columbia, Vancouver, BC, 2005. D. Rickenmann. Dynamics and Geomorphology of Mountain Rivers, Lecture Notes in Earth Sciences, volume 52, chapter Bedload transport and discharge in the Erlenbach stream, pages 53–66. Springer, New York, 1994. AR Robinson. Vortex tube sand trap. Transactions of the American Society of Civil Engineers, 127:391–425, 1962. S.E. Ryan and C. A. Troendle. Measuring bedload with handheld samplers in coarse-grained mountain channels. Stream Notes, USDA Forest Service, 3:1–4, 1999. DA Sear, W. Damon, DJ Booker, and DG Anderson. A load cell based continuous recording bedload trap. Earth Surface Processes and Landforms, 25(6):659– 672, 2000. R.W. Sobocinski, T.E. Cerling, S. J. Morrison, and I.L. Larsen. Sediment transport in a small stream based on cs-137 inventories of the bed load fraction. Water Resource Research, 26(6):1177–1187, 1990. J. Tunnicliffe, A.S. Gottesfeld, and M. Mohamed. High resolution measurement of bedload transport. Hydrological Processes, 14(15):2631–2643, 2000. Peter R Wilcock. Two-fraction model of initial sediment motion in gravel-bed rivers. Science, 280(5362):410–412, April 1998. Peter R Wilcock and Joanna C Crowe. Surface-based transport model for mixedsize sediment. Journal of Hydraulic Engineering, 129(2):120–128, February 2003. Peter R. Wilcock and Brian W. McArdell. Surface-based fractional transport rates: Mobilization thresholds and partial transport of a sand-gravel sediment. Water Resources Research, 29(4):1297–1312, 1993. Peter R. Wilcock and Brian W. McArdell. Partial transport of a sand/gravel sediment. Water Resources Research, 33(1):235–245, 1997. 86  J.J. Williams, P.D. Thorne, and A.D. Heathershaw. Comparisons between acoustic measurements and predictions of the bedload transport of marine gravels. Sedimentology, 36(6):973–979, 1989. A.E. Zimmermann, M. Church, and M.A. Hassan. Video-based gravel transport measurements with a flume mounted light table. Earth Surface Processes and Landforms, 33(14):2285 – 2296, 2008.  87  Appendix A  Sieve Data  88  Appendix A. Sieve Data  Table A.1: February 4, 2010 Mixed grain size flume run sieve results  89  Appendix A. Sieve Data  Table A.2: February 11, 2010 Mixed grain size flume run sieve results  90  Appendix A. Sieve Data  Table A.3: March 12, 2010 Mixed grain size flume run sieve results  91  Appendix A. Sieve Data  Table A.4: March 9, 2010 Mixed grain size flume run sieve results  Table A.5: November 4, 2010 11-16mm single grain size flume run sieve results  92  Appendix A. Sieve Data  Table A.6: November 24, 2010 22-32mm single grain size flume run sieve results  Table A.7: December 5, 2010 32-45mm single grain size flume run sieve results  93  Appendix A. Sieve Data  Table A.8: December 12, 2010 8-11mm single grain size flume run sieve results  Table A.9: December 12, 2010 32-45mm single grain size flume run sieve results  94  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0105187/manifest

Comment

Related Items