UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Decadal-scale evolution of Elwha River downstream of Glines Canyon Dam : perspectives from numerical… De Rego, Kathryn Grace 2018

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

Item Metadata

Download

Media
24-ubc_2018_may_de_rego_kathryn.pdf [ 22.12MB ]
Metadata
JSON: 24-1.0364666.json
JSON-LD: 24-1.0364666-ld.json
RDF/XML (Pretty): 24-1.0364666-rdf.xml
RDF/JSON: 24-1.0364666-rdf.json
Turtle: 24-1.0364666-turtle.txt
N-Triples: 24-1.0364666-rdf-ntriples.txt
Original Record: 24-1.0364666-source.json
Full Text
24-1.0364666-fulltext.txt
Citation
24-1.0364666.ris

Full Text

Decadal-scale evolution of Elwha River downstream of GlinesCanyon damPerspectives from numerical modelingbyKathryn Grace De RegoMA, Geography and Mediaeval History, University of St. Andrews, 2010MSc, Earth and Quaternary Sciences, Indiana State University, 2012A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Geography)The University of British Columbia(Vancouver)March 2018c© Kathryn Grace De Rego, 2018AbstractThe goal of this study is to understand the legacy of dams on river channel evolution. Mostmajor world rivers are dammed, and these features have pervasive impacts on downstreamgeomorphology. Dam removals have become a popular restoration technique. However, lit-tle is known about how rivers respond to dam removal on long timescales, especially withregards to sediment exchanges between the channel and floodplain. We examine how damemplacement and removal have affected channel stability and migration along Elwha River, acobble-bedded wandering stream. Two dams were built on the river in the early 20th century,blocking sediment supply to the reaches below them. The dams were removed between 2011and 2014.A numerical model, MAST-1D, is adapted to simulate channel evolution on the set ofreaches between the two dams. New representations of bank erosion, vegetation encroach-ment, and avulsion are developed to make the model suitable for cobble-bedded streams. Inthe model, channel width and migration oscillate between a range of values, increasing afteravulsions due to reorganization of channel geometry. The model is successful at simulatingchannel change during the sediment-starved period following dam emplacement. While itreplicates the general pattern of channel change following dam removal, the simulations un-derestimate the competence of the system to export the initial pulse of sediment from theformer reservoir deposit. Constraining the volume and caliber of sediment supply from thereservoir is crucial for predicting sediment deposition and storage downstream.Model simulations indicate that dam emplacement results in channel armoring, which re-duces the competence of the flow to undercut bank toes, reducing the migration rate and lead-ing to net channel narrowing. Both field and model data show that activation of floodplainchannels via avulsion and, to a lesser extent, bank erosion, were responsible for increased lev-els of channel-floodplain exchange during the post-removal period. We predict that in thefuture, Elwha River will be more laterally unstable than it was in the 20th century, both due tothe legacy of the dam removal and because of climate change.iiLay SummaryThis thesis is one of many studies that seek to understand the environmental impact of dams.The sand, gravels, and cobbles that are found on the banks and bottoms of rivers move overtime, causing the river to migrate. Dams trap sediment behind them and drastically change thestructure of the channel downstream. If the dams are removed, then that sediment is suddenlyreleased, creating a risky environment for downstream communities and ecosystems.We used a numerical model to characterize how construction and removal of a dam onElwha River will affect the movement of sediment downstream. We found that the channelmigrates a lot less when the dam is place because it is coarser and harder to erode. We ex-pect the river to migrate more in the future both because sediment supply from upstream hasreturned and because of changes in flow due to climate change.iiiPrefaceMuch of the content of this thesis is based on an adaptation MAST-1D, a numerical modeldeveloped by J.W. Lauer, E. Viparelli, H. Pie`gay, and C. Li. I have made significant changesto the model, which include those described in Chapter 2 as well as algorithms to allow themodel to simulate hydrographs and other bug fixing. I also customized model parametersand inputs to Elwha River. J. Walden, under the supervision of J.W. Lauer, performed thephotosieving analysis in Chapter 3. The remainder of the thesis is my original intellectualwork.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation and research questions . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Research approach and available tools . . . . . . . . . . . . . . . . . . . . . . . . 51.3 Thesis organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 A decadal-scale numerical model for low-sinuosity, cobble-bedded rivers . . . . . 82.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.3 Model framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.3.1 General structure and model procedure . . . . . . . . . . . . . . . . . . . 122.3.2 Lateral exchange and width change . . . . . . . . . . . . . . . . . . . . . 152.3.3 Avulsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.4 Model behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.4.1 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30v3 The impact of sediment supply on channel stability along Elwha River, Washing-ton following dam emplacement and removal . . . . . . . . . . . . . . . . . . . . . . 323.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3 Study location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.4.1 MAST-1D setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.4.2 Model confirmation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.4.3 Other confirmation data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.5.1 Model performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.5.2 Channel evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533.6.1 MAST-1D confirmation–successes and failures . . . . . . . . . . . . . . . 543.6.2 Impact of sediment supply on channel stability . . . . . . . . . . . . . . 593.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614 Elwha in the 21st century: can we use the past to predict the future? . . . . . . . . . 624.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634.3 Elwha River in context: a review of the decadal-scale climatology and sedimentsupply regimes on Pacific Northwest rivers . . . . . . . . . . . . . . . . . . . . . 654.3.1 Decadal-scale climate variability in the Pacific Northwest and its effectson hydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.3.2 Sediment supply of large Pacific Northwestern rivers draining moun-tainous catchments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.4 Elwha past and future: Monte Carlo simulations of channel evolution withvarying sediment supply and climatic regimes . . . . . . . . . . . . . . . . . . . 724.4.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 895 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94Appendix A MAST-1D model description . . . . . . . . . . . . . . . . . . . . . . . . . . 108A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108A.2 Model procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110A.3 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111viA.3.1 Hydraulics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111A.3.2 Sediment transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114A.3.3 Width change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118A.3.4 Sediment reservoir exchanges . . . . . . . . . . . . . . . . . . . . . . . . . 119A.3.5 Substrate maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124A.4 Initial and boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125A.4.1 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125A.4.2 Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126A.5 Variable list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Appendix B MAST-1D parameters and calibration . . . . . . . . . . . . . . . . . . . . 132Appendix C Field data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139C.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139C.2 Subsurface bulk sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141C.3 Photosieving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145C.4 Survey data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149Appendix D Air photo information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196viiList of TablesTable 2.1 Select initial conditions for model runs . . . . . . . . . . . . . . . . . . . . . . 22Table 3.1 Boundary conditions for model runs . . . . . . . . . . . . . . . . . . . . . . . 38Table 3.2 Summary of model runs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Table 3.3 Air photos used in the analysis with available accompanying data . . . . . . 43Table 4.1 Hydrological parameters for three climatic periods. . . . . . . . . . . . . . . . 67Table 4.2 Monte Carlo simulation sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72Table 4.3 Metrics of effective discharge. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Table A.1 MAST-1D list of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Table B.1 Initial conditions for simulations from Chapter 2 . . . . . . . . . . . . . . . . 133Table B.2 Initial conditions for simulations from Chapter 3 . . . . . . . . . . . . . . . . 133Table B.3 Initial conditions for simulations from Chapter 4 . . . . . . . . . . . . . . . . 135Table B.4 Floodplain and substrate initial conditions . . . . . . . . . . . . . . . . . . . . 135Table B.5 Calibration parameters and other constants . . . . . . . . . . . . . . . . . . . 136Table C.1 Bulk samples on point bar heads . . . . . . . . . . . . . . . . . . . . . . . . . . 142Table C.2 Bulk samples on cutbank toes and collapse deposits . . . . . . . . . . . . . . 143Table C.3 Bulk samples on other surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . 144Table C.4 Photosieved grainsize data for point bar heads taken from photos . . . . . . 145Table C.5 Photosieved grainsize data for bank toes taken from photos . . . . . . . . . . 147Table C.6 Photosieved grainsize data corresponding to bulk sample locations . . . . . 149Table C.7 Survey data key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150Table C.8 Survey data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151Table D.1 Air photos with available accompanying data . . . . . . . . . . . . . . . . . . 197viiiList of FiguresFigure 1.1 Conceptual diagram of geomorphic processes and appropriate numericalmodels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Figure 2.1 Gravel-cobble cutbank along Elwha River, Washington, USA . . . . . . . . . 12Figure 2.2 Schematic showing MAST-1D reservoirs and exchanges . . . . . . . . . . . 13Figure 2.3 Schematic showing bank erosion and encroachment . . . . . . . . . . . . . . 14Figure 2.4 Conceptual model of bank erosion along a coarse, cohesionless bank . . . . 16Figure 2.5 Example of a local avulsion on Elwha River, Washington . . . . . . . . . . . 19Figure 2.6 Schematic of the MAST-1D avulsion algorithm . . . . . . . . . . . . . . . . . 21Figure 2.7 Sample of repeating hydrograph used in model runs . . . . . . . . . . . . . 22Figure 2.8 Channel characteristics over time . . . . . . . . . . . . . . . . . . . . . . . . . 23Figure 2.9 Rates of bank erosion and vegetation encroachment . . . . . . . . . . . . . . 24Figure 2.10 Evolution of channel width over time for different intial widths . . . . . . . 25Figure 2.11 Channel width for runs with and without an upstream supply of sediment 25Figure 2.12 Effect of flow sequencing on channel width . . . . . . . . . . . . . . . . . . . 26Figure 2.13 Sensitivity of channel width to model parameters . . . . . . . . . . . . . . . 27Figure 2.14 Sensitivity of migration rate to model parameters . . . . . . . . . . . . . . . 28Figure 3.1 Map of Elwha River basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Figure 3.2 Longitudinal profile of Elwha River and the study area . . . . . . . . . . . . 36Figure 3.3 Composite subsurface grain size distributions of Elwha River cutbank andpoint bar head deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 3.4 Channel width measured from air photos and calculated in MAST-1D . . . 45Figure 3.5 Channel outlines for Elwha River before and after dam removal . . . . . . . 47Figure 3.6 New channel and floodplain creation in MAST-1D vs. air photos . . . . . . 48Figure 3.7 Comparison of modeled D50 with field data . . . . . . . . . . . . . . . . . . . 49Figure 3.8 Total (channel and floodplain) storage within the study area . . . . . . . . . 49Figure 3.9 Evolution of channel characteristics over time . . . . . . . . . . . . . . . . . 51Figure 3.10 Average annual bank erosion and encroachment rates for simulations withand without dam emplacement/removal . . . . . . . . . . . . . . . . . . . . 51Figure 3.11 Total simulated bank erosion compared to annual peak daily discharge . . 52ixFigure 3.12 Temporally-averaged rates of new channel formation . . . . . . . . . . . . . 53Figure 3.13 Annual rates of channel widening and narrowing plotted against the lengthof time between air photo pairs . . . . . . . . . . . . . . . . . . . . . . . . . . 55Figure 3.14 Residuals of modeled channel widening and narrowing rates . . . . . . . . 55Figure 3.15 Impact of sediment supply caliber on storage within the study area . . . . . 58Figure 4.1 Metrics of hydrologic change . . . . . . . . . . . . . . . . . . . . . . . . . . . 68Figure 4.2 Monthly mean daily discharge for three climatic periods . . . . . . . . . . . 69Figure 4.3 Flood frequency plot for the three hydrologic periods . . . . . . . . . . . . . 69Figure 4.4 Sediment supply boundary conditions for the Monte Carlo simulations . . 73Figure 4.5 Schematic showing the migration rate in relation to channel widening andnarrowing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75Figure 4.6 Evolution over time of the channel and floodplain for runs S1-Pall and S2-Pall 77Figure 4.7 Correlation between floodplain D50, channel width (Bc), and high flow events 78Figure 4.8 Geomorphic metrics as a function of sediment supply and hydrologic regime 79Figure 4.9 Examples of effective discharge plots with bimodal distributions . . . . . . 80Figure 4.10 Frequency that flow exceeds effective discharge metrics . . . . . . . . . . . . 81Figure 4.11 Response time to changes in hydrologic regime . . . . . . . . . . . . . . . . 83Figure 4.12 Deposition of fine material on new point bar surfaces . . . . . . . . . . . . . 85Figure 4.13 Average daily sediment transport by month for sediment supply scenarioS3, divided into phases of transport described by Carling (1988) . . . . . . . 87Figure 4.14 The ratio of average annual sediment transport to channel migration as afunction of the average nival discharge . . . . . . . . . . . . . . . . . . . . . 90Figure A.1 ModelScheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Figure B.1 Map with node locations for simulations in Chapter 3 . . . . . . . . . . . . . 134Figure B.2 Calibration of Bc for nodes with a slope of 0.0069 in Chapters 2 and 3 . . . . 135Figure B.3 Calibration of Bc for nodes with a slope of 0.0081 in Chapter 3 . . . . . . . . 137Figure B.4 Calibration of Bc for nodes with a slope of 0.0074 in Chapter 4 . . . . . . . . 138Figure B.5 Calibration of C for dam removal simulations . . . . . . . . . . . . . . . . . 138Figure C.1 Map with sample locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140xAcknowledgmentsThe Renaissance scientist Sir Isaac Newton famously wrote that ‘if [he has] seen further, it is bystanding upon the shoulders of giants.’ My own foray into science has conjured an image farless majestic. Doing a PhD sometimes felt like trying to balance on the tip of a pyramid in themiddle of an earthquake. Despite two decades of schooling, I was not prepared for how hardit would be to do independent research. Even though I moved around a lot before Vancouver,I underestimated the difficulty of transitioning to life in a big city. But my pyramid is strongand sturdy, and I was able to stand in the end.My advisors and committee formed the top of the pyramid, providing the core knowledgeneeded to complete this PhD and also creating a sound surface to stand on. Marwan Hassanwas my primary source of guidance; he made sure that I was on the right track and steeredme back when I went astray. If it weren’t for his vision and encouragement at crucial stages,I don’t think I would’ve ever finished. Brett Eaton was the ‘voice of reason.’ He has a talentfor explaining concepts in terms that make sense, and the best advice I’ve received–both withregards to my project and for graduate school in general–came from Brett. Wes Lauer wentabove and beyond the call of committee member. I’m so grateful for his patience and theamount of effort he put into my programme: explaining MAST-1D and helping generate ideas,introducing me to other researchers, and imparting incredible foresight. In 2015, Wes joinedme in the field and made me collect a bunch of samples I didn’t think I had time for. Everysingle one of them proved crucial when I wrote up this thesis.A pyramid isn’t very tall without a ton of bricks, and the scope of my knowledge would beshort indeed without the support and inspiration from my many labmates. Our conversationsin lab meetings really helped shape my ideas. More importantly, the long PhD journey wasenjoyable because I shared it with such kind-hearted and insightful ‘travel companions.’ I’lladmit I won’t entirely miss getting my beer served in mason jars, but I will fondly remembersharing stories and laughing over the absurdities of academia. Special shout-outs go to CarlesFerrer-Boix and Shawn Chartrand, who were especially generous with their expertise overthe years, and to Leo King, who was an encouraging and empathetic office-mate and friend,particularly during the dreaded ‘final stage.’My field assistants quite literally dug up the stone for this metaphorical pyramid. It turnsout that doing 1000 kg bulk samples on swift, cobble-bedded rivers isn’t easy, but Rose Bea-gley, Aron Zahradka, and Lawrence Bird made it doable with their outdoor expertise, drive,xiand sheer muscle. Jane Walden was my right-hand woman during the 2015 field season andbeyond. Her enthusiasm, outdoor prowess, dependability, and friendship made that fiveweeks on the river both successful and incredibly fun. And not all digging requires shovelsand buckets. A huge part of research involves finding computer bugs and making technologywork efficiently. Vincent Kujala responded to all my computer woes with patience and deter-mination, and I would’ve never finished my model runs without his bottomless database ofcomputer tips and tricks. He was never able to train me to do updates regularly, but everytime I see the update pop-up on my personal computer, I feel the appropriate level of guilt.As a ‘professional student,’ I spent the entirety of my 20s at university. In that decade,I made very little money but struck the jackpot on wonderful friends. They have been themortar of my grad school life, holding it together and giving it shape and consistency as Igrew. When I was running into walls with code, it was nice to be able to vent my frustra-tions as Badger the reckless gnome druid. Even in the most challenging times, catching upwith friends–whether in person or by phone or Skype or letter–gave me an immense senseof joy and contentment. Seeing friendships grow stronger despite the obstacles of time anddistance has been one of the most rewarding experiences of adulthood, and the support andencouragement I’ve received has made me a better scientist, and, more importantly, a betterperson.I am also incredibly fortunate to have inherited a large, loving family. The older I get, themore I value the unconditional love I’ve received from both sides of the family. Grandma Ivy,Uncle David, and Auntie Danette were integral to the success of my time in Vancouver; theyprovided me with a legal address, car help, unlimited portions of delicious food, and a warmand loving place to escape when I was feeling homesick. It’s not always easy living far fromfamily, but Tara, Steve, Can Can, Cha Cha, and Ping Pong have given me a better home than Icould have ever asked for.I’ve also been blessed with a multitude of wonderful Christophers. Christopher Quick wasmy first friend in Vancouver and my closest; he has been loyal, understanding, and support-ive through good times and bad. Christopher De Rego excelled in his role as the practicaltwin and gave me good doses of common sense during some pivotal moments. ChristopherSemiao entered my life somewhat late into my PhD journey. He has become the skeleton ofmy pyramid–the steel rebar put in during renovations to make it earthquake-proof. Whenmy model crashed for the 100th time, Chris S. drove across the city at midnight to give me ashoulder to cry on. Then he did the same thing when it crashed for the 101st time. And the102nd. I saw the light at the end of the tunnel because Chris was holding a lantern there. Hiscontributions are endless, but he wants to be especially recognized for some chauffeur servicesthat made the timely completion of this Acknowledgements section possible.One time in middle school, I was driving home from the commissary with my dad in hisold blue Ford Ranger. He told me I would be the more difficult teenager. He wasn’t entirelywrong. My parents raised my brother and me to to think independently and to strive to con-xiitribute to society in great ways. I repaid them by moving as far away as possible, developinga worldview completely different from theirs, and choosing a financially unstable lifestyle.Yet despite everything, Mom and Dad have been my greatest cheerleaders. Their ‘I love you’is what formed the base of my pyramid, and the values they taught me have proven moreimportant than anything else I’ve ever learned.If you take the time to read this thesis, you’ll notice that I use the word ‘we,’ even thoughI am the sole listed author. Science written in the third person is soulless, and using ‘I’ feltweird and inaccurate. They say it takes a village; thanks for being mine.xiiiDedicationTo my family–past, present, and future.xivChapter 1Introduction1.1 Motivation and research questionsEarth’s surface is a dynamic composite of erosional and depositional landforms. The greatdiversity of our landscapes are built on this canvas; sediments and sediment movement formthe foundation for our ecosystems and the structure over which we have built our civilizations.Rivers and floodplains in particular have been instrumental to human development as theycreate fertile agricultural land and convenient means of transport, trade, and energy. Riversprovide us with aquatic resources such as fish, and hydropower supplies electricity to millionsof people. However, they are also dangerous; flooding is frequent and destructive, and erosionputs structures at risk. Since we live so intimately with alluvial landscapes, an understandingof how sediment is routed through them is crucial to developing safe, lasting infrastructurewhile protecting our precious natural resources.Church (2002) notes that the transfer of sediment through the landscape on geomorphicallysignificant timescales is characterized by the dynamics of sediment storage. The ‘sedimentcascade,’ or movement of material through storage reservoirs, is responsible for the spatialand temporal distribution of sediment transport through drainage basins. Sediment tends totravel in a diffusive nature, both because of stochastic variability in particle velocity (Martinet al., 2012) and because of differences in transportability related to size (Church, 2002). Riversact as the ‘plumbers’ of drainage basins in that they route material in source areas, filtering iton the way through selective transport (Gomez et al., 2001; Hoey and Ferguson, 1994), depositionand re-entrainment in temporary reservoirs such as point bars and floodplains (Church andSlaymaker, 1989), and physical and chemical weathering (Heller et al., 2001; Sklar et al., 2006).The kinetic energy required to transport sediment is provided by the flow of water, which isfundamentally dependent on climate but also modulated by land cover, elevation, and localgeology. Together, the hydrologic and sediment regimes determine the shape and size of rivers(Lane, 1955; Leopold and Maddock, 1953; Eaton et al., 2004) and the rate at which they routesediment between reservoirs (Constantine et al., 2014).11.1. Motivation and research questionsHumans have been modifying the sediment cascade for millennia and are now arguablythe greatest geomorphic agents on Earth (Hooke, 1994, 2000). We have, for example, alteredflow routing and channel geometry by building canals and drainage systems (Hooke, 2000;Lauer et al., 2017), changed the frequency of flooding by modifying the permeability of landcover (e.g. Bledsoe and Watson, 2001), increased the supply of sediment to large floodplainsvia agricultural soil erosion (Hassan et al., 2017), and reduced channel-floodplain connectivitythrough bank protection, levees, and channelization. But one of the greatest human impactson the landscape has been the construction of dams. Streams have been dammed for hundredsof years to store water and generate power. During the early 20th century, dam building onlarge rivers became common. In his census of dams in the United States, Graf (1999) foundthat every major river contains at least one, and that they store about 5000 m3 of water perAmerican. These dams have left a pervasive imprint on the sediment cascade; they oftenalter the sediment transport regime by cutting off supply, reducing peak flows and increasingbase flow (Andrews, 1986; Graf , 2006; Magilligan and Nislow, 2005). Reducing the sedimentsupply to a river causes channel stabilization, which lowers rates of flooding and migrationand decreases biodiversity (James and Singer, 2008; Gregory and Park, 1974; Kloehn et al., 2008;Konrad et al., 2011; Poff et al., 2007; Williams and Wolman, 1984). The rate and timing of dambuilding has varied on a global scale. Many developing countries are experiencing eras oflarge dam construction, with massive projects planned that will impact some of the world’slargest rivers (Nones et al., 2013; Rubin et al., 2015). In the US, however, dam constructionpeaked between the 1950s to 1970s, and has been in decline since (Graf , 1999). In fact, many ofthe dams constructed during the first half of the 20th century have surpassed their design livesand pose risks to nearby communities. The need to manage aging structures, combined withincreasing concern about endangered species and the desire to restore ‘natural’ conditions tobasins, has made dam removal an increasingly popular river restoration technique (Grant,2001). To date, over 1000 dams have been removed in the United States, most within the pastdecade, and the trend is expected to continue for the foreseeable future (O’Connor et al., 2015).Most dams disrupt the sediment cascade by trapping incoming sediment loads in the reser-voirs behind them. When the dams are removed, that sediment is released to downstreamreaches, often catastrophically. Few dam removals have been studied scientifically, so ourknowledge of the evolution of rivers and floodplains after the sediment ‘faucets’ have beenturned back on is based on a handful of recent studies (e.g. Burroughs et al., 2009; East et al.,2015; Major et al., 2012; Warrick et al., 2015; Wilcox et al., 2014). It appears that channels arevery efficient in transporting fine sediment, with most of it leaving the basin regardless of theflow regime during removal (East et al., 2015; Grant and Lewis, 2015; Major et al., 2012). Less isknown about bed material, which is important for determining channel shape and morphol-ogy. So far, it seems as though most coarse material is deposited immediately downstream ofthe dam and has little morphologic impact farther downstream (Grant and Lewis, 2015; Majoret al., 2012), although it is not clear how generalizable this finding is, particularly with regards21.1. Motivation and research questionsto larger rivers with active floodplains. Most available data is collected on rivers within thefirst few months to years after the removal. While we have gained valuable insights into howlarge pulses of sediment affect channel morphology, virtually nothing is known about howthese streams will respond on decadal timescales as the formerly sediment-starved channelsadjust to transport higher loads. Potential responses include greater flood risk and more fre-quent channel instability. The working hypothesis implicit in many dam removal projects isthat the channel returns to its pre-dam state quickly. However, conceptual models of riverresponse to dam removals include channel widening and an increase in migration rate, andvery few studies have quantified these effects on real systems (Burroughs et al., 2009; Doyleet al., 2002; Major et al., 2012).The goal of this study is to assess the decadal-scale impact of dam emplacement and re-moval on the sediment cascade for Elwha River. The Elwha is a steep, cobble-bedded river onthe north end of the Olympic Peninsula in Washington, USA. Like many rivers in the PacificNorthwest, it was dammed in the early 20th century to supply hydropower in the wake ofgrowing industrial demand. Mapes (2013) has recounted a history of the power projects andtheir role in the success of the nearby town of Port Angeles. Construction of the 32 meterhigh Elwha Dam was initiated by Thomas Aldwell, who saw the steep and powerful riveras a profitible resource for hydropower. The dam was completed in 1913, and, as expected,soon brought business to Port Angeles, elevating it from a sleepy pioneer town to a bustlingindustrial port. The Puget Sound Mills and Timber Company arrived in 1914, and was fol-lowed by the Washington Pulp and Paper Company. Demand from the paper industry led tothe construction of the 64 m high Glines Canyon Dam 14 km upstream from Elwha Dam in1927. Business waned in the mid-20th century, and by 1949, most remaining customers usedthe regional power grid and only the paper mill still extracted electricity from the two dams.Because of a deal with the state fish commissioner, Thomas Aldwell was exempted frombuilding a fish passage structure on Elwha Dam, which was (and still is) a legal requirementfor any dam in the state of Washington. Prior to dam construction, the river provided habi-tat for all five species of Pacific salmon as well as trout, eulachon, and lamprey. The damblocked 113 km of habitat, and the absence of anadromous fish led to nutrient deficiencies inthe Elwha catchment (Munn et al., 1999). By the late 20th century, the population of salmondownstream of Elwha Dam was only about 1% of that in the early 20th century (Departmentof the Interior, 1995; Duda et al., 2011b). The history of dam decommissioning is recounted indetail by Mapes (2013) and is briefly summarized here. Efforts to remove the dams began in1986, when the Lower Elwha Klallam Tribe and environmental organizations both petitionedthe Federal Energy Regulatory Commission (FERC) for dam removal, noting that since GlinesCanyon Dam lay within Olympic National Park, it could not legally be relicensed. The ideainitially recieved little support from the federal government. However, following passage ofthe Electric Consumers Protection Act (also in 1986), environmental regulations became morestringent on hydropower projects. It became cheaper to remove Glines Canyon and Elwha31.1. Motivation and research questionsDams than to update them to reflect the new standards. The bill calling for removal of the twodams was passed in 1992. After several years of political turmoil, funding for the removalswas negotiated, and demolition began in September 2011.The Elwha River project included funding for scientific research before and in the fewyears following dam removal in order to assess its impact on wildlife biology (e.g. Pess et al.,2008; Jenkins et al., 2015), riparian and coastal ecology (e.g. Duda et al., 2011a; Foley et al., 2015;Morley et al., 2008), and geomorphology (e.g. Draut et al., 2011; Draut and Ritchie, 2015; Eastet al., 2015; Magirl et al., 2015). Most of the latter has been focused on documenting the fateand impact of the initial sediment pulse that was released downstream of Elwha and GlinesCanyon Dams in 2011 and 2012. At this early stage in Elwha River’s recovery, there has onlybeen brief speculative focus on the decadal-scale processes, including long-term migrationrates, channel-floodplain connectivity, and the sensitivity of these processes to climatically-driven hydrologic variability.This thesis is centered around the fundamental question:How has dam emplacement and removal impacted Elwha River geomorphology betweenthe former Glines Canyon and Elwha Dam sites on decadal timescales?We are particularly interested in the impact of the sediment supply disturbances on processesof channel-floodplain coupling, which include migration, width change, avulsion, and thecompetence of the channel to evacuate sediment both from the former reservoir deposits andfrom its banks. This leads to a number of sub-questions:1. How do steep, coarse, cobble-bedded rivers like the Elwha behave on decadal timescalesunder steady sediment and discharge regimes?2. What processes lead to channel width change, and how are they impacted by channelimpoundment?3. What are the main processes that contributed to geomorphic change during periods ofsediment starvation and sediment excess? Can the same processes explain channel evolutionduring both sediment supply scenarios?4. What is the long-term legacy of the dams on Elwha River, and how long will the effects last?5. What impact, if any, does decadal-scale climate variability have on channel evolution, andis the effect different before and after dam removal?41.2. Research approach and available toolsFigure 1.1: Conceptual diagram of geomorphic processes and appropriate numericalmodels. Based on a similar conceptualization by Church (2008).1.2 Research approach and available toolsTo answer the research questions listed above, we use a numerical modeling approach. Nu-merical modeling is useful for a number of reasons. When properly calibrated and verified, itis useful for projecting past and future periods for which there is little to no field data. By com-paring calculated output to field data, we are able to test whether certain model assumptionsappear to be valid, and this can provide information on which processes are most importantto the system. Finally, numerical models are valuable tools for hypothesis testing and can bean efficient way of generating ideas and prioritizing field observation.All models are approximations of reality. It is important to select numerical models thathave suitable underlying assumptions for the system in question and which consider the ap-propriate spatial and temporal scales. A schematic showing the range of geomorphic features,51.3. Thesis organizationand corresponding models, relevant on common spatial and temporal scales is presented inFigure 1.1. In this study, we are most concerned with processes on the multi-reach scale (mul-tiple kms and through sections of river that may behave in unique ways) and timescales rang-ing from about 5 to 150 years. Two and three-dimensional deterministic models offer detailedrepresentations of reach-scale landforms, and many incorporate channel-floodplain coupling(Darby and Van de Wiel, 2003; Nelson et al., 2003). However, they are too computationally ex-pensive to run for decadal timescales without making assumptions about the discharge regimethat are overly simplistic for our research questions. Over long timescales, it is common to useeither regime or landscape evolution models. The former attempt to predict channel dimen-sions using a single ‘channel-forming’ flow and steady state sediment flux and caliber (e.g.Lane, 1955; Eaton et al., 2004). Since dam emplacement and removal involves changes to thegoverning sediment regime, these models are inappropriate for our purposes. Landscape evo-lution models have been used to interpret large-scale disturbances, but they usually rely onregime relations when predicting lateral change (Tucker and Hancock, 2010).In practice, most numerical modeling projects on multi-reach, decadal timescales involvethe use of 1-dimensional models (Lauer et al., 2016). This presents us with a problem: most 1-dimensional models neglect either lateral or longitudinal fluxes (e.g. Czuba et al., 2012; Fergusonand Church, 2009; Konrad, 2012), but, as noted above, both are important in the context of damemplacement and removal. This issue has not gone unnoticed; Doyle et al. (2002) remarkedupon how changes in channel width are incorporated into conceptual models of dam removal,but most analysts use models that assume fixed banks when assessing system response. Whilesome 1-D models are attempting to incorporate both longitudinal and lateral processes (e.g.Parker et al., 2011; Eke et al., 2014), they are designed for fine-grained lowland systems with lowgradients that function very differently to the high-energy systems in the Pacific Northwest.Therefore, an additional goal of this project is to develop a decadal-scale numerical model thatis appropriate for steep, cobble-bedded rivers and their lateral and longitudinal complexity.1.3 Thesis organizationThe thesis is organized into four parts. In Chapter 2, we explain in more detail why appropri-ate decadal-scale numerical models are lacking for coarse-bedded, steep rivers like the Elwha.A 1-dimensional model is proposed that incorporates both longitudinal and lateral change.There are two novel components. The first is that we include algorithms for bank erosion andvegetation growth that together allow the channel to exchange sediment with the floodplainwhile adjusting its width and capacity. The bank function is based on theories of bank stabil-ity for systems with floodplains composed of primarily gravel and cobble-sized material. Asimple function representing avulsion, based on the accumulation of excess sediment in thechannel, is also added to the model to simulate channel-floodplain coupling resulting fromthe abandonment and activation of channels. These additions represent what we hypothesizeare integral processes operating on the coarse, steep, and active rivers of the Pacific North-61.3. Thesis organizationwest. We use the model to qualify how these systems behave under very simple governinghydrologic and sediment supply regimes.Chapter 3 is devoted to confirmation of the model. We simulate evolution of Elwha Riverbetween 1918 and 2016, which includes both emplacement and removal of Glines CanyonDam. Results from the model are compared with a suite of field and remotely sensed data.By noting where the it was successful and where it failed, we are able to test the whether themodel assumptions are valid for periods of sediment starvation and excess. Comparing andcontrasting simulations with and without representation of the dams sheds light on the impactof sediment supply disturbances on channel stability.The long-term evolution of Elwha River is explored in Chapter 4. The goal of this sectionis to characterize how rates of sediment transport and channel-floodplain coupling may af-fect the long-term recovery to pre-dam levels. We consider two sources of disturbance: thedecadal-scale legacy of the dams and climate variability. In order to quantify the range of vari-ability in model response, a Monte-Carlo approach is used, with discharge treated stochasti-cally. We examine metrics of sediment transport and channel stability for a variety of sedi-ment supply scenarios, including one representing Elwha’s history of damming. Our analy-sis shows the importance of hydrology in modulating how river channels interact with theirfloodplain reservoirs and can be used to guide future field data collection.Finally, in the Conclusion (Chapter 5), we discuss how decadal-scale modeling has allowedus to better understand the potental legacy of dam building and removal on the sedimentcascade. Remaining gaps are identified and future research directions proposed. This workincreases understanding of sediment movement through human landscapes, but there is stillso much to learn.7Chapter 2A decadal-scale numerical model forlow-sinuosity, cobble-bedded rivers2.1 SummaryEven though width change is one of the most important responses of rivers to changes ingoverning conditions, many numerical models assume that banks are immovable. We haveadapted MAST-1D, a reach-scale numerical model, to simulate the relevant decadal-scale pro-cesses for coarse (gravel-cobble bedded) multithreaded rivers. The model has separate func-tions for bank erosion and vegetation encroachment, allowing for width change. Bank erosionis a function of the mobility and transport capacity for large, structurally-important grainswhich protect the bank toe. Vegetation growth is linearly proportional to channel width andoccurs during conditions of low shear stress. In addition, MAST-1D simulates local, reach-scale avulsions, which occur when aggradation causes channel depth to drop below a thresh-old. The behavior of MAST-1D was assessed using simple boundary conditions. When theannual hydrograph and sediment supply regime are kept constant, the channel width, migra-tion rate, and sediment transport rate oscillate on decadal timescales. The time period betweenoscillations is dependent on the frequency of local avulsions, which are most sensitive to sed-iment supply and the size of coarse particles. Our simulations suggest that internal decadalscale variability is an inherent feature of coarse, wandering rivers and that it is closely coupledwith reach-scale sediment storage and evacuation. Traditional regime approaches to charac-terizing channel dimensions may be too simplistic for alluvial systems where avulsion is anatural, frequent process.2.2 IntroductionIt is a well-established principle that alluvial channels are composed of self-formed boundariesthat evolve in response to governing flow and sediment supply regimes. In their pioneeringwork, Leopold and Maddock (1953) introduced the idea that river dimensions are correlated with82.2. Introductiona single ‘formative’ discharge, which they defined as the bankfull discharge but which hasalso been associated with the flow responsible for most sediment transport (Andrews, 1980;Wolman and Miller, 1960). Hydraulic geometry has persisted as one of the key concepts ingeomorphology and still influences the way in which many researchers approach channelwidth. It is an approximation of reality; channel width fluctuates through time, respondingto the sequencing of flood events and natural variability in sediment supply (e.g. Baker, 1977;Lisle, 1982). This is especially true for gravel-bedded rivers with a wandering morphology,which are multi-threaded and often characterized by flashy flood regimes. The time it takesfor width to recover between floods is dependent on the rate at which vegetation colonizeschannel bars, which depends on climate and flood frequency (Wolman and Gerson, 1978). Widthchange can affect sediment transport and bed evolution in a variety of ways. Widening resultsin a higher flood conveyance, reducing the shear stress for a given flow. However, it also leadsto bank erosion and a wider zone over which sediment is mobile, increasing supply. The rateof channel widening is related to channel migration and the frequency of floodplain turnover,which have implications for riparian ecology (Kloehn et al., 2008; Konrad, 2012; O’Connor et al.,2003). Despite the importance of these processes in the context of bank erosion and floodrisk, many modeling projects on multi-reach, decadal scales either neglect width change (e.g.Czuba et al., 2012; Ferguson and Church, 2009; Gomez et al., 2009; Verhaar et al., 2008) or assume aconstant width-discharge relation (Viparelli et al., 2011).The processes that maintain active channel width in alluvial, wandering rivers can be sum-marized by three factors: 1) the ability of the flow to scour banks and bars, 2) activation or re-activation of floodplain surfaces via avulsion, and 3) the rate of vegetation encroachment ontobare sediment surfaces. Traditionally, the net effect of these factors have been explained byregime models. Both empirical (e.g. Leopold and Maddock, 1953) and analytical (e.g. Millar andQuick, 1993; Eaton et al., 2004) approaches have successfully predicted channel dimensions onthe reach scale in terms of a set of governing conditions such as a channel-forming discharge,bank strength, and sediment load. The primary caveat of these models is that they assume thechannel is in equilibrium with the boundary conditions. As such, they are inappropriate toolsfor use on rivers that are actively adjusting to a disturbance (such as a large flood or change tosediment supply), and they do not consider the propagation of the disturbance downstream.Numerical models offer the opportunity to assess both the spatial and temporal responsesof channels to changes in the boundary conditions. Models featuring width change began ap-pearing in the early 1990s and focused on the fluvial-driven processes leading to mass failureof banks composed of uniform sediment size (Darby and Thorne, 1996; Mosselman, 1992; Pizzuto,1990). In 1993, a task force was set up by the American Society of Civil Engineers to summarizewidth change modeling efforts and identify areas of improvement (ASCE Task Committee onHydraulics and of River Width Adjustment, 1998a,b). Since then, models have grown to incor-porate a wider and more realistic set of bank erosion processes, including sediment transportof bank material (Carroll et al., 2004; Rinaldi et al., 2008), the effect of groundwater fluxes in92.2. Introductionreducing bank stability (Higson and Singer, 2015; Rinaldi et al., 2008), bank protection from co-hesive slumps of floodplain material (Parker et al., 2011), and channel widening via incision(Cantelli et al., 2007; Cui et al., 2006). Many of these advances have been aided in large part byimprovements in 2D modeling (Darby and Van de Wiel, 2003; Nelson et al., 2003).Numerical models addressing channel avulsion fall into two general types–1-3D morpho-dynamic models of channel bifurcations and cellular-based models of entire channel-floodplainnetworks (for review, see Hajek and Wolinsky, 2012). The former focus on the geomorphic con-ditions that lead to preferential flow and sediment transport in different distributary channels,including local variations in slope, complex flow at island heads, and upstream meanderingand bar deposition (Kleinhans et al., 2008; Miori et al., 2006). To date, most assume that channelwidth in each distributary remains constant, although Miori et al. (2006) used regime relationsto characterize width change as a result of channel shifting. Cellular models (e.g. Jerolmack andPaola, 2007) have qualitatively and quantitatively reproduced long-term avulsion behavior ob-served in the field, including so-called ‘local’ avulsion, where a segment of channel forges anew path, then rejoins the existing channel a short distance downstream.Understanding the influence of vegetation on fluvial processes has significantly advancedour knowledge of bank and bar erosion (Abernethy and Rutherfurd, 2001; Beechie et al., 2006;Fetherston et al., 1995; Micheli and Kirchner, 2002; Perona et al., 2012), deposition (Friedman et al.,1996), and channel morphology (Gran and Paola, 2001; Tsujimoto, 1999). There has been a flurryof research on the interactions between flow, sediment transport, and riparian ecosystems (seereviews by Camporeale et al., 2013; Corenblit et al., 2007; Greet et al., 2011; Osterkamp and Hupp,2010). Particular emphasis has been put on numerically modeling the influence of hydrologicprocesses on riparian growth and succession (e.g. Camporeale and Ridolfi, 2007; Tealdi et al.,2011). However, the sediment transport and bed evolution models that take vegetation intoaccount (e.g. Tsujimoto, 1999; Van De Wiel and Darby, 2004) generally only consider what Cam-poreale et al. (2013) refers to as its ‘passive’ role: in other words, they simulate the effect ofvegetation on flow resistance and deposition, but do not account for vegetation encroach-ment onto bare sediment surfaces. This is a major impediment for modeling channels longerthan a couple years, after which vegetation growth becomes a major factor in width change(O’Connor et al., 2003; Williams and Wolman, 1984). In addition, they neglect the transport anddeposition of wood, which is one of the main drivers of channel morphology (Bertoldi et al.,2009; Fetherston et al., 1995; O’Connor et al., 2003) on rivers in forested landscapes.Despite improvements in our ability to quantify bank processes, few existing models areapplicable on the annual-decadal, multi-reach scales relevant to many contemporary man-agement issues. Two- and three-dimensional models are becoming increasingly successfulin capturing river erosion and deposition, but most do not consider vegetation growth andare currently limited to rather small spatial and temporal scales. Some numerical models havebeen applied to decadal and longer timescales (e.g. Eke et al., 2014; Higson and Singer, 2015), butonly on the scale of a single bend. Tunnicliffe and Church (2015) developed a 1-D model for the102.3. Model frameworkmulti-reach scale, but their width change function relies on an optimality criterion, which isdesigned to characterize channels in equilibrium. Landscape evolution and/or cellular mod-els can be applicable to decadal timescales, but they generally either assume a constant widthfor the active channel belt or rely on hydraulic geometry and regime equations that are notappropriate for rivers adjusting to changing governing conditions such as sediment supply orhydrologic regime (Jerolmack and Paola, 2007; Tucker and Hancock, 2010). It is therefore unsur-prising that so many projects use one-dimensional, fixed bank numerical models to simulatesystems that are expected to experience width change.We aim to fill this gap by adapting MAST-1D, a 1-dimensional, reach-scale bed evolutionmodel, to allow for channel width change and local channel avulsion. Bank erosion and veg-etation encroachment are modeled as two separate processes, following the approach used inthe model by Parker et al. (2011) and its decendents. Our version of MAST-1D is designed forlow-sinuosity, gravel-cobble bedded rivers where cohesive material constitutes a negligibleportion of the bank and the channel belt is dominated by small, local avulsions rather thanmeander extension. Channel stability is dictated by the mobility of the largest size fractionsin the alluvial deposits. Our model calculatates size-specific sediment transport, allowing forbed elevation and grainsize to evolve. It is therefore an appropriate tool for exploring graveland cobble-bed systems undergoing changes to sediment supply or hydrology. In this chapter,we describe the model and its behavior. In Chapter 3, the model is applied to Elwha River,where two dams were emplaced and removed a century later, to examine the effect of sedi-ment supply on channel evolution.2.3 Model frameworkMAST-1D is a one-dimensional bed evolution model designed to simulate channel and flood-plain exchange over decadal and longer timescales. It is unique in that it allows for the size-specific exchange of sediment between the floodplain and the channel through both channelmigration and overbank flooding. Details on the original version of MAST-1D can be found inLauer et al. (2016). We have modified MAST-1D to better represent coarse, wandering rivers.Our model is limited to systems where the floodplain material is composed primarily of graveland cobble, there is negligible cohesive material in the cutbank (Figure 2.1), and channel sinu-osity is low.Three primary changes have been made to MAST-1D: 1) hydraulics are calculated witha daily discharge series instead of a flow duration curve; 2) lateral channel change is repre-sented by two processes–channel widening and narrowing–so that the width is allowed tochange over time; and 3) a simple avulsion function acts as an additional source of channel-floodplain exchange. We have summarized the basic model procedure and provided detailson the channel width change and avulsion procedures below. A full model description can bein Appendix A.112.3. Model frameworkFigure 2.1: Gravel-cobble cutbank along Elwha River, Washington, USA. the bank is com-posed entirely of non-cohesive material and is protected by an armored toe deposit.2.3.1 General structure and model procedureIn MAST-1D, the river is conceptualized as a linear series of model cells, each of which con-tains a set of reservoirs: a channel bed (active layer), point bar, floodplain, and channel andfloodplain substrates (Figure 2.2). Each has a characteristic geometry, volume, and grain sizedistribution, which is modified as sediment is exchanged with other reservoirs due to trans-port, channel widening or narrowing, and avulsion. Like other 1D bed evolution models, theoutgoing sediment load from an upstream model cell becomes the supply to the downstreamnode.Each model cell is long enough to incorporate a reach-sized portion of channel. In otherwords, the sediment exchanges between reservoirs represent an average over multiple bendsand bar sequences. The channel shape is assumed to be rectangular. This neglects the impor-tant influence of channel morphology on hydraulics and sediment transport. Ferguson (2003)has suggested that using a depth-averaged shear stress (i.e. a rectangular channel) underes-timates channel-wide transport because sediment transport scales non-linearly with flow. Hederived a mathematical solution for incorporating channel shape into one-dimensional sedi-ment transport calculations. His method involves integrating a sediment transport formulaover a range of depths. While this works well for simple formulae (such as the Meyer-Peterand Mu¨ller equation he used), relations designed for transport of sediment mixtures, which in-clude hiding functions, yield significantly more complicated integrals that lose physical mean-ing. In addition, large floods capable of transforming channel morphology may occur multiple122.3. Model frameworkFigure 2.2: Schematic showing MAST-1D reservoirs and exchanges within one model cell.Red-filled arrows denote exchanges that are determined by width change.times over decadal timescales, and therefore channel shape cannot be assumed constant. Fur-thermore, our model does not account for the transport and deposition of large wood, whichcan affect channel shape as much as sediment transport (Abbe and Montgomery, 1996). Forthese reasons, we believe that using a rectangular channel is justified, even though it imposesa significant loss of realism.MAST-1D iterates through 5 main processes. First, hydraulics are calculated using thestandard step method applied to the backwater equation, assuming steady, gradually-variedsub-critical flow. The flow area is divided into two segments–the channel and floodplain.Bedload is then calculated with a form of the Wilcock and Crowe (2003) equation that has beenoptimized for large, cobble-bedded rivers by Gaeuman et al. (2009). It is assumed that anywashload in the channel passes through the system, although during flood conditions somewashload and fine bed material may deposit on the floodplain. Next, lateral exchanges ofsediment to and from the channel are calculated, as described below. Sediment transport andlateral exchange rates are used to calculate the change in bed elevation using a modified form132.3. Model frameworkFigure 2.3: Schematic showing reservoir exchanges. Arrows and dashed boxes show sed-iment fluxes. Bank erosion (a.) causes material from the floodplain and floodplainsubstrate to be added to the channel zone, increasing the width of the channel. Veg-etation encroachment (b.) leads to channel narrowing, as channel and point barsediment is incorporated into the floodplain zone.142.3. Model frameworkof the Exner equation:∆z∆t=1Bc(1− λ) ∗Im + Qs,in −Qs,out∆x(2.1)where ∆z∆t is the rate of bed elevation change, Bc is channel width, λ is bed porosity, ∆x isthe channel length, Im is the incoming flux due to lateral change, and Qs,in and Qs,out arethe sediment transport rates for the upstream supply and load. Mass conservation is thenapplied on a size-specific basis to each reservoir (active layer, channel, and substrates), and thegrainsize distributions and geometries are updated to reflect incoming and outgoing sedimentfluxes. If the channel aggrades beyond a threshold, avulsion occurs, transfering sedimentbetween the channel and floodplain and lowering the bed elevation (see below).2.3.2 Lateral exchange and width changeIn alluvial rivers, active channel width is maintained by two processes: the scour of banksby the flow, which acts to widen the channel, and the encroachment of floodplain vegetation,which leads to narrowing of the active channel margin. The interaction between these twoprocesses results in channel migration and an exchange of sediment to and from the flood-plain.In MAST-1D, lateral exchange describes the magnitude of sediment fluxes between chan-nel and floodplain reservoirs within each model cell (see Figure 2.3; floodplain and floodplainsubstrate reservoirs are teal and channel reservoirs–active layer and channel substrate–are ma-genta). The volumes of sediment entering and exiting the channel per unit time are functionsof bank erosion (E) and narrowing caused by vegetation (N), whereIm =E∆t∗ LF ∗ ∆x (2.2)andOm =N∆t∗ LPB ∗ ∆x (2.3)where LF is the height of the floodplain above the substrate and LPB is the height of the pointbar (a constant). The rate of channel width change is a function of the magnitudes of bankerosion and point bar and vegetation growth (Parker et al., 2011):∆Bc∆t=E∆t+N∆t(2.4)where∆Bc∆trepresents the total rate of channel width change. Note thatN∆tis negative.Bank erosionOur simple model of channel widening only relates bank erosion to sediment transport capac-ity. Parker et al. (2011) notes how bank retreat in natural rivers is held in check by a protective152.3. Model frameworkFigure 2.4: Conceptual model of bank erosion along a coarse, cohesionless bank. I. Thebank is protected by flow by an unconsolidated toe deposit with an armored sur-face. During flows with high enough energy, the armor layer on the bank toe isentrained. II. The protective bank toe deposit is transported away. III. Bank materialis eroded away, either by dislodgement and avalanching of individual particles or byslip failure. IV. When the supply of bank material exceeds the capacity to transformit, a new toe develops, over which a new armor layer forms as the flood recedes.layer on the bank toe. In the fine-grained, lowland streams considered in their study, thisprotection took the form of slump blocks of cohesive material that fall into the river duringbank retreat and cap the bank toe. Our hypothesis is that in coarse-bedded, non-cohesivefloodplains, an armor of large grains protect the bank toe and modulate rates of bank erosion(Figure 2.1). Scour results when the shear stress of the flow is sufficient to entrain and trans-port particles from the near-bank region. A conceptual model for bank retreat is presented inFigure 2.4. In order for erosion to occur, the coarse surface grains armoring the bank toe mustbe transported (Stage I in Figure 2.4). The initiation of bank erosion is therefore a functionof the entrainment condition for near-bank particles (Nanson and Hickin, 1986), which dependon fluid forces acting on the bed and grain-bed collisions (Sutherland, 1967). In particular, thelargest grains must be reaching the threshold for full mobility (MacKenzie and Eaton, 2017) inorder to break up the armor layer on the bank toe and allow it to be transported away (StageII). Stage III is initiated once the protective toe is gone. Bank retreat occurs, by entrainment andavalanching of individual particles and by shear failure (see ASCE Task Committee on Hydraulicsand of River Width Adjustment, 1998a). When the sediment supplied from the bank exceeds thetransport capacity in the near-bank region, a new toe develops, becoming armored as the flowwanes (Stage IV; and refer to Thorne, 1982).Because MAST-1D uses a rectangular cross-section, and because the spatial resolution isdesigned to be large (multiple channel widths and meander bends), our widening functionuses channel-averaged hydraulic and sediment transport metrics to calculate bank erosion.162.3. Model frameworkBoth field and experimental data support this approach for self-formed channels with cohe-sionless banks. Nanson and Hickin (1986) argue that bank erosion is a function of the bed shearstress, as sediment transport on the bed leads to the undercutting and subsequent collapse ofbanks. They find that stream power, along with the shear force of bank material (which is afunction of grainsize) explain most of the variability in channel migration rates along severalCanadian rivers. In fact, stream power appears correlated to bank erosion rates on a variety ofsystems (Krapesch et al., 2011; Nicoll and Hickin, 2010). These studies suggest that the magnitudeof bank erosion can be related to the channel-averaged strength of flow. The disadvantage tothis approach is that it does not take into account increased shear stress along the banks due tochannel curvature. The model is therefore most appropriate for channels with low sinuosity,where a straight channel assumption is justifiable. In addition, MAST-1D does not account forretreat that occurs when banks become oversteepened as a result of degradation (e.g. Cantelliet al., 2007), nor does it consider mass failure resulting from a build-up of pore pressure innear-bank deposits (Higson and Singer, 2015).Our approach for the initiation of bank erosion stems from the work of MacKenzie and Eaton(2017), who find that lateral channel instability in their laboratory channel occurred whenthe largest grains in the sediment mixture were fully mobile on the bed. In natural, poorly-sorted cobble-bedded rivers, it is likely that full mobility of the coarse fraction is rarely, if ever,achieved. Therefore, widening in our modified version of MAST-1D is occurs when a supply-normalized unit transport rate of the upper tail of the grainsize distribution, qsCmax, exceeds athreshold, qscr. Our physical interpretation is that, beyond qsCmax, the grains on the bank toeare sufficiently mobile to break up the armor layer, allowing the toe to be scoured away (StageII in Figure 2.4). We define the supply-normalized unit coarse transport rate asqsCmax = qsC/ fC (2.5)where qsC is the unit sediment transport rate of the coarse end of the surface sediment mixtureand fC is the fraction of that group of sizes present in the bed. qsC is calculated via the Gaeumanet al. (2009) version of the Wilcock and Crowe (2003) function, though any fractional sedimenttransport equation will work. It is directly proportional to fC (see Equation 2 in Gaeumanet al., 2009). Therefore, qsCmax represents the transport rate expected with an unlimited supplyof coarse sediment. Equation A.43 is equivalent to the fractional transport scaled to the bedsurface distribution described by Wilcock and McArdell (1993), who use it to identify thresholdsbetween partial and full transport. There is currently no straightforward way to determine thethreshold unit transport rate qscr. As a first step, it should be estmated for each system, ideallyby comparing sediment transport calculations to field data of bank erosion.Once bank erosion is initiated (Phase III in Figure 2.4), floodplain sediment mixes with theactive layer adjacent to the bank, and the magnitude of bank erosion depends on the abilityof the flow to transport this near-bank sediment. When coarse sediment supply from thebank exceeds the transport capacity, it will build up along the bank toe and protect it from172.3. Model frameworkfurther erosion. The near-bank sediment transport capacity, qsNB, is a function of the grainsizedistribution of the near bank region, which is defined byfi,NB = α f fi + (1− α f ) fi,FP (2.6)where fi,NB is the near-bank fraction of size class i, fi is the fraction in the active layer, fi,FPis the fraction in the floodplain, and α f is a mixing constant that ranges between 0 and 1.Larger values of α f make bank erosion more dependent on the grainsize distribution of thebed and therefore more sensitive to sediment supply. qsNB is calculated using the Gaeumanet al. (or other fractional) sediment transport relation, with fi,NB as the grainsize distribution.The portion of qsi,NB that transports coarse floodplain material, qsC,FP, isqsC,FP =qsC,NBfC,NBfC,FP(1− α f ) (2.7)where qsC,NB is the unit coarse sediment transport rate of the near-bank mixture and fC,FP isthe fraction of coarse material in the floodplain. The bank erosion rate isE∆t=0, qsCmax ≤ qscr(qsC,FP)/(LF ∗ fC,FP), qsCmax > qscr (2.8)where LF is the bank (floodplain) height.Vegetation encroachmentChannel narrowing results from multiple interrelated processes, including deposition on bars,degradation leading to the development of benches, and encroachment of vegetation ontoexposed surfaces. One of the weaknesses of using a simple rectangular cross-section is thatlateral variability in deposition cannot be modeled. Here we focus on channel narrowing dueto vegetation encroachment. The relationship between vegetation growth and hydrology iscomplicated, with the germination and ultimate success of seedlings dependent on the spatialand temporal availability of soil moisture (Camporeale et al., 2013; Pasquale et al., 2012), floodsequencing (Camporeale and Ridolfi, 2007; Perona et al., 2012), the availability of in-channel wood,which act as nurse logs (Bertoldi et al., 2009; Fetherston et al., 1995), and species (Robertson andAugspurger, 1999). As a first step, channel narrowing only occurs during relatively low flowsin MAST-1D. We assume that the magnitude of vegetation encroachment is proportional to thearea of unvegetated point bar surface (this is the same approach used by Konrad, 2012). Therate of encroachment is treated as a constant, αn:N∆t=−αn ∗ (Bc − Bmin), τ < τr0, τ ≥ τr (2.9)182.3. Model frameworkFigure 2.5: Example of a local avulsion on Elwha River, Washington roughly 2 km down-stream of Glines Canyon Dam. In 1994 (a), the channel is composed of two branches.An avulsion takes place between 1994 and 2006 (b), creating a third, narrow channelabout halfway down the bifurcation. By 2009 (c), the old channel has lost all flowand been colonized by vegetation. Active channel margins are outlined in blue.Bmin is a constant user-defined minimum width and Bc − Bmin represents the unvegetatedpoint bar. τr represents a reference shear stress, below which flow is low enough to leave sur-faces exposed for colonization. As theories of vegetation colonization become more advanced,this algorithm may be improved.2.3.3 AvulsionAvulsion is an inherent feature of river systems in which the rate of bed aggradation outpacesthe timescale over which the channel can sequester that sediment into the floodplain via lateralmigration (Jerolmack and Mohrig, 2007). In forested rivers, local avulsions occur where thechannel is blocked by log jams (Gottesfeld and Gottesfeld, 1990; O’Connor et al., 2003). A casestudy of a typical channel evolution sequence following a local avulsion is presented in Figure2.5. In the period between air photos in Figures 2.5a and b, flow from the downstream half ofthe left fork of a bifurcating stream is captured by a new channel that cuts across a mid-channelisland, joining with the right fork a few hundred meters downstream. The old channel is notabandoned immediately, although it narrows. After a few to several years (Figure 2.5), the newchannel captures all flow from the left fork (at most stages), it widens, and the old channel is192.3. Model frameworkcolonized by vegetation and incorporated into the floodplain.In MAST-1D, we characterize avulsion in a much simpler way by treating it as an instanta-neous process. It is triggered in model cells experiencing high levels of aggradation where thebed elevation approaches that of the floodplain. This is similar to the approach taken by Jerol-mack and Paola (2007), although their cellular model accounts for levees and defines initiationof avulsion when the channel aggrades above the floodplain. A schematic of the MAST-1Dalgorithm is presented in Figure 2.6. The floodplain (teal) is composed of flood channels thatinitially lay above the primary conveyance zone (i.e. the active layer, Figure 2.6a). Note that inMAST-1D, the floodplain is treated as a single rectangular reservoir, though we show a floodchannel in Figure 2.6 to clarify the process. As the channel (displayed in magenta) aggrades,the height of the bank, L f , decreases. When L f dips below a threshold value Lt, avulsion isinitiated (Figure 2.6b). The bed elevation lowers by a spacing constant La as part of the channelinhabits the floodplain channel (Figure 2.6c):znew = zold − La (2.10)zold is the pre-avulsion bed elevation and znew is the resulting elevation (Figure 2.6d). Thesurface of the new channel becomes active, so that the volume of floodplain material added tothe active layer for size class i (ALin,i) isALin,i = αa ∗ Bc ∗ LAL ∗ fi,FP ∗ ∆x (2.11)where αa is the fraction of channel that avulses, LAL is the thickness of the active layer, and fi,FPis the fraction of size class i in the floodplain. We make the simplification that the abandonedportion of channel becomes vegetated immediately, so that the volume of channel sedimentsequestered into the floodplain reservoir (FPin,i) isFPin,i = [αa ∗ Bc ∗ LAL ∗ fi,AL + Bc ∗ La ∗ fi,SC] ∗ ∆x (2.12)where fi,AL is the fraction of size class i in the active layer and fSC is the fraction in the chan-nel portion of the substrate. To conserve mass, we must assume that the aggraded materialin the ‘non-avulsed’ portion of channel (the channel substrate) enters the floodplain. There-fore, even though we are assuming that the overall channel width does not change during theavulsion, slightly more sediment is sequestered into the floodplain than that which becomesnew channel. It is also important to note that our avulsion threshold does not account for thevery important influence of log jams, which are known to cause avulsions forested wanderingrivers. We assume that the magnitude of lag jamming is proportional to the aggradation rateof sediment. This is a portion of the model that can be improved.After the avulsion, the boundary between the floodplain, active layer, and substrate reser-voirs are adjusted to reflect the new LF (which represents channel depth). Details of this pro-202.3. Model frameworkFigure 2.6: Conceptual diagram of avulsion in MAST-1D with relevant parameters. a.river setup prior to avulsion. The channel elevation is lower than that of the flood-plain channel. b. river setup at the threshold for avulsion. The channel (magenta)has aggraded so that it is within a threshold length (Lt) of the floodplain height.c. river setup after avulsion. The depth of the channel has lowered by La. Oldchannel material is sequestered into the floodplain, while floodplain material is in-corporated into the active layer. d. Longitudinal profile before (black line) and after(orange line) an avulsion on node n.212.4. Model behaviorFigure 2.7: Sample of repeating hydrograph used in model runs.Table 2.1: Select initial conditions for model runsParameter Value SourceBank height 1.86 m derived from Castro and Jackson (2001)Width 81 m Model calibrationGradient 0.0069 DEMThickness of overbank material 0.14 m Bank surveyD50 68 mm Bulk SamplingD90 265 mm Bulk Samplingcedure can be found in Appendix A.2.4 Model behaviorSimple simulations were performed to demonstrate the general behavior of MAST-1D. Theinput parameters are loosely based on Elwha River, a wandering cobble-bedded stream witha drainage area of about 850 m3/s. The primary initial conditions are listed in Table 2.1. A fulllist can be found in Appendix B. Sediment was supplied at capacity.The fractions and transport rates of coarse material in the active layer and floodplain arecalculated by summing the fractions for three boulder size classes:fC = f256−362 + f362−512 + f512−1024 (2.13)andqsC = qs256−362 + qs362−512 + qs512−1024 (2.14)where the subscripts refer to the bounds of the size classes used in mm. These classes approx-imately reflect all material including and above the channel D90 (Table 2.1).When run for over two centuries with a repeating hydrograph (Figure 2.7), avulsions oc-cur every few decades (Figure 2.8a). The other channel characteristics oscillate at the same222.4. Model behaviorFigure 2.8: Channel characteristics over time for a single node. a) width of material ex-changed during avulsion, b) annual channel migration rate, calculated as the meanof the annual rates of narrowing and widening, c) channel width, d) sediment trans-port rate during the maximum discharge, e) median channel grainsize. Parameters:qscr: 10−7; α f : 0.55; τr: 32 m3/s; αn: 0.055; Bmin: 40 m; Lt: 0.75 m; αa: 1; La: 1 m. SeeAppendix B for other parameters.232.4. Model behaviorFigure 2.9: Rates of bank erosion and vegetation encroachment. See Figure 2.8 for param-eters.timescale. The migration rate (2.8b) declines gradually from around 1.5 m/yr to about 1 m/yr,then rapidly rises following an avulsion due to the increase in channel depth. Decadal vari-ability is greater for bank erosion than vegetation encroachment, for which rates for all yearsfall within a 0.4 m/y range (Figure 2.9). For roughly half the time, bank erosion out-paces veg-etation encroachment, and channel width is increasing (Figure 2.8c). As bank erosion declines,encroachment becomes the dominant mode of channel migration, and width declines until thenext avulsion event. Spikes in the rate of sediment transport (Figure 2.8d) and bed coarsening(Figure 2.8e) occur at the same time as avulsions. In between these events, sediment transportdeclines slowly as channel capacity is reduced due to aggradation. The D50 fluctuates betweena range of about 8 mm.The trends presented in Figure 2.8 hold regardless of the initial channel dimensions. Widthis plotted twice a year, during the low flow period and after the largest flow, in Figure 2.10.The three series represent runs which are identical except for the starting width. Annually,width varies by roughly a half meter as narrowing occurs during low flows and bank erosionwidens the channel during the upper tail of the hydrograph. All runs evolve to a similar rangeof widths, which slowly declines through time.Sediment supply affects the evolution of channel width and the frequency of avulsion (Fig-ure 2.11). When upstream sediment is supplied at capacity, the channel avulses roughly twiceas often than if there is no upstream supply, and fluctuates between a smaller range of widths.When sediment supply is cut off, width decreases slowly on centennial timescales, althoughavulsions cause a spike in sediment supply that lead to temporary channel widening. Ourmodel suggests that rivers are much less sensitive to flow sequencing then they are to sedimentsupply. To test the effect of hydrograph shape on the evolution of channel width, MAST-1Dwas run using the discharge record from the 2007 water year. For one run, the actual sequenceof discharges were maintained, and the hydrograph was repeated for 100 years. The same setof discharges were then ordered sequentually (similar to the hydrograph presented in Figure242.4. Model behaviorFigure 2.10: Evolution of channel width over time for different intial widths. For param-eters, refer to Figure 2.8.Figure 2.11: Channel width for runs with and without an upstream supply of sediment.2.7) and repeated for the same time frame. Channel width for the two runs is plotted in Figure2.12. The shape of the hydrograph appears to change the magnitude, but not amplitude, ofthe response. Even though the run with the ‘natural’ hydrograph evolves to a slightly lowerwidth, avulsion occurs at a similar frequency.In summary, MAST-1D predicts that coarse-bedded, wandering alluvial systems exhibitcyclical, autogenic behavior. The channel fills with sediment, which reduces its capacity andcauses it to fill faster. Eventually, a local avulsion occurs, which deepens the channel andincreases its capacity again. This process is maintained regardless of initial channel width,sediment supply, or flow sequencing.252.4. Model behaviorFigure 2.12: Effect of flow sequencing on channel width. Flow is calculated using by re-peating daily flows from the 2007 water year. For the ‘natural’ run, the true orderof flow events was preserved, while discharges were put in ascending order for the‘sorted’ run.2.4.1 Sensitivity analysisA sensitivity analysis was performed for the bank erosion, vegetation encroachment, and avul-sion parameters. The results are presented in Figure 2.13 for channel width and in Figure 2.14for channel migration. Variation in most parameters has a limited impact on channel widthand avulsion frequency. Only for Cmax (the coarse particle size class) did a variation of 10%affect width by more than that amount. When the bounds of the three size classes that we havedefined as the ‘coarse’ range are each lowered by 10%, the model evolves so that the range ofchannel widths is nearly 10 m higher, the migration rate is about a third higher than the baserun, and avulsions occur more frequently. The opposite occurs when Cmax is 10% coarser.The channel migration rate is moderately sensitive to the floodplain mixing coefficient(α f ), narrowing coefficient (αn), and minimum width (Bcmin), which each vary by roughly 10%when they are raised/lowered by that amount. Increasing α f implies that less of the near-banksediment in transport is sourced from the banks, lowering the rate of channel widening forany given flow in which bank erosion is initiated. When α f is low, the implication is a higherbank-sediment transport rate and therefore a higher widening rate. Since channel narrowingis linearly related to αn, increasing it causes higher migration rates, and vice versa. Vegetationencroachment rates are inversely proportional to Bmin.Surprisingly, neither channel width nor migration rates are sensitive to the avulsion thresh-old (Lt), avulsion mixing parameter (αa), or avulsion depth (La). It is important to note thatthe upstream boundary condition was set so that sediment was supplied at capacity for thesensitivity analysis runs. The choice of parameter may become more important in conditionsof sediment starvation or excess, where the difference in grainsize between the active layerand floodplain is more pronounced.262.5. DiscussionFigure 2.13: Sensitivity of channel width to model parameters. Solid teal lines representthe base run, yellow dashed line represents a 10% increase in the parameter, andmagenta dotted lines denote a 10% decrease in the parameter value. Base run: qscr:10−7; α f : 0.55; τr: 32 m3/s; αn: 0.055; Bmin: 40 m; Lt: 0.75 m; αa: 0.2; La: 1 m.2.5 DiscussionTheories of river dynamics have traditionally favored lowland, migrating channels, yet weoften apply numerical models based on these theories to rivers behaving according to a dif-ferent set of processes. Conceptual models for wandering, cobble-bedded rivers indicate thatbank erosion during large floods and avulsion are the key ways in which the floodplain andchannel exchange sediment (O’Connor et al., 2003; Konrad, 2012). However, there is a paucity ofdecadal-scale numerical models that take both of these processes into account. In MAST-1D,we have made simplifying assumptions about channel geometry, morphology, and reach-scaleprocesses in order to explore the long-term dynamics of avulsion and width change.272.5. DiscussionFigure 2.14: Sensitivity of migration rate to model parameters. Solid teal lines representthe base run, yellow dashed line represents a 10% increase in the parameter, andmagenta dotted lines denote a 10% decrease in the parameter value. Base run: qscr:10−7; α f : 0.55; τr: 32 m3/s; αn: 0.055; Bmin: 40 m; Lt: 0.75 m; αa: 0.2; La: 1 m.Our model suggests that autogenic variability is inherent to rivers with erodable banks andthe ability to shift channels. Decadal-scale oscillations in migration rate, sediment transportcapacity, and channel width appear to be modulated by the frequency of avulsions (Figure2.8). When the channel avulses, a fresh supply of relatively fine floodplain sediment is imme-diately available for transport and the channel becomes deeper, causing higher shear stressesduring large flows. This increases the mobility of the channel and causes a spike in the sedi-ment transport rate (Figure 2.8d). The bed quickly armors as the fine sediment is selectivelytransported (Figure 2.8e). Over a few decades, the sediment transport and migration ratesslowly decline, trending to a more stable channel until the next avulsion event.282.5. DiscussionThere are two major assumptions in our avulsion model that impact its behavior. The firstis that floodplain sediment is mixed with the active layer following an avulsion. The physicalinterpretation of this is that the sediment in side channels have a high connectivity to theprimary channel in the reach. Since the majority of avulsions and side channels appear to besub-reach scale (Jerolmack and Paola, 2007, also see Figure 2.5), it is likely that they can readilyexchange sediment. The other assumption is that floodplain channels inhabited following anavulsion are not armored. Avulsing channels tend to favor former channel locations (Jerolmackand Paola, 2007), so it is possible that new channels are already armored and do not provide apulse of fresh sediment supply (Figure 2.8d). However, floodplain channels are often activatedduring high discharge events, and these may accumulate sediment that is readily releasedwhen the channel captures a higher percentage of flow during an avulsion.Channel width is a function of the relative intensities of bank erosion and vegetation en-croachment. Its oscillatory behavior on decadal timescales (Figure 2.8c) is a result of two pro-cesses. The abrupt increase in shear stress and mobility following avulsion leads to higherrates of bank erosion, causing widening. As the channel widens, its competence per unitchannel decreases because the flow is spread over a larger surface area. In addition, lateralsediment supply via bank erosion accelerates aggradation, further reducing the capacity (andshear stress) of the channel during high flows. Bank erosion rates subsequently decrease.When they become lower than the rate of vegetation encroachment, the channel begins tonarrow. A narrower channel means that there is less exposed sediment available for coloniza-tion, and the rate of encroachment declines slightly until the next avulsion event. The resultsshown in Figure 2.9 suggest that bank erosion is more sensitive to autogenic fluctuations inbed mobility than vegetation.Our MAST-1D simulations suggest that sediment supply has a large impact on timescalesof autogenic adjustments: when supply is high (at capacity), avulsion is more frequent thanwhen supply is low (no upstream supply, Figure 2.11). An excess of sediment exists in thechannel when upstream supply is at capacity because bank erosion also acts as a source of sed-iment supply. While vegetation encroachment sequesters some of this excess sediment, it doesnot directly lead to a change in bed elevation. Therefore, the channel is in a constant state ofnet aggradation. In reality, sediment supply to gravel-bed rivers can be episodic, and periodsof aggradation followed by stability may be the norm. In addition, abrasion and weathering ofgrains, processes not accounted for in MAST-1D, may increase mobility of individual particlesand counter aggradation.Interestingly, MAST-1D predicts that the channel still aggrades (albeit at a much lowerrate) when upstream supply is cut off, even though the channel armors. This is at odds withthe commonly-held view that sediment starved conditions generally lead to degradation (e.g.Galay, 1983; Williams and Wolman, 1984). Much of our theory about channel profile changestems from observations of rivers with self-formed substrates. In the Pacific Northwest, manyrivers are reworking floodplain surfaces that were created during glacial and para-glacial con-292.6. Conclusionditions, and are under-fit to transport the largest particles in the channel at rates necessaryto cause substantial bed elevation change (Hassan et al., 2014). Therefore, it is plausible thatduring conditions of sediment starvation, fine material supplied via bank erosion is quicklywinnowed away while the remaining coarse floodplain gravels, cobbles, and boulders accu-mulate in the channel and contribute to the armor layer. The long term response of the riverwould then be a slow net loss of finer floodplain material instead of incision. While field stud-ies of wandering rivers generally associate low sediment supply with channel stability (e.g.Konrad et al., 2011; Pohl, 2004), our MAST-1D results suggest that bank erosion and avulsion dostill occur during sediment starvation, though at much lower rates. This is supported by fieldevidence; the avulsion presented in Figure 2.5 occurred along the sediment-starved ElwhaRiver, where a dam 2 km upstream had cut off supply to the reach for nearly 100 years.We adapted MAST-1D with the hypothesis that coarse cobble-boulder sized sediment formsan armor layer on the bank toe that protects the bank from frequent erosion. Our model qual-itatively mimics the behavior of the experimental channels described in MacKenzie and Eaton(2017): increasing the size of the largest grains in the sediment mixture leads to less lateralmigration and a narrower channel (Figures 2.13 and 2.14). In fact, model behavior is moresensitive by a large margin to the coarse grainsize fraction (Cmax) than any of the other pa-rameters tested. Our simplistic conceptual model of non-cohesive banks may leave out keycomponents that contribute to the initiation and magnitude of bank erosion. In particular, weneglect the role of vegetation and its interaction with bank material. Both field (Beechie et al.,2006) and numerical (Eaton and Giles, 2009) evidence suggest that vegetation plays little rolein bank strength for larger channels, whose bank toes extend much deeper than the rootingdepth. However, MAST-1D will overestimate bank erosion in channels narrower than about15 m (Beechie et al., 2006), where vegetation becomes significant. In addition, improvementscould be made to MAST-1D to include the important role of woody debris in initiating channelavulsion and providing suitable terrain for vegetation growth.2.6 ConclusionTraditionally, geomorphologists have operated under the assumption that rivers can be ade-quately characterized by a single ‘regime’ width, and it is common to numerically model chan-nel processes without dynamically linking them to the floodplain. We developed MAST-1D tosimulate coarse-bedded, wandering rivers, where width changes are frequent and morpholog-ically important. These types of rivers are unique in that local, reach-scale avulsions and bankerosion both act as sources of sediment to the channel. The model suggests that the dominantdischarge concept overlooks inherent decadal-scale variability, even when flow and sedimentsupply regimes are in a steady state. Channel aggradation leads to avulsion, which locallyincreases channel depth (and shear stress), increases sediment mobility, and affects channelmigration rates. Changes to boundary conditions, such as a reduction in sediment supply, al-ter the timescale and magnitude at which wandering rivers operate but do not fundamentally302.6. Conclusionchange the dominant fluvial processes.MAST-1D serves as a tool for analyzing and predicting river behavior on the scale ofdecades to centuries. While we have included most relevant parameters for this timescale,future effort should be focused on modeling the role of vegetation, particularly as it relatesto log-jam formation and colonization of point bar surfaces, and to including algorithms forabrasion and floodplain weathering, which impact long-term sediment mobility. In addition,our understanding of long-term channel evolution would benefit from more observations oflocal avulsion events. Future field and laboratory campaigns should focus on quantifying thefluctuation of channel depth following channel reorganization.31Chapter 3The impact of sediment supply onchannel stability along Elwha River,Washington following damemplacement and removal3.1 SummaryTwo hydropower dams, which had been in place on Elwha River for nearly one hundred years,were removed beginning in 2011. The channel, which had coarsened and become more sta-ble as a result of sediment starvation, experienced increased rates of bank erosion and newchannel activation following the removal. Our research attempts to understand the impact ofsediment supply on channel stability. We are especially interested in whether the same ge-omorphic processes that led to increased stability following dam emplacement can explainpatterns of channel change in the opposite trajectory after the removal. We used MAST-1D,a one-dimensional size-specific bed evolution model which accounts for bank erosion, vege-tation encroachment, and local channel avulsion to simulate reach-scale evolution of ElwhaRiver over its entire history of dam emplacement and removal. The model treats bank erosionas a function of sediment mobility, but does not account for hydraulic changes related to barformation. MAST-1D was able to reproduce field data of particle size, channel width, andrates of channel widening and narrowing over the period in which the dams were in place.Our observations and modeling suggest that channel armoring led to decreased rates of bankerosion, and that reductions in channel width slowed the rate of vegetation encroachment, asless bar space was available for colonization. Reductions in sediment supply decreased thefrequency of avulsion. These factors combined to increase the overall stability of the channel.Simulations in MAST-1D are consistent with increased rates of bank erosion and avulsion thatwere observed in the field following dam removal, but the model underestimated the ability323.2. Introductionof Elwha River to export the pulse of sediment released from the upstream reservoir. It islikely that local channel morphology, especially the impact of bar deposition and curvature onlateral variability in flow strength, play an important role in characterizing channel instabilityin this period. Information on the caliber of sediment supply from the upstream reservoir,as well as a better understanding of vegetation processes, can increase the success of futuremodeling studies and improve our understanding of landscape recovery from dams.3.2 IntroductionThe Glines Canyon and Elwha Dams, which had cut off sediment supply to Elwha River fornearly a century, were torn down between 2011 and 2014, releasing several million cubic me-ters of sediment downstream. Within the span of a few years, the reaches between the damsunderwent a regime shift from a sediment starved, armored system to a supply-rich, unstableconfiguration (Draut et al., 2011; East et al., 2015; Kloehn et al., 2008; Konrad, 2009; Pohl, 2004).Elwha River may take decades to recover from the sediment pulse (East et al., 2015) and to ad-just to a regime with a higher sediment supply. While other unregulated regional rivers offera glimpse of how the Elwha may evolve (Kloehn et al., 2008), it is unclear how long recoverywill take and how dynamic it will be in the future (Draut et al., 2011).While the case of Elwha River is unusual in its scale, the geomorphic processes related tosediment supply reductions and pulses are observed on many gravel and cobble-bed rivers.Streams respond to adjustments in governing conditions with changes in particle size, width,and depth. These factors affect the capacity of the channel and its ability to transport sedi-ment. Most river bed surfaces coarsen as a result of low sediment supply (e.g. Hassan et al.,2006; Williams and Wolman, 1984), and many experience channel narrowing due to flood peakreduction and encroachment of vegetation onto formally active channel surfaces (e.g. Gordonand Meentemeyer, 2006; Swanson et al., 2011). The Piave and Brenta Rivers in Italy narrowedas a result of declining sediment supply due to gravel mining, then widened after miningstopped (Kaless et al., 2014). Similarly, Kondolf et al. (2002) found that land use changes in riversof vastly different scales resulted in width change–again, increased sediment load to the riverled to widening, and decreased load resulted in narrowing.Pulses of sediment from dam removals, landslides, and other disturbances can increasesediment transport rates, change the texture of the bed surface, and alter channel morphologyon timescales of several months to decades and longer (Czuba et al., 2012; Major et al., 2012;Wilcox et al., 2014). While ample research has focused on the evolution of sediment pulsesthrough translation and dispersion (e.g. Cui et al., 2003; East et al., 2015; Lisle et al., 2001), muchless is known about the feedbacks between sediment supply and channel stability during ex-treme events. Large inputs of sediment can lead to deposition on channel bars, which con-stricts flow near the banks and can lead to bank erosion. In addition, if a sediment pulse iscomposed of fine material, it can increase the mobility of the bed surface, leading to instability(Wilcock and Crowe, 2003). While Madej et al. (2009) observed channel widening on Redwood333.3. Study locationCreek, California following a sediment pulse caused by a large flood, Tullos and Wang (2014)found that little widening occurred on Dahan River, Taiwan after a dam failure. More work isneeded on the factors that lead to channel instability during extreme sediment supply events.Assessing the impact of sediment supply on channel stability in the field is challengingfor a number of reasons. Geomorphic processes often operate on timescales much longer thanavailable field observations. Decadal-scale studies generally rely on aerial photographs, whichvary widely in quality and are not always available at convenient time intervals. It can bedifficult to distinguish the impact of sediment supply from that of hydrology, when both leadto morphologic changes observed in photographs. It is for this reason that many researchersand managers turn to numerical modeling to explain field observations.Here we use MAST-1D, a reach-scale, one-dimensional bed evolution model, to assess thefeedbacks between sediment supply and channel stability on Elwha River over its century-long history of dam emplacement and subsequent removal. Our objectives are twofold: wefirst assess whether MAST-1D can adequately reproduce observed geometric and texturalchanges both over the period of sediment starvation when the dams were in place and duringthe sediment pulse following removal. We then analyze how these changes in sediment supplyhave impacted channel evolution. Our focus is on changes to rates of floodplain-channel inter-action. While channel coarsening and narrowing has been observed on many dammed chan-nels, most previous research on the subject (e.g. Konrad et al., 2011; Williams and Wolman, 1984)considers rivers where the flow regime has been significantly altered; therefore, the influenceof sediment supply cannot be separated from that of the flow regime. The Glines Canyon Damremoval is the largest to date in the United States, and the magnitude of the sediment pulseis unique among other projects. As we usually lack detailed before-and-after observations onnatural sediment pulses, feedbacks between channel stability and events of this magnitudeare largely still open questions. While the results of our study are only strictly applicable toElwha River between Glines Canyon Dam and Elwha Dams, they should be relevant to thoseinterested in the effects of sediment supply on channels, particularly in gravel-bedded, moun-tain rivers. In addition, they may prove useful for improving channel geometry algorithms forlandscape evolution models.3.3 Study locationElwha River drains 833 km2 of primarily steep terrain from the northern flanks of the OlympicMountains to the Strait of Juan de Fuca (Figure 3.1). The basin straddles the Hurricane RidgeFault, which separates two major Eocene-aged terranes (Tabor and Cady, 1978). The upperportion of the basin drains the highly deformed Olympic Sedimentary Complex, containingmetasedimentary rocks, while downstream the river flows through the Coast Range Terraine,composed primarily of basalt and sandstone (Brandon et al., 1998; Tabor and Cady, 1978). Theregion is uplifting at a rate of 0.28 mm/yr (Brandon et al., 1998), providing steep terrain suscep-tible to landsliding and debris flows (Montgomery and Brandon, 2002). During the Last Glacial343.3. Study locationFigure 3.1: Map of Elwha River basin showing locations detailed in the text. Black tri-anges denote US Geological Survey stream gauges. Red shading in the main mapdelineates alluvial surfaces (channel and floodplain) between Glines Canyon andElwha Dams. Shading in the inset shows the entire Elwha Basin.Maximum, the valley was overrun by both an alpine glacier from the south and from the Juande Fuca Lobe of the Cordilleran Ice Sheet from the north. Retreat began about 14.5k years ago,after which glacio-lacustrine and outwash deposits filled much of the lower valley (Easterbrook,1986; Polenz et al., 2004). Elwha River incised during the Early Holocene following isostatic re-bound along the coast (Mosher and Hewitt, 2004), leaving local bluffs that act as major sedimentsources to the delta. Modern alluvial clasts are composed of a variety of lithologies, sourcedfrom both the local bedrock and from granitic sources within the former glacial extent.Annual precipitation in the basin varies between ∼5600 mm in the headwaters and ∼1500353.3. Study locationFigure 3.2: Longitudinal profile of a. Elwha River and b. the study area extracted fromDEMs collected prior to dam removal. Red box in a. denotes the study area.mm in the rainshadow at the mouth (Munn et al., 1999). Elwha River has a rainfall-dominatedhybrid hydrologic regime with a bimodal hydrograph (Reidy-Liermann et al., 2012). Peak flowsrange between∼130 and 1200 m3/s and generally occur in late autumn or winter. A secondaryrunoff peak occurs during the spring snowmelt season. Mean discharge over the period ofrecord (1896-2015) is 43 m3/s.The upper 83% of the watershed is within Olympic National Park and is nearly pristine.During the early 20th century, two hydropower dams were constructed on the lower portionof the river to accomodate growth of nearby Port Angeles, WA. The 33 m Elwha Dam, formingLake Aldwell, was completed in 1913 and was followed in 1927 by the 64 m tall Glines CanyonDam and adjacent Lake Mills reservoir. The reservoirs trapped 3 ± 0.8 and 16 ± 2 millionm3 of sediment, respectively (Bountry et al., 2011). Prior to 1975, hydrograph alteration forpower generation would have influenced peak flows but likely had a minor impact on dailydischarge. From 1977-2011, the dams were primarily managed as run-of-the-river (see Dudaet al., 2011b). Both dams were removed between 2011-2014.Our study focuses on the Middle Elwha, the set of reaches between the two former dams(Figure 3.1). A 1.5 km bedrock canyon confines the channel immediately downstream of GlinesCanyon Dam, except for an ∼700 m long reach where a small floodplain has developed. Gra-dient is steep, at about 0.011 (Figure 3.2). The next 3 km are less steep (gradient is about 0.008)and have a gravel-dominated, laterally-active anabranching morphology (Knighton, 1998). Thelast 4 km are nearly straight with a gradient of about 0.006, and bedrock outcrops limit flood-plain development and migration in localized reaches.Apart from the dams, human alteration has been limited to minor projects on residential363.4. Methodsproperties. Prior to dam removal, the sediment-starved bed was coarse and stable (Draut et al.,2011; Pohl, 2004), although floodplain-channel interaction continued on a reduced scale (Kloehnet al., 2008). Removal of both dams began in September 2011. Elwha Dam was removedincrementally during a six-month period. The Glines Canyon Dam occurred over a longerperiod that extended from 2011-2014. The full removal schedule can be found in Randle et al.(2015) and East et al. (2015) and is only briefly summarized here. Pieces of the dam weredemolished in stages, and during the first year, only suspended sediment passed downstreaminto the study area. In October 2012, the first pulse of bed material spilled over the dam. Itwas composed primarily of sand and fine gravel (Draut and Ritchie, 2015). The pulse reducedsurface particle size, and was associated with minor channel widening (East et al., 2015). Peakflows during the first two years following the removal were abnormally low; even so, 90% ofthe 5 million m3 of sediment released from the two reservoirs in that time was transportedinto the Strait of Juan de Fuca (Magirl et al., 2015; Randle et al., 2015; Warrick et al., 2015). Damremoval was halted between November 2012 and September 2013. A second pulse of bedmaterial entered the study area when removal re-commenced and the final blast occurred.Removal was completed in summer 2014. The first major flow events following dam removaloccurred in the winter of the 2015 water year. Two floods with recurrence intervals of about2 years caused extensive flooding, despite their modest magnitudes. We observed that theelevation of flow on the floodplain locally reached levels recorded during the largest flood onrecord in the pre-removal period. Early in the 2016 water year, a 30 year flood passed throughthe system, causing additional flooding and infrastructure damage.3.4 Methods3.4.1 MAST-1D setupMAST-1D is a one-dimensional bed evolution model designed to simulate channel and flood-plain exchange over decadal and longer timescales. The model space is divided into a seriesof nodes. Every node contains a set of reservoirs, each with an evolving geometry and grainsize distribution. Size-specific vertical and lateral exchanges of sediment occur between reser-voirs. Aggradation and degradation lead to exchanges between the channel bed, underlyingsubstrate, and sediment load. Lateral mixing between the floodplain and channel occurs viamigration, overbank flooding, and avulsion. Sediment is exchanged between model nodes viadownstream variability in sediment transport capacity.MAST-1D iterates through 5 main processes. First, hydraulics are calculated for a two-part cross-section that includes rectangular channel and floodplain units. We use the standardstep method applied to the backwater equation, assuming steady, gradually-varied sub-criticalflow. Bedload is then calculated with the set of equations presented by Gaeuman et al. (2009),which were calibrated to a large, cobble-bedded river. Next, lateral exchanges of sediment toand from the channel are calculated. Two processes contribute to migration–bank erosion and373.4. MethodsTable 3.1: Boundary conditions for model runsPre-Glines Pre-Removal Post-Removal1918-1927 1927-2011 2011-2016Between damsUpstream sediment supply Sediment rating curve 0 Decay functionDownstream WSE 30 m 30 m Linear reduction/stage-discharge rating curveControl runUpstream sediment supply Sediment rating curve Sediment rating curve Sediment rating curveDownstream WSE Stage-discharge ratingcurveStage-discharge ratingcurveStage-discharge ratingcurvevegetation encroachment. Bank erosion is initiated when the mobility of the coarse (boulder-sized) fraction of the channel surface exceeds a threshold, and its magnitude is a function ofthe sediment transport rate of coarse particles in the channel surface and floodplain. Vegeta-tion encroachment occurs during periods of low shear stress, and its rate is proportional tochannel width. Sediment transport and lateral exchange rates are used to calculate the changein bed elevation using a 1-D form of the Exner equation. Mass conservation is then applied ona size-specific basis to each reservoir (active layer, channel, and substrates), and the grain sizedistributions and geometries are updated to reflect incoming and outgoing sediment fluxes.If the channel aggrades to the extent that it is approaching the height of the banks, avulsionoccurs, lowering the bed elevation and exchanging sediment between the channel and flood-plain. Full details on MAST-1D can be found in Chapter 2 and in Appendix A.There are several potentially important processes not accounted for in our implementationof MAST-1D. Lateral variability in flow hydraulics due to curvature and bar growth is disre-garded, as is temporal variability in channel sinuosity. In addition, the model does not includethe influence of large woody debris and log jams, which can be instrumental in trapping sedi-ment and initiating avulsion.Boundary conditions and run sequencesMAST-1D requires a flow record and two boundary conditions–an upstream sediment sup-ply and a downstream water surface elevation. For the flow, we use daily discharge data for1918-2016 from the USGS Gauge at McDonald Bridge (12045500), located in a bedrock canyonroughly halfway between the former Glines Canyon and Elwha dams (Figure 3.1). We dividethe record into three periods, which are summarized in Table 3.1. The first, which we term the‘Pre-Glines period,’ spans between 1918-1927 and includes the period between emplacementof Elwha and Glines Canyon Dams. We do not simulate the first five years of backwater effectscaused by the Elwha Dam (1913-1918) because the discharge record does not exist for that pe-riod. For the Pre-Glines period, the downstream water surface elevation is set at the heightof the Elwha Dam, 30 m above the initial bed elevation. The upstream bedload was supplied383.4. Methodsat capacity, which we calculated using a 2-zone channel/floodplain cross-section with dimen-sions that have been calibrated to field measurements of long-term sediment supply (this isexplained further below). The suspended sediment load was determined by an empericalrelation developed for Elwha River by Konrad (2009):Qs,s = 10−5 ∗ (0.78Q)2.5 (3.1)where Qs,s is the suspended transport rate in kgs−1 and Q is the discharge at the McDonaldBridge gauge in m3s−1. We assume that grains smaller than 0.5 mm always travel in suspen-sion, which is generally the case in samples collected by Curran et al. (2009). During overbankflooding, larger particles may become suspended and deposit on the floodplain. We set thetrapping efficiency of the floodplain at 20% for both coarse and fine grains. All fine (<5 mm)sediment not deposited on the floodplain is supplied to the next downstream node. Coarsegrains that are not deposited on the floodplain are assumed to re-enter the channel and areaccounted for in the bedload calculation.The second, or Pre-Removal period, starts when Glines Canyon Dam is installed in 1927and lasts until dam removal begins in 2011. We assume that all sediment was trapped be-hind the dam, and set the sediment supply to zero (Curran et al. (2009) estimates that 14% ofsuspended sediment flows past the dam, but we assume this has negligible influence on thebed).The third (Post-Removal) period commences following initiation of dam removal in Septem-ber 2011. We simulated the removal of Elwha Dam in the model by incrementally loweringthe downstream water surface elevation over the 6 month period over which the removal tookplace. After that, a stage-discharge rating curve was developed to set the boundary. The stageis calculated for each flow using the 2-zone cross section and assuming normal flow.In our model, the pulse of upstream sediment supply following removal of Glines CanyonDam is released in October 2012, when bed material began spilling over the dam site. Wedo not account for sediment released during the period between September 2011 and Octo-ber 2012, which was exclusively washload and most of which passed through the study areawithout depositing (East et al., 2015). To simulate the pulse, an exponential decay function ofthe formQs,i = Qc,i(1+ Ce−tλ ) (3.2)is used. Qs,i is the sediment feed for bedload size class i, Qc,i is the size-specific sedimenttransport capacity (see below), t is the time elapsed since the dam removal, and C and λ areconstants that determine the shape of the curve.C and λ were calibrated using measured rates of Lake Mills reservoir denudation andsediment load estimates upstream of Lake Mills provided by the US Bureau of Reclamationand US Geological Survey for the 2012-2016 water years. The rates represent the total sedimentload, so assumptions must be made on the partitioning of sediment sizes. Sedimentology393.4. Methodswork by Draut and Ritchie (2015) indicates that most sediment deposited before October 2013was under 8 mm. However, the load coarsened over time; a subsurface sample collectedon a bar just upstream of the dam site in August 2015 show that grains as large as 181 mmwere available for transport, and cobble-size pieces of dam were observed throughout thestudy area. To mimic this coarsening, the sediment pulse is divided into two phases. BetweenOctober 2012 and October 2013, the size-specific transport capacity (Qc) is truncated at 8 mm,so only sand and fine gravel is supplied. The value of C (hereafter C1) is adjusted accordingly.After October 2013, the full grain size distribution of Qc, is used, and a new value of C (C2) iscalibrated.The partitioning of the sediment pulse between suspended load and bedload is anothersource of uncertainty. No information is available on the percentage of sand that traveled insuspension. To capture the range of possible responses, we perform three runs with differentvalues for the percent of suspended sediment supplied during the Post-Removal period (Table3.2). The parameters and boundary conditions for the Pre-Glines and Pre-Removal periods areidentical for each of these runs (see ‘Between Dams’ in Table 3.1). In the first scenario, run R1,77% of the total load is suspended. This is the ratio reported by Curran et al. (2009), whocollected suspended and bedload samples upstream of Lake Mills from 2006-2007. In theirassessment of Lake Mills denudation, Randle et al. (2015) divided the deposit into fine (silt andclay) and coarse (sand and coarser) material and found that the former only constituted 29%of material evacuated between 2012 and 2013. Therefore, for the second scenario (R2), weassume that the suspended load is 29% of the total load. This is likely an underestimate, sincesand travels in suspension along Elwha River. For the third scenario (R3), we use stratigraphicdata collected from the Lake Mills reservoir between 1989 and 1994 by Gilbert and Link (1995).They estimated volumes and size distributions for each delta unit. We are assuming that thedelta topset (the uppermost layer, which is composed of coarse sediment) represents the bedmaterial captured in Lake Mills. The fraction of bedload is calculated as the volume of thedelta topset divided by the total volume of all reservoir deposits (excluding tributary fans).This comes out to 16% bedload and 84% suspended load and is likely an underestimate, sinceit does not consider bed material sediment available from tributaries. While the uncertaintyin our estimates of sediment supply following the dam removal is high, the three scenariosshould provide a sense of the range of possible responses.For each sediment supply scenario, sediment supply was calculated using Equation 4.2 forthe 2013-2016 water years. C and λ were adjusted until the time series of calculated sedimentsupply was similar to that provided by the USGS/USBR. A plot of the calibrations is presentedin Appendix B.To separate the influence of the dams from hydrologic variability, a control simulation,C4, is run using the same initial conditions as the other three runs but supplying upstreamsediment using a rating curve (see below). In C4, the downstream water surface elevation iscalculated via a stage-discharge rating curve (see above and refer to Table 3.1).403.4. MethodsTable 3.2: Summary of model runsRun %sus.*Source** C1 C2 τ***R1 77 Average of measurements by Curran et al. (2009) 100 42 365R2 29 Coarse/fine partition by Randle et al. (2015) 310 140 365R3 84 Pre-removal Lake Mills reservoir measurements 70 30 365C4 varies Control run, supply determined by rating curves - - -*Percent of load comprised of suspended sediment during Post-Removal period**Source used to determine the partitioning between suspended and bedload during the Post-Removal period***In daysInitial conditions and model calibrationThe study area is divided into 21 cells, each with unique valley widths and sinuosities mea-sured from air photos. In MAST-1D, sinuosity is assumed to remain constant through time,and each cell should incorporate several meander bends. The small length of our study area,as well as the need to capture longitudinal variability in the response of the dams, necessitateda relaxation of this rule, and our cells are roughly 5 channel widths long, shorter than whatwould be considered ‘reach scale.’Model cells fall into one of two types: ‘canyon’ and ‘alluvial.’ Canyon cells, which rep-resent bedrock canyon reaches, have fixed banks (the channel cannot migrate or avulse). Tosimulate bedrock, these cells are not allowed to degrade 0.2 m lower than the initial elevation(potential aggradation is unlimited). Channel width was measured from air photos (see be-low), and initial slope was calculated from a 0.5 meter resolution DEM provided by the USGeological Survey. The initial channel depth was set at an arbitrary value high enough toprevent overbank flooding.For alluvial cells, lateral channel fluxes and unlimited aggradation and/or degradationmay occur. The study area was divided into two sections based on slopes measured from theDEM. Model cells in the upstream segment, which stretches from 2 km downstream of theformer Glines Canyon Dam to river km 17 (Figure 3.2), were given initial slopes of 0.0081.Alluvial cells downstream of river km 17 were assigned an initial slope of 0.0069. For eachsection, the initial channel width was calibrated using the Lake Mills reservoir survey data ofGilbert and Link (1995, see above). An annual bedload sediment supply was calculated fromthe reservoir data using volumes of the delta topset and tributary fan deposits. These sed-iments represent the long-term natural bed material supply to Elwha River. We computedbedload sediment transport capacity for the period between 1927 and 1994 using the 2-partcross-section and Gaeuman et al. (2009) equation, and adjusted channel width until the annualaverage sediment yield was within 5% of that measured in the reservoir. The grain size distri-bution used in the calculation is adapted from a composite of bulk subsurface sediment sam-ples of bank toe and collapse deposits collected between Glines Canyon Dam and the Straitof Juan de Fuca. We found that the grain size distributions of coarse-layer bank deposits aresimilar to those of point bar heads within the channel (Figure 3.3). We added a 1% lag of sedi-413.4. Methodsment in the 512-1024 mm size range to account for large boulders observed on riffle crests butnot collected in the bank samples. A comparison of the grain size distributions of the modeledload and reservoir bed material is presented in Appendix B. The two distributions are similar,indicating that the calibration is adequate. Initial channel widths are plotted in Figure 3.4.They represent the width that which is able to transport the observed sediment load given themeasured flow regime and assuming static banks and is roughly analogous to an empericallyderived regime width. All alluvial cells were assigned the same initial channel depth of 1.86m, which was determined using a regional hydraulic geometry relation developed by Castroand Jackson (2001).The initial channel geometry and grain size distribution were used to derive a sedimentrating curve for the upstream boundary condition. We used the dimensions of the downstreamalluvial section (81 m channel width and slope of 0.0069) for numerical stability, althoughcalculations using the upstream alluvial section yields similar results. The rating curve wascalculated using a 2-part channel/floodplain cross-section, using daily discharge data fromthe USGS gauge at McDonald Bridge (Figure 3.1). Qc,i was extracted from the rating curve foreach day.All model cells start with the same initial grain size distributions. The channel GSD isthe same used to calculate incoming bedload supply and calibrate initial channel width (seeabove). The initial floodplain grain size distribution is a function of both channel materialsizes and the thickness of fine overbank deposits. The overbank thickness on the floodplainwas estimated from measurements of bank stratigraphy collected along eroding banks duringlow flow (see Appendix C). The thickness of bed material in the banks was calculated as thedifference between the channel depth and overbank thickness and was held as a constant.MAST-1D calculates the inital floodplain grain size so that the system is at perfect steady statebetween lateral deposition and lateral erosion in all size classes, given duration-averaged flowdata and a constant migration rate (see the Appendix A for details).A detailed list of the initial conditions and model parameters, as well a description of thecalibration procedure, can be found in Appendix B.3.4.2 Model confirmationAirphoto analysisIn order to assess the performance of MAST-1D, channel width and rates of channel-floodplainexchange were calculated from aerial photographs. Channel and valley margins were delin-eated. Table 3.3 lists the available photos. The channel is defined here as unvegetated surfaceswithin the floodplain (wetted channel plus bars and visible paleochannels), which providedconsistency over photos captured during different discharges. All photos were viewed at aresolution of 1:1500 during the digitization process to ensure consistency. The resolution andquality of the air photos improved over time, and this likely introduced systematic error, as423.4. MethodsFigure 3.3: Composite subsurface grain size distributions of Elwha River cutbank andpoint bar head deposits. Shading represents the standard deviation of all samplesfor each landform.Table 3.3: Air photos used in the analysis with available accompanying dataPhoto date Source R/S RE(m)TE(m)1976** National Park Service* - 7 12.21981 National Park Service* - 15 181994-09-21 USGS DOQ 1:12000 3.9 10.72006-04-01 USDA NAIP 1 m 5 11.22009-10-08 USDA NAIP 1 m 5 11.22013-08-31 USDA NAIP 1 m 5 11.22014-12-03 USGS/National Park Service 0.05 m - -2015-06-04 USGS/National Park Service 0.05 m - -2016-08-11 USGS/National Park Service 0.05 m - -*Air photos digitized by author.**Coverage of air photo does not extend to whole study areaR/S Resolution or scaleRE Registration errorTE Total error (registration + digitization)more floodplain channels are visible in the most recent, high resolution photos (Draut et al.,2011). In addition, identifying the break between vegetated and unvegetated surfaces can bedifficult, both in shadowed areas and on point bars where vegetation density varies. For thesereasons, we conservatively estimate that digitization error is 10 m. To calculate total error (TE),433.4. Methodswe follow the method of Draut et al. (2008), whereTE =√RE2 + DE2 (3.3)RE is the registration error and DE is the digitization error. TE is listed in Table 3.3 for photoswhere the registration error is available. Floodplain margins were delineated using a combi-nation of air photos, digitized geological maps (Tabor and Cady, 1978), and, where available,LiDAR data.The digitized air photos were divided into segments that aligned with the MAST-1D modelcells. Channel and valley centerlines were produced for each photo. The sinuosity of each riversegment was calculated for each photo, and the average of all available years was input intoMAST-1D. Channel widths for each segment were calculated as the quotient of the channelarea and centerline length on each photo. The area of floodplain eroded or created betweensets of air photos was measured by using the Union tool in ESRI ArcMap. No attempt wasmade to distinguish channel widening via bank erosion in the air photos from new channelcreated following an avulsion or activation of a floodplain channel. In other words, an avul-sion that results in a bare gravel/cobble surface would appear to have led to widening in ouranalysis. Therefore, when comparing the output of MAST-1D to channel change measuredfrom the air photos, the area of new channel creation calculated from the air photos is equiv-alent to the sum of the areas of modeled bank erosion and avulsed channel, and the area ofvegetation encroachment is the sum of areas of channel narrowing and avulsion.3.4.3 Other confirmation dataOther field data were collected and acquired from outside sources to calibrate and verifythe model. Wolman pebble counts of point bar head deposits collected near the end of thePre-Removal period were provided by the US Geological Survey (USGS) and National Atmo-spheric and Oceanic Administration (NOAA). To quantify Post-Removal particle sizes, digitalphotographs of point bar heads were taken using a GoPro camera in September-October 2015.Particle size information was extracted from the images at Seattle University using DigitalGravelometer software. The size distributions extracted from the photos were truncated atabout 32 mm as recommended by the software. Details on the field and lab procedure areprovided in Appendix C.Details on the post-removal sediment budget through the 2016 water year were providedby the USGS. This information included sediment flux exiting the Lake Mills reservoir, storagewithin the sediment area, and flux out of the reach. In addition, reach storage between 2011and 2013 was provided in East et al. (2015). Details on the calculation methods are found inEast et al. (2015). The sediment budget is presented in metric tons. To convert the MAST-1Dvolumes (in m3) into tons, we assume a sediment density of 2.7 tm−3.443.5. ResultsFigure 3.4: Channel width plotted against channel coordinate as measured from air pho-tos and calculated in MAST-1D for a) 1981, b) 2009, and c) 2016. The initial modelcondition (i.e. channel width in 1919) is plotted in orange for comparison. ‘X’ marksrefer to bedrock canyons and were modeled with fixed banks in MAST-1D. GlinesCanyon Dam is at river km 21.5.3.5 ResultsThe results are divided into two parts. First, MAST-1D output is compared to the air photo andfield data to assess the performance of the model. Then, rates of modeled channel evolutionare presented.453.5. Results3.5.1 Model performanceHere we compare evolution of Elwha River simulated by MAST-1D to field and air photodata to assess its success in replicating observed spatial and temporal patterns of geomorphicchange. We focus on metrics of channel width, widening and vegetation encroachment, grainsize, and channel storage. Modeled and measured channel width as a function of channelcoordinate is presented in Figure 3.4 for select years, along with the initial model condition forcomparison. MAST-1D replicates the spatial pattern well for the Pre-Removal period (Figures3.4b and c). While width is overestimated for most reaches in 1981, the longitudinal trendholds over the majority of the study channel. MAST-1D does not perform well between riverkms 12.5 and 15, where a bedrock canyon around river km 13.5 and the upstream end of theLake Aldwell delta affect hydraulics and channel evolution. During the Post-Removal period(Figure 3.4d), all runs show similar trends in width, which match the air photo data in mostreaches, with exceptions being at river km 18, where MAST-1D neglected to predict a 60 mincrease in width, and at river km 14.5, just upstream of a bedrock canyon.The creation of new channel falls under two categories: bank erosion in the current channeland activation/re-activation of former floodplain channels. Both processes were relevant onElwha River, especially following dam removal. Channel outlines derived from the airphotosare presented in Figure 3.5 for the post-removal period. Most channel change occurred in theupstream half of the study area, which was characterized both by creation of new channelvia re-activation of floodplain surfaces and by widening of the main channel. In the down-stream half of the study area, bank erosion occurred, but the channel remained primarilysingle-threaded.Widening (bank erosion plus avulsion) rates from run R3 is compared to net change mea-sured from the air photos in Figure 3.6 (the trends were similar for all dam removal scenarios).Each point represents the spatially-averaged annual rate of channel widening or narrowingover the time interval between sequential air photos, which ranges between under 1 to 13years (see Table 3.3). Error bars denote the standard deviation of rates for each model cell/airphoto reach, which range between 240 and 918 m in valley length. The predictive successof MAST-1D is very different for widening and narrowing. While there is a large amount ofscatter in the spatial variability of widening, the majority of the rates are within an order ofmagnitude of aerial photo measurements, and a trend between modeled and measured datais visible. MAST-1D overpredicts channel widening for most years. Errors are smaller for thePre-Removal period than they are Post-Removal; the percent difference for the former rangesbetween 7 and 77%, while modeled Post-Removal rates are 26-121% different from estimatesfrom the air photos.Our implementation of MAST-1D does not replicate channel narrowing with high suc-cess (Figure 3.6b). While Pre-Removal spatially-averaged narrowing rates measured from theaerial photos span nearly two orders of magnitude, modeled rates all fall between 1 and 3m/yr. In addition, while there is a lot of spatial variability in air photo measurements (as463.5. ResultsFigure 3.5: Channel outlines for Elwha River just before (2009) and following dam re-movaldepicted by the large horizontal error bars), MAST-1D predicts very little longitudinal vari-ability in vegetation encroachment. The model predicts much higher spatial variability inPost-Removal narrowing than the air photos show, mainly due to a higher occurrence of nar-rowing due to avulsion than is observed in the latter.The MAST-1D simulation adequately replicated bed coarsening due to sediment starva-tion following the closure of Glines Canyon Dam and fining due to the progradation of theLake Aldwell delta at river km 12 during the Pre-Removal period (Figure 3.7a). Apart fromthe bedrock canyons at the upstream end of the study area (river kms 19-21), where the D50 isunderestimated by over a factor of 4, model output is within the range of USGS and NOAA473.5. ResultsFigure 3.6: The relationship between a) annual widening (bank erosion plus new channelcreation) and b) vegetation encroachment as predicted in model run R3 vs thosemeasured from air photos. Each marker represents the mean rate of all model cellsover one pair of dates, while the error bars delineate the standard deviation. Someerror bars have been cut off for display. Dotted lines are one order of magnitudeaway from the 1:1 (solid) line.pebble counts. As is apparent from Figure 3.7b, the truncated 2015 D50 for all three MAST-1Dpredictions is slightly lower than that calculated from photosieving for the upstream reaches.At the downstream half of the study area, runs R1 and R3 are within the lower end of ob-served values and predict the recovery of bed material grain size following the sediment pulsereleased from Lake Mills.Total (channel plus floodplain) sediment storage between the former Glines Canyon Damand the upstream end of the former Lake Aldwell deposit during the Post-Removal period isplotted for runs R1-R3 in Figure 3.8 along with field and remotely sensed data from the USGSand US Bureau of Reclamation. All three modeled post-removal sediment supply scenariosoverpredict reach storage by at least a factor of 4. However, in R1 and R3, much of the storedmaterial consists of fine sediment sourced from the suspended load, which has little influenceon channel geomorphology. While the general temporal pattern of sediment storage is similarfor all runs, the model is highly sensitive to the supply of bed material. Total storage scaleswith the proportion of the incoming sediment pulse consisting of bedload.According to the field data, channel storage increased between 2012 and 2014, then de-creased slightly between 2014 and 2016. All three model runs show an increase in storage upto 2014, but none are able to replicate the subsequent net export of sediment. R3 providesthe best fit to the data; the projected amount of sediment exported from the study area is 13million t, 20% lower than the USGS/USBR calculations, but simulated bed material storage iswithin the error range in 2013 and just above it in 2016.483.5. ResultsFigure 3.7: D50 for a) the end of the Pre-Removal period and b) 2015, four years after damremoval. Markers represent a) Wolman counts provided by the USGS and NOAAand b) photosieved samples. MAST-1D output was truncated at 32 mm in b) to makeit comparable to the photosieved samples.Figure 3.8: Total (channel and floodplain) storage within the study area. For the modelruns, solid lines represent storage for all material (fine material and bed material),while the dotted lines show storage of bed material (sizes >0.5 mm) only. Storagemeasured from field data (black line) is sourced from East et al. (2015) for 2013 andfrom unpublished data from the USGS/USBR for all other dates. It represents totalstorage of all size classes. Uncertainty estimates, where available, are represented byerror bars. The contribution of bed material to total storage has not been quantifiedin the field.493.5. Results3.5.2 Channel evolutionHere we present output from the MAST-1D simulations that show the effect of sediment sup-ply on channel evolution. In runs R1-R3, the sediment supply regime is impacted by damemplacement and removal. Run C4 represents the control case where there is no disturbanceto sediment supply. We focus on the surface particle size and channel widening and narrowingprocesses.The sediment supply regime imposed by the dam impacted the transport rate and calibreof bed surface material (Figure 3.9). A comparison of the annual sediment yield between thedammed and control run for the Pre-Glines and Pre-Removal period (Figure 3.9a) reveals that,as expected, the removal of upstream sediment supply following dam emplacement leads to areduction in the sediment load. It takes about 15 years for the sediment regime to adjust, afterwhich the yield for the dammed simulation remains under 20% of the control run. The channelsurface particle size evolves on a similar timescale (Figure 3.9c); the dammed reach coarsenedrapidly until 1935 then continued to increase slowly throughout the rest of the run, while thecontrol fluctuated within a range of less than 10 mm throughout the entire run, fining slightlyafter the 1970s.During the Post-Removal period (Figure 3.9b and d), the sediment yield declined expo-nentially, mimicking the sediment supply for R1 and R3. The total mass of sediment exitingthe modeled channel is inversely proportional to the amount of bedload supplied; runs R1and R3 each experience yields approaching 2.5 million m3, while yields in R2 are less than halfthat. The particle size for all three runs drops into the sand/fine gravel range during the first(fine) sediment pulse, then coarsens during the second pulse. While the D50 for runs R1 andR3 approach that of the control run by 2015, R2 does not recover.Bank erosion and vegetation encroachment as calculated by the model were impacted bysediment supply disturbances caused by the dam. Average rates for runs R3 and C4 are pre-sented in Figure 3.10. The trends for the other runs are similar to R3. The bank erosion ratesfor both runs are variable on an annual timescale and are dependent on the magnitude ofpeak flow events. They were highest during the 1970s-1990s, when a greater proportion oflarge flood events occurred compared to the mid-20th century. After about 1945, slightly lesserosion is predicted during any given flow in R3 than in C4 during the Pre-Removal period.Annually-averaged erosion rates are compared with annual peak daily discharges in Fig-ure 3.11 (runs R3 and C4 are presented). The control data (grey points) represent the range ofbank erosion that could be expected in a system with sediment supplied at capacity. Duringthe Pre-Removal period, the modeled erosion rates are within the range of those expected in asystem that is not supply-limited, despite the fact that rates in R3 are lower for any given flood(Figure 3.10). However, avulsion is much less common in the sediment starved system. Time-averaged annual widening rates are presented in Figure 3.12. The solid teal series representthe total widening rate, which can be divided into widening via bank erosion (the magentadotted series) and avulsion. In the dam run (R3), nearly all widening is due to bank erosion,503.5. ResultsFigure 3.9: Evolution of channel characteristics over time. Magenta series denote the sim-ulations modeling the dams (R1-R3), while teal series represent the control run C4.a) annual sediment yield during the Pre-Glines and Pre-Removal periods, b) yieldfor the Post-Removal period, c) reach-average channel surface D50 for the Pre-Damand Pre-Removal periods, and d) reach-average channel surface D50 for the Post-Removal period. Sediment yield is defined as the total sediment (suspended plusbedload) passing into the upstream end of the former Lake Aldwell reservoir.Figure 3.10: Average annual bank erosion and encroachment rates for simulations withand without dam emplacement/removal. Channel narrowing is represented asnegative for display. Line a. delineates 1927, the year Glines Canyon Dam wasclosed, while line b. denotes the introduction of bedload sediment into the MiddleRiver in late 2012.513.5. ResultsFigure 3.11: Total bank erosion compared to annual peak daily discharge for each wa-ter year. Data are calculated from the 1919-2016, 1927-2011, and 2012-2016 wateryears for the control, pre-removal period, post-removal period, respectively. Pointsrepresent the median bank erosion of all model cells, and error bars delineate thestandard deviation. Run R3 is used for the dam simulations. The trends in theother runs are similar.while avulsion accounts for about a quarter of new channel in C4.Following dam removal , annual erosion rates increase to over 15 m/yr for R3, while theyremain below 5 m/yr for C4 (Figure 3.10). For the first two years following release of the bedmaterial sediment pulses, peak flows were lower than the annual flood. These flows appearto have been near the threshold of bank instability and the erosion rate is within the rangeexpected for the low and at capacity sediment supply regime. (Figure 3.11). Rates are muchhigher for the next two years; spatially averaged bank erosion exceeds 12 m and is within therange expected for flows nearly twice as high in the supply-limited and at-capacity systems .Averaged over the entire Post-Removal period, the bank erosion rates range between 2.5and 11 m/yr, 1.25-5x higher than the control run (Figure 3.12b). In most reaches, rates aresimilar for R1, R2, and R3, suggesting that the proportion of bed material in the pulse has onlya marginal impact on rates of bank erosion. However, the avulsion rate varies both longitu-dinally within individual runs and between them. For all simulations, avulsion rates are at amaximum at the first alluvial model cell downstream of the dam (river km 19.7) then decreaseuntil about river km 16, after which the rates are negligible. The avulsion rate in R2 is overtwice high as in the other two runs; avulsion accounts for nearly all the new channel formationin the upstream half of the channel in that run, while in R1 and R3 avulsion and bank erosionare roughly equally dominant.Annually-averaged rates of channel narrowing (Figure 3.10) exhibit much less year-to-yearvariability than widening. In run C3, simulated annual narrowing rates range between 0 and 4m/yr. Narrowing rates for the dammed and control run begin to diverge about 20 years after523.6. DiscussionFigure 3.12: Temporally-averaged rates of new channel formation a) the Pre-Removal pe-riod b) the Post-Removal period, and c) the control simulation with sediment sup-plied at capacity (run C4). Solid teal series denote the total widening, while bankerosion rates are plotted in dashed magenta.dam emplacement, with the former leveling off at a rate roughly 1 m/yr lower than the latterdue to a narrower channel and less bar space for colonization. The narrowing rate increasesfollowing dam removal, but at a lower magnitude than bank erosion.3.6 DiscussionThe closure of Glines Canyon Dam in 1927 marked the beginning of a regime shift for thedownstream reaches of Elwha River from a system that was connected to frequent sourcesof sediment to one that was sediment-starved and geomorphically stable. Following dam re-moval, the river has both responded to a large disturbance–in the form of a 16-million tonsediment pulse–and begun its transition into a new, more supply-rich regime. Our goal was tocharacterize these transitions by identifying the key spatial and temporal adjustments to chan-nel stability. By using MAST-1D to simulate different sediment supply scenarios, we were ableto compare the modeled evolution of Elwha River to a hypothetical scenario where sedimentsupply remained consistent through time to quantify the effect of supply on geomorphic be-havior. Numerical modeling is also useful in that it can shed light on the relative importanceof various channel behaviors which may be fundamentally different in process but lead to533.6. Discussionsimilar landform adjustments. Identifying areas where MAST-1D is successful in replicatingfield data and areas where it fails leads to insight on the range of geomorphic responses tovariability in sediment supply.3.6.1 MAST-1D confirmation–successes and failuresOne fundamental assumption of MAST-1D (and other 1D models) is that channel evolutioncan be characterized by reach-average sediment fluxes that are calculated using channel-averagehydraulics. This simplified representation appears adequate for the Pre-Removal period. Themodel predicted channel D50 values that are within the range of field data for all reaches ex-cept for a series of bedrock canyons immediately downstream of the dam (Figure 3.7). Thecanyons have slopes and channel morphologies outside of the range of what the sedimenttransport equation was designed for.MAST-1D also successfully replicated pre-removal channel width (Figure 3.4) and, to alesser extent, channel widening (Figure 3.6). Scatter between model and air photo data inthe latter may be partially due to systematic error in the air photo measurements. Rivers aremore likely to reoccupy portions of channel that were recently abandoned, meaning that overdecadal timescales the net movement of the channel margin is less than the total movement(Konrad, 2012). Comparing channel margins between sets of aerial photographs generates thenet channel change, while MAST-1D is only able to calculate the total change. Therefore, allelse being equal, the difference between narrowing and widening for the two methods shouldincrease over time. To test this, we plotted the average migration rate calculated via air photosas a function of the length of time between sequential air photos (Figure 3.13). The photosdepict Elwha River upstream of Glines Canyon Dam and were chosen because the patternof sediment supply in the study area (between the dams) is similar to that which we expectto observe from the systematic error. Details on these photos can be found in Appendix D.There appears to be a weak exponential relationship between the rate of channel movementand the length of time between air photos, especially with regards to channel widening. Thesignificance of the relations cannot be tested because they are non-linear.If this error is significant over the reach between the two dams, then MAST-1D shouldoverpredict rates of widening and narrowing compared to the air photos during periods withlong gaps between photos. Residuals of the data in Figure 3.6 are plotted in Figure 3.14. Theyare exclusively negative for air photo measurements spanning four years or more, as wouldbe expected if channel and floodplain deposits reoccupied their former territory between airphotos. However, there is no relation for photos taken fewer than four years apart. Whileusing rates of channel change calculated from aerial photography to calibrate MAST-1D prob-ably introduces systematic error, it is likely small in comparison to the uncertainty related tocharacterizing sediment supply.MAST-1D replicated channel characteristics in the Post-Removal period with limited suc-cess. Our air photo analysis shows that Elwha River widened in two ways following the initial543.6. DiscussionFigure 3.13: Annual rates of channel widening and narrowing plotted against the lengthof time between air photo pairs. Markers show the mean rate for all model cell-s/reaches and error bars mark the standard deviation. Regression lines were fittedwith least-squares non-linear regression to the mean values.Figure 3.14: Residuals between modeled rates of channel widening and narrowing andthose measured from air photos for the study area. Magenta circles represent chan-nel narrowing, while teal triangles show widening.pulse of bedload sediment released past Glines Canyon Dam in 2012 (Figure 3.5). Aggrada-tion in the main channel caused flow and sediment to be diverted into floodplain channels,reactivating them (East et al., 2015). Then, significant bank erosion occurred following low-moderate flood events in 2014-2015. Most channel widening–both via floodplain activationand bank erosion–occurred in the alluvial reaches in the upstream half of the study area (riverkms 17-20). While width increased 10-20 m downstream of river km 17 (and about 60 m justupstream of the Lake Aldwell delta), the channel remained primarily single-threaded. MAST-1D captures the correct spatial pattern. The equivalent of floodplain activation in the modelis avulsion. It is initiated by the same process–channel aggradation–although in MAST-1D,avulsion does not immediately lead to net width change, because the assumption is made that553.6. Discussionactivation of floodplain surfaces leads to an equal amount of channel abandonment. In thefield, post-removal floodplain activation was not necessarily accompanied by an equal rate ofchannel abandonment. For the most part, channel width five years after removal was initi-ated is predicted correctly (Figure 3.4), although the contribution of bank erosion to channelwidening is overestimated in reaches with significant floodplain channel reactivation.The model performs poorly in two places. The first is at river km 13, where MAST-1Doverpredicts width by an order of 2. The reach is just upstream of a bedrock canyon, anda small part of its outer bend flows adjacent to a bedrock outcrop. This may have limitedbank erosion. The model also fails near river km 18 where sinuosity increased following damremoval.Reachwide, MAST-1D overpredicts the total amount of post-removal widening up to 2016and fails to capture the timing (hence the scatter in Figure 3.6). It also underestimates theability of Elwha River to export the pulse of sediment out of the system, leading to levels ofsystem storage that are too high (Figure 3.8) and a grain size distribution slightly too fine inthe upstream upstream portion of the channel (Figure 3.7). The severity of this failing dependson whether excess sediment stored in the modeled system is composed primarily of washloador bed material. In MAST-1D, suspended sediment cannot be deposited in the channel; anystorage of washload material is constrained to the floodplain overbank zone. As the floodplainreservior is very large, it should be able to sequester the excess sediment without having asignificant impact on modeled rates of channel change. Underestimating the competence ofthe channel to evacuate bed material may have a large impact on the modeled measures ofstability. Channel storage leads to aggradation, and thus more avulsion. Finer sediment inthe channel would also lead to increased sediment mobility on the near-bank (Wilcock andCrowe, 2003), presumably increasing bank erosion. These processes may partially explain theoverestimation of widening presented in Figure 3.6.Part of the reason for the discrepancy between modeled and measured storage may bebecause the sediment transport equation we selected (Gaeuman et al., 2009) was calibrated to areach downstream from a dam, and may be ill-equipped to deal with large sand loads. It is alsopossible that cross-section variability not accounted for in MAST-1D impacts post-removalevolution. The model does not account for the influence of channel morphology, which willhave a first-order impact on sediment transport dynamics. Flow concentration in the thalwegwill locally increase the sediment transport rate and may increase the total competence of thechannel. Ferguson (2003) estimates that gravel bed rivers with one or more deep thalwegs mayhave transport rates at least five times higher than rectangular channels because of the non-linear relationship between shear stress and sediment transport rate. Flow concentration alongthalwegs will also increase shear stress along the channel margin, leading higher rates of bankerosion. This effect is enhanced by deposition of bars on the opposite bank. In fact, part of thereason MAST-1D underestimates grain size several years after dam removal may be becauseit does not account for the role that channel morphology plays in partitioning sediment into563.6. Discussioncoarse patches on bar heads and in the thalweg and fine patches on bar tails.It is more likely that the discrepancy between modeled and measured storage is due to aninadequate representation of sediment supply at the upstream boundary condition. The pri-mary source of error in our model is due to the large uncertainty surrounding the grain size ofsediment supplied from Lake Mills. Randle et al. (2015) estimated that 71% of the material evac-uated from the former Lake Mills was sand size or larger between 2012 and 2013. However,it is unclear how much of the sand traveled in suspension. Our simulations are very sensitiveto the partitioning of supply between the bed and the suspended loads because the former ismore likely to end up stored within the system. The results presented in Figure 3.8 suggestthat most material traveled in suspension; run R3, for which we assumed that 84% of the totalload is suspended, matched the field data best. In fact, if we assume that most sediment storedin the mainstem and floodplain channels is composed of bed material, then R3 replicates thesystem quite well; bed material storage is only slightly higher than the range of variability inthe field data.The error bars around the field data presented in Figure 3.8 may underestimate the actualuncertainty regarding sediment storage in the study area. Warrick et al. (2015) used the reachstorage analysis performed by East et al. (2015), among other data, to compile a sedimentbudget for Elwha River between 2011 and 2013. They found a 1.7 Mt disparity between mea-surements of sediment stored between Glines Canyon and Elwha Dams (the East et al. (2015)data) and calculations of storage for the same area derived from DEM differencing estimates(for incoming supply from the reservoir) and sediment monitoring stations (for incoming fluxupstream of Glines Canyon dam and outgoing flux downstream of the dams). Sediment fluxout of the study area as calculated using the East et al. (2015) data is 40% higher than measuredat the downstream gauge. A new range of error reflecting this discrepancy is shaded in greyin Figure 3.15b. We assumed that it remained at 40% for all years. Total storage from R1 fitswell within the range. Warrick et al. (2015) suggests that the discrepancy relates to error in datafrom the sediment monitoring stations rather than the East et al. (2015) data. Still, the possibil-ity remains that at least part of the disparity between modeled and measured storage is due tounderestimation of the latter.To our knowledge, no data apart from that presented in Randle et al. (2015) exists regardingthe caliber of sediment supplied to Elwha River downstream of Glines Canyon Dam. In orderto assess the sensitivity of the model to the grain size distribution of the sediment supply, were-ran simulation R1 with a slightly finer sediment pulse (hereafter termed ‘Finer GSD’). Themethod described in Section 3.4.1 was used to determine the upstream boundary condition.However, instead of inputing the size distribution derived from bank material (Figure 3.3) intothe sediment transport capacity calculation, we used the distribution from a bulk sample col-lected on a point bar head just upstream of the former Glines Canyon Dam (Figure 3.15a; alsosee Appendix C). The values for C1 and C2 input into Equation 4.2 were 44 and 17, respectively.The fractions of sand and fine gravel supplied to the system was similar for ‘Finer GSD’ and573.6. DiscussionFigure 3.15: Impact of sediment supply caliber on storage within the study area. a. Sub-surface grain size distribution for the composite cutbank sample (black solid line)and sample of a point bar head immediately upstream of the Glines Canyon Damcollected in September 2015 (solid magenta line). Dotted lines represent the time-averaged GSD of upstream sediment supply. The dotted black line refers to modelrun R3, the dotted magenta line to run ‘Finer GSD’ and the dashed magenta lineto ‘Finer pulse.’ b. Sediment storage. ‘Finer GSD’ represents a run that was iden-tical to R3, except a finer grain size distribution was used to calibrate the sedimentpulse. ‘Finer pulse’ is similar to the ‘Finer GSD’ run, except the grainsize distri-bution of the upstream boundary condition was truncated at 8 mm for the entirepost-removal period, instead of just the first year. See Figure 3.8 and the text forfurther details.R3, but the fractions of coarse gravel and cobble in the former were slightly finer (Figure 3.15a).This difference between the two runs is marginal in terms of channel storage (Figure 3.15b).However, if we assume that the sediment load for ‘Finer GSD’ remains truncated at 8 mm (andC1 is used in Equation 4.2 for the entire duration of the run), then the temporal pattern of bedmaterial storage closely resembles that of the field data. This seems to suggest that, at leastfor the first five years following dam removal, bed material evacuated from the former LakeMills reservoir was primarily composed of sand and fine gravel, although cobble-size chunksof concrete from Glines Canyon Dam were observed throughout the study area. Constrainingupstream sediment supply is crucial for determining whether our sediment flux calculationsare reliable.Another impediment to modeling width change on decadal timescales appears to be thelack of a reliable, simple framework for characterizing vegetation dynamics. For MAST-1D,the assumption was made that narrowing occurred at low shear stresses. While the overalltimescale of response to the emplacement of Glines Canyon Dam seems reasonable comparedto the response time of other systems (see below), the narrowing function was not able to583.6. Discussionreplicate rates of vegetation encroachment measured from the aerial photographs (Figure 3.6).3.6.2 Impact of sediment supply on channel stabilityBank erosion and avulsionThe MAST-1D modeling suggests that upstream sediment supply has a first-order impact onchannel stability. Rates of both bank erosion and avulsion were lower in the runs simulatingdam emplacement compared to the control run (Figures 3.11 and 3.12). A coarsening of thebed texture during the Pre-Removal period (Figure 3.9c) reduces the mobility of near-bankchannel sediment, so that the armor layer protecting bank toe deposits are less likely to bebroken and that, if removed, the flow is not able to move as much sediment away from thebank. This means that large flow events contribute to less bank erosion (Figure 3.10). Whilethis response is conditioned by the way bank erosion is characterized in the model, Konradet al. (2011) made a similar observation while analyzing aerial photos taken before and afteremplacement of a dam on Green River. They found that the discharges that best correlatedwith floodplain turnover increased following emplacement of the dam and suggested thathigher flows were necessary to destabilize the banks because the bed had armored. However,damming on the Green River led to a reduction in peak flows, which Konrad et al. (2011) sug-gest were more important in reducing migration rates than armoring. The results depictedin Figure 3.10 suggest that a reduction in bank erosion can occur even when the hydrographundergoes minimal alteration.Sediment starvation also lowers the magnitude of channel deposition, which decreases thefrequency of avulsion (Figure 3.12). Since new channels act as a source of sediment supply,a reduction in avulsion further decreases the amount of available sediment to the reach. Ourmodeling suggests that emplacement of Glines Canyon Dam led to increased channel stabilityon Elwha River by armoring the channel (therefore leading to lower mobility in the near-bankregion) and by reducing the occurrence of destabilizing depositional features in the channel.Both of these processes act as a positive feedback, because they further reduce sediment supplyfrom the floodplain. Our findings complement those of Draut et al. (2011), who observedthat Elwha River showed little geomorphic response just downstream of the Elwha Dam toa 40 year flood in 2007, even though the same flood led to significant channel change bothupstream of both dams and near the river mouth, where till bluffs on the channel margins actas an additional source of sediment supply.The relationship between sediment supply and channel widening is less straightforwardduring the Post-Removal period. The pulse of sediment reduced grainsize (Figure 3.9), in-creasing the mobility of the near-bank region (Figure 3.11 and leading to channel widening(Figure 3.4). However, as discussed above, the processes by which this occurred are morecomplicated than can be explained by reach-wide patterns of channel mobility and deposition,and they are highly sensitive to the upstream boundary conditions. Our modeling suggests593.6. Discussionthat the river is most sensitive to incoming sediment supply on the alluvial reaches closest tothe dam (Figure 3.12). While comparable increases in rates of bank erosion were experiencedthroughout the study channel, the amount of avulsed channel increased with proximity tothe dam. It was in the upstream half of the study area that the runs R1-R3 differed most; thehigher the supply of bedload, the more avulsion that occurred. This suggests that the reachesclosest to the dam act as the primary sinks for excess bed material sediment. Consistent de-position will lead to more avulsion but also keep bank heights relatively low, so that floodingis more frequent. This partitioning of shear stress between the channel and floodplain furtherdecreases the transport capacity of the channel, acting as a positive feedback mechanism thatencourages more deposition.The proximity of alluvial channel reaches to the source of the sediment pulse likely impactsthe influence of the pulse on channel stability. Of particular importance, as suggested byFigure 3.12, is the presence of active floodplain surfaces accessible via avulsion, which canact as loci of deposition. This may partially explain why the response of the banks to damremovals has been less noticeable (or not mentioned) on other mountain rivers. Bank erosionwas not reported on the White Salmon River up to two years following removal of Condit Dam(Wilcox et al., 2014) or on the Sandy River after Marmot Dam was removed (Major et al., 2012).Both of these rivers flow through at least 3 km of bedrock canyon, which may have acted as abuffer to the initial sediment pulse. Elwha River flows through broad alluvial valleys within 2km of the dam.Vegetation encroachmentModeled rates of channel narrowing are not as sensitive to sediment supply or decadal-scalehydrologic variability as rates of bank erosion. During the Pre-Removal period, the modelednarrowing rate gradually lowered, eventually stabilizing after about two or three decades (Fig-ure 3.10). It did not begin to significantly diverge from the control run until a couple decadesafter dam emplacement. Drought in the 1930s may have been the primary cause of channelnarrowing during this period. It is also possible that the model is responding to the initialcondition. While there is scant data on channel margins for this period along Elwha River,evidence from other rivers suggests they narrow on a similar timescale. Williams and Wolman(1984) found that channels experiencing narrowing following dam emplacement tended toadjust after a few decades, although they found that most change occurred within the firstdecade. Merritt and Cooper (2000) similarly found that the initial width adjustment took abouta decade. The difference between these studies and Elwha River is that the reductions in chan-nel width in the latter were caused mainly by a temporary increase in the rate of vegetationencroachment due to a reduction in peak flows, which left more channel surface area availablefor colonization. The rate of vegetation encroachment decreases in our simulations as the totalamount of point bar available for colonization decreases. This suggests that channel widthchange on Elwha River, which did not experience major flow alteration, was driven more by603.7. Conclusionstabilization or destabilization of the eroding banks than by any major change in vegetationdynamics.That being said, MAST-1D is not particularly successful in reproducing the rates of vegeta-tion encroachment observed in air photos for the study reach (Figure 3.6a). The model assumesthat channel narrowing occurs solely as a result of low shear stress, which is contradicted bya number of studies (e.g. Bertoldi et al., 2009; Camporeale et al., 2013; Perona et al., 2009). Oneoption is to ignore the dependence of vegetation growth on discharge and set channel nar-rowing as a linear function of available bar space, the approach taken by Davidson (2016) andKonrad (2012). However, this ignores the importance of flow variability–particularly the abilityof low-moderate flood events to scour fresh vegetation (Perona et al., 2012). Perhaps stochasticapproaches, such as that adopted by Camporeale and Ridolfi (2007) can prove useful. In addi-tion, MAST-1D (as well as nearly all bed evolution models) fail to consider the importanceof in-channel wood, whose distribution impacts bar stability, seedling germination, and avul-sion.3.7 ConclusionDecadal-scale numerical modeling of Elwha River demonstrates the impact of sediment sup-ply disturbances imposed by the emplacement and removal of Glines Canyon Dam on channelstability, which we characterize in terms of bank erosion and avulsion. We have shown that thereduction in bed mobility and channel deposition caused by sediment starvation can explainthe reach-scale increase in channel stability following emplacement of the dam. It is clear thatthe sediment pulse following dam removal led to channel instability, which manifested itselfin increases in width and avulsion. But the geomorphic processes responsible for the increasein instability are more complex than those for the decrease. Local effects of channel morphol-ogy on the flow field likely play just as large a role in initiating bank erosion and avulsion asreach-wide patterns of bed mobility and channel storage.Dam removal is becoming an increasingly popular form of river restoration. Decadal-scalenumerical experiments of dam-influenced systems like Elwha River offer valuable insight onhow to interpret geomorphic adjustments following large-scale anthropogenic disturbancesand can identify missing pieces of the puzzle. Our modeling efforts have suggested that quan-tifying the caliber of sediment supply, and how it evolves over time, should be prioritized,as uncertainty related to sediment mobility precluded us from explaining observed patternsin sediment yield. In addition, while our ability to predict bank erosion on the reach scaleis improving, much work is still needed to successfully quantify vegetation processes. On awhole, this study demonstrates the importance of considering the impact of sediment supplyon lateral exchange processes, particularly on decadal timescales.61Chapter 4Elwha in the 21st century: can we usethe past to predict the future?4.1 SummaryThe question of whether or not Elwha River, and other medium-large streams undergoingdam removal, will evolve to resemble 20th century rivers not affected by direct anthropogenicalteration is relevant to water resource managers. We consider two questions: how will ElwhaRiver respond to the pulse of sediment following dam removal on decadal timescales? andis it appropriate to assume hydrologic stationarity in light of decadal-scale climatic variabilityand ongoing climate change?We present a brief literature review of decadal-scale climatic and sediment supply regimesin the Pacific Northwest to put Elwha River in context. Elwha is a rainfall-dominated hybridstream with a secondary nival peak that varies in strength over time. Three large-scale climatephases have been identified in Elwha River discharge data between 1925 and 2016 that alignwell with phases of the Pacific Decadal Oscillation and with regional trends in flood magni-tude. Phase 1, which lasts between 1925 and 1946, is characterized by low annual water yieldsand flood events. During Phase 2 (1947-1976), abundant precipitation and cool temperatureslead to a hydrologic regime characterized by high overall water yield and a strong snowmeltperiod in the spring. In Period 3 (1977-2016), the strength of the nival period decreases andoverall water yield decreases, with a concurrent increase in the magnitude of flood events.Pacific Northwest rivers have high sediment loads due to abundant supply from upliftingmountains, relict glacial sediments, and, locally, extensive floodplains. In high-order trunkstreams, supply can be considered primarily continuous, but high-magnitude events may re-lease pulses into the system.We use a Monte Carlo modeling approach to explore how dam removal and hydrologicregimes affect sediment transport and channel stability over the course of decades. Simula-tions with sediment supply regimes representing dam emplacement and removal are com-624.2. Introductionpared to model runs in which sediment was supplied at capacity. While the channel grainsizeof the post-removal channel recovered within a few decades, channel width and migrationrate were significantly higher than the capacity run because the pulse of sediment releasedduring the dam removal deposited in the floodplain, creating a fine sediment reservoir thatwas more easily erodible. Both sediment transport and channel stability were sensitive to cli-mate, with Period 3 being more flood-driven than Periods 1 and 2. Channel stability is linkedto the strength of the snowmelt season. In hybrid streams with abundant sediment supply andcoarse banks, nival flows are able to efficiently transport sediment without leading to bank in-stability. With declining snowpack projected for the future, we expect the strength of the nivalflow to decrease on Elwha River, making it even more laterally unstable and flood-driven.Using channel measurements from regional rivers dating to the 20th century will likely un-derestimate future channel width and migration.4.2 IntroductionIn many resource management and restoration projects, it is necessary to predict future sed-iment loads and bank stability in order to plan for flood and erosion risk and to maintainecosystem health. Often, projects involve catchments in which the sediment supply and/orthe flow regime are subject to anthropogenic alteration, so that the previous sediment trans-port regime will not be a good predictor of future behavior. Elwha River represents one suchcase. After almost 100 years of sediment starvation following emplacement of Glines CanyonDam, more than 9 million tons of sediment were released during the largest dam removal inhistory (Warrick et al., 2015). One of the important questions–both on Elwha River and forother dam removal projects–is how long the pulse of sediment will affect channel dynamics.Analysis of recent medium to large scale dam removals in the Pacific Northwest seem tosuggest that the majority of reservoir sediment can be evacuated without extreme aggrada-tion downstream. The White Salmon River, Washington aggraded when Condit dam wasremoved, but the channel subsequently incised nearly to its pre-removal elevation within 15days (Wilcox et al., 2014). Aggradation downstream of Sandy River in Oregon was confined tothe first 2 km downstream of Marmot Dam, and deposition elsewhere was limited primarily topool-filling and bar accretion (Major et al., 2012). Accumulation was also limited to pools alongthe Rogue River, Oregon, following removal of the Savage Rapids Dam (Bountry et al., 2013).In the latter two cases, at least half of the reservoir sediment was evacuated during the yearfollowing the removal, even though no peak flow events reached the 2-yr recurrence interval.This has led Major et al. (2012) to suggest that the sequencing of flow events does not have asignificant effect on the evolution of the reservoir, which is largely driven by slope. Indeed,based on elevation measurements taken on the channel thalweg, Elwha River recovered fromthe initial pulse of sediment within two years, despite abnormally low peak flows during thedam removal (East et al., 2015).However, on larger rivers with active floodplains, there is evidence that sediment pulses634.2. Introductiontravel more slowly. For example, Czuba et al. (2012) suggests that a pulse of sediment origi-nating from a series of rockfalls and debris avalanches on Mt. Rainier in 1963 is still causingaggradation along the White River in Washington. Since dam removal (especially on mediumto large streams) is a recent restoration technique, the impact of former reservoir deposits onlong-term sediment supply is still unclear.In addition, predicting future channel behavior is confounded by climatic variability. Sed-iment transport is dependent on the amount of excess force applied to the river bed and banksby the flow. Therefore, determining the characteristic hydrologic regime is an essential stepfor assessing any fluvial system. It is common practice to use relations developed from flowfrequency analyses to predict future discharge conditions. Frequency anaysis is often basedon the assumption of stationarity, the condition in which the mean and variance of the datarecord are constant through time. The magnitude of annual peak flows are treated as indepen-dent events, when in reality persistent climatic cycles and trends influence the magnitude andfrequency of floods.Hydrologists acknowledge that flood frequency records are non-stationary on decadaltimescales because of natural variability and global climate change (Khaliq et al., 2006; Millyet al., 2008; Salas and Obeysekera, 2014). Some researchers (e.g. Kiem et al., 2003; Cunderlik andBurn, 2003; Whitfield et al., 2010, c.f. Villarini et al., 2009) warn against extrapolating recordsencompassing only a single climatic phase or using historical data to predict flow in futureclimates. However, others argue that stationarity is an acceptable assumption given the un-certainty in flow frequency analysis (e.g. Serinaldi and Kilsby, 2015). Matalas (1997) defends theusefulness of stationarity by noting that while the frequency of events is likely to change inthe future, their magnitudes will not. But what are the implications of non-stationary flowfrequencies on sediment transport regimes? While climate variability has been acknowledgedin the context of bed material sediment transport (McLean and Church, 1999), we are not awareof a study that quantifies its effect, even though bed material transport affects channel mor-phology. Ferrer-Boix and Hassan (2015) have found that, in an experimental channel, the fre-quency of large discharge events affects the amount of fine sediment winnowing and thus thesediment transport rate. Flood frequencies also characterize the effective discharge, a geomor-phologic metric often used in place of a full hydrograph (Basso et al., 2015).The purpose of this chapter is to speculate on whether Elwha River will evolve to resembleregional rivers not subject to direct anthropogenic disturbance. In order to do so, we mustconsider both the future sediment supply and flow regimes. This leads to two questions: whatis the decadal-scale impact of dam removal on Elwha River? and does climatic variability havea large impact on the sediment transport regime and stability of the channel? Our approach istwofold. First, we contextualize Elwha River with a brief review of decadal-scale climate andsediment supply in the Pacific Northwest. We then use a Monte Carlo approach to examinethe range of geomorphic responses to dam removal as compared to ‘natural’ sediment supplyconditions. We impose observed 20th century climate regimes and reflect on whether future644.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riverssediment transport and channel stability can adequately be predicted using past hydrologicdata.4.3 Elwha River in context: a review of the decadal-scaleclimatology and sediment supply regimes on Pacific Northwestrivers4.3.1 Decadal-scale climate variability in the Pacific Northwest and its effects onhydrologyClimate in the Pacific Northwest (PNW) is driven largely by energy fluxes to and from theNorth Pacific Ocean. Much of the interannual variability is explained by the well-known El-Nin˜o-Southern Oscillation (ENSO) phenomenon, where relatively warm, dry winters are char-acteristic of El Nin˜o years and the opposite occur during La Nin˜a events. A lower-frequencyclimatic pattern was identified in the 1990s (e.g. Trenberth, 1990; Ebbesmeyer et al., 1991; Hareand Francis, 1995; Zhang et al., 1997). Mantua et al. (1997) performed the Emperical Orthog-onal Functions (EOF) technique on a North Pacific sea surface temperature anamoly (SSTA)dataset to indentify individual modes of variability. He found that the first principle com-ponent, which he termed the Pacific Decadal Oscillation (PDO), was correlated with trendsin river discharge and salmon productivity in the Pacific Northwest and Alaska. Since then,the PDO has been used widely as a decadal-scale climatic metric and correlated with regionaltemperature, precipitation, runoff, and streamflow (for reviews, see Mantua and Hare, 2002,and Whitfield et al. 2010). Unlike ENSO, in which El Nin˜os and La Nin˜as occur between pe-riods of ‘neutral’ conditions, the PDO is a bimodal index that switches between two regimes.During positive phases, the PNW is warmer and drier, while a cooler, wetter climate is charac-teristic of negative phases. Only two full cycles have been identified over the modern period;negative phases lasted between the 1890s and 1925 and from 1947-1977, and positive phaseslasted from 1925-1947 and 1977-~1998 (Mantua and Hare, 2002; Zhang et al., 1997).Over the last few years, it has become clear that the PDO signal captures elements of sev-eral coupled atmospheric and oceanic processes, comprising white noise from atmosphericforcing and red noise caused by ocean ‘memory’ (Newman et al., 2003, 2016). The atmosphericforcing comes from stochastic fluctuations of the Aleutian Low, a persistent low pressure sys-tem around the convergence of the Polar and Ferrell Cells, and from atmospheric teleconnec-tions of ENSO. A decadal periodicity emerges partially because these signals are reddenedby the thermal inertia of the ocean and from the seasonal height variation of its mixed sur-face layer. In addition, ocean temperature and pressure are affected by fluctuations at theboundary of the major gyres in the North Pacific, which create oceanic Rossby waves thattake a few years to a decade to cross the ocean (Newman et al., 2016). Despite evidence ofPDO-correlated regime shifts in many biophysical data, including marine ecosystems (Mantuaet al., 1997; Litzow, 2006), snowpack (Mote, 2003), drought (McCabe et al., 2004) and stream-654.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riversflow (Slaymaker, 1972; Mantua et al., 1997; Pagano and Garen, 2005; Whitfield et al., 2010; Fleminget al., 2007; Bowling et al., 2000; Moore, 1996; Reidy-Liermann et al., 2012), climate scientists areincreasingly critical of the notion that phase shifts actually occur in the ocean. The PDO indexonly captures a portion of North Pacific variability (Overland et al., 2008), and some believethat regime shifts are merely coincidences arising from representing multiple processes usinga single metric (Newman et al., 2016). Overland et al. (2008) argues that North Pacific variabilityis inherently random with a regime-like character and autocorrelation that cause multi-yeardeviations from a single century-long mean, but found that the 1945 and 1977 shifts are statis-tically significant.Despite its flaws, the PDO index has been a useful tool for linking climate to decadal vari-ability in streamflow. Temperature appears to be the most important determinant of dischargepatterns during the winter and in low-elevation mountain basins, with warmer periods re-sulting in less precipitation contributing to the snowpack (Liu et al., 2013; Mote, 2006). Moore(1996) found that temperature was the primary control on discharge response to the Paci-fic/North American teleconnection pattern (which is highly correlated to the PDO); duringstrong (warm) phases of the PNA, there was significantly less snow-water equivalent, leadingto more discharge during the winter and weaker spring freshets. The response of precipitationto PDO-captured processes varies more widely than temperature regionally but is relativelyconsistent on local scales, with the response of streams complicated by the interplay betweenglacial dynamics, land use patterns, and hydrologic regime (Whitfield et al., 2010). Fleming et al.(2007) compared discharge patterns between different states of the PDO for many rivers incoastal northwest Washington and southwest British Columbia. They found that the effectsof precipitation anomalies manifest most strongly during the freshet; in pluvial streams, neg-ative PDO years experience more discharge in the winter than positive PDO periods, whilein nival streams the signal is lagged until the snowmelt season. Hybrid streams are affectedduring both seasons and the effect on these systems is the largest, as they are more sensitiveto snowpack fluctuations. In fact, Fleming and his collegues found that some pluvial streamsbecame hybrid systems during the negative PDO because of the accumulation of snowpack athigher elevations.A climatic regime shift occurred between 1976 and 1977, which resulted in the transitionfrom a negative to positive PDO but also seemed to reflect different patterns of oceanic spatialand temporal variability. Bond et al. (2003) did principle component analysis on North PacificSSTA data and found that since the late 1980s, the time series has been dominated by thesecond PC (termed the ‘Victoria Pattern’), not the PDO (the first PC). Fish catch data fromLitzow (2006) shows that, after 1976-77, marine ecosystems in the Gulf of Alaska had a spatio-temporal structure similar to that predicted by Bond et al. (2003). Piechota et al. (1997) foundthat the signature of ENSO events in streamflow records after 1976 are distinguishable fromthose before the regime shift, and Hamlet and Lettenmaier (2007) have suggested that climaticvariability since around 1973 has led to increased variance in runoff and that these trends may664.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riversTable 4.1: Hydrological parameters for three climatic periods.Annual Yield (m3/yr) σ/µ 2-year flood* (m3/s)Period 1 (1925-1946) 1.2*109 5.3*107 310Period 2 (1947-1976) 1.5*109 4.1*107 377Period 3 (1977-2016) 1.3*109 5.7*107 480σ Variance of annual water yieldµ Mean of annual water yield*See Figure 4.3 for details on flood frequency analysis.lead to higher flood risks.Streamflow is also correlated to large-scale increases in temperature that are not (directly)related to the PDO. In general, hemispheric-scale hydrologic models show that snowpack hasdecreased with rising temperatures throughout the Western US since the early 20th century,albeit with some decadal fluctuations (Mote et al., 2005). Principle component analysis ondrought indices by McCabe et al. (2004) suggest that both the PDO and North American tem-perature explain variance during dry periods, corroborating model findings. In the PNW, bothnumerical modeling (Hamlet and Lettenmaier, 2007) and time series analysis (Luce and Holden,2009; Mote, 2003) reveal that increasing temperatures since the mid-20th century have con-tributed to reduced spring snowpack, more runoff in the cool season, and lower dischargesduring the dry season. EOF analysis on outputs from Global Circulation Models (GCMs) withmid-range levels of greenhouse gas emissions suggest that an increase in SST will become theleading principle component in all regions of the North Pacific sometime within the first halfof the 21st century (Overland and Wang, 2007). In the PNW, snowpack will become more sensi-tive to temperature, and April snowpack is expected to decrease up to 70% by the 2080s (Elsneret al., 2010; Mote, 2006). Some lower-elevation nival streams are expected to transition to hy-brid regimes, while other hybrid streams will likely become purely pluvial (Reidy-Liermannet al., 2012).Elwha River hydrologyElwha River drains temperate rainforest on the north face of the Olympic Mountains. Mostof the catchment is contained within Olympic National Park and is unaffected by land usechange. Annual precipitation in the basin varies between ∼5500 mm in the headwaters and∼1500 mm in the rainshadow at the mouth (Munn et al., 1999). Snow is the primary formof precipitation during winter in the headwater areas, but precipitation quickly transitions torain with declining elevation. As such, the river has a rainfall-dominated hybrid hydrologicregime with a bimodal hydrograph (Reidy-Liermann et al., 2012). Peak flows range between∼130 and 1200 m3/s and generally occur in late autumn or winter. A secondary runoff peakoccurs during the spring snowmelt season. Mean discharge over the period of record (1896-2015) is 43 m3/s.Elwha River discharge patterns have been influenced by decadal-scale climatic variability674.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riversFigure 4.1: Metrics of hydrologic change. a. cumulative departure of ElwhaRiver annual water yield from the mean. The three hydrologic peri-ods listed in Table 4.1 are highlighted with shading. Yields were cal-culated from daily discharge data from USGS gauge 12045500 at Mc-Donald Bridge (https://waterdata.usgs.gov/usa/nwis/uv?12045500). b.the monthly Pacific Decadal Oscillation index. Data was downloadedfrom the Joint Institute for the Study of the Atmosphere and Ocean(http://research.jisao.washington.edu/pdo/PDO.latest.txt)(Table 4.1). The cumulative departure from the mean of annual water yield for the period ofcontinuous daily recording, 1919-2016, is presented alongside the PDO index in Figure 4.1.Inflection points correlate well with regional shifts in the PDO, and the record can be dividedinto three periods. During Period 1, which aligns with the positive PDO phase between 1925and 1946, the average annual yield was 1.2*109 m3, 11% lower than the 1925-2016 average. Wa-ter yields in the winter and nival seasons were roughly equal (Figure 4.2a). The flow regimeabruptly shifted between Periods 1 and 2 as the PDO entered a negative phase (see the inflec-tion point in Figure 4.1a). During this phase (1947-1976) average annual water yields were1.5*109 m3, 8% higher than average. Much of the increased flow originated from snowmeltand was released in June (Figure 4.2b). After the climatic shift of 1976-1977, annual yieldsdecreased as a whole (mean yield is 1.3*109 m3/s) but display greater interannual variability(Figure 4.1). Average monthly flow in Period 3 is similar to Period 2 during the winter, butnival flows are considerably lower (Figure 4.2c).Over half of the annual peak flows occurred during December or January in all three cli-matic phases. But the timing and magnitude of flood events differ between the three periods.10-14% of peak flow events occurred during the nival period (April-June) in Periods 1 and2, while all peak events fell between October and March in Period 3. In addition, peak flowmagnitude increases through time (Figure 4.3): the two-year flood is over 100 m3/s higher inPeriod 3 than in Periods 1 and 2 (Table 4.1). Both annual peak flows and annual peak dailyflows in Period 3 are statistically different from Periods 1 and 2 at the 90% confidence levelusing the students’ t test (Period 3 is only different from Period 1 at the 95% level). Peak flows684.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riversFigure 4.2: Monthly mean daily discharge for a. Period 1 (1925-1946), b. Period 2 (1946-1976), and c. Period 3 (1977-2016).Figure 4.3: Flood frequency plot for the three hydrologic periods (refer to Table 4.1).Frequency analysis was done with the US Geological Survey ‘PeakFQ’ software(https://water.usgs.gov/software/PeakFQ/). A log-Pearson type III distributionwas used to define the flood series. Further details on the methods are described inIACWD (1982).during Periods 1 and 2 are not statistically different.In summary, Elwha River is a hybrid stream with peak flows occuring primarily in winterbut with a significant nival period that was slightly stronger before the 1976-1977 climatic shift.Period 1 (1925-1946) is characterized by low water yields and flood events. Discharges duringPeriod 2 (1947-1976) reflect the negative phase of the PDO, and average annual water yieldsare greatest during this period. Flows following 1977 (Period 3) have been more variable withlarger flood events than the other two periods.694.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest rivers4.3.2 Sediment supply of large Pacific Northwestern rivers draining mountainouscatchmentsThe Pacific Northwest is adjacent to a subduction zone at the margin of the North Ameri-can and Juan de Fuca plates. Most streams in the region originate in the tectonically-activeCascade, Olympic, or Coast mountain ranges. High denudation rates caused by uplift are ex-pressed primarily in the frequency of debris flows, which are efficient in transporting sedimentfrom areas of high relief into storage and eventual transport at lower elevations (Montgomeryand Brandon, 2002). In addition, outwash from both Pleistocene-age and modern glaciers addto the high sediment loads in northern portions of the region and in basins with high relief.In many catchments, high-order (trunk) streams have been aggrading in the long term due toincreased sediment supply from easily erodable volcanic material (Czuba et al., 2012; Guthrieet al., 2012; Jordan and Slaymaker, 1991). All of these conditions have created supply-rich streamswhere channels have low armor ratios and have evolved to transport large amounts of sedi-ment (Pfeiffer et al., 2017).Sediment flux entering high-order streams has been filtered by storage zones within thecatchments. Benda and Dunne (1997a,b) used numerical modeling to characterize the natureof sediment routing within channel networks in Oregon. They show that random pulses ofsediment supplied to the stream network via debris flows are conditioned by differential ratesof transport and storage. In high order streams, stochastic pulses originating throughout thebasin merge so that supply is nearly continuous. Local storage zones also provide consistentsources of sediment in rivers with high rates of floodplain turnover (e.g. Beechie et al., 2006;O’Connor et al., 2003) or where rivers erode into bluffs composed of Pleistocene-age glaciallandforms (Draut et al., 2011). However, fluctuations in supply are noticable in trunk streamswhen disturbance events such as forest fires or heavy storm events affect large portions of thebasin (Benda and Dunne, 1997a) or where single sediment pulses are high enough in magnitudeto disrupt the system for a long period of time (Czuba et al., 2012).Large-scale fluctuations in sediment supply may also be correlated to climate variability.More chronic wet and stormy conditions or increases in the occurrence of forest fire could bothincrease the frequency of landslides and debris flows (Benda and Dunne, 1997a; Cannon and De-Graff , 2009), but differences in recharge rate may complicate the relationship (Jakob et al., 2005).Long-term records of debris flows from the Alps show that their frequency has been sensitiveto decadal-scale climate variability in the past, although future temperature and precipita-tion projections for the region favor increases in debris flow magnatude rather than frequency(Stoffel et al., 2008, 2014). Increased flow and exposed sediment resulting from melting glaciersis another possible source, but Czuba et al. (2012) did not find a correlation between glacialretreat and sediment supply on Mt. Rainier. The influence of decadal-scale climate variabilityand change on sediment supply is still largely uncertain, and it is likely than any signals aremasked by the episodic nature of sediment delivery in headwater areas. Naik and Jay (2011)and Inman and Jenkins (1999) have both found that suspended sediment loads during climat-704.3. Elwha River in context: a review of the decadal-scale climatology and sediment supply regimes onPacific Northwest riversically cool periods were larger than during warm phases on the Columbia River and in Cali-fornia rivers, respectively. Both studies relied on sediment rating curves, so the signals reflectchanges in discharge and may have little to do with sediment supply. However, in large, highorder streams, sediment supply sources may be extensive enough that supply correlates moreclosely to discharge than in higher areas of the catchment.Sediment supply on Elwha RiverThe 833 km2 Elwha River basin drains the rugged northern slopes of the tectonically-activeOlympic Mountains. Uplift occurs at a rate of 0.28 mm/yr (Brandon et al., 1998), providingsteep terrain susceptible to frequent landsliding and debris flows (Montgomery and Brandon,2002). The basin is dissected by the Hurricane Ridge Fault, which separates the highly de-formed Olympic Sedimentary Complex, containing metasedimentary rocks, from the CoastRange Terraine, which is composed primarily of basalt and sandstone (Brandon et al., 1998;Tabor and Cady, 1978). Only a small portion of the basin is glaciated, and pro-glacial materiallikely acts as a negligible source of sediment. Elwha River is divided into a series of subbasins,each separated by steep bedrock canyons. Within each subbasin, the river flows within bothnarrow and broad floodplains, which act as continuous sources of sediment supply (except forlocalized reaches, where the channel flows against bedrock outcrops at the valley margins). Inthe lower portions of the valley, till, glacio-lacustrine, and outwash deposits underly bluffsranging from a few to tens of meters high. These are significant sources of sediment in thelocalized reaches where they occur (Draut et al., 2011).Since Elwha River is a high order stream, and because supply from the actively migratingchannel receives ample supply from the banks, it is likely that the supply rate relative to dis-charge is rather consistent. However, disturbance events can provide pulses of sediment. In1967, a landslide dam breach several km upstream of Glines Canyon released a wave of sed-iment that deposited a 0.25 km2 fan in the channel (Tabor, 1987). The influence of the depositon modern sediment loads is unknown, although it was reworked locally for at least 15 yearsfollowing the breach (Acker et al., 2008). The time-averaged sediment supply rate has increasedover the last three decades: Bountry et al. (2011) found that the mean annual sedimentation ratein the Lake Mills reservoir was 47% higher from 1994-2010 than from 1927-1994. Whether theincrease is related to climate, either directly or indirectly, is unclear.The largest disturbance to Elwha River’s sediment supply regime was from the emplace-ment and removal of the Glines Canyon and Elwha Dams, which cut off supply to downstreamreaches from the early 20th century until 2011. Between 2011 and 2013, 10.5 million tons of sed-iment were released from behind the dams (Warrick et al., 2015). Terraces of a few to severalmeters remain in the former reservoirs. The deposits have become vegetated and more stable;however, they will likely act as a continuous source of sediment for decades to centuries.714.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesTable 4.2: Monte Carlo simulation sets. Each set is comprised of 200 runs of 150 years.Sediment supply scenarioDischarge S1 S2 S3scenario (with dam) (at capacity) (episodic)P1 (1925-1946) S1-P1 S2-P1 S3-P1P2 (1947-1976) S1-P2 S2-P2 S3-P2P3 (1977-2016) S1-P3 S2-P3 S3-P3Pall (1925-2016) S1-Pall S2-Pall S3-PallP123 (cycles through P1-P3) S1-P123 - -4.4 Elwha past and future: Monte Carlo simulations of channelevolution with varying sediment supply and climatic regimes4.4.1 MethodsA numerical modeling campaign was designed to answer the two main questions of this study:1) will the post-dam Elwha River evolve to the same state as it would if the dam never existed,and 2) in the context of geomorphic modeling, does an assumption of hydrologic stationarityadequately represent historical time periods with different mean climates? We adopt a MonteCarlo approach to simulate the range of possible river responses to different sediment supplyand hydrologic regimes.Channel evolution between Glines Canyon and Elwha Dams was modeled using MAST-1D. Verification of the model is presented in Chapter 3. In order to minimize computation time,we assume that the reach is fully alluvial and longitudinally homogenous. The model param-eters can be found in Appendix B. We performed 13 sets of model runs, each comprised of 200individual simulations spanning 150 years (1918-2068). To add stochasticity to the model, thedaily discharge record was created by randomly sampling annual hydrographs from differentclimatic periods (see below). The sets are listed in Table 4.2.Sediment supply scenariosThree different sediment supply regimes are used (Figure 4.4). For each, the sediment trans-port capacity is calculated using a form of the Wilcock and Crowe (2003) equation modified forlarge, cobble-bedded streams by Gaeuman et al. (2009). Hydraulics are calculated assumingnormal flow in a two-part cross-section that includes a rectangular channel and floodplain.The sediment supply is a function of the capacity:Qs,i = CQc,i (4.1)where Qs,i is the sediment supply for size class i, Qc,i is the sediment transport capacity for thatclass, and C is a supply modifier. The first scenario (S1) represents emplacement and removalof Glines Canyon Dam. Sediment supply is set at capacity (C = 1) for 9 years and then reduced724.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.4: Sediment supply boundary conditions for the Monte Carlo simulations. a)S1. The supply coefficient C becomes zero following dam emplacement in 1927 (S1-D) and jumps to 30 before declining exponentially to simulate the pulse of post-removal bed material beginning in 2012 (S1-PD). b) S2. Constant feed; C remains at1 throughout the run. c) S3. Stochastic sediment supply. The value of C is selectedat random from a positively-skewed log-normal distribution with a mean value of1.to zero in 1927 to simulate emplacement of the dam. The pulse following removal is modeledusing an exponential decay curveQs,i = Qc,i(1+ Cre−tλ ) (4.2)where t is the time in days since the removal, Cr = 30, and λ = 365 days. (Figure 4.4a). SeeChapter 3 for a description of the calibration and grainsize distribution of the pulse.For the second set of model runs (S2), sediment supply is set at capacity (C = 1, Figure4.4b). These simulations represent evolution of Elwha River without influence from the dams.We are assuming that sediment is supplied at capacity. While this is probably reasonable giventhe decadal timescale of the study and since Elwha River is a high-order stream in a supply-734.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesrich basin (Benda and Dunne, 1997b; Montgomery and Brandon, 2002), we also perform a third setof runs (S3) where sediment is supplied episodically to see whether the mode of sediment de-livery impacts channel evolution. For each discharge, the value of C is selected randomly froma log-normal distribution with a mean of 1 and standard deviation of 0.9 (Figure 4.4c). Withthis distribution, sediment is supplied below capacity most of the time, but pulses of sedimentperiodically pass through the system. This sediment supply scenario likely overestimates theperiod of time that the channel is sediment-starved but will provide an end-member case.Discharge scenariosTo examine the impact of historic climate regimes on channel evolution, each sediment supplyscenario was run with four different discharge regimes. The first three (P1-P3) correspond tothe three periods outlined in Section 4.3.1. For each run, a discharge record is created usingobserved flow data from the USGS gauge at McDonald bridge. Water years within each period(i.e. 1925-1946 for P1, 1947-1976 for P2, and 1977-2016 for P3) are selected at random andcomposited to create a 150 year record. Flow sequences within each water year remain intact,so that the annual hydrograph is maintained. A fourth scenario, Pall, is run with water yearsselected from all three climatic periods (1925-2016). Pall represents the condition of hydrologicstationarity.To examine the response time of the simulated channels to changes in climate regime, dis-charge scenario P123 was run with supply scenario S1 (with the dam). In P123, the climatestate shifts over the course of the run. Modeled time was divided into three periods. Forthe first 29 modeled years (1918-1946), discharges were sampled from the water years occur-ing in P1. Discharges were extracted from P2 between 1947 and 1976. For the final 91 years(1977-2068), water years from P3 are used.Geomorphic metricsThe effective discharge, bankfull discharge, and migration rate are used to quantify the mod-eled sediment transport regime and degree of channel stability. The effective discharge (Qe f f ),introduced by Wolman and Miller (1960), is defined as the flow that, when integrated over longtimescales, transports the most sediment. Wolman and Miller, as well as many others afterthem (e.g. Andrews, 2015; Emmett and Wolman, 2001; Torizzo and Pitlick, 2004), have found thatthe effective discharge typically occurs frequently and is less than or close to the bankfull dis-charge, suggesting that it is related to the channel-forming flow. However, others have foundthat plots of sediment transport against flow frequency tend to have a bimodal distribution–inother words, both frequent and more rare (recurrence intervals on the order of years-decades)flows are geomorphically important (e.g. Lenzi et al., 2006; Phillips, 2002). To fully appreciatehow the dominant flow responds to changes in the flow regime, the Q50 is also calculated. It isthe discharge below which half of the duration-averaged sediment is transported (Sichingab-ula, 1999) and is a more useful metric than Qe f f when the frequency/transport relationship is744.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.5: Schematic showing the relationship between the migration rate (m) and a. awidening channel, b. a narrowing channel, and c. a channel where widening andnarrowing is occurring concurrently. The solid lines show the channel position andcenterline at time t, while the shading and dotted line represent the channel at t− 1.bimodal. To calculate Qe f f and Q50, discharge data must be binned. So as not to underesti-mate the importance of extreme flood events, we follow the method of Hassan et al. (2014) andchoose bins that are the resolution of our discharge data (100 ft3/s or 2.8 m3/s).To see how the Q50 relates to channel-forming conditions, it is also normalized by the bank-full discharge Qb (Emmett and Wolman, 2001), which we calculate using the Manning EquationQb =1nAR2/3S1/2 (4.3)where n is the Manning roughness coefficient, A is channel area, R is the hydraulic radius,and S is slope. Manning’s n is calculated with a modified form of the Strickler relation thataccounts for roughness from form drag and sinuosityn = k1(0.013D1/6 + k2) (4.4)where D is a characteristic grainsize which we set to 2D65, following Wilcock et al. (2009), k1is a sinuosity multiplier and k2 is a form drag factor. Using a D65 of 112 mm (from the initialgrainsize distribution) and setting k1 to 1.15 and k2 to 0.0066 yields an n value of 0.044, whichis close to values measured on similar rivers in the field by Barnes (1967). We are assuming forthese calculations that D is constant, but since it is taken to the 1/6 power in Equation 4.4, Qbis not very sensitive to it.We define the channel migration rate m as the lateral movement of the channel centerlineover a defined period of time. Migration can occur as a result of channel widening, narrowing,or both (Figure 4.5). MAST-1D does not keep track of left and right banks. We assume that allwidening occurs on one bank and all narrowing on the other. The migration rate is simply the754.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesaverage of movement from widening and narrowing:m = 0.5(−∆Bc,n∆t+∆Bc,w∆t)(4.5)where ∆Bc,n∆t is the rate of channel narrowing (which is negative by definition) and∆Bc,w∆t is therate of widening.To test whether or not geomorphic processes on Elwha River can be considered station-ary despite changes in the governing climatic conditions, metrics of effective and bankfulldischarge from runs with the three climatic periods (P1-P3) are each tested for equal centraltendency with Pall using Welch’s t-test.4.4.2 ResultsThe results are divided into two parts. First, the results from S1-Pall and S2-Pall are presentedto demonstrate the differences in channel evolution between dammed and constant sedimentsupply conditions. Then, we focus on the impact of climate on the effective discharge andmigration rate for all sediment supply scenarios.Evolution of dammed and at-capacity channel-floodplain systemsS1-Pall behaved differently from S2-Pall, both before and after the dam removal. Evolutionof several key variables is presented in Figure 4.6. Following dam emplacement, both runshave similar width ranges for about 15 years, when the dammed channel begins to narrow ata faster rate (Figure 4.6a). By the time dam removal occurs, the two populations are nearlydistinct. Following dam removal, channel widening occurs. While width for individual runsfluctuates based on the flow record, the median width for all runs stabilizes at just under 100m after about 15 years. It never evolves back to the pre-disturbance condition (representedby the at capacity supply scenario); S2-Pall very gradually narrows, finally reaching a medianwidth that is 30% lower than the post-dam channel.The range of widths after 150 years is much wider for S1-Pall than for S2-Pall. Chan-nel width is negatively correlated to the grainsize distribution of the post-removal floodplain(Figure 4.6c), which is in turn dependent on flow sequencing during the dam removal. Therelationship between channel width, floodplain D50, and flow sequencing is presented in Fig-ure 4.7 for S1-Pall. Both parameters were averaged for each run from the years after 2040,for which the median of all runs is stable. 35% of the variability in channel width can be ex-plained by the grainsize distribution of the floodplain (p << 0.001). The correlation of flood-plain grainsize and channel width to geomorphically-significant flows over time is shown inFigure 4.7b. The y-axis shows the r2 value for least-squares linear regression for channel widthand floodplain D50 as functions of the cumulative volume of flow above 150 m3/s. In otherwords, flows occuring between 2012 and 2013 were summed for 2013, flows occuring between2012 and 2014 were summed for 2014, and so on so that at the end of the run in 2068, the764.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.6: Evolution over time of the channel and floodplain for runs with dam emplac-ment/removal (S1-Pall) and sediment supplied at capacity (S2-Pall). The shadingdelineates the 5th-95th percentile of all 200 runs, while the line denotes the median.Grey horizontal lines delineate dam emplacement in 1927 and removal in 2012. a.channel width. b. channel D50. c. floodplain D50. d. channel migration rate.regression is performed against all flows >150 m3/s during the post-removal period. With aconstant source of sediment (S2-Pall), the correlation between both channel width and flood-plain grainsize increases as the regression considers more recent flows. When all >150 m3/sflows between 2012 and 2068 are considered, they explain 30 and 50% of the variability infloodplain D50 and width, respectively. In contrast, the high flow events just after dam re-moval had a lasting impact on channel form for S1-Pall. The sequence of flows during thefirst two years following dam removal explain almost 70% of the variability in floodplain D50.The correlation declines over time, and flows after 2030 have virtually no impact on floodplainsediment. The correlation between flow and channel width rises with increasingly recent flowhistory after 2015, but unlike in S2-Pall, channel width is more correlated to the two yearsfollowing dam removal than it is to the following 30 years.The channel migration rate is less correlated with floodplain grainsize (r2 = 0.20, p <<.001). Migration rates for the Monte Carlo simulations are similar for S1-Pall and S2-Pall,774.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.7: Correlation between floodplain D50, channel width (Bc), and high flow events.a. Regression for channel width as a function of floodplain D50 for S1-Pall. Eachpoint represents one model simulation. Channel width and floodplain D50 wereboth averaged over all nodes. Values for both variables are averaged for the periodafter 2040. b. Regression of channel width and floodplain (FP) D50 as a function ofcumulative volume of flow >150 m3/s. See text for details. Grey shading shows theperiod over which Bc and FP D50 were averaged.although the range of values is lower during the dammed period and higher after the removal(Figure 4.6d). Unlike the floodplain, the grainsize of the channel (Figure 4.6b) recovered to thedam-free, constant supply state within 50 years after the removal.Effect of climate on sediment transport and channel stabilityDifferences in the hydrologic regime led to unique regime conditions for all three climatic pe-riods (Figure 4.8). P1, which has the lowest average annual water yield and peak flows (Table4.1), has the lowest sediment yield and migration rate in all sediment supply scenarios. Thechannel is much more dynamic in P2; it has the highest annual sediment yield in all sedimentsupply scenarios except for the dammed period in S1, and its median migration rate is 10-35%higher than P1. Channels under the P3 climate regime experience the highest rates of lateralinstability; the migration rate is nearly double that in P1. In addition, channels in P3 maintaina higher width/depth ratio than P1 and P2 (Figure 4.8i-l), although the ranges for all hydrol-ogy scenarios overlap quite a bit within each sediment supply scenario. The largest differenceoccurs in the sediment-starved post-dam channel, where over 75% of model runs in P3 hadwidth-depth ratios higher than the other discharge groups. However, despite differences inlateral stability and sediment transport competence, bankfull discharges for the three periods784.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.8: Geomorphic metrics as a function of sediment supply and hydrologic regime.Note the scale differences on the y-axis. Upper headings refer to the sediment sup-ply regime: S1-dam is the pre-dam removal period (1919-2011) and S1-post dam isthe post-removal period (2012-2068). For S2 and S3 (sediment supplied at capac-ity and episodically, respectively), the whole model period is used. a-d. Averageannual sediment yield, e-h. Average annual channel migration rate. i-l. Channelwidth/depth ratio at the end of the modeled period.794.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.9: Examples of effective discharge plots with bimodal distributions. a. A sin-gle run from S3-Pall. Effective discharge is 103 m3/s. b. A single run from S3-P3.Effective discharge is 231 m3/s.occur between 1-4 times per year (Figure 4.10). The ranges of bankfull recurrence intervalsoverlap for all climate periods in each sediment supply scenario except for the dammed pe-riod in S1 (based on a oneway ANOVA test, they are only statistically indistinguishable forS1-PD).The relationship between discharge and cumulative decadal-scale transport in the MAST-1D simulations has a bimodal distribution as described by Phillips (2002). Examples of effectivedischarge plots from two simulations are presented in Figure 4.9. Much of the sediment istransported during flows of around 100 m3/s, which occur on average around 18 times a year(Figure 4.10d). However, large peaks also occur at bankfull and flood discharges. In Figure4.9a, the former peak is more dominant and the effective discharge is low. In Figure 4.9b asingle flood flow bin transports the largest percentage of sediment, and the effective dischargeonly occurs about once a year, even though more sediment as a whole is transported duringfrequent events.Metrics of effective discharge are presented in Table 4.3. Under undisturbed sediment sup-ply conditions (S2 and S3), discharge/yield relations for most runs in P1 and P2 are similar tothat presented in Figure 4.9a; the effective discharge is around 100 m3/s and it occurs roughly3-20 days per year (Figure 4.10). P2 has a slightly higher Q50 than P1, but the Q50/Qb ratio isnearly identical for both periods, with the Q50 occurring at about half bankfull flow. On theother hand, the sediment transport regime for P3 resembles that of Figure 4.9b. The effectivedischarge is over twice that of P1 and P2, and it has a recurrence interval of about 1 year. TheQ50 for P3 represents conditions closer to bankfull than in the other two periods.Since P1/P2 and P3 show distinct patterns of effective discharge, the simulation assuming804.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesFigure 4.10: Frequency that flow exceeds effective discharge metrics for the effective dis-charge (a-d), discharge over which 50% of sediment is transported (e-h), and thebankfull discharge (i-l). See Figure 4.8 for a description of the sediment supplyregimes. Note differences in y-axis scaling. A p value of 0.003 corresponds to theevent occurring once a year.814.4. Elwha past and future: Monte Carlo simulations of channel evolution with varying sedimentsupply and climatic regimesTable 4.3: Metrics of effective discharge. The value listed is the median of the individualsimulations. Bold values for P1-P3 denote populations that are significantly differentfrom Pall at the 99.9% confidence level.Sediment supply scenarioRun P1 P2 P3 PallS1 DQe f f 452 511 508 477Q50 206 200 290 252Qb 84 113 133 132Q50/Qb 2.40 1.90 2.17 1.89S1 PDQe f f 92 95 355 103Q50 101 125 185 140Qb 224 300 365 312Q50/Qb 0.46 0.43 0.50 0.46S2Qe f f 140 106 282 111Q50 117 135 175 149Qb 202 220 232 204Q50/Qb 0.58 0.61 0.75 0.72S3Qe f f 103 106 282 101Q50 111 127 161 136Qb 201 224 257 231Q50/Qb 0.55 0.55 0.63 0.58D Prior to dam removal (1919-2011)PD Post-dam removal (2012-2068)hydrologic stationarity (Pall) is a poor representation of the sediment transport regime. Forundisturbed sediment supply regimes (S2 and S3), Pall aligns most closely with a transportregime dominated by frequent flow events; it is statistically indistinguishable from P1 and P2for about half the effective discharge metrics. It fails to characterize the high-flow dominatedregime of P3. Performance is worse in supply regime S1. The effective discharge for Pall issomewhat similar to P2, but is statistically different from both P1 and P3 in nearly all metrics.Our modeling suggests that channels respond to changes between the three hydrologicregimes within a decadal timescale. The median migration rates and channel widths fromsimulation set S1-P123 are plotted against time. The other hydrologic scenarios are shownfor comparison. The channels responded to the change from P1 to P2 rapidly. The medianmigration rate for S1-P123 reached that of the P2 run immediately, and channel width adjustedin less than 10 years. The response time from P2 to P3 was longer, with the S1-P123 migrationrate and channel width both taking a couple decades to reach that of P3.In our simulations, dam emplacement has a large impact on channel regime characteristicsand on the effective discharge. When sediment supply is low, larger, less frequent flood eventsare the dominant form of sediment transport. Qe f f for the dammed S1 period is up to 5 times824.5. DiscussionFigure 4.11: Response time of a. the channel migration rate and b. channel width tochanges in hydrologic regime for sediment supply scenario S1 (dam emplacementand removal). Series represent the median from each Monte Carlo run. In P1-P3and Pall, a single hydrologic regime spanned the entire run. In P123, the hydro-logic regime switched from P1-P2 and P2 to P3. Shading represents the divisionbetween the three hydrologic periods.higher than for ‘natural’ S2 and S3 sediment supply regimes (Table 4.3). It is similar for allthree climate regimes, but occurs most frequently during P3, for which large floods are morecommon (Figure 4.10a; also see Figure 4.3). As a result, sediment yield for P3 surpasses thatof P2, whose discharge regime is less efficient at transporting large grains (Figure 4.8a). Damemplacement also led to a reduction in the bankfull discharge (Table 4.3); our simulations sug-gest that flooding occurred more frequently by almost an order of magnitude (Figure 4.10i-j).Following dam removal, the sediment transport regime adjusted so that patterns of sedimentyield during the post-dam period in S1 more closely resemble those of S2 and S3 (Figure 4.8b).However, channel geometry does not fully recover to pre-dam conditions; the width/depthratio remains higher (Figure 4.8j) and flooding occurs slightly less frequently (Figure 4.10b).4.5 DiscussionDams have left a geomorphic legacy on the landscape by fragmenting the routing of sedimentthrough basins and creating storage loci behind reservoirs. As an increasing number of damsare removed, basin continuity will be restored, but the former reservoir deposits may persistfor decades or centuries. Whether rivers will fully recover from damming, or whether theywill adopt a new steady state, is still largely an open question, especially in the context ofglobal climate change.Our Monte Carlo simulations suggest the latter, at least on decadal timescales. One of theassumptions of our implementation of MAST-1D is that sediment in the floodplain composesa single, homogenous reservoir. In other words, any sediment entering the floodplain adjusts834.5. Discussionthe grainsize distribution of the entire floodplain, including the banks. During the periodwhen Glines Canyon Dam was in place, the channel became armored, and sediment enteringthe floodplain via channel migration was coarser. A coarse floodplain and channel bottomimpeded erosion of the protective bank toe, causing a drop in channel migration rate and anarrower channel (Figure 4.6). Following dam removal, bed material was finer than the bulkfloodplain mixture, at least for the first year following the release of bed material from theformer reservoir (Draut and Ritchie, 2015). Much of the sediment went into secondary stor-age on the floodplain, via overbank deposition, avulsion and floodplain channel reactivation(see Chapter 3). We predicted that the floodplain becomes finer and that the banks becomemore erodable. With lower bank strength, the flow is able to maintain a wider channel (Eatonet al., 2004; Millar and Quick, 1993, Figure 4.6a). Since floodplain turnover is much slowerthan turnover in the channel, the grainsize in the former did not recover after 50 years whilethe channel active layer adjusted to reflect the long-term sediment supply caliber after a fewdecades (Figure 4.6b and c).The reservoir model is an oversimplification of floodplain dynamics. In reality, the sizedistribution of existing eroding banks does not change, and any increase in erodability onthese deposits is due to rising channel mobility that enables erosion of the bank toe and anincrease in local shear stress caused by bar growth on the opposite bank. In addition, oursimulations overestimate post-removal storage, particularly of fine material (see discussionof this issue in Chapter 3). The major drop in floodplain particle size shown in Figure 4.6cis primarily due to deposition of suspended sediment during overbank flooding. Suspendedsediment lowers the overall grainsize distribution of the floodplain, but has little geomorphicinfluence. However, because the bank erosion algorithm depends on the fraction of coarsematerial in the floodplain (see Chapter 2), the fine fraction will have increased modeled ratesof bank erosion. This likely explains why the channel width and migration rate fail to returnto pre-dam levels.However, rivers tend to re-occupy recently abandoned surfaces more frequently than theyerode older floodplain material (Jerolmack and Paola, 2007; Konrad, 2012). Therefore, bank sta-bility in the decades following dam removal is dependent on the particle size of the depositsthat are stored in the modern channel. Suspended material does deposit on channel marginsand point bars (an example is provided in Figure 4.12). If current bar deposits contain pre-dominately fine material and are incorporated into the floodplain, they may create patches offloodplain with lower bank strength than older deposits, changing the regime channel dimen-sions and contributing to increased instability on decadal timescales. While our simulationscertainly overestimate the magnitude of floodplain fining, our results generate a hypothesisthat can be tested in the field. Indeed, a channel survey conducted in 2015 revealed that ElwhaRiver is already cutting into fresh bar deposits (refer to Appendix C) which appear to lackboulder-sized sediment currently found in modern cutbank deposits.As shown in Figure 4.7, our simulations suggest that the amount of floodplain fining, and844.5. DiscussionFigure 4.12: Deposition of fine material on new point bar surfaces. Much of the materialoriginated as suspended load and is much finer than the adjacent eroding bank.as a consequence the long-term effects on channel width and migration, is directly related tothe magnitude of high flow events in the couple years following dam removal. Equation 4.2implies that, the higher the flow during the initial phase of dam removal, the more reservoirsediment that is eroded. Increased sediment supply leads to more overbank deposition andhigher avulsion rates (see Chapter 3), which introduce more fine sediment into the floodplain.This contradicts the findings of Major et al. (2012), who postulate that the flow magnitude dur-ing dam removal affected short-term timing of sediment flux to Sandy River, Oregon follow-ing removal of Marmot Dam, but had little impact on the long-term evolution of the reservoir.However, Marmot Dam was removed in one stage and eroded primarily through knickpointretreat. The Glines Canyon Dam was removed in stages over the course of two years, whichallowed the channel to migrate across the reservoir deposit (Randle et al., 2015). The migrationrate, which is proportional to flow strength, will have affected the net amount of sedimentreleased from the reservoir.The three years following dam removal on Elwha River were abnormally dry, with peakflows all below the two year flood. The Elwha channel was able to export about 90% of the854.5. Discussionsediment released from Lake Mills, and East et al. (2015) have predicted that around 0.3 Mtremain in the channel and roughly 0.2 Mt were deposited in the floodplain. This sedimentwill likely have a limited effect on future floodplain dynamics, and the decadal-scale widthand migration rate will probably lower than our modeled range, particularly since we over-estimated the amount of fine channel storage. However, we hypothesize that the hydrologicregime during dam removal does affect the evolution of the channel. A large flood during orshortly after dam removal can erode large swaths of the former reservoir deposit. Most of thebedload component of the pulse will enter long-term storage on bars downstream of the damand may eventually become incorporated into the floodplain, altering its composition.We have shown that decadal-scale channel stability is sensitive to large inputs of fine sed-iment into the floodplain. However, there is still uncertainty regarding the actual supply ofthis sediment from upstream reservoirs. Little is known about the partitioning of sedimentbetween the suspended and bed loads, or about how the caliber of sediment supply evolvesover time as fine reservoirs are exhausted. These factors will control how the river behaves ondecadal timescales, and future field campaigns should prioritize characterizing the caliber ofdeposits remaining within the former reservoir as well as on point bars downstream.The range of migration rates we calculated for Elwha River are somewhat lower than thosefound on other regional rivers. While we predicted migration rates of between 1-2 m/yr forthe at capacity (S2) and stochastic (S3) supply scenarios (Figure 4.8e-h), O’Connor et al. (2003)found that migration rates for rivers draining the western side of the Olympic peninsula were2-12 times higher. It is true that our numerical model does not account for many local bankerosion processes. But Elwha River is also straighter and steeper than other regional rivers,and local climate is slightly different. These factors may partially explain the discrepancy.Even if the governing factors were more similar, Elwha River today cannot be expected tohave the same channel exchange rates to those measured over the course of the 20th centurybecause the flow regime will reflect a different climate. O’Connor et al. (2003) used historicmaps and photos that go as far back as the late 1800s and incorporate multiple phases of thePacific Decadal Oscillation. Our modeling shows that small adjustments to the hydrologicregime can have a moderate impact on sediment transport. In all cases except for S1-dam,the channel geometry appears to be adjusted to the flow regime. For S1-post dam, S2, andS3, median bankfull discharge occurs 1-2 times per year for all discharge scenarios (Figure4.10j-l). This approximately corresponds to the one year peak flow event (see Figure 4.3),which aligns with regional bankfull flow frequencies for maritime mountains in the PacificNorthwest (Castro and Jackson, 2001).Our modeling suggests that adjustment to new hydrologic regimes can occur within afew years, which is short enough time to make the change relevant on decadal timescales(Figure 4.11). Metrics of channel forming discharge are different for the three periods andreflect unique geomorphic regimes. The bankfull discharge scales with the frequency of largefloods which are capable of mobilizing the banks, regardless of sediment supply regime. But864.5. DiscussionFigure 4.13: Average daily sediment transport by month, divided into phases of trans-port described by Carling (1988). Phase 1 represents movement of fine sedimentwinnowed from the bed surface. During Phase 2, most of the active layer is intransport, but the coarsest grains are mostly immobile and maintain channel sta-bility. The entire bed is in transport during Phase 3. See text for details. a. Period1; b. Period 2; c. Period 3.the relationship between bankfull flow and effective discharge is different for P3 than for theearlier two periods (Table 4.3), representing a shift in the nature of the sediment transportregime from one driven by modest, frequent flow events to one dominated more by largeflood events (except in the case of sediment-starved S1-D, for which exceptionally high flowsare required to mobilize the bed).In order to see the significance of this shift on channel stability, it is useful to divide sed-iment transport events by their ability to mobilize the active layer. Carling (1988) identifiedthree different phases of sediment transport. During Phase 1 transport, most of the bed isstatic, and mobilized sediment is comprised of fine grains winnowed from the bed surface.Phase 2 transport occurs when most grains are at least partially mobile, but the largest particlesdo not move. Significant transport occurs within the active layer, but there is little geomorphicchange. Only during Phase 3, when flow is strong enough to mobilize the structural grains,does the bed become restructured. Distributions of transport phases, averaged by month forruns with supply scenario S3, are presented in Figure 4.13 for the three climatic periods. Wedefined the threshold between Phase 1 and Phase 2 as the point at which channel shear stress874.5. Discussionis 1.5 times that needed to entrain a 54 mm particle, which roughly corresponds to the the D50.This implies that the D50 is well above the entrainment threshold, but not fully mobile, whichWilcock and McArdell (1993) found occurs at roughly twice the shear stress required to mobi-lize the particle. For our calculations, Phase 3 occurs when the channel shear stress exceeds 1.5times the entrainment threshold for a 300 mm particle, which roughly corresponds to the D90.In our simulations, most sediment transport occurs during Phase 2, regardless of the climaticperiod. But Phase 3 transport constitutes a slightly higher percentage of the total load for P3(10% as opposed to 7% for both P1 and P2). Qe f f describes Phase 3 transport for P3, while itoccurs during Phase 2 for P1 and P2.The increased importance of high-magnitude flood events manifests itself in greater geo-morphic instability. Even though the average annual water and sediment yield are both lowerin P3 than P2 (Table 4.1 and Figure 4.8b-d), P3 maintains a significantly higher width-depthratio and migration rate. It is also possible that scour-and-fill episodes are more intense dur-ing P3, although the spatial and temporal resolution of MAST-1D is too coarse to characterizethese processes well. Battin et al. (2007) suggest that larger peak flows may increase bed scourin the winter, jeopardizing salmon eggs during the incubation period. McKean and Tonina(2013) point out that only small portions of the bed are mobile during floods and that largerfloods will not put salmon redds at risk. Neither studies considered the effect of increasingflood peaks on channel width and bank erosion, which act as a control on channel competenceand a source of sediment supply.The difference in channel stability between the three climatic periods is due in large partto the distribution of annual water yield between the winter and nival seasons. During thecool phase of the PDO (P2), much of the winter precipitation that would lead to high floods iscaptured in the basin as snow and released more slowly in the spring. While nival flows areefficient in transporting sediment (Figure 4.13b), in Elwha River’s hybrid regime, they are notpowerful enough to lead to Phase 3 transport. The ratio between mean annual sediment yieldand migration rate is plotted as a function of average discharge during the nival period inFigure 4.14. In the two sediment supply scenarios without Glines Canyon Dam, P2 transportsroughly 50% more sediment for every unit of bank movement. P1 and P3, which are bothcharacterized by weak snowmelt flows, have nearly the same flux/migration ratio, despitethe fact that P3 is more driven by low-frequency flood flows. Large floods were rare duringP1, but frequent events were smaller as well, so that the reduction in channel migration ismatched by a low sediment yield. In P2, a winter season with abundant transport and bankerosion is followed by nival flows that convey sediment through the channel without leadingto bank instability.The implication is that, at least for hybrid streams with high sediment supply, the bed ma-terial sediment transport regime is essentially de-coupled from the channel forming discharge.The channel is shaped by the magnitude and frequency of flood peaks, while sediment trans-port is more closely related to total water yield. The relationship between the two is dependent884.6. Conclusionon temperature, which dictates the partitioning of precipitation between rain and snow. How-ever, the importance of the snowmelt season varies depending on sediment availability andlandscape history. For supply scenario S1, the difference in flux/migration ratio between P2and the other hydrologic regimes is much lower than in S2 and S3. During the dammed period,nival flows contribute little to overall sediment transport because the bed surface is armoredand only winter flood flows are able to mobilize the coarse fraction. Following dam removal,near-bank sediments are more mobile, and nival flows can feasibly lead to bank erosion.It appears that the assumption of hydrologic stationarity is problematic in the context ofdecadal-scale geomorphic processes. The range of output for the Pall simulations fell betweenthe three climate periods and acts as an average for the historic period. However, there was nocyclic pattern; each period had a unique set of governing parameters which led to differencesin the flows responsible for transporting the most sediment, the amount of channel instabil-ity, and the relationship between sediment yield and bank erosion. There is no evidence toindicate that the future hydrologic regime will resemble any of the periods captured in thehistoric record. Reidy-Liermann et al. (2012) predicts that many basins in the region may experi-ence larger flood peaks in the future as the result of decreasing snowpack and the consequenttransition to more flashy, rainfall-dominated hydrologic regimes. They anticipate that by 2040,Elwha River will have transitioned into a completely pluvial system. Our analysis suggeststhat the future channel will be dominated even more by flood flows and that it will be lesslaterally stable. Using the 20th century to predict the 21st may lead to an underestimation offlood and erosion risk.4.6 ConclusionOur objective was to consider whether Elwha River, and other hypothetical systems under-going dam removal, will evolve to resemble past undisturbed systems. Our analysis suggeststhat, at least for high-order trunk streams with active channel/floodplain coupling, the answermay depend on the amount of sediment that is sequestered into the floodplain while the initialpulse of reservoir sediment is moving through the system. When the sediment pulse is muchfiner than the floodplain, it can create patches that are easily erodible, leading to higher chan-nel widths and migration rates that may persist for decades. While it is common to quantifythe movement of the sediment pulse through the system by measuring elevation change in thethalweg, it may be more appropriate on decadal timescales to consider the material depositedon bars and in floodplain channels. To do so, more information is also needed on the caliberof sediment supply.Hybrid streams in the Pacific Northwest will likely behave differently in the 21st centurythan in the 20th. On Elwha River, the sediment transport regime transitioned over the course ofthe 20th century from being dominated by modest, very frequent flows, to being shaped moreby flood events. Our simulations show that this transition would probably have occurred tosome extent even if the dams had not been in place. It appears that as snowpack decreases894.6. ConclusionFigure 4.14: The ratio of average annual sediment transport (in thousands of m3) to chan-nel migration (in m/y) as a function of the average nival discharge. Pre- and post-dam removal periods in S1 are denoted by circles and squares, respectively. Tri-angles represent S2, and S3 is plotted in diamonds. Colors represent the climaticperiods and are the same as Figures 4.8 and 4.10.and disappears, channel instability and coupling with the floodplain will increase.Geomorphologists have a long tradition of conceptualizing rivers as being governed by arelatively static set of environmental forcings, where channels are fluctuating around regimedimensions that are characterized by a single channel-forming discharge and homogeneousparticle size. Our analysis suggests that this framework is questionable on human timescales,in the context of natural and anthropogenically altered sediment supply and streamflow vari-ability.90Chapter 5ConclusionOver the past century, dams and the reservoirs behind them have become one of the mostpervasive features of Earth’s landscapes. One of the new frontiers in geomorphology is learn-ing how landscapes evolve after dams have been removed. The purpose of this thesis was toexplore the decadal-scale legacy of dams on Elwha River. In particular, we were interestedin examining processes related to channel/floodplain coupling, which determine how mate-rial of various sizes travels through the sediment cascade. To do so, a numerical model wasadapted that we suggest captures the most relevant processes operating on large, wandering,low-sinuosity cobble-bedded streams over decadal timescales. Our hypothesis, explained inChapter 2, was that bank stability is determined by the ability of the flow to mobilize thelarge structural grains near the channel margins. Non-cohesive banks are protected by a banktoe, and when that toe is scoured away, bank erosion continues while the channel has the ca-pacity to transport large grains. A crucial element to our model is the assumption that thechannel-wide shear stress is an adequate proxy for near-bank flow conditions. Widening viabank erosion is countered by channel narrowing due to vegetation growth on exposed channelsurfaces. We use channel-wide shear stress as a proxy for river flow, and assume that narrow-ing occurs when the shear stress is low. Even given the simplest characterization of govern-ing conditions–steady sediment supply and hydrologic regimes–channel width and migrationrate are not constant. Instead, they fluctuate within a narrow range. This is because wander-ing rivers often avulse, reoccupying old locations and scouring floodplain channels that oncecarried only flood flow. Our modeling in Chapter 2 suggests that when the river avulses, thedeeper channel becomes more effective at scouring banks.As is demonstrated in Chapter 3, this model of mobility-driven bank erosion appears toadequately describe channel-floodplain coupling on the sediment-starved Elwha River whenGlines Canyon and Elwha Dams were in place. Comparison between model output and fielddata show that the former is able to reproduce observed channel width and grainsize. GlinesCanyon Dam blocked sediment supply to the study area, creating a new reservoir behind thedam. The exchange of sediment between the channel and floodplain slowed, both due to adecrease in the number of avulsions and because channel coarsening led to a reduction in the91frequency and magnitude of bank erosion events. We found in Chapter 4 that sediment trans-port was dominated by large, winter flood events during this period. While flows during thesnowmelt period transport up to about a third of the annual yield in supply-rich conditions,they are unable to mobilize the coarse bed during the dammed period.According to our model, channel-floodplain exchange increases significantly during thefirst decade following dam removal. The primary flux is due to increases in floodplain storagecaused by avulsion. This is corroborated by air photo analysis, which reveals that for the up-stream half of the study area, most widening was due to the activation of floodplain channels.Channel width and rates of bank erosion also increased. However, the poor fit between mod-eled and field rates of channel movement and sediment flux during the post-removal periodsuggests that our assumption that channel-wide hydraulics are suitable proxies for near-bankflow strength is not justified during periods of exceptionally high sediment supply.Unlike other shorter-term studies (East et al., 2015; Major et al., 2012), our decadal-scaleperspective from Chapter 4 seems to suggest that future channel evolution can be dependenton the flow regime during the first two years of dam removal if large flood events duringsediment pulses result in ample deposition of sediment on the floodplain. If this sedimentis finer than the underlying material, it can increase the mobility of the floodplain, leadingto higher magnitudes of channel instability on decadal timescales. However, our analysis isbased on major assumptions about the caliber of sediment supply and the efficiency of thefloodplain in trapping suspended sediment. Regardless of the impact of sediment supply,future flow regimes will influence how stable channels are in the future; declining snowpackand increasing winter peak flows expected for many Pacific Northwest basins will likely leadto higher rates of migration and more transport events that break up surface layers in thechannel.Our study revealed that a rather simple characterization of lateral flux, using channel-wide flow metrics, is sufficient to explain quite a bit of the variablility in observed channelwidth and migration, especially during periods of low sediment supply. However, the currentmodel is missing key reach-scale processes that affect performance during periods of sedimentexcess. This includes the role of log jams, which can act as in-channel storage reservoirs andcause avulsion. In the future, the influence of channel morphology in partitioning flow andsediment should be considered. In addition, more realistic algorithms for vegetation growthwill help quantify the important effect of channel narrowing on sequestering sediment withinthe floodplain.Numerical models are only as good as the data that back them. As we have shown inChapter 2, our model is highly sensitive to grain size. Its underestimation of the competenceof Elwha River to evacuate the pulse of sediment following dam removal (see Chapter 3) mayreflect uncertainty in the upstream boundary condition more than any process deficiencies inthe model. Quantifying sediment supply and particle size is one of the most important–andmost challenging–tasks for any geomorphic study. We suggest that future field studies on dam92removals prioritize campaigns to quantify the magnitude and, more importantly, the size, ofsediment pulses.It appears that dams leave a lasting legacy on the sediment cascade. They divide naturalstorage centers into two parts: a fine reservoir behind the dam composed of sediment reflect-ing the long-term sediment load, and a coarser channel/floodplain system downstream of thedam that is starved of fine material. Remixing of the reservoirs following dam removal is aprocess that occurs over decadal timescales. Overbank deposition, avulsion, and vegetationgrowth all reintroduce reservoir sediment back into the downstream floodplain. Most adjust-ment occurs within the few years after the removal, but patches of old and new material mightpersist on the order of decades. This means that rates of bank erosion and sediment supplyto the channel can be affected over long timescales. Channel-floodplain coupling is ultimatelydetermined by channel processes, which are highly sensitive to sediment supply and channelmorphology.So far, most studies of dam removals have focused on the short-term impact of the initialsediment pulse on the channel profile and planform. There has been little consideration tosediment exchange between the channel and floodplain–neither in terms of additional sed-iment supply via bank erosion or with regards to the long-term sequestration of sediment.We have shown that the legacy of dams does not end after the initial wave of sediment hasdissipated. Our 1-D reservoir-based approach has generated general hypotheses about howchannel-floodplain coupling may impact channel stability. More observations are needed toquantify how dam emplacement and removal affect the patchiness of sediment deposits–andwhether erosion of these deposits conform to existing conceptual and numerical models ofbank stability. In addition, our research has only touched on the many responses of rivers tosediment pulses. We have considered the characteristics of one reservoir. Future work shouldexplore the various responses of channel stability to sediment inputs of different magnitudesand calibers.93BibliographyAbbe, T. B., and D. R. Montgomery (1996), Large woody debris jams, channel hydraulics andhabitat formation in large rivers, Regulated Rivers Research & Management, 12(23), 201–221.Abernethy, B., and I. D. Rutherfurd (2001), The distribution and strength of riparian tree rootsin relation to riverbank reinforcement, Hydrological processes, 15(1), 63–79.Acker, S. A., T. J. Beechie, and P. B. Shafroth (2008), Effects of a natural dam-break flood ongeomorphology and vegetation on the elwha river, washington, usa.Andrews, E. D. (1980), Effective and bankfull discharges of streams in the yampa river basin,colorado and wyoming, Journal of Hydrology, 46(3-4), 311–330.Andrews, E. D. (1986), Downstream effects of flaming gorge reservoir on the green river, col-orado and utah, Geological Society of America Bulletin, 97(8), 1012–1023.Andrews, S. G. C., E. Mortenson (2015), Evaluation of alternative streamflows to improveaquatic and riparian habitat in the walker river basin, nevada, Journal of the Nevada WaterResources Association.ASCE Task Committee on Hydraulics, B. M., and M. of River Width Adjustment (1998a), Riverwidth adjustment. i: Processes and mechanisms, Journal of Hydraulic Engineering, ASCE, 124.ASCE Task Committee on Hydraulics, B. M., and M. of River Width Adjustment (1998b), Riverwidth adjustment. ii: Modeling., J. Hydr. Engrg., ASCE, 124.Baker, V. R. (1977), Stream-channel response to floods, with examples from central texas, Geo-logical Society of America Bulletin, 88(8), 1057–1071.Barnes, H. H. (1967), Roughness characteristics of natural channels, Tech. rep., US Govt. Print.Off.,.Basso, S., A. Frascati, M. Marani, M. Schirmer, and G. Botter (2015), Climatic and landscapecontrols on effective discharge, Geophysical Research Letters, 42(20), 8441–8447.Battin, J., M. W. Wiley, M. H. Ruckelshaus, R. N. Palmer, E. Korb, K. K. Bartz, and H. Imaki(2007), Projected impacts of climate change on salmon habitat restoration, Proceedings of thenational academy of sciences, 104(16), 6720–6725.Beechie, T. J., M. Liermann, M. M. Pollock, S. Baker, and J. Davies (2006), Channel pattern andriver-floodplain dynamics in forested mountain river systems, Geomorphology, 78(1), 124–141.94BIBLIOGRAPHYBenda, L., and T. Dunne (1997a), Stochastic forcing of sediment routing and storage in channelnetworks, Water Resources Research, 33(12), 2865–2880.Benda, L., and T. Dunne (1997b), Stochastic forcing of sediment supply to channel networksfrom landsliding and debris flow, Water Resources Research, 33(12), 2849–2863.Bertoldi, W., A. Gurnell, N. Surian, K. Tockner, L. Zanoni, L. Ziliani, and G. Zolezzi (2009),Understanding reference processes: linkages between river flows, sediment dynamics andvegetated landforms along the tagliamento river, italy, River Research and Applications, 25(5),501–516.Bledsoe, B. P., and C. C. Watson (2001), Effects of urbanization on channel instability, JAWRAJournal of the American Water Resources Association, 37(2), 255–270.Bond, N., J. Overland, M. Spillane, and P. Stabeno (2003), Recent shifts in the state of the northpacific, Geophysical Research Letters, 30(23).Bountry, J., R. Ferrari, K. Wille, and T. Randle (2011), Survey report for lake mills and lake ald-well on the elwha river, washington, Tech. rep., SRH-2010-23, US Department of the Interior,Bureau of Reclamation, Denver, CO.Bountry, J. A., Y. G. Lai, and T. J. Randle (2013), Sediment impacts from the savage rapids damremoval, rogue river, oregon, Reviews in Engineering Geology, 21, 93–104.Bowling, L. C., P. Storck, and D. P. Lettenmaier (2000), Hydrologic effects of logging in westernwashington, united states, Water Resources Research, 36(11), 3223–3240.Brandon, M. T., M. K. Roden-Tice, and J. I. Garver (1998), Late cenozoic exhumation of the cas-cadia accretionary wedge in the olympic mountains, northwest washington state, GeologicalSociety of America Bulletin, 110(8), 985–1009.Brunner, G. W. (2016), Hec-ras river analysis system. hydraulic reference manual. version 5.0.,Tech. rep., US Army Corps of Engineers Hydrologic Engineering Center.Bunte, K., and S. R. Abt (2001), Sampling surface and subsurface particle-size distributions inwadable gravel-and cobble-bed streams for analyses in sediment transport, hydraulics, andstreambed monitoring.Burroughs, B. A., D. B. Hayes, K. D. Klomp, J. F. Hansen, and J. Mistak (2009), Effects ofstronach dam removal on fluvial geomorphology in the pine river, michigan, united states,Geomorphology, 110(3), 96–107.Camporeale, C., and L. Ridolfi (2007), Noise-induced phenomena in riparian vegetation dy-namics, Geophysical Research Letters, 34(18).Camporeale, C., E. Perucca, L. Ridolfi, and A. Gurnell (2013), Modeling the interactions be-tween river morphodynamics and riparian vegetation, Reviews of Geophysics, 51(3), 379–414.Cannon, S. H., and J. DeGraff (2009), The increasing wildfire and post-fire debris-flow threatin western usa, and implications for consequences of climate change, in Landslides–disasterrisk reduction, pp. 177–190, Springer.Cantelli, A., M. Wong, G. Parker, and C. Paola (2007), Numerical model linking bed and bankevolution of incisional channel created by dam removal, Water Resources Research, 43(7).95BIBLIOGRAPHYCarling, P. (1988), The concept of dominant discharge applied to two gravel-bed streams inrelation to channel stability thresholds, Earth Surface Processes and Landforms, 13(4), 355–367.Carroll, R., J. Warwick, A. James, and J. Miller (2004), Modeling erosion and overbank de-position during extreme flood conditions on the carson river, nevada, Journal of Hydrology,297(1), 1–21.Castro, J. M., and P. L. Jackson (2001), Bankfull discharge recurrence intervals and regionalhydraulic geometry relationships: patterns in the pacific northwest, usa, JAWRA Journal ofthe American Water Resources Association, 37(5), 1249–1262.Church, M. (2002), Fluvial sediment transfer in cold regions, in Landscapes of transition, pp.93–117, Springer.Church, M. (2008), Gravel Bed Rivers 6, Volume 11, chap. Multiple scales in rivers, pp. 3–32,Elsevier.Church, M., and O. Slaymaker (1989), Disequilibrium of holocene sediment yield in glaciatedbritish columbia, Nature, 337(2), 452–454.Church, M. A., D. McLean, and J. Wolcott (1987), River bed gravels: sampling and analysis,Sediment Transport in Gravel-Bed Rivers. John Wiley and Sons New York. 1987. p 43-88, 17 fig, 3tab, 50 ref.Constantine, J. A., T. Dunne, J. Ahmed, C. Legleiter, and E. D. Lazarus (2014), Sediment sup-ply as a driver of river meandering and floodplain evolution in the amazon basin, NatureGeoscience, 7(12), 899–903.Corenblit, D., E. Tabacchi, J. Steiger, and A. M. Gurnell (2007), Reciprocal interactions andadjustments between fluvial landforms and vegetation dynamics in river corridors: a reviewof complementary approaches, Earth-Science Reviews, 84(1), 56–86.Cui, Y., G. Parker, J. Pizzuto, and T. E. Lisle (2003), Sediment pulses in mountain rivers: 2.comparison between experiments and numerical predictions, Water Resources Research, 39(9).Cui, Y., G. Parker, C. Braudrick, W. E. Dietrich, and B. Cluer (2006), Dam removal expressassessment models (dream). part 1: model development and validation, Journal of HydraulicResearch, 44(3), 291–307.Cunderlik, J. M., and D. H. Burn (2003), Non-stationary pooled flood frequency analysis, Jour-nal of Hydrology, 276(1), 210–223.Curran, C. A., C. P. Konrad, J. L. Higgins, and M. K. Bryant (2009), Estimates of sedimentload prior to dam removal in the elwha river, clallam county, washington, Tech. rep., USGeological Survey.Czuba, J. A., C. S. Magirl, C. R. Czuba, C. A. Curran, K. H. Johnson, T. D. Olsen, H. K. Kimball,and C. C. Gish (2012), Geomorphic analysis of the river response to sedimentation downstream ofMount Rainier, Washington, US Department of the Interior, US Geological Survey.Darby, S. E., and C. R. Thorne (1996), Development and testing of riverbank-stability analysis,Journal of hydraulic engineering, 122(8), 443–454.96BIBLIOGRAPHYDarby, S. E., and M. J. Van de Wiel (2003), Models in fluvial geomorphology, Tools in fluvialgeomorphology, pp. 501–537.Davidson, S. (2016), Modeling disturbance and channel evolution in mountain streams, Ph.D.thesis, University of British Columbia.Department of the Interior (1995), Elwha river ecosystem restoration implementation, finalenvironmental impact statement. nps d-271a.Dietrich, W. E., T. Dunne, N. F. Humphrey, L. M. Reid, et al. (1982), Construction of sedimentbudgets for drainage basins, Sediment Budgets in Forested Drainage Basins. United States ForestService Gen. Tech. Rep. PNW-141. p5-23.Doyle, M. W., E. H. Stanley, and J. M. Harbor (2002), Geomorphic analogies for assessingprobable channel response to dam removal, JAWRA Journal of the American Water ResourcesAssociation, 38(6), 1567–1579.Draut, A., and A. C. Ritchie (2015), Sedimentology of new fluvial deposits on the elwha river,washington, usa, formed during large-scale dam removal, River Research and Applications,31(1), 42–61.Draut, A. E., J. B. Logan, R. E. McCoy, M. McHenry, and J. A. Warrick (2008), Channel evolutionon the lower elwha river, washington, 1939-2006, Tech. rep., Geological Survey (US).Draut, A. E., J. B. Logan, and M. C. Mastin (2011), Channel evolution on the dammed elwhariver, washington, usa, Geomorphology, 127(1), 71–87.Duda, J., H. J. Coe, S. Morley, and K. Kloehn (2011a), Establishing spatial trends in waterchemistry and stable isotopes (δ15n and δ13c) in the elwha river prior to dam removal andsalmon recolonization, River research and applications, 27(10), 1169–1181.Duda, J. J., J. A. Warrick, and C. S. Magirl (2011b), Coastal habitats of the elwha river,washington-biological and physical patterns and processes prior to dam removal, Tech. rep.,US Geological Survey.East, A. E., G. R. Pess, J. A. Bountry, C. S. Magirl, A. C. Ritchie, J. B. Logan, T. J. Randle,M. C. Mastin, J. T. Minear, J. J. Duda, et al. (2015), Large-scale dam removal on the elwhariver, washington, usa: River channel and floodplain geomorphic change, Geomorphology,228, 765–786.Easterbrook, D. J. (1986), Stratigraphy and chronology of quaternary deposits of the pugetlowland and olympic mountains of washington and the cascade mountains of washingtonand oregon, Quaternary Science Reviews, 5, 145–159.Eaton, B. C., and T. R. Giles (2009), Assessing the effect of vegetation-related bank strengthon channel morphology and stability in gravel-bed streams using numerical models, EarthSurface Processes and Landforms, 34(5), 712–724.Eaton, B. C., M. Church, and R. G. Millar (2004), Rational regime model of alluvial channelmorphology and response, Earth Surface Processes and Landforms, 29(4), 511–529.Ebbesmeyer, C. C., D. R. Cayan, D. R. McLain, F. H. Nichols, D. H. Peterson, and K. T. Red-mond (1991), 1976 step in the pacific climate: forty environmental changes between 1968-1975 and 1977-1984.97BIBLIOGRAPHYEke, E., G. Parker, and Y. Shimizu (2014), Numerical modeling of erosional and depositionalbank processes in migrating river bends with self-formed width: Morphodynamics of barpush and bank pull, Journal of Geophysical Research: Earth Surface, 119(7), 1455–1483.Elsner, M. M., L. Cuo, N. Voisin, J. S. Deems, A. F. Hamlet, J. A. Vano, K. E. Mickelson, S.-Y. Lee,and D. P. Lettenmaier (2010), Implications of 21st century climate change for the hydrologyof washington state, Climatic Change, 102(1-2), 225–260.Emmett, W. W., and M. G. Wolman (2001), Effective discharge and gravel-bed rivers, EarthSurface Processes and Landforms, 26(13), 1369–1380.Ferguson, R. (2003), The missing dimension: effects of lateral variation on 1-d calculations offluvial bedload transport, Geomorphology, 56(1), 1–14.Ferguson, R., and M. Church (2009), A critical perspective on 1-d modeling of river processes:Gravel load and aggradation in lower fraser river, Water Resources Research, 45(11).Ferrer-Boix, C., and M. A. Hassan (2015), Influence of the sediment supply texture on morpho-logical adjustments in gravel-bed rivers, Water Resources Research, 50(11), 8868–8890.Fetherston, K. L., R. J. Naiman, and R. E. Bilby (1995), Large woody debris, physical process,and riparian forest development in montane river networks of the pacific northwest, Geo-morphology, 13(1-4), 133–144.Fleming, S. W., P. H. Whitfield, R. Moore, and E. J. Quilty (2007), Regime-dependent stream-flow sensitivities to pacific climate modes cross the georgia–puget transboundary ecoregion,Hydrological Processes, 21(24), 3264–3287.Foley, M. M., J. J. Duda, M. M. Beirne, R. Paradis, A. Ritchie, and J. A. Warrick (2015), Rapidwater quality change in the elwha river estuary complex during dam removal, Limnologyand Oceanography, 60(5), 1719–1732.Friedman, J. M., W. R. Osterkamp, and W. M. Lewis (1996), Channel narrowing and vegetationdevelopment following a great plains flood, Ecology, 77(7), 2167–2181.Gaeuman, D., E. Andrews, A. Krause, and W. Smith (2009), Predicting fractional bed loadtransport rates: Application of the wilcock-crowe equations to a regulated gravel bed river,Water resources research, 45(6).Galay, V. (1983), Causes of river bed degradation, Water resources research, 19(5), 1057–1090.Gilbert, J., and R. Link (1995), Alluvium distribution in lake mills, glines canyon project andlake aldwell, elwha project, washington, Elwha Technical Series PN-95-4, August, Boise, ID.Gomez, B., B. J. Rosser, D. H. Peacock, D. M. Hicks, and J. A. Palmer (2001), Downstream finingin a rapidly aggrading gravel bed river, Water Resources Research, 37(6), 1813–1823.Gomez, B., Y. Cui, A. Kettner, D. Peacock, and J. Syvitski (2009), Simulating changes to thesediment transport regime of the waipaoa river, new zealand, driven by climate change inthe twenty-first century, Global and Planetary Change, 67(3), 153–166.Gordon, E., and R. K. Meentemeyer (2006), Effects of dam operation and land use on streamchannel morphology and riparian vegetation, Geomorphology, 82(3), 412–429.98BIBLIOGRAPHYGottesfeld, A. S., and L. M. J. Gottesfeld (1990), Floodplain dynamics of a wandering river,dendrochronology of the morice river, british columbia, canada, Geomorphology, 3(2), 159–179.Graf, W. L. (1999), Dam nation: A geographic census of american dams and their large-scalehydrologic impacts, Water resources research, 35(4), 1305–1311.Graf, W. L. (2006), Downstream hydrologic and geomorphic effects of large dams on americanrivers, Geomorphology, 79(3), 336–360.Gran, K., and C. Paola (2001), Riparian vegetation controls on braided stream dynamics, WaterResources Research, 37(12), 3275–3283.Grant, G. (2001), Dam removal: Panacea or pandora for rivers?, Hydrological Processes, 15(8),1531–1532.Grant, G. E., and S. L. Lewis (2015), The remains of the dam: What have we learned from15 years of us dam removals?, in Engineering Geology for Society and Territory-Volume 3, pp.31–35, Springer.Greet, J., J. Angus Webb, and R. D. Cousens (2011), The importance of seasonal flow timing forriparian vegetation dynamics: a systematic review using causal criteria analysis, FreshwaterBiology, 56(7), 1231–1247.Gregory, K., and C. Park (1974), Adjustment of river channel capacity downstream from areservoir, Water Resources Research, 10(4), 870–873.Guthrie, R., P. Friele, K. Allstadt, N. Roberts, S. Evans, K. Delaney, D. Roche, J. Clague, andM. Jakob (2012), The 6 august 2010 mount meager rock slide-debris flow, coast mountains,british columbia: characteristics, dynamics, and implications for hazard and risk assess-ment., Natural Hazards & Earth System Sciences, 12(5).Hajek, E. A., and M. A. Wolinsky (2012), Simplified process modeling of river avulsion andalluvial architecture: Connecting models and field data, Sedimentary Geology, 257, 1–30.Hamlet, A. F., and D. P. Lettenmaier (2007), Effects of 20th century warming and climate vari-ability on flood risk in the western us, Water Resources Research, 43(6).Hare, S., and R. Francis (1995), Climate change and salmon production in the northeast pacificocean, Canadian Special Publication of Fisheries and Aquatic Sciences, pp. 357–372.Hassan, M., L. Roberge, M. Church, M. More, S. Donner, J. Leach, and K. Ali (2017), What arethe contemporary sources of sediment in the mississippi river?, Geophysical Research Letters,44(17), 8919–8924.Hassan, M. A., R. Egozi, and G. Parker (2006), Experiments on the effect of hydrograph char-acteristics on vertical grain sorting in gravel bed rivers, Water Resources Research, 42(9).Hassan, M. A., D. Brayshaw, Y. Alila, and E. Andrews (2014), Effective discharge in small for-merly glaciated mountain streams of british columbia: Limitations and implications, WaterResources Research, 50(5), 4440–4458.99BIBLIOGRAPHYHeller, P. L., P. E. Beland, N. F. Humphrey, S. K. Konrad, R. M. Lynds, M. E. McMillan,K. E. Valentine, Y. A. Widman, and D. J. Furbish (2001), Paradox of downstream fining andweathering-rind formation in the lower hoh river, olympic peninsula, washington, Geology,29(11), 971–974.Higson, J. L., and M. B. Singer (2015), The impact of the streamflow hydrograph on sedimentsupply from terrace erosion, Geomorphology, 248, 475–488.Hoey, T. B., and R. Ferguson (1994), Numerical simulation of downstream fining by selectivetransport in gravel bed rivers: Model development and illustration, Water resources research,30(7), 2251–2260.Hooke, R. L. (1994), On the efficacy of humans as geomorphic agents, GSA Today, 4(9), 217.Hooke, R. L. (2000), On the history of humans as geomorphic agents, Geology, 28(9), 843–846.Inman, D. L., and S. A. Jenkins (1999), Climate change and the episodicity of sediment flux ofsmall california rivers, The Journal of geology, 107(3), 251–270.Interagency Advisory Committee on Water Data (1982), Guidelines for determining flood flowfrequency: Bulletin 17B of the Hydrology Subcommittee, Office of Water Data Coordination,Reston, VA.Jakob, M., M. Bovis, and M. Oden (2005), The significance of channel recharge rates for esti-mating debris-flow magnitude and frequency, Earth Surface Processes and Landforms, 30(6),755–766.James, L. A., and M. B. Singer (2008), Development of the lower sacramento valley flood-control system: Historical perspective, Natural Hazards Review, 9(3), 125–135.Jenkins, K. J., N. Chelgren, K. Sager-Fradkin, P. Happe, and M. J. Adams (2015), Occupancypatterns of mammals and lentic amphibians in the elwha river riparian zone before damremoval, River research and applications, 31(2), 193–206.Jerolmack, D. J., and D. Mohrig (2007), Conditions for branching in depositional rivers, Geol-ogy, 35(5), 463–466.Jerolmack, D. J., and C. Paola (2007), Complexity in a cellular model of river avulsion, Geomor-phology, 91(3), 259–270.Jordan, P., and O. Slaymaker (1991), Holocene sediment production in lillooet river basin,british colombia: A sediment budget approach, Ge´ographie physique et Quaternaire, 45(1), 45–57.Kaless, G., L. Mao, and M. Lenzi (2014), Regime theories in gravel-bed rivers: models, con-trolling variables, and applications in disturbed italian rivers, Hydrological Processes, 28(4),2348–2360.Khaliq, M., T. Ouarda, J.-C. Ondo, P. Gachon, and B. Bobe´e (2006), Frequency analysis of a se-quence of dependent and/or non-stationary hydro-meteorological observations: A review,Journal of hydrology, 329(3), 534–552.Kiem, A. S., S. W. Franks, and G. Kuczera (2003), Multi-decadal variability of flood risk, Geo-physical Research Letters, 30(2).100BIBLIOGRAPHYKleinhans, M., H. Jagers, E. Mosselman, and C. Sloff (2008), Bifurcation dynamics and avulsionduration in meandering rivers by one-dimensional and three-dimensional models, Waterresources research, 44(8).Kloehn, K. K., T. J. Beechie, S. A. Morley, H. J. Coe, and J. J. Duda (2008), Influence of dams onriver-floodplain dynamics in the elwha river, washington, Northwest Science, 82(sp1), 224–235.Knighton, D. (1998), Fluvial forms and processes: London, Edward Arnold.Kondolf, G., H. Pie´gay, and N. Landon (2002), Channel response to increased and decreasedbedload supply from land use change: contrasts between two catchments, Geomorphology,45(1), 35–51.Konrad, C., H. Berge, R. Fuerstenberg, K. Steff, T. Olsen, and J. Guyenet (2011), Channel dy-namics in the middle green river, washington, from 1936 to 2002, Northwest Science, 85(1),1–14.Konrad, C. P. (2009), Simulating the recovery of suspended sediment transport and river-bedstability in response to dam removal on the elwha river, washington, Ecological Engineering,35(7), 1104–1115.Konrad, C. P. (2012), Reoccupation of floodplains by rivers and its relation to the age structureof floodplain vegetation, Journal of Geophysical Research: Biogeosciences (2005–2012), 117(G4).Krapesch, G., C. Hauer, and H. Habersack (2011), Scale orientated analysis of river widthchanges due to extreme flood hazards, Natural Hazards and Earth System Sciences, 11(8), 2137.Lane, E. (1955), Design of stable channels, Transactions of the American Society of Civil Engineers,120, 1234–1279.Lauer, J., and G. Parker (2008a), Modeling framework for sediment deposition, storage, andevacuation in the floodplain of a meandering river: Theory, Water Resources Research, 44.Lauer, J. W., and G. Parker (2008b), Net local removal of floodplain sediment by river meandermigration, Geomorphology, 96(1), 123–149.Lauer, J. W., E. Viparelli, and H. Pie´gay (2016), Morphodynamics and sediment tracers in 1-d (mast-1d): 1-d sediment transport that includes exchange with an off-channel sedimentreservoir, Advances in Water Resources, 93, 135–149.Lauer, J. W., C. Echterling, C. Lenhart, P. Belmont, and R. Rausch (2017), Air-photo basedchange in channel width in the minnesota river basin: Modes of adjustment and implica-tions for sediment budget, Geomorphology, 297, 170–184.Lenzi, M., L. Mao, and F. Comiti (2006), Effective discharge for sediment transport in a moun-tain river: computational approaches and geomorphic effectiveness, Journal of Hydrology,326(1), 257–276.Leopold, L., and T. J. Maddock (1953), The hydraulic geometry of stream channels and somephysiographic implications.Lisle, T. E. (1982), Effects of aggradation and degradation on riffle-pool morphology in naturalgravel channels, northwestern california, Water Resources Research, 18(6), 1643–1651.101BIBLIOGRAPHYLisle, T. E., Y. Cui, G. Parker, J. E. Pizzuto, and A. M. Dodd (2001), The dominance of disper-sion in the evolution of bed material waves in gravel-bed rivers, Earth Surface Processes andLandforms, 26(13), 1409–1420.Litzow, M. A. (2006), Climate regime shifts and community reorganization in the gulf of alaska:how do recent shifts compare with 1976/1977?, ICES Journal of Marine Science: Journal duConseil, 63(8), 1386–1396.Liu, M., J. C. Adam, and A. F. Hamlet (2013), Spatial-temporal variations of evapotranspirationand runoff/precipitation ratios responding to the changing climate in the pacific northwestduring 1921-2006, Journal of Geophysical Research: Atmospheres, 118(2), 380–394.Luce, C. H., and Z. A. Holden (2009), Declining annual streamflow distributions in the pacificnorthwest united states, 1948–2006, Geophysical Research Letters, 36(16).MacKenzie, L. G., and B. C. Eaton (2017), Large grains matter: contrasting bed stability andmorphodynamics during two nearly identical experiments, Earth Surface Processes and Land-forms.Madej, M. A., D. G. Sutherland, T. E. Lisle, and B. Pryor (2009), Channel responses to varyingsediment input: A flume experiment modeled after redwood creek, california, Geomorphol-ogy, 103(4), 507–519.Magilligan, F. J., and K. H. Nislow (2005), Changes in hydrologic regime by dams, Geomorphol-ogy, 71(1), 61–78.Magirl, C. S., R. C. Hilldale, C. A. Curran, J. J. Duda, T. D. Straub, M. Domanski, and J. R. Fore-man (2015), Large-scale dam removal on the elwha river, washington, usa: Fluvial sedimentload, Geomorphology, 246, 669–686.Major, J. J., J. E. OConnor, C. J. Podolak, M. K. Keith, G. E. Grant, K. R. Spicer, S. Pittman, H. M.Bragg, J. R. Wallick, D. Q. Tanner, et al. (2012), Geomorphic response of the Sandy River, Oregon,to removal of Marmot Dam, US Geological Survey.Mantua, N. J., and S. R. Hare (2002), The pacific decadal oscillation, Journal of oceanography,58(1), 35–44.Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis (1997), A pacific inter-decadal climate oscillation with impacts on salmon production, Bulletin of the american Me-teorological Society, 78(6), 1069–1079.Mapes, L. (2013), Elwha: A river reborn, Mountaineers Books.Martin, R. L., D. J. Jerolmack, and R. Schumer (2012), The physical basis for anomalous diffu-sion in bed load transport, Journal of Geophysical Research: Earth Surface (2003–2012), 117(F1).Matalas, N. C. (1997), Stochastic hydrology in the context of climate change, in Climate Changeand Water Resources Planning Criteria, pp. 89–101, Springer.McCabe, G. J., M. A. Palecki, and J. L. Betancourt (2004), Pacific and atlantic ocean influenceson multidecadal drought frequency in the united states, Proceedings of the National Academyof Sciences, 101(12), 4136–4141.102BIBLIOGRAPHYMcKean, J., and D. Tonina (2013), Bed stability in unconfined gravel bed mountain streams:With implications for salmon spawning viability in future climates, Journal of GeophysicalResearch: Earth Surface, 118(3), 1227–1240.McLean, D. G., and M. Church (1999), Sediment transport along lower fraser river: 2. estimatesbased on the long-term gravel budget, Water Resources Research, 35(8), 2549–2559.Merritt, D. M., and D. J. Cooper (2000), Riparian vegetation and channel change in responseto river regulation: a comparative study of regulated and unregulated streams in the greenriver basin, usa, River Research and Applications, 16(6), 543–564.Micheli, E., and J. Kirchner (2002), Effects of wet meadow riparian vegetation on streambankerosion. 2. measurements of vegetated bank strength and consequences for failure mechan-ics, Earth Surface Processes and Landforms, 27(7), 687–697.Millar, R. G., and M. C. Quick (1993), Effect of bank stability on geometry of gravel rivers,Journal of Hydraulic Engineering, 119(12), 1343–1363.Milly, P. C., J. Betancourt, M. Falkenmark, R. M. Hirsch, Z. W. Kundzewicz, D. P. Letten-maier, and R. J. Stouffer (2008), Stationarity is dead: Whither water management?, Science,319(5863), 573–574.Miori, S., R. Repetto, and M. Tubino (2006), A one-dimensional model of bifurcations in gravelbed channels with erodible banks, Water Resources Research, 42(11).Montgomery, D. R., and M. T. Brandon (2002), Topographic controls on erosion rates in tecton-ically active mountain ranges, Earth and Planetary Science Letters, 201(3), 481–489.Moore, R. (1996), Snowpack and runoff responses to climatic variability, southern coast moun-tains, british columbia, Northwest Science, 70(4), 321–333.Morley, S. A., J. J. Duda, H. J. Coe, K. K. Kloehn, and M. L. McHenry (2008), Benthic inverte-brates and periphyton in the elwha river basin: current conditions and predicted responseto dam removal, Northwest Science, 82, 179–196.Mosher, D. C., and A. T. Hewitt (2004), Late quaternary deglaciation and sea-level history ofeastern juan de fuca strait, cascadia, Quaternary International, 121(1), 23–39.Mosselman, E. (1992), Mathematical modeling of morphological processes in rivers with erodi-ble banks, Unpublished Ph. D. Thesis, Delft University of Technology.Mote, P. W. (2003), Trends in snow water equivalent in the pacific northwest and their climaticcauses, Geophysical Research Letters, 30(12).Mote, P. W. (2006), Climate-driven variability and trends in mountain snowpack in westernnorth america*, Journal of Climate, 19(23), 6209–6220.Mote, P. W., A. F. Hamlet, M. P. Clark, and D. P. Lettenmaier (2005), Declining mountain snow-pack in western north america*, Bulletin of the American meteorological Society, 86(1), 39–49.Munn, M. D., R. Black, A. Haggland, M. Hummling, and R. Huffman (1999), An assessment ofstream habitat and nutrients in the elwha river basin: implications for restoration, Tech. rep.,US Dept. of the Interior, US Geological Survey; Branch of Information Services [distributor],.103BIBLIOGRAPHYNaik, P. K., and D. A. Jay (2011), Distinguishing human and climate influences on the columbiariver: changes in mean flow and sediment transport, Journal of hydrology, 404(3), 259–277.Nanson, G. C., and E. J. Hickin (1986), A statistical analysis of bank erosion and channel mi-gration in western canada, Geological Society of America Bulletin, 97(4), 497–504.Nelson, J. M., J. P. Bennett, and S. M. Wiele (2003), Flow and sediment-transport modeling,Tools in fluvial geomorphology, 18, 539–576.Newman, M., G. P. Compo, and M. A. Alexander (2003), Enso-forced variability of the pacificdecadal oscillation, Journal of Climate, 16(23), 3853–3857.Newman, M., M. A. Alexander, T. R. Ault, K. M. Cobb, C. Deser, E. Di Lorenzo, N. J. Mantua,A. J. Miller, S. Minobe, H. Nakamura, et al. (2016), The pacific decadal oscillation, revisited,Journal of Climate, (2016).Nicoll, T. J., and E. J. Hickin (2010), Planform geometry and channel migration of confinedmeandering rivers on the canadian prairies, Geomorphology, 116(1), 37–47.Nones, M., P. Ronco, and G. Di Silvio (2013), Modelling the impact of large impoundments onthe lower zambezi river, International journal of river basin management, 11(2), 221–236.O’Connor, J. E., M. A. Jones, and T. L. Haluska (2003), Flood plain and channel dynamics ofthe quinault and queets rivers, washington, usa, Geomorphology, 51(1), 31–59.O’Connor, J. E., J. J. Duda, and G. E. Grant (2015), 1000 dams down and counting, Science,348(6234), 496–497.Osterkamp, W., and C. Hupp (2010), Fluvial processes and vegetationglimpses of the past, thepresent, and perhaps the future, Geomorphology, 116(3), 274–285.Overland, J., S. Rodionov, S. Minobe, and N. Bond (2008), North pacific regime shifts: Defini-tions, issues and recent transitions, Progress in Oceanography, 77(2), 92–102.Overland, J. E., and M. Wang (2007), Future climate of the north pacific ocean, Eos, TransactionsAmerican Geophysical Union, 88(16), 178–182.Pagano, T., and D. Garen (2005), A recent increase in western us streamflow variability andpersistence, Journal of Hydrometeorology, 6(2), 173–179.Parker, G., Y. Shimizu, G. Wilkerson, E. Eke, J. Abad, J. Lauer, C. Paola, W. Dietrich, andV. Voller (2011), A new framework for modeling the migration of meandering rivers, EarthSurface Processes and Landforms, 36(1), 70–86.Pasquale, N., P. Perona, R. Francis, and P. Burlando (2012), Effects of streamflow variability onthe vertical root density distribution of willow cutting experiments, Ecological engineering,40, 167–172.Perona, P., P. Molnar, M. Savina, and P. Burlando (2009), An observation-based stochasticmodel for sediment and vegetation dynamics in the floodplain of an alpine braided river,Water Resources Research, 45(9).104BIBLIOGRAPHYPerona, P., P. Molnar, B. Crouzy, E. Perucca, Z. Jiang, S. McLelland, D. Wu¨thrich, K. Edmaier,R. Francis, C. Camporeale, et al. (2012), Biomass selection by floods and related timescales:Part 1. experimental observations, Advances in Water Resources, 39, 85–96.Pess, G. R., M. L. McHenry, T. J. Beechie, and J. Davies (2008), Biological impacts of the elwhariver dams and potential salmonid responses to dam removal, Northwest Science, 82, 72–90.Pfeiffer, A. M., N. J. Finnegan, and J. K. Willenbring (2017), Sediment supply controls equi-librium channel geometry in gravel rivers, Proceedings of the National Academy of Sciences,114(13), 3346–3351.Phillips, J. D. (2002), Geomorphic impacts of flash flooding in a forested headwater basin,Journal of Hydrology, 269(3), 236–250.Piechota, T. C., J. A. Dracup, and R. G. Fovell (1997), Western us streamflow and atmosphericcirculation patterns during el nin˜o-southern oscillation, Journal of Hydrology, 201(1), 249–271.Pizzuto, J. E. (1990), Numerical simulation of gravel river widening, Water Resources Research,26(9), 1971–1980.Poff, N. L., J. D. Olden, D. M. Merritt, and D. M. Pepin (2007), Homogenization of regional riverdynamics by dams and global biodiversity implications, Proceedings of the National Academyof Sciences, 104(14), 5732–5737.Pohl, M. (2004), Channel bed mobility downstream from the elwha dams, washington, TheProfessional Geographer, 56(3), 422–431.Polenz, M., K. W. Wegmann, and H. W. Schasse (2004), Geologic map of the Elwha and Ange-les Point 7.5-minute quadrangles, Clallam County, Washington, Washington State Division ofGeology and Earth Resources.Randle, T. J., J. A. Bountry, A. Ritchie, and K. Wille (2015), Large-scale dam removal on theelwha river, washington, usa: Erosion of reservoir sediment, Geomorphology, 246, 709–728.Reidy-Liermann, C., J. D. Olden, T. Beechie, M. J. Kennard, P. Skidmore, C. Konrad, andH. Imaki (2012), Hydrogeomorphic classification of washington state rivers to supportemerging environmental flow management strategies, River Research and Applications, 28(9),1340–1358.Rinaldi, M., B. Mengoni, L. Luppi, S. E. Darby, and E. Mosselman (2008), Numerical simulationof hydrodynamics and bank erosion in a river bend, Water Resources Research, 44(9).Robertson, K. M., and C. K. Augspurger (1999), Geomorphic processes and spatial patterns ofprimary forest succession on the bogue chitto river, usa, Journal of Ecology, 87(6), 1052–1063.Rubin, Z. K., G. M. Kondolf, and P. A. Carling (2015), Anticipated geomorphic impacts frommekong basin dam construction, International Journal of River Basin Management, 13(1), 105–121.Salas, J. D., and J. Obeysekera (2014), Revisiting the concepts of return period and risk fornonstationary hydrologic extreme events, Journal of Hydrologic Engineering, 19(3), 554–568.Serinaldi, F., and C. G. Kilsby (2015), Stationarity is undead: Uncertainty dominates the distri-bution of extremes, Advances in Water Resources, 77, 17–36.105BIBLIOGRAPHYSichingabula, H. M. (1999), Magnitude-frequency characteristics of effective discharge for sus-pended sediment transport, fraser river, british columbia, canada, Hydrological processes,13(9), 1361–1380.Sklar, L. S., W. E. Dietrich, E. Foufoula-Georgiou, B. Lashermes, and D. Bellugi (2006), Dogravel bed river size distributions record channel network structure?, Water resources re-search, 42(6).Slaymaker, H. O. (1972), Recent fluctuations in the mean discharge of the fraser river, Contem-porary Geography: Research Trends. BC Geographical Series, (16).Stoffel, M., D. Conus, M. A. Grichting, I. Lie`vre, and G. Maıˆtre (2008), Unraveling the patternsof late holocene debris-flow activity on a cone in the swiss alps: chronology, environmentand implications for the future, Global and Planetary Change, 60(3), 222–234.Stoffel, M., T. Mendlik, M. Schneuwly-Bollschweiler, and A. Gobiet (2014), Possible impacts ofclimate change on debris-flow activity in the swiss alps, Climatic change, 122(1-2), 141–155.Sutherland, A. J. (1967), Proposed mechanism for sediment entrainment by turbulent flows,Journal of Geophysical Research, 72(24), 6183–6194.Swanson, B. J., G. A. Meyer, and J. E. Coonrod (2011), Historical channel narrowing alongthe rio grande near albuquerque, new mexico in response to peak discharge reductions andengineering: magnitude and uncertainty of change from air photo measurements, EarthSurface Processes and Landforms, 36(7), 885–900.Tabor, R. W. (1987), Geology of Olympic National Park, Pacific Northwest National Parks &Forests Association.Tabor, R. W., and W. M. Cady (1978), Geologic map of the olympic peninsula, washington,Tech. rep.Tealdi, S., C. Camporeale, and L. Ridolfi (2011), Modeling the impact of river damming onriparian vegetation, Journal of hydrology, 396(3), 302–312.Thorne, C. (1982), Gravel-bed rivers, chap. Processes and mechanisms of river bank erosion, pp.227–271, John Wiley & Sons, Inc.Torizzo, M., and J. Pitlick (2004), Magnitude-frequency of bed load transport in mountainstreams in colorado, Journal of Hydrology, 290(1), 137–151.Trenberth, K. E. (1990), Recent observed interdecadal climate changes in the northern hemi-sphere, Bulletin of the American Meteorological Society, 71(7), 988–993.Tsujimoto, T. (1999), Fluvial processes in streams with vegetation, Journal of hydraulic research,37(6), 789–803.Tucker, G. E., and G. R. Hancock (2010), Modelling landscape evolution, Earth Surface Processesand Landforms, 35(1), 28–50.Tullos, D., and H.-W. Wang (2014), Morphological responses and sediment processes followinga typhoon-induced dam failure, dahan river, taiwan, Earth Surface Processes and Landforms,39(2), 245–258.106BIBLIOGRAPHYTunnicliffe, J. F., and M. Church (2015), A 1-d morphodynamic model of postglacial valleyincision, Journal of Geophysical Research: Earth Surface, 120(11), 2253–2279.Van De Wiel, M. J., and S. E. Darby (2004), Numerical modeling of bed topography and bankerosion along tree-lined meandering rivers, Riparian vegetation and fluvial geomorphology, pp.267–282.Verhaar, P. M., P. M. Biron, R. I. Ferguson, and T. B. Hoey (2008), A modified morphodynamicmodel for investigating the response of rivers to short-term climate change, Geomorphology,101(4), 674–682.Villarini, G., F. Serinaldi, J. A. Smith, and W. F. Krajewski (2009), On the stationarity of annualflood peaks in the continental united states during the 20th century, Water Resources Research,45(8).Viparelli, E., D. Gaeuman, P. Wilcock, and G. Parker (2011), A model to predict the evolutionof a gravel bed river under an imposed cyclic hydrograph and its application to the trinityriver, Water Resources Research, 47(2).Warrick, J. A., J. A. Bountry, A. E. East, C. S. Magirl, T. J. Randle, G. Gelfenbaum, A. C. Ritchie,G. R. Pess, V. Leung, and J. J. Duda (2015), Large-scale dam removal on the elwha river,washington, usa: source-to-sink sediment budget and synthesis, Geomorphology, 246, 729–750.Whitfield, P. H., R. Moore, S. W. Fleming, and A. Zawadzki (2010), Pacific decadal oscillationand the hydroclimatology of western canadareview and prospects, Canadian Water ResourcesJournal, 35(1), 1–28.Wilcock, P., J. Pitlick, and Y. Cui (2009), Sediment transport primer: estimating bed-materialtransport in gravel-bed rivers.Wilcock, P. R., and J. C. Crowe (2003), Surface-based transport model for mixed-size sediment,Journal of Hydraulic Engineering, 129(2), 120–128.Wilcock, P. R., and B. W. McArdell (1993), Surface-based fractional transport rates: Mobiliza-tion thresholds and partial transport of a sand-gravel sediment, Water Resources Research,29(4), 1297–1312.Wilcox, A. C., J. E. O’Connor, and J. J. Major (2014), Rapid reservoir erosion, hyperconcentratedflow, and downstream deposition triggered by breaching of 38 m tall condit dam, whitesalmon river, washington, Journal of Geophysical Research: Earth Surface, 119(6), 1376–1394.Williams, G. P., and M. G. Wolman (1984), Downstream effects of dams on alluvial rivers.Wolman, M., and J. Miller (1960), Magnitude and frequency of forces in geomorphic processes,Journal of Geology, 68, 54–74.Wolman, M. G., and R. Gerson (1978), Relative scales of time and effectiveness of climate inwatershed geomorphology, Earth Surface Processes and Landforms, 3(2), 189–208.Zhang, Y., J. M. Wallace, and D. S. Battisti (1997), Enso-like interdecadal variability: 1900-93,Journal of climate, 10(5), 1004–1020.107Appendix AMAST-1D model descriptionA.1 IntroductionMAST-1D (morphodynamic and sediment tracers in 1-D) is a bed evolution model where thechannel and floodplain are coupled. Details can be found in the original publications (Lauerand Parker, 2008a,b; Lauer et al., 2016).MAST-1D is designed to model long spatial (10s to 100s km) and temporal (decades-millenia) timescales where bank erosion and channel migration allow for channel sedimentto be sourced and stored within the floodplain. The channel (active layer), substrate, andfloodplain are treated as a set of reservoirs, each with a characteristic geometry, volume, andgrainsize distribution. Mass is conserved within each reservoir on a size-specific basis. Chan-nel exchange occurs between reservoirs via longitudinal sediment transport, bank erosion,channel narrowing, avulsion, and bed elevation change, all of which are functions of an im-posed water discharge.A model schematic is presented in Figure A.1. The model space is structured into a seriesof nodes aligned in the longitudinal direction. During each time step, the outgoing sedimentload in the upstream node is calculated. It is a function of the sediment transport capacityof the active layer reservoir and the depositional properties of the floodplain. When the flowis high enough to overtop the banks, some sediment is deposited in the floodplain reservoiras overbank material. The rest is transported downstream and becomes the incoming fluxto the next node. Once transport is calculated, lateral exchanges of sediment between reser-voirs are characterized for each node. When flow strength is low, we assume that the bed isstable and that vegetation is able to grow on channel surfaces. This leads to transfer of sed-iment into the floodplain from a point bar deposit, which is composed of material from theactive layer reservoir and the sediment load. The point bar is assumed to have a single con-stant height. Floodplain sediment is transferred to the active layer reservoir via bank erosionwhen flow is strong enough to mobilize bank material. Net fluxes to and from the active layer,from the incoming and outgoing sediment load and from bank erosion, are calculated to de-108A.1. IntroductionFigure A.1: Model schematictermine changes in bed elevation (z) and channel width (Bc). When input to the active layerexceeds output, aggradation occurs and the underlying channed substrate increases in thick-ness. When sediment transport capacity is greater than supply to the active layer, the channeldegrades and substrate material is incorporated into the active layer.In MAST-1D, flow strength is calculated from an imposed discharge, which can be repre-sented either by a stepped flow duration curve or a hydrograph. When using a the latter, flowstrength, sediment transport, and reservoir exchanges are determined using each dischargevalue in turn. Each flow is imposed over a number of timesteps that cumulatively equal thetemporal resolution of the discharge record (e.g. if a daily discharge record is used and eachmodel time step is 0.25 days, then 4 time steps will be performed for each discharge). Whena flow duration curve is used, the discharge record is divided into bins that are assigned atime-averaged duration. Flow strength, sediment transport capacity, and reservoir exchangesare determined for each discharge in turn, and the total flux for each exchange is the duration-weighted average of all imposed flows. While the flow duration curve does not account fortemporal variability in the hydrologic regime, it is more numerically stable and allows MAST-1D to run much faster. For both flow algorithms, we assume that the channel is rectangularand that sediment distribution within each reservoir is spatially homogenous.Full details of the model are divided into three sections. First, the model steps are listed in109A.2. Model procedureorder with references to the equations used. Then, the governing equations for flow, sedimenttransport, and reservoir exchange are presented in detail. Finally, methods for determininginitial model conditions are highlighted.A.2 Model procedureThe model steps proceed as follows:1. Boundary conditions are set using Equations A.88-A.90. The upstream boundary is animposed size-specific sediment load. The downstream water surface elevation is im-posed, either as a constant (i.e. in the case of a downstream control such as a dam) or asa function of discharge.2. The floodplain number (Lauer and Parker, 2008a), which is a parameter that determinesthe ability of the floodplain to trap overbank material, is either calculated with EquationsA.91-A.92 or set by the user. The initial floodplain grainsize distribution is set to reflectlong-term steady-state conditions with Equations A.93-A.94.3. Hydraulics are calculated for the entire reach using the standard step backwater method(Equations A.1-A.16).4. Bedload transport for the upstream-most node is determined with Equations A.17-A.25.5. Equations A.43-A.47 are used to calculate rates of channel widening and narrowing.6. Lateral reservoir exchange rates are calculated with Equations A.48-A.49 and A.51-A.58.These include exchanges between the floodplain and active layer due to bank erosionand channel narrowing.7. The Exner equation (Equation A.59) is solved to determine whether the channel is ag-grading or degrading. Vertical exchange rates between the substrate, active layer, andfloodplain reservoirs are determined using Equations A.60-A.64.8. The suspended sediment concentration and deposition rate is determined with Equa-tions A.26-A.28 and A.40.9. Volumes and grainsize distributions of the reservoirs are updated using Equations A.70-A.72.10. Equations A.74-A.78 are used to update channel geometry. If the conditions for avulsionare met, Equations A.83-A.87 are used to adjust reservoir dimensions and grainsizes.110A.3. Governing equations11. Equations A.79-A.82 are used to split or combine substrate layers. If the thickness of theuppermost substrate layer becomes thicker than a threshold, it is split into two strati-graphic units. If the thickness dips below a threshold, it is combined with the strati-graphic unit below it.12. If the net amount of channel degradation drops below a critical value and the node des-ignated as bedrock-influenced, then the node becomes partly alluvial. In future steps,partly alluvial transport is calculated as a function of the bed that is alluvial. When thevolume of sediment in the active layer of the partly-alluvial node reaches its capacity,then the node becomes fully alluvial.13. The sediment transport rates Qs,i and Qs,w are set as the feed for the next node in thedownstream direction. Steps 4-12 are repeated for all nodes.14. The boundary conditions are changed if needed by the user. Steps 3-13 are repeated forthe specified number of timesteps or discharges.A.3 Governing equationsA.3.1 HydraulicsHydraulics are calculated using the standard-step method applied to the backwater equation,assuming steady, gradually-varied, subcritical flow. The water surface elevation (WSE) of thedownstream-most node is provided as a boundary condition (see Section A.4.1), and conser-vation of energy is used to determine the WSE of the next node upstream. The WSE of thesecond node is then used to calculate the third node upstream, and so on. The procedure wasadapted from that used in the HEC-RAS model (Brunner, 2016).Calculation of flow depth and velocityHydraulics between each node are calculated using the 1-dimensional form of the BernoulliEquation:z2 + y2 +αv,2v¯222g= z1 + y1 +αv,1v¯212g+ he (A.1)The subscript 1 denotes the downstream node, while 2 represents the upstream node. z and yare bed elevation and flow depth, respectively, and WSE = z+ y. v¯ is average velocity over thenode cross-section, g represents gravitational acceleration, αv is a weighting coefficient thataccounts for the partitioning of average velocity between the channel and floodplain, and hedenotes the energy loss between the upstream and downstream nodes.The channel cross-section is divided into two parts: the channel and the floodplain, eachwith a characteristic roughness. The mean kinetic energy head (αvv¯222g ) is defined as the discharge-111A.3. Governing equationsweighted average between the velocity heads of the two sections:αvv¯22g=Qcv2c2g + Q fv2f2gQc + Q f(A.2)where Qc and Q f are the discharges over the channel and floodplain zones, respectively, andvc and v f are the corresponding velocities. If we rewrite v in terms of Q and flow area A:v =QA(A.3)andv¯ =QAc + A f, (A.4)then we can use the Manning Equation:Q = KS1/2f (A.5)to reformulate equation A.2 in terms of area and conveyance K, which is defined asK =1nAR2/3 (A.6)R is the hydraulic radius, estimated as the flow depth y (assuming a wide channel) and n isthe Manning coefficient. For the channel, it is calculated using a modified form of the Stricklerrelation:nc = Cn,m(0.0146D1/665 + Cn,a) (A.7)where D65 is the 65th percentile of the bed material grainsize distribution and Cn,a and Cn,mare user-defined constants that account for roughness due to form drag and sinuosity. TheManning coefficient for the floodplain in each node is a user-defined constant.The energy slope (S f ) is assumed to be constant between the two nodes and is defined as:S f =( Q1 + Q2Kc,1 + K f ,1 + Kc,2 + K f ,2)2(A.8)Using values of A and K calculated for the channel and floodplain, we can solve for α:αv =(Ac + A f )2[K3cA2c+K3fA2f](Kc + K f )2(A.9)The energy loss (he) between the upstream and downstream nodes is a function of both frictionand expansion/contraction. We are ignoring the effects of expansion and contraction so that112A.3. Governing equationshe depends on friction only:he =∆x(Qc,1 + Qc,2) + ∆xχ(Q f ,1 + Q f ,2)Qc,1 + Q f ,1 + Qc,2 + Q f ,2(A.10)where ∆x is the length of channel in the node and χ is the channel sinuosity. Finally, EquationsA.4, A.9, and A.10 can be substituted into Equation A.1 to solve for the WSE of the upstreamnode, z2 + y2.Iteration procedureEquation A.1 cannot be solved using direct methods, so iteration must be used to converge onthe proper WSE. An initial guess is made for y2 (usually as the flow depth from the previoustimestep). Then the respective channel area for both the upstream and downstream nodes arecalculated asAc = y ∗ Bc (A.11)and the floodplain areas areA f = B f [y− (L f − La)] (A.12)where Bc is the channel width, B f is the width of the floodplain, L f is the height of the flood-plain, and La is the thickness of the active layer.Equation A.1 is solved for y2, and the error is calculated as the difference between the inputand output y2 values. The error divided by a user-defined stabilizing term is subtracted fromthe input y2 term to create the y2 for the next iteration. This continues until the error is lessthan .001 m.Finally, the values from the final iteration are input into the Manning Equation, resultingin the friction slope and discharge for the node:S f = (QKc + K f)2, (A.13)Qc = Q(KcKc + K f), (A.14)Q f = Q−Qc, (A.15)andy f = yc − (L f + La) (A.16)where y f is the flow depth on the floodplain. Channel velocity (Vc) is calculated using Equa-tion A.3, with the Qc and Ac.113A.3. Governing equationsA.3.2 Sediment transportThere are two forms of sediment transport in MAST-1D, bedload transport and suspendedload. In this implementation, it is assumed that all silt and clay (termed washload here) travelsas suspension, regardless of discharge. Sand and gravel may also travel in suspension and bedeposited on the floodplain depending on the flow conditions.BedloadBedload transport is calculated using the Gaeuman et al. (2009) equations, which are a form ofthe Wilcock and Crowe equations that are suitable for large, cobble-bed streams. The sedimenttransport rate is a function of the excess channel-wide shear stress over a grainsize-dependentthreshold. The shear stress exerted on the sediment grains (the skin friction, or τ′) is calculatedfollowing the method in the BAGS Primer (Wilcock et al., 2009) asτ′ = 0.00148ρ ∗ g ∗ (2S f ∗ D65)0.25V1.5c (A.17)where ρ is the density of water, set at 1000 kg/m3 and Vc is channel velocity. The dimensionlessreference shear stress for the mean particle size τ∗rm isτ∗rm = 0.03+0.0221+ e7.1σSG−1.66(A.18)σSG is the standard deviation of the sediment grainsize on the psi scale. This is converted intoa dimensional reference shear stress (τrm) using the Shields equation:τrm = τ∗rm(ρs − ρ)gDg (A.19)where ρs is sediment density and Dg is the mean particle size. A hiding function is used tocalculate the reference shear stress for each particle size:τri = τrm(DiD50)b(A.20)where τri is the reference shear stress and Di is the grain diameter for size class i andb =0.71+ e1.9−Di/3Dg(A.21)where Dg is the mean grain size. The dimensionless transport rate for each size class i dependson φi, the ratio between the shear stress and the reference shear stress for that size, whereφi =τ′τri(A.22)114A.3. Governing equationsThe equation isw∗i =0.002φ7.5i , φi < 1.3514(1− 0.894φ0.5i)4.5, φi ≥ 1.35(A.23)The fractional transport rate is then put into dimensional form and multiplied by its fractionin the bed:qs,i = fiw∗i u∗3Bcρg(ρs − ρ) (A.24)where qs,i is the sediment transport rate for size class i and u∗ is the shear velocity, which isu∗ =(τρ)0.5(A.25)WashloadSediment can enter the suspended load in two ways: 1) via bank erosion, and 2) as incomingload from upstream. It is assumed that washload sediment is neither entrained from nordeposited on the channel bed. It may be deposited on the point bar and thus transferredto the floodplain through lateral migration, or it may be deposited directly on the floodplainthrough overbank deposition (see Section A.3.4). Washload sediment is not entrained fromthe floodplain; the only mechanism for moving washload from the floodplain to the channelis through bank erosion. The amount of deposition on the floodplain is a function of thesediment concentration in the overbank flow scaled by a constant floodplain number (SectionA.4.2):dw =FCQ fB f(A.26)where dw is the average amount of washload sediment deposited on the floodplain per unitchannel length, F is the floodplain number and C is the suspended sediment concentration,C =qw,inQc(A.27)where qw,in is the incoming suspended sediment from the upstream node or boundary feed.The suspended sediment transport rate (qw) is calculated via conservation of mass:qw = ( fSAL,w Iv,SAL,w + fFP,wE∆t+ fPB,wN∆t+ qw,in − dw∆x) (A.28)where Iv,SAL,w is the incoming substrate sediment due to vertical channel change (SectionA.3.4), E∆t andN∆t are rates of widening and narrowing, respectively, and ∆x is the length ofthe node. ( fSAL,w, fFP,w, and fPB,w refer to the fractions of mud in the active layer, floodplain,and point bar, respectively). Note that N∆t is negative.115A.3. Governing equationsSuspended sand and gravelIf flooding occurs and turbulence is strong enough so that the diffusive forces lifting parti-cles exceed gravitational forces, some sediment travels in suspension and is deposited on thefloodplain. The gravitational forces acting on the grain are characterized by its settling veloc-ity, which is calculated using the emperical formulation derived by Dietrich et al. (1982). Thevelocity is determined via a dimensionless parameter, D∗, which quantifies the ratio betweenthe gravitational force acting on the particle and the viscous properties of the flow:D∗ = (ρs − ρ)(D ∗ 10−3)3gρν2(A.29)where ν is the kinematic viscosity. The dimensionless velocity, W∗i , isW∗i =10a, D∗2 > 0.5D∗25832 , D∗2 ≤ 0.05(A.30)where a isa = 10−3.76715+1.92944log(D∗)−0.09815[log(D∗)]2−0.00575[log(D∗)]3+0.00056[log(D∗)]4 (A.31)The settling velocity vb,i isvb,i =( (ρs − ρ)ρW∗i gν)1/3(A.32)To determine the amount of sediment that deposits on the floodplain, the proportion oftotal suspended sediment that is transported overbank must be calculated. To do so, a Rouseprofile is created. We assume that sediment in the bottom 5% of the profile travels as bedload.By this definition, the suspended sediment transport rate within the channel that occurs belowthe top of the bank isqs,b,i =∫ (LF−LAL)/y0.050.05(1− z)0.95zZdz (A.33)where y is the flow depth in the channel and Z is the Rouse number,Z =vb,iκu∗(A.34)where κ is the von Karman constant (0.4). The sediment transport rate above the level of thefloodplain isqs,o,i =∫ 1(LF−LAL)/y0.05(1− z)0.95zZdz (A.35)The total proportion of overbank suspended sediment that is transported above the level ofthe banks (Po) isPo =0.951− (LF − LAL)/y ∗qs,o,iqs,o,i + qs,b,i(A.36)116A.3. Governing equationsEquations A.33 and A.35 are discretized into 20 segments.Only a portion of sand and gravel in any given size class travels in suspension. The restsaltates along the bed. We define the former portion with a constant, αFS, which ranges be-tween 0 and 1. The sediment concentration of size class i in the overbank water column isCi =qs,i,inαFSPoQc(A.37)where qs,i,in is the incoming sediment feed in size class i. The fraction of overbank sedimentthat deposits on the floodplain per unit channel length isdi =FbedCiQ fB f(A.38)where Fbed is the floodplain number for bed material, a constant.Calculation of total transport and floodplain depositionFor each timestep, the total bedload sediment transport rates per size class, Qs,i, is the weightedsum of the rates for each flow in the duration curve, so thatQs,i =n∑j=1qi,j pj (A.39)where qi,j is the size-specific transport rate and pj is the flow frequency for flow j, and n is thenumber of discharges in the flow duration curve. When running MAST-1D with a hydrograph,n = 1. Qs,i is calculated using Equation A.24. The total suspended sediment transport rate(Qw) isQs,w =n∑j=1qw,j (A.40)Equation A.28 is used to calculate qw,j.The overbank deposition rates dw and di are also duration-averaged sums:dw =n∑j=1dw,j pj (A.41)anddi =n∑j=1di,j pj (A.42)where dw,j and di,j are solved for using Equations A.26 andA.38, respectively.117A.3. Governing equationsA.3.3 Width changeThere are two components to width change that result in sediment exchanges: channel widen-ing via erosion and narrowing from vegetation encroachment onto bars. When rates of erosionand vegetation growth are not equal, width change occurs. When they are equal, there is mi-gration but no net change in width. The governing equations are briefly described here. Fulldetails on the theory and rationale can be found in Chapter 2.Channel wideningOur simple model of channel widening only relates bank erosion to sediment transport capac-ity. Channel widening occurs when a supply-normalized unit transport rate of the upper tail ofthe grainsize distribution, qsCmax, exceeds a threshold, qscr. We define the supply-normalizedunit coarse transport rate asqsCmax = qsC/ fC (A.43)where qsC is the unit sediment transport rate of the coarse end of the surface sediment mixtureand fC is the fraction of that group of sizes present in the bed. qsC is calculated via the equa-tions in Section A.3.2. qsCmax represents the transport rate expected with an unlimited supplyof coarse sediment. There is currently no straightforward way to determine the threshold unittransport rate qscr, and for now it is a user-defined constant.Once bank erosion is initiated (qsCmax > qscr, floodplain sediment mixes with the activelayer adjacent to the bank, and the magnitude of bank erosion depends on the ability of theflow to transport this near-bank sediment. When coarse sediment supply from the bank ex-ceeds the transport capacity, it will build up along the bank toe and protect it from furthererosion. The near-bank sediment transport capacity, qsNB, is a function of the grainsize distri-bution of the near bank region, which is defined byfi,NB = α f fi + (1− α f ) fi,FP (A.44)where fi,NB is the near-bank fraction of size class i, fi is the fraction in the active layer, fi,FP isthe fraction in the floodplain, and α f is a mixing constant that ranges between 0 and 1. qsNBis calculated using the bedload relations outlined in Section A.3.2, with fi,NB as the grainsizedistribution. The portion of qsi,NB that transports coarse floodplain material, qsC,FP, isqsC,FP =qsC,NBfC,NBfC,FP(1− α f ) (A.45)where qsC,NB is the unit coarse sediment transport rate of the near-bank mixture and fC,FP isthe fraction of coarse material in the floodplain. The bank erosion rate (E/∆t) isE∆t=0, qsCmax ≤ qscr(qsC,FP)/(LF ∗ fC,FP), qsCmax > qscr (A.46)118A.3. Governing equationsChannel narrowingThe narrowing function is: Channel narrowing results from multiple interrelated processes,including deposition on bars, degradation leading to the development of benches, and en-croachment of vegetation onto exposed surfaces. We assume that channel narrowing only oc-curs during relatively low flows. The rate of vegetation enroachment is treated as a constant,αn:N∆t=−αn ∗ (Bc − Bmin), τ < τr0, τ ≥ τr (A.47)Bmin is a constant user-defined minimum width and Bc − Bmin represents the unvegetatedpoint bar. τr represents a reference shear stress, below which flow is low enough to leavesurfaces exposed for colonization.A.3.4 Sediment reservoir exchangesThere are five sediment reservoir types in MAST-1D: the load, active layer, floodplain, chan-nel substrate, and floodplain substrate. There are multiple layers of substrate, and layers maybe added, removed, and combined, depending on the evolution of the bed. Substrate is ac-counted for in two zones: one under the channel region and the other beneath the floodplain.The size-specific amount of sediment for all reservoirs except the load is determined by a con-servation of mass equation:∆Vr, i∆t= (1− λ)∆Sr,i∆t(A.48)where Vr is the new volume of material size class i in reservoir r, λ is porosity, and ∆Sr,i is thechange in storage of sediment in a given size class. ∆Sr,i is calculated as the difference betweenthe inputs (I) and the outputs (O) during each timestep. For the floodplain and substrate types,∆Sr,i = (Im,r,i + Iv,r,i)− (Om,r,i +Ov,r,i) (A.49)where Im,r,i and Om,r,i are inputs and outputs due to net erosion and Iv,r,i and Ov,r,i are in-puts and outputs due to the vertical change in the position of the bed. The active layer hasadditional terms because it exchanges material with the sediment load:∆SAL = (Im,r,i + Iv,r,i + Qs,in,i)− (Om,r,i +Ov,r,i + Qs,i) (A.50)Qs,in,i is the bedload feed and Qs,i is the bedload for size i.The mass balance for the sediment load is described in terms of suspended sediment dis-charge, Equations A.28 and A.40. I and O terms for each reservoir are described in more detailbelow.119A.3. Governing equationsLateral exchangesLateral reservoir exchanges are driven by channel migration. The size distribution of substrateunderlying the channel may be different from that below the floodplain because of selectivedeposition onto the channel and point bar and the size-specific supply of sediment from up-stream nodes. Therefore, there are two substrate reservoirs, one each for the channel andfloodplain. As the channel moves across the floodplain, it lay above old floodplain substrate,which becomes incorporated into the channel substrate. It also abandons a portion of both itsunderlying substrate, which mixes into the floodplain substrate reservoir. For each grainsizeclass i in the substrate,Im,SF,i = Om,SC,i = −(1− λ) N∆t LS fSC,i∆x (A.51)andOm,SF,i = Im,SC,i = (1− λ) E∆t LS fSF,i∆x (A.52)LS is the height of the substrate. The subscript SF represents the floodplain portion of thesubstrate and SC denotes the channel substrate. Washload from the floodplain (subscript FP)goes straight into the sediment load (subscript L) and does not interact with the active layer:Om,FP,w = Im,L,w = (1− λ) E∆t LF fw∆x (A.53)where the subscript w denotes the size class traveling as suspended load. Bed material-sizedsediment that erodes from the floodplain is exchanged directly with the active layer:Om,FP,i = Im,AL,i = (1− λ) E∆t LS fFP,i∆x (A.54)The subscript AL refers to the active layer.Inputs to the floodplain from the active layer and load occur in two ways: from overbankdeposition and from channel narrowing. The latter is modulated by a point bar reservoir,which has a grainsize distribution and height, LPB. The fraction of pointbar that is composedof washload sediment isfPB,w = 1−(1+k¯Qs,wQs,b)−1(A.55)where Qs,b is the duration-averaged bed material load (∑Qs,i) and k¯ is a user-defined relation-ship between the proportion between suspended and bedload in the load and that proportionon the point bar. The export of suspended sediment from the load and the input into thefloodplain via vegetation encroachment and overbank deposition becomesOm,AL,w = Im,FP,w = (1− λ) fPB,wLPB N∆t∆x + dw∆x (A.56)120A.3. Governing equationsThe input of each size class of bed material into the floodplain due to vegetation encroachmentand overbank deposition is described asOm,AL,i = Im,FP,i = (1− λ) fPB,iLPB N∆t∆x + di∆x (A.57)wherefPB,i = (1− fPB,w)[αbar fAL,i + (1− αbar) fQs,i] (A.58)and αbar is the proportion of point bar bed material sediment that is sourced from the activelayer as opposed to the load.Vertical exchangesVertical reservoir exchanges (subscript v) are driven by the Exner equation, where∆z∆t=1Bc(1− λ) ∗(∑(Im,AL,i + Qs, f )−Qs∆x(A.59)∆z∆t is the rate of bed elevation change and Qs, f is the total incoming bedload feed and Qs isthe computed load, with is exported to the next downstream node. If the channel is aggrading(∆z∆t > 0), then the uppermost substrate channel and floodplain layers receive bed materialsediment from the active layer and floodplain, respectively:Iv,SC,i = Ov,AL,i =∆z∆t Bc∆x(1− λ)(αbed fAL,i + (1− αbed fL,i), z > 00, z ≤ 0 (A.60)where αbed is the proportion of sediment entering the substrate from the bed vs. the bedload,the subscript v refers to vertical exchange, andIv,SF,i = Ov,FP,i =∆z∆t ∆xχ B f (1− λ) fFP,i, z > 00, z ≤ 0 (A.61)If the bed is degrading (∆z∆t < 0), then the uppermost substrate layers provides sediment to theactive layer and active floodplain:Ov,SC,i = Iv,AL,i =−∆z∆t Bc∆x(1− λ) fSC,i, z < 00, z ≥ 0 (A.62)andOv,SF,i = Iv,F,i =−∆z∆t ∆xχ Bc(1− λ) fSF,i, z < 00, z ≥ 0 (A.63)121A.3. Governing equationsIt is assumed that washload sediment does not infiltrate into the bed during aggradation (i.e.Iv,SC,w = Ov,AL,w = 0). However, during degradation, fine sediment from the uppermostsubstrate layer is entrained and enters the load:Ov,SC,w = Iv,L,w =−∆z∆t Bc∆x(1− λ) fSC,w, z < 00, z ≥ 0 (A.64)Fine sediment in the floodplain is exchanged with the uppermost floodplain substrate in thesame way as bed material, using Equations A.61 and A.63, but replacing i with w.Exchanges in bedrock channelsThe user may specify nodes that are underlain by non-erodable material such as bedrock. Thechannel is only allowed to degrade to a user-defined threshold, after which ∆z∆t is set at 0 and thechannel becomes ‘partly alluvial.’ Conservation of mass is maintained by adjusting the totalvolume of the active layer instead of sourcing material from the substrate. In partly-alluvialnodes, washload sediment may be evacuated from the active layer when total sediment inputsexceed outputs:Qs,adj,w = Qs,in,w − fw,AL ∗ (Qs,in −Qs) (A.65)Equation A.65 ensures that the active layer grainsize distribution of a partly-alluvial node doesnot become dominated by fine sediment. The change in washload volume in the active layer(∆SAL,w) is∆SAL,w = (Im,r,w + Qs,in,w)− (Om,r,w + Qs,adj,w) (A.66)When the inputs to a partly-alluvial node exceed outputs, the size-specific bedload exiting thenode is adjusted to fill the active layer with bed material sediment:∆SAL,i = (Im,r,i + Qs,in,i)− (Om,r,i + Qs,adj,i) (A.67)whereQs,adj,i = Qs,in,i − (Qs,in −Qs)[ fi,ALαpa + fi,L(1− αpa)] (A.68)Qs,adj,i is the adjusted sediment load for size class i, Qs,in and Qs,out are the total sediment feedand load, respectively, and αpa is the ratio between the volume of a fully alluvial active layerand the current volume:αpa =VALLALBc∆x(A.69)Equations A.67 and A.66 replace Equation A.50 in partly alluvial nodes. When αpa is greaterthan or equal to 1, the bed is no longer partially alluvial, and bed elevation changes may occuragain.122A.3. Governing equationsReservoir geometry and grainsize distributionsFor each timestep, the volume for each reservoir size class (including the suspended load)is calculated by multiplying the result from Equation A.48 by the length of the timestep andadding it to the initial volume:Vr,i = V0,r,i +∆Vr,i∆t∆t (A.70)where V0,r,i is the volume after the previous timestep of length t (and, in the case of the finesediment fraction, i is replaced with w). The total volume Vr isVr =n∑i=1Vr,i +Vr,w (A.71)and the size fractions arefr,i =Vr,iVr(A.72)for the bed material load andfr,w =Vr,wVr(A.73)for the washload. Reservoir volumes and size fractions are updated during each timestepfor the active layer (AL), floodplain (F), and substrate layers (S). Channel width is calculatedbased on the encroachment and erosion rates:Bc = Bc,0 +(N∆t+E∆t)∆t (A.74)where Bc,0 is the previous channel width. The floodplain width isB f = B f ,0 −(N∆t+E∆t)χ∆t (A.75)where B f ,0 is the previous floodplain width. The floodplain height then becomesL f = VF(B f∆xχ)−1(A.76)The height of the uppermost substrate layer is a function of the vertical bed change:Ls = Ls,0 +∆z∆t∆t (A.77)where Ls,0 is the previous upper substrate height. The heights of deeper substrate layers donot change. The new bed elevation z isz = z0 +∆z∆t∆t (A.78)123A.3. Governing equationswhere z0 is the previous bed elevation.A.3.5 Substrate maintenanceAll substrate layers are initially set at a uniform thickness, LS,0. Substrate layers are split orcombined when the thickness of the uppermost layer, LS, exceeds or drops below thicknessthresholds or when aggradation begins to approach the height of the floodplain.StratigraphyIn order to preserve the stratigraphy of subsurface deposits, the substrate reservoirs are mod-ified as the river aggrades and degrades. If the river aggrades over a defined threshold (Lsp),the uppermost substrate layer is split in two, creating a new stratigraphic layer. The thicknessof this new layer isLs,new = Ls − Ls,0 (A.79)and the thickness of the old layer becomes Ls,0. New volumes are calculated for the layers asVr = LsBc∆x (A.80)where r represents the channel (SC) and floodplain (SF) substrate reservoirs. The grainsizedistribution of the new layers is the same as in the respective parent layers.If the uppermost substrate layer thickness is reduced below a threshold due to degradation,that layer is combined with the layer below it. The thickness of the new layer becomesLs,new = Ls + Ls,−1 (A.81)where Ls,−1 is the thickness of the lower layer. The grainsize distribution of the combined layeris a weighted average of the two layersfS,i = fS,iLs + fS,−1,iLS,−1 (A.82)where fS,i and fS,−1,i are the fractions for size i of the upper and lower layers, respectively. Thevolume of the reservoir is solved using Equation A.80.AvulsionAvulsion (the rapid shift of the dominant channel to a new location) is common in alluvialrivers when the channel is blocked by sediment, large wood, or other obstructions. In MAST-1D, avulsion is triggered in model nodes experiencing high levels of aggradation where thebed elevation approaches that of the floodplain. The implication is that avulsions occur inareas of persistent sediment deposition (such as deltas). When the floodplain height (L f ) dips124A.4. Initial and boundary conditionsbelow a user-defined threshold value, the bed elevation lowers by a spacing constant (Lav):znew = zold − Lav (A.83)where zold is the pre-avulsion bed elevation and znew is the resulting elevation. The surface ofthe new channel becomes active, so that the volume of floodplain material added to the activelayer for size class i (ALin,i) in the avulsed node isALin,i = αa ∗ Bc ∗ LAL ∗ fi,FP ∗ ∆x (A.84)where αa is the fraction of channel that avulses, LAL is the thickness of the active layer, andfi,FP is the fraction of size class i in the floodplain. We make the simplification that the aban-doned portion of channel becomes vegetated immediately. Channel substrate and the avulsedportion of the active layer are incorporated into the floodplain. Floodplain material is mixedinto the active layer to represent the surface material forming the base of the new channel. Thevolume of channel sediment sequestered into the floodplain reservoir (FPin,i) isFPin,i = [αa ∗ Bc ∗ LAL ∗ fi,AL + Bc ∗ La ∗ fi,SC] ∗ ∆x (A.85)The grainsize distribution of the active layer is adjusted to incorporate floodplain materialunder the new channel:fAL,i = αa fFP,i + (1− αa) fAL,i (A.86)Sediment from the substrate and the old active layer are incorporated into the floodplain toconserve mass and a new floodplain volume is calculated:VFP,i = VFP,old,i +VSC,i +VSF,i +VAL,old,i −VAL,new,i (A.87)where VAL,old,i and VAL,new,i are the old and new active layer size-specific volumes, respectivelyand VFPold,i is the old floodplain volume. If the uppermost substrate layer is lower or higherthan Lav, then it is either split or combined using the methodology in Section A.3.5 beforeperforming EquationA.87 so that LFP = Lav.A.4 Initial and boundary conditionsA.4.1 Boundary conditionsAs a 1-dimensional model, MAST-1D requires two boundary conditions: an upstream sedi-ment supply, and a downstream hydraulic boundary.125A.4. Initial and boundary conditionsSediment supplyBy default, the upstream sediment feed is a user-defined proportion of the bedload sedimentcapacity, whereQs,i,in = C f ,bQs,i,cap (A.88)andQs,w,in = C f ,wn∑j=1Qs,i,in (A.89)Qs,i,in and Qs,w,in are the total size-specific feed rates for the bedload and suspended load,respectively, Qs,i,cap is the size-specific bedload transport capacity, calculated using EquationsA.17-A.25 and a user-specified size distribution, and C f ,b is proportion of capacity that is inputas bedload feed. Suspended feed is assumed to be a set multiple of initial beload capacity,modulated by the constant C f ,w.The user may allow C f ,b and C f ,w to change over time, either directly or via a function.When a flow duration curve is used, Qs,i,cap is determined using the duration-weighted sumof sediment transport rate for all flows in the curve. When hydraulics are calculated using ahydrograph, Qs,i,cap is recalculated for each new flow, using the initial geometric and sedimentconditions. In other words, a sediment capacity rating curve is used.HydraulicsThe water surface elevation is set at the downstream boundary. The user has the option ofsetting the boundary as a constant or changing it manually (for example, if the modeled reachends at a reservoir or shoreline). If the water surface elevation is not known, it is calculatedassuming normal flow conditions and a wide channel with the Manning equation:z + y = z +(ncQBcS0.5c)3/5(A.90)where Sc is the channel bed slope. If y is greater than the channel depth (L f − La), then theflow depth is solved for iteratively using Equations A.1-A.10. ∆x is set at 100 m, and iterationscontinue until the upper and lower channel depths (y1 and y2) converge.When a flow duration curve is known, the boundary water surface elevation is solved foreach flow at the beginning of the run. If a hydrograph is used, then the boundary condition issolved for each discharge using the initial channel geometry and sediment conditions.A.4.2 Initial conditionsThe initial channel geometry and grainsize distribution of the active layer are supplied by theuser. Given these conditions, as well as the upstream sediment boundary and sediment mixingparameters, the floodplain grainsize distribution and floodplain number are calculated so thatthe model river would be in equilibrium if channel width and migration rate were constant.126A.5. Variable listFloodplain numberThe floodplain number F determines the proportion of the suspended load that is depositedduring each timestep. The depth of fine sediment on the floodplain during the initial condition(Lw) is assumed to be equivalent to the depth during equilibrium and is calculated asLw,0 = L f − (La + Lpb) (A.91)The floodplain number then is the proportion that the sediment concentration would haveto be reduced to reproduce Lw within an average floodplain reworking time (B f /m), assum-ing that mud is being transported at the capacity determined in Equation A.89 and given aconstant user-defined migration rate m:F =Lw,0md f ull(A.92)where d f ull is the average suspended sediment deposition rate per unit floodplain if the entiresediment load were deposited. The floodplain number can also be set manually by the user.Initial floodplain grainsize distributionThe initial grainsize distribution for the floodplain and floodplain substrate reservoirs is acombiniation of the distributions of the active layer and point bar:fF,i,0 =fAL,i,0Lal + fPB,i,0LpbL f(A.93)andfF,i,0 =fAL,w,0Lal + fPB,w,0Lpb + Lw,0L f(A.94)The subscript 0 refers to the values of the variables at the initial condition.A.5 Variable listTable A.1: MAST-1D list of variablesVariable Unit DescriptionA m2 total area of flowAc m2 area of flow for channelA f m2 area of flow for floodplainAL - active layerBc m channel width127A.5. Variable listTable A.1: MAST-1D list of variablesVariable Unit DescriptionB f m floodplain widthBmin m minimum channel widthC - washload sediment concentrationC f ,b - bedload sediment feed boundary capacity pa-rameterCn,a - addition constant for Manning’s nCn,m - multiplier for Manning’s nC f ,w - suspended sediment feed boundary capacityparameterCi - concentration of size class i in overbank flowCmax - size classes of ‘coarse’ particles, roughly greaterthan the D90dw m2/s average washload deposition rate per unitlength on floodplaindi m2/s average suspended sand/gravel depositionrate per unit length on floodplain for size classid f ull m2/s average suspended sediment deposition rateper unit length on floodplain if entire sedimentload were depositedDg m mean grain sizeDi m, mm grain size at percentile iE m bank erosionF - floodplain numberFbed - floodplain number for bed materialFP - floodplainfi - fraction of size i in sediment mixturefi,NB - fraction of size i in near-bank sediment mixturefw - fraction of suspended-size sediment in mixtureg m/s2 gravitational accelerationhe m energy head lossI - reservoir inputsk¯ - coefficient of suspended and bedload sedimentin the load and on the point barK - conveyance128A.5. Variable listTable A.1: MAST-1D list of variablesVariable Unit DescriptionKc - conveyance for channelK f - conveyance for floodplainLa m thickness of active layerLav m bed lowering during avulsionL f m floodplain heightLpb m thickness of point barLs m substrate thicknessLw m thickness of fine sediment on floodplain duringinitial conditionN m vegetation encroachmentnc - Manning’s n for channeln f - Manning’s n for floodplainO - outputsPB - point barpj - flow frequency for flow jPo - proportion of suspended load for size class ithat is overbankqs,b,i - Rouse integral for in-channel portion of flowqs,b,i - Rouse integral for overbank portion of flowqsCmax m3/s transport rate for the coarse fractionqscr - mobility threshold for bank erosionQ m3/s total dischargeQc m3/s channel dischargeQ f m3/s floodplain dischargeQs,adj,i m3/s adjusted fractional sediment load for partially-alluvial channelQs,b,i m3/s proportion of suspended sand/gravel travelingwithin the banksQs, f m3/s total bedload sediment feedQs,i m3/s total sediment load for size i over durationcurveQs,in,i m3/s total sediment feed for size i over durationcurveQs,o,i m3/s proportion of suspended sand/gravel travelingabove the banks129A.5. Variable listTable A.1: MAST-1D list of variablesVariable Unit DescriptionQs,w m3/s total suspended sediment load over durationcurveqs,i m3/s channel-wide sediment load for size class iqw m3/s suspended sediment loadqw,in m3/s suspended sediment feedS m3 storage of sediment in reservoirSC - channel substrateSc - channel bed slopeS f - energy slopeSF - floodplain substrateu∗ m/s shear velocityv¯ m/s cross-sectional average velocityvc m/s flow velocity in channelv f m/s flow velocity on floodplainV m3 volumew∗i - dimensionless transport for size ix m channel-wise coordinatey m flow depth in the channelz m bed elevationZ − Rouse numberαa - portion of channel that avulsesαbar - fraction of point bar bed material sedimentsourced from active layerαbed - fraction of sediment entering substrate frombed vs. bedloadαe - channel widening coefficientα f - proportion of active layer transport in near-bank regionαn - channel narrowing coefficientαpa - fraction between volume of partly-alluvial fullyalluvial active layerαv - weighting coefficient for average velocityλ - porosityν m2/s kinematic viscosityρ kg/m3 density of water130A.5. Variable listTable A.1: MAST-1D list of variablesVariable Unit Descriptionρs kg/m3 sediment densityσSG - standard deviation of sediment mixture on psiscaleτ′ skin friction (shear stress on grains)τcr N/m2 shear stress needed to entrain reference D84τr N/m2 reference shear stress below which vegetationencroachment occursτ∗rm - dimensionless reference shear stress for themean particle sizeτrm N/m2 reference shear stress for the mean particle sizeτri N/m2 reference shear stress for size class iφi - ratio between skin friction and reference shearstress of size classχ - channel sinuosity131Appendix BMAST-1D parameters and calibrationHere we present initial conditions and constants for the MAST-1D simulations. For Chapters2 and 4, all nodes are initially homogenous. For Chapter 3, the study area was broken into 21segments that correspond to morphologically similar portions of channel (Figure B.1). Initialconditions are presented in Tables B.1 to B.3. Reaches with ‘Canyon’ designations do notmigrate and cannot degrade by more than 0.2 m. Slopes were extracted from a DEM frombefore dam removal. Sinuosity is kept constant in MAST-1D. Values are averages of sinuositiescalculated from available air photos (see Chapter 3). Valley widths presented in Table B.2represent valley area divided by the valley centerline for each node. Valley margins weredigitized using a LiDAR DEM in conjunction with air photos. Valley widths in Tables B.1 andB.3 were assumed spatially constant.Initial channel widths were calibrated so that the average annual sediment transport capac-ity, calcuated using a duration curve derived from daily discharge from Elwha River between1927 and 1994, matched the annually-averaged sediment load derived from measurements ofaccumulation in Lake Mills from 1994. The calibrations, along with a comparision of sedimentload GSDs calculated via MAST-1D and measured in the reservoir, are presented in FiguresB.2-B.4.Other initial conditions that are eqivalent for all runs are presented in Table B.4. The initialfloodplain depth was taken from an emperical hydrologic geometry relation for coastal PacificNorthwest rivers presented by Castro and Jackson (2001). The initial thickness of overbanksediment on the floodplain is averaged from measurements from a bank survey conducted in2015 (see Appendix C). Other constants and calibration parameters are in Table B.5.Calibrations of the dam removal sediment supply paramter C for various sediment supplyscenarios described in see Chapter 3 are presented in Figure B.5. Field data were extractedfrom unpublished structure-from-motion analysis of air photos of the former reservoir andwere provided by the US Geological Survey.132Table B.1: Initial conditions for simulations from Chapter 2Node Valleylength (m)Valleywidth (m)Canyon Bc (m) ∆x(m) Sinuosity Slope0 4902 581 No 81 5000 1.02 0.00691, 4902 581 No 81 5000 1.02 0.00692 4902 581 No 81 5000 1.02 0.00693 4902 581 No 81 5000 1.02 0.00694 4902 581 No 81 5000 1.02 0.0069Table B.2: Initial conditions for simulations from Chapter 3Node** Valleylength (m)Valleywidth (m)Canyon Bc (m) ∆x(m) Sinuosity Slope0 237 25 Yes 14.4 238 1.01 0.0221, 680 180 Yes 50.6 731 1.08 0.00812 592 41 Yes 32.5 663 1.12 0.0153 737 340 No 111 779 1.06 0.00814 662 397 No 111 731 1.10 0.00815 871 523 No 111 918 1.05 0.00816 616 187 No 111 790 1.28 0.00817 686 504 No 81 735 1.07 0.00698 538 502 No 81 587 1.09 0.00699 704 361 No 81 718 1.02 0.006910 704 358 No 81 671 0.95 0.006911 223 61 Yes 28 235 1.06 0.006912 574 344 No 81 608 1.06 0.006913 647 334 No 81 668 1.03 0.006914 653 427 No 81 - 1.07 0.006915 502 516 No 81 - 1.07* 0.006916 614 439 No 81 - 1.07* 0.006917 502 316 No 81 - 1.07* 0.006918 819 92 Yes 81 - 1.07* 0.006919 741 209 No 81 - 1.07* 0.006920 797 189 No 81 - 1.07* 0.0069*Aldwell sinuosities are average of all upstream nodes**Node count begins at Glines Canyon Dam and increase downstream133Figure B.1: Map with node locations for simulations in Chapter 3134Table B.3: Initial conditions for simulations from Chapter 4Node Valleylength (m)Valleywidth (m)Canyon Bc (m) ∆x(m) Sinuosity Slope0 2234 594 No 94 2279 1.02 0.00741, 2234 594 No 94 2279 1.02 0.00742 2234 594 No 94 2279 1.02 0.00743 2234 594 No 94 2279 1.02 0.00744 2234 594 No 94 2279 1.02 0.00745 2234 594 No 94 2279 1.02 0.0074Table B.4: Floodplain and substrate initial conditionsVariable Unit Description ValueL f m floodplain height 2.26Ls m substrate thickness 1.0Lw m thickness of fine sediment on flood-plain during initial condition0.14Figure B.2: Calibration of Bc for nodes with a slope of 0.0069 in Chapters 2 and 3135Table B.5: Calibration parameters and other constants. These values were used for allruns unless stated otherwise in the text.Variable Unit Description ValueBmin m minimum channel width 40Cn,a - addition constant for Manning’s n 0.0066Cn,m - multiplier for Manning’s n 1.15Cmax - coarse size classes for bank erosionalgorithms256-1024mmF - floodplain number for mud 0.2Fbed - floodplain number for bed material 0.75g m/s2 gravitational acceleration 9.81k¯ - coefficient of suspended and bed-load sediment in the load and onthe point bar10∗ ∗ −6La m thickness of active layer 0.4Lav m bed lowering during avulsion 1.0Lc m critical floodplain height for avul-sion0.75Lpb m thickness of point bar 1.72Lw m thickness of fine sediment on flood-plain during initial condition 0.14n f - Manning’s n for floodplain 0.1qscr - mobility threshold for bank erosion 10−6αa - avulsion exchange parameter 0.1αbar - fraction of point bar bed materialsediment sourced from active layer1.0αbed - fraction of sediment entering sub-strate from bed vs. bedload0.4αe - bank mobility coefficient 10∗ ∗ −6α f - fraction of channel contribution totransport in near-bank zone0.55αn - channel narrowing coefficient 0.05αpa - fraction between volume of partly-alluvial fully alluvial active layer0.5λ - porosity 0.5ν m2/s kinematic viscosity 1/1300000ρ kg/m3 density of water 1000ρs kg/m3 sediment density 2650τc,n N/m2 critical shear stress for channel nar-rowing32.0136Figure B.3: Calibration of Bc for nodes with a slope of 0.0081 in Chapter 3137Figure B.4: [Calibration of Bc for nodes with a slope of 0.0074 in Chapter 4Figure B.5: Calibration of C for dam removal simulations138Appendix CField dataC.1 IntroductionBank surveying and sediment sampling were conducted August-September 2015 to charac-terize the amount and size of sediment supplied from the floodplain. Elwha River dischargeduring the period ranged between 6 and 27 m3/s but were typically around 7 m3/s. FigureC.1 shows the locations of the samples. Both subsurface bulk sampling and photosieving areused to characterize the grainsize of the channel and floodplain. Sampling methods and re-sults are presented in Sections C.2 and C.3, respectively. Surveying of the bank stratigraphy isdescribed in Section C.4.139C.1.IntroductionFigure C.1: Map with locations of sediment samples and survey line140C.2. Subsurface bulk samplingC.2 Subsurface bulk samplingBulk subsurface sediment samples were collected at 8 point bar heads, 6 cutbank collapsedeposits/toes, and on other point bar locations from immediately above Glines Canyon Damto the Strait of Juan de Fuca. 1 sample of Lake Mills reservoir material was also extracted.We followed the guidelines outlined in Church et al. (1987) and Bunte and Abt (2001) as closelyas possible. Church et al. (1987) recommends that samples sizes are large enough so that thecoarsest size class constitutes no more than 5% of the samples mass for rivers with grains>128mm. This was not feasible on the cobble-bedded Elwha, and our individual samples typicallycontain ~20% of the coarsest grains. Samples sizes were typically about 500 kg. When all pointbar head subsurface samples are composited, the largest (360-512 mm) grains make up 12% ofthe 2808 kg sample. For bank subsurface samples, the composite is 2983 kg and the 360-512mm size class makes up 11% of the sample. Subsurface grainsize data for point bar heads arein Table C.1 and bank deposit grainsizes are listed in Table C.2. Other bulk samples, includingthose on the middle and tails of point bars and in Lake Mills reservoir are presented in TableC.3.141C.2. Subsurface bulk samplingTable C.1: Bulk samples on point bar headsELW02 ELW06 ELW01 ELW09 ELW03 ELW08 ELW07Easting 455741 456544 457092 458027 458826 458315 458115Northing 5319583 5323014 5324770 5326777 5329405 5331708 5332339Size0.0625 0.62 1.11 0.52 0.46 0.41 0.86 0.750.0884 1.02 1.54 0.77 0.69 0.78 2.18 1.260.125 1.43 1.66 0.96 0.79 1.06 2.25 1.820.177 2.22 2.10 1.29 1.07 1.66 3.81 2.780.25 3.11 2.75 1.76 1.46 2.26 5.99 3.940.354 4.66 4.09 2.77 2.50 3.24 9.59 5.540.5 6.00 5.40 4.07 3.99 4.38 11.69 6.870.707 7.19 6.66 5.67 6.73 5.96 12.82 8.441 8.33 7.89 7.42 10.84 7.80 13.56 10.901.41 9.72 9.43 9.69 15.41 9.69 14.66 14.772 11.17 10.95 12.12 18.96 11.27 16.13 18.722.83 13.58 13.26 15.54 22.92 13.29 19.06 24.104 16.23 15.73 19.21 26.47 15.63 22.54 28.725.66 19.03 17.89 22.89 29.68 18.28 26.53 33.368 21.96 20.22 26.90 32.97 21.35 29.18 37.7211.3 24.61 23.50 31.29 36.48 24.53 33.40 42.4016 27.98 26.30 36.99 40.41 28.26 39.50 47.2222.6 31.36 30.82 43.34 47.11 32.13 45.31 53.3832 38.62 35.40 52.93 53.16 35.87 49.29 58.5245.3 44.69 40.05 60.75 61.60 41.75 57.29 69.8064 50.95 46.13 70.99 70.90 48.73 67.60 82.0390.5 57.24 51.89 79.39 81.26 55.68 77.80 88.14128 65.74 58.26 89.44 88.52 61.63 93.24 94.26181 75.10 61.81 95.27 100.00 73.25 100.00 100.00256 88.48 72.90 95.27 100.00 89.16 100.00 100.00362 88.48 82.54 100.00 100.00 89.16 100.00 100.00512 100.00 100.00 100.00 100.00 100.00 100.00 100.00D16 3.88 4.18 2.95 1.50 4.20 1.94 1.58D50 60.73 80.79 28.78 26.70 68.19 33.00 18.71D84 227.95 372.70 106.10 103.14 228.80 104.03 71.57142C.2. Subsurface bulk samplingTable C.2: Bulk samples on cutbank toes and collapse depositsELW02 ELW06 ELW03 ELW01 ELW08 ELW07Easting 455850 456520 458772 457092 458267 458198Northing 5319375 5322916 5329334 5324770 5331886 5332397Size0.0625 0.33 0.11 1.81 0.27 0.08 0.050.0884 0.72 0.25 3.14 0.55 0.41 0.120.125 0.86 0.44 4.36 0.79 0.69 0.190.177 1.22 0.82 5.95 1.18 0.92 0.370.25 1.71 1.27 7.12 1.68 1.40 0.630.354 2.71 2.10 8.32 2.69 1.95 1.070.5 3.98 3.08 9.44 4.03 2.68 1.460.707 6.05 4.29 10.89 6.16 4.12 1.851 9.51 5.77 12.35 9.07 6.50 2.311.41 14.25 7.58 13.56 12.55 9.40 2.892 18.26 9.21 14.40 15.35 11.81 3.402.83 22.97 11.52 15.46 18.65 14.96 4.044 26.76 13.91 16.68 21.44 18.56 4.785.66 30.06 16.40 17.94 24.58 23.12 5.468 32.57 19.09 20.54 28.04 25.62 6.3211.3 35.86 21.55 23.12 31.58 28.94 7.5716 38.68 24.20 25.89 36.00 32.65 9.4022.6 41.97 27.13 30.28 40.74 36.96 11.5932 46.98 28.27 36.33 44.13 43.42 15.1645.3 54.10 31.86 43.26 49.73 52.61 24.9364 61.69 36.29 51.56 58.37 62.07 36.5490.5 70.45 42.60 64.65 68.47 71.84 50.22128 81.88 47.59 80.63 78.92 85.00 60.65181 93.88 57.09 90.66 88.84 100.00 69.65256 100.00 72.79 95.67 91.95 100.00 91.10362 100.00 83.20 100.00 100.00 100.00 100.00512 100.00 100.00 100.00 100.00 100.00 100.00D16 1.65 5.35 3.30 2.14 3.13 32.97D50 37.06 139.76 59.96 45.75 41.02 90.01D84 136.09 368.09 143.81 152.87 124.69 228.24143C.2. Subsurface bulk samplingTable C.3: Bulk samples on other surfaces. The suffix PBH refers to the upper portion ofthe point bar. MI represents the middle of the point bar, and TL refers to the tail. TRrefers to a sample from a terrace deposit in the former Lake Mills reservoir.ELW05TR ELW04PBH ELW04PBT ELW07MI ELW07TLEasting - 455132 455134 458181 458094Northing - 5316515 5316507 5332443 5332573Size0.0625 0.11 0.18 0.16 0.28 0.440.0884 0.41 0.32 0.32 0.55 0.740.125 0.41 0.43 0.49 0.85 1.080.177 0.82 0.62 0.86 1.45 1.760.25 1.54 0.93 1.53 2.31 2.780.354 3.94 1.74 3.23 3.49 4.270.5 8.09 2.93 5.55 4.67 6.940.707 15.20 4.42 8.03 6.84 11.461 25.85 6.01 10.18 10.81 17.451.41 35.59 7.78 12.16 15.54 23.872 45.55 9.48 13.99 18.91 28.782.83 54.31 12.15 16.64 22.40 34.524 60.48 15.34 19.84 25.14 40.275.66 65.56 18.75 23.46 27.82 45.358 70.24 22.52 27.81 30.74 50.1511.3 74.87 26.82 32.33 33.72 56.3916 80.98 32.43 38.91 37.76 65.0222.6 86.78 37.40 46.66 42.73 76.1532 90.01 44.92 56.01 52.51 89.0345.3 95.01 53.01 65.52 61.19 98.7164 97.94 59.07 76.82 71.75 100.0090.5 100.00 65.98 94.65 85.04 100.00128 100.00 76.96 100.00 95.82 100.00181 100.00 91.37 100.00 100.00 100.00256 100.00 100.00 100.00 100.00 100.00362 100.00 100.00 100.00 100.00 100.00512 100.00 100.00 100.00 100.00 100.00D16 0.73 4.28 2.60 1.48 0.92D50 2.39 39.78 25.61 29.28 7.91D84 19.17 151.61 73.59 88.08 27.95144C.3. PhotosievingC.3 PhotosievingTo characterize longitudinal variability in grainsize, photos were taken of sediment on pointbar heads and cutbank toe deposits using a Go-Pro camera with a fish-eye lens. A samplingsquare ranging between 2x2 and 3x3 meters was delineated on the ground (most sampleswere 2x2 m). The camera was held at the edge of the grid with a slight tilt at a height of 2 m.Wet areas and deposits with abundant woody debris and vegetation were avoided whereverpossible.Data on sediment size was extracted from the photos by Jane Walden at Seattle Universityusing Digital Gravelometer software (http://www.sedimetrics.com). The software automati-cally corrects for camera tilt. We did not correct for the fisheye lens, but the samples were inthe middle of the photo and we expect error from distortion to be small relative to total error.We followed software recommendations and truncated the grainsize distribution at approxi-mately 32 mm.Sediment size metrics extracted from the photos are presented in Table C.4 for point barheads, Table C.5 for cutbanks, and Table C.6 for point bar and bank photos corresponding tobulk sample locations.Table C.4: Photosieved grainsize data for point bar heads taken from photosSample ID Coord* (m) D16 (mm) D50 (mm) D84 (mm)PH88 752.55 64.96 151.88 257.25PH100 1245.41 61.76 142.37 301.47PH90 1317.35 68.28 145.86 233.48PH102 1349.04 46.76 109.91 518.13PH92 1449.63 61.19 142.7 344.44PH94 1798.32 63.66 122.89 262.99PH104 2094.89 62.35 149.66 250.26PH106 2332.02 63.73 131.96 312.46PH107 2481.68 74.12 145.44 216.06PH108 2944.37 69.1 156.53 399.9PH97 3033.98 75.89 152.5 254.92PH121 3202.84 58.58 140.51 299.25PH109 3328.42 86.97 188.83 341.12PH119 3560.37 81.33 170.37 269.2PH117 4167.84 76.48 144.37 226.95PH115 4292.19 86.09 197.59 292.23PH114 4465.62 79.57 189.58 276.31145C.3. PhotosievingTable C.4: Photosieved grainsize data for point bar heads taken from photosSample ID Coord* (m) D16 (mm) D50 (mm) D84 (mm)PH112 4626.86 69.23 153.23 246.81PH111 4961.23 52.49 108.93 176.78PH84 5274.56 74.99 174.56 319.65PH86 5707.68 73.93 161.24 241.33PH28 8338.11 51.35 87.86 158.21PH27 8636.2 60.5 110.95 163.8PH22 8846.52 57.23 106.8 179.96PH26 8863.58 49.28 85.53 125.3PH24 8924.24 50.35 77.91 165.54PH21 8939.78 57.54 108.85 205.66PH23 9102.24 61.28 123.83 183.5PH20 9130.59 52.93 100.46 155.65PH11 9600.59 54.83 110.33 210.76PH13 10230.61 83.99 183.57 260.97PH14 10401.3 43.31 83.72 178.39PH15 10619.84 66.29 137.97 222.57PH17 10807.6 66.48 143.51 262.43PH19 11081.92 79.07 172.32 398.53PH01 11423.9 70.25 148.08 262.53PH03 11523.27 70.06 161.97 405.05PH08 11795.76 99.46 192.07 348.29PH07 12092.94 78.01 173.39 298.4PH06 12219.13 49.78 83.97 164.87PH83 12465.71 61.6 120.67 265.98PH81 12743.08 96.1 247.0 394.51PH77 13278.61 76.78 208.51 360.48PH76 13841.58 98.14 233.32 349.23PH75 14098.83 85.16 201.31 333.9PH73 14322.25 58.21 168.28 293.02PH71 14746.22 75.7 181.37 319.34PH69 14936.11 78.83 156.31 361.44PH68 15058.64 89.6 182.75 300.8PH49 15443.61 61.94 181.6 284.1PH50 15547.54 84.81 198.99 348.23PH66 15906.29 78.57 182.93 284.95146C.3. PhotosievingTable C.4: Photosieved grainsize data for point bar heads taken from photosSample ID Coord* (m) D16 (mm) D50 (mm) D84 (mm)PH52 16060.52 76.17 135.58 251.59PH54 16442.74 70.04 181.5 525.43PH55 16646.04 78.4 185.56 380.15PH58 17112.08 123.14 270.47 432.57PH60 17252.9 70.21 186.13 369.86PH63 17910.96 89.44 219.02 275.97PH45 18229.48 74.51 180.33 371.26PH44 18389.19 72.5 176.92 336.85PH30 18859.2 60.76 141.37 293.61PH31 18974.41 75.58 207.35 338.78PH39 18996.66 98.53 353.45 722.54PH38 19032.93 77.22 166.71 305.91PH33 19155.16 83.93 208.05 376.51PH36 19255.13 97.31 191.25 274.59PH34 19522.44 76.48 157.08 308.13PH48 20424.95 83.16 197.06 499.98PH46 20730.97 80.07 187.72 328.63PH87 21515.89 248.13PH123 23774.4 86.86 192.04 432.67PH122 24547.07 76.7 158.67 341.01*Distance upstream of Strait of Juan de FucaTable C.5: Photosieved grainsize data for bank toes taken from photosSample ID Coord* (m) D16 (mm) D50 (mm) D84 (mm)PHC87 548.64 49.55 108.67 192.28PHC89 868.68 82.48 164.56 269.15PHC99 1205.18 43.07 104.26 213.45PHC101 1321.61 56.33 252.61 431.69PHC91 1357.88 58.17 119.01 253.57PHC96 1828.8 64.91 163.48 300.01PHC103 1871.47 67.78 135.2 233.38PHC105 2316.48 44.93 98.89 155.85147C.3. PhotosievingTable C.5: Photosieved grainsize data for bank toes taken from photosSample ID Coord* (m) D16 (mm) D50 (mm) D84 (mm)PHC120 3276.9 62.03 115.17 185.05PHC116 4311.7 45.27 99.25 194.57PHC113 4511.34 65.77 141.57 262.34PHC110 4824.98 52.87 136.53 258.68PHC85 5340.71 102.54 199.59 461.17PHC16 10789.92 70.24 164.39 238.34PHC18 11097.77 83.07 182.63 332.52PHC10 11864.34 69.09 158.99 314.05PHC09 12130.13 71.18 145.72 280.99PHC82 12696.14 65.18 136.31 246.67PHC80 12911.02 64.17 139.61 258.18PHC79 12989.97 60.01 125.67 223.97PHC78 13315.19 82.28 179.67 335.37PHC74 14173.2 65.47 178.73 279.04PHC72 14707.21 44.04 87.19 157.38PHC70 14853.21 56.83 133.03 241.9PHC67 15193.67 59.94 161.34 355.46PHC65 15524.07 71.47 181.16 385.27PHC51 15731.64 62.68 190.0 319.93PHC53 16308.93 93.7 182.13 348.83PHC56 16729.56 64.7 150.31 352.46PHC57 16881.04 81.63 196.68 466.07PHC59 17244.97 53.22 161.49 295.26PHC61 17511.98 56.5 113.49 237.43PHC62 17853.66 54.23 113.24 218.98PHC64 17976.8 69.53 233.11 389.78PHC43 18221.25 72.25 182.49 322.71PHC42 18539.46 65.56 133.25 223.72PHC41 18552.57 95.6 210.22 392.49PHC29 18850.66 64.42 145.0 232.74PHC32 19110.96 68.99 168.13 285.11PHC35 19404.18 60.28 131.37 243.79PHC47 20594.12 84.32 250.53 450.77*Distance upstream of Strait of Juan de Fuca148C.4. Survey dataTable C.6: Photosieved grainsize data corresponding to bulk sample locationsSample ID D16 (mm) D50 (mm) D84 (mm)ELW01PB 177.02 77.68 284.45ELW09PB 69.08 36.57 114.87ELW03PB 179.39 62.47 301.6ELW06PB 173.54 80.5 282.58ELW02PB 197.98 83.09 294.88ELW08PB 80.25 39.82 122.5ELW07MI(1/3) 137.95 48.45 184.21ELW07MI(2/3) 138.93 57.44 185.19ELW07MI(3/3) 127.07 61.38 238.07ELW07TL 64.12 30.64 105.12ELW04PBT 108.61 53.01 212.58ELW04PBH 133.73 59.88 221.55ELW05TR 105.62 53.76 247.51ELW01CB 51.23 104.85 194.94ELW03CB 82.5 187.53 401.48ELW06CB 78.98 216.37 350.77ELW02CB 65.19 119.83 215.38ELW08CB 34.31 57.32 86.2C.4 Survey dataTo characterize the nature of sediment supply from the banks, we performed a semi-quantitativesurvey of cutbanks. We walked most alluvial and partly-alluvial portions of the channel, start-ing from just downstream of the canyon below Glines Canyon Dam and ending at the Straitof Juan de Fuca (Figure C.1). We noted bank stratigraphy at intervals of 10-20 m. For mostlocations, the height of each stratigraphic unit was estimated from a distance of several metersby eye, and, where necessary, aided by binoculars. Error in these measurements is likely up toabout a meter. In select locations, measurements were made at the cutbank using surveyingtape. Measurements began at the water surface and do not account for submerged bank toedeposits. We also noted presence of large wood along the bank. In addition, basic channelmorphology was noted (pool, riffle, plane bed, etc).The key used to describe morphologic units is presented in Table C.7 and the survey datais in Table C.8. Numbers represent elevation above the water surface in meters. For example,an entry with a CT value of 1 and a CE of 2 represents a location where a 1 meter exposure ofcobble bank is visible above a 1 meter cobble toe deposit. Measurements made using the sur-149C.4. Survey dataTable C.7: Survey data keyA Former Lake Aldwell reservoirAP No bank exposure. Morphology determined by air photos/ground truthing.B Feature present on both banksBD Immobile boulders in channelBK Exposed bankBR Bedrock exposureCB Cobble barFCE Fine cobble/gravel exposureFCT Fine cobble/gravel toeCE Cobble exposureCT Cobble toeFB Fine (fine gravel/sand) barGE Gravel exposureGT Gravel toeL Left bank. *Secondary bank (same location as entry above it)LWD Large woody debris on bankM Middle ElwhaMO Morphologic feature (riffle, pool, or plane bed)OB Overbank depositP PoolPA PaleochannelPB Plane bedPH Photo IDPr PresentRE Reservoir depositsR Right bank. *Secondary bank (same location as entry above it)RI RiffleRR Rip-rapRH ReachS Stagnant waterSP Spacing between observations (m)tall Deposit of reservoir fines; too tall to measure (at least 3 m)TL Till/outwash exposurevey tape are presented in bold. The table is ordered from upstream to downstream. Locationsthat were sampled for photosieving (C.3) are noted. The coordinates for these points can befound in Tables C.4 and C.4.150C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRPH46 M 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - - 0.5 - - - - - - - - - - - -M 20 L RI - - - - - - - - Pr - - - - - -M 20 L RI - 2 - - - - - - - - - - - - -M 20 L RI - 2 - - - - - - - - - - - - -M 20 L RI - 2 - - - - - - - - - - - - -PHC47 M 20 L RI - 1.65 - - - - - - - - - - - - -M 20 L RI - 0.5 - - - - - - - - - - - - -M 20 L P - 0.25 - - - - - - - - - - - - -M 20 L P - 0.25 - - - - - - - - - - - - -M 20 L P - 0.1 - - - - - - - - - - - - -M 20 L P - 0.5 - - - - - - - - - - - - -PH48 M 20 L RI - - - - - - - - - - 1 - Pr - -PH34 M 20 L RI - 1.5 - - - - - - Pr - - - - - -M 20 L RI - 1 - - - - - - Pr - - - - - -M 20 L RI - - - - - - - - Pr - - - - - -M 20 L RI - 0.5 - - - - - - - - - - - - -PH36 M 20 L RI - 2,1 - - - - - 1.2 - - - - - - -M 20 L P - 1 1.75 - - - - 2 - - - - - - -M 20 L P - 1 1.75 - - - - 2 - - - - - - -PHC35 M 20 L P - 1.78 - - - - - 2.08 - - - - - - -M 20 L P - 1 1.45 - - - - 1.5 - - 1 - - - -151C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L P - - - - - - - 1.5 - - 1 - - - -M 20 L P - 1.2 1.45 - - - - 1.5 - - - - - - -M 20 L P - 0.5 - - - - - 1 - - - - - - -M 20 L P - - - - - - - 1 - - 1 - - - -M 20 L P - 1 - - - - - 1.1 - - - - - - -M 20 R RI - - - - - - - - Pr - 1 - - - -M 20 R RI - - - - - - - - - - 1 - - - -M 20 R-PA Dry - - - - - - - 1.5 - - 1 - - - -M 20 R-PA S - 0.5 - - - - - 1 - - 1 - - - -PH33 M 20 R S - - - - - - - - - - - - - - -M 20 R-PA S - - - - - - - - - - - - - - -M 20 R-PA S - 1.5 - - - - - 2.5 - - - - - - -M 20 R-PA S - 1 - - - - - 2 - - 1 - - - -M 20 R-PA S - 1 - - - - - 2 - - - - - - -PHC32 M 20 R-PA P - 1.2 1.95 - - - - 2.45 - - - - - - -M 20 L P - 0.75 - - - - - - - - - - - - -M 20 R-PA P - - - - - - - - - - 1 - - - -M 20 L RI - 1 1.5 - - - - - - - - - - - -M 20 R-PA P - - - - - - - 1 - - - - - - -M 20 L RI - 1 1.3 - - - - - - - - - - - -M 20 R-PA P - 0.75 - - - - - 1.5 - - - - - - -M 20 R-PA RI - 0.5 1 - - - - 1.1 - - - - - - -M 20 L RI - - 1.2,.75- - - - 1 - - - - - - -PH31 M 20 R-PA P - 0.5 2.3 - - - - 2.5 - - - - - - -152C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L P - 1 - - - - - 1.5 - - - - - - -M 20 L P - - 0.5 - - - - 1 - - 1 - - - -M 20 L P - - 0.75 - - - - 1 - - - - - - -PHC29 M 20 R-PA P - 0.95 1.65 - - - - 1.85 - - - - - - -M 20 R-PA P - - 0.3 - - - - - - - - - - - -M 20 R-PA*P - 1.5 - - - - - 1.6 - - - - - - -M 20 L RI - 1 - - - - - 1.3 - - - - - - -M 20 R-PA*RI - - - - - - - - - - 1 - - - -M 20 L RI - 1 - - - - - 1.2 - - 1 - - - -M 20 R* P - - - - - - - - Pr - - - - - -PH30 M 20 L RI - 1 - - - - - 1.2 - - 1 - - - -M 20 L RI - - 0.5 - - - - 0.6 1 - - - - - -M 20 L P - - 0.5 - - - - - - - - - - - -M 20 L P - - - - - - - - 0.5 - 1 - - - -M 20 L P - - - - - - - - 1 - - - - - -M 20 R RI - - 0.1 - - - - - - - - - - - -M 20 R RI - - - - - - - - - - - Pr - - -M 20 R RI - - - - - - - - - - - Pr - - -M 20 R RI - - - - - - - - - - - Pr - - -M 20 R RI - - - - - - - - - - - Pr - - -M 20 R P - - - - - - - - - - - Pr - - -M 20 R P - - - - - - - - - - - Pr - - -153C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R P - - - - - - - - - - - Pr - - -M 20 R P - - - - - - - - - - - Pr - - -M 20 R P - - - - - - - - - - - Pr - - -M 20 R P - - - - - - - - - - - Pr - - -M 20 R P - 0.3 - - - - - - - - - - - - -M 20 R P - 1 2 - - - - 2.1 - - - - - - -M 20 R P - - - - - - - - - - 1 - - - -M 20 R RI - 1 - - - - - - - - 1 - - - -PH44 M 20 R RI - - - - - - - - - - 1 - - - -M 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - - - - - - - - 0.5 - - - - - -M 20 R P - - 0.5 - - - - 0.6 - - - - - - -M 20 R P - - - - - - - - - - 1 - - - -M 20 R P - - 1 - - - - - - - 1 - - - -M 20 R P - 0.5 - - - - - - - - - - - - -M 20 R P - - - - - - - - 0.75 - - - - - -M 20 R P - - - - - - - - 1 - - - - - -PH45 M 20 L RI - 0.3 - - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - -M 20 L S - - - - - - - 0.3 - - - - - - -M 20 L S - - - - - - - 0.3 - - - - - - -M 20 L S - - - - - - - - - - - - - - -M 20 L S - - - - - - - - - - - - - - -M 20 L S - - - - - - - - - - - - - - -154C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRPHC37 M 20 R S - 0.5 0.75 - - - - - - - - - - - -M 20 L S - - - - - - - - - - - - - - -PH38 M 20 L S - - - - - - - - - - - - - - -M 20 R S - 0.75 - - - - - 1 - - - - - - -PH39 M 20 R S - 1.2 1.4 - - - - 1.5 - - - - - - -M 20 R S - 1.4 - - - - - 1.5 - - - - - - -M 20 R S - - 0.4 - - - - - - - 1 - - - -M 20 R S - 1.4 - - - - - - - - - - - - -M 20 R S - - 1 - - - - - - - - - - - -PHC40 M 20 R S - - 1 - - - - - - - - - - - -M 20 R S - - 0.1 - - - - - - - 1 - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R Dry - - - - - - - - - - 1 - - - -M 20 R Dry - - - - - - - - - - - - - - -M 20 R Dry - - - - - - - - - - - - - - -M 20 R Dry - - - - - - - - - - - - - - -PHC41 M 20 L Dry - 0.9 1.35 - - - - - - - - - - - -M 20 L S - - 1 - - - - - - - - - - - -M 20 L S - - 1.5 - - - - - - - - - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R S - - 0.5 - - - - - - - - - - - -M 20 R S - 1 - - - - - - - - - - - - -M 20 R S - 0.75 - - - - - - - - - - - - -155C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R S - 1 - - - - - - - - - - - - -M 20 R S - - 1.5 - - - - - - - 1 - - - -PHC41 M 20 L S - 1.55 2.1 - - - - - - - - - - - -PHC42 M 20 R S - 2 2.2 - - - - - - - - - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R S - - - - - - - - - - 1 - - - -M 20 R S - - - - - - - - - - - - - - -M 20 R S - - - - - - - - - - - - - - -M 20 L S - 0.5 - - - - - - - - - - - - -M 20 L S - 0.2 - - - - - - - - 1 - - - -M 20 L S - 1 - - - - - - - - - - - - -M 20 L S - 1.5 1.7 - - - - - - - - - - - -M 20 L S - 0.5 - - - - - - - - - - - - -M 20 L S - 0.5 - - - - - - - - - - - - -M 20 L S - - - - - - - - - - 1 - - - -PHC43 M 20 L RI - 1.15 - - - - - - - - 1.55 - - - -M 20 R P - - - - 1.5 - - - - - 1 - - - -M 20 R P - - - - 1.5 0.5 - - - - - - - - -M 20 R P - - - - 1.5 - - - - - 1 - - - -M 20 R P - - 0.5 - - - - - - - - - - - -M 20 R P - 1 - - - - - - - - - - - - -M 20 R P - 1 1.3 - - - - - - - - - - - -M 20 R RI - 1 1.5 - - - - - - - - - - - -156C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R RI - 1 - - - - - - - - - - - - -M 20 R RI - 1 - - - - - - - - 1 - - - -M 20 R P - 1.5 - - - - - - - - - - - - -M 20 R P - 1.5 - - - - - - - - - - - - -M 20 R P - 1.5 - - - - - - - - - - - - -M 20 R P - 1 - - - - - - - - - - - - -M 20 R P - - - - - - - - - - 1 - - - -PHC64 M 20 R P - 1.84 - - - - - - - - - - - - -M 20 R P - - - - - - - - - - - - - - -PH63 M 20 R RI - 3 - - - - - - - - - - - - -M 20 R RI - 2.5 - - - - - - - - - - - - -M 20 R P - 2 - - - - - - - - 1 - - - -M 20 R P - 1.5 2.5 - - - - - - - - - - - -PHC62 M 20 R P - 1.25 2.15 - - - - - - - - - - - -M 20 R P - 0.3 0.7 - - - - - - - - - - - -M 20 R P - 0.5 0.6 - - - - - - - - - - - -M 20 R RI - - 0.5 - - - - - - - - - - - -M 20 R RI - - 0.75 - - - - - - - 1 - - - -M 20 R P - - 0.5 - - - - - - - 1 - - - -M 20 R P - - 0.5 - - - - - - - 1 - - - -M 20 R P - - - - 1 - - - - - - - - - -M 20 R P - - 0.3 - - - - - - - 1 - - - -M 20 R P - 0.5 0.7 - - - - 1 - - - - - - -M 20 R P - 0.5 0.7 - - - - 1 - - - - - - -157C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R P - - 0.5 - - - - - - - - - - - -M 20 R P - 1 - - - - - - - - - - - - -M 20 R P - - 0.5 - - - - - - - 1 - - - -M 20 R P - - - - - - - - - - 1 - - - -M 20 L RI - - - 0.3 - - - - - - - - - - -M 20 L RI - - - 0.75,1.5- - - - - - - - - - -M 20 L RI - 1,1.6 1.5 - - - - - - - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -PHC61 M 20 L P - 1.68 2.2 - - - - - - - - - - - -M 20 L P - 1 1.3 - - - - - - - - - - - -M 20 L P - - - 2 - - - - - - - - - - -M 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - - 0.3 - - - - - 1 - - - - - -M 20 R RI - - - - - - - - 1.5 - - - - - -M 20 R RI - - - - - - - - - - 1 - - - -M 20 L RI - - 0.5 - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - Pr - -M 20 L P - - - - - - - - - - - - Pr - -M 20 L P - - - - - - - - - - - - Pr - -M 20 L P - - - - - - - - - - - - Pr - -PH60 M 20 L RI - - - - - - - - - - - - Pr - -PHC59 M 20 R P - - 1.44 - - - - 0.8 - - - - - - -M 20 R P - - 0.75 - - - - 0.2 - - - - - - -158C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R P - - 0.3 - - - - 0.4 - - - - - - -M 20 R P - - - - 0.5 - - - - - 1 - - - -M 20 R P - - 0.5 - - - - 2.5 - - - - - - -M 20 R P - - - - - - - 0.3 - - 1 - - - -M 20 L RI - - - - - - - - 0.3 - - - - - -M 20 L RI - - - - 0.3 - - - - - - - - - -PH58 M 20 L RI - - - - - - - - 0.75 - - - - - -M 20 L RI - - - - - - - - 1 - - - - - -M 20 L RI - - - - - - - - 1 - - - - - -M 20 L RI - - - - - - - - 1 - - - - - -M 20 R RI - - - - - - - - - - 1 - - - -M 20 L* RI - - - - - - - - - - - - Pr - -M 20 R RI - 1 1.5 - - - - - - - - - - - -M 20 L* RI - 0.75 - - - - - - - - - - - - -M 20 R RI - 1.5 - - - - - - - - - - - - -M 20 L* RI - - - - - - - - - - - - Pr - -M 20 R RI - 1.5 - - - - - - - - - - - - -M 20 L* RI - - - - - - - - - - - - Pr - -M 20 R RI - 1 - - - - - - - - - - - - -M 20 L* RI - 0.5 - - - - - - - - - - - - -M 20 R RI - 0.5 - - - - - - - - - - - - -M 20 L* RI - - - - - - - - - - - - Pr - -PHC57 M 20 R RI - 0.45 1.02 - - - - - - - - - - - -M 20 L* RI - - - - - - - - - - - - Pr - -159C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R RI - - - - - - - - - - 1 - - - -M 20 L* RI - - - - - - - - - - - - Pr - -M 20 R RI - 0.75 - - - - - - - - - - - - -M 20 L* RI - 1 - - - - - - - - - - Pr - -M 20 R RI - 1.5 - - - - - - - - - - - - -M 20 L RI - - - - - - - - - - - - Pr - -M 20 L P - - 2 - - - - 2.1 - - - - - - -M 20 L P - 1.5 2 - - - - - - - - - - - -PHC56 M 20 L P - 1.5 - - 1.92 - - - - - - - - - -M 20 L RI - 0.5 - - - - - - - - - - - - -M 20 R RI - - - - - - - - - - - - - - -PH55 M 20 R RI - 0.5 - - - - - - - - - - - - -M 20 R RI - 0.5 - - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrPH54 M 20 R RI - - - - - - - - - - - - - - -M 20 R RI - - - - - - - - - - 1 - - - -160C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R RI - - 0.3 - - - - - - - - - - - -M 20 L RI - - - - - - - - - - - - - - -M 20 L P - 0.3 - - - - - - - - - - - - -M 20 L P - 0.5 - - - - - - - - - - - - -M 20 L P - - - - - - - - - - 1 - - - -PHC53 M 20 L P - 1.64 - - - - - - - - - - - - -M 20 L P - 0.75 - - - - - - - - - - - - -M 20 L P - 0.75 - - - - - - - - - - - - -M 20 R RI - 1.5 2.5 - - - - 2.6 - - - - - - -M 20 R RI - 1.5 2.5 - - - - 2.6 - - - - - - -M 20 R RI - 1.5 2.5 - - - - 2.6 - - - - - - -M 20 R RI - 1.5 2.5 - - - - 2.6 - - - - - - -M 20 R RI - 1.5 2.5 - - - - 2.6 - - - - - - -M 20 R RI - 2 3 - - - - 3.1 - - - - - - -M 20 R P - 2 3 - - - - 3.1 - - - - - - -M 20 R P - 0.5 - - - - - - - - 1 - - - -PH52 M 20 R RI - 1 - - - - - - - - 1 - - - -M 20 R RI - 1 2 - - - - 2.1 - - - - - - -M 20 R RI - 1 2 - - - - 2.1 - - - - - - -M 20 R RI - 1 - - - - - - - - 1 - - - -M 20 R RI - 1.5 1 - - - - 2.1 - - - - - - -M 20 R RI - - 1 - - - - - - - 1 - - - -M 20 R RI - - 0.5 - - - - - - - - - - - -M 20 R RI - - - - - - - - 0.75 - - - - - -161C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L P - 0.3 - - - - - - - - 1 - - - -M 20 L P - 1 - - - - - - - - 1 - - - -M 20 L P - 1.5 2 - - - - 2.2 - - - - - - -M 20 L P - 1 2 - - - - 2.5 - - - - - - -M 20 R P - 1.5 - - - - - 1.6 - - - - - - -M 20 R P - 2 3 - - - - 3.1 - - - - - - -M 20 R P - 1 - - - - - - - - 1 - - - -PHC51 M 20 R P - 1.64 - - - - - 1.94 - - - - - - -M 20 R P - 1 - - - - - - - - - - - - -M 20 R RI - 1.5 2 - - - - 2.1 - - - - - - -M 20 R RI - 1 - - - - - - - - - - - - -M 20 R RI - - 0.5 - - - - - - - 1 - - - -M 20 R RI - - - - - - - - - - 1 - - - -M 20 R RI - 0.5 - - - - - 0.6 - - - - - - -M 20 R RI - - - - - - - - - - - - Pr - -PH50 M 20 R RI - 1 - - - - - - - - - Pr - - -M 20 R P - 1 1.5 - - - - 1.6 - - - Pr - - -M 20 R P - 1 - - - - - 1.1 - - - - - - -M 20 R P - 1 - - - - - 2.5 - - - - - - -M 20 R P - 1 - - - - - - - - - - - - -PH49 M 20 R RI - 1 - - - - - - - - - - - - -PHC65 M 20 R RI - 1.64 - - - - - 2.24 - - 1 - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -162C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M AP L PB - 0.5 - - - - - - - - - - - - -M 20 L PB - 0.5 1.5 - - - - - - - 1 - - - -M 20 L PB - 0.5 1.5 - - - - - - - 1 - - - -M 20 L PB - 1 2 - - - - 2.1 - - - - - - -M 20 L PB - - 2 - - - - - - - 1 - - - -M 20 L PB - 1 - - - - - 1.1 - - 1 - - - -M 20 L PB - 1.5 - - - - - - - - - - - - -PHC67 M 20 L PB - 1.2 1.66 - - - - - - - - - - - -M 20 L PB - 1 - - - - - 1.1 - - 1 - - - -M 20 L PB - 1.5 - - - - - 1.6 - - - - - - -PH68 M 20 L PB - 0.5 - - - - - 0.6 - - 1 - - - -M 20 L PB - 1.5 0.5 - - - - - - - - - - - -M 20 L PB - 0.5 - - - - - - - - 1 - - - -M 20 L PB - 0.5 - - - - - - - - 1 - - - -M 20 L RI - 0.5 1 - - - - - - - 1 - - - -PH69 M 20 L RI - - - - - - - - - - 1 - - - -M 20 L P - 1 - - - - - - - - 1 - - - -163C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R P - - - - - - - - 1.5 - - - - - -M 20 R P - - - - - - - - 0.15 - - - - - -PHC70 M 20 L P - 1.54 2.2 - - - - 2.3 - - - - - - -M 20 R* P - - - - - - - - 1.75 - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -M 20 R* P - - - - - - - - 1.75 - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -M 20 R* P - 1 - - - - - - - - - - - - -M 20 L P - - - - - - - - - - 1 - - - -M 20 R* RI - 0.3 - - - - - - - - - - - - -PH71 M 20 L RI - 0.5 - - - - - - - - - - - - -M 20 L RI - 1.5 - - - - - 1.6 - - - - - - -M 20 L RI - 0.5 - - - - - 2.5 - - - - - - -M 20 R* RI - 1.5 - - - - - - - - - - - - -M 20 L RI - - 1 - - - - - - - - - - - -M 20 R* RI - 0.5 - - 1.5 - - - - - - - - - -M 20 L RI - 1 - - - - - - - - - - - - -M 20 R RI - 1 - - - - - - - - - - - - -M 20 L RI - 1 - - - - - - - - 1 - - - -M 20 R RI - 1.5 - - - - - - - - - - - - -M 20 L P - 1.5 - - - - - 1.6 - - - - - - -M 20 R P - - 0.5 - - - - 0.6 - - - - - - -M 20 L P - - 1 - - - - - - - 1 - - - -M 20 R P - - - - - - - - - - 1 - - - -164C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L P - - 1.5 - - - - 1.6 - - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -PHC72 M 20 R RI - 1.58 2 - - - - - - - - - - - -M 20 L RI - 2 - - - - - - - - - - - - -M 20 L P - 1.5 - - - - - - - - 1 - - - -M 20 R P - 1.5 - - - - - - - - - - - - -M 20 L P - - 0.5 - - - - - - - - - - - -M 20 L P - - 0.5 - - - - - - - 1 - - - -M 20 L P - 1.5 - - - - - 1.6 - - - - - - -M 20 R P - 1.5 - - - - - - - - - - - - -M 20 L P - - - - - - - - 0.5 - - - - - -M 20 L P - - 0.5 - - - - - - - - - - - -M 20 L P - - 0.3 - - - - - - - - - - - -M 20 L P - - - - - - - - - - - - - - -M 20 R P - - 0.3 - - - - - - - - - - - -M 20 R P - - 0.3 - - - - - - - - - - - -PH73 M 20 R RI - - - - - - - - - - - - - - -M 20 R RI - - - - - - - - - - - - - - -M 20 R P - - - - - - - - - - - - - - -M 20 R P - - 0.3 - - - - - - - - - - - -M 20 R P - - 0.3 - - - - - - - - - - - -M 20 L P - - - 1.5 - - - - - - - - - - -M 20 L P - 1 - - - - - - - - - - - - -PHC74 M 20 L P - 1.76 - - - - - 1.77 - - - - - - -165C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L RI - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - Pr166C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrPHC78 M 20 L RI - 1.34 1.86 - - - - 2.2 - - - - - - -M 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - 0.5 - - - - - - - - - - - - -M 20 L RI - - 0.5 - - - - - - - 1 - - - -M 20 L RI - 1 - - - - - 1.3 - - - - - - -M 20 L RI - 1.5 - - - - - 1.8 - - - - - - -M 20 L RI - 1.5 - - - - - 2.5 - - - - - - -M 20 L RI - 1.25 - - - - - 2 - - - - - - -M 20 L RI - - - - - - - 0.5 - - 1 - - - -M 20 L RI - - - 0.5 - - - - - - - - - - -167C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L RI - - - - - - - 2.5 - - - - - - -M 20 L RI - 0.5 - - - - - 2 - - - - - - -M 20 L P - 0.5 - - - - - - - - 1 - - - -M 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - 6 PrM 20 L P - - - - - - - - - - - - - 6 PrM 20 L P - - - - - - - - - - - - - 6 PrM 20 L RI - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - 1 - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 L P - - - - - - - - - - - - - - PrM 20 R P - 2 - - - - - 2.3 - - - - - - -PHC79 M 20 R P - 1.86 - - - - - 2.32 - - - - - - -M 20 L P - - - - - - - - - - - - - - PrM 20 R P - - 1.5 - - - - - - - - - - - -M 20 L P - 0.75 - - - - - - - - - - - - -M 20 R P - 2 - - - - - - - - - - - - -M 20 L P - 1.5 - - - - - - - - - - - - -M 20 R P - 0.75 1.25 - - - - 1.75 - - - - - - -M 20 L P - 1.5 - - - - - 1.7 - - - - - - -M 20 R P - 0.5 - - - - - - - - 1 - - - -PHC80 M 20 L P - 2 - - - - - - - - - - - - -M 20 L P - 1.5 2 - - - - - - - - - - - -M 20 L RI - 0.75 - - - - - 1.5 - - - - - - -168C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRM 20 L RI - - - - - - - - - - 1 - - - -M 20 L RI - 1 - - - - - 1.5 - - - - - - -M 20 L RI - 1.5 2 - - - - 2.1 - - - - - - -PH81 M 20 L RI - 2 3 - - - - - - - - - - - -M 20 L P - 0.5 2 - - - - 2.1 - - - - - - -M 20 L P - 1.5 - - 2.25 - - 2.35 - - - - - - -M 20 L P - 1 - - - - - 2 - - - - - - -PHC82 M 20 L P - 1.8 2.4 - - - - - - - - - - - -M 20 L RI - 0.5 1 - - - - 1.1 - - - - - - -M 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - - - - - - - - PrM 20 R P - - - - - - - 0.5 - - - - - - -M 20 R P - - - - - - - - 1.5 - - - - - -M 20 R P - - - - - - - - - - 1 - - - -PH83 M 20 R RI - - - 1 - - - 1.3 - - - - - - -M 20 R RI - 2.5 - - - - - - - - - - - - -M 20 R RI - 2.5 - - - - - - - - - - - - -M 20 R RI - - - - - - - - - - 1 - - - -M 20 R RI - - - 0.75 - - - - - - - - - - -M 20 R RI - - - - - - - 3 - - - - - - -M 20 R RI - - - - - - - 3 - - - - - - -M 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - PrM 20 R RI - - - - - - - - - - - - - - Pr169C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRPH06 A 10 P - - - - - - - - - - - - - - -A 10 P - - - - - - - - - - 1 - - - -A 10 P - - - - - - - - - - 1 - - - -A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - - - - - -PHC09 A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - - - - - -A 10 P - - 4 - - - - - - - 1 - - - -A 10 P - - 4 - - - - - - - 1 - - - -A 10 P - - 4 - - - - - - - 1 - - - -A 10 P - - 3.5 - - - - - - - 1 - - - -PH07 A 10 RI - 2 3.5 - - - - - - - - - - - -A 10 RI - 1 - - - - - - - - 1 - - - -A 10 RI - - - - - - - - - - 1 - - - -A 10 RI - - - - - - - - - - 1 - - - -A 10 RI - - - - - - - - - - 1 - - - -A 10 RI - 0.5 - - - - - - - - 1 - - - -A 10 P - 0.5 - - - - - - - - 1 - - - -A 10 P - 0.5 - - - - - - - - 1 - - - -A 10 RI - 1 - - - - 3 - - - 1 - - - -A 10 RI - - - - - - 3 - - - 1 - - - -170C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 10 RI - - - - - - 3 - - - 1 - - - -A 10 RI - - - - - - 3 - - - 1 - - - -A 10 P - - - - - - - - - - 1 - - - -A 10 P - - - - - - 2.7 - - - - - - - -A 10 P - - - - - - 2.7 - - - - - - - -A 10 P - 1 - - - - 2.5 - - - - - - - -A 10 P - 1 3 - - - - - - - - - - - -A 10 RI - 1 3 - - - - - - - - - - - -PHC10 A 10 RI - 1 3 - - - - - - - - - - - -A 10 RI - 1 2.5 - - - - - - - - - - - -A 10 RI - 1 2.5 - - - - - - - - - - - -A 10 RI - 1 3 - - - - - - - - - - - -A 10 RI - 1 3 - - - - - - - - - - - -A 10 RI - 1 3 - - - - - - - - - - - -A 10 RI - 1 3 - - - - - - - - - - - -PH08 A 10 RI - 1 3 - - - - - - - - - - - -A 10 RI - - - - - - - - - - - - - - -PH05 A 10 P - - - - - - - - 1 - 1 - - - -A 10 P - - - - - - - - 1 - 1 - - - -A 10 P - - - - - - - - 1.5 - 1 - - - -A 10 RI - - - - - - - - 1 - 1 - - - -A 10 RI - - - - - - - - 1 - - - - - -A 10 RI - - - - - - - - 1 - - - - - -A 10 RI - - - - - - - - PrB - - - - - -171C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 10 RI - - - - - - - - PrB - - - - - -A 10 RI - - - - - - - - PrB - - - - - -A 10 RI - - - - - - - - PrB - - - - - -A 10 RI - - - - - - - - PrB - - - - - -PH04 A 10 RI - - - - - - - - - - 1 - - - PrA 10 P - - - - - - - - - - 1 - - - PrA 10 P - - - - - - - - - - 1 - - - PrA 10 P - - - - - - - - - - 1 - - - PrA 10 P - - - - - - 0.3 - - - 1 - - - -A 10 RI - - - - - - - - 1.5 - - - - - -A 10 RI - - - - - - - - 1.5 - 1 - - - -A 10 P - - - - - - - - 1.5 - - - - - -A 10 P - - - - - - - - 1.5 - - - - - -A 10 P - - - - - - - - 1 - - - - - -A 10 P - - - - - - - - 1 - - - - - -A 10 P - - - - - - - - PrB - - - - - -A 10 P - - - - - - - - PrB - - - - - -PH03 A 10 P - - - - - - - - - - - - - - -A 10 RI - - - - - - - - - - 1 - - - -A 10 RI - 0.1 - - - - - - - - 1 - - - -A 10 RI - 1.5 - - - - - - - - 1 - - - -A 10 RI - 1 - - - - - - - - 1 - - - -A 10 RI - 0.5 - - - - - - - - - - - - -A 10 RI - 0.5 - - - - - - - - - - - - -172C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 10 RI - 0.2 - - - - - - - - 1 - - - -A 10 RI - 0.2 - - - - - - - - - - - - -A 10 RI - 0.2 - - - - - - - - - - - - -A 10 RI - - - - - - 3 - - - - - - - -PH01 A 10 RI - - - - - 1 - - - - - - - - -A 10 P - 0.1 - - - - - - - - - - - - -A 10 P - 4 - - - - - - - - - - - - -A 10 P - - - - - - 4 - - - - - - - -A 10 P - - - - - - 5 - - - - - - - -A 10 P - - - - - - 3 - - - 1 - - - -A 10 P - - - - - - 5.4 - - - 1 - - - -A 10 P - - - - - - 4.4 - - - - - - - -A 10 P - 0.6 - - - - - - - - 1 - - - -A 10 P - 1 - - - - 4.3 - - - 1 - - - -A 10 P - 0.5 - - - - 4.2 - - - - - - - -A 10 P - - - - - - - - - - 1 - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -173C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -A 10 P tall - - - - - - - - - - - - - -PH19 A 20 RI tall - 1 - - - - - - - - - - - -PHC18 A 20 RI tall - 1 - - - - - - - - - - - -A 20 RI tall - 1 - - - - - - - - - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 P - - - - - - - - 0.75 - - - - - -A 20 P - - - - - - - - 1 - - - - - -A 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrPH17 A 20 RI - 1.5 - - - - - - - - 1 - - - -174C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 RI - 2.5 - - - - - - - - - - - - -A 20 RI - 2.5 - - - - - - - - 1 - - - -A 20 RI - 2.5 - - - - - - - - - - - - -PHC16 A 20 RI - 2 - - - - - - - - - - - - -A 20 RI - 2 - - - - - - - - 1 - - - -A 20 RI - 1.5 - - - - - - - - 1 - - - -A 20 RI - - - - - - - - 1 - 1 - - - -A 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrPH15 A 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -PH14 A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -175C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - 0.8 - - - - - -PH13 A 20 RI - - - - - - - - 0.8 - - - - - -A 20 P - - - - - - - - 0.8 - - - - - -A 20 P - - - - - - - - 0.8 - - - - - -A 20 P - - - - - - - - 0.8 - - - - - -A 20 RI - - - - - - - - 0.8 - - - - - -A 20 RI - - - - - - - - 0.8 - - - - - -PH12 A 20 RI - - - - - - - - 0.8 - - - - - -A 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - 1 - - - - - - - - - - - - PrA 20 P - 1 - - - - - - - - - - - - -A 20 P - 1.5 - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - Pr176C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrPH11 A 20 RI - - - - - - - - - - - - - - PrPH20 A 20 RI - - - - - - - - - - 1 - - - -A 20 P - - - - - - - - - - 1 - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - 1 - - - -A 20 P - - - - - - - - - - 1 - - - -A 20 RI - - - - - - - - - - - - - - -A 20 RI - - - - - - - - - - - - - - PrPH21 A 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - Pr177C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrPH22 A 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - -PH23 A 20 RI - - - - - - - - - - - - - - -A 20 RI - - - - - - - - 0.2 - - - - - -A 20 RI - - - - - - - - 0.5 - - - - - -A 20 RI - - - - - - - - 1 - - - - - -A 20 RI - - - - - - - - 1 - - - - - -A 20 RI - - - - - - - - 1.5 - - - - - -A 20 RI - - - - - - - - - - - - - 1 -A 20 P - - - - - - - - - - - - - 1 -PH24 A 20 RI - - - - - - - - - - - - - 1 -PHC25 A 20 RI - - - - - - - - - - - - - 1 -A 20 P - - - - - - - - 0.5 - - - - - -PH26 A 20 RI - - - - - - - - 1 - - - - - -A 20 RI - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 0.5 - - - - - -A 20 RI - - - - - - - - - - - - - - Pr178C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 RI - - - - - - - - - - - - - - PrPH27 A 20 RI - - - - - - - - - - - - - - PrA 20 P - - - - - - - - - - - - - - PrA 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 0.5 - - - - - -A 20 P - - - - - - - - 1 - - - - - -A 20 P - - - - - - - - 1 - - - - - -A 20 P - - - - - - - - 1 - - - - - -A 20 P - - - - - - - - 1 - - - - - -A 20 P - - - - - - - - 1 - - - - - -PH28 A 20 RI - - - - - - - - - - - - - - -A 20 RI - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - - -A 20 P - - - - - - - - - - - - - 2 -A 20 P - - - - - - - - - - - - - 2 -A 20 P - - - - - - - - - - - - - 1 -A 20 P - - - - - - - - - - - - - 1 -A 20 P - - - - - - - - - - - - - 1 -A 20 P - - - - - - - - - - - - - 2 -A 20 P - - - - - - - - 1.5 - - - - - -A 20 P - - - - - - - - 1.5 - - - - - -179C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRA 20 P - - - - - - - - 2 - - - - - -L 20 L RI - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -L 20 L S - - - - - - - - - - - Pr - - -PHC85 L 20 L P - 1.87 3.1 - - - - - - - - - - - -L 20 R* P - - - - - - - - - - - Pr - - -L 20 L P - 1 1.5 - - - - 1.6 - - - - - - -L 20 R* P - - - - - - - - - - - Pr - - -L 20 L RI - 1 - - - - - 1.3 - - - - - - -L 20 R* RI - - - - - - - - - - - Pr - - -L 20 L RI - - - - - - - - - - - - - - PrPH84 L 20 R* RI - - - - - - - - 0.75 - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 R* P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - - - - - - - PrL 20 R* P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - - - - - - - PrL 20 R* P - - - - - - - - 1 - - - - - -L 20 R RI - - - - - - - - - - - Pr - - -180C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R P - - - - - - - - - - - - - - PrL 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -181C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 R RI - - - - - - - - - - - Pr - - -L 20 L P - - - - - - - - - - - - - - PrL 20 L P - - - - - - - - - - - - - - PrL 20 L P - - - - - - - - - - - - - - PrL 20 L P - - - - - - - - - - - - - - PrL 20 L P - - - - - - - - - - - - - - PrL 20 L P - - 0.6 - - - - - - - - - - - -L 20 L P - - 1 - - - - - - - - - - - -L 20 R P - 2 - - - - - - - - - - - - -L 20 R P - 2 - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - Pr - - -L 20 R P - - - - - - - - - - - Pr - - -L 20 L P - - 0.5 - - - - 0.7 - - - - - - -PH111 L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 L P - - 0.5 - - - - - - - - - - - -L 20 L P - - 1.5 - - - - - - - - - - - -L 20 L P - - 2.5 - - - - - - - - - - - -L 20 L P - - 3 - - - - - - - - - - - -L 20 L RI - 1.5 - - - - 2 1.75 - - - - - - -L 20 L* RI - 1.5 - - - - - - - - 1 - - - -182C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRPHC110 L 20 R RI - 0.99 3.09 - - - - - - - - - - - -L 20 L* RI - 0.75 - - - - - - - - - - - - -L 20 R RI - 1.5 3 - - - - - - - - - - - -L 20 L* RI - - - - - - - - - - - - - 1 -L 20 R RI - 2 - - - - - - - - - - - - -L 20 L* RI - - - - - - - - - - - - - 1 -L 20 R RI - 1.25 2 - - - - - - - - - - - -L 20 L* RI - - - - - - - - - - - - - 1 -L 20 R RI - 0.5 1 - - - - 1.1 - - - - - - -L 20 L* RI - - - - - - - - - - - - - 1 -L 20 R RI - 0.5 1 - - - - 1.1 - - - - - - -L 20 L* RI - - - - - - - - - - - - - 1 -L 20 R RI - 0.5 - - - - - 0.7 - - 1 - - - -PH112 L 20 L RI - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - - - 1 - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - 1.5 - - - - - 1.25 - - 1 - - - -L 20 L P - 1 - - - - - 2 - - - - - - -L 20 R P - - - - - - - - - 1 - - - - -PHC113 L 20 L RI - 1.78 - - - - - 1.98 - - - - - - -L 20 R* RI - - - - - - - - - 1.5 - - - - -PH114 L 20 L RI - 1.75 1.85 - - - - 2 - - - - - - -L 20 R* RI - - - - - - - - - 1.5 - - - - -183C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L P - - - - - - - - - - 1 - - - -L 20 R* RI - - - - - - - - - - 1 - - - -L 20 L RI - - - 1.25 - - - - - - - - - - -L 20 R* RI - - - - - - - - - 1 - - - - -L 20 R RI - - - - - - - - - 1.5 - - - - -L 20 R RI - - - - - - - - - 1 - - - - -L 20 R RI - - - - - - - - - 0.3 - - - - -L 20 R RI - 1.75 2 - - - - - - - - - - - -PH115/PHC116 L 20 R RI - 2.24 2.81 - - - - - - - - - - - -L 20 R P - 1 2 - - - - - - - - - - - -L 20 R P - 1 1.75 - - - - - - - - - - - -L 20 R P - 1 1.5 - - - - 1.6 - - - - - - -L 20 R P - 1 - - - - - - - - 1 - - - -PH117 L 20 R RI - - - - - - - - - - - - - - -L 20 L RI - - - - - - - - - - - - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 L RI - - - - - - - - 1.5 - - - - - -L 20 L RI - 0.75 1.5 - - - - - - - 1 - - - -L 20 L P - 1 - - - - - 1.5 - - - - - - -L 20 L P - - - 1 - - - 1.5 - - - - - - -L 20 L P - - 1.5 - - - - 1.7 - - - - - - -L 20 L P - 2 3 - - - - - - - - - - - -L 20 L P - 1.5 3 - - - - - - - - - - - -184C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRPHC118 L 20 L RI - 1.4 3.2 - - - - 4.2 - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - 0.5 - - - - - 0.75 - - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 R RI - - - - - - - 0.7 0.5 - - - - - -L 20 R P - - 0.5 - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - 0.5 - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - 1 - - - - - - - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - - - - - - - - -L 20 R RI - - - - - - - - - - - - - - -PH119 L 20 L P - - 0.5 0.25 - - - - - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - 1 - - - - - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - 0.3 - - - - - - - - - - - - -L 20 L P - 0.3 0.75 - - - - - - - - - - - -L 20 L P - 0.75 1.5 - - - - - - - - - - - -185C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L RI - - - - 1.5 - - - - - - - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 L RI - - - - - - - - - - 1 - - - -PH109 L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 0.5 - - - - - -L 20 L P - - - - - - - - - 0.5 - - - - -L 20 L P - - - - - - - - - 0.5 - - - - -L 20 L P - - 0.3 - - - - - - - - - - - -L 20 L P - - 0.75 - - - - - - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R P - - - - 0.5 - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R RI - - - - 0.5 - - - - - - - - - -L 20 R RI - - - - 0.3 - - - - - - - - - -L 20 R RI - - - - 0.3 - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -PH108 L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - 0.5 - - - - - - - - - - -L 20 R RI - - - - - - - - - - - - - - -186C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L P - - - - - - - - - 1 1 - - - -L 20 L P - - - - - - - - - 0.75 - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - 1 - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - 1.5 - - - - - 1.6 - - - - - - -L 20 R P - - 0.5 - - - - - - - - - - - -L 20 R P - 1.5 - - - - - 1.6 - - - - - - -L 20 R P - 1.5 - - - - - 1.6 - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - 2 - - - - - - - - - - - - -L 20 R P - 2 - - - - - - - - - - - - -L 20 R P - - - - - - - 1 - - - - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -PH107 L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - 1 - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - 0.2 - - - - 0.5 - - - - - - -187C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 R P - - 1 - - - - 2 - - - - - - -PH106 L 20 R RI - - 1.5 - - - - 2 - - - - - - -PHC105 L 20 R RI - 1.02 1.67 - - - - 2.22 - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - 1.75 - - - - 2 - - - - - - -L 20 R RI - - 0.5 - - - - 0.75 - - - - - - -L 20 L RI - - - - - - - - - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - 1 - - - - - 1.3 - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - - - - - - - -PH104 L 20 L P - 1 - - - - - 1.5 - - - - - - -L 20 L RI - 0.3 - - - - - 1 - - - - - - -L 20 L RI - - 0.2 - - - - 1 - - - - - - -L 20 L P - 1.5 - - - - - - - - - - - - -L 20 L P - 1.5 - - - - - - - - - - - - -L 20 L P - 0.2 - - - - - 1 - - - - - - -L 20 L P - 0.5 - - - - - - - - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 L P - 0.3 - - - - - - - - 1 - - - -L 20 L RI - 0.3 - - - - - 1 - - - - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 L P - - 0.5 - - - - - - - - - - - -PHC103 L 20 R P - 1.46 - - - - - - - - - - - - -188C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L P - - - - - - - 1 - - - - - - -L 20 L P - - .5,2 - - - - 1 - - - - - - -L 20 L P - - 1 - - - - - - - - - - - -L 20 L P - - 1 - - - - - - - - - - - -L 20 L P - - - 1 - - - - - - - - - - -L 20 L P - - 1 - - - - - - - - - - - -L 20 L RI - - 0.5 - - - - - - - - - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - - 0.5 - - - - -L 20 L P - - - - - - - 1 - - - - - - -L 20 L P - - - - - - - 1 - - - - - - -L 20 L P - - - - - - - 1 - - 1 - - - -L 20 L P - - - - - - - 0.3 - - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 L P - - - - - - - 0.3 - - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - 1 - - 1 - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - 1 - - - - - - -L 20 R P - - - - - - - - - 0.75 - - - - -L 20 R P - - - - - - 0.3 1 - - - - - - -PH102 L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -189C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - - - - - - - -PHC101 L 20 L RI - 0.66 1.24 - - - - 1.68 - - - - - - -L 20 L P - - - - - - - - - - - - - - -L 20 L P - - 0.5 - - - - 0.7 - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -PH100 L 20 R RI - - - - 0.3 - - - - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - 2.2 - - - 2.5 - - - - - - -L 20 L RI - - - 0.75 - - - 1 - - - - - - -PHC99 L 20 L RI - 0.82 - - - - - 0.86 - - - - - - -L 20 L P - 0.3 - - - - - 1.5 - - 1 - - - -L 20 L P - 0.3 - - - - - 0.75 - - 1 - - - -L 20 L P - 0.3 - - - - - 0.75 - - - - - - -L 20 L P - - - - - - - - - 2 - - - - -L 20 L P - - - - - - - - - 1 - - - - -L 20 L P - - - - - - - - - 0.1 - - - - -L 20 L P - - - - - - - - - 0.5 - - - - -L 20 L P - - - - - - - 0.5 - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 L P - 1.5 - - - - - 1.7 - - - - - - -PHC120 L 20 L RI - 1.29 1.72 - - - - 2.33 - - - - - - -190C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L RI - 0.5 - - - - - 1.25 - - - - - - -L 20 L RI - - - - - - - - - - - - - - PrL 20 L P - - - - - - - - - - - - - - PrL 20 L RI - - - - - - - - - - - - - - PrPH121 L 20 L RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - - - - - - - - - - - - - -L 20 R RI - - - 0.75 - - - 0.85 - - - - - - -L 20 R RI - - 0.5 - - - - 0.75 - - - - - - -L 20 R P - - - - - - - 1 - - 1 - - - -L 20 R P - 0.75 - - - - - - - - - - - - -L 20 R P - - - 0.5 - - - - - - - - - - -L 20 R P - - - 0.5 - - - - - - - - - - -PHC98 L 20 R RI - 1.06 1.98 - - - - - - - - - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - 0.5 - - - - - - - 1 - - - -L 20 R RI - 0.75 - - - - - 1 - - - - - - -PH97 L 20 R RI - - - - - - - - - - - - - - -L 20 R P - - - 0.5 - - - 1.25 - - 1 - - - -L 20 R RI - - - - - - - 1 - - 1 - - - -L 20 R RI - - - 1 - - - 1.5 - - - - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - 0.75 - - - - - - - - - - -L 20 R RI - - - 0.5 - - - 0.75 - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -191C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L P - - - - - - - - - 1.5 - - - - -L 20 L RI - - - - - - - - - 0.5 - - - - -L 20 L RI - - - 1 - - - 2.5 - - - - - - -L 20 L P - - - 2 2.5 - - - - - - - - - -L 20 L RI - - - - - - - 2.5 - - - - - - -L 20 L RI - - - 0.5 - - - 2.5 - - - - - - -L 20 L RI - - - - - - 2.5 2.25 - - - - - - -L 20 L RI - 1.75 2 - - - - - - - - - - - -L 20 L RI - 0.5 - - - - - 1.5 - - - - - - -L 20 L RI - - - - - - - 1.5 - - 1 - - - -PHC96 L 20 L RI - - - - - - - 1.5 - - 1 - - - -L 20 L RI - 1.6 - - - - - 2.7 - - - - - - -L 20 L P - 1.75 2 - - - - 3 - - - - - - -L 20 R P - - - - - - - 1 - - 1 - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - - 0.5 - - - - -PH94 L 20 R P - - - - - - - - - - - - - - -L 20 R RI - - - - - 1 - - - - - - - - -PHC93 L 20 R P - 1.1 1.38 - - - - - - - - - - - -L 20 R P - - 0.5 - - - - 0.6 - - - - - - -L 20 R P - - 0.5 - - - - - - - - - - - -L 20 R P - - - 1 - - - - - - - - - - -L 20 R P - - - 1 - - - - - - - - - - -PH92 L 20 R RI - - - - - - - - - - 1 - - - -192C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L RI - - - - - - - - - - - - - - -L 20 L RI - - - - - - - - - - - - - - -L 20 L RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - - - - 1 - - - -L 20 R RI - - - - - - - 0.5 - - - - - - -L 20 R RI - - 0.3 - - - - 0.5 - - - - - - -L 20 R RI - - - - - - - 0.75 - - 1 - - - -L 20 R RI - - 0.5 - - - - 1 - - - - - - -L 20 R RI - 1.25 - - - - - - - - - - - - -PHC91 L 20 R RI - 1.2 - - - - - 1.46 - - - - - - -PH90 L 20 R RI - - 1 - - - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - 1 - - - - - - - - - - - -L 20 R P - - 1 - - - - - - - - - - - -L 20 R P - - - - - - - - - - - - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - - - - 1 - - - -L 20 R P - - - - - - - 0.75 - - - - - - -L 20 L RI - - - - - - 0.3 - - - - - - - -L 20 L RI - - - - - - - - - 0.5 - - - - -L 20 L RI - - - - - - - - - 0.5 - - - - -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -193C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 1 -L 20 L RI - - - - - - - - - - - - - 1 -L 20 L RI - - - - - - - - - - - - - 1 -L 20 L P - - - - - - - - - - - - - 2 -L 20 L P - - - - - - - - - - - - - 3 -PH88 L 20 L RI - 0.5 - - - - - - - - - - - - -L 20 L RI - 0.5 - - - - - - - - - - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - - - 1 - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 R RI - - - - - - 0.75 - - - 1 - - - -L 20 R RI - - - 0.5 - - - 1.25 - - - - - - -L 20 R P - - - - - 0.75 - 1 - - - - - - -PHC87 L 20 R P - - 1.3 - - - - 1.72 - - - - - - -194C.4.SurveydataTable C.8: Survey dataPH RH SP BK MO RE CT CE FCT FCE GT GE OB CB FB LWDRR BD TL BRL 20 R P - - 0.5 - - - - 1 - - - - - - -L 20 R P - - 0.75 - - - - 1.5 - - - - - - -L 20 R P - - - - 1.5 - - - - - - - - - -L 20 R P - 0.5 - - - - - - - - - - - - -L 20 R P - - - - - - - 1.5 - - 1 - - - -L 20 R P - 0.5 - - - - - 1.5 - - - - - - -L 20 R P - 0.5 - - - - - 1.5 - - - - - - -L 20 R P - 0.5 - - - - - 1.5 - - - - - - -L 20 L P - - - - - - - 1 - - - - - - -L 20 L P - - - - - - - - 2 - - - - - -L 20 L P - - - - - - - - 2 - - - - - -L 20 L P - - - - - - - - 1.5 - - - - - -L 20 L P - - - - - - - - 1.5 - - - - - -L 20 L P - - - - - - - - 1.5 - - - - - -L 20 L P - - - - - - - - 1.5 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 1 - - - - - -L 20 L P - - - - - - - - 0.5 - - - - - -L 20 L P - - - - - - - 1 - - - - - - -195Appendix DAir photo informationAvailable information on air photos used for the analysis in Chapter 3 are presented in TableD.1. These include photos for the Middle Elwha, which are also presented in Table 3.3 inChapter 3, and for the Upper Elwha.196Table D.1: Air photos with available accompanying dataPhoto date Source R/S RE(m)TE(m)ME UE1939 NOAA Fisheries* - - - - X1968 NOAA Fisheries* - - - - X1976** National Park Service*** - 7 12.2 X -1976 NOAA Fisheries* - - - - X1981 National Park Service*** - 15 18 X -1981 NOAA Fisheries* - - - - X1990-09-04 USGS DOQ 1:12500 2 10.2 - X1994-09-21 USGS DOQ 1:12000 3.9 10.7 X -2000 NOAA Fisheries* - - - - X2006-04-01 USDA NAIP 1 m 5 11.2 X X2009-10-08 USDA NAIP 1 m 5 11.2 X X2014-12-30 USGS/National Park Service 0.05 m - - X X2015-04-16 USGS/National Park Service 0.05 m - - X -2016-08-11 USGS/National Park Service 0.05 m - - X X*Air photos digitized by NOAA Fisheries.***Air photos digitized by author.**Coverage of air photo does not extend to whole study areaR/S Resolution or scaleRE Registration errorTE Total error (registration + digitization)MR Photo available for the Middle Elwha (between the dams)UR Photo available for the Upper Elwha (upstream of Glines Canyon)197

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.24.1-0364666/manifest

Comment

Related Items